Fixed MTP to work with TWRP

fs/ocfs2/Kconfig

@@ -0,0 +1,76 @@
+config OCFS2_FS
+	tristate "OCFS2 file system support"
+	depends on NET && SYSFS && CONFIGFS_FS
+	select JBD2
+	select CRC32
+	select QUOTA
+	select QUOTA_TREE
+	select FS_POSIX_ACL
+	help
+	  OCFS2 is a general purpose extent based shared disk cluster file
+	  system with many similarities to ext3. It supports 64 bit inode
+	  numbers, and has automatically extending metadata groups which may
+	  also make it attractive for non-clustered use.
+
+	  You'll want to install the ocfs2-tools package in order to at least
+	  get "mount.ocfs2".
+
+	  Project web page:    http://oss.oracle.com/projects/ocfs2
+	  Tools web page:      http://oss.oracle.com/projects/ocfs2-tools
+	  OCFS2 mailing lists: http://oss.oracle.com/projects/ocfs2/mailman/
+
+	  For more information on OCFS2, see the file
+	  <file:Documentation/filesystems/ocfs2.txt>.
+
+config OCFS2_FS_O2CB
+	tristate "O2CB Kernelspace Clustering"
+	depends on OCFS2_FS
+	default y
+	help
+	  OCFS2 includes a simple kernelspace clustering package, the OCFS2
+	  Cluster Base.  It only requires a very small userspace component
+	  to configure it. This comes with the standard ocfs2-tools package.
+	  O2CB is limited to maintaining a cluster for OCFS2 file systems.
+	  It cannot manage any other cluster applications.
+
+	  It is always safe to say Y here, as the clustering method is
+	  run-time selectable.
+
+config OCFS2_FS_USERSPACE_CLUSTER
+	tristate "OCFS2 Userspace Clustering"
+	depends on OCFS2_FS && DLM
+	default y
+	help
+	  This option will allow OCFS2 to use userspace clustering services
+	  in conjunction with the DLM in fs/dlm.  If you are using a
+	  userspace cluster manager, say Y here.
+
+	  It is safe to say Y, as the clustering method is run-time
+	  selectable.
+
+config OCFS2_FS_STATS
+	bool "OCFS2 statistics"
+	depends on OCFS2_FS && DEBUG_FS
+	default y
+	help
+	  This option allows some fs statistics to be captured. Enabling
+	  this option may increase the memory consumption.
+
+config OCFS2_DEBUG_MASKLOG
+	bool "OCFS2 logging support"
+	depends on OCFS2_FS
+	default y
+	help
+	  The ocfs2 filesystem has an extensive logging system.  The system
+	  allows selection of events to log via files in /sys/o2cb/logmask/.
+	  This option will enlarge your kernel, but it allows debugging of
+	  ocfs2 filesystem issues.
+
+config OCFS2_DEBUG_FS
+	bool "OCFS2 expensive checks"
+	depends on OCFS2_FS
+	default n
+	help
+	  This option will enable expensive consistency checks. Enable
+	  this option for debugging only as it is likely to decrease
+	  performance of the filesystem.

fs/ocfs2/Makefile

@@ -0,0 +1,53 @@
+ccflags-y := -Ifs/ocfs2
+
+ccflags-y += -DCATCH_BH_JBD_RACES
+
+obj-$(CONFIG_OCFS2_FS) += 	\
+	ocfs2.o			\
+	ocfs2_stackglue.o
+
+obj-$(CONFIG_OCFS2_FS_O2CB) += ocfs2_stack_o2cb.o
+obj-$(CONFIG_OCFS2_FS_USERSPACE_CLUSTER) += ocfs2_stack_user.o
+
+ocfs2-objs := \
+	alloc.o 		\
+	aops.o 			\
+	blockcheck.o		\
+	buffer_head_io.o	\
+	dcache.o 		\
+	dir.o 			\
+	dlmglue.o 		\
+	export.o 		\
+	extent_map.o 		\
+	file.o 			\
+	heartbeat.o 		\
+	inode.o 		\
+	ioctl.o 		\
+	journal.o 		\
+	localalloc.o 		\
+	locks.o			\
+	mmap.o 			\
+	namei.o 		\
+	refcounttree.o		\
+	reservations.o		\
+	move_extents.o		\
+	resize.o		\
+	slot_map.o 		\
+	suballoc.o 		\
+	super.o 		\
+	symlink.o 		\
+	sysfile.o 		\
+	uptodate.o		\
+	quota_local.o		\
+	quota_global.o		\
+	xattr.o			\
+	acl.o
+
+ocfs2_stackglue-objs := stackglue.o
+ocfs2_stack_o2cb-objs := stack_o2cb.o
+ocfs2_stack_user-objs := stack_user.o
+
+obj-$(CONFIG_OCFS2_FS) += dlmfs/
+# cluster/ is always needed when OCFS2_FS for masklog support
+obj-$(CONFIG_OCFS2_FS) += cluster/
+obj-$(CONFIG_OCFS2_FS_O2CB) += dlm/

fs/ocfs2/acl.c

@@ -0,0 +1,312 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * acl.c
+ *
+ * Copyright (C) 2004, 2008 Oracle.  All rights reserved.
+ *
+ * CREDITS:
+ * Lots of code in this file is copy from linux/fs/ext3/acl.c.
+ * Copyright (C) 2001-2003 Andreas Gruenbacher, <agruen@suse.de>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/string.h>
+
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+#include "alloc.h"
+#include "dlmglue.h"
+#include "file.h"
+#include "inode.h"
+#include "journal.h"
+#include "ocfs2_fs.h"
+
+#include "xattr.h"
+#include "acl.h"
+
+/*
+ * Convert from xattr value to acl struct.
+ */
+static struct posix_acl *ocfs2_acl_from_xattr(const void *value, size_t size)
+{
+	int n, count;
+	struct posix_acl *acl;
+
+	if (!value)
+		return NULL;
+	if (size < sizeof(struct posix_acl_entry))
+		return ERR_PTR(-EINVAL);
+
+	count = size / sizeof(struct posix_acl_entry);
+
+	acl = posix_acl_alloc(count, GFP_NOFS);
+	if (!acl)
+		return ERR_PTR(-ENOMEM);
+	for (n = 0; n < count; n++) {
+		struct ocfs2_acl_entry *entry =
+			(struct ocfs2_acl_entry *)value;
+
+		acl->a_entries[n].e_tag  = le16_to_cpu(entry->e_tag);
+		acl->a_entries[n].e_perm = le16_to_cpu(entry->e_perm);
+		switch(acl->a_entries[n].e_tag) {
+		case ACL_USER:
+			acl->a_entries[n].e_uid =
+				make_kuid(&init_user_ns,
+					  le32_to_cpu(entry->e_id));
+			break;
+		case ACL_GROUP:
+			acl->a_entries[n].e_gid =
+				make_kgid(&init_user_ns,
+					  le32_to_cpu(entry->e_id));
+			break;
+		default:
+			break;
+		}
+		value += sizeof(struct posix_acl_entry);
+
+	}
+	return acl;
+}
+
+/*
+ * Convert acl struct to xattr value.
+ */
+static void *ocfs2_acl_to_xattr(const struct posix_acl *acl, size_t *size)
+{
+	struct ocfs2_acl_entry *entry = NULL;
+	char *ocfs2_acl;
+	size_t n;
+
+	*size = acl->a_count * sizeof(struct posix_acl_entry);
+
+	ocfs2_acl = kmalloc(*size, GFP_NOFS);
+	if (!ocfs2_acl)
+		return ERR_PTR(-ENOMEM);
+
+	entry = (struct ocfs2_acl_entry *)ocfs2_acl;
+	for (n = 0; n < acl->a_count; n++, entry++) {
+		entry->e_tag  = cpu_to_le16(acl->a_entries[n].e_tag);
+		entry->e_perm = cpu_to_le16(acl->a_entries[n].e_perm);
+		switch(acl->a_entries[n].e_tag) {
+		case ACL_USER:
+			entry->e_id = cpu_to_le32(
+				from_kuid(&init_user_ns,
+					  acl->a_entries[n].e_uid));
+			break;
+		case ACL_GROUP:
+			entry->e_id = cpu_to_le32(
+				from_kgid(&init_user_ns,
+					  acl->a_entries[n].e_gid));
+			break;
+		default:
+			entry->e_id = cpu_to_le32(ACL_UNDEFINED_ID);
+			break;
+		}
+	}
+	return ocfs2_acl;
+}
+
+static struct posix_acl *ocfs2_get_acl_nolock(struct inode *inode,
+					      int type,
+					      struct buffer_head *di_bh)
+{
+	int name_index;
+	char *value = NULL;
+	struct posix_acl *acl;
+	int retval;
+
+	switch (type) {
+	case ACL_TYPE_ACCESS:
+		name_index = OCFS2_XATTR_INDEX_POSIX_ACL_ACCESS;
+		break;
+	case ACL_TYPE_DEFAULT:
+		name_index = OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT;
+		break;
+	default:
+		return ERR_PTR(-EINVAL);
+	}
+
+	retval = ocfs2_xattr_get_nolock(inode, di_bh, name_index, "", NULL, 0);
+	if (retval > 0) {
+		value = kmalloc(retval, GFP_NOFS);
+		if (!value)
+			return ERR_PTR(-ENOMEM);
+		retval = ocfs2_xattr_get_nolock(inode, di_bh, name_index,
+						"", value, retval);
+	}
+
+	if (retval > 0)
+		acl = ocfs2_acl_from_xattr(value, retval);
+	else if (retval == -ENODATA || retval == 0)
+		acl = NULL;
+	else
+		acl = ERR_PTR(retval);
+
+	kfree(value);
+
+	return acl;
+}
+
+/*
+ * Helper function to set i_mode in memory and disk. Some call paths
+ * will not have di_bh or a journal handle to pass, in which case it
+ * will create it's own.
+ */
+static int ocfs2_acl_set_mode(struct inode *inode, struct buffer_head *di_bh,
+			      handle_t *handle, umode_t new_mode)
+{
+	int ret, commit_handle = 0;
+	struct ocfs2_dinode *di;
+
+	if (di_bh == NULL) {
+		ret = ocfs2_read_inode_block(inode, &di_bh);
+		if (ret) {
+			mlog_errno(ret);
+			goto out;
+		}
+	} else
+		get_bh(di_bh);
+
+	if (handle == NULL) {
+		handle = ocfs2_start_trans(OCFS2_SB(inode->i_sb),
+					   OCFS2_INODE_UPDATE_CREDITS);
+		if (IS_ERR(handle)) {
+			ret = PTR_ERR(handle);
+			mlog_errno(ret);
+			goto out_brelse;
+		}
+
+		commit_handle = 1;
+	}
+
+	di = (struct ocfs2_dinode *)di_bh->b_data;
+	ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (ret) {
+		mlog_errno(ret);
+		goto out_commit;
+	}
+
+	inode->i_mode = new_mode;
+	inode->i_ctime = CURRENT_TIME;
+	di->i_mode = cpu_to_le16(inode->i_mode);
+	di->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec);
+	di->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
+	ocfs2_update_inode_fsync_trans(handle, inode, 0);
+
+	ocfs2_journal_dirty(handle, di_bh);
+
+out_commit:
+	if (commit_handle)
+		ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle);
+out_brelse:
+	brelse(di_bh);
+out:
+	return ret;
+}
+
+/*
+ * Set the access or default ACL of an inode.
+ */
+int ocfs2_set_acl(handle_t *handle,
+			 struct inode *inode,
+			 struct buffer_head *di_bh,
+			 int type,
+			 struct posix_acl *acl,
+			 struct ocfs2_alloc_context *meta_ac,
+			 struct ocfs2_alloc_context *data_ac)
+{
+	int name_index;
+	void *value = NULL;
+	size_t size = 0;
+	int ret;
+
+	if (S_ISLNK(inode->i_mode))
+		return -EOPNOTSUPP;
+
+	switch (type) {
+	case ACL_TYPE_ACCESS:
+		name_index = OCFS2_XATTR_INDEX_POSIX_ACL_ACCESS;
+		if (acl) {
+			umode_t mode = inode->i_mode;
+			ret = posix_acl_equiv_mode(acl, &mode);
+			if (ret < 0)
+				return ret;
+			else {
+				if (ret == 0)
+					acl = NULL;
+
+				ret = ocfs2_acl_set_mode(inode, di_bh,
+							 handle, mode);
+				if (ret)
+					return ret;
+
+			}
+		}
+		break;
+	case ACL_TYPE_DEFAULT:
+		name_index = OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT;
+		if (!S_ISDIR(inode->i_mode))
+			return acl ? -EACCES : 0;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	if (acl) {
+		value = ocfs2_acl_to_xattr(acl, &size);
+		if (IS_ERR(value))
+			return (int)PTR_ERR(value);
+	}
+
+	if (handle)
+		ret = ocfs2_xattr_set_handle(handle, inode, di_bh, name_index,
+					     "", value, size, 0,
+					     meta_ac, data_ac);
+	else
+		ret = ocfs2_xattr_set(inode, name_index, "", value, size, 0);
+
+	kfree(value);
+
+	return ret;
+}
+
+int ocfs2_iop_set_acl(struct inode *inode, struct posix_acl *acl, int type)
+{
+	return ocfs2_set_acl(NULL, inode, NULL, type, acl, NULL, NULL);
+}
+
+struct posix_acl *ocfs2_iop_get_acl(struct inode *inode, int type)
+{
+	struct ocfs2_super *osb;
+	struct buffer_head *di_bh = NULL;
+	struct posix_acl *acl;
+	int ret = -EAGAIN;
+
+	osb = OCFS2_SB(inode->i_sb);
+	if (!(osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL))
+		return NULL;
+
+	ret = ocfs2_read_inode_block(inode, &di_bh);
+	if (ret < 0)
+		return ERR_PTR(ret);
+
+	acl = ocfs2_get_acl_nolock(inode, type, di_bh);
+
+	brelse(di_bh);
+
+	return acl;
+}

fs/ocfs2/acl.h

@@ -0,0 +1,39 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * acl.h
+ *
+ * Copyright (C) 2004, 2008 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#ifndef OCFS2_ACL_H
+#define OCFS2_ACL_H
+
+#include <linux/posix_acl_xattr.h>
+
+struct ocfs2_acl_entry {
+	__le16 e_tag;
+	__le16 e_perm;
+	__le32 e_id;
+};
+
+struct posix_acl *ocfs2_iop_get_acl(struct inode *inode, int type);
+int ocfs2_iop_set_acl(struct inode *inode, struct posix_acl *acl, int type);
+int ocfs2_set_acl(handle_t *handle,
+			 struct inode *inode,
+			 struct buffer_head *di_bh,
+			 int type,
+			 struct posix_acl *acl,
+			 struct ocfs2_alloc_context *meta_ac,
+			 struct ocfs2_alloc_context *data_ac);
+
+#endif /* OCFS2_ACL_H */

fs/ocfs2/alloc.h

@@ -0,0 +1,324 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * alloc.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_ALLOC_H
+#define OCFS2_ALLOC_H
+
+
+/*
+ * For xattr tree leaf, we limit the leaf byte size to be 64K.
+ */
+#define OCFS2_MAX_XATTR_TREE_LEAF_SIZE 65536
+
+/*
+ * ocfs2_extent_tree and ocfs2_extent_tree_operations are used to abstract
+ * the b-tree operations in ocfs2. Now all the b-tree operations are not
+ * limited to ocfs2_dinode only. Any data which need to allocate clusters
+ * to store can use b-tree. And it only needs to implement its ocfs2_extent_tree
+ * and operation.
+ *
+ * ocfs2_extent_tree becomes the first-class object for extent tree
+ * manipulation.  Callers of the alloc.c code need to fill it via one of
+ * the ocfs2_init_*_extent_tree() operations below.
+ *
+ * ocfs2_extent_tree contains info for the root of the b-tree, it must have a
+ * root ocfs2_extent_list and a root_bh so that they can be used in the b-tree
+ * functions.  It needs the ocfs2_caching_info structure associated with
+ * I/O on the tree.  With metadata ecc, we now call different journal_access
+ * functions for each type of metadata, so it must have the
+ * root_journal_access function.
+ * ocfs2_extent_tree_operations abstract the normal operations we do for
+ * the root of extent b-tree.
+ */
+struct ocfs2_extent_tree_operations;
+struct ocfs2_extent_tree {
+	struct ocfs2_extent_tree_operations	*et_ops;
+	struct buffer_head			*et_root_bh;
+	struct ocfs2_extent_list		*et_root_el;
+	struct ocfs2_caching_info		*et_ci;
+	ocfs2_journal_access_func		et_root_journal_access;
+	void					*et_object;
+	unsigned int				et_max_leaf_clusters;
+};
+
+/*
+ * ocfs2_init_*_extent_tree() will fill an ocfs2_extent_tree from the
+ * specified object buffer.
+ */
+void ocfs2_init_dinode_extent_tree(struct ocfs2_extent_tree *et,
+				   struct ocfs2_caching_info *ci,
+				   struct buffer_head *bh);
+void ocfs2_init_xattr_tree_extent_tree(struct ocfs2_extent_tree *et,
+				       struct ocfs2_caching_info *ci,
+				       struct buffer_head *bh);
+struct ocfs2_xattr_value_buf;
+void ocfs2_init_xattr_value_extent_tree(struct ocfs2_extent_tree *et,
+					struct ocfs2_caching_info *ci,
+					struct ocfs2_xattr_value_buf *vb);
+void ocfs2_init_dx_root_extent_tree(struct ocfs2_extent_tree *et,
+				    struct ocfs2_caching_info *ci,
+				    struct buffer_head *bh);
+void ocfs2_init_refcount_extent_tree(struct ocfs2_extent_tree *et,
+				     struct ocfs2_caching_info *ci,
+				     struct buffer_head *bh);
+
+/*
+ * Read an extent block into *bh.  If *bh is NULL, a bh will be
+ * allocated.  This is a cached read.  The extent block will be validated
+ * with ocfs2_validate_extent_block().
+ */
+int ocfs2_read_extent_block(struct ocfs2_caching_info *ci, u64 eb_blkno,
+			    struct buffer_head **bh);
+
+struct ocfs2_alloc_context;
+int ocfs2_insert_extent(handle_t *handle,
+			struct ocfs2_extent_tree *et,
+			u32 cpos,
+			u64 start_blk,
+			u32 new_clusters,
+			u8 flags,
+			struct ocfs2_alloc_context *meta_ac);
+
+enum ocfs2_alloc_restarted {
+	RESTART_NONE = 0,
+	RESTART_TRANS,
+	RESTART_META
+};
+int ocfs2_add_clusters_in_btree(handle_t *handle,
+				struct ocfs2_extent_tree *et,
+				u32 *logical_offset,
+				u32 clusters_to_add,
+				int mark_unwritten,
+				struct ocfs2_alloc_context *data_ac,
+				struct ocfs2_alloc_context *meta_ac,
+				enum ocfs2_alloc_restarted *reason_ret);
+struct ocfs2_cached_dealloc_ctxt;
+struct ocfs2_path;
+int ocfs2_split_extent(handle_t *handle,
+		       struct ocfs2_extent_tree *et,
+		       struct ocfs2_path *path,
+		       int split_index,
+		       struct ocfs2_extent_rec *split_rec,
+		       struct ocfs2_alloc_context *meta_ac,
+		       struct ocfs2_cached_dealloc_ctxt *dealloc);
+int ocfs2_mark_extent_written(struct inode *inode,
+			      struct ocfs2_extent_tree *et,
+			      handle_t *handle, u32 cpos, u32 len, u32 phys,
+			      struct ocfs2_alloc_context *meta_ac,
+			      struct ocfs2_cached_dealloc_ctxt *dealloc);
+int ocfs2_change_extent_flag(handle_t *handle,
+			     struct ocfs2_extent_tree *et,
+			     u32 cpos, u32 len, u32 phys,
+			     struct ocfs2_alloc_context *meta_ac,
+			     struct ocfs2_cached_dealloc_ctxt *dealloc,
+			     int new_flags, int clear_flags);
+int ocfs2_remove_extent(handle_t *handle, struct ocfs2_extent_tree *et,
+			u32 cpos, u32 len,
+			struct ocfs2_alloc_context *meta_ac,
+			struct ocfs2_cached_dealloc_ctxt *dealloc);
+int ocfs2_remove_btree_range(struct inode *inode,
+			     struct ocfs2_extent_tree *et,
+			     u32 cpos, u32 phys_cpos, u32 len, int flags,
+			     struct ocfs2_cached_dealloc_ctxt *dealloc,
+			     u64 refcount_loc);
+
+int ocfs2_num_free_extents(struct ocfs2_super *osb,
+			   struct ocfs2_extent_tree *et);
+
+/*
+ * how many new metadata chunks would an allocation need at maximum?
+ *
+ * Please note that the caller must make sure that root_el is the root
+ * of extent tree. So for an inode, it should be &fe->id2.i_list. Otherwise
+ * the result may be wrong.
+ */
+static inline int ocfs2_extend_meta_needed(struct ocfs2_extent_list *root_el)
+{
+	/*
+	 * Rather than do all the work of determining how much we need
+	 * (involves a ton of reads and locks), just ask for the
+	 * maximal limit.  That's a tree depth shift.  So, one block for
+	 * level of the tree (current l_tree_depth), one block for the
+	 * new tree_depth==0 extent_block, and one block at the new
+	 * top-of-the tree.
+	 */
+	return le16_to_cpu(root_el->l_tree_depth) + 2;
+}
+
+void ocfs2_dinode_new_extent_list(struct inode *inode, struct ocfs2_dinode *di);
+void ocfs2_set_inode_data_inline(struct inode *inode, struct ocfs2_dinode *di);
+int ocfs2_convert_inline_data_to_extents(struct inode *inode,
+					 struct buffer_head *di_bh);
+
+int ocfs2_truncate_log_init(struct ocfs2_super *osb);
+void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb);
+void ocfs2_schedule_truncate_log_flush(struct ocfs2_super *osb,
+				       int cancel);
+int ocfs2_flush_truncate_log(struct ocfs2_super *osb);
+int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb,
+				      int slot_num,
+				      struct ocfs2_dinode **tl_copy);
+int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb,
+					 struct ocfs2_dinode *tl_copy);
+int ocfs2_truncate_log_needs_flush(struct ocfs2_super *osb);
+int ocfs2_truncate_log_append(struct ocfs2_super *osb,
+			      handle_t *handle,
+			      u64 start_blk,
+			      unsigned int num_clusters);
+int __ocfs2_flush_truncate_log(struct ocfs2_super *osb);
+
+/*
+ * Process local structure which describes the block unlinks done
+ * during an operation. This is populated via
+ * ocfs2_cache_block_dealloc().
+ *
+ * ocfs2_run_deallocs() should be called after the potentially
+ * de-allocating routines. No journal handles should be open, and most
+ * locks should have been dropped.
+ */
+struct ocfs2_cached_dealloc_ctxt {
+	struct ocfs2_per_slot_free_list		*c_first_suballocator;
+	struct ocfs2_cached_block_free 		*c_global_allocator;
+};
+static inline void ocfs2_init_dealloc_ctxt(struct ocfs2_cached_dealloc_ctxt *c)
+{
+	c->c_first_suballocator = NULL;
+	c->c_global_allocator = NULL;
+}
+int ocfs2_cache_cluster_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
+				u64 blkno, unsigned int bit);
+int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
+			      int type, int slot, u64 suballoc, u64 blkno,
+			      unsigned int bit);
+static inline int ocfs2_dealloc_has_cluster(struct ocfs2_cached_dealloc_ctxt *c)
+{
+	return c->c_global_allocator != NULL;
+}
+int ocfs2_run_deallocs(struct ocfs2_super *osb,
+		       struct ocfs2_cached_dealloc_ctxt *ctxt);
+
+struct ocfs2_truncate_context {
+	struct ocfs2_cached_dealloc_ctxt tc_dealloc;
+	int tc_ext_alloc_locked; /* is it cluster locked? */
+	/* these get destroyed once it's passed to ocfs2_commit_truncate. */
+	struct buffer_head *tc_last_eb_bh;
+};
+
+int ocfs2_zero_range_for_truncate(struct inode *inode, handle_t *handle,
+				  u64 range_start, u64 range_end);
+int ocfs2_commit_truncate(struct ocfs2_super *osb,
+			  struct inode *inode,
+			  struct buffer_head *di_bh);
+int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
+			  unsigned int start, unsigned int end, int trunc);
+
+int ocfs2_find_leaf(struct ocfs2_caching_info *ci,
+		    struct ocfs2_extent_list *root_el, u32 cpos,
+		    struct buffer_head **leaf_bh);
+int ocfs2_search_extent_list(struct ocfs2_extent_list *el, u32 v_cluster);
+
+int ocfs2_trim_fs(struct super_block *sb, struct fstrim_range *range);
+/*
+ * Helper function to look at the # of clusters in an extent record.
+ */
+static inline unsigned int ocfs2_rec_clusters(struct ocfs2_extent_list *el,
+					      struct ocfs2_extent_rec *rec)
+{
+	/*
+	 * Cluster count in extent records is slightly different
+	 * between interior nodes and leaf nodes. This is to support
+	 * unwritten extents which need a flags field in leaf node
+	 * records, thus shrinking the available space for a clusters
+	 * field.
+	 */
+	if (el->l_tree_depth)
+		return le32_to_cpu(rec->e_int_clusters);
+	else
+		return le16_to_cpu(rec->e_leaf_clusters);
+}
+
+/*
+ * This is only valid for leaf nodes, which are the only ones that can
+ * have empty extents anyway.
+ */
+static inline int ocfs2_is_empty_extent(struct ocfs2_extent_rec *rec)
+{
+	return !rec->e_leaf_clusters;
+}
+
+int ocfs2_grab_pages(struct inode *inode, loff_t start, loff_t end,
+		     struct page **pages, int *num);
+void ocfs2_map_and_dirty_page(struct inode *inode, handle_t *handle,
+			      unsigned int from, unsigned int to,
+			      struct page *page, int zero, u64 *phys);
+/*
+ * Structures which describe a path through a btree, and functions to
+ * manipulate them.
+ *
+ * The idea here is to be as generic as possible with the tree
+ * manipulation code.
+ */
+struct ocfs2_path_item {
+	struct buffer_head		*bh;
+	struct ocfs2_extent_list	*el;
+};
+
+#define OCFS2_MAX_PATH_DEPTH	5
+
+struct ocfs2_path {
+	int				p_tree_depth;
+	ocfs2_journal_access_func	p_root_access;
+	struct ocfs2_path_item		p_node[OCFS2_MAX_PATH_DEPTH];
+};
+
+#define path_root_bh(_path) ((_path)->p_node[0].bh)
+#define path_root_el(_path) ((_path)->p_node[0].el)
+#define path_root_access(_path)((_path)->p_root_access)
+#define path_leaf_bh(_path) ((_path)->p_node[(_path)->p_tree_depth].bh)
+#define path_leaf_el(_path) ((_path)->p_node[(_path)->p_tree_depth].el)
+#define path_num_items(_path) ((_path)->p_tree_depth + 1)
+
+void ocfs2_reinit_path(struct ocfs2_path *path, int keep_root);
+void ocfs2_free_path(struct ocfs2_path *path);
+int ocfs2_find_path(struct ocfs2_caching_info *ci,
+		    struct ocfs2_path *path,
+		    u32 cpos);
+struct ocfs2_path *ocfs2_new_path_from_path(struct ocfs2_path *path);
+struct ocfs2_path *ocfs2_new_path_from_et(struct ocfs2_extent_tree *et);
+int ocfs2_path_bh_journal_access(handle_t *handle,
+				 struct ocfs2_caching_info *ci,
+				 struct ocfs2_path *path,
+				 int idx);
+int ocfs2_journal_access_path(struct ocfs2_caching_info *ci,
+			      handle_t *handle,
+			      struct ocfs2_path *path);
+int ocfs2_find_cpos_for_right_leaf(struct super_block *sb,
+				   struct ocfs2_path *path, u32 *cpos);
+int ocfs2_find_cpos_for_left_leaf(struct super_block *sb,
+				  struct ocfs2_path *path, u32 *cpos);
+int ocfs2_find_subtree_root(struct ocfs2_extent_tree *et,
+			    struct ocfs2_path *left,
+			    struct ocfs2_path *right);
+#endif /* OCFS2_ALLOC_H */

fs/ocfs2/aops.h

@@ -0,0 +1,105 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2002, 2004, 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_AOPS_H
+#define OCFS2_AOPS_H
+
+#include <linux/aio.h>
+
+handle_t *ocfs2_start_walk_page_trans(struct inode *inode,
+							 struct page *page,
+							 unsigned from,
+							 unsigned to);
+
+int ocfs2_map_page_blocks(struct page *page, u64 *p_blkno,
+			  struct inode *inode, unsigned int from,
+			  unsigned int to, int new);
+
+void ocfs2_unlock_and_free_pages(struct page **pages, int num_pages);
+
+int walk_page_buffers(	handle_t *handle,
+			struct buffer_head *head,
+			unsigned from,
+			unsigned to,
+			int *partial,
+			int (*fn)(	handle_t *handle,
+					struct buffer_head *bh));
+
+int ocfs2_write_end_nolock(struct address_space *mapping,
+			   loff_t pos, unsigned len, unsigned copied,
+			   struct page *page, void *fsdata);
+
+int ocfs2_write_begin_nolock(struct file *filp,
+			     struct address_space *mapping,
+			     loff_t pos, unsigned len, unsigned flags,
+			     struct page **pagep, void **fsdata,
+			     struct buffer_head *di_bh, struct page *mmap_page);
+
+int ocfs2_read_inline_data(struct inode *inode, struct page *page,
+			   struct buffer_head *di_bh);
+int ocfs2_size_fits_inline_data(struct buffer_head *di_bh, u64 new_size);
+
+int ocfs2_get_block(struct inode *inode, sector_t iblock,
+		    struct buffer_head *bh_result, int create);
+/* all ocfs2_dio_end_io()'s fault */
+#define ocfs2_iocb_is_rw_locked(iocb) \
+	test_bit(0, (unsigned long *)&iocb->private)
+static inline void ocfs2_iocb_set_rw_locked(struct kiocb *iocb, int level)
+{
+	set_bit(0, (unsigned long *)&iocb->private);
+	if (level)
+		set_bit(1, (unsigned long *)&iocb->private);
+	else
+		clear_bit(1, (unsigned long *)&iocb->private);
+}
+
+/*
+ * Using a named enum representing lock types in terms of #N bit stored in
+ * iocb->private, which is going to be used for communication between
+ * ocfs2_dio_end_io() and ocfs2_file_aio_write/read().
+ */
+enum ocfs2_iocb_lock_bits {
+	OCFS2_IOCB_RW_LOCK = 0,
+	OCFS2_IOCB_RW_LOCK_LEVEL,
+	OCFS2_IOCB_SEM,
+	OCFS2_IOCB_UNALIGNED_IO,
+	OCFS2_IOCB_NUM_LOCKS
+};
+
+#define ocfs2_iocb_clear_rw_locked(iocb) \
+	clear_bit(OCFS2_IOCB_RW_LOCK, (unsigned long *)&iocb->private)
+#define ocfs2_iocb_rw_locked_level(iocb) \
+	test_bit(OCFS2_IOCB_RW_LOCK_LEVEL, (unsigned long *)&iocb->private)
+#define ocfs2_iocb_set_sem_locked(iocb) \
+	set_bit(OCFS2_IOCB_SEM, (unsigned long *)&iocb->private)
+#define ocfs2_iocb_clear_sem_locked(iocb) \
+	clear_bit(OCFS2_IOCB_SEM, (unsigned long *)&iocb->private)
+#define ocfs2_iocb_is_sem_locked(iocb) \
+	test_bit(OCFS2_IOCB_SEM, (unsigned long *)&iocb->private)
+
+#define ocfs2_iocb_set_unaligned_aio(iocb) \
+	set_bit(OCFS2_IOCB_UNALIGNED_IO, (unsigned long *)&iocb->private)
+#define ocfs2_iocb_clear_unaligned_aio(iocb) \
+	clear_bit(OCFS2_IOCB_UNALIGNED_IO, (unsigned long *)&iocb->private)
+#define ocfs2_iocb_is_unaligned_aio(iocb) \
+	test_bit(OCFS2_IOCB_UNALIGNED_IO, (unsigned long *)&iocb->private)
+
+#endif /* OCFS2_FILE_H */

fs/ocfs2/blockcheck.c

@@ -0,0 +1,647 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * blockcheck.c
+ *
+ * Checksum and ECC codes for the OCFS2 userspace library.
+ *
+ * Copyright (C) 2006, 2008 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License, version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/crc32.h>
+#include <linux/buffer_head.h>
+#include <linux/bitops.h>
+#include <linux/debugfs.h>
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <asm/byteorder.h>
+
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "blockcheck.h"
+
+
+/*
+ * We use the following conventions:
+ *
+ * d = # data bits
+ * p = # parity bits
+ * c = # total code bits (d + p)
+ */
+
+
+/*
+ * Calculate the bit offset in the hamming code buffer based on the bit's
+ * offset in the data buffer.  Since the hamming code reserves all
+ * power-of-two bits for parity, the data bit number and the code bit
+ * number are offset by all the parity bits beforehand.
+ *
+ * Recall that bit numbers in hamming code are 1-based.  This function
+ * takes the 0-based data bit from the caller.
+ *
+ * An example.  Take bit 1 of the data buffer.  1 is a power of two (2^0),
+ * so it's a parity bit.  2 is a power of two (2^1), so it's a parity bit.
+ * 3 is not a power of two.  So bit 1 of the data buffer ends up as bit 3
+ * in the code buffer.
+ *
+ * The caller can pass in *p if it wants to keep track of the most recent
+ * number of parity bits added.  This allows the function to start the
+ * calculation at the last place.
+ */
+static unsigned int calc_code_bit(unsigned int i, unsigned int *p_cache)
+{
+	unsigned int b, p = 0;
+
+	/*
+	 * Data bits are 0-based, but we're talking code bits, which
+	 * are 1-based.
+	 */
+	b = i + 1;
+
+	/* Use the cache if it is there */
+	if (p_cache)
+		p = *p_cache;
+        b += p;
+
+	/*
+	 * For every power of two below our bit number, bump our bit.
+	 *
+	 * We compare with (b + 1) because we have to compare with what b
+	 * would be _if_ it were bumped up by the parity bit.  Capice?
+	 *
+	 * p is set above.
+	 */
+	for (; (1 << p) < (b + 1); p++)
+		b++;
+
+	if (p_cache)
+		*p_cache = p;
+
+	return b;
+}
+
+/*
+ * This is the low level encoder function.  It can be called across
+ * multiple hunks just like the crc32 code.  'd' is the number of bits
+ * _in_this_hunk_.  nr is the bit offset of this hunk.  So, if you had
+ * two 512B buffers, you would do it like so:
+ *
+ * parity = ocfs2_hamming_encode(0, buf1, 512 * 8, 0);
+ * parity = ocfs2_hamming_encode(parity, buf2, 512 * 8, 512 * 8);
+ *
+ * If you just have one buffer, use ocfs2_hamming_encode_block().
+ */
+u32 ocfs2_hamming_encode(u32 parity, void *data, unsigned int d, unsigned int nr)
+{
+	unsigned int i, b, p = 0;
+
+	BUG_ON(!d);
+
+	/*
+	 * b is the hamming code bit number.  Hamming code specifies a
+	 * 1-based array, but C uses 0-based.  So 'i' is for C, and 'b' is
+	 * for the algorithm.
+	 *
+	 * The i++ in the for loop is so that the start offset passed
+	 * to ocfs2_find_next_bit_set() is one greater than the previously
+	 * found bit.
+	 */
+	for (i = 0; (i = ocfs2_find_next_bit(data, d, i)) < d; i++)
+	{
+		/*
+		 * i is the offset in this hunk, nr + i is the total bit
+		 * offset.
+		 */
+		b = calc_code_bit(nr + i, &p);
+
+		/*
+		 * Data bits in the resultant code are checked by
+		 * parity bits that are part of the bit number
+		 * representation.  Huh?
+		 *
+		 * <wikipedia href="http://en.wikipedia.org/wiki/Hamming_code">
+		 * In other words, the parity bit at position 2^k
+		 * checks bits in positions having bit k set in
+		 * their binary representation.  Conversely, for
+		 * instance, bit 13, i.e. 1101(2), is checked by
+		 * bits 1000(2) = 8, 0100(2)=4 and 0001(2) = 1.
+		 * </wikipedia>
+		 *
+		 * Note that 'k' is the _code_ bit number.  'b' in
+		 * our loop.
+		 */
+		parity ^= b;
+	}
+
+	/* While the data buffer was treated as little endian, the
+	 * return value is in host endian. */
+	return parity;
+}
+
+u32 ocfs2_hamming_encode_block(void *data, unsigned int blocksize)
+{
+	return ocfs2_hamming_encode(0, data, blocksize * 8, 0);
+}
+
+/*
+ * Like ocfs2_hamming_encode(), this can handle hunks.  nr is the bit
+ * offset of the current hunk.  If bit to be fixed is not part of the
+ * current hunk, this does nothing.
+ *
+ * If you only have one hunk, use ocfs2_hamming_fix_block().
+ */
+void ocfs2_hamming_fix(void *data, unsigned int d, unsigned int nr,
+		       unsigned int fix)
+{
+	unsigned int i, b;
+
+	BUG_ON(!d);
+
+	/*
+	 * If the bit to fix has an hweight of 1, it's a parity bit.  One
+	 * busted parity bit is its own error.  Nothing to do here.
+	 */
+	if (hweight32(fix) == 1)
+		return;
+
+	/*
+	 * nr + d is the bit right past the data hunk we're looking at.
+	 * If fix after that, nothing to do
+	 */
+	if (fix >= calc_code_bit(nr + d, NULL))
+		return;
+
+	/*
+	 * nr is the offset in the data hunk we're starting at.  Let's
+	 * start b at the offset in the code buffer.  See hamming_encode()
+	 * for a more detailed description of 'b'.
+	 */
+	b = calc_code_bit(nr, NULL);
+	/* If the fix is before this hunk, nothing to do */
+	if (fix < b)
+		return;
+
+	for (i = 0; i < d; i++, b++)
+	{
+		/* Skip past parity bits */
+		while (hweight32(b) == 1)
+			b++;
+
+		/*
+		 * i is the offset in this data hunk.
+		 * nr + i is the offset in the total data buffer.
+		 * b is the offset in the total code buffer.
+		 *
+		 * Thus, when b == fix, bit i in the current hunk needs
+		 * fixing.
+		 */
+		if (b == fix)
+		{
+			if (ocfs2_test_bit(i, data))
+				ocfs2_clear_bit(i, data);
+			else
+				ocfs2_set_bit(i, data);
+			break;
+		}
+	}
+}
+
+void ocfs2_hamming_fix_block(void *data, unsigned int blocksize,
+			     unsigned int fix)
+{
+	ocfs2_hamming_fix(data, blocksize * 8, 0, fix);
+}
+
+
+/*
+ * Debugfs handling.
+ */
+
+#ifdef CONFIG_DEBUG_FS
+
+static int blockcheck_u64_get(void *data, u64 *val)
+{
+	*val = *(u64 *)data;
+	return 0;
+}
+DEFINE_SIMPLE_ATTRIBUTE(blockcheck_fops, blockcheck_u64_get, NULL, "%llu\n");
+
+static struct dentry *blockcheck_debugfs_create(const char *name,
+						struct dentry *parent,
+						u64 *value)
+{
+	return debugfs_create_file(name, S_IFREG | S_IRUSR, parent, value,
+				   &blockcheck_fops);
+}
+
+static void ocfs2_blockcheck_debug_remove(struct ocfs2_blockcheck_stats *stats)
+{
+	if (stats) {
+		debugfs_remove(stats->b_debug_check);
+		stats->b_debug_check = NULL;
+		debugfs_remove(stats->b_debug_failure);
+		stats->b_debug_failure = NULL;
+		debugfs_remove(stats->b_debug_recover);
+		stats->b_debug_recover = NULL;
+		debugfs_remove(stats->b_debug_dir);
+		stats->b_debug_dir = NULL;
+	}
+}
+
+static int ocfs2_blockcheck_debug_install(struct ocfs2_blockcheck_stats *stats,
+					  struct dentry *parent)
+{
+	int rc = -EINVAL;
+
+	if (!stats)
+		goto out;
+
+	stats->b_debug_dir = debugfs_create_dir("blockcheck", parent);
+	if (!stats->b_debug_dir)
+		goto out;
+
+	stats->b_debug_check =
+		blockcheck_debugfs_create("blocks_checked",
+					  stats->b_debug_dir,
+					  &stats->b_check_count);
+
+	stats->b_debug_failure =
+		blockcheck_debugfs_create("checksums_failed",
+					  stats->b_debug_dir,
+					  &stats->b_failure_count);
+
+	stats->b_debug_recover =
+		blockcheck_debugfs_create("ecc_recoveries",
+					  stats->b_debug_dir,
+					  &stats->b_recover_count);
+	if (stats->b_debug_check && stats->b_debug_failure &&
+	    stats->b_debug_recover)
+		rc = 0;
+
+out:
+	if (rc)
+		ocfs2_blockcheck_debug_remove(stats);
+	return rc;
+}
+#else
+static inline int ocfs2_blockcheck_debug_install(struct ocfs2_blockcheck_stats *stats,
+						 struct dentry *parent)
+{
+	return 0;
+}
+
+static inline void ocfs2_blockcheck_debug_remove(struct ocfs2_blockcheck_stats *stats)
+{
+}
+#endif  /* CONFIG_DEBUG_FS */
+
+/* Always-called wrappers for starting and stopping the debugfs files */
+int ocfs2_blockcheck_stats_debugfs_install(struct ocfs2_blockcheck_stats *stats,
+					   struct dentry *parent)
+{
+	return ocfs2_blockcheck_debug_install(stats, parent);
+}
+
+void ocfs2_blockcheck_stats_debugfs_remove(struct ocfs2_blockcheck_stats *stats)
+{
+	ocfs2_blockcheck_debug_remove(stats);
+}
+
+static void ocfs2_blockcheck_inc_check(struct ocfs2_blockcheck_stats *stats)
+{
+	u64 new_count;
+
+	if (!stats)
+		return;
+
+	spin_lock(&stats->b_lock);
+	stats->b_check_count++;
+	new_count = stats->b_check_count;
+	spin_unlock(&stats->b_lock);
+
+	if (!new_count)
+		mlog(ML_NOTICE, "Block check count has wrapped\n");
+}
+
+static void ocfs2_blockcheck_inc_failure(struct ocfs2_blockcheck_stats *stats)
+{
+	u64 new_count;
+
+	if (!stats)
+		return;
+
+	spin_lock(&stats->b_lock);
+	stats->b_failure_count++;
+	new_count = stats->b_failure_count;
+	spin_unlock(&stats->b_lock);
+
+	if (!new_count)
+		mlog(ML_NOTICE, "Checksum failure count has wrapped\n");
+}
+
+static void ocfs2_blockcheck_inc_recover(struct ocfs2_blockcheck_stats *stats)
+{
+	u64 new_count;
+
+	if (!stats)
+		return;
+
+	spin_lock(&stats->b_lock);
+	stats->b_recover_count++;
+	new_count = stats->b_recover_count;
+	spin_unlock(&stats->b_lock);
+
+	if (!new_count)
+		mlog(ML_NOTICE, "ECC recovery count has wrapped\n");
+}
+
+
+
+/*
+ * These are the low-level APIs for using the ocfs2_block_check structure.
+ */
+
+/*
+ * This function generates check information for a block.
+ * data is the block to be checked.  bc is a pointer to the
+ * ocfs2_block_check structure describing the crc32 and the ecc.
+ *
+ * bc should be a pointer inside data, as the function will
+ * take care of zeroing it before calculating the check information.  If
+ * bc does not point inside data, the caller must make sure any inline
+ * ocfs2_block_check structures are zeroed.
+ *
+ * The data buffer must be in on-disk endian (little endian for ocfs2).
+ * bc will be filled with little-endian values and will be ready to go to
+ * disk.
+ */
+void ocfs2_block_check_compute(void *data, size_t blocksize,
+			       struct ocfs2_block_check *bc)
+{
+	u32 crc;
+	u32 ecc;
+
+	memset(bc, 0, sizeof(struct ocfs2_block_check));
+
+	crc = crc32_le(~0, data, blocksize);
+	ecc = ocfs2_hamming_encode_block(data, blocksize);
+
+	/*
+	 * No ecc'd ocfs2 structure is larger than 4K, so ecc will be no
+	 * larger than 16 bits.
+	 */
+	BUG_ON(ecc > USHRT_MAX);
+
+	bc->bc_crc32e = cpu_to_le32(crc);
+	bc->bc_ecc = cpu_to_le16((u16)ecc);
+}
+
+/*
+ * This function validates existing check information.  Like _compute,
+ * the function will take care of zeroing bc before calculating check codes.
+ * If bc is not a pointer inside data, the caller must have zeroed any
+ * inline ocfs2_block_check structures.
+ *
+ * Again, the data passed in should be the on-disk endian.
+ */
+int ocfs2_block_check_validate(void *data, size_t blocksize,
+			       struct ocfs2_block_check *bc,
+			       struct ocfs2_blockcheck_stats *stats)
+{
+	int rc = 0;
+	u32 bc_crc32e;
+	u16 bc_ecc;
+	u32 crc, ecc;
+
+	ocfs2_blockcheck_inc_check(stats);
+
+	bc_crc32e = le32_to_cpu(bc->bc_crc32e);
+	bc_ecc = le16_to_cpu(bc->bc_ecc);
+
+	memset(bc, 0, sizeof(struct ocfs2_block_check));
+
+	/* Fast path - if the crc32 validates, we're good to go */
+	crc = crc32_le(~0, data, blocksize);
+	if (crc == bc_crc32e)
+		goto out;
+
+	ocfs2_blockcheck_inc_failure(stats);
+	mlog(ML_ERROR,
+	     "CRC32 failed: stored: 0x%x, computed 0x%x. Applying ECC.\n",
+	     (unsigned int)bc_crc32e, (unsigned int)crc);
+
+	/* Ok, try ECC fixups */
+	ecc = ocfs2_hamming_encode_block(data, blocksize);
+	ocfs2_hamming_fix_block(data, blocksize, ecc ^ bc_ecc);
+
+	/* And check the crc32 again */
+	crc = crc32_le(~0, data, blocksize);
+	if (crc == bc_crc32e) {
+		ocfs2_blockcheck_inc_recover(stats);
+		goto out;
+	}
+
+	mlog(ML_ERROR, "Fixed CRC32 failed: stored: 0x%x, computed 0x%x\n",
+	     (unsigned int)bc_crc32e, (unsigned int)crc);
+
+	rc = -EIO;
+
+out:
+	bc->bc_crc32e = cpu_to_le32(bc_crc32e);
+	bc->bc_ecc = cpu_to_le16(bc_ecc);
+
+	return rc;
+}
+
+/*
+ * This function generates check information for a list of buffer_heads.
+ * bhs is the blocks to be checked.  bc is a pointer to the
+ * ocfs2_block_check structure describing the crc32 and the ecc.
+ *
+ * bc should be a pointer inside data, as the function will
+ * take care of zeroing it before calculating the check information.  If
+ * bc does not point inside data, the caller must make sure any inline
+ * ocfs2_block_check structures are zeroed.
+ *
+ * The data buffer must be in on-disk endian (little endian for ocfs2).
+ * bc will be filled with little-endian values and will be ready to go to
+ * disk.
+ */
+void ocfs2_block_check_compute_bhs(struct buffer_head **bhs, int nr,
+				   struct ocfs2_block_check *bc)
+{
+	int i;
+	u32 crc, ecc;
+
+	BUG_ON(nr < 0);
+
+	if (!nr)
+		return;
+
+	memset(bc, 0, sizeof(struct ocfs2_block_check));
+
+	for (i = 0, crc = ~0, ecc = 0; i < nr; i++) {
+		crc = crc32_le(crc, bhs[i]->b_data, bhs[i]->b_size);
+		/*
+		 * The number of bits in a buffer is obviously b_size*8.
+		 * The offset of this buffer is b_size*i, so the bit offset
+		 * of this buffer is b_size*8*i.
+		 */
+		ecc = (u16)ocfs2_hamming_encode(ecc, bhs[i]->b_data,
+						bhs[i]->b_size * 8,
+						bhs[i]->b_size * 8 * i);
+	}
+
+	/*
+	 * No ecc'd ocfs2 structure is larger than 4K, so ecc will be no
+	 * larger than 16 bits.
+	 */
+	BUG_ON(ecc > USHRT_MAX);
+
+	bc->bc_crc32e = cpu_to_le32(crc);
+	bc->bc_ecc = cpu_to_le16((u16)ecc);
+}
+
+/*
+ * This function validates existing check information on a list of
+ * buffer_heads.  Like _compute_bhs, the function will take care of
+ * zeroing bc before calculating check codes.  If bc is not a pointer
+ * inside data, the caller must have zeroed any inline
+ * ocfs2_block_check structures.
+ *
+ * Again, the data passed in should be the on-disk endian.
+ */
+int ocfs2_block_check_validate_bhs(struct buffer_head **bhs, int nr,
+				   struct ocfs2_block_check *bc,
+				   struct ocfs2_blockcheck_stats *stats)
+{
+	int i, rc = 0;
+	u32 bc_crc32e;
+	u16 bc_ecc;
+	u32 crc, ecc, fix;
+
+	BUG_ON(nr < 0);
+
+	if (!nr)
+		return 0;
+
+	ocfs2_blockcheck_inc_check(stats);
+
+	bc_crc32e = le32_to_cpu(bc->bc_crc32e);
+	bc_ecc = le16_to_cpu(bc->bc_ecc);
+
+	memset(bc, 0, sizeof(struct ocfs2_block_check));
+
+	/* Fast path - if the crc32 validates, we're good to go */
+	for (i = 0, crc = ~0; i < nr; i++)
+		crc = crc32_le(crc, bhs[i]->b_data, bhs[i]->b_size);
+	if (crc == bc_crc32e)
+		goto out;
+
+	ocfs2_blockcheck_inc_failure(stats);
+	mlog(ML_ERROR,
+	     "CRC32 failed: stored: %u, computed %u.  Applying ECC.\n",
+	     (unsigned int)bc_crc32e, (unsigned int)crc);
+
+	/* Ok, try ECC fixups */
+	for (i = 0, ecc = 0; i < nr; i++) {
+		/*
+		 * The number of bits in a buffer is obviously b_size*8.
+		 * The offset of this buffer is b_size*i, so the bit offset
+		 * of this buffer is b_size*8*i.
+		 */
+		ecc = (u16)ocfs2_hamming_encode(ecc, bhs[i]->b_data,
+						bhs[i]->b_size * 8,
+						bhs[i]->b_size * 8 * i);
+	}
+	fix = ecc ^ bc_ecc;
+	for (i = 0; i < nr; i++) {
+		/*
+		 * Try the fix against each buffer.  It will only affect
+		 * one of them.
+		 */
+		ocfs2_hamming_fix(bhs[i]->b_data, bhs[i]->b_size * 8,
+				  bhs[i]->b_size * 8 * i, fix);
+	}
+
+	/* And check the crc32 again */
+	for (i = 0, crc = ~0; i < nr; i++)
+		crc = crc32_le(crc, bhs[i]->b_data, bhs[i]->b_size);
+	if (crc == bc_crc32e) {
+		ocfs2_blockcheck_inc_recover(stats);
+		goto out;
+	}
+
+	mlog(ML_ERROR, "Fixed CRC32 failed: stored: %u, computed %u\n",
+	     (unsigned int)bc_crc32e, (unsigned int)crc);
+
+	rc = -EIO;
+
+out:
+	bc->bc_crc32e = cpu_to_le32(bc_crc32e);
+	bc->bc_ecc = cpu_to_le16(bc_ecc);
+
+	return rc;
+}
+
+/*
+ * These are the main API.  They check the superblock flag before
+ * calling the underlying operations.
+ *
+ * They expect the buffer(s) to be in disk format.
+ */
+void ocfs2_compute_meta_ecc(struct super_block *sb, void *data,
+			    struct ocfs2_block_check *bc)
+{
+	if (ocfs2_meta_ecc(OCFS2_SB(sb)))
+		ocfs2_block_check_compute(data, sb->s_blocksize, bc);
+}
+
+int ocfs2_validate_meta_ecc(struct super_block *sb, void *data,
+			    struct ocfs2_block_check *bc)
+{
+	int rc = 0;
+	struct ocfs2_super *osb = OCFS2_SB(sb);
+
+	if (ocfs2_meta_ecc(osb))
+		rc = ocfs2_block_check_validate(data, sb->s_blocksize, bc,
+						&osb->osb_ecc_stats);
+
+	return rc;
+}
+
+void ocfs2_compute_meta_ecc_bhs(struct super_block *sb,
+				struct buffer_head **bhs, int nr,
+				struct ocfs2_block_check *bc)
+{
+	if (ocfs2_meta_ecc(OCFS2_SB(sb)))
+		ocfs2_block_check_compute_bhs(bhs, nr, bc);
+}
+
+int ocfs2_validate_meta_ecc_bhs(struct super_block *sb,
+				struct buffer_head **bhs, int nr,
+				struct ocfs2_block_check *bc)
+{
+	int rc = 0;
+	struct ocfs2_super *osb = OCFS2_SB(sb);
+
+	if (ocfs2_meta_ecc(osb))
+		rc = ocfs2_block_check_validate_bhs(bhs, nr, bc,
+						    &osb->osb_ecc_stats);
+
+	return rc;
+}
+

fs/ocfs2/blockcheck.h

@@ -0,0 +1,107 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * blockcheck.h
+ *
+ * Checksum and ECC codes for the OCFS2 userspace library.
+ *
+ * Copyright (C) 2004, 2008 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License, version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#ifndef OCFS2_BLOCKCHECK_H
+#define OCFS2_BLOCKCHECK_H
+
+
+/* Count errors and error correction from blockcheck.c */
+struct ocfs2_blockcheck_stats {
+	spinlock_t b_lock;
+	u64 b_check_count;	/* Number of blocks we've checked */
+	u64 b_failure_count;	/* Number of failed checksums */
+	u64 b_recover_count;	/* Number of blocks fixed by ecc */
+
+	/*
+	 * debugfs entries, used if this is passed to
+	 * ocfs2_blockcheck_stats_debugfs_install()
+	 */
+	struct dentry *b_debug_dir;	/* Parent of the debugfs  files */
+	struct dentry *b_debug_check;	/* Exposes b_check_count */
+	struct dentry *b_debug_failure;	/* Exposes b_failure_count */
+	struct dentry *b_debug_recover;	/* Exposes b_recover_count */
+};
+
+
+/* High level block API */
+void ocfs2_compute_meta_ecc(struct super_block *sb, void *data,
+			    struct ocfs2_block_check *bc);
+int ocfs2_validate_meta_ecc(struct super_block *sb, void *data,
+			    struct ocfs2_block_check *bc);
+void ocfs2_compute_meta_ecc_bhs(struct super_block *sb,
+				struct buffer_head **bhs, int nr,
+				struct ocfs2_block_check *bc);
+int ocfs2_validate_meta_ecc_bhs(struct super_block *sb,
+				struct buffer_head **bhs, int nr,
+				struct ocfs2_block_check *bc);
+
+/* Lower level API */
+void ocfs2_block_check_compute(void *data, size_t blocksize,
+			       struct ocfs2_block_check *bc);
+int ocfs2_block_check_validate(void *data, size_t blocksize,
+			       struct ocfs2_block_check *bc,
+			       struct ocfs2_blockcheck_stats *stats);
+void ocfs2_block_check_compute_bhs(struct buffer_head **bhs, int nr,
+				   struct ocfs2_block_check *bc);
+int ocfs2_block_check_validate_bhs(struct buffer_head **bhs, int nr,
+				   struct ocfs2_block_check *bc,
+				   struct ocfs2_blockcheck_stats *stats);
+
+/* Debug Initialization */
+int ocfs2_blockcheck_stats_debugfs_install(struct ocfs2_blockcheck_stats *stats,
+					   struct dentry *parent);
+void ocfs2_blockcheck_stats_debugfs_remove(struct ocfs2_blockcheck_stats *stats);
+
+/*
+ * Hamming code functions
+ */
+
+/*
+ * Encoding hamming code parity bits for a buffer.
+ *
+ * This is the low level encoder function.  It can be called across
+ * multiple hunks just like the crc32 code.  'd' is the number of bits
+ * _in_this_hunk_.  nr is the bit offset of this hunk.  So, if you had
+ * two 512B buffers, you would do it like so:
+ *
+ * parity = ocfs2_hamming_encode(0, buf1, 512 * 8, 0);
+ * parity = ocfs2_hamming_encode(parity, buf2, 512 * 8, 512 * 8);
+ *
+ * If you just have one buffer, use ocfs2_hamming_encode_block().
+ */
+u32 ocfs2_hamming_encode(u32 parity, void *data, unsigned int d,
+			 unsigned int nr);
+/*
+ * Fix a buffer with a bit error.  The 'fix' is the original parity
+ * xor'd with the parity calculated now.
+ *
+ * Like ocfs2_hamming_encode(), this can handle hunks.  nr is the bit
+ * offset of the current hunk.  If bit to be fixed is not part of the
+ * current hunk, this does nothing.
+ *
+ * If you only have one buffer, use ocfs2_hamming_fix_block().
+ */
+void ocfs2_hamming_fix(void *data, unsigned int d, unsigned int nr,
+		       unsigned int fix);
+
+/* Convenience wrappers for a single buffer of data */
+extern u32 ocfs2_hamming_encode_block(void *data, unsigned int blocksize);
+extern void ocfs2_hamming_fix_block(void *data, unsigned int blocksize,
+				    unsigned int fix);
+#endif

fs/ocfs2/buffer_head_io.c

@@ -0,0 +1,427 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * io.c
+ *
+ * Buffer cache handling
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/highmem.h>
+
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "inode.h"
+#include "journal.h"
+#include "uptodate.h"
+#include "buffer_head_io.h"
+#include "ocfs2_trace.h"
+
+/*
+ * Bits on bh->b_state used by ocfs2.
+ *
+ * These MUST be after the JBD2 bits.  Hence, we use BH_JBDPrivateStart.
+ */
+enum ocfs2_state_bits {
+	BH_NeedsValidate = BH_JBDPrivateStart,
+};
+
+/* Expand the magic b_state functions */
+BUFFER_FNS(NeedsValidate, needs_validate);
+
+int ocfs2_write_block(struct ocfs2_super *osb, struct buffer_head *bh,
+		      struct ocfs2_caching_info *ci)
+{
+	int ret = 0;
+
+	trace_ocfs2_write_block((unsigned long long)bh->b_blocknr, ci);
+
+	BUG_ON(bh->b_blocknr < OCFS2_SUPER_BLOCK_BLKNO);
+	BUG_ON(buffer_jbd(bh));
+
+	/* No need to check for a soft readonly file system here. non
+	 * journalled writes are only ever done on system files which
+	 * can get modified during recovery even if read-only. */
+	if (ocfs2_is_hard_readonly(osb)) {
+		ret = -EROFS;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	ocfs2_metadata_cache_io_lock(ci);
+
+	lock_buffer(bh);
+	set_buffer_uptodate(bh);
+
+	/* remove from dirty list before I/O. */
+	clear_buffer_dirty(bh);
+
+	get_bh(bh); /* for end_buffer_write_sync() */
+	bh->b_end_io = end_buffer_write_sync;
+	submit_bh(WRITE, bh);
+
+	wait_on_buffer(bh);
+
+	if (buffer_uptodate(bh)) {
+		ocfs2_set_buffer_uptodate(ci, bh);
+	} else {
+		/* We don't need to remove the clustered uptodate
+		 * information for this bh as it's not marked locally
+		 * uptodate. */
+		ret = -EIO;
+		mlog_errno(ret);
+	}
+
+	ocfs2_metadata_cache_io_unlock(ci);
+out:
+	return ret;
+}
+
+int ocfs2_read_blocks_sync(struct ocfs2_super *osb, u64 block,
+			   unsigned int nr, struct buffer_head *bhs[])
+{
+	int status = 0;
+	unsigned int i;
+	struct buffer_head *bh;
+
+	trace_ocfs2_read_blocks_sync((unsigned long long)block, nr);
+
+	if (!nr)
+		goto bail;
+
+	for (i = 0 ; i < nr ; i++) {
+		if (bhs[i] == NULL) {
+			bhs[i] = sb_getblk(osb->sb, block++);
+			if (bhs[i] == NULL) {
+				status = -ENOMEM;
+				mlog_errno(status);
+				goto bail;
+			}
+		}
+		bh = bhs[i];
+
+		if (buffer_jbd(bh)) {
+			trace_ocfs2_read_blocks_sync_jbd(
+					(unsigned long long)bh->b_blocknr);
+			continue;
+		}
+
+		if (buffer_dirty(bh)) {
+			/* This should probably be a BUG, or
+			 * at least return an error. */
+			mlog(ML_ERROR,
+			     "trying to sync read a dirty "
+			     "buffer! (blocknr = %llu), skipping\n",
+			     (unsigned long long)bh->b_blocknr);
+			continue;
+		}
+
+		lock_buffer(bh);
+		if (buffer_jbd(bh)) {
+			mlog(ML_ERROR,
+			     "block %llu had the JBD bit set "
+			     "while I was in lock_buffer!",
+			     (unsigned long long)bh->b_blocknr);
+			BUG();
+		}
+
+		clear_buffer_uptodate(bh);
+		get_bh(bh); /* for end_buffer_read_sync() */
+		bh->b_end_io = end_buffer_read_sync;
+		submit_bh(READ, bh);
+	}
+
+	for (i = nr; i > 0; i--) {
+		bh = bhs[i - 1];
+
+		/* No need to wait on the buffer if it's managed by JBD. */
+		if (!buffer_jbd(bh))
+			wait_on_buffer(bh);
+
+		if (!buffer_uptodate(bh)) {
+			/* Status won't be cleared from here on out,
+			 * so we can safely record this and loop back
+			 * to cleanup the other buffers. */
+			status = -EIO;
+			put_bh(bh);
+			bhs[i - 1] = NULL;
+		}
+	}
+
+bail:
+	return status;
+}
+
+int ocfs2_read_blocks(struct ocfs2_caching_info *ci, u64 block, int nr,
+		      struct buffer_head *bhs[], int flags,
+		      int (*validate)(struct super_block *sb,
+				      struct buffer_head *bh))
+{
+	int status = 0;
+	int i, ignore_cache = 0;
+	struct buffer_head *bh;
+	struct super_block *sb = ocfs2_metadata_cache_get_super(ci);
+
+	trace_ocfs2_read_blocks_begin(ci, (unsigned long long)block, nr, flags);
+
+	BUG_ON(!ci);
+	BUG_ON((flags & OCFS2_BH_READAHEAD) &&
+	       (flags & OCFS2_BH_IGNORE_CACHE));
+
+	if (bhs == NULL) {
+		status = -EINVAL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	if (nr < 0) {
+		mlog(ML_ERROR, "asked to read %d blocks!\n", nr);
+		status = -EINVAL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	if (nr == 0) {
+		status = 0;
+		goto bail;
+	}
+
+	ocfs2_metadata_cache_io_lock(ci);
+	for (i = 0 ; i < nr ; i++) {
+		if (bhs[i] == NULL) {
+			bhs[i] = sb_getblk(sb, block++);
+			if (bhs[i] == NULL) {
+				ocfs2_metadata_cache_io_unlock(ci);
+				status = -ENOMEM;
+				mlog_errno(status);
+				goto bail;
+			}
+		}
+		bh = bhs[i];
+		ignore_cache = (flags & OCFS2_BH_IGNORE_CACHE);
+
+		/* There are three read-ahead cases here which we need to
+		 * be concerned with. All three assume a buffer has
+		 * previously been submitted with OCFS2_BH_READAHEAD
+		 * and it hasn't yet completed I/O.
+		 *
+		 * 1) The current request is sync to disk. This rarely
+		 *    happens these days, and never when performance
+		 *    matters - the code can just wait on the buffer
+		 *    lock and re-submit.
+		 *
+		 * 2) The current request is cached, but not
+		 *    readahead. ocfs2_buffer_uptodate() will return
+		 *    false anyway, so we'll wind up waiting on the
+		 *    buffer lock to do I/O. We re-check the request
+		 *    with after getting the lock to avoid a re-submit.
+		 *
+		 * 3) The current request is readahead (and so must
+		 *    also be a caching one). We short circuit if the
+		 *    buffer is locked (under I/O) and if it's in the
+		 *    uptodate cache. The re-check from #2 catches the
+		 *    case that the previous read-ahead completes just
+		 *    before our is-it-in-flight check.
+		 */
+
+		if (!ignore_cache && !ocfs2_buffer_uptodate(ci, bh)) {
+			trace_ocfs2_read_blocks_from_disk(
+			     (unsigned long long)bh->b_blocknr,
+			     (unsigned long long)ocfs2_metadata_cache_owner(ci));
+			/* We're using ignore_cache here to say
+			 * "go to disk" */
+			ignore_cache = 1;
+		}
+
+		trace_ocfs2_read_blocks_bh((unsigned long long)bh->b_blocknr,
+			ignore_cache, buffer_jbd(bh), buffer_dirty(bh));
+
+		if (buffer_jbd(bh)) {
+			continue;
+		}
+
+		if (ignore_cache) {
+			if (buffer_dirty(bh)) {
+				/* This should probably be a BUG, or
+				 * at least return an error. */
+				continue;
+			}
+
+			/* A read-ahead request was made - if the
+			 * buffer is already under read-ahead from a
+			 * previously submitted request than we are
+			 * done here. */
+			if ((flags & OCFS2_BH_READAHEAD)
+			    && ocfs2_buffer_read_ahead(ci, bh))
+				continue;
+
+			lock_buffer(bh);
+			if (buffer_jbd(bh)) {
+#ifdef CATCH_BH_JBD_RACES
+				mlog(ML_ERROR, "block %llu had the JBD bit set "
+					       "while I was in lock_buffer!",
+				     (unsigned long long)bh->b_blocknr);
+				BUG();
+#else
+				unlock_buffer(bh);
+				continue;
+#endif
+			}
+
+			/* Re-check ocfs2_buffer_uptodate() as a
+			 * previously read-ahead buffer may have
+			 * completed I/O while we were waiting for the
+			 * buffer lock. */
+			if (!(flags & OCFS2_BH_IGNORE_CACHE)
+			    && !(flags & OCFS2_BH_READAHEAD)
+			    && ocfs2_buffer_uptodate(ci, bh)) {
+				unlock_buffer(bh);
+				continue;
+			}
+
+			clear_buffer_uptodate(bh);
+			get_bh(bh); /* for end_buffer_read_sync() */
+			if (validate)
+				set_buffer_needs_validate(bh);
+			bh->b_end_io = end_buffer_read_sync;
+			submit_bh(READ, bh);
+			continue;
+		}
+	}
+
+	status = 0;
+
+	for (i = (nr - 1); i >= 0; i--) {
+		bh = bhs[i];
+
+		if (!(flags & OCFS2_BH_READAHEAD)) {
+			/* We know this can't have changed as we hold the
+			 * owner sem. Avoid doing any work on the bh if the
+			 * journal has it. */
+			if (!buffer_jbd(bh))
+				wait_on_buffer(bh);
+
+			if (!buffer_uptodate(bh)) {
+				/* Status won't be cleared from here on out,
+				 * so we can safely record this and loop back
+				 * to cleanup the other buffers. Don't need to
+				 * remove the clustered uptodate information
+				 * for this bh as it's not marked locally
+				 * uptodate. */
+				status = -EIO;
+				put_bh(bh);
+				bhs[i] = NULL;
+				continue;
+			}
+
+			if (buffer_needs_validate(bh)) {
+				/* We never set NeedsValidate if the
+				 * buffer was held by the journal, so
+				 * that better not have changed */
+				BUG_ON(buffer_jbd(bh));
+				clear_buffer_needs_validate(bh);
+				status = validate(sb, bh);
+				if (status) {
+					put_bh(bh);
+					bhs[i] = NULL;
+					continue;
+				}
+			}
+		}
+
+		/* Always set the buffer in the cache, even if it was
+		 * a forced read, or read-ahead which hasn't yet
+		 * completed. */
+		ocfs2_set_buffer_uptodate(ci, bh);
+	}
+	ocfs2_metadata_cache_io_unlock(ci);
+
+	trace_ocfs2_read_blocks_end((unsigned long long)block, nr,
+				    flags, ignore_cache);
+
+bail:
+
+	return status;
+}
+
+/* Check whether the blkno is the super block or one of the backups. */
+static void ocfs2_check_super_or_backup(struct super_block *sb,
+					sector_t blkno)
+{
+	int i;
+	u64 backup_blkno;
+
+	if (blkno == OCFS2_SUPER_BLOCK_BLKNO)
+		return;
+
+	for (i = 0; i < OCFS2_MAX_BACKUP_SUPERBLOCKS; i++) {
+		backup_blkno = ocfs2_backup_super_blkno(sb, i);
+		if (backup_blkno == blkno)
+			return;
+	}
+
+	BUG();
+}
+
+/*
+ * Write super block and backups doesn't need to collaborate with journal,
+ * so we don't need to lock ip_io_mutex and ci doesn't need to bea passed
+ * into this function.
+ */
+int ocfs2_write_super_or_backup(struct ocfs2_super *osb,
+				struct buffer_head *bh)
+{
+	int ret = 0;
+	struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
+
+	BUG_ON(buffer_jbd(bh));
+	ocfs2_check_super_or_backup(osb->sb, bh->b_blocknr);
+
+	if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb)) {
+		ret = -EROFS;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	lock_buffer(bh);
+	set_buffer_uptodate(bh);
+
+	/* remove from dirty list before I/O. */
+	clear_buffer_dirty(bh);
+
+	get_bh(bh); /* for end_buffer_write_sync() */
+	bh->b_end_io = end_buffer_write_sync;
+	ocfs2_compute_meta_ecc(osb->sb, bh->b_data, &di->i_check);
+	submit_bh(WRITE, bh);
+
+	wait_on_buffer(bh);
+
+	if (!buffer_uptodate(bh)) {
+		ret = -EIO;
+		mlog_errno(ret);
+	}
+
+out:
+	return ret;
+}

fs/ocfs2/buffer_head_io.h

@@ -0,0 +1,77 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ocfs2_buffer_head.h
+ *
+ * Buffer cache handling functions defined
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_BUFFER_HEAD_IO_H
+#define OCFS2_BUFFER_HEAD_IO_H
+
+#include <linux/buffer_head.h>
+
+void ocfs2_end_buffer_io_sync(struct buffer_head *bh,
+			     int uptodate);
+
+int ocfs2_write_block(struct ocfs2_super          *osb,
+		      struct buffer_head  *bh,
+		      struct ocfs2_caching_info   *ci);
+int ocfs2_read_blocks_sync(struct ocfs2_super *osb, u64 block,
+			   unsigned int nr, struct buffer_head *bhs[]);
+
+/*
+ * If not NULL, validate() will be called on a buffer that is freshly
+ * read from disk.  It will not be called if the buffer was in cache.
+ * Note that if validate() is being used for this buffer, it needs to
+ * be set even for a READAHEAD call, as it marks the buffer for later
+ * validation.
+ */
+int ocfs2_read_blocks(struct ocfs2_caching_info *ci, u64 block, int nr,
+		      struct buffer_head *bhs[], int flags,
+		      int (*validate)(struct super_block *sb,
+				      struct buffer_head *bh));
+
+int ocfs2_write_super_or_backup(struct ocfs2_super *osb,
+				struct buffer_head *bh);
+
+#define OCFS2_BH_IGNORE_CACHE      1
+#define OCFS2_BH_READAHEAD         8
+
+static inline int ocfs2_read_block(struct ocfs2_caching_info *ci, u64 off,
+				   struct buffer_head **bh,
+				   int (*validate)(struct super_block *sb,
+						   struct buffer_head *bh))
+{
+	int status = 0;
+
+	if (bh == NULL) {
+		printk("ocfs2: bh == NULL\n");
+		status = -EINVAL;
+		goto bail;
+	}
+
+	status = ocfs2_read_blocks(ci, off, 1, bh, 0, validate);
+
+bail:
+	return status;
+}
+
+#endif /* OCFS2_BUFFER_HEAD_IO_H */

fs/ocfs2/cluster/Makefile

@@ -0,0 +1,4 @@
+obj-$(CONFIG_OCFS2_FS) += ocfs2_nodemanager.o
+
+ocfs2_nodemanager-objs := heartbeat.o masklog.o sys.o nodemanager.o \
+	quorum.o tcp.o netdebug.o

fs/ocfs2/cluster/heartbeat.h

@@ -0,0 +1,90 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * heartbeat.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+#ifndef O2CLUSTER_HEARTBEAT_H
+#define O2CLUSTER_HEARTBEAT_H
+
+#include "ocfs2_heartbeat.h"
+
+#define O2HB_REGION_TIMEOUT_MS		2000
+
+#define O2HB_MAX_REGION_NAME_LEN	32
+
+/* number of changes to be seen as live */
+#define O2HB_LIVE_THRESHOLD	   2
+/* number of equal samples to be seen as dead */
+extern unsigned int o2hb_dead_threshold;
+#define O2HB_DEFAULT_DEAD_THRESHOLD	   31
+/* Otherwise MAX_WRITE_TIMEOUT will be zero... */
+#define O2HB_MIN_DEAD_THRESHOLD	  2
+#define O2HB_MAX_WRITE_TIMEOUT_MS (O2HB_REGION_TIMEOUT_MS * (o2hb_dead_threshold - 1))
+
+#define O2HB_CB_MAGIC		0x51d1e4ec
+
+/* callback stuff */
+enum o2hb_callback_type {
+	O2HB_NODE_DOWN_CB = 0,
+	O2HB_NODE_UP_CB,
+	O2HB_NUM_CB
+};
+
+struct o2nm_node;
+typedef void (o2hb_cb_func)(struct o2nm_node *, int, void *);
+
+struct o2hb_callback_func {
+	u32			hc_magic;
+	struct list_head	hc_item;
+	o2hb_cb_func		*hc_func;
+	void			*hc_data;
+	int			hc_priority;
+	enum o2hb_callback_type hc_type;
+};
+
+struct config_group *o2hb_alloc_hb_set(void);
+void o2hb_free_hb_set(struct config_group *group);
+
+void o2hb_setup_callback(struct o2hb_callback_func *hc,
+			 enum o2hb_callback_type type,
+			 o2hb_cb_func *func,
+			 void *data,
+			 int priority);
+int o2hb_register_callback(const char *region_uuid,
+			   struct o2hb_callback_func *hc);
+void o2hb_unregister_callback(const char *region_uuid,
+			      struct o2hb_callback_func *hc);
+void o2hb_fill_node_map(unsigned long *map,
+			unsigned bytes);
+void o2hb_exit(void);
+int o2hb_init(void);
+int o2hb_check_node_heartbeating(u8 node_num);
+int o2hb_check_node_heartbeating_no_sem(u8 node_num);
+int o2hb_check_node_heartbeating_from_callback(u8 node_num);
+int o2hb_check_local_node_heartbeating(void);
+void o2hb_stop_all_regions(void);
+int o2hb_get_all_regions(char *region_uuids, u8 numregions);
+int o2hb_global_heartbeat_active(void);
+
+#endif /* O2CLUSTER_HEARTBEAT_H */

fs/ocfs2/cluster/masklog.c

@@ -0,0 +1,155 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2004, 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/proc_fs.h>
+#include <linux/seq_file.h>
+#include <linux/string.h>
+#include <asm/uaccess.h>
+
+#include "masklog.h"
+
+struct mlog_bits mlog_and_bits = MLOG_BITS_RHS(MLOG_INITIAL_AND_MASK);
+EXPORT_SYMBOL_GPL(mlog_and_bits);
+struct mlog_bits mlog_not_bits = MLOG_BITS_RHS(0);
+EXPORT_SYMBOL_GPL(mlog_not_bits);
+
+static ssize_t mlog_mask_show(u64 mask, char *buf)
+{
+	char *state;
+
+	if (__mlog_test_u64(mask, mlog_and_bits))
+		state = "allow";
+	else if (__mlog_test_u64(mask, mlog_not_bits))
+		state = "deny";
+	else
+		state = "off";
+
+	return snprintf(buf, PAGE_SIZE, "%s\n", state);
+}
+
+static ssize_t mlog_mask_store(u64 mask, const char *buf, size_t count)
+{
+	if (!strncasecmp(buf, "allow", 5)) {
+		__mlog_set_u64(mask, mlog_and_bits);
+		__mlog_clear_u64(mask, mlog_not_bits);
+	} else if (!strncasecmp(buf, "deny", 4)) {
+		__mlog_set_u64(mask, mlog_not_bits);
+		__mlog_clear_u64(mask, mlog_and_bits);
+	} else if (!strncasecmp(buf, "off", 3)) {
+		__mlog_clear_u64(mask, mlog_not_bits);
+		__mlog_clear_u64(mask, mlog_and_bits);
+	} else
+		return -EINVAL;
+
+	return count;
+}
+
+struct mlog_attribute {
+	struct attribute attr;
+	u64 mask;
+};
+
+#define to_mlog_attr(_attr) container_of(_attr, struct mlog_attribute, attr)
+
+#define define_mask(_name) {			\
+	.attr = {				\
+		.name = #_name,			\
+		.mode = S_IRUGO | S_IWUSR,	\
+	},					\
+	.mask = ML_##_name,			\
+}
+
+static struct mlog_attribute mlog_attrs[MLOG_MAX_BITS] = {
+	define_mask(TCP),
+	define_mask(MSG),
+	define_mask(SOCKET),
+	define_mask(HEARTBEAT),
+	define_mask(HB_BIO),
+	define_mask(DLMFS),
+	define_mask(DLM),
+	define_mask(DLM_DOMAIN),
+	define_mask(DLM_THREAD),
+	define_mask(DLM_MASTER),
+	define_mask(DLM_RECOVERY),
+	define_mask(DLM_GLUE),
+	define_mask(VOTE),
+	define_mask(CONN),
+	define_mask(QUORUM),
+	define_mask(BASTS),
+	define_mask(CLUSTER),
+	define_mask(ERROR),
+	define_mask(NOTICE),
+	define_mask(KTHREAD),
+};
+
+static struct attribute *mlog_attr_ptrs[MLOG_MAX_BITS] = {NULL, };
+
+static ssize_t mlog_show(struct kobject *obj, struct attribute *attr,
+			 char *buf)
+{
+	struct mlog_attribute *mlog_attr = to_mlog_attr(attr);
+
+	return mlog_mask_show(mlog_attr->mask, buf);
+}
+
+static ssize_t mlog_store(struct kobject *obj, struct attribute *attr,
+			  const char *buf, size_t count)
+{
+	struct mlog_attribute *mlog_attr = to_mlog_attr(attr);
+
+	return mlog_mask_store(mlog_attr->mask, buf, count);
+}
+
+static const struct sysfs_ops mlog_attr_ops = {
+	.show  = mlog_show,
+	.store = mlog_store,
+};
+
+static struct kobj_type mlog_ktype = {
+	.default_attrs = mlog_attr_ptrs,
+	.sysfs_ops     = &mlog_attr_ops,
+};
+
+static struct kset mlog_kset = {
+	.kobj   = {.ktype = &mlog_ktype},
+};
+
+int mlog_sys_init(struct kset *o2cb_kset)
+{
+	int i = 0;
+
+	while (mlog_attrs[i].attr.mode) {
+		mlog_attr_ptrs[i] = &mlog_attrs[i].attr;
+		i++;
+	}
+	mlog_attr_ptrs[i] = NULL;
+
+	kobject_set_name(&mlog_kset.kobj, "logmask");
+	mlog_kset.kobj.kset = o2cb_kset;
+	return kset_register(&mlog_kset);
+}
+
+void mlog_sys_shutdown(void)
+{
+	kset_unregister(&mlog_kset);
+}

fs/ocfs2/cluster/masklog.h

@@ -0,0 +1,220 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef O2CLUSTER_MASKLOG_H
+#define O2CLUSTER_MASKLOG_H
+
+/*
+ * For now this is a trivial wrapper around printk() that gives the critical
+ * ability to enable sets of debugging output at run-time.  In the future this
+ * will almost certainly be redirected to relayfs so that it can pay a
+ * substantially lower heisenberg tax.
+ *
+ * Callers associate the message with a bitmask and a global bitmask is
+ * maintained with help from /proc.  If any of the bits match the message is
+ * output.
+ *
+ * We must have efficient bit tests on i386 and it seems gcc still emits crazy
+ * code for the 64bit compare.  It emits very good code for the dual unsigned
+ * long tests, though, completely avoiding tests that can never pass if the
+ * caller gives a constant bitmask that fills one of the longs with all 0s.  So
+ * the desire is to have almost all of the calls decided on by comparing just
+ * one of the longs.  This leads to having infrequently given bits that are
+ * frequently matched in the high bits.
+ *
+ * _ERROR and _NOTICE are used for messages that always go to the console and
+ * have appropriate KERN_ prefixes.  We wrap these in our function instead of
+ * just calling printk() so that this can eventually make its way through
+ * relayfs along with the debugging messages.  Everything else gets KERN_DEBUG.
+ * The inline tests and macro dance give GCC the opportunity to quite cleverly
+ * only emit the appropriage printk() when the caller passes in a constant
+ * mask, as is almost always the case.
+ *
+ * All this bitmask nonsense is managed from the files under
+ * /sys/fs/o2cb/logmask/.  Reading the files gives a straightforward
+ * indication of which bits are allowed (allow) or denied (off/deny).
+ * 	ENTRY deny
+ * 	EXIT deny
+ * 	TCP off
+ * 	MSG off
+ * 	SOCKET off
+ * 	ERROR allow
+ * 	NOTICE allow
+ *
+ * Writing changes the state of a given bit and requires a strictly formatted
+ * single write() call:
+ *
+ * 	write(fd, "allow", 5);
+ *
+ * Echoing allow/deny/off string into the logmask files can flip the bits
+ * on or off as expected; here is the bash script for example:
+ *
+ * log_mask="/sys/fs/o2cb/log_mask"
+ * for node in ENTRY EXIT TCP MSG SOCKET ERROR NOTICE; do
+ *	echo allow >"$log_mask"/"$node"
+ * done
+ *
+ * The debugfs.ocfs2 tool can also flip the bits with the -l option:
+ *
+ * debugfs.ocfs2 -l TCP allow
+ */
+
+/* for task_struct */
+#include <linux/sched.h>
+
+/* bits that are frequently given and infrequently matched in the low word */
+/* NOTE: If you add a flag, you need to also update masklog.c! */
+#define ML_TCP		0x0000000000000001ULL /* net cluster/tcp.c */
+#define ML_MSG		0x0000000000000002ULL /* net network messages */
+#define ML_SOCKET	0x0000000000000004ULL /* net socket lifetime */
+#define ML_HEARTBEAT	0x0000000000000008ULL /* hb all heartbeat tracking */
+#define ML_HB_BIO	0x0000000000000010ULL /* hb io tracing */
+#define ML_DLMFS	0x0000000000000020ULL /* dlm user dlmfs */
+#define ML_DLM		0x0000000000000040ULL /* dlm general debugging */
+#define ML_DLM_DOMAIN	0x0000000000000080ULL /* dlm domain debugging */
+#define ML_DLM_THREAD	0x0000000000000100ULL /* dlm domain thread */
+#define ML_DLM_MASTER	0x0000000000000200ULL /* dlm master functions */
+#define ML_DLM_RECOVERY	0x0000000000000400ULL /* dlm master functions */
+#define ML_DLM_GLUE	0x0000000000000800ULL /* ocfs2 dlm glue layer */
+#define ML_VOTE		0x0000000000001000ULL /* ocfs2 node messaging  */
+#define ML_CONN		0x0000000000002000ULL /* net connection management */
+#define ML_QUORUM	0x0000000000004000ULL /* net connection quorum */
+#define ML_BASTS	0x0000000000008000ULL /* dlmglue asts and basts */
+#define ML_CLUSTER	0x0000000000010000ULL /* cluster stack */
+
+/* bits that are infrequently given and frequently matched in the high word */
+#define ML_ERROR	0x1000000000000000ULL /* sent to KERN_ERR */
+#define ML_NOTICE	0x2000000000000000ULL /* setn to KERN_NOTICE */
+#define ML_KTHREAD	0x4000000000000000ULL /* kernel thread activity */
+
+#define MLOG_INITIAL_AND_MASK (ML_ERROR|ML_NOTICE)
+#ifndef MLOG_MASK_PREFIX
+#define MLOG_MASK_PREFIX 0
+#endif
+
+/*
+ * When logging is disabled, force the bit test to 0 for anything other
+ * than errors and notices, allowing gcc to remove the code completely.
+ * When enabled, allow all masks.
+ */
+#if defined(CONFIG_OCFS2_DEBUG_MASKLOG)
+#define ML_ALLOWED_BITS ~0
+#else
+#define ML_ALLOWED_BITS (ML_ERROR|ML_NOTICE)
+#endif
+
+#define MLOG_MAX_BITS 64
+
+struct mlog_bits {
+	unsigned long words[MLOG_MAX_BITS / BITS_PER_LONG];
+};
+
+extern struct mlog_bits mlog_and_bits, mlog_not_bits;
+
+#if BITS_PER_LONG == 32
+
+#define __mlog_test_u64(mask, bits)			\
+	( (u32)(mask & 0xffffffff) & bits.words[0] || 	\
+	  ((u64)(mask) >> 32) & bits.words[1] )
+#define __mlog_set_u64(mask, bits) do {			\
+	bits.words[0] |= (u32)(mask & 0xffffffff);	\
+       	bits.words[1] |= (u64)(mask) >> 32;		\
+} while (0)
+#define __mlog_clear_u64(mask, bits) do {		\
+	bits.words[0] &= ~((u32)(mask & 0xffffffff));	\
+       	bits.words[1] &= ~((u64)(mask) >> 32);		\
+} while (0)
+#define MLOG_BITS_RHS(mask) {				\
+	{						\
+		[0] = (u32)(mask & 0xffffffff),		\
+		[1] = (u64)(mask) >> 32,		\
+	}						\
+}
+
+#else /* 32bit long above, 64bit long below */
+
+#define __mlog_test_u64(mask, bits)	((mask) & bits.words[0])
+#define __mlog_set_u64(mask, bits) do {		\
+	bits.words[0] |= (mask);		\
+} while (0)
+#define __mlog_clear_u64(mask, bits) do {	\
+	bits.words[0] &= ~(mask);		\
+} while (0)
+#define MLOG_BITS_RHS(mask) { { (mask) } }
+
+#endif
+
+/*
+ * smp_processor_id() "helpfully" screams when called outside preemptible
+ * regions in current kernels.  sles doesn't have the variants that don't
+ * scream.  just do this instead of trying to guess which we're building
+ * against.. *sigh*.
+ */
+#define __mlog_cpu_guess ({		\
+	unsigned long _cpu = get_cpu();	\
+	put_cpu();			\
+	_cpu;				\
+})
+
+/* In the following two macros, the whitespace after the ',' just
+ * before ##args is intentional. Otherwise, gcc 2.95 will eat the
+ * previous token if args expands to nothing.
+ */
+#define __mlog_printk(level, fmt, args...)				\
+	printk(level "(%s,%u,%lu):%s:%d " fmt, current->comm,		\
+	       task_pid_nr(current), __mlog_cpu_guess,			\
+	       __PRETTY_FUNCTION__, __LINE__ , ##args)
+
+#define mlog(mask, fmt, args...) do {					\
+	u64 __m = MLOG_MASK_PREFIX | (mask);				\
+	if ((__m & ML_ALLOWED_BITS) &&					\
+	    __mlog_test_u64(__m, mlog_and_bits) &&			\
+	    !__mlog_test_u64(__m, mlog_not_bits)) {			\
+		if (__m & ML_ERROR)					\
+			__mlog_printk(KERN_ERR, "ERROR: "fmt , ##args);	\
+		else if (__m & ML_NOTICE)				\
+			__mlog_printk(KERN_NOTICE, fmt , ##args);	\
+		else __mlog_printk(KERN_INFO, fmt , ##args);		\
+	}								\
+} while (0)
+
+#define mlog_errno(st) do {						\
+	int _st = (st);							\
+	if (_st != -ERESTARTSYS && _st != -EINTR &&			\
+	    _st != AOP_TRUNCATED_PAGE && _st != -ENOSPC &&		\
+	    _st != -EDQUOT)						\
+		mlog(ML_ERROR, "status = %lld\n", (long long)_st);	\
+} while (0)
+
+#define mlog_bug_on_msg(cond, fmt, args...) do {			\
+	if (cond) {							\
+		mlog(ML_ERROR, "bug expression: " #cond "\n");		\
+		mlog(ML_ERROR, fmt, ##args);				\
+		BUG();							\
+	}								\
+} while (0)
+
+#include <linux/kobject.h>
+#include <linux/sysfs.h>
+int mlog_sys_init(struct kset *o2cb_subsys);
+void mlog_sys_shutdown(void);
+
+#endif /* O2CLUSTER_MASKLOG_H */

fs/ocfs2/cluster/netdebug.c

@@ -0,0 +1,539 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * netdebug.c
+ *
+ * debug functionality for o2net
+ *
+ * Copyright (C) 2005, 2008 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+#ifdef CONFIG_DEBUG_FS
+
+#include <linux/module.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/idr.h>
+#include <linux/kref.h>
+#include <linux/seq_file.h>
+#include <linux/debugfs.h>
+
+#include <linux/uaccess.h>
+
+#include "tcp.h"
+#include "nodemanager.h"
+#define MLOG_MASK_PREFIX ML_TCP
+#include "masklog.h"
+
+#include "tcp_internal.h"
+
+#define O2NET_DEBUG_DIR		"o2net"
+#define SC_DEBUG_NAME		"sock_containers"
+#define NST_DEBUG_NAME		"send_tracking"
+#define STATS_DEBUG_NAME	"stats"
+#define NODES_DEBUG_NAME	"connected_nodes"
+
+#define SHOW_SOCK_CONTAINERS	0
+#define SHOW_SOCK_STATS		1
+
+static struct dentry *o2net_dentry;
+static struct dentry *sc_dentry;
+static struct dentry *nst_dentry;
+static struct dentry *stats_dentry;
+static struct dentry *nodes_dentry;
+
+static DEFINE_SPINLOCK(o2net_debug_lock);
+
+static LIST_HEAD(sock_containers);
+static LIST_HEAD(send_tracking);
+
+void o2net_debug_add_nst(struct o2net_send_tracking *nst)
+{
+	spin_lock(&o2net_debug_lock);
+	list_add(&nst->st_net_debug_item, &send_tracking);
+	spin_unlock(&o2net_debug_lock);
+}
+
+void o2net_debug_del_nst(struct o2net_send_tracking *nst)
+{
+	spin_lock(&o2net_debug_lock);
+	if (!list_empty(&nst->st_net_debug_item))
+		list_del_init(&nst->st_net_debug_item);
+	spin_unlock(&o2net_debug_lock);
+}
+
+static struct o2net_send_tracking
+			*next_nst(struct o2net_send_tracking *nst_start)
+{
+	struct o2net_send_tracking *nst, *ret = NULL;
+
+	assert_spin_locked(&o2net_debug_lock);
+
+	list_for_each_entry(nst, &nst_start->st_net_debug_item,
+			    st_net_debug_item) {
+		/* discover the head of the list */
+		if (&nst->st_net_debug_item == &send_tracking)
+			break;
+
+		/* use st_task to detect real nsts in the list */
+		if (nst->st_task != NULL) {
+			ret = nst;
+			break;
+		}
+	}
+
+	return ret;
+}
+
+static void *nst_seq_start(struct seq_file *seq, loff_t *pos)
+{
+	struct o2net_send_tracking *nst, *dummy_nst = seq->private;
+
+	spin_lock(&o2net_debug_lock);
+	nst = next_nst(dummy_nst);
+	spin_unlock(&o2net_debug_lock);
+
+	return nst;
+}
+
+static void *nst_seq_next(struct seq_file *seq, void *v, loff_t *pos)
+{
+	struct o2net_send_tracking *nst, *dummy_nst = seq->private;
+
+	spin_lock(&o2net_debug_lock);
+	nst = next_nst(dummy_nst);
+	list_del_init(&dummy_nst->st_net_debug_item);
+	if (nst)
+		list_add(&dummy_nst->st_net_debug_item,
+			 &nst->st_net_debug_item);
+	spin_unlock(&o2net_debug_lock);
+
+	return nst; /* unused, just needs to be null when done */
+}
+
+static int nst_seq_show(struct seq_file *seq, void *v)
+{
+	struct o2net_send_tracking *nst, *dummy_nst = seq->private;
+	ktime_t now;
+	s64 sock, send, status;
+
+	spin_lock(&o2net_debug_lock);
+	nst = next_nst(dummy_nst);
+	if (!nst)
+		goto out;
+
+	now = ktime_get();
+	sock = ktime_to_us(ktime_sub(now, nst->st_sock_time));
+	send = ktime_to_us(ktime_sub(now, nst->st_send_time));
+	status = ktime_to_us(ktime_sub(now, nst->st_status_time));
+
+	/* get_task_comm isn't exported.  oh well. */
+	seq_printf(seq, "%p:\n"
+		   "  pid:          %lu\n"
+		   "  tgid:         %lu\n"
+		   "  process name: %s\n"
+		   "  node:         %u\n"
+		   "  sc:           %p\n"
+		   "  message id:   %d\n"
+		   "  message type: %u\n"
+		   "  message key:  0x%08x\n"
+		   "  sock acquiry: %lld usecs ago\n"
+		   "  send start:   %lld usecs ago\n"
+		   "  wait start:   %lld usecs ago\n",
+		   nst, (unsigned long)task_pid_nr(nst->st_task),
+		   (unsigned long)nst->st_task->tgid,
+		   nst->st_task->comm, nst->st_node,
+		   nst->st_sc, nst->st_id, nst->st_msg_type,
+		   nst->st_msg_key,
+		   (long long)sock,
+		   (long long)send,
+		   (long long)status);
+
+out:
+	spin_unlock(&o2net_debug_lock);
+
+	return 0;
+}
+
+static void nst_seq_stop(struct seq_file *seq, void *v)
+{
+}
+
+static const struct seq_operations nst_seq_ops = {
+	.start = nst_seq_start,
+	.next = nst_seq_next,
+	.stop = nst_seq_stop,
+	.show = nst_seq_show,
+};
+
+static int nst_fop_open(struct inode *inode, struct file *file)
+{
+	struct o2net_send_tracking *dummy_nst;
+
+	dummy_nst = __seq_open_private(file, &nst_seq_ops, sizeof(*dummy_nst));
+	if (!dummy_nst)
+		return -ENOMEM;
+	o2net_debug_add_nst(dummy_nst);
+
+	return 0;
+}
+
+static int nst_fop_release(struct inode *inode, struct file *file)
+{
+	struct seq_file *seq = file->private_data;
+	struct o2net_send_tracking *dummy_nst = seq->private;
+
+	o2net_debug_del_nst(dummy_nst);
+	return seq_release_private(inode, file);
+}
+
+static const struct file_operations nst_seq_fops = {
+	.open = nst_fop_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = nst_fop_release,
+};
+
+void o2net_debug_add_sc(struct o2net_sock_container *sc)
+{
+	spin_lock(&o2net_debug_lock);
+	list_add(&sc->sc_net_debug_item, &sock_containers);
+	spin_unlock(&o2net_debug_lock);
+}
+
+void o2net_debug_del_sc(struct o2net_sock_container *sc)
+{
+	spin_lock(&o2net_debug_lock);
+	list_del_init(&sc->sc_net_debug_item);
+	spin_unlock(&o2net_debug_lock);
+}
+
+struct o2net_sock_debug {
+	int dbg_ctxt;
+	struct o2net_sock_container *dbg_sock;
+};
+
+static struct o2net_sock_container
+			*next_sc(struct o2net_sock_container *sc_start)
+{
+	struct o2net_sock_container *sc, *ret = NULL;
+
+	assert_spin_locked(&o2net_debug_lock);
+
+	list_for_each_entry(sc, &sc_start->sc_net_debug_item,
+			    sc_net_debug_item) {
+		/* discover the head of the list miscast as a sc */
+		if (&sc->sc_net_debug_item == &sock_containers)
+			break;
+
+		/* use sc_page to detect real scs in the list */
+		if (sc->sc_page != NULL) {
+			ret = sc;
+			break;
+		}
+	}
+
+	return ret;
+}
+
+static void *sc_seq_start(struct seq_file *seq, loff_t *pos)
+{
+	struct o2net_sock_debug *sd = seq->private;
+	struct o2net_sock_container *sc, *dummy_sc = sd->dbg_sock;
+
+	spin_lock(&o2net_debug_lock);
+	sc = next_sc(dummy_sc);
+	spin_unlock(&o2net_debug_lock);
+
+	return sc;
+}
+
+static void *sc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
+{
+	struct o2net_sock_debug *sd = seq->private;
+	struct o2net_sock_container *sc, *dummy_sc = sd->dbg_sock;
+
+	spin_lock(&o2net_debug_lock);
+	sc = next_sc(dummy_sc);
+	list_del_init(&dummy_sc->sc_net_debug_item);
+	if (sc)
+		list_add(&dummy_sc->sc_net_debug_item, &sc->sc_net_debug_item);
+	spin_unlock(&o2net_debug_lock);
+
+	return sc; /* unused, just needs to be null when done */
+}
+
+#ifdef CONFIG_OCFS2_FS_STATS
+# define sc_send_count(_s)		((_s)->sc_send_count)
+# define sc_recv_count(_s)		((_s)->sc_recv_count)
+# define sc_tv_acquiry_total_ns(_s)	(ktime_to_ns((_s)->sc_tv_acquiry_total))
+# define sc_tv_send_total_ns(_s)	(ktime_to_ns((_s)->sc_tv_send_total))
+# define sc_tv_status_total_ns(_s)	(ktime_to_ns((_s)->sc_tv_status_total))
+# define sc_tv_process_total_ns(_s)	(ktime_to_ns((_s)->sc_tv_process_total))
+#else
+# define sc_send_count(_s)		(0U)
+# define sc_recv_count(_s)		(0U)
+# define sc_tv_acquiry_total_ns(_s)	(0LL)
+# define sc_tv_send_total_ns(_s)	(0LL)
+# define sc_tv_status_total_ns(_s)	(0LL)
+# define sc_tv_process_total_ns(_s)	(0LL)
+#endif
+
+/* So that debugfs.ocfs2 can determine which format is being used */
+#define O2NET_STATS_STR_VERSION		1
+static void sc_show_sock_stats(struct seq_file *seq,
+			       struct o2net_sock_container *sc)
+{
+	if (!sc)
+		return;
+
+	seq_printf(seq, "%d,%u,%lu,%lld,%lld,%lld,%lu,%lld\n", O2NET_STATS_STR_VERSION,
+		   sc->sc_node->nd_num, (unsigned long)sc_send_count(sc),
+		   (long long)sc_tv_acquiry_total_ns(sc),
+		   (long long)sc_tv_send_total_ns(sc),
+		   (long long)sc_tv_status_total_ns(sc),
+		   (unsigned long)sc_recv_count(sc),
+		   (long long)sc_tv_process_total_ns(sc));
+}
+
+static void sc_show_sock_container(struct seq_file *seq,
+				   struct o2net_sock_container *sc)
+{
+	struct inet_sock *inet = NULL;
+	__be32 saddr = 0, daddr = 0;
+	__be16 sport = 0, dport = 0;
+
+	if (!sc)
+		return;
+
+	if (sc->sc_sock) {
+		inet = inet_sk(sc->sc_sock->sk);
+		/* the stack's structs aren't sparse endian clean */
+		saddr = (__force __be32)inet->inet_saddr;
+		daddr = (__force __be32)inet->inet_daddr;
+		sport = (__force __be16)inet->inet_sport;
+		dport = (__force __be16)inet->inet_dport;
+	}
+
+	/* XXX sigh, inet-> doesn't have sparse annotation so any
+	 * use of it here generates a warning with -Wbitwise */
+	seq_printf(seq, "%p:\n"
+		   "  krefs:           %d\n"
+		   "  sock:            %pI4:%u -> "
+				      "%pI4:%u\n"
+		   "  remote node:     %s\n"
+		   "  page off:        %zu\n"
+		   "  handshake ok:    %u\n"
+		   "  timer:           %lld usecs\n"
+		   "  data ready:      %lld usecs\n"
+		   "  advance start:   %lld usecs\n"
+		   "  advance stop:    %lld usecs\n"
+		   "  func start:      %lld usecs\n"
+		   "  func stop:       %lld usecs\n"
+		   "  func key:        0x%08x\n"
+		   "  func type:       %u\n",
+		   sc,
+		   atomic_read(&sc->sc_kref.refcount),
+		   &saddr, inet ? ntohs(sport) : 0,
+		   &daddr, inet ? ntohs(dport) : 0,
+		   sc->sc_node->nd_name,
+		   sc->sc_page_off,
+		   sc->sc_handshake_ok,
+		   (long long)ktime_to_us(sc->sc_tv_timer),
+		   (long long)ktime_to_us(sc->sc_tv_data_ready),
+		   (long long)ktime_to_us(sc->sc_tv_advance_start),
+		   (long long)ktime_to_us(sc->sc_tv_advance_stop),
+		   (long long)ktime_to_us(sc->sc_tv_func_start),
+		   (long long)ktime_to_us(sc->sc_tv_func_stop),
+		   sc->sc_msg_key,
+		   sc->sc_msg_type);
+}
+
+static int sc_seq_show(struct seq_file *seq, void *v)
+{
+	struct o2net_sock_debug *sd = seq->private;
+	struct o2net_sock_container *sc, *dummy_sc = sd->dbg_sock;
+
+	spin_lock(&o2net_debug_lock);
+	sc = next_sc(dummy_sc);
+
+	if (sc) {
+		if (sd->dbg_ctxt == SHOW_SOCK_CONTAINERS)
+			sc_show_sock_container(seq, sc);
+		else
+			sc_show_sock_stats(seq, sc);
+	}
+
+	spin_unlock(&o2net_debug_lock);
+
+	return 0;
+}
+
+static void sc_seq_stop(struct seq_file *seq, void *v)
+{
+}
+
+static const struct seq_operations sc_seq_ops = {
+	.start = sc_seq_start,
+	.next = sc_seq_next,
+	.stop = sc_seq_stop,
+	.show = sc_seq_show,
+};
+
+static int sc_common_open(struct file *file, int ctxt)
+{
+	struct o2net_sock_debug *sd;
+	struct o2net_sock_container *dummy_sc;
+
+	dummy_sc = kzalloc(sizeof(*dummy_sc), GFP_KERNEL);
+	if (!dummy_sc)
+		return -ENOMEM;
+
+	sd = __seq_open_private(file, &sc_seq_ops, sizeof(*sd));
+	if (!sd) {
+		kfree(dummy_sc);
+		return -ENOMEM;
+	}
+
+	sd->dbg_ctxt = ctxt;
+	sd->dbg_sock = dummy_sc;
+
+	o2net_debug_add_sc(dummy_sc);
+
+	return 0;
+}
+
+static int sc_fop_release(struct inode *inode, struct file *file)
+{
+	struct seq_file *seq = file->private_data;
+	struct o2net_sock_debug *sd = seq->private;
+	struct o2net_sock_container *dummy_sc = sd->dbg_sock;
+
+	o2net_debug_del_sc(dummy_sc);
+	return seq_release_private(inode, file);
+}
+
+static int stats_fop_open(struct inode *inode, struct file *file)
+{
+	return sc_common_open(file, SHOW_SOCK_STATS);
+}
+
+static const struct file_operations stats_seq_fops = {
+	.open = stats_fop_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = sc_fop_release,
+};
+
+static int sc_fop_open(struct inode *inode, struct file *file)
+{
+	return sc_common_open(file, SHOW_SOCK_CONTAINERS);
+}
+
+static const struct file_operations sc_seq_fops = {
+	.open = sc_fop_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = sc_fop_release,
+};
+
+static int o2net_fill_bitmap(char *buf, int len)
+{
+	unsigned long map[BITS_TO_LONGS(O2NM_MAX_NODES)];
+	int i = -1, out = 0;
+
+	o2net_fill_node_map(map, sizeof(map));
+
+	while ((i = find_next_bit(map, O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES)
+		out += snprintf(buf + out, PAGE_SIZE - out, "%d ", i);
+	out += snprintf(buf + out, PAGE_SIZE - out, "\n");
+
+	return out;
+}
+
+static int nodes_fop_open(struct inode *inode, struct file *file)
+{
+	char *buf;
+
+	buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
+	if (!buf)
+		return -ENOMEM;
+
+	i_size_write(inode, o2net_fill_bitmap(buf, PAGE_SIZE));
+
+	file->private_data = buf;
+
+	return 0;
+}
+
+static int o2net_debug_release(struct inode *inode, struct file *file)
+{
+	kfree(file->private_data);
+	return 0;
+}
+
+static ssize_t o2net_debug_read(struct file *file, char __user *buf,
+				size_t nbytes, loff_t *ppos)
+{
+	return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
+				       i_size_read(file->f_mapping->host));
+}
+
+static const struct file_operations nodes_fops = {
+	.open		= nodes_fop_open,
+	.release	= o2net_debug_release,
+	.read		= o2net_debug_read,
+	.llseek		= generic_file_llseek,
+};
+
+void o2net_debugfs_exit(void)
+{
+	debugfs_remove(nodes_dentry);
+	debugfs_remove(stats_dentry);
+	debugfs_remove(sc_dentry);
+	debugfs_remove(nst_dentry);
+	debugfs_remove(o2net_dentry);
+}
+
+int o2net_debugfs_init(void)
+{
+	umode_t mode = S_IFREG|S_IRUSR;
+
+	o2net_dentry = debugfs_create_dir(O2NET_DEBUG_DIR, NULL);
+	if (o2net_dentry)
+		nst_dentry = debugfs_create_file(NST_DEBUG_NAME, mode,
+					o2net_dentry, NULL, &nst_seq_fops);
+	if (nst_dentry)
+		sc_dentry = debugfs_create_file(SC_DEBUG_NAME, mode,
+					o2net_dentry, NULL, &sc_seq_fops);
+	if (sc_dentry)
+		stats_dentry = debugfs_create_file(STATS_DEBUG_NAME, mode,
+					o2net_dentry, NULL, &stats_seq_fops);
+	if (stats_dentry)
+		nodes_dentry = debugfs_create_file(NODES_DEBUG_NAME, mode,
+					o2net_dentry, NULL, &nodes_fops);
+	if (nodes_dentry)
+		return 0;
+
+	o2net_debugfs_exit();
+	mlog_errno(-ENOMEM);
+	return -ENOMEM;
+}
+
+#endif	/* CONFIG_DEBUG_FS */

fs/ocfs2/cluster/nodemanager.c

@@ -0,0 +1,987 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2004, 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/slab.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/configfs.h>
+
+#include "tcp.h"
+#include "nodemanager.h"
+#include "heartbeat.h"
+#include "masklog.h"
+#include "sys.h"
+
+/* for now we operate under the assertion that there can be only one
+ * cluster active at a time.  Changing this will require trickling
+ * cluster references throughout where nodes are looked up */
+struct o2nm_cluster *o2nm_single_cluster = NULL;
+
+char *o2nm_fence_method_desc[O2NM_FENCE_METHODS] = {
+		"reset",	/* O2NM_FENCE_RESET */
+		"panic",	/* O2NM_FENCE_PANIC */
+};
+
+struct o2nm_node *o2nm_get_node_by_num(u8 node_num)
+{
+	struct o2nm_node *node = NULL;
+
+	if (node_num >= O2NM_MAX_NODES || o2nm_single_cluster == NULL)
+		goto out;
+
+	read_lock(&o2nm_single_cluster->cl_nodes_lock);
+	node = o2nm_single_cluster->cl_nodes[node_num];
+	if (node)
+		config_item_get(&node->nd_item);
+	read_unlock(&o2nm_single_cluster->cl_nodes_lock);
+out:
+	return node;
+}
+EXPORT_SYMBOL_GPL(o2nm_get_node_by_num);
+
+int o2nm_configured_node_map(unsigned long *map, unsigned bytes)
+{
+	struct o2nm_cluster *cluster = o2nm_single_cluster;
+
+	BUG_ON(bytes < (sizeof(cluster->cl_nodes_bitmap)));
+
+	if (cluster == NULL)
+		return -EINVAL;
+
+	read_lock(&cluster->cl_nodes_lock);
+	memcpy(map, cluster->cl_nodes_bitmap, sizeof(cluster->cl_nodes_bitmap));
+	read_unlock(&cluster->cl_nodes_lock);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(o2nm_configured_node_map);
+
+static struct o2nm_node *o2nm_node_ip_tree_lookup(struct o2nm_cluster *cluster,
+						  __be32 ip_needle,
+						  struct rb_node ***ret_p,
+						  struct rb_node **ret_parent)
+{
+	struct rb_node **p = &cluster->cl_node_ip_tree.rb_node;
+	struct rb_node *parent = NULL;
+	struct o2nm_node *node, *ret = NULL;
+
+	while (*p) {
+		int cmp;
+
+		parent = *p;
+		node = rb_entry(parent, struct o2nm_node, nd_ip_node);
+
+		cmp = memcmp(&ip_needle, &node->nd_ipv4_address,
+				sizeof(ip_needle));
+		if (cmp < 0)
+			p = &(*p)->rb_left;
+		else if (cmp > 0)
+			p = &(*p)->rb_right;
+		else {
+			ret = node;
+			break;
+		}
+	}
+
+	if (ret_p != NULL)
+		*ret_p = p;
+	if (ret_parent != NULL)
+		*ret_parent = parent;
+
+	return ret;
+}
+
+struct o2nm_node *o2nm_get_node_by_ip(__be32 addr)
+{
+	struct o2nm_node *node = NULL;
+	struct o2nm_cluster *cluster = o2nm_single_cluster;
+
+	if (cluster == NULL)
+		goto out;
+
+	read_lock(&cluster->cl_nodes_lock);
+	node = o2nm_node_ip_tree_lookup(cluster, addr, NULL, NULL);
+	if (node)
+		config_item_get(&node->nd_item);
+	read_unlock(&cluster->cl_nodes_lock);
+
+out:
+	return node;
+}
+EXPORT_SYMBOL_GPL(o2nm_get_node_by_ip);
+
+void o2nm_node_put(struct o2nm_node *node)
+{
+	config_item_put(&node->nd_item);
+}
+EXPORT_SYMBOL_GPL(o2nm_node_put);
+
+void o2nm_node_get(struct o2nm_node *node)
+{
+	config_item_get(&node->nd_item);
+}
+EXPORT_SYMBOL_GPL(o2nm_node_get);
+
+u8 o2nm_this_node(void)
+{
+	u8 node_num = O2NM_MAX_NODES;
+
+	if (o2nm_single_cluster && o2nm_single_cluster->cl_has_local)
+		node_num = o2nm_single_cluster->cl_local_node;
+
+	return node_num;
+}
+EXPORT_SYMBOL_GPL(o2nm_this_node);
+
+/* node configfs bits */
+
+static struct o2nm_cluster *to_o2nm_cluster(struct config_item *item)
+{
+	return item ?
+		container_of(to_config_group(item), struct o2nm_cluster,
+			     cl_group)
+		: NULL;
+}
+
+static struct o2nm_node *to_o2nm_node(struct config_item *item)
+{
+	return item ? container_of(item, struct o2nm_node, nd_item) : NULL;
+}
+
+static void o2nm_node_release(struct config_item *item)
+{
+	struct o2nm_node *node = to_o2nm_node(item);
+	kfree(node);
+}
+
+static ssize_t o2nm_node_num_read(struct o2nm_node *node, char *page)
+{
+	return sprintf(page, "%d\n", node->nd_num);
+}
+
+static struct o2nm_cluster *to_o2nm_cluster_from_node(struct o2nm_node *node)
+{
+	/* through the first node_set .parent
+	 * mycluster/nodes/mynode == o2nm_cluster->o2nm_node_group->o2nm_node */
+	return to_o2nm_cluster(node->nd_item.ci_parent->ci_parent);
+}
+
+enum {
+	O2NM_NODE_ATTR_NUM = 0,
+	O2NM_NODE_ATTR_PORT,
+	O2NM_NODE_ATTR_ADDRESS,
+	O2NM_NODE_ATTR_LOCAL,
+};
+
+static ssize_t o2nm_node_num_write(struct o2nm_node *node, const char *page,
+				   size_t count)
+{
+	struct o2nm_cluster *cluster = to_o2nm_cluster_from_node(node);
+	unsigned long tmp;
+	char *p = (char *)page;
+
+	tmp = simple_strtoul(p, &p, 0);
+	if (!p || (*p && (*p != '\n')))
+		return -EINVAL;
+
+	if (tmp >= O2NM_MAX_NODES)
+		return -ERANGE;
+
+	/* once we're in the cl_nodes tree networking can look us up by
+	 * node number and try to use our address and port attributes
+	 * to connect to this node.. make sure that they've been set
+	 * before writing the node attribute? */
+	if (!test_bit(O2NM_NODE_ATTR_ADDRESS, &node->nd_set_attributes) ||
+	    !test_bit(O2NM_NODE_ATTR_PORT, &node->nd_set_attributes))
+		return -EINVAL; /* XXX */
+
+	write_lock(&cluster->cl_nodes_lock);
+	if (cluster->cl_nodes[tmp])
+		p = NULL;
+	else  {
+		cluster->cl_nodes[tmp] = node;
+		node->nd_num = tmp;
+		set_bit(tmp, cluster->cl_nodes_bitmap);
+	}
+	write_unlock(&cluster->cl_nodes_lock);
+	if (p == NULL)
+		return -EEXIST;
+
+	return count;
+}
+static ssize_t o2nm_node_ipv4_port_read(struct o2nm_node *node, char *page)
+{
+	return sprintf(page, "%u\n", ntohs(node->nd_ipv4_port));
+}
+
+static ssize_t o2nm_node_ipv4_port_write(struct o2nm_node *node,
+					 const char *page, size_t count)
+{
+	unsigned long tmp;
+	char *p = (char *)page;
+
+	tmp = simple_strtoul(p, &p, 0);
+	if (!p || (*p && (*p != '\n')))
+		return -EINVAL;
+
+	if (tmp == 0)
+		return -EINVAL;
+	if (tmp >= (u16)-1)
+		return -ERANGE;
+
+	node->nd_ipv4_port = htons(tmp);
+
+	return count;
+}
+
+static ssize_t o2nm_node_ipv4_address_read(struct o2nm_node *node, char *page)
+{
+	return sprintf(page, "%pI4\n", &node->nd_ipv4_address);
+}
+
+static ssize_t o2nm_node_ipv4_address_write(struct o2nm_node *node,
+					    const char *page,
+					    size_t count)
+{
+	struct o2nm_cluster *cluster = to_o2nm_cluster_from_node(node);
+	int ret, i;
+	struct rb_node **p, *parent;
+	unsigned int octets[4];
+	__be32 ipv4_addr = 0;
+
+	ret = sscanf(page, "%3u.%3u.%3u.%3u", &octets[3], &octets[2],
+		     &octets[1], &octets[0]);
+	if (ret != 4)
+		return -EINVAL;
+
+	for (i = 0; i < ARRAY_SIZE(octets); i++) {
+		if (octets[i] > 255)
+			return -ERANGE;
+		be32_add_cpu(&ipv4_addr, octets[i] << (i * 8));
+	}
+
+	ret = 0;
+	write_lock(&cluster->cl_nodes_lock);
+	if (o2nm_node_ip_tree_lookup(cluster, ipv4_addr, &p, &parent))
+		ret = -EEXIST;
+	else {
+		rb_link_node(&node->nd_ip_node, parent, p);
+		rb_insert_color(&node->nd_ip_node, &cluster->cl_node_ip_tree);
+	}
+	write_unlock(&cluster->cl_nodes_lock);
+	if (ret)
+		return ret;
+
+	memcpy(&node->nd_ipv4_address, &ipv4_addr, sizeof(ipv4_addr));
+
+	return count;
+}
+
+static ssize_t o2nm_node_local_read(struct o2nm_node *node, char *page)
+{
+	return sprintf(page, "%d\n", node->nd_local);
+}
+
+static ssize_t o2nm_node_local_write(struct o2nm_node *node, const char *page,
+				     size_t count)
+{
+	struct o2nm_cluster *cluster = to_o2nm_cluster_from_node(node);
+	unsigned long tmp;
+	char *p = (char *)page;
+	ssize_t ret;
+
+	tmp = simple_strtoul(p, &p, 0);
+	if (!p || (*p && (*p != '\n')))
+		return -EINVAL;
+
+	tmp = !!tmp; /* boolean of whether this node wants to be local */
+
+	/* setting local turns on networking rx for now so we require having
+	 * set everything else first */
+	if (!test_bit(O2NM_NODE_ATTR_ADDRESS, &node->nd_set_attributes) ||
+	    !test_bit(O2NM_NODE_ATTR_NUM, &node->nd_set_attributes) ||
+	    !test_bit(O2NM_NODE_ATTR_PORT, &node->nd_set_attributes))
+		return -EINVAL; /* XXX */
+
+	/* the only failure case is trying to set a new local node
+	 * when a different one is already set */
+	if (tmp && tmp == cluster->cl_has_local &&
+	    cluster->cl_local_node != node->nd_num)
+		return -EBUSY;
+
+	/* bring up the rx thread if we're setting the new local node. */
+	if (tmp && !cluster->cl_has_local) {
+		ret = o2net_start_listening(node);
+		if (ret)
+			return ret;
+	}
+
+	if (!tmp && cluster->cl_has_local &&
+	    cluster->cl_local_node == node->nd_num) {
+		o2net_stop_listening(node);
+		cluster->cl_local_node = O2NM_INVALID_NODE_NUM;
+	}
+
+	node->nd_local = tmp;
+	if (node->nd_local) {
+		cluster->cl_has_local = tmp;
+		cluster->cl_local_node = node->nd_num;
+	}
+
+	return count;
+}
+
+struct o2nm_node_attribute {
+	struct configfs_attribute attr;
+	ssize_t (*show)(struct o2nm_node *, char *);
+	ssize_t (*store)(struct o2nm_node *, const char *, size_t);
+};
+
+static struct o2nm_node_attribute o2nm_node_attr_num = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "num",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= o2nm_node_num_read,
+	.store	= o2nm_node_num_write,
+};
+
+static struct o2nm_node_attribute o2nm_node_attr_ipv4_port = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "ipv4_port",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= o2nm_node_ipv4_port_read,
+	.store	= o2nm_node_ipv4_port_write,
+};
+
+static struct o2nm_node_attribute o2nm_node_attr_ipv4_address = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "ipv4_address",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= o2nm_node_ipv4_address_read,
+	.store	= o2nm_node_ipv4_address_write,
+};
+
+static struct o2nm_node_attribute o2nm_node_attr_local = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "local",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= o2nm_node_local_read,
+	.store	= o2nm_node_local_write,
+};
+
+static struct configfs_attribute *o2nm_node_attrs[] = {
+	[O2NM_NODE_ATTR_NUM] = &o2nm_node_attr_num.attr,
+	[O2NM_NODE_ATTR_PORT] = &o2nm_node_attr_ipv4_port.attr,
+	[O2NM_NODE_ATTR_ADDRESS] = &o2nm_node_attr_ipv4_address.attr,
+	[O2NM_NODE_ATTR_LOCAL] = &o2nm_node_attr_local.attr,
+	NULL,
+};
+
+static int o2nm_attr_index(struct configfs_attribute *attr)
+{
+	int i;
+	for (i = 0; i < ARRAY_SIZE(o2nm_node_attrs); i++) {
+		if (attr == o2nm_node_attrs[i])
+			return i;
+	}
+	BUG();
+	return 0;
+}
+
+static ssize_t o2nm_node_show(struct config_item *item,
+			      struct configfs_attribute *attr,
+			      char *page)
+{
+	struct o2nm_node *node = to_o2nm_node(item);
+	struct o2nm_node_attribute *o2nm_node_attr =
+		container_of(attr, struct o2nm_node_attribute, attr);
+	ssize_t ret = 0;
+
+	if (o2nm_node_attr->show)
+		ret = o2nm_node_attr->show(node, page);
+	return ret;
+}
+
+static ssize_t o2nm_node_store(struct config_item *item,
+			       struct configfs_attribute *attr,
+			       const char *page, size_t count)
+{
+	struct o2nm_node *node = to_o2nm_node(item);
+	struct o2nm_node_attribute *o2nm_node_attr =
+		container_of(attr, struct o2nm_node_attribute, attr);
+	ssize_t ret;
+	int attr_index = o2nm_attr_index(attr);
+
+	if (o2nm_node_attr->store == NULL) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	if (test_bit(attr_index, &node->nd_set_attributes))
+		return -EBUSY;
+
+	ret = o2nm_node_attr->store(node, page, count);
+	if (ret < count)
+		goto out;
+
+	set_bit(attr_index, &node->nd_set_attributes);
+out:
+	return ret;
+}
+
+static struct configfs_item_operations o2nm_node_item_ops = {
+	.release		= o2nm_node_release,
+	.show_attribute		= o2nm_node_show,
+	.store_attribute	= o2nm_node_store,
+};
+
+static struct config_item_type o2nm_node_type = {
+	.ct_item_ops	= &o2nm_node_item_ops,
+	.ct_attrs	= o2nm_node_attrs,
+	.ct_owner	= THIS_MODULE,
+};
+
+/* node set */
+
+struct o2nm_node_group {
+	struct config_group ns_group;
+	/* some stuff? */
+};
+
+#if 0
+static struct o2nm_node_group *to_o2nm_node_group(struct config_group *group)
+{
+	return group ?
+		container_of(group, struct o2nm_node_group, ns_group)
+		: NULL;
+}
+#endif
+
+struct o2nm_cluster_attribute {
+	struct configfs_attribute attr;
+	ssize_t (*show)(struct o2nm_cluster *, char *);
+	ssize_t (*store)(struct o2nm_cluster *, const char *, size_t);
+};
+
+static ssize_t o2nm_cluster_attr_write(const char *page, ssize_t count,
+                                       unsigned int *val)
+{
+	unsigned long tmp;
+	char *p = (char *)page;
+
+	tmp = simple_strtoul(p, &p, 0);
+	if (!p || (*p && (*p != '\n')))
+		return -EINVAL;
+
+	if (tmp == 0)
+		return -EINVAL;
+	if (tmp >= (u32)-1)
+		return -ERANGE;
+
+	*val = tmp;
+
+	return count;
+}
+
+static ssize_t o2nm_cluster_attr_idle_timeout_ms_read(
+	struct o2nm_cluster *cluster, char *page)
+{
+	return sprintf(page, "%u\n", cluster->cl_idle_timeout_ms);
+}
+
+static ssize_t o2nm_cluster_attr_idle_timeout_ms_write(
+	struct o2nm_cluster *cluster, const char *page, size_t count)
+{
+	ssize_t ret;
+	unsigned int val;
+
+	ret =  o2nm_cluster_attr_write(page, count, &val);
+
+	if (ret > 0) {
+		if (cluster->cl_idle_timeout_ms != val
+			&& o2net_num_connected_peers()) {
+			mlog(ML_NOTICE,
+			     "o2net: cannot change idle timeout after "
+			     "the first peer has agreed to it."
+			     "  %d connected peers\n",
+			     o2net_num_connected_peers());
+			ret = -EINVAL;
+		} else if (val <= cluster->cl_keepalive_delay_ms) {
+			mlog(ML_NOTICE, "o2net: idle timeout must be larger "
+			     "than keepalive delay\n");
+			ret = -EINVAL;
+		} else {
+			cluster->cl_idle_timeout_ms = val;
+		}
+	}
+
+	return ret;
+}
+
+static ssize_t o2nm_cluster_attr_keepalive_delay_ms_read(
+	struct o2nm_cluster *cluster, char *page)
+{
+	return sprintf(page, "%u\n", cluster->cl_keepalive_delay_ms);
+}
+
+static ssize_t o2nm_cluster_attr_keepalive_delay_ms_write(
+	struct o2nm_cluster *cluster, const char *page, size_t count)
+{
+	ssize_t ret;
+	unsigned int val;
+
+	ret =  o2nm_cluster_attr_write(page, count, &val);
+
+	if (ret > 0) {
+		if (cluster->cl_keepalive_delay_ms != val
+		    && o2net_num_connected_peers()) {
+			mlog(ML_NOTICE,
+			     "o2net: cannot change keepalive delay after"
+			     " the first peer has agreed to it."
+			     "  %d connected peers\n",
+			     o2net_num_connected_peers());
+			ret = -EINVAL;
+		} else if (val >= cluster->cl_idle_timeout_ms) {
+			mlog(ML_NOTICE, "o2net: keepalive delay must be "
+			     "smaller than idle timeout\n");
+			ret = -EINVAL;
+		} else {
+			cluster->cl_keepalive_delay_ms = val;
+		}
+	}
+
+	return ret;
+}
+
+static ssize_t o2nm_cluster_attr_reconnect_delay_ms_read(
+	struct o2nm_cluster *cluster, char *page)
+{
+	return sprintf(page, "%u\n", cluster->cl_reconnect_delay_ms);
+}
+
+static ssize_t o2nm_cluster_attr_reconnect_delay_ms_write(
+	struct o2nm_cluster *cluster, const char *page, size_t count)
+{
+	return o2nm_cluster_attr_write(page, count,
+	                               &cluster->cl_reconnect_delay_ms);
+}
+
+static ssize_t o2nm_cluster_attr_fence_method_read(
+	struct o2nm_cluster *cluster, char *page)
+{
+	ssize_t ret = 0;
+
+	if (cluster)
+		ret = sprintf(page, "%s\n",
+			      o2nm_fence_method_desc[cluster->cl_fence_method]);
+	return ret;
+}
+
+static ssize_t o2nm_cluster_attr_fence_method_write(
+	struct o2nm_cluster *cluster, const char *page, size_t count)
+{
+	unsigned int i;
+
+	if (page[count - 1] != '\n')
+		goto bail;
+
+	for (i = 0; i < O2NM_FENCE_METHODS; ++i) {
+		if (count != strlen(o2nm_fence_method_desc[i]) + 1)
+			continue;
+		if (strncasecmp(page, o2nm_fence_method_desc[i], count - 1))
+			continue;
+		if (cluster->cl_fence_method != i) {
+			printk(KERN_INFO "ocfs2: Changing fence method to %s\n",
+			       o2nm_fence_method_desc[i]);
+			cluster->cl_fence_method = i;
+		}
+		return count;
+	}
+
+bail:
+	return -EINVAL;
+}
+
+static struct o2nm_cluster_attribute o2nm_cluster_attr_idle_timeout_ms = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "idle_timeout_ms",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= o2nm_cluster_attr_idle_timeout_ms_read,
+	.store	= o2nm_cluster_attr_idle_timeout_ms_write,
+};
+
+static struct o2nm_cluster_attribute o2nm_cluster_attr_keepalive_delay_ms = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "keepalive_delay_ms",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= o2nm_cluster_attr_keepalive_delay_ms_read,
+	.store	= o2nm_cluster_attr_keepalive_delay_ms_write,
+};
+
+static struct o2nm_cluster_attribute o2nm_cluster_attr_reconnect_delay_ms = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "reconnect_delay_ms",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= o2nm_cluster_attr_reconnect_delay_ms_read,
+	.store	= o2nm_cluster_attr_reconnect_delay_ms_write,
+};
+
+static struct o2nm_cluster_attribute o2nm_cluster_attr_fence_method = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "fence_method",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= o2nm_cluster_attr_fence_method_read,
+	.store	= o2nm_cluster_attr_fence_method_write,
+};
+
+static struct configfs_attribute *o2nm_cluster_attrs[] = {
+	&o2nm_cluster_attr_idle_timeout_ms.attr,
+	&o2nm_cluster_attr_keepalive_delay_ms.attr,
+	&o2nm_cluster_attr_reconnect_delay_ms.attr,
+	&o2nm_cluster_attr_fence_method.attr,
+	NULL,
+};
+static ssize_t o2nm_cluster_show(struct config_item *item,
+                                 struct configfs_attribute *attr,
+                                 char *page)
+{
+	struct o2nm_cluster *cluster = to_o2nm_cluster(item);
+	struct o2nm_cluster_attribute *o2nm_cluster_attr =
+		container_of(attr, struct o2nm_cluster_attribute, attr);
+	ssize_t ret = 0;
+
+	if (o2nm_cluster_attr->show)
+		ret = o2nm_cluster_attr->show(cluster, page);
+	return ret;
+}
+
+static ssize_t o2nm_cluster_store(struct config_item *item,
+                                  struct configfs_attribute *attr,
+                                  const char *page, size_t count)
+{
+	struct o2nm_cluster *cluster = to_o2nm_cluster(item);
+	struct o2nm_cluster_attribute *o2nm_cluster_attr =
+		container_of(attr, struct o2nm_cluster_attribute, attr);
+	ssize_t ret;
+
+	if (o2nm_cluster_attr->store == NULL) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	ret = o2nm_cluster_attr->store(cluster, page, count);
+	if (ret < count)
+		goto out;
+out:
+	return ret;
+}
+
+static struct config_item *o2nm_node_group_make_item(struct config_group *group,
+						     const char *name)
+{
+	struct o2nm_node *node = NULL;
+
+	if (strlen(name) > O2NM_MAX_NAME_LEN)
+		return ERR_PTR(-ENAMETOOLONG);
+
+	node = kzalloc(sizeof(struct o2nm_node), GFP_KERNEL);
+	if (node == NULL)
+		return ERR_PTR(-ENOMEM);
+
+	strcpy(node->nd_name, name); /* use item.ci_namebuf instead? */
+	config_item_init_type_name(&node->nd_item, name, &o2nm_node_type);
+	spin_lock_init(&node->nd_lock);
+
+	mlog(ML_CLUSTER, "o2nm: Registering node %s\n", name);
+
+	return &node->nd_item;
+}
+
+static void o2nm_node_group_drop_item(struct config_group *group,
+				      struct config_item *item)
+{
+	struct o2nm_node *node = to_o2nm_node(item);
+	struct o2nm_cluster *cluster = to_o2nm_cluster(group->cg_item.ci_parent);
+
+	o2net_disconnect_node(node);
+
+	if (cluster->cl_has_local &&
+	    (cluster->cl_local_node == node->nd_num)) {
+		cluster->cl_has_local = 0;
+		cluster->cl_local_node = O2NM_INVALID_NODE_NUM;
+		o2net_stop_listening(node);
+	}
+
+	/* XXX call into net to stop this node from trading messages */
+
+	write_lock(&cluster->cl_nodes_lock);
+
+	/* XXX sloppy */
+	if (node->nd_ipv4_address)
+		rb_erase(&node->nd_ip_node, &cluster->cl_node_ip_tree);
+
+	/* nd_num might be 0 if the node number hasn't been set.. */
+	if (cluster->cl_nodes[node->nd_num] == node) {
+		cluster->cl_nodes[node->nd_num] = NULL;
+		clear_bit(node->nd_num, cluster->cl_nodes_bitmap);
+	}
+	write_unlock(&cluster->cl_nodes_lock);
+
+	mlog(ML_CLUSTER, "o2nm: Unregistered node %s\n",
+	     config_item_name(&node->nd_item));
+
+	config_item_put(item);
+}
+
+static struct configfs_group_operations o2nm_node_group_group_ops = {
+	.make_item	= o2nm_node_group_make_item,
+	.drop_item	= o2nm_node_group_drop_item,
+};
+
+static struct config_item_type o2nm_node_group_type = {
+	.ct_group_ops	= &o2nm_node_group_group_ops,
+	.ct_owner	= THIS_MODULE,
+};
+
+/* cluster */
+
+static void o2nm_cluster_release(struct config_item *item)
+{
+	struct o2nm_cluster *cluster = to_o2nm_cluster(item);
+
+	kfree(cluster->cl_group.default_groups);
+	kfree(cluster);
+}
+
+static struct configfs_item_operations o2nm_cluster_item_ops = {
+	.release	= o2nm_cluster_release,
+	.show_attribute		= o2nm_cluster_show,
+	.store_attribute	= o2nm_cluster_store,
+};
+
+static struct config_item_type o2nm_cluster_type = {
+	.ct_item_ops	= &o2nm_cluster_item_ops,
+	.ct_attrs	= o2nm_cluster_attrs,
+	.ct_owner	= THIS_MODULE,
+};
+
+/* cluster set */
+
+struct o2nm_cluster_group {
+	struct configfs_subsystem cs_subsys;
+	/* some stuff? */
+};
+
+#if 0
+static struct o2nm_cluster_group *to_o2nm_cluster_group(struct config_group *group)
+{
+	return group ?
+		container_of(to_configfs_subsystem(group), struct o2nm_cluster_group, cs_subsys)
+	       : NULL;
+}
+#endif
+
+static struct config_group *o2nm_cluster_group_make_group(struct config_group *group,
+							  const char *name)
+{
+	struct o2nm_cluster *cluster = NULL;
+	struct o2nm_node_group *ns = NULL;
+	struct config_group *o2hb_group = NULL, *ret = NULL;
+	void *defs = NULL;
+
+	/* this runs under the parent dir's i_mutex; there can be only
+	 * one caller in here at a time */
+	if (o2nm_single_cluster)
+		return ERR_PTR(-ENOSPC);
+
+	cluster = kzalloc(sizeof(struct o2nm_cluster), GFP_KERNEL);
+	ns = kzalloc(sizeof(struct o2nm_node_group), GFP_KERNEL);
+	defs = kcalloc(3, sizeof(struct config_group *), GFP_KERNEL);
+	o2hb_group = o2hb_alloc_hb_set();
+	if (cluster == NULL || ns == NULL || o2hb_group == NULL || defs == NULL)
+		goto out;
+
+	config_group_init_type_name(&cluster->cl_group, name,
+				    &o2nm_cluster_type);
+	config_group_init_type_name(&ns->ns_group, "node",
+				    &o2nm_node_group_type);
+
+	cluster->cl_group.default_groups = defs;
+	cluster->cl_group.default_groups[0] = &ns->ns_group;
+	cluster->cl_group.default_groups[1] = o2hb_group;
+	cluster->cl_group.default_groups[2] = NULL;
+	rwlock_init(&cluster->cl_nodes_lock);
+	cluster->cl_node_ip_tree = RB_ROOT;
+	cluster->cl_reconnect_delay_ms = O2NET_RECONNECT_DELAY_MS_DEFAULT;
+	cluster->cl_idle_timeout_ms    = O2NET_IDLE_TIMEOUT_MS_DEFAULT;
+	cluster->cl_keepalive_delay_ms = O2NET_KEEPALIVE_DELAY_MS_DEFAULT;
+	cluster->cl_fence_method       = O2NM_FENCE_RESET;
+
+	ret = &cluster->cl_group;
+	o2nm_single_cluster = cluster;
+
+out:
+	if (ret == NULL) {
+		kfree(cluster);
+		kfree(ns);
+		o2hb_free_hb_set(o2hb_group);
+		kfree(defs);
+		ret = ERR_PTR(-ENOMEM);
+	}
+
+	return ret;
+}
+
+static void o2nm_cluster_group_drop_item(struct config_group *group, struct config_item *item)
+{
+	struct o2nm_cluster *cluster = to_o2nm_cluster(item);
+	int i;
+	struct config_item *killme;
+
+	BUG_ON(o2nm_single_cluster != cluster);
+	o2nm_single_cluster = NULL;
+
+	for (i = 0; cluster->cl_group.default_groups[i]; i++) {
+		killme = &cluster->cl_group.default_groups[i]->cg_item;
+		cluster->cl_group.default_groups[i] = NULL;
+		config_item_put(killme);
+	}
+
+	config_item_put(item);
+}
+
+static struct configfs_group_operations o2nm_cluster_group_group_ops = {
+	.make_group	= o2nm_cluster_group_make_group,
+	.drop_item	= o2nm_cluster_group_drop_item,
+};
+
+static struct config_item_type o2nm_cluster_group_type = {
+	.ct_group_ops	= &o2nm_cluster_group_group_ops,
+	.ct_owner	= THIS_MODULE,
+};
+
+static struct o2nm_cluster_group o2nm_cluster_group = {
+	.cs_subsys = {
+		.su_group = {
+			.cg_item = {
+				.ci_namebuf = "cluster",
+				.ci_type = &o2nm_cluster_group_type,
+			},
+		},
+	},
+};
+
+int o2nm_depend_item(struct config_item *item)
+{
+	return configfs_depend_item(&o2nm_cluster_group.cs_subsys, item);
+}
+
+void o2nm_undepend_item(struct config_item *item)
+{
+	configfs_undepend_item(&o2nm_cluster_group.cs_subsys, item);
+}
+
+int o2nm_depend_this_node(void)
+{
+	int ret = 0;
+	struct o2nm_node *local_node;
+
+	local_node = o2nm_get_node_by_num(o2nm_this_node());
+	if (!local_node) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	ret = o2nm_depend_item(&local_node->nd_item);
+	o2nm_node_put(local_node);
+
+out:
+	return ret;
+}
+
+void o2nm_undepend_this_node(void)
+{
+	struct o2nm_node *local_node;
+
+	local_node = o2nm_get_node_by_num(o2nm_this_node());
+	BUG_ON(!local_node);
+
+	o2nm_undepend_item(&local_node->nd_item);
+	o2nm_node_put(local_node);
+}
+
+
+static void __exit exit_o2nm(void)
+{
+	/* XXX sync with hb callbacks and shut down hb? */
+	o2net_unregister_hb_callbacks();
+	configfs_unregister_subsystem(&o2nm_cluster_group.cs_subsys);
+	o2cb_sys_shutdown();
+
+	o2net_exit();
+	o2hb_exit();
+}
+
+static int __init init_o2nm(void)
+{
+	int ret = -1;
+
+	ret = o2hb_init();
+	if (ret)
+		goto out;
+
+	ret = o2net_init();
+	if (ret)
+		goto out_o2hb;
+
+	ret = o2net_register_hb_callbacks();
+	if (ret)
+		goto out_o2net;
+
+	config_group_init(&o2nm_cluster_group.cs_subsys.su_group);
+	mutex_init(&o2nm_cluster_group.cs_subsys.su_mutex);
+	ret = configfs_register_subsystem(&o2nm_cluster_group.cs_subsys);
+	if (ret) {
+		printk(KERN_ERR "nodemanager: Registration returned %d\n", ret);
+		goto out_callbacks;
+	}
+
+	ret = o2cb_sys_init();
+	if (!ret)
+		goto out;
+
+	configfs_unregister_subsystem(&o2nm_cluster_group.cs_subsys);
+out_callbacks:
+	o2net_unregister_hb_callbacks();
+out_o2net:
+	o2net_exit();
+out_o2hb:
+	o2hb_exit();
+out:
+	return ret;
+}
+
+MODULE_AUTHOR("Oracle");
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("OCFS2 cluster management");
+
+module_init(init_o2nm)
+module_exit(exit_o2nm)

fs/ocfs2/cluster/nodemanager.h

@@ -0,0 +1,88 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * nodemanager.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+#ifndef O2CLUSTER_NODEMANAGER_H
+#define O2CLUSTER_NODEMANAGER_H
+
+#include "ocfs2_nodemanager.h"
+
+/* This totally doesn't belong here. */
+#include <linux/configfs.h>
+#include <linux/rbtree.h>
+
+enum o2nm_fence_method {
+	O2NM_FENCE_RESET	= 0,
+	O2NM_FENCE_PANIC,
+	O2NM_FENCE_METHODS,	/* Number of fence methods */
+};
+
+struct o2nm_node {
+	spinlock_t		nd_lock;
+	struct config_item	nd_item;
+	char			nd_name[O2NM_MAX_NAME_LEN+1]; /* replace? */
+	__u8			nd_num;
+	/* only one address per node, as attributes, for now. */
+	__be32			nd_ipv4_address;
+	__be16			nd_ipv4_port;
+	struct rb_node		nd_ip_node;
+	/* there can be only one local node for now */
+	int			nd_local;
+
+	unsigned long		nd_set_attributes;
+};
+
+struct o2nm_cluster {
+	struct config_group	cl_group;
+	unsigned		cl_has_local:1;
+	u8			cl_local_node;
+	rwlock_t		cl_nodes_lock;
+	struct o2nm_node  	*cl_nodes[O2NM_MAX_NODES];
+	struct rb_root		cl_node_ip_tree;
+	unsigned int		cl_idle_timeout_ms;
+	unsigned int		cl_keepalive_delay_ms;
+	unsigned int		cl_reconnect_delay_ms;
+	enum o2nm_fence_method	cl_fence_method;
+
+	/* this bitmap is part of a hack for disk bitmap.. will go eventually. - zab */
+	unsigned long	cl_nodes_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
+};
+
+extern struct o2nm_cluster *o2nm_single_cluster;
+
+u8 o2nm_this_node(void);
+
+int o2nm_configured_node_map(unsigned long *map, unsigned bytes);
+struct o2nm_node *o2nm_get_node_by_num(u8 node_num);
+struct o2nm_node *o2nm_get_node_by_ip(__be32 addr);
+void o2nm_node_get(struct o2nm_node *node);
+void o2nm_node_put(struct o2nm_node *node);
+
+int o2nm_depend_item(struct config_item *item);
+void o2nm_undepend_item(struct config_item *item);
+int o2nm_depend_this_node(void);
+void o2nm_undepend_this_node(void);
+
+#endif /* O2CLUSTER_NODEMANAGER_H */

fs/ocfs2/cluster/ocfs2_heartbeat.h

@@ -0,0 +1,38 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ocfs2_heartbeat.h
+ *
+ * On-disk structures for ocfs2_heartbeat
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef _OCFS2_HEARTBEAT_H
+#define _OCFS2_HEARTBEAT_H
+
+struct o2hb_disk_heartbeat_block {
+	__le64 hb_seq;
+	__u8  hb_node;
+	__u8  hb_pad1[3];
+	__le32 hb_cksum;
+	__le64 hb_generation;
+	__le32 hb_dead_ms;
+};
+
+#endif /* _OCFS2_HEARTBEAT_H */

fs/ocfs2/cluster/ocfs2_nodemanager.h

@@ -0,0 +1,45 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ocfs2_nodemanager.h
+ *
+ * Header describing the interface between userspace and the kernel
+ * for the ocfs2_nodemanager module.
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+#ifndef _OCFS2_NODEMANAGER_H
+#define _OCFS2_NODEMANAGER_H
+
+#define O2NM_API_VERSION	5
+
+#define O2NM_MAX_NODES		255
+#define O2NM_INVALID_NODE_NUM	255
+
+/* host name, group name, cluster name all 64 bytes */
+#define O2NM_MAX_NAME_LEN        64    // __NEW_UTS_LEN
+
+/*
+ * Maximum number of global heartbeat regions allowed.
+ * **CAUTION**  Changing this number will break dlm compatibility.
+ */
+#define O2NM_MAX_REGIONS	32
+
+#endif /* _OCFS2_NODEMANAGER_H */

fs/ocfs2/cluster/quorum.c

@@ -0,0 +1,340 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ *
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+/* This quorum hack is only here until we transition to some more rational
+ * approach that is driven from userspace.  Honest.  No foolin'.
+ *
+ * Imagine two nodes lose network connectivity to each other but they're still
+ * up and operating in every other way.  Presumably a network timeout indicates
+ * that a node is broken and should be recovered.  They can't both recover each
+ * other and both carry on without serialising their access to the file system.
+ * They need to decide who is authoritative.  Now extend that problem to
+ * arbitrary groups of nodes losing connectivity between each other.
+ *
+ * So we declare that a node which has given up on connecting to a majority
+ * of nodes who are still heartbeating will fence itself.
+ *
+ * There are huge opportunities for races here.  After we give up on a node's
+ * connection we need to wait long enough to give heartbeat an opportunity
+ * to declare the node as truly dead.  We also need to be careful with the
+ * race between when we see a node start heartbeating and when we connect
+ * to it.
+ *
+ * So nodes that are in this transtion put a hold on the quorum decision
+ * with a counter.  As they fall out of this transition they drop the count
+ * and if they're the last, they fire off the decision.
+ */
+#include <linux/kernel.h>
+#include <linux/workqueue.h>
+#include <linux/reboot.h>
+
+#include "heartbeat.h"
+#include "nodemanager.h"
+#define MLOG_MASK_PREFIX ML_QUORUM
+#include "masklog.h"
+#include "quorum.h"
+
+static struct o2quo_state {
+	spinlock_t		qs_lock;
+	struct work_struct	qs_work;
+	int			qs_pending;
+	int			qs_heartbeating;
+	unsigned long		qs_hb_bm[BITS_TO_LONGS(O2NM_MAX_NODES)];
+	int			qs_connected;
+	unsigned long		qs_conn_bm[BITS_TO_LONGS(O2NM_MAX_NODES)];
+	int			qs_holds;
+	unsigned long		qs_hold_bm[BITS_TO_LONGS(O2NM_MAX_NODES)];
+} o2quo_state;
+
+/* this is horribly heavy-handed.  It should instead flip the file
+ * system RO and call some userspace script. */
+static void o2quo_fence_self(void)
+{
+	/* panic spins with interrupts enabled.  with preempt
+	 * threads can still schedule, etc, etc */
+	o2hb_stop_all_regions();
+
+	switch (o2nm_single_cluster->cl_fence_method) {
+	case O2NM_FENCE_PANIC:
+		panic("*** ocfs2 is very sorry to be fencing this system by "
+		      "panicing ***\n");
+		break;
+	default:
+		WARN_ON(o2nm_single_cluster->cl_fence_method >=
+			O2NM_FENCE_METHODS);
+	case O2NM_FENCE_RESET:
+		printk(KERN_ERR "*** ocfs2 is very sorry to be fencing this "
+		       "system by restarting ***\n");
+		emergency_restart();
+		break;
+	};
+}
+
+/* Indicate that a timeout occurred on a hearbeat region write. The
+ * other nodes in the cluster may consider us dead at that time so we
+ * want to "fence" ourselves so that we don't scribble on the disk
+ * after they think they've recovered us. This can't solve all
+ * problems related to writeout after recovery but this hack can at
+ * least close some of those gaps. When we have real fencing, this can
+ * go away as our node would be fenced externally before other nodes
+ * begin recovery. */
+void o2quo_disk_timeout(void)
+{
+	o2quo_fence_self();
+}
+
+static void o2quo_make_decision(struct work_struct *work)
+{
+	int quorum;
+	int lowest_hb, lowest_reachable = 0, fence = 0;
+	struct o2quo_state *qs = &o2quo_state;
+
+	spin_lock(&qs->qs_lock);
+
+	lowest_hb = find_first_bit(qs->qs_hb_bm, O2NM_MAX_NODES);
+	if (lowest_hb != O2NM_MAX_NODES)
+		lowest_reachable = test_bit(lowest_hb, qs->qs_conn_bm);
+
+	mlog(0, "heartbeating: %d, connected: %d, "
+	     "lowest: %d (%sreachable)\n", qs->qs_heartbeating,
+	     qs->qs_connected, lowest_hb, lowest_reachable ? "" : "un");
+
+	if (!test_bit(o2nm_this_node(), qs->qs_hb_bm) ||
+	    qs->qs_heartbeating == 1)
+		goto out;
+
+	if (qs->qs_heartbeating & 1) {
+		/* the odd numbered cluster case is straight forward --
+		 * if we can't talk to the majority we're hosed */
+		quorum = (qs->qs_heartbeating + 1)/2;
+		if (qs->qs_connected < quorum) {
+			mlog(ML_ERROR, "fencing this node because it is "
+			     "only connected to %u nodes and %u is needed "
+			     "to make a quorum out of %u heartbeating nodes\n",
+			     qs->qs_connected, quorum,
+			     qs->qs_heartbeating);
+			fence = 1;
+		}
+	} else {
+		/* the even numbered cluster adds the possibility of each half
+		 * of the cluster being able to talk amongst themselves.. in
+		 * that case we're hosed if we can't talk to the group that has
+		 * the lowest numbered node */
+		quorum = qs->qs_heartbeating / 2;
+		if (qs->qs_connected < quorum) {
+			mlog(ML_ERROR, "fencing this node because it is "
+			     "only connected to %u nodes and %u is needed "
+			     "to make a quorum out of %u heartbeating nodes\n",
+			     qs->qs_connected, quorum,
+			     qs->qs_heartbeating);
+			fence = 1;
+		}
+		else if ((qs->qs_connected == quorum) &&
+			 !lowest_reachable) {
+			mlog(ML_ERROR, "fencing this node because it is "
+			     "connected to a half-quorum of %u out of %u "
+			     "nodes which doesn't include the lowest active "
+			     "node %u\n", quorum, qs->qs_heartbeating,
+			     lowest_hb);
+			fence = 1;
+		}
+	}
+
+out:
+	if (fence) {
+		spin_unlock(&qs->qs_lock);
+		o2quo_fence_self();
+	} else {
+		mlog(ML_NOTICE, "not fencing this node, heartbeating: %d, "
+			"connected: %d, lowest: %d (%sreachable)\n",
+			qs->qs_heartbeating, qs->qs_connected, lowest_hb,
+			lowest_reachable ? "" : "un");
+		spin_unlock(&qs->qs_lock);
+
+	}
+
+}
+
+static void o2quo_set_hold(struct o2quo_state *qs, u8 node)
+{
+	assert_spin_locked(&qs->qs_lock);
+
+	if (!test_and_set_bit(node, qs->qs_hold_bm)) {
+		qs->qs_holds++;
+		mlog_bug_on_msg(qs->qs_holds == O2NM_MAX_NODES,
+			        "node %u\n", node);
+		mlog(0, "node %u, %d total\n", node, qs->qs_holds);
+	}
+}
+
+static void o2quo_clear_hold(struct o2quo_state *qs, u8 node)
+{
+	assert_spin_locked(&qs->qs_lock);
+
+	if (test_and_clear_bit(node, qs->qs_hold_bm)) {
+		mlog(0, "node %u, %d total\n", node, qs->qs_holds - 1);
+		if (--qs->qs_holds == 0) {
+			if (qs->qs_pending) {
+				qs->qs_pending = 0;
+				schedule_work(&qs->qs_work);
+			}
+		}
+		mlog_bug_on_msg(qs->qs_holds < 0, "node %u, holds %d\n",
+				node, qs->qs_holds);
+	}
+}
+
+/* as a node comes up we delay the quorum decision until we know the fate of
+ * the connection.  the hold will be droped in conn_up or hb_down.  it might be
+ * perpetuated by con_err until hb_down.  if we already have a conn, we might
+ * be dropping a hold that conn_up got. */
+void o2quo_hb_up(u8 node)
+{
+	struct o2quo_state *qs = &o2quo_state;
+
+	spin_lock(&qs->qs_lock);
+
+	qs->qs_heartbeating++;
+	mlog_bug_on_msg(qs->qs_heartbeating == O2NM_MAX_NODES,
+		        "node %u\n", node);
+	mlog_bug_on_msg(test_bit(node, qs->qs_hb_bm), "node %u\n", node);
+	set_bit(node, qs->qs_hb_bm);
+
+	mlog(0, "node %u, %d total\n", node, qs->qs_heartbeating);
+
+	if (!test_bit(node, qs->qs_conn_bm))
+		o2quo_set_hold(qs, node);
+	else
+		o2quo_clear_hold(qs, node);
+
+	spin_unlock(&qs->qs_lock);
+}
+
+/* hb going down releases any holds we might have had due to this node from
+ * conn_up, conn_err, or hb_up */
+void o2quo_hb_down(u8 node)
+{
+	struct o2quo_state *qs = &o2quo_state;
+
+	spin_lock(&qs->qs_lock);
+
+	qs->qs_heartbeating--;
+	mlog_bug_on_msg(qs->qs_heartbeating < 0,
+			"node %u, %d heartbeating\n",
+			node, qs->qs_heartbeating);
+	mlog_bug_on_msg(!test_bit(node, qs->qs_hb_bm), "node %u\n", node);
+	clear_bit(node, qs->qs_hb_bm);
+
+	mlog(0, "node %u, %d total\n", node, qs->qs_heartbeating);
+
+	o2quo_clear_hold(qs, node);
+
+	spin_unlock(&qs->qs_lock);
+}
+
+/* this tells us that we've decided that the node is still heartbeating
+ * even though we've lost it's conn.  it must only be called after conn_err
+ * and indicates that we must now make a quorum decision in the future,
+ * though we might be doing so after waiting for holds to drain.  Here
+ * we'll be dropping the hold from conn_err. */
+void o2quo_hb_still_up(u8 node)
+{
+	struct o2quo_state *qs = &o2quo_state;
+
+	spin_lock(&qs->qs_lock);
+
+	mlog(0, "node %u\n", node);
+
+	qs->qs_pending = 1;
+	o2quo_clear_hold(qs, node);
+
+	spin_unlock(&qs->qs_lock);
+}
+
+/* This is analogous to hb_up.  as a node's connection comes up we delay the
+ * quorum decision until we see it heartbeating.  the hold will be droped in
+ * hb_up or hb_down.  it might be perpetuated by con_err until hb_down.  if
+ * it's already heartbeating we might be dropping a hold that conn_up got.
+ * */
+void o2quo_conn_up(u8 node)
+{
+	struct o2quo_state *qs = &o2quo_state;
+
+	spin_lock(&qs->qs_lock);
+
+	qs->qs_connected++;
+	mlog_bug_on_msg(qs->qs_connected == O2NM_MAX_NODES,
+		        "node %u\n", node);
+	mlog_bug_on_msg(test_bit(node, qs->qs_conn_bm), "node %u\n", node);
+	set_bit(node, qs->qs_conn_bm);
+
+	mlog(0, "node %u, %d total\n", node, qs->qs_connected);
+
+	if (!test_bit(node, qs->qs_hb_bm))
+		o2quo_set_hold(qs, node);
+	else
+		o2quo_clear_hold(qs, node);
+
+	spin_unlock(&qs->qs_lock);
+}
+
+/* we've decided that we won't ever be connecting to the node again.  if it's
+ * still heartbeating we grab a hold that will delay decisions until either the
+ * node stops heartbeating from hb_down or the caller decides that the node is
+ * still up and calls still_up */
+void o2quo_conn_err(u8 node)
+{
+	struct o2quo_state *qs = &o2quo_state;
+
+	spin_lock(&qs->qs_lock);
+
+	if (test_bit(node, qs->qs_conn_bm)) {
+		qs->qs_connected--;
+		mlog_bug_on_msg(qs->qs_connected < 0,
+				"node %u, connected %d\n",
+				node, qs->qs_connected);
+
+		clear_bit(node, qs->qs_conn_bm);
+	}
+
+	mlog(0, "node %u, %d total\n", node, qs->qs_connected);
+
+	if (test_bit(node, qs->qs_hb_bm))
+		o2quo_set_hold(qs, node);
+
+	spin_unlock(&qs->qs_lock);
+}
+
+void o2quo_init(void)
+{
+	struct o2quo_state *qs = &o2quo_state;
+
+	spin_lock_init(&qs->qs_lock);
+	INIT_WORK(&qs->qs_work, o2quo_make_decision);
+}
+
+void o2quo_exit(void)
+{
+	struct o2quo_state *qs = &o2quo_state;
+
+	flush_work(&qs->qs_work);
+}

fs/ocfs2/cluster/quorum.h

@@ -0,0 +1,36 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+#ifndef O2CLUSTER_QUORUM_H
+#define O2CLUSTER_QUORUM_H
+
+void o2quo_init(void);
+void o2quo_exit(void);
+
+void o2quo_hb_up(u8 node);
+void o2quo_hb_down(u8 node);
+void o2quo_hb_still_up(u8 node);
+void o2quo_conn_up(u8 node);
+void o2quo_conn_err(u8 node);
+void o2quo_disk_timeout(void);
+
+#endif /* O2CLUSTER_QUORUM_H */

fs/ocfs2/cluster/sys.c

@@ -0,0 +1,82 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * sys.c
+ *
+ * OCFS2 cluster sysfs interface
+ *
+ * Copyright (C) 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation,
+ * version 2 of the License.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/kobject.h>
+#include <linux/sysfs.h>
+#include <linux/fs.h>
+
+#include "ocfs2_nodemanager.h"
+#include "masklog.h"
+#include "sys.h"
+
+
+static ssize_t version_show(struct kobject *kobj, struct kobj_attribute *attr,
+			    char *buf)
+{
+	return snprintf(buf, PAGE_SIZE, "%u\n", O2NM_API_VERSION);
+}
+static struct kobj_attribute attr_version =
+	__ATTR(interface_revision, S_IRUGO, version_show, NULL);
+
+static struct attribute *o2cb_attrs[] = {
+	&attr_version.attr,
+	NULL,
+};
+
+static struct attribute_group o2cb_attr_group = {
+	.attrs = o2cb_attrs,
+};
+
+static struct kset *o2cb_kset;
+
+void o2cb_sys_shutdown(void)
+{
+	mlog_sys_shutdown();
+	kset_unregister(o2cb_kset);
+}
+
+int o2cb_sys_init(void)
+{
+	int ret;
+
+	o2cb_kset = kset_create_and_add("o2cb", NULL, fs_kobj);
+	if (!o2cb_kset)
+		return -ENOMEM;
+
+	ret = sysfs_create_group(&o2cb_kset->kobj, &o2cb_attr_group);
+	if (ret)
+		goto error;
+
+	ret = mlog_sys_init(o2cb_kset);
+	if (ret)
+		goto error;
+	return 0;
+error:
+	kset_unregister(o2cb_kset);
+	return ret;
+}

fs/ocfs2/cluster/sys.h

@@ -0,0 +1,33 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * sys.h
+ *
+ * Function prototypes for o2cb sysfs interface
+ *
+ * Copyright (C) 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation,
+ * version 2 of the License.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+#ifndef O2CLUSTER_SYS_H
+#define O2CLUSTER_SYS_H
+
+void o2cb_sys_shutdown(void);
+int o2cb_sys_init(void);
+
+#endif /* O2CLUSTER_SYS_H */

fs/ocfs2/cluster/tcp.h

@@ -0,0 +1,155 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * tcp.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+#ifndef O2CLUSTER_TCP_H
+#define O2CLUSTER_TCP_H
+
+#include <linux/socket.h>
+#ifdef __KERNEL__
+#include <net/sock.h>
+#include <linux/tcp.h>
+#else
+#include <sys/socket.h>
+#endif
+#include <linux/inet.h>
+#include <linux/in.h>
+
+struct o2net_msg
+{
+	__be16 magic;
+	__be16 data_len;
+	__be16 msg_type;
+	__be16 pad1;
+	__be32 sys_status;
+	__be32 status;
+	__be32 key;
+	__be32 msg_num;
+	__u8  buf[0];
+};
+
+typedef int (o2net_msg_handler_func)(struct o2net_msg *msg, u32 len, void *data,
+				     void **ret_data);
+typedef void (o2net_post_msg_handler_func)(int status, void *data,
+					   void *ret_data);
+
+#define O2NET_MAX_PAYLOAD_BYTES  (4096 - sizeof(struct o2net_msg))
+
+/* same as hb delay, we're waiting for another node to recognize our hb */
+#define O2NET_RECONNECT_DELAY_MS_DEFAULT	2000
+
+#define O2NET_KEEPALIVE_DELAY_MS_DEFAULT	2000
+#define O2NET_IDLE_TIMEOUT_MS_DEFAULT		30000
+
+#define O2NET_TCP_USER_TIMEOUT			0x7fffffff
+
+/* TODO: figure this out.... */
+static inline int o2net_link_down(int err, struct socket *sock)
+{
+	if (sock) {
+		if (sock->sk->sk_state != TCP_ESTABLISHED &&
+	    	    sock->sk->sk_state != TCP_CLOSE_WAIT)
+			return 1;
+	}
+
+	if (err >= 0)
+		return 0;
+	switch (err) {
+		/* ????????????????????????? */
+		case -ERESTARTSYS:
+		case -EBADF:
+		/* When the server has died, an ICMP port unreachable
+		 * message prompts ECONNREFUSED. */
+		case -ECONNREFUSED:
+		case -ENOTCONN:
+		case -ECONNRESET:
+		case -EPIPE:
+			return 1;
+	}
+	return 0;
+}
+
+enum {
+	O2NET_DRIVER_UNINITED,
+	O2NET_DRIVER_READY,
+};
+
+int o2net_send_message(u32 msg_type, u32 key, void *data, u32 len,
+		       u8 target_node, int *status);
+int o2net_send_message_vec(u32 msg_type, u32 key, struct kvec *vec,
+			   size_t veclen, u8 target_node, int *status);
+
+int o2net_register_handler(u32 msg_type, u32 key, u32 max_len,
+			   o2net_msg_handler_func *func, void *data,
+			   o2net_post_msg_handler_func *post_func,
+			   struct list_head *unreg_list);
+void o2net_unregister_handler_list(struct list_head *list);
+
+void o2net_fill_node_map(unsigned long *map, unsigned bytes);
+
+struct o2nm_node;
+int o2net_register_hb_callbacks(void);
+void o2net_unregister_hb_callbacks(void);
+int o2net_start_listening(struct o2nm_node *node);
+void o2net_stop_listening(struct o2nm_node *node);
+void o2net_disconnect_node(struct o2nm_node *node);
+int o2net_num_connected_peers(void);
+
+int o2net_init(void);
+void o2net_exit(void);
+
+struct o2net_send_tracking;
+struct o2net_sock_container;
+
+#ifdef CONFIG_DEBUG_FS
+int o2net_debugfs_init(void);
+void o2net_debugfs_exit(void);
+void o2net_debug_add_nst(struct o2net_send_tracking *nst);
+void o2net_debug_del_nst(struct o2net_send_tracking *nst);
+void o2net_debug_add_sc(struct o2net_sock_container *sc);
+void o2net_debug_del_sc(struct o2net_sock_container *sc);
+#else
+static inline int o2net_debugfs_init(void)
+{
+	return 0;
+}
+static inline void o2net_debugfs_exit(void)
+{
+}
+static inline void o2net_debug_add_nst(struct o2net_send_tracking *nst)
+{
+}
+static inline void o2net_debug_del_nst(struct o2net_send_tracking *nst)
+{
+}
+static inline void o2net_debug_add_sc(struct o2net_sock_container *sc)
+{
+}
+static inline void o2net_debug_del_sc(struct o2net_sock_container *sc)
+{
+}
+#endif	/* CONFIG_DEBUG_FS */
+
+#endif /* O2CLUSTER_TCP_H */

fs/ocfs2/cluster/tcp_internal.h

@@ -0,0 +1,242 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef O2CLUSTER_TCP_INTERNAL_H
+#define O2CLUSTER_TCP_INTERNAL_H
+
+#define O2NET_MSG_MAGIC           ((u16)0xfa55)
+#define O2NET_MSG_STATUS_MAGIC    ((u16)0xfa56)
+#define O2NET_MSG_KEEP_REQ_MAGIC  ((u16)0xfa57)
+#define O2NET_MSG_KEEP_RESP_MAGIC ((u16)0xfa58)
+
+/* we're delaying our quorum decision so that heartbeat will have timed
+ * out truly dead nodes by the time we come around to making decisions
+ * on their number */
+#define O2NET_QUORUM_DELAY_MS	((o2hb_dead_threshold + 2) * O2HB_REGION_TIMEOUT_MS)
+
+/*
+ * This version number represents quite a lot, unfortunately.  It not
+ * only represents the raw network message protocol on the wire but also
+ * locking semantics of the file system using the protocol.  It should
+ * be somewhere else, I'm sure, but right now it isn't.
+ *
+ * With version 11, we separate out the filesystem locking portion.  The
+ * filesystem now has a major.minor version it negotiates.  Version 11
+ * introduces this negotiation to the o2dlm protocol, and as such the
+ * version here in tcp_internal.h should not need to be bumped for
+ * filesystem locking changes.
+ *
+ * New in version 11
+ * 	- Negotiation of filesystem locking in the dlm join.
+ *
+ * New in version 10:
+ * 	- Meta/data locks combined
+ *
+ * New in version 9:
+ * 	- All votes removed
+ *
+ * New in version 8:
+ * 	- Replace delete inode votes with a cluster lock
+ *
+ * New in version 7:
+ * 	- DLM join domain includes the live nodemap
+ *
+ * New in version 6:
+ * 	- DLM lockres remote refcount fixes.
+ *
+ * New in version 5:
+ * 	- Network timeout checking protocol
+ *
+ * New in version 4:
+ * 	- Remove i_generation from lock names for better stat performance.
+ *
+ * New in version 3:
+ * 	- Replace dentry votes with a cluster lock
+ *
+ * New in version 2:
+ * 	- full 64 bit i_size in the metadata lock lvbs
+ * 	- introduction of "rw" lock and pushing meta/data locking down
+ */
+#define O2NET_PROTOCOL_VERSION 11ULL
+struct o2net_handshake {
+	__be64	protocol_version;
+	__be64	connector_id;
+	__be32  o2hb_heartbeat_timeout_ms;
+	__be32  o2net_idle_timeout_ms;
+	__be32  o2net_keepalive_delay_ms;
+	__be32  o2net_reconnect_delay_ms;
+};
+
+struct o2net_node {
+	/* this is never called from int/bh */
+	spinlock_t			nn_lock;
+
+	/* set the moment an sc is allocated and a connect is started */
+	struct o2net_sock_container	*nn_sc;
+	/* _valid is only set after the handshake passes and tx can happen */
+	unsigned			nn_sc_valid:1;
+	/* if this is set tx just returns it */
+	int				nn_persistent_error;
+	/* It is only set to 1 after the idle time out. */
+	atomic_t			nn_timeout;
+
+	/* threads waiting for an sc to arrive wait on the wq for generation
+	 * to increase.  it is increased when a connecting socket succeeds
+	 * or fails or when an accepted socket is attached. */
+	wait_queue_head_t		nn_sc_wq;
+
+	struct idr			nn_status_idr;
+	struct list_head		nn_status_list;
+
+	/* connects are attempted from when heartbeat comes up until either hb
+	 * goes down, the node is unconfigured, no connect attempts succeed
+	 * before O2NET_CONN_IDLE_DELAY, or a connect succeeds.  connect_work
+	 * is queued from set_nn_state both from hb up and from itself if a
+	 * connect attempt fails and so can be self-arming.  shutdown is
+	 * careful to first mark the nn such that no connects will be attempted
+	 * before canceling delayed connect work and flushing the queue. */
+	struct delayed_work		nn_connect_work;
+	unsigned long			nn_last_connect_attempt;
+
+	/* this is queued as nodes come up and is canceled when a connection is
+	 * established.  this expiring gives up on the node and errors out
+	 * transmits */
+	struct delayed_work		nn_connect_expired;
+
+	/* after we give up on a socket we wait a while before deciding
+	 * that it is still heartbeating and that we should do some
+	 * quorum work */
+	struct delayed_work		nn_still_up;
+};
+
+struct o2net_sock_container {
+	struct kref		sc_kref;
+	/* the next two are valid for the life time of the sc */
+	struct socket		*sc_sock;
+	struct o2nm_node	*sc_node;
+
+	/* all of these sc work structs hold refs on the sc while they are
+	 * queued.  they should not be able to ref a freed sc.  the teardown
+	 * race is with o2net_wq destruction in o2net_stop_listening() */
+
+	/* rx and connect work are generated from socket callbacks.  sc
+	 * shutdown removes the callbacks and then flushes the work queue */
+	struct work_struct	sc_rx_work;
+	struct work_struct	sc_connect_work;
+	/* shutdown work is triggered in two ways.  the simple way is
+	 * for a code path calls ensure_shutdown which gets a lock, removes
+	 * the sc from the nn, and queues the work.  in this case the
+	 * work is single-shot.  the work is also queued from a sock
+	 * callback, though, and in this case the work will find the sc
+	 * still on the nn and will call ensure_shutdown itself.. this
+	 * ends up triggering the shutdown work again, though nothing
+	 * will be done in that second iteration.  so work queue teardown
+	 * has to be careful to remove the sc from the nn before waiting
+	 * on the work queue so that the shutdown work doesn't remove the
+	 * sc and rearm itself.
+	 */
+	struct work_struct	sc_shutdown_work;
+
+	struct timer_list	sc_idle_timeout;
+	struct delayed_work	sc_keepalive_work;
+
+	unsigned		sc_handshake_ok:1;
+
+	struct page 		*sc_page;
+	size_t			sc_page_off;
+
+	/* original handlers for the sockets */
+	void			(*sc_state_change)(struct sock *sk);
+	void			(*sc_data_ready)(struct sock *sk);
+
+	u32			sc_msg_key;
+	u16			sc_msg_type;
+
+#ifdef CONFIG_DEBUG_FS
+	struct list_head        sc_net_debug_item;
+	ktime_t			sc_tv_timer;
+	ktime_t			sc_tv_data_ready;
+	ktime_t			sc_tv_advance_start;
+	ktime_t			sc_tv_advance_stop;
+	ktime_t			sc_tv_func_start;
+	ktime_t			sc_tv_func_stop;
+#endif
+#ifdef CONFIG_OCFS2_FS_STATS
+	ktime_t			sc_tv_acquiry_total;
+	ktime_t			sc_tv_send_total;
+	ktime_t			sc_tv_status_total;
+	u32			sc_send_count;
+	u32			sc_recv_count;
+	ktime_t			sc_tv_process_total;
+#endif
+	struct mutex		sc_send_lock;
+};
+
+struct o2net_msg_handler {
+	struct rb_node		nh_node;
+	u32			nh_max_len;
+	u32			nh_msg_type;
+	u32			nh_key;
+	o2net_msg_handler_func	*nh_func;
+	o2net_msg_handler_func	*nh_func_data;
+	o2net_post_msg_handler_func
+				*nh_post_func;
+	struct kref		nh_kref;
+	struct list_head	nh_unregister_item;
+};
+
+enum o2net_system_error {
+	O2NET_ERR_NONE = 0,
+	O2NET_ERR_NO_HNDLR,
+	O2NET_ERR_OVERFLOW,
+	O2NET_ERR_DIED,
+	O2NET_ERR_MAX
+};
+
+struct o2net_status_wait {
+	enum o2net_system_error	ns_sys_status;
+	s32			ns_status;
+	int			ns_id;
+	wait_queue_head_t	ns_wq;
+	struct list_head	ns_node_item;
+};
+
+#ifdef CONFIG_DEBUG_FS
+/* just for state dumps */
+struct o2net_send_tracking {
+	struct list_head		st_net_debug_item;
+	struct task_struct		*st_task;
+	struct o2net_sock_container	*st_sc;
+	u32				st_id;
+	u32				st_msg_type;
+	u32				st_msg_key;
+	u8				st_node;
+	ktime_t				st_sock_time;
+	ktime_t				st_send_time;
+	ktime_t				st_status_time;
+};
+#else
+struct o2net_send_tracking {
+	u32	dummy;
+};
+#endif	/* CONFIG_DEBUG_FS */
+
+#endif /* O2CLUSTER_TCP_INTERNAL_H */

fs/ocfs2/dcache.c

@@ -0,0 +1,476 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dcache.c
+ *
+ * dentry cache handling code
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/namei.h>
+
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "dcache.h"
+#include "dlmglue.h"
+#include "file.h"
+#include "inode.h"
+#include "ocfs2_trace.h"
+
+void ocfs2_dentry_attach_gen(struct dentry *dentry)
+{
+	unsigned long gen =
+		OCFS2_I(dentry->d_parent->d_inode)->ip_dir_lock_gen;
+	BUG_ON(dentry->d_inode);
+	dentry->d_fsdata = (void *)gen;
+}
+
+
+static int ocfs2_dentry_revalidate(struct dentry *dentry, unsigned int flags)
+{
+	struct inode *inode;
+	int ret = 0;    /* if all else fails, just return false */
+	struct ocfs2_super *osb;
+
+	if (flags & LOOKUP_RCU)
+		return -ECHILD;
+
+	inode = dentry->d_inode;
+	osb = OCFS2_SB(dentry->d_sb);
+
+	trace_ocfs2_dentry_revalidate(dentry, dentry->d_name.len,
+				      dentry->d_name.name);
+
+	/* For a negative dentry -
+	 * check the generation number of the parent and compare with the
+	 * one stored in the inode.
+	 */
+	if (inode == NULL) {
+		unsigned long gen = (unsigned long) dentry->d_fsdata;
+		unsigned long pgen;
+		spin_lock(&dentry->d_lock);
+		pgen = OCFS2_I(dentry->d_parent->d_inode)->ip_dir_lock_gen;
+		spin_unlock(&dentry->d_lock);
+		trace_ocfs2_dentry_revalidate_negative(dentry->d_name.len,
+						       dentry->d_name.name,
+						       pgen, gen);
+		if (gen != pgen)
+			goto bail;
+		goto valid;
+	}
+
+	BUG_ON(!osb);
+
+	if (inode == osb->root_inode || is_bad_inode(inode))
+		goto bail;
+
+	spin_lock(&OCFS2_I(inode)->ip_lock);
+	/* did we or someone else delete this inode? */
+	if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_DELETED) {
+		spin_unlock(&OCFS2_I(inode)->ip_lock);
+		trace_ocfs2_dentry_revalidate_delete(
+				(unsigned long long)OCFS2_I(inode)->ip_blkno);
+		goto bail;
+	}
+	spin_unlock(&OCFS2_I(inode)->ip_lock);
+
+	/*
+	 * We don't need a cluster lock to test this because once an
+	 * inode nlink hits zero, it never goes back.
+	 */
+	if (inode->i_nlink == 0) {
+		trace_ocfs2_dentry_revalidate_orphaned(
+			(unsigned long long)OCFS2_I(inode)->ip_blkno,
+			S_ISDIR(inode->i_mode));
+		goto bail;
+	}
+
+	/*
+	 * If the last lookup failed to create dentry lock, let us
+	 * redo it.
+	 */
+	if (!dentry->d_fsdata) {
+		trace_ocfs2_dentry_revalidate_nofsdata(
+				(unsigned long long)OCFS2_I(inode)->ip_blkno);
+		goto bail;
+	}
+
+valid:
+	ret = 1;
+
+bail:
+	trace_ocfs2_dentry_revalidate_ret(ret);
+	return ret;
+}
+
+static int ocfs2_match_dentry(struct dentry *dentry,
+			      u64 parent_blkno,
+			      int skip_unhashed)
+{
+	struct inode *parent;
+
+	/*
+	 * ocfs2_lookup() does a d_splice_alias() _before_ attaching
+	 * to the lock data, so we skip those here, otherwise
+	 * ocfs2_dentry_attach_lock() will get its original dentry
+	 * back.
+	 */
+	if (!dentry->d_fsdata)
+		return 0;
+
+	if (!dentry->d_parent)
+		return 0;
+
+	if (skip_unhashed && d_unhashed(dentry))
+		return 0;
+
+	parent = dentry->d_parent->d_inode;
+	/* Negative parent dentry? */
+	if (!parent)
+		return 0;
+
+	/* Name is in a different directory. */
+	if (OCFS2_I(parent)->ip_blkno != parent_blkno)
+		return 0;
+
+	return 1;
+}
+
+/*
+ * Walk the inode alias list, and find a dentry which has a given
+ * parent. ocfs2_dentry_attach_lock() wants to find _any_ alias as it
+ * is looking for a dentry_lock reference. The downconvert thread is
+ * looking to unhash aliases, so we allow it to skip any that already
+ * have that property.
+ */
+struct dentry *ocfs2_find_local_alias(struct inode *inode,
+				      u64 parent_blkno,
+				      int skip_unhashed)
+{
+	struct dentry *dentry;
+
+	spin_lock(&inode->i_lock);
+	hlist_for_each_entry(dentry, &inode->i_dentry, d_u.d_alias) {
+		spin_lock(&dentry->d_lock);
+		if (ocfs2_match_dentry(dentry, parent_blkno, skip_unhashed)) {
+			trace_ocfs2_find_local_alias(dentry->d_name.len,
+						     dentry->d_name.name);
+
+			dget_dlock(dentry);
+			spin_unlock(&dentry->d_lock);
+			spin_unlock(&inode->i_lock);
+			return dentry;
+		}
+		spin_unlock(&dentry->d_lock);
+	}
+	spin_unlock(&inode->i_lock);
+	return NULL;
+}
+
+DEFINE_SPINLOCK(dentry_attach_lock);
+
+/*
+ * Attach this dentry to a cluster lock.
+ *
+ * Dentry locks cover all links in a given directory to a particular
+ * inode. We do this so that ocfs2 can build a lock name which all
+ * nodes in the cluster can agree on at all times. Shoving full names
+ * in the cluster lock won't work due to size restrictions. Covering
+ * links inside of a directory is a good compromise because it still
+ * allows us to use the parent directory lock to synchronize
+ * operations.
+ *
+ * Call this function with the parent dir semaphore and the parent dir
+ * cluster lock held.
+ *
+ * The dir semaphore will protect us from having to worry about
+ * concurrent processes on our node trying to attach a lock at the
+ * same time.
+ *
+ * The dir cluster lock (held at either PR or EX mode) protects us
+ * from unlink and rename on other nodes.
+ *
+ * A dput() can happen asynchronously due to pruning, so we cover
+ * attaching and detaching the dentry lock with a
+ * dentry_attach_lock.
+ *
+ * A node which has done lookup on a name retains a protected read
+ * lock until final dput. If the user requests and unlink or rename,
+ * the protected read is upgraded to an exclusive lock. Other nodes
+ * who have seen the dentry will then be informed that they need to
+ * downgrade their lock, which will involve d_delete on the
+ * dentry. This happens in ocfs2_dentry_convert_worker().
+ */
+int ocfs2_dentry_attach_lock(struct dentry *dentry,
+			     struct inode *inode,
+			     u64 parent_blkno)
+{
+	int ret;
+	struct dentry *alias;
+	struct ocfs2_dentry_lock *dl = dentry->d_fsdata;
+
+	trace_ocfs2_dentry_attach_lock(dentry->d_name.len, dentry->d_name.name,
+				       (unsigned long long)parent_blkno, dl);
+
+	/*
+	 * Negative dentry. We ignore these for now.
+	 *
+	 * XXX: Could we can improve ocfs2_dentry_revalidate() by
+	 * tracking these?
+	 */
+	if (!inode)
+		return 0;
+
+	if (!dentry->d_inode && dentry->d_fsdata) {
+		/* Converting a negative dentry to positive
+		   Clear dentry->d_fsdata */
+		dentry->d_fsdata = dl = NULL;
+	}
+
+	if (dl) {
+		mlog_bug_on_msg(dl->dl_parent_blkno != parent_blkno,
+				" \"%.*s\": old parent: %llu, new: %llu\n",
+				dentry->d_name.len, dentry->d_name.name,
+				(unsigned long long)parent_blkno,
+				(unsigned long long)dl->dl_parent_blkno);
+		return 0;
+	}
+
+	alias = ocfs2_find_local_alias(inode, parent_blkno, 0);
+	if (alias) {
+		/*
+		 * Great, an alias exists, which means we must have a
+		 * dentry lock already. We can just grab the lock off
+		 * the alias and add it to the list.
+		 *
+		 * We're depending here on the fact that this dentry
+		 * was found and exists in the dcache and so must have
+		 * a reference to the dentry_lock because we can't
+		 * race creates. Final dput() cannot happen on it
+		 * since we have it pinned, so our reference is safe.
+		 */
+		dl = alias->d_fsdata;
+		mlog_bug_on_msg(!dl, "parent %llu, ino %llu\n",
+				(unsigned long long)parent_blkno,
+				(unsigned long long)OCFS2_I(inode)->ip_blkno);
+
+		mlog_bug_on_msg(dl->dl_parent_blkno != parent_blkno,
+				" \"%.*s\": old parent: %llu, new: %llu\n",
+				dentry->d_name.len, dentry->d_name.name,
+				(unsigned long long)parent_blkno,
+				(unsigned long long)dl->dl_parent_blkno);
+
+		trace_ocfs2_dentry_attach_lock_found(dl->dl_lockres.l_name,
+				(unsigned long long)parent_blkno,
+				(unsigned long long)OCFS2_I(inode)->ip_blkno);
+
+		goto out_attach;
+	}
+
+	/*
+	 * There are no other aliases
+	 */
+	dl = kmalloc(sizeof(*dl), GFP_NOFS);
+	if (!dl) {
+		ret = -ENOMEM;
+		mlog_errno(ret);
+		return ret;
+	}
+
+	dl->dl_count = 0;
+	/*
+	 * Does this have to happen below, for all attaches, in case
+	 * the struct inode gets blown away by the downconvert thread?
+	 */
+	dl->dl_inode = igrab(inode);
+	dl->dl_parent_blkno = parent_blkno;
+	ocfs2_dentry_lock_res_init(dl, parent_blkno, inode);
+
+out_attach:
+	spin_lock(&dentry_attach_lock);
+	dentry->d_fsdata = dl;
+	dl->dl_count++;
+	spin_unlock(&dentry_attach_lock);
+
+	/*
+	 * This actually gets us our PRMODE level lock. From now on,
+	 * we'll have a notification if one of these names is
+	 * destroyed on another node.
+	 */
+	ret = ocfs2_dentry_lock(dentry, 0);
+	if (!ret)
+		ocfs2_dentry_unlock(dentry, 0);
+	else
+		mlog_errno(ret);
+
+	/*
+	 * In case of error, manually free the allocation and do the iput().
+	 * We need to do this because error here means no d_instantiate(),
+	 * which means iput() will not be called during dput(dentry).
+	 */
+	if (ret < 0 && !alias) {
+		ocfs2_lock_res_free(&dl->dl_lockres);
+		BUG_ON(dl->dl_count != 1);
+		spin_lock(&dentry_attach_lock);
+		dentry->d_fsdata = NULL;
+		spin_unlock(&dentry_attach_lock);
+		kfree(dl);
+		iput(inode);
+	}
+
+	dput(alias);
+
+	return ret;
+}
+
+/*
+ * ocfs2_dentry_iput() and friends.
+ *
+ * At this point, our particular dentry is detached from the inodes
+ * alias list, so there's no way that the locking code can find it.
+ *
+ * The interesting stuff happens when we determine that our lock needs
+ * to go away because this is the last subdir alias in the
+ * system. This function needs to handle a couple things:
+ *
+ * 1) Synchronizing lock shutdown with the downconvert threads. This
+ *    is already handled for us via the lockres release drop function
+ *    called in ocfs2_release_dentry_lock()
+ *
+ * 2) A race may occur when we're doing our lock shutdown and
+ *    another process wants to create a new dentry lock. Right now we
+ *    let them race, which means that for a very short while, this
+ *    node might have two locks on a lock resource. This should be a
+ *    problem though because one of them is in the process of being
+ *    thrown out.
+ */
+static void ocfs2_drop_dentry_lock(struct ocfs2_super *osb,
+				   struct ocfs2_dentry_lock *dl)
+{
+	iput(dl->dl_inode);
+	ocfs2_simple_drop_lockres(osb, &dl->dl_lockres);
+	ocfs2_lock_res_free(&dl->dl_lockres);
+	kfree(dl);
+}
+
+void ocfs2_dentry_lock_put(struct ocfs2_super *osb,
+			   struct ocfs2_dentry_lock *dl)
+{
+	int unlock = 0;
+
+	BUG_ON(dl->dl_count == 0);
+
+	spin_lock(&dentry_attach_lock);
+	dl->dl_count--;
+	unlock = !dl->dl_count;
+	spin_unlock(&dentry_attach_lock);
+
+	if (unlock)
+		ocfs2_drop_dentry_lock(osb, dl);
+}
+
+static void ocfs2_dentry_iput(struct dentry *dentry, struct inode *inode)
+{
+	struct ocfs2_dentry_lock *dl = dentry->d_fsdata;
+
+	if (!dl) {
+		/*
+		 * No dentry lock is ok if we're disconnected or
+		 * unhashed.
+		 */
+		if (!(dentry->d_flags & DCACHE_DISCONNECTED) &&
+		    !d_unhashed(dentry)) {
+			unsigned long long ino = 0ULL;
+			if (inode)
+				ino = (unsigned long long)OCFS2_I(inode)->ip_blkno;
+			mlog(ML_ERROR, "Dentry is missing cluster lock. "
+			     "inode: %llu, d_flags: 0x%x, d_name: %.*s\n",
+			     ino, dentry->d_flags, dentry->d_name.len,
+			     dentry->d_name.name);
+		}
+
+		goto out;
+	}
+
+	mlog_bug_on_msg(dl->dl_count == 0, "dentry: %.*s, count: %u\n",
+			dentry->d_name.len, dentry->d_name.name,
+			dl->dl_count);
+
+	ocfs2_dentry_lock_put(OCFS2_SB(dentry->d_sb), dl);
+
+out:
+	iput(inode);
+}
+
+/*
+ * d_move(), but keep the locks in sync.
+ *
+ * When we are done, "dentry" will have the parent dir and name of
+ * "target", which will be thrown away.
+ *
+ * We manually update the lock of "dentry" if need be.
+ *
+ * "target" doesn't have it's dentry lock touched - we allow the later
+ * dput() to handle this for us.
+ *
+ * This is called during ocfs2_rename(), while holding parent
+ * directory locks. The dentries have already been deleted on other
+ * nodes via ocfs2_remote_dentry_delete().
+ *
+ * Normally, the VFS handles the d_move() for the file system, after
+ * the ->rename() callback. OCFS2 wants to handle this internally, so
+ * the new lock can be created atomically with respect to the cluster.
+ */
+void ocfs2_dentry_move(struct dentry *dentry, struct dentry *target,
+		       struct inode *old_dir, struct inode *new_dir)
+{
+	int ret;
+	struct ocfs2_super *osb = OCFS2_SB(old_dir->i_sb);
+	struct inode *inode = dentry->d_inode;
+
+	/*
+	 * Move within the same directory, so the actual lock info won't
+	 * change.
+	 *
+	 * XXX: Is there any advantage to dropping the lock here?
+	 */
+	if (old_dir == new_dir)
+		goto out_move;
+
+	ocfs2_dentry_lock_put(osb, dentry->d_fsdata);
+
+	dentry->d_fsdata = NULL;
+	ret = ocfs2_dentry_attach_lock(dentry, inode, OCFS2_I(new_dir)->ip_blkno);
+	if (ret)
+		mlog_errno(ret);
+
+out_move:
+	d_move(dentry, target);
+}
+
+const struct dentry_operations ocfs2_dentry_ops = {
+	.d_revalidate		= ocfs2_dentry_revalidate,
+	.d_iput			= ocfs2_dentry_iput,
+};

fs/ocfs2/dcache.h

@@ -0,0 +1,59 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dcache.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_DCACHE_H
+#define OCFS2_DCACHE_H
+
+extern const struct dentry_operations ocfs2_dentry_ops;
+
+struct ocfs2_dentry_lock {
+	unsigned int		dl_count;
+	u64			dl_parent_blkno;
+
+	/*
+	 * The ocfs2_dentry_lock keeps an inode reference until
+	 * dl_lockres has been destroyed. This is usually done in
+	 * ->d_iput() anyway, so there should be minimal impact.
+	 */
+	struct inode		*dl_inode;
+	struct ocfs2_lock_res	dl_lockres;
+};
+
+int ocfs2_dentry_attach_lock(struct dentry *dentry, struct inode *inode,
+			     u64 parent_blkno);
+
+void ocfs2_dentry_lock_put(struct ocfs2_super *osb,
+			   struct ocfs2_dentry_lock *dl);
+
+struct dentry *ocfs2_find_local_alias(struct inode *inode, u64 parent_blkno,
+				      int skip_unhashed);
+
+void ocfs2_dentry_move(struct dentry *dentry, struct dentry *target,
+		       struct inode *old_dir, struct inode *new_dir);
+
+extern spinlock_t dentry_attach_lock;
+void ocfs2_dentry_attach_gen(struct dentry *dentry);
+
+#endif /* OCFS2_DCACHE_H */

fs/ocfs2/dir.h

@@ -0,0 +1,116 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dir.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_DIR_H
+#define OCFS2_DIR_H
+
+struct ocfs2_dx_hinfo {
+	u32	major_hash;
+	u32	minor_hash;
+};
+
+struct ocfs2_dir_lookup_result {
+	struct buffer_head		*dl_leaf_bh;	/* Unindexed leaf
+							 * block */
+	struct ocfs2_dir_entry		*dl_entry;	/* Target dirent in
+							 * unindexed leaf */
+
+	struct buffer_head		*dl_dx_root_bh;	/* Root of indexed
+							 * tree */
+
+	struct buffer_head		*dl_dx_leaf_bh;	/* Indexed leaf block */
+	struct ocfs2_dx_entry		*dl_dx_entry;	/* Target dx_entry in
+							 * indexed leaf */
+	struct ocfs2_dx_hinfo		dl_hinfo;	/* Name hash results */
+
+	struct buffer_head		*dl_prev_leaf_bh;/* Previous entry in
+							  * dir free space
+							  * list. NULL if
+							  * previous entry is
+							  * dx root block. */
+};
+
+void ocfs2_free_dir_lookup_result(struct ocfs2_dir_lookup_result *res);
+
+int ocfs2_find_entry(const char *name, int namelen,
+		     struct inode *dir,
+		     struct ocfs2_dir_lookup_result *lookup);
+int ocfs2_delete_entry(handle_t *handle,
+		       struct inode *dir,
+		       struct ocfs2_dir_lookup_result *res);
+int __ocfs2_add_entry(handle_t *handle,
+		      struct inode *dir,
+		      const char *name, int namelen,
+		      struct inode *inode, u64 blkno,
+		      struct buffer_head *parent_fe_bh,
+		      struct ocfs2_dir_lookup_result *lookup);
+static inline int ocfs2_add_entry(handle_t *handle,
+				  struct dentry *dentry,
+				  struct inode *inode, u64 blkno,
+				  struct buffer_head *parent_fe_bh,
+				  struct ocfs2_dir_lookup_result *lookup)
+{
+	return __ocfs2_add_entry(handle, dentry->d_parent->d_inode,
+				 dentry->d_name.name, dentry->d_name.len,
+				 inode, blkno, parent_fe_bh, lookup);
+}
+int ocfs2_update_entry(struct inode *dir, handle_t *handle,
+		       struct ocfs2_dir_lookup_result *res,
+		       struct inode *new_entry_inode);
+
+int ocfs2_check_dir_for_entry(struct inode *dir,
+			      const char *name,
+			      int namelen);
+int ocfs2_empty_dir(struct inode *inode);
+
+int ocfs2_find_files_on_disk(const char *name,
+			     int namelen,
+			     u64 *blkno,
+			     struct inode *inode,
+			     struct ocfs2_dir_lookup_result *res);
+int ocfs2_lookup_ino_from_name(struct inode *dir, const char *name,
+			       int namelen, u64 *blkno);
+int ocfs2_readdir(struct file *file, struct dir_context *ctx);
+int ocfs2_dir_foreach(struct inode *inode, struct dir_context *ctx);
+int ocfs2_prepare_dir_for_insert(struct ocfs2_super *osb,
+				 struct inode *dir,
+				 struct buffer_head *parent_fe_bh,
+				 const char *name,
+				 int namelen,
+				 struct ocfs2_dir_lookup_result *lookup);
+struct ocfs2_alloc_context;
+int ocfs2_fill_new_dir(struct ocfs2_super *osb,
+		       handle_t *handle,
+		       struct inode *parent,
+		       struct inode *inode,
+		       struct buffer_head *fe_bh,
+		       struct ocfs2_alloc_context *data_ac,
+		       struct ocfs2_alloc_context *meta_ac);
+
+int ocfs2_dx_dir_truncate(struct inode *dir, struct buffer_head *di_bh);
+
+struct ocfs2_dir_block_trailer *ocfs2_dir_trailer_from_size(int blocksize,
+							    void *data);
+#endif /* OCFS2_DIR_H */

fs/ocfs2/dlm/Makefile

@@ -0,0 +1,7 @@
+ccflags-y := -Ifs/ocfs2
+
+obj-$(CONFIG_OCFS2_FS_O2CB) += ocfs2_dlm.o
+
+ocfs2_dlm-objs := dlmdomain.o dlmdebug.o dlmthread.o dlmrecovery.o \
+	dlmmaster.o dlmast.o dlmconvert.o dlmlock.o dlmunlock.o
+

fs/ocfs2/dlm/dlmapi.h

@@ -0,0 +1,220 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dlmapi.h
+ *
+ * externally exported dlm interfaces
+ *
+ * Copyright (C) 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+#ifndef DLMAPI_H
+#define DLMAPI_H
+
+struct dlm_lock;
+struct dlm_ctxt;
+
+/* NOTE: changes made to this enum should be reflected in dlmdebug.c */
+enum dlm_status {
+	DLM_NORMAL = 0,           /*  0: request in progress */
+	DLM_GRANTED,              /*  1: request granted */
+	DLM_DENIED,               /*  2: request denied */
+	DLM_DENIED_NOLOCKS,       /*  3: request denied, out of system resources */
+	DLM_WORKING,              /*  4: async request in progress */
+	DLM_BLOCKED,              /*  5: lock request blocked */
+	DLM_BLOCKED_ORPHAN,       /*  6: lock request blocked by a orphan lock*/
+	DLM_DENIED_GRACE_PERIOD,  /*  7: topological change in progress */
+	DLM_SYSERR,               /*  8: system error */
+	DLM_NOSUPPORT,            /*  9: unsupported */
+	DLM_CANCELGRANT,          /* 10: can't cancel convert: already granted */
+	DLM_IVLOCKID,             /* 11: bad lockid */
+	DLM_SYNC,                 /* 12: synchronous request granted */
+	DLM_BADTYPE,              /* 13: bad resource type */
+	DLM_BADRESOURCE,          /* 14: bad resource handle */
+	DLM_MAXHANDLES,           /* 15: no more resource handles */
+	DLM_NOCLINFO,             /* 16: can't contact cluster manager */
+	DLM_NOLOCKMGR,            /* 17: can't contact lock manager */
+	DLM_NOPURGED,             /* 18: can't contact purge daemon */
+	DLM_BADARGS,              /* 19: bad api args */
+	DLM_VOID,                 /* 20: no status */
+	DLM_NOTQUEUED,            /* 21: NOQUEUE was specified and request failed */
+	DLM_IVBUFLEN,             /* 22: invalid resource name length */
+	DLM_CVTUNGRANT,           /* 23: attempted to convert ungranted lock */
+	DLM_BADPARAM,             /* 24: invalid lock mode specified */
+	DLM_VALNOTVALID,          /* 25: value block has been invalidated */
+	DLM_REJECTED,             /* 26: request rejected, unrecognized client */
+	DLM_ABORT,                /* 27: blocked lock request cancelled */
+	DLM_CANCEL,               /* 28: conversion request cancelled */
+	DLM_IVRESHANDLE,          /* 29: invalid resource handle */
+	DLM_DEADLOCK,             /* 30: deadlock recovery refused this request */
+	DLM_DENIED_NOASTS,        /* 31: failed to allocate AST */
+	DLM_FORWARD,              /* 32: request must wait for primary's response */
+	DLM_TIMEOUT,              /* 33: timeout value for lock has expired */
+	DLM_IVGROUPID,            /* 34: invalid group specification */
+	DLM_VERS_CONFLICT,        /* 35: version conflicts prevent request handling */
+	DLM_BAD_DEVICE_PATH,      /* 36: Locks device does not exist or path wrong */
+	DLM_NO_DEVICE_PERMISSION, /* 37: Client has insufficient pers for device */
+	DLM_NO_CONTROL_DEVICE,    /* 38: Cannot set options on opened device */
+
+	DLM_RECOVERING,           /* 39: extension, allows caller to fail a lock
+				     request if it is being recovered */
+	DLM_MIGRATING,            /* 40: extension, allows caller to fail a lock
+				     request if it is being migrated */
+	DLM_MAXSTATS,             /* 41: upper limit for return code validation */
+};
+
+/* for pretty-printing dlm_status error messages */
+const char *dlm_errmsg(enum dlm_status err);
+/* for pretty-printing dlm_status error names */
+const char *dlm_errname(enum dlm_status err);
+
+/* Eventually the DLM will use standard errno values, but in the
+ * meantime this lets us track dlm errors as they bubble up. When we
+ * bring its error reporting into line with the rest of the stack,
+ * these can just be replaced with calls to mlog_errno. */
+#define dlm_error(st) do {						\
+	if ((st) != DLM_RECOVERING &&					\
+	    (st) != DLM_MIGRATING &&					\
+	    (st) != DLM_FORWARD)					\
+		mlog(ML_ERROR, "dlm status = %s\n", dlm_errname((st)));	\
+} while (0)
+
+#define DLM_LKSB_UNUSED1           0x01
+#define DLM_LKSB_PUT_LVB           0x02
+#define DLM_LKSB_GET_LVB           0x04
+#define DLM_LKSB_UNUSED2           0x08
+#define DLM_LKSB_UNUSED3           0x10
+#define DLM_LKSB_UNUSED4           0x20
+#define DLM_LKSB_UNUSED5           0x40
+#define DLM_LKSB_UNUSED6           0x80
+
+#define DLM_LVB_LEN  64
+
+/* Callers are only allowed access to the lvb and status members of
+ * this struct. */
+struct dlm_lockstatus {
+	enum dlm_status status;
+	u32 flags;
+	struct dlm_lock *lockid;
+	char lvb[DLM_LVB_LEN];
+};
+
+/* Valid lock modes. */
+#define LKM_IVMODE      (-1)            /* invalid mode */
+#define LKM_NLMODE      0               /* null lock */
+#define LKM_CRMODE      1               /* concurrent read    unsupported */
+#define LKM_CWMODE      2               /* concurrent write   unsupported */
+#define LKM_PRMODE      3               /* protected read */
+#define LKM_PWMODE      4               /* protected write    unsupported */
+#define LKM_EXMODE      5               /* exclusive */
+#define LKM_MAXMODE     5
+#define LKM_MODEMASK    0xff
+
+/* Flags passed to dlmlock and dlmunlock:
+ * reserved: flags used by the "real" dlm
+ * only a few are supported by this dlm
+ * (U) = unsupported by ocfs2 dlm */
+#define LKM_ORPHAN       0x00000010  /* this lock is orphanable (U) */
+#define LKM_PARENTABLE   0x00000020  /* this lock was orphaned (U) */
+#define LKM_BLOCK        0x00000040  /* blocking lock request (U) */
+#define LKM_LOCAL        0x00000080  /* local lock request */
+#define LKM_VALBLK       0x00000100  /* lock value block request */
+#define LKM_NOQUEUE      0x00000200  /* non blocking request */
+#define LKM_CONVERT      0x00000400  /* conversion request */
+#define LKM_NODLCKWT     0x00000800  /* this lock wont deadlock (U) */
+#define LKM_UNLOCK       0x00001000  /* deallocate this lock */
+#define LKM_CANCEL       0x00002000  /* cancel conversion request */
+#define LKM_DEQALL       0x00004000  /* remove all locks held by proc (U) */
+#define LKM_INVVALBLK    0x00008000  /* invalidate lock value block */
+#define LKM_SYNCSTS      0x00010000  /* return synchronous status if poss (U) */
+#define LKM_TIMEOUT      0x00020000  /* lock request contains timeout (U) */
+#define LKM_SNGLDLCK     0x00040000  /* request can self-deadlock (U) */
+#define LKM_FINDLOCAL    0x00080000  /* find local lock request (U) */
+#define LKM_PROC_OWNED   0x00100000  /* owned by process, not group (U) */
+#define LKM_XID          0x00200000  /* use transaction id for deadlock (U) */
+#define LKM_XID_CONFLICT 0x00400000  /* do not allow lock inheritance (U) */
+#define LKM_FORCE        0x00800000  /* force unlock flag */
+#define LKM_REVVALBLK    0x01000000  /* temporary solution: re-validate
+					lock value block (U) */
+/* unused */
+#define LKM_UNUSED1      0x00000001  /* unused */
+#define LKM_UNUSED2      0x00000002  /* unused */
+#define LKM_UNUSED3      0x00000004  /* unused */
+#define LKM_UNUSED4      0x00000008  /* unused */
+#define LKM_UNUSED5      0x02000000  /* unused */
+#define LKM_UNUSED6      0x04000000  /* unused */
+#define LKM_UNUSED7      0x08000000  /* unused */
+
+/* ocfs2 extensions: internal only
+ * should never be used by caller */
+#define LKM_MIGRATION    0x10000000  /* extension: lockres is to be migrated
+					to another node */
+#define LKM_PUT_LVB      0x20000000  /* extension: lvb is being passed
+					should be applied to lockres */
+#define LKM_GET_LVB      0x40000000  /* extension: lvb should be copied
+					from lockres when lock is granted */
+#define LKM_RECOVERY     0x80000000  /* extension: flag for recovery lock
+					used to avoid recovery rwsem */
+
+
+typedef void (dlm_astlockfunc_t)(void *);
+typedef void (dlm_bastlockfunc_t)(void *, int);
+typedef void (dlm_astunlockfunc_t)(void *, enum dlm_status);
+
+enum dlm_status dlmlock(struct dlm_ctxt *dlm,
+			int mode,
+			struct dlm_lockstatus *lksb,
+			int flags,
+			const char *name,
+			int namelen,
+			dlm_astlockfunc_t *ast,
+			void *data,
+			dlm_bastlockfunc_t *bast);
+
+enum dlm_status dlmunlock(struct dlm_ctxt *dlm,
+			  struct dlm_lockstatus *lksb,
+			  int flags,
+			  dlm_astunlockfunc_t *unlockast,
+			  void *data);
+
+struct dlm_protocol_version {
+	u8 pv_major;
+	u8 pv_minor;
+};
+struct dlm_ctxt * dlm_register_domain(const char *domain, u32 key,
+				      struct dlm_protocol_version *fs_proto);
+
+void dlm_unregister_domain(struct dlm_ctxt *dlm);
+
+void dlm_print_one_lock(struct dlm_lock *lockid);
+
+typedef void (dlm_eviction_func)(int, void *);
+struct dlm_eviction_cb {
+	struct list_head        ec_item;
+	dlm_eviction_func       *ec_func;
+	void                    *ec_data;
+};
+void dlm_setup_eviction_cb(struct dlm_eviction_cb *cb,
+			   dlm_eviction_func *f,
+			   void *data);
+void dlm_register_eviction_cb(struct dlm_ctxt *dlm,
+			      struct dlm_eviction_cb *cb);
+void dlm_unregister_eviction_cb(struct dlm_eviction_cb *cb);
+
+#endif /* DLMAPI_H */

fs/ocfs2/dlm/dlmast.c

@@ -0,0 +1,500 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dlmast.c
+ *
+ * AST and BAST functionality for local and remote nodes
+ *
+ * Copyright (C) 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/highmem.h>
+#include <linux/init.h>
+#include <linux/sysctl.h>
+#include <linux/random.h>
+#include <linux/blkdev.h>
+#include <linux/socket.h>
+#include <linux/inet.h>
+#include <linux/spinlock.h>
+
+
+#include "cluster/heartbeat.h"
+#include "cluster/nodemanager.h"
+#include "cluster/tcp.h"
+
+#include "dlmapi.h"
+#include "dlmcommon.h"
+
+#define MLOG_MASK_PREFIX ML_DLM
+#include "cluster/masklog.h"
+
+static void dlm_update_lvb(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
+			   struct dlm_lock *lock);
+static int dlm_should_cancel_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock);
+
+/* Should be called as an ast gets queued to see if the new
+ * lock level will obsolete a pending bast.
+ * For example, if dlm_thread queued a bast for an EX lock that
+ * was blocking another EX, but before sending the bast the
+ * lock owner downconverted to NL, the bast is now obsolete.
+ * Only the ast should be sent.
+ * This is needed because the lock and convert paths can queue
+ * asts out-of-band (not waiting for dlm_thread) in order to
+ * allow for LKM_NOQUEUE to get immediate responses. */
+static int dlm_should_cancel_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
+{
+	assert_spin_locked(&dlm->ast_lock);
+	assert_spin_locked(&lock->spinlock);
+
+	if (lock->ml.highest_blocked == LKM_IVMODE)
+		return 0;
+	BUG_ON(lock->ml.highest_blocked == LKM_NLMODE);
+
+	if (lock->bast_pending &&
+	    list_empty(&lock->bast_list))
+		/* old bast already sent, ok */
+		return 0;
+
+	if (lock->ml.type == LKM_EXMODE)
+		/* EX blocks anything left, any bast still valid */
+		return 0;
+	else if (lock->ml.type == LKM_NLMODE)
+		/* NL blocks nothing, no reason to send any bast, cancel it */
+		return 1;
+	else if (lock->ml.highest_blocked != LKM_EXMODE)
+		/* PR only blocks EX */
+		return 1;
+
+	return 0;
+}
+
+void __dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
+{
+	struct dlm_lock_resource *res;
+
+	BUG_ON(!dlm);
+	BUG_ON(!lock);
+
+	res = lock->lockres;
+
+	assert_spin_locked(&dlm->ast_lock);
+
+	if (!list_empty(&lock->ast_list)) {
+		mlog(ML_ERROR, "%s: res %.*s, lock %u:%llu, "
+		     "AST list not empty, pending %d, newlevel %d\n",
+		     dlm->name, res->lockname.len, res->lockname.name,
+		     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
+		     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
+		     lock->ast_pending, lock->ml.type);
+		BUG();
+	}
+	if (lock->ast_pending)
+		mlog(0, "%s: res %.*s, lock %u:%llu, AST getting flushed\n",
+		     dlm->name, res->lockname.len, res->lockname.name,
+		     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
+		     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)));
+
+	/* putting lock on list, add a ref */
+	dlm_lock_get(lock);
+	spin_lock(&lock->spinlock);
+
+	/* check to see if this ast obsoletes the bast */
+	if (dlm_should_cancel_bast(dlm, lock)) {
+		mlog(0, "%s: res %.*s, lock %u:%llu, Cancelling BAST\n",
+		     dlm->name, res->lockname.len, res->lockname.name,
+		     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
+		     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)));
+		lock->bast_pending = 0;
+		list_del_init(&lock->bast_list);
+		lock->ml.highest_blocked = LKM_IVMODE;
+		/* removing lock from list, remove a ref.  guaranteed
+		 * this won't be the last ref because of the get above,
+		 * so res->spinlock will not be taken here */
+		dlm_lock_put(lock);
+		/* free up the reserved bast that we are cancelling.
+		 * guaranteed that this will not be the last reserved
+		 * ast because *both* an ast and a bast were reserved
+		 * to get to this point.  the res->spinlock will not be
+		 * taken here */
+		dlm_lockres_release_ast(dlm, res);
+	}
+	list_add_tail(&lock->ast_list, &dlm->pending_asts);
+	lock->ast_pending = 1;
+	spin_unlock(&lock->spinlock);
+}
+
+void dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
+{
+	BUG_ON(!dlm);
+	BUG_ON(!lock);
+
+	spin_lock(&dlm->ast_lock);
+	__dlm_queue_ast(dlm, lock);
+	spin_unlock(&dlm->ast_lock);
+}
+
+
+void __dlm_queue_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
+{
+	struct dlm_lock_resource *res;
+
+	BUG_ON(!dlm);
+	BUG_ON(!lock);
+
+	assert_spin_locked(&dlm->ast_lock);
+
+	res = lock->lockres;
+
+	BUG_ON(!list_empty(&lock->bast_list));
+	if (lock->bast_pending)
+		mlog(0, "%s: res %.*s, lock %u:%llu, BAST getting flushed\n",
+		     dlm->name, res->lockname.len, res->lockname.name,
+		     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
+		     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)));
+
+	/* putting lock on list, add a ref */
+	dlm_lock_get(lock);
+	spin_lock(&lock->spinlock);
+	list_add_tail(&lock->bast_list, &dlm->pending_basts);
+	lock->bast_pending = 1;
+	spin_unlock(&lock->spinlock);
+}
+
+void dlm_queue_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
+{
+	BUG_ON(!dlm);
+	BUG_ON(!lock);
+
+	spin_lock(&dlm->ast_lock);
+	__dlm_queue_bast(dlm, lock);
+	spin_unlock(&dlm->ast_lock);
+}
+
+static void dlm_update_lvb(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
+			   struct dlm_lock *lock)
+{
+	struct dlm_lockstatus *lksb = lock->lksb;
+	BUG_ON(!lksb);
+
+	/* only updates if this node masters the lockres */
+	spin_lock(&res->spinlock);
+	if (res->owner == dlm->node_num) {
+		/* check the lksb flags for the direction */
+		if (lksb->flags & DLM_LKSB_GET_LVB) {
+			mlog(0, "getting lvb from lockres for %s node\n",
+				  lock->ml.node == dlm->node_num ? "master" :
+				  "remote");
+			memcpy(lksb->lvb, res->lvb, DLM_LVB_LEN);
+		}
+		/* Do nothing for lvb put requests - they should be done in
+ 		 * place when the lock is downconverted - otherwise we risk
+ 		 * racing gets and puts which could result in old lvb data
+ 		 * being propagated. We leave the put flag set and clear it
+ 		 * here. In the future we might want to clear it at the time
+ 		 * the put is actually done.
+		 */
+	}
+	spin_unlock(&res->spinlock);
+
+	/* reset any lvb flags on the lksb */
+	lksb->flags &= ~(DLM_LKSB_PUT_LVB|DLM_LKSB_GET_LVB);
+}
+
+void dlm_do_local_ast(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
+		      struct dlm_lock *lock)
+{
+	dlm_astlockfunc_t *fn;
+	struct dlm_lockstatus *lksb;
+
+	mlog(0, "%s: res %.*s, lock %u:%llu, Local AST\n", dlm->name,
+	     res->lockname.len, res->lockname.name,
+	     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
+	     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)));
+
+	lksb = lock->lksb;
+	fn = lock->ast;
+	BUG_ON(lock->ml.node != dlm->node_num);
+
+	dlm_update_lvb(dlm, res, lock);
+	(*fn)(lock->astdata);
+}
+
+
+int dlm_do_remote_ast(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
+		      struct dlm_lock *lock)
+{
+	int ret;
+	struct dlm_lockstatus *lksb;
+	int lksbflags;
+
+	mlog(0, "%s: res %.*s, lock %u:%llu, Remote AST\n", dlm->name,
+	     res->lockname.len, res->lockname.name,
+	     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
+	     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)));
+
+	lksb = lock->lksb;
+	BUG_ON(lock->ml.node == dlm->node_num);
+
+	lksbflags = lksb->flags;
+	dlm_update_lvb(dlm, res, lock);
+
+	/* lock request came from another node
+	 * go do the ast over there */
+	ret = dlm_send_proxy_ast(dlm, res, lock, lksbflags);
+	return ret;
+}
+
+void dlm_do_local_bast(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
+		       struct dlm_lock *lock, int blocked_type)
+{
+	dlm_bastlockfunc_t *fn = lock->bast;
+
+	BUG_ON(lock->ml.node != dlm->node_num);
+
+	mlog(0, "%s: res %.*s, lock %u:%llu, Local BAST, blocked %d\n",
+	     dlm->name, res->lockname.len, res->lockname.name,
+	     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
+	     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
+	     blocked_type);
+
+	(*fn)(lock->astdata, blocked_type);
+}
+
+
+
+int dlm_proxy_ast_handler(struct o2net_msg *msg, u32 len, void *data,
+			  void **ret_data)
+{
+	int ret;
+	unsigned int locklen;
+	struct dlm_ctxt *dlm = data;
+	struct dlm_lock_resource *res = NULL;
+	struct dlm_lock *lock = NULL;
+	struct dlm_proxy_ast *past = (struct dlm_proxy_ast *) msg->buf;
+	char *name;
+	struct list_head *head = NULL;
+	__be64 cookie;
+	u32 flags;
+	u8 node;
+
+	if (!dlm_grab(dlm)) {
+		dlm_error(DLM_REJECTED);
+		return DLM_REJECTED;
+	}
+
+	mlog_bug_on_msg(!dlm_domain_fully_joined(dlm),
+			"Domain %s not fully joined!\n", dlm->name);
+
+	name = past->name;
+	locklen = past->namelen;
+	cookie = past->cookie;
+	flags = be32_to_cpu(past->flags);
+	node = past->node_idx;
+
+	if (locklen > DLM_LOCKID_NAME_MAX) {
+		ret = DLM_IVBUFLEN;
+		mlog(ML_ERROR, "Invalid name length (%d) in proxy ast "
+		     "handler!\n", locklen);
+		goto leave;
+	}
+
+	if ((flags & (LKM_PUT_LVB|LKM_GET_LVB)) ==
+	     (LKM_PUT_LVB|LKM_GET_LVB)) {
+		mlog(ML_ERROR, "Both PUT and GET lvb specified, (0x%x)\n",
+		     flags);
+		ret = DLM_BADARGS;
+		goto leave;
+	}
+
+	mlog(0, "lvb: %s\n", flags & LKM_PUT_LVB ? "put lvb" :
+		  (flags & LKM_GET_LVB ? "get lvb" : "none"));
+
+	mlog(0, "type=%d, blocked_type=%d\n", past->type, past->blocked_type);
+
+	if (past->type != DLM_AST &&
+	    past->type != DLM_BAST) {
+		mlog(ML_ERROR, "Unknown ast type! %d, cookie=%u:%llu"
+		     "name=%.*s, node=%u\n", past->type,
+		     dlm_get_lock_cookie_node(be64_to_cpu(cookie)),
+		     dlm_get_lock_cookie_seq(be64_to_cpu(cookie)),
+		     locklen, name, node);
+		ret = DLM_IVLOCKID;
+		goto leave;
+	}
+
+	res = dlm_lookup_lockres(dlm, name, locklen);
+	if (!res) {
+		mlog(0, "Got %sast for unknown lockres! cookie=%u:%llu, "
+		     "name=%.*s, node=%u\n", (past->type == DLM_AST ? "" : "b"),
+		     dlm_get_lock_cookie_node(be64_to_cpu(cookie)),
+		     dlm_get_lock_cookie_seq(be64_to_cpu(cookie)),
+		     locklen, name, node);
+		ret = DLM_IVLOCKID;
+		goto leave;
+	}
+
+	/* cannot get a proxy ast message if this node owns it */
+	BUG_ON(res->owner == dlm->node_num);
+
+	mlog(0, "%s: res %.*s\n", dlm->name, res->lockname.len,
+	     res->lockname.name);
+
+	spin_lock(&res->spinlock);
+	if (res->state & DLM_LOCK_RES_RECOVERING) {
+		mlog(0, "Responding with DLM_RECOVERING!\n");
+		ret = DLM_RECOVERING;
+		goto unlock_out;
+	}
+	if (res->state & DLM_LOCK_RES_MIGRATING) {
+		mlog(0, "Responding with DLM_MIGRATING!\n");
+		ret = DLM_MIGRATING;
+		goto unlock_out;
+	}
+	/* try convert queue for both ast/bast */
+	head = &res->converting;
+	lock = NULL;
+	list_for_each_entry(lock, head, list) {
+		if (lock->ml.cookie == cookie)
+			goto do_ast;
+	}
+
+	/* if not on convert, try blocked for ast, granted for bast */
+	if (past->type == DLM_AST)
+		head = &res->blocked;
+	else
+		head = &res->granted;
+
+	list_for_each_entry(lock, head, list) {
+		if (lock->ml.cookie == cookie)
+			goto do_ast;
+	}
+
+	mlog(0, "Got %sast for unknown lock! cookie=%u:%llu, name=%.*s, "
+	     "node=%u\n", past->type == DLM_AST ? "" : "b",
+	     dlm_get_lock_cookie_node(be64_to_cpu(cookie)),
+	     dlm_get_lock_cookie_seq(be64_to_cpu(cookie)),
+	     locklen, name, node);
+
+	ret = DLM_NORMAL;
+unlock_out:
+	spin_unlock(&res->spinlock);
+	goto leave;
+
+do_ast:
+	ret = DLM_NORMAL;
+	if (past->type == DLM_AST) {
+		/* do not alter lock refcount.  switching lists. */
+		list_move_tail(&lock->list, &res->granted);
+		mlog(0, "%s: res %.*s, lock %u:%llu, Granted type %d => %d\n",
+		     dlm->name, res->lockname.len, res->lockname.name,
+		     dlm_get_lock_cookie_node(be64_to_cpu(cookie)),
+		     dlm_get_lock_cookie_seq(be64_to_cpu(cookie)),
+		     lock->ml.type, lock->ml.convert_type);
+
+		if (lock->ml.convert_type != LKM_IVMODE) {
+			lock->ml.type = lock->ml.convert_type;
+			lock->ml.convert_type = LKM_IVMODE;
+		} else {
+			// should already be there....
+		}
+
+		lock->lksb->status = DLM_NORMAL;
+
+		/* if we requested the lvb, fetch it into our lksb now */
+		if (flags & LKM_GET_LVB) {
+			BUG_ON(!(lock->lksb->flags & DLM_LKSB_GET_LVB));
+			memcpy(lock->lksb->lvb, past->lvb, DLM_LVB_LEN);
+		}
+	}
+	spin_unlock(&res->spinlock);
+
+	if (past->type == DLM_AST)
+		dlm_do_local_ast(dlm, res, lock);
+	else
+		dlm_do_local_bast(dlm, res, lock, past->blocked_type);
+
+leave:
+	if (res)
+		dlm_lockres_put(res);
+
+	dlm_put(dlm);
+	return ret;
+}
+
+
+
+int dlm_send_proxy_ast_msg(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
+			   struct dlm_lock *lock, int msg_type,
+			   int blocked_type, int flags)
+{
+	int ret = 0;
+	struct dlm_proxy_ast past;
+	struct kvec vec[2];
+	size_t veclen = 1;
+	int status;
+
+	mlog(0, "%s: res %.*s, to %u, type %d, blocked_type %d\n", dlm->name,
+	     res->lockname.len, res->lockname.name, lock->ml.node, msg_type,
+	     blocked_type);
+
+	memset(&past, 0, sizeof(struct dlm_proxy_ast));
+	past.node_idx = dlm->node_num;
+	past.type = msg_type;
+	past.blocked_type = blocked_type;
+	past.namelen = res->lockname.len;
+	memcpy(past.name, res->lockname.name, past.namelen);
+	past.cookie = lock->ml.cookie;
+
+	vec[0].iov_len = sizeof(struct dlm_proxy_ast);
+	vec[0].iov_base = &past;
+	if (flags & DLM_LKSB_GET_LVB) {
+		be32_add_cpu(&past.flags, LKM_GET_LVB);
+		vec[1].iov_len = DLM_LVB_LEN;
+		vec[1].iov_base = lock->lksb->lvb;
+		veclen++;
+	}
+
+	ret = o2net_send_message_vec(DLM_PROXY_AST_MSG, dlm->key, vec, veclen,
+				     lock->ml.node, &status);
+	if (ret < 0)
+		mlog(ML_ERROR, "%s: res %.*s, error %d send AST to node %u\n",
+		     dlm->name, res->lockname.len, res->lockname.name, ret,
+		     lock->ml.node);
+	else {
+		if (status == DLM_RECOVERING) {
+			mlog(ML_ERROR, "sent AST to node %u, it thinks this "
+			     "node is dead!\n", lock->ml.node);
+			BUG();
+		} else if (status == DLM_MIGRATING) {
+			mlog(ML_ERROR, "sent AST to node %u, it returned "
+			     "DLM_MIGRATING!\n", lock->ml.node);
+			BUG();
+		} else if (status != DLM_NORMAL && status != DLM_IVLOCKID) {
+			mlog(ML_ERROR, "AST to node %u returned %d!\n",
+			     lock->ml.node, status);
+			/* ignore it */
+		}
+		ret = 0;
+	}
+	return ret;
+}

fs/ocfs2/dlm/dlmconvert.c

@@ -0,0 +1,544 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dlmconvert.c
+ *
+ * underlying calls for lock conversion
+ *
+ * Copyright (C) 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/highmem.h>
+#include <linux/init.h>
+#include <linux/sysctl.h>
+#include <linux/random.h>
+#include <linux/blkdev.h>
+#include <linux/socket.h>
+#include <linux/inet.h>
+#include <linux/spinlock.h>
+
+
+#include "cluster/heartbeat.h"
+#include "cluster/nodemanager.h"
+#include "cluster/tcp.h"
+
+#include "dlmapi.h"
+#include "dlmcommon.h"
+
+#include "dlmconvert.h"
+
+#define MLOG_MASK_PREFIX ML_DLM
+#include "cluster/masklog.h"
+
+/* NOTE: __dlmconvert_master is the only function in here that
+ * needs a spinlock held on entry (res->spinlock) and it is the
+ * only one that holds a lock on exit (res->spinlock).
+ * All other functions in here need no locks and drop all of
+ * the locks that they acquire. */
+static enum dlm_status __dlmconvert_master(struct dlm_ctxt *dlm,
+					   struct dlm_lock_resource *res,
+					   struct dlm_lock *lock, int flags,
+					   int type, int *call_ast,
+					   int *kick_thread);
+static enum dlm_status dlm_send_remote_convert_request(struct dlm_ctxt *dlm,
+					   struct dlm_lock_resource *res,
+					   struct dlm_lock *lock, int flags, int type);
+
+/*
+ * this is only called directly by dlmlock(), and only when the
+ * local node is the owner of the lockres
+ * locking:
+ *   caller needs:  none
+ *   taken:         takes and drops res->spinlock
+ *   held on exit:  none
+ * returns: see __dlmconvert_master
+ */
+enum dlm_status dlmconvert_master(struct dlm_ctxt *dlm,
+				  struct dlm_lock_resource *res,
+				  struct dlm_lock *lock, int flags, int type)
+{
+	int call_ast = 0, kick_thread = 0;
+	enum dlm_status status;
+
+	spin_lock(&res->spinlock);
+	/* we are not in a network handler, this is fine */
+	__dlm_wait_on_lockres(res);
+	__dlm_lockres_reserve_ast(res);
+	res->state |= DLM_LOCK_RES_IN_PROGRESS;
+
+	status = __dlmconvert_master(dlm, res, lock, flags, type,
+				     &call_ast, &kick_thread);
+
+	res->state &= ~DLM_LOCK_RES_IN_PROGRESS;
+	spin_unlock(&res->spinlock);
+	wake_up(&res->wq);
+	if (status != DLM_NORMAL && status != DLM_NOTQUEUED)
+		dlm_error(status);
+
+	/* either queue the ast or release it */
+	if (call_ast)
+		dlm_queue_ast(dlm, lock);
+	else
+		dlm_lockres_release_ast(dlm, res);
+
+	if (kick_thread)
+		dlm_kick_thread(dlm, res);
+
+	return status;
+}
+
+/* performs lock conversion at the lockres master site
+ * locking:
+ *   caller needs:  res->spinlock
+ *   taken:         takes and drops lock->spinlock
+ *   held on exit:  res->spinlock
+ * returns: DLM_NORMAL, DLM_NOTQUEUED, DLM_DENIED
+ *   call_ast: whether ast should be called for this lock
+ *   kick_thread: whether dlm_kick_thread should be called
+ */
+static enum dlm_status __dlmconvert_master(struct dlm_ctxt *dlm,
+					   struct dlm_lock_resource *res,
+					   struct dlm_lock *lock, int flags,
+					   int type, int *call_ast,
+					   int *kick_thread)
+{
+	enum dlm_status status = DLM_NORMAL;
+	struct dlm_lock *tmplock=NULL;
+
+	assert_spin_locked(&res->spinlock);
+
+	mlog(0, "type=%d, convert_type=%d, new convert_type=%d\n",
+	     lock->ml.type, lock->ml.convert_type, type);
+
+	spin_lock(&lock->spinlock);
+
+	/* already converting? */
+	if (lock->ml.convert_type != LKM_IVMODE) {
+		mlog(ML_ERROR, "attempted to convert a lock with a lock "
+		     "conversion pending\n");
+		status = DLM_DENIED;
+		goto unlock_exit;
+	}
+
+	/* must be on grant queue to convert */
+	if (!dlm_lock_on_list(&res->granted, lock)) {
+		mlog(ML_ERROR, "attempted to convert a lock not on grant "
+		     "queue\n");
+		status = DLM_DENIED;
+		goto unlock_exit;
+	}
+
+	if (flags & LKM_VALBLK) {
+		switch (lock->ml.type) {
+			case LKM_EXMODE:
+				/* EX + LKM_VALBLK + convert == set lvb */
+				mlog(0, "will set lvb: converting %s->%s\n",
+				     dlm_lock_mode_name(lock->ml.type),
+				     dlm_lock_mode_name(type));
+				lock->lksb->flags |= DLM_LKSB_PUT_LVB;
+				break;
+			case LKM_PRMODE:
+			case LKM_NLMODE:
+				/* refetch if new level is not NL */
+				if (type > LKM_NLMODE) {
+					mlog(0, "will fetch new value into "
+					     "lvb: converting %s->%s\n",
+					     dlm_lock_mode_name(lock->ml.type),
+					     dlm_lock_mode_name(type));
+					lock->lksb->flags |= DLM_LKSB_GET_LVB;
+				} else {
+					mlog(0, "will NOT fetch new value "
+					     "into lvb: converting %s->%s\n",
+					     dlm_lock_mode_name(lock->ml.type),
+					     dlm_lock_mode_name(type));
+					flags &= ~(LKM_VALBLK);
+				}
+				break;
+		}
+	}
+
+
+	/* in-place downconvert? */
+	if (type <= lock->ml.type)
+		goto grant;
+
+	/* upconvert from here on */
+	status = DLM_NORMAL;
+	list_for_each_entry(tmplock, &res->granted, list) {
+		if (tmplock == lock)
+			continue;
+		if (!dlm_lock_compatible(tmplock->ml.type, type))
+			goto switch_queues;
+	}
+
+	list_for_each_entry(tmplock, &res->converting, list) {
+		if (!dlm_lock_compatible(tmplock->ml.type, type))
+			goto switch_queues;
+		/* existing conversion requests take precedence */
+		if (!dlm_lock_compatible(tmplock->ml.convert_type, type))
+			goto switch_queues;
+	}
+
+	/* fall thru to grant */
+
+grant:
+	mlog(0, "res %.*s, granting %s lock\n", res->lockname.len,
+	     res->lockname.name, dlm_lock_mode_name(type));
+	/* immediately grant the new lock type */
+	lock->lksb->status = DLM_NORMAL;
+	if (lock->ml.node == dlm->node_num)
+		mlog(0, "doing in-place convert for nonlocal lock\n");
+	lock->ml.type = type;
+	if (lock->lksb->flags & DLM_LKSB_PUT_LVB)
+		memcpy(res->lvb, lock->lksb->lvb, DLM_LVB_LEN);
+
+	status = DLM_NORMAL;
+	*call_ast = 1;
+	goto unlock_exit;
+
+switch_queues:
+	if (flags & LKM_NOQUEUE) {
+		mlog(0, "failed to convert NOQUEUE lock %.*s from "
+		     "%d to %d...\n", res->lockname.len, res->lockname.name,
+		     lock->ml.type, type);
+		status = DLM_NOTQUEUED;
+		goto unlock_exit;
+	}
+	mlog(0, "res %.*s, queueing...\n", res->lockname.len,
+	     res->lockname.name);
+
+	lock->ml.convert_type = type;
+	/* do not alter lock refcount.  switching lists. */
+	list_move_tail(&lock->list, &res->converting);
+
+unlock_exit:
+	spin_unlock(&lock->spinlock);
+	if (status == DLM_DENIED) {
+		__dlm_print_one_lock_resource(res);
+	}
+	if (status == DLM_NORMAL)
+		*kick_thread = 1;
+	return status;
+}
+
+void dlm_revert_pending_convert(struct dlm_lock_resource *res,
+				struct dlm_lock *lock)
+{
+	/* do not alter lock refcount.  switching lists. */
+	list_move_tail(&lock->list, &res->granted);
+	lock->ml.convert_type = LKM_IVMODE;
+	lock->lksb->flags &= ~(DLM_LKSB_GET_LVB|DLM_LKSB_PUT_LVB);
+}
+
+/* messages the master site to do lock conversion
+ * locking:
+ *   caller needs:  none
+ *   taken:         takes and drops res->spinlock, uses DLM_LOCK_RES_IN_PROGRESS
+ *   held on exit:  none
+ * returns: DLM_NORMAL, DLM_RECOVERING, status from remote node
+ */
+enum dlm_status dlmconvert_remote(struct dlm_ctxt *dlm,
+				  struct dlm_lock_resource *res,
+				  struct dlm_lock *lock, int flags, int type)
+{
+	enum dlm_status status;
+
+	mlog(0, "type=%d, convert_type=%d, busy=%d\n", lock->ml.type,
+	     lock->ml.convert_type, res->state & DLM_LOCK_RES_IN_PROGRESS);
+
+	spin_lock(&res->spinlock);
+	if (res->state & DLM_LOCK_RES_RECOVERING) {
+		mlog(0, "bailing out early since res is RECOVERING "
+		     "on secondary queue\n");
+		/* __dlm_print_one_lock_resource(res); */
+		status = DLM_RECOVERING;
+		goto bail;
+	}
+	/* will exit this call with spinlock held */
+	__dlm_wait_on_lockres(res);
+
+	if (lock->ml.convert_type != LKM_IVMODE) {
+		__dlm_print_one_lock_resource(res);
+		mlog(ML_ERROR, "converting a remote lock that is already "
+		     "converting! (cookie=%u:%llu, conv=%d)\n",
+		     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
+		     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
+		     lock->ml.convert_type);
+		status = DLM_DENIED;
+		goto bail;
+	}
+	res->state |= DLM_LOCK_RES_IN_PROGRESS;
+	/* move lock to local convert queue */
+	/* do not alter lock refcount.  switching lists. */
+	list_move_tail(&lock->list, &res->converting);
+	lock->convert_pending = 1;
+	lock->ml.convert_type = type;
+
+	if (flags & LKM_VALBLK) {
+		if (lock->ml.type == LKM_EXMODE) {
+			flags |= LKM_PUT_LVB;
+			lock->lksb->flags |= DLM_LKSB_PUT_LVB;
+		} else {
+			if (lock->ml.convert_type == LKM_NLMODE)
+				flags &= ~LKM_VALBLK;
+			else {
+				flags |= LKM_GET_LVB;
+				lock->lksb->flags |= DLM_LKSB_GET_LVB;
+			}
+		}
+	}
+	spin_unlock(&res->spinlock);
+
+	/* no locks held here.
+	 * need to wait for a reply as to whether it got queued or not. */
+	status = dlm_send_remote_convert_request(dlm, res, lock, flags, type);
+
+	spin_lock(&res->spinlock);
+	res->state &= ~DLM_LOCK_RES_IN_PROGRESS;
+	lock->convert_pending = 0;
+	/* if it failed, move it back to granted queue */
+	if (status != DLM_NORMAL) {
+		if (status != DLM_NOTQUEUED)
+			dlm_error(status);
+		dlm_revert_pending_convert(res, lock);
+	}
+bail:
+	spin_unlock(&res->spinlock);
+
+	/* TODO: should this be a wake_one? */
+	/* wake up any IN_PROGRESS waiters */
+	wake_up(&res->wq);
+
+	return status;
+}
+
+/* sends DLM_CONVERT_LOCK_MSG to master site
+ * locking:
+ *   caller needs:  none
+ *   taken:         none
+ *   held on exit:  none
+ * returns: DLM_NOLOCKMGR, status from remote node
+ */
+static enum dlm_status dlm_send_remote_convert_request(struct dlm_ctxt *dlm,
+					   struct dlm_lock_resource *res,
+					   struct dlm_lock *lock, int flags, int type)
+{
+	struct dlm_convert_lock convert;
+	int tmpret;
+	enum dlm_status ret;
+	int status = 0;
+	struct kvec vec[2];
+	size_t veclen = 1;
+
+	mlog(0, "%.*s\n", res->lockname.len, res->lockname.name);
+
+	memset(&convert, 0, sizeof(struct dlm_convert_lock));
+	convert.node_idx = dlm->node_num;
+	convert.requested_type = type;
+	convert.cookie = lock->ml.cookie;
+	convert.namelen = res->lockname.len;
+	convert.flags = cpu_to_be32(flags);
+	memcpy(convert.name, res->lockname.name, convert.namelen);
+
+	vec[0].iov_len = sizeof(struct dlm_convert_lock);
+	vec[0].iov_base = &convert;
+
+	if (flags & LKM_PUT_LVB) {
+		/* extra data to send if we are updating lvb */
+		vec[1].iov_len = DLM_LVB_LEN;
+		vec[1].iov_base = lock->lksb->lvb;
+		veclen++;
+	}
+
+	tmpret = o2net_send_message_vec(DLM_CONVERT_LOCK_MSG, dlm->key,
+					vec, veclen, res->owner, &status);
+	if (tmpret >= 0) {
+		// successfully sent and received
+		ret = status;  // this is already a dlm_status
+		if (ret == DLM_RECOVERING) {
+			mlog(0, "node %u returned DLM_RECOVERING from convert "
+			     "message!\n", res->owner);
+		} else if (ret == DLM_MIGRATING) {
+			mlog(0, "node %u returned DLM_MIGRATING from convert "
+			     "message!\n", res->owner);
+		} else if (ret == DLM_FORWARD) {
+			mlog(0, "node %u returned DLM_FORWARD from convert "
+			     "message!\n", res->owner);
+		} else if (ret != DLM_NORMAL && ret != DLM_NOTQUEUED)
+			dlm_error(ret);
+	} else {
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", tmpret, DLM_CONVERT_LOCK_MSG, dlm->key,
+		     res->owner);
+		if (dlm_is_host_down(tmpret)) {
+			/* instead of logging the same network error over
+			 * and over, sleep here and wait for the heartbeat
+			 * to notice the node is dead.  times out after 5s. */
+			dlm_wait_for_node_death(dlm, res->owner,
+						DLM_NODE_DEATH_WAIT_MAX);
+			ret = DLM_RECOVERING;
+			mlog(0, "node %u died so returning DLM_RECOVERING "
+			     "from convert message!\n", res->owner);
+		} else {
+			ret = dlm_err_to_dlm_status(tmpret);
+		}
+	}
+
+	return ret;
+}
+
+/* handler for DLM_CONVERT_LOCK_MSG on master site
+ * locking:
+ *   caller needs:  none
+ *   taken:         takes and drop res->spinlock
+ *   held on exit:  none
+ * returns: DLM_NORMAL, DLM_IVLOCKID, DLM_BADARGS,
+ *          status from __dlmconvert_master
+ */
+int dlm_convert_lock_handler(struct o2net_msg *msg, u32 len, void *data,
+			     void **ret_data)
+{
+	struct dlm_ctxt *dlm = data;
+	struct dlm_convert_lock *cnv = (struct dlm_convert_lock *)msg->buf;
+	struct dlm_lock_resource *res = NULL;
+	struct dlm_lock *lock = NULL;
+	struct dlm_lock *tmp_lock;
+	struct dlm_lockstatus *lksb;
+	enum dlm_status status = DLM_NORMAL;
+	u32 flags;
+	int call_ast = 0, kick_thread = 0, ast_reserved = 0, wake = 0;
+
+	if (!dlm_grab(dlm)) {
+		dlm_error(DLM_REJECTED);
+		return DLM_REJECTED;
+	}
+
+	mlog_bug_on_msg(!dlm_domain_fully_joined(dlm),
+			"Domain %s not fully joined!\n", dlm->name);
+
+	if (cnv->namelen > DLM_LOCKID_NAME_MAX) {
+		status = DLM_IVBUFLEN;
+		dlm_error(status);
+		goto leave;
+	}
+
+	flags = be32_to_cpu(cnv->flags);
+
+	if ((flags & (LKM_PUT_LVB|LKM_GET_LVB)) ==
+	     (LKM_PUT_LVB|LKM_GET_LVB)) {
+		mlog(ML_ERROR, "both PUT and GET lvb specified\n");
+		status = DLM_BADARGS;
+		goto leave;
+	}
+
+	mlog(0, "lvb: %s\n", flags & LKM_PUT_LVB ? "put lvb" :
+	     (flags & LKM_GET_LVB ? "get lvb" : "none"));
+
+	status = DLM_IVLOCKID;
+	res = dlm_lookup_lockres(dlm, cnv->name, cnv->namelen);
+	if (!res) {
+		dlm_error(status);
+		goto leave;
+	}
+
+	spin_lock(&res->spinlock);
+	status = __dlm_lockres_state_to_status(res);
+	if (status != DLM_NORMAL) {
+		spin_unlock(&res->spinlock);
+		dlm_error(status);
+		goto leave;
+	}
+	list_for_each_entry(tmp_lock, &res->granted, list) {
+		if (tmp_lock->ml.cookie == cnv->cookie &&
+		    tmp_lock->ml.node == cnv->node_idx) {
+			lock = tmp_lock;
+			dlm_lock_get(lock);
+			break;
+		}
+	}
+	spin_unlock(&res->spinlock);
+	if (!lock) {
+		status = DLM_IVLOCKID;
+		mlog(ML_ERROR, "did not find lock to convert on grant queue! "
+			       "cookie=%u:%llu\n",
+		     dlm_get_lock_cookie_node(be64_to_cpu(cnv->cookie)),
+		     dlm_get_lock_cookie_seq(be64_to_cpu(cnv->cookie)));
+		dlm_print_one_lock_resource(res);
+		goto leave;
+	}
+
+	/* found the lock */
+	lksb = lock->lksb;
+
+	/* see if caller needed to get/put lvb */
+	if (flags & LKM_PUT_LVB) {
+		BUG_ON(lksb->flags & (DLM_LKSB_PUT_LVB|DLM_LKSB_GET_LVB));
+		lksb->flags |= DLM_LKSB_PUT_LVB;
+		memcpy(&lksb->lvb[0], &cnv->lvb[0], DLM_LVB_LEN);
+	} else if (flags & LKM_GET_LVB) {
+		BUG_ON(lksb->flags & (DLM_LKSB_PUT_LVB|DLM_LKSB_GET_LVB));
+		lksb->flags |= DLM_LKSB_GET_LVB;
+	}
+
+	spin_lock(&res->spinlock);
+	status = __dlm_lockres_state_to_status(res);
+	if (status == DLM_NORMAL) {
+		__dlm_lockres_reserve_ast(res);
+		ast_reserved = 1;
+		res->state |= DLM_LOCK_RES_IN_PROGRESS;
+		status = __dlmconvert_master(dlm, res, lock, flags,
+					     cnv->requested_type,
+					     &call_ast, &kick_thread);
+		res->state &= ~DLM_LOCK_RES_IN_PROGRESS;
+		wake = 1;
+	}
+	spin_unlock(&res->spinlock);
+	if (wake)
+		wake_up(&res->wq);
+
+	if (status != DLM_NORMAL) {
+		if (status != DLM_NOTQUEUED)
+			dlm_error(status);
+		lksb->flags &= ~(DLM_LKSB_GET_LVB|DLM_LKSB_PUT_LVB);
+	}
+
+leave:
+	if (lock)
+		dlm_lock_put(lock);
+
+	/* either queue the ast or release it, if reserved */
+	if (call_ast)
+		dlm_queue_ast(dlm, lock);
+	else if (ast_reserved)
+		dlm_lockres_release_ast(dlm, res);
+
+	if (kick_thread)
+		dlm_kick_thread(dlm, res);
+
+	if (res)
+		dlm_lockres_put(res);
+
+	dlm_put(dlm);
+
+	return status;
+}

fs/ocfs2/dlm/dlmconvert.h

@@ -0,0 +1,35 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dlmconvert.h
+ *
+ * Copyright (C) 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+#ifndef DLMCONVERT_H
+#define DLMCONVERT_H
+
+enum dlm_status dlmconvert_master(struct dlm_ctxt *dlm,
+				  struct dlm_lock_resource *res,
+				  struct dlm_lock *lock, int flags, int type);
+enum dlm_status dlmconvert_remote(struct dlm_ctxt *dlm,
+				  struct dlm_lock_resource *res,
+				  struct dlm_lock *lock, int flags, int type);
+
+#endif

fs/ocfs2/dlm/dlmdebug.h

@@ -0,0 +1,81 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dlmdebug.h
+ *
+ * Copyright (C) 2008 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+#ifndef DLMDEBUG_H
+#define DLMDEBUG_H
+
+void dlm_print_one_mle(struct dlm_master_list_entry *mle);
+
+#ifdef CONFIG_DEBUG_FS
+
+struct dlm_debug_ctxt {
+	struct kref debug_refcnt;
+	struct dentry *debug_state_dentry;
+	struct dentry *debug_lockres_dentry;
+	struct dentry *debug_mle_dentry;
+	struct dentry *debug_purgelist_dentry;
+};
+
+struct debug_lockres {
+	int dl_len;
+	char *dl_buf;
+	struct dlm_ctxt *dl_ctxt;
+	struct dlm_lock_resource *dl_res;
+};
+
+int dlm_debug_init(struct dlm_ctxt *dlm);
+void dlm_debug_shutdown(struct dlm_ctxt *dlm);
+
+int dlm_create_debugfs_subroot(struct dlm_ctxt *dlm);
+void dlm_destroy_debugfs_subroot(struct dlm_ctxt *dlm);
+
+int dlm_create_debugfs_root(void);
+void dlm_destroy_debugfs_root(void);
+
+#else
+
+static inline int dlm_debug_init(struct dlm_ctxt *dlm)
+{
+	return 0;
+}
+static inline void dlm_debug_shutdown(struct dlm_ctxt *dlm)
+{
+}
+static inline int dlm_create_debugfs_subroot(struct dlm_ctxt *dlm)
+{
+	return 0;
+}
+static inline void dlm_destroy_debugfs_subroot(struct dlm_ctxt *dlm)
+{
+}
+static inline int dlm_create_debugfs_root(void)
+{
+	return 0;
+}
+static inline void dlm_destroy_debugfs_root(void)
+{
+}
+
+#endif	/* CONFIG_DEBUG_FS */
+#endif	/* DLMDEBUG_H */

fs/ocfs2/dlm/dlmdomain.h

@@ -0,0 +1,36 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dlmdomain.h
+ *
+ * Copyright (C) 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+#ifndef DLMDOMAIN_H
+#define DLMDOMAIN_H
+
+extern spinlock_t dlm_domain_lock;
+extern struct list_head dlm_domains;
+
+int dlm_joined(struct dlm_ctxt *dlm);
+int dlm_shutting_down(struct dlm_ctxt *dlm);
+void dlm_fire_domain_eviction_callbacks(struct dlm_ctxt *dlm,
+					int node_num);
+
+#endif

fs/ocfs2/dlm/dlmlock.c

@@ -0,0 +1,761 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dlmlock.c
+ *
+ * underlying calls for lock creation
+ *
+ * Copyright (C) 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+#include <linux/init.h>
+#include <linux/sysctl.h>
+#include <linux/random.h>
+#include <linux/blkdev.h>
+#include <linux/socket.h>
+#include <linux/inet.h>
+#include <linux/spinlock.h>
+#include <linux/delay.h>
+
+
+#include "cluster/heartbeat.h"
+#include "cluster/nodemanager.h"
+#include "cluster/tcp.h"
+
+#include "dlmapi.h"
+#include "dlmcommon.h"
+
+#include "dlmconvert.h"
+
+#define MLOG_MASK_PREFIX ML_DLM
+#include "cluster/masklog.h"
+
+static struct kmem_cache *dlm_lock_cache;
+
+static DEFINE_SPINLOCK(dlm_cookie_lock);
+static u64 dlm_next_cookie = 1;
+
+static enum dlm_status dlm_send_remote_lock_request(struct dlm_ctxt *dlm,
+					       struct dlm_lock_resource *res,
+					       struct dlm_lock *lock, int flags);
+static void dlm_init_lock(struct dlm_lock *newlock, int type,
+			  u8 node, u64 cookie);
+static void dlm_lock_release(struct kref *kref);
+static void dlm_lock_detach_lockres(struct dlm_lock *lock);
+
+int dlm_init_lock_cache(void)
+{
+	dlm_lock_cache = kmem_cache_create("o2dlm_lock",
+					   sizeof(struct dlm_lock),
+					   0, SLAB_HWCACHE_ALIGN, NULL);
+	if (dlm_lock_cache == NULL)
+		return -ENOMEM;
+	return 0;
+}
+
+void dlm_destroy_lock_cache(void)
+{
+	if (dlm_lock_cache)
+		kmem_cache_destroy(dlm_lock_cache);
+}
+
+/* Tell us whether we can grant a new lock request.
+ * locking:
+ *   caller needs:  res->spinlock
+ *   taken:         none
+ *   held on exit:  none
+ * returns: 1 if the lock can be granted, 0 otherwise.
+ */
+static int dlm_can_grant_new_lock(struct dlm_lock_resource *res,
+				  struct dlm_lock *lock)
+{
+	struct dlm_lock *tmplock;
+
+	list_for_each_entry(tmplock, &res->granted, list) {
+		if (!dlm_lock_compatible(tmplock->ml.type, lock->ml.type))
+			return 0;
+	}
+
+	list_for_each_entry(tmplock, &res->converting, list) {
+		if (!dlm_lock_compatible(tmplock->ml.type, lock->ml.type))
+			return 0;
+		if (!dlm_lock_compatible(tmplock->ml.convert_type,
+					 lock->ml.type))
+			return 0;
+	}
+
+	return 1;
+}
+
+/* performs lock creation at the lockres master site
+ * locking:
+ *   caller needs:  none
+ *   taken:         takes and drops res->spinlock
+ *   held on exit:  none
+ * returns: DLM_NORMAL, DLM_NOTQUEUED
+ */
+static enum dlm_status dlmlock_master(struct dlm_ctxt *dlm,
+				      struct dlm_lock_resource *res,
+				      struct dlm_lock *lock, int flags)
+{
+	int call_ast = 0, kick_thread = 0;
+	enum dlm_status status = DLM_NORMAL;
+
+	mlog(0, "type=%d\n", lock->ml.type);
+
+	spin_lock(&res->spinlock);
+	/* if called from dlm_create_lock_handler, need to
+	 * ensure it will not sleep in dlm_wait_on_lockres */
+	status = __dlm_lockres_state_to_status(res);
+	if (status != DLM_NORMAL &&
+	    lock->ml.node != dlm->node_num) {
+		/* erf.  state changed after lock was dropped. */
+		spin_unlock(&res->spinlock);
+		dlm_error(status);
+		return status;
+	}
+	__dlm_wait_on_lockres(res);
+	__dlm_lockres_reserve_ast(res);
+
+	if (dlm_can_grant_new_lock(res, lock)) {
+		mlog(0, "I can grant this lock right away\n");
+		/* got it right away */
+		lock->lksb->status = DLM_NORMAL;
+		status = DLM_NORMAL;
+		dlm_lock_get(lock);
+		list_add_tail(&lock->list, &res->granted);
+
+		/* for the recovery lock, we can't allow the ast
+		 * to be queued since the dlmthread is already
+		 * frozen.  but the recovery lock is always locked
+		 * with LKM_NOQUEUE so we do not need the ast in
+		 * this special case */
+		if (!dlm_is_recovery_lock(res->lockname.name,
+					  res->lockname.len)) {
+			kick_thread = 1;
+			call_ast = 1;
+		} else {
+			mlog(0, "%s: returning DLM_NORMAL to "
+			     "node %u for reco lock\n", dlm->name,
+			     lock->ml.node);
+		}
+	} else {
+		/* for NOQUEUE request, unless we get the
+		 * lock right away, return DLM_NOTQUEUED */
+		if (flags & LKM_NOQUEUE) {
+			status = DLM_NOTQUEUED;
+			if (dlm_is_recovery_lock(res->lockname.name,
+						 res->lockname.len)) {
+				mlog(0, "%s: returning NOTQUEUED to "
+				     "node %u for reco lock\n", dlm->name,
+				     lock->ml.node);
+			}
+		} else {
+			status = DLM_NORMAL;
+			dlm_lock_get(lock);
+			list_add_tail(&lock->list, &res->blocked);
+			kick_thread = 1;
+		}
+	}
+
+	spin_unlock(&res->spinlock);
+	wake_up(&res->wq);
+
+	/* either queue the ast or release it */
+	if (call_ast)
+		dlm_queue_ast(dlm, lock);
+	else
+		dlm_lockres_release_ast(dlm, res);
+
+	dlm_lockres_calc_usage(dlm, res);
+	if (kick_thread)
+		dlm_kick_thread(dlm, res);
+
+	return status;
+}
+
+void dlm_revert_pending_lock(struct dlm_lock_resource *res,
+			     struct dlm_lock *lock)
+{
+	/* remove from local queue if it failed */
+	list_del_init(&lock->list);
+	lock->lksb->flags &= ~DLM_LKSB_GET_LVB;
+}
+
+
+/*
+ * locking:
+ *   caller needs:  none
+ *   taken:         takes and drops res->spinlock
+ *   held on exit:  none
+ * returns: DLM_DENIED, DLM_RECOVERING, or net status
+ */
+static enum dlm_status dlmlock_remote(struct dlm_ctxt *dlm,
+				      struct dlm_lock_resource *res,
+				      struct dlm_lock *lock, int flags)
+{
+	enum dlm_status status = DLM_DENIED;
+	int lockres_changed = 1;
+
+	mlog(0, "type=%d, lockres %.*s, flags = 0x%x\n",
+	     lock->ml.type, res->lockname.len,
+	     res->lockname.name, flags);
+
+	/*
+	 * Wait if resource is getting recovered, remastered, etc.
+	 * If the resource was remastered and new owner is self, then exit.
+	 */
+	spin_lock(&res->spinlock);
+	__dlm_wait_on_lockres(res);
+	if (res->owner == dlm->node_num) {
+		spin_unlock(&res->spinlock);
+		return DLM_RECOVERING;
+	}
+	res->state |= DLM_LOCK_RES_IN_PROGRESS;
+
+	/* add lock to local (secondary) queue */
+	dlm_lock_get(lock);
+	list_add_tail(&lock->list, &res->blocked);
+	lock->lock_pending = 1;
+	spin_unlock(&res->spinlock);
+
+	/* spec seems to say that you will get DLM_NORMAL when the lock
+	 * has been queued, meaning we need to wait for a reply here. */
+	status = dlm_send_remote_lock_request(dlm, res, lock, flags);
+
+	spin_lock(&res->spinlock);
+	res->state &= ~DLM_LOCK_RES_IN_PROGRESS;
+	lock->lock_pending = 0;
+	if (status != DLM_NORMAL) {
+		if (status == DLM_RECOVERING &&
+		    dlm_is_recovery_lock(res->lockname.name,
+					 res->lockname.len)) {
+			/* recovery lock was mastered by dead node.
+			 * we need to have calc_usage shoot down this
+			 * lockres and completely remaster it. */
+			mlog(0, "%s: recovery lock was owned by "
+			     "dead node %u, remaster it now.\n",
+			     dlm->name, res->owner);
+		} else if (status != DLM_NOTQUEUED) {
+			/*
+			 * DO NOT call calc_usage, as this would unhash
+			 * the remote lockres before we ever get to use
+			 * it.  treat as if we never made any change to
+			 * the lockres.
+			 */
+			lockres_changed = 0;
+			dlm_error(status);
+		}
+		dlm_revert_pending_lock(res, lock);
+		dlm_lock_put(lock);
+	} else if (dlm_is_recovery_lock(res->lockname.name,
+					res->lockname.len)) {
+		/* special case for the $RECOVERY lock.
+		 * there will never be an AST delivered to put
+		 * this lock on the proper secondary queue
+		 * (granted), so do it manually. */
+		mlog(0, "%s: $RECOVERY lock for this node (%u) is "
+		     "mastered by %u; got lock, manually granting (no ast)\n",
+		     dlm->name, dlm->node_num, res->owner);
+		list_move_tail(&lock->list, &res->granted);
+	}
+	spin_unlock(&res->spinlock);
+
+	if (lockres_changed)
+		dlm_lockres_calc_usage(dlm, res);
+
+	wake_up(&res->wq);
+	return status;
+}
+
+
+/* for remote lock creation.
+ * locking:
+ *   caller needs:  none, but need res->state & DLM_LOCK_RES_IN_PROGRESS
+ *   taken:         none
+ *   held on exit:  none
+ * returns: DLM_NOLOCKMGR, or net status
+ */
+static enum dlm_status dlm_send_remote_lock_request(struct dlm_ctxt *dlm,
+					       struct dlm_lock_resource *res,
+					       struct dlm_lock *lock, int flags)
+{
+	struct dlm_create_lock create;
+	int tmpret, status = 0;
+	enum dlm_status ret;
+
+	memset(&create, 0, sizeof(create));
+	create.node_idx = dlm->node_num;
+	create.requested_type = lock->ml.type;
+	create.cookie = lock->ml.cookie;
+	create.namelen = res->lockname.len;
+	create.flags = cpu_to_be32(flags);
+	memcpy(create.name, res->lockname.name, create.namelen);
+
+	tmpret = o2net_send_message(DLM_CREATE_LOCK_MSG, dlm->key, &create,
+				    sizeof(create), res->owner, &status);
+	if (tmpret >= 0) {
+		ret = status;
+		if (ret == DLM_REJECTED) {
+			mlog(ML_ERROR, "%s: res %.*s, Stale lockres no longer "
+			     "owned by node %u. That node is coming back up "
+			     "currently.\n", dlm->name, create.namelen,
+			     create.name, res->owner);
+			dlm_print_one_lock_resource(res);
+			BUG();
+		}
+	} else {
+		mlog(ML_ERROR, "%s: res %.*s, Error %d send CREATE LOCK to "
+		     "node %u\n", dlm->name, create.namelen, create.name,
+		     tmpret, res->owner);
+		if (dlm_is_host_down(tmpret))
+			ret = DLM_RECOVERING;
+		else
+			ret = dlm_err_to_dlm_status(tmpret);
+	}
+
+	return ret;
+}
+
+void dlm_lock_get(struct dlm_lock *lock)
+{
+	kref_get(&lock->lock_refs);
+}
+
+void dlm_lock_put(struct dlm_lock *lock)
+{
+	kref_put(&lock->lock_refs, dlm_lock_release);
+}
+
+static void dlm_lock_release(struct kref *kref)
+{
+	struct dlm_lock *lock;
+
+	lock = container_of(kref, struct dlm_lock, lock_refs);
+
+	BUG_ON(!list_empty(&lock->list));
+	BUG_ON(!list_empty(&lock->ast_list));
+	BUG_ON(!list_empty(&lock->bast_list));
+	BUG_ON(lock->ast_pending);
+	BUG_ON(lock->bast_pending);
+
+	dlm_lock_detach_lockres(lock);
+
+	if (lock->lksb_kernel_allocated) {
+		mlog(0, "freeing kernel-allocated lksb\n");
+		kfree(lock->lksb);
+	}
+	kmem_cache_free(dlm_lock_cache, lock);
+}
+
+/* associate a lock with it's lockres, getting a ref on the lockres */
+void dlm_lock_attach_lockres(struct dlm_lock *lock,
+			     struct dlm_lock_resource *res)
+{
+	dlm_lockres_get(res);
+	lock->lockres = res;
+}
+
+/* drop ref on lockres, if there is still one associated with lock */
+static void dlm_lock_detach_lockres(struct dlm_lock *lock)
+{
+	struct dlm_lock_resource *res;
+
+	res = lock->lockres;
+	if (res) {
+		lock->lockres = NULL;
+		mlog(0, "removing lock's lockres reference\n");
+		dlm_lockres_put(res);
+	}
+}
+
+static void dlm_init_lock(struct dlm_lock *newlock, int type,
+			  u8 node, u64 cookie)
+{
+	INIT_LIST_HEAD(&newlock->list);
+	INIT_LIST_HEAD(&newlock->ast_list);
+	INIT_LIST_HEAD(&newlock->bast_list);
+	spin_lock_init(&newlock->spinlock);
+	newlock->ml.type = type;
+	newlock->ml.convert_type = LKM_IVMODE;
+	newlock->ml.highest_blocked = LKM_IVMODE;
+	newlock->ml.node = node;
+	newlock->ml.pad1 = 0;
+	newlock->ml.list = 0;
+	newlock->ml.flags = 0;
+	newlock->ast = NULL;
+	newlock->bast = NULL;
+	newlock->astdata = NULL;
+	newlock->ml.cookie = cpu_to_be64(cookie);
+	newlock->ast_pending = 0;
+	newlock->bast_pending = 0;
+	newlock->convert_pending = 0;
+	newlock->lock_pending = 0;
+	newlock->unlock_pending = 0;
+	newlock->cancel_pending = 0;
+	newlock->lksb_kernel_allocated = 0;
+
+	kref_init(&newlock->lock_refs);
+}
+
+struct dlm_lock * dlm_new_lock(int type, u8 node, u64 cookie,
+			       struct dlm_lockstatus *lksb)
+{
+	struct dlm_lock *lock;
+	int kernel_allocated = 0;
+
+	lock = kmem_cache_zalloc(dlm_lock_cache, GFP_NOFS);
+	if (!lock)
+		return NULL;
+
+	if (!lksb) {
+		/* zero memory only if kernel-allocated */
+		lksb = kzalloc(sizeof(*lksb), GFP_NOFS);
+		if (!lksb) {
+			kmem_cache_free(dlm_lock_cache, lock);
+			return NULL;
+		}
+		kernel_allocated = 1;
+	}
+
+	dlm_init_lock(lock, type, node, cookie);
+	if (kernel_allocated)
+		lock->lksb_kernel_allocated = 1;
+	lock->lksb = lksb;
+	lksb->lockid = lock;
+	return lock;
+}
+
+/* handler for lock creation net message
+ * locking:
+ *   caller needs:  none
+ *   taken:         takes and drops res->spinlock
+ *   held on exit:  none
+ * returns: DLM_NORMAL, DLM_SYSERR, DLM_IVLOCKID, DLM_NOTQUEUED
+ */
+int dlm_create_lock_handler(struct o2net_msg *msg, u32 len, void *data,
+			    void **ret_data)
+{
+	struct dlm_ctxt *dlm = data;
+	struct dlm_create_lock *create = (struct dlm_create_lock *)msg->buf;
+	struct dlm_lock_resource *res = NULL;
+	struct dlm_lock *newlock = NULL;
+	struct dlm_lockstatus *lksb = NULL;
+	enum dlm_status status = DLM_NORMAL;
+	char *name;
+	unsigned int namelen;
+
+	BUG_ON(!dlm);
+
+	if (!dlm_grab(dlm))
+		return DLM_REJECTED;
+
+	name = create->name;
+	namelen = create->namelen;
+	status = DLM_REJECTED;
+	if (!dlm_domain_fully_joined(dlm)) {
+		mlog(ML_ERROR, "Domain %s not fully joined, but node %u is "
+		     "sending a create_lock message for lock %.*s!\n",
+		     dlm->name, create->node_idx, namelen, name);
+		dlm_error(status);
+		goto leave;
+	}
+
+	status = DLM_IVBUFLEN;
+	if (namelen > DLM_LOCKID_NAME_MAX) {
+		dlm_error(status);
+		goto leave;
+	}
+
+	status = DLM_SYSERR;
+	newlock = dlm_new_lock(create->requested_type,
+			       create->node_idx,
+			       be64_to_cpu(create->cookie), NULL);
+	if (!newlock) {
+		dlm_error(status);
+		goto leave;
+	}
+
+	lksb = newlock->lksb;
+
+	if (be32_to_cpu(create->flags) & LKM_GET_LVB) {
+		lksb->flags |= DLM_LKSB_GET_LVB;
+		mlog(0, "set DLM_LKSB_GET_LVB flag\n");
+	}
+
+	status = DLM_IVLOCKID;
+	res = dlm_lookup_lockres(dlm, name, namelen);
+	if (!res) {
+		dlm_error(status);
+		goto leave;
+	}
+
+	spin_lock(&res->spinlock);
+	status = __dlm_lockres_state_to_status(res);
+	spin_unlock(&res->spinlock);
+
+	if (status != DLM_NORMAL) {
+		mlog(0, "lockres recovering/migrating/in-progress\n");
+		goto leave;
+	}
+
+	dlm_lock_attach_lockres(newlock, res);
+
+	status = dlmlock_master(dlm, res, newlock, be32_to_cpu(create->flags));
+leave:
+	if (status != DLM_NORMAL)
+		if (newlock)
+			dlm_lock_put(newlock);
+
+	if (res)
+		dlm_lockres_put(res);
+
+	dlm_put(dlm);
+
+	return status;
+}
+
+
+/* fetch next node-local (u8 nodenum + u56 cookie) into u64 */
+static inline void dlm_get_next_cookie(u8 node_num, u64 *cookie)
+{
+	u64 tmpnode = node_num;
+
+	/* shift single byte of node num into top 8 bits */
+	tmpnode <<= 56;
+
+	spin_lock(&dlm_cookie_lock);
+	*cookie = (dlm_next_cookie | tmpnode);
+	if (++dlm_next_cookie & 0xff00000000000000ull) {
+		mlog(0, "This node's cookie will now wrap!\n");
+		dlm_next_cookie = 1;
+	}
+	spin_unlock(&dlm_cookie_lock);
+}
+
+enum dlm_status dlmlock(struct dlm_ctxt *dlm, int mode,
+			struct dlm_lockstatus *lksb, int flags,
+			const char *name, int namelen, dlm_astlockfunc_t *ast,
+			void *data, dlm_bastlockfunc_t *bast)
+{
+	enum dlm_status status;
+	struct dlm_lock_resource *res = NULL;
+	struct dlm_lock *lock = NULL;
+	int convert = 0, recovery = 0;
+
+	/* yes this function is a mess.
+	 * TODO: clean this up.  lots of common code in the
+	 *       lock and convert paths, especially in the retry blocks */
+	if (!lksb) {
+		dlm_error(DLM_BADARGS);
+		return DLM_BADARGS;
+	}
+
+	status = DLM_BADPARAM;
+	if (mode != LKM_EXMODE && mode != LKM_PRMODE && mode != LKM_NLMODE) {
+		dlm_error(status);
+		goto error;
+	}
+
+	if (flags & ~LKM_VALID_FLAGS) {
+		dlm_error(status);
+		goto error;
+	}
+
+	convert = (flags & LKM_CONVERT);
+	recovery = (flags & LKM_RECOVERY);
+
+	if (recovery &&
+	    (!dlm_is_recovery_lock(name, namelen) || convert) ) {
+		dlm_error(status);
+		goto error;
+	}
+	if (convert && (flags & LKM_LOCAL)) {
+		mlog(ML_ERROR, "strange LOCAL convert request!\n");
+		goto error;
+	}
+
+	if (convert) {
+		/* CONVERT request */
+
+		/* if converting, must pass in a valid dlm_lock */
+		lock = lksb->lockid;
+		if (!lock) {
+			mlog(ML_ERROR, "NULL lock pointer in convert "
+			     "request\n");
+			goto error;
+		}
+
+		res = lock->lockres;
+		if (!res) {
+			mlog(ML_ERROR, "NULL lockres pointer in convert "
+			     "request\n");
+			goto error;
+		}
+		dlm_lockres_get(res);
+
+		/* XXX: for ocfs2 purposes, the ast/bast/astdata/lksb are
+	 	 * static after the original lock call.  convert requests will
+		 * ensure that everything is the same, or return DLM_BADARGS.
+	 	 * this means that DLM_DENIED_NOASTS will never be returned.
+	 	 */
+		if (lock->lksb != lksb || lock->ast != ast ||
+		    lock->bast != bast || lock->astdata != data) {
+			status = DLM_BADARGS;
+			mlog(ML_ERROR, "new args:  lksb=%p, ast=%p, bast=%p, "
+			     "astdata=%p\n", lksb, ast, bast, data);
+			mlog(ML_ERROR, "orig args: lksb=%p, ast=%p, bast=%p, "
+			     "astdata=%p\n", lock->lksb, lock->ast,
+			     lock->bast, lock->astdata);
+			goto error;
+		}
+retry_convert:
+		dlm_wait_for_recovery(dlm);
+
+		if (res->owner == dlm->node_num)
+			status = dlmconvert_master(dlm, res, lock, flags, mode);
+		else
+			status = dlmconvert_remote(dlm, res, lock, flags, mode);
+		if (status == DLM_RECOVERING || status == DLM_MIGRATING ||
+		    status == DLM_FORWARD) {
+			/* for now, see how this works without sleeping
+			 * and just retry right away.  I suspect the reco
+			 * or migration will complete fast enough that
+			 * no waiting will be necessary */
+			mlog(0, "retrying convert with migration/recovery/"
+			     "in-progress\n");
+			msleep(100);
+			goto retry_convert;
+		}
+	} else {
+		u64 tmpcookie;
+
+		/* LOCK request */
+		status = DLM_BADARGS;
+		if (!name) {
+			dlm_error(status);
+			goto error;
+		}
+
+		status = DLM_IVBUFLEN;
+		if (namelen > DLM_LOCKID_NAME_MAX || namelen < 1) {
+			dlm_error(status);
+			goto error;
+		}
+
+		dlm_get_next_cookie(dlm->node_num, &tmpcookie);
+		lock = dlm_new_lock(mode, dlm->node_num, tmpcookie, lksb);
+		if (!lock) {
+			dlm_error(status);
+			goto error;
+		}
+
+		if (!recovery)
+			dlm_wait_for_recovery(dlm);
+
+		/* find or create the lock resource */
+		res = dlm_get_lock_resource(dlm, name, namelen, flags);
+		if (!res) {
+			status = DLM_IVLOCKID;
+			dlm_error(status);
+			goto error;
+		}
+
+		mlog(0, "type=%d, flags = 0x%x\n", mode, flags);
+		mlog(0, "creating lock: lock=%p res=%p\n", lock, res);
+
+		dlm_lock_attach_lockres(lock, res);
+		lock->ast = ast;
+		lock->bast = bast;
+		lock->astdata = data;
+
+retry_lock:
+		if (flags & LKM_VALBLK) {
+			mlog(0, "LKM_VALBLK passed by caller\n");
+
+			/* LVB requests for non PR, PW or EX locks are
+			 * ignored. */
+			if (mode < LKM_PRMODE)
+				flags &= ~LKM_VALBLK;
+			else {
+				flags |= LKM_GET_LVB;
+				lock->lksb->flags |= DLM_LKSB_GET_LVB;
+			}
+		}
+
+		if (res->owner == dlm->node_num)
+			status = dlmlock_master(dlm, res, lock, flags);
+		else
+			status = dlmlock_remote(dlm, res, lock, flags);
+
+		if (status == DLM_RECOVERING || status == DLM_MIGRATING ||
+		    status == DLM_FORWARD) {
+			msleep(100);
+			if (recovery) {
+				if (status != DLM_RECOVERING)
+					goto retry_lock;
+				/* wait to see the node go down, then
+				 * drop down and allow the lockres to
+				 * get cleaned up.  need to remaster. */
+				dlm_wait_for_node_death(dlm, res->owner,
+						DLM_NODE_DEATH_WAIT_MAX);
+			} else {
+				dlm_wait_for_recovery(dlm);
+				goto retry_lock;
+			}
+		}
+
+		/* Inflight taken in dlm_get_lock_resource() is dropped here */
+		spin_lock(&res->spinlock);
+		dlm_lockres_drop_inflight_ref(dlm, res);
+		spin_unlock(&res->spinlock);
+
+		dlm_lockres_calc_usage(dlm, res);
+		dlm_kick_thread(dlm, res);
+
+		if (status != DLM_NORMAL) {
+			lock->lksb->flags &= ~DLM_LKSB_GET_LVB;
+			if (status != DLM_NOTQUEUED)
+				dlm_error(status);
+			goto error;
+		}
+	}
+
+error:
+	if (status != DLM_NORMAL) {
+		if (lock && !convert)
+			dlm_lock_put(lock);
+		// this is kind of unnecessary
+		lksb->status = status;
+	}
+
+	/* put lockres ref from the convert path
+	 * or from dlm_get_lock_resource */
+	if (res)
+		dlm_lockres_put(res);
+
+	return status;
+}
+EXPORT_SYMBOL_GPL(dlmlock);

fs/ocfs2/dlm/dlmthread.c

@@ -0,0 +1,756 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dlmthread.c
+ *
+ * standalone DLM module
+ *
+ * Copyright (C) 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/highmem.h>
+#include <linux/init.h>
+#include <linux/sysctl.h>
+#include <linux/random.h>
+#include <linux/blkdev.h>
+#include <linux/socket.h>
+#include <linux/inet.h>
+#include <linux/timer.h>
+#include <linux/kthread.h>
+#include <linux/delay.h>
+
+
+#include "cluster/heartbeat.h"
+#include "cluster/nodemanager.h"
+#include "cluster/tcp.h"
+
+#include "dlmapi.h"
+#include "dlmcommon.h"
+#include "dlmdomain.h"
+
+#define MLOG_MASK_PREFIX (ML_DLM|ML_DLM_THREAD)
+#include "cluster/masklog.h"
+
+static int dlm_thread(void *data);
+static void dlm_flush_asts(struct dlm_ctxt *dlm);
+
+#define dlm_lock_is_remote(dlm, lock)     ((lock)->ml.node != (dlm)->node_num)
+
+/* will exit holding res->spinlock, but may drop in function */
+/* waits until flags are cleared on res->state */
+void __dlm_wait_on_lockres_flags(struct dlm_lock_resource *res, int flags)
+{
+	DECLARE_WAITQUEUE(wait, current);
+
+	assert_spin_locked(&res->spinlock);
+
+	add_wait_queue(&res->wq, &wait);
+repeat:
+	set_current_state(TASK_UNINTERRUPTIBLE);
+	if (res->state & flags) {
+		spin_unlock(&res->spinlock);
+		schedule();
+		spin_lock(&res->spinlock);
+		goto repeat;
+	}
+	remove_wait_queue(&res->wq, &wait);
+	__set_current_state(TASK_RUNNING);
+}
+
+int __dlm_lockres_has_locks(struct dlm_lock_resource *res)
+{
+	if (list_empty(&res->granted) &&
+	    list_empty(&res->converting) &&
+	    list_empty(&res->blocked))
+		return 0;
+	return 1;
+}
+
+/* "unused": the lockres has no locks, is not on the dirty list,
+ * has no inflight locks (in the gap between mastery and acquiring
+ * the first lock), and has no bits in its refmap.
+ * truly ready to be freed. */
+int __dlm_lockres_unused(struct dlm_lock_resource *res)
+{
+	int bit;
+
+	assert_spin_locked(&res->spinlock);
+
+	if (__dlm_lockres_has_locks(res))
+		return 0;
+
+	/* Locks are in the process of being created */
+	if (res->inflight_locks)
+		return 0;
+
+	if (!list_empty(&res->dirty) || res->state & DLM_LOCK_RES_DIRTY)
+		return 0;
+
+	if (res->state & DLM_LOCK_RES_RECOVERING)
+		return 0;
+
+	/* Another node has this resource with this node as the master */
+	bit = find_next_bit(res->refmap, O2NM_MAX_NODES, 0);
+	if (bit < O2NM_MAX_NODES)
+		return 0;
+
+	return 1;
+}
+
+
+/* Call whenever you may have added or deleted something from one of
+ * the lockres queue's. This will figure out whether it belongs on the
+ * unused list or not and does the appropriate thing. */
+void __dlm_lockres_calc_usage(struct dlm_ctxt *dlm,
+			      struct dlm_lock_resource *res)
+{
+	assert_spin_locked(&dlm->spinlock);
+	assert_spin_locked(&res->spinlock);
+
+	if (__dlm_lockres_unused(res)){
+		if (list_empty(&res->purge)) {
+			mlog(0, "%s: Adding res %.*s to purge list\n",
+			     dlm->name, res->lockname.len, res->lockname.name);
+
+			res->last_used = jiffies;
+			dlm_lockres_get(res);
+			list_add_tail(&res->purge, &dlm->purge_list);
+			dlm->purge_count++;
+		}
+	} else if (!list_empty(&res->purge)) {
+		mlog(0, "%s: Removing res %.*s from purge list\n",
+		     dlm->name, res->lockname.len, res->lockname.name);
+
+		list_del_init(&res->purge);
+		dlm_lockres_put(res);
+		dlm->purge_count--;
+	}
+}
+
+void dlm_lockres_calc_usage(struct dlm_ctxt *dlm,
+			    struct dlm_lock_resource *res)
+{
+	spin_lock(&dlm->spinlock);
+	spin_lock(&res->spinlock);
+
+	__dlm_lockres_calc_usage(dlm, res);
+
+	spin_unlock(&res->spinlock);
+	spin_unlock(&dlm->spinlock);
+}
+
+static void dlm_purge_lockres(struct dlm_ctxt *dlm,
+			     struct dlm_lock_resource *res)
+{
+	int master;
+	int ret = 0;
+
+	assert_spin_locked(&dlm->spinlock);
+	assert_spin_locked(&res->spinlock);
+
+	master = (res->owner == dlm->node_num);
+
+	mlog(0, "%s: Purging res %.*s, master %d\n", dlm->name,
+	     res->lockname.len, res->lockname.name, master);
+
+	if (!master) {
+		res->state |= DLM_LOCK_RES_DROPPING_REF;
+		/* drop spinlock...  retake below */
+		spin_unlock(&res->spinlock);
+		spin_unlock(&dlm->spinlock);
+
+		spin_lock(&res->spinlock);
+		/* This ensures that clear refmap is sent after the set */
+		__dlm_wait_on_lockres_flags(res, DLM_LOCK_RES_SETREF_INPROG);
+		spin_unlock(&res->spinlock);
+
+		/* clear our bit from the master's refmap, ignore errors */
+		ret = dlm_drop_lockres_ref(dlm, res);
+		if (ret < 0) {
+			if (!dlm_is_host_down(ret))
+				BUG();
+		}
+		spin_lock(&dlm->spinlock);
+		spin_lock(&res->spinlock);
+	}
+
+	if (!list_empty(&res->purge)) {
+		mlog(0, "%s: Removing res %.*s from purgelist, master %d\n",
+		     dlm->name, res->lockname.len, res->lockname.name, master);
+		list_del_init(&res->purge);
+		dlm_lockres_put(res);
+		dlm->purge_count--;
+	}
+
+	if (!__dlm_lockres_unused(res)) {
+		mlog(ML_ERROR, "%s: res %.*s in use after deref\n",
+		     dlm->name, res->lockname.len, res->lockname.name);
+		__dlm_print_one_lock_resource(res);
+		BUG();
+	}
+
+	__dlm_unhash_lockres(dlm, res);
+
+	/* lockres is not in the hash now.  drop the flag and wake up
+	 * any processes waiting in dlm_get_lock_resource. */
+	if (!master) {
+		res->state &= ~DLM_LOCK_RES_DROPPING_REF;
+		spin_unlock(&res->spinlock);
+		wake_up(&res->wq);
+	} else
+		spin_unlock(&res->spinlock);
+}
+
+static void dlm_run_purge_list(struct dlm_ctxt *dlm,
+			       int purge_now)
+{
+	unsigned int run_max, unused;
+	unsigned long purge_jiffies;
+	struct dlm_lock_resource *lockres;
+
+	spin_lock(&dlm->spinlock);
+	run_max = dlm->purge_count;
+
+	while(run_max && !list_empty(&dlm->purge_list)) {
+		run_max--;
+
+		lockres = list_entry(dlm->purge_list.next,
+				     struct dlm_lock_resource, purge);
+
+		spin_lock(&lockres->spinlock);
+
+		purge_jiffies = lockres->last_used +
+			msecs_to_jiffies(DLM_PURGE_INTERVAL_MS);
+
+		/* Make sure that we want to be processing this guy at
+		 * this time. */
+		if (!purge_now && time_after(purge_jiffies, jiffies)) {
+			/* Since resources are added to the purge list
+			 * in tail order, we can stop at the first
+			 * unpurgable resource -- anyone added after
+			 * him will have a greater last_used value */
+			spin_unlock(&lockres->spinlock);
+			break;
+		}
+
+		/* Status of the lockres *might* change so double
+		 * check. If the lockres is unused, holding the dlm
+		 * spinlock will prevent people from getting and more
+		 * refs on it. */
+		unused = __dlm_lockres_unused(lockres);
+		if (!unused ||
+		    (lockres->state & DLM_LOCK_RES_MIGRATING) ||
+		    (lockres->inflight_assert_workers != 0)) {
+			mlog(0, "%s: res %.*s is in use or being remastered, "
+			     "used %d, state %d, assert master workers %u\n",
+			     dlm->name, lockres->lockname.len,
+			     lockres->lockname.name,
+			     !unused, lockres->state,
+			     lockres->inflight_assert_workers);
+			list_move_tail(&lockres->purge, &dlm->purge_list);
+			spin_unlock(&lockres->spinlock);
+			continue;
+		}
+
+		dlm_lockres_get(lockres);
+
+		dlm_purge_lockres(dlm, lockres);
+
+		dlm_lockres_put(lockres);
+
+		/* Avoid adding any scheduling latencies */
+		cond_resched_lock(&dlm->spinlock);
+	}
+
+	spin_unlock(&dlm->spinlock);
+}
+
+static void dlm_shuffle_lists(struct dlm_ctxt *dlm,
+			      struct dlm_lock_resource *res)
+{
+	struct dlm_lock *lock, *target;
+	int can_grant = 1;
+
+	/*
+	 * Because this function is called with the lockres
+	 * spinlock, and because we know that it is not migrating/
+	 * recovering/in-progress, it is fine to reserve asts and
+	 * basts right before queueing them all throughout
+	 */
+	assert_spin_locked(&dlm->ast_lock);
+	assert_spin_locked(&res->spinlock);
+	BUG_ON((res->state & (DLM_LOCK_RES_MIGRATING|
+			      DLM_LOCK_RES_RECOVERING|
+			      DLM_LOCK_RES_IN_PROGRESS)));
+
+converting:
+	if (list_empty(&res->converting))
+		goto blocked;
+	mlog(0, "%s: res %.*s has locks on the convert queue\n", dlm->name,
+	     res->lockname.len, res->lockname.name);
+
+	target = list_entry(res->converting.next, struct dlm_lock, list);
+	if (target->ml.convert_type == LKM_IVMODE) {
+		mlog(ML_ERROR, "%s: res %.*s converting lock to invalid mode\n",
+		     dlm->name, res->lockname.len, res->lockname.name);
+		BUG();
+	}
+	list_for_each_entry(lock, &res->granted, list) {
+		if (lock==target)
+			continue;
+		if (!dlm_lock_compatible(lock->ml.type,
+					 target->ml.convert_type)) {
+			can_grant = 0;
+			/* queue the BAST if not already */
+			if (lock->ml.highest_blocked == LKM_IVMODE) {
+				__dlm_lockres_reserve_ast(res);
+				__dlm_queue_bast(dlm, lock);
+			}
+			/* update the highest_blocked if needed */
+			if (lock->ml.highest_blocked < target->ml.convert_type)
+				lock->ml.highest_blocked =
+					target->ml.convert_type;
+		}
+	}
+
+	list_for_each_entry(lock, &res->converting, list) {
+		if (lock==target)
+			continue;
+		if (!dlm_lock_compatible(lock->ml.type,
+					 target->ml.convert_type)) {
+			can_grant = 0;
+			if (lock->ml.highest_blocked == LKM_IVMODE) {
+				__dlm_lockres_reserve_ast(res);
+				__dlm_queue_bast(dlm, lock);
+			}
+			if (lock->ml.highest_blocked < target->ml.convert_type)
+				lock->ml.highest_blocked =
+					target->ml.convert_type;
+		}
+	}
+
+	/* we can convert the lock */
+	if (can_grant) {
+		spin_lock(&target->spinlock);
+		BUG_ON(target->ml.highest_blocked != LKM_IVMODE);
+
+		mlog(0, "%s: res %.*s, AST for Converting lock %u:%llu, type "
+		     "%d => %d, node %u\n", dlm->name, res->lockname.len,
+		     res->lockname.name,
+		     dlm_get_lock_cookie_node(be64_to_cpu(target->ml.cookie)),
+		     dlm_get_lock_cookie_seq(be64_to_cpu(target->ml.cookie)),
+		     target->ml.type,
+		     target->ml.convert_type, target->ml.node);
+
+		target->ml.type = target->ml.convert_type;
+		target->ml.convert_type = LKM_IVMODE;
+		list_move_tail(&target->list, &res->granted);
+
+		BUG_ON(!target->lksb);
+		target->lksb->status = DLM_NORMAL;
+
+		spin_unlock(&target->spinlock);
+
+		__dlm_lockres_reserve_ast(res);
+		__dlm_queue_ast(dlm, target);
+		/* go back and check for more */
+		goto converting;
+	}
+
+blocked:
+	if (list_empty(&res->blocked))
+		goto leave;
+	target = list_entry(res->blocked.next, struct dlm_lock, list);
+
+	list_for_each_entry(lock, &res->granted, list) {
+		if (lock==target)
+			continue;
+		if (!dlm_lock_compatible(lock->ml.type, target->ml.type)) {
+			can_grant = 0;
+			if (lock->ml.highest_blocked == LKM_IVMODE) {
+				__dlm_lockres_reserve_ast(res);
+				__dlm_queue_bast(dlm, lock);
+			}
+			if (lock->ml.highest_blocked < target->ml.type)
+				lock->ml.highest_blocked = target->ml.type;
+		}
+	}
+
+	list_for_each_entry(lock, &res->converting, list) {
+		if (lock==target)
+			continue;
+		if (!dlm_lock_compatible(lock->ml.type, target->ml.type)) {
+			can_grant = 0;
+			if (lock->ml.highest_blocked == LKM_IVMODE) {
+				__dlm_lockres_reserve_ast(res);
+				__dlm_queue_bast(dlm, lock);
+			}
+			if (lock->ml.highest_blocked < target->ml.type)
+				lock->ml.highest_blocked = target->ml.type;
+		}
+	}
+
+	/* we can grant the blocked lock (only
+	 * possible if converting list empty) */
+	if (can_grant) {
+		spin_lock(&target->spinlock);
+		BUG_ON(target->ml.highest_blocked != LKM_IVMODE);
+
+		mlog(0, "%s: res %.*s, AST for Blocked lock %u:%llu, type %d, "
+		     "node %u\n", dlm->name, res->lockname.len,
+		     res->lockname.name,
+		     dlm_get_lock_cookie_node(be64_to_cpu(target->ml.cookie)),
+		     dlm_get_lock_cookie_seq(be64_to_cpu(target->ml.cookie)),
+		     target->ml.type, target->ml.node);
+
+		/* target->ml.type is already correct */
+		list_move_tail(&target->list, &res->granted);
+
+		BUG_ON(!target->lksb);
+		target->lksb->status = DLM_NORMAL;
+
+		spin_unlock(&target->spinlock);
+
+		__dlm_lockres_reserve_ast(res);
+		__dlm_queue_ast(dlm, target);
+		/* go back and check for more */
+		goto converting;
+	}
+
+leave:
+	return;
+}
+
+/* must have NO locks when calling this with res !=NULL * */
+void dlm_kick_thread(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
+{
+	if (res) {
+		spin_lock(&dlm->spinlock);
+		spin_lock(&res->spinlock);
+		__dlm_dirty_lockres(dlm, res);
+		spin_unlock(&res->spinlock);
+		spin_unlock(&dlm->spinlock);
+	}
+	wake_up(&dlm->dlm_thread_wq);
+}
+
+void __dlm_dirty_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
+{
+	assert_spin_locked(&dlm->spinlock);
+	assert_spin_locked(&res->spinlock);
+
+	/* don't shuffle secondary queues */
+	if ((res->owner == dlm->node_num)) {
+		if (res->state & (DLM_LOCK_RES_MIGRATING |
+				  DLM_LOCK_RES_BLOCK_DIRTY))
+		    return;
+
+		if (list_empty(&res->dirty)) {
+			/* ref for dirty_list */
+			dlm_lockres_get(res);
+			list_add_tail(&res->dirty, &dlm->dirty_list);
+			res->state |= DLM_LOCK_RES_DIRTY;
+		}
+	}
+
+	mlog(0, "%s: res %.*s\n", dlm->name, res->lockname.len,
+	     res->lockname.name);
+}
+
+
+/* Launch the NM thread for the mounted volume */
+int dlm_launch_thread(struct dlm_ctxt *dlm)
+{
+	mlog(0, "Starting dlm_thread...\n");
+
+	dlm->dlm_thread_task = kthread_run(dlm_thread, dlm, "dlm_thread");
+	if (IS_ERR(dlm->dlm_thread_task)) {
+		mlog_errno(PTR_ERR(dlm->dlm_thread_task));
+		dlm->dlm_thread_task = NULL;
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+void dlm_complete_thread(struct dlm_ctxt *dlm)
+{
+	if (dlm->dlm_thread_task) {
+		mlog(ML_KTHREAD, "Waiting for dlm thread to exit\n");
+		kthread_stop(dlm->dlm_thread_task);
+		dlm->dlm_thread_task = NULL;
+	}
+}
+
+static int dlm_dirty_list_empty(struct dlm_ctxt *dlm)
+{
+	int empty;
+
+	spin_lock(&dlm->spinlock);
+	empty = list_empty(&dlm->dirty_list);
+	spin_unlock(&dlm->spinlock);
+
+	return empty;
+}
+
+static void dlm_flush_asts(struct dlm_ctxt *dlm)
+{
+	int ret;
+	struct dlm_lock *lock;
+	struct dlm_lock_resource *res;
+	u8 hi;
+
+	spin_lock(&dlm->ast_lock);
+	while (!list_empty(&dlm->pending_asts)) {
+		lock = list_entry(dlm->pending_asts.next,
+				  struct dlm_lock, ast_list);
+		/* get an extra ref on lock */
+		dlm_lock_get(lock);
+		res = lock->lockres;
+		mlog(0, "%s: res %.*s, Flush AST for lock %u:%llu, type %d, "
+		     "node %u\n", dlm->name, res->lockname.len,
+		     res->lockname.name,
+		     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
+		     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
+		     lock->ml.type, lock->ml.node);
+
+		BUG_ON(!lock->ast_pending);
+
+		/* remove from list (including ref) */
+		list_del_init(&lock->ast_list);
+		dlm_lock_put(lock);
+		spin_unlock(&dlm->ast_lock);
+
+		if (lock->ml.node != dlm->node_num) {
+			ret = dlm_do_remote_ast(dlm, res, lock);
+			if (ret < 0)
+				mlog_errno(ret);
+		} else
+			dlm_do_local_ast(dlm, res, lock);
+
+		spin_lock(&dlm->ast_lock);
+
+		/* possible that another ast was queued while
+		 * we were delivering the last one */
+		if (!list_empty(&lock->ast_list)) {
+			mlog(0, "%s: res %.*s, AST queued while flushing last "
+			     "one\n", dlm->name, res->lockname.len,
+			     res->lockname.name);
+		} else
+			lock->ast_pending = 0;
+
+		/* drop the extra ref.
+		 * this may drop it completely. */
+		dlm_lock_put(lock);
+		dlm_lockres_release_ast(dlm, res);
+	}
+
+	while (!list_empty(&dlm->pending_basts)) {
+		lock = list_entry(dlm->pending_basts.next,
+				  struct dlm_lock, bast_list);
+		/* get an extra ref on lock */
+		dlm_lock_get(lock);
+		res = lock->lockres;
+
+		BUG_ON(!lock->bast_pending);
+
+		/* get the highest blocked lock, and reset */
+		spin_lock(&lock->spinlock);
+		BUG_ON(lock->ml.highest_blocked <= LKM_IVMODE);
+		hi = lock->ml.highest_blocked;
+		lock->ml.highest_blocked = LKM_IVMODE;
+		spin_unlock(&lock->spinlock);
+
+		/* remove from list (including ref) */
+		list_del_init(&lock->bast_list);
+		dlm_lock_put(lock);
+		spin_unlock(&dlm->ast_lock);
+
+		mlog(0, "%s: res %.*s, Flush BAST for lock %u:%llu, "
+		     "blocked %d, node %u\n",
+		     dlm->name, res->lockname.len, res->lockname.name,
+		     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
+		     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
+		     hi, lock->ml.node);
+
+		if (lock->ml.node != dlm->node_num) {
+			ret = dlm_send_proxy_bast(dlm, res, lock, hi);
+			if (ret < 0)
+				mlog_errno(ret);
+		} else
+			dlm_do_local_bast(dlm, res, lock, hi);
+
+		spin_lock(&dlm->ast_lock);
+
+		/* possible that another bast was queued while
+		 * we were delivering the last one */
+		if (!list_empty(&lock->bast_list)) {
+			mlog(0, "%s: res %.*s, BAST queued while flushing last "
+			     "one\n", dlm->name, res->lockname.len,
+			     res->lockname.name);
+		} else
+			lock->bast_pending = 0;
+
+		/* drop the extra ref.
+		 * this may drop it completely. */
+		dlm_lock_put(lock);
+		dlm_lockres_release_ast(dlm, res);
+	}
+	wake_up(&dlm->ast_wq);
+	spin_unlock(&dlm->ast_lock);
+}
+
+
+#define DLM_THREAD_TIMEOUT_MS (4 * 1000)
+#define DLM_THREAD_MAX_DIRTY  100
+#define DLM_THREAD_MAX_ASTS   10
+
+static int dlm_thread(void *data)
+{
+	struct dlm_lock_resource *res;
+	struct dlm_ctxt *dlm = data;
+	unsigned long timeout = msecs_to_jiffies(DLM_THREAD_TIMEOUT_MS);
+
+	mlog(0, "dlm thread running for %s...\n", dlm->name);
+
+	while (!kthread_should_stop()) {
+		int n = DLM_THREAD_MAX_DIRTY;
+
+		/* dlm_shutting_down is very point-in-time, but that
+		 * doesn't matter as we'll just loop back around if we
+		 * get false on the leading edge of a state
+		 * transition. */
+		dlm_run_purge_list(dlm, dlm_shutting_down(dlm));
+
+		/* We really don't want to hold dlm->spinlock while
+		 * calling dlm_shuffle_lists on each lockres that
+		 * needs to have its queues adjusted and AST/BASTs
+		 * run.  So let's pull each entry off the dirty_list
+		 * and drop dlm->spinlock ASAP.  Once off the list,
+		 * res->spinlock needs to be taken again to protect
+		 * the queues while calling dlm_shuffle_lists.  */
+		spin_lock(&dlm->spinlock);
+		while (!list_empty(&dlm->dirty_list)) {
+			int delay = 0;
+			res = list_entry(dlm->dirty_list.next,
+					 struct dlm_lock_resource, dirty);
+
+			/* peel a lockres off, remove it from the list,
+			 * unset the dirty flag and drop the dlm lock */
+			BUG_ON(!res);
+			dlm_lockres_get(res);
+
+			spin_lock(&res->spinlock);
+			/* We clear the DLM_LOCK_RES_DIRTY state once we shuffle lists below */
+			list_del_init(&res->dirty);
+			spin_unlock(&res->spinlock);
+			spin_unlock(&dlm->spinlock);
+			/* Drop dirty_list ref */
+			dlm_lockres_put(res);
+
+		 	/* lockres can be re-dirtied/re-added to the
+			 * dirty_list in this gap, but that is ok */
+
+			spin_lock(&dlm->ast_lock);
+			spin_lock(&res->spinlock);
+			if (res->owner != dlm->node_num) {
+				__dlm_print_one_lock_resource(res);
+				mlog(ML_ERROR, "%s: inprog %d, mig %d, reco %d,"
+				     " dirty %d\n", dlm->name,
+				     !!(res->state & DLM_LOCK_RES_IN_PROGRESS),
+				     !!(res->state & DLM_LOCK_RES_MIGRATING),
+				     !!(res->state & DLM_LOCK_RES_RECOVERING),
+				     !!(res->state & DLM_LOCK_RES_DIRTY));
+			}
+			BUG_ON(res->owner != dlm->node_num);
+
+			/* it is now ok to move lockreses in these states
+			 * to the dirty list, assuming that they will only be
+			 * dirty for a short while. */
+			BUG_ON(res->state & DLM_LOCK_RES_MIGRATING);
+			if (res->state & (DLM_LOCK_RES_IN_PROGRESS |
+					  DLM_LOCK_RES_RECOVERING)) {
+				/* move it to the tail and keep going */
+				res->state &= ~DLM_LOCK_RES_DIRTY;
+				spin_unlock(&res->spinlock);
+				spin_unlock(&dlm->ast_lock);
+				mlog(0, "%s: res %.*s, inprogress, delay list "
+				     "shuffle, state %d\n", dlm->name,
+				     res->lockname.len, res->lockname.name,
+				     res->state);
+				delay = 1;
+				goto in_progress;
+			}
+
+			/* at this point the lockres is not migrating/
+			 * recovering/in-progress.  we have the lockres
+			 * spinlock and do NOT have the dlm lock.
+			 * safe to reserve/queue asts and run the lists. */
+
+			/* called while holding lockres lock */
+			dlm_shuffle_lists(dlm, res);
+			res->state &= ~DLM_LOCK_RES_DIRTY;
+			spin_unlock(&res->spinlock);
+			spin_unlock(&dlm->ast_lock);
+
+			dlm_lockres_calc_usage(dlm, res);
+
+in_progress:
+
+			spin_lock(&dlm->spinlock);
+			/* if the lock was in-progress, stick
+			 * it on the back of the list */
+			if (delay) {
+				spin_lock(&res->spinlock);
+				__dlm_dirty_lockres(dlm, res);
+				spin_unlock(&res->spinlock);
+			}
+			dlm_lockres_put(res);
+
+			/* unlikely, but we may need to give time to
+			 * other tasks */
+			if (!--n) {
+				mlog(0, "%s: Throttling dlm thread\n",
+				     dlm->name);
+				break;
+			}
+		}
+
+		spin_unlock(&dlm->spinlock);
+		dlm_flush_asts(dlm);
+
+		/* yield and continue right away if there is more work to do */
+		if (!n) {
+			cond_resched();
+			continue;
+		}
+
+		wait_event_interruptible_timeout(dlm->dlm_thread_wq,
+						 !dlm_dirty_list_empty(dlm) ||
+						 kthread_should_stop(),
+						 timeout);
+	}
+
+	mlog(0, "quitting DLM thread\n");
+	return 0;
+}

fs/ocfs2/dlm/dlmunlock.c

@@ -0,0 +1,698 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dlmunlock.c
+ *
+ * underlying calls for unlocking locks
+ *
+ * Copyright (C) 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/highmem.h>
+#include <linux/init.h>
+#include <linux/sysctl.h>
+#include <linux/random.h>
+#include <linux/blkdev.h>
+#include <linux/socket.h>
+#include <linux/inet.h>
+#include <linux/spinlock.h>
+#include <linux/delay.h>
+
+#include "cluster/heartbeat.h"
+#include "cluster/nodemanager.h"
+#include "cluster/tcp.h"
+
+#include "dlmapi.h"
+#include "dlmcommon.h"
+
+#define MLOG_MASK_PREFIX ML_DLM
+#include "cluster/masklog.h"
+
+#define DLM_UNLOCK_FREE_LOCK           0x00000001
+#define DLM_UNLOCK_CALL_AST            0x00000002
+#define DLM_UNLOCK_REMOVE_LOCK         0x00000004
+#define DLM_UNLOCK_REGRANT_LOCK        0x00000008
+#define DLM_UNLOCK_CLEAR_CONVERT_TYPE  0x00000010
+
+
+static enum dlm_status dlm_get_cancel_actions(struct dlm_ctxt *dlm,
+					      struct dlm_lock_resource *res,
+					      struct dlm_lock *lock,
+					      struct dlm_lockstatus *lksb,
+					      int *actions);
+static enum dlm_status dlm_get_unlock_actions(struct dlm_ctxt *dlm,
+					      struct dlm_lock_resource *res,
+					      struct dlm_lock *lock,
+					      struct dlm_lockstatus *lksb,
+					      int *actions);
+
+static enum dlm_status dlm_send_remote_unlock_request(struct dlm_ctxt *dlm,
+						 struct dlm_lock_resource *res,
+						 struct dlm_lock *lock,
+						 struct dlm_lockstatus *lksb,
+						 int flags,
+						 u8 owner);
+
+
+/*
+ * according to the spec:
+ * http://opendlm.sourceforge.net/cvsmirror/opendlm/docs/dlmbook_final.pdf
+ *
+ *  flags & LKM_CANCEL != 0: must be converting or blocked
+ *  flags & LKM_CANCEL == 0: must be granted
+ *
+ * So to unlock a converting lock, you must first cancel the
+ * convert (passing LKM_CANCEL in flags), then call the unlock
+ * again (with no LKM_CANCEL in flags).
+ */
+
+
+/*
+ * locking:
+ *   caller needs:  none
+ *   taken:         res->spinlock and lock->spinlock taken and dropped
+ *   held on exit:  none
+ * returns: DLM_NORMAL, DLM_NOLOCKMGR, status from network
+ * all callers should have taken an extra ref on lock coming in
+ */
+static enum dlm_status dlmunlock_common(struct dlm_ctxt *dlm,
+					struct dlm_lock_resource *res,
+					struct dlm_lock *lock,
+					struct dlm_lockstatus *lksb,
+					int flags, int *call_ast,
+					int master_node)
+{
+	enum dlm_status status;
+	int actions = 0;
+	int in_use;
+        u8 owner;
+
+	mlog(0, "master_node = %d, valblk = %d\n", master_node,
+	     flags & LKM_VALBLK);
+
+	if (master_node)
+		BUG_ON(res->owner != dlm->node_num);
+	else
+		BUG_ON(res->owner == dlm->node_num);
+
+	spin_lock(&dlm->ast_lock);
+	/* We want to be sure that we're not freeing a lock
+	 * that still has AST's pending... */
+	in_use = !list_empty(&lock->ast_list);
+	spin_unlock(&dlm->ast_lock);
+	if (in_use && !(flags & LKM_CANCEL)) {
+	       mlog(ML_ERROR, "lockres %.*s: Someone is calling dlmunlock "
+		    "while waiting for an ast!", res->lockname.len,
+		    res->lockname.name);
+		return DLM_BADPARAM;
+	}
+
+	spin_lock(&res->spinlock);
+	if (res->state & DLM_LOCK_RES_IN_PROGRESS) {
+		if (master_node && !(flags & LKM_CANCEL)) {
+			mlog(ML_ERROR, "lockres in progress!\n");
+			spin_unlock(&res->spinlock);
+			return DLM_FORWARD;
+		}
+		/* ok for this to sleep if not in a network handler */
+		__dlm_wait_on_lockres(res);
+		res->state |= DLM_LOCK_RES_IN_PROGRESS;
+	}
+	spin_lock(&lock->spinlock);
+
+	if (res->state & DLM_LOCK_RES_RECOVERING) {
+		status = DLM_RECOVERING;
+		goto leave;
+	}
+
+	if (res->state & DLM_LOCK_RES_MIGRATING) {
+		status = DLM_MIGRATING;
+		goto leave;
+	}
+
+	/* see above for what the spec says about
+	 * LKM_CANCEL and the lock queue state */
+	if (flags & LKM_CANCEL)
+		status = dlm_get_cancel_actions(dlm, res, lock, lksb, &actions);
+	else
+		status = dlm_get_unlock_actions(dlm, res, lock, lksb, &actions);
+
+	if (status != DLM_NORMAL && (status != DLM_CANCELGRANT || !master_node))
+		goto leave;
+
+	/* By now this has been masked out of cancel requests. */
+	if (flags & LKM_VALBLK) {
+		/* make the final update to the lvb */
+		if (master_node)
+			memcpy(res->lvb, lksb->lvb, DLM_LVB_LEN);
+		else
+			flags |= LKM_PUT_LVB; /* let the send function
+					       * handle it. */
+	}
+
+	if (!master_node) {
+		owner = res->owner;
+		/* drop locks and send message */
+		if (flags & LKM_CANCEL)
+			lock->cancel_pending = 1;
+		else
+			lock->unlock_pending = 1;
+		spin_unlock(&lock->spinlock);
+		spin_unlock(&res->spinlock);
+		status = dlm_send_remote_unlock_request(dlm, res, lock, lksb,
+							flags, owner);
+		spin_lock(&res->spinlock);
+		spin_lock(&lock->spinlock);
+		/* if the master told us the lock was already granted,
+		 * let the ast handle all of these actions */
+		if (status == DLM_CANCELGRANT) {
+			actions &= ~(DLM_UNLOCK_REMOVE_LOCK|
+				     DLM_UNLOCK_REGRANT_LOCK|
+				     DLM_UNLOCK_CLEAR_CONVERT_TYPE);
+		} else if (status == DLM_RECOVERING ||
+			   status == DLM_MIGRATING ||
+			   status == DLM_FORWARD ||
+			   status == DLM_NOLOCKMGR
+			   ) {
+			/* must clear the actions because this unlock
+			 * is about to be retried.  cannot free or do
+			 * any list manipulation. */
+			mlog(0, "%s:%.*s: clearing actions, %s\n",
+			     dlm->name, res->lockname.len,
+			     res->lockname.name,
+			     status==DLM_RECOVERING?"recovering":
+			     (status==DLM_MIGRATING?"migrating":
+				(status == DLM_FORWARD ? "forward" :
+						"nolockmanager")));
+			actions = 0;
+		}
+		if (flags & LKM_CANCEL)
+			lock->cancel_pending = 0;
+		else
+			lock->unlock_pending = 0;
+
+	}
+
+	/* get an extra ref on lock.  if we are just switching
+	 * lists here, we dont want the lock to go away. */
+	dlm_lock_get(lock);
+
+	if (actions & DLM_UNLOCK_REMOVE_LOCK) {
+		list_del_init(&lock->list);
+		dlm_lock_put(lock);
+	}
+	if (actions & DLM_UNLOCK_REGRANT_LOCK) {
+		dlm_lock_get(lock);
+		list_add_tail(&lock->list, &res->granted);
+	}
+	if (actions & DLM_UNLOCK_CLEAR_CONVERT_TYPE) {
+		mlog(0, "clearing convert_type at %smaster node\n",
+		     master_node ? "" : "non-");
+		lock->ml.convert_type = LKM_IVMODE;
+	}
+
+	/* remove the extra ref on lock */
+	dlm_lock_put(lock);
+
+leave:
+	res->state &= ~DLM_LOCK_RES_IN_PROGRESS;
+	if (!dlm_lock_on_list(&res->converting, lock))
+		BUG_ON(lock->ml.convert_type != LKM_IVMODE);
+	else
+		BUG_ON(lock->ml.convert_type == LKM_IVMODE);
+	spin_unlock(&lock->spinlock);
+	spin_unlock(&res->spinlock);
+	wake_up(&res->wq);
+
+	/* let the caller's final dlm_lock_put handle the actual kfree */
+	if (actions & DLM_UNLOCK_FREE_LOCK) {
+		/* this should always be coupled with list removal */
+		BUG_ON(!(actions & DLM_UNLOCK_REMOVE_LOCK));
+		mlog(0, "lock %u:%llu should be gone now! refs=%d\n",
+		     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
+		     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
+		     atomic_read(&lock->lock_refs.refcount)-1);
+		dlm_lock_put(lock);
+	}
+	if (actions & DLM_UNLOCK_CALL_AST)
+		*call_ast = 1;
+
+	/* if cancel or unlock succeeded, lvb work is done */
+	if (status == DLM_NORMAL)
+		lksb->flags &= ~(DLM_LKSB_PUT_LVB|DLM_LKSB_GET_LVB);
+
+	return status;
+}
+
+void dlm_commit_pending_unlock(struct dlm_lock_resource *res,
+			       struct dlm_lock *lock)
+{
+	/* leave DLM_LKSB_PUT_LVB on the lksb so any final
+	 * update of the lvb will be sent to the new master */
+	list_del_init(&lock->list);
+}
+
+void dlm_commit_pending_cancel(struct dlm_lock_resource *res,
+			       struct dlm_lock *lock)
+{
+	list_move_tail(&lock->list, &res->granted);
+	lock->ml.convert_type = LKM_IVMODE;
+}
+
+
+static inline enum dlm_status dlmunlock_master(struct dlm_ctxt *dlm,
+					  struct dlm_lock_resource *res,
+					  struct dlm_lock *lock,
+					  struct dlm_lockstatus *lksb,
+					  int flags,
+					  int *call_ast)
+{
+	return dlmunlock_common(dlm, res, lock, lksb, flags, call_ast, 1);
+}
+
+static inline enum dlm_status dlmunlock_remote(struct dlm_ctxt *dlm,
+					  struct dlm_lock_resource *res,
+					  struct dlm_lock *lock,
+					  struct dlm_lockstatus *lksb,
+					  int flags, int *call_ast)
+{
+	return dlmunlock_common(dlm, res, lock, lksb, flags, call_ast, 0);
+}
+
+/*
+ * locking:
+ *   caller needs:  none
+ *   taken:         none
+ *   held on exit:  none
+ * returns: DLM_NORMAL, DLM_NOLOCKMGR, status from network
+ */
+static enum dlm_status dlm_send_remote_unlock_request(struct dlm_ctxt *dlm,
+						 struct dlm_lock_resource *res,
+						 struct dlm_lock *lock,
+						 struct dlm_lockstatus *lksb,
+						 int flags,
+						 u8 owner)
+{
+	struct dlm_unlock_lock unlock;
+	int tmpret;
+	enum dlm_status ret;
+	int status = 0;
+	struct kvec vec[2];
+	size_t veclen = 1;
+
+	mlog(0, "%.*s\n", res->lockname.len, res->lockname.name);
+
+	if (owner == dlm->node_num) {
+		/* ended up trying to contact ourself.  this means
+		 * that the lockres had been remote but became local
+		 * via a migration.  just retry it, now as local */
+		mlog(0, "%s:%.*s: this node became the master due to a "
+		     "migration, re-evaluate now\n", dlm->name,
+		     res->lockname.len, res->lockname.name);
+		return DLM_FORWARD;
+	}
+
+	memset(&unlock, 0, sizeof(unlock));
+	unlock.node_idx = dlm->node_num;
+	unlock.flags = cpu_to_be32(flags);
+	unlock.cookie = lock->ml.cookie;
+	unlock.namelen = res->lockname.len;
+	memcpy(unlock.name, res->lockname.name, unlock.namelen);
+
+	vec[0].iov_len = sizeof(struct dlm_unlock_lock);
+	vec[0].iov_base = &unlock;
+
+	if (flags & LKM_PUT_LVB) {
+		/* extra data to send if we are updating lvb */
+		vec[1].iov_len = DLM_LVB_LEN;
+		vec[1].iov_base = lock->lksb->lvb;
+		veclen++;
+	}
+
+	tmpret = o2net_send_message_vec(DLM_UNLOCK_LOCK_MSG, dlm->key,
+					vec, veclen, owner, &status);
+	if (tmpret >= 0) {
+		// successfully sent and received
+		if (status == DLM_FORWARD)
+			mlog(0, "master was in-progress.  retry\n");
+		ret = status;
+	} else {
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", tmpret, DLM_UNLOCK_LOCK_MSG, dlm->key, owner);
+		if (dlm_is_host_down(tmpret)) {
+			/* NOTE: this seems strange, but it is what we want.
+			 * when the master goes down during a cancel or
+			 * unlock, the recovery code completes the operation
+			 * as if the master had not died, then passes the
+			 * updated state to the recovery master.  this thread
+			 * just needs to finish out the operation and call
+			 * the unlockast. */
+			if (dlm_is_node_dead(dlm, owner))
+				ret = DLM_NORMAL;
+			else
+				ret = DLM_NOLOCKMGR;
+		} else {
+			/* something bad.  this will BUG in ocfs2 */
+			ret = dlm_err_to_dlm_status(tmpret);
+		}
+	}
+
+	return ret;
+}
+
+/*
+ * locking:
+ *   caller needs:  none
+ *   taken:         takes and drops res->spinlock
+ *   held on exit:  none
+ * returns: DLM_NORMAL, DLM_BADARGS, DLM_IVLOCKID,
+ *          return value from dlmunlock_master
+ */
+int dlm_unlock_lock_handler(struct o2net_msg *msg, u32 len, void *data,
+			    void **ret_data)
+{
+	struct dlm_ctxt *dlm = data;
+	struct dlm_unlock_lock *unlock = (struct dlm_unlock_lock *)msg->buf;
+	struct dlm_lock_resource *res = NULL;
+	struct dlm_lock *lock = NULL;
+	enum dlm_status status = DLM_NORMAL;
+	int found = 0, i;
+	struct dlm_lockstatus *lksb = NULL;
+	int ignore;
+	u32 flags;
+	struct list_head *queue;
+
+	flags = be32_to_cpu(unlock->flags);
+
+	if (flags & LKM_GET_LVB) {
+		mlog(ML_ERROR, "bad args!  GET_LVB specified on unlock!\n");
+		return DLM_BADARGS;
+	}
+
+	if ((flags & (LKM_PUT_LVB|LKM_CANCEL)) == (LKM_PUT_LVB|LKM_CANCEL)) {
+		mlog(ML_ERROR, "bad args!  cannot modify lvb on a CANCEL "
+		     "request!\n");
+		return DLM_BADARGS;
+	}
+
+	if (unlock->namelen > DLM_LOCKID_NAME_MAX) {
+		mlog(ML_ERROR, "Invalid name length in unlock handler!\n");
+		return DLM_IVBUFLEN;
+	}
+
+	if (!dlm_grab(dlm))
+		return DLM_REJECTED;
+
+	mlog_bug_on_msg(!dlm_domain_fully_joined(dlm),
+			"Domain %s not fully joined!\n", dlm->name);
+
+	mlog(0, "lvb: %s\n", flags & LKM_PUT_LVB ? "put lvb" : "none");
+
+	res = dlm_lookup_lockres(dlm, unlock->name, unlock->namelen);
+	if (!res) {
+		/* We assume here that a no lock resource simply means
+		 * it was migrated away and destroyed before the other
+		 * node could detect it. */
+		mlog(0, "returning DLM_FORWARD -- res no longer exists\n");
+		status = DLM_FORWARD;
+		goto not_found;
+	}
+
+	queue=&res->granted;
+	found = 0;
+	spin_lock(&res->spinlock);
+	if (res->state & DLM_LOCK_RES_RECOVERING) {
+		spin_unlock(&res->spinlock);
+		mlog(0, "returning DLM_RECOVERING\n");
+		status = DLM_RECOVERING;
+		goto leave;
+	}
+
+	if (res->state & DLM_LOCK_RES_MIGRATING) {
+		spin_unlock(&res->spinlock);
+		mlog(0, "returning DLM_MIGRATING\n");
+		status = DLM_MIGRATING;
+		goto leave;
+	}
+
+	if (res->owner != dlm->node_num) {
+		spin_unlock(&res->spinlock);
+		mlog(0, "returning DLM_FORWARD -- not master\n");
+		status = DLM_FORWARD;
+		goto leave;
+	}
+
+	for (i=0; i<3; i++) {
+		list_for_each_entry(lock, queue, list) {
+			if (lock->ml.cookie == unlock->cookie &&
+		    	    lock->ml.node == unlock->node_idx) {
+				dlm_lock_get(lock);
+				found = 1;
+				break;
+			}
+		}
+		if (found)
+			break;
+		/* scan granted -> converting -> blocked queues */
+		queue++;
+	}
+	spin_unlock(&res->spinlock);
+	if (!found) {
+		status = DLM_IVLOCKID;
+		goto not_found;
+	}
+
+	/* lock was found on queue */
+	lksb = lock->lksb;
+	if (flags & (LKM_VALBLK|LKM_PUT_LVB) &&
+	    lock->ml.type != LKM_EXMODE)
+		flags &= ~(LKM_VALBLK|LKM_PUT_LVB);
+
+	/* unlockast only called on originating node */
+	if (flags & LKM_PUT_LVB) {
+		lksb->flags |= DLM_LKSB_PUT_LVB;
+		memcpy(&lksb->lvb[0], &unlock->lvb[0], DLM_LVB_LEN);
+	}
+
+	/* if this is in-progress, propagate the DLM_FORWARD
+	 * all the way back out */
+	status = dlmunlock_master(dlm, res, lock, lksb, flags, &ignore);
+	if (status == DLM_FORWARD)
+		mlog(0, "lockres is in progress\n");
+
+	if (flags & LKM_PUT_LVB)
+		lksb->flags &= ~DLM_LKSB_PUT_LVB;
+
+	dlm_lockres_calc_usage(dlm, res);
+	dlm_kick_thread(dlm, res);
+
+not_found:
+	if (!found)
+		mlog(ML_ERROR, "failed to find lock to unlock! "
+			       "cookie=%u:%llu\n",
+		     dlm_get_lock_cookie_node(be64_to_cpu(unlock->cookie)),
+		     dlm_get_lock_cookie_seq(be64_to_cpu(unlock->cookie)));
+	else
+		dlm_lock_put(lock);
+
+leave:
+	if (res)
+		dlm_lockres_put(res);
+
+	dlm_put(dlm);
+
+	return status;
+}
+
+
+static enum dlm_status dlm_get_cancel_actions(struct dlm_ctxt *dlm,
+					      struct dlm_lock_resource *res,
+					      struct dlm_lock *lock,
+					      struct dlm_lockstatus *lksb,
+					      int *actions)
+{
+	enum dlm_status status;
+
+	if (dlm_lock_on_list(&res->blocked, lock)) {
+		/* cancel this outright */
+		status = DLM_NORMAL;
+		*actions = (DLM_UNLOCK_CALL_AST |
+			    DLM_UNLOCK_REMOVE_LOCK);
+	} else if (dlm_lock_on_list(&res->converting, lock)) {
+		/* cancel the request, put back on granted */
+		status = DLM_NORMAL;
+		*actions = (DLM_UNLOCK_CALL_AST |
+			    DLM_UNLOCK_REMOVE_LOCK |
+			    DLM_UNLOCK_REGRANT_LOCK |
+			    DLM_UNLOCK_CLEAR_CONVERT_TYPE);
+	} else if (dlm_lock_on_list(&res->granted, lock)) {
+		/* too late, already granted. */
+		status = DLM_CANCELGRANT;
+		*actions = DLM_UNLOCK_CALL_AST;
+	} else {
+		mlog(ML_ERROR, "lock to cancel is not on any list!\n");
+		status = DLM_IVLOCKID;
+		*actions = 0;
+	}
+	return status;
+}
+
+static enum dlm_status dlm_get_unlock_actions(struct dlm_ctxt *dlm,
+					      struct dlm_lock_resource *res,
+					      struct dlm_lock *lock,
+					      struct dlm_lockstatus *lksb,
+					      int *actions)
+{
+	enum dlm_status status;
+
+	/* unlock request */
+	if (!dlm_lock_on_list(&res->granted, lock)) {
+		status = DLM_DENIED;
+		dlm_error(status);
+		*actions = 0;
+	} else {
+		/* unlock granted lock */
+		status = DLM_NORMAL;
+		*actions = (DLM_UNLOCK_FREE_LOCK |
+			    DLM_UNLOCK_CALL_AST |
+			    DLM_UNLOCK_REMOVE_LOCK);
+	}
+	return status;
+}
+
+/* there seems to be no point in doing this async
+ * since (even for the remote case) there is really
+ * no work to queue up... so just do it and fire the
+ * unlockast by hand when done... */
+enum dlm_status dlmunlock(struct dlm_ctxt *dlm, struct dlm_lockstatus *lksb,
+			  int flags, dlm_astunlockfunc_t *unlockast, void *data)
+{
+	enum dlm_status status;
+	struct dlm_lock_resource *res;
+	struct dlm_lock *lock = NULL;
+	int call_ast, is_master;
+
+	if (!lksb) {
+		dlm_error(DLM_BADARGS);
+		return DLM_BADARGS;
+	}
+
+	if (flags & ~(LKM_CANCEL | LKM_VALBLK | LKM_INVVALBLK)) {
+		dlm_error(DLM_BADPARAM);
+		return DLM_BADPARAM;
+	}
+
+	if ((flags & (LKM_VALBLK | LKM_CANCEL)) == (LKM_VALBLK | LKM_CANCEL)) {
+		mlog(0, "VALBLK given with CANCEL: ignoring VALBLK\n");
+		flags &= ~LKM_VALBLK;
+	}
+
+	if (!lksb->lockid || !lksb->lockid->lockres) {
+		dlm_error(DLM_BADPARAM);
+		return DLM_BADPARAM;
+	}
+
+	lock = lksb->lockid;
+	BUG_ON(!lock);
+	dlm_lock_get(lock);
+
+	res = lock->lockres;
+	BUG_ON(!res);
+	dlm_lockres_get(res);
+retry:
+	call_ast = 0;
+	/* need to retry up here because owner may have changed */
+	mlog(0, "lock=%p res=%p\n", lock, res);
+
+	spin_lock(&res->spinlock);
+	is_master = (res->owner == dlm->node_num);
+	if (flags & LKM_VALBLK && lock->ml.type != LKM_EXMODE)
+		flags &= ~LKM_VALBLK;
+	spin_unlock(&res->spinlock);
+
+	if (is_master) {
+		status = dlmunlock_master(dlm, res, lock, lksb, flags,
+					  &call_ast);
+		mlog(0, "done calling dlmunlock_master: returned %d, "
+		     "call_ast is %d\n", status, call_ast);
+	} else {
+		status = dlmunlock_remote(dlm, res, lock, lksb, flags,
+					  &call_ast);
+		mlog(0, "done calling dlmunlock_remote: returned %d, "
+		     "call_ast is %d\n", status, call_ast);
+	}
+
+	if (status == DLM_RECOVERING ||
+	    status == DLM_MIGRATING ||
+	    status == DLM_FORWARD ||
+	    status == DLM_NOLOCKMGR) {
+
+		/* We want to go away for a tiny bit to allow recovery
+		 * / migration to complete on this resource. I don't
+		 * know of any wait queue we could sleep on as this
+		 * may be happening on another node. Perhaps the
+		 * proper solution is to queue up requests on the
+		 * other end? */
+
+		/* do we want to yield(); ?? */
+		msleep(50);
+
+		mlog(0, "retrying unlock due to pending recovery/"
+		     "migration/in-progress/reconnect\n");
+		goto retry;
+	}
+
+	if (call_ast) {
+		mlog(0, "calling unlockast(%p, %d)\n", data, status);
+		if (is_master) {
+			/* it is possible that there is one last bast
+			 * pending.  make sure it is flushed, then
+			 * call the unlockast.
+			 * not an issue if this is a mastered remotely,
+			 * since this lock has been removed from the
+			 * lockres queues and cannot be found. */
+			dlm_kick_thread(dlm, NULL);
+			wait_event(dlm->ast_wq,
+				   dlm_lock_basts_flushed(dlm, lock));
+		}
+		(*unlockast)(data, status);
+	}
+
+	if (status == DLM_CANCELGRANT)
+		status = DLM_NORMAL;
+
+	if (status == DLM_NORMAL) {
+		mlog(0, "kicking the thread\n");
+		dlm_kick_thread(dlm, res);
+	} else
+		dlm_error(status);
+
+	dlm_lockres_calc_usage(dlm, res);
+	dlm_lockres_put(res);
+	dlm_lock_put(lock);
+
+	mlog(0, "returning status=%d!\n", status);
+	return status;
+}
+EXPORT_SYMBOL_GPL(dlmunlock);
+

fs/ocfs2/dlmfs/Makefile

@@ -0,0 +1,5 @@
+ccflags-y := -Ifs/ocfs2
+
+obj-$(CONFIG_OCFS2_FS) += ocfs2_dlmfs.o
+
+ocfs2_dlmfs-objs := userdlm.o dlmfs.o

fs/ocfs2/dlmfs/dlmfs.c

@@ -0,0 +1,704 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dlmfs.c
+ *
+ * Code which implements the kernel side of a minimal userspace
+ * interface to our DLM. This file handles the virtual file system
+ * used for communication with userspace. Credit should go to ramfs,
+ * which was a template for the fs side of this module.
+ *
+ * Copyright (C) 2003, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+/* Simple VFS hooks based on: */
+/*
+ * Resizable simple ram filesystem for Linux.
+ *
+ * Copyright (C) 2000 Linus Torvalds.
+ *               2000 Transmeta Corp.
+ */
+
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/pagemap.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+#include <linux/init.h>
+#include <linux/string.h>
+#include <linux/backing-dev.h>
+#include <linux/poll.h>
+
+#include <asm/uaccess.h>
+
+#include "stackglue.h"
+#include "userdlm.h"
+
+#define MLOG_MASK_PREFIX ML_DLMFS
+#include "cluster/masklog.h"
+
+
+static const struct super_operations dlmfs_ops;
+static const struct file_operations dlmfs_file_operations;
+static const struct inode_operations dlmfs_dir_inode_operations;
+static const struct inode_operations dlmfs_root_inode_operations;
+static const struct inode_operations dlmfs_file_inode_operations;
+static struct kmem_cache *dlmfs_inode_cache;
+
+struct workqueue_struct *user_dlm_worker;
+
+
+
+/*
+ * These are the ABI capabilities of dlmfs.
+ *
+ * Over time, dlmfs has added some features that were not part of the
+ * initial ABI.  Unfortunately, some of these features are not detectable
+ * via standard usage.  For example, Linux's default poll always returns
+ * POLLIN, so there is no way for a caller of poll(2) to know when dlmfs
+ * added poll support.  Instead, we provide this list of new capabilities.
+ *
+ * Capabilities is a read-only attribute.  We do it as a module parameter
+ * so we can discover it whether dlmfs is built in, loaded, or even not
+ * loaded.
+ *
+ * The ABI features are local to this machine's dlmfs mount.  This is
+ * distinct from the locking protocol, which is concerned with inter-node
+ * interaction.
+ *
+ * Capabilities:
+ * - bast	: POLLIN against the file descriptor of a held lock
+ *		  signifies a bast fired on the lock.
+ */
+#define DLMFS_CAPABILITIES "bast stackglue"
+static int param_set_dlmfs_capabilities(const char *val,
+					struct kernel_param *kp)
+{
+	printk(KERN_ERR "%s: readonly parameter\n", kp->name);
+	return -EINVAL;
+}
+static int param_get_dlmfs_capabilities(char *buffer,
+					struct kernel_param *kp)
+{
+	return strlcpy(buffer, DLMFS_CAPABILITIES,
+		       strlen(DLMFS_CAPABILITIES) + 1);
+}
+module_param_call(capabilities, param_set_dlmfs_capabilities,
+		  param_get_dlmfs_capabilities, NULL, 0444);
+MODULE_PARM_DESC(capabilities, DLMFS_CAPABILITIES);
+
+
+/*
+ * decodes a set of open flags into a valid lock level and a set of flags.
+ * returns < 0 if we have invalid flags
+ * flags which mean something to us:
+ * O_RDONLY -> PRMODE level
+ * O_WRONLY -> EXMODE level
+ *
+ * O_NONBLOCK -> NOQUEUE
+ */
+static int dlmfs_decode_open_flags(int open_flags,
+				   int *level,
+				   int *flags)
+{
+	if (open_flags & (O_WRONLY|O_RDWR))
+		*level = DLM_LOCK_EX;
+	else
+		*level = DLM_LOCK_PR;
+
+	*flags = 0;
+	if (open_flags & O_NONBLOCK)
+		*flags |= DLM_LKF_NOQUEUE;
+
+	return 0;
+}
+
+static int dlmfs_file_open(struct inode *inode,
+			   struct file *file)
+{
+	int status, level, flags;
+	struct dlmfs_filp_private *fp = NULL;
+	struct dlmfs_inode_private *ip;
+
+	if (S_ISDIR(inode->i_mode))
+		BUG();
+
+	mlog(0, "open called on inode %lu, flags 0x%x\n", inode->i_ino,
+		file->f_flags);
+
+	status = dlmfs_decode_open_flags(file->f_flags, &level, &flags);
+	if (status < 0)
+		goto bail;
+
+	/* We don't want to honor O_APPEND at read/write time as it
+	 * doesn't make sense for LVB writes. */
+	file->f_flags &= ~O_APPEND;
+
+	fp = kmalloc(sizeof(*fp), GFP_NOFS);
+	if (!fp) {
+		status = -ENOMEM;
+		goto bail;
+	}
+	fp->fp_lock_level = level;
+
+	ip = DLMFS_I(inode);
+
+	status = user_dlm_cluster_lock(&ip->ip_lockres, level, flags);
+	if (status < 0) {
+		/* this is a strange error to return here but I want
+		 * to be able userspace to be able to distinguish a
+		 * valid lock request from one that simply couldn't be
+		 * granted. */
+		if (flags & DLM_LKF_NOQUEUE && status == -EAGAIN)
+			status = -ETXTBSY;
+		kfree(fp);
+		goto bail;
+	}
+
+	file->private_data = fp;
+bail:
+	return status;
+}
+
+static int dlmfs_file_release(struct inode *inode,
+			      struct file *file)
+{
+	int level, status;
+	struct dlmfs_inode_private *ip = DLMFS_I(inode);
+	struct dlmfs_filp_private *fp = file->private_data;
+
+	if (S_ISDIR(inode->i_mode))
+		BUG();
+
+	mlog(0, "close called on inode %lu\n", inode->i_ino);
+
+	status = 0;
+	if (fp) {
+		level = fp->fp_lock_level;
+		if (level != DLM_LOCK_IV)
+			user_dlm_cluster_unlock(&ip->ip_lockres, level);
+
+		kfree(fp);
+		file->private_data = NULL;
+	}
+
+	return 0;
+}
+
+/*
+ * We do ->setattr() just to override size changes.  Our size is the size
+ * of the LVB and nothing else.
+ */
+static int dlmfs_file_setattr(struct dentry *dentry, struct iattr *attr)
+{
+	int error;
+	struct inode *inode = dentry->d_inode;
+
+	attr->ia_valid &= ~ATTR_SIZE;
+	error = inode_change_ok(inode, attr);
+	if (error)
+		return error;
+
+	setattr_copy(inode, attr);
+	mark_inode_dirty(inode);
+	return 0;
+}
+
+static unsigned int dlmfs_file_poll(struct file *file, poll_table *wait)
+{
+	int event = 0;
+	struct inode *inode = file_inode(file);
+	struct dlmfs_inode_private *ip = DLMFS_I(inode);
+
+	poll_wait(file, &ip->ip_lockres.l_event, wait);
+
+	spin_lock(&ip->ip_lockres.l_lock);
+	if (ip->ip_lockres.l_flags & USER_LOCK_BLOCKED)
+		event = POLLIN | POLLRDNORM;
+	spin_unlock(&ip->ip_lockres.l_lock);
+
+	return event;
+}
+
+static ssize_t dlmfs_file_read(struct file *filp,
+			       char __user *buf,
+			       size_t count,
+			       loff_t *ppos)
+{
+	int bytes_left;
+	ssize_t readlen, got;
+	char *lvb_buf;
+	struct inode *inode = file_inode(filp);
+
+	mlog(0, "inode %lu, count = %zu, *ppos = %llu\n",
+		inode->i_ino, count, *ppos);
+
+	if (*ppos >= i_size_read(inode))
+		return 0;
+
+	if (!count)
+		return 0;
+
+	if (!access_ok(VERIFY_WRITE, buf, count))
+		return -EFAULT;
+
+	/* don't read past the lvb */
+	if ((count + *ppos) > i_size_read(inode))
+		readlen = i_size_read(inode) - *ppos;
+	else
+		readlen = count;
+
+	lvb_buf = kmalloc(readlen, GFP_NOFS);
+	if (!lvb_buf)
+		return -ENOMEM;
+
+	got = user_dlm_read_lvb(inode, lvb_buf, readlen);
+	if (got) {
+		BUG_ON(got != readlen);
+		bytes_left = __copy_to_user(buf, lvb_buf, readlen);
+		readlen -= bytes_left;
+	} else
+		readlen = 0;
+
+	kfree(lvb_buf);
+
+	*ppos = *ppos + readlen;
+
+	mlog(0, "read %zd bytes\n", readlen);
+	return readlen;
+}
+
+static ssize_t dlmfs_file_write(struct file *filp,
+				const char __user *buf,
+				size_t count,
+				loff_t *ppos)
+{
+	int bytes_left;
+	ssize_t writelen;
+	char *lvb_buf;
+	struct inode *inode = file_inode(filp);
+
+	mlog(0, "inode %lu, count = %zu, *ppos = %llu\n",
+		inode->i_ino, count, *ppos);
+
+	if (*ppos >= i_size_read(inode))
+		return -ENOSPC;
+
+	if (!count)
+		return 0;
+
+	if (!access_ok(VERIFY_READ, buf, count))
+		return -EFAULT;
+
+	/* don't write past the lvb */
+	if ((count + *ppos) > i_size_read(inode))
+		writelen = i_size_read(inode) - *ppos;
+	else
+		writelen = count - *ppos;
+
+	lvb_buf = kmalloc(writelen, GFP_NOFS);
+	if (!lvb_buf)
+		return -ENOMEM;
+
+	bytes_left = copy_from_user(lvb_buf, buf, writelen);
+	writelen -= bytes_left;
+	if (writelen)
+		user_dlm_write_lvb(inode, lvb_buf, writelen);
+
+	kfree(lvb_buf);
+
+	*ppos = *ppos + writelen;
+	mlog(0, "wrote %zd bytes\n", writelen);
+	return writelen;
+}
+
+static void dlmfs_init_once(void *foo)
+{
+	struct dlmfs_inode_private *ip =
+		(struct dlmfs_inode_private *) foo;
+
+	ip->ip_conn = NULL;
+	ip->ip_parent = NULL;
+
+	inode_init_once(&ip->ip_vfs_inode);
+}
+
+static struct inode *dlmfs_alloc_inode(struct super_block *sb)
+{
+	struct dlmfs_inode_private *ip;
+
+	ip = kmem_cache_alloc(dlmfs_inode_cache, GFP_NOFS);
+	if (!ip)
+		return NULL;
+
+	return &ip->ip_vfs_inode;
+}
+
+static void dlmfs_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	kmem_cache_free(dlmfs_inode_cache, DLMFS_I(inode));
+}
+
+static void dlmfs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, dlmfs_i_callback);
+}
+
+static void dlmfs_evict_inode(struct inode *inode)
+{
+	int status;
+	struct dlmfs_inode_private *ip;
+
+	clear_inode(inode);
+
+	mlog(0, "inode %lu\n", inode->i_ino);
+
+	ip = DLMFS_I(inode);
+
+	if (S_ISREG(inode->i_mode)) {
+		status = user_dlm_destroy_lock(&ip->ip_lockres);
+		if (status < 0)
+			mlog_errno(status);
+		iput(ip->ip_parent);
+		goto clear_fields;
+	}
+
+	mlog(0, "we're a directory, ip->ip_conn = 0x%p\n", ip->ip_conn);
+	/* we must be a directory. If required, lets unregister the
+	 * dlm context now. */
+	if (ip->ip_conn)
+		user_dlm_unregister(ip->ip_conn);
+clear_fields:
+	ip->ip_parent = NULL;
+	ip->ip_conn = NULL;
+}
+
+static struct backing_dev_info dlmfs_backing_dev_info = {
+	.name		= "ocfs2-dlmfs",
+	.ra_pages	= 0,	/* No readahead */
+	.capabilities	= BDI_CAP_NO_ACCT_AND_WRITEBACK,
+};
+
+static struct inode *dlmfs_get_root_inode(struct super_block *sb)
+{
+	struct inode *inode = new_inode(sb);
+	umode_t mode = S_IFDIR | 0755;
+
+	if (inode) {
+		inode->i_ino = get_next_ino();
+		inode_init_owner(inode, NULL, mode);
+		inode->i_mapping->backing_dev_info = &dlmfs_backing_dev_info;
+		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
+		inc_nlink(inode);
+
+		inode->i_fop = &simple_dir_operations;
+		inode->i_op = &dlmfs_root_inode_operations;
+	}
+
+	return inode;
+}
+
+static struct inode *dlmfs_get_inode(struct inode *parent,
+				     struct dentry *dentry,
+				     umode_t mode)
+{
+	struct super_block *sb = parent->i_sb;
+	struct inode * inode = new_inode(sb);
+	struct dlmfs_inode_private *ip;
+
+	if (!inode)
+		return NULL;
+
+	inode->i_ino = get_next_ino();
+	inode_init_owner(inode, parent, mode);
+	inode->i_mapping->backing_dev_info = &dlmfs_backing_dev_info;
+	inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
+
+	ip = DLMFS_I(inode);
+	ip->ip_conn = DLMFS_I(parent)->ip_conn;
+
+	switch (mode & S_IFMT) {
+	default:
+		/* for now we don't support anything other than
+		 * directories and regular files. */
+		BUG();
+		break;
+	case S_IFREG:
+		inode->i_op = &dlmfs_file_inode_operations;
+		inode->i_fop = &dlmfs_file_operations;
+
+		i_size_write(inode,  DLM_LVB_LEN);
+
+		user_dlm_lock_res_init(&ip->ip_lockres, dentry);
+
+		/* released at clear_inode time, this insures that we
+		 * get to drop the dlm reference on each lock *before*
+		 * we call the unregister code for releasing parent
+		 * directories. */
+		ip->ip_parent = igrab(parent);
+		BUG_ON(!ip->ip_parent);
+		break;
+	case S_IFDIR:
+		inode->i_op = &dlmfs_dir_inode_operations;
+		inode->i_fop = &simple_dir_operations;
+
+		/* directory inodes start off with i_nlink ==
+		 * 2 (for "." entry) */
+		inc_nlink(inode);
+		break;
+	}
+	return inode;
+}
+
+/*
+ * File creation. Allocate an inode, and we're done..
+ */
+/* SMP-safe */
+static int dlmfs_mkdir(struct inode * dir,
+		       struct dentry * dentry,
+		       umode_t mode)
+{
+	int status;
+	struct inode *inode = NULL;
+	struct qstr *domain = &dentry->d_name;
+	struct dlmfs_inode_private *ip;
+	struct ocfs2_cluster_connection *conn;
+
+	mlog(0, "mkdir %.*s\n", domain->len, domain->name);
+
+	/* verify that we have a proper domain */
+	if (domain->len >= GROUP_NAME_MAX) {
+		status = -EINVAL;
+		mlog(ML_ERROR, "invalid domain name for directory.\n");
+		goto bail;
+	}
+
+	inode = dlmfs_get_inode(dir, dentry, mode | S_IFDIR);
+	if (!inode) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	ip = DLMFS_I(inode);
+
+	conn = user_dlm_register(domain);
+	if (IS_ERR(conn)) {
+		status = PTR_ERR(conn);
+		mlog(ML_ERROR, "Error %d could not register domain \"%.*s\"\n",
+		     status, domain->len, domain->name);
+		goto bail;
+	}
+	ip->ip_conn = conn;
+
+	inc_nlink(dir);
+	d_instantiate(dentry, inode);
+	dget(dentry);	/* Extra count - pin the dentry in core */
+
+	status = 0;
+bail:
+	if (status < 0)
+		iput(inode);
+	return status;
+}
+
+static int dlmfs_create(struct inode *dir,
+			struct dentry *dentry,
+			umode_t mode,
+			bool excl)
+{
+	int status = 0;
+	struct inode *inode;
+	struct qstr *name = &dentry->d_name;
+
+	mlog(0, "create %.*s\n", name->len, name->name);
+
+	/* verify name is valid and doesn't contain any dlm reserved
+	 * characters */
+	if (name->len >= USER_DLM_LOCK_ID_MAX_LEN ||
+	    name->name[0] == '$') {
+		status = -EINVAL;
+		mlog(ML_ERROR, "invalid lock name, %.*s\n", name->len,
+		     name->name);
+		goto bail;
+	}
+
+	inode = dlmfs_get_inode(dir, dentry, mode | S_IFREG);
+	if (!inode) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	d_instantiate(dentry, inode);
+	dget(dentry);	/* Extra count - pin the dentry in core */
+bail:
+	return status;
+}
+
+static int dlmfs_unlink(struct inode *dir,
+			struct dentry *dentry)
+{
+	int status;
+	struct inode *inode = dentry->d_inode;
+
+	mlog(0, "unlink inode %lu\n", inode->i_ino);
+
+	/* if there are no current holders, or none that are waiting
+	 * to acquire a lock, this basically destroys our lockres. */
+	status = user_dlm_destroy_lock(&DLMFS_I(inode)->ip_lockres);
+	if (status < 0) {
+		mlog(ML_ERROR, "unlink %.*s, error %d from destroy\n",
+		     dentry->d_name.len, dentry->d_name.name, status);
+		goto bail;
+	}
+	status = simple_unlink(dir, dentry);
+bail:
+	return status;
+}
+
+static int dlmfs_fill_super(struct super_block * sb,
+			    void * data,
+			    int silent)
+{
+	sb->s_maxbytes = MAX_LFS_FILESIZE;
+	sb->s_blocksize = PAGE_CACHE_SIZE;
+	sb->s_blocksize_bits = PAGE_CACHE_SHIFT;
+	sb->s_magic = DLMFS_MAGIC;
+	sb->s_op = &dlmfs_ops;
+	sb->s_root = d_make_root(dlmfs_get_root_inode(sb));
+	if (!sb->s_root)
+		return -ENOMEM;
+	return 0;
+}
+
+static const struct file_operations dlmfs_file_operations = {
+	.open		= dlmfs_file_open,
+	.release	= dlmfs_file_release,
+	.poll		= dlmfs_file_poll,
+	.read		= dlmfs_file_read,
+	.write		= dlmfs_file_write,
+	.llseek		= default_llseek,
+};
+
+static const struct inode_operations dlmfs_dir_inode_operations = {
+	.create		= dlmfs_create,
+	.lookup		= simple_lookup,
+	.unlink		= dlmfs_unlink,
+};
+
+/* this way we can restrict mkdir to only the toplevel of the fs. */
+static const struct inode_operations dlmfs_root_inode_operations = {
+	.lookup		= simple_lookup,
+	.mkdir		= dlmfs_mkdir,
+	.rmdir		= simple_rmdir,
+};
+
+static const struct super_operations dlmfs_ops = {
+	.statfs		= simple_statfs,
+	.alloc_inode	= dlmfs_alloc_inode,
+	.destroy_inode	= dlmfs_destroy_inode,
+	.evict_inode	= dlmfs_evict_inode,
+	.drop_inode	= generic_delete_inode,
+};
+
+static const struct inode_operations dlmfs_file_inode_operations = {
+	.getattr	= simple_getattr,
+	.setattr	= dlmfs_file_setattr,
+};
+
+static struct dentry *dlmfs_mount(struct file_system_type *fs_type,
+	int flags, const char *dev_name, void *data)
+{
+	return mount_nodev(fs_type, flags, data, dlmfs_fill_super);
+}
+
+static struct file_system_type dlmfs_fs_type = {
+	.owner		= THIS_MODULE,
+	.name		= "ocfs2_dlmfs",
+	.mount		= dlmfs_mount,
+	.kill_sb	= kill_litter_super,
+};
+MODULE_ALIAS_FS("ocfs2_dlmfs");
+
+static int __init init_dlmfs_fs(void)
+{
+	int status;
+	int cleanup_inode = 0, cleanup_worker = 0;
+
+	status = bdi_init(&dlmfs_backing_dev_info);
+	if (status)
+		return status;
+
+	dlmfs_inode_cache = kmem_cache_create("dlmfs_inode_cache",
+				sizeof(struct dlmfs_inode_private),
+				0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
+					SLAB_MEM_SPREAD),
+				dlmfs_init_once);
+	if (!dlmfs_inode_cache) {
+		status = -ENOMEM;
+		goto bail;
+	}
+	cleanup_inode = 1;
+
+	user_dlm_worker = create_singlethread_workqueue("user_dlm");
+	if (!user_dlm_worker) {
+		status = -ENOMEM;
+		goto bail;
+	}
+	cleanup_worker = 1;
+
+	user_dlm_set_locking_protocol();
+	status = register_filesystem(&dlmfs_fs_type);
+bail:
+	if (status) {
+		if (cleanup_inode)
+			kmem_cache_destroy(dlmfs_inode_cache);
+		if (cleanup_worker)
+			destroy_workqueue(user_dlm_worker);
+		bdi_destroy(&dlmfs_backing_dev_info);
+	} else
+		printk("OCFS2 User DLM kernel interface loaded\n");
+	return status;
+}
+
+static void __exit exit_dlmfs_fs(void)
+{
+	unregister_filesystem(&dlmfs_fs_type);
+
+	flush_workqueue(user_dlm_worker);
+	destroy_workqueue(user_dlm_worker);
+
+	/*
+	 * Make sure all delayed rcu free inodes are flushed before we
+	 * destroy cache.
+	 */
+	rcu_barrier();
+	kmem_cache_destroy(dlmfs_inode_cache);
+
+	bdi_destroy(&dlmfs_backing_dev_info);
+}
+
+MODULE_AUTHOR("Oracle");
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("OCFS2 DLM-Filesystem");
+
+module_init(init_dlmfs_fs)
+module_exit(exit_dlmfs_fs)

fs/ocfs2/dlmfs/userdlm.c

@@ -0,0 +1,688 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * userdlm.c
+ *
+ * Code which implements the kernel side of a minimal userspace
+ * interface to our DLM.
+ *
+ * Many of the functions here are pared down versions of dlmglue.c
+ * functions.
+ *
+ * Copyright (C) 2003, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/signal.h>
+
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/crc32.h>
+
+#include "ocfs2_lockingver.h"
+#include "stackglue.h"
+#include "userdlm.h"
+
+#define MLOG_MASK_PREFIX ML_DLMFS
+#include "cluster/masklog.h"
+
+
+static inline struct user_lock_res *user_lksb_to_lock_res(struct ocfs2_dlm_lksb *lksb)
+{
+	return container_of(lksb, struct user_lock_res, l_lksb);
+}
+
+static inline int user_check_wait_flag(struct user_lock_res *lockres,
+				       int flag)
+{
+	int ret;
+
+	spin_lock(&lockres->l_lock);
+	ret = lockres->l_flags & flag;
+	spin_unlock(&lockres->l_lock);
+
+	return ret;
+}
+
+static inline void user_wait_on_busy_lock(struct user_lock_res *lockres)
+
+{
+	wait_event(lockres->l_event,
+		   !user_check_wait_flag(lockres, USER_LOCK_BUSY));
+}
+
+static inline void user_wait_on_blocked_lock(struct user_lock_res *lockres)
+
+{
+	wait_event(lockres->l_event,
+		   !user_check_wait_flag(lockres, USER_LOCK_BLOCKED));
+}
+
+/* I heart container_of... */
+static inline struct ocfs2_cluster_connection *
+cluster_connection_from_user_lockres(struct user_lock_res *lockres)
+{
+	struct dlmfs_inode_private *ip;
+
+	ip = container_of(lockres,
+			  struct dlmfs_inode_private,
+			  ip_lockres);
+	return ip->ip_conn;
+}
+
+static struct inode *
+user_dlm_inode_from_user_lockres(struct user_lock_res *lockres)
+{
+	struct dlmfs_inode_private *ip;
+
+	ip = container_of(lockres,
+			  struct dlmfs_inode_private,
+			  ip_lockres);
+	return &ip->ip_vfs_inode;
+}
+
+static inline void user_recover_from_dlm_error(struct user_lock_res *lockres)
+{
+	spin_lock(&lockres->l_lock);
+	lockres->l_flags &= ~USER_LOCK_BUSY;
+	spin_unlock(&lockres->l_lock);
+}
+
+#define user_log_dlm_error(_func, _stat, _lockres) do {			\
+	mlog(ML_ERROR, "Dlm error %d while calling %s on "		\
+		"resource %.*s\n", _stat, _func,			\
+		_lockres->l_namelen, _lockres->l_name); 		\
+} while (0)
+
+/* WARNING: This function lives in a world where the only three lock
+ * levels are EX, PR, and NL. It *will* have to be adjusted when more
+ * lock types are added. */
+static inline int user_highest_compat_lock_level(int level)
+{
+	int new_level = DLM_LOCK_EX;
+
+	if (level == DLM_LOCK_EX)
+		new_level = DLM_LOCK_NL;
+	else if (level == DLM_LOCK_PR)
+		new_level = DLM_LOCK_PR;
+	return new_level;
+}
+
+static void user_ast(struct ocfs2_dlm_lksb *lksb)
+{
+	struct user_lock_res *lockres = user_lksb_to_lock_res(lksb);
+	int status;
+
+	mlog(ML_BASTS, "AST fired for lockres %.*s, level %d => %d\n",
+	     lockres->l_namelen, lockres->l_name, lockres->l_level,
+	     lockres->l_requested);
+
+	spin_lock(&lockres->l_lock);
+
+	status = ocfs2_dlm_lock_status(&lockres->l_lksb);
+	if (status) {
+		mlog(ML_ERROR, "lksb status value of %u on lockres %.*s\n",
+		     status, lockres->l_namelen, lockres->l_name);
+		spin_unlock(&lockres->l_lock);
+		return;
+	}
+
+	mlog_bug_on_msg(lockres->l_requested == DLM_LOCK_IV,
+			"Lockres %.*s, requested ivmode. flags 0x%x\n",
+			lockres->l_namelen, lockres->l_name, lockres->l_flags);
+
+	/* we're downconverting. */
+	if (lockres->l_requested < lockres->l_level) {
+		if (lockres->l_requested <=
+		    user_highest_compat_lock_level(lockres->l_blocking)) {
+			lockres->l_blocking = DLM_LOCK_NL;
+			lockres->l_flags &= ~USER_LOCK_BLOCKED;
+		}
+	}
+
+	lockres->l_level = lockres->l_requested;
+	lockres->l_requested = DLM_LOCK_IV;
+	lockres->l_flags |= USER_LOCK_ATTACHED;
+	lockres->l_flags &= ~USER_LOCK_BUSY;
+
+	spin_unlock(&lockres->l_lock);
+
+	wake_up(&lockres->l_event);
+}
+
+static inline void user_dlm_grab_inode_ref(struct user_lock_res *lockres)
+{
+	struct inode *inode;
+	inode = user_dlm_inode_from_user_lockres(lockres);
+	if (!igrab(inode))
+		BUG();
+}
+
+static void user_dlm_unblock_lock(struct work_struct *work);
+
+static void __user_dlm_queue_lockres(struct user_lock_res *lockres)
+{
+	if (!(lockres->l_flags & USER_LOCK_QUEUED)) {
+		user_dlm_grab_inode_ref(lockres);
+
+		INIT_WORK(&lockres->l_work, user_dlm_unblock_lock);
+
+		queue_work(user_dlm_worker, &lockres->l_work);
+		lockres->l_flags |= USER_LOCK_QUEUED;
+	}
+}
+
+static void __user_dlm_cond_queue_lockres(struct user_lock_res *lockres)
+{
+	int queue = 0;
+
+	if (!(lockres->l_flags & USER_LOCK_BLOCKED))
+		return;
+
+	switch (lockres->l_blocking) {
+	case DLM_LOCK_EX:
+		if (!lockres->l_ex_holders && !lockres->l_ro_holders)
+			queue = 1;
+		break;
+	case DLM_LOCK_PR:
+		if (!lockres->l_ex_holders)
+			queue = 1;
+		break;
+	default:
+		BUG();
+	}
+
+	if (queue)
+		__user_dlm_queue_lockres(lockres);
+}
+
+static void user_bast(struct ocfs2_dlm_lksb *lksb, int level)
+{
+	struct user_lock_res *lockres = user_lksb_to_lock_res(lksb);
+
+	mlog(ML_BASTS, "BAST fired for lockres %.*s, blocking %d, level %d\n",
+	     lockres->l_namelen, lockres->l_name, level, lockres->l_level);
+
+	spin_lock(&lockres->l_lock);
+	lockres->l_flags |= USER_LOCK_BLOCKED;
+	if (level > lockres->l_blocking)
+		lockres->l_blocking = level;
+
+	__user_dlm_queue_lockres(lockres);
+	spin_unlock(&lockres->l_lock);
+
+	wake_up(&lockres->l_event);
+}
+
+static void user_unlock_ast(struct ocfs2_dlm_lksb *lksb, int status)
+{
+	struct user_lock_res *lockres = user_lksb_to_lock_res(lksb);
+
+	mlog(ML_BASTS, "UNLOCK AST fired for lockres %.*s, flags 0x%x\n",
+	     lockres->l_namelen, lockres->l_name, lockres->l_flags);
+
+	if (status)
+		mlog(ML_ERROR, "dlm returns status %d\n", status);
+
+	spin_lock(&lockres->l_lock);
+	/* The teardown flag gets set early during the unlock process,
+	 * so test the cancel flag to make sure that this ast isn't
+	 * for a concurrent cancel. */
+	if (lockres->l_flags & USER_LOCK_IN_TEARDOWN
+	    && !(lockres->l_flags & USER_LOCK_IN_CANCEL)) {
+		lockres->l_level = DLM_LOCK_IV;
+	} else if (status == DLM_CANCELGRANT) {
+		/* We tried to cancel a convert request, but it was
+		 * already granted. Don't clear the busy flag - the
+		 * ast should've done this already. */
+		BUG_ON(!(lockres->l_flags & USER_LOCK_IN_CANCEL));
+		lockres->l_flags &= ~USER_LOCK_IN_CANCEL;
+		goto out_noclear;
+	} else {
+		BUG_ON(!(lockres->l_flags & USER_LOCK_IN_CANCEL));
+		/* Cancel succeeded, we want to re-queue */
+		lockres->l_requested = DLM_LOCK_IV; /* cancel an
+						    * upconvert
+						    * request. */
+		lockres->l_flags &= ~USER_LOCK_IN_CANCEL;
+		/* we want the unblock thread to look at it again
+		 * now. */
+		if (lockres->l_flags & USER_LOCK_BLOCKED)
+			__user_dlm_queue_lockres(lockres);
+	}
+
+	lockres->l_flags &= ~USER_LOCK_BUSY;
+out_noclear:
+	spin_unlock(&lockres->l_lock);
+
+	wake_up(&lockres->l_event);
+}
+
+/*
+ * This is the userdlmfs locking protocol version.
+ *
+ * See fs/ocfs2/dlmglue.c for more details on locking versions.
+ */
+static struct ocfs2_locking_protocol user_dlm_lproto = {
+	.lp_max_version = {
+		.pv_major = OCFS2_LOCKING_PROTOCOL_MAJOR,
+		.pv_minor = OCFS2_LOCKING_PROTOCOL_MINOR,
+	},
+	.lp_lock_ast		= user_ast,
+	.lp_blocking_ast	= user_bast,
+	.lp_unlock_ast		= user_unlock_ast,
+};
+
+static inline void user_dlm_drop_inode_ref(struct user_lock_res *lockres)
+{
+	struct inode *inode;
+	inode = user_dlm_inode_from_user_lockres(lockres);
+	iput(inode);
+}
+
+static void user_dlm_unblock_lock(struct work_struct *work)
+{
+	int new_level, status;
+	struct user_lock_res *lockres =
+		container_of(work, struct user_lock_res, l_work);
+	struct ocfs2_cluster_connection *conn =
+		cluster_connection_from_user_lockres(lockres);
+
+	mlog(0, "lockres %.*s\n", lockres->l_namelen, lockres->l_name);
+
+	spin_lock(&lockres->l_lock);
+
+	mlog_bug_on_msg(!(lockres->l_flags & USER_LOCK_QUEUED),
+			"Lockres %.*s, flags 0x%x\n",
+			lockres->l_namelen, lockres->l_name, lockres->l_flags);
+
+	/* notice that we don't clear USER_LOCK_BLOCKED here. If it's
+	 * set, we want user_ast clear it. */
+	lockres->l_flags &= ~USER_LOCK_QUEUED;
+
+	/* It's valid to get here and no longer be blocked - if we get
+	 * several basts in a row, we might be queued by the first
+	 * one, the unblock thread might run and clear the queued
+	 * flag, and finally we might get another bast which re-queues
+	 * us before our ast for the downconvert is called. */
+	if (!(lockres->l_flags & USER_LOCK_BLOCKED)) {
+		mlog(ML_BASTS, "lockres %.*s USER_LOCK_BLOCKED\n",
+		     lockres->l_namelen, lockres->l_name);
+		spin_unlock(&lockres->l_lock);
+		goto drop_ref;
+	}
+
+	if (lockres->l_flags & USER_LOCK_IN_TEARDOWN) {
+		mlog(ML_BASTS, "lockres %.*s USER_LOCK_IN_TEARDOWN\n",
+		     lockres->l_namelen, lockres->l_name);
+		spin_unlock(&lockres->l_lock);
+		goto drop_ref;
+	}
+
+	if (lockres->l_flags & USER_LOCK_BUSY) {
+		if (lockres->l_flags & USER_LOCK_IN_CANCEL) {
+			mlog(ML_BASTS, "lockres %.*s USER_LOCK_IN_CANCEL\n",
+			     lockres->l_namelen, lockres->l_name);
+			spin_unlock(&lockres->l_lock);
+			goto drop_ref;
+		}
+
+		lockres->l_flags |= USER_LOCK_IN_CANCEL;
+		spin_unlock(&lockres->l_lock);
+
+		status = ocfs2_dlm_unlock(conn, &lockres->l_lksb,
+					  DLM_LKF_CANCEL);
+		if (status)
+			user_log_dlm_error("ocfs2_dlm_unlock", status, lockres);
+		goto drop_ref;
+	}
+
+	/* If there are still incompat holders, we can exit safely
+	 * without worrying about re-queueing this lock as that will
+	 * happen on the last call to user_cluster_unlock. */
+	if ((lockres->l_blocking == DLM_LOCK_EX)
+	    && (lockres->l_ex_holders || lockres->l_ro_holders)) {
+		spin_unlock(&lockres->l_lock);
+		mlog(ML_BASTS, "lockres %.*s, EX/PR Holders %u,%u\n",
+		     lockres->l_namelen, lockres->l_name,
+		     lockres->l_ex_holders, lockres->l_ro_holders);
+		goto drop_ref;
+	}
+
+	if ((lockres->l_blocking == DLM_LOCK_PR)
+	    && lockres->l_ex_holders) {
+		spin_unlock(&lockres->l_lock);
+		mlog(ML_BASTS, "lockres %.*s, EX Holders %u\n",
+		     lockres->l_namelen, lockres->l_name,
+		     lockres->l_ex_holders);
+		goto drop_ref;
+	}
+
+	/* yay, we can downconvert now. */
+	new_level = user_highest_compat_lock_level(lockres->l_blocking);
+	lockres->l_requested = new_level;
+	lockres->l_flags |= USER_LOCK_BUSY;
+	mlog(ML_BASTS, "lockres %.*s, downconvert %d => %d\n",
+	     lockres->l_namelen, lockres->l_name, lockres->l_level, new_level);
+	spin_unlock(&lockres->l_lock);
+
+	/* need lock downconvert request now... */
+	status = ocfs2_dlm_lock(conn, new_level, &lockres->l_lksb,
+				DLM_LKF_CONVERT|DLM_LKF_VALBLK,
+				lockres->l_name,
+				lockres->l_namelen);
+	if (status) {
+		user_log_dlm_error("ocfs2_dlm_lock", status, lockres);
+		user_recover_from_dlm_error(lockres);
+	}
+
+drop_ref:
+	user_dlm_drop_inode_ref(lockres);
+}
+
+static inline void user_dlm_inc_holders(struct user_lock_res *lockres,
+					int level)
+{
+	switch(level) {
+	case DLM_LOCK_EX:
+		lockres->l_ex_holders++;
+		break;
+	case DLM_LOCK_PR:
+		lockres->l_ro_holders++;
+		break;
+	default:
+		BUG();
+	}
+}
+
+/* predict what lock level we'll be dropping down to on behalf
+ * of another node, and return true if the currently wanted
+ * level will be compatible with it. */
+static inline int
+user_may_continue_on_blocked_lock(struct user_lock_res *lockres,
+				  int wanted)
+{
+	BUG_ON(!(lockres->l_flags & USER_LOCK_BLOCKED));
+
+	return wanted <= user_highest_compat_lock_level(lockres->l_blocking);
+}
+
+int user_dlm_cluster_lock(struct user_lock_res *lockres,
+			  int level,
+			  int lkm_flags)
+{
+	int status, local_flags;
+	struct ocfs2_cluster_connection *conn =
+		cluster_connection_from_user_lockres(lockres);
+
+	if (level != DLM_LOCK_EX &&
+	    level != DLM_LOCK_PR) {
+		mlog(ML_ERROR, "lockres %.*s: invalid request!\n",
+		     lockres->l_namelen, lockres->l_name);
+		status = -EINVAL;
+		goto bail;
+	}
+
+	mlog(ML_BASTS, "lockres %.*s, level %d, flags = 0x%x\n",
+	     lockres->l_namelen, lockres->l_name, level, lkm_flags);
+
+again:
+	if (signal_pending(current)) {
+		status = -ERESTARTSYS;
+		goto bail;
+	}
+
+	spin_lock(&lockres->l_lock);
+
+	/* We only compare against the currently granted level
+	 * here. If the lock is blocked waiting on a downconvert,
+	 * we'll get caught below. */
+	if ((lockres->l_flags & USER_LOCK_BUSY) &&
+	    (level > lockres->l_level)) {
+		/* is someone sitting in dlm_lock? If so, wait on
+		 * them. */
+		spin_unlock(&lockres->l_lock);
+
+		user_wait_on_busy_lock(lockres);
+		goto again;
+	}
+
+	if ((lockres->l_flags & USER_LOCK_BLOCKED) &&
+	    (!user_may_continue_on_blocked_lock(lockres, level))) {
+		/* is the lock is currently blocked on behalf of
+		 * another node */
+		spin_unlock(&lockres->l_lock);
+
+		user_wait_on_blocked_lock(lockres);
+		goto again;
+	}
+
+	if (level > lockres->l_level) {
+		local_flags = lkm_flags | DLM_LKF_VALBLK;
+		if (lockres->l_level != DLM_LOCK_IV)
+			local_flags |= DLM_LKF_CONVERT;
+
+		lockres->l_requested = level;
+		lockres->l_flags |= USER_LOCK_BUSY;
+		spin_unlock(&lockres->l_lock);
+
+		BUG_ON(level == DLM_LOCK_IV);
+		BUG_ON(level == DLM_LOCK_NL);
+
+		/* call dlm_lock to upgrade lock now */
+		status = ocfs2_dlm_lock(conn, level, &lockres->l_lksb,
+					local_flags, lockres->l_name,
+					lockres->l_namelen);
+		if (status) {
+			if ((lkm_flags & DLM_LKF_NOQUEUE) &&
+			    (status != -EAGAIN))
+				user_log_dlm_error("ocfs2_dlm_lock",
+						   status, lockres);
+			user_recover_from_dlm_error(lockres);
+			goto bail;
+		}
+
+		user_wait_on_busy_lock(lockres);
+		goto again;
+	}
+
+	user_dlm_inc_holders(lockres, level);
+	spin_unlock(&lockres->l_lock);
+
+	status = 0;
+bail:
+	return status;
+}
+
+static inline void user_dlm_dec_holders(struct user_lock_res *lockres,
+					int level)
+{
+	switch(level) {
+	case DLM_LOCK_EX:
+		BUG_ON(!lockres->l_ex_holders);
+		lockres->l_ex_holders--;
+		break;
+	case DLM_LOCK_PR:
+		BUG_ON(!lockres->l_ro_holders);
+		lockres->l_ro_holders--;
+		break;
+	default:
+		BUG();
+	}
+}
+
+void user_dlm_cluster_unlock(struct user_lock_res *lockres,
+			     int level)
+{
+	if (level != DLM_LOCK_EX &&
+	    level != DLM_LOCK_PR) {
+		mlog(ML_ERROR, "lockres %.*s: invalid request!\n",
+		     lockres->l_namelen, lockres->l_name);
+		return;
+	}
+
+	spin_lock(&lockres->l_lock);
+	user_dlm_dec_holders(lockres, level);
+	__user_dlm_cond_queue_lockres(lockres);
+	spin_unlock(&lockres->l_lock);
+}
+
+void user_dlm_write_lvb(struct inode *inode,
+			const char *val,
+			unsigned int len)
+{
+	struct user_lock_res *lockres = &DLMFS_I(inode)->ip_lockres;
+	char *lvb;
+
+	BUG_ON(len > DLM_LVB_LEN);
+
+	spin_lock(&lockres->l_lock);
+
+	BUG_ON(lockres->l_level < DLM_LOCK_EX);
+	lvb = ocfs2_dlm_lvb(&lockres->l_lksb);
+	memcpy(lvb, val, len);
+
+	spin_unlock(&lockres->l_lock);
+}
+
+ssize_t user_dlm_read_lvb(struct inode *inode,
+			  char *val,
+			  unsigned int len)
+{
+	struct user_lock_res *lockres = &DLMFS_I(inode)->ip_lockres;
+	char *lvb;
+	ssize_t ret = len;
+
+	BUG_ON(len > DLM_LVB_LEN);
+
+	spin_lock(&lockres->l_lock);
+
+	BUG_ON(lockres->l_level < DLM_LOCK_PR);
+	if (ocfs2_dlm_lvb_valid(&lockres->l_lksb)) {
+		lvb = ocfs2_dlm_lvb(&lockres->l_lksb);
+		memcpy(val, lvb, len);
+	} else
+		ret = 0;
+
+	spin_unlock(&lockres->l_lock);
+	return ret;
+}
+
+void user_dlm_lock_res_init(struct user_lock_res *lockres,
+			    struct dentry *dentry)
+{
+	memset(lockres, 0, sizeof(*lockres));
+
+	spin_lock_init(&lockres->l_lock);
+	init_waitqueue_head(&lockres->l_event);
+	lockres->l_level = DLM_LOCK_IV;
+	lockres->l_requested = DLM_LOCK_IV;
+	lockres->l_blocking = DLM_LOCK_IV;
+
+	/* should have been checked before getting here. */
+	BUG_ON(dentry->d_name.len >= USER_DLM_LOCK_ID_MAX_LEN);
+
+	memcpy(lockres->l_name,
+	       dentry->d_name.name,
+	       dentry->d_name.len);
+	lockres->l_namelen = dentry->d_name.len;
+}
+
+int user_dlm_destroy_lock(struct user_lock_res *lockres)
+{
+	int status = -EBUSY;
+	struct ocfs2_cluster_connection *conn =
+		cluster_connection_from_user_lockres(lockres);
+
+	mlog(ML_BASTS, "lockres %.*s\n", lockres->l_namelen, lockres->l_name);
+
+	spin_lock(&lockres->l_lock);
+	if (lockres->l_flags & USER_LOCK_IN_TEARDOWN) {
+		spin_unlock(&lockres->l_lock);
+		return 0;
+	}
+
+	lockres->l_flags |= USER_LOCK_IN_TEARDOWN;
+
+	while (lockres->l_flags & USER_LOCK_BUSY) {
+		spin_unlock(&lockres->l_lock);
+
+		user_wait_on_busy_lock(lockres);
+
+		spin_lock(&lockres->l_lock);
+	}
+
+	if (lockres->l_ro_holders || lockres->l_ex_holders) {
+		spin_unlock(&lockres->l_lock);
+		goto bail;
+	}
+
+	status = 0;
+	if (!(lockres->l_flags & USER_LOCK_ATTACHED)) {
+		spin_unlock(&lockres->l_lock);
+		goto bail;
+	}
+
+	lockres->l_flags &= ~USER_LOCK_ATTACHED;
+	lockres->l_flags |= USER_LOCK_BUSY;
+	spin_unlock(&lockres->l_lock);
+
+	status = ocfs2_dlm_unlock(conn, &lockres->l_lksb, DLM_LKF_VALBLK);
+	if (status) {
+		user_log_dlm_error("ocfs2_dlm_unlock", status, lockres);
+		goto bail;
+	}
+
+	user_wait_on_busy_lock(lockres);
+
+	status = 0;
+bail:
+	return status;
+}
+
+static void user_dlm_recovery_handler_noop(int node_num,
+					   void *recovery_data)
+{
+	/* We ignore recovery events */
+	return;
+}
+
+void user_dlm_set_locking_protocol(void)
+{
+	ocfs2_stack_glue_set_max_proto_version(&user_dlm_lproto.lp_max_version);
+}
+
+struct ocfs2_cluster_connection *user_dlm_register(struct qstr *name)
+{
+	int rc;
+	struct ocfs2_cluster_connection *conn;
+
+	rc = ocfs2_cluster_connect_agnostic(name->name, name->len,
+					    &user_dlm_lproto,
+					    user_dlm_recovery_handler_noop,
+					    NULL, &conn);
+	if (rc)
+		mlog_errno(rc);
+
+	return rc ? ERR_PTR(rc) : conn;
+}
+
+void user_dlm_unregister(struct ocfs2_cluster_connection *conn)
+{
+	ocfs2_cluster_disconnect(conn, 0);
+}

fs/ocfs2/dlmfs/userdlm.h

@@ -0,0 +1,113 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * userdlm.h
+ *
+ * Userspace dlm defines
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+
+#ifndef USERDLM_H
+#define USERDLM_H
+
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/workqueue.h>
+
+/* user_lock_res->l_flags flags. */
+#define USER_LOCK_ATTACHED      (0x00000001) /* we have initialized
+					       * the lvb */
+#define USER_LOCK_BUSY          (0x00000002) /* we are currently in
+					       * dlm_lock */
+#define USER_LOCK_BLOCKED       (0x00000004) /* blocked waiting to
+					      * downconvert*/
+#define USER_LOCK_IN_TEARDOWN   (0x00000008) /* we're currently
+					      * destroying this
+					      * lock. */
+#define USER_LOCK_QUEUED        (0x00000010) /* lock is on the
+					      * workqueue */
+#define USER_LOCK_IN_CANCEL     (0x00000020)
+
+struct user_lock_res {
+	spinlock_t               l_lock;
+
+	int                      l_flags;
+
+#define USER_DLM_LOCK_ID_MAX_LEN  32
+	char                     l_name[USER_DLM_LOCK_ID_MAX_LEN];
+	int                      l_namelen;
+	int                      l_level;
+	unsigned int             l_ro_holders;
+	unsigned int             l_ex_holders;
+	struct ocfs2_dlm_lksb    l_lksb;
+
+	int                      l_requested;
+	int                      l_blocking;
+
+	wait_queue_head_t        l_event;
+
+	struct work_struct       l_work;
+};
+
+extern struct workqueue_struct *user_dlm_worker;
+
+void user_dlm_lock_res_init(struct user_lock_res *lockres,
+			    struct dentry *dentry);
+int user_dlm_destroy_lock(struct user_lock_res *lockres);
+int user_dlm_cluster_lock(struct user_lock_res *lockres,
+			  int level,
+			  int lkm_flags);
+void user_dlm_cluster_unlock(struct user_lock_res *lockres,
+			     int level);
+void user_dlm_write_lvb(struct inode *inode,
+			const char *val,
+			unsigned int len);
+ssize_t user_dlm_read_lvb(struct inode *inode,
+			  char *val,
+			  unsigned int len);
+struct ocfs2_cluster_connection *user_dlm_register(struct qstr *name);
+void user_dlm_unregister(struct ocfs2_cluster_connection *conn);
+void user_dlm_set_locking_protocol(void);
+
+struct dlmfs_inode_private {
+	struct ocfs2_cluster_connection	*ip_conn;
+
+	struct user_lock_res ip_lockres; /* unused for directories. */
+	struct inode         *ip_parent;
+
+	struct inode         ip_vfs_inode;
+};
+
+static inline struct dlmfs_inode_private *
+DLMFS_I(struct inode *inode)
+{
+        return container_of(inode,
+			    struct dlmfs_inode_private,
+			    ip_vfs_inode);
+}
+
+struct dlmfs_filp_private {
+	int                  fp_lock_level;
+};
+
+#define DLMFS_MAGIC	0x76a9f425
+
+#endif /* USERDLM_H */

fs/ocfs2/dlmglue.h

@@ -0,0 +1,173 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dlmglue.h
+ *
+ * description here
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+
+#ifndef DLMGLUE_H
+#define DLMGLUE_H
+
+#include "dcache.h"
+
+#define OCFS2_LVB_VERSION 5
+
+struct ocfs2_meta_lvb {
+	__u8         lvb_version;
+	__u8         lvb_reserved0;
+	__be16       lvb_idynfeatures;
+	__be32       lvb_iclusters;
+	__be32       lvb_iuid;
+	__be32       lvb_igid;
+	__be64       lvb_iatime_packed;
+	__be64       lvb_ictime_packed;
+	__be64       lvb_imtime_packed;
+	__be64       lvb_isize;
+	__be16       lvb_imode;
+	__be16       lvb_inlink;
+	__be32       lvb_iattr;
+	__be32       lvb_igeneration;
+	__be32       lvb_reserved2;
+};
+
+#define OCFS2_QINFO_LVB_VERSION 1
+
+struct ocfs2_qinfo_lvb {
+	__u8	lvb_version;
+	__u8	lvb_reserved[3];
+	__be32	lvb_bgrace;
+	__be32	lvb_igrace;
+	__be32	lvb_syncms;
+	__be32	lvb_blocks;
+	__be32	lvb_free_blk;
+	__be32	lvb_free_entry;
+};
+
+#define OCFS2_ORPHAN_LVB_VERSION 1
+
+struct ocfs2_orphan_scan_lvb {
+	__u8	lvb_version;
+	__u8	lvb_reserved[3];
+	__be32	lvb_os_seqno;
+};
+
+/* ocfs2_inode_lock_full() 'arg_flags' flags */
+/* don't wait on recovery. */
+#define OCFS2_META_LOCK_RECOVERY	(0x01)
+/* Instruct the dlm not to queue ourselves on the other node. */
+#define OCFS2_META_LOCK_NOQUEUE		(0x02)
+/* don't block waiting for the downconvert thread, instead return -EAGAIN */
+#define OCFS2_LOCK_NONBLOCK		(0x04)
+
+/* Locking subclasses of inode cluster lock */
+enum {
+	OI_LS_NORMAL = 0,
+	OI_LS_PARENT,
+	OI_LS_RENAME1,
+	OI_LS_RENAME2,
+	OI_LS_REFLINK_TARGET,
+};
+
+int ocfs2_dlm_init(struct ocfs2_super *osb);
+void ocfs2_dlm_shutdown(struct ocfs2_super *osb, int hangup_pending);
+void ocfs2_lock_res_init_once(struct ocfs2_lock_res *res);
+void ocfs2_inode_lock_res_init(struct ocfs2_lock_res *res,
+			       enum ocfs2_lock_type type,
+			       unsigned int generation,
+			       struct inode *inode);
+void ocfs2_dentry_lock_res_init(struct ocfs2_dentry_lock *dl,
+				u64 parent, struct inode *inode);
+struct ocfs2_file_private;
+void ocfs2_file_lock_res_init(struct ocfs2_lock_res *lockres,
+			      struct ocfs2_file_private *fp);
+struct ocfs2_mem_dqinfo;
+void ocfs2_qinfo_lock_res_init(struct ocfs2_lock_res *lockres,
+                               struct ocfs2_mem_dqinfo *info);
+void ocfs2_refcount_lock_res_init(struct ocfs2_lock_res *lockres,
+				  struct ocfs2_super *osb, u64 ref_blkno,
+				  unsigned int generation);
+void ocfs2_lock_res_free(struct ocfs2_lock_res *res);
+int ocfs2_create_new_inode_locks(struct inode *inode);
+int ocfs2_drop_inode_locks(struct inode *inode);
+int ocfs2_rw_lock(struct inode *inode, int write);
+void ocfs2_rw_unlock(struct inode *inode, int write);
+int ocfs2_open_lock(struct inode *inode);
+int ocfs2_try_open_lock(struct inode *inode, int write);
+void ocfs2_open_unlock(struct inode *inode);
+int ocfs2_inode_lock_atime(struct inode *inode,
+			  struct vfsmount *vfsmnt,
+			  int *level);
+int ocfs2_inode_lock_full_nested(struct inode *inode,
+			 struct buffer_head **ret_bh,
+			 int ex,
+			 int arg_flags,
+			 int subclass);
+int ocfs2_inode_lock_with_page(struct inode *inode,
+			      struct buffer_head **ret_bh,
+			      int ex,
+			      struct page *page);
+/* Variants without special locking class or flags */
+#define ocfs2_inode_lock_full(i, r, e, f)\
+		ocfs2_inode_lock_full_nested(i, r, e, f, OI_LS_NORMAL)
+#define ocfs2_inode_lock_nested(i, b, e, s)\
+		ocfs2_inode_lock_full_nested(i, b, e, 0, s)
+/* 99% of the time we don't want to supply any additional flags --
+ * those are for very specific cases only. */
+#define ocfs2_inode_lock(i, b, e) ocfs2_inode_lock_full_nested(i, b, e, 0, OI_LS_NORMAL)
+void ocfs2_inode_unlock(struct inode *inode,
+		       int ex);
+int ocfs2_super_lock(struct ocfs2_super *osb,
+		     int ex);
+void ocfs2_super_unlock(struct ocfs2_super *osb,
+			int ex);
+int ocfs2_orphan_scan_lock(struct ocfs2_super *osb, u32 *seqno);
+void ocfs2_orphan_scan_unlock(struct ocfs2_super *osb, u32 seqno);
+
+int ocfs2_rename_lock(struct ocfs2_super *osb);
+void ocfs2_rename_unlock(struct ocfs2_super *osb);
+int ocfs2_nfs_sync_lock(struct ocfs2_super *osb, int ex);
+void ocfs2_nfs_sync_unlock(struct ocfs2_super *osb, int ex);
+int ocfs2_dentry_lock(struct dentry *dentry, int ex);
+void ocfs2_dentry_unlock(struct dentry *dentry, int ex);
+int ocfs2_file_lock(struct file *file, int ex, int trylock);
+void ocfs2_file_unlock(struct file *file);
+int ocfs2_qinfo_lock(struct ocfs2_mem_dqinfo *oinfo, int ex);
+void ocfs2_qinfo_unlock(struct ocfs2_mem_dqinfo *oinfo, int ex);
+struct ocfs2_refcount_tree;
+int ocfs2_refcount_lock(struct ocfs2_refcount_tree *ref_tree, int ex);
+void ocfs2_refcount_unlock(struct ocfs2_refcount_tree *ref_tree, int ex);
+
+
+void ocfs2_mark_lockres_freeing(struct ocfs2_super *osb,
+				struct ocfs2_lock_res *lockres);
+void ocfs2_simple_drop_lockres(struct ocfs2_super *osb,
+			       struct ocfs2_lock_res *lockres);
+
+/* for the downconvert thread */
+void ocfs2_wake_downconvert_thread(struct ocfs2_super *osb);
+
+struct ocfs2_dlm_debug *ocfs2_new_dlm_debug(void);
+void ocfs2_put_dlm_debug(struct ocfs2_dlm_debug *dlm_debug);
+
+/* To set the locking protocol on module initialization */
+void ocfs2_set_locking_protocol(void);
+#endif	/* DLMGLUE_H */

fs/ocfs2/export.c

@@ -0,0 +1,271 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * export.c
+ *
+ * Functions to facilitate NFS exporting
+ *
+ * Copyright (C) 2002, 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "dir.h"
+#include "dlmglue.h"
+#include "dcache.h"
+#include "export.h"
+#include "inode.h"
+
+#include "buffer_head_io.h"
+#include "suballoc.h"
+#include "ocfs2_trace.h"
+
+struct ocfs2_inode_handle
+{
+	u64 ih_blkno;
+	u32 ih_generation;
+};
+
+static struct dentry *ocfs2_get_dentry(struct super_block *sb,
+		struct ocfs2_inode_handle *handle)
+{
+	struct inode *inode;
+	struct ocfs2_super *osb = OCFS2_SB(sb);
+	u64 blkno = handle->ih_blkno;
+	int status, set;
+	struct dentry *result;
+
+	trace_ocfs2_get_dentry_begin(sb, handle, (unsigned long long)blkno);
+
+	if (blkno == 0) {
+		result = ERR_PTR(-ESTALE);
+		goto bail;
+	}
+
+	inode = ocfs2_ilookup(sb, blkno);
+	/*
+	 * If the inode exists in memory, we only need to check it's
+	 * generation number
+	 */
+	if (inode)
+		goto check_gen;
+
+	/*
+	 * This will synchronize us against ocfs2_delete_inode() on
+	 * all nodes
+	 */
+	status = ocfs2_nfs_sync_lock(osb, 1);
+	if (status < 0) {
+		mlog(ML_ERROR, "getting nfs sync lock(EX) failed %d\n", status);
+		goto check_err;
+	}
+
+	status = ocfs2_test_inode_bit(osb, blkno, &set);
+	trace_ocfs2_get_dentry_test_bit(status, set);
+	if (status < 0) {
+		if (status == -EINVAL) {
+			/*
+			 * The blkno NFS gave us doesn't even show up
+			 * as an inode, we return -ESTALE to be
+			 * nice
+			 */
+			status = -ESTALE;
+		} else
+			mlog(ML_ERROR, "test inode bit failed %d\n", status);
+		goto unlock_nfs_sync;
+	}
+
+	/* If the inode allocator bit is clear, this inode must be stale */
+	if (!set) {
+		status = -ESTALE;
+		goto unlock_nfs_sync;
+	}
+
+	inode = ocfs2_iget(osb, blkno, 0, 0);
+
+unlock_nfs_sync:
+	ocfs2_nfs_sync_unlock(osb, 1);
+
+check_err:
+	if (status < 0) {
+		if (status == -ESTALE) {
+			trace_ocfs2_get_dentry_stale((unsigned long long)blkno,
+						     handle->ih_generation);
+		}
+		result = ERR_PTR(status);
+		goto bail;
+	}
+
+	if (IS_ERR(inode)) {
+		mlog_errno(PTR_ERR(inode));
+		result = (void *)inode;
+		goto bail;
+	}
+
+check_gen:
+	if (handle->ih_generation != inode->i_generation) {
+		iput(inode);
+		trace_ocfs2_get_dentry_generation((unsigned long long)blkno,
+						  handle->ih_generation,
+						  inode->i_generation);
+		result = ERR_PTR(-ESTALE);
+		goto bail;
+	}
+
+	result = d_obtain_alias(inode);
+	if (IS_ERR(result))
+		mlog_errno(PTR_ERR(result));
+
+bail:
+	trace_ocfs2_get_dentry_end(result);
+	return result;
+}
+
+static struct dentry *ocfs2_get_parent(struct dentry *child)
+{
+	int status;
+	u64 blkno;
+	struct dentry *parent;
+	struct inode *dir = child->d_inode;
+
+	trace_ocfs2_get_parent(child, child->d_name.len, child->d_name.name,
+			       (unsigned long long)OCFS2_I(dir)->ip_blkno);
+
+	status = ocfs2_inode_lock(dir, NULL, 0);
+	if (status < 0) {
+		if (status != -ENOENT)
+			mlog_errno(status);
+		parent = ERR_PTR(status);
+		goto bail;
+	}
+
+	status = ocfs2_lookup_ino_from_name(dir, "..", 2, &blkno);
+	if (status < 0) {
+		parent = ERR_PTR(-ENOENT);
+		goto bail_unlock;
+	}
+
+	parent = d_obtain_alias(ocfs2_iget(OCFS2_SB(dir->i_sb), blkno, 0, 0));
+
+bail_unlock:
+	ocfs2_inode_unlock(dir, 0);
+
+bail:
+	trace_ocfs2_get_parent_end(parent);
+
+	return parent;
+}
+
+static int ocfs2_encode_fh(struct inode *inode, u32 *fh_in, int *max_len,
+			   struct inode *parent)
+{
+	int len = *max_len;
+	int type = 1;
+	u64 blkno;
+	u32 generation;
+	__le32 *fh = (__force __le32 *) fh_in;
+
+#ifdef TRACE_HOOKS_ARE_NOT_BRAINDEAD_IN_YOUR_OPINION
+#error "You go ahead and fix that mess, then.  Somehow"
+	trace_ocfs2_encode_fh_begin(dentry, dentry->d_name.len,
+				    dentry->d_name.name,
+				    fh, len, connectable);
+#endif
+
+	if (parent && (len < 6)) {
+		*max_len = 6;
+		type = FILEID_INVALID;
+		goto bail;
+	} else if (len < 3) {
+		*max_len = 3;
+		type = FILEID_INVALID;
+		goto bail;
+	}
+
+	blkno = OCFS2_I(inode)->ip_blkno;
+	generation = inode->i_generation;
+
+	trace_ocfs2_encode_fh_self((unsigned long long)blkno, generation);
+
+	len = 3;
+	fh[0] = cpu_to_le32((u32)(blkno >> 32));
+	fh[1] = cpu_to_le32((u32)(blkno & 0xffffffff));
+	fh[2] = cpu_to_le32(generation);
+
+	if (parent) {
+		blkno = OCFS2_I(parent)->ip_blkno;
+		generation = parent->i_generation;
+
+		fh[3] = cpu_to_le32((u32)(blkno >> 32));
+		fh[4] = cpu_to_le32((u32)(blkno & 0xffffffff));
+		fh[5] = cpu_to_le32(generation);
+
+		len = 6;
+		type = 2;
+
+		trace_ocfs2_encode_fh_parent((unsigned long long)blkno,
+					     generation);
+	}
+
+	*max_len = len;
+
+bail:
+	trace_ocfs2_encode_fh_type(type);
+	return type;
+}
+
+static struct dentry *ocfs2_fh_to_dentry(struct super_block *sb,
+		struct fid *fid, int fh_len, int fh_type)
+{
+	struct ocfs2_inode_handle handle;
+
+	if (fh_len < 3 || fh_type > 2)
+		return NULL;
+
+	handle.ih_blkno = (u64)le32_to_cpu(fid->raw[0]) << 32;
+	handle.ih_blkno |= (u64)le32_to_cpu(fid->raw[1]);
+	handle.ih_generation = le32_to_cpu(fid->raw[2]);
+	return ocfs2_get_dentry(sb, &handle);
+}
+
+static struct dentry *ocfs2_fh_to_parent(struct super_block *sb,
+		struct fid *fid, int fh_len, int fh_type)
+{
+	struct ocfs2_inode_handle parent;
+
+	if (fh_type != 2 || fh_len < 6)
+		return NULL;
+
+	parent.ih_blkno = (u64)le32_to_cpu(fid->raw[3]) << 32;
+	parent.ih_blkno |= (u64)le32_to_cpu(fid->raw[4]);
+	parent.ih_generation = le32_to_cpu(fid->raw[5]);
+	return ocfs2_get_dentry(sb, &parent);
+}
+
+const struct export_operations ocfs2_export_ops = {
+	.encode_fh	= ocfs2_encode_fh,
+	.fh_to_dentry	= ocfs2_fh_to_dentry,
+	.fh_to_parent	= ocfs2_fh_to_parent,
+	.get_parent	= ocfs2_get_parent,
+};

fs/ocfs2/export.h

@@ -0,0 +1,33 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * export.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_EXPORT_H
+#define OCFS2_EXPORT_H
+
+#include <linux/exportfs.h>
+
+extern const struct export_operations ocfs2_export_ops;
+
+#endif /* OCFS2_EXPORT_H */

fs/ocfs2/extent_map.c

@@ -0,0 +1,994 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * extent_map.c
+ *
+ * Block/Cluster mapping functions
+ *
+ * Copyright (C) 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License, version 2,  as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/types.h>
+#include <linux/fiemap.h>
+
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "dlmglue.h"
+#include "extent_map.h"
+#include "inode.h"
+#include "super.h"
+#include "symlink.h"
+#include "ocfs2_trace.h"
+
+#include "buffer_head_io.h"
+
+/*
+ * The extent caching implementation is intentionally trivial.
+ *
+ * We only cache a small number of extents stored directly on the
+ * inode, so linear order operations are acceptable. If we ever want
+ * to increase the size of the extent map, then these algorithms must
+ * get smarter.
+ */
+
+void ocfs2_extent_map_init(struct inode *inode)
+{
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+
+	oi->ip_extent_map.em_num_items = 0;
+	INIT_LIST_HEAD(&oi->ip_extent_map.em_list);
+}
+
+static void __ocfs2_extent_map_lookup(struct ocfs2_extent_map *em,
+				      unsigned int cpos,
+				      struct ocfs2_extent_map_item **ret_emi)
+{
+	unsigned int range;
+	struct ocfs2_extent_map_item *emi;
+
+	*ret_emi = NULL;
+
+	list_for_each_entry(emi, &em->em_list, ei_list) {
+		range = emi->ei_cpos + emi->ei_clusters;
+
+		if (cpos >= emi->ei_cpos && cpos < range) {
+			list_move(&emi->ei_list, &em->em_list);
+
+			*ret_emi = emi;
+			break;
+		}
+	}
+}
+
+static int ocfs2_extent_map_lookup(struct inode *inode, unsigned int cpos,
+				   unsigned int *phys, unsigned int *len,
+				   unsigned int *flags)
+{
+	unsigned int coff;
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+	struct ocfs2_extent_map_item *emi;
+
+	spin_lock(&oi->ip_lock);
+
+	__ocfs2_extent_map_lookup(&oi->ip_extent_map, cpos, &emi);
+	if (emi) {
+		coff = cpos - emi->ei_cpos;
+		*phys = emi->ei_phys + coff;
+		if (len)
+			*len = emi->ei_clusters - coff;
+		if (flags)
+			*flags = emi->ei_flags;
+	}
+
+	spin_unlock(&oi->ip_lock);
+
+	if (emi == NULL)
+		return -ENOENT;
+
+	return 0;
+}
+
+/*
+ * Forget about all clusters equal to or greater than cpos.
+ */
+void ocfs2_extent_map_trunc(struct inode *inode, unsigned int cpos)
+{
+	struct ocfs2_extent_map_item *emi, *n;
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+	struct ocfs2_extent_map *em = &oi->ip_extent_map;
+	LIST_HEAD(tmp_list);
+	unsigned int range;
+
+	spin_lock(&oi->ip_lock);
+	list_for_each_entry_safe(emi, n, &em->em_list, ei_list) {
+		if (emi->ei_cpos >= cpos) {
+			/* Full truncate of this record. */
+			list_move(&emi->ei_list, &tmp_list);
+			BUG_ON(em->em_num_items == 0);
+			em->em_num_items--;
+			continue;
+		}
+
+		range = emi->ei_cpos + emi->ei_clusters;
+		if (range > cpos) {
+			/* Partial truncate */
+			emi->ei_clusters = cpos - emi->ei_cpos;
+		}
+	}
+	spin_unlock(&oi->ip_lock);
+
+	list_for_each_entry_safe(emi, n, &tmp_list, ei_list) {
+		list_del(&emi->ei_list);
+		kfree(emi);
+	}
+}
+
+/*
+ * Is any part of emi2 contained within emi1
+ */
+static int ocfs2_ei_is_contained(struct ocfs2_extent_map_item *emi1,
+				 struct ocfs2_extent_map_item *emi2)
+{
+	unsigned int range1, range2;
+
+	/*
+	 * Check if logical start of emi2 is inside emi1
+	 */
+	range1 = emi1->ei_cpos + emi1->ei_clusters;
+	if (emi2->ei_cpos >= emi1->ei_cpos && emi2->ei_cpos < range1)
+		return 1;
+
+	/*
+	 * Check if logical end of emi2 is inside emi1
+	 */
+	range2 = emi2->ei_cpos + emi2->ei_clusters;
+	if (range2 > emi1->ei_cpos && range2 <= range1)
+		return 1;
+
+	return 0;
+}
+
+static void ocfs2_copy_emi_fields(struct ocfs2_extent_map_item *dest,
+				  struct ocfs2_extent_map_item *src)
+{
+	dest->ei_cpos = src->ei_cpos;
+	dest->ei_phys = src->ei_phys;
+	dest->ei_clusters = src->ei_clusters;
+	dest->ei_flags = src->ei_flags;
+}
+
+/*
+ * Try to merge emi with ins. Returns 1 if merge succeeds, zero
+ * otherwise.
+ */
+static int ocfs2_try_to_merge_extent_map(struct ocfs2_extent_map_item *emi,
+					 struct ocfs2_extent_map_item *ins)
+{
+	/*
+	 * Handle contiguousness
+	 */
+	if (ins->ei_phys == (emi->ei_phys + emi->ei_clusters) &&
+	    ins->ei_cpos == (emi->ei_cpos + emi->ei_clusters) &&
+	    ins->ei_flags == emi->ei_flags) {
+		emi->ei_clusters += ins->ei_clusters;
+		return 1;
+	} else if ((ins->ei_phys + ins->ei_clusters) == emi->ei_phys &&
+		   (ins->ei_cpos + ins->ei_clusters) == emi->ei_cpos &&
+		   ins->ei_flags == emi->ei_flags) {
+		emi->ei_phys = ins->ei_phys;
+		emi->ei_cpos = ins->ei_cpos;
+		emi->ei_clusters += ins->ei_clusters;
+		return 1;
+	}
+
+	/*
+	 * Overlapping extents - this shouldn't happen unless we've
+	 * split an extent to change it's flags. That is exceedingly
+	 * rare, so there's no sense in trying to optimize it yet.
+	 */
+	if (ocfs2_ei_is_contained(emi, ins) ||
+	    ocfs2_ei_is_contained(ins, emi)) {
+		ocfs2_copy_emi_fields(emi, ins);
+		return 1;
+	}
+
+	/* No merge was possible. */
+	return 0;
+}
+
+/*
+ * In order to reduce complexity on the caller, this insert function
+ * is intentionally liberal in what it will accept.
+ *
+ * The only rule is that the truncate call *must* be used whenever
+ * records have been deleted. This avoids inserting overlapping
+ * records with different physical mappings.
+ */
+void ocfs2_extent_map_insert_rec(struct inode *inode,
+				 struct ocfs2_extent_rec *rec)
+{
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+	struct ocfs2_extent_map *em = &oi->ip_extent_map;
+	struct ocfs2_extent_map_item *emi, *new_emi = NULL;
+	struct ocfs2_extent_map_item ins;
+
+	ins.ei_cpos = le32_to_cpu(rec->e_cpos);
+	ins.ei_phys = ocfs2_blocks_to_clusters(inode->i_sb,
+					       le64_to_cpu(rec->e_blkno));
+	ins.ei_clusters = le16_to_cpu(rec->e_leaf_clusters);
+	ins.ei_flags = rec->e_flags;
+
+search:
+	spin_lock(&oi->ip_lock);
+
+	list_for_each_entry(emi, &em->em_list, ei_list) {
+		if (ocfs2_try_to_merge_extent_map(emi, &ins)) {
+			list_move(&emi->ei_list, &em->em_list);
+			spin_unlock(&oi->ip_lock);
+			goto out;
+		}
+	}
+
+	/*
+	 * No item could be merged.
+	 *
+	 * Either allocate and add a new item, or overwrite the last recently
+	 * inserted.
+	 */
+
+	if (em->em_num_items < OCFS2_MAX_EXTENT_MAP_ITEMS) {
+		if (new_emi == NULL) {
+			spin_unlock(&oi->ip_lock);
+
+			new_emi = kmalloc(sizeof(*new_emi), GFP_NOFS);
+			if (new_emi == NULL)
+				goto out;
+
+			goto search;
+		}
+
+		ocfs2_copy_emi_fields(new_emi, &ins);
+		list_add(&new_emi->ei_list, &em->em_list);
+		em->em_num_items++;
+		new_emi = NULL;
+	} else {
+		BUG_ON(list_empty(&em->em_list) || em->em_num_items == 0);
+		emi = list_entry(em->em_list.prev,
+				 struct ocfs2_extent_map_item, ei_list);
+		list_move(&emi->ei_list, &em->em_list);
+		ocfs2_copy_emi_fields(emi, &ins);
+	}
+
+	spin_unlock(&oi->ip_lock);
+
+out:
+	kfree(new_emi);
+}
+
+static int ocfs2_last_eb_is_empty(struct inode *inode,
+				  struct ocfs2_dinode *di)
+{
+	int ret, next_free;
+	u64 last_eb_blk = le64_to_cpu(di->i_last_eb_blk);
+	struct buffer_head *eb_bh = NULL;
+	struct ocfs2_extent_block *eb;
+	struct ocfs2_extent_list *el;
+
+	ret = ocfs2_read_extent_block(INODE_CACHE(inode), last_eb_blk, &eb_bh);
+	if (ret) {
+		mlog_errno(ret);
+		goto out;
+	}
+
+	eb = (struct ocfs2_extent_block *) eb_bh->b_data;
+	el = &eb->h_list;
+
+	if (el->l_tree_depth) {
+		ocfs2_error(inode->i_sb,
+			    "Inode %lu has non zero tree depth in "
+			    "leaf block %llu\n", inode->i_ino,
+			    (unsigned long long)eb_bh->b_blocknr);
+		ret = -EROFS;
+		goto out;
+	}
+
+	next_free = le16_to_cpu(el->l_next_free_rec);
+
+	if (next_free == 0 ||
+	    (next_free == 1 && ocfs2_is_empty_extent(&el->l_recs[0])))
+		ret = 1;
+
+out:
+	brelse(eb_bh);
+	return ret;
+}
+
+/*
+ * Return the 1st index within el which contains an extent start
+ * larger than v_cluster.
+ */
+static int ocfs2_search_for_hole_index(struct ocfs2_extent_list *el,
+				       u32 v_cluster)
+{
+	int i;
+	struct ocfs2_extent_rec *rec;
+
+	for(i = 0; i < le16_to_cpu(el->l_next_free_rec); i++) {
+		rec = &el->l_recs[i];
+
+		if (v_cluster < le32_to_cpu(rec->e_cpos))
+			break;
+	}
+
+	return i;
+}
+
+/*
+ * Figure out the size of a hole which starts at v_cluster within the given
+ * extent list.
+ *
+ * If there is no more allocation past v_cluster, we return the maximum
+ * cluster size minus v_cluster.
+ *
+ * If we have in-inode extents, then el points to the dinode list and
+ * eb_bh is NULL. Otherwise, eb_bh should point to the extent block
+ * containing el.
+ */
+int ocfs2_figure_hole_clusters(struct ocfs2_caching_info *ci,
+			       struct ocfs2_extent_list *el,
+			       struct buffer_head *eb_bh,
+			       u32 v_cluster,
+			       u32 *num_clusters)
+{
+	int ret, i;
+	struct buffer_head *next_eb_bh = NULL;
+	struct ocfs2_extent_block *eb, *next_eb;
+
+	i = ocfs2_search_for_hole_index(el, v_cluster);
+
+	if (i == le16_to_cpu(el->l_next_free_rec) && eb_bh) {
+		eb = (struct ocfs2_extent_block *)eb_bh->b_data;
+
+		/*
+		 * Check the next leaf for any extents.
+		 */
+
+		if (le64_to_cpu(eb->h_next_leaf_blk) == 0ULL)
+			goto no_more_extents;
+
+		ret = ocfs2_read_extent_block(ci,
+					      le64_to_cpu(eb->h_next_leaf_blk),
+					      &next_eb_bh);
+		if (ret) {
+			mlog_errno(ret);
+			goto out;
+		}
+
+		next_eb = (struct ocfs2_extent_block *)next_eb_bh->b_data;
+		el = &next_eb->h_list;
+		i = ocfs2_search_for_hole_index(el, v_cluster);
+	}
+
+no_more_extents:
+	if (i == le16_to_cpu(el->l_next_free_rec)) {
+		/*
+		 * We're at the end of our existing allocation. Just
+		 * return the maximum number of clusters we could
+		 * possibly allocate.
+		 */
+		*num_clusters = UINT_MAX - v_cluster;
+	} else {
+		*num_clusters = le32_to_cpu(el->l_recs[i].e_cpos) - v_cluster;
+	}
+
+	ret = 0;
+out:
+	brelse(next_eb_bh);
+	return ret;
+}
+
+static int ocfs2_get_clusters_nocache(struct inode *inode,
+				      struct buffer_head *di_bh,
+				      u32 v_cluster, unsigned int *hole_len,
+				      struct ocfs2_extent_rec *ret_rec,
+				      unsigned int *is_last)
+{
+	int i, ret, tree_height, len;
+	struct ocfs2_dinode *di;
+	struct ocfs2_extent_block *uninitialized_var(eb);
+	struct ocfs2_extent_list *el;
+	struct ocfs2_extent_rec *rec;
+	struct buffer_head *eb_bh = NULL;
+
+	memset(ret_rec, 0, sizeof(*ret_rec));
+	if (is_last)
+		*is_last = 0;
+
+	di = (struct ocfs2_dinode *) di_bh->b_data;
+	el = &di->id2.i_list;
+	tree_height = le16_to_cpu(el->l_tree_depth);
+
+	if (tree_height > 0) {
+		ret = ocfs2_find_leaf(INODE_CACHE(inode), el, v_cluster,
+				      &eb_bh);
+		if (ret) {
+			mlog_errno(ret);
+			goto out;
+		}
+
+		eb = (struct ocfs2_extent_block *) eb_bh->b_data;
+		el = &eb->h_list;
+
+		if (el->l_tree_depth) {
+			ocfs2_error(inode->i_sb,
+				    "Inode %lu has non zero tree depth in "
+				    "leaf block %llu\n", inode->i_ino,
+				    (unsigned long long)eb_bh->b_blocknr);
+			ret = -EROFS;
+			goto out;
+		}
+	}
+
+	i = ocfs2_search_extent_list(el, v_cluster);
+	if (i == -1) {
+		/*
+		 * Holes can be larger than the maximum size of an
+		 * extent, so we return their lengths in a separate
+		 * field.
+		 */
+		if (hole_len) {
+			ret = ocfs2_figure_hole_clusters(INODE_CACHE(inode),
+							 el, eb_bh,
+							 v_cluster, &len);
+			if (ret) {
+				mlog_errno(ret);
+				goto out;
+			}
+
+			*hole_len = len;
+		}
+		goto out_hole;
+	}
+
+	rec = &el->l_recs[i];
+
+	BUG_ON(v_cluster < le32_to_cpu(rec->e_cpos));
+
+	if (!rec->e_blkno) {
+		ocfs2_error(inode->i_sb, "Inode %lu has bad extent "
+			    "record (%u, %u, 0)", inode->i_ino,
+			    le32_to_cpu(rec->e_cpos),
+			    ocfs2_rec_clusters(el, rec));
+		ret = -EROFS;
+		goto out;
+	}
+
+	*ret_rec = *rec;
+
+	/*
+	 * Checking for last extent is potentially expensive - we
+	 * might have to look at the next leaf over to see if it's
+	 * empty.
+	 *
+	 * The first two checks are to see whether the caller even
+	 * cares for this information, and if the extent is at least
+	 * the last in it's list.
+	 *
+	 * If those hold true, then the extent is last if any of the
+	 * additional conditions hold true:
+	 *  - Extent list is in-inode
+	 *  - Extent list is right-most
+	 *  - Extent list is 2nd to rightmost, with empty right-most
+	 */
+	if (is_last) {
+		if (i == (le16_to_cpu(el->l_next_free_rec) - 1)) {
+			if (tree_height == 0)
+				*is_last = 1;
+			else if (eb->h_blkno == di->i_last_eb_blk)
+				*is_last = 1;
+			else if (eb->h_next_leaf_blk == di->i_last_eb_blk) {
+				ret = ocfs2_last_eb_is_empty(inode, di);
+				if (ret < 0) {
+					mlog_errno(ret);
+					goto out;
+				}
+				if (ret == 1)
+					*is_last = 1;
+			}
+		}
+	}
+
+out_hole:
+	ret = 0;
+out:
+	brelse(eb_bh);
+	return ret;
+}
+
+static void ocfs2_relative_extent_offsets(struct super_block *sb,
+					  u32 v_cluster,
+					  struct ocfs2_extent_rec *rec,
+					  u32 *p_cluster, u32 *num_clusters)
+
+{
+	u32 coff = v_cluster - le32_to_cpu(rec->e_cpos);
+
+	*p_cluster = ocfs2_blocks_to_clusters(sb, le64_to_cpu(rec->e_blkno));
+	*p_cluster = *p_cluster + coff;
+
+	if (num_clusters)
+		*num_clusters = le16_to_cpu(rec->e_leaf_clusters) - coff;
+}
+
+int ocfs2_xattr_get_clusters(struct inode *inode, u32 v_cluster,
+			     u32 *p_cluster, u32 *num_clusters,
+			     struct ocfs2_extent_list *el,
+			     unsigned int *extent_flags)
+{
+	int ret = 0, i;
+	struct buffer_head *eb_bh = NULL;
+	struct ocfs2_extent_block *eb;
+	struct ocfs2_extent_rec *rec;
+	u32 coff;
+
+	if (el->l_tree_depth) {
+		ret = ocfs2_find_leaf(INODE_CACHE(inode), el, v_cluster,
+				      &eb_bh);
+		if (ret) {
+			mlog_errno(ret);
+			goto out;
+		}
+
+		eb = (struct ocfs2_extent_block *) eb_bh->b_data;
+		el = &eb->h_list;
+
+		if (el->l_tree_depth) {
+			ocfs2_error(inode->i_sb,
+				    "Inode %lu has non zero tree depth in "
+				    "xattr leaf block %llu\n", inode->i_ino,
+				    (unsigned long long)eb_bh->b_blocknr);
+			ret = -EROFS;
+			goto out;
+		}
+	}
+
+	i = ocfs2_search_extent_list(el, v_cluster);
+	if (i == -1) {
+		ret = -EROFS;
+		mlog_errno(ret);
+		goto out;
+	} else {
+		rec = &el->l_recs[i];
+		BUG_ON(v_cluster < le32_to_cpu(rec->e_cpos));
+
+		if (!rec->e_blkno) {
+			ocfs2_error(inode->i_sb, "Inode %lu has bad extent "
+				    "record (%u, %u, 0) in xattr", inode->i_ino,
+				    le32_to_cpu(rec->e_cpos),
+				    ocfs2_rec_clusters(el, rec));
+			ret = -EROFS;
+			goto out;
+		}
+		coff = v_cluster - le32_to_cpu(rec->e_cpos);
+		*p_cluster = ocfs2_blocks_to_clusters(inode->i_sb,
+						    le64_to_cpu(rec->e_blkno));
+		*p_cluster = *p_cluster + coff;
+		if (num_clusters)
+			*num_clusters = ocfs2_rec_clusters(el, rec) - coff;
+
+		if (extent_flags)
+			*extent_flags = rec->e_flags;
+	}
+out:
+	if (eb_bh)
+		brelse(eb_bh);
+	return ret;
+}
+
+int ocfs2_get_clusters(struct inode *inode, u32 v_cluster,
+		       u32 *p_cluster, u32 *num_clusters,
+		       unsigned int *extent_flags)
+{
+	int ret;
+	unsigned int uninitialized_var(hole_len), flags = 0;
+	struct buffer_head *di_bh = NULL;
+	struct ocfs2_extent_rec rec;
+
+	if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
+		ret = -ERANGE;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	ret = ocfs2_extent_map_lookup(inode, v_cluster, p_cluster,
+				      num_clusters, extent_flags);
+	if (ret == 0)
+		goto out;
+
+	ret = ocfs2_read_inode_block(inode, &di_bh);
+	if (ret) {
+		mlog_errno(ret);
+		goto out;
+	}
+
+	ret = ocfs2_get_clusters_nocache(inode, di_bh, v_cluster, &hole_len,
+					 &rec, NULL);
+	if (ret) {
+		mlog_errno(ret);
+		goto out;
+	}
+
+	if (rec.e_blkno == 0ULL) {
+		/*
+		 * A hole was found. Return some canned values that
+		 * callers can key on. If asked for, num_clusters will
+		 * be populated with the size of the hole.
+		 */
+		*p_cluster = 0;
+		if (num_clusters) {
+			*num_clusters = hole_len;
+		}
+	} else {
+		ocfs2_relative_extent_offsets(inode->i_sb, v_cluster, &rec,
+					      p_cluster, num_clusters);
+		flags = rec.e_flags;
+
+		ocfs2_extent_map_insert_rec(inode, &rec);
+	}
+
+	if (extent_flags)
+		*extent_flags = flags;
+
+out:
+	brelse(di_bh);
+	return ret;
+}
+
+/*
+ * This expects alloc_sem to be held. The allocation cannot change at
+ * all while the map is in the process of being updated.
+ */
+int ocfs2_extent_map_get_blocks(struct inode *inode, u64 v_blkno, u64 *p_blkno,
+				u64 *ret_count, unsigned int *extent_flags)
+{
+	int ret;
+	int bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1);
+	u32 cpos, num_clusters, p_cluster;
+	u64 boff = 0;
+
+	cpos = ocfs2_blocks_to_clusters(inode->i_sb, v_blkno);
+
+	ret = ocfs2_get_clusters(inode, cpos, &p_cluster, &num_clusters,
+				 extent_flags);
+	if (ret) {
+		mlog_errno(ret);
+		goto out;
+	}
+
+	/*
+	 * p_cluster == 0 indicates a hole.
+	 */
+	if (p_cluster) {
+		boff = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster);
+		boff += (v_blkno & (u64)(bpc - 1));
+	}
+
+	*p_blkno = boff;
+
+	if (ret_count) {
+		*ret_count = ocfs2_clusters_to_blocks(inode->i_sb, num_clusters);
+		*ret_count -= v_blkno & (u64)(bpc - 1);
+	}
+
+out:
+	return ret;
+}
+
+/*
+ * The ocfs2_fiemap_inline() may be a little bit misleading, since
+ * it not only handles the fiemap for inlined files, but also deals
+ * with the fast symlink, cause they have no difference for extent
+ * mapping per se.
+ */
+static int ocfs2_fiemap_inline(struct inode *inode, struct buffer_head *di_bh,
+			       struct fiemap_extent_info *fieinfo,
+			       u64 map_start)
+{
+	int ret;
+	unsigned int id_count;
+	struct ocfs2_dinode *di;
+	u64 phys;
+	u32 flags = FIEMAP_EXTENT_DATA_INLINE|FIEMAP_EXTENT_LAST;
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+
+	di = (struct ocfs2_dinode *)di_bh->b_data;
+	if (ocfs2_inode_is_fast_symlink(inode))
+		id_count = ocfs2_fast_symlink_chars(inode->i_sb);
+	else
+		id_count = le16_to_cpu(di->id2.i_data.id_count);
+
+	if (map_start < id_count) {
+		phys = oi->ip_blkno << inode->i_sb->s_blocksize_bits;
+		if (ocfs2_inode_is_fast_symlink(inode))
+			phys += offsetof(struct ocfs2_dinode, id2.i_symlink);
+		else
+			phys += offsetof(struct ocfs2_dinode,
+					 id2.i_data.id_data);
+
+		ret = fiemap_fill_next_extent(fieinfo, 0, phys, id_count,
+					      flags);
+		if (ret < 0)
+			return ret;
+	}
+
+	return 0;
+}
+
+#define OCFS2_FIEMAP_FLAGS	(FIEMAP_FLAG_SYNC)
+
+int ocfs2_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
+		 u64 map_start, u64 map_len)
+{
+	int ret, is_last;
+	u32 mapping_end, cpos;
+	unsigned int hole_size;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	u64 len_bytes, phys_bytes, virt_bytes;
+	struct buffer_head *di_bh = NULL;
+	struct ocfs2_extent_rec rec;
+
+	ret = fiemap_check_flags(fieinfo, OCFS2_FIEMAP_FLAGS);
+	if (ret)
+		return ret;
+
+	ret = ocfs2_inode_lock(inode, &di_bh, 0);
+	if (ret) {
+		mlog_errno(ret);
+		goto out;
+	}
+
+	down_read(&OCFS2_I(inode)->ip_alloc_sem);
+
+	/*
+	 * Handle inline-data and fast symlink separately.
+	 */
+	if ((OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) ||
+	    ocfs2_inode_is_fast_symlink(inode)) {
+		ret = ocfs2_fiemap_inline(inode, di_bh, fieinfo, map_start);
+		goto out_unlock;
+	}
+
+	cpos = map_start >> osb->s_clustersize_bits;
+	mapping_end = ocfs2_clusters_for_bytes(inode->i_sb,
+					       map_start + map_len);
+	is_last = 0;
+	while (cpos < mapping_end && !is_last) {
+		u32 fe_flags;
+
+		ret = ocfs2_get_clusters_nocache(inode, di_bh, cpos,
+						 &hole_size, &rec, &is_last);
+		if (ret) {
+			mlog_errno(ret);
+			goto out_unlock;
+		}
+
+		if (rec.e_blkno == 0ULL) {
+			cpos += hole_size;
+			continue;
+		}
+
+		fe_flags = 0;
+		if (rec.e_flags & OCFS2_EXT_UNWRITTEN)
+			fe_flags |= FIEMAP_EXTENT_UNWRITTEN;
+		if (rec.e_flags & OCFS2_EXT_REFCOUNTED)
+			fe_flags |= FIEMAP_EXTENT_SHARED;
+		if (is_last)
+			fe_flags |= FIEMAP_EXTENT_LAST;
+		len_bytes = (u64)le16_to_cpu(rec.e_leaf_clusters) << osb->s_clustersize_bits;
+		phys_bytes = le64_to_cpu(rec.e_blkno) << osb->sb->s_blocksize_bits;
+		virt_bytes = (u64)le32_to_cpu(rec.e_cpos) << osb->s_clustersize_bits;
+
+		ret = fiemap_fill_next_extent(fieinfo, virt_bytes, phys_bytes,
+					      len_bytes, fe_flags);
+		if (ret)
+			break;
+
+		cpos = le32_to_cpu(rec.e_cpos)+ le16_to_cpu(rec.e_leaf_clusters);
+	}
+
+	if (ret > 0)
+		ret = 0;
+
+out_unlock:
+	brelse(di_bh);
+
+	up_read(&OCFS2_I(inode)->ip_alloc_sem);
+
+	ocfs2_inode_unlock(inode, 0);
+out:
+
+	return ret;
+}
+
+int ocfs2_seek_data_hole_offset(struct file *file, loff_t *offset, int whence)
+{
+	struct inode *inode = file->f_mapping->host;
+	int ret;
+	unsigned int is_last = 0, is_data = 0;
+	u16 cs_bits = OCFS2_SB(inode->i_sb)->s_clustersize_bits;
+	u32 cpos, cend, clen, hole_size;
+	u64 extoff, extlen;
+	struct buffer_head *di_bh = NULL;
+	struct ocfs2_extent_rec rec;
+
+	BUG_ON(whence != SEEK_DATA && whence != SEEK_HOLE);
+
+	ret = ocfs2_inode_lock(inode, &di_bh, 0);
+	if (ret) {
+		mlog_errno(ret);
+		goto out;
+	}
+
+	down_read(&OCFS2_I(inode)->ip_alloc_sem);
+
+	if (*offset >= i_size_read(inode)) {
+		ret = -ENXIO;
+		goto out_unlock;
+	}
+
+	if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
+		if (whence == SEEK_HOLE)
+			*offset = i_size_read(inode);
+		goto out_unlock;
+	}
+
+	clen = 0;
+	cpos = *offset >> cs_bits;
+	cend = ocfs2_clusters_for_bytes(inode->i_sb, i_size_read(inode));
+
+	while (cpos < cend && !is_last) {
+		ret = ocfs2_get_clusters_nocache(inode, di_bh, cpos, &hole_size,
+						 &rec, &is_last);
+		if (ret) {
+			mlog_errno(ret);
+			goto out_unlock;
+		}
+
+		extoff = cpos;
+		extoff <<= cs_bits;
+
+		if (rec.e_blkno == 0ULL) {
+			clen = hole_size;
+			is_data = 0;
+		} else {
+			clen = le16_to_cpu(rec.e_leaf_clusters) -
+				(cpos - le32_to_cpu(rec.e_cpos));
+			is_data = (rec.e_flags & OCFS2_EXT_UNWRITTEN) ?  0 : 1;
+		}
+
+		if ((!is_data && whence == SEEK_HOLE) ||
+		    (is_data && whence == SEEK_DATA)) {
+			if (extoff > *offset)
+				*offset = extoff;
+			goto out_unlock;
+		}
+
+		if (!is_last)
+			cpos += clen;
+	}
+
+	if (whence == SEEK_HOLE) {
+		extoff = cpos;
+		extoff <<= cs_bits;
+		extlen = clen;
+		extlen <<=  cs_bits;
+
+		if ((extoff + extlen) > i_size_read(inode))
+			extlen = i_size_read(inode) - extoff;
+		extoff += extlen;
+		if (extoff > *offset)
+			*offset = extoff;
+		goto out_unlock;
+	}
+
+	ret = -ENXIO;
+
+out_unlock:
+
+	brelse(di_bh);
+
+	up_read(&OCFS2_I(inode)->ip_alloc_sem);
+
+	ocfs2_inode_unlock(inode, 0);
+out:
+	return ret;
+}
+
+int ocfs2_read_virt_blocks(struct inode *inode, u64 v_block, int nr,
+			   struct buffer_head *bhs[], int flags,
+			   int (*validate)(struct super_block *sb,
+					   struct buffer_head *bh))
+{
+	int rc = 0;
+	u64 p_block, p_count;
+	int i, count, done = 0;
+
+	trace_ocfs2_read_virt_blocks(
+	     inode, (unsigned long long)v_block, nr, bhs, flags,
+	     validate);
+
+	if (((v_block + nr - 1) << inode->i_sb->s_blocksize_bits) >=
+	    i_size_read(inode)) {
+		BUG_ON(!(flags & OCFS2_BH_READAHEAD));
+		goto out;
+	}
+
+	while (done < nr) {
+		down_read(&OCFS2_I(inode)->ip_alloc_sem);
+		rc = ocfs2_extent_map_get_blocks(inode, v_block + done,
+						 &p_block, &p_count, NULL);
+		up_read(&OCFS2_I(inode)->ip_alloc_sem);
+		if (rc) {
+			mlog_errno(rc);
+			break;
+		}
+
+		if (!p_block) {
+			rc = -EIO;
+			mlog(ML_ERROR,
+			     "Inode #%llu contains a hole at offset %llu\n",
+			     (unsigned long long)OCFS2_I(inode)->ip_blkno,
+			     (unsigned long long)(v_block + done) <<
+			     inode->i_sb->s_blocksize_bits);
+			break;
+		}
+
+		count = nr - done;
+		if (p_count < count)
+			count = p_count;
+
+		/*
+		 * If the caller passed us bhs, they should have come
+		 * from a previous readahead call to this function.  Thus,
+		 * they should have the right b_blocknr.
+		 */
+		for (i = 0; i < count; i++) {
+			if (!bhs[done + i])
+				continue;
+			BUG_ON(bhs[done + i]->b_blocknr != (p_block + i));
+		}
+
+		rc = ocfs2_read_blocks(INODE_CACHE(inode), p_block, count,
+				       bhs + done, flags, validate);
+		if (rc) {
+			mlog_errno(rc);
+			break;
+		}
+		done += count;
+	}
+
+out:
+	return rc;
+}
+
+

fs/ocfs2/extent_map.h

@@ -0,0 +1,92 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * extent_map.h
+ *
+ * In-memory file extent mappings for OCFS2.
+ *
+ * Copyright (C) 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License, version 2,  as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef _EXTENT_MAP_H
+#define _EXTENT_MAP_H
+
+struct ocfs2_extent_map_item {
+	unsigned int			ei_cpos;
+	unsigned int			ei_phys;
+	unsigned int			ei_clusters;
+	unsigned int			ei_flags;
+
+	struct list_head		ei_list;
+};
+
+#define OCFS2_MAX_EXTENT_MAP_ITEMS			3
+struct ocfs2_extent_map {
+	unsigned int			em_num_items;
+	struct list_head		em_list;
+};
+
+void ocfs2_extent_map_init(struct inode *inode);
+void ocfs2_extent_map_trunc(struct inode *inode, unsigned int cluster);
+void ocfs2_extent_map_insert_rec(struct inode *inode,
+				 struct ocfs2_extent_rec *rec);
+
+int ocfs2_get_clusters(struct inode *inode, u32 v_cluster, u32 *p_cluster,
+		       u32 *num_clusters, unsigned int *extent_flags);
+int ocfs2_extent_map_get_blocks(struct inode *inode, u64 v_blkno, u64 *p_blkno,
+				u64 *ret_count, unsigned int *extent_flags);
+
+int ocfs2_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
+		 u64 map_start, u64 map_len);
+
+int ocfs2_seek_data_hole_offset(struct file *file, loff_t *offset, int origin);
+
+int ocfs2_xattr_get_clusters(struct inode *inode, u32 v_cluster,
+			     u32 *p_cluster, u32 *num_clusters,
+			     struct ocfs2_extent_list *el,
+			     unsigned int *extent_flags);
+
+int ocfs2_read_virt_blocks(struct inode *inode, u64 v_block, int nr,
+			   struct buffer_head *bhs[], int flags,
+			   int (*validate)(struct super_block *sb,
+					   struct buffer_head *bh));
+int ocfs2_figure_hole_clusters(struct ocfs2_caching_info *ci,
+			       struct ocfs2_extent_list *el,
+			       struct buffer_head *eb_bh,
+			       u32 v_cluster,
+			       u32 *num_clusters);
+static inline int ocfs2_read_virt_block(struct inode *inode, u64 v_block,
+					struct buffer_head **bh,
+					int (*validate)(struct super_block *sb,
+							struct buffer_head *bh))
+{
+	int status = 0;
+
+	if (bh == NULL) {
+		printk("ocfs2: bh == NULL\n");
+		status = -EINVAL;
+		goto bail;
+	}
+
+	status = ocfs2_read_virt_blocks(inode, v_block, 1, bh, 0, validate);
+
+bail:
+	return status;
+}
+
+
+#endif  /* _EXTENT_MAP_H */

fs/ocfs2/file.h

@@ -0,0 +1,76 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * file.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_FILE_H
+#define OCFS2_FILE_H
+
+extern const struct file_operations ocfs2_fops;
+extern const struct file_operations ocfs2_dops;
+extern const struct file_operations ocfs2_fops_no_plocks;
+extern const struct file_operations ocfs2_dops_no_plocks;
+extern const struct inode_operations ocfs2_file_iops;
+extern const struct inode_operations ocfs2_special_file_iops;
+struct ocfs2_alloc_context;
+enum ocfs2_alloc_restarted;
+
+struct ocfs2_file_private {
+	struct file		*fp_file;
+	struct mutex		fp_mutex;
+	struct ocfs2_lock_res	fp_flock;
+};
+
+int ocfs2_add_inode_data(struct ocfs2_super *osb,
+			 struct inode *inode,
+			 u32 *logical_offset,
+			 u32 clusters_to_add,
+			 int mark_unwritten,
+			 struct buffer_head *fe_bh,
+			 handle_t *handle,
+			 struct ocfs2_alloc_context *data_ac,
+			 struct ocfs2_alloc_context *meta_ac,
+			 enum ocfs2_alloc_restarted *reason_ret);
+int ocfs2_simple_size_update(struct inode *inode,
+			     struct buffer_head *di_bh,
+			     u64 new_i_size);
+int ocfs2_extend_no_holes(struct inode *inode, struct buffer_head *di_bh,
+			  u64 new_i_size, u64 zero_to);
+int ocfs2_zero_extend(struct inode *inode, struct buffer_head *di_bh,
+		      loff_t zero_to);
+int ocfs2_setattr(struct dentry *dentry, struct iattr *attr);
+int ocfs2_getattr(struct vfsmount *mnt, struct dentry *dentry,
+		  struct kstat *stat);
+int ocfs2_permission(struct inode *inode, int mask);
+
+int ocfs2_should_update_atime(struct inode *inode,
+			      struct vfsmount *vfsmnt);
+int ocfs2_update_inode_atime(struct inode *inode,
+			     struct buffer_head *bh);
+
+int ocfs2_change_file_space(struct file *file, unsigned int cmd,
+			    struct ocfs2_space_resv *sr);
+
+int ocfs2_check_range_for_refcount(struct inode *inode, loff_t pos,
+				   size_t count);
+#endif /* OCFS2_FILE_H */

fs/ocfs2/heartbeat.c

@@ -0,0 +1,134 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * heartbeat.c
+ *
+ * Register ourselves with the heartbaet service, keep our node maps
+ * up to date, and fire off recovery when needed.
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/highmem.h>
+
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "heartbeat.h"
+#include "inode.h"
+#include "journal.h"
+#include "ocfs2_trace.h"
+
+#include "buffer_head_io.h"
+
+static inline void __ocfs2_node_map_set_bit(struct ocfs2_node_map *map,
+					    int bit);
+static inline void __ocfs2_node_map_clear_bit(struct ocfs2_node_map *map,
+					      int bit);
+
+/* special case -1 for now
+ * TODO: should *really* make sure the calling func never passes -1!!  */
+static void ocfs2_node_map_init(struct ocfs2_node_map *map)
+{
+	map->num_nodes = OCFS2_NODE_MAP_MAX_NODES;
+	memset(map->map, 0, BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES) *
+	       sizeof(unsigned long));
+}
+
+void ocfs2_init_node_maps(struct ocfs2_super *osb)
+{
+	spin_lock_init(&osb->node_map_lock);
+	ocfs2_node_map_init(&osb->osb_recovering_orphan_dirs);
+}
+
+void ocfs2_do_node_down(int node_num, void *data)
+{
+	struct ocfs2_super *osb = data;
+
+	BUG_ON(osb->node_num == node_num);
+
+	trace_ocfs2_do_node_down(node_num);
+
+	if (!osb->cconn) {
+		/*
+		 * No cluster connection means we're not even ready to
+		 * participate yet.  We check the slots after the cluster
+		 * comes up, so we will notice the node death then.  We
+		 * can safely ignore it here.
+		 */
+		return;
+	}
+
+	ocfs2_recovery_thread(osb, node_num);
+}
+
+static inline void __ocfs2_node_map_set_bit(struct ocfs2_node_map *map,
+					    int bit)
+{
+	set_bit(bit, map->map);
+}
+
+void ocfs2_node_map_set_bit(struct ocfs2_super *osb,
+			    struct ocfs2_node_map *map,
+			    int bit)
+{
+	if (bit==-1)
+		return;
+	BUG_ON(bit >= map->num_nodes);
+	spin_lock(&osb->node_map_lock);
+	__ocfs2_node_map_set_bit(map, bit);
+	spin_unlock(&osb->node_map_lock);
+}
+
+static inline void __ocfs2_node_map_clear_bit(struct ocfs2_node_map *map,
+					      int bit)
+{
+	clear_bit(bit, map->map);
+}
+
+void ocfs2_node_map_clear_bit(struct ocfs2_super *osb,
+			      struct ocfs2_node_map *map,
+			      int bit)
+{
+	if (bit==-1)
+		return;
+	BUG_ON(bit >= map->num_nodes);
+	spin_lock(&osb->node_map_lock);
+	__ocfs2_node_map_clear_bit(map, bit);
+	spin_unlock(&osb->node_map_lock);
+}
+
+int ocfs2_node_map_test_bit(struct ocfs2_super *osb,
+			    struct ocfs2_node_map *map,
+			    int bit)
+{
+	int ret;
+	if (bit >= map->num_nodes) {
+		mlog(ML_ERROR, "bit=%d map->num_nodes=%d\n", bit, map->num_nodes);
+		BUG();
+	}
+	spin_lock(&osb->node_map_lock);
+	ret = test_bit(bit, map->map);
+	spin_unlock(&osb->node_map_lock);
+	return ret;
+}
+

fs/ocfs2/heartbeat.h

@@ -0,0 +1,45 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * heartbeat.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_HEARTBEAT_H
+#define OCFS2_HEARTBEAT_H
+
+void ocfs2_init_node_maps(struct ocfs2_super *osb);
+
+void ocfs2_do_node_down(int node_num, void *data);
+
+/* node map functions - used to keep track of mounted and in-recovery
+ * nodes. */
+void ocfs2_node_map_set_bit(struct ocfs2_super *osb,
+			    struct ocfs2_node_map *map,
+			    int bit);
+void ocfs2_node_map_clear_bit(struct ocfs2_super *osb,
+			      struct ocfs2_node_map *map,
+			      int bit);
+int ocfs2_node_map_test_bit(struct ocfs2_super *osb,
+			    struct ocfs2_node_map *map,
+			    int bit);
+
+#endif /* OCFS2_HEARTBEAT_H */

fs/ocfs2/inode.h

@@ -0,0 +1,186 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * inode.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_INODE_H
+#define OCFS2_INODE_H
+
+#include "extent_map.h"
+
+/* OCFS2 Inode Private Data */
+struct ocfs2_inode_info
+{
+	u64			ip_blkno;
+
+	struct ocfs2_lock_res		ip_rw_lockres;
+	struct ocfs2_lock_res		ip_inode_lockres;
+	struct ocfs2_lock_res		ip_open_lockres;
+
+	/* protects allocation changes on this inode. */
+	struct rw_semaphore		ip_alloc_sem;
+
+	/* protects extended attribute changes on this inode */
+	struct rw_semaphore		ip_xattr_sem;
+
+	/* Number of outstanding AIO's which are not page aligned */
+	struct mutex			ip_unaligned_aio;
+
+	/* These fields are protected by ip_lock */
+	spinlock_t			ip_lock;
+	u32				ip_open_count;
+	struct list_head		ip_io_markers;
+	u32				ip_clusters;
+
+	u16				ip_dyn_features;
+	struct mutex			ip_io_mutex;
+	u32				ip_flags; /* see below */
+	u32				ip_attr; /* inode attributes */
+
+	/* protected by recovery_lock. */
+	struct inode			*ip_next_orphan;
+
+	struct ocfs2_caching_info	ip_metadata_cache;
+	struct ocfs2_extent_map		ip_extent_map;
+	struct inode			vfs_inode;
+	struct jbd2_inode		ip_jinode;
+
+	u32				ip_dir_start_lookup;
+
+	/* Only valid if the inode is the dir. */
+	u32				ip_last_used_slot;
+	u64				ip_last_used_group;
+	u32				ip_dir_lock_gen;
+
+	struct ocfs2_alloc_reservation	ip_la_data_resv;
+
+	/*
+	 * Transactions that contain inode's metadata needed to complete
+	 * fsync and fdatasync, respectively.
+	 */
+	tid_t i_sync_tid;
+	tid_t i_datasync_tid;
+};
+
+/*
+ * Flags for the ip_flags field
+ */
+/* System file inodes  */
+#define OCFS2_INODE_SYSTEM_FILE		0x00000001
+#define OCFS2_INODE_JOURNAL		0x00000002
+#define OCFS2_INODE_BITMAP		0x00000004
+/* This inode has been wiped from disk */
+#define OCFS2_INODE_DELETED		0x00000008
+/* Has the inode been orphaned on another node?
+ *
+ * This hints to ocfs2_drop_inode that it should clear i_nlink before
+ * continuing.
+ *
+ * We *only* set this on unlink vote from another node. If the inode
+ * was locally orphaned, then we're sure of the state and don't need
+ * to twiddle i_nlink later - it's either zero or not depending on
+ * whether our unlink succeeded. Otherwise we got this from a node
+ * whose intention was to orphan the inode, however he may have
+ * crashed, failed etc, so we let ocfs2_drop_inode zero the value and
+ * rely on ocfs2_delete_inode to sort things out under the proper
+ * cluster locks.
+ */
+#define OCFS2_INODE_MAYBE_ORPHANED	0x00000010
+/* Does someone have the file open O_DIRECT */
+#define OCFS2_INODE_OPEN_DIRECT		0x00000020
+/* Tell the inode wipe code it's not in orphan dir */
+#define OCFS2_INODE_SKIP_ORPHAN_DIR     0x00000040
+
+static inline struct ocfs2_inode_info *OCFS2_I(struct inode *inode)
+{
+	return container_of(inode, struct ocfs2_inode_info, vfs_inode);
+}
+
+#define INODE_JOURNAL(i) (OCFS2_I(i)->ip_flags & OCFS2_INODE_JOURNAL)
+#define SET_INODE_JOURNAL(i) (OCFS2_I(i)->ip_flags |= OCFS2_INODE_JOURNAL)
+
+extern struct kmem_cache *ocfs2_inode_cache;
+
+extern const struct address_space_operations ocfs2_aops;
+extern const struct ocfs2_caching_operations ocfs2_inode_caching_ops;
+
+static inline struct ocfs2_caching_info *INODE_CACHE(struct inode *inode)
+{
+	return &OCFS2_I(inode)->ip_metadata_cache;
+}
+
+void ocfs2_evict_inode(struct inode *inode);
+int ocfs2_drop_inode(struct inode *inode);
+
+/* Flags for ocfs2_iget() */
+#define OCFS2_FI_FLAG_SYSFILE		0x1
+#define OCFS2_FI_FLAG_ORPHAN_RECOVERY	0x2
+struct inode *ocfs2_ilookup(struct super_block *sb, u64 feoff);
+struct inode *ocfs2_iget(struct ocfs2_super *osb, u64 feoff, unsigned flags,
+			 int sysfile_type);
+int ocfs2_inode_init_private(struct inode *inode);
+int ocfs2_inode_revalidate(struct dentry *dentry);
+void ocfs2_populate_inode(struct inode *inode, struct ocfs2_dinode *fe,
+			  int create_ino);
+void ocfs2_read_inode(struct inode *inode);
+void ocfs2_read_inode2(struct inode *inode, void *opaque);
+ssize_t ocfs2_rw_direct(int rw, struct file *filp, char *buf,
+			size_t size, loff_t *offp);
+void ocfs2_sync_blockdev(struct super_block *sb);
+void ocfs2_refresh_inode(struct inode *inode,
+			 struct ocfs2_dinode *fe);
+int ocfs2_mark_inode_dirty(handle_t *handle,
+			   struct inode *inode,
+			   struct buffer_head *bh);
+struct buffer_head *ocfs2_bread(struct inode *inode,
+				int block, int *err, int reada);
+
+void ocfs2_set_inode_flags(struct inode *inode);
+void ocfs2_get_inode_flags(struct ocfs2_inode_info *oi);
+
+static inline blkcnt_t ocfs2_inode_sector_count(struct inode *inode)
+{
+	int c_to_s_bits = OCFS2_SB(inode->i_sb)->s_clustersize_bits - 9;
+
+	return (blkcnt_t)OCFS2_I(inode)->ip_clusters << c_to_s_bits;
+}
+
+/* Validate that a bh contains a valid inode */
+int ocfs2_validate_inode_block(struct super_block *sb,
+			       struct buffer_head *bh);
+/*
+ * Read an inode block into *bh.  If *bh is NULL, a bh will be allocated.
+ * This is a cached read.  The inode will be validated with
+ * ocfs2_validate_inode_block().
+ */
+int ocfs2_read_inode_block(struct inode *inode, struct buffer_head **bh);
+/* The same, but can be passed OCFS2_BH_* flags */
+int ocfs2_read_inode_block_full(struct inode *inode, struct buffer_head **bh,
+				int flags);
+
+static inline struct ocfs2_inode_info *cache_info_to_inode(struct ocfs2_caching_info *ci)
+{
+	return container_of(ci, struct ocfs2_inode_info, ip_metadata_cache);
+}
+
+#endif /* OCFS2_INODE_H */

fs/ocfs2/ioctl.h

@@ -0,0 +1,16 @@
+/*
+ * ioctl.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2006 Herbert Poetzl
+ *
+ */
+
+#ifndef OCFS2_IOCTL_PROTO_H
+#define OCFS2_IOCTL_PROTO_H
+
+long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg);
+long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg);
+
+#endif /* OCFS2_IOCTL_PROTO_H */

fs/ocfs2/journal.h

@@ -0,0 +1,640 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * journal.h
+ *
+ * Defines journalling api and structures.
+ *
+ * Copyright (C) 2003, 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_JOURNAL_H
+#define OCFS2_JOURNAL_H
+
+#include <linux/fs.h>
+#include <linux/jbd2.h>
+
+enum ocfs2_journal_state {
+	OCFS2_JOURNAL_FREE = 0,
+	OCFS2_JOURNAL_LOADED,
+	OCFS2_JOURNAL_IN_SHUTDOWN,
+};
+
+struct ocfs2_super;
+struct ocfs2_dinode;
+
+/*
+ * The recovery_list is a simple linked list of node numbers to recover.
+ * It is protected by the recovery_lock.
+ */
+
+struct ocfs2_recovery_map {
+	unsigned int rm_used;
+	unsigned int *rm_entries;
+};
+
+
+struct ocfs2_journal {
+	enum ocfs2_journal_state   j_state;    /* Journals current state   */
+
+	journal_t                 *j_journal; /* The kernels journal type */
+	struct inode              *j_inode;   /* Kernel inode pointing to
+					       * this journal             */
+	struct ocfs2_super        *j_osb;     /* pointer to the super
+					       * block for the node
+					       * we're currently
+					       * running on -- not
+					       * necessarily the super
+					       * block from the node
+					       * which we usually run
+					       * from (recovery,
+					       * etc)                     */
+	struct buffer_head        *j_bh;      /* Journal disk inode block */
+	atomic_t                  j_num_trans; /* Number of transactions
+					        * currently in the system. */
+	spinlock_t                j_lock;
+	unsigned long             j_trans_id;
+	struct rw_semaphore       j_trans_barrier;
+	wait_queue_head_t         j_checkpointed;
+
+	/* both fields protected by j_lock*/
+	struct list_head          j_la_cleanups;
+	struct work_struct        j_recovery_work;
+};
+
+extern spinlock_t trans_inc_lock;
+
+/* wrap j_trans_id so we never have it equal to zero. */
+static inline unsigned long ocfs2_inc_trans_id(struct ocfs2_journal *j)
+{
+	unsigned long old_id;
+	spin_lock(&trans_inc_lock);
+	old_id = j->j_trans_id++;
+	if (unlikely(!j->j_trans_id))
+		j->j_trans_id = 1;
+	spin_unlock(&trans_inc_lock);
+	return old_id;
+}
+
+static inline void ocfs2_set_ci_lock_trans(struct ocfs2_journal *journal,
+					   struct ocfs2_caching_info *ci)
+{
+	spin_lock(&trans_inc_lock);
+	ci->ci_last_trans = journal->j_trans_id;
+	spin_unlock(&trans_inc_lock);
+}
+
+/* Used to figure out whether it's safe to drop a metadata lock on an
+ * cached object. Returns true if all the object's changes have been
+ * checkpointed to disk. You should be holding the spinlock on the
+ * metadata lock while calling this to be sure that nobody can take
+ * the lock and put it on another transaction. */
+static inline int ocfs2_ci_fully_checkpointed(struct ocfs2_caching_info *ci)
+{
+	int ret;
+	struct ocfs2_journal *journal =
+		OCFS2_SB(ocfs2_metadata_cache_get_super(ci))->journal;
+
+	spin_lock(&trans_inc_lock);
+	ret = time_after(journal->j_trans_id, ci->ci_last_trans);
+	spin_unlock(&trans_inc_lock);
+	return ret;
+}
+
+/* convenience function to check if an object backed by struct
+ * ocfs2_caching_info  is still new (has never hit disk) Will do you a
+ * favor and set created_trans = 0 when you've
+ * been checkpointed.  returns '1' if the ci is still new. */
+static inline int ocfs2_ci_is_new(struct ocfs2_caching_info *ci)
+{
+	int ret;
+	struct ocfs2_journal *journal =
+		OCFS2_SB(ocfs2_metadata_cache_get_super(ci))->journal;
+
+	spin_lock(&trans_inc_lock);
+	ret = !(time_after(journal->j_trans_id, ci->ci_created_trans));
+	if (!ret)
+		ci->ci_created_trans = 0;
+	spin_unlock(&trans_inc_lock);
+	return ret;
+}
+
+/* Wrapper for inodes so we can check system files */
+static inline int ocfs2_inode_is_new(struct inode *inode)
+{
+	/* System files are never "new" as they're written out by
+	 * mkfs. This helps us early during mount, before we have the
+	 * journal open and j_trans_id could be junk. */
+	if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SYSTEM_FILE)
+		return 0;
+
+	return ocfs2_ci_is_new(INODE_CACHE(inode));
+}
+
+static inline void ocfs2_ci_set_new(struct ocfs2_super *osb,
+				    struct ocfs2_caching_info *ci)
+{
+	spin_lock(&trans_inc_lock);
+	ci->ci_created_trans = osb->journal->j_trans_id;
+	spin_unlock(&trans_inc_lock);
+}
+
+/* Exported only for the journal struct init code in super.c. Do not call. */
+void ocfs2_orphan_scan_init(struct ocfs2_super *osb);
+void ocfs2_orphan_scan_start(struct ocfs2_super *osb);
+void ocfs2_orphan_scan_stop(struct ocfs2_super *osb);
+void ocfs2_orphan_scan_exit(struct ocfs2_super *osb);
+
+void ocfs2_complete_recovery(struct work_struct *work);
+void ocfs2_wait_for_recovery(struct ocfs2_super *osb);
+
+int ocfs2_recovery_init(struct ocfs2_super *osb);
+void ocfs2_recovery_exit(struct ocfs2_super *osb);
+
+int ocfs2_compute_replay_slots(struct ocfs2_super *osb);
+/*
+ *  Journal Control:
+ *  Initialize, Load, Shutdown, Wipe a journal.
+ *
+ *  ocfs2_journal_init     - Initialize journal structures in the OSB.
+ *  ocfs2_journal_load     - Load the given journal off disk. Replay it if
+ *                          there's transactions still in there.
+ *  ocfs2_journal_shutdown - Shutdown a journal, this will flush all
+ *                          uncommitted, uncheckpointed transactions.
+ *  ocfs2_journal_wipe     - Wipe transactions from a journal. Optionally
+ *                          zero out each block.
+ *  ocfs2_recovery_thread  - Perform recovery on a node. osb is our own osb.
+ *  ocfs2_mark_dead_nodes - Start recovery on nodes we won't get a heartbeat
+ *                          event on.
+ *  ocfs2_start_checkpoint - Kick the commit thread to do a checkpoint.
+ */
+void   ocfs2_set_journal_params(struct ocfs2_super *osb);
+int    ocfs2_journal_init(struct ocfs2_journal *journal,
+			  int *dirty);
+void   ocfs2_journal_shutdown(struct ocfs2_super *osb);
+int    ocfs2_journal_wipe(struct ocfs2_journal *journal,
+			  int full);
+int    ocfs2_journal_load(struct ocfs2_journal *journal, int local,
+			  int replayed);
+int    ocfs2_check_journals_nolocks(struct ocfs2_super *osb);
+void   ocfs2_recovery_thread(struct ocfs2_super *osb,
+			     int node_num);
+int    ocfs2_mark_dead_nodes(struct ocfs2_super *osb);
+void   ocfs2_complete_mount_recovery(struct ocfs2_super *osb);
+void ocfs2_complete_quota_recovery(struct ocfs2_super *osb);
+
+static inline void ocfs2_start_checkpoint(struct ocfs2_super *osb)
+{
+	wake_up(&osb->checkpoint_event);
+}
+
+static inline void ocfs2_checkpoint_inode(struct inode *inode)
+{
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+	if (ocfs2_mount_local(osb))
+		return;
+
+	if (!ocfs2_ci_fully_checkpointed(INODE_CACHE(inode))) {
+		/* WARNING: This only kicks off a single
+		 * checkpoint. If someone races you and adds more
+		 * metadata to the journal, you won't know, and will
+		 * wind up waiting *a lot* longer than necessary. Right
+		 * now we only use this in clear_inode so that's
+		 * OK. */
+		ocfs2_start_checkpoint(osb);
+
+		wait_event(osb->journal->j_checkpointed,
+			   ocfs2_ci_fully_checkpointed(INODE_CACHE(inode)));
+	}
+}
+
+/*
+ *  Transaction Handling:
+ *  Manage the lifetime of a transaction handle.
+ *
+ *  ocfs2_start_trans      - Begin a transaction. Give it an upper estimate of
+ *                          the number of blocks that will be changed during
+ *                          this handle.
+ *  ocfs2_commit_trans - Complete a handle. It might return -EIO if
+ *                       the journal was aborted. The majority of paths don't
+ *                       check the return value as an error there comes too
+ *                       late to do anything (and will be picked up in a
+ *                       later transaction).
+ *  ocfs2_extend_trans     - Extend a handle by nblocks credits. This may
+ *                          commit the handle to disk in the process, but will
+ *                          not release any locks taken during the transaction.
+ *  ocfs2_journal_access* - Notify the handle that we want to journal this
+ *                          buffer. Will have to call ocfs2_journal_dirty once
+ *                          we've actually dirtied it. Type is one of . or .
+ *                          Always call the specific flavor of
+ *                          ocfs2_journal_access_*() unless you intend to
+ *                          manage the checksum by hand.
+ *  ocfs2_journal_dirty    - Mark a journalled buffer as having dirty data.
+ *  ocfs2_jbd2_file_inode  - Mark an inode so that its data goes out before
+ *                           the current handle commits.
+ */
+
+/* You must always start_trans with a number of buffs > 0, but it's
+ * perfectly legal to go through an entire transaction without having
+ * dirtied any buffers. */
+handle_t		    *ocfs2_start_trans(struct ocfs2_super *osb,
+					       int max_buffs);
+int			     ocfs2_commit_trans(struct ocfs2_super *osb,
+						handle_t *handle);
+int			     ocfs2_extend_trans(handle_t *handle, int nblocks);
+int			     ocfs2_allocate_extend_trans(handle_t *handle,
+						int thresh);
+
+/*
+ * Define an arbitrary limit for the amount of data we will anticipate
+ * writing to any given transaction.  For unbounded transactions such as
+ * fallocate(2) we can write more than this, but we always
+ * start off at the maximum transaction size and grow the transaction
+ * optimistically as we go.
+ */
+#define OCFS2_MAX_TRANS_DATA	64U
+
+/*
+ * Create access is for when we get a newly created buffer and we're
+ * not gonna read it off disk, but rather fill it ourselves.  Right
+ * now, we don't do anything special with this (it turns into a write
+ * request), but this is a good placeholder in case we do...
+ *
+ * Write access is for when we read a block off disk and are going to
+ * modify it. This way the journalling layer knows it may need to make
+ * a copy of that block (if it's part of another, uncommitted
+ * transaction) before we do so.
+ */
+#define OCFS2_JOURNAL_ACCESS_CREATE 0
+#define OCFS2_JOURNAL_ACCESS_WRITE  1
+#define OCFS2_JOURNAL_ACCESS_UNDO   2
+
+
+/* ocfs2_inode */
+int ocfs2_journal_access_di(handle_t *handle, struct ocfs2_caching_info *ci,
+			    struct buffer_head *bh, int type);
+/* ocfs2_extent_block */
+int ocfs2_journal_access_eb(handle_t *handle, struct ocfs2_caching_info *ci,
+			    struct buffer_head *bh, int type);
+/* ocfs2_refcount_block */
+int ocfs2_journal_access_rb(handle_t *handle, struct ocfs2_caching_info *ci,
+			    struct buffer_head *bh, int type);
+/* ocfs2_group_desc */
+int ocfs2_journal_access_gd(handle_t *handle, struct ocfs2_caching_info *ci,
+			    struct buffer_head *bh, int type);
+/* ocfs2_xattr_block */
+int ocfs2_journal_access_xb(handle_t *handle, struct ocfs2_caching_info *ci,
+			    struct buffer_head *bh, int type);
+/* quota blocks */
+int ocfs2_journal_access_dq(handle_t *handle, struct ocfs2_caching_info *ci,
+			    struct buffer_head *bh, int type);
+/* dirblock */
+int ocfs2_journal_access_db(handle_t *handle, struct ocfs2_caching_info *ci,
+			    struct buffer_head *bh, int type);
+/* ocfs2_dx_root_block */
+int ocfs2_journal_access_dr(handle_t *handle, struct ocfs2_caching_info *ci,
+			    struct buffer_head *bh, int type);
+/* ocfs2_dx_leaf */
+int ocfs2_journal_access_dl(handle_t *handle, struct ocfs2_caching_info *ci,
+			    struct buffer_head *bh, int type);
+/* Anything that has no ecc */
+int ocfs2_journal_access(handle_t *handle, struct ocfs2_caching_info *ci,
+			 struct buffer_head *bh, int type);
+
+/*
+ * A word about the journal_access/journal_dirty "dance". It is
+ * entirely legal to journal_access a buffer more than once (as long
+ * as the access type is the same -- I'm not sure what will happen if
+ * access type is different but this should never happen anyway) It is
+ * also legal to journal_dirty a buffer more than once. In fact, you
+ * can even journal_access a buffer after you've done a
+ * journal_access/journal_dirty pair. The only thing you cannot do
+ * however, is journal_dirty a buffer which you haven't yet passed to
+ * journal_access at least once.
+ *
+ * That said, 99% of the time this doesn't matter and this is what the
+ * path looks like:
+ *
+ *	<read a bh>
+ *	ocfs2_journal_access(handle, bh,	OCFS2_JOURNAL_ACCESS_WRITE);
+ *	<modify the bh>
+ * 	ocfs2_journal_dirty(handle, bh);
+ */
+void ocfs2_journal_dirty(handle_t *handle, struct buffer_head *bh);
+
+/*
+ *  Credit Macros:
+ *  Convenience macros to calculate number of credits needed.
+ *
+ *  For convenience sake, I have a set of macros here which calculate
+ *  the *maximum* number of sectors which will be changed for various
+ *  metadata updates.
+ */
+
+/* simple file updates like chmod, etc. */
+#define OCFS2_INODE_UPDATE_CREDITS 1
+
+/* extended attribute block update */
+#define OCFS2_XATTR_BLOCK_UPDATE_CREDITS 1
+
+/* Update of a single quota block */
+#define OCFS2_QUOTA_BLOCK_UPDATE_CREDITS 1
+
+/* global quotafile inode update, data block */
+#define OCFS2_QINFO_WRITE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + \
+				   OCFS2_QUOTA_BLOCK_UPDATE_CREDITS)
+
+#define OCFS2_LOCAL_QINFO_WRITE_CREDITS OCFS2_QUOTA_BLOCK_UPDATE_CREDITS
+/*
+ * The two writes below can accidentally see global info dirty due
+ * to set_info() quotactl so make them prepared for the writes.
+ */
+/* quota data block, global info */
+/* Write to local quota file */
+#define OCFS2_QWRITE_CREDITS (OCFS2_QINFO_WRITE_CREDITS + \
+			      OCFS2_QUOTA_BLOCK_UPDATE_CREDITS)
+
+/* global quota data block, local quota data block, global quota inode,
+ * global quota info */
+#define OCFS2_QSYNC_CREDITS (OCFS2_QINFO_WRITE_CREDITS + \
+			     2 * OCFS2_QUOTA_BLOCK_UPDATE_CREDITS)
+
+static inline int ocfs2_quota_trans_credits(struct super_block *sb)
+{
+	int credits = 0;
+
+	if (OCFS2_HAS_RO_COMPAT_FEATURE(sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA))
+		credits += OCFS2_QWRITE_CREDITS;
+	if (OCFS2_HAS_RO_COMPAT_FEATURE(sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA))
+		credits += OCFS2_QWRITE_CREDITS;
+	return credits;
+}
+
+/* group extend. inode update and last group update. */
+#define OCFS2_GROUP_EXTEND_CREDITS	(OCFS2_INODE_UPDATE_CREDITS + 1)
+
+/* group add. inode update and the new group update. */
+#define OCFS2_GROUP_ADD_CREDITS	(OCFS2_INODE_UPDATE_CREDITS + 1)
+
+/* get one bit out of a suballocator: dinode + group descriptor +
+ * prev. group desc. if we relink. */
+#define OCFS2_SUBALLOC_ALLOC (3)
+
+static inline int ocfs2_inline_to_extents_credits(struct super_block *sb)
+{
+	return OCFS2_SUBALLOC_ALLOC + OCFS2_INODE_UPDATE_CREDITS +
+	       ocfs2_quota_trans_credits(sb);
+}
+
+/* dinode + group descriptor update. We don't relink on free yet. */
+#define OCFS2_SUBALLOC_FREE  (2)
+
+#define OCFS2_TRUNCATE_LOG_UPDATE OCFS2_INODE_UPDATE_CREDITS
+#define OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC (OCFS2_SUBALLOC_FREE 		      \
+					 + OCFS2_TRUNCATE_LOG_UPDATE)
+
+static inline int ocfs2_remove_extent_credits(struct super_block *sb)
+{
+	return OCFS2_TRUNCATE_LOG_UPDATE + OCFS2_INODE_UPDATE_CREDITS +
+	       ocfs2_quota_trans_credits(sb);
+}
+
+/* data block for new dir/symlink, allocation of directory block, dx_root
+ * update for free list */
+#define OCFS2_DIR_LINK_ADDITIONAL_CREDITS (1 + OCFS2_SUBALLOC_ALLOC + 1)
+
+static inline int ocfs2_add_dir_index_credits(struct super_block *sb)
+{
+	/* 1 block for index, 2 allocs (data, metadata), 1 clusters
+	 * worth of blocks for initial extent. */
+	return 1 + 2 * OCFS2_SUBALLOC_ALLOC +
+		ocfs2_clusters_to_blocks(sb, 1);
+}
+
+/* parent fe, parent block, new file entry, index leaf, inode alloc fe, inode
+ * alloc group descriptor + mkdir/symlink blocks + dir blocks + xattr
+ * blocks + quota update */
+static inline int ocfs2_mknod_credits(struct super_block *sb, int is_dir,
+				      int xattr_credits)
+{
+	int dir_credits = OCFS2_DIR_LINK_ADDITIONAL_CREDITS;
+
+	if (is_dir)
+		dir_credits += ocfs2_add_dir_index_credits(sb);
+
+	return 4 + OCFS2_SUBALLOC_ALLOC + dir_credits + xattr_credits +
+	       ocfs2_quota_trans_credits(sb);
+}
+
+/* local alloc metadata change + main bitmap updates */
+#define OCFS2_WINDOW_MOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS                 \
+				  + OCFS2_SUBALLOC_ALLOC + OCFS2_SUBALLOC_FREE)
+
+/* used when we don't need an allocation change for a dir extend. One
+ * for the dinode, one for the new block. */
+#define OCFS2_SIMPLE_DIR_EXTEND_CREDITS (2)
+
+/* file update (nlink, etc) + directory mtime/ctime + dir entry block + quota
+ * update on dir + index leaf + dx root update for free list +
+ * previous dirblock update in the free list */
+static inline int ocfs2_link_credits(struct super_block *sb)
+{
+	return 2*OCFS2_INODE_UPDATE_CREDITS + 4 +
+	       ocfs2_quota_trans_credits(sb);
+}
+
+/* inode + dir inode (if we unlink a dir), + dir entry block + orphan
+ * dir inode link + dir inode index leaf + dir index root */
+static inline int ocfs2_unlink_credits(struct super_block *sb)
+{
+	/* The quota update from ocfs2_link_credits is unused here... */
+	return 2 * OCFS2_INODE_UPDATE_CREDITS + 3 + ocfs2_link_credits(sb);
+}
+
+/* dinode + orphan dir dinode + inode alloc dinode + orphan dir entry +
+ * inode alloc group descriptor + orphan dir index root +
+ * orphan dir index leaf */
+#define OCFS2_DELETE_INODE_CREDITS (3 * OCFS2_INODE_UPDATE_CREDITS + 4)
+
+/* dinode update, old dir dinode update, new dir dinode update, old
+ * dir dir entry, new dir dir entry, dir entry update for renaming
+ * directory + target unlink + 3 x dir index leaves */
+static inline int ocfs2_rename_credits(struct super_block *sb)
+{
+	return 3 * OCFS2_INODE_UPDATE_CREDITS + 6 + ocfs2_unlink_credits(sb);
+}
+
+/* global bitmap dinode, group desc., relinked group,
+ * suballocator dinode, group desc., relinked group,
+ * dinode, xattr block */
+#define OCFS2_XATTR_BLOCK_CREATE_CREDITS (OCFS2_SUBALLOC_ALLOC * 2 + \
+					  + OCFS2_INODE_UPDATE_CREDITS \
+					  + OCFS2_XATTR_BLOCK_UPDATE_CREDITS)
+
+/* inode update, removal of dx root block from allocator */
+#define OCFS2_DX_ROOT_REMOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS +	\
+				      OCFS2_SUBALLOC_FREE)
+
+static inline int ocfs2_calc_dxi_expand_credits(struct super_block *sb)
+{
+	int credits = 1 + OCFS2_SUBALLOC_ALLOC;
+
+	credits += ocfs2_clusters_to_blocks(sb, 1);
+	credits += ocfs2_quota_trans_credits(sb);
+
+	return credits;
+}
+
+/* inode update, new refcount block and its allocation credits. */
+#define OCFS2_REFCOUNT_TREE_CREATE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1 \
+					    + OCFS2_SUBALLOC_ALLOC)
+
+/* inode and the refcount block update. */
+#define OCFS2_REFCOUNT_TREE_SET_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1)
+
+/*
+ * inode and the refcount block update.
+ * It doesn't include the credits for sub alloc change.
+ * So if we need to free the bit, OCFS2_SUBALLOC_FREE needs to be added.
+ */
+#define OCFS2_REFCOUNT_TREE_REMOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1)
+
+/* 2 metadata alloc, 2 new blocks and root refcount block */
+#define OCFS2_EXPAND_REFCOUNT_TREE_CREDITS (OCFS2_SUBALLOC_ALLOC * 2 + 3)
+
+/*
+ * Please note that the caller must make sure that root_el is the root
+ * of extent tree. So for an inode, it should be &fe->id2.i_list. Otherwise
+ * the result may be wrong.
+ */
+static inline int ocfs2_calc_extend_credits(struct super_block *sb,
+					    struct ocfs2_extent_list *root_el)
+{
+	int bitmap_blocks, sysfile_bitmap_blocks, extent_blocks;
+
+	/* bitmap dinode, group desc. + relinked group. */
+	bitmap_blocks = OCFS2_SUBALLOC_ALLOC;
+
+	/* we might need to shift tree depth so lets assume an
+	 * absolute worst case of complete fragmentation.  Even with
+	 * that, we only need one update for the dinode, and then
+	 * however many metadata chunks needed * a remaining suballoc
+	 * alloc. */
+	sysfile_bitmap_blocks = 1 +
+		(OCFS2_SUBALLOC_ALLOC - 1) * ocfs2_extend_meta_needed(root_el);
+
+	/* this does not include *new* metadata blocks, which are
+	 * accounted for in sysfile_bitmap_blocks. root_el +
+	 * prev. last_eb_blk + blocks along edge of tree.
+	 * calc_symlink_credits passes because we just need 1
+	 * credit for the dinode there. */
+	extent_blocks = 1 + 1 + le16_to_cpu(root_el->l_tree_depth);
+
+	return bitmap_blocks + sysfile_bitmap_blocks + extent_blocks +
+	       ocfs2_quota_trans_credits(sb);
+}
+
+static inline int ocfs2_calc_symlink_credits(struct super_block *sb)
+{
+	int blocks = ocfs2_mknod_credits(sb, 0, 0);
+
+	/* links can be longer than one block so we may update many
+	 * within our single allocated extent. */
+	blocks += ocfs2_clusters_to_blocks(sb, 1);
+
+	return blocks + ocfs2_quota_trans_credits(sb);
+}
+
+static inline int ocfs2_calc_group_alloc_credits(struct super_block *sb,
+						 unsigned int cpg)
+{
+	int blocks;
+	int bitmap_blocks = OCFS2_SUBALLOC_ALLOC + 1;
+	/* parent inode update + new block group header + bitmap inode update
+	   + bitmap blocks affected */
+	blocks = 1 + 1 + 1 + bitmap_blocks;
+	return blocks;
+}
+
+/*
+ * Allocating a discontiguous block group requires the credits from
+ * ocfs2_calc_group_alloc_credits() as well as enough credits to fill
+ * the group descriptor's extent list.  The caller already has started
+ * the transaction with ocfs2_calc_group_alloc_credits().  They extend
+ * it with these credits.
+ */
+static inline int ocfs2_calc_bg_discontig_credits(struct super_block *sb)
+{
+	return ocfs2_extent_recs_per_gd(sb);
+}
+
+static inline int ocfs2_calc_tree_trunc_credits(struct super_block *sb,
+						unsigned int clusters_to_del,
+						struct ocfs2_dinode *fe,
+						struct ocfs2_extent_list *last_el)
+{
+ 	/* for dinode + all headers in this pass + update to next leaf */
+	u16 next_free = le16_to_cpu(last_el->l_next_free_rec);
+	u16 tree_depth = le16_to_cpu(fe->id2.i_list.l_tree_depth);
+	int credits = 1 + tree_depth + 1;
+	int i;
+
+	i = next_free - 1;
+	BUG_ON(i < 0);
+
+	/* We may be deleting metadata blocks, so metadata alloc dinode +
+	   one desc. block for each possible delete. */
+	if (tree_depth && next_free == 1 &&
+	    ocfs2_rec_clusters(last_el, &last_el->l_recs[i]) == clusters_to_del)
+		credits += 1 + tree_depth;
+
+	/* update to the truncate log. */
+	credits += OCFS2_TRUNCATE_LOG_UPDATE;
+
+	credits += ocfs2_quota_trans_credits(sb);
+
+	return credits;
+}
+
+static inline int ocfs2_jbd2_file_inode(handle_t *handle, struct inode *inode)
+{
+	return jbd2_journal_file_inode(handle, &OCFS2_I(inode)->ip_jinode);
+}
+
+static inline int ocfs2_begin_ordered_truncate(struct inode *inode,
+					       loff_t new_size)
+{
+	return jbd2_journal_begin_ordered_truncate(
+				OCFS2_SB(inode->i_sb)->journal->j_journal,
+				&OCFS2_I(inode)->ip_jinode,
+				new_size);
+}
+
+static inline void ocfs2_update_inode_fsync_trans(handle_t *handle,
+						  struct inode *inode,
+						  int datasync)
+{
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+
+	oi->i_sync_tid = handle->h_transaction->t_tid;
+	if (datasync)
+		oi->i_datasync_tid = handle->h_transaction->t_tid;
+}
+
+#endif /* OCFS2_JOURNAL_H */

fs/ocfs2/localalloc.h

@@ -0,0 +1,68 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * localalloc.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_LOCALALLOC_H
+#define OCFS2_LOCALALLOC_H
+
+int ocfs2_load_local_alloc(struct ocfs2_super *osb);
+
+void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb);
+
+void ocfs2_la_set_sizes(struct ocfs2_super *osb, int requested_mb);
+unsigned int ocfs2_la_default_mb(struct ocfs2_super *osb);
+
+int ocfs2_begin_local_alloc_recovery(struct ocfs2_super *osb,
+				     int node_num,
+				     struct ocfs2_dinode **alloc_copy);
+
+int ocfs2_complete_local_alloc_recovery(struct ocfs2_super *osb,
+					struct ocfs2_dinode *alloc);
+
+int ocfs2_alloc_should_use_local(struct ocfs2_super *osb,
+				 u64 bits);
+
+struct ocfs2_alloc_context;
+int ocfs2_reserve_local_alloc_bits(struct ocfs2_super *osb,
+				   u32 bits_wanted,
+				   struct ocfs2_alloc_context *ac);
+
+int ocfs2_claim_local_alloc_bits(struct ocfs2_super *osb,
+				 handle_t *handle,
+				 struct ocfs2_alloc_context *ac,
+				 u32 bits_wanted,
+				 u32 *bit_off,
+				 u32 *num_bits);
+
+int ocfs2_free_local_alloc_bits(struct ocfs2_super *osb,
+				handle_t *handle,
+				struct ocfs2_alloc_context *ac,
+				u32 bit_off,
+				u32 num_bits);
+
+void ocfs2_local_alloc_seen_free_bits(struct ocfs2_super *osb,
+				      unsigned int num_clusters);
+void ocfs2_la_enable_worker(struct work_struct *work);
+
+#endif /* OCFS2_LOCALALLOC_H */

fs/ocfs2/locks.c

@@ -0,0 +1,141 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * locks.c
+ *
+ * Userspace file locking support
+ *
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/fcntl.h>
+
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "dlmglue.h"
+#include "file.h"
+#include "inode.h"
+#include "locks.h"
+
+static int ocfs2_do_flock(struct file *file, struct inode *inode,
+			  int cmd, struct file_lock *fl)
+{
+	int ret = 0, level = 0, trylock = 0;
+	struct ocfs2_file_private *fp = file->private_data;
+	struct ocfs2_lock_res *lockres = &fp->fp_flock;
+
+	if (fl->fl_type == F_WRLCK)
+		level = 1;
+	if (!IS_SETLKW(cmd))
+		trylock = 1;
+
+	mutex_lock(&fp->fp_mutex);
+
+	if (lockres->l_flags & OCFS2_LOCK_ATTACHED &&
+	    lockres->l_level > LKM_NLMODE) {
+		int old_level = 0;
+
+		if (lockres->l_level == LKM_EXMODE)
+			old_level = 1;
+
+		if (level == old_level)
+			goto out;
+
+		/*
+		 * Converting an existing lock is not guaranteed to be
+		 * atomic, so we can get away with simply unlocking
+		 * here and allowing the lock code to try at the new
+		 * level.
+		 */
+
+		flock_lock_file_wait(file,
+				     &(struct file_lock){.fl_type = F_UNLCK});
+
+		ocfs2_file_unlock(file);
+	}
+
+	ret = ocfs2_file_lock(file, level, trylock);
+	if (ret) {
+		if (ret == -EAGAIN && trylock)
+			ret = -EWOULDBLOCK;
+		else
+			mlog_errno(ret);
+		goto out;
+	}
+
+	ret = flock_lock_file_wait(file, fl);
+	if (ret)
+		ocfs2_file_unlock(file);
+
+out:
+	mutex_unlock(&fp->fp_mutex);
+
+	return ret;
+}
+
+static int ocfs2_do_funlock(struct file *file, int cmd, struct file_lock *fl)
+{
+	int ret;
+	struct ocfs2_file_private *fp = file->private_data;
+
+	mutex_lock(&fp->fp_mutex);
+	ocfs2_file_unlock(file);
+	ret = flock_lock_file_wait(file, fl);
+	mutex_unlock(&fp->fp_mutex);
+
+	return ret;
+}
+
+/*
+ * Overall flow of ocfs2_flock() was influenced by gfs2_flock().
+ */
+int ocfs2_flock(struct file *file, int cmd, struct file_lock *fl)
+{
+	struct inode *inode = file->f_mapping->host;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+	if (!(fl->fl_flags & FL_FLOCK))
+		return -ENOLCK;
+	if (__mandatory_lock(inode))
+		return -ENOLCK;
+
+	if ((osb->s_mount_opt & OCFS2_MOUNT_LOCALFLOCKS) ||
+	    ocfs2_mount_local(osb))
+		return flock_lock_file_wait(file, fl);
+
+	if (fl->fl_type == F_UNLCK)
+		return ocfs2_do_funlock(file, cmd, fl);
+	else
+		return ocfs2_do_flock(file, inode, cmd, fl);
+}
+
+int ocfs2_lock(struct file *file, int cmd, struct file_lock *fl)
+{
+	struct inode *inode = file->f_mapping->host;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+	if (!(fl->fl_flags & FL_POSIX))
+		return -ENOLCK;
+	if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
+		return -ENOLCK;
+
+	return ocfs2_plock(osb->cconn, OCFS2_I(inode)->ip_blkno, file, cmd, fl);
+}

fs/ocfs2/locks.h

@@ -0,0 +1,32 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * locks.h
+ *
+ * Function prototypes for Userspace file locking support
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_LOCKS_H
+#define OCFS2_LOCKS_H
+
+int ocfs2_flock(struct file *file, int cmd, struct file_lock *fl);
+int ocfs2_lock(struct file *file, int cmd, struct file_lock *fl);
+
+#endif /* OCFS2_LOCKS_H */

fs/ocfs2/mmap.c

@@ -0,0 +1,194 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * mmap.c
+ *
+ * Code to deal with the mess that is clustered mmap.
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/highmem.h>
+#include <linux/pagemap.h>
+#include <linux/uio.h>
+#include <linux/signal.h>
+#include <linux/rbtree.h>
+
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "aops.h"
+#include "dlmglue.h"
+#include "file.h"
+#include "inode.h"
+#include "mmap.h"
+#include "super.h"
+#include "ocfs2_trace.h"
+
+
+static int ocfs2_fault(struct vm_area_struct *area, struct vm_fault *vmf)
+{
+	sigset_t oldset;
+	int ret;
+
+	ocfs2_block_signals(&oldset);
+	ret = filemap_fault(area, vmf);
+	ocfs2_unblock_signals(&oldset);
+
+	trace_ocfs2_fault(OCFS2_I(area->vm_file->f_mapping->host)->ip_blkno,
+			  area, vmf->page, vmf->pgoff);
+	return ret;
+}
+
+static int __ocfs2_page_mkwrite(struct file *file, struct buffer_head *di_bh,
+				struct page *page)
+{
+	int ret = VM_FAULT_NOPAGE;
+	struct inode *inode = file_inode(file);
+	struct address_space *mapping = inode->i_mapping;
+	loff_t pos = page_offset(page);
+	unsigned int len = PAGE_CACHE_SIZE;
+	pgoff_t last_index;
+	struct page *locked_page = NULL;
+	void *fsdata;
+	loff_t size = i_size_read(inode);
+
+	last_index = (size - 1) >> PAGE_CACHE_SHIFT;
+
+	/*
+	 * There are cases that lead to the page no longer bebongs to the
+	 * mapping.
+	 * 1) pagecache truncates locally due to memory pressure.
+	 * 2) pagecache truncates when another is taking EX lock against 
+	 * inode lock. see ocfs2_data_convert_worker.
+	 * 
+	 * The i_size check doesn't catch the case where nodes truncated and
+	 * then re-extended the file. We'll re-check the page mapping after
+	 * taking the page lock inside of ocfs2_write_begin_nolock().
+	 *
+	 * Let VM retry with these cases.
+	 */
+	if ((page->mapping != inode->i_mapping) ||
+	    (!PageUptodate(page)) ||
+	    (page_offset(page) >= size))
+		goto out;
+
+	/*
+	 * Call ocfs2_write_begin() and ocfs2_write_end() to take
+	 * advantage of the allocation code there. We pass a write
+	 * length of the whole page (chopped to i_size) to make sure
+	 * the whole thing is allocated.
+	 *
+	 * Since we know the page is up to date, we don't have to
+	 * worry about ocfs2_write_begin() skipping some buffer reads
+	 * because the "write" would invalidate their data.
+	 */
+	if (page->index == last_index)
+		len = ((size - 1) & ~PAGE_CACHE_MASK) + 1;
+
+	ret = ocfs2_write_begin_nolock(file, mapping, pos, len, 0, &locked_page,
+				       &fsdata, di_bh, page);
+	if (ret) {
+		if (ret != -ENOSPC)
+			mlog_errno(ret);
+		if (ret == -ENOMEM)
+			ret = VM_FAULT_OOM;
+		else
+			ret = VM_FAULT_SIGBUS;
+		goto out;
+	}
+
+	if (!locked_page) {
+		ret = VM_FAULT_NOPAGE;
+		goto out;
+	}
+	ret = ocfs2_write_end_nolock(mapping, pos, len, len, locked_page,
+				     fsdata);
+	BUG_ON(ret != len);
+	ret = VM_FAULT_LOCKED;
+out:
+	return ret;
+}
+
+static int ocfs2_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
+{
+	struct page *page = vmf->page;
+	struct inode *inode = file_inode(vma->vm_file);
+	struct buffer_head *di_bh = NULL;
+	sigset_t oldset;
+	int ret;
+
+	sb_start_pagefault(inode->i_sb);
+	ocfs2_block_signals(&oldset);
+
+	/*
+	 * The cluster locks taken will block a truncate from another
+	 * node. Taking the data lock will also ensure that we don't
+	 * attempt page truncation as part of a downconvert.
+	 */
+	ret = ocfs2_inode_lock(inode, &di_bh, 1);
+	if (ret < 0) {
+		mlog_errno(ret);
+		goto out;
+	}
+
+	/*
+	 * The alloc sem should be enough to serialize with
+	 * ocfs2_truncate_file() changing i_size as well as any thread
+	 * modifying the inode btree.
+	 */
+	down_write(&OCFS2_I(inode)->ip_alloc_sem);
+
+	ret = __ocfs2_page_mkwrite(vma->vm_file, di_bh, page);
+
+	up_write(&OCFS2_I(inode)->ip_alloc_sem);
+
+	brelse(di_bh);
+	ocfs2_inode_unlock(inode, 1);
+
+out:
+	ocfs2_unblock_signals(&oldset);
+	sb_end_pagefault(inode->i_sb);
+	return ret;
+}
+
+static const struct vm_operations_struct ocfs2_file_vm_ops = {
+	.fault		= ocfs2_fault,
+	.page_mkwrite	= ocfs2_page_mkwrite,
+	.remap_pages	= generic_file_remap_pages,
+};
+
+int ocfs2_mmap(struct file *file, struct vm_area_struct *vma)
+{
+	int ret = 0, lock_level = 0;
+
+	ret = ocfs2_inode_lock_atime(file_inode(file),
+				    file->f_path.mnt, &lock_level);
+	if (ret < 0) {
+		mlog_errno(ret);
+		goto out;
+	}
+	ocfs2_inode_unlock(file_inode(file), lock_level);
+out:
+	vma->vm_ops = &ocfs2_file_vm_ops;
+	return 0;
+}
+

fs/ocfs2/mmap.h

@@ -0,0 +1,6 @@
+#ifndef OCFS2_MMAP_H
+#define OCFS2_MMAP_H
+
+int ocfs2_mmap(struct file *file, struct vm_area_struct *vma);
+
+#endif  /* OCFS2_MMAP_H */

fs/ocfs2/move_extents.h

@@ -0,0 +1,22 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * move_extents.h
+ *
+ * Copyright (C) 2011 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+#ifndef OCFS2_MOVE_EXTENTS_H
+#define OCFS2_MOVE_EXTENTS_H
+
+int ocfs2_ioctl_move_extents(struct file *filp,  void __user *argp);
+
+#endif /* OCFS2_MOVE_EXTENTS_H */

fs/ocfs2/namei.h

@@ -0,0 +1,45 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * namei.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_NAMEI_H
+#define OCFS2_NAMEI_H
+
+extern const struct inode_operations ocfs2_dir_iops;
+
+struct dentry *ocfs2_get_parent(struct dentry *child);
+
+int ocfs2_orphan_del(struct ocfs2_super *osb,
+		     handle_t *handle,
+		     struct inode *orphan_dir_inode,
+		     struct inode *inode,
+		     struct buffer_head *orphan_dir_bh);
+int ocfs2_create_inode_in_orphan(struct inode *dir,
+				 int mode,
+				 struct inode **new_inode);
+int ocfs2_mv_orphaned_inode_to_new(struct inode *dir,
+				   struct inode *new_inode,
+				   struct dentry *new_dentry);
+
+#endif /* OCFS2_NAMEI_H */

fs/ocfs2/ocfs1_fs_compat.h

@@ -0,0 +1,109 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ocfs1_fs_compat.h
+ *
+ * OCFS1 volume header definitions.  OCFS2 creates valid but unmountable
+ * OCFS1 volume headers on the first two sectors of an OCFS2 volume.
+ * This allows an OCFS1 volume to see the partition and cleanly fail to
+ * mount it.
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License, version 2,  as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef _OCFS1_FS_COMPAT_H
+#define _OCFS1_FS_COMPAT_H
+
+#define OCFS1_MAX_VOL_SIGNATURE_LEN          128
+#define OCFS1_MAX_MOUNT_POINT_LEN            128
+#define OCFS1_MAX_VOL_ID_LENGTH               16
+#define OCFS1_MAX_VOL_LABEL_LEN               64
+#define OCFS1_MAX_CLUSTER_NAME_LEN            64
+
+#define OCFS1_MAJOR_VERSION              (2)
+#define OCFS1_MINOR_VERSION              (0)
+#define OCFS1_VOLUME_SIGNATURE		 "OracleCFS"
+
+/*
+ * OCFS1 superblock.  Lives at sector 0.
+ */
+struct ocfs1_vol_disk_hdr
+{
+/*00*/	__u32 minor_version;
+	__u32 major_version;
+/*08*/	__u8 signature[OCFS1_MAX_VOL_SIGNATURE_LEN];
+/*88*/	__u8 mount_point[OCFS1_MAX_MOUNT_POINT_LEN];
+/*108*/	__u64 serial_num;
+/*110*/	__u64 device_size;
+	__u64 start_off;
+/*120*/	__u64 bitmap_off;
+	__u64 publ_off;
+/*130*/	__u64 vote_off;
+	__u64 root_bitmap_off;
+/*140*/	__u64 data_start_off;
+	__u64 root_bitmap_size;
+/*150*/	__u64 root_off;
+	__u64 root_size;
+/*160*/	__u64 cluster_size;
+	__u64 num_nodes;
+/*170*/	__u64 num_clusters;
+	__u64 dir_node_size;
+/*180*/	__u64 file_node_size;
+	__u64 internal_off;
+/*190*/	__u64 node_cfg_off;
+	__u64 node_cfg_size;
+/*1A0*/	__u64 new_cfg_off;
+	__u32 prot_bits;
+	__s32 excl_mount;
+/*1B0*/
+};
+
+
+struct ocfs1_disk_lock
+{
+/*00*/	__u32 curr_master;
+	__u8 file_lock;
+	__u8 compat_pad[3];  /* Not in original definition.  Used to
+				make the already existing alignment
+				explicit */
+	__u64 last_write_time;
+/*10*/	__u64 last_read_time;
+	__u32 writer_node_num;
+	__u32 reader_node_num;
+/*20*/	__u64 oin_node_map;
+	__u64 dlock_seq_num;
+/*30*/
+};
+
+/*
+ * OCFS1 volume label.  Lives at sector 1.
+ */
+struct ocfs1_vol_label
+{
+/*00*/	struct ocfs1_disk_lock disk_lock;
+/*30*/	__u8 label[OCFS1_MAX_VOL_LABEL_LEN];
+/*70*/	__u16 label_len;
+/*72*/	__u8 vol_id[OCFS1_MAX_VOL_ID_LENGTH];
+/*82*/	__u16 vol_id_len;
+/*84*/	__u8 cluster_name[OCFS1_MAX_CLUSTER_NAME_LEN];
+/*A4*/	__u16 cluster_name_len;
+/*A6*/
+};
+
+
+#endif /* _OCFS1_FS_COMPAT_H */
+

fs/ocfs2/ocfs2.h

@@ -0,0 +1,888 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ocfs2.h
+ *
+ * Defines macros and structures used in OCFS2
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_H
+#define OCFS2_H
+
+#include <linux/spinlock.h>
+#include <linux/sched.h>
+#include <linux/wait.h>
+#include <linux/list.h>
+#include <linux/llist.h>
+#include <linux/rbtree.h>
+#include <linux/workqueue.h>
+#include <linux/kref.h>
+#include <linux/mutex.h>
+#include <linux/lockdep.h>
+#include <linux/jbd2.h>
+
+/* For union ocfs2_dlm_lksb */
+#include "stackglue.h"
+
+#include "ocfs2_fs.h"
+#include "ocfs2_lockid.h"
+#include "ocfs2_ioctl.h"
+
+/* For struct ocfs2_blockcheck_stats */
+#include "blockcheck.h"
+
+#include "reservations.h"
+
+/* Caching of metadata buffers */
+
+/* Most user visible OCFS2 inodes will have very few pieces of
+ * metadata, but larger files (including bitmaps, etc) must be taken
+ * into account when designing an access scheme. We allow a small
+ * amount of inlined blocks to be stored on an array and grow the
+ * structure into a rb tree when necessary. */
+#define OCFS2_CACHE_INFO_MAX_ARRAY 2
+
+/* Flags for ocfs2_caching_info */
+
+enum ocfs2_caching_info_flags {
+	/* Indicates that the metadata cache is using the inline array */
+	OCFS2_CACHE_FL_INLINE	= 1<<1,
+};
+
+struct ocfs2_caching_operations;
+struct ocfs2_caching_info {
+	/*
+	 * The parent structure provides the locks, but because the
+	 * parent structure can differ, it provides locking operations
+	 * to struct ocfs2_caching_info.
+	 */
+	const struct ocfs2_caching_operations *ci_ops;
+
+	/* next two are protected by trans_inc_lock */
+	/* which transaction were we created on? Zero if none. */
+	unsigned long		ci_created_trans;
+	/* last transaction we were a part of. */
+	unsigned long		ci_last_trans;
+
+	/* Cache structures */
+	unsigned int		ci_flags;
+	unsigned int		ci_num_cached;
+	union {
+	sector_t	ci_array[OCFS2_CACHE_INFO_MAX_ARRAY];
+		struct rb_root	ci_tree;
+	} ci_cache;
+};
+/*
+ * Need this prototype here instead of in uptodate.h because journal.h
+ * uses it.
+ */
+struct super_block *ocfs2_metadata_cache_get_super(struct ocfs2_caching_info *ci);
+
+/* this limits us to 256 nodes
+ * if we need more, we can do a kmalloc for the map */
+#define OCFS2_NODE_MAP_MAX_NODES    256
+struct ocfs2_node_map {
+	u16 num_nodes;
+	unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)];
+};
+
+enum ocfs2_ast_action {
+	OCFS2_AST_INVALID = 0,
+	OCFS2_AST_ATTACH,
+	OCFS2_AST_CONVERT,
+	OCFS2_AST_DOWNCONVERT,
+};
+
+/* actions for an unlockast function to take. */
+enum ocfs2_unlock_action {
+	OCFS2_UNLOCK_INVALID = 0,
+	OCFS2_UNLOCK_CANCEL_CONVERT,
+	OCFS2_UNLOCK_DROP_LOCK,
+};
+
+/* ocfs2_lock_res->l_flags flags. */
+#define OCFS2_LOCK_ATTACHED      (0x00000001) /* we have initialized
+					       * the lvb */
+#define OCFS2_LOCK_BUSY          (0x00000002) /* we are currently in
+					       * dlm_lock */
+#define OCFS2_LOCK_BLOCKED       (0x00000004) /* blocked waiting to
+					       * downconvert*/
+#define OCFS2_LOCK_LOCAL         (0x00000008) /* newly created inode */
+#define OCFS2_LOCK_NEEDS_REFRESH (0x00000010)
+#define OCFS2_LOCK_REFRESHING    (0x00000020)
+#define OCFS2_LOCK_INITIALIZED   (0x00000040) /* track initialization
+					       * for shutdown paths */
+#define OCFS2_LOCK_FREEING       (0x00000080) /* help dlmglue track
+					       * when to skip queueing
+					       * a lock because it's
+					       * about to be
+					       * dropped. */
+#define OCFS2_LOCK_QUEUED        (0x00000100) /* queued for downconvert */
+#define OCFS2_LOCK_NOCACHE       (0x00000200) /* don't use a holder count */
+#define OCFS2_LOCK_PENDING       (0x00000400) /* This lockres is pending a
+						 call to dlm_lock.  Only
+						 exists with BUSY set. */
+#define OCFS2_LOCK_UPCONVERT_FINISHING (0x00000800) /* blocks the dc thread
+						     * from downconverting
+						     * before the upconvert
+						     * has completed */
+
+struct ocfs2_lock_res_ops;
+
+typedef void (*ocfs2_lock_callback)(int status, unsigned long data);
+
+#ifdef CONFIG_OCFS2_FS_STATS
+struct ocfs2_lock_stats {
+	u64		ls_total;	/* Total wait in NSEC */
+	u32		ls_gets;	/* Num acquires */
+	u32		ls_fail;	/* Num failed acquires */
+
+	/* Storing max wait in usecs saves 24 bytes per inode */
+	u32		ls_max;		/* Max wait in USEC */
+};
+#endif
+
+struct ocfs2_lock_res {
+	void                    *l_priv;
+	struct ocfs2_lock_res_ops *l_ops;
+
+
+	struct list_head         l_blocked_list;
+	struct list_head         l_mask_waiters;
+
+	unsigned long		 l_flags;
+	char                     l_name[OCFS2_LOCK_ID_MAX_LEN];
+	unsigned int             l_ro_holders;
+	unsigned int             l_ex_holders;
+	signed char		 l_level;
+	signed char		 l_requested;
+	signed char		 l_blocking;
+
+	/* Data packed - type enum ocfs2_lock_type */
+	unsigned char            l_type;
+
+	/* used from AST/BAST funcs. */
+	/* Data packed - enum type ocfs2_ast_action */
+	unsigned char            l_action;
+	/* Data packed - enum type ocfs2_unlock_action */
+	unsigned char            l_unlock_action;
+	unsigned int             l_pending_gen;
+
+	spinlock_t               l_lock;
+
+	struct ocfs2_dlm_lksb    l_lksb;
+
+	wait_queue_head_t        l_event;
+
+	struct list_head         l_debug_list;
+
+#ifdef CONFIG_OCFS2_FS_STATS
+	struct ocfs2_lock_stats  l_lock_prmode;		/* PR mode stats */
+	u32                      l_lock_refresh;	/* Disk refreshes */
+	struct ocfs2_lock_stats  l_lock_exmode;		/* EX mode stats */
+#endif
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+	struct lockdep_map	 l_lockdep_map;
+#endif
+};
+
+enum ocfs2_orphan_scan_state {
+	ORPHAN_SCAN_ACTIVE,
+	ORPHAN_SCAN_INACTIVE
+};
+
+struct ocfs2_orphan_scan {
+	struct mutex 		os_lock;
+	struct ocfs2_super 	*os_osb;
+	struct ocfs2_lock_res 	os_lockres;     /* lock to synchronize scans */
+	struct delayed_work 	os_orphan_scan_work;
+	struct timespec		os_scantime;  /* time this node ran the scan */
+	u32			os_count;      /* tracks node specific scans */
+	u32  			os_seqno;       /* tracks cluster wide scans */
+	atomic_t		os_state;              /* ACTIVE or INACTIVE */
+};
+
+struct ocfs2_dlm_debug {
+	struct kref d_refcnt;
+	struct dentry *d_locking_state;
+	struct list_head d_lockres_tracking;
+};
+
+enum ocfs2_vol_state
+{
+	VOLUME_INIT = 0,
+	VOLUME_MOUNTED,
+	VOLUME_MOUNTED_QUOTAS,
+	VOLUME_DISMOUNTED,
+	VOLUME_DISABLED
+};
+
+struct ocfs2_alloc_stats
+{
+	atomic_t moves;
+	atomic_t local_data;
+	atomic_t bitmap_data;
+	atomic_t bg_allocs;
+	atomic_t bg_extends;
+};
+
+enum ocfs2_local_alloc_state
+{
+	OCFS2_LA_UNUSED = 0,	/* Local alloc will never be used for
+				 * this mountpoint. */
+	OCFS2_LA_ENABLED,	/* Local alloc is in use. */
+	OCFS2_LA_THROTTLED,	/* Local alloc is in use, but number
+				 * of bits has been reduced. */
+	OCFS2_LA_DISABLED	/* Local alloc has temporarily been
+				 * disabled. */
+};
+
+enum ocfs2_mount_options
+{
+	OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Local heartbeat */
+	OCFS2_MOUNT_BARRIER = 1 << 1,	/* Use block barriers */
+	OCFS2_MOUNT_NOINTR  = 1 << 2,   /* Don't catch signals */
+	OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */
+	OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */
+	OCFS2_MOUNT_LOCALFLOCKS = 1 << 5, /* No cluster aware user file locks */
+	OCFS2_MOUNT_NOUSERXATTR = 1 << 6, /* No user xattr */
+	OCFS2_MOUNT_INODE64 = 1 << 7,	/* Allow inode numbers > 2^32 */
+	OCFS2_MOUNT_POSIX_ACL = 1 << 8,	/* Force POSIX access control lists */
+	OCFS2_MOUNT_NO_POSIX_ACL = 1 << 9,	/* Disable POSIX access
+						   control lists */
+	OCFS2_MOUNT_USRQUOTA = 1 << 10, /* We support user quotas */
+	OCFS2_MOUNT_GRPQUOTA = 1 << 11, /* We support group quotas */
+	OCFS2_MOUNT_COHERENCY_BUFFERED = 1 << 12, /* Allow concurrent O_DIRECT
+						     writes */
+	OCFS2_MOUNT_HB_NONE = 1 << 13, /* No heartbeat */
+	OCFS2_MOUNT_HB_GLOBAL = 1 << 14, /* Global heartbeat */
+};
+
+#define OCFS2_OSB_SOFT_RO	0x0001
+#define OCFS2_OSB_HARD_RO	0x0002
+#define OCFS2_OSB_ERROR_FS	0x0004
+#define OCFS2_DEFAULT_ATIME_QUANTUM	60
+
+struct ocfs2_journal;
+struct ocfs2_slot_info;
+struct ocfs2_recovery_map;
+struct ocfs2_replay_map;
+struct ocfs2_quota_recovery;
+struct ocfs2_super
+{
+	struct task_struct *commit_task;
+	struct super_block *sb;
+	struct inode *root_inode;
+	struct inode *sys_root_inode;
+	struct inode *global_system_inodes[NUM_GLOBAL_SYSTEM_INODES];
+	struct inode **local_system_inodes;
+
+	struct ocfs2_slot_info *slot_info;
+
+	u32 *slot_recovery_generations;
+
+	spinlock_t node_map_lock;
+
+	u64 root_blkno;
+	u64 system_dir_blkno;
+	u64 bitmap_blkno;
+	u32 bitmap_cpg;
+	u8 *uuid;
+	char *uuid_str;
+	u32 uuid_hash;
+	u8 *vol_label;
+	u64 first_cluster_group_blkno;
+	u32 fs_generation;
+
+	u32 s_feature_compat;
+	u32 s_feature_incompat;
+	u32 s_feature_ro_compat;
+
+	/* Protects s_next_generation, osb_flags and s_inode_steal_slot.
+	 * Could protect more on osb as it's very short lived.
+	 */
+	spinlock_t osb_lock;
+	u32 s_next_generation;
+	unsigned long osb_flags;
+	s16 s_inode_steal_slot;
+	s16 s_meta_steal_slot;
+	atomic_t s_num_inodes_stolen;
+	atomic_t s_num_meta_stolen;
+
+	unsigned long s_mount_opt;
+	unsigned int s_atime_quantum;
+
+	unsigned int max_slots;
+	unsigned int node_num;
+	int slot_num;
+	int preferred_slot;
+	int s_sectsize_bits;
+	int s_clustersize;
+	int s_clustersize_bits;
+	unsigned int s_xattr_inline_size;
+
+	atomic_t vol_state;
+	struct mutex recovery_lock;
+	struct ocfs2_recovery_map *recovery_map;
+	struct ocfs2_replay_map *replay_map;
+	struct task_struct *recovery_thread_task;
+	int disable_recovery;
+	wait_queue_head_t checkpoint_event;
+	struct ocfs2_journal *journal;
+	unsigned long osb_commit_interval;
+
+	struct delayed_work		la_enable_wq;
+
+	/*
+	 * Must hold local alloc i_mutex and osb->osb_lock to change
+	 * local_alloc_bits. Reads can be done under either lock.
+	 */
+	unsigned int local_alloc_bits;
+	unsigned int local_alloc_default_bits;
+	/* osb_clusters_at_boot can become stale! Do not trust it to
+	 * be up to date. */
+	unsigned int osb_clusters_at_boot;
+
+	enum ocfs2_local_alloc_state local_alloc_state; /* protected
+							 * by osb_lock */
+
+	struct buffer_head *local_alloc_bh;
+
+	u64 la_last_gd;
+
+	struct ocfs2_reservation_map	osb_la_resmap;
+
+	unsigned int	osb_resv_level;
+	unsigned int	osb_dir_resv_level;
+
+	/* Next three fields are for local node slot recovery during
+	 * mount. */
+	int dirty;
+	struct ocfs2_dinode *local_alloc_copy;
+	struct ocfs2_quota_recovery *quota_rec;
+
+	struct ocfs2_blockcheck_stats osb_ecc_stats;
+	struct ocfs2_alloc_stats alloc_stats;
+	char dev_str[20];		/* "major,minor" of the device */
+
+	u8 osb_stackflags;
+
+	char osb_cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
+	char osb_cluster_name[OCFS2_CLUSTER_NAME_LEN + 1];
+	struct ocfs2_cluster_connection *cconn;
+	struct ocfs2_lock_res osb_super_lockres;
+	struct ocfs2_lock_res osb_rename_lockres;
+	struct ocfs2_lock_res osb_nfs_sync_lockres;
+	struct ocfs2_dlm_debug *osb_dlm_debug;
+
+	struct dentry *osb_debug_root;
+	struct dentry *osb_ctxt;
+
+	wait_queue_head_t recovery_event;
+
+	spinlock_t dc_task_lock;
+	struct task_struct *dc_task;
+	wait_queue_head_t dc_event;
+	unsigned long dc_wake_sequence;
+	unsigned long dc_work_sequence;
+
+	/*
+	 * Any thread can add locks to the list, but the downconvert
+	 * thread is the only one allowed to remove locks. Any change
+	 * to this rule requires updating
+	 * ocfs2_downconvert_thread_do_work().
+	 */
+	struct list_head blocked_lock_list;
+	unsigned long blocked_lock_count;
+
+	/* List of dquot structures to drop last reference to */
+	struct llist_head dquot_drop_list;
+	struct work_struct dquot_drop_work;
+
+	wait_queue_head_t		osb_mount_event;
+
+	/* Truncate log info */
+	struct inode			*osb_tl_inode;
+	struct buffer_head		*osb_tl_bh;
+	struct delayed_work		osb_truncate_log_wq;
+	atomic_t			osb_tl_disable;
+	/*
+	 * How many clusters in our truncate log.
+	 * It must be protected by osb_tl_inode->i_mutex.
+	 */
+	unsigned int truncated_clusters;
+
+	struct ocfs2_node_map		osb_recovering_orphan_dirs;
+	unsigned int			*osb_orphan_wipes;
+	wait_queue_head_t		osb_wipe_event;
+
+	struct ocfs2_orphan_scan	osb_orphan_scan;
+
+	/* used to protect metaecc calculation check of xattr. */
+	spinlock_t osb_xattr_lock;
+
+	unsigned int			osb_dx_mask;
+	u32				osb_dx_seed[4];
+
+	/* the group we used to allocate inodes. */
+	u64				osb_inode_alloc_group;
+
+	/* rb tree root for refcount lock. */
+	struct rb_root	osb_rf_lock_tree;
+	struct ocfs2_refcount_tree *osb_ref_tree_lru;
+
+	struct mutex system_file_mutex;
+};
+
+#define OCFS2_SB(sb)	    ((struct ocfs2_super *)(sb)->s_fs_info)
+
+/* Useful typedef for passing around journal access functions */
+typedef int (*ocfs2_journal_access_func)(handle_t *handle,
+					 struct ocfs2_caching_info *ci,
+					 struct buffer_head *bh, int type);
+
+static inline int ocfs2_should_order_data(struct inode *inode)
+{
+	if (!S_ISREG(inode->i_mode))
+		return 0;
+	if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)
+		return 0;
+	return 1;
+}
+
+static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb)
+{
+	if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC)
+		return 1;
+	return 0;
+}
+
+static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb)
+{
+	/*
+	 * Support for sparse files is a pre-requisite
+	 */
+	if (!ocfs2_sparse_alloc(osb))
+		return 0;
+
+	if (osb->s_feature_ro_compat & OCFS2_FEATURE_RO_COMPAT_UNWRITTEN)
+		return 1;
+	return 0;
+}
+
+static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb)
+{
+	if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA)
+		return 1;
+	return 0;
+}
+
+static inline int ocfs2_supports_xattr(struct ocfs2_super *osb)
+{
+	if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR)
+		return 1;
+	return 0;
+}
+
+static inline int ocfs2_meta_ecc(struct ocfs2_super *osb)
+{
+	if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_META_ECC)
+		return 1;
+	return 0;
+}
+
+static inline int ocfs2_supports_indexed_dirs(struct ocfs2_super *osb)
+{
+	if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INDEXED_DIRS)
+		return 1;
+	return 0;
+}
+
+static inline int ocfs2_supports_discontig_bg(struct ocfs2_super *osb)
+{
+	if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG)
+		return 1;
+	return 0;
+}
+
+static inline unsigned int ocfs2_link_max(struct ocfs2_super *osb)
+{
+	if (ocfs2_supports_indexed_dirs(osb))
+		return OCFS2_DX_LINK_MAX;
+	return OCFS2_LINK_MAX;
+}
+
+static inline unsigned int ocfs2_read_links_count(struct ocfs2_dinode *di)
+{
+	u32 nlink = le16_to_cpu(di->i_links_count);
+	u32 hi = le16_to_cpu(di->i_links_count_hi);
+
+	if (di->i_dyn_features & cpu_to_le16(OCFS2_INDEXED_DIR_FL))
+		nlink |= (hi << OCFS2_LINKS_HI_SHIFT);
+
+	return nlink;
+}
+
+static inline void ocfs2_set_links_count(struct ocfs2_dinode *di, u32 nlink)
+{
+	u16 lo, hi;
+
+	lo = nlink;
+	hi = nlink >> OCFS2_LINKS_HI_SHIFT;
+
+	di->i_links_count = cpu_to_le16(lo);
+	di->i_links_count_hi = cpu_to_le16(hi);
+}
+
+static inline void ocfs2_add_links_count(struct ocfs2_dinode *di, int n)
+{
+	u32 links = ocfs2_read_links_count(di);
+
+	links += n;
+
+	ocfs2_set_links_count(di, links);
+}
+
+static inline int ocfs2_refcount_tree(struct ocfs2_super *osb)
+{
+	if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE)
+		return 1;
+	return 0;
+}
+
+/* set / clear functions because cluster events can make these happen
+ * in parallel so we want the transitions to be atomic. this also
+ * means that any future flags osb_flags must be protected by spinlock
+ * too! */
+static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
+				      unsigned long flag)
+{
+	spin_lock(&osb->osb_lock);
+	osb->osb_flags |= flag;
+	spin_unlock(&osb->osb_lock);
+}
+
+static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
+				     int hard)
+{
+	spin_lock(&osb->osb_lock);
+	osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
+	if (hard)
+		osb->osb_flags |= OCFS2_OSB_HARD_RO;
+	else
+		osb->osb_flags |= OCFS2_OSB_SOFT_RO;
+	spin_unlock(&osb->osb_lock);
+}
+
+static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
+{
+	int ret;
+
+	spin_lock(&osb->osb_lock);
+	ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
+	spin_unlock(&osb->osb_lock);
+
+	return ret;
+}
+
+static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
+{
+	int ret;
+
+	spin_lock(&osb->osb_lock);
+	ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
+	spin_unlock(&osb->osb_lock);
+
+	return ret;
+}
+
+static inline int ocfs2_clusterinfo_valid(struct ocfs2_super *osb)
+{
+	return (osb->s_feature_incompat &
+		(OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK |
+		 OCFS2_FEATURE_INCOMPAT_CLUSTERINFO));
+}
+
+static inline int ocfs2_userspace_stack(struct ocfs2_super *osb)
+{
+	if (ocfs2_clusterinfo_valid(osb) &&
+	    memcmp(osb->osb_cluster_stack, OCFS2_CLASSIC_CLUSTER_STACK,
+		   OCFS2_STACK_LABEL_LEN))
+		return 1;
+	return 0;
+}
+
+static inline int ocfs2_o2cb_stack(struct ocfs2_super *osb)
+{
+	if (ocfs2_clusterinfo_valid(osb) &&
+	    !memcmp(osb->osb_cluster_stack, OCFS2_CLASSIC_CLUSTER_STACK,
+		   OCFS2_STACK_LABEL_LEN))
+		return 1;
+	return 0;
+}
+
+static inline int ocfs2_cluster_o2cb_global_heartbeat(struct ocfs2_super *osb)
+{
+	return ocfs2_o2cb_stack(osb) &&
+		(osb->osb_stackflags & OCFS2_CLUSTER_O2CB_GLOBAL_HEARTBEAT);
+}
+
+static inline int ocfs2_mount_local(struct ocfs2_super *osb)
+{
+	return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT);
+}
+
+static inline int ocfs2_uses_extended_slot_map(struct ocfs2_super *osb)
+{
+	return (osb->s_feature_incompat &
+		OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP);
+}
+
+
+#define OCFS2_IS_VALID_DINODE(ptr)					\
+	(!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
+
+#define OCFS2_IS_VALID_EXTENT_BLOCK(ptr)				\
+	(!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
+
+#define OCFS2_IS_VALID_GROUP_DESC(ptr)					\
+	(!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
+
+
+#define OCFS2_IS_VALID_XATTR_BLOCK(ptr)					\
+	(!strcmp((ptr)->xb_signature, OCFS2_XATTR_BLOCK_SIGNATURE))
+
+#define OCFS2_IS_VALID_DIR_TRAILER(ptr)					\
+	(!strcmp((ptr)->db_signature, OCFS2_DIR_TRAILER_SIGNATURE))
+
+#define OCFS2_IS_VALID_DX_ROOT(ptr)					\
+	(!strcmp((ptr)->dr_signature, OCFS2_DX_ROOT_SIGNATURE))
+
+#define OCFS2_IS_VALID_DX_LEAF(ptr)					\
+	(!strcmp((ptr)->dl_signature, OCFS2_DX_LEAF_SIGNATURE))
+
+#define OCFS2_IS_VALID_REFCOUNT_BLOCK(ptr)				\
+	(!strcmp((ptr)->rf_signature, OCFS2_REFCOUNT_BLOCK_SIGNATURE))
+
+static inline unsigned long ino_from_blkno(struct super_block *sb,
+					   u64 blkno)
+{
+	return (unsigned long)(blkno & (u64)ULONG_MAX);
+}
+
+static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
+					   u32 clusters)
+{
+	int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
+		sb->s_blocksize_bits;
+
+	return (u64)clusters << c_to_b_bits;
+}
+
+static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
+					   u64 blocks)
+{
+	int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
+		sb->s_blocksize_bits;
+
+	return (u32)(blocks >> b_to_c_bits);
+}
+
+static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
+						    u64 bytes)
+{
+	int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
+	unsigned int clusters;
+
+	bytes += OCFS2_SB(sb)->s_clustersize - 1;
+	/* OCFS2 just cannot have enough clusters to overflow this */
+	clusters = (unsigned int)(bytes >> cl_bits);
+
+	return clusters;
+}
+
+static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
+					 u64 bytes)
+{
+	bytes += sb->s_blocksize - 1;
+	return bytes >> sb->s_blocksize_bits;
+}
+
+static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
+					  u32 clusters)
+{
+	return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
+}
+
+static inline u64 ocfs2_block_to_cluster_start(struct super_block *sb,
+					       u64 blocks)
+{
+	int bits = OCFS2_SB(sb)->s_clustersize_bits - sb->s_blocksize_bits;
+	unsigned int clusters;
+
+	clusters = ocfs2_blocks_to_clusters(sb, blocks);
+	return (u64)clusters << bits;
+}
+
+static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
+						u64 bytes)
+{
+	int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
+	unsigned int clusters;
+
+	clusters = ocfs2_clusters_for_bytes(sb, bytes);
+	return (u64)clusters << cl_bits;
+}
+
+static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
+					      u64 bytes)
+{
+	u64 blocks;
+
+        blocks = ocfs2_blocks_for_bytes(sb, bytes);
+	return blocks << sb->s_blocksize_bits;
+}
+
+static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
+{
+	return (unsigned long)((bytes + 511) >> 9);
+}
+
+static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb,
+							unsigned long pg_index)
+{
+	u32 clusters = pg_index;
+	unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
+
+	if (unlikely(PAGE_CACHE_SHIFT > cbits))
+		clusters = pg_index << (PAGE_CACHE_SHIFT - cbits);
+	else if (PAGE_CACHE_SHIFT < cbits)
+		clusters = pg_index >> (cbits - PAGE_CACHE_SHIFT);
+
+	return clusters;
+}
+
+/*
+ * Find the 1st page index which covers the given clusters.
+ */
+static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb,
+							u32 clusters)
+{
+	unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
+        pgoff_t index = clusters;
+
+	if (PAGE_CACHE_SHIFT > cbits) {
+		index = (pgoff_t)clusters >> (PAGE_CACHE_SHIFT - cbits);
+	} else if (PAGE_CACHE_SHIFT < cbits) {
+		index = (pgoff_t)clusters << (cbits - PAGE_CACHE_SHIFT);
+	}
+
+	return index;
+}
+
+static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb)
+{
+	unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
+	unsigned int pages_per_cluster = 1;
+
+	if (PAGE_CACHE_SHIFT < cbits)
+		pages_per_cluster = 1 << (cbits - PAGE_CACHE_SHIFT);
+
+	return pages_per_cluster;
+}
+
+static inline unsigned int ocfs2_megabytes_to_clusters(struct super_block *sb,
+						       unsigned int megs)
+{
+	BUILD_BUG_ON(OCFS2_MAX_CLUSTERSIZE > 1048576);
+
+	return megs << (20 - OCFS2_SB(sb)->s_clustersize_bits);
+}
+
+static inline unsigned int ocfs2_clusters_to_megabytes(struct super_block *sb,
+						       unsigned int clusters)
+{
+	return clusters >> (20 - OCFS2_SB(sb)->s_clustersize_bits);
+}
+
+static inline void _ocfs2_set_bit(unsigned int bit, unsigned long *bitmap)
+{
+	__set_bit_le(bit, bitmap);
+}
+#define ocfs2_set_bit(bit, addr) _ocfs2_set_bit((bit), (unsigned long *)(addr))
+
+static inline void _ocfs2_clear_bit(unsigned int bit, unsigned long *bitmap)
+{
+	__clear_bit_le(bit, bitmap);
+}
+#define ocfs2_clear_bit(bit, addr) _ocfs2_clear_bit((bit), (unsigned long *)(addr))
+
+#define ocfs2_test_bit test_bit_le
+#define ocfs2_find_next_zero_bit find_next_zero_bit_le
+#define ocfs2_find_next_bit find_next_bit_le
+
+static inline void *correct_addr_and_bit_unaligned(int *bit, void *addr)
+{
+#if BITS_PER_LONG == 64
+	*bit += ((unsigned long) addr & 7UL) << 3;
+	addr = (void *) ((unsigned long) addr & ~7UL);
+#elif BITS_PER_LONG == 32
+	*bit += ((unsigned long) addr & 3UL) << 3;
+	addr = (void *) ((unsigned long) addr & ~3UL);
+#else
+#error "how many bits you are?!"
+#endif
+	return addr;
+}
+
+static inline void ocfs2_set_bit_unaligned(int bit, void *bitmap)
+{
+	bitmap = correct_addr_and_bit_unaligned(&bit, bitmap);
+	ocfs2_set_bit(bit, bitmap);
+}
+
+static inline void ocfs2_clear_bit_unaligned(int bit, void *bitmap)
+{
+	bitmap = correct_addr_and_bit_unaligned(&bit, bitmap);
+	ocfs2_clear_bit(bit, bitmap);
+}
+
+static inline int ocfs2_test_bit_unaligned(int bit, void *bitmap)
+{
+	bitmap = correct_addr_and_bit_unaligned(&bit, bitmap);
+	return ocfs2_test_bit(bit, bitmap);
+}
+
+static inline int ocfs2_find_next_zero_bit_unaligned(void *bitmap, int max,
+							int start)
+{
+	int fix = 0, ret, tmpmax;
+	bitmap = correct_addr_and_bit_unaligned(&fix, bitmap);
+	tmpmax = max + fix;
+	start += fix;
+
+	ret = ocfs2_find_next_zero_bit(bitmap, tmpmax, start) - fix;
+	if (ret > max)
+		return max;
+	return ret;
+}
+
+#endif  /* OCFS2_H */
+

fs/ocfs2/ocfs2_ioctl.h

@@ -0,0 +1,242 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ocfs2_ioctl.h
+ *
+ * Defines OCFS2 ioctls.
+ *
+ * Copyright (C) 2010 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License, version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#ifndef OCFS2_IOCTL_H
+#define OCFS2_IOCTL_H
+
+/*
+ * ioctl commands
+ */
+#define OCFS2_IOC_GETFLAGS	FS_IOC_GETFLAGS
+#define OCFS2_IOC_SETFLAGS	FS_IOC_SETFLAGS
+#define OCFS2_IOC32_GETFLAGS	FS_IOC32_GETFLAGS
+#define OCFS2_IOC32_SETFLAGS	FS_IOC32_SETFLAGS
+
+/*
+ * Space reservation / allocation / free ioctls and argument structure
+ * are designed to be compatible with XFS.
+ *
+ * ALLOCSP* and FREESP* are not and will never be supported, but are
+ * included here for completeness.
+ */
+struct ocfs2_space_resv {
+	__s16		l_type;
+	__s16		l_whence;
+	__s64		l_start;
+	__s64		l_len;		/* len == 0 means until end of file */
+	__s32		l_sysid;
+	__u32		l_pid;
+	__s32		l_pad[4];	/* reserve area			    */
+};
+
+#define OCFS2_IOC_ALLOCSP		_IOW ('X', 10, struct ocfs2_space_resv)
+#define OCFS2_IOC_FREESP		_IOW ('X', 11, struct ocfs2_space_resv)
+#define OCFS2_IOC_RESVSP		_IOW ('X', 40, struct ocfs2_space_resv)
+#define OCFS2_IOC_UNRESVSP	_IOW ('X', 41, struct ocfs2_space_resv)
+#define OCFS2_IOC_ALLOCSP64	_IOW ('X', 36, struct ocfs2_space_resv)
+#define OCFS2_IOC_FREESP64	_IOW ('X', 37, struct ocfs2_space_resv)
+#define OCFS2_IOC_RESVSP64	_IOW ('X', 42, struct ocfs2_space_resv)
+#define OCFS2_IOC_UNRESVSP64	_IOW ('X', 43, struct ocfs2_space_resv)
+
+/* Used to pass group descriptor data when online resize is done */
+struct ocfs2_new_group_input {
+	__u64 group;		/* Group descriptor's blkno. */
+	__u32 clusters;		/* Total number of clusters in this group */
+	__u32 frees;		/* Total free clusters in this group */
+	__u16 chain;		/* Chain for this group */
+	__u16 reserved1;
+	__u32 reserved2;
+};
+
+#define OCFS2_IOC_GROUP_EXTEND	_IOW('o', 1, int)
+#define OCFS2_IOC_GROUP_ADD	_IOW('o', 2,struct ocfs2_new_group_input)
+#define OCFS2_IOC_GROUP_ADD64	_IOW('o', 3,struct ocfs2_new_group_input)
+
+/* Used to pass 2 file names to reflink. */
+struct reflink_arguments {
+	__u64 old_path;
+	__u64 new_path;
+	__u64 preserve;
+};
+#define OCFS2_IOC_REFLINK	_IOW('o', 4, struct reflink_arguments)
+
+/* Following definitions dedicated for ocfs2_info_request ioctls. */
+#define OCFS2_INFO_MAX_REQUEST		(50)
+#define OCFS2_TEXT_UUID_LEN		(OCFS2_VOL_UUID_LEN * 2)
+
+/* Magic number of all requests */
+#define OCFS2_INFO_MAGIC		(0x4F32494E)
+
+/*
+ * Always try to separate info request into small pieces to
+ * guarantee the backward&forward compatibility.
+ */
+struct ocfs2_info {
+	__u64 oi_requests;	/* Array of __u64 pointers to requests */
+	__u32 oi_count;		/* Number of requests in info_requests */
+	__u32 oi_pad;
+};
+
+struct ocfs2_info_request {
+/*00*/	__u32 ir_magic;	/* Magic number */
+	__u32 ir_code;	/* Info request code */
+	__u32 ir_size;	/* Size of request */
+	__u32 ir_flags;	/* Request flags */
+/*10*/			/* Request specific fields */
+};
+
+struct ocfs2_info_clustersize {
+	struct ocfs2_info_request ic_req;
+	__u32 ic_clustersize;
+	__u32 ic_pad;
+};
+
+struct ocfs2_info_blocksize {
+	struct ocfs2_info_request ib_req;
+	__u32 ib_blocksize;
+	__u32 ib_pad;
+};
+
+struct ocfs2_info_maxslots {
+	struct ocfs2_info_request im_req;
+	__u32 im_max_slots;
+	__u32 im_pad;
+};
+
+struct ocfs2_info_label {
+	struct ocfs2_info_request il_req;
+	__u8	il_label[OCFS2_MAX_VOL_LABEL_LEN];
+} __attribute__ ((packed));
+
+struct ocfs2_info_uuid {
+	struct ocfs2_info_request iu_req;
+	__u8	iu_uuid_str[OCFS2_TEXT_UUID_LEN + 1];
+} __attribute__ ((packed));
+
+struct ocfs2_info_fs_features {
+	struct ocfs2_info_request if_req;
+	__u32 if_compat_features;
+	__u32 if_incompat_features;
+	__u32 if_ro_compat_features;
+	__u32 if_pad;
+};
+
+struct ocfs2_info_journal_size {
+	struct ocfs2_info_request ij_req;
+	__u64 ij_journal_size;
+};
+
+struct ocfs2_info_freeinode {
+	struct ocfs2_info_request ifi_req;
+	struct ocfs2_info_local_freeinode {
+		__u64 lfi_total;
+		__u64 lfi_free;
+	} ifi_stat[OCFS2_MAX_SLOTS];
+	__u32 ifi_slotnum; /* out */
+	__u32 ifi_pad;
+};
+
+#define OCFS2_INFO_MAX_HIST     (32)
+
+struct ocfs2_info_freefrag {
+	struct ocfs2_info_request iff_req;
+	struct ocfs2_info_freefrag_stats { /* (out) */
+		struct ocfs2_info_free_chunk_list {
+			__u32 fc_chunks[OCFS2_INFO_MAX_HIST];
+			__u32 fc_clusters[OCFS2_INFO_MAX_HIST];
+		} ffs_fc_hist;
+		__u32 ffs_clusters;
+		__u32 ffs_free_clusters;
+		__u32 ffs_free_chunks;
+		__u32 ffs_free_chunks_real;
+		__u32 ffs_min; /* Minimum free chunksize in clusters */
+		__u32 ffs_max;
+		__u32 ffs_avg;
+		__u32 ffs_pad;
+	} iff_ffs;
+	__u32 iff_chunksize; /* chunksize in clusters(in) */
+	__u32 iff_pad;
+};
+
+/* Codes for ocfs2_info_request */
+enum ocfs2_info_type {
+	OCFS2_INFO_CLUSTERSIZE = 1,
+	OCFS2_INFO_BLOCKSIZE,
+	OCFS2_INFO_MAXSLOTS,
+	OCFS2_INFO_LABEL,
+	OCFS2_INFO_UUID,
+	OCFS2_INFO_FS_FEATURES,
+	OCFS2_INFO_JOURNAL_SIZE,
+	OCFS2_INFO_FREEINODE,
+	OCFS2_INFO_FREEFRAG,
+	OCFS2_INFO_NUM_TYPES
+};
+
+/* Flags for struct ocfs2_info_request */
+/* Filled by the caller */
+#define OCFS2_INFO_FL_NON_COHERENT	(0x00000001)	/* Cluster coherency not
+							   required. This is a hint.
+							   It is up to ocfs2 whether
+							   the request can be fulfilled
+							   without locking. */
+/* Filled by ocfs2 */
+#define OCFS2_INFO_FL_FILLED		(0x40000000)	/* Filesystem understood
+							   this request and
+							   filled in the answer */
+
+#define OCFS2_INFO_FL_ERROR		(0x80000000)	/* Error happened during
+							   request handling. */
+
+#define OCFS2_IOC_INFO		_IOR('o', 5, struct ocfs2_info)
+
+struct ocfs2_move_extents {
+/* All values are in bytes */
+	/* in */
+	__u64 me_start;		/* Virtual start in the file to move */
+	__u64 me_len;		/* Length of the extents to be moved */
+	__u64 me_goal;		/* Physical offset of the goal,
+				   it's in block unit */
+	__u64 me_threshold;	/* Maximum distance from goal or threshold
+				   for auto defragmentation */
+	__u64 me_flags;		/* Flags for the operation:
+				 * - auto defragmentation.
+				 * - refcount,xattr cases.
+				 */
+	/* out */
+	__u64 me_moved_len;	/* Moved/defraged length */
+	__u64 me_new_offset;	/* Resulting physical location */
+	__u32 me_reserved[2];	/* Reserved for futhure */
+};
+
+#define OCFS2_MOVE_EXT_FL_AUTO_DEFRAG	(0x00000001)	/* Kernel manages to
+							   claim new clusters
+							   as the goal place
+							   for extents moving */
+#define OCFS2_MOVE_EXT_FL_PART_DEFRAG	(0x00000002)	/* Allow partial extent
+							   moving, is to make
+							   movement less likely
+							   to fail, may make fs
+							   even more fragmented */
+#define OCFS2_MOVE_EXT_FL_COMPLETE	(0x00000004)	/* Move or defragmenation
+							   completely gets done.
+							 */
+
+#define OCFS2_IOC_MOVE_EXT	_IOW('o', 6, struct ocfs2_move_extents)
+
+#endif /* OCFS2_IOCTL_H */

fs/ocfs2/ocfs2_lockid.h

@@ -0,0 +1,128 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ocfs2_lockid.h
+ *
+ * Defines OCFS2 lockid bits.
+ *
+ * Copyright (C) 2002, 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_LOCKID_H
+#define OCFS2_LOCKID_H
+
+/* lock ids are made up in the following manner:
+ * name[0]     --> type
+ * name[1-6]   --> 6 pad characters, reserved for now
+ * name[7-22]  --> block number, expressed in hex as 16 chars
+ * name[23-30] --> i_generation, expressed in hex 8 chars
+ * name[31]    --> '\0' */
+#define OCFS2_LOCK_ID_MAX_LEN  32
+#define OCFS2_LOCK_ID_PAD "000000"
+
+#define OCFS2_DENTRY_LOCK_INO_START 18
+
+enum ocfs2_lock_type {
+	OCFS2_LOCK_TYPE_META = 0,
+	OCFS2_LOCK_TYPE_DATA,
+	OCFS2_LOCK_TYPE_SUPER,
+	OCFS2_LOCK_TYPE_RENAME,
+	OCFS2_LOCK_TYPE_RW,
+	OCFS2_LOCK_TYPE_DENTRY,
+	OCFS2_LOCK_TYPE_OPEN,
+	OCFS2_LOCK_TYPE_FLOCK,
+	OCFS2_LOCK_TYPE_QINFO,
+	OCFS2_LOCK_TYPE_NFS_SYNC,
+	OCFS2_LOCK_TYPE_ORPHAN_SCAN,
+	OCFS2_LOCK_TYPE_REFCOUNT,
+	OCFS2_NUM_LOCK_TYPES
+};
+
+static inline char ocfs2_lock_type_char(enum ocfs2_lock_type type)
+{
+	char c;
+	switch (type) {
+		case OCFS2_LOCK_TYPE_META:
+			c = 'M';
+			break;
+		case OCFS2_LOCK_TYPE_DATA:
+			c = 'D';
+			break;
+		case OCFS2_LOCK_TYPE_SUPER:
+			c = 'S';
+			break;
+		case OCFS2_LOCK_TYPE_RENAME:
+			c = 'R';
+			break;
+		case OCFS2_LOCK_TYPE_RW:
+			c = 'W';
+			break;
+		case OCFS2_LOCK_TYPE_DENTRY:
+			c = 'N';
+			break;
+		case OCFS2_LOCK_TYPE_OPEN:
+			c = 'O';
+			break;
+		case OCFS2_LOCK_TYPE_FLOCK:
+			c = 'F';
+			break;
+		case OCFS2_LOCK_TYPE_QINFO:
+			c = 'Q';
+			break;
+		case OCFS2_LOCK_TYPE_NFS_SYNC:
+			c = 'Y';
+			break;
+		case OCFS2_LOCK_TYPE_ORPHAN_SCAN:
+			c = 'P';
+			break;
+		case OCFS2_LOCK_TYPE_REFCOUNT:
+			c = 'T';
+			break;
+		default:
+			c = '\0';
+	}
+
+	return c;
+}
+
+static char *ocfs2_lock_type_strings[] = {
+	[OCFS2_LOCK_TYPE_META] = "Meta",
+	[OCFS2_LOCK_TYPE_DATA] = "Data",
+	[OCFS2_LOCK_TYPE_SUPER] = "Super",
+	[OCFS2_LOCK_TYPE_RENAME] = "Rename",
+	/* Need to differntiate from [R]ename.. serializing writes is the
+	 * important job it does, anyway. */
+	[OCFS2_LOCK_TYPE_RW] = "Write/Read",
+	[OCFS2_LOCK_TYPE_DENTRY] = "Dentry",
+	[OCFS2_LOCK_TYPE_OPEN] = "Open",
+	[OCFS2_LOCK_TYPE_FLOCK] = "Flock",
+	[OCFS2_LOCK_TYPE_QINFO] = "Quota",
+	[OCFS2_LOCK_TYPE_NFS_SYNC] = "NFSSync",
+	[OCFS2_LOCK_TYPE_ORPHAN_SCAN] = "OrphanScan",
+	[OCFS2_LOCK_TYPE_REFCOUNT] = "Refcount",
+};
+
+static inline const char *ocfs2_lock_type_string(enum ocfs2_lock_type type)
+{
+#ifdef __KERNEL__
+	BUG_ON(type >= OCFS2_NUM_LOCK_TYPES);
+#endif
+	return ocfs2_lock_type_strings[type];
+}
+
+#endif  /* OCFS2_LOCKID_H */

fs/ocfs2/ocfs2_lockingver.h

@@ -0,0 +1,32 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ocfs2_lockingver.h
+ *
+ * Defines OCFS2 Locking version values.
+ *
+ * Copyright (C) 2008 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License, version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#ifndef OCFS2_LOCKINGVER_H
+#define OCFS2_LOCKINGVER_H
+
+/*
+ * The protocol version for ocfs2 cluster locking.  See dlmglue.c for
+ * more details.
+ *
+ * 1.0 - Initial locking version from ocfs2 1.4.
+ */
+#define OCFS2_LOCKING_PROTOCOL_MAJOR 1
+#define OCFS2_LOCKING_PROTOCOL_MINOR 0
+
+#endif  /* OCFS2_LOCKINGVER_H */

fs/ocfs2/quota.h

@@ -0,0 +1,122 @@
+/*
+ * quota.h for OCFS2
+ *
+ * On disk quota structures for local and global quota file, in-memory
+ * structures.
+ *
+ */
+
+#ifndef _OCFS2_QUOTA_H
+#define _OCFS2_QUOTA_H
+
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/quota.h>
+#include <linux/list.h>
+#include <linux/dqblk_qtree.h>
+
+#include "ocfs2.h"
+
+/* Number of quota types we support */
+#define OCFS2_MAXQUOTAS 2
+
+/*
+ * In-memory structures
+ */
+struct ocfs2_dquot {
+	struct dquot dq_dquot;	/* Generic VFS dquot */
+	loff_t dq_local_off;	/* Offset in the local quota file */
+	u64 dq_local_phys_blk;	/* Physical block carrying quota structure */
+	struct ocfs2_quota_chunk *dq_chunk;	/* Chunk dquot is in */
+	unsigned int dq_use_count;	/* Number of nodes having reference to this entry in global quota file */
+	s64 dq_origspace;	/* Last globally synced space usage */
+	s64 dq_originodes;	/* Last globally synced inode usage */
+	struct llist_node list;	/* Member of list of dquots to drop */
+};
+
+/* Description of one chunk to recover in memory */
+struct ocfs2_recovery_chunk {
+	struct list_head rc_list;	/* List of chunks */
+	int rc_chunk;			/* Chunk number */
+	unsigned long *rc_bitmap;	/* Bitmap of entries to recover */
+};
+
+struct ocfs2_quota_recovery {
+	struct list_head r_list[OCFS2_MAXQUOTAS];	/* List of chunks to recover */
+};
+
+/* In-memory structure with quota header information */
+struct ocfs2_mem_dqinfo {
+	unsigned int dqi_type;		/* Quota type this structure describes */
+	unsigned int dqi_chunks;	/* Number of chunks in local quota file */
+	unsigned int dqi_blocks;	/* Number of blocks allocated for local quota file */
+	unsigned int dqi_syncms;	/* How often should we sync with other nodes */
+	struct list_head dqi_chunk;	/* List of chunks */
+	struct inode *dqi_gqinode;	/* Global quota file inode */
+	struct ocfs2_lock_res dqi_gqlock;	/* Lock protecting quota information structure */
+	struct buffer_head *dqi_gqi_bh;	/* Buffer head with global quota file inode - set only if inode lock is obtained */
+	int dqi_gqi_count;		/* Number of holders of dqi_gqi_bh */
+	u64 dqi_giblk;			/* Number of block with global information header */
+	struct buffer_head *dqi_lqi_bh;	/* Buffer head with local quota file inode */
+	struct buffer_head *dqi_libh;	/* Buffer with local information header */
+	struct qtree_mem_dqinfo dqi_gi;	/* Info about global file */
+	struct delayed_work dqi_sync_work;	/* Work for syncing dquots */
+	struct ocfs2_quota_recovery *dqi_rec;	/* Pointer to recovery
+						 * information, in case we
+						 * enable quotas on file
+						 * needing it */
+};
+
+static inline struct ocfs2_dquot *OCFS2_DQUOT(struct dquot *dquot)
+{
+	return container_of(dquot, struct ocfs2_dquot, dq_dquot);
+}
+
+struct ocfs2_quota_chunk {
+	struct list_head qc_chunk;	/* List of quotafile chunks */
+	int qc_num;			/* Number of quota chunk */
+	struct buffer_head *qc_headerbh;	/* Buffer head with chunk header */
+};
+
+extern struct kmem_cache *ocfs2_dquot_cachep;
+extern struct kmem_cache *ocfs2_qf_chunk_cachep;
+
+extern struct qtree_fmt_operations ocfs2_global_ops;
+
+struct ocfs2_quota_recovery *ocfs2_begin_quota_recovery(
+				struct ocfs2_super *osb, int slot_num);
+int ocfs2_finish_quota_recovery(struct ocfs2_super *osb,
+				struct ocfs2_quota_recovery *rec,
+				int slot_num);
+void ocfs2_free_quota_recovery(struct ocfs2_quota_recovery *rec);
+ssize_t ocfs2_quota_read(struct super_block *sb, int type, char *data,
+			 size_t len, loff_t off);
+ssize_t ocfs2_quota_write(struct super_block *sb, int type,
+			  const char *data, size_t len, loff_t off);
+int ocfs2_global_read_info(struct super_block *sb, int type);
+int ocfs2_global_write_info(struct super_block *sb, int type);
+int ocfs2_global_read_dquot(struct dquot *dquot);
+int __ocfs2_sync_dquot(struct dquot *dquot, int freeing);
+static inline int ocfs2_sync_dquot(struct dquot *dquot)
+{
+	return __ocfs2_sync_dquot(dquot, 0);
+}
+static inline int ocfs2_global_release_dquot(struct dquot *dquot)
+{
+	return __ocfs2_sync_dquot(dquot, 1);
+}
+
+int ocfs2_lock_global_qf(struct ocfs2_mem_dqinfo *oinfo, int ex);
+void ocfs2_unlock_global_qf(struct ocfs2_mem_dqinfo *oinfo, int ex);
+int ocfs2_validate_quota_block(struct super_block *sb, struct buffer_head *bh);
+int ocfs2_read_quota_phys_block(struct inode *inode, u64 p_block,
+				struct buffer_head **bh);
+int ocfs2_create_local_dquot(struct dquot *dquot);
+int ocfs2_local_release_dquot(handle_t *handle, struct dquot *dquot);
+int ocfs2_local_write_dquot(struct dquot *dquot);
+void ocfs2_drop_dquot_refs(struct work_struct *work);
+
+extern const struct dquot_operations ocfs2_quota_operations;
+extern struct quota_format_type ocfs2_quota_format;
+
+#endif /* _OCFS2_QUOTA_H */

fs/ocfs2/quota_global.c

@@ -0,0 +1,971 @@
+/*
+ *  Implementation of operations over global quota file
+ */
+#include <linux/spinlock.h>
+#include <linux/fs.h>
+#include <linux/slab.h>
+#include <linux/quota.h>
+#include <linux/quotaops.h>
+#include <linux/dqblk_qtree.h>
+#include <linux/jiffies.h>
+#include <linux/writeback.h>
+#include <linux/workqueue.h>
+#include <linux/llist.h>
+
+#include <cluster/masklog.h>
+
+#include "ocfs2_fs.h"
+#include "ocfs2.h"
+#include "alloc.h"
+#include "blockcheck.h"
+#include "inode.h"
+#include "journal.h"
+#include "file.h"
+#include "sysfile.h"
+#include "dlmglue.h"
+#include "uptodate.h"
+#include "super.h"
+#include "buffer_head_io.h"
+#include "quota.h"
+#include "ocfs2_trace.h"
+
+/*
+ * Locking of quotas with OCFS2 is rather complex. Here are rules that
+ * should be obeyed by all the functions:
+ * - any write of quota structure (either to local or global file) is protected
+ *   by dqio_mutex or dquot->dq_lock.
+ * - any modification of global quota file holds inode cluster lock, i_mutex,
+ *   and ip_alloc_sem of the global quota file (achieved by
+ *   ocfs2_lock_global_qf). It also has to hold qinfo_lock.
+ * - an allocation of new blocks for local quota file is protected by
+ *   its ip_alloc_sem
+ *
+ * A rough sketch of locking dependencies (lf = local file, gf = global file):
+ * Normal filesystem operation:
+ *   start_trans -> dqio_mutex -> write to lf
+ * Syncing of local and global file:
+ *   ocfs2_lock_global_qf -> start_trans -> dqio_mutex -> qinfo_lock ->
+ *     write to gf
+ *						       -> write to lf
+ * Acquire dquot for the first time:
+ *   dq_lock -> ocfs2_lock_global_qf -> qinfo_lock -> read from gf
+ *				     -> alloc space for gf
+ *				     -> start_trans -> qinfo_lock -> write to gf
+ *	     -> ip_alloc_sem of lf -> alloc space for lf
+ *	     -> write to lf
+ * Release last reference to dquot:
+ *   dq_lock -> ocfs2_lock_global_qf -> start_trans -> qinfo_lock -> write to gf
+ *	     -> write to lf
+ * Note that all the above operations also hold the inode cluster lock of lf.
+ * Recovery:
+ *   inode cluster lock of recovered lf
+ *     -> read bitmaps -> ip_alloc_sem of lf
+ *     -> ocfs2_lock_global_qf -> start_trans -> dqio_mutex -> qinfo_lock ->
+ *        write to gf
+ */
+
+static void qsync_work_fn(struct work_struct *work);
+
+static void ocfs2_global_disk2memdqb(struct dquot *dquot, void *dp)
+{
+	struct ocfs2_global_disk_dqblk *d = dp;
+	struct mem_dqblk *m = &dquot->dq_dqb;
+
+	/* Update from disk only entries not set by the admin */
+	if (!test_bit(DQ_LASTSET_B + QIF_ILIMITS_B, &dquot->dq_flags)) {
+		m->dqb_ihardlimit = le64_to_cpu(d->dqb_ihardlimit);
+		m->dqb_isoftlimit = le64_to_cpu(d->dqb_isoftlimit);
+	}
+	if (!test_bit(DQ_LASTSET_B + QIF_INODES_B, &dquot->dq_flags))
+		m->dqb_curinodes = le64_to_cpu(d->dqb_curinodes);
+	if (!test_bit(DQ_LASTSET_B + QIF_BLIMITS_B, &dquot->dq_flags)) {
+		m->dqb_bhardlimit = le64_to_cpu(d->dqb_bhardlimit);
+		m->dqb_bsoftlimit = le64_to_cpu(d->dqb_bsoftlimit);
+	}
+	if (!test_bit(DQ_LASTSET_B + QIF_SPACE_B, &dquot->dq_flags))
+		m->dqb_curspace = le64_to_cpu(d->dqb_curspace);
+	if (!test_bit(DQ_LASTSET_B + QIF_BTIME_B, &dquot->dq_flags))
+		m->dqb_btime = le64_to_cpu(d->dqb_btime);
+	if (!test_bit(DQ_LASTSET_B + QIF_ITIME_B, &dquot->dq_flags))
+		m->dqb_itime = le64_to_cpu(d->dqb_itime);
+	OCFS2_DQUOT(dquot)->dq_use_count = le32_to_cpu(d->dqb_use_count);
+}
+
+static void ocfs2_global_mem2diskdqb(void *dp, struct dquot *dquot)
+{
+	struct ocfs2_global_disk_dqblk *d = dp;
+	struct mem_dqblk *m = &dquot->dq_dqb;
+
+	d->dqb_id = cpu_to_le32(from_kqid(&init_user_ns, dquot->dq_id));
+	d->dqb_use_count = cpu_to_le32(OCFS2_DQUOT(dquot)->dq_use_count);
+	d->dqb_ihardlimit = cpu_to_le64(m->dqb_ihardlimit);
+	d->dqb_isoftlimit = cpu_to_le64(m->dqb_isoftlimit);
+	d->dqb_curinodes = cpu_to_le64(m->dqb_curinodes);
+	d->dqb_bhardlimit = cpu_to_le64(m->dqb_bhardlimit);
+	d->dqb_bsoftlimit = cpu_to_le64(m->dqb_bsoftlimit);
+	d->dqb_curspace = cpu_to_le64(m->dqb_curspace);
+	d->dqb_btime = cpu_to_le64(m->dqb_btime);
+	d->dqb_itime = cpu_to_le64(m->dqb_itime);
+	d->dqb_pad1 = d->dqb_pad2 = 0;
+}
+
+static int ocfs2_global_is_id(void *dp, struct dquot *dquot)
+{
+	struct ocfs2_global_disk_dqblk *d = dp;
+	struct ocfs2_mem_dqinfo *oinfo =
+			sb_dqinfo(dquot->dq_sb, dquot->dq_id.type)->dqi_priv;
+
+	if (qtree_entry_unused(&oinfo->dqi_gi, dp))
+		return 0;
+
+	return qid_eq(make_kqid(&init_user_ns, dquot->dq_id.type,
+				le32_to_cpu(d->dqb_id)),
+		      dquot->dq_id);
+}
+
+struct qtree_fmt_operations ocfs2_global_ops = {
+	.mem2disk_dqblk = ocfs2_global_mem2diskdqb,
+	.disk2mem_dqblk = ocfs2_global_disk2memdqb,
+	.is_id = ocfs2_global_is_id,
+};
+
+int ocfs2_validate_quota_block(struct super_block *sb, struct buffer_head *bh)
+{
+	struct ocfs2_disk_dqtrailer *dqt =
+		ocfs2_block_dqtrailer(sb->s_blocksize, bh->b_data);
+
+	trace_ocfs2_validate_quota_block((unsigned long long)bh->b_blocknr);
+
+	BUG_ON(!buffer_uptodate(bh));
+
+	/*
+	 * If the ecc fails, we return the error but otherwise
+	 * leave the filesystem running.  We know any error is
+	 * local to this block.
+	 */
+	return ocfs2_validate_meta_ecc(sb, bh->b_data, &dqt->dq_check);
+}
+
+int ocfs2_read_quota_phys_block(struct inode *inode, u64 p_block,
+				struct buffer_head **bhp)
+{
+	int rc;
+
+	*bhp = NULL;
+	rc = ocfs2_read_blocks(INODE_CACHE(inode), p_block, 1, bhp, 0,
+			       ocfs2_validate_quota_block);
+	if (rc)
+		mlog_errno(rc);
+	return rc;
+}
+
+/* Read data from global quotafile - avoid pagecache and such because we cannot
+ * afford acquiring the locks... We use quota cluster lock to serialize
+ * operations. Caller is responsible for acquiring it. */
+ssize_t ocfs2_quota_read(struct super_block *sb, int type, char *data,
+			 size_t len, loff_t off)
+{
+	struct ocfs2_mem_dqinfo *oinfo = sb_dqinfo(sb, type)->dqi_priv;
+	struct inode *gqinode = oinfo->dqi_gqinode;
+	loff_t i_size = i_size_read(gqinode);
+	int offset = off & (sb->s_blocksize - 1);
+	sector_t blk = off >> sb->s_blocksize_bits;
+	int err = 0;
+	struct buffer_head *bh;
+	size_t toread, tocopy;
+	u64 pblock = 0, pcount = 0;
+
+	if (off > i_size)
+		return 0;
+	if (off + len > i_size)
+		len = i_size - off;
+	toread = len;
+	while (toread > 0) {
+		tocopy = min_t(size_t, (sb->s_blocksize - offset), toread);
+		if (!pcount) {
+			err = ocfs2_extent_map_get_blocks(gqinode, blk, &pblock,
+							  &pcount, NULL);
+			if (err) {
+				mlog_errno(err);
+				return err;
+			}
+		} else {
+			pcount--;
+			pblock++;
+		}
+		bh = NULL;
+		err = ocfs2_read_quota_phys_block(gqinode, pblock, &bh);
+		if (err) {
+			mlog_errno(err);
+			return err;
+		}
+		memcpy(data, bh->b_data + offset, tocopy);
+		brelse(bh);
+		offset = 0;
+		toread -= tocopy;
+		data += tocopy;
+		blk++;
+	}
+	return len;
+}
+
+/* Write to quotafile (we know the transaction is already started and has
+ * enough credits) */
+ssize_t ocfs2_quota_write(struct super_block *sb, int type,
+			  const char *data, size_t len, loff_t off)
+{
+	struct mem_dqinfo *info = sb_dqinfo(sb, type);
+	struct ocfs2_mem_dqinfo *oinfo = info->dqi_priv;
+	struct inode *gqinode = oinfo->dqi_gqinode;
+	int offset = off & (sb->s_blocksize - 1);
+	sector_t blk = off >> sb->s_blocksize_bits;
+	int err = 0, new = 0, ja_type;
+	struct buffer_head *bh = NULL;
+	handle_t *handle = journal_current_handle();
+	u64 pblock, pcount;
+
+	if (!handle) {
+		mlog(ML_ERROR, "Quota write (off=%llu, len=%llu) cancelled "
+		     "because transaction was not started.\n",
+		     (unsigned long long)off, (unsigned long long)len);
+		return -EIO;
+	}
+	if (len > sb->s_blocksize - OCFS2_QBLK_RESERVED_SPACE - offset) {
+		WARN_ON(1);
+		len = sb->s_blocksize - OCFS2_QBLK_RESERVED_SPACE - offset;
+	}
+
+	if (i_size_read(gqinode) < off + len) {
+		loff_t rounded_end =
+				ocfs2_align_bytes_to_blocks(sb, off + len);
+
+		/* Space is already allocated in ocfs2_acquire_dquot() */
+		err = ocfs2_simple_size_update(gqinode,
+					       oinfo->dqi_gqi_bh,
+					       rounded_end);
+		if (err < 0)
+			goto out;
+		new = 1;
+	}
+	err = ocfs2_extent_map_get_blocks(gqinode, blk, &pblock, &pcount, NULL);
+	if (err) {
+		mlog_errno(err);
+		goto out;
+	}
+	/* Not rewriting whole block? */
+	if ((offset || len < sb->s_blocksize - OCFS2_QBLK_RESERVED_SPACE) &&
+	    !new) {
+		err = ocfs2_read_quota_phys_block(gqinode, pblock, &bh);
+		ja_type = OCFS2_JOURNAL_ACCESS_WRITE;
+	} else {
+		bh = sb_getblk(sb, pblock);
+		if (!bh)
+			err = -ENOMEM;
+		ja_type = OCFS2_JOURNAL_ACCESS_CREATE;
+	}
+	if (err) {
+		mlog_errno(err);
+		goto out;
+	}
+	lock_buffer(bh);
+	if (new)
+		memset(bh->b_data, 0, sb->s_blocksize);
+	memcpy(bh->b_data + offset, data, len);
+	flush_dcache_page(bh->b_page);
+	set_buffer_uptodate(bh);
+	unlock_buffer(bh);
+	ocfs2_set_buffer_uptodate(INODE_CACHE(gqinode), bh);
+	err = ocfs2_journal_access_dq(handle, INODE_CACHE(gqinode), bh,
+				      ja_type);
+	if (err < 0) {
+		brelse(bh);
+		goto out;
+	}
+	ocfs2_journal_dirty(handle, bh);
+	brelse(bh);
+out:
+	if (err) {
+		mlog_errno(err);
+		return err;
+	}
+	gqinode->i_version++;
+	ocfs2_mark_inode_dirty(handle, gqinode, oinfo->dqi_gqi_bh);
+	return len;
+}
+
+int ocfs2_lock_global_qf(struct ocfs2_mem_dqinfo *oinfo, int ex)
+{
+	int status;
+	struct buffer_head *bh = NULL;
+
+	status = ocfs2_inode_lock(oinfo->dqi_gqinode, &bh, ex);
+	if (status < 0)
+		return status;
+	spin_lock(&dq_data_lock);
+	if (!oinfo->dqi_gqi_count++)
+		oinfo->dqi_gqi_bh = bh;
+	else
+		WARN_ON(bh != oinfo->dqi_gqi_bh);
+	spin_unlock(&dq_data_lock);
+	if (ex) {
+		mutex_lock(&oinfo->dqi_gqinode->i_mutex);
+		down_write(&OCFS2_I(oinfo->dqi_gqinode)->ip_alloc_sem);
+	} else {
+		down_read(&OCFS2_I(oinfo->dqi_gqinode)->ip_alloc_sem);
+	}
+	return 0;
+}
+
+void ocfs2_unlock_global_qf(struct ocfs2_mem_dqinfo *oinfo, int ex)
+{
+	if (ex) {
+		up_write(&OCFS2_I(oinfo->dqi_gqinode)->ip_alloc_sem);
+		mutex_unlock(&oinfo->dqi_gqinode->i_mutex);
+	} else {
+		up_read(&OCFS2_I(oinfo->dqi_gqinode)->ip_alloc_sem);
+	}
+	ocfs2_inode_unlock(oinfo->dqi_gqinode, ex);
+	brelse(oinfo->dqi_gqi_bh);
+	spin_lock(&dq_data_lock);
+	if (!--oinfo->dqi_gqi_count)
+		oinfo->dqi_gqi_bh = NULL;
+	spin_unlock(&dq_data_lock);
+}
+
+/* Read information header from global quota file */
+int ocfs2_global_read_info(struct super_block *sb, int type)
+{
+	struct inode *gqinode = NULL;
+	unsigned int ino[OCFS2_MAXQUOTAS] = { USER_QUOTA_SYSTEM_INODE,
+					      GROUP_QUOTA_SYSTEM_INODE };
+	struct ocfs2_global_disk_dqinfo dinfo;
+	struct mem_dqinfo *info = sb_dqinfo(sb, type);
+	struct ocfs2_mem_dqinfo *oinfo = info->dqi_priv;
+	u64 pcount;
+	int status;
+
+	/* Read global header */
+	gqinode = ocfs2_get_system_file_inode(OCFS2_SB(sb), ino[type],
+			OCFS2_INVALID_SLOT);
+	if (!gqinode) {
+		mlog(ML_ERROR, "failed to get global quota inode (type=%d)\n",
+			type);
+		status = -EINVAL;
+		goto out_err;
+	}
+	oinfo->dqi_gi.dqi_sb = sb;
+	oinfo->dqi_gi.dqi_type = type;
+	ocfs2_qinfo_lock_res_init(&oinfo->dqi_gqlock, oinfo);
+	oinfo->dqi_gi.dqi_entry_size = sizeof(struct ocfs2_global_disk_dqblk);
+	oinfo->dqi_gi.dqi_ops = &ocfs2_global_ops;
+	oinfo->dqi_gqi_bh = NULL;
+	oinfo->dqi_gqi_count = 0;
+	oinfo->dqi_gqinode = gqinode;
+	status = ocfs2_lock_global_qf(oinfo, 0);
+	if (status < 0) {
+		mlog_errno(status);
+		goto out_err;
+	}
+
+	status = ocfs2_extent_map_get_blocks(gqinode, 0, &oinfo->dqi_giblk,
+					     &pcount, NULL);
+	if (status < 0)
+		goto out_unlock;
+
+	status = ocfs2_qinfo_lock(oinfo, 0);
+	if (status < 0)
+		goto out_unlock;
+	status = sb->s_op->quota_read(sb, type, (char *)&dinfo,
+				      sizeof(struct ocfs2_global_disk_dqinfo),
+				      OCFS2_GLOBAL_INFO_OFF);
+	ocfs2_qinfo_unlock(oinfo, 0);
+	ocfs2_unlock_global_qf(oinfo, 0);
+	if (status != sizeof(struct ocfs2_global_disk_dqinfo)) {
+		mlog(ML_ERROR, "Cannot read global quota info (%d).\n",
+		     status);
+		if (status >= 0)
+			status = -EIO;
+		mlog_errno(status);
+		goto out_err;
+	}
+	info->dqi_bgrace = le32_to_cpu(dinfo.dqi_bgrace);
+	info->dqi_igrace = le32_to_cpu(dinfo.dqi_igrace);
+	oinfo->dqi_syncms = le32_to_cpu(dinfo.dqi_syncms);
+	oinfo->dqi_gi.dqi_blocks = le32_to_cpu(dinfo.dqi_blocks);
+	oinfo->dqi_gi.dqi_free_blk = le32_to_cpu(dinfo.dqi_free_blk);
+	oinfo->dqi_gi.dqi_free_entry = le32_to_cpu(dinfo.dqi_free_entry);
+	oinfo->dqi_gi.dqi_blocksize_bits = sb->s_blocksize_bits;
+	oinfo->dqi_gi.dqi_usable_bs = sb->s_blocksize -
+						OCFS2_QBLK_RESERVED_SPACE;
+	oinfo->dqi_gi.dqi_qtree_depth = qtree_depth(&oinfo->dqi_gi);
+	INIT_DELAYED_WORK(&oinfo->dqi_sync_work, qsync_work_fn);
+	schedule_delayed_work(&oinfo->dqi_sync_work,
+			      msecs_to_jiffies(oinfo->dqi_syncms));
+
+out_err:
+	return status;
+out_unlock:
+	ocfs2_unlock_global_qf(oinfo, 0);
+	mlog_errno(status);
+	goto out_err;
+}
+
+/* Write information to global quota file. Expects exlusive lock on quota
+ * file inode and quota info */
+static int __ocfs2_global_write_info(struct super_block *sb, int type)
+{
+	struct mem_dqinfo *info = sb_dqinfo(sb, type);
+	struct ocfs2_mem_dqinfo *oinfo = info->dqi_priv;
+	struct ocfs2_global_disk_dqinfo dinfo;
+	ssize_t size;
+
+	spin_lock(&dq_data_lock);
+	info->dqi_flags &= ~DQF_INFO_DIRTY;
+	dinfo.dqi_bgrace = cpu_to_le32(info->dqi_bgrace);
+	dinfo.dqi_igrace = cpu_to_le32(info->dqi_igrace);
+	spin_unlock(&dq_data_lock);
+	dinfo.dqi_syncms = cpu_to_le32(oinfo->dqi_syncms);
+	dinfo.dqi_blocks = cpu_to_le32(oinfo->dqi_gi.dqi_blocks);
+	dinfo.dqi_free_blk = cpu_to_le32(oinfo->dqi_gi.dqi_free_blk);
+	dinfo.dqi_free_entry = cpu_to_le32(oinfo->dqi_gi.dqi_free_entry);
+	size = sb->s_op->quota_write(sb, type, (char *)&dinfo,
+				     sizeof(struct ocfs2_global_disk_dqinfo),
+				     OCFS2_GLOBAL_INFO_OFF);
+	if (size != sizeof(struct ocfs2_global_disk_dqinfo)) {
+		mlog(ML_ERROR, "Cannot write global quota info structure\n");
+		if (size >= 0)
+			size = -EIO;
+		return size;
+	}
+	return 0;
+}
+
+int ocfs2_global_write_info(struct super_block *sb, int type)
+{
+	int err;
+	struct ocfs2_mem_dqinfo *info = sb_dqinfo(sb, type)->dqi_priv;
+
+	err = ocfs2_qinfo_lock(info, 1);
+	if (err < 0)
+		return err;
+	err = __ocfs2_global_write_info(sb, type);
+	ocfs2_qinfo_unlock(info, 1);
+	return err;
+}
+
+static int ocfs2_global_qinit_alloc(struct super_block *sb, int type)
+{
+	struct ocfs2_mem_dqinfo *oinfo = sb_dqinfo(sb, type)->dqi_priv;
+
+	/*
+	 * We may need to allocate tree blocks and a leaf block but not the
+	 * root block
+	 */
+	return oinfo->dqi_gi.dqi_qtree_depth;
+}
+
+static int ocfs2_calc_global_qinit_credits(struct super_block *sb, int type)
+{
+	/* We modify all the allocated blocks, tree root, info block and
+	 * the inode */
+	return (ocfs2_global_qinit_alloc(sb, type) + 2) *
+			OCFS2_QUOTA_BLOCK_UPDATE_CREDITS + 1;
+}
+
+/* Sync local information about quota modifications with global quota file.
+ * Caller must have started the transaction and obtained exclusive lock for
+ * global quota file inode */
+int __ocfs2_sync_dquot(struct dquot *dquot, int freeing)
+{
+	int err, err2;
+	struct super_block *sb = dquot->dq_sb;
+	int type = dquot->dq_id.type;
+	struct ocfs2_mem_dqinfo *info = sb_dqinfo(sb, type)->dqi_priv;
+	struct ocfs2_global_disk_dqblk dqblk;
+	s64 spacechange, inodechange;
+	time_t olditime, oldbtime;
+
+	err = sb->s_op->quota_read(sb, type, (char *)&dqblk,
+				   sizeof(struct ocfs2_global_disk_dqblk),
+				   dquot->dq_off);
+	if (err != sizeof(struct ocfs2_global_disk_dqblk)) {
+		if (err >= 0) {
+			mlog(ML_ERROR, "Short read from global quota file "
+				       "(%u read)\n", err);
+			err = -EIO;
+		}
+		goto out;
+	}
+
+	/* Update space and inode usage. Get also other information from
+	 * global quota file so that we don't overwrite any changes there.
+	 * We are */
+	spin_lock(&dq_data_lock);
+	spacechange = dquot->dq_dqb.dqb_curspace -
+					OCFS2_DQUOT(dquot)->dq_origspace;
+	inodechange = dquot->dq_dqb.dqb_curinodes -
+					OCFS2_DQUOT(dquot)->dq_originodes;
+	olditime = dquot->dq_dqb.dqb_itime;
+	oldbtime = dquot->dq_dqb.dqb_btime;
+	ocfs2_global_disk2memdqb(dquot, &dqblk);
+	trace_ocfs2_sync_dquot(from_kqid(&init_user_ns, dquot->dq_id),
+			       dquot->dq_dqb.dqb_curspace,
+			       (long long)spacechange,
+			       dquot->dq_dqb.dqb_curinodes,
+			       (long long)inodechange);
+	if (!test_bit(DQ_LASTSET_B + QIF_SPACE_B, &dquot->dq_flags))
+		dquot->dq_dqb.dqb_curspace += spacechange;
+	if (!test_bit(DQ_LASTSET_B + QIF_INODES_B, &dquot->dq_flags))
+		dquot->dq_dqb.dqb_curinodes += inodechange;
+	/* Set properly space grace time... */
+	if (dquot->dq_dqb.dqb_bsoftlimit &&
+	    dquot->dq_dqb.dqb_curspace > dquot->dq_dqb.dqb_bsoftlimit) {
+		if (!test_bit(DQ_LASTSET_B + QIF_BTIME_B, &dquot->dq_flags) &&
+		    oldbtime > 0) {
+			if (dquot->dq_dqb.dqb_btime > 0)
+				dquot->dq_dqb.dqb_btime =
+					min(dquot->dq_dqb.dqb_btime, oldbtime);
+			else
+				dquot->dq_dqb.dqb_btime = oldbtime;
+		}
+	} else {
+		dquot->dq_dqb.dqb_btime = 0;
+		clear_bit(DQ_BLKS_B, &dquot->dq_flags);
+	}
+	/* Set properly inode grace time... */
+	if (dquot->dq_dqb.dqb_isoftlimit &&
+	    dquot->dq_dqb.dqb_curinodes > dquot->dq_dqb.dqb_isoftlimit) {
+		if (!test_bit(DQ_LASTSET_B + QIF_ITIME_B, &dquot->dq_flags) &&
+		    olditime > 0) {
+			if (dquot->dq_dqb.dqb_itime > 0)
+				dquot->dq_dqb.dqb_itime =
+					min(dquot->dq_dqb.dqb_itime, olditime);
+			else
+				dquot->dq_dqb.dqb_itime = olditime;
+		}
+	} else {
+		dquot->dq_dqb.dqb_itime = 0;
+		clear_bit(DQ_INODES_B, &dquot->dq_flags);
+	}
+	/* All information is properly updated, clear the flags */
+	__clear_bit(DQ_LASTSET_B + QIF_SPACE_B, &dquot->dq_flags);
+	__clear_bit(DQ_LASTSET_B + QIF_INODES_B, &dquot->dq_flags);
+	__clear_bit(DQ_LASTSET_B + QIF_BLIMITS_B, &dquot->dq_flags);
+	__clear_bit(DQ_LASTSET_B + QIF_ILIMITS_B, &dquot->dq_flags);
+	__clear_bit(DQ_LASTSET_B + QIF_BTIME_B, &dquot->dq_flags);
+	__clear_bit(DQ_LASTSET_B + QIF_ITIME_B, &dquot->dq_flags);
+	OCFS2_DQUOT(dquot)->dq_origspace = dquot->dq_dqb.dqb_curspace;
+	OCFS2_DQUOT(dquot)->dq_originodes = dquot->dq_dqb.dqb_curinodes;
+	spin_unlock(&dq_data_lock);
+	err = ocfs2_qinfo_lock(info, freeing);
+	if (err < 0) {
+		mlog(ML_ERROR, "Failed to lock quota info, losing quota write"
+			       " (type=%d, id=%u)\n", dquot->dq_id.type,
+			       (unsigned)from_kqid(&init_user_ns, dquot->dq_id));
+		goto out;
+	}
+	if (freeing)
+		OCFS2_DQUOT(dquot)->dq_use_count--;
+	err = qtree_write_dquot(&info->dqi_gi, dquot);
+	if (err < 0)
+		goto out_qlock;
+	if (freeing && !OCFS2_DQUOT(dquot)->dq_use_count) {
+		err = qtree_release_dquot(&info->dqi_gi, dquot);
+		if (info_dirty(sb_dqinfo(sb, type))) {
+			err2 = __ocfs2_global_write_info(sb, type);
+			if (!err)
+				err = err2;
+		}
+	}
+out_qlock:
+	ocfs2_qinfo_unlock(info, freeing);
+out:
+	if (err < 0)
+		mlog_errno(err);
+	return err;
+}
+
+/*
+ *  Functions for periodic syncing of dquots with global file
+ */
+static int ocfs2_sync_dquot_helper(struct dquot *dquot, unsigned long type)
+{
+	handle_t *handle;
+	struct super_block *sb = dquot->dq_sb;
+	struct ocfs2_mem_dqinfo *oinfo = sb_dqinfo(sb, type)->dqi_priv;
+	struct ocfs2_super *osb = OCFS2_SB(sb);
+	int status = 0;
+
+	trace_ocfs2_sync_dquot_helper(from_kqid(&init_user_ns, dquot->dq_id),
+				      dquot->dq_id.type,
+				      type, sb->s_id);
+	if (type != dquot->dq_id.type)
+		goto out;
+	status = ocfs2_lock_global_qf(oinfo, 1);
+	if (status < 0)
+		goto out;
+
+	handle = ocfs2_start_trans(osb, OCFS2_QSYNC_CREDITS);
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		mlog_errno(status);
+		goto out_ilock;
+	}
+	mutex_lock(&sb_dqopt(sb)->dqio_mutex);
+	status = ocfs2_sync_dquot(dquot);
+	if (status < 0)
+		mlog_errno(status);
+	/* We have to write local structure as well... */
+	status = ocfs2_local_write_dquot(dquot);
+	if (status < 0)
+		mlog_errno(status);
+	mutex_unlock(&sb_dqopt(sb)->dqio_mutex);
+	ocfs2_commit_trans(osb, handle);
+out_ilock:
+	ocfs2_unlock_global_qf(oinfo, 1);
+out:
+	return status;
+}
+
+static void qsync_work_fn(struct work_struct *work)
+{
+	struct ocfs2_mem_dqinfo *oinfo = container_of(work,
+						      struct ocfs2_mem_dqinfo,
+						      dqi_sync_work.work);
+	struct super_block *sb = oinfo->dqi_gqinode->i_sb;
+
+	dquot_scan_active(sb, ocfs2_sync_dquot_helper, oinfo->dqi_type);
+	schedule_delayed_work(&oinfo->dqi_sync_work,
+			      msecs_to_jiffies(oinfo->dqi_syncms));
+}
+
+/*
+ *  Wrappers for generic quota functions
+ */
+
+static int ocfs2_write_dquot(struct dquot *dquot)
+{
+	handle_t *handle;
+	struct ocfs2_super *osb = OCFS2_SB(dquot->dq_sb);
+	int status = 0;
+
+	trace_ocfs2_write_dquot(from_kqid(&init_user_ns, dquot->dq_id),
+				dquot->dq_id.type);
+
+	handle = ocfs2_start_trans(osb, OCFS2_QWRITE_CREDITS);
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		mlog_errno(status);
+		goto out;
+	}
+	mutex_lock(&sb_dqopt(dquot->dq_sb)->dqio_mutex);
+	status = ocfs2_local_write_dquot(dquot);
+	mutex_unlock(&sb_dqopt(dquot->dq_sb)->dqio_mutex);
+	ocfs2_commit_trans(osb, handle);
+out:
+	return status;
+}
+
+static int ocfs2_calc_qdel_credits(struct super_block *sb, int type)
+{
+	struct ocfs2_mem_dqinfo *oinfo = sb_dqinfo(sb, type)->dqi_priv;
+	/*
+	 * We modify tree, leaf block, global info, local chunk header,
+	 * global and local inode; OCFS2_QINFO_WRITE_CREDITS already
+	 * accounts for inode update
+	 */
+	return (oinfo->dqi_gi.dqi_qtree_depth + 2) *
+	       OCFS2_QUOTA_BLOCK_UPDATE_CREDITS +
+	       OCFS2_QINFO_WRITE_CREDITS +
+	       OCFS2_INODE_UPDATE_CREDITS;
+}
+
+void ocfs2_drop_dquot_refs(struct work_struct *work)
+{
+	struct ocfs2_super *osb = container_of(work, struct ocfs2_super,
+					       dquot_drop_work);
+	struct llist_node *list;
+	struct ocfs2_dquot *odquot, *next_odquot;
+
+	list = llist_del_all(&osb->dquot_drop_list);
+	llist_for_each_entry_safe(odquot, next_odquot, list, list) {
+		/* Drop the reference we acquired in ocfs2_dquot_release() */
+		dqput(&odquot->dq_dquot);
+	}
+}
+
+/*
+ * Called when the last reference to dquot is dropped. If we are called from
+ * downconvert thread, we cannot do all the handling here because grabbing
+ * quota lock could deadlock (the node holding the quota lock could need some
+ * other cluster lock to proceed but with blocked downconvert thread we cannot
+ * release any lock).
+ */
+static int ocfs2_release_dquot(struct dquot *dquot)
+{
+	handle_t *handle;
+	struct ocfs2_mem_dqinfo *oinfo =
+			sb_dqinfo(dquot->dq_sb, dquot->dq_id.type)->dqi_priv;
+	struct ocfs2_super *osb = OCFS2_SB(dquot->dq_sb);
+	int status = 0;
+
+	trace_ocfs2_release_dquot(from_kqid(&init_user_ns, dquot->dq_id),
+				  dquot->dq_id.type);
+
+	mutex_lock(&dquot->dq_lock);
+	/* Check whether we are not racing with some other dqget() */
+	if (atomic_read(&dquot->dq_count) > 1)
+		goto out;
+	/* Running from downconvert thread? Postpone quota processing to wq */
+	if (current == osb->dc_task) {
+		/*
+		 * Grab our own reference to dquot and queue it for delayed
+		 * dropping.  Quota code rechecks after calling
+		 * ->release_dquot() and won't free dquot structure.
+		 */
+		dqgrab(dquot);
+		/* First entry on list -> queue work */
+		if (llist_add(&OCFS2_DQUOT(dquot)->list, &osb->dquot_drop_list))
+			queue_work(ocfs2_wq, &osb->dquot_drop_work);
+		goto out;
+	}
+	status = ocfs2_lock_global_qf(oinfo, 1);
+	if (status < 0)
+		goto out;
+	handle = ocfs2_start_trans(osb,
+		ocfs2_calc_qdel_credits(dquot->dq_sb, dquot->dq_id.type));
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		mlog_errno(status);
+		goto out_ilock;
+	}
+
+	status = ocfs2_global_release_dquot(dquot);
+	if (status < 0) {
+		mlog_errno(status);
+		goto out_trans;
+	}
+	status = ocfs2_local_release_dquot(handle, dquot);
+	/*
+	 * If we fail here, we cannot do much as global structure is
+	 * already released. So just complain...
+	 */
+	if (status < 0)
+		mlog_errno(status);
+	/*
+	 * Clear dq_off so that we search for the structure in quota file next
+	 * time we acquire it. The structure might be deleted and reallocated
+	 * elsewhere by another node while our dquot structure is on freelist.
+	 */
+	dquot->dq_off = 0;
+	clear_bit(DQ_ACTIVE_B, &dquot->dq_flags);
+out_trans:
+	ocfs2_commit_trans(osb, handle);
+out_ilock:
+	ocfs2_unlock_global_qf(oinfo, 1);
+out:
+	mutex_unlock(&dquot->dq_lock);
+	if (status)
+		mlog_errno(status);
+	return status;
+}
+
+/*
+ * Read global dquot structure from disk or create it if it does
+ * not exist. Also update use count of the global structure and
+ * create structure in node-local quota file.
+ */
+static int ocfs2_acquire_dquot(struct dquot *dquot)
+{
+	int status = 0, err;
+	int ex = 0;
+	struct super_block *sb = dquot->dq_sb;
+	struct ocfs2_super *osb = OCFS2_SB(sb);
+	int type = dquot->dq_id.type;
+	struct ocfs2_mem_dqinfo *info = sb_dqinfo(sb, type)->dqi_priv;
+	struct inode *gqinode = info->dqi_gqinode;
+	int need_alloc = ocfs2_global_qinit_alloc(sb, type);
+	handle_t *handle;
+
+	trace_ocfs2_acquire_dquot(from_kqid(&init_user_ns, dquot->dq_id),
+				  type);
+	mutex_lock(&dquot->dq_lock);
+	/*
+	 * We need an exclusive lock, because we're going to update use count
+	 * and instantiate possibly new dquot structure
+	 */
+	status = ocfs2_lock_global_qf(info, 1);
+	if (status < 0)
+		goto out;
+	status = ocfs2_qinfo_lock(info, 0);
+	if (status < 0)
+		goto out_dq;
+	/*
+	 * We always want to read dquot structure from disk because we don't
+	 * know what happened with it while it was on freelist.
+	 */
+	status = qtree_read_dquot(&info->dqi_gi, dquot);
+	ocfs2_qinfo_unlock(info, 0);
+	if (status < 0)
+		goto out_dq;
+
+	OCFS2_DQUOT(dquot)->dq_use_count++;
+	OCFS2_DQUOT(dquot)->dq_origspace = dquot->dq_dqb.dqb_curspace;
+	OCFS2_DQUOT(dquot)->dq_originodes = dquot->dq_dqb.dqb_curinodes;
+	if (!dquot->dq_off) {	/* No real quota entry? */
+		ex = 1;
+		/*
+		 * Add blocks to quota file before we start a transaction since
+		 * locking allocators ranks above a transaction start
+		 */
+		WARN_ON(journal_current_handle());
+		status = ocfs2_extend_no_holes(gqinode, NULL,
+			i_size_read(gqinode) + (need_alloc << sb->s_blocksize_bits),
+			i_size_read(gqinode));
+		if (status < 0)
+			goto out_dq;
+	}
+
+	handle = ocfs2_start_trans(osb,
+				   ocfs2_calc_global_qinit_credits(sb, type));
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		goto out_dq;
+	}
+	status = ocfs2_qinfo_lock(info, ex);
+	if (status < 0)
+		goto out_trans;
+	status = qtree_write_dquot(&info->dqi_gi, dquot);
+	if (ex && info_dirty(sb_dqinfo(sb, type))) {
+		err = __ocfs2_global_write_info(sb, type);
+		if (!status)
+			status = err;
+	}
+	ocfs2_qinfo_unlock(info, ex);
+out_trans:
+	ocfs2_commit_trans(osb, handle);
+out_dq:
+	ocfs2_unlock_global_qf(info, 1);
+	if (status < 0)
+		goto out;
+
+	status = ocfs2_create_local_dquot(dquot);
+	if (status < 0)
+		goto out;
+	set_bit(DQ_ACTIVE_B, &dquot->dq_flags);
+out:
+	mutex_unlock(&dquot->dq_lock);
+	if (status)
+		mlog_errno(status);
+	return status;
+}
+
+static int ocfs2_mark_dquot_dirty(struct dquot *dquot)
+{
+	unsigned long mask = (1 << (DQ_LASTSET_B + QIF_ILIMITS_B)) |
+			     (1 << (DQ_LASTSET_B + QIF_BLIMITS_B)) |
+			     (1 << (DQ_LASTSET_B + QIF_INODES_B)) |
+			     (1 << (DQ_LASTSET_B + QIF_SPACE_B)) |
+			     (1 << (DQ_LASTSET_B + QIF_BTIME_B)) |
+			     (1 << (DQ_LASTSET_B + QIF_ITIME_B));
+	int sync = 0;
+	int status;
+	struct super_block *sb = dquot->dq_sb;
+	int type = dquot->dq_id.type;
+	struct ocfs2_mem_dqinfo *oinfo = sb_dqinfo(sb, type)->dqi_priv;
+	handle_t *handle;
+	struct ocfs2_super *osb = OCFS2_SB(sb);
+
+	trace_ocfs2_mark_dquot_dirty(from_kqid(&init_user_ns, dquot->dq_id),
+				     type);
+
+	/* In case user set some limits, sync dquot immediately to global
+	 * quota file so that information propagates quicker */
+	spin_lock(&dq_data_lock);
+	if (dquot->dq_flags & mask)
+		sync = 1;
+	spin_unlock(&dq_data_lock);
+	/* This is a slight hack but we can't afford getting global quota
+	 * lock if we already have a transaction started. */
+	if (!sync || journal_current_handle()) {
+		status = ocfs2_write_dquot(dquot);
+		goto out;
+	}
+	status = ocfs2_lock_global_qf(oinfo, 1);
+	if (status < 0)
+		goto out;
+	handle = ocfs2_start_trans(osb, OCFS2_QSYNC_CREDITS);
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		mlog_errno(status);
+		goto out_ilock;
+	}
+	mutex_lock(&sb_dqopt(sb)->dqio_mutex);
+	status = ocfs2_sync_dquot(dquot);
+	if (status < 0) {
+		mlog_errno(status);
+		goto out_dlock;
+	}
+	/* Now write updated local dquot structure */
+	status = ocfs2_local_write_dquot(dquot);
+out_dlock:
+	mutex_unlock(&sb_dqopt(sb)->dqio_mutex);
+	ocfs2_commit_trans(osb, handle);
+out_ilock:
+	ocfs2_unlock_global_qf(oinfo, 1);
+out:
+	if (status)
+		mlog_errno(status);
+	return status;
+}
+
+/* This should happen only after set_dqinfo(). */
+static int ocfs2_write_info(struct super_block *sb, int type)
+{
+	handle_t *handle;
+	int status = 0;
+	struct ocfs2_mem_dqinfo *oinfo = sb_dqinfo(sb, type)->dqi_priv;
+
+	status = ocfs2_lock_global_qf(oinfo, 1);
+	if (status < 0)
+		goto out;
+	handle = ocfs2_start_trans(OCFS2_SB(sb), OCFS2_QINFO_WRITE_CREDITS);
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		mlog_errno(status);
+		goto out_ilock;
+	}
+	status = dquot_commit_info(sb, type);
+	ocfs2_commit_trans(OCFS2_SB(sb), handle);
+out_ilock:
+	ocfs2_unlock_global_qf(oinfo, 1);
+out:
+	if (status)
+		mlog_errno(status);
+	return status;
+}
+
+static struct dquot *ocfs2_alloc_dquot(struct super_block *sb, int type)
+{
+	struct ocfs2_dquot *dquot =
+				kmem_cache_zalloc(ocfs2_dquot_cachep, GFP_NOFS);
+
+	if (!dquot)
+		return NULL;
+	return &dquot->dq_dquot;
+}
+
+static void ocfs2_destroy_dquot(struct dquot *dquot)
+{
+	kmem_cache_free(ocfs2_dquot_cachep, dquot);
+}
+
+const struct dquot_operations ocfs2_quota_operations = {
+	/* We never make dquot dirty so .write_dquot is never called */
+	.acquire_dquot	= ocfs2_acquire_dquot,
+	.release_dquot	= ocfs2_release_dquot,
+	.mark_dirty	= ocfs2_mark_dquot_dirty,
+	.write_info	= ocfs2_write_info,
+	.alloc_dquot	= ocfs2_alloc_dquot,
+	.destroy_dquot	= ocfs2_destroy_dquot,
+};

fs/ocfs2/refcounttree.h

@@ -0,0 +1,118 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * refcounttree.h
+ *
+ * Copyright (C) 2009 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+#ifndef OCFS2_REFCOUNTTREE_H
+#define OCFS2_REFCOUNTTREE_H
+
+struct ocfs2_refcount_tree {
+	struct rb_node rf_node;
+	u64 rf_blkno;
+	u32 rf_generation;
+	struct kref rf_getcnt;
+	struct rw_semaphore rf_sem;
+	struct ocfs2_lock_res rf_lockres;
+	int rf_removed;
+
+	/* the following 4 fields are used by caching_info. */
+	spinlock_t rf_lock;
+	struct ocfs2_caching_info rf_ci;
+	struct mutex rf_io_mutex;
+	struct super_block *rf_sb;
+};
+
+void ocfs2_purge_refcount_trees(struct ocfs2_super *osb);
+int ocfs2_lock_refcount_tree(struct ocfs2_super *osb, u64 ref_blkno, int rw,
+			     struct ocfs2_refcount_tree **tree,
+			     struct buffer_head **ref_bh);
+void ocfs2_unlock_refcount_tree(struct ocfs2_super *osb,
+				struct ocfs2_refcount_tree *tree,
+				int rw);
+
+int ocfs2_decrease_refcount(struct inode *inode,
+			    handle_t *handle, u32 cpos, u32 len,
+			    struct ocfs2_alloc_context *meta_ac,
+			    struct ocfs2_cached_dealloc_ctxt *dealloc,
+			    int delete);
+int ocfs2_prepare_refcount_change_for_del(struct inode *inode,
+					  u64 refcount_loc,
+					  u64 phys_blkno,
+					  u32 clusters,
+					  int *credits,
+					  int *ref_blocks);
+int ocfs2_refcount_cow(struct inode *inode,
+		       struct buffer_head *di_bh,
+		       u32 cpos, u32 write_len, u32 max_cpos);
+
+typedef int (ocfs2_post_refcount_func)(struct inode *inode,
+				       handle_t *handle,
+				       void *para);
+/*
+ * Some refcount caller need to do more work after we modify the data b-tree
+ * during refcount operation(including CoW and add refcount flag), and make the
+ * transaction complete. So it must give us this structure so that we can do it
+ * within our transaction.
+ *
+ */
+struct ocfs2_post_refcount {
+	int credits;			/* credits it need for journal. */
+	ocfs2_post_refcount_func *func;	/* real function. */
+	void *para;
+};
+
+int ocfs2_refcounted_xattr_delete_need(struct inode *inode,
+				       struct ocfs2_caching_info *ref_ci,
+				       struct buffer_head *ref_root_bh,
+				       struct ocfs2_xattr_value_root *xv,
+				       int *meta_add, int *credits);
+int ocfs2_refcount_cow_xattr(struct inode *inode,
+			     struct ocfs2_dinode *di,
+			     struct ocfs2_xattr_value_buf *vb,
+			     struct ocfs2_refcount_tree *ref_tree,
+			     struct buffer_head *ref_root_bh,
+			     u32 cpos, u32 write_len,
+			     struct ocfs2_post_refcount *post);
+int ocfs2_duplicate_clusters_by_page(handle_t *handle,
+				     struct inode *inode,
+				     u32 cpos, u32 old_cluster,
+				     u32 new_cluster, u32 new_len);
+int ocfs2_duplicate_clusters_by_jbd(handle_t *handle,
+				    struct inode *inode,
+				    u32 cpos, u32 old_cluster,
+				    u32 new_cluster, u32 new_len);
+int ocfs2_cow_sync_writeback(struct super_block *sb,
+			     struct inode *inode,
+			     u32 cpos, u32 num_clusters);
+int ocfs2_add_refcount_flag(struct inode *inode,
+			    struct ocfs2_extent_tree *data_et,
+			    struct ocfs2_caching_info *ref_ci,
+			    struct buffer_head *ref_root_bh,
+			    u32 cpos, u32 p_cluster, u32 num_clusters,
+			    struct ocfs2_cached_dealloc_ctxt *dealloc,
+			    struct ocfs2_post_refcount *post);
+int ocfs2_remove_refcount_tree(struct inode *inode, struct buffer_head *di_bh);
+int ocfs2_try_remove_refcount_tree(struct inode *inode,
+				   struct buffer_head *di_bh);
+int ocfs2_increase_refcount(handle_t *handle,
+			    struct ocfs2_caching_info *ci,
+			    struct buffer_head *ref_root_bh,
+			    u64 cpos, u32 len,
+			    struct ocfs2_alloc_context *meta_ac,
+			    struct ocfs2_cached_dealloc_ctxt *dealloc);
+int ocfs2_reflink_ioctl(struct inode *inode,
+			const char __user *oldname,
+			const char __user *newname,
+			bool preserve);
+#endif /* OCFS2_REFCOUNTTREE_H */

fs/ocfs2/reservations.c

@@ -0,0 +1,839 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * reservations.c
+ *
+ * Allocation reservations implementation
+ *
+ * Some code borrowed from fs/ext3/balloc.c and is:
+ *
+ * Copyright (C) 1992, 1993, 1994, 1995
+ * Remy Card (card@masi.ibp.fr)
+ * Laboratoire MASI - Institut Blaise Pascal
+ * Universite Pierre et Marie Curie (Paris VI)
+ *
+ * The rest is copyright (C) 2010 Novell.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/highmem.h>
+#include <linux/bitops.h>
+#include <linux/list.h>
+
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+#include "ocfs2_trace.h"
+
+#ifdef CONFIG_OCFS2_DEBUG_FS
+#define OCFS2_CHECK_RESERVATIONS
+#endif
+
+DEFINE_SPINLOCK(resv_lock);
+
+#define	OCFS2_MIN_RESV_WINDOW_BITS	8
+#define	OCFS2_MAX_RESV_WINDOW_BITS	1024
+
+int ocfs2_dir_resv_allowed(struct ocfs2_super *osb)
+{
+	return (osb->osb_resv_level && osb->osb_dir_resv_level);
+}
+
+static unsigned int ocfs2_resv_window_bits(struct ocfs2_reservation_map *resmap,
+					   struct ocfs2_alloc_reservation *resv)
+{
+	struct ocfs2_super *osb = resmap->m_osb;
+	unsigned int bits;
+
+	if (!(resv->r_flags & OCFS2_RESV_FLAG_DIR)) {
+		/* 8, 16, 32, 64, 128, 256, 512, 1024 */
+		bits = 4 << osb->osb_resv_level;
+	} else {
+		bits = 4 << osb->osb_dir_resv_level;
+	}
+	return bits;
+}
+
+static inline unsigned int ocfs2_resv_end(struct ocfs2_alloc_reservation *resv)
+{
+	if (resv->r_len)
+		return resv->r_start + resv->r_len - 1;
+	return resv->r_start;
+}
+
+static inline int ocfs2_resv_empty(struct ocfs2_alloc_reservation *resv)
+{
+	return !!(resv->r_len == 0);
+}
+
+static inline int ocfs2_resmap_disabled(struct ocfs2_reservation_map *resmap)
+{
+	if (resmap->m_osb->osb_resv_level == 0)
+		return 1;
+	return 0;
+}
+
+static void ocfs2_dump_resv(struct ocfs2_reservation_map *resmap)
+{
+	struct ocfs2_super *osb = resmap->m_osb;
+	struct rb_node *node;
+	struct ocfs2_alloc_reservation *resv;
+	int i = 0;
+
+	mlog(ML_NOTICE, "Dumping resmap for device %s. Bitmap length: %u\n",
+	     osb->dev_str, resmap->m_bitmap_len);
+
+	node = rb_first(&resmap->m_reservations);
+	while (node) {
+		resv = rb_entry(node, struct ocfs2_alloc_reservation, r_node);
+
+		mlog(ML_NOTICE, "start: %u\tend: %u\tlen: %u\tlast_start: %u"
+		     "\tlast_len: %u\n", resv->r_start,
+		     ocfs2_resv_end(resv), resv->r_len, resv->r_last_start,
+		     resv->r_last_len);
+
+		node = rb_next(node);
+		i++;
+	}
+
+	mlog(ML_NOTICE, "%d reservations found. LRU follows\n", i);
+
+	i = 0;
+	list_for_each_entry(resv, &resmap->m_lru, r_lru) {
+		mlog(ML_NOTICE, "LRU(%d) start: %u\tend: %u\tlen: %u\t"
+		     "last_start: %u\tlast_len: %u\n", i, resv->r_start,
+		     ocfs2_resv_end(resv), resv->r_len, resv->r_last_start,
+		     resv->r_last_len);
+
+		i++;
+	}
+}
+
+#ifdef OCFS2_CHECK_RESERVATIONS
+static int ocfs2_validate_resmap_bits(struct ocfs2_reservation_map *resmap,
+				      int i,
+				      struct ocfs2_alloc_reservation *resv)
+{
+	char *disk_bitmap = resmap->m_disk_bitmap;
+	unsigned int start = resv->r_start;
+	unsigned int end = ocfs2_resv_end(resv);
+
+	while (start <= end) {
+		if (ocfs2_test_bit(start, disk_bitmap)) {
+			mlog(ML_ERROR,
+			     "reservation %d covers an allocated area "
+			     "starting at bit %u!\n", i, start);
+			return 1;
+		}
+
+		start++;
+	}
+	return 0;
+}
+
+static void ocfs2_check_resmap(struct ocfs2_reservation_map *resmap)
+{
+	unsigned int off = 0;
+	int i = 0;
+	struct rb_node *node;
+	struct ocfs2_alloc_reservation *resv;
+
+	node = rb_first(&resmap->m_reservations);
+	while (node) {
+		resv = rb_entry(node, struct ocfs2_alloc_reservation, r_node);
+
+		if (i > 0 && resv->r_start <= off) {
+			mlog(ML_ERROR, "reservation %d has bad start off!\n",
+			     i);
+			goto bad;
+		}
+
+		if (resv->r_len == 0) {
+			mlog(ML_ERROR, "reservation %d has no length!\n",
+			     i);
+			goto bad;
+		}
+
+		if (resv->r_start > ocfs2_resv_end(resv)) {
+			mlog(ML_ERROR, "reservation %d has invalid range!\n",
+			     i);
+			goto bad;
+		}
+
+		if (ocfs2_resv_end(resv) >= resmap->m_bitmap_len) {
+			mlog(ML_ERROR, "reservation %d extends past bitmap!\n",
+			     i);
+			goto bad;
+		}
+
+		if (ocfs2_validate_resmap_bits(resmap, i, resv))
+			goto bad;
+
+		off = ocfs2_resv_end(resv);
+		node = rb_next(node);
+
+		i++;
+	}
+	return;
+
+bad:
+	ocfs2_dump_resv(resmap);
+	BUG();
+}
+#else
+static inline void ocfs2_check_resmap(struct ocfs2_reservation_map *resmap)
+{
+
+}
+#endif
+
+void ocfs2_resv_init_once(struct ocfs2_alloc_reservation *resv)
+{
+	memset(resv, 0, sizeof(*resv));
+	INIT_LIST_HEAD(&resv->r_lru);
+}
+
+void ocfs2_resv_set_type(struct ocfs2_alloc_reservation *resv,
+			 unsigned int flags)
+{
+	BUG_ON(flags & ~OCFS2_RESV_TYPES);
+
+	resv->r_flags |= flags;
+}
+
+int ocfs2_resmap_init(struct ocfs2_super *osb,
+		      struct ocfs2_reservation_map *resmap)
+{
+	memset(resmap, 0, sizeof(*resmap));
+
+	resmap->m_osb = osb;
+	resmap->m_reservations = RB_ROOT;
+	/* m_bitmap_len is initialized to zero by the above memset. */
+	INIT_LIST_HEAD(&resmap->m_lru);
+
+	return 0;
+}
+
+static void ocfs2_resv_mark_lru(struct ocfs2_reservation_map *resmap,
+				struct ocfs2_alloc_reservation *resv)
+{
+	assert_spin_locked(&resv_lock);
+
+	if (!list_empty(&resv->r_lru))
+		list_del_init(&resv->r_lru);
+
+	list_add_tail(&resv->r_lru, &resmap->m_lru);
+}
+
+static void __ocfs2_resv_trunc(struct ocfs2_alloc_reservation *resv)
+{
+	resv->r_len = 0;
+	resv->r_start = 0;
+}
+
+static void ocfs2_resv_remove(struct ocfs2_reservation_map *resmap,
+			      struct ocfs2_alloc_reservation *resv)
+{
+	if (resv->r_flags & OCFS2_RESV_FLAG_INUSE) {
+		list_del_init(&resv->r_lru);
+		rb_erase(&resv->r_node, &resmap->m_reservations);
+		resv->r_flags &= ~OCFS2_RESV_FLAG_INUSE;
+	}
+}
+
+static void __ocfs2_resv_discard(struct ocfs2_reservation_map *resmap,
+				 struct ocfs2_alloc_reservation *resv)
+{
+	assert_spin_locked(&resv_lock);
+
+	__ocfs2_resv_trunc(resv);
+	/*
+	 * last_len and last_start no longer make sense if
+	 * we're changing the range of our allocations.
+	 */
+	resv->r_last_len = resv->r_last_start = 0;
+
+	ocfs2_resv_remove(resmap, resv);
+}
+
+/* does nothing if 'resv' is null */
+void ocfs2_resv_discard(struct ocfs2_reservation_map *resmap,
+			struct ocfs2_alloc_reservation *resv)
+{
+	if (resv) {
+		spin_lock(&resv_lock);
+		__ocfs2_resv_discard(resmap, resv);
+		spin_unlock(&resv_lock);
+	}
+}
+
+static void ocfs2_resmap_clear_all_resv(struct ocfs2_reservation_map *resmap)
+{
+	struct rb_node *node;
+	struct ocfs2_alloc_reservation *resv;
+
+	assert_spin_locked(&resv_lock);
+
+	while ((node = rb_last(&resmap->m_reservations)) != NULL) {
+		resv = rb_entry(node, struct ocfs2_alloc_reservation, r_node);
+
+		__ocfs2_resv_discard(resmap, resv);
+	}
+}
+
+void ocfs2_resmap_restart(struct ocfs2_reservation_map *resmap,
+			  unsigned int clen, char *disk_bitmap)
+{
+	if (ocfs2_resmap_disabled(resmap))
+		return;
+
+	spin_lock(&resv_lock);
+
+	ocfs2_resmap_clear_all_resv(resmap);
+	resmap->m_bitmap_len = clen;
+	resmap->m_disk_bitmap = disk_bitmap;
+
+	spin_unlock(&resv_lock);
+}
+
+void ocfs2_resmap_uninit(struct ocfs2_reservation_map *resmap)
+{
+	/* Does nothing for now. Keep this around for API symmetry */
+}
+
+static void ocfs2_resv_insert(struct ocfs2_reservation_map *resmap,
+			      struct ocfs2_alloc_reservation *new)
+{
+	struct rb_root *root = &resmap->m_reservations;
+	struct rb_node *parent = NULL;
+	struct rb_node **p = &root->rb_node;
+	struct ocfs2_alloc_reservation *tmp;
+
+	assert_spin_locked(&resv_lock);
+
+	trace_ocfs2_resv_insert(new->r_start, new->r_len);
+
+	while (*p) {
+		parent = *p;
+
+		tmp = rb_entry(parent, struct ocfs2_alloc_reservation, r_node);
+
+		if (new->r_start < tmp->r_start) {
+			p = &(*p)->rb_left;
+
+			/*
+			 * This is a good place to check for
+			 * overlapping reservations.
+			 */
+			BUG_ON(ocfs2_resv_end(new) >= tmp->r_start);
+		} else if (new->r_start > ocfs2_resv_end(tmp)) {
+			p = &(*p)->rb_right;
+		} else {
+			/* This should never happen! */
+			mlog(ML_ERROR, "Duplicate reservation window!\n");
+			BUG();
+		}
+	}
+
+	rb_link_node(&new->r_node, parent, p);
+	rb_insert_color(&new->r_node, root);
+	new->r_flags |= OCFS2_RESV_FLAG_INUSE;
+
+	ocfs2_resv_mark_lru(resmap, new);
+
+	ocfs2_check_resmap(resmap);
+}
+
+/**
+ * ocfs2_find_resv_lhs() - find the window which contains goal
+ * @resmap: reservation map to search
+ * @goal: which bit to search for
+ *
+ * If a window containing that goal is not found, we return the window
+ * which comes before goal. Returns NULL on empty rbtree or no window
+ * before goal.
+ */
+static struct ocfs2_alloc_reservation *
+ocfs2_find_resv_lhs(struct ocfs2_reservation_map *resmap, unsigned int goal)
+{
+	struct ocfs2_alloc_reservation *resv = NULL;
+	struct ocfs2_alloc_reservation *prev_resv = NULL;
+	struct rb_node *node = resmap->m_reservations.rb_node;
+
+	assert_spin_locked(&resv_lock);
+
+	if (!node)
+		return NULL;
+
+	node = rb_first(&resmap->m_reservations);
+	while (node) {
+		resv = rb_entry(node, struct ocfs2_alloc_reservation, r_node);
+
+		if (resv->r_start <= goal && ocfs2_resv_end(resv) >= goal)
+			break;
+
+		/* Check if we overshot the reservation just before goal? */
+		if (resv->r_start > goal) {
+			resv = prev_resv;
+			break;
+		}
+
+		prev_resv = resv;
+		node = rb_next(node);
+	}
+
+	return resv;
+}
+
+/*
+ * We are given a range within the bitmap, which corresponds to a gap
+ * inside the reservations tree (search_start, search_len). The range
+ * can be anything from the whole bitmap, to a gap between
+ * reservations.
+ *
+ * The start value of *rstart is insignificant.
+ *
+ * This function searches the bitmap range starting at search_start
+ * with length search_len for a set of contiguous free bits. We try
+ * to find up to 'wanted' bits, but can sometimes return less.
+ *
+ * Returns the length of allocation, 0 if no free bits are found.
+ *
+ * *cstart and *clen will also be populated with the result.
+ */
+static int ocfs2_resmap_find_free_bits(struct ocfs2_reservation_map *resmap,
+				       unsigned int wanted,
+				       unsigned int search_start,
+				       unsigned int search_len,
+				       unsigned int *rstart,
+				       unsigned int *rlen)
+{
+	void *bitmap = resmap->m_disk_bitmap;
+	unsigned int best_start, best_len = 0;
+	int offset, start, found;
+
+	trace_ocfs2_resmap_find_free_bits_begin(search_start, search_len,
+						wanted, resmap->m_bitmap_len);
+
+	found = best_start = best_len = 0;
+
+	start = search_start;
+	while ((offset = ocfs2_find_next_zero_bit(bitmap, resmap->m_bitmap_len,
+						 start)) != -1) {
+		/* Search reached end of the region */
+		if (offset >= (search_start + search_len))
+			break;
+
+		if (offset == start) {
+			/* we found a zero */
+			found++;
+			/* move start to the next bit to test */
+			start++;
+		} else {
+			/* got a zero after some ones */
+			found = 1;
+			start = offset + 1;
+		}
+		if (found > best_len) {
+			best_len = found;
+			best_start = start - found;
+		}
+
+		if (found >= wanted)
+			break;
+	}
+
+	if (best_len == 0)
+		return 0;
+
+	if (best_len >= wanted)
+		best_len = wanted;
+
+	*rlen = best_len;
+	*rstart = best_start;
+
+	trace_ocfs2_resmap_find_free_bits_end(best_start, best_len);
+
+	return *rlen;
+}
+
+static void __ocfs2_resv_find_window(struct ocfs2_reservation_map *resmap,
+				     struct ocfs2_alloc_reservation *resv,
+				     unsigned int goal, unsigned int wanted)
+{
+	struct rb_root *root = &resmap->m_reservations;
+	unsigned int gap_start, gap_end, gap_len;
+	struct ocfs2_alloc_reservation *prev_resv, *next_resv;
+	struct rb_node *prev, *next;
+	unsigned int cstart, clen;
+	unsigned int best_start = 0, best_len = 0;
+
+	/*
+	 * Nasty cases to consider:
+	 *
+	 * - rbtree is empty
+	 * - our window should be first in all reservations
+	 * - our window should be last in all reservations
+	 * - need to make sure we don't go past end of bitmap
+	 */
+	trace_ocfs2_resv_find_window_begin(resv->r_start, ocfs2_resv_end(resv),
+					   goal, wanted, RB_EMPTY_ROOT(root));
+
+	assert_spin_locked(&resv_lock);
+
+	if (RB_EMPTY_ROOT(root)) {
+		/*
+		 * Easiest case - empty tree. We can just take
+		 * whatever window of free bits we want.
+		 */
+		clen = ocfs2_resmap_find_free_bits(resmap, wanted, goal,
+						   resmap->m_bitmap_len - goal,
+						   &cstart, &clen);
+
+		/*
+		 * This should never happen - the local alloc window
+		 * will always have free bits when we're called.
+		 */
+		BUG_ON(goal == 0 && clen == 0);
+
+		if (clen == 0)
+			return;
+
+		resv->r_start = cstart;
+		resv->r_len = clen;
+
+		ocfs2_resv_insert(resmap, resv);
+		return;
+	}
+
+	prev_resv = ocfs2_find_resv_lhs(resmap, goal);
+
+	if (prev_resv == NULL) {
+		/*
+		 * A NULL here means that the search code couldn't
+		 * find a window that starts before goal.
+		 *
+		 * However, we can take the first window after goal,
+		 * which is also by definition, the leftmost window in
+		 * the entire tree. If we can find free bits in the
+		 * gap between goal and the LHS window, then the
+		 * reservation can safely be placed there.
+		 *
+		 * Otherwise we fall back to a linear search, checking
+		 * the gaps in between windows for a place to
+		 * allocate.
+		 */
+
+		next = rb_first(root);
+		next_resv = rb_entry(next, struct ocfs2_alloc_reservation,
+				     r_node);
+
+		/*
+		 * The search should never return such a window. (see
+		 * comment above
+		 */
+		if (next_resv->r_start <= goal) {
+			mlog(ML_ERROR, "goal: %u next_resv: start %u len %u\n",
+			     goal, next_resv->r_start, next_resv->r_len);
+			ocfs2_dump_resv(resmap);
+			BUG();
+		}
+
+		clen = ocfs2_resmap_find_free_bits(resmap, wanted, goal,
+						   next_resv->r_start - goal,
+						   &cstart, &clen);
+		if (clen) {
+			best_len = clen;
+			best_start = cstart;
+			if (best_len == wanted)
+				goto out_insert;
+		}
+
+		prev_resv = next_resv;
+		next_resv = NULL;
+	}
+
+	trace_ocfs2_resv_find_window_prev(prev_resv->r_start,
+					  ocfs2_resv_end(prev_resv));
+
+	prev = &prev_resv->r_node;
+
+	/* Now we do a linear search for a window, starting at 'prev_rsv' */
+	while (1) {
+		next = rb_next(prev);
+		if (next) {
+			next_resv = rb_entry(next,
+					     struct ocfs2_alloc_reservation,
+					     r_node);
+
+			gap_start = ocfs2_resv_end(prev_resv) + 1;
+			gap_end = next_resv->r_start - 1;
+			gap_len = gap_end - gap_start + 1;
+		} else {
+			/*
+			 * We're at the rightmost edge of the
+			 * tree. See if a reservation between this
+			 * window and the end of the bitmap will work.
+			 */
+			gap_start = ocfs2_resv_end(prev_resv) + 1;
+			gap_len = resmap->m_bitmap_len - gap_start;
+			gap_end = resmap->m_bitmap_len - 1;
+		}
+
+		trace_ocfs2_resv_find_window_next(next ? next_resv->r_start: -1,
+					next ? ocfs2_resv_end(next_resv) : -1);
+		/*
+		 * No need to check this gap if we have already found
+		 * a larger region of free bits.
+		 */
+		if (gap_len <= best_len)
+			goto next_resv;
+
+		clen = ocfs2_resmap_find_free_bits(resmap, wanted, gap_start,
+						   gap_len, &cstart, &clen);
+		if (clen == wanted) {
+			best_len = clen;
+			best_start = cstart;
+			goto out_insert;
+		} else if (clen > best_len) {
+			best_len = clen;
+			best_start = cstart;
+		}
+
+next_resv:
+		if (!next)
+			break;
+
+		prev = next;
+		prev_resv = rb_entry(prev, struct ocfs2_alloc_reservation,
+				     r_node);
+	}
+
+out_insert:
+	if (best_len) {
+		resv->r_start = best_start;
+		resv->r_len = best_len;
+		ocfs2_resv_insert(resmap, resv);
+	}
+}
+
+static void ocfs2_cannibalize_resv(struct ocfs2_reservation_map *resmap,
+				   struct ocfs2_alloc_reservation *resv,
+				   unsigned int wanted)
+{
+	struct ocfs2_alloc_reservation *lru_resv;
+	int tmpwindow = !!(resv->r_flags & OCFS2_RESV_FLAG_TMP);
+	unsigned int min_bits;
+
+	if (!tmpwindow)
+		min_bits = ocfs2_resv_window_bits(resmap, resv) >> 1;
+	else
+		min_bits = wanted; /* We at know the temp window will use all
+				    * of these bits */
+
+	/*
+	 * Take the first reservation off the LRU as our 'target'. We
+	 * don't try to be smart about it. There might be a case for
+	 * searching based on size but I don't have enough data to be
+	 * sure. --Mark (3/16/2010)
+	 */
+	lru_resv = list_first_entry(&resmap->m_lru,
+				    struct ocfs2_alloc_reservation, r_lru);
+
+	trace_ocfs2_cannibalize_resv_begin(lru_resv->r_start,
+					   lru_resv->r_len,
+					   ocfs2_resv_end(lru_resv));
+
+	/*
+	 * Cannibalize (some or all) of the target reservation and
+	 * feed it to the current window.
+	 */
+	if (lru_resv->r_len <= min_bits) {
+		/*
+		 * Discard completely if size is less than or equal to a
+		 * reasonable threshold - 50% of window bits for non temporary
+		 * windows.
+		 */
+		resv->r_start = lru_resv->r_start;
+		resv->r_len = lru_resv->r_len;
+
+		__ocfs2_resv_discard(resmap, lru_resv);
+	} else {
+		unsigned int shrink;
+		if (tmpwindow)
+			shrink = min_bits;
+		else
+			shrink = lru_resv->r_len / 2;
+
+		lru_resv->r_len -= shrink;
+
+		resv->r_start = ocfs2_resv_end(lru_resv) + 1;
+		resv->r_len = shrink;
+	}
+
+	trace_ocfs2_cannibalize_resv_end(resv->r_start, ocfs2_resv_end(resv),
+					 resv->r_len, resv->r_last_start,
+					 resv->r_last_len);
+
+	ocfs2_resv_insert(resmap, resv);
+}
+
+static void ocfs2_resv_find_window(struct ocfs2_reservation_map *resmap,
+				   struct ocfs2_alloc_reservation *resv,
+				   unsigned int wanted)
+{
+	unsigned int goal = 0;
+
+	BUG_ON(!ocfs2_resv_empty(resv));
+
+	/*
+	 * Begin by trying to get a window as close to the previous
+	 * one as possible. Using the most recent allocation as a
+	 * start goal makes sense.
+	 */
+	if (resv->r_last_len) {
+		goal = resv->r_last_start + resv->r_last_len;
+		if (goal >= resmap->m_bitmap_len)
+			goal = 0;
+	}
+
+	__ocfs2_resv_find_window(resmap, resv, goal, wanted);
+
+	/* Search from last alloc didn't work, try once more from beginning. */
+	if (ocfs2_resv_empty(resv) && goal != 0)
+		__ocfs2_resv_find_window(resmap, resv, 0, wanted);
+
+	if (ocfs2_resv_empty(resv)) {
+		/*
+		 * Still empty? Pull oldest one off the LRU, remove it from
+		 * tree, put this one in it's place.
+		 */
+		ocfs2_cannibalize_resv(resmap, resv, wanted);
+	}
+
+	BUG_ON(ocfs2_resv_empty(resv));
+}
+
+int ocfs2_resmap_resv_bits(struct ocfs2_reservation_map *resmap,
+			   struct ocfs2_alloc_reservation *resv,
+			   int *cstart, int *clen)
+{
+	if (resv == NULL || ocfs2_resmap_disabled(resmap))
+		return -ENOSPC;
+
+	spin_lock(&resv_lock);
+
+	if (ocfs2_resv_empty(resv)) {
+		/*
+		 * We don't want to over-allocate for temporary
+		 * windows. Otherwise, we run the risk of fragmenting the
+		 * allocation space.
+		 */
+		unsigned int wanted = ocfs2_resv_window_bits(resmap, resv);
+
+		if ((resv->r_flags & OCFS2_RESV_FLAG_TMP) || wanted < *clen)
+			wanted = *clen;
+
+		/*
+		 * Try to get a window here. If it works, we must fall
+		 * through and test the bitmap . This avoids some
+		 * ping-ponging of windows due to non-reserved space
+		 * being allocation before we initialize a window for
+		 * that inode.
+		 */
+		ocfs2_resv_find_window(resmap, resv, wanted);
+		trace_ocfs2_resmap_resv_bits(resv->r_start, resv->r_len);
+	}
+
+	BUG_ON(ocfs2_resv_empty(resv));
+
+	*cstart = resv->r_start;
+	*clen = resv->r_len;
+
+	spin_unlock(&resv_lock);
+	return 0;
+}
+
+static void
+	ocfs2_adjust_resv_from_alloc(struct ocfs2_reservation_map *resmap,
+				     struct ocfs2_alloc_reservation *resv,
+				     unsigned int start, unsigned int end)
+{
+	unsigned int rhs = 0;
+	unsigned int old_end = ocfs2_resv_end(resv);
+
+	BUG_ON(start != resv->r_start || old_end < end);
+
+	/*
+	 * Completely used? We can remove it then.
+	 */
+	if (old_end == end) {
+		__ocfs2_resv_discard(resmap, resv);
+		return;
+	}
+
+	rhs = old_end - end;
+
+	/*
+	 * This should have been trapped above.
+	 */
+	BUG_ON(rhs == 0);
+
+	resv->r_start = end + 1;
+	resv->r_len = old_end - resv->r_start + 1;
+}
+
+void ocfs2_resmap_claimed_bits(struct ocfs2_reservation_map *resmap,
+			       struct ocfs2_alloc_reservation *resv,
+			       u32 cstart, u32 clen)
+{
+	unsigned int cend = cstart + clen - 1;
+
+	if (resmap == NULL || ocfs2_resmap_disabled(resmap))
+		return;
+
+	if (resv == NULL)
+		return;
+
+	BUG_ON(cstart != resv->r_start);
+
+	spin_lock(&resv_lock);
+
+	trace_ocfs2_resmap_claimed_bits_begin(cstart, cend, clen, resv->r_start,
+					      ocfs2_resv_end(resv), resv->r_len,
+					      resv->r_last_start,
+					      resv->r_last_len);
+
+	BUG_ON(cstart < resv->r_start);
+	BUG_ON(cstart > ocfs2_resv_end(resv));
+	BUG_ON(cend > ocfs2_resv_end(resv));
+
+	ocfs2_adjust_resv_from_alloc(resmap, resv, cstart, cend);
+	resv->r_last_start = cstart;
+	resv->r_last_len = clen;
+
+	/*
+	 * May have been discarded above from
+	 * ocfs2_adjust_resv_from_alloc().
+	 */
+	if (!ocfs2_resv_empty(resv))
+		ocfs2_resv_mark_lru(resmap, resv);
+
+	trace_ocfs2_resmap_claimed_bits_end(resv->r_start, ocfs2_resv_end(resv),
+					    resv->r_len, resv->r_last_start,
+					    resv->r_last_len);
+
+	ocfs2_check_resmap(resmap);
+
+	spin_unlock(&resv_lock);
+}

fs/ocfs2/reservations.h

@@ -0,0 +1,159 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * reservations.h
+ *
+ * Allocation reservations function prototypes and structures.
+ *
+ * Copyright (C) 2010 Novell.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#ifndef	OCFS2_RESERVATIONS_H
+#define	OCFS2_RESERVATIONS_H
+
+#include <linux/rbtree.h>
+
+#define OCFS2_DEFAULT_RESV_LEVEL	2
+#define OCFS2_MAX_RESV_LEVEL	9
+#define OCFS2_MIN_RESV_LEVEL	0
+
+struct ocfs2_alloc_reservation {
+	struct rb_node	r_node;
+
+	unsigned int	r_start;	/* Beginning of current window */
+	unsigned int	r_len;		/* Length of the window */
+
+	unsigned int	r_last_len;	/* Length of most recent alloc */
+	unsigned int	r_last_start;	/* Start of most recent alloc */
+	struct list_head	r_lru;	/* LRU list head */
+
+	unsigned int	r_flags;
+};
+
+#define	OCFS2_RESV_FLAG_INUSE	0x01	/* Set when r_node is part of a btree */
+#define	OCFS2_RESV_FLAG_TMP	0x02	/* Temporary reservation, will be
+					 * destroyed immedately after use */
+#define	OCFS2_RESV_FLAG_DIR	0x04	/* Reservation is for an unindexed
+					 * directory btree */
+
+struct ocfs2_reservation_map {
+	struct rb_root		m_reservations;
+	char			*m_disk_bitmap;
+
+	struct ocfs2_super	*m_osb;
+
+	/* The following are not initialized to meaningful values until a disk
+	 * bitmap is provided. */
+	u32			m_bitmap_len;	/* Number of valid
+						 * bits available */
+
+	struct list_head	m_lru;		/* LRU of reservations
+						 * structures. */
+
+};
+
+void ocfs2_resv_init_once(struct ocfs2_alloc_reservation *resv);
+
+#define OCFS2_RESV_TYPES	(OCFS2_RESV_FLAG_TMP|OCFS2_RESV_FLAG_DIR)
+void ocfs2_resv_set_type(struct ocfs2_alloc_reservation *resv,
+			 unsigned int flags);
+
+int ocfs2_dir_resv_allowed(struct ocfs2_super *osb);
+
+/**
+ * ocfs2_resv_discard() - truncate a reservation
+ * @resmap:
+ * @resv: the reservation to truncate.
+ *
+ * After this function is called, the reservation will be empty, and
+ * unlinked from the rbtree.
+ */
+void ocfs2_resv_discard(struct ocfs2_reservation_map *resmap,
+			struct ocfs2_alloc_reservation *resv);
+
+
+/**
+ * ocfs2_resmap_init() - Initialize fields of a reservations bitmap
+ * @resmap: struct ocfs2_reservation_map to initialize
+ * @obj: unused for now
+ * @ops: unused for now
+ * @max_bitmap_bytes: Maximum size of the bitmap (typically blocksize)
+ *
+ * Only possible return value other than '0' is -ENOMEM for failure to
+ * allocation mirror bitmap.
+ */
+int ocfs2_resmap_init(struct ocfs2_super *osb,
+		      struct ocfs2_reservation_map *resmap);
+
+/**
+ * ocfs2_resmap_restart() - "restart" a reservation bitmap
+ * @resmap: reservations bitmap
+ * @clen: Number of valid bits in the bitmap
+ * @disk_bitmap: the disk bitmap this resmap should refer to.
+ *
+ * Re-initialize the parameters of a reservation bitmap. This is
+ * useful for local alloc window slides.
+ *
+ * This function will call ocfs2_trunc_resv against all existing
+ * reservations. A future version will recalculate existing
+ * reservations based on the new bitmap.
+ */
+void ocfs2_resmap_restart(struct ocfs2_reservation_map *resmap,
+			  unsigned int clen, char *disk_bitmap);
+
+/**
+ * ocfs2_resmap_uninit() - uninitialize a reservation bitmap structure
+ * @resmap: the struct ocfs2_reservation_map to uninitialize
+ */
+void ocfs2_resmap_uninit(struct ocfs2_reservation_map *resmap);
+
+/**
+ * ocfs2_resmap_resv_bits() - Return still-valid reservation bits
+ * @resmap: reservations bitmap
+ * @resv: reservation to base search from
+ * @cstart: start of proposed allocation
+ * @clen: length (in clusters) of proposed allocation
+ *
+ * Using the reservation data from resv, this function will compare
+ * resmap and resmap->m_disk_bitmap to determine what part (if any) of
+ * the reservation window is still clear to use. If resv is empty,
+ * this function will try to allocate a window for it.
+ *
+ * On success, zero is returned and the valid allocation area is set in cstart
+ * and clen.
+ *
+ * Returns -ENOSPC if reservations are disabled.
+ */
+int ocfs2_resmap_resv_bits(struct ocfs2_reservation_map *resmap,
+			   struct ocfs2_alloc_reservation *resv,
+			   int *cstart, int *clen);
+
+/**
+ * ocfs2_resmap_claimed_bits() - Tell the reservation code that bits were used.
+ * @resmap: reservations bitmap
+ * @resv: optional reservation to recalulate based on new bitmap
+ * @cstart: start of allocation in clusters
+ * @clen: end of allocation in clusters.
+ *
+ * Tell the reservation code that bits were used to fulfill allocation in
+ * resmap. The bits don't have to have been part of any existing
+ * reservation. But we must always call this function when bits are claimed.
+ * Internally, the reservations code will use this information to mark the
+ * reservations bitmap. If resv is passed, it's next allocation window will be
+ * calculated. It also expects that 'cstart' is the same as we passed back
+ * from ocfs2_resmap_resv_bits().
+ */
+void ocfs2_resmap_claimed_bits(struct ocfs2_reservation_map *resmap,
+			       struct ocfs2_alloc_reservation *resv,
+			       u32 cstart, u32 clen);
+
+#endif	/* OCFS2_RESERVATIONS_H */

fs/ocfs2/resize.c

@@ -0,0 +1,589 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * resize.c
+ *
+ * volume resize.
+ * Inspired by ext3/resize.c.
+ *
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "dlmglue.h"
+#include "inode.h"
+#include "journal.h"
+#include "super.h"
+#include "sysfile.h"
+#include "uptodate.h"
+#include "ocfs2_trace.h"
+
+#include "buffer_head_io.h"
+#include "suballoc.h"
+#include "resize.h"
+
+/*
+ * Check whether there are new backup superblocks exist
+ * in the last group. If there are some, mark them or clear
+ * them in the bitmap.
+ *
+ * Return how many backups we find in the last group.
+ */
+static u16 ocfs2_calc_new_backup_super(struct inode *inode,
+				       struct ocfs2_group_desc *gd,
+				       u16 cl_cpg,
+				       int set)
+{
+	int i;
+	u16 backups = 0;
+	u32 cluster;
+	u64 blkno, gd_blkno, lgd_blkno = le64_to_cpu(gd->bg_blkno);
+
+	for (i = 0; i < OCFS2_MAX_BACKUP_SUPERBLOCKS; i++) {
+		blkno = ocfs2_backup_super_blkno(inode->i_sb, i);
+		cluster = ocfs2_blocks_to_clusters(inode->i_sb, blkno);
+
+		gd_blkno = ocfs2_which_cluster_group(inode, cluster);
+		if (gd_blkno < lgd_blkno)
+			continue;
+		else if (gd_blkno > lgd_blkno)
+			break;
+
+		if (set)
+			ocfs2_set_bit(cluster % cl_cpg,
+				      (unsigned long *)gd->bg_bitmap);
+		else
+			ocfs2_clear_bit(cluster % cl_cpg,
+					(unsigned long *)gd->bg_bitmap);
+		backups++;
+	}
+
+	return backups;
+}
+
+static int ocfs2_update_last_group_and_inode(handle_t *handle,
+					     struct inode *bm_inode,
+					     struct buffer_head *bm_bh,
+					     struct buffer_head *group_bh,
+					     u32 first_new_cluster,
+					     int new_clusters)
+{
+	int ret = 0;
+	struct ocfs2_super *osb = OCFS2_SB(bm_inode->i_sb);
+	struct ocfs2_dinode *fe = (struct ocfs2_dinode *) bm_bh->b_data;
+	struct ocfs2_chain_list *cl = &fe->id2.i_chain;
+	struct ocfs2_chain_rec *cr;
+	struct ocfs2_group_desc *group;
+	u16 chain, num_bits, backups = 0;
+	u16 cl_bpc = le16_to_cpu(cl->cl_bpc);
+	u16 cl_cpg = le16_to_cpu(cl->cl_cpg);
+
+	trace_ocfs2_update_last_group_and_inode(new_clusters,
+						first_new_cluster);
+
+	ret = ocfs2_journal_access_gd(handle, INODE_CACHE(bm_inode),
+				      group_bh, OCFS2_JOURNAL_ACCESS_WRITE);
+	if (ret < 0) {
+		mlog_errno(ret);
+		goto out;
+	}
+
+	group = (struct ocfs2_group_desc *)group_bh->b_data;
+
+	/* update the group first. */
+	num_bits = new_clusters * cl_bpc;
+	le16_add_cpu(&group->bg_bits, num_bits);
+	le16_add_cpu(&group->bg_free_bits_count, num_bits);
+
+	/*
+	 * check whether there are some new backup superblocks exist in
+	 * this group and update the group bitmap accordingly.
+	 */
+	if (OCFS2_HAS_COMPAT_FEATURE(osb->sb,
+				     OCFS2_FEATURE_COMPAT_BACKUP_SB)) {
+		backups = ocfs2_calc_new_backup_super(bm_inode,
+						     group,
+						     cl_cpg, 1);
+		le16_add_cpu(&group->bg_free_bits_count, -1 * backups);
+	}
+
+	ocfs2_journal_dirty(handle, group_bh);
+
+	/* update the inode accordingly. */
+	ret = ocfs2_journal_access_di(handle, INODE_CACHE(bm_inode), bm_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (ret < 0) {
+		mlog_errno(ret);
+		goto out_rollback;
+	}
+
+	chain = le16_to_cpu(group->bg_chain);
+	cr = (&cl->cl_recs[chain]);
+	le32_add_cpu(&cr->c_total, num_bits);
+	le32_add_cpu(&cr->c_free, num_bits);
+	le32_add_cpu(&fe->id1.bitmap1.i_total, num_bits);
+	le32_add_cpu(&fe->i_clusters, new_clusters);
+
+	if (backups) {
+		le32_add_cpu(&cr->c_free, -1 * backups);
+		le32_add_cpu(&fe->id1.bitmap1.i_used, backups);
+	}
+
+	spin_lock(&OCFS2_I(bm_inode)->ip_lock);
+	OCFS2_I(bm_inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
+	le64_add_cpu(&fe->i_size, (u64)new_clusters << osb->s_clustersize_bits);
+	spin_unlock(&OCFS2_I(bm_inode)->ip_lock);
+	i_size_write(bm_inode, le64_to_cpu(fe->i_size));
+
+	ocfs2_journal_dirty(handle, bm_bh);
+
+out_rollback:
+	if (ret < 0) {
+		ocfs2_calc_new_backup_super(bm_inode,
+					    group,
+					    cl_cpg, 0);
+		le16_add_cpu(&group->bg_free_bits_count, backups);
+		le16_add_cpu(&group->bg_bits, -1 * num_bits);
+		le16_add_cpu(&group->bg_free_bits_count, -1 * num_bits);
+	}
+out:
+	if (ret)
+		mlog_errno(ret);
+	return ret;
+}
+
+static int update_backups(struct inode * inode, u32 clusters, char *data)
+{
+	int i, ret = 0;
+	u32 cluster;
+	u64 blkno;
+	struct buffer_head *backup = NULL;
+	struct ocfs2_dinode *backup_di = NULL;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+	/* calculate the real backups we need to update. */
+	for (i = 0; i < OCFS2_MAX_BACKUP_SUPERBLOCKS; i++) {
+		blkno = ocfs2_backup_super_blkno(inode->i_sb, i);
+		cluster = ocfs2_blocks_to_clusters(inode->i_sb, blkno);
+		if (cluster > clusters)
+			break;
+
+		ret = ocfs2_read_blocks_sync(osb, blkno, 1, &backup);
+		if (ret < 0) {
+			mlog_errno(ret);
+			break;
+		}
+
+		memcpy(backup->b_data, data, inode->i_sb->s_blocksize);
+
+		backup_di = (struct ocfs2_dinode *)backup->b_data;
+		backup_di->i_blkno = cpu_to_le64(blkno);
+
+		ret = ocfs2_write_super_or_backup(osb, backup);
+		brelse(backup);
+		backup = NULL;
+		if (ret < 0) {
+			mlog_errno(ret);
+			break;
+		}
+	}
+
+	return ret;
+}
+
+static void ocfs2_update_super_and_backups(struct inode *inode,
+					   int new_clusters)
+{
+	int ret;
+	u32 clusters = 0;
+	struct buffer_head *super_bh = NULL;
+	struct ocfs2_dinode *super_di = NULL;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+	/*
+	 * update the superblock last.
+	 * It doesn't matter if the write failed.
+	 */
+	ret = ocfs2_read_blocks_sync(osb, OCFS2_SUPER_BLOCK_BLKNO, 1,
+				     &super_bh);
+	if (ret < 0) {
+		mlog_errno(ret);
+		goto out;
+	}
+
+	super_di = (struct ocfs2_dinode *)super_bh->b_data;
+	le32_add_cpu(&super_di->i_clusters, new_clusters);
+	clusters = le32_to_cpu(super_di->i_clusters);
+
+	ret = ocfs2_write_super_or_backup(osb, super_bh);
+	if (ret < 0) {
+		mlog_errno(ret);
+		goto out;
+	}
+
+	if (OCFS2_HAS_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_COMPAT_BACKUP_SB))
+		ret = update_backups(inode, clusters, super_bh->b_data);
+
+out:
+	brelse(super_bh);
+	if (ret)
+		printk(KERN_WARNING "ocfs2: Failed to update super blocks on %s"
+			" during fs resize. This condition is not fatal,"
+			" but fsck.ocfs2 should be run to fix it\n",
+			osb->dev_str);
+	return;
+}
+
+/*
+ * Extend the filesystem to the new number of clusters specified.  This entry
+ * point is only used to extend the current filesystem to the end of the last
+ * existing group.
+ */
+int ocfs2_group_extend(struct inode * inode, int new_clusters)
+{
+	int ret;
+	handle_t *handle;
+	struct buffer_head *main_bm_bh = NULL;
+	struct buffer_head *group_bh = NULL;
+	struct inode *main_bm_inode = NULL;
+	struct ocfs2_dinode *fe = NULL;
+	struct ocfs2_group_desc *group = NULL;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	u16 cl_bpc;
+	u32 first_new_cluster;
+	u64 lgd_blkno;
+
+	if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
+		return -EROFS;
+
+	if (new_clusters < 0)
+		return -EINVAL;
+	else if (new_clusters == 0)
+		return 0;
+
+	main_bm_inode = ocfs2_get_system_file_inode(osb,
+						    GLOBAL_BITMAP_SYSTEM_INODE,
+						    OCFS2_INVALID_SLOT);
+	if (!main_bm_inode) {
+		ret = -EINVAL;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	mutex_lock(&main_bm_inode->i_mutex);
+
+	ret = ocfs2_inode_lock(main_bm_inode, &main_bm_bh, 1);
+	if (ret < 0) {
+		mlog_errno(ret);
+		goto out_mutex;
+	}
+
+	fe = (struct ocfs2_dinode *)main_bm_bh->b_data;
+
+	/* main_bm_bh is validated by inode read inside ocfs2_inode_lock(),
+	 * so any corruption is a code bug. */
+	BUG_ON(!OCFS2_IS_VALID_DINODE(fe));
+
+	if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
+		ocfs2_group_bitmap_size(osb->sb, 0,
+					osb->s_feature_incompat) * 8) {
+		mlog(ML_ERROR, "The disk is too old and small. "
+		     "Force to do offline resize.");
+		ret = -EINVAL;
+		goto out_unlock;
+	}
+
+	first_new_cluster = le32_to_cpu(fe->i_clusters);
+	lgd_blkno = ocfs2_which_cluster_group(main_bm_inode,
+					      first_new_cluster - 1);
+
+	ret = ocfs2_read_group_descriptor(main_bm_inode, fe, lgd_blkno,
+					  &group_bh);
+	if (ret < 0) {
+		mlog_errno(ret);
+		goto out_unlock;
+	}
+	group = (struct ocfs2_group_desc *)group_bh->b_data;
+
+	cl_bpc = le16_to_cpu(fe->id2.i_chain.cl_bpc);
+	if (le16_to_cpu(group->bg_bits) / cl_bpc + new_clusters >
+		le16_to_cpu(fe->id2.i_chain.cl_cpg)) {
+		ret = -EINVAL;
+		goto out_unlock;
+	}
+
+
+	trace_ocfs2_group_extend(
+	     (unsigned long long)le64_to_cpu(group->bg_blkno), new_clusters);
+
+	handle = ocfs2_start_trans(osb, OCFS2_GROUP_EXTEND_CREDITS);
+	if (IS_ERR(handle)) {
+		mlog_errno(PTR_ERR(handle));
+		ret = -EINVAL;
+		goto out_unlock;
+	}
+
+	/* update the last group descriptor and inode. */
+	ret = ocfs2_update_last_group_and_inode(handle, main_bm_inode,
+						main_bm_bh, group_bh,
+						first_new_cluster,
+						new_clusters);
+	if (ret) {
+		mlog_errno(ret);
+		goto out_commit;
+	}
+
+	ocfs2_update_super_and_backups(main_bm_inode, new_clusters);
+
+out_commit:
+	ocfs2_commit_trans(osb, handle);
+out_unlock:
+	brelse(group_bh);
+	brelse(main_bm_bh);
+
+	ocfs2_inode_unlock(main_bm_inode, 1);
+
+out_mutex:
+	mutex_unlock(&main_bm_inode->i_mutex);
+	iput(main_bm_inode);
+
+out:
+	return ret;
+}
+
+static int ocfs2_check_new_group(struct inode *inode,
+				 struct ocfs2_dinode *di,
+				 struct ocfs2_new_group_input *input,
+				 struct buffer_head *group_bh)
+{
+	int ret;
+	struct ocfs2_group_desc *gd =
+		(struct ocfs2_group_desc *)group_bh->b_data;
+	u16 cl_bpc = le16_to_cpu(di->id2.i_chain.cl_bpc);
+
+	ret = ocfs2_check_group_descriptor(inode->i_sb, di, group_bh);
+	if (ret)
+		goto out;
+
+	ret = -EINVAL;
+	if (le16_to_cpu(gd->bg_chain) != input->chain)
+		mlog(ML_ERROR, "Group descriptor # %llu has bad chain %u "
+		     "while input has %u set.\n",
+		     (unsigned long long)le64_to_cpu(gd->bg_blkno),
+		     le16_to_cpu(gd->bg_chain), input->chain);
+	else if (le16_to_cpu(gd->bg_bits) != input->clusters * cl_bpc)
+		mlog(ML_ERROR, "Group descriptor # %llu has bit count %u but "
+		     "input has %u clusters set\n",
+		     (unsigned long long)le64_to_cpu(gd->bg_blkno),
+		     le16_to_cpu(gd->bg_bits), input->clusters);
+	else if (le16_to_cpu(gd->bg_free_bits_count) != input->frees * cl_bpc)
+		mlog(ML_ERROR, "Group descriptor # %llu has free bit count %u "
+		     "but it should have %u set\n",
+		     (unsigned long long)le64_to_cpu(gd->bg_blkno),
+		     le16_to_cpu(gd->bg_bits),
+		     input->frees * cl_bpc);
+	else
+		ret = 0;
+
+out:
+	return ret;
+}
+
+static int ocfs2_verify_group_and_input(struct inode *inode,
+					struct ocfs2_dinode *di,
+					struct ocfs2_new_group_input *input,
+					struct buffer_head *group_bh)
+{
+	u16 cl_count = le16_to_cpu(di->id2.i_chain.cl_count);
+	u16 cl_cpg = le16_to_cpu(di->id2.i_chain.cl_cpg);
+	u16 next_free = le16_to_cpu(di->id2.i_chain.cl_next_free_rec);
+	u32 cluster = ocfs2_blocks_to_clusters(inode->i_sb, input->group);
+	u32 total_clusters = le32_to_cpu(di->i_clusters);
+	int ret = -EINVAL;
+
+	if (cluster < total_clusters)
+		mlog(ML_ERROR, "add a group which is in the current volume.\n");
+	else if (input->chain >= cl_count)
+		mlog(ML_ERROR, "input chain exceeds the limit.\n");
+	else if (next_free != cl_count && next_free != input->chain)
+		mlog(ML_ERROR,
+		     "the add group should be in chain %u\n", next_free);
+	else if (total_clusters + input->clusters < total_clusters)
+		mlog(ML_ERROR, "add group's clusters overflow.\n");
+	else if (input->clusters > cl_cpg)
+		mlog(ML_ERROR, "the cluster exceeds the maximum of a group\n");
+	else if (input->frees > input->clusters)
+		mlog(ML_ERROR, "the free cluster exceeds the total clusters\n");
+	else if (total_clusters % cl_cpg != 0)
+		mlog(ML_ERROR,
+		     "the last group isn't full. Use group extend first.\n");
+	else if (input->group != ocfs2_which_cluster_group(inode, cluster))
+		mlog(ML_ERROR, "group blkno is invalid\n");
+	else if ((ret = ocfs2_check_new_group(inode, di, input, group_bh)))
+		mlog(ML_ERROR, "group descriptor check failed.\n");
+	else
+		ret = 0;
+
+	return ret;
+}
+
+/* Add a new group descriptor to global_bitmap. */
+int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input)
+{
+	int ret;
+	handle_t *handle;
+	struct buffer_head *main_bm_bh = NULL;
+	struct inode *main_bm_inode = NULL;
+	struct ocfs2_dinode *fe = NULL;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	struct buffer_head *group_bh = NULL;
+	struct ocfs2_group_desc *group = NULL;
+	struct ocfs2_chain_list *cl;
+	struct ocfs2_chain_rec *cr;
+	u16 cl_bpc;
+	u64 bg_ptr;
+
+	if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
+		return -EROFS;
+
+	main_bm_inode = ocfs2_get_system_file_inode(osb,
+						    GLOBAL_BITMAP_SYSTEM_INODE,
+						    OCFS2_INVALID_SLOT);
+	if (!main_bm_inode) {
+		ret = -EINVAL;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	mutex_lock(&main_bm_inode->i_mutex);
+
+	ret = ocfs2_inode_lock(main_bm_inode, &main_bm_bh, 1);
+	if (ret < 0) {
+		mlog_errno(ret);
+		goto out_mutex;
+	}
+
+	fe = (struct ocfs2_dinode *)main_bm_bh->b_data;
+
+	if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
+		ocfs2_group_bitmap_size(osb->sb, 0,
+					osb->s_feature_incompat) * 8) {
+		mlog(ML_ERROR, "The disk is too old and small."
+		     " Force to do offline resize.");
+		ret = -EINVAL;
+		goto out_unlock;
+	}
+
+	ret = ocfs2_read_blocks_sync(osb, input->group, 1, &group_bh);
+	if (ret < 0) {
+		mlog(ML_ERROR, "Can't read the group descriptor # %llu "
+		     "from the device.", (unsigned long long)input->group);
+		goto out_unlock;
+	}
+
+	ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode), group_bh);
+
+	ret = ocfs2_verify_group_and_input(main_bm_inode, fe, input, group_bh);
+	if (ret) {
+		mlog_errno(ret);
+		goto out_free_group_bh;
+	}
+
+	trace_ocfs2_group_add((unsigned long long)input->group,
+			       input->chain, input->clusters, input->frees);
+
+	handle = ocfs2_start_trans(osb, OCFS2_GROUP_ADD_CREDITS);
+	if (IS_ERR(handle)) {
+		mlog_errno(PTR_ERR(handle));
+		ret = -EINVAL;
+		goto out_free_group_bh;
+	}
+
+	cl_bpc = le16_to_cpu(fe->id2.i_chain.cl_bpc);
+	cl = &fe->id2.i_chain;
+	cr = &cl->cl_recs[input->chain];
+
+	ret = ocfs2_journal_access_gd(handle, INODE_CACHE(main_bm_inode),
+				      group_bh, OCFS2_JOURNAL_ACCESS_WRITE);
+	if (ret < 0) {
+		mlog_errno(ret);
+		goto out_commit;
+	}
+
+	group = (struct ocfs2_group_desc *)group_bh->b_data;
+	bg_ptr = le64_to_cpu(group->bg_next_group);
+	group->bg_next_group = cr->c_blkno;
+	ocfs2_journal_dirty(handle, group_bh);
+
+	ret = ocfs2_journal_access_di(handle, INODE_CACHE(main_bm_inode),
+				      main_bm_bh, OCFS2_JOURNAL_ACCESS_WRITE);
+	if (ret < 0) {
+		group->bg_next_group = cpu_to_le64(bg_ptr);
+		mlog_errno(ret);
+		goto out_commit;
+	}
+
+	if (input->chain == le16_to_cpu(cl->cl_next_free_rec)) {
+		le16_add_cpu(&cl->cl_next_free_rec, 1);
+		memset(cr, 0, sizeof(struct ocfs2_chain_rec));
+	}
+
+	cr->c_blkno = cpu_to_le64(input->group);
+	le32_add_cpu(&cr->c_total, input->clusters * cl_bpc);
+	le32_add_cpu(&cr->c_free, input->frees * cl_bpc);
+
+	le32_add_cpu(&fe->id1.bitmap1.i_total, input->clusters *cl_bpc);
+	le32_add_cpu(&fe->id1.bitmap1.i_used,
+		     (input->clusters - input->frees) * cl_bpc);
+	le32_add_cpu(&fe->i_clusters, input->clusters);
+
+	ocfs2_journal_dirty(handle, main_bm_bh);
+
+	spin_lock(&OCFS2_I(main_bm_inode)->ip_lock);
+	OCFS2_I(main_bm_inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
+	le64_add_cpu(&fe->i_size, (u64)input->clusters << osb->s_clustersize_bits);
+	spin_unlock(&OCFS2_I(main_bm_inode)->ip_lock);
+	i_size_write(main_bm_inode, le64_to_cpu(fe->i_size));
+
+	ocfs2_update_super_and_backups(main_bm_inode, input->clusters);
+
+out_commit:
+	ocfs2_commit_trans(osb, handle);
+
+out_free_group_bh:
+	brelse(group_bh);
+
+out_unlock:
+	brelse(main_bm_bh);
+
+	ocfs2_inode_unlock(main_bm_inode, 1);
+
+out_mutex:
+	mutex_unlock(&main_bm_inode->i_mutex);
+	iput(main_bm_inode);
+
+out:
+	return ret;
+}

fs/ocfs2/resize.h

@@ -0,0 +1,32 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * resize.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_RESIZE_H
+#define OCFS2_RESIZE_H
+
+int ocfs2_group_extend(struct inode * inode, int new_clusters);
+int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input);
+
+#endif /* OCFS2_RESIZE_H */

fs/ocfs2/slot_map.c

@@ -0,0 +1,538 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * slot_map.c
+ *
+ *
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "dlmglue.h"
+#include "extent_map.h"
+#include "heartbeat.h"
+#include "inode.h"
+#include "slot_map.h"
+#include "super.h"
+#include "sysfile.h"
+#include "ocfs2_trace.h"
+
+#include "buffer_head_io.h"
+
+
+struct ocfs2_slot {
+	int sl_valid;
+	unsigned int sl_node_num;
+};
+
+struct ocfs2_slot_info {
+	int si_extended;
+	int si_slots_per_block;
+	struct inode *si_inode;
+	unsigned int si_blocks;
+	struct buffer_head **si_bh;
+	unsigned int si_num_slots;
+	struct ocfs2_slot *si_slots;
+};
+
+
+static int __ocfs2_node_num_to_slot(struct ocfs2_slot_info *si,
+				    unsigned int node_num);
+
+static void ocfs2_invalidate_slot(struct ocfs2_slot_info *si,
+				  int slot_num)
+{
+	BUG_ON((slot_num < 0) || (slot_num >= si->si_num_slots));
+	si->si_slots[slot_num].sl_valid = 0;
+}
+
+static void ocfs2_set_slot(struct ocfs2_slot_info *si,
+			   int slot_num, unsigned int node_num)
+{
+	BUG_ON((slot_num < 0) || (slot_num >= si->si_num_slots));
+
+	si->si_slots[slot_num].sl_valid = 1;
+	si->si_slots[slot_num].sl_node_num = node_num;
+}
+
+/* This version is for the extended slot map */
+static void ocfs2_update_slot_info_extended(struct ocfs2_slot_info *si)
+{
+	int b, i, slotno;
+	struct ocfs2_slot_map_extended *se;
+
+	slotno = 0;
+	for (b = 0; b < si->si_blocks; b++) {
+		se = (struct ocfs2_slot_map_extended *)si->si_bh[b]->b_data;
+		for (i = 0;
+		     (i < si->si_slots_per_block) &&
+		     (slotno < si->si_num_slots);
+		     i++, slotno++) {
+			if (se->se_slots[i].es_valid)
+				ocfs2_set_slot(si, slotno,
+					       le32_to_cpu(se->se_slots[i].es_node_num));
+			else
+				ocfs2_invalidate_slot(si, slotno);
+		}
+	}
+}
+
+/*
+ * Post the slot information on disk into our slot_info struct.
+ * Must be protected by osb_lock.
+ */
+static void ocfs2_update_slot_info_old(struct ocfs2_slot_info *si)
+{
+	int i;
+	struct ocfs2_slot_map *sm;
+
+	sm = (struct ocfs2_slot_map *)si->si_bh[0]->b_data;
+
+	for (i = 0; i < si->si_num_slots; i++) {
+		if (le16_to_cpu(sm->sm_slots[i]) == (u16)OCFS2_INVALID_SLOT)
+			ocfs2_invalidate_slot(si, i);
+		else
+			ocfs2_set_slot(si, i, le16_to_cpu(sm->sm_slots[i]));
+	}
+}
+
+static void ocfs2_update_slot_info(struct ocfs2_slot_info *si)
+{
+	/*
+	 * The slot data will have been refreshed when ocfs2_super_lock
+	 * was taken.
+	 */
+	if (si->si_extended)
+		ocfs2_update_slot_info_extended(si);
+	else
+		ocfs2_update_slot_info_old(si);
+}
+
+int ocfs2_refresh_slot_info(struct ocfs2_super *osb)
+{
+	int ret;
+	struct ocfs2_slot_info *si = osb->slot_info;
+
+	if (si == NULL)
+		return 0;
+
+	BUG_ON(si->si_blocks == 0);
+	BUG_ON(si->si_bh == NULL);
+
+	trace_ocfs2_refresh_slot_info(si->si_blocks);
+
+	/*
+	 * We pass -1 as blocknr because we expect all of si->si_bh to
+	 * be !NULL.  Thus, ocfs2_read_blocks() will ignore blocknr.  If
+	 * this is not true, the read of -1 (UINT64_MAX) will fail.
+	 */
+	ret = ocfs2_read_blocks(INODE_CACHE(si->si_inode), -1, si->si_blocks,
+				si->si_bh, OCFS2_BH_IGNORE_CACHE, NULL);
+	if (ret == 0) {
+		spin_lock(&osb->osb_lock);
+		ocfs2_update_slot_info(si);
+		spin_unlock(&osb->osb_lock);
+	}
+
+	return ret;
+}
+
+/* post the our slot info stuff into it's destination bh and write it
+ * out. */
+static void ocfs2_update_disk_slot_extended(struct ocfs2_slot_info *si,
+					    int slot_num,
+					    struct buffer_head **bh)
+{
+	int blkind = slot_num / si->si_slots_per_block;
+	int slotno = slot_num % si->si_slots_per_block;
+	struct ocfs2_slot_map_extended *se;
+
+	BUG_ON(blkind >= si->si_blocks);
+
+	se = (struct ocfs2_slot_map_extended *)si->si_bh[blkind]->b_data;
+	se->se_slots[slotno].es_valid = si->si_slots[slot_num].sl_valid;
+	if (si->si_slots[slot_num].sl_valid)
+		se->se_slots[slotno].es_node_num =
+			cpu_to_le32(si->si_slots[slot_num].sl_node_num);
+	*bh = si->si_bh[blkind];
+}
+
+static void ocfs2_update_disk_slot_old(struct ocfs2_slot_info *si,
+				       int slot_num,
+				       struct buffer_head **bh)
+{
+	int i;
+	struct ocfs2_slot_map *sm;
+
+	sm = (struct ocfs2_slot_map *)si->si_bh[0]->b_data;
+	for (i = 0; i < si->si_num_slots; i++) {
+		if (si->si_slots[i].sl_valid)
+			sm->sm_slots[i] =
+				cpu_to_le16(si->si_slots[i].sl_node_num);
+		else
+			sm->sm_slots[i] = cpu_to_le16(OCFS2_INVALID_SLOT);
+	}
+	*bh = si->si_bh[0];
+}
+
+static int ocfs2_update_disk_slot(struct ocfs2_super *osb,
+				  struct ocfs2_slot_info *si,
+				  int slot_num)
+{
+	int status;
+	struct buffer_head *bh;
+
+	spin_lock(&osb->osb_lock);
+	if (si->si_extended)
+		ocfs2_update_disk_slot_extended(si, slot_num, &bh);
+	else
+		ocfs2_update_disk_slot_old(si, slot_num, &bh);
+	spin_unlock(&osb->osb_lock);
+
+	status = ocfs2_write_block(osb, bh, INODE_CACHE(si->si_inode));
+	if (status < 0)
+		mlog_errno(status);
+
+	return status;
+}
+
+/*
+ * Calculate how many bytes are needed by the slot map.  Returns
+ * an error if the slot map file is too small.
+ */
+static int ocfs2_slot_map_physical_size(struct ocfs2_super *osb,
+					struct inode *inode,
+					unsigned long long *bytes)
+{
+	unsigned long long bytes_needed;
+
+	if (ocfs2_uses_extended_slot_map(osb)) {
+		bytes_needed = osb->max_slots *
+			sizeof(struct ocfs2_extended_slot);
+	} else {
+		bytes_needed = osb->max_slots * sizeof(__le16);
+	}
+	if (bytes_needed > i_size_read(inode)) {
+		mlog(ML_ERROR,
+		     "Slot map file is too small!  (size %llu, needed %llu)\n",
+		     i_size_read(inode), bytes_needed);
+		return -ENOSPC;
+	}
+
+	*bytes = bytes_needed;
+	return 0;
+}
+
+/* try to find global node in the slot info. Returns -ENOENT
+ * if nothing is found. */
+static int __ocfs2_node_num_to_slot(struct ocfs2_slot_info *si,
+				    unsigned int node_num)
+{
+	int i, ret = -ENOENT;
+
+	for(i = 0; i < si->si_num_slots; i++) {
+		if (si->si_slots[i].sl_valid &&
+		    (node_num == si->si_slots[i].sl_node_num)) {
+			ret = i;
+			break;
+		}
+	}
+
+	return ret;
+}
+
+static int __ocfs2_find_empty_slot(struct ocfs2_slot_info *si,
+				   int preferred)
+{
+	int i, ret = -ENOSPC;
+
+	if ((preferred >= 0) && (preferred < si->si_num_slots)) {
+		if (!si->si_slots[preferred].sl_valid) {
+			ret = preferred;
+			goto out;
+		}
+	}
+
+	for(i = 0; i < si->si_num_slots; i++) {
+		if (!si->si_slots[i].sl_valid) {
+			ret = i;
+			break;
+		}
+	}
+out:
+	return ret;
+}
+
+int ocfs2_node_num_to_slot(struct ocfs2_super *osb, unsigned int node_num)
+{
+	int slot;
+	struct ocfs2_slot_info *si = osb->slot_info;
+
+	spin_lock(&osb->osb_lock);
+	slot = __ocfs2_node_num_to_slot(si, node_num);
+	spin_unlock(&osb->osb_lock);
+
+	return slot;
+}
+
+int ocfs2_slot_to_node_num_locked(struct ocfs2_super *osb, int slot_num,
+				  unsigned int *node_num)
+{
+	struct ocfs2_slot_info *si = osb->slot_info;
+
+	assert_spin_locked(&osb->osb_lock);
+
+	BUG_ON(slot_num < 0);
+	BUG_ON(slot_num > osb->max_slots);
+
+	if (!si->si_slots[slot_num].sl_valid)
+		return -ENOENT;
+
+	*node_num = si->si_slots[slot_num].sl_node_num;
+	return 0;
+}
+
+static void __ocfs2_free_slot_info(struct ocfs2_slot_info *si)
+{
+	unsigned int i;
+
+	if (si == NULL)
+		return;
+
+	if (si->si_inode)
+		iput(si->si_inode);
+	if (si->si_bh) {
+		for (i = 0; i < si->si_blocks; i++) {
+			if (si->si_bh[i]) {
+				brelse(si->si_bh[i]);
+				si->si_bh[i] = NULL;
+			}
+		}
+		kfree(si->si_bh);
+	}
+
+	kfree(si);
+}
+
+int ocfs2_clear_slot(struct ocfs2_super *osb, int slot_num)
+{
+	struct ocfs2_slot_info *si = osb->slot_info;
+
+	if (si == NULL)
+		return 0;
+
+	spin_lock(&osb->osb_lock);
+	ocfs2_invalidate_slot(si, slot_num);
+	spin_unlock(&osb->osb_lock);
+
+	return ocfs2_update_disk_slot(osb, osb->slot_info, slot_num);
+}
+
+static int ocfs2_map_slot_buffers(struct ocfs2_super *osb,
+				  struct ocfs2_slot_info *si)
+{
+	int status = 0;
+	u64 blkno;
+	unsigned long long blocks, bytes = 0;
+	unsigned int i;
+	struct buffer_head *bh;
+
+	status = ocfs2_slot_map_physical_size(osb, si->si_inode, &bytes);
+	if (status)
+		goto bail;
+
+	blocks = ocfs2_blocks_for_bytes(si->si_inode->i_sb, bytes);
+	BUG_ON(blocks > UINT_MAX);
+	si->si_blocks = blocks;
+	if (!si->si_blocks)
+		goto bail;
+
+	if (si->si_extended)
+		si->si_slots_per_block =
+			(osb->sb->s_blocksize /
+			 sizeof(struct ocfs2_extended_slot));
+	else
+		si->si_slots_per_block = osb->sb->s_blocksize / sizeof(__le16);
+
+	/* The size checks above should ensure this */
+	BUG_ON((osb->max_slots / si->si_slots_per_block) > blocks);
+
+	trace_ocfs2_map_slot_buffers(bytes, si->si_blocks);
+
+	si->si_bh = kcalloc(si->si_blocks, sizeof(struct buffer_head *),
+			    GFP_KERNEL);
+	if (!si->si_bh) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	for (i = 0; i < si->si_blocks; i++) {
+		status = ocfs2_extent_map_get_blocks(si->si_inode, i,
+						     &blkno, NULL, NULL);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+
+		trace_ocfs2_map_slot_buffers_block((unsigned long long)blkno, i);
+
+		bh = NULL;  /* Acquire a fresh bh */
+		status = ocfs2_read_blocks(INODE_CACHE(si->si_inode), blkno,
+					   1, &bh, OCFS2_BH_IGNORE_CACHE, NULL);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+
+		si->si_bh[i] = bh;
+	}
+
+bail:
+	return status;
+}
+
+int ocfs2_init_slot_info(struct ocfs2_super *osb)
+{
+	int status;
+	struct inode *inode = NULL;
+	struct ocfs2_slot_info *si;
+
+	si = kzalloc(sizeof(struct ocfs2_slot_info) +
+		     (sizeof(struct ocfs2_slot) * osb->max_slots),
+		     GFP_KERNEL);
+	if (!si) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	si->si_extended = ocfs2_uses_extended_slot_map(osb);
+	si->si_num_slots = osb->max_slots;
+	si->si_slots = (struct ocfs2_slot *)((char *)si +
+					     sizeof(struct ocfs2_slot_info));
+
+	inode = ocfs2_get_system_file_inode(osb, SLOT_MAP_SYSTEM_INODE,
+					    OCFS2_INVALID_SLOT);
+	if (!inode) {
+		status = -EINVAL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	si->si_inode = inode;
+	status = ocfs2_map_slot_buffers(osb, si);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	osb->slot_info = (struct ocfs2_slot_info *)si;
+bail:
+	if (status < 0 && si)
+		__ocfs2_free_slot_info(si);
+
+	return status;
+}
+
+void ocfs2_free_slot_info(struct ocfs2_super *osb)
+{
+	struct ocfs2_slot_info *si = osb->slot_info;
+
+	osb->slot_info = NULL;
+	__ocfs2_free_slot_info(si);
+}
+
+int ocfs2_find_slot(struct ocfs2_super *osb)
+{
+	int status;
+	int slot;
+	struct ocfs2_slot_info *si;
+
+	si = osb->slot_info;
+
+	spin_lock(&osb->osb_lock);
+	ocfs2_update_slot_info(si);
+
+	/* search for ourselves first and take the slot if it already
+	 * exists. Perhaps we need to mark this in a variable for our
+	 * own journal recovery? Possibly not, though we certainly
+	 * need to warn to the user */
+	slot = __ocfs2_node_num_to_slot(si, osb->node_num);
+	if (slot < 0) {
+		/* if no slot yet, then just take 1st available
+		 * one. */
+		slot = __ocfs2_find_empty_slot(si, osb->preferred_slot);
+		if (slot < 0) {
+			spin_unlock(&osb->osb_lock);
+			mlog(ML_ERROR, "no free slots available!\n");
+			status = -EINVAL;
+			goto bail;
+		}
+	} else
+		printk(KERN_INFO "ocfs2: Slot %d on device (%s) was already "
+		       "allocated to this node!\n", slot, osb->dev_str);
+
+	ocfs2_set_slot(si, slot, osb->node_num);
+	osb->slot_num = slot;
+	spin_unlock(&osb->osb_lock);
+
+	trace_ocfs2_find_slot(osb->slot_num);
+
+	status = ocfs2_update_disk_slot(osb, si, osb->slot_num);
+	if (status < 0)
+		mlog_errno(status);
+
+bail:
+	return status;
+}
+
+void ocfs2_put_slot(struct ocfs2_super *osb)
+{
+	int status, slot_num;
+	struct ocfs2_slot_info *si = osb->slot_info;
+
+	if (!si)
+		return;
+
+	spin_lock(&osb->osb_lock);
+	ocfs2_update_slot_info(si);
+
+	slot_num = osb->slot_num;
+	ocfs2_invalidate_slot(si, osb->slot_num);
+	osb->slot_num = OCFS2_INVALID_SLOT;
+	spin_unlock(&osb->osb_lock);
+
+	status = ocfs2_update_disk_slot(osb, si, slot_num);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+bail:
+	ocfs2_free_slot_info(osb);
+}
+

fs/ocfs2/slot_map.h

@@ -0,0 +1,44 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * slotmap.h
+ *
+ * description here
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+
+#ifndef SLOTMAP_H
+#define SLOTMAP_H
+
+int ocfs2_init_slot_info(struct ocfs2_super *osb);
+void ocfs2_free_slot_info(struct ocfs2_super *osb);
+
+int ocfs2_find_slot(struct ocfs2_super *osb);
+void ocfs2_put_slot(struct ocfs2_super *osb);
+
+int ocfs2_refresh_slot_info(struct ocfs2_super *osb);
+
+int ocfs2_node_num_to_slot(struct ocfs2_super *osb, unsigned int node_num);
+int ocfs2_slot_to_node_num_locked(struct ocfs2_super *osb, int slot_num,
+				  unsigned int *node_num);
+
+int ocfs2_clear_slot(struct ocfs2_super *osb, int slot_num);
+
+#endif

fs/ocfs2/stack_o2cb.c

@@ -0,0 +1,449 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * stack_o2cb.c
+ *
+ * Code which interfaces ocfs2 with the o2cb stack.
+ *
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation, version 2.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#include <linux/kernel.h>
+#include <linux/crc32.h>
+#include <linux/slab.h>
+#include <linux/module.h>
+
+/* Needed for AOP_TRUNCATED_PAGE in mlog_errno() */
+#include <linux/fs.h>
+
+#include "cluster/masklog.h"
+#include "cluster/nodemanager.h"
+#include "cluster/heartbeat.h"
+#include "cluster/tcp.h"
+
+#include "stackglue.h"
+
+struct o2dlm_private {
+	struct dlm_eviction_cb op_eviction_cb;
+};
+
+static struct ocfs2_stack_plugin o2cb_stack;
+
+/* These should be identical */
+#if (DLM_LOCK_IV != LKM_IVMODE)
+# error Lock modes do not match
+#endif
+#if (DLM_LOCK_NL != LKM_NLMODE)
+# error Lock modes do not match
+#endif
+#if (DLM_LOCK_CR != LKM_CRMODE)
+# error Lock modes do not match
+#endif
+#if (DLM_LOCK_CW != LKM_CWMODE)
+# error Lock modes do not match
+#endif
+#if (DLM_LOCK_PR != LKM_PRMODE)
+# error Lock modes do not match
+#endif
+#if (DLM_LOCK_PW != LKM_PWMODE)
+# error Lock modes do not match
+#endif
+#if (DLM_LOCK_EX != LKM_EXMODE)
+# error Lock modes do not match
+#endif
+static inline int mode_to_o2dlm(int mode)
+{
+	BUG_ON(mode > LKM_MAXMODE);
+
+	return mode;
+}
+
+#define map_flag(_generic, _o2dlm)		\
+	if (flags & (_generic)) {		\
+		flags &= ~(_generic);		\
+		o2dlm_flags |= (_o2dlm);	\
+	}
+static int flags_to_o2dlm(u32 flags)
+{
+	int o2dlm_flags = 0;
+
+	map_flag(DLM_LKF_NOQUEUE, LKM_NOQUEUE);
+	map_flag(DLM_LKF_CANCEL, LKM_CANCEL);
+	map_flag(DLM_LKF_CONVERT, LKM_CONVERT);
+	map_flag(DLM_LKF_VALBLK, LKM_VALBLK);
+	map_flag(DLM_LKF_IVVALBLK, LKM_INVVALBLK);
+	map_flag(DLM_LKF_ORPHAN, LKM_ORPHAN);
+	map_flag(DLM_LKF_FORCEUNLOCK, LKM_FORCE);
+	map_flag(DLM_LKF_TIMEOUT, LKM_TIMEOUT);
+	map_flag(DLM_LKF_LOCAL, LKM_LOCAL);
+
+	/* map_flag() should have cleared every flag passed in */
+	BUG_ON(flags != 0);
+
+	return o2dlm_flags;
+}
+#undef map_flag
+
+/*
+ * Map an o2dlm status to standard errno values.
+ *
+ * o2dlm only uses a handful of these, and returns even fewer to the
+ * caller. Still, we try to assign sane values to each error.
+ *
+ * The following value pairs have special meanings to dlmglue, thus
+ * the right hand side needs to stay unique - never duplicate the
+ * mapping elsewhere in the table!
+ *
+ * DLM_NORMAL:		0
+ * DLM_NOTQUEUED:	-EAGAIN
+ * DLM_CANCELGRANT:	-EBUSY
+ * DLM_CANCEL:		-DLM_ECANCEL
+ */
+/* Keep in sync with dlmapi.h */
+static int status_map[] = {
+	[DLM_NORMAL]			= 0,		/* Success */
+	[DLM_GRANTED]			= -EINVAL,
+	[DLM_DENIED]			= -EACCES,
+	[DLM_DENIED_NOLOCKS]		= -EACCES,
+	[DLM_WORKING]			= -EACCES,
+	[DLM_BLOCKED]			= -EINVAL,
+	[DLM_BLOCKED_ORPHAN]		= -EINVAL,
+	[DLM_DENIED_GRACE_PERIOD]	= -EACCES,
+	[DLM_SYSERR]			= -ENOMEM,	/* It is what it is */
+	[DLM_NOSUPPORT]			= -EPROTO,
+	[DLM_CANCELGRANT]		= -EBUSY,	/* Cancel after grant */
+	[DLM_IVLOCKID]			= -EINVAL,
+	[DLM_SYNC]			= -EINVAL,
+	[DLM_BADTYPE]			= -EINVAL,
+	[DLM_BADRESOURCE]		= -EINVAL,
+	[DLM_MAXHANDLES]		= -ENOMEM,
+	[DLM_NOCLINFO]			= -EINVAL,
+	[DLM_NOLOCKMGR]			= -EINVAL,
+	[DLM_NOPURGED]			= -EINVAL,
+	[DLM_BADARGS]			= -EINVAL,
+	[DLM_VOID]			= -EINVAL,
+	[DLM_NOTQUEUED]			= -EAGAIN,	/* Trylock failed */
+	[DLM_IVBUFLEN]			= -EINVAL,
+	[DLM_CVTUNGRANT]		= -EPERM,
+	[DLM_BADPARAM]			= -EINVAL,
+	[DLM_VALNOTVALID]		= -EINVAL,
+	[DLM_REJECTED]			= -EPERM,
+	[DLM_ABORT]			= -EINVAL,
+	[DLM_CANCEL]			= -DLM_ECANCEL,	/* Successful cancel */
+	[DLM_IVRESHANDLE]		= -EINVAL,
+	[DLM_DEADLOCK]			= -EDEADLK,
+	[DLM_DENIED_NOASTS]		= -EINVAL,
+	[DLM_FORWARD]			= -EINVAL,
+	[DLM_TIMEOUT]			= -ETIMEDOUT,
+	[DLM_IVGROUPID]			= -EINVAL,
+	[DLM_VERS_CONFLICT]		= -EOPNOTSUPP,
+	[DLM_BAD_DEVICE_PATH]		= -ENOENT,
+	[DLM_NO_DEVICE_PERMISSION]	= -EPERM,
+	[DLM_NO_CONTROL_DEVICE]		= -ENOENT,
+	[DLM_RECOVERING]		= -ENOTCONN,
+	[DLM_MIGRATING]			= -ERESTART,
+	[DLM_MAXSTATS]			= -EINVAL,
+};
+
+static int dlm_status_to_errno(enum dlm_status status)
+{
+	BUG_ON(status < 0 || status >= ARRAY_SIZE(status_map));
+
+	return status_map[status];
+}
+
+static void o2dlm_lock_ast_wrapper(void *astarg)
+{
+	struct ocfs2_dlm_lksb *lksb = astarg;
+
+	lksb->lksb_conn->cc_proto->lp_lock_ast(lksb);
+}
+
+static void o2dlm_blocking_ast_wrapper(void *astarg, int level)
+{
+	struct ocfs2_dlm_lksb *lksb = astarg;
+
+	lksb->lksb_conn->cc_proto->lp_blocking_ast(lksb, level);
+}
+
+static void o2dlm_unlock_ast_wrapper(void *astarg, enum dlm_status status)
+{
+	struct ocfs2_dlm_lksb *lksb = astarg;
+	int error = dlm_status_to_errno(status);
+
+	/*
+	 * In o2dlm, you can get both the lock_ast() for the lock being
+	 * granted and the unlock_ast() for the CANCEL failing.  A
+	 * successful cancel sends DLM_NORMAL here.  If the
+	 * lock grant happened before the cancel arrived, you get
+	 * DLM_CANCELGRANT.
+	 *
+	 * There's no need for the double-ast.  If we see DLM_CANCELGRANT,
+	 * we just ignore it.  We expect the lock_ast() to handle the
+	 * granted lock.
+	 */
+	if (status == DLM_CANCELGRANT)
+		return;
+
+	lksb->lksb_conn->cc_proto->lp_unlock_ast(lksb, error);
+}
+
+static int o2cb_dlm_lock(struct ocfs2_cluster_connection *conn,
+			 int mode,
+			 struct ocfs2_dlm_lksb *lksb,
+			 u32 flags,
+			 void *name,
+			 unsigned int namelen)
+{
+	enum dlm_status status;
+	int o2dlm_mode = mode_to_o2dlm(mode);
+	int o2dlm_flags = flags_to_o2dlm(flags);
+	int ret;
+
+	status = dlmlock(conn->cc_lockspace, o2dlm_mode, &lksb->lksb_o2dlm,
+			 o2dlm_flags, name, namelen,
+			 o2dlm_lock_ast_wrapper, lksb,
+			 o2dlm_blocking_ast_wrapper);
+	ret = dlm_status_to_errno(status);
+	return ret;
+}
+
+static int o2cb_dlm_unlock(struct ocfs2_cluster_connection *conn,
+			   struct ocfs2_dlm_lksb *lksb,
+			   u32 flags)
+{
+	enum dlm_status status;
+	int o2dlm_flags = flags_to_o2dlm(flags);
+	int ret;
+
+	status = dlmunlock(conn->cc_lockspace, &lksb->lksb_o2dlm,
+			   o2dlm_flags, o2dlm_unlock_ast_wrapper, lksb);
+	ret = dlm_status_to_errno(status);
+	return ret;
+}
+
+static int o2cb_dlm_lock_status(struct ocfs2_dlm_lksb *lksb)
+{
+	return dlm_status_to_errno(lksb->lksb_o2dlm.status);
+}
+
+/*
+ * o2dlm aways has a "valid" LVB. If the dlm loses track of the LVB
+ * contents, it will zero out the LVB.  Thus the caller can always trust
+ * the contents.
+ */
+static int o2cb_dlm_lvb_valid(struct ocfs2_dlm_lksb *lksb)
+{
+	return 1;
+}
+
+static void *o2cb_dlm_lvb(struct ocfs2_dlm_lksb *lksb)
+{
+	return (void *)(lksb->lksb_o2dlm.lvb);
+}
+
+static void o2cb_dump_lksb(struct ocfs2_dlm_lksb *lksb)
+{
+	dlm_print_one_lock(lksb->lksb_o2dlm.lockid);
+}
+
+/*
+ * Check if this node is heartbeating and is connected to all other
+ * heartbeating nodes.
+ */
+static int o2cb_cluster_check(void)
+{
+	u8 node_num;
+	int i;
+	unsigned long hbmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
+	unsigned long netmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
+
+	node_num = o2nm_this_node();
+	if (node_num == O2NM_MAX_NODES) {
+		printk(KERN_ERR "o2cb: This node has not been configured.\n");
+		return -EINVAL;
+	}
+
+	/*
+	 * o2dlm expects o2net sockets to be created. If not, then
+	 * dlm_join_domain() fails with a stack of errors which are both cryptic
+	 * and incomplete. The idea here is to detect upfront whether we have
+	 * managed to connect to all nodes or not. If not, then list the nodes
+	 * to allow the user to check the configuration (incorrect IP, firewall,
+	 * etc.) Yes, this is racy. But its not the end of the world.
+	 */
+#define	O2CB_MAP_STABILIZE_COUNT	60
+	for (i = 0; i < O2CB_MAP_STABILIZE_COUNT; ++i) {
+		o2hb_fill_node_map(hbmap, sizeof(hbmap));
+		if (!test_bit(node_num, hbmap)) {
+			printk(KERN_ERR "o2cb: %s heartbeat has not been "
+			       "started.\n", (o2hb_global_heartbeat_active() ?
+					      "Global" : "Local"));
+			return -EINVAL;
+		}
+		o2net_fill_node_map(netmap, sizeof(netmap));
+		/* Force set the current node to allow easy compare */
+		set_bit(node_num, netmap);
+		if (!memcmp(hbmap, netmap, sizeof(hbmap)))
+			return 0;
+		if (i < O2CB_MAP_STABILIZE_COUNT)
+			msleep(1000);
+	}
+
+	printk(KERN_ERR "o2cb: This node could not connect to nodes:");
+	i = -1;
+	while ((i = find_next_bit(hbmap, O2NM_MAX_NODES,
+				  i + 1)) < O2NM_MAX_NODES) {
+		if (!test_bit(i, netmap))
+			printk(" %u", i);
+	}
+	printk(".\n");
+
+	return -ENOTCONN;
+}
+
+/*
+ * Called from the dlm when it's about to evict a node. This is how the
+ * classic stack signals node death.
+ */
+static void o2dlm_eviction_cb(int node_num, void *data)
+{
+	struct ocfs2_cluster_connection *conn = data;
+
+	printk(KERN_NOTICE "o2cb: o2dlm has evicted node %d from domain %.*s\n",
+	       node_num, conn->cc_namelen, conn->cc_name);
+
+	conn->cc_recovery_handler(node_num, conn->cc_recovery_data);
+}
+
+static int o2cb_cluster_connect(struct ocfs2_cluster_connection *conn)
+{
+	int rc = 0;
+	u32 dlm_key;
+	struct dlm_ctxt *dlm;
+	struct o2dlm_private *priv;
+	struct dlm_protocol_version fs_version;
+
+	BUG_ON(conn == NULL);
+	BUG_ON(conn->cc_proto == NULL);
+
+	/* Ensure cluster stack is up and all nodes are connected */
+	rc = o2cb_cluster_check();
+	if (rc) {
+		printk(KERN_ERR "o2cb: Cluster check failed. Fix errors "
+		       "before retrying.\n");
+		goto out;
+	}
+
+	priv = kzalloc(sizeof(struct o2dlm_private), GFP_KERNEL);
+	if (!priv) {
+		rc = -ENOMEM;
+		goto out_free;
+	}
+
+	/* This just fills the structure in.  It is safe to pass conn. */
+	dlm_setup_eviction_cb(&priv->op_eviction_cb, o2dlm_eviction_cb,
+			      conn);
+
+	conn->cc_private = priv;
+
+	/* used by the dlm code to make message headers unique, each
+	 * node in this domain must agree on this. */
+	dlm_key = crc32_le(0, conn->cc_name, conn->cc_namelen);
+	fs_version.pv_major = conn->cc_version.pv_major;
+	fs_version.pv_minor = conn->cc_version.pv_minor;
+
+	dlm = dlm_register_domain(conn->cc_name, dlm_key, &fs_version);
+	if (IS_ERR(dlm)) {
+		rc = PTR_ERR(dlm);
+		mlog_errno(rc);
+		goto out_free;
+	}
+
+	conn->cc_version.pv_major = fs_version.pv_major;
+	conn->cc_version.pv_minor = fs_version.pv_minor;
+	conn->cc_lockspace = dlm;
+
+	dlm_register_eviction_cb(dlm, &priv->op_eviction_cb);
+
+out_free:
+	if (rc)
+		kfree(conn->cc_private);
+
+out:
+	return rc;
+}
+
+static int o2cb_cluster_disconnect(struct ocfs2_cluster_connection *conn)
+{
+	struct dlm_ctxt *dlm = conn->cc_lockspace;
+	struct o2dlm_private *priv = conn->cc_private;
+
+	dlm_unregister_eviction_cb(&priv->op_eviction_cb);
+	conn->cc_private = NULL;
+	kfree(priv);
+
+	dlm_unregister_domain(dlm);
+	conn->cc_lockspace = NULL;
+
+	return 0;
+}
+
+static int o2cb_cluster_this_node(struct ocfs2_cluster_connection *conn,
+				  unsigned int *node)
+{
+	int node_num;
+
+	node_num = o2nm_this_node();
+	if (node_num == O2NM_INVALID_NODE_NUM)
+		return -ENOENT;
+
+	if (node_num >= O2NM_MAX_NODES)
+		return -EOVERFLOW;
+
+	*node = node_num;
+	return 0;
+}
+
+static struct ocfs2_stack_operations o2cb_stack_ops = {
+	.connect	= o2cb_cluster_connect,
+	.disconnect	= o2cb_cluster_disconnect,
+	.this_node	= o2cb_cluster_this_node,
+	.dlm_lock	= o2cb_dlm_lock,
+	.dlm_unlock	= o2cb_dlm_unlock,
+	.lock_status	= o2cb_dlm_lock_status,
+	.lvb_valid	= o2cb_dlm_lvb_valid,
+	.lock_lvb	= o2cb_dlm_lvb,
+	.dump_lksb	= o2cb_dump_lksb,
+};
+
+static struct ocfs2_stack_plugin o2cb_stack = {
+	.sp_name	= "o2cb",
+	.sp_ops		= &o2cb_stack_ops,
+	.sp_owner	= THIS_MODULE,
+};
+
+static int __init o2cb_stack_init(void)
+{
+	return ocfs2_stack_glue_register(&o2cb_stack);
+}
+
+static void __exit o2cb_stack_exit(void)
+{
+	ocfs2_stack_glue_unregister(&o2cb_stack);
+}
+
+MODULE_AUTHOR("Oracle");
+MODULE_DESCRIPTION("ocfs2 driver for the classic o2cb stack");
+MODULE_LICENSE("GPL");
+module_init(o2cb_stack_init);
+module_exit(o2cb_stack_exit);

fs/ocfs2/stackglue.c

@@ -0,0 +1,748 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * stackglue.c
+ *
+ * Code which implements an OCFS2 specific interface to underlying
+ * cluster stacks.
+ *
+ * Copyright (C) 2007, 2009 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation, version 2.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#include <linux/list.h>
+#include <linux/spinlock.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/kmod.h>
+#include <linux/fs.h>
+#include <linux/kobject.h>
+#include <linux/sysfs.h>
+#include <linux/sysctl.h>
+
+#include "ocfs2_fs.h"
+
+#include "stackglue.h"
+
+#define OCFS2_STACK_PLUGIN_O2CB		"o2cb"
+#define OCFS2_STACK_PLUGIN_USER		"user"
+#define OCFS2_MAX_HB_CTL_PATH		256
+
+static struct ocfs2_protocol_version locking_max_version;
+static DEFINE_SPINLOCK(ocfs2_stack_lock);
+static LIST_HEAD(ocfs2_stack_list);
+static char cluster_stack_name[OCFS2_STACK_LABEL_LEN + 1];
+static char ocfs2_hb_ctl_path[OCFS2_MAX_HB_CTL_PATH] = "/sbin/ocfs2_hb_ctl";
+
+/*
+ * The stack currently in use.  If not null, active_stack->sp_count > 0,
+ * the module is pinned, and the locking protocol cannot be changed.
+ */
+static struct ocfs2_stack_plugin *active_stack;
+
+static struct ocfs2_stack_plugin *ocfs2_stack_lookup(const char *name)
+{
+	struct ocfs2_stack_plugin *p;
+
+	assert_spin_locked(&ocfs2_stack_lock);
+
+	list_for_each_entry(p, &ocfs2_stack_list, sp_list) {
+		if (!strcmp(p->sp_name, name))
+			return p;
+	}
+
+	return NULL;
+}
+
+static int ocfs2_stack_driver_request(const char *stack_name,
+				      const char *plugin_name)
+{
+	int rc;
+	struct ocfs2_stack_plugin *p;
+
+	spin_lock(&ocfs2_stack_lock);
+
+	/*
+	 * If the stack passed by the filesystem isn't the selected one,
+	 * we can't continue.
+	 */
+	if (strcmp(stack_name, cluster_stack_name)) {
+		rc = -EBUSY;
+		goto out;
+	}
+
+	if (active_stack) {
+		/*
+		 * If the active stack isn't the one we want, it cannot
+		 * be selected right now.
+		 */
+		if (!strcmp(active_stack->sp_name, plugin_name))
+			rc = 0;
+		else
+			rc = -EBUSY;
+		goto out;
+	}
+
+	p = ocfs2_stack_lookup(plugin_name);
+	if (!p || !try_module_get(p->sp_owner)) {
+		rc = -ENOENT;
+		goto out;
+	}
+
+	active_stack = p;
+	rc = 0;
+
+out:
+	/* If we found it, pin it */
+	if (!rc)
+		active_stack->sp_count++;
+
+	spin_unlock(&ocfs2_stack_lock);
+	return rc;
+}
+
+/*
+ * This function looks up the appropriate stack and makes it active.  If
+ * there is no stack, it tries to load it.  It will fail if the stack still
+ * cannot be found.  It will also fail if a different stack is in use.
+ */
+static int ocfs2_stack_driver_get(const char *stack_name)
+{
+	int rc;
+	char *plugin_name = OCFS2_STACK_PLUGIN_O2CB;
+
+	/*
+	 * Classic stack does not pass in a stack name.  This is
+	 * compatible with older tools as well.
+	 */
+	if (!stack_name || !*stack_name)
+		stack_name = OCFS2_STACK_PLUGIN_O2CB;
+
+	if (strlen(stack_name) != OCFS2_STACK_LABEL_LEN) {
+		printk(KERN_ERR
+		       "ocfs2 passed an invalid cluster stack label: \"%s\"\n",
+		       stack_name);
+		return -EINVAL;
+	}
+
+	/* Anything that isn't the classic stack is a user stack */
+	if (strcmp(stack_name, OCFS2_STACK_PLUGIN_O2CB))
+		plugin_name = OCFS2_STACK_PLUGIN_USER;
+
+	rc = ocfs2_stack_driver_request(stack_name, plugin_name);
+	if (rc == -ENOENT) {
+		request_module("ocfs2_stack_%s", plugin_name);
+		rc = ocfs2_stack_driver_request(stack_name, plugin_name);
+	}
+
+	if (rc == -ENOENT) {
+		printk(KERN_ERR
+		       "ocfs2: Cluster stack driver \"%s\" cannot be found\n",
+		       plugin_name);
+	} else if (rc == -EBUSY) {
+		printk(KERN_ERR
+		       "ocfs2: A different cluster stack is in use\n");
+	}
+
+	return rc;
+}
+
+static void ocfs2_stack_driver_put(void)
+{
+	spin_lock(&ocfs2_stack_lock);
+	BUG_ON(active_stack == NULL);
+	BUG_ON(active_stack->sp_count == 0);
+
+	active_stack->sp_count--;
+	if (!active_stack->sp_count) {
+		module_put(active_stack->sp_owner);
+		active_stack = NULL;
+	}
+	spin_unlock(&ocfs2_stack_lock);
+}
+
+int ocfs2_stack_glue_register(struct ocfs2_stack_plugin *plugin)
+{
+	int rc;
+
+	spin_lock(&ocfs2_stack_lock);
+	if (!ocfs2_stack_lookup(plugin->sp_name)) {
+		plugin->sp_count = 0;
+		plugin->sp_max_proto = locking_max_version;
+		list_add(&plugin->sp_list, &ocfs2_stack_list);
+		printk(KERN_INFO "ocfs2: Registered cluster interface %s\n",
+		       plugin->sp_name);
+		rc = 0;
+	} else {
+		printk(KERN_ERR "ocfs2: Stack \"%s\" already registered\n",
+		       plugin->sp_name);
+		rc = -EEXIST;
+	}
+	spin_unlock(&ocfs2_stack_lock);
+
+	return rc;
+}
+EXPORT_SYMBOL_GPL(ocfs2_stack_glue_register);
+
+void ocfs2_stack_glue_unregister(struct ocfs2_stack_plugin *plugin)
+{
+	struct ocfs2_stack_plugin *p;
+
+	spin_lock(&ocfs2_stack_lock);
+	p = ocfs2_stack_lookup(plugin->sp_name);
+	if (p) {
+		BUG_ON(p != plugin);
+		BUG_ON(plugin == active_stack);
+		BUG_ON(plugin->sp_count != 0);
+		list_del_init(&plugin->sp_list);
+		printk(KERN_INFO "ocfs2: Unregistered cluster interface %s\n",
+		       plugin->sp_name);
+	} else {
+		printk(KERN_ERR "Stack \"%s\" is not registered\n",
+		       plugin->sp_name);
+	}
+	spin_unlock(&ocfs2_stack_lock);
+}
+EXPORT_SYMBOL_GPL(ocfs2_stack_glue_unregister);
+
+void ocfs2_stack_glue_set_max_proto_version(struct ocfs2_protocol_version *max_proto)
+{
+	struct ocfs2_stack_plugin *p;
+
+	spin_lock(&ocfs2_stack_lock);
+	if (memcmp(max_proto, &locking_max_version,
+		   sizeof(struct ocfs2_protocol_version))) {
+		BUG_ON(locking_max_version.pv_major != 0);
+
+		locking_max_version = *max_proto;
+		list_for_each_entry(p, &ocfs2_stack_list, sp_list) {
+			p->sp_max_proto = locking_max_version;
+		}
+	}
+	spin_unlock(&ocfs2_stack_lock);
+}
+EXPORT_SYMBOL_GPL(ocfs2_stack_glue_set_max_proto_version);
+
+
+/*
+ * The ocfs2_dlm_lock() and ocfs2_dlm_unlock() functions take no argument
+ * for the ast and bast functions.  They will pass the lksb to the ast
+ * and bast.  The caller can wrap the lksb with their own structure to
+ * get more information.
+ */
+int ocfs2_dlm_lock(struct ocfs2_cluster_connection *conn,
+		   int mode,
+		   struct ocfs2_dlm_lksb *lksb,
+		   u32 flags,
+		   void *name,
+		   unsigned int namelen)
+{
+	if (!lksb->lksb_conn)
+		lksb->lksb_conn = conn;
+	else
+		BUG_ON(lksb->lksb_conn != conn);
+	return active_stack->sp_ops->dlm_lock(conn, mode, lksb, flags,
+					      name, namelen);
+}
+EXPORT_SYMBOL_GPL(ocfs2_dlm_lock);
+
+int ocfs2_dlm_unlock(struct ocfs2_cluster_connection *conn,
+		     struct ocfs2_dlm_lksb *lksb,
+		     u32 flags)
+{
+	BUG_ON(lksb->lksb_conn == NULL);
+
+	return active_stack->sp_ops->dlm_unlock(conn, lksb, flags);
+}
+EXPORT_SYMBOL_GPL(ocfs2_dlm_unlock);
+
+int ocfs2_dlm_lock_status(struct ocfs2_dlm_lksb *lksb)
+{
+	return active_stack->sp_ops->lock_status(lksb);
+}
+EXPORT_SYMBOL_GPL(ocfs2_dlm_lock_status);
+
+int ocfs2_dlm_lvb_valid(struct ocfs2_dlm_lksb *lksb)
+{
+	return active_stack->sp_ops->lvb_valid(lksb);
+}
+EXPORT_SYMBOL_GPL(ocfs2_dlm_lvb_valid);
+
+void *ocfs2_dlm_lvb(struct ocfs2_dlm_lksb *lksb)
+{
+	return active_stack->sp_ops->lock_lvb(lksb);
+}
+EXPORT_SYMBOL_GPL(ocfs2_dlm_lvb);
+
+void ocfs2_dlm_dump_lksb(struct ocfs2_dlm_lksb *lksb)
+{
+	active_stack->sp_ops->dump_lksb(lksb);
+}
+EXPORT_SYMBOL_GPL(ocfs2_dlm_dump_lksb);
+
+int ocfs2_stack_supports_plocks(void)
+{
+	return active_stack && active_stack->sp_ops->plock;
+}
+EXPORT_SYMBOL_GPL(ocfs2_stack_supports_plocks);
+
+/*
+ * ocfs2_plock() can only be safely called if
+ * ocfs2_stack_supports_plocks() returned true
+ */
+int ocfs2_plock(struct ocfs2_cluster_connection *conn, u64 ino,
+		struct file *file, int cmd, struct file_lock *fl)
+{
+	WARN_ON_ONCE(active_stack->sp_ops->plock == NULL);
+	if (active_stack->sp_ops->plock)
+		return active_stack->sp_ops->plock(conn, ino, file, cmd, fl);
+	return -EOPNOTSUPP;
+}
+EXPORT_SYMBOL_GPL(ocfs2_plock);
+
+int ocfs2_cluster_connect(const char *stack_name,
+			  const char *cluster_name,
+			  int cluster_name_len,
+			  const char *group,
+			  int grouplen,
+			  struct ocfs2_locking_protocol *lproto,
+			  void (*recovery_handler)(int node_num,
+						   void *recovery_data),
+			  void *recovery_data,
+			  struct ocfs2_cluster_connection **conn)
+{
+	int rc = 0;
+	struct ocfs2_cluster_connection *new_conn;
+
+	BUG_ON(group == NULL);
+	BUG_ON(conn == NULL);
+	BUG_ON(recovery_handler == NULL);
+
+	if (grouplen > GROUP_NAME_MAX) {
+		rc = -EINVAL;
+		goto out;
+	}
+
+	if (memcmp(&lproto->lp_max_version, &locking_max_version,
+		   sizeof(struct ocfs2_protocol_version))) {
+		rc = -EINVAL;
+		goto out;
+	}
+
+	new_conn = kzalloc(sizeof(struct ocfs2_cluster_connection),
+			   GFP_KERNEL);
+	if (!new_conn) {
+		rc = -ENOMEM;
+		goto out;
+	}
+
+	strlcpy(new_conn->cc_name, group, GROUP_NAME_MAX + 1);
+	new_conn->cc_namelen = grouplen;
+	if (cluster_name_len)
+		strlcpy(new_conn->cc_cluster_name, cluster_name,
+			CLUSTER_NAME_MAX + 1);
+	new_conn->cc_cluster_name_len = cluster_name_len;
+	new_conn->cc_recovery_handler = recovery_handler;
+	new_conn->cc_recovery_data = recovery_data;
+
+	new_conn->cc_proto = lproto;
+	/* Start the new connection at our maximum compatibility level */
+	new_conn->cc_version = lproto->lp_max_version;
+
+	/* This will pin the stack driver if successful */
+	rc = ocfs2_stack_driver_get(stack_name);
+	if (rc)
+		goto out_free;
+
+	rc = active_stack->sp_ops->connect(new_conn);
+	if (rc) {
+		ocfs2_stack_driver_put();
+		goto out_free;
+	}
+
+	*conn = new_conn;
+
+out_free:
+	if (rc)
+		kfree(new_conn);
+
+out:
+	return rc;
+}
+EXPORT_SYMBOL_GPL(ocfs2_cluster_connect);
+
+/* The caller will ensure all nodes have the same cluster stack */
+int ocfs2_cluster_connect_agnostic(const char *group,
+				   int grouplen,
+				   struct ocfs2_locking_protocol *lproto,
+				   void (*recovery_handler)(int node_num,
+							    void *recovery_data),
+				   void *recovery_data,
+				   struct ocfs2_cluster_connection **conn)
+{
+	char *stack_name = NULL;
+
+	if (cluster_stack_name[0])
+		stack_name = cluster_stack_name;
+	return ocfs2_cluster_connect(stack_name, NULL, 0, group, grouplen,
+				     lproto, recovery_handler, recovery_data,
+				     conn);
+}
+EXPORT_SYMBOL_GPL(ocfs2_cluster_connect_agnostic);
+
+/* If hangup_pending is 0, the stack driver will be dropped */
+int ocfs2_cluster_disconnect(struct ocfs2_cluster_connection *conn,
+			     int hangup_pending)
+{
+	int ret;
+
+	BUG_ON(conn == NULL);
+
+	ret = active_stack->sp_ops->disconnect(conn);
+
+	/* XXX Should we free it anyway? */
+	if (!ret) {
+		kfree(conn);
+		if (!hangup_pending)
+			ocfs2_stack_driver_put();
+	}
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(ocfs2_cluster_disconnect);
+
+/*
+ * Leave the group for this filesystem.  This is executed by a userspace
+ * program (stored in ocfs2_hb_ctl_path).
+ */
+static void ocfs2_leave_group(const char *group)
+{
+	int ret;
+	char *argv[5], *envp[3];
+
+	argv[0] = ocfs2_hb_ctl_path;
+	argv[1] = "-K";
+	argv[2] = "-u";
+	argv[3] = (char *)group;
+	argv[4] = NULL;
+
+	/* minimal command environment taken from cpu_run_sbin_hotplug */
+	envp[0] = "HOME=/";
+	envp[1] = "PATH=/sbin:/bin:/usr/sbin:/usr/bin";
+	envp[2] = NULL;
+
+	ret = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_PROC);
+	if (ret < 0) {
+		printk(KERN_ERR
+		       "ocfs2: Error %d running user helper "
+		       "\"%s %s %s %s\"\n",
+		       ret, argv[0], argv[1], argv[2], argv[3]);
+	}
+}
+
+/*
+ * Hangup is a required post-umount.  ocfs2-tools software expects the
+ * filesystem to call "ocfs2_hb_ctl" during unmount.  This happens
+ * regardless of whether the DLM got started, so we can't do it
+ * in ocfs2_cluster_disconnect().  The ocfs2_leave_group() function does
+ * the actual work.
+ */
+void ocfs2_cluster_hangup(const char *group, int grouplen)
+{
+	BUG_ON(group == NULL);
+	BUG_ON(group[grouplen] != '\0');
+
+	ocfs2_leave_group(group);
+
+	/* cluster_disconnect() was called with hangup_pending==1 */
+	ocfs2_stack_driver_put();
+}
+EXPORT_SYMBOL_GPL(ocfs2_cluster_hangup);
+
+int ocfs2_cluster_this_node(struct ocfs2_cluster_connection *conn,
+			    unsigned int *node)
+{
+	return active_stack->sp_ops->this_node(conn, node);
+}
+EXPORT_SYMBOL_GPL(ocfs2_cluster_this_node);
+
+
+/*
+ * Sysfs bits
+ */
+
+static ssize_t ocfs2_max_locking_protocol_show(struct kobject *kobj,
+					       struct kobj_attribute *attr,
+					       char *buf)
+{
+	ssize_t ret = 0;
+
+	spin_lock(&ocfs2_stack_lock);
+	if (locking_max_version.pv_major)
+		ret = snprintf(buf, PAGE_SIZE, "%u.%u\n",
+			       locking_max_version.pv_major,
+			       locking_max_version.pv_minor);
+	spin_unlock(&ocfs2_stack_lock);
+
+	return ret;
+}
+
+static struct kobj_attribute ocfs2_attr_max_locking_protocol =
+	__ATTR(max_locking_protocol, S_IRUGO,
+	       ocfs2_max_locking_protocol_show, NULL);
+
+static ssize_t ocfs2_loaded_cluster_plugins_show(struct kobject *kobj,
+						 struct kobj_attribute *attr,
+						 char *buf)
+{
+	ssize_t ret = 0, total = 0, remain = PAGE_SIZE;
+	struct ocfs2_stack_plugin *p;
+
+	spin_lock(&ocfs2_stack_lock);
+	list_for_each_entry(p, &ocfs2_stack_list, sp_list) {
+		ret = snprintf(buf, remain, "%s\n",
+			       p->sp_name);
+		if (ret < 0) {
+			total = ret;
+			break;
+		}
+		if (ret == remain) {
+			/* snprintf() didn't fit */
+			total = -E2BIG;
+			break;
+		}
+		total += ret;
+		remain -= ret;
+	}
+	spin_unlock(&ocfs2_stack_lock);
+
+	return total;
+}
+
+static struct kobj_attribute ocfs2_attr_loaded_cluster_plugins =
+	__ATTR(loaded_cluster_plugins, S_IRUGO,
+	       ocfs2_loaded_cluster_plugins_show, NULL);
+
+static ssize_t ocfs2_active_cluster_plugin_show(struct kobject *kobj,
+						struct kobj_attribute *attr,
+						char *buf)
+{
+	ssize_t ret = 0;
+
+	spin_lock(&ocfs2_stack_lock);
+	if (active_stack) {
+		ret = snprintf(buf, PAGE_SIZE, "%s\n",
+			       active_stack->sp_name);
+		if (ret == PAGE_SIZE)
+			ret = -E2BIG;
+	}
+	spin_unlock(&ocfs2_stack_lock);
+
+	return ret;
+}
+
+static struct kobj_attribute ocfs2_attr_active_cluster_plugin =
+	__ATTR(active_cluster_plugin, S_IRUGO,
+	       ocfs2_active_cluster_plugin_show, NULL);
+
+static ssize_t ocfs2_cluster_stack_show(struct kobject *kobj,
+					struct kobj_attribute *attr,
+					char *buf)
+{
+	ssize_t ret;
+	spin_lock(&ocfs2_stack_lock);
+	ret = snprintf(buf, PAGE_SIZE, "%s\n", cluster_stack_name);
+	spin_unlock(&ocfs2_stack_lock);
+
+	return ret;
+}
+
+static ssize_t ocfs2_cluster_stack_store(struct kobject *kobj,
+					 struct kobj_attribute *attr,
+					 const char *buf, size_t count)
+{
+	size_t len = count;
+	ssize_t ret;
+
+	if (len == 0)
+		return len;
+
+	if (buf[len - 1] == '\n')
+		len--;
+
+	if ((len != OCFS2_STACK_LABEL_LEN) ||
+	    (strnlen(buf, len) != len))
+		return -EINVAL;
+
+	spin_lock(&ocfs2_stack_lock);
+	if (active_stack) {
+		if (!strncmp(buf, cluster_stack_name, len))
+			ret = count;
+		else
+			ret = -EBUSY;
+	} else {
+		memcpy(cluster_stack_name, buf, len);
+		ret = count;
+	}
+	spin_unlock(&ocfs2_stack_lock);
+
+	return ret;
+}
+
+
+static struct kobj_attribute ocfs2_attr_cluster_stack =
+	__ATTR(cluster_stack, S_IRUGO | S_IWUSR,
+	       ocfs2_cluster_stack_show,
+	       ocfs2_cluster_stack_store);
+
+
+
+static ssize_t ocfs2_dlm_recover_show(struct kobject *kobj,
+					struct kobj_attribute *attr,
+					char *buf)
+{
+	return snprintf(buf, PAGE_SIZE, "1\n");
+}
+
+static struct kobj_attribute ocfs2_attr_dlm_recover_support =
+	__ATTR(dlm_recover_callback_support, S_IRUGO,
+	       ocfs2_dlm_recover_show, NULL);
+
+static struct attribute *ocfs2_attrs[] = {
+	&ocfs2_attr_max_locking_protocol.attr,
+	&ocfs2_attr_loaded_cluster_plugins.attr,
+	&ocfs2_attr_active_cluster_plugin.attr,
+	&ocfs2_attr_cluster_stack.attr,
+	&ocfs2_attr_dlm_recover_support.attr,
+	NULL,
+};
+
+static struct attribute_group ocfs2_attr_group = {
+	.attrs = ocfs2_attrs,
+};
+
+static struct kset *ocfs2_kset;
+
+static void ocfs2_sysfs_exit(void)
+{
+	kset_unregister(ocfs2_kset);
+}
+
+static int ocfs2_sysfs_init(void)
+{
+	int ret;
+
+	ocfs2_kset = kset_create_and_add("ocfs2", NULL, fs_kobj);
+	if (!ocfs2_kset)
+		return -ENOMEM;
+
+	ret = sysfs_create_group(&ocfs2_kset->kobj, &ocfs2_attr_group);
+	if (ret)
+		goto error;
+
+	return 0;
+
+error:
+	kset_unregister(ocfs2_kset);
+	return ret;
+}
+
+/*
+ * Sysctl bits
+ *
+ * The sysctl lives at /proc/sys/fs/ocfs2/nm/hb_ctl_path.  The 'nm' doesn't
+ * make as much sense in a multiple cluster stack world, but it's safer
+ * and easier to preserve the name.
+ */
+
+#define FS_OCFS2_NM		1
+
+static struct ctl_table ocfs2_nm_table[] = {
+	{
+		.procname	= "hb_ctl_path",
+		.data		= ocfs2_hb_ctl_path,
+		.maxlen		= OCFS2_MAX_HB_CTL_PATH,
+		.mode		= 0644,
+		.proc_handler	= proc_dostring,
+	},
+	{ }
+};
+
+static struct ctl_table ocfs2_mod_table[] = {
+	{
+		.procname	= "nm",
+		.data		= NULL,
+		.maxlen		= 0,
+		.mode		= 0555,
+		.child		= ocfs2_nm_table
+	},
+	{ }
+};
+
+static struct ctl_table ocfs2_kern_table[] = {
+	{
+		.procname	= "ocfs2",
+		.data		= NULL,
+		.maxlen		= 0,
+		.mode		= 0555,
+		.child		= ocfs2_mod_table
+	},
+	{ }
+};
+
+static struct ctl_table ocfs2_root_table[] = {
+	{
+		.procname	= "fs",
+		.data		= NULL,
+		.maxlen		= 0,
+		.mode		= 0555,
+		.child		= ocfs2_kern_table
+	},
+	{ }
+};
+
+static struct ctl_table_header *ocfs2_table_header;
+
+
+/*
+ * Initialization
+ */
+
+static int __init ocfs2_stack_glue_init(void)
+{
+	strcpy(cluster_stack_name, OCFS2_STACK_PLUGIN_O2CB);
+
+	ocfs2_table_header = register_sysctl_table(ocfs2_root_table);
+	if (!ocfs2_table_header) {
+		printk(KERN_ERR
+		       "ocfs2 stack glue: unable to register sysctl\n");
+		return -ENOMEM; /* or something. */
+	}
+
+	return ocfs2_sysfs_init();
+}
+
+static void __exit ocfs2_stack_glue_exit(void)
+{
+	memset(&locking_max_version, 0,
+	       sizeof(struct ocfs2_protocol_version));
+	locking_max_version.pv_major = 0;
+	locking_max_version.pv_minor = 0;
+	ocfs2_sysfs_exit();
+	if (ocfs2_table_header)
+		unregister_sysctl_table(ocfs2_table_header);
+}
+
+MODULE_AUTHOR("Oracle");
+MODULE_DESCRIPTION("ocfs2 cluter stack glue layer");
+MODULE_LICENSE("GPL");
+module_init(ocfs2_stack_glue_init);
+module_exit(ocfs2_stack_glue_exit);

fs/ocfs2/stackglue.h

@@ -0,0 +1,301 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * stackglue.h
+ *
+ * Glue to the underlying cluster stack.
+ *
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation, version 2.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+
+#ifndef STACKGLUE_H
+#define STACKGLUE_H
+
+#include <linux/types.h>
+#include <linux/list.h>
+#include <linux/dlmconstants.h>
+
+#include "dlm/dlmapi.h"
+#include <linux/dlm.h>
+
+/* Needed for plock-related prototypes */
+struct file;
+struct file_lock;
+
+/*
+ * dlmconstants.h does not have a LOCAL flag.  We hope to remove it
+ * some day, but right now we need it.  Let's fake it.  This value is larger
+ * than any flag in dlmconstants.h.
+ */
+#define DLM_LKF_LOCAL		0x00100000
+
+/*
+ * This shadows DLM_LOCKSPACE_LEN in fs/dlm/dlm_internal.h.  That probably
+ * wants to be in a public header.
+ */
+#define GROUP_NAME_MAX		64
+
+/* This shadows  OCFS2_CLUSTER_NAME_LEN */
+#define CLUSTER_NAME_MAX	16
+
+
+/*
+ * ocfs2_protocol_version changes when ocfs2 does something different in
+ * its inter-node behavior.  See dlmglue.c for more information.
+ */
+struct ocfs2_protocol_version {
+	u8 pv_major;
+	u8 pv_minor;
+};
+
+/*
+ * The dlm_lockstatus struct includes lvb space, but the dlm_lksb struct only
+ * has a pointer to separately allocated lvb space.  This struct exists only to
+ * include in the lksb union to make space for a combined dlm_lksb and lvb.
+ */
+struct fsdlm_lksb_plus_lvb {
+	struct dlm_lksb lksb;
+	char lvb[DLM_LVB_LEN];
+};
+
+/*
+ * A union of all lock status structures.  We define it here so that the
+ * size of the union is known.  Lock status structures are embedded in
+ * ocfs2 inodes.
+ */
+struct ocfs2_cluster_connection;
+struct ocfs2_dlm_lksb {
+	 union {
+		 struct dlm_lockstatus lksb_o2dlm;
+		 struct dlm_lksb lksb_fsdlm;
+		 struct fsdlm_lksb_plus_lvb padding;
+	 };
+	 struct ocfs2_cluster_connection *lksb_conn;
+};
+
+/*
+ * The ocfs2_locking_protocol defines the handlers called on ocfs2's behalf.
+ */
+struct ocfs2_locking_protocol {
+	struct ocfs2_protocol_version lp_max_version;
+	void (*lp_lock_ast)(struct ocfs2_dlm_lksb *lksb);
+	void (*lp_blocking_ast)(struct ocfs2_dlm_lksb *lksb, int level);
+	void (*lp_unlock_ast)(struct ocfs2_dlm_lksb *lksb, int error);
+};
+
+
+/*
+ * A cluster connection.  Mostly opaque to ocfs2, the connection holds
+ * state for the underlying stack.  ocfs2 does use cc_version to determine
+ * locking compatibility.
+ */
+struct ocfs2_cluster_connection {
+	char cc_name[GROUP_NAME_MAX + 1];
+	int cc_namelen;
+	char cc_cluster_name[CLUSTER_NAME_MAX + 1];
+	int cc_cluster_name_len;
+	struct ocfs2_protocol_version cc_version;
+	struct ocfs2_locking_protocol *cc_proto;
+	void (*cc_recovery_handler)(int node_num, void *recovery_data);
+	void *cc_recovery_data;
+	void *cc_lockspace;
+	void *cc_private;
+};
+
+/*
+ * Each cluster stack implements the stack operations structure.  Not used
+ * in the ocfs2 code, the stackglue code translates generic cluster calls
+ * into stack operations.
+ */
+struct ocfs2_stack_operations {
+	/*
+	 * The fs code calls ocfs2_cluster_connect() to attach a new
+	 * filesystem to the cluster stack.  The ->connect() op is passed
+	 * an ocfs2_cluster_connection with the name and recovery field
+	 * filled in.
+	 *
+	 * The stack must set up any notification mechanisms and create
+	 * the filesystem lockspace in the DLM.  The lockspace should be
+	 * stored on cc_lockspace.  Any other information can be stored on
+	 * cc_private.
+	 *
+	 * ->connect() must not return until it is guaranteed that
+	 *
+	 *  - Node down notifications for the filesystem will be received
+	 *    and passed to conn->cc_recovery_handler().
+	 *  - Locking requests for the filesystem will be processed.
+	 */
+	int (*connect)(struct ocfs2_cluster_connection *conn);
+
+	/*
+	 * The fs code calls ocfs2_cluster_disconnect() when a filesystem
+	 * no longer needs cluster services.  All DLM locks have been
+	 * dropped, and recovery notification is being ignored by the
+	 * fs code.  The stack must disengage from the DLM and discontinue
+	 * recovery notification.
+	 *
+	 * Once ->disconnect() has returned, the connection structure will
+	 * be freed.  Thus, a stack must not return from ->disconnect()
+	 * until it will no longer reference the conn pointer.
+	 *
+	 * Once this call returns, the stack glue will be dropping this
+	 * connection's reference on the module.
+	 */
+	int (*disconnect)(struct ocfs2_cluster_connection *conn);
+
+	/*
+	 * ->this_node() returns the cluster's unique identifier for the
+	 * local node.
+	 */
+	int (*this_node)(struct ocfs2_cluster_connection *conn,
+			 unsigned int *node);
+
+	/*
+	 * Call the underlying dlm lock function.  The ->dlm_lock()
+	 * callback should convert the flags and mode as appropriate.
+	 *
+	 * ast and bast functions are not part of the call because the
+	 * stack will likely want to wrap ast and bast calls before passing
+	 * them to stack->sp_proto.  There is no astarg.  The lksb will
+	 * be passed back to the ast and bast functions.  The caller can
+	 * use this to find their object.
+	 */
+	int (*dlm_lock)(struct ocfs2_cluster_connection *conn,
+			int mode,
+			struct ocfs2_dlm_lksb *lksb,
+			u32 flags,
+			void *name,
+			unsigned int namelen);
+
+	/*
+	 * Call the underlying dlm unlock function.  The ->dlm_unlock()
+	 * function should convert the flags as appropriate.
+	 *
+	 * The unlock ast is not passed, as the stack will want to wrap
+	 * it before calling stack->sp_proto->lp_unlock_ast().  There is
+	 * no astarg.  The lksb will be passed back to the unlock ast
+	 * function.  The caller can use this to find their object.
+	 */
+	int (*dlm_unlock)(struct ocfs2_cluster_connection *conn,
+			  struct ocfs2_dlm_lksb *lksb,
+			  u32 flags);
+
+	/*
+	 * Return the status of the current lock status block.  The fs
+	 * code should never dereference the union.  The ->lock_status()
+	 * callback pulls out the stack-specific lksb, converts the status
+	 * to a proper errno, and returns it.
+	 */
+	int (*lock_status)(struct ocfs2_dlm_lksb *lksb);
+
+	/*
+	 * Return non-zero if the LVB is valid.
+	 */
+	int (*lvb_valid)(struct ocfs2_dlm_lksb *lksb);
+
+	/*
+	 * Pull the lvb pointer off of the stack-specific lksb.
+	 */
+	void *(*lock_lvb)(struct ocfs2_dlm_lksb *lksb);
+
+	/*
+	 * Cluster-aware posix locks
+	 *
+	 * This is NULL for stacks which do not support posix locks.
+	 */
+	int (*plock)(struct ocfs2_cluster_connection *conn,
+		     u64 ino,
+		     struct file *file,
+		     int cmd,
+		     struct file_lock *fl);
+
+	/*
+	 * This is an optoinal debugging hook.  If provided, the
+	 * stack can dump debugging information about this lock.
+	 */
+	void (*dump_lksb)(struct ocfs2_dlm_lksb *lksb);
+};
+
+/*
+ * Each stack plugin must describe itself by registering a
+ * ocfs2_stack_plugin structure.  This is only seen by stackglue and the
+ * stack driver.
+ */
+struct ocfs2_stack_plugin {
+	char *sp_name;
+	struct ocfs2_stack_operations *sp_ops;
+	struct module *sp_owner;
+
+	/* These are managed by the stackglue code. */
+	struct list_head sp_list;
+	unsigned int sp_count;
+	struct ocfs2_protocol_version sp_max_proto;
+};
+
+
+/* Used by the filesystem */
+int ocfs2_cluster_connect(const char *stack_name,
+			  const char *cluster_name,
+			  int cluster_name_len,
+			  const char *group,
+			  int grouplen,
+			  struct ocfs2_locking_protocol *lproto,
+			  void (*recovery_handler)(int node_num,
+						   void *recovery_data),
+			  void *recovery_data,
+			  struct ocfs2_cluster_connection **conn);
+/*
+ * Used by callers that don't store their stack name.  They must ensure
+ * all nodes have the same stack.
+ */
+int ocfs2_cluster_connect_agnostic(const char *group,
+				   int grouplen,
+				   struct ocfs2_locking_protocol *lproto,
+				   void (*recovery_handler)(int node_num,
+							    void *recovery_data),
+				   void *recovery_data,
+				   struct ocfs2_cluster_connection **conn);
+int ocfs2_cluster_disconnect(struct ocfs2_cluster_connection *conn,
+			     int hangup_pending);
+void ocfs2_cluster_hangup(const char *group, int grouplen);
+int ocfs2_cluster_this_node(struct ocfs2_cluster_connection *conn,
+			    unsigned int *node);
+
+struct ocfs2_lock_res;
+int ocfs2_dlm_lock(struct ocfs2_cluster_connection *conn,
+		   int mode,
+		   struct ocfs2_dlm_lksb *lksb,
+		   u32 flags,
+		   void *name,
+		   unsigned int namelen);
+int ocfs2_dlm_unlock(struct ocfs2_cluster_connection *conn,
+		     struct ocfs2_dlm_lksb *lksb,
+		     u32 flags);
+
+int ocfs2_dlm_lock_status(struct ocfs2_dlm_lksb *lksb);
+int ocfs2_dlm_lvb_valid(struct ocfs2_dlm_lksb *lksb);
+void *ocfs2_dlm_lvb(struct ocfs2_dlm_lksb *lksb);
+void ocfs2_dlm_dump_lksb(struct ocfs2_dlm_lksb *lksb);
+
+int ocfs2_stack_supports_plocks(void);
+int ocfs2_plock(struct ocfs2_cluster_connection *conn, u64 ino,
+		struct file *file, int cmd, struct file_lock *fl);
+
+void ocfs2_stack_glue_set_max_proto_version(struct ocfs2_protocol_version *max_proto);
+
+
+/* Used by stack plugins */
+int ocfs2_stack_glue_register(struct ocfs2_stack_plugin *plugin);
+void ocfs2_stack_glue_unregister(struct ocfs2_stack_plugin *plugin);
+
+#endif  /* STACKGLUE_H */