Fixed MTP to work with TWRP

arch/ia64/mm/Makefile

@@ -0,0 +1,11 @@
+#
+# Makefile for the ia64-specific parts of the memory manager.
+#
+
+obj-y := init.o fault.o tlb.o extable.o ioremap.o
+
+obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
+obj-$(CONFIG_NUMA)	   += numa.o
+obj-$(CONFIG_DISCONTIGMEM) += discontig.o
+obj-$(CONFIG_SPARSEMEM)	   += discontig.o
+obj-$(CONFIG_FLATMEM)	   += contig.o

arch/ia64/mm/contig.c

@@ -0,0 +1,278 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1998-2003 Hewlett-Packard Co
+ *	David Mosberger-Tang <davidm@hpl.hp.com>
+ *	Stephane Eranian <eranian@hpl.hp.com>
+ * Copyright (C) 2000, Rohit Seth <rohit.seth@intel.com>
+ * Copyright (C) 1999 VA Linux Systems
+ * Copyright (C) 1999 Walt Drummond <drummond@valinux.com>
+ * Copyright (C) 2003 Silicon Graphics, Inc. All rights reserved.
+ *
+ * Routines used by ia64 machines with contiguous (or virtually contiguous)
+ * memory.
+ */
+#include <linux/bootmem.h>
+#include <linux/efi.h>
+#include <linux/memblock.h>
+#include <linux/mm.h>
+#include <linux/nmi.h>
+#include <linux/swap.h>
+
+#include <asm/meminit.h>
+#include <asm/pgalloc.h>
+#include <asm/pgtable.h>
+#include <asm/sections.h>
+#include <asm/mca.h>
+
+#ifdef CONFIG_VIRTUAL_MEM_MAP
+static unsigned long max_gap;
+#endif
+
+/* physical address where the bootmem map is located */
+unsigned long bootmap_start;
+
+/**
+ * find_bootmap_location - callback to find a memory area for the bootmap
+ * @start: start of region
+ * @end: end of region
+ * @arg: unused callback data
+ *
+ * Find a place to put the bootmap and return its starting address in
+ * bootmap_start.  This address must be page-aligned.
+ */
+static int __init
+find_bootmap_location (u64 start, u64 end, void *arg)
+{
+	u64 needed = *(unsigned long *)arg;
+	u64 range_start, range_end, free_start;
+	int i;
+
+#if IGNORE_PFN0
+	if (start == PAGE_OFFSET) {
+		start += PAGE_SIZE;
+		if (start >= end)
+			return 0;
+	}
+#endif
+
+	free_start = PAGE_OFFSET;
+
+	for (i = 0; i < num_rsvd_regions; i++) {
+		range_start = max(start, free_start);
+		range_end   = min(end, rsvd_region[i].start & PAGE_MASK);
+
+		free_start = PAGE_ALIGN(rsvd_region[i].end);
+
+		if (range_end <= range_start)
+			continue; /* skip over empty range */
+
+		if (range_end - range_start >= needed) {
+			bootmap_start = __pa(range_start);
+			return -1;	/* done */
+		}
+
+		/* nothing more available in this segment */
+		if (range_end == end)
+			return 0;
+	}
+	return 0;
+}
+
+#ifdef CONFIG_SMP
+static void *cpu_data;
+/**
+ * per_cpu_init - setup per-cpu variables
+ *
+ * Allocate and setup per-cpu data areas.
+ */
+void *per_cpu_init(void)
+{
+	static bool first_time = true;
+	void *cpu0_data = __cpu0_per_cpu;
+	unsigned int cpu;
+
+	if (!first_time)
+		goto skip;
+	first_time = false;
+
+	/*
+	 * get_free_pages() cannot be used before cpu_init() done.
+	 * BSP allocates PERCPU_PAGE_SIZE bytes for all possible CPUs
+	 * to avoid that AP calls get_zeroed_page().
+	 */
+	for_each_possible_cpu(cpu) {
+		void *src = cpu == 0 ? cpu0_data : __phys_per_cpu_start;
+
+		memcpy(cpu_data, src, __per_cpu_end - __per_cpu_start);
+		__per_cpu_offset[cpu] = (char *)cpu_data - __per_cpu_start;
+		per_cpu(local_per_cpu_offset, cpu) = __per_cpu_offset[cpu];
+
+		/*
+		 * percpu area for cpu0 is moved from the __init area
+		 * which is setup by head.S and used till this point.
+		 * Update ar.k3.  This move is ensures that percpu
+		 * area for cpu0 is on the correct node and its
+		 * virtual address isn't insanely far from other
+		 * percpu areas which is important for congruent
+		 * percpu allocator.
+		 */
+		if (cpu == 0)
+			ia64_set_kr(IA64_KR_PER_CPU_DATA, __pa(cpu_data) -
+				    (unsigned long)__per_cpu_start);
+
+		cpu_data += PERCPU_PAGE_SIZE;
+	}
+skip:
+	return __per_cpu_start + __per_cpu_offset[smp_processor_id()];
+}
+
+static inline void
+alloc_per_cpu_data(void)
+{
+	cpu_data = __alloc_bootmem(PERCPU_PAGE_SIZE * num_possible_cpus(),
+				   PERCPU_PAGE_SIZE, __pa(MAX_DMA_ADDRESS));
+}
+
+/**
+ * setup_per_cpu_areas - setup percpu areas
+ *
+ * Arch code has already allocated and initialized percpu areas.  All
+ * this function has to do is to teach the determined layout to the
+ * dynamic percpu allocator, which happens to be more complex than
+ * creating whole new ones using helpers.
+ */
+void __init
+setup_per_cpu_areas(void)
+{
+	struct pcpu_alloc_info *ai;
+	struct pcpu_group_info *gi;
+	unsigned int cpu;
+	ssize_t static_size, reserved_size, dyn_size;
+	int rc;
+
+	ai = pcpu_alloc_alloc_info(1, num_possible_cpus());
+	if (!ai)
+		panic("failed to allocate pcpu_alloc_info");
+	gi = &ai->groups[0];
+
+	/* units are assigned consecutively to possible cpus */
+	for_each_possible_cpu(cpu)
+		gi->cpu_map[gi->nr_units++] = cpu;
+
+	/* set parameters */
+	static_size = __per_cpu_end - __per_cpu_start;
+	reserved_size = PERCPU_MODULE_RESERVE;
+	dyn_size = PERCPU_PAGE_SIZE - static_size - reserved_size;
+	if (dyn_size < 0)
+		panic("percpu area overflow static=%zd reserved=%zd\n",
+		      static_size, reserved_size);
+
+	ai->static_size		= static_size;
+	ai->reserved_size	= reserved_size;
+	ai->dyn_size		= dyn_size;
+	ai->unit_size		= PERCPU_PAGE_SIZE;
+	ai->atom_size		= PAGE_SIZE;
+	ai->alloc_size		= PERCPU_PAGE_SIZE;
+
+	rc = pcpu_setup_first_chunk(ai, __per_cpu_start + __per_cpu_offset[0]);
+	if (rc)
+		panic("failed to setup percpu area (err=%d)", rc);
+
+	pcpu_free_alloc_info(ai);
+}
+#else
+#define alloc_per_cpu_data() do { } while (0)
+#endif /* CONFIG_SMP */
+
+/**
+ * find_memory - setup memory map
+ *
+ * Walk the EFI memory map and find usable memory for the system, taking
+ * into account reserved areas.
+ */
+void __init
+find_memory (void)
+{
+	unsigned long bootmap_size;
+
+	reserve_memory();
+
+	/* first find highest page frame number */
+	min_low_pfn = ~0UL;
+	max_low_pfn = 0;
+	efi_memmap_walk(find_max_min_low_pfn, NULL);
+	max_pfn = max_low_pfn;
+	/* how many bytes to cover all the pages */
+	bootmap_size = bootmem_bootmap_pages(max_pfn) << PAGE_SHIFT;
+
+	/* look for a location to hold the bootmap */
+	bootmap_start = ~0UL;
+	efi_memmap_walk(find_bootmap_location, &bootmap_size);
+	if (bootmap_start == ~0UL)
+		panic("Cannot find %ld bytes for bootmap\n", bootmap_size);
+
+	bootmap_size = init_bootmem_node(NODE_DATA(0),
+			(bootmap_start >> PAGE_SHIFT), 0, max_pfn);
+
+	/* Free all available memory, then mark bootmem-map as being in use. */
+	efi_memmap_walk(filter_rsvd_memory, free_bootmem);
+	reserve_bootmem(bootmap_start, bootmap_size, BOOTMEM_DEFAULT);
+
+	find_initrd();
+
+	alloc_per_cpu_data();
+}
+
+/*
+ * Set up the page tables.
+ */
+
+void __init
+paging_init (void)
+{
+	unsigned long max_dma;
+	unsigned long max_zone_pfns[MAX_NR_ZONES];
+
+	memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
+#ifdef CONFIG_ZONE_DMA
+	max_dma = virt_to_phys((void *) MAX_DMA_ADDRESS) >> PAGE_SHIFT;
+	max_zone_pfns[ZONE_DMA] = max_dma;
+#endif
+	max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
+
+#ifdef CONFIG_VIRTUAL_MEM_MAP
+	efi_memmap_walk(filter_memory, register_active_ranges);
+	efi_memmap_walk(find_largest_hole, (u64 *)&max_gap);
+	if (max_gap < LARGE_GAP) {
+		vmem_map = (struct page *) 0;
+		free_area_init_nodes(max_zone_pfns);
+	} else {
+		unsigned long map_size;
+
+		/* allocate virtual_mem_map */
+
+		map_size = PAGE_ALIGN(ALIGN(max_low_pfn, MAX_ORDER_NR_PAGES) *
+			sizeof(struct page));
+		VMALLOC_END -= map_size;
+		vmem_map = (struct page *) VMALLOC_END;
+		efi_memmap_walk(create_mem_map_page_table, NULL);
+
+		/*
+		 * alloc_node_mem_map makes an adjustment for mem_map
+		 * which isn't compatible with vmem_map.
+		 */
+		NODE_DATA(0)->node_mem_map = vmem_map +
+			find_min_pfn_with_active_regions();
+		free_area_init_nodes(max_zone_pfns);
+
+		printk("Virtual mem_map starts at 0x%p\n", mem_map);
+	}
+#else /* !CONFIG_VIRTUAL_MEM_MAP */
+	memblock_add_node(0, PFN_PHYS(max_low_pfn), 0);
+	free_area_init_nodes(max_zone_pfns);
+#endif /* !CONFIG_VIRTUAL_MEM_MAP */
+	zero_page_memmap_ptr = virt_to_page(ia64_imva(empty_zero_page));
+}

arch/ia64/mm/discontig.c

@@ -0,0 +1,764 @@
+/*
+ * Copyright (c) 2000, 2003 Silicon Graphics, Inc.  All rights reserved.
+ * Copyright (c) 2001 Intel Corp.
+ * Copyright (c) 2001 Tony Luck <tony.luck@intel.com>
+ * Copyright (c) 2002 NEC Corp.
+ * Copyright (c) 2002 Kimio Suganuma <k-suganuma@da.jp.nec.com>
+ * Copyright (c) 2004 Silicon Graphics, Inc
+ *	Russ Anderson <rja@sgi.com>
+ *	Jesse Barnes <jbarnes@sgi.com>
+ *	Jack Steiner <steiner@sgi.com>
+ */
+
+/*
+ * Platform initialization for Discontig Memory
+ */
+
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/nmi.h>
+#include <linux/swap.h>
+#include <linux/bootmem.h>
+#include <linux/acpi.h>
+#include <linux/efi.h>
+#include <linux/nodemask.h>
+#include <linux/slab.h>
+#include <asm/pgalloc.h>
+#include <asm/tlb.h>
+#include <asm/meminit.h>
+#include <asm/numa.h>
+#include <asm/sections.h>
+
+/*
+ * Track per-node information needed to setup the boot memory allocator, the
+ * per-node areas, and the real VM.
+ */
+struct early_node_data {
+	struct ia64_node_data *node_data;
+	unsigned long pernode_addr;
+	unsigned long pernode_size;
+#ifdef CONFIG_ZONE_DMA
+	unsigned long num_dma_physpages;
+#endif
+	unsigned long min_pfn;
+	unsigned long max_pfn;
+};
+
+static struct early_node_data mem_data[MAX_NUMNODES] __initdata;
+static nodemask_t memory_less_mask __initdata;
+
+pg_data_t *pgdat_list[MAX_NUMNODES];
+
+/*
+ * To prevent cache aliasing effects, align per-node structures so that they
+ * start at addresses that are strided by node number.
+ */
+#define MAX_NODE_ALIGN_OFFSET	(32 * 1024 * 1024)
+#define NODEDATA_ALIGN(addr, node)						\
+	((((addr) + 1024*1024-1) & ~(1024*1024-1)) + 				\
+	     (((node)*PERCPU_PAGE_SIZE) & (MAX_NODE_ALIGN_OFFSET - 1)))
+
+/**
+ * build_node_maps - callback to setup bootmem structs for each node
+ * @start: physical start of range
+ * @len: length of range
+ * @node: node where this range resides
+ *
+ * We allocate a struct bootmem_data for each piece of memory that we wish to
+ * treat as a virtually contiguous block (i.e. each node). Each such block
+ * must start on an %IA64_GRANULE_SIZE boundary, so we round the address down
+ * if necessary.  Any non-existent pages will simply be part of the virtual
+ * memmap.  We also update min_low_pfn and max_low_pfn here as we receive
+ * memory ranges from the caller.
+ */
+static int __init build_node_maps(unsigned long start, unsigned long len,
+				  int node)
+{
+	unsigned long spfn, epfn, end = start + len;
+	struct bootmem_data *bdp = &bootmem_node_data[node];
+
+	epfn = GRANULEROUNDUP(end) >> PAGE_SHIFT;
+	spfn = GRANULEROUNDDOWN(start) >> PAGE_SHIFT;
+
+	if (!bdp->node_low_pfn) {
+		bdp->node_min_pfn = spfn;
+		bdp->node_low_pfn = epfn;
+	} else {
+		bdp->node_min_pfn = min(spfn, bdp->node_min_pfn);
+		bdp->node_low_pfn = max(epfn, bdp->node_low_pfn);
+	}
+
+	return 0;
+}
+
+/**
+ * early_nr_cpus_node - return number of cpus on a given node
+ * @node: node to check
+ *
+ * Count the number of cpus on @node.  We can't use nr_cpus_node() yet because
+ * acpi_boot_init() (which builds the node_to_cpu_mask array) hasn't been
+ * called yet.  Note that node 0 will also count all non-existent cpus.
+ */
+static int __meminit early_nr_cpus_node(int node)
+{
+	int cpu, n = 0;
+
+	for_each_possible_early_cpu(cpu)
+		if (node == node_cpuid[cpu].nid)
+			n++;
+
+	return n;
+}
+
+/**
+ * compute_pernodesize - compute size of pernode data
+ * @node: the node id.
+ */
+static unsigned long __meminit compute_pernodesize(int node)
+{
+	unsigned long pernodesize = 0, cpus;
+
+	cpus = early_nr_cpus_node(node);
+	pernodesize += PERCPU_PAGE_SIZE * cpus;
+	pernodesize += node * L1_CACHE_BYTES;
+	pernodesize += L1_CACHE_ALIGN(sizeof(pg_data_t));
+	pernodesize += L1_CACHE_ALIGN(sizeof(struct ia64_node_data));
+	pernodesize += L1_CACHE_ALIGN(sizeof(pg_data_t));
+	pernodesize = PAGE_ALIGN(pernodesize);
+	return pernodesize;
+}
+
+/**
+ * per_cpu_node_setup - setup per-cpu areas on each node
+ * @cpu_data: per-cpu area on this node
+ * @node: node to setup
+ *
+ * Copy the static per-cpu data into the region we just set aside and then
+ * setup __per_cpu_offset for each CPU on this node.  Return a pointer to
+ * the end of the area.
+ */
+static void *per_cpu_node_setup(void *cpu_data, int node)
+{
+#ifdef CONFIG_SMP
+	int cpu;
+
+	for_each_possible_early_cpu(cpu) {
+		void *src = cpu == 0 ? __cpu0_per_cpu : __phys_per_cpu_start;
+
+		if (node != node_cpuid[cpu].nid)
+			continue;
+
+		memcpy(__va(cpu_data), src, __per_cpu_end - __per_cpu_start);
+		__per_cpu_offset[cpu] = (char *)__va(cpu_data) -
+			__per_cpu_start;
+
+		/*
+		 * percpu area for cpu0 is moved from the __init area
+		 * which is setup by head.S and used till this point.
+		 * Update ar.k3.  This move is ensures that percpu
+		 * area for cpu0 is on the correct node and its
+		 * virtual address isn't insanely far from other
+		 * percpu areas which is important for congruent
+		 * percpu allocator.
+		 */
+		if (cpu == 0)
+			ia64_set_kr(IA64_KR_PER_CPU_DATA,
+				    (unsigned long)cpu_data -
+				    (unsigned long)__per_cpu_start);
+
+		cpu_data += PERCPU_PAGE_SIZE;
+	}
+#endif
+	return cpu_data;
+}
+
+#ifdef CONFIG_SMP
+/**
+ * setup_per_cpu_areas - setup percpu areas
+ *
+ * Arch code has already allocated and initialized percpu areas.  All
+ * this function has to do is to teach the determined layout to the
+ * dynamic percpu allocator, which happens to be more complex than
+ * creating whole new ones using helpers.
+ */
+void __init setup_per_cpu_areas(void)
+{
+	struct pcpu_alloc_info *ai;
+	struct pcpu_group_info *uninitialized_var(gi);
+	unsigned int *cpu_map;
+	void *base;
+	unsigned long base_offset;
+	unsigned int cpu;
+	ssize_t static_size, reserved_size, dyn_size;
+	int node, prev_node, unit, nr_units, rc;
+
+	ai = pcpu_alloc_alloc_info(MAX_NUMNODES, nr_cpu_ids);
+	if (!ai)
+		panic("failed to allocate pcpu_alloc_info");
+	cpu_map = ai->groups[0].cpu_map;
+
+	/* determine base */
+	base = (void *)ULONG_MAX;
+	for_each_possible_cpu(cpu)
+		base = min(base,
+			   (void *)(__per_cpu_offset[cpu] + __per_cpu_start));
+	base_offset = (void *)__per_cpu_start - base;
+
+	/* build cpu_map, units are grouped by node */
+	unit = 0;
+	for_each_node(node)
+		for_each_possible_cpu(cpu)
+			if (node == node_cpuid[cpu].nid)
+				cpu_map[unit++] = cpu;
+	nr_units = unit;
+
+	/* set basic parameters */
+	static_size = __per_cpu_end - __per_cpu_start;
+	reserved_size = PERCPU_MODULE_RESERVE;
+	dyn_size = PERCPU_PAGE_SIZE - static_size - reserved_size;
+	if (dyn_size < 0)
+		panic("percpu area overflow static=%zd reserved=%zd\n",
+		      static_size, reserved_size);
+
+	ai->static_size		= static_size;
+	ai->reserved_size	= reserved_size;
+	ai->dyn_size		= dyn_size;
+	ai->unit_size		= PERCPU_PAGE_SIZE;
+	ai->atom_size		= PAGE_SIZE;
+	ai->alloc_size		= PERCPU_PAGE_SIZE;
+
+	/*
+	 * CPUs are put into groups according to node.  Walk cpu_map
+	 * and create new groups at node boundaries.
+	 */
+	prev_node = -1;
+	ai->nr_groups = 0;
+	for (unit = 0; unit < nr_units; unit++) {
+		cpu = cpu_map[unit];
+		node = node_cpuid[cpu].nid;
+
+		if (node == prev_node) {
+			gi->nr_units++;
+			continue;
+		}
+		prev_node = node;
+
+		gi = &ai->groups[ai->nr_groups++];
+		gi->nr_units		= 1;
+		gi->base_offset		= __per_cpu_offset[cpu] + base_offset;
+		gi->cpu_map		= &cpu_map[unit];
+	}
+
+	rc = pcpu_setup_first_chunk(ai, base);
+	if (rc)
+		panic("failed to setup percpu area (err=%d)", rc);
+
+	pcpu_free_alloc_info(ai);
+}
+#endif
+
+/**
+ * fill_pernode - initialize pernode data.
+ * @node: the node id.
+ * @pernode: physical address of pernode data
+ * @pernodesize: size of the pernode data
+ */
+static void __init fill_pernode(int node, unsigned long pernode,
+	unsigned long pernodesize)
+{
+	void *cpu_data;
+	int cpus = early_nr_cpus_node(node);
+	struct bootmem_data *bdp = &bootmem_node_data[node];
+
+	mem_data[node].pernode_addr = pernode;
+	mem_data[node].pernode_size = pernodesize;
+	memset(__va(pernode), 0, pernodesize);
+
+	cpu_data = (void *)pernode;
+	pernode += PERCPU_PAGE_SIZE * cpus;
+	pernode += node * L1_CACHE_BYTES;
+
+	pgdat_list[node] = __va(pernode);
+	pernode += L1_CACHE_ALIGN(sizeof(pg_data_t));
+
+	mem_data[node].node_data = __va(pernode);
+	pernode += L1_CACHE_ALIGN(sizeof(struct ia64_node_data));
+
+	pgdat_list[node]->bdata = bdp;
+	pernode += L1_CACHE_ALIGN(sizeof(pg_data_t));
+
+	cpu_data = per_cpu_node_setup(cpu_data, node);
+
+	return;
+}
+
+/**
+ * find_pernode_space - allocate memory for memory map and per-node structures
+ * @start: physical start of range
+ * @len: length of range
+ * @node: node where this range resides
+ *
+ * This routine reserves space for the per-cpu data struct, the list of
+ * pg_data_ts and the per-node data struct.  Each node will have something like
+ * the following in the first chunk of addr. space large enough to hold it.
+ *
+ *    ________________________
+ *   |                        |
+ *   |~~~~~~~~~~~~~~~~~~~~~~~~| <-- NODEDATA_ALIGN(start, node) for the first
+ *   |    PERCPU_PAGE_SIZE *  |     start and length big enough
+ *   |    cpus_on_this_node   | Node 0 will also have entries for all non-existent cpus.
+ *   |------------------------|
+ *   |   local pg_data_t *    |
+ *   |------------------------|
+ *   |  local ia64_node_data  |
+ *   |------------------------|
+ *   |          ???           |
+ *   |________________________|
+ *
+ * Once this space has been set aside, the bootmem maps are initialized.  We
+ * could probably move the allocation of the per-cpu and ia64_node_data space
+ * outside of this function and use alloc_bootmem_node(), but doing it here
+ * is straightforward and we get the alignments we want so...
+ */
+static int __init find_pernode_space(unsigned long start, unsigned long len,
+				     int node)
+{
+	unsigned long spfn, epfn;
+	unsigned long pernodesize = 0, pernode, pages, mapsize;
+	struct bootmem_data *bdp = &bootmem_node_data[node];
+
+	spfn = start >> PAGE_SHIFT;
+	epfn = (start + len) >> PAGE_SHIFT;
+
+	pages = bdp->node_low_pfn - bdp->node_min_pfn;
+	mapsize = bootmem_bootmap_pages(pages) << PAGE_SHIFT;
+
+	/*
+	 * Make sure this memory falls within this node's usable memory
+	 * since we may have thrown some away in build_maps().
+	 */
+	if (spfn < bdp->node_min_pfn || epfn > bdp->node_low_pfn)
+		return 0;
+
+	/* Don't setup this node's local space twice... */
+	if (mem_data[node].pernode_addr)
+		return 0;
+
+	/*
+	 * Calculate total size needed, incl. what's necessary
+	 * for good alignment and alias prevention.
+	 */
+	pernodesize = compute_pernodesize(node);
+	pernode = NODEDATA_ALIGN(start, node);
+
+	/* Is this range big enough for what we want to store here? */
+	if (start + len > (pernode + pernodesize + mapsize))
+		fill_pernode(node, pernode, pernodesize);
+
+	return 0;
+}
+
+/**
+ * free_node_bootmem - free bootmem allocator memory for use
+ * @start: physical start of range
+ * @len: length of range
+ * @node: node where this range resides
+ *
+ * Simply calls the bootmem allocator to free the specified ranged from
+ * the given pg_data_t's bdata struct.  After this function has been called
+ * for all the entries in the EFI memory map, the bootmem allocator will
+ * be ready to service allocation requests.
+ */
+static int __init free_node_bootmem(unsigned long start, unsigned long len,
+				    int node)
+{
+	free_bootmem_node(pgdat_list[node], start, len);
+
+	return 0;
+}
+
+/**
+ * reserve_pernode_space - reserve memory for per-node space
+ *
+ * Reserve the space used by the bootmem maps & per-node space in the boot
+ * allocator so that when we actually create the real mem maps we don't
+ * use their memory.
+ */
+static void __init reserve_pernode_space(void)
+{
+	unsigned long base, size, pages;
+	struct bootmem_data *bdp;
+	int node;
+
+	for_each_online_node(node) {
+		pg_data_t *pdp = pgdat_list[node];
+
+		if (node_isset(node, memory_less_mask))
+			continue;
+
+		bdp = pdp->bdata;
+
+		/* First the bootmem_map itself */
+		pages = bdp->node_low_pfn - bdp->node_min_pfn;
+		size = bootmem_bootmap_pages(pages) << PAGE_SHIFT;
+		base = __pa(bdp->node_bootmem_map);
+		reserve_bootmem_node(pdp, base, size, BOOTMEM_DEFAULT);
+
+		/* Now the per-node space */
+		size = mem_data[node].pernode_size;
+		base = __pa(mem_data[node].pernode_addr);
+		reserve_bootmem_node(pdp, base, size, BOOTMEM_DEFAULT);
+	}
+}
+
+static void __meminit scatter_node_data(void)
+{
+	pg_data_t **dst;
+	int node;
+
+	/*
+	 * for_each_online_node() can't be used at here.
+	 * node_online_map is not set for hot-added nodes at this time,
+	 * because we are halfway through initialization of the new node's
+	 * structures.  If for_each_online_node() is used, a new node's
+	 * pg_data_ptrs will be not initialized. Instead of using it,
+	 * pgdat_list[] is checked.
+	 */
+	for_each_node(node) {
+		if (pgdat_list[node]) {
+			dst = LOCAL_DATA_ADDR(pgdat_list[node])->pg_data_ptrs;
+			memcpy(dst, pgdat_list, sizeof(pgdat_list));
+		}
+	}
+}
+
+/**
+ * initialize_pernode_data - fixup per-cpu & per-node pointers
+ *
+ * Each node's per-node area has a copy of the global pg_data_t list, so
+ * we copy that to each node here, as well as setting the per-cpu pointer
+ * to the local node data structure.  The active_cpus field of the per-node
+ * structure gets setup by the platform_cpu_init() function later.
+ */
+static void __init initialize_pernode_data(void)
+{
+	int cpu, node;
+
+	scatter_node_data();
+
+#ifdef CONFIG_SMP
+	/* Set the node_data pointer for each per-cpu struct */
+	for_each_possible_early_cpu(cpu) {
+		node = node_cpuid[cpu].nid;
+		per_cpu(ia64_cpu_info, cpu).node_data =
+			mem_data[node].node_data;
+	}
+#else
+	{
+		struct cpuinfo_ia64 *cpu0_cpu_info;
+		cpu = 0;
+		node = node_cpuid[cpu].nid;
+		cpu0_cpu_info = (struct cpuinfo_ia64 *)(__phys_per_cpu_start +
+			((char *)&ia64_cpu_info - __per_cpu_start));
+		cpu0_cpu_info->node_data = mem_data[node].node_data;
+	}
+#endif /* CONFIG_SMP */
+}
+
+/**
+ * memory_less_node_alloc - * attempt to allocate memory on the best NUMA slit
+ * 	node but fall back to any other node when __alloc_bootmem_node fails
+ *	for best.
+ * @nid: node id
+ * @pernodesize: size of this node's pernode data
+ */
+static void __init *memory_less_node_alloc(int nid, unsigned long pernodesize)
+{
+	void *ptr = NULL;
+	u8 best = 0xff;
+	int bestnode = -1, node, anynode = 0;
+
+	for_each_online_node(node) {
+		if (node_isset(node, memory_less_mask))
+			continue;
+		else if (node_distance(nid, node) < best) {
+			best = node_distance(nid, node);
+			bestnode = node;
+		}
+		anynode = node;
+	}
+
+	if (bestnode == -1)
+		bestnode = anynode;
+
+	ptr = __alloc_bootmem_node(pgdat_list[bestnode], pernodesize,
+		PERCPU_PAGE_SIZE, __pa(MAX_DMA_ADDRESS));
+
+	return ptr;
+}
+
+/**
+ * memory_less_nodes - allocate and initialize CPU only nodes pernode
+ *	information.
+ */
+static void __init memory_less_nodes(void)
+{
+	unsigned long pernodesize;
+	void *pernode;
+	int node;
+
+	for_each_node_mask(node, memory_less_mask) {
+		pernodesize = compute_pernodesize(node);
+		pernode = memory_less_node_alloc(node, pernodesize);
+		fill_pernode(node, __pa(pernode), pernodesize);
+	}
+
+	return;
+}
+
+/**
+ * find_memory - walk the EFI memory map and setup the bootmem allocator
+ *
+ * Called early in boot to setup the bootmem allocator, and to
+ * allocate the per-cpu and per-node structures.
+ */
+void __init find_memory(void)
+{
+	int node;
+
+	reserve_memory();
+
+	if (num_online_nodes() == 0) {
+		printk(KERN_ERR "node info missing!\n");
+		node_set_online(0);
+	}
+
+	nodes_or(memory_less_mask, memory_less_mask, node_online_map);
+	min_low_pfn = -1;
+	max_low_pfn = 0;
+
+	/* These actually end up getting called by call_pernode_memory() */
+	efi_memmap_walk(filter_rsvd_memory, build_node_maps);
+	efi_memmap_walk(filter_rsvd_memory, find_pernode_space);
+	efi_memmap_walk(find_max_min_low_pfn, NULL);
+
+	for_each_online_node(node)
+		if (bootmem_node_data[node].node_low_pfn) {
+			node_clear(node, memory_less_mask);
+			mem_data[node].min_pfn = ~0UL;
+		}
+
+	efi_memmap_walk(filter_memory, register_active_ranges);
+
+	/*
+	 * Initialize the boot memory maps in reverse order since that's
+	 * what the bootmem allocator expects
+	 */
+	for (node = MAX_NUMNODES - 1; node >= 0; node--) {
+		unsigned long pernode, pernodesize, map;
+		struct bootmem_data *bdp;
+
+		if (!node_online(node))
+			continue;
+		else if (node_isset(node, memory_less_mask))
+			continue;
+
+		bdp = &bootmem_node_data[node];
+		pernode = mem_data[node].pernode_addr;
+		pernodesize = mem_data[node].pernode_size;
+		map = pernode + pernodesize;
+
+		init_bootmem_node(pgdat_list[node],
+				  map>>PAGE_SHIFT,
+				  bdp->node_min_pfn,
+				  bdp->node_low_pfn);
+	}
+
+	efi_memmap_walk(filter_rsvd_memory, free_node_bootmem);
+
+	reserve_pernode_space();
+	memory_less_nodes();
+	initialize_pernode_data();
+
+	max_pfn = max_low_pfn;
+
+	find_initrd();
+}
+
+#ifdef CONFIG_SMP
+/**
+ * per_cpu_init - setup per-cpu variables
+ *
+ * find_pernode_space() does most of this already, we just need to set
+ * local_per_cpu_offset
+ */
+void *per_cpu_init(void)
+{
+	int cpu;
+	static int first_time = 1;
+
+	if (first_time) {
+		first_time = 0;
+		for_each_possible_early_cpu(cpu)
+			per_cpu(local_per_cpu_offset, cpu) = __per_cpu_offset[cpu];
+	}
+
+	return __per_cpu_start + __per_cpu_offset[smp_processor_id()];
+}
+#endif /* CONFIG_SMP */
+
+/**
+ * call_pernode_memory - use SRAT to call callback functions with node info
+ * @start: physical start of range
+ * @len: length of range
+ * @arg: function to call for each range
+ *
+ * efi_memmap_walk() knows nothing about layout of memory across nodes. Find
+ * out to which node a block of memory belongs.  Ignore memory that we cannot
+ * identify, and split blocks that run across multiple nodes.
+ *
+ * Take this opportunity to round the start address up and the end address
+ * down to page boundaries.
+ */
+void call_pernode_memory(unsigned long start, unsigned long len, void *arg)
+{
+	unsigned long rs, re, end = start + len;
+	void (*func)(unsigned long, unsigned long, int);
+	int i;
+
+	start = PAGE_ALIGN(start);
+	end &= PAGE_MASK;
+	if (start >= end)
+		return;
+
+	func = arg;
+
+	if (!num_node_memblks) {
+		/* No SRAT table, so assume one node (node 0) */
+		if (start < end)
+			(*func)(start, end - start, 0);
+		return;
+	}
+
+	for (i = 0; i < num_node_memblks; i++) {
+		rs = max(start, node_memblk[i].start_paddr);
+		re = min(end, node_memblk[i].start_paddr +
+			 node_memblk[i].size);
+
+		if (rs < re)
+			(*func)(rs, re - rs, node_memblk[i].nid);
+
+		if (re == end)
+			break;
+	}
+}
+
+/**
+ * count_node_pages - callback to build per-node memory info structures
+ * @start: physical start of range
+ * @len: length of range
+ * @node: node where this range resides
+ *
+ * Each node has it's own number of physical pages, DMAable pages, start, and
+ * end page frame number.  This routine will be called by call_pernode_memory()
+ * for each piece of usable memory and will setup these values for each node.
+ * Very similar to build_maps().
+ */
+static __init int count_node_pages(unsigned long start, unsigned long len, int node)
+{
+	unsigned long end = start + len;
+
+#ifdef CONFIG_ZONE_DMA
+	if (start <= __pa(MAX_DMA_ADDRESS))
+		mem_data[node].num_dma_physpages +=
+			(min(end, __pa(MAX_DMA_ADDRESS)) - start) >>PAGE_SHIFT;
+#endif
+	start = GRANULEROUNDDOWN(start);
+	end = GRANULEROUNDUP(end);
+	mem_data[node].max_pfn = max(mem_data[node].max_pfn,
+				     end >> PAGE_SHIFT);
+	mem_data[node].min_pfn = min(mem_data[node].min_pfn,
+				     start >> PAGE_SHIFT);
+
+	return 0;
+}
+
+/**
+ * paging_init - setup page tables
+ *
+ * paging_init() sets up the page tables for each node of the system and frees
+ * the bootmem allocator memory for general use.
+ */
+void __init paging_init(void)
+{
+	unsigned long max_dma;
+	unsigned long pfn_offset = 0;
+	unsigned long max_pfn = 0;
+	int node;
+	unsigned long max_zone_pfns[MAX_NR_ZONES];
+
+	max_dma = virt_to_phys((void *) MAX_DMA_ADDRESS) >> PAGE_SHIFT;
+
+	efi_memmap_walk(filter_rsvd_memory, count_node_pages);
+
+	sparse_memory_present_with_active_regions(MAX_NUMNODES);
+	sparse_init();
+
+#ifdef CONFIG_VIRTUAL_MEM_MAP
+	VMALLOC_END -= PAGE_ALIGN(ALIGN(max_low_pfn, MAX_ORDER_NR_PAGES) *
+		sizeof(struct page));
+	vmem_map = (struct page *) VMALLOC_END;
+	efi_memmap_walk(create_mem_map_page_table, NULL);
+	printk("Virtual mem_map starts at 0x%p\n", vmem_map);
+#endif
+
+	for_each_online_node(node) {
+		pfn_offset = mem_data[node].min_pfn;
+
+#ifdef CONFIG_VIRTUAL_MEM_MAP
+		NODE_DATA(node)->node_mem_map = vmem_map + pfn_offset;
+#endif
+		if (mem_data[node].max_pfn > max_pfn)
+			max_pfn = mem_data[node].max_pfn;
+	}
+
+	memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
+#ifdef CONFIG_ZONE_DMA
+	max_zone_pfns[ZONE_DMA] = max_dma;
+#endif
+	max_zone_pfns[ZONE_NORMAL] = max_pfn;
+	free_area_init_nodes(max_zone_pfns);
+
+	zero_page_memmap_ptr = virt_to_page(ia64_imva(empty_zero_page));
+}
+
+#ifdef CONFIG_MEMORY_HOTPLUG
+pg_data_t *arch_alloc_nodedata(int nid)
+{
+	unsigned long size = compute_pernodesize(nid);
+
+	return kzalloc(size, GFP_KERNEL);
+}
+
+void arch_free_nodedata(pg_data_t *pgdat)
+{
+	kfree(pgdat);
+}
+
+void arch_refresh_nodedata(int update_node, pg_data_t *update_pgdat)
+{
+	pgdat_list[update_node] = update_pgdat;
+	scatter_node_data();
+}
+#endif
+
+#ifdef CONFIG_SPARSEMEM_VMEMMAP
+int __meminit vmemmap_populate(unsigned long start, unsigned long end, int node)
+{
+	return vmemmap_populate_basepages(start, end, node);
+}
+
+void vmemmap_free(unsigned long start, unsigned long end)
+{
+}
+#endif

arch/ia64/mm/extable.c

@@ -0,0 +1,115 @@
+/*
+ * Kernel exception handling table support.  Derived from arch/alpha/mm/extable.c.
+ *
+ * Copyright (C) 1998, 1999, 2001-2002, 2004 Hewlett-Packard Co
+ *	David Mosberger-Tang <davidm@hpl.hp.com>
+ */
+
+#include <linux/sort.h>
+
+#include <asm/uaccess.h>
+#include <linux/module.h>
+
+static int cmp_ex(const void *a, const void *b)
+{
+	const struct exception_table_entry *l = a, *r = b;
+	u64 lip = (u64) &l->addr + l->addr;
+	u64 rip = (u64) &r->addr + r->addr;
+
+	/* avoid overflow */
+	if (lip > rip)
+		return 1;
+	if (lip < rip)
+		return -1;
+	return 0;
+}
+
+static void swap_ex(void *a, void *b, int size)
+{
+	struct exception_table_entry *l = a, *r = b, tmp;
+	u64 delta = (u64) r - (u64) l;
+
+	tmp = *l;
+	l->addr = r->addr + delta;
+	l->cont = r->cont + delta;
+	r->addr = tmp.addr - delta;
+	r->cont = tmp.cont - delta;
+}
+
+/*
+ * Sort the exception table. It's usually already sorted, but there
+ * may be unordered entries due to multiple text sections (such as the
+ * .init text section). Note that the exception-table-entries contain
+ * location-relative addresses, which requires a bit of care during
+ * sorting to avoid overflows in the offset members (e.g., it would
+ * not be safe to make a temporary copy of an exception-table entry on
+ * the stack, because the stack may be more than 2GB away from the
+ * exception-table).
+ */
+void sort_extable (struct exception_table_entry *start,
+		   struct exception_table_entry *finish)
+{
+	sort(start, finish - start, sizeof(struct exception_table_entry),
+	     cmp_ex, swap_ex);
+}
+
+static inline unsigned long ex_to_addr(const struct exception_table_entry *x)
+{
+	return (unsigned long)&x->addr + x->addr;
+}
+
+#ifdef CONFIG_MODULES
+/*
+ * Any entry referring to the module init will be at the beginning or
+ * the end.
+ */
+void trim_init_extable(struct module *m)
+{
+	/*trim the beginning*/
+	while (m->num_exentries &&
+	       within_module_init(ex_to_addr(&m->extable[0]), m)) {
+		m->extable++;
+		m->num_exentries--;
+	}
+	/*trim the end*/
+	while (m->num_exentries &&
+	       within_module_init(ex_to_addr(&m->extable[m->num_exentries-1]),
+				  m))
+		m->num_exentries--;
+}
+#endif /* CONFIG_MODULES */
+
+const struct exception_table_entry *
+search_extable (const struct exception_table_entry *first,
+		const struct exception_table_entry *last,
+		unsigned long ip)
+{
+	const struct exception_table_entry *mid;
+	unsigned long mid_ip;
+	long diff;
+
+        while (first <= last) {
+		mid = &first[(last - first)/2];
+		mid_ip = (u64) &mid->addr + mid->addr;
+		diff = mid_ip - ip;
+                if (diff == 0)
+                        return mid;
+                else if (diff < 0)
+                        first = mid + 1;
+                else
+                        last = mid - 1;
+        }
+        return NULL;
+}
+
+void
+ia64_handle_exception (struct pt_regs *regs, const struct exception_table_entry *e)
+{
+	long fix = (u64) &e->cont + e->cont;
+
+	regs->r8 = -EFAULT;
+	if (fix & 4)
+		regs->r9 = 0;
+	regs->cr_iip = fix & ~0xf;
+	ia64_psr(regs)->ri = fix & 0x3;		/* set continuation slot number */
+}

arch/ia64/mm/fault.c

@@ -0,0 +1,308 @@
+/*
+ * MMU fault handling support.
+ *
+ * Copyright (C) 1998-2002 Hewlett-Packard Co
+ *	David Mosberger-Tang <davidm@hpl.hp.com>
+ */
+#include <linux/sched.h>
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/interrupt.h>
+#include <linux/kprobes.h>
+#include <linux/kdebug.h>
+#include <linux/prefetch.h>
+
+#include <asm/pgtable.h>
+#include <asm/processor.h>
+#include <asm/uaccess.h>
+
+extern int die(char *, struct pt_regs *, long);
+
+#ifdef CONFIG_KPROBES
+static inline int notify_page_fault(struct pt_regs *regs, int trap)
+{
+	int ret = 0;
+
+	if (!user_mode(regs)) {
+		/* kprobe_running() needs smp_processor_id() */
+		preempt_disable();
+		if (kprobe_running() && kprobe_fault_handler(regs, trap))
+			ret = 1;
+		preempt_enable();
+	}
+
+	return ret;
+}
+#else
+static inline int notify_page_fault(struct pt_regs *regs, int trap)
+{
+	return 0;
+}
+#endif
+
+/*
+ * Return TRUE if ADDRESS points at a page in the kernel's mapped segment
+ * (inside region 5, on ia64) and that page is present.
+ */
+static int
+mapped_kernel_page_is_present (unsigned long address)
+{
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+	pte_t *ptep, pte;
+
+	pgd = pgd_offset_k(address);
+	if (pgd_none(*pgd) || pgd_bad(*pgd))
+		return 0;
+
+	pud = pud_offset(pgd, address);
+	if (pud_none(*pud) || pud_bad(*pud))
+		return 0;
+
+	pmd = pmd_offset(pud, address);
+	if (pmd_none(*pmd) || pmd_bad(*pmd))
+		return 0;
+
+	ptep = pte_offset_kernel(pmd, address);
+	if (!ptep)
+		return 0;
+
+	pte = *ptep;
+	return pte_present(pte);
+}
+
+#	define VM_READ_BIT	0
+#	define VM_WRITE_BIT	1
+#	define VM_EXEC_BIT	2
+
+void __kprobes
+ia64_do_page_fault (unsigned long address, unsigned long isr, struct pt_regs *regs)
+{
+	int signal = SIGSEGV, code = SEGV_MAPERR;
+	struct vm_area_struct *vma, *prev_vma;
+	struct mm_struct *mm = current->mm;
+	struct siginfo si;
+	unsigned long mask;
+	int fault;
+	unsigned int flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
+
+	mask = ((((isr >> IA64_ISR_X_BIT) & 1UL) << VM_EXEC_BIT)
+		| (((isr >> IA64_ISR_W_BIT) & 1UL) << VM_WRITE_BIT));
+
+	/* mmap_sem is performance critical.... */
+	prefetchw(&mm->mmap_sem);
+
+	/*
+	 * If we're in an interrupt or have no user context, we must not take the fault..
+	 */
+	if (in_atomic() || !mm)
+		goto no_context;
+
+#ifdef CONFIG_VIRTUAL_MEM_MAP
+	/*
+	 * If fault is in region 5 and we are in the kernel, we may already
+	 * have the mmap_sem (pfn_valid macro is called during mmap). There
+	 * is no vma for region 5 addr's anyway, so skip getting the semaphore
+	 * and go directly to the exception handling code.
+	 */
+
+	if ((REGION_NUMBER(address) == 5) && !user_mode(regs))
+		goto bad_area_no_up;
+#endif
+
+	/*
+	 * This is to handle the kprobes on user space access instructions
+	 */
+	if (notify_page_fault(regs, TRAP_BRKPT))
+		return;
+
+	if (user_mode(regs))
+		flags |= FAULT_FLAG_USER;
+	if (mask & VM_WRITE)
+		flags |= FAULT_FLAG_WRITE;
+retry:
+	down_read(&mm->mmap_sem);
+
+	vma = find_vma_prev(mm, address, &prev_vma);
+	if (!vma && !prev_vma )
+		goto bad_area;
+
+        /*
+         * find_vma_prev() returns vma such that address < vma->vm_end or NULL
+         *
+         * May find no vma, but could be that the last vm area is the
+         * register backing store that needs to expand upwards, in
+         * this case vma will be null, but prev_vma will ne non-null
+         */
+        if (( !vma && prev_vma ) || (address < vma->vm_start) )
+		goto check_expansion;
+
+  good_area:
+	code = SEGV_ACCERR;
+
+	/* OK, we've got a good vm_area for this memory area.  Check the access permissions: */
+
+#	if (((1 << VM_READ_BIT) != VM_READ || (1 << VM_WRITE_BIT) != VM_WRITE) \
+	    || (1 << VM_EXEC_BIT) != VM_EXEC)
+#		error File is out of sync with <linux/mm.h>.  Please update.
+#	endif
+
+	if (((isr >> IA64_ISR_R_BIT) & 1UL) && (!(vma->vm_flags & (VM_READ | VM_WRITE))))
+		goto bad_area;
+
+	if ((vma->vm_flags & mask) != mask)
+		goto bad_area;
+
+	/*
+	 * If for any reason at all we couldn't handle the fault, make
+	 * sure we exit gracefully rather than endlessly redo the
+	 * fault.
+	 */
+	fault = handle_mm_fault(mm, vma, address, flags);
+
+	if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
+		return;
+
+	if (unlikely(fault & VM_FAULT_ERROR)) {
+		/*
+		 * We ran out of memory, or some other thing happened
+		 * to us that made us unable to handle the page fault
+		 * gracefully.
+		 */
+		if (fault & VM_FAULT_OOM) {
+			goto out_of_memory;
+		} else if (fault & VM_FAULT_SIGSEGV) {
+			goto bad_area;
+		} else if (fault & VM_FAULT_SIGBUS) {
+			signal = SIGBUS;
+			goto bad_area;
+		}
+		BUG();
+	}
+
+	if (flags & FAULT_FLAG_ALLOW_RETRY) {
+		if (fault & VM_FAULT_MAJOR)
+			current->maj_flt++;
+		else
+			current->min_flt++;
+		if (fault & VM_FAULT_RETRY) {
+			flags &= ~FAULT_FLAG_ALLOW_RETRY;
+			flags |= FAULT_FLAG_TRIED;
+
+			 /* No need to up_read(&mm->mmap_sem) as we would
+			 * have already released it in __lock_page_or_retry
+			 * in mm/filemap.c.
+			 */
+
+			goto retry;
+		}
+	}
+
+	up_read(&mm->mmap_sem);
+	return;
+
+  check_expansion:
+	if (!(prev_vma && (prev_vma->vm_flags & VM_GROWSUP) && (address == prev_vma->vm_end))) {
+		if (!vma)
+			goto bad_area;
+		if (!(vma->vm_flags & VM_GROWSDOWN))
+			goto bad_area;
+		if (REGION_NUMBER(address) != REGION_NUMBER(vma->vm_start)
+		    || REGION_OFFSET(address) >= RGN_MAP_LIMIT)
+			goto bad_area;
+		if (expand_stack(vma, address))
+			goto bad_area;
+	} else {
+		vma = prev_vma;
+		if (REGION_NUMBER(address) != REGION_NUMBER(vma->vm_start)
+		    || REGION_OFFSET(address) >= RGN_MAP_LIMIT)
+			goto bad_area;
+		/*
+		 * Since the register backing store is accessed sequentially,
+		 * we disallow growing it by more than a page at a time.
+		 */
+		if (address > vma->vm_end + PAGE_SIZE - sizeof(long))
+			goto bad_area;
+		if (expand_upwards(vma, address))
+			goto bad_area;
+	}
+	goto good_area;
+
+  bad_area:
+	up_read(&mm->mmap_sem);
+#ifdef CONFIG_VIRTUAL_MEM_MAP
+  bad_area_no_up:
+#endif
+	if ((isr & IA64_ISR_SP)
+	    || ((isr & IA64_ISR_NA) && (isr & IA64_ISR_CODE_MASK) == IA64_ISR_CODE_LFETCH))
+	{
+		/*
+		 * This fault was due to a speculative load or lfetch.fault, set the "ed"
+		 * bit in the psr to ensure forward progress.  (Target register will get a
+		 * NaT for ld.s, lfetch will be canceled.)
+		 */
+		ia64_psr(regs)->ed = 1;
+		return;
+	}
+	if (user_mode(regs)) {
+		si.si_signo = signal;
+		si.si_errno = 0;
+		si.si_code = code;
+		si.si_addr = (void __user *) address;
+		si.si_isr = isr;
+		si.si_flags = __ISR_VALID;
+		force_sig_info(signal, &si, current);
+		return;
+	}
+
+  no_context:
+	if ((isr & IA64_ISR_SP)
+	    || ((isr & IA64_ISR_NA) && (isr & IA64_ISR_CODE_MASK) == IA64_ISR_CODE_LFETCH))
+	{
+		/*
+		 * This fault was due to a speculative load or lfetch.fault, set the "ed"
+		 * bit in the psr to ensure forward progress.  (Target register will get a
+		 * NaT for ld.s, lfetch will be canceled.)
+		 */
+		ia64_psr(regs)->ed = 1;
+		return;
+	}
+
+	/*
+	 * Since we have no vma's for region 5, we might get here even if the address is
+	 * valid, due to the VHPT walker inserting a non present translation that becomes
+	 * stale. If that happens, the non present fault handler already purged the stale
+	 * translation, which fixed the problem. So, we check to see if the translation is
+	 * valid, and return if it is.
+	 */
+	if (REGION_NUMBER(address) == 5 && mapped_kernel_page_is_present(address))
+		return;
+
+	if (ia64_done_with_exception(regs))
+		return;
+
+	/*
+	 * Oops. The kernel tried to access some bad page. We'll have to terminate things
+	 * with extreme prejudice.
+	 */
+	bust_spinlocks(1);
+
+	if (address < PAGE_SIZE)
+		printk(KERN_ALERT "Unable to handle kernel NULL pointer dereference (address %016lx)\n", address);
+	else
+		printk(KERN_ALERT "Unable to handle kernel paging request at "
+		       "virtual address %016lx\n", address);
+	if (die("Oops", regs, isr))
+		regs = NULL;
+	bust_spinlocks(0);
+	if (regs)
+		do_exit(SIGKILL);
+	return;
+
+  out_of_memory:
+	up_read(&mm->mmap_sem);
+	if (!user_mode(regs))
+		goto no_context;
+	pagefault_out_of_memory();
+}

arch/ia64/mm/hugetlbpage.c

@@ -0,0 +1,199 @@
+/*
+ * IA-64 Huge TLB Page Support for Kernel.
+ *
+ * Copyright (C) 2002-2004 Rohit Seth <rohit.seth@intel.com>
+ * Copyright (C) 2003-2004 Ken Chen <kenneth.w.chen@intel.com>
+ *
+ * Sep, 2003: add numa support
+ * Feb, 2004: dynamic hugetlb page size via boot parameter
+ */
+
+#include <linux/init.h>
+#include <linux/fs.h>
+#include <linux/mm.h>
+#include <linux/hugetlb.h>
+#include <linux/pagemap.h>
+#include <linux/module.h>
+#include <linux/sysctl.h>
+#include <linux/log2.h>
+#include <asm/mman.h>
+#include <asm/pgalloc.h>
+#include <asm/tlb.h>
+#include <asm/tlbflush.h>
+
+unsigned int hpage_shift = HPAGE_SHIFT_DEFAULT;
+EXPORT_SYMBOL(hpage_shift);
+
+pte_t *
+huge_pte_alloc(struct mm_struct *mm, unsigned long addr, unsigned long sz)
+{
+	unsigned long taddr = htlbpage_to_page(addr);
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+	pte_t *pte = NULL;
+
+	pgd = pgd_offset(mm, taddr);
+	pud = pud_alloc(mm, pgd, taddr);
+	if (pud) {
+		pmd = pmd_alloc(mm, pud, taddr);
+		if (pmd)
+			pte = pte_alloc_map(mm, NULL, pmd, taddr);
+	}
+	return pte;
+}
+
+pte_t *
+huge_pte_offset (struct mm_struct *mm, unsigned long addr)
+{
+	unsigned long taddr = htlbpage_to_page(addr);
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+	pte_t *pte = NULL;
+
+	pgd = pgd_offset(mm, taddr);
+	if (pgd_present(*pgd)) {
+		pud = pud_offset(pgd, taddr);
+		if (pud_present(*pud)) {
+			pmd = pmd_offset(pud, taddr);
+			if (pmd_present(*pmd))
+				pte = pte_offset_map(pmd, taddr);
+		}
+	}
+
+	return pte;
+}
+
+int huge_pmd_unshare(struct mm_struct *mm, unsigned long *addr, pte_t *ptep)
+{
+	return 0;
+}
+
+#define mk_pte_huge(entry) { pte_val(entry) |= _PAGE_P; }
+
+/*
+ * Don't actually need to do any preparation, but need to make sure
+ * the address is in the right region.
+ */
+int prepare_hugepage_range(struct file *file,
+			unsigned long addr, unsigned long len)
+{
+	if (len & ~HPAGE_MASK)
+		return -EINVAL;
+	if (addr & ~HPAGE_MASK)
+		return -EINVAL;
+	if (REGION_NUMBER(addr) != RGN_HPAGE)
+		return -EINVAL;
+
+	return 0;
+}
+
+struct page *follow_huge_addr(struct mm_struct *mm, unsigned long addr, int write)
+{
+	struct page *page;
+	pte_t *ptep;
+
+	if (REGION_NUMBER(addr) != RGN_HPAGE)
+		return ERR_PTR(-EINVAL);
+
+	ptep = huge_pte_offset(mm, addr);
+	if (!ptep || pte_none(*ptep))
+		return NULL;
+	page = pte_page(*ptep);
+	page += ((addr & ~HPAGE_MASK) >> PAGE_SHIFT);
+	return page;
+}
+int pmd_huge(pmd_t pmd)
+{
+	return 0;
+}
+
+int pud_huge(pud_t pud)
+{
+	return 0;
+}
+
+void hugetlb_free_pgd_range(struct mmu_gather *tlb,
+			unsigned long addr, unsigned long end,
+			unsigned long floor, unsigned long ceiling)
+{
+	/*
+	 * This is called to free hugetlb page tables.
+	 *
+	 * The offset of these addresses from the base of the hugetlb
+	 * region must be scaled down by HPAGE_SIZE/PAGE_SIZE so that
+	 * the standard free_pgd_range will free the right page tables.
+	 *
+	 * If floor and ceiling are also in the hugetlb region, they
+	 * must likewise be scaled down; but if outside, left unchanged.
+	 */
+
+	addr = htlbpage_to_page(addr);
+	end  = htlbpage_to_page(end);
+	if (REGION_NUMBER(floor) == RGN_HPAGE)
+		floor = htlbpage_to_page(floor);
+	if (REGION_NUMBER(ceiling) == RGN_HPAGE)
+		ceiling = htlbpage_to_page(ceiling);
+
+	free_pgd_range(tlb, addr, end, floor, ceiling);
+}
+
+unsigned long hugetlb_get_unmapped_area(struct file *file, unsigned long addr, unsigned long len,
+		unsigned long pgoff, unsigned long flags)
+{
+	struct vm_unmapped_area_info info;
+
+	if (len > RGN_MAP_LIMIT)
+		return -ENOMEM;
+	if (len & ~HPAGE_MASK)
+		return -EINVAL;
+
+	/* Handle MAP_FIXED */
+	if (flags & MAP_FIXED) {
+		if (prepare_hugepage_range(file, addr, len))
+			return -EINVAL;
+		return addr;
+	}
+
+	/* This code assumes that RGN_HPAGE != 0. */
+	if ((REGION_NUMBER(addr) != RGN_HPAGE) || (addr & (HPAGE_SIZE - 1)))
+		addr = HPAGE_REGION_BASE;
+
+	info.flags = 0;
+	info.length = len;
+	info.low_limit = addr;
+	info.high_limit = HPAGE_REGION_BASE + RGN_MAP_LIMIT;
+	info.align_mask = PAGE_MASK & (HPAGE_SIZE - 1);
+	info.align_offset = 0;
+	return vm_unmapped_area(&info);
+}
+
+static int __init hugetlb_setup_sz(char *str)
+{
+	u64 tr_pages;
+	unsigned long long size;
+
+	if (ia64_pal_vm_page_size(&tr_pages, NULL) != 0)
+		/*
+		 * shouldn't happen, but just in case.
+		 */
+		tr_pages = 0x15557000UL;
+
+	size = memparse(str, &str);
+	if (*str || !is_power_of_2(size) || !(tr_pages & size) ||
+		size <= PAGE_SIZE ||
+		size >= (1UL << PAGE_SHIFT << MAX_ORDER)) {
+		printk(KERN_WARNING "Invalid huge page size specified\n");
+		return 1;
+	}
+
+	hpage_shift = __ffs(size);
+	/*
+	 * boot cpu already executed ia64_mmu_init, and has HPAGE_SHIFT_DEFAULT
+	 * override here with new page shift.
+	 */
+	ia64_set_rr(HPAGE_REGION_BASE, hpage_shift << 2);
+	return 0;
+}
+early_param("hugepagesz", hugetlb_setup_sz);

arch/ia64/mm/init.c

@@ -0,0 +1,766 @@
+/*
+ * Initialize MMU support.
+ *
+ * Copyright (C) 1998-2003 Hewlett-Packard Co
+ *	David Mosberger-Tang <davidm@hpl.hp.com>
+ */
+#include <linux/kernel.h>
+#include <linux/init.h>
+
+#include <linux/bootmem.h>
+#include <linux/efi.h>
+#include <linux/elf.h>
+#include <linux/memblock.h>
+#include <linux/mm.h>
+#include <linux/mmzone.h>
+#include <linux/module.h>
+#include <linux/personality.h>
+#include <linux/reboot.h>
+#include <linux/slab.h>
+#include <linux/swap.h>
+#include <linux/proc_fs.h>
+#include <linux/bitops.h>
+#include <linux/kexec.h>
+
+#include <asm/dma.h>
+#include <asm/io.h>
+#include <asm/machvec.h>
+#include <asm/numa.h>
+#include <asm/patch.h>
+#include <asm/pgalloc.h>
+#include <asm/sal.h>
+#include <asm/sections.h>
+#include <asm/tlb.h>
+#include <asm/uaccess.h>
+#include <asm/unistd.h>
+#include <asm/mca.h>
+#include <asm/paravirt.h>
+
+extern void ia64_tlb_init (void);
+
+unsigned long MAX_DMA_ADDRESS = PAGE_OFFSET + 0x100000000UL;
+
+#ifdef CONFIG_VIRTUAL_MEM_MAP
+unsigned long VMALLOC_END = VMALLOC_END_INIT;
+EXPORT_SYMBOL(VMALLOC_END);
+struct page *vmem_map;
+EXPORT_SYMBOL(vmem_map);
+#endif
+
+struct page *zero_page_memmap_ptr;	/* map entry for zero page */
+EXPORT_SYMBOL(zero_page_memmap_ptr);
+
+void
+__ia64_sync_icache_dcache (pte_t pte)
+{
+	unsigned long addr;
+	struct page *page;
+
+	page = pte_page(pte);
+	addr = (unsigned long) page_address(page);
+
+	if (test_bit(PG_arch_1, &page->flags))
+		return;				/* i-cache is already coherent with d-cache */
+
+	flush_icache_range(addr, addr + (PAGE_SIZE << compound_order(page)));
+	set_bit(PG_arch_1, &page->flags);	/* mark page as clean */
+}
+
+/*
+ * Since DMA is i-cache coherent, any (complete) pages that were written via
+ * DMA can be marked as "clean" so that lazy_mmu_prot_update() doesn't have to
+ * flush them when they get mapped into an executable vm-area.
+ */
+void
+dma_mark_clean(void *addr, size_t size)
+{
+	unsigned long pg_addr, end;
+
+	pg_addr = PAGE_ALIGN((unsigned long) addr);
+	end = (unsigned long) addr + size;
+	while (pg_addr + PAGE_SIZE <= end) {
+		struct page *page = virt_to_page(pg_addr);
+		set_bit(PG_arch_1, &page->flags);
+		pg_addr += PAGE_SIZE;
+	}
+}
+
+inline void
+ia64_set_rbs_bot (void)
+{
+	unsigned long stack_size = rlimit_max(RLIMIT_STACK) & -16;
+
+	if (stack_size > MAX_USER_STACK_SIZE)
+		stack_size = MAX_USER_STACK_SIZE;
+	current->thread.rbs_bot = PAGE_ALIGN(current->mm->start_stack - stack_size);
+}
+
+/*
+ * This performs some platform-dependent address space initialization.
+ * On IA-64, we want to setup the VM area for the register backing
+ * store (which grows upwards) and install the gateway page which is
+ * used for signal trampolines, etc.
+ */
+void
+ia64_init_addr_space (void)
+{
+	struct vm_area_struct *vma;
+
+	ia64_set_rbs_bot();
+
+	/*
+	 * If we're out of memory and kmem_cache_alloc() returns NULL, we simply ignore
+	 * the problem.  When the process attempts to write to the register backing store
+	 * for the first time, it will get a SEGFAULT in this case.
+	 */
+	vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
+	if (vma) {
+		INIT_LIST_HEAD(&vma->anon_vma_chain);
+		vma->vm_mm = current->mm;
+		vma->vm_start = current->thread.rbs_bot & PAGE_MASK;
+		vma->vm_end = vma->vm_start + PAGE_SIZE;
+		vma->vm_flags = VM_DATA_DEFAULT_FLAGS|VM_GROWSUP|VM_ACCOUNT;
+		vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
+		down_write(&current->mm->mmap_sem);
+		if (insert_vm_struct(current->mm, vma)) {
+			up_write(&current->mm->mmap_sem);
+			kmem_cache_free(vm_area_cachep, vma);
+			return;
+		}
+		up_write(&current->mm->mmap_sem);
+	}
+
+	/* map NaT-page at address zero to speed up speculative dereferencing of NULL: */
+	if (!(current->personality & MMAP_PAGE_ZERO)) {
+		vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
+		if (vma) {
+			INIT_LIST_HEAD(&vma->anon_vma_chain);
+			vma->vm_mm = current->mm;
+			vma->vm_end = PAGE_SIZE;
+			vma->vm_page_prot = __pgprot(pgprot_val(PAGE_READONLY) | _PAGE_MA_NAT);
+			vma->vm_flags = VM_READ | VM_MAYREAD | VM_IO |
+					VM_DONTEXPAND | VM_DONTDUMP;
+			down_write(&current->mm->mmap_sem);
+			if (insert_vm_struct(current->mm, vma)) {
+				up_write(&current->mm->mmap_sem);
+				kmem_cache_free(vm_area_cachep, vma);
+				return;
+			}
+			up_write(&current->mm->mmap_sem);
+		}
+	}
+}
+
+void
+free_initmem (void)
+{
+	free_reserved_area(ia64_imva(__init_begin), ia64_imva(__init_end),
+			   -1, "unused kernel");
+}
+
+void __init
+free_initrd_mem (unsigned long start, unsigned long end)
+{
+	/*
+	 * EFI uses 4KB pages while the kernel can use 4KB or bigger.
+	 * Thus EFI and the kernel may have different page sizes. It is
+	 * therefore possible to have the initrd share the same page as
+	 * the end of the kernel (given current setup).
+	 *
+	 * To avoid freeing/using the wrong page (kernel sized) we:
+	 *	- align up the beginning of initrd
+	 *	- align down the end of initrd
+	 *
+	 *  |             |
+	 *  |=============| a000
+	 *  |             |
+	 *  |             |
+	 *  |             | 9000
+	 *  |/////////////|
+	 *  |/////////////|
+	 *  |=============| 8000
+	 *  |///INITRD////|
+	 *  |/////////////|
+	 *  |/////////////| 7000
+	 *  |             |
+	 *  |KKKKKKKKKKKKK|
+	 *  |=============| 6000
+	 *  |KKKKKKKKKKKKK|
+	 *  |KKKKKKKKKKKKK|
+	 *  K=kernel using 8KB pages
+	 *
+	 * In this example, we must free page 8000 ONLY. So we must align up
+	 * initrd_start and keep initrd_end as is.
+	 */
+	start = PAGE_ALIGN(start);
+	end = end & PAGE_MASK;
+
+	if (start < end)
+		printk(KERN_INFO "Freeing initrd memory: %ldkB freed\n", (end - start) >> 10);
+
+	for (; start < end; start += PAGE_SIZE) {
+		if (!virt_addr_valid(start))
+			continue;
+		free_reserved_page(virt_to_page(start));
+	}
+}
+
+/*
+ * This installs a clean page in the kernel's page table.
+ */
+static struct page * __init
+put_kernel_page (struct page *page, unsigned long address, pgprot_t pgprot)
+{
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+	pte_t *pte;
+
+	if (!PageReserved(page))
+		printk(KERN_ERR "put_kernel_page: page at 0x%p not in reserved memory\n",
+		       page_address(page));
+
+	pgd = pgd_offset_k(address);		/* note: this is NOT pgd_offset()! */
+
+	{
+		pud = pud_alloc(&init_mm, pgd, address);
+		if (!pud)
+			goto out;
+		pmd = pmd_alloc(&init_mm, pud, address);
+		if (!pmd)
+			goto out;
+		pte = pte_alloc_kernel(pmd, address);
+		if (!pte)
+			goto out;
+		if (!pte_none(*pte))
+			goto out;
+		set_pte(pte, mk_pte(page, pgprot));
+	}
+  out:
+	/* no need for flush_tlb */
+	return page;
+}
+
+static void __init
+setup_gate (void)
+{
+	void *gate_section;
+	struct page *page;
+
+	/*
+	 * Map the gate page twice: once read-only to export the ELF
+	 * headers etc. and once execute-only page to enable
+	 * privilege-promotion via "epc":
+	 */
+	gate_section = paravirt_get_gate_section();
+	page = virt_to_page(ia64_imva(gate_section));
+	put_kernel_page(page, GATE_ADDR, PAGE_READONLY);
+#ifdef HAVE_BUGGY_SEGREL
+	page = virt_to_page(ia64_imva(gate_section + PAGE_SIZE));
+	put_kernel_page(page, GATE_ADDR + PAGE_SIZE, PAGE_GATE);
+#else
+	put_kernel_page(page, GATE_ADDR + PERCPU_PAGE_SIZE, PAGE_GATE);
+	/* Fill in the holes (if any) with read-only zero pages: */
+	{
+		unsigned long addr;
+
+		for (addr = GATE_ADDR + PAGE_SIZE;
+		     addr < GATE_ADDR + PERCPU_PAGE_SIZE;
+		     addr += PAGE_SIZE)
+		{
+			put_kernel_page(ZERO_PAGE(0), addr,
+					PAGE_READONLY);
+			put_kernel_page(ZERO_PAGE(0), addr + PERCPU_PAGE_SIZE,
+					PAGE_READONLY);
+		}
+	}
+#endif
+	ia64_patch_gate();
+}
+
+static struct vm_area_struct gate_vma;
+
+static int __init gate_vma_init(void)
+{
+	gate_vma.vm_mm = NULL;
+	gate_vma.vm_start = FIXADDR_USER_START;
+	gate_vma.vm_end = FIXADDR_USER_END;
+	gate_vma.vm_flags = VM_READ | VM_MAYREAD | VM_EXEC | VM_MAYEXEC;
+	gate_vma.vm_page_prot = __P101;
+
+	return 0;
+}
+__initcall(gate_vma_init);
+
+struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
+{
+	return &gate_vma;
+}
+
+int in_gate_area_no_mm(unsigned long addr)
+{
+	if ((addr >= FIXADDR_USER_START) && (addr < FIXADDR_USER_END))
+		return 1;
+	return 0;
+}
+
+int in_gate_area(struct mm_struct *mm, unsigned long addr)
+{
+	return in_gate_area_no_mm(addr);
+}
+
+void ia64_mmu_init(void *my_cpu_data)
+{
+	unsigned long pta, impl_va_bits;
+	extern void tlb_init(void);
+
+#ifdef CONFIG_DISABLE_VHPT
+#	define VHPT_ENABLE_BIT	0
+#else
+#	define VHPT_ENABLE_BIT	1
+#endif
+
+	/*
+	 * Check if the virtually mapped linear page table (VMLPT) overlaps with a mapped
+	 * address space.  The IA-64 architecture guarantees that at least 50 bits of
+	 * virtual address space are implemented but if we pick a large enough page size
+	 * (e.g., 64KB), the mapped address space is big enough that it will overlap with
+	 * VMLPT.  I assume that once we run on machines big enough to warrant 64KB pages,
+	 * IMPL_VA_MSB will be significantly bigger, so this is unlikely to become a
+	 * problem in practice.  Alternatively, we could truncate the top of the mapped
+	 * address space to not permit mappings that would overlap with the VMLPT.
+	 * --davidm 00/12/06
+	 */
+#	define pte_bits			3
+#	define mapped_space_bits	(3*(PAGE_SHIFT - pte_bits) + PAGE_SHIFT)
+	/*
+	 * The virtual page table has to cover the entire implemented address space within
+	 * a region even though not all of this space may be mappable.  The reason for
+	 * this is that the Access bit and Dirty bit fault handlers perform
+	 * non-speculative accesses to the virtual page table, so the address range of the
+	 * virtual page table itself needs to be covered by virtual page table.
+	 */
+#	define vmlpt_bits		(impl_va_bits - PAGE_SHIFT + pte_bits)
+#	define POW2(n)			(1ULL << (n))
+
+	impl_va_bits = ffz(~(local_cpu_data->unimpl_va_mask | (7UL << 61)));
+
+	if (impl_va_bits < 51 || impl_va_bits > 61)
+		panic("CPU has bogus IMPL_VA_MSB value of %lu!\n", impl_va_bits - 1);
+	/*
+	 * mapped_space_bits - PAGE_SHIFT is the total number of ptes we need,
+	 * which must fit into "vmlpt_bits - pte_bits" slots. Second half of
+	 * the test makes sure that our mapped space doesn't overlap the
+	 * unimplemented hole in the middle of the region.
+	 */
+	if ((mapped_space_bits - PAGE_SHIFT > vmlpt_bits - pte_bits) ||
+	    (mapped_space_bits > impl_va_bits - 1))
+		panic("Cannot build a big enough virtual-linear page table"
+		      " to cover mapped address space.\n"
+		      " Try using a smaller page size.\n");
+
+
+	/* place the VMLPT at the end of each page-table mapped region: */
+	pta = POW2(61) - POW2(vmlpt_bits);
+
+	/*
+	 * Set the (virtually mapped linear) page table address.  Bit
+	 * 8 selects between the short and long format, bits 2-7 the
+	 * size of the table, and bit 0 whether the VHPT walker is
+	 * enabled.
+	 */
+	ia64_set_pta(pta | (0 << 8) | (vmlpt_bits << 2) | VHPT_ENABLE_BIT);
+
+	ia64_tlb_init();
+
+#ifdef	CONFIG_HUGETLB_PAGE
+	ia64_set_rr(HPAGE_REGION_BASE, HPAGE_SHIFT << 2);
+	ia64_srlz_d();
+#endif
+}
+
+#ifdef CONFIG_VIRTUAL_MEM_MAP
+int vmemmap_find_next_valid_pfn(int node, int i)
+{
+	unsigned long end_address, hole_next_pfn;
+	unsigned long stop_address;
+	pg_data_t *pgdat = NODE_DATA(node);
+
+	end_address = (unsigned long) &vmem_map[pgdat->node_start_pfn + i];
+	end_address = PAGE_ALIGN(end_address);
+	stop_address = (unsigned long) &vmem_map[pgdat_end_pfn(pgdat)];
+
+	do {
+		pgd_t *pgd;
+		pud_t *pud;
+		pmd_t *pmd;
+		pte_t *pte;
+
+		pgd = pgd_offset_k(end_address);
+		if (pgd_none(*pgd)) {
+			end_address += PGDIR_SIZE;
+			continue;
+		}
+
+		pud = pud_offset(pgd, end_address);
+		if (pud_none(*pud)) {
+			end_address += PUD_SIZE;
+			continue;
+		}
+
+		pmd = pmd_offset(pud, end_address);
+		if (pmd_none(*pmd)) {
+			end_address += PMD_SIZE;
+			continue;
+		}
+
+		pte = pte_offset_kernel(pmd, end_address);
+retry_pte:
+		if (pte_none(*pte)) {
+			end_address += PAGE_SIZE;
+			pte++;
+			if ((end_address < stop_address) &&
+			    (end_address != ALIGN(end_address, 1UL << PMD_SHIFT)))
+				goto retry_pte;
+			continue;
+		}
+		/* Found next valid vmem_map page */
+		break;
+	} while (end_address < stop_address);
+
+	end_address = min(end_address, stop_address);
+	end_address = end_address - (unsigned long) vmem_map + sizeof(struct page) - 1;
+	hole_next_pfn = end_address / sizeof(struct page);
+	return hole_next_pfn - pgdat->node_start_pfn;
+}
+
+int __init create_mem_map_page_table(u64 start, u64 end, void *arg)
+{
+	unsigned long address, start_page, end_page;
+	struct page *map_start, *map_end;
+	int node;
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+	pte_t *pte;
+
+	map_start = vmem_map + (__pa(start) >> PAGE_SHIFT);
+	map_end   = vmem_map + (__pa(end) >> PAGE_SHIFT);
+
+	start_page = (unsigned long) map_start & PAGE_MASK;
+	end_page = PAGE_ALIGN((unsigned long) map_end);
+	node = paddr_to_nid(__pa(start));
+
+	for (address = start_page; address < end_page; address += PAGE_SIZE) {
+		pgd = pgd_offset_k(address);
+		if (pgd_none(*pgd))
+			pgd_populate(&init_mm, pgd, alloc_bootmem_pages_node(NODE_DATA(node), PAGE_SIZE));
+		pud = pud_offset(pgd, address);
+
+		if (pud_none(*pud))
+			pud_populate(&init_mm, pud, alloc_bootmem_pages_node(NODE_DATA(node), PAGE_SIZE));
+		pmd = pmd_offset(pud, address);
+
+		if (pmd_none(*pmd))
+			pmd_populate_kernel(&init_mm, pmd, alloc_bootmem_pages_node(NODE_DATA(node), PAGE_SIZE));
+		pte = pte_offset_kernel(pmd, address);
+
+		if (pte_none(*pte))
+			set_pte(pte, pfn_pte(__pa(alloc_bootmem_pages_node(NODE_DATA(node), PAGE_SIZE)) >> PAGE_SHIFT,
+					     PAGE_KERNEL));
+	}
+	return 0;
+}
+
+struct memmap_init_callback_data {
+	struct page *start;
+	struct page *end;
+	int nid;
+	unsigned long zone;
+};
+
+static int __meminit
+virtual_memmap_init(u64 start, u64 end, void *arg)
+{
+	struct memmap_init_callback_data *args;
+	struct page *map_start, *map_end;
+
+	args = (struct memmap_init_callback_data *) arg;
+	map_start = vmem_map + (__pa(start) >> PAGE_SHIFT);
+	map_end   = vmem_map + (__pa(end) >> PAGE_SHIFT);
+
+	if (map_start < args->start)
+		map_start = args->start;
+	if (map_end > args->end)
+		map_end = args->end;
+
+	/*
+	 * We have to initialize "out of bounds" struct page elements that fit completely
+	 * on the same pages that were allocated for the "in bounds" elements because they
+	 * may be referenced later (and found to be "reserved").
+	 */
+	map_start -= ((unsigned long) map_start & (PAGE_SIZE - 1)) / sizeof(struct page);
+	map_end += ((PAGE_ALIGN((unsigned long) map_end) - (unsigned long) map_end)
+		    / sizeof(struct page));
+
+	if (map_start < map_end)
+		memmap_init_zone((unsigned long)(map_end - map_start),
+				 args->nid, args->zone, page_to_pfn(map_start),
+				 MEMMAP_EARLY);
+	return 0;
+}
+
+void __meminit
+memmap_init (unsigned long size, int nid, unsigned long zone,
+	     unsigned long start_pfn)
+{
+	if (!vmem_map)
+		memmap_init_zone(size, nid, zone, start_pfn, MEMMAP_EARLY);
+	else {
+		struct page *start;
+		struct memmap_init_callback_data args;
+
+		start = pfn_to_page(start_pfn);
+		args.start = start;
+		args.end = start + size;
+		args.nid = nid;
+		args.zone = zone;
+
+		efi_memmap_walk(virtual_memmap_init, &args);
+	}
+}
+
+int
+ia64_pfn_valid (unsigned long pfn)
+{
+	char byte;
+	struct page *pg = pfn_to_page(pfn);
+
+	return     (__get_user(byte, (char __user *) pg) == 0)
+		&& ((((u64)pg & PAGE_MASK) == (((u64)(pg + 1) - 1) & PAGE_MASK))
+			|| (__get_user(byte, (char __user *) (pg + 1) - 1) == 0));
+}
+EXPORT_SYMBOL(ia64_pfn_valid);
+
+int __init find_largest_hole(u64 start, u64 end, void *arg)
+{
+	u64 *max_gap = arg;
+
+	static u64 last_end = PAGE_OFFSET;
+
+	/* NOTE: this algorithm assumes efi memmap table is ordered */
+
+	if (*max_gap < (start - last_end))
+		*max_gap = start - last_end;
+	last_end = end;
+	return 0;
+}
+
+#endif /* CONFIG_VIRTUAL_MEM_MAP */
+
+int __init register_active_ranges(u64 start, u64 len, int nid)
+{
+	u64 end = start + len;
+
+#ifdef CONFIG_KEXEC
+	if (start > crashk_res.start && start < crashk_res.end)
+		start = crashk_res.end;
+	if (end > crashk_res.start && end < crashk_res.end)
+		end = crashk_res.start;
+#endif
+
+	if (start < end)
+		memblock_add_node(__pa(start), end - start, nid);
+	return 0;
+}
+
+int
+find_max_min_low_pfn (u64 start, u64 end, void *arg)
+{
+	unsigned long pfn_start, pfn_end;
+#ifdef CONFIG_FLATMEM
+	pfn_start = (PAGE_ALIGN(__pa(start))) >> PAGE_SHIFT;
+	pfn_end = (PAGE_ALIGN(__pa(end - 1))) >> PAGE_SHIFT;
+#else
+	pfn_start = GRANULEROUNDDOWN(__pa(start)) >> PAGE_SHIFT;
+	pfn_end = GRANULEROUNDUP(__pa(end - 1)) >> PAGE_SHIFT;
+#endif
+	min_low_pfn = min(min_low_pfn, pfn_start);
+	max_low_pfn = max(max_low_pfn, pfn_end);
+	return 0;
+}
+
+/*
+ * Boot command-line option "nolwsys" can be used to disable the use of any light-weight
+ * system call handler.  When this option is in effect, all fsyscalls will end up bubbling
+ * down into the kernel and calling the normal (heavy-weight) syscall handler.  This is
+ * useful for performance testing, but conceivably could also come in handy for debugging
+ * purposes.
+ */
+
+static int nolwsys __initdata;
+
+static int __init
+nolwsys_setup (char *s)
+{
+	nolwsys = 1;
+	return 1;
+}
+
+__setup("nolwsys", nolwsys_setup);
+
+void __init
+mem_init (void)
+{
+	int i;
+
+	BUG_ON(PTRS_PER_PGD * sizeof(pgd_t) != PAGE_SIZE);
+	BUG_ON(PTRS_PER_PMD * sizeof(pmd_t) != PAGE_SIZE);
+	BUG_ON(PTRS_PER_PTE * sizeof(pte_t) != PAGE_SIZE);
+
+#ifdef CONFIG_PCI
+	/*
+	 * This needs to be called _after_ the command line has been parsed but _before_
+	 * any drivers that may need the PCI DMA interface are initialized or bootmem has
+	 * been freed.
+	 */
+	platform_dma_init();
+#endif
+
+#ifdef CONFIG_FLATMEM
+	BUG_ON(!mem_map);
+#endif
+
+	set_max_mapnr(max_low_pfn);
+	high_memory = __va(max_low_pfn * PAGE_SIZE);
+	free_all_bootmem();
+	mem_init_print_info(NULL);
+
+	/*
+	 * For fsyscall entrpoints with no light-weight handler, use the ordinary
+	 * (heavy-weight) handler, but mark it by setting bit 0, so the fsyscall entry
+	 * code can tell them apart.
+	 */
+	for (i = 0; i < NR_syscalls; ++i) {
+		extern unsigned long sys_call_table[NR_syscalls];
+		unsigned long *fsyscall_table = paravirt_get_fsyscall_table();
+
+		if (!fsyscall_table[i] || nolwsys)
+			fsyscall_table[i] = sys_call_table[i] | 1;
+	}
+	setup_gate();
+}
+
+#ifdef CONFIG_MEMORY_HOTPLUG
+int arch_add_memory(int nid, u64 start, u64 size)
+{
+	pg_data_t *pgdat;
+	struct zone *zone;
+	unsigned long start_pfn = start >> PAGE_SHIFT;
+	unsigned long nr_pages = size >> PAGE_SHIFT;
+	int ret;
+
+	pgdat = NODE_DATA(nid);
+
+	zone = pgdat->node_zones +
+		zone_for_memory(nid, start, size, ZONE_NORMAL);
+	ret = __add_pages(nid, zone, start_pfn, nr_pages);
+
+	if (ret)
+		printk("%s: Problem encountered in __add_pages() as ret=%d\n",
+		       __func__,  ret);
+
+	return ret;
+}
+
+#ifdef CONFIG_MEMORY_HOTREMOVE
+int arch_remove_memory(u64 start, u64 size)
+{
+	unsigned long start_pfn = start >> PAGE_SHIFT;
+	unsigned long nr_pages = size >> PAGE_SHIFT;
+	struct zone *zone;
+	int ret;
+
+	zone = page_zone(pfn_to_page(start_pfn));
+	ret = __remove_pages(zone, start_pfn, nr_pages);
+	if (ret)
+		pr_warn("%s: Problem encountered in __remove_pages() as"
+			" ret=%d\n", __func__,  ret);
+
+	return ret;
+}
+#endif
+#endif
+
+/*
+ * Even when CONFIG_IA32_SUPPORT is not enabled it is
+ * useful to have the Linux/x86 domain registered to
+ * avoid an attempted module load when emulators call
+ * personality(PER_LINUX32). This saves several milliseconds
+ * on each such call.
+ */
+static struct exec_domain ia32_exec_domain;
+
+static int __init
+per_linux32_init(void)
+{
+	ia32_exec_domain.name = "Linux/x86";
+	ia32_exec_domain.handler = NULL;
+	ia32_exec_domain.pers_low = PER_LINUX32;
+	ia32_exec_domain.pers_high = PER_LINUX32;
+	ia32_exec_domain.signal_map = default_exec_domain.signal_map;
+	ia32_exec_domain.signal_invmap = default_exec_domain.signal_invmap;
+	register_exec_domain(&ia32_exec_domain);
+
+	return 0;
+}
+
+__initcall(per_linux32_init);
+
+/**
+ * show_mem - give short summary of memory stats
+ *
+ * Shows a simple page count of reserved and used pages in the system.
+ * For discontig machines, it does this on a per-pgdat basis.
+ */
+void show_mem(unsigned int filter)
+{
+	int total_reserved = 0;
+	unsigned long total_present = 0;
+	pg_data_t *pgdat;
+
+	printk(KERN_INFO "Mem-info:\n");
+	show_free_areas(filter);
+	printk(KERN_INFO "Node memory in pages:\n");
+	for_each_online_pgdat(pgdat) {
+		unsigned long present;
+		unsigned long flags;
+		int reserved = 0;
+		int nid = pgdat->node_id;
+		int zoneid;
+
+		if (skip_free_areas_node(filter, nid))
+			continue;
+		pgdat_resize_lock(pgdat, &flags);
+
+		for (zoneid = 0; zoneid < MAX_NR_ZONES; zoneid++) {
+			struct zone *zone = &pgdat->node_zones[zoneid];
+			if (!populated_zone(zone))
+				continue;
+
+			reserved += zone->present_pages - zone->managed_pages;
+		}
+		present = pgdat->node_present_pages;
+
+		pgdat_resize_unlock(pgdat, &flags);
+		total_present += present;
+		total_reserved += reserved;
+		printk(KERN_INFO "Node %4d:  RAM: %11ld, rsvd: %8d, ",
+		       nid, present, reserved);
+	}
+	printk(KERN_INFO "%ld pages of RAM\n", total_present);
+	printk(KERN_INFO "%d reserved pages\n", total_reserved);
+	printk(KERN_INFO "Total of %ld pages in page table cache\n",
+	       quicklist_total_size());
+	printk(KERN_INFO "%ld free buffer pages\n", nr_free_buffer_pages());
+}

arch/ia64/mm/ioremap.c

@@ -0,0 +1,125 @@
+/*
+ * (c) Copyright 2006, 2007 Hewlett-Packard Development Company, L.P.
+ *	Bjorn Helgaas <bjorn.helgaas@hp.com>
+ *
+ * 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.
+ */
+
+#include <linux/compiler.h>
+#include <linux/module.h>
+#include <linux/efi.h>
+#include <linux/io.h>
+#include <linux/vmalloc.h>
+#include <asm/io.h>
+#include <asm/meminit.h>
+
+static inline void __iomem *
+__ioremap_uc(unsigned long phys_addr)
+{
+	return (void __iomem *) (__IA64_UNCACHED_OFFSET | phys_addr);
+}
+
+void __iomem *
+early_ioremap (unsigned long phys_addr, unsigned long size)
+{
+	u64 attr;
+	attr = kern_mem_attribute(phys_addr, size);
+	if (attr & EFI_MEMORY_WB)
+		return (void __iomem *) phys_to_virt(phys_addr);
+	return __ioremap_uc(phys_addr);
+}
+
+void __iomem *
+ioremap (unsigned long phys_addr, unsigned long size)
+{
+	void __iomem *addr;
+	struct vm_struct *area;
+	unsigned long offset;
+	pgprot_t prot;
+	u64 attr;
+	unsigned long gran_base, gran_size;
+	unsigned long page_base;
+
+	/*
+	 * For things in kern_memmap, we must use the same attribute
+	 * as the rest of the kernel.  For more details, see
+	 * Documentation/ia64/aliasing.txt.
+	 */
+	attr = kern_mem_attribute(phys_addr, size);
+	if (attr & EFI_MEMORY_WB)
+		return (void __iomem *) phys_to_virt(phys_addr);
+	else if (attr & EFI_MEMORY_UC)
+		return __ioremap_uc(phys_addr);
+
+	/*
+	 * Some chipsets don't support UC access to memory.  If
+	 * WB is supported for the whole granule, we prefer that.
+	 */
+	gran_base = GRANULEROUNDDOWN(phys_addr);
+	gran_size = GRANULEROUNDUP(phys_addr + size) - gran_base;
+	if (efi_mem_attribute(gran_base, gran_size) & EFI_MEMORY_WB)
+		return (void __iomem *) phys_to_virt(phys_addr);
+
+	/*
+	 * WB is not supported for the whole granule, so we can't use
+	 * the region 7 identity mapping.  If we can safely cover the
+	 * area with kernel page table mappings, we can use those
+	 * instead.
+	 */
+	page_base = phys_addr & PAGE_MASK;
+	size = PAGE_ALIGN(phys_addr + size) - page_base;
+	if (efi_mem_attribute(page_base, size) & EFI_MEMORY_WB) {
+		prot = PAGE_KERNEL;
+
+		/*
+		 * Mappings have to be page-aligned
+		 */
+		offset = phys_addr & ~PAGE_MASK;
+		phys_addr &= PAGE_MASK;
+
+		/*
+		 * Ok, go for it..
+		 */
+		area = get_vm_area(size, VM_IOREMAP);
+		if (!area)
+			return NULL;
+
+		area->phys_addr = phys_addr;
+		addr = (void __iomem *) area->addr;
+		if (ioremap_page_range((unsigned long) addr,
+				(unsigned long) addr + size, phys_addr, prot)) {
+			vunmap((void __force *) addr);
+			return NULL;
+		}
+
+		return (void __iomem *) (offset + (char __iomem *)addr);
+	}
+
+	return __ioremap_uc(phys_addr);
+}
+EXPORT_SYMBOL(ioremap);
+
+void __iomem *
+ioremap_nocache (unsigned long phys_addr, unsigned long size)
+{
+	if (kern_mem_attribute(phys_addr, size) & EFI_MEMORY_WB)
+		return NULL;
+
+	return __ioremap_uc(phys_addr);
+}
+EXPORT_SYMBOL(ioremap_nocache);
+
+void
+early_iounmap (volatile void __iomem *addr, unsigned long size)
+{
+}
+
+void
+iounmap (volatile void __iomem *addr)
+{
+	if (REGION_NUMBER(addr) == RGN_GATE)
+		vunmap((void *) ((unsigned long) addr & PAGE_MASK));
+}
+EXPORT_SYMBOL(iounmap);

arch/ia64/mm/numa.c

@@ -0,0 +1,110 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * This file contains NUMA specific variables and functions which can
+ * be split away from DISCONTIGMEM and are used on NUMA machines with
+ * contiguous memory.
+ * 
+ *                         2002/08/07 Erich Focht <efocht@ess.nec.de>
+ */
+
+#include <linux/cpu.h>
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/node.h>
+#include <linux/init.h>
+#include <linux/bootmem.h>
+#include <linux/module.h>
+#include <asm/mmzone.h>
+#include <asm/numa.h>
+
+
+/*
+ * The following structures are usually initialized by ACPI or
+ * similar mechanisms and describe the NUMA characteristics of the machine.
+ */
+int num_node_memblks;
+struct node_memblk_s node_memblk[NR_NODE_MEMBLKS];
+struct node_cpuid_s node_cpuid[NR_CPUS] =
+	{ [0 ... NR_CPUS-1] = { .phys_id = 0, .nid = NUMA_NO_NODE } };
+
+/*
+ * This is a matrix with "distances" between nodes, they should be
+ * proportional to the memory access latency ratios.
+ */
+u8 numa_slit[MAX_NUMNODES * MAX_NUMNODES];
+
+/* Identify which cnode a physical address resides on */
+int
+paddr_to_nid(unsigned long paddr)
+{
+	int	i;
+
+	for (i = 0; i < num_node_memblks; i++)
+		if (paddr >= node_memblk[i].start_paddr &&
+		    paddr < node_memblk[i].start_paddr + node_memblk[i].size)
+			break;
+
+	return (i < num_node_memblks) ? node_memblk[i].nid : (num_node_memblks ? -1 : 0);
+}
+
+#if defined(CONFIG_SPARSEMEM) && defined(CONFIG_NUMA)
+/*
+ * Because of holes evaluate on section limits.
+ * If the section of memory exists, then return the node where the section
+ * resides.  Otherwise return node 0 as the default.  This is used by
+ * SPARSEMEM to allocate the SPARSEMEM sectionmap on the NUMA node where
+ * the section resides.
+ */
+int __meminit __early_pfn_to_nid(unsigned long pfn)
+{
+	int i, section = pfn >> PFN_SECTION_SHIFT, ssec, esec;
+	/*
+	 * NOTE: The following SMP-unsafe globals are only used early in boot
+	 * when the kernel is running single-threaded.
+	 */
+	static int __meminitdata last_ssec, last_esec;
+	static int __meminitdata last_nid;
+
+	if (section >= last_ssec && section < last_esec)
+		return last_nid;
+
+	for (i = 0; i < num_node_memblks; i++) {
+		ssec = node_memblk[i].start_paddr >> PA_SECTION_SHIFT;
+		esec = (node_memblk[i].start_paddr + node_memblk[i].size +
+			((1L << PA_SECTION_SHIFT) - 1)) >> PA_SECTION_SHIFT;
+		if (section >= ssec && section < esec) {
+			last_ssec = ssec;
+			last_esec = esec;
+			last_nid = node_memblk[i].nid;
+			return node_memblk[i].nid;
+		}
+	}
+
+	return -1;
+}
+
+void numa_clear_node(int cpu)
+{
+	unmap_cpu_from_node(cpu, NUMA_NO_NODE);
+}
+
+#ifdef CONFIG_MEMORY_HOTPLUG
+/*
+ *  SRAT information is stored in node_memblk[], then we can use SRAT
+ *  information at memory-hot-add if necessary.
+ */
+
+int memory_add_physaddr_to_nid(u64 addr)
+{
+	int nid = paddr_to_nid(addr);
+	if (nid < 0)
+		return 0;
+	return nid;
+}
+
+EXPORT_SYMBOL_GPL(memory_add_physaddr_to_nid);
+#endif
+#endif

arch/ia64/mm/tlb.c

@@ -0,0 +1,561 @@
+/*
+ * TLB support routines.
+ *
+ * Copyright (C) 1998-2001, 2003 Hewlett-Packard Co
+ *	David Mosberger-Tang <davidm@hpl.hp.com>
+ *
+ * 08/02/00 A. Mallick <asit.k.mallick@intel.com>
+ *		Modified RID allocation for SMP
+ *          Goutham Rao <goutham.rao@intel.com>
+ *              IPI based ptc implementation and A-step IPI implementation.
+ * Rohit Seth <rohit.seth@intel.com>
+ * Ken Chen <kenneth.w.chen@intel.com>
+ * Christophe de Dinechin <ddd@hp.com>: Avoid ptc.e on memory allocation
+ * Copyright (C) 2007 Intel Corp
+ *	Fenghua Yu <fenghua.yu@intel.com>
+ *	Add multiple ptc.g/ptc.ga instruction support in global tlb purge.
+ */
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/smp.h>
+#include <linux/mm.h>
+#include <linux/bootmem.h>
+#include <linux/slab.h>
+
+#include <asm/delay.h>
+#include <asm/mmu_context.h>
+#include <asm/pgalloc.h>
+#include <asm/pal.h>
+#include <asm/tlbflush.h>
+#include <asm/dma.h>
+#include <asm/processor.h>
+#include <asm/sal.h>
+#include <asm/tlb.h>
+
+static struct {
+	u64 mask;		/* mask of supported purge page-sizes */
+	unsigned long max_bits;	/* log2 of largest supported purge page-size */
+} purge;
+
+struct ia64_ctx ia64_ctx = {
+	.lock =	__SPIN_LOCK_UNLOCKED(ia64_ctx.lock),
+	.next =	1,
+	.max_ctx = ~0U
+};
+
+DEFINE_PER_CPU(u8, ia64_need_tlb_flush);
+DEFINE_PER_CPU(u8, ia64_tr_num);  /*Number of TR slots in current processor*/
+DEFINE_PER_CPU(u8, ia64_tr_used); /*Max Slot number used by kernel*/
+
+struct ia64_tr_entry *ia64_idtrs[NR_CPUS];
+
+/*
+ * Initializes the ia64_ctx.bitmap array based on max_ctx+1.
+ * Called after cpu_init() has setup ia64_ctx.max_ctx based on
+ * maximum RID that is supported by boot CPU.
+ */
+void __init
+mmu_context_init (void)
+{
+	ia64_ctx.bitmap = alloc_bootmem((ia64_ctx.max_ctx+1)>>3);
+	ia64_ctx.flushmap = alloc_bootmem((ia64_ctx.max_ctx+1)>>3);
+}
+
+/*
+ * Acquire the ia64_ctx.lock before calling this function!
+ */
+void
+wrap_mmu_context (struct mm_struct *mm)
+{
+	int i, cpu;
+	unsigned long flush_bit;
+
+	for (i=0; i <= ia64_ctx.max_ctx / BITS_PER_LONG; i++) {
+		flush_bit = xchg(&ia64_ctx.flushmap[i], 0);
+		ia64_ctx.bitmap[i] ^= flush_bit;
+	}
+ 
+	/* use offset at 300 to skip daemons */
+	ia64_ctx.next = find_next_zero_bit(ia64_ctx.bitmap,
+				ia64_ctx.max_ctx, 300);
+	ia64_ctx.limit = find_next_bit(ia64_ctx.bitmap,
+				ia64_ctx.max_ctx, ia64_ctx.next);
+
+	/*
+	 * can't call flush_tlb_all() here because of race condition
+	 * with O(1) scheduler [EF]
+	 */
+	cpu = get_cpu(); /* prevent preemption/migration */
+	for_each_online_cpu(i)
+		if (i != cpu)
+			per_cpu(ia64_need_tlb_flush, i) = 1;
+	put_cpu();
+	local_flush_tlb_all();
+}
+
+/*
+ * Implement "spinaphores" ... like counting semaphores, but they
+ * spin instead of sleeping.  If there are ever any other users for
+ * this primitive it can be moved up to a spinaphore.h header.
+ */
+struct spinaphore {
+	unsigned long	ticket;
+	unsigned long	serve;
+};
+
+static inline void spinaphore_init(struct spinaphore *ss, int val)
+{
+	ss->ticket = 0;
+	ss->serve = val;
+}
+
+static inline void down_spin(struct spinaphore *ss)
+{
+	unsigned long t = ia64_fetchadd(1, &ss->ticket, acq), serve;
+
+	if (time_before(t, ss->serve))
+		return;
+
+	ia64_invala();
+
+	for (;;) {
+		asm volatile ("ld8.c.nc %0=[%1]" : "=r"(serve) : "r"(&ss->serve) : "memory");
+		if (time_before(t, serve))
+			return;
+		cpu_relax();
+	}
+}
+
+static inline void up_spin(struct spinaphore *ss)
+{
+	ia64_fetchadd(1, &ss->serve, rel);
+}
+
+static struct spinaphore ptcg_sem;
+static u16 nptcg = 1;
+static int need_ptcg_sem = 1;
+static int toolatetochangeptcgsem = 0;
+
+/*
+ * Kernel parameter "nptcg=" overrides max number of concurrent global TLB
+ * purges which is reported from either PAL or SAL PALO.
+ *
+ * We don't have sanity checking for nptcg value. It's the user's responsibility
+ * for valid nptcg value on the platform. Otherwise, kernel may hang in some
+ * cases.
+ */
+static int __init
+set_nptcg(char *str)
+{
+	int value = 0;
+
+	get_option(&str, &value);
+	setup_ptcg_sem(value, NPTCG_FROM_KERNEL_PARAMETER);
+
+	return 1;
+}
+
+__setup("nptcg=", set_nptcg);
+
+/*
+ * Maximum number of simultaneous ptc.g purges in the system can
+ * be defined by PAL_VM_SUMMARY (in which case we should take
+ * the smallest value for any cpu in the system) or by the PAL
+ * override table (in which case we should ignore the value from
+ * PAL_VM_SUMMARY).
+ *
+ * Kernel parameter "nptcg=" overrides maximum number of simultanesous ptc.g
+ * purges defined in either PAL_VM_SUMMARY or PAL override table. In this case,
+ * we should ignore the value from either PAL_VM_SUMMARY or PAL override table.
+ *
+ * Complicating the logic here is the fact that num_possible_cpus()
+ * isn't fully setup until we start bringing cpus online.
+ */
+void
+setup_ptcg_sem(int max_purges, int nptcg_from)
+{
+	static int kp_override;
+	static int palo_override;
+	static int firstcpu = 1;
+
+	if (toolatetochangeptcgsem) {
+		if (nptcg_from == NPTCG_FROM_PAL && max_purges == 0)
+			BUG_ON(1 < nptcg);
+		else
+			BUG_ON(max_purges < nptcg);
+		return;
+	}
+
+	if (nptcg_from == NPTCG_FROM_KERNEL_PARAMETER) {
+		kp_override = 1;
+		nptcg = max_purges;
+		goto resetsema;
+	}
+	if (kp_override) {
+		need_ptcg_sem = num_possible_cpus() > nptcg;
+		return;
+	}
+
+	if (nptcg_from == NPTCG_FROM_PALO) {
+		palo_override = 1;
+
+		/* In PALO max_purges == 0 really means it! */
+		if (max_purges == 0)
+			panic("Whoa! Platform does not support global TLB purges.\n");
+		nptcg = max_purges;
+		if (nptcg == PALO_MAX_TLB_PURGES) {
+			need_ptcg_sem = 0;
+			return;
+		}
+		goto resetsema;
+	}
+	if (palo_override) {
+		if (nptcg != PALO_MAX_TLB_PURGES)
+			need_ptcg_sem = (num_possible_cpus() > nptcg);
+		return;
+	}
+
+	/* In PAL_VM_SUMMARY max_purges == 0 actually means 1 */
+	if (max_purges == 0) max_purges = 1;
+
+	if (firstcpu) {
+		nptcg = max_purges;
+		firstcpu = 0;
+	}
+	if (max_purges < nptcg)
+		nptcg = max_purges;
+	if (nptcg == PAL_MAX_PURGES) {
+		need_ptcg_sem = 0;
+		return;
+	} else
+		need_ptcg_sem = (num_possible_cpus() > nptcg);
+
+resetsema:
+	spinaphore_init(&ptcg_sem, max_purges);
+}
+
+void
+ia64_global_tlb_purge (struct mm_struct *mm, unsigned long start,
+		       unsigned long end, unsigned long nbits)
+{
+	struct mm_struct *active_mm = current->active_mm;
+
+	toolatetochangeptcgsem = 1;
+
+	if (mm != active_mm) {
+		/* Restore region IDs for mm */
+		if (mm && active_mm) {
+			activate_context(mm);
+		} else {
+			flush_tlb_all();
+			return;
+		}
+	}
+
+	if (need_ptcg_sem)
+		down_spin(&ptcg_sem);
+
+	do {
+		/*
+		 * Flush ALAT entries also.
+		 */
+		ia64_ptcga(start, (nbits << 2));
+		ia64_srlz_i();
+		start += (1UL << nbits);
+	} while (start < end);
+
+	if (need_ptcg_sem)
+		up_spin(&ptcg_sem);
+
+        if (mm != active_mm) {
+                activate_context(active_mm);
+        }
+}
+
+void
+local_flush_tlb_all (void)
+{
+	unsigned long i, j, flags, count0, count1, stride0, stride1, addr;
+
+	addr    = local_cpu_data->ptce_base;
+	count0  = local_cpu_data->ptce_count[0];
+	count1  = local_cpu_data->ptce_count[1];
+	stride0 = local_cpu_data->ptce_stride[0];
+	stride1 = local_cpu_data->ptce_stride[1];
+
+	local_irq_save(flags);
+	for (i = 0; i < count0; ++i) {
+		for (j = 0; j < count1; ++j) {
+			ia64_ptce(addr);
+			addr += stride1;
+		}
+		addr += stride0;
+	}
+	local_irq_restore(flags);
+	ia64_srlz_i();			/* srlz.i implies srlz.d */
+}
+
+void
+flush_tlb_range (struct vm_area_struct *vma, unsigned long start,
+		 unsigned long end)
+{
+	struct mm_struct *mm = vma->vm_mm;
+	unsigned long size = end - start;
+	unsigned long nbits;
+
+#ifndef CONFIG_SMP
+	if (mm != current->active_mm) {
+		mm->context = 0;
+		return;
+	}
+#endif
+
+	nbits = ia64_fls(size + 0xfff);
+	while (unlikely (((1UL << nbits) & purge.mask) == 0) &&
+			(nbits < purge.max_bits))
+		++nbits;
+	if (nbits > purge.max_bits)
+		nbits = purge.max_bits;
+	start &= ~((1UL << nbits) - 1);
+
+	preempt_disable();
+#ifdef CONFIG_SMP
+	if (mm != current->active_mm || cpumask_weight(mm_cpumask(mm)) != 1) {
+		platform_global_tlb_purge(mm, start, end, nbits);
+		preempt_enable();
+		return;
+	}
+#endif
+	do {
+		ia64_ptcl(start, (nbits<<2));
+		start += (1UL << nbits);
+	} while (start < end);
+	preempt_enable();
+	ia64_srlz_i();			/* srlz.i implies srlz.d */
+}
+EXPORT_SYMBOL(flush_tlb_range);
+
+void ia64_tlb_init(void)
+{
+	ia64_ptce_info_t uninitialized_var(ptce_info); /* GCC be quiet */
+	u64 tr_pgbits;
+	long status;
+	pal_vm_info_1_u_t vm_info_1;
+	pal_vm_info_2_u_t vm_info_2;
+	int cpu = smp_processor_id();
+
+	if ((status = ia64_pal_vm_page_size(&tr_pgbits, &purge.mask)) != 0) {
+		printk(KERN_ERR "PAL_VM_PAGE_SIZE failed with status=%ld; "
+		       "defaulting to architected purge page-sizes.\n", status);
+		purge.mask = 0x115557000UL;
+	}
+	purge.max_bits = ia64_fls(purge.mask);
+
+	ia64_get_ptce(&ptce_info);
+	local_cpu_data->ptce_base = ptce_info.base;
+	local_cpu_data->ptce_count[0] = ptce_info.count[0];
+	local_cpu_data->ptce_count[1] = ptce_info.count[1];
+	local_cpu_data->ptce_stride[0] = ptce_info.stride[0];
+	local_cpu_data->ptce_stride[1] = ptce_info.stride[1];
+
+	local_flush_tlb_all();	/* nuke left overs from bootstrapping... */
+	status = ia64_pal_vm_summary(&vm_info_1, &vm_info_2);
+
+	if (status) {
+		printk(KERN_ERR "ia64_pal_vm_summary=%ld\n", status);
+		per_cpu(ia64_tr_num, cpu) = 8;
+		return;
+	}
+	per_cpu(ia64_tr_num, cpu) = vm_info_1.pal_vm_info_1_s.max_itr_entry+1;
+	if (per_cpu(ia64_tr_num, cpu) >
+				(vm_info_1.pal_vm_info_1_s.max_dtr_entry+1))
+		per_cpu(ia64_tr_num, cpu) =
+				vm_info_1.pal_vm_info_1_s.max_dtr_entry+1;
+	if (per_cpu(ia64_tr_num, cpu) > IA64_TR_ALLOC_MAX) {
+		static int justonce = 1;
+		per_cpu(ia64_tr_num, cpu) = IA64_TR_ALLOC_MAX;
+		if (justonce) {
+			justonce = 0;
+			printk(KERN_DEBUG "TR register number exceeds "
+			       "IA64_TR_ALLOC_MAX!\n");
+		}
+	}
+}
+
+/*
+ * is_tr_overlap
+ *
+ * Check overlap with inserted TRs.
+ */
+static int is_tr_overlap(struct ia64_tr_entry *p, u64 va, u64 log_size)
+{
+	u64 tr_log_size;
+	u64 tr_end;
+	u64 va_rr = ia64_get_rr(va);
+	u64 va_rid = RR_TO_RID(va_rr);
+	u64 va_end = va + (1<<log_size) - 1;
+
+	if (va_rid != RR_TO_RID(p->rr))
+		return 0;
+	tr_log_size = (p->itir & 0xff) >> 2;
+	tr_end = p->ifa + (1<<tr_log_size) - 1;
+
+	if (va > tr_end || p->ifa > va_end)
+		return 0;
+	return 1;
+
+}
+
+/*
+ * ia64_insert_tr in virtual mode. Allocate a TR slot
+ *
+ * target_mask : 0x1 : itr, 0x2 : dtr, 0x3 : idtr
+ *
+ * va 	: virtual address.
+ * pte 	: pte entries inserted.
+ * log_size: range to be covered.
+ *
+ * Return value:  <0 :  error No.
+ *
+ *		  >=0 : slot number allocated for TR.
+ * Must be called with preemption disabled.
+ */
+int ia64_itr_entry(u64 target_mask, u64 va, u64 pte, u64 log_size)
+{
+	int i, r;
+	unsigned long psr;
+	struct ia64_tr_entry *p;
+	int cpu = smp_processor_id();
+
+	if (!ia64_idtrs[cpu]) {
+		ia64_idtrs[cpu] = kmalloc(2 * IA64_TR_ALLOC_MAX *
+				sizeof (struct ia64_tr_entry), GFP_KERNEL);
+		if (!ia64_idtrs[cpu])
+			return -ENOMEM;
+	}
+	r = -EINVAL;
+	/*Check overlap with existing TR entries*/
+	if (target_mask & 0x1) {
+		p = ia64_idtrs[cpu];
+		for (i = IA64_TR_ALLOC_BASE; i <= per_cpu(ia64_tr_used, cpu);
+								i++, p++) {
+			if (p->pte & 0x1)
+				if (is_tr_overlap(p, va, log_size)) {
+					printk(KERN_DEBUG "Overlapped Entry"
+						"Inserted for TR Reigster!!\n");
+					goto out;
+			}
+		}
+	}
+	if (target_mask & 0x2) {
+		p = ia64_idtrs[cpu] + IA64_TR_ALLOC_MAX;
+		for (i = IA64_TR_ALLOC_BASE; i <= per_cpu(ia64_tr_used, cpu);
+								i++, p++) {
+			if (p->pte & 0x1)
+				if (is_tr_overlap(p, va, log_size)) {
+					printk(KERN_DEBUG "Overlapped Entry"
+						"Inserted for TR Reigster!!\n");
+					goto out;
+				}
+		}
+	}
+
+	for (i = IA64_TR_ALLOC_BASE; i < per_cpu(ia64_tr_num, cpu); i++) {
+		switch (target_mask & 0x3) {
+		case 1:
+			if (!((ia64_idtrs[cpu] + i)->pte & 0x1))
+				goto found;
+			continue;
+		case 2:
+			if (!((ia64_idtrs[cpu] + IA64_TR_ALLOC_MAX + i)->pte & 0x1))
+				goto found;
+			continue;
+		case 3:
+			if (!((ia64_idtrs[cpu] + i)->pte & 0x1) &&
+			    !((ia64_idtrs[cpu] + IA64_TR_ALLOC_MAX + i)->pte & 0x1))
+				goto found;
+			continue;
+		default:
+			r = -EINVAL;
+			goto out;
+		}
+	}
+found:
+	if (i >= per_cpu(ia64_tr_num, cpu))
+		return -EBUSY;
+
+	/*Record tr info for mca hander use!*/
+	if (i > per_cpu(ia64_tr_used, cpu))
+		per_cpu(ia64_tr_used, cpu) = i;
+
+	psr = ia64_clear_ic();
+	if (target_mask & 0x1) {
+		ia64_itr(0x1, i, va, pte, log_size);
+		ia64_srlz_i();
+		p = ia64_idtrs[cpu] + i;
+		p->ifa = va;
+		p->pte = pte;
+		p->itir = log_size << 2;
+		p->rr = ia64_get_rr(va);
+	}
+	if (target_mask & 0x2) {
+		ia64_itr(0x2, i, va, pte, log_size);
+		ia64_srlz_i();
+		p = ia64_idtrs[cpu] + IA64_TR_ALLOC_MAX + i;
+		p->ifa = va;
+		p->pte = pte;
+		p->itir = log_size << 2;
+		p->rr = ia64_get_rr(va);
+	}
+	ia64_set_psr(psr);
+	r = i;
+out:
+	return r;
+}
+EXPORT_SYMBOL_GPL(ia64_itr_entry);
+
+/*
+ * ia64_purge_tr
+ *
+ * target_mask: 0x1: purge itr, 0x2 : purge dtr, 0x3 purge idtr.
+ * slot: slot number to be freed.
+ *
+ * Must be called with preemption disabled.
+ */
+void ia64_ptr_entry(u64 target_mask, int slot)
+{
+	int cpu = smp_processor_id();
+	int i;
+	struct ia64_tr_entry *p;
+
+	if (slot < IA64_TR_ALLOC_BASE || slot >= per_cpu(ia64_tr_num, cpu))
+		return;
+
+	if (target_mask & 0x1) {
+		p = ia64_idtrs[cpu] + slot;
+		if ((p->pte&0x1) && is_tr_overlap(p, p->ifa, p->itir>>2)) {
+			p->pte = 0;
+			ia64_ptr(0x1, p->ifa, p->itir>>2);
+			ia64_srlz_i();
+		}
+	}
+
+	if (target_mask & 0x2) {
+		p = ia64_idtrs[cpu] + IA64_TR_ALLOC_MAX + slot;
+		if ((p->pte & 0x1) && is_tr_overlap(p, p->ifa, p->itir>>2)) {
+			p->pte = 0;
+			ia64_ptr(0x2, p->ifa, p->itir>>2);
+			ia64_srlz_i();
+		}
+	}
+
+	for (i = per_cpu(ia64_tr_used, cpu); i >= IA64_TR_ALLOC_BASE; i--) {
+		if (((ia64_idtrs[cpu] + i)->pte & 0x1) ||
+		    ((ia64_idtrs[cpu] + IA64_TR_ALLOC_MAX + i)->pte & 0x1))
+			break;
+	}
+	per_cpu(ia64_tr_used, cpu) = i;
+}
+EXPORT_SYMBOL_GPL(ia64_ptr_entry);