Fixed MTP to work with TWRP

arch/arm/vfp/Makefile

@@ -0,0 +1,15 @@
+#
+# linux/arch/arm/vfp/Makefile
+#
+# Copyright (C) 2001 ARM Limited
+#
+
+# ccflags-y := -DDEBUG
+# asflags-y := -DDEBUG
+
+KBUILD_AFLAGS	:=$(KBUILD_AFLAGS:-msoft-float=-Wa,-mfpu=softvfp+vfp -mfloat-abi=soft)
+LDFLAGS		+=--no-warn-mismatch
+
+obj-y			+= vfp.o
+
+vfp-$(CONFIG_VFP)	+= vfpmodule.o entry.o vfphw.o vfpsingle.o vfpdouble.o

arch/arm/vfp/entry.S

@@ -0,0 +1,59 @@
+/*
+ *  linux/arch/arm/vfp/entry.S
+ *
+ *  Copyright (C) 2004 ARM Limited.
+ *  Written by Deep Blue Solutions Limited.
+ *
+ * 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/init.h>
+#include <linux/linkage.h>
+#include <asm/thread_info.h>
+#include <asm/vfpmacros.h>
+#include <asm/assembler.h>
+#include <asm/asm-offsets.h>
+
+@ VFP entry point.
+@
+@  r0  = instruction opcode (32-bit ARM or two 16-bit Thumb)
+@  r2  = PC value to resume execution after successful emulation
+@  r9  = normal "successful" return address
+@  r10 = this threads thread_info structure
+@  lr  = unrecognised instruction return address
+@  IRQs enabled.
+@
+ENTRY(do_vfp)
+	inc_preempt_count r10, r4
+ 	ldr	r4, .LCvfp
+	ldr	r11, [r10, #TI_CPU]	@ CPU number
+	add	r10, r10, #TI_VFPSTATE	@ r10 = workspace
+	ldr	pc, [r4]		@ call VFP entry point
+ENDPROC(do_vfp)
+
+ENTRY(vfp_null_entry)
+	dec_preempt_count_ti r10, r4
+	ret	lr
+ENDPROC(vfp_null_entry)
+
+	.align	2
+.LCvfp:
+	.word	vfp_vector
+
+@ This code is called if the VFP does not exist. It needs to flag the
+@ failure to the VFP initialisation code.
+
+	__INIT
+ENTRY(vfp_testing_entry)
+	dec_preempt_count_ti r10, r4
+	ldr	r0, VFP_arch_address
+	str	r0, [r0]		@ set to non-zero value
+	ret	r9			@ we have handled the fault
+ENDPROC(vfp_testing_entry)
+
+	.align	2
+VFP_arch_address:
+	.word	VFP_arch
+
+	__FINIT

arch/arm/vfp/vfp.h

@@ -0,0 +1,380 @@
+/*
+ *  linux/arch/arm/vfp/vfp.h
+ *
+ *  Copyright (C) 2004 ARM Limited.
+ *  Written by Deep Blue Solutions Limited.
+ *
+ * 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.
+ */
+
+static inline u32 vfp_shiftright32jamming(u32 val, unsigned int shift)
+{
+	if (shift) {
+		if (shift < 32)
+			val = val >> shift | ((val << (32 - shift)) != 0);
+		else
+			val = val != 0;
+	}
+	return val;
+}
+
+static inline u64 vfp_shiftright64jamming(u64 val, unsigned int shift)
+{
+	if (shift) {
+		if (shift < 64)
+			val = val >> shift | ((val << (64 - shift)) != 0);
+		else
+			val = val != 0;
+	}
+	return val;
+}
+
+static inline u32 vfp_hi64to32jamming(u64 val)
+{
+	u32 v;
+
+	asm(
+	"cmp	%Q1, #1		@ vfp_hi64to32jamming\n\t"
+	"movcc	%0, %R1\n\t"
+	"orrcs	%0, %R1, #1"
+	: "=r" (v) : "r" (val) : "cc");
+
+	return v;
+}
+
+static inline void add128(u64 *resh, u64 *resl, u64 nh, u64 nl, u64 mh, u64 ml)
+{
+	asm(	"adds	%Q0, %Q2, %Q4\n\t"
+		"adcs	%R0, %R2, %R4\n\t"
+		"adcs	%Q1, %Q3, %Q5\n\t"
+		"adc	%R1, %R3, %R5"
+	    : "=r" (nl), "=r" (nh)
+	    : "0" (nl), "1" (nh), "r" (ml), "r" (mh)
+	    : "cc");
+	*resh = nh;
+	*resl = nl;
+}
+
+static inline void sub128(u64 *resh, u64 *resl, u64 nh, u64 nl, u64 mh, u64 ml)
+{
+	asm(	"subs	%Q0, %Q2, %Q4\n\t"
+		"sbcs	%R0, %R2, %R4\n\t"
+		"sbcs	%Q1, %Q3, %Q5\n\t"
+		"sbc	%R1, %R3, %R5\n\t"
+	    : "=r" (nl), "=r" (nh)
+	    : "0" (nl), "1" (nh), "r" (ml), "r" (mh)
+	    : "cc");
+	*resh = nh;
+	*resl = nl;
+}
+
+static inline void mul64to128(u64 *resh, u64 *resl, u64 n, u64 m)
+{
+	u32 nh, nl, mh, ml;
+	u64 rh, rma, rmb, rl;
+
+	nl = n;
+	ml = m;
+	rl = (u64)nl * ml;
+
+	nh = n >> 32;
+	rma = (u64)nh * ml;
+
+	mh = m >> 32;
+	rmb = (u64)nl * mh;
+	rma += rmb;
+
+	rh = (u64)nh * mh;
+	rh += ((u64)(rma < rmb) << 32) + (rma >> 32);
+
+	rma <<= 32;
+	rl += rma;
+	rh += (rl < rma);
+
+	*resl = rl;
+	*resh = rh;
+}
+
+static inline void shift64left(u64 *resh, u64 *resl, u64 n)
+{
+	*resh = n >> 63;
+	*resl = n << 1;
+}
+
+static inline u64 vfp_hi64multiply64(u64 n, u64 m)
+{
+	u64 rh, rl;
+	mul64to128(&rh, &rl, n, m);
+	return rh | (rl != 0);
+}
+
+static inline u64 vfp_estimate_div128to64(u64 nh, u64 nl, u64 m)
+{
+	u64 mh, ml, remh, reml, termh, terml, z;
+
+	if (nh >= m)
+		return ~0ULL;
+	mh = m >> 32;
+	if (mh << 32 <= nh) {
+		z = 0xffffffff00000000ULL;
+	} else {
+		z = nh;
+		do_div(z, mh);
+		z <<= 32;
+	}
+	mul64to128(&termh, &terml, m, z);
+	sub128(&remh, &reml, nh, nl, termh, terml);
+	ml = m << 32;
+	while ((s64)remh < 0) {
+		z -= 0x100000000ULL;
+		add128(&remh, &reml, remh, reml, mh, ml);
+	}
+	remh = (remh << 32) | (reml >> 32);
+	if (mh << 32 <= remh) {
+		z |= 0xffffffff;
+	} else {
+		do_div(remh, mh);
+		z |= remh;
+	}
+	return z;
+}
+
+/*
+ * Operations on unpacked elements
+ */
+#define vfp_sign_negate(sign)	(sign ^ 0x8000)
+
+/*
+ * Single-precision
+ */
+struct vfp_single {
+	s16	exponent;
+	u16	sign;
+	u32	significand;
+};
+
+extern s32 vfp_get_float(unsigned int reg);
+extern void vfp_put_float(s32 val, unsigned int reg);
+
+/*
+ * VFP_SINGLE_MANTISSA_BITS - number of bits in the mantissa
+ * VFP_SINGLE_EXPONENT_BITS - number of bits in the exponent
+ * VFP_SINGLE_LOW_BITS - number of low bits in the unpacked significand
+ *  which are not propagated to the float upon packing.
+ */
+#define VFP_SINGLE_MANTISSA_BITS	(23)
+#define VFP_SINGLE_EXPONENT_BITS	(8)
+#define VFP_SINGLE_LOW_BITS		(32 - VFP_SINGLE_MANTISSA_BITS - 2)
+#define VFP_SINGLE_LOW_BITS_MASK	((1 << VFP_SINGLE_LOW_BITS) - 1)
+
+/*
+ * The bit in an unpacked float which indicates that it is a quiet NaN
+ */
+#define VFP_SINGLE_SIGNIFICAND_QNAN	(1 << (VFP_SINGLE_MANTISSA_BITS - 1 + VFP_SINGLE_LOW_BITS))
+
+/*
+ * Operations on packed single-precision numbers
+ */
+#define vfp_single_packed_sign(v)	((v) & 0x80000000)
+#define vfp_single_packed_negate(v)	((v) ^ 0x80000000)
+#define vfp_single_packed_abs(v)	((v) & ~0x80000000)
+#define vfp_single_packed_exponent(v)	(((v) >> VFP_SINGLE_MANTISSA_BITS) & ((1 << VFP_SINGLE_EXPONENT_BITS) - 1))
+#define vfp_single_packed_mantissa(v)	((v) & ((1 << VFP_SINGLE_MANTISSA_BITS) - 1))
+
+/*
+ * Unpack a single-precision float.  Note that this returns the magnitude
+ * of the single-precision float mantissa with the 1. if necessary,
+ * aligned to bit 30.
+ */
+static inline void vfp_single_unpack(struct vfp_single *s, s32 val)
+{
+	u32 significand;
+
+	s->sign = vfp_single_packed_sign(val) >> 16,
+	s->exponent = vfp_single_packed_exponent(val);
+
+	significand = (u32) val;
+	significand = (significand << (32 - VFP_SINGLE_MANTISSA_BITS)) >> 2;
+	if (s->exponent && s->exponent != 255)
+		significand |= 0x40000000;
+	s->significand = significand;
+}
+
+/*
+ * Re-pack a single-precision float.  This assumes that the float is
+ * already normalised such that the MSB is bit 30, _not_ bit 31.
+ */
+static inline s32 vfp_single_pack(struct vfp_single *s)
+{
+	u32 val;
+	val = (s->sign << 16) +
+	      (s->exponent << VFP_SINGLE_MANTISSA_BITS) +
+	      (s->significand >> VFP_SINGLE_LOW_BITS);
+	return (s32)val;
+}
+
+#define VFP_NUMBER		(1<<0)
+#define VFP_ZERO		(1<<1)
+#define VFP_DENORMAL		(1<<2)
+#define VFP_INFINITY		(1<<3)
+#define VFP_NAN			(1<<4)
+#define VFP_NAN_SIGNAL		(1<<5)
+
+#define VFP_QNAN		(VFP_NAN)
+#define VFP_SNAN		(VFP_NAN|VFP_NAN_SIGNAL)
+
+static inline int vfp_single_type(struct vfp_single *s)
+{
+	int type = VFP_NUMBER;
+	if (s->exponent == 255) {
+		if (s->significand == 0)
+			type = VFP_INFINITY;
+		else if (s->significand & VFP_SINGLE_SIGNIFICAND_QNAN)
+			type = VFP_QNAN;
+		else
+			type = VFP_SNAN;
+	} else if (s->exponent == 0) {
+		if (s->significand == 0)
+			type |= VFP_ZERO;
+		else
+			type |= VFP_DENORMAL;
+	}
+	return type;
+}
+
+#ifndef DEBUG
+#define vfp_single_normaliseround(sd,vsd,fpscr,except,func) __vfp_single_normaliseround(sd,vsd,fpscr,except)
+u32 __vfp_single_normaliseround(int sd, struct vfp_single *vs, u32 fpscr, u32 exceptions);
+#else
+u32 vfp_single_normaliseround(int sd, struct vfp_single *vs, u32 fpscr, u32 exceptions, const char *func);
+#endif
+
+/*
+ * Double-precision
+ */
+struct vfp_double {
+	s16	exponent;
+	u16	sign;
+	u64	significand;
+};
+
+/*
+ * VFP_REG_ZERO is a special register number for vfp_get_double
+ * which returns (double)0.0.  This is useful for the compare with
+ * zero instructions.
+ */
+#ifdef CONFIG_VFPv3
+#define VFP_REG_ZERO	32
+#else
+#define VFP_REG_ZERO	16
+#endif
+extern u64 vfp_get_double(unsigned int reg);
+extern void vfp_put_double(u64 val, unsigned int reg);
+
+#define VFP_DOUBLE_MANTISSA_BITS	(52)
+#define VFP_DOUBLE_EXPONENT_BITS	(11)
+#define VFP_DOUBLE_LOW_BITS		(64 - VFP_DOUBLE_MANTISSA_BITS - 2)
+#define VFP_DOUBLE_LOW_BITS_MASK	((1 << VFP_DOUBLE_LOW_BITS) - 1)
+
+/*
+ * The bit in an unpacked double which indicates that it is a quiet NaN
+ */
+#define VFP_DOUBLE_SIGNIFICAND_QNAN	(1ULL << (VFP_DOUBLE_MANTISSA_BITS - 1 + VFP_DOUBLE_LOW_BITS))
+
+/*
+ * Operations on packed single-precision numbers
+ */
+#define vfp_double_packed_sign(v)	((v) & (1ULL << 63))
+#define vfp_double_packed_negate(v)	((v) ^ (1ULL << 63))
+#define vfp_double_packed_abs(v)	((v) & ~(1ULL << 63))
+#define vfp_double_packed_exponent(v)	(((v) >> VFP_DOUBLE_MANTISSA_BITS) & ((1 << VFP_DOUBLE_EXPONENT_BITS) - 1))
+#define vfp_double_packed_mantissa(v)	((v) & ((1ULL << VFP_DOUBLE_MANTISSA_BITS) - 1))
+
+/*
+ * Unpack a double-precision float.  Note that this returns the magnitude
+ * of the double-precision float mantissa with the 1. if necessary,
+ * aligned to bit 62.
+ */
+static inline void vfp_double_unpack(struct vfp_double *s, s64 val)
+{
+	u64 significand;
+
+	s->sign = vfp_double_packed_sign(val) >> 48;
+	s->exponent = vfp_double_packed_exponent(val);
+
+	significand = (u64) val;
+	significand = (significand << (64 - VFP_DOUBLE_MANTISSA_BITS)) >> 2;
+	if (s->exponent && s->exponent != 2047)
+		significand |= (1ULL << 62);
+	s->significand = significand;
+}
+
+/*
+ * Re-pack a double-precision float.  This assumes that the float is
+ * already normalised such that the MSB is bit 30, _not_ bit 31.
+ */
+static inline s64 vfp_double_pack(struct vfp_double *s)
+{
+	u64 val;
+	val = ((u64)s->sign << 48) +
+	      ((u64)s->exponent << VFP_DOUBLE_MANTISSA_BITS) +
+	      (s->significand >> VFP_DOUBLE_LOW_BITS);
+	return (s64)val;
+}
+
+static inline int vfp_double_type(struct vfp_double *s)
+{
+	int type = VFP_NUMBER;
+	if (s->exponent == 2047) {
+		if (s->significand == 0)
+			type = VFP_INFINITY;
+		else if (s->significand & VFP_DOUBLE_SIGNIFICAND_QNAN)
+			type = VFP_QNAN;
+		else
+			type = VFP_SNAN;
+	} else if (s->exponent == 0) {
+		if (s->significand == 0)
+			type |= VFP_ZERO;
+		else
+			type |= VFP_DENORMAL;
+	}
+	return type;
+}
+
+u32 vfp_double_normaliseround(int dd, struct vfp_double *vd, u32 fpscr, u32 exceptions, const char *func);
+
+u32 vfp_estimate_sqrt_significand(u32 exponent, u32 significand);
+
+/*
+ * A special flag to tell the normalisation code not to normalise.
+ */
+#define VFP_NAN_FLAG	0x100
+
+/*
+ * A bit pattern used to indicate the initial (unset) value of the
+ * exception mask, in case nothing handles an instruction.  This
+ * doesn't include the NAN flag, which get masked out before
+ * we check for an error.
+ */
+#define VFP_EXCEPTION_ERROR	((u32)-1 & ~VFP_NAN_FLAG)
+
+/*
+ * A flag to tell vfp instruction type.
+ *  OP_SCALAR - this operation always operates in scalar mode
+ *  OP_SD - the instruction exceptionally writes to a single precision result.
+ *  OP_DD - the instruction exceptionally writes to a double precision result.
+ *  OP_SM - the instruction exceptionally reads from a single precision operand.
+ */
+#define OP_SCALAR	(1 << 0)
+#define OP_SD		(1 << 1)
+#define OP_DD		(1 << 1)
+#define OP_SM		(1 << 2)
+
+struct op {
+	u32 (* const fn)(int dd, int dn, int dm, u32 fpscr);
+	u32 flags;
+};
+
+extern void vfp_save_state(void *location, u32 fpexc);

arch/arm/vfp/vfphw.S

@@ -0,0 +1,317 @@
+/*
+ *  linux/arch/arm/vfp/vfphw.S
+ *
+ *  Copyright (C) 2004 ARM Limited.
+ *  Written by Deep Blue Solutions Limited.
+ *
+ * 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 code is called from the kernel's undefined instruction trap.
+ * r9 holds the return address for successful handling.
+ * lr holds the return address for unrecognised instructions.
+ * r10 points at the start of the private FP workspace in the thread structure
+ * sp points to a struct pt_regs (as defined in include/asm/proc/ptrace.h)
+ */
+#include <linux/init.h>
+#include <linux/linkage.h>
+#include <asm/thread_info.h>
+#include <asm/vfpmacros.h>
+#include <linux/kern_levels.h>
+#include <asm/assembler.h>
+#include <asm/asm-offsets.h>
+
+	.macro	DBGSTR, str
+#ifdef DEBUG
+	stmfd	sp!, {r0-r3, ip, lr}
+	ldr	r0, =1f
+	bl	printk
+	ldmfd	sp!, {r0-r3, ip, lr}
+
+	.pushsection .rodata, "a"
+1:	.ascii	KERN_DEBUG "VFP: \str\n"
+	.byte	0
+	.previous
+#endif
+	.endm
+
+	.macro  DBGSTR1, str, arg
+#ifdef DEBUG
+	stmfd	sp!, {r0-r3, ip, lr}
+	mov	r1, \arg
+	ldr	r0, =1f
+	bl	printk
+	ldmfd	sp!, {r0-r3, ip, lr}
+
+	.pushsection .rodata, "a"
+1:	.ascii	KERN_DEBUG "VFP: \str\n"
+	.byte	0
+	.previous
+#endif
+	.endm
+
+	.macro  DBGSTR3, str, arg1, arg2, arg3
+#ifdef DEBUG
+	stmfd	sp!, {r0-r3, ip, lr}
+	mov	r3, \arg3
+	mov	r2, \arg2
+	mov	r1, \arg1
+	ldr	r0, =1f
+	bl	printk
+	ldmfd	sp!, {r0-r3, ip, lr}
+
+	.pushsection .rodata, "a"
+1:	.ascii	KERN_DEBUG "VFP: \str\n"
+	.byte	0
+	.previous
+#endif
+	.endm
+
+
+@ VFP hardware support entry point.
+@
+@  r0  = instruction opcode (32-bit ARM or two 16-bit Thumb)
+@  r2  = PC value to resume execution after successful emulation
+@  r9  = normal "successful" return address
+@  r10 = vfp_state union
+@  r11 = CPU number
+@  lr  = unrecognised instruction return address
+@  IRQs enabled.
+ENTRY(vfp_support_entry)
+	DBGSTR3	"instr %08x pc %08x state %p", r0, r2, r10
+
+	ldr	r3, [sp, #S_PSR]	@ Neither lazy restore nor FP exceptions
+	and	r3, r3, #MODE_MASK	@ are supported in kernel mode
+	teq	r3, #USR_MODE
+	bne	vfp_kmode_exception	@ Returns through lr
+
+	VFPFMRX	r1, FPEXC		@ Is the VFP enabled?
+	DBGSTR1	"fpexc %08x", r1
+	tst	r1, #FPEXC_EN
+	bne	look_for_VFP_exceptions	@ VFP is already enabled
+
+	DBGSTR1 "enable %x", r10
+	ldr	r3, vfp_current_hw_state_address
+	orr	r1, r1, #FPEXC_EN	@ user FPEXC has the enable bit set
+	ldr	r4, [r3, r11, lsl #2]	@ vfp_current_hw_state pointer
+	bic	r5, r1, #FPEXC_EX	@ make sure exceptions are disabled
+	cmp	r4, r10			@ this thread owns the hw context?
+#ifndef CONFIG_SMP
+	@ For UP, checking that this thread owns the hw context is
+	@ sufficient to determine that the hardware state is valid.
+	beq	vfp_hw_state_valid
+
+	@ On UP, we lazily save the VFP context.  As a different
+	@ thread wants ownership of the VFP hardware, save the old
+	@ state if there was a previous (valid) owner.
+
+	VFPFMXR	FPEXC, r5		@ enable VFP, disable any pending
+					@ exceptions, so we can get at the
+					@ rest of it
+
+	DBGSTR1	"save old state %p", r4
+	cmp	r4, #0			@ if the vfp_current_hw_state is NULL
+	beq	vfp_reload_hw		@ then the hw state needs reloading
+	VFPFSTMIA r4, r5		@ save the working registers
+	VFPFMRX	r5, FPSCR		@ current status
+#ifndef CONFIG_CPU_FEROCEON
+	tst	r1, #FPEXC_EX		@ is there additional state to save?
+	beq	1f
+	VFPFMRX	r6, FPINST		@ FPINST (only if FPEXC.EX is set)
+	tst	r1, #FPEXC_FP2V		@ is there an FPINST2 to read?
+	beq	1f
+	VFPFMRX	r8, FPINST2		@ FPINST2 if needed (and present)
+1:
+#endif
+	stmia	r4, {r1, r5, r6, r8}	@ save FPEXC, FPSCR, FPINST, FPINST2
+vfp_reload_hw:
+
+#else
+	@ For SMP, if this thread does not own the hw context, then we
+	@ need to reload it.  No need to save the old state as on SMP,
+	@ we always save the state when we switch away from a thread.
+	bne	vfp_reload_hw
+
+	@ This thread has ownership of the current hardware context.
+	@ However, it may have been migrated to another CPU, in which
+	@ case the saved state is newer than the hardware context.
+	@ Check this by looking at the CPU number which the state was
+	@ last loaded onto.
+	ldr	ip, [r10, #VFP_CPU]
+	teq	ip, r11
+	beq	vfp_hw_state_valid
+
+vfp_reload_hw:
+	@ We're loading this threads state into the VFP hardware. Update
+	@ the CPU number which contains the most up to date VFP context.
+	str	r11, [r10, #VFP_CPU]
+
+	VFPFMXR	FPEXC, r5		@ enable VFP, disable any pending
+					@ exceptions, so we can get at the
+					@ rest of it
+#endif
+
+	DBGSTR1	"load state %p", r10
+	str	r10, [r3, r11, lsl #2]	@ update the vfp_current_hw_state pointer
+					@ Load the saved state back into the VFP
+	VFPFLDMIA r10, r5		@ reload the working registers while
+					@ FPEXC is in a safe state
+	ldmia	r10, {r1, r5, r6, r8}	@ load FPEXC, FPSCR, FPINST, FPINST2
+#ifndef CONFIG_CPU_FEROCEON
+	tst	r1, #FPEXC_EX		@ is there additional state to restore?
+	beq	1f
+	VFPFMXR	FPINST, r6		@ restore FPINST (only if FPEXC.EX is set)
+	tst	r1, #FPEXC_FP2V		@ is there an FPINST2 to write?
+	beq	1f
+	VFPFMXR	FPINST2, r8		@ FPINST2 if needed (and present)
+1:
+#endif
+	VFPFMXR	FPSCR, r5		@ restore status
+
+@ The context stored in the VFP hardware is up to date with this thread
+vfp_hw_state_valid:
+	tst	r1, #FPEXC_EX
+	bne	process_exception	@ might as well handle the pending
+					@ exception before retrying branch
+					@ out before setting an FPEXC that
+					@ stops us reading stuff
+	VFPFMXR	FPEXC, r1		@ Restore FPEXC last
+	sub	r2, r2, #4		@ Retry current instruction - if Thumb
+	str	r2, [sp, #S_PC]		@ mode it's two 16-bit instructions,
+					@ else it's one 32-bit instruction, so
+					@ always subtract 4 from the following
+					@ instruction address.
+	dec_preempt_count_ti r10, r4
+	ret	r9			@ we think we have handled things
+
+
+look_for_VFP_exceptions:
+	@ Check for synchronous or asynchronous exception
+	tst	r1, #FPEXC_EX | FPEXC_DEX
+	bne	process_exception
+	@ On some implementations of the VFP subarch 1, setting FPSCR.IXE
+	@ causes all the CDP instructions to be bounced synchronously without
+	@ setting the FPEXC.EX bit
+	VFPFMRX	r5, FPSCR
+	tst	r5, #FPSCR_IXE
+	bne	process_exception
+
+	@ Fall into hand on to next handler - appropriate coproc instr
+	@ not recognised by VFP
+
+	DBGSTR	"not VFP"
+	dec_preempt_count_ti r10, r4
+	ret	lr
+
+process_exception:
+	DBGSTR	"bounce"
+	mov	r2, sp			@ nothing stacked - regdump is at TOS
+	mov	lr, r9			@ setup for a return to the user code.
+
+	@ Now call the C code to package up the bounce to the support code
+	@   r0 holds the trigger instruction
+	@   r1 holds the FPEXC value
+	@   r2 pointer to register dump
+	b	VFP_bounce		@ we have handled this - the support
+					@ code will raise an exception if
+					@ required. If not, the user code will
+					@ retry the faulted instruction
+ENDPROC(vfp_support_entry)
+
+ENTRY(vfp_save_state)
+	@ Save the current VFP state
+	@ r0 - save location
+	@ r1 - FPEXC
+	DBGSTR1	"save VFP state %p", r0
+	VFPFSTMIA r0, r2		@ save the working registers
+	VFPFMRX	r2, FPSCR		@ current status
+	tst	r1, #FPEXC_EX		@ is there additional state to save?
+	beq	1f
+	VFPFMRX	r3, FPINST		@ FPINST (only if FPEXC.EX is set)
+	tst	r1, #FPEXC_FP2V		@ is there an FPINST2 to read?
+	beq	1f
+	VFPFMRX	r12, FPINST2		@ FPINST2 if needed (and present)
+1:
+	stmia	r0, {r1, r2, r3, r12}	@ save FPEXC, FPSCR, FPINST, FPINST2
+	ret	lr
+ENDPROC(vfp_save_state)
+
+	.align
+vfp_current_hw_state_address:
+	.word	vfp_current_hw_state
+
+	.macro	tbl_branch, base, tmp, shift
+#ifdef CONFIG_THUMB2_KERNEL
+	adr	\tmp, 1f
+	add	\tmp, \tmp, \base, lsl \shift
+	ret	\tmp
+#else
+	add	pc, pc, \base, lsl \shift
+	mov	r0, r0
+#endif
+1:
+	.endm
+
+ENTRY(vfp_get_float)
+	tbl_branch r0, r3, #3
+	.irp	dr,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15
+1:	mrc	p10, 0, r0, c\dr, c0, 0	@ fmrs	r0, s0
+	ret	lr
+	.org	1b + 8
+1:	mrc	p10, 0, r0, c\dr, c0, 4	@ fmrs	r0, s1
+	ret	lr
+	.org	1b + 8
+	.endr
+ENDPROC(vfp_get_float)
+
+ENTRY(vfp_put_float)
+	tbl_branch r1, r3, #3
+	.irp	dr,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15
+1:	mcr	p10, 0, r0, c\dr, c0, 0	@ fmsr	r0, s0
+	ret	lr
+	.org	1b + 8
+1:	mcr	p10, 0, r0, c\dr, c0, 4	@ fmsr	r0, s1
+	ret	lr
+	.org	1b + 8
+	.endr
+ENDPROC(vfp_put_float)
+
+ENTRY(vfp_get_double)
+	tbl_branch r0, r3, #3
+	.irp	dr,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15
+1:	fmrrd	r0, r1, d\dr
+	ret	lr
+	.org	1b + 8
+	.endr
+#ifdef CONFIG_VFPv3
+	@ d16 - d31 registers
+	.irp	dr,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15
+1:	mrrc	p11, 3, r0, r1, c\dr	@ fmrrd	r0, r1, d\dr
+	ret	lr
+	.org	1b + 8
+	.endr
+#endif
+
+	@ virtual register 16 (or 32 if VFPv3) for compare with zero
+	mov	r0, #0
+	mov	r1, #0
+	ret	lr
+ENDPROC(vfp_get_double)
+
+ENTRY(vfp_put_double)
+	tbl_branch r2, r3, #3
+	.irp	dr,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15
+1:	fmdrr	d\dr, r0, r1
+	ret	lr
+	.org	1b + 8
+	.endr
+#ifdef CONFIG_VFPv3
+	@ d16 - d31 registers
+	.irp	dr,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15
+1:	mcrr	p11, 3, r0, r1, c\dr	@ fmdrr	r0, r1, d\dr
+	ret	lr
+	.org	1b + 8
+	.endr
+#endif
+ENDPROC(vfp_put_double)

arch/arm/vfp/vfpinstr.h

@@ -0,0 +1,88 @@
+/*
+ *  linux/arch/arm/vfp/vfpinstr.h
+ *
+ *  Copyright (C) 2004 ARM Limited.
+ *  Written by Deep Blue Solutions Limited.
+ *
+ * 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.
+ *
+ * VFP instruction masks.
+ */
+#define INST_CPRTDO(inst)	(((inst) & 0x0f000000) == 0x0e000000)
+#define INST_CPRT(inst)		((inst) & (1 << 4))
+#define INST_CPRT_L(inst)	((inst) & (1 << 20))
+#define INST_CPRT_Rd(inst)	(((inst) & (15 << 12)) >> 12)
+#define INST_CPRT_OP(inst)	(((inst) >> 21) & 7)
+#define INST_CPNUM(inst)	((inst) & 0xf00)
+#define CPNUM(cp)		((cp) << 8)
+
+#define FOP_MASK	(0x00b00040)
+#define FOP_FMAC	(0x00000000)
+#define FOP_FNMAC	(0x00000040)
+#define FOP_FMSC	(0x00100000)
+#define FOP_FNMSC	(0x00100040)
+#define FOP_FMUL	(0x00200000)
+#define FOP_FNMUL	(0x00200040)
+#define FOP_FADD	(0x00300000)
+#define FOP_FSUB	(0x00300040)
+#define FOP_FDIV	(0x00800000)
+#define FOP_EXT		(0x00b00040)
+
+#define FOP_TO_IDX(inst)	((inst & 0x00b00000) >> 20 | (inst & (1 << 6)) >> 4)
+
+#define FEXT_MASK	(0x000f0080)
+#define FEXT_FCPY	(0x00000000)
+#define FEXT_FABS	(0x00000080)
+#define FEXT_FNEG	(0x00010000)
+#define FEXT_FSQRT	(0x00010080)
+#define FEXT_FCMP	(0x00040000)
+#define FEXT_FCMPE	(0x00040080)
+#define FEXT_FCMPZ	(0x00050000)
+#define FEXT_FCMPEZ	(0x00050080)
+#define FEXT_FCVT	(0x00070080)
+#define FEXT_FUITO	(0x00080000)
+#define FEXT_FSITO	(0x00080080)
+#define FEXT_FTOUI	(0x000c0000)
+#define FEXT_FTOUIZ	(0x000c0080)
+#define FEXT_FTOSI	(0x000d0000)
+#define FEXT_FTOSIZ	(0x000d0080)
+
+#define FEXT_TO_IDX(inst)	((inst & 0x000f0000) >> 15 | (inst & (1 << 7)) >> 7)
+
+#define vfp_get_sd(inst)	((inst & 0x0000f000) >> 11 | (inst & (1 << 22)) >> 22)
+#define vfp_get_dd(inst)	((inst & 0x0000f000) >> 12 | (inst & (1 << 22)) >> 18)
+#define vfp_get_sm(inst)	((inst & 0x0000000f) << 1 | (inst & (1 << 5)) >> 5)
+#define vfp_get_dm(inst)	((inst & 0x0000000f) | (inst & (1 << 5)) >> 1)
+#define vfp_get_sn(inst)	((inst & 0x000f0000) >> 15 | (inst & (1 << 7)) >> 7)
+#define vfp_get_dn(inst)	((inst & 0x000f0000) >> 16 | (inst & (1 << 7)) >> 3)
+
+#define vfp_single(inst)	(((inst) & 0x0000f00) == 0xa00)
+
+#define FPSCR_N	(1 << 31)
+#define FPSCR_Z	(1 << 30)
+#define FPSCR_C (1 << 29)
+#define FPSCR_V	(1 << 28)
+
+/*
+ * Since we aren't building with -mfpu=vfp, we need to code
+ * these instructions using their MRC/MCR equivalents.
+ */
+#define vfpreg(_vfp_) #_vfp_
+
+#define fmrx(_vfp_) ({			\
+	u32 __v;			\
+	asm("mrc p10, 7, %0, " vfpreg(_vfp_) ", cr0, 0 @ fmrx	%0, " #_vfp_	\
+	    : "=r" (__v) : : "cc");	\
+	__v;				\
+ })
+
+#define fmxr(_vfp_,_var_)		\
+	asm("mcr p10, 7, %0, " vfpreg(_vfp_) ", cr0, 0 @ fmxr	" #_vfp_ ", %0"	\
+	   : : "r" (_var_) : "cc")
+
+u32 vfp_single_cpdo(u32 inst, u32 fpscr);
+u32 vfp_single_cprt(u32 inst, u32 fpscr, struct pt_regs *regs);
+
+u32 vfp_double_cpdo(u32 inst, u32 fpscr);

arch/arm/vfp/vfpmodule.c

@@ -0,0 +1,801 @@
+/*
+ *  linux/arch/arm/vfp/vfpmodule.c
+ *
+ *  Copyright (C) 2004 ARM Limited.
+ *  Written by Deep Blue Solutions Limited.
+ *
+ * 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/types.h>
+#include <linux/cpu.h>
+#include <linux/cpu_pm.h>
+#include <linux/hardirq.h>
+#include <linux/kernel.h>
+#include <linux/notifier.h>
+#include <linux/signal.h>
+#include <linux/sched.h>
+#include <linux/smp.h>
+#include <linux/init.h>
+#include <linux/uaccess.h>
+#include <linux/user.h>
+#include <linux/export.h>
+
+#include <asm/cp15.h>
+#include <asm/cputype.h>
+#include <asm/system_info.h>
+#include <asm/thread_notify.h>
+#include <asm/vfp.h>
+
+#include "vfpinstr.h"
+#include "vfp.h"
+
+/*
+ * Our undef handlers (in entry.S)
+ */
+void vfp_testing_entry(void);
+void vfp_support_entry(void);
+void vfp_null_entry(void);
+
+void (*vfp_vector)(void) = vfp_null_entry;
+
+/*
+ * Dual-use variable.
+ * Used in startup: set to non-zero if VFP checks fail
+ * After startup, holds VFP architecture
+ */
+unsigned int VFP_arch;
+
+/*
+ * The pointer to the vfpstate structure of the thread which currently
+ * owns the context held in the VFP hardware, or NULL if the hardware
+ * context is invalid.
+ *
+ * For UP, this is sufficient to tell which thread owns the VFP context.
+ * However, for SMP, we also need to check the CPU number stored in the
+ * saved state too to catch migrations.
+ */
+union vfp_state *vfp_current_hw_state[NR_CPUS];
+
+/*
+ * Is 'thread's most up to date state stored in this CPUs hardware?
+ * Must be called from non-preemptible context.
+ */
+static bool vfp_state_in_hw(unsigned int cpu, struct thread_info *thread)
+{
+#ifdef CONFIG_SMP
+	if (thread->vfpstate.hard.cpu != cpu)
+		return false;
+#endif
+	return vfp_current_hw_state[cpu] == &thread->vfpstate;
+}
+
+/*
+ * Force a reload of the VFP context from the thread structure.  We do
+ * this by ensuring that access to the VFP hardware is disabled, and
+ * clear vfp_current_hw_state.  Must be called from non-preemptible context.
+ */
+static void vfp_force_reload(unsigned int cpu, struct thread_info *thread)
+{
+	if (vfp_state_in_hw(cpu, thread)) {
+		fmxr(FPEXC, fmrx(FPEXC) & ~FPEXC_EN);
+		vfp_current_hw_state[cpu] = NULL;
+	}
+#ifdef CONFIG_SMP
+	thread->vfpstate.hard.cpu = NR_CPUS;
+#endif
+}
+
+/*
+ * Per-thread VFP initialization.
+ */
+static void vfp_thread_flush(struct thread_info *thread)
+{
+	union vfp_state *vfp = &thread->vfpstate;
+	unsigned int cpu;
+
+	/*
+	 * Disable VFP to ensure we initialize it first.  We must ensure
+	 * that the modification of vfp_current_hw_state[] and hardware
+	 * disable are done for the same CPU and without preemption.
+	 *
+	 * Do this first to ensure that preemption won't overwrite our
+	 * state saving should access to the VFP be enabled at this point.
+	 */
+	cpu = get_cpu();
+	if (vfp_current_hw_state[cpu] == vfp)
+		vfp_current_hw_state[cpu] = NULL;
+	fmxr(FPEXC, fmrx(FPEXC) & ~FPEXC_EN);
+	put_cpu();
+
+	memset(vfp, 0, sizeof(union vfp_state));
+
+	vfp->hard.fpexc = FPEXC_EN;
+	vfp->hard.fpscr = FPSCR_ROUND_NEAREST;
+#ifdef CONFIG_SMP
+	vfp->hard.cpu = NR_CPUS;
+#endif
+}
+
+static void vfp_thread_exit(struct thread_info *thread)
+{
+	/* release case: Per-thread VFP cleanup. */
+	union vfp_state *vfp = &thread->vfpstate;
+	unsigned int cpu = get_cpu();
+
+	if (vfp_current_hw_state[cpu] == vfp)
+		vfp_current_hw_state[cpu] = NULL;
+	put_cpu();
+}
+
+static void vfp_thread_copy(struct thread_info *thread)
+{
+	struct thread_info *parent = current_thread_info();
+
+	vfp_sync_hwstate(parent);
+	thread->vfpstate = parent->vfpstate;
+#ifdef CONFIG_SMP
+	thread->vfpstate.hard.cpu = NR_CPUS;
+#endif
+}
+
+/*
+ * When this function is called with the following 'cmd's, the following
+ * is true while this function is being run:
+ *  THREAD_NOFTIFY_SWTICH:
+ *   - the previously running thread will not be scheduled onto another CPU.
+ *   - the next thread to be run (v) will not be running on another CPU.
+ *   - thread->cpu is the local CPU number
+ *   - not preemptible as we're called in the middle of a thread switch
+ *  THREAD_NOTIFY_FLUSH:
+ *   - the thread (v) will be running on the local CPU, so
+ *	v === current_thread_info()
+ *   - thread->cpu is the local CPU number at the time it is accessed,
+ *	but may change at any time.
+ *   - we could be preempted if tree preempt rcu is enabled, so
+ *	it is unsafe to use thread->cpu.
+ *  THREAD_NOTIFY_EXIT
+ *   - the thread (v) will be running on the local CPU, so
+ *	v === current_thread_info()
+ *   - thread->cpu is the local CPU number at the time it is accessed,
+ *	but may change at any time.
+ *   - we could be preempted if tree preempt rcu is enabled, so
+ *	it is unsafe to use thread->cpu.
+ */
+static int vfp_notifier(struct notifier_block *self, unsigned long cmd, void *v)
+{
+	struct thread_info *thread = v;
+	u32 fpexc;
+#ifdef CONFIG_SMP
+	unsigned int cpu;
+#endif
+
+	switch (cmd) {
+	case THREAD_NOTIFY_SWITCH:
+		fpexc = fmrx(FPEXC);
+
+#ifdef CONFIG_SMP
+		cpu = thread->cpu;
+
+		/*
+		 * On SMP, if VFP is enabled, save the old state in
+		 * case the thread migrates to a different CPU. The
+		 * restoring is done lazily.
+		 */
+		if ((fpexc & FPEXC_EN) && vfp_current_hw_state[cpu])
+			vfp_save_state(vfp_current_hw_state[cpu], fpexc);
+#endif
+
+		/*
+		 * Always disable VFP so we can lazily save/restore the
+		 * old state.
+		 */
+		fmxr(FPEXC, fpexc & ~FPEXC_EN);
+		break;
+
+	case THREAD_NOTIFY_FLUSH:
+		vfp_thread_flush(thread);
+		break;
+
+	case THREAD_NOTIFY_EXIT:
+		vfp_thread_exit(thread);
+		break;
+
+	case THREAD_NOTIFY_COPY:
+		vfp_thread_copy(thread);
+		break;
+	}
+
+	return NOTIFY_DONE;
+}
+
+static struct notifier_block vfp_notifier_block = {
+	.notifier_call	= vfp_notifier,
+};
+
+/*
+ * Raise a SIGFPE for the current process.
+ * sicode describes the signal being raised.
+ */
+static void vfp_raise_sigfpe(unsigned int sicode, struct pt_regs *regs)
+{
+	siginfo_t info;
+
+	memset(&info, 0, sizeof(info));
+
+	info.si_signo = SIGFPE;
+	info.si_code = sicode;
+	info.si_addr = (void __user *)(instruction_pointer(regs) - 4);
+
+	/*
+	 * This is the same as NWFPE, because it's not clear what
+	 * this is used for
+	 */
+	current->thread.error_code = 0;
+	current->thread.trap_no = 6;
+
+	send_sig_info(SIGFPE, &info, current);
+}
+
+static void vfp_panic(char *reason, u32 inst)
+{
+	int i;
+
+	pr_err("VFP: Error: %s\n", reason);
+	pr_err("VFP: EXC 0x%08x SCR 0x%08x INST 0x%08x\n",
+		fmrx(FPEXC), fmrx(FPSCR), inst);
+	for (i = 0; i < 32; i += 2)
+		pr_err("VFP: s%2u: 0x%08x s%2u: 0x%08x\n",
+		       i, vfp_get_float(i), i+1, vfp_get_float(i+1));
+}
+
+/*
+ * Process bitmask of exception conditions.
+ */
+static void vfp_raise_exceptions(u32 exceptions, u32 inst, u32 fpscr, struct pt_regs *regs)
+{
+	int si_code = 0;
+
+	pr_debug("VFP: raising exceptions %08x\n", exceptions);
+
+	if (exceptions == VFP_EXCEPTION_ERROR) {
+		vfp_panic("unhandled bounce", inst);
+		vfp_raise_sigfpe(0, regs);
+		return;
+	}
+
+	/*
+	 * If any of the status flags are set, update the FPSCR.
+	 * Comparison instructions always return at least one of
+	 * these flags set.
+	 */
+	if (exceptions & (FPSCR_N|FPSCR_Z|FPSCR_C|FPSCR_V))
+		fpscr &= ~(FPSCR_N|FPSCR_Z|FPSCR_C|FPSCR_V);
+
+	fpscr |= exceptions;
+
+	fmxr(FPSCR, fpscr);
+
+#define RAISE(stat,en,sig)				\
+	if (exceptions & stat && fpscr & en)		\
+		si_code = sig;
+
+	/*
+	 * These are arranged in priority order, least to highest.
+	 */
+	RAISE(FPSCR_DZC, FPSCR_DZE, FPE_FLTDIV);
+	RAISE(FPSCR_IXC, FPSCR_IXE, FPE_FLTRES);
+	RAISE(FPSCR_UFC, FPSCR_UFE, FPE_FLTUND);
+	RAISE(FPSCR_OFC, FPSCR_OFE, FPE_FLTOVF);
+	RAISE(FPSCR_IOC, FPSCR_IOE, FPE_FLTINV);
+
+	if (si_code)
+		vfp_raise_sigfpe(si_code, regs);
+}
+
+/*
+ * Emulate a VFP instruction.
+ */
+static u32 vfp_emulate_instruction(u32 inst, u32 fpscr, struct pt_regs *regs)
+{
+	u32 exceptions = VFP_EXCEPTION_ERROR;
+
+	pr_debug("VFP: emulate: INST=0x%08x SCR=0x%08x\n", inst, fpscr);
+
+	if (INST_CPRTDO(inst)) {
+		if (!INST_CPRT(inst)) {
+			/*
+			 * CPDO
+			 */
+			if (vfp_single(inst)) {
+				exceptions = vfp_single_cpdo(inst, fpscr);
+			} else {
+				exceptions = vfp_double_cpdo(inst, fpscr);
+			}
+		} else {
+			/*
+			 * A CPRT instruction can not appear in FPINST2, nor
+			 * can it cause an exception.  Therefore, we do not
+			 * have to emulate it.
+			 */
+		}
+	} else {
+		/*
+		 * A CPDT instruction can not appear in FPINST2, nor can
+		 * it cause an exception.  Therefore, we do not have to
+		 * emulate it.
+		 */
+	}
+	return exceptions & ~VFP_NAN_FLAG;
+}
+
+/*
+ * Package up a bounce condition.
+ */
+void VFP_bounce(u32 trigger, u32 fpexc, struct pt_regs *regs)
+{
+	u32 fpscr, orig_fpscr, fpsid, exceptions;
+
+	pr_debug("VFP: bounce: trigger %08x fpexc %08x\n", trigger, fpexc);
+
+	/*
+	 * At this point, FPEXC can have the following configuration:
+	 *
+	 *  EX DEX IXE
+	 *  0   1   x   - synchronous exception
+	 *  1   x   0   - asynchronous exception
+	 *  1   x   1   - sychronous on VFP subarch 1 and asynchronous on later
+	 *  0   0   1   - synchronous on VFP9 (non-standard subarch 1
+	 *                implementation), undefined otherwise
+	 *
+	 * Clear various bits and enable access to the VFP so we can
+	 * handle the bounce.
+	 */
+	fmxr(FPEXC, fpexc & ~(FPEXC_EX|FPEXC_DEX|FPEXC_FP2V|FPEXC_VV|FPEXC_TRAP_MASK));
+
+	fpsid = fmrx(FPSID);
+	orig_fpscr = fpscr = fmrx(FPSCR);
+
+	/*
+	 * Check for the special VFP subarch 1 and FPSCR.IXE bit case
+	 */
+	if ((fpsid & FPSID_ARCH_MASK) == (1 << FPSID_ARCH_BIT)
+	    && (fpscr & FPSCR_IXE)) {
+		/*
+		 * Synchronous exception, emulate the trigger instruction
+		 */
+		goto emulate;
+	}
+
+	if (fpexc & FPEXC_EX) {
+#ifndef CONFIG_CPU_FEROCEON
+		/*
+		 * Asynchronous exception. The instruction is read from FPINST
+		 * and the interrupted instruction has to be restarted.
+		 */
+		trigger = fmrx(FPINST);
+		regs->ARM_pc -= 4;
+#endif
+	} else if (!(fpexc & FPEXC_DEX)) {
+		/*
+		 * Illegal combination of bits. It can be caused by an
+		 * unallocated VFP instruction but with FPSCR.IXE set and not
+		 * on VFP subarch 1.
+		 */
+		 vfp_raise_exceptions(VFP_EXCEPTION_ERROR, trigger, fpscr, regs);
+		goto exit;
+	}
+
+	/*
+	 * Modify fpscr to indicate the number of iterations remaining.
+	 * If FPEXC.EX is 0, FPEXC.DEX is 1 and the FPEXC.VV bit indicates
+	 * whether FPEXC.VECITR or FPSCR.LEN is used.
+	 */
+	if (fpexc & (FPEXC_EX | FPEXC_VV)) {
+		u32 len;
+
+		len = fpexc + (1 << FPEXC_LENGTH_BIT);
+
+		fpscr &= ~FPSCR_LENGTH_MASK;
+		fpscr |= (len & FPEXC_LENGTH_MASK) << (FPSCR_LENGTH_BIT - FPEXC_LENGTH_BIT);
+	}
+
+	/*
+	 * Handle the first FP instruction.  We used to take note of the
+	 * FPEXC bounce reason, but this appears to be unreliable.
+	 * Emulate the bounced instruction instead.
+	 */
+	exceptions = vfp_emulate_instruction(trigger, fpscr, regs);
+	if (exceptions)
+		vfp_raise_exceptions(exceptions, trigger, orig_fpscr, regs);
+
+	/*
+	 * If there isn't a second FP instruction, exit now. Note that
+	 * the FPEXC.FP2V bit is valid only if FPEXC.EX is 1.
+	 */
+	if ((fpexc & (FPEXC_EX | FPEXC_FP2V)) != (FPEXC_EX | FPEXC_FP2V))
+		goto exit;
+
+	/*
+	 * The barrier() here prevents fpinst2 being read
+	 * before the condition above.
+	 */
+	barrier();
+	trigger = fmrx(FPINST2);
+
+ emulate:
+	exceptions = vfp_emulate_instruction(trigger, orig_fpscr, regs);
+	if (exceptions)
+		vfp_raise_exceptions(exceptions, trigger, orig_fpscr, regs);
+ exit:
+	preempt_enable();
+}
+
+static void vfp_enable(void *unused)
+{
+	u32 access;
+
+	BUG_ON(preemptible());
+	access = get_copro_access();
+
+	/*
+	 * Enable full access to VFP (cp10 and cp11)
+	 */
+	set_copro_access(access | CPACC_FULL(10) | CPACC_FULL(11));
+}
+
+#ifdef CONFIG_CPU_PM
+static int vfp_pm_suspend(void)
+{
+	struct thread_info *ti = current_thread_info();
+	u32 fpexc = fmrx(FPEXC);
+
+	/* if vfp is on, then save state for resumption */
+	if (fpexc & FPEXC_EN) {
+		pr_debug("%s: saving vfp state\n", __func__);
+		vfp_save_state(&ti->vfpstate, fpexc);
+
+		/* disable, just in case */
+		fmxr(FPEXC, fmrx(FPEXC) & ~FPEXC_EN);
+	} else if (vfp_current_hw_state[ti->cpu]) {
+#ifndef CONFIG_SMP
+		fmxr(FPEXC, fpexc | FPEXC_EN);
+		vfp_save_state(vfp_current_hw_state[ti->cpu], fpexc);
+		fmxr(FPEXC, fpexc);
+#endif
+	}
+
+	/* clear any information we had about last context state */
+	vfp_current_hw_state[ti->cpu] = NULL;
+
+	return 0;
+}
+
+static void vfp_pm_resume(void)
+{
+	/* ensure we have access to the vfp */
+	vfp_enable(NULL);
+
+	/* and disable it to ensure the next usage restores the state */
+	fmxr(FPEXC, fmrx(FPEXC) & ~FPEXC_EN);
+}
+
+static int vfp_cpu_pm_notifier(struct notifier_block *self, unsigned long cmd,
+	void *v)
+{
+	switch (cmd) {
+	case CPU_PM_ENTER:
+		vfp_pm_suspend();
+		break;
+	case CPU_PM_ENTER_FAILED:
+	case CPU_PM_EXIT:
+		vfp_pm_resume();
+		break;
+	}
+	return NOTIFY_OK;
+}
+
+static struct notifier_block vfp_cpu_pm_notifier_block = {
+	.notifier_call = vfp_cpu_pm_notifier,
+};
+
+static void vfp_pm_init(void)
+{
+	cpu_pm_register_notifier(&vfp_cpu_pm_notifier_block);
+}
+
+#else
+static inline void vfp_pm_init(void) { }
+#endif /* CONFIG_CPU_PM */
+
+/*
+ * Ensure that the VFP state stored in 'thread->vfpstate' is up to date
+ * with the hardware state.
+ */
+void vfp_sync_hwstate(struct thread_info *thread)
+{
+	unsigned int cpu = get_cpu();
+
+	if (vfp_state_in_hw(cpu, thread)) {
+		u32 fpexc = fmrx(FPEXC);
+
+		/*
+		 * Save the last VFP state on this CPU.
+		 */
+		fmxr(FPEXC, fpexc | FPEXC_EN);
+		vfp_save_state(&thread->vfpstate, fpexc | FPEXC_EN);
+		fmxr(FPEXC, fpexc);
+	}
+
+	put_cpu();
+}
+
+/* Ensure that the thread reloads the hardware VFP state on the next use. */
+void vfp_flush_hwstate(struct thread_info *thread)
+{
+	unsigned int cpu = get_cpu();
+
+	vfp_force_reload(cpu, thread);
+
+	put_cpu();
+}
+
+/*
+ * Save the current VFP state into the provided structures and prepare
+ * for entry into a new function (signal handler).
+ */
+int vfp_preserve_user_clear_hwstate(struct user_vfp __user *ufp,
+				    struct user_vfp_exc __user *ufp_exc)
+{
+	struct thread_info *thread = current_thread_info();
+	struct vfp_hard_struct *hwstate = &thread->vfpstate.hard;
+	int err = 0;
+
+	/* Ensure that the saved hwstate is up-to-date. */
+	vfp_sync_hwstate(thread);
+
+	/*
+	 * Copy the floating point registers. There can be unused
+	 * registers see asm/hwcap.h for details.
+	 */
+	err |= __copy_to_user(&ufp->fpregs, &hwstate->fpregs,
+			      sizeof(hwstate->fpregs));
+	/*
+	 * Copy the status and control register.
+	 */
+	__put_user_error(hwstate->fpscr, &ufp->fpscr, err);
+
+	/*
+	 * Copy the exception registers.
+	 */
+	__put_user_error(hwstate->fpexc, &ufp_exc->fpexc, err);
+	__put_user_error(hwstate->fpinst, &ufp_exc->fpinst, err);
+	__put_user_error(hwstate->fpinst2, &ufp_exc->fpinst2, err);
+
+	if (err)
+		return -EFAULT;
+
+	/* Ensure that VFP is disabled. */
+	vfp_flush_hwstate(thread);
+
+	/*
+	 * As per the PCS, clear the length and stride bits for function
+	 * entry.
+	 */
+	hwstate->fpscr &= ~(FPSCR_LENGTH_MASK | FPSCR_STRIDE_MASK);
+	return 0;
+}
+
+/* Sanitise and restore the current VFP state from the provided structures. */
+int vfp_restore_user_hwstate(struct user_vfp __user *ufp,
+			     struct user_vfp_exc __user *ufp_exc)
+{
+	struct thread_info *thread = current_thread_info();
+	struct vfp_hard_struct *hwstate = &thread->vfpstate.hard;
+	unsigned long fpexc;
+	int err = 0;
+
+	/* Disable VFP to avoid corrupting the new thread state. */
+	vfp_flush_hwstate(thread);
+
+	/*
+	 * Copy the floating point registers. There can be unused
+	 * registers see asm/hwcap.h for details.
+	 */
+	err |= __copy_from_user(&hwstate->fpregs, &ufp->fpregs,
+				sizeof(hwstate->fpregs));
+	/*
+	 * Copy the status and control register.
+	 */
+	__get_user_error(hwstate->fpscr, &ufp->fpscr, err);
+
+	/*
+	 * Sanitise and restore the exception registers.
+	 */
+	__get_user_error(fpexc, &ufp_exc->fpexc, err);
+
+	/* Ensure the VFP is enabled. */
+	fpexc |= FPEXC_EN;
+
+	/* Ensure FPINST2 is invalid and the exception flag is cleared. */
+	fpexc &= ~(FPEXC_EX | FPEXC_FP2V);
+	hwstate->fpexc = fpexc;
+
+	__get_user_error(hwstate->fpinst, &ufp_exc->fpinst, err);
+	__get_user_error(hwstate->fpinst2, &ufp_exc->fpinst2, err);
+
+	return err ? -EFAULT : 0;
+}
+
+/*
+ * VFP hardware can lose all context when a CPU goes offline.
+ * As we will be running in SMP mode with CPU hotplug, we will save the
+ * hardware state at every thread switch.  We clear our held state when
+ * a CPU has been killed, indicating that the VFP hardware doesn't contain
+ * a threads VFP state.  When a CPU starts up, we re-enable access to the
+ * VFP hardware.
+ *
+ * Both CPU_DYING and CPU_STARTING are called on the CPU which
+ * is being offlined/onlined.
+ */
+static int vfp_hotplug(struct notifier_block *b, unsigned long action,
+	void *hcpu)
+{
+	if (action == CPU_DYING || action == CPU_DYING_FROZEN)
+		vfp_current_hw_state[(long)hcpu] = NULL;
+	else if (action == CPU_STARTING || action == CPU_STARTING_FROZEN)
+		vfp_enable(NULL);
+	return NOTIFY_OK;
+}
+
+void vfp_kmode_exception(void)
+{
+	/*
+	 * If we reach this point, a floating point exception has been raised
+	 * while running in kernel mode. If the NEON/VFP unit was enabled at the
+	 * time, it means a VFP instruction has been issued that requires
+	 * software assistance to complete, something which is not currently
+	 * supported in kernel mode.
+	 * If the NEON/VFP unit was disabled, and the location pointed to below
+	 * is properly preceded by a call to kernel_neon_begin(), something has
+	 * caused the task to be scheduled out and back in again. In this case,
+	 * rebuilding and running with CONFIG_DEBUG_ATOMIC_SLEEP enabled should
+	 * be helpful in localizing the problem.
+	 */
+	if (fmrx(FPEXC) & FPEXC_EN)
+		pr_crit("BUG: unsupported FP instruction in kernel mode\n");
+	else
+		pr_crit("BUG: FP instruction issued in kernel mode with FP unit disabled\n");
+}
+
+#ifdef CONFIG_KERNEL_MODE_NEON
+
+/*
+ * Kernel-side NEON support functions
+ */
+void kernel_neon_begin(void)
+{
+	struct thread_info *thread = current_thread_info();
+	unsigned int cpu;
+	u32 fpexc;
+
+	/*
+	 * Kernel mode NEON is only allowed outside of interrupt context
+	 * with preemption disabled. This will make sure that the kernel
+	 * mode NEON register contents never need to be preserved.
+	 */
+	BUG_ON(in_interrupt());
+	cpu = get_cpu();
+
+	fpexc = fmrx(FPEXC) | FPEXC_EN;
+	fmxr(FPEXC, fpexc);
+
+	/*
+	 * Save the userland NEON/VFP state. Under UP,
+	 * the owner could be a task other than 'current'
+	 */
+	if (vfp_state_in_hw(cpu, thread))
+		vfp_save_state(&thread->vfpstate, fpexc);
+#ifndef CONFIG_SMP
+	else if (vfp_current_hw_state[cpu] != NULL)
+		vfp_save_state(vfp_current_hw_state[cpu], fpexc);
+#endif
+	vfp_current_hw_state[cpu] = NULL;
+}
+EXPORT_SYMBOL(kernel_neon_begin);
+
+void kernel_neon_end(void)
+{
+	/* Disable the NEON/VFP unit. */
+	fmxr(FPEXC, fmrx(FPEXC) & ~FPEXC_EN);
+	put_cpu();
+}
+EXPORT_SYMBOL(kernel_neon_end);
+
+#endif /* CONFIG_KERNEL_MODE_NEON */
+
+/*
+ * VFP support code initialisation.
+ */
+static int __init vfp_init(void)
+{
+	unsigned int vfpsid;
+	unsigned int cpu_arch = cpu_architecture();
+
+	if (cpu_arch >= CPU_ARCH_ARMv6)
+		on_each_cpu(vfp_enable, NULL, 1);
+
+	/*
+	 * First check that there is a VFP that we can use.
+	 * The handler is already setup to just log calls, so
+	 * we just need to read the VFPSID register.
+	 */
+	vfp_vector = vfp_testing_entry;
+	barrier();
+	vfpsid = fmrx(FPSID);
+	barrier();
+	vfp_vector = vfp_null_entry;
+
+	pr_info("VFP support v0.3: ");
+	if (VFP_arch)
+		pr_cont("not present\n");
+	else if (vfpsid & FPSID_NODOUBLE) {
+		pr_cont("no double precision support\n");
+	} else {
+		hotcpu_notifier(vfp_hotplug, 0);
+
+		VFP_arch = (vfpsid & FPSID_ARCH_MASK) >> FPSID_ARCH_BIT;  /* Extract the architecture version */
+		pr_cont("implementor %02x architecture %d part %02x variant %x rev %x\n",
+			(vfpsid & FPSID_IMPLEMENTER_MASK) >> FPSID_IMPLEMENTER_BIT,
+			(vfpsid & FPSID_ARCH_MASK) >> FPSID_ARCH_BIT,
+			(vfpsid & FPSID_PART_MASK) >> FPSID_PART_BIT,
+			(vfpsid & FPSID_VARIANT_MASK) >> FPSID_VARIANT_BIT,
+			(vfpsid & FPSID_REV_MASK) >> FPSID_REV_BIT);
+
+		vfp_vector = vfp_support_entry;
+
+		thread_register_notifier(&vfp_notifier_block);
+		vfp_pm_init();
+
+		/*
+		 * We detected VFP, and the support code is
+		 * in place; report VFP support to userspace.
+		 */
+		elf_hwcap |= HWCAP_VFP;
+#ifdef CONFIG_VFPv3
+		if (VFP_arch >= 2) {
+			elf_hwcap |= HWCAP_VFPv3;
+
+			/*
+			 * Check for VFPv3 D16 and VFPv4 D16.  CPUs in
+			 * this configuration only have 16 x 64bit
+			 * registers.
+			 */
+			if (((fmrx(MVFR0) & MVFR0_A_SIMD_MASK)) == 1)
+				elf_hwcap |= HWCAP_VFPv3D16; /* also v4-D16 */
+			else
+				elf_hwcap |= HWCAP_VFPD32;
+		}
+#endif
+		/*
+		 * Check for the presence of the Advanced SIMD
+		 * load/store instructions, integer and single
+		 * precision floating point operations. Only check
+		 * for NEON if the hardware has the MVFR registers.
+		 */
+		if ((read_cpuid_id() & 0x000f0000) == 0x000f0000) {
+#ifdef CONFIG_NEON
+			if ((fmrx(MVFR1) & 0x000fff00) == 0x00011100)
+				elf_hwcap |= HWCAP_NEON;
+#endif
+#ifdef CONFIG_VFPv3
+			if ((fmrx(MVFR1) & 0xf0000000) == 0x10000000)
+				elf_hwcap |= HWCAP_VFPv4;
+#endif
+		}
+	}
+	return 0;
+}
+
+core_initcall(vfp_init);