x86: fix copy_user on x86

Switch copy_user_generic_string(), copy_user_generic_unrolled() and
__copy_user_nocache() from custom tail handlers to generic
copy_user_tail_handle().

Signed-off-by: Vitaly Mayatskikh <v.mayatskih@gmail.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

2 files changed, 275 insertions, 435 deletions

diff --git a/arch/x86/lib/copy_user_64.S b/arch/x86/lib/copy_user_64.S
index 7eaaf0123b4..e5afb4ad3f1 100644
--- a/arch/x86/lib/copy_user_64.S
+++ b/arch/x86/lib/copy_user_64.S
@@ -1,8 +1,10 @@
-/* Copyright 2002 Andi Kleen, SuSE Labs.
+/*
+ * Copyright 2008 Vitaly Mayatskikh <vmayatsk@redhat.com>
+ * Copyright 2002 Andi Kleen, SuSE Labs.
  * Subject to the GNU Public License v2.
- * 
- * Functions to copy from and to user space.		
- */		 
+ *
+ * Functions to copy from and to user space.
+ */
 
 #include <linux/linkage.h>
 #include <asm/dwarf2.h>
@@ -20,60 +22,88 @@
 	.long \orig-1f	/* by default jump to orig */
 1:
 	.section .altinstr_replacement,"ax"
-2:	.byte 0xe9	             /* near jump with 32bit immediate */
+2:	.byte 0xe9			/* near jump with 32bit immediate */
 	.long \alt-1b /* offset */   /* or alternatively to alt */
 	.previous
 	.section .altinstructions,"a"
 	.align 8
 	.quad  0b
 	.quad  2b
-	.byte  \feature		     /* when feature is set */
+	.byte  \feature			/* when feature is set */
 	.byte  5
 	.byte  5
 	.previous
 	.endm
 
-/* Standard copy_to_user with segment limit checking */		
+	.macro ALIGN_DESTINATION
+#ifdef FIX_ALIGNMENT
+	/* check for bad alignment of destination */
+	movl %edi,%ecx
+	andl $7,%ecx
+	jz 102f				/* already aligned */
+	subl $8,%ecx
+	negl %ecx
+	subl %ecx,%edx
+100:	movb (%rsi),%al
+101:	movb %al,(%rdi)
+	incq %rsi
+	incq %rdi
+	decl %ecx
+	jnz 100b
+102:
+	.section .fixup,"ax"
+103:	addl %r8d,%edx			/* ecx is zerorest also */
+	jmp copy_user_handle_tail
+	.previous
+
+	.section __ex_table,"a"
+	.align 8
+	.quad 100b,103b
+	.quad 101b,103b
+	.previous
+#endif
+	.endm
+
+/* Standard copy_to_user with segment limit checking */
 ENTRY(copy_to_user)
 	CFI_STARTPROC
 	GET_THREAD_INFO(%rax)
 	movq %rdi,%rcx
 	addq %rdx,%rcx
-	jc  bad_to_user
+	jc bad_to_user
 	cmpq TI_addr_limit(%rax),%rcx
 	jae bad_to_user
-	xorl %eax,%eax	/* clear zero flag */
 	ALTERNATIVE_JUMP X86_FEATURE_REP_GOOD,copy_user_generic_unrolled,copy_user_generic_string
 	CFI_ENDPROC
 
-ENTRY(copy_user_generic)
+/* Standard copy_from_user with segment limit checking */
+ENTRY(copy_from_user)
 	CFI_STARTPROC
-	movl $1,%ecx	/* set zero flag */
+	GET_THREAD_INFO(%rax)
+	movq %rsi,%rcx
+	addq %rdx,%rcx
+	jc bad_from_user
+	cmpq TI_addr_limit(%rax),%rcx
+	jae bad_from_user
 	ALTERNATIVE_JUMP X86_FEATURE_REP_GOOD,copy_user_generic_unrolled,copy_user_generic_string
 	CFI_ENDPROC
+ENDPROC(copy_from_user)
 
-ENTRY(__copy_from_user_inatomic)
+ENTRY(copy_user_generic)
 	CFI_STARTPROC
-	xorl %ecx,%ecx	/* clear zero flag */
 	ALTERNATIVE_JUMP X86_FEATURE_REP_GOOD,copy_user_generic_unrolled,copy_user_generic_string
 	CFI_ENDPROC
+ENDPROC(copy_user_generic)
 
-/* Standard copy_from_user with segment limit checking */	
-ENTRY(copy_from_user)
+ENTRY(__copy_from_user_inatomic)
 	CFI_STARTPROC
-	GET_THREAD_INFO(%rax)
-	movq %rsi,%rcx
-	addq %rdx,%rcx
-	jc  bad_from_user
-	cmpq TI_addr_limit(%rax),%rcx
-	jae  bad_from_user
-	movl $1,%ecx	/* set zero flag */
 	ALTERNATIVE_JUMP X86_FEATURE_REP_GOOD,copy_user_generic_unrolled,copy_user_generic_string
 	CFI_ENDPROC
-ENDPROC(copy_from_user)
-	
+ENDPROC(__copy_from_user_inatomic)
+
 	.section .fixup,"ax"
 	/* must zero dest */
+ENTRY(bad_from_user)
 bad_from_user:
 	CFI_STARTPROC
 	movl %edx,%ecx
@@ -81,271 +111,158 @@ bad_from_user:
 	rep
 	stosb
 bad_to_user:
-	movl	%edx,%eax
+	movl %edx,%eax
 	ret
 	CFI_ENDPROC
-END(bad_from_user)
+ENDPROC(bad_from_user)
 	.previous
-	
-		
+
 /*
  * copy_user_generic_unrolled - memory copy with exception handling.
- * This version is for CPUs like P4 that don't have efficient micro code for rep movsq
- * 	
- * Input:	
+ * This version is for CPUs like P4 that don't have efficient micro
+ * code for rep movsq
+ *
+ * Input:
  * rdi destination
  * rsi source
  * rdx count
- * ecx zero flag -- if true zero destination on error
  *
- * Output:		
- * eax uncopied bytes or 0 if successful.
+ * Output:
+ * eax uncopied bytes or 0 if successfull.
  */
 ENTRY(copy_user_generic_unrolled)
 	CFI_STARTPROC
-	pushq %rbx
-	CFI_ADJUST_CFA_OFFSET 8
-	CFI_REL_OFFSET rbx, 0
-	pushq %rcx
-	CFI_ADJUST_CFA_OFFSET 8
-	CFI_REL_OFFSET rcx, 0
-	xorl %eax,%eax		/*zero for the exception handler */
-
-#ifdef FIX_ALIGNMENT
-	/* check for bad alignment of destination */
-	movl %edi,%ecx
-	andl $7,%ecx
-	jnz  .Lbad_alignment
-.Lafter_bad_alignment:
-#endif
-
-	movq %rdx,%rcx
-
-	movl $64,%ebx
-	shrq $6,%rdx
-	decq %rdx
-	js   .Lhandle_tail
-
-	.p2align 4
-.Lloop:
-.Ls1:	movq (%rsi),%r11
-.Ls2:	movq 1*8(%rsi),%r8
-.Ls3:	movq 2*8(%rsi),%r9
-.Ls4:	movq 3*8(%rsi),%r10
-.Ld1:	movq %r11,(%rdi)
-.Ld2:	movq %r8,1*8(%rdi)
-.Ld3:	movq %r9,2*8(%rdi)
-.Ld4:	movq %r10,3*8(%rdi)
-
-.Ls5:	movq 4*8(%rsi),%r11
-.Ls6:	movq 5*8(%rsi),%r8
-.Ls7:	movq 6*8(%rsi),%r9
-.Ls8:	movq 7*8(%rsi),%r10
-.Ld5:	movq %r11,4*8(%rdi)
-.Ld6:	movq %r8,5*8(%rdi)
-.Ld7:	movq %r9,6*8(%rdi)
-.Ld8:	movq %r10,7*8(%rdi)
-
-	decq %rdx
-
+	cmpl $8,%edx
+	jb 20f		/* less then 8 bytes, go to byte copy loop */
+	ALIGN_DESTINATION
+	movl %edx,%ecx
+	andl $63,%edx
+	shrl $6,%ecx
+	jz 17f
+1:	movq (%rsi),%r8
+2:	movq 1*8(%rsi),%r9
+3:	movq 2*8(%rsi),%r10
+4:	movq 3*8(%rsi),%r11
+5:	movq %r8,(%rdi)
+6:	movq %r9,1*8(%rdi)
+7:	movq %r10,2*8(%rdi)
+8:	movq %r11,3*8(%rdi)
+9:	movq 4*8(%rsi),%r8
+10:	movq 5*8(%rsi),%r9
+11:	movq 6*8(%rsi),%r10
+12:	movq 7*8(%rsi),%r11
+13:	movq %r8,4*8(%rdi)
+14:	movq %r9,5*8(%rdi)
+15:	movq %r10,6*8(%rdi)
+16:	movq %r11,7*8(%rdi)
 	leaq 64(%rsi),%rsi
 	leaq 64(%rdi),%rdi
-
-	jns  .Lloop
-
-	.p2align 4
-.Lhandle_tail:
-	movl %ecx,%edx
-	andl $63,%ecx
-	shrl $3,%ecx
-	jz   .Lhandle_7
-	movl $8,%ebx
-	.p2align 4
-.Lloop_8:
-.Ls9:	movq (%rsi),%r8
-.Ld9:	movq %r8,(%rdi)
 	decl %ecx
-	leaq 8(%rdi),%rdi
+	jnz 1b
+17:	movl %edx,%ecx
+	andl $7,%edx
+	shrl $3,%ecx
+	jz 20f
+18:	movq (%rsi),%r8
+19:	movq %r8,(%rdi)
 	leaq 8(%rsi),%rsi
-	jnz .Lloop_8
-
-.Lhandle_7:
+	leaq 8(%rdi),%rdi
+	decl %ecx
+	jnz 18b
+20:	andl %edx,%edx
+	jz 23f
 	movl %edx,%ecx
-	andl $7,%ecx
-	jz   .Lende
-	.p2align 4
-.Lloop_1:
-.Ls10:	movb (%rsi),%bl
-.Ld10:	movb %bl,(%rdi)
-	incq %rdi
+21:	movb (%rsi),%al
+22:	movb %al,(%rdi)
 	incq %rsi
+	incq %rdi
 	decl %ecx
-	jnz .Lloop_1
-
-	CFI_REMEMBER_STATE
-.Lende:
-	popq %rcx
-	CFI_ADJUST_CFA_OFFSET -8
-	CFI_RESTORE rcx
-	popq %rbx
-	CFI_ADJUST_CFA_OFFSET -8
-	CFI_RESTORE rbx
+	jnz 21b
+23:	xor %eax,%eax
 	ret
-	CFI_RESTORE_STATE
 
-#ifdef FIX_ALIGNMENT
-	/* align destination */
-	.p2align 4
-.Lbad_alignment:
-	movl $8,%r9d
-	subl %ecx,%r9d
-	movl %r9d,%ecx
-	cmpq %r9,%rdx
-	jz   .Lhandle_7
-	js   .Lhandle_7
-.Lalign_1:
-.Ls11:	movb (%rsi),%bl
-.Ld11:	movb %bl,(%rdi)
-	incq %rsi
-	incq %rdi
-	decl %ecx
-	jnz .Lalign_1
-	subq %r9,%rdx
-	jmp .Lafter_bad_alignment
-#endif
+	.section .fixup,"ax"
+30:	shll $6,%ecx
+	addl %ecx,%edx
+	jmp 60f
+40:	leal (%edx,%ecx,8),%edx
+	jmp 60f
+50:	movl %ecx,%edx
+60:	jmp copy_user_handle_tail /* ecx is zerorest also */
+	.previous
 
-	/* table sorted by exception address */
 	.section __ex_table,"a"
 	.align 8
-	.quad .Ls1,.Ls1e	/* Ls1-Ls4 have copied zero bytes */
-	.quad .Ls2,.Ls1e
-	.quad .Ls3,.Ls1e
-	.quad .Ls4,.Ls1e
-	.quad .Ld1,.Ls1e	/* Ld1-Ld4 have copied 0-24 bytes */
-	.quad .Ld2,.Ls2e
-	.quad .Ld3,.Ls3e
-	.quad .Ld4,.Ls4e
-	.quad .Ls5,.Ls5e	/* Ls5-Ls8 have copied 32 bytes */
-	.quad .Ls6,.Ls5e
-	.quad .Ls7,.Ls5e
-	.quad .Ls8,.Ls5e
-	.quad .Ld5,.Ls5e	/* Ld5-Ld8 have copied 32-56 bytes */
-	.quad .Ld6,.Ls6e
-	.quad .Ld7,.Ls7e
-	.quad .Ld8,.Ls8e
-	.quad .Ls9,.Le_quad
-	.quad .Ld9,.Le_quad
-	.quad .Ls10,.Le_byte
-	.quad .Ld10,.Le_byte
-#ifdef FIX_ALIGNMENT
-	.quad .Ls11,.Lzero_rest
-	.quad .Ld11,.Lzero_rest
-#endif
-	.quad .Le5,.Le_zero
+	.quad 1b,30b
+	.quad 2b,30b
+	.quad 3b,30b
+	.quad 4b,30b
+	.quad 5b,30b
+	.quad 6b,30b
+	.quad 7b,30b
+	.quad 8b,30b
+	.quad 9b,30b
+	.quad 10b,30b
+	.quad 11b,30b
+	.quad 12b,30b
+	.quad 13b,30b
+	.quad 14b,30b
+	.quad 15b,30b
+	.quad 16b,30b
+	.quad 18b,40b
+	.quad 19b,40b
+	.quad 21b,50b
+	.quad 22b,50b
 	.previous
-
-	/* eax: zero, ebx: 64 */
-.Ls1e: 	addl $8,%eax		/* eax is bytes left uncopied within the loop (Ls1e: 64 .. Ls8e: 8) */
-.Ls2e: 	addl $8,%eax
-.Ls3e: 	addl $8,%eax
-.Ls4e: 	addl $8,%eax
-.Ls5e: 	addl $8,%eax
-.Ls6e: 	addl $8,%eax
-.Ls7e: 	addl $8,%eax
-.Ls8e: 	addl $8,%eax
-	addq %rbx,%rdi	/* +64 */
-	subq %rax,%rdi  /* correct destination with computed offset */
-
-	shlq $6,%rdx	/* loop counter * 64 (stride length) */
-	addq %rax,%rdx	/* add offset to loopcnt */
-	andl $63,%ecx	/* remaining bytes */
-	addq %rcx,%rdx	/* add them */
-	jmp .Lzero_rest
-
-	/* exception on quad word loop in tail handling */
-	/* ecx:	loopcnt/8, %edx: length, rdi: correct */
-.Le_quad:
-	shll $3,%ecx
-	andl $7,%edx
-	addl %ecx,%edx
-	/* edx: bytes to zero, rdi: dest, eax:zero */
-.Lzero_rest:
-	cmpl $0,(%rsp)
-	jz   .Le_zero
-	movq %rdx,%rcx
-.Le_byte:
-	xorl %eax,%eax
-.Le5:	rep
-	stosb
-	/* when there is another exception while zeroing the rest just return */
-.Le_zero:
-	movq %rdx,%rax
-	jmp .Lende
 	CFI_ENDPROC
-ENDPROC(copy_user_generic)
+ENDPROC(copy_user_generic_unrolled)
 
-
-	/* Some CPUs run faster using the string copy instructions.
-	   This is also a lot simpler. Use them when possible.
-	   Patch in jmps to this code instead of copying it fully
-	   to avoid unwanted aliasing in the exception tables. */
-
- /* rdi	destination
-  * rsi source
-  * rdx count
-  * ecx zero flag
-  *
-  * Output:
-  * eax uncopied bytes or 0 if successfull.
-  *
-  * Only 4GB of copy is supported. This shouldn't be a problem
-  * because the kernel normally only writes from/to page sized chunks
-  * even if user space passed a longer buffer.
-  * And more would be dangerous because both Intel and AMD have
-  * errata with rep movsq > 4GB. If someone feels the need to fix
-  * this please consider this.
-  */
+/* Some CPUs run faster using the string copy instructions.
+ * This is also a lot simpler. Use them when possible.
+ *
+ * Only 4GB of copy is supported. This shouldn't be a problem
+ * because the kernel normally only writes from/to page sized chunks
+ * even if user space passed a longer buffer.
+ * And more would be dangerous because both Intel and AMD have
+ * errata with rep movsq > 4GB. If someone feels the need to fix
+ * this please consider this.
+ *
+ * Input:
+ * rdi destination
+ * rsi source
+ * rdx count
+ *
+ * Output:
+ * eax uncopied bytes or 0 if successful.
+ */
 ENTRY(copy_user_generic_string)
 	CFI_STARTPROC
-	movl %ecx,%r8d		/* save zero flag */
+	andl %edx,%edx
+	jz 4f
+	cmpl $8,%edx
+	jb 2f		/* less than 8 bytes, go to byte copy loop */
+	ALIGN_DESTINATION
 	movl %edx,%ecx
 	shrl $3,%ecx
-	andl $7,%edx	
-	jz   10f
-1:	rep 
-	movsq 
-	movl %edx,%ecx
-2:	rep
-	movsb
-9:	movl %ecx,%eax
-	ret
-
-	/* multiple of 8 byte */
-10:	rep
+	andl $7,%edx
+1:	rep
 	movsq
-	xor %eax,%eax
+2:	movl %edx,%ecx
+3:	rep
+	movsb
+4:	xorl %eax,%eax
 	ret
 
-	/* exception handling */
-3:      lea (%rdx,%rcx,8),%rax	/* exception on quad loop */
-	jmp 6f
-5:	movl %ecx,%eax		/* exception on byte loop */
-	/* eax: left over bytes */
-6:	testl %r8d,%r8d		/* zero flag set? */
-	jz 7f
-	movl %eax,%ecx		/* initialize x86 loop counter */
-	push %rax
-	xorl %eax,%eax
-8:	rep
-	stosb 			/* zero the rest */
-11:	pop %rax
-7:	ret
-	CFI_ENDPROC
-END(copy_user_generic_c)
+	.section .fixup,"ax"
+11:	leal (%edx,%ecx,8),%ecx
+12:	movl %ecx,%edx		/* ecx is zerorest also */
+	jmp copy_user_handle_tail
+	.previous
 
 	.section __ex_table,"a"
-	.quad 1b,3b
-	.quad 2b,5b
-	.quad 8b,11b
-	.quad 10b,3b
+	.align 8
+	.quad 1b,11b
+	.quad 3b,12b
 	.previous
+	CFI_ENDPROC
+ENDPROC(copy_user_generic_string)
diff --git a/arch/x86/lib/copy_user_nocache_64.S b/arch/x86/lib/copy_user_nocache_64.S
index 9d3d1ab8376..93353d6a526 100644
--- a/arch/x86/lib/copy_user_nocache_64.S
+++ b/arch/x86/lib/copy_user_nocache_64.S
@@ -1,4 +1,6 @@
-/* Copyright 2002 Andi Kleen, SuSE Labs.
+/*
+ * Copyright 2008 Vitaly Mayatskikh <vmayatsk@redhat.com>
+ * Copyright 2002 Andi Kleen, SuSE Labs.
  * Subject to the GNU Public License v2.
  *
  * Functions to copy from and to user space.
@@ -12,204 +14,125 @@
 #include <asm/current.h>
 #include <asm/asm-offsets.h>
 #include <asm/thread_info.h>
-#include <asm/cpufeature.h>
-
-/*
- * copy_user_nocache - Uncached memory copy with exception handling
- * This will force destination/source out of cache for more performance.
- *
- * Input:
- * rdi destination
- * rsi source
- * rdx count
- * rcx zero flag	when 1 zero on exception
- *
- * Output:
- * eax uncopied bytes or 0 if successful.
- */
-ENTRY(__copy_user_nocache)
-	CFI_STARTPROC
-	pushq %rbx
-	CFI_ADJUST_CFA_OFFSET 8
-	CFI_REL_OFFSET rbx, 0
-	pushq %rcx		/* save zero flag */
-	CFI_ADJUST_CFA_OFFSET 8
-	CFI_REL_OFFSET rcx, 0
-
-	xorl %eax,%eax		/* zero for the exception handler */
 
+	.macro ALIGN_DESTINATION
 #ifdef FIX_ALIGNMENT
 	/* check for bad alignment of destination */
 	movl %edi,%ecx
 	andl $7,%ecx
-	jnz  .Lbad_alignment
-.Lafter_bad_alignment:
-#endif
-
-	movq %rdx,%rcx
-
-	movl $64,%ebx
-	shrq $6,%rdx
-	decq %rdx
-	js   .Lhandle_tail
-
-	.p2align 4
-.Lloop:
-.Ls1:	movq (%rsi),%r11
-.Ls2:	movq 1*8(%rsi),%r8
-.Ls3:	movq 2*8(%rsi),%r9
-.Ls4:	movq 3*8(%rsi),%r10
-.Ld1:	movnti %r11,(%rdi)
-.Ld2:	movnti %r8,1*8(%rdi)
-.Ld3:	movnti %r9,2*8(%rdi)
-.Ld4:	movnti %r10,3*8(%rdi)
-
-.Ls5:	movq 4*8(%rsi),%r11
-.Ls6:	movq 5*8(%rsi),%r8
-.Ls7:	movq 6*8(%rsi),%r9
-.Ls8:	movq 7*8(%rsi),%r10
-.Ld5:	movnti %r11,4*8(%rdi)
-.Ld6:	movnti %r8,5*8(%rdi)
-.Ld7:	movnti %r9,6*8(%rdi)
-.Ld8:	movnti %r10,7*8(%rdi)
+	jz 102f				/* already aligned */
+	subl $8,%ecx
+	negl %ecx
+	subl %ecx,%edx
+100:	movb (%rsi),%al
+101:	movb %al,(%rdi)
+	incq %rsi
+	incq %rdi
+	decl %ecx
+	jnz 100b
+102:
+	.section .fixup,"ax"
+103:	addl %r8d,%edx			/* ecx is zerorest also */
+	jmp copy_user_handle_tail
+	.previous
 
-	dec  %rdx
+	.section __ex_table,"a"
+	.align 8
+	.quad 100b,103b
+	.quad 101b,103b
+	.previous
+#endif
+	.endm
 
+/*
+ * copy_user_nocache - Uncached memory copy with exception handling
+ * This will force destination/source out of cache for more performance.
+ */
+ENTRY(__copy_user_nocache)
+	CFI_STARTPROC
+	cmpl $8,%edx
+	jb 20f		/* less then 8 bytes, go to byte copy loop */
+	ALIGN_DESTINATION
+	movl %edx,%ecx
+	andl $63,%edx
+	shrl $6,%ecx
+	jz 17f
+1:	movq (%rsi),%r8
+2:	movq 1*8(%rsi),%r9
+3:	movq 2*8(%rsi),%r10
+4:	movq 3*8(%rsi),%r11
+5:	movnti %r8,(%rdi)
+6:	movnti %r9,1*8(%rdi)
+7:	movnti %r10,2*8(%rdi)
+8:	movnti %r11,3*8(%rdi)
+9:	movq 4*8(%rsi),%r8
+10:	movq 5*8(%rsi),%r9
+11:	movq 6*8(%rsi),%r10
+12:	movq 7*8(%rsi),%r11
+13:	movnti %r8,4*8(%rdi)
+14:	movnti %r9,5*8(%rdi)
+15:	movnti %r10,6*8(%rdi)
+16:	movnti %r11,7*8(%rdi)
 	leaq 64(%rsi),%rsi
 	leaq 64(%rdi),%rdi
-
-	jns  .Lloop
-
-	.p2align 4
-.Lhandle_tail:
-	movl %ecx,%edx
-	andl $63,%ecx
-	shrl $3,%ecx
-	jz   .Lhandle_7
-	movl $8,%ebx
-	.p2align 4
-.Lloop_8:
-.Ls9:	movq (%rsi),%r8
-.Ld9:	movnti %r8,(%rdi)
 	decl %ecx
-	leaq 8(%rdi),%rdi
+	jnz 1b
+17:	movl %edx,%ecx
+	andl $7,%edx
+	shrl $3,%ecx
+	jz 20f
+18:	movq (%rsi),%r8
+19:	movnti %r8,(%rdi)
 	leaq 8(%rsi),%rsi
-	jnz .Lloop_8
-
-.Lhandle_7:
+	leaq 8(%rdi),%rdi
+	decl %ecx
+	jnz 18b
+20:	andl %edx,%edx
+	jz 23f
 	movl %edx,%ecx
-	andl $7,%ecx
-	jz   .Lende
-	.p2align 4
-.Lloop_1:
-.Ls10:	movb (%rsi),%bl
-.Ld10:	movb %bl,(%rdi)
-	incq %rdi
+21:	movb (%rsi),%al
+22:	movb %al,(%rdi)
 	incq %rsi
+	incq %rdi
 	decl %ecx
-	jnz .Lloop_1
-
-	CFI_REMEMBER_STATE
-.Lende:
-	popq %rcx
-	CFI_ADJUST_CFA_OFFSET -8
-	CFI_RESTORE %rcx
-	popq %rbx
-	CFI_ADJUST_CFA_OFFSET -8
-	CFI_RESTORE rbx
+	jnz 21b
+23:	xorl %eax,%eax
 	sfence
 	ret
-	CFI_RESTORE_STATE
 
-#ifdef FIX_ALIGNMENT
-	/* align destination */
-	.p2align 4
-.Lbad_alignment:
-	movl $8,%r9d
-	subl %ecx,%r9d
-	movl %r9d,%ecx
-	cmpq %r9,%rdx
-	jz   .Lhandle_7
-	js   .Lhandle_7
-.Lalign_1:
-.Ls11:	movb (%rsi),%bl
-.Ld11:	movb %bl,(%rdi)
-	incq %rsi
-	incq %rdi
-	decl %ecx
-	jnz .Lalign_1
-	subq %r9,%rdx
-	jmp .Lafter_bad_alignment
-#endif
+	.section .fixup,"ax"
+30:	shll $6,%ecx
+	addl %ecx,%edx
+	jmp 60f
+40:	leal (%edx,%ecx,8),%edx
+	jmp 60f
+50:	movl %ecx,%edx
+60:	sfence
+	movl %r8d,%ecx
+	jmp copy_user_handle_tail
+	.previous
 
-	/* table sorted by exception address */
 	.section __ex_table,"a"
-	.align 8
-	.quad .Ls1,.Ls1e	/* .Ls[1-4] - 0 bytes copied */
-	.quad .Ls2,.Ls1e
-	.quad .Ls3,.Ls1e
-	.quad .Ls4,.Ls1e
-	.quad .Ld1,.Ls1e	/* .Ld[1-4] - 0..24 bytes coped */
-	.quad .Ld2,.Ls2e
-	.quad .Ld3,.Ls3e
-	.quad .Ld4,.Ls4e
-	.quad .Ls5,.Ls5e	/* .Ls[5-8] - 32 bytes copied */
-	.quad .Ls6,.Ls5e
-	.quad .Ls7,.Ls5e
-	.quad .Ls8,.Ls5e
-	.quad .Ld5,.Ls5e	/* .Ld[5-8] - 32..56 bytes copied */
-	.quad .Ld6,.Ls6e
-	.quad .Ld7,.Ls7e
-	.quad .Ld8,.Ls8e
-	.quad .Ls9,.Le_quad
-	.quad .Ld9,.Le_quad
-	.quad .Ls10,.Le_byte
-	.quad .Ld10,.Le_byte
-#ifdef FIX_ALIGNMENT
-	.quad .Ls11,.Lzero_rest
-	.quad .Ld11,.Lzero_rest
-#endif
-	.quad .Le5,.Le_zero
+	.quad 1b,30b
+	.quad 2b,30b
+	.quad 3b,30b
+	.quad 4b,30b
+	.quad 5b,30b
+	.quad 6b,30b
+	.quad 7b,30b
+	.quad 8b,30b
+	.quad 9b,30b
+	.quad 10b,30b
+	.quad 11b,30b
+	.quad 12b,30b
+	.quad 13b,30b
+	.quad 14b,30b
+	.quad 15b,30b
+	.quad 16b,30b
+	.quad 18b,40b
+	.quad 19b,40b
+	.quad 21b,50b
+	.quad 22b,50b
 	.previous
-
-	/* eax: zero, ebx: 64 */
-.Ls1e: 	addl $8,%eax	/* eax: bytes left uncopied: Ls1e: 64 .. Ls8e: 8 */
-.Ls2e: 	addl $8,%eax
-.Ls3e: 	addl $8,%eax
-.Ls4e: 	addl $8,%eax
-.Ls5e: 	addl $8,%eax
-.Ls6e: 	addl $8,%eax
-.Ls7e: 	addl $8,%eax
-.Ls8e: 	addl $8,%eax
-	addq %rbx,%rdi	/* +64 */
-	subq %rax,%rdi  /* correct destination with computed offset */
-
-	shlq $6,%rdx	/* loop counter * 64 (stride length) */
-	addq %rax,%rdx	/* add offset to loopcnt */
-	andl $63,%ecx	/* remaining bytes */
-	addq %rcx,%rdx	/* add them */
-	jmp .Lzero_rest
-
-	/* exception on quad word loop in tail handling */
-	/* ecx:	loopcnt/8, %edx: length, rdi: correct */
-.Le_quad:
-	shll $3,%ecx
-	andl $7,%edx
-	addl %ecx,%edx
-	/* edx: bytes to zero, rdi: dest, eax:zero */
-.Lzero_rest:
-	cmpl $0,(%rsp)	/* zero flag set? */
-	jz   .Le_zero
-	movq %rdx,%rcx
-.Le_byte:
-	xorl %eax,%eax
-.Le5:	rep
-	stosb
-	/* when there is another exception while zeroing the rest just return */
-.Le_zero:
-	movq %rdx,%rax
-	jmp .Lende
 	CFI_ENDPROC
 ENDPROC(__copy_user_nocache)
-
-