linux/arch/x86/lib/copy_user_64.S
Paolo Abeni 236222d393 x86/uaccess: Optimize copy_user_enhanced_fast_string() for short strings
According to the Intel datasheet, the REP MOVSB instruction
exposes a pretty heavy setup cost (50 ticks), which hurts
short string copy operations.

This change tries to avoid this cost by calling the explicit
loop available in the unrolled code for strings shorter
than 64 bytes.

The 64 bytes cutoff value is arbitrary from the code logic
point of view - it has been selected based on measurements,
as the largest value that still ensures a measurable gain.

Micro benchmarks of the __copy_from_user() function with
lengths in the [0-63] range show this performance gain
(shorter the string, larger the gain):

 - in the [55%-4%] range on Intel Xeon(R) CPU E5-2690 v4
 - in the [72%-9%] range on Intel Core i7-4810MQ

Other tested CPUs - namely Intel Atom S1260 and AMD Opteron
8216 - show no difference, because they do not expose the
ERMS feature bit.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Alan Cox <gnomes@lxorguk.ukuu.org.uk>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Hannes Frederic Sowa <hannes@stressinduktion.org>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/4533a1d101fd460f80e21329a34928fad521c1d4.1498744345.git.pabeni@redhat.com
[ Clarified the changelog. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-06-30 09:52:51 +02:00

345 lines
7.7 KiB
ArmAsm

/*
* 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.
*/
#include <linux/linkage.h>
#include <asm/current.h>
#include <asm/asm-offsets.h>
#include <asm/thread_info.h>
#include <asm/cpufeatures.h>
#include <asm/alternative-asm.h>
#include <asm/asm.h>
#include <asm/smap.h>
#include <asm/export.h>
/*
* 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:
* rdi destination
* rsi source
* rdx count
*
* Output:
* eax uncopied bytes or 0 if successful.
*/
ENTRY(copy_user_generic_unrolled)
ASM_STAC
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 .L_copy_short_string
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
decl %ecx
jnz 1b
.L_copy_short_string:
movl %edx,%ecx
andl $7,%edx
shrl $3,%ecx
jz 20f
18: movq (%rsi),%r8
19: movq %r8,(%rdi)
leaq 8(%rsi),%rsi
leaq 8(%rdi),%rdi
decl %ecx
jnz 18b
20: andl %edx,%edx
jz 23f
movl %edx,%ecx
21: movb (%rsi),%al
22: movb %al,(%rdi)
incq %rsi
incq %rdi
decl %ecx
jnz 21b
23: xor %eax,%eax
ASM_CLAC
ret
.section .fixup,"ax"
30: shll $6,%ecx
addl %ecx,%edx
jmp 60f
40: leal (%rdx,%rcx,8),%edx
jmp 60f
50: movl %ecx,%edx
60: jmp copy_user_handle_tail /* ecx is zerorest also */
.previous
_ASM_EXTABLE(1b,30b)
_ASM_EXTABLE(2b,30b)
_ASM_EXTABLE(3b,30b)
_ASM_EXTABLE(4b,30b)
_ASM_EXTABLE(5b,30b)
_ASM_EXTABLE(6b,30b)
_ASM_EXTABLE(7b,30b)
_ASM_EXTABLE(8b,30b)
_ASM_EXTABLE(9b,30b)
_ASM_EXTABLE(10b,30b)
_ASM_EXTABLE(11b,30b)
_ASM_EXTABLE(12b,30b)
_ASM_EXTABLE(13b,30b)
_ASM_EXTABLE(14b,30b)
_ASM_EXTABLE(15b,30b)
_ASM_EXTABLE(16b,30b)
_ASM_EXTABLE(18b,40b)
_ASM_EXTABLE(19b,40b)
_ASM_EXTABLE(21b,50b)
_ASM_EXTABLE(22b,50b)
ENDPROC(copy_user_generic_unrolled)
EXPORT_SYMBOL(copy_user_generic_unrolled)
/* 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)
ASM_STAC
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
1: rep
movsq
2: movl %edx,%ecx
3: rep
movsb
xorl %eax,%eax
ASM_CLAC
ret
.section .fixup,"ax"
11: leal (%rdx,%rcx,8),%ecx
12: movl %ecx,%edx /* ecx is zerorest also */
jmp copy_user_handle_tail
.previous
_ASM_EXTABLE(1b,11b)
_ASM_EXTABLE(3b,12b)
ENDPROC(copy_user_generic_string)
EXPORT_SYMBOL(copy_user_generic_string)
/*
* Some CPUs are adding enhanced REP MOVSB/STOSB instructions.
* It's recommended to use enhanced REP MOVSB/STOSB if it's enabled.
*
* Input:
* rdi destination
* rsi source
* rdx count
*
* Output:
* eax uncopied bytes or 0 if successful.
*/
ENTRY(copy_user_enhanced_fast_string)
ASM_STAC
cmpl $64,%edx
jb .L_copy_short_string /* less then 64 bytes, avoid the costly 'rep' */
movl %edx,%ecx
1: rep
movsb
xorl %eax,%eax
ASM_CLAC
ret
.section .fixup,"ax"
12: movl %ecx,%edx /* ecx is zerorest also */
jmp copy_user_handle_tail
.previous
_ASM_EXTABLE(1b,12b)
ENDPROC(copy_user_enhanced_fast_string)
EXPORT_SYMBOL(copy_user_enhanced_fast_string)
/*
* copy_user_nocache - Uncached memory copy with exception handling
* This will force destination out of cache for more performance.
*
* Note: Cached memory copy is used when destination or size is not
* naturally aligned. That is:
* - Require 8-byte alignment when size is 8 bytes or larger.
* - Require 4-byte alignment when size is 4 bytes.
*/
ENTRY(__copy_user_nocache)
ASM_STAC
/* If size is less than 8 bytes, go to 4-byte copy */
cmpl $8,%edx
jb .L_4b_nocache_copy_entry
/* If destination is not 8-byte aligned, "cache" copy to align it */
ALIGN_DESTINATION
/* Set 4x8-byte copy count and remainder */
movl %edx,%ecx
andl $63,%edx
shrl $6,%ecx
jz .L_8b_nocache_copy_entry /* jump if count is 0 */
/* Perform 4x8-byte nocache loop-copy */
.L_4x8b_nocache_copy_loop:
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
decl %ecx
jnz .L_4x8b_nocache_copy_loop
/* Set 8-byte copy count and remainder */
.L_8b_nocache_copy_entry:
movl %edx,%ecx
andl $7,%edx
shrl $3,%ecx
jz .L_4b_nocache_copy_entry /* jump if count is 0 */
/* Perform 8-byte nocache loop-copy */
.L_8b_nocache_copy_loop:
20: movq (%rsi),%r8
21: movnti %r8,(%rdi)
leaq 8(%rsi),%rsi
leaq 8(%rdi),%rdi
decl %ecx
jnz .L_8b_nocache_copy_loop
/* If no byte left, we're done */
.L_4b_nocache_copy_entry:
andl %edx,%edx
jz .L_finish_copy
/* If destination is not 4-byte aligned, go to byte copy: */
movl %edi,%ecx
andl $3,%ecx
jnz .L_1b_cache_copy_entry
/* Set 4-byte copy count (1 or 0) and remainder */
movl %edx,%ecx
andl $3,%edx
shrl $2,%ecx
jz .L_1b_cache_copy_entry /* jump if count is 0 */
/* Perform 4-byte nocache copy: */
30: movl (%rsi),%r8d
31: movnti %r8d,(%rdi)
leaq 4(%rsi),%rsi
leaq 4(%rdi),%rdi
/* If no bytes left, we're done: */
andl %edx,%edx
jz .L_finish_copy
/* Perform byte "cache" loop-copy for the remainder */
.L_1b_cache_copy_entry:
movl %edx,%ecx
.L_1b_cache_copy_loop:
40: movb (%rsi),%al
41: movb %al,(%rdi)
incq %rsi
incq %rdi
decl %ecx
jnz .L_1b_cache_copy_loop
/* Finished copying; fence the prior stores */
.L_finish_copy:
xorl %eax,%eax
ASM_CLAC
sfence
ret
.section .fixup,"ax"
.L_fixup_4x8b_copy:
shll $6,%ecx
addl %ecx,%edx
jmp .L_fixup_handle_tail
.L_fixup_8b_copy:
lea (%rdx,%rcx,8),%rdx
jmp .L_fixup_handle_tail
.L_fixup_4b_copy:
lea (%rdx,%rcx,4),%rdx
jmp .L_fixup_handle_tail
.L_fixup_1b_copy:
movl %ecx,%edx
.L_fixup_handle_tail:
sfence
jmp copy_user_handle_tail
.previous
_ASM_EXTABLE(1b,.L_fixup_4x8b_copy)
_ASM_EXTABLE(2b,.L_fixup_4x8b_copy)
_ASM_EXTABLE(3b,.L_fixup_4x8b_copy)
_ASM_EXTABLE(4b,.L_fixup_4x8b_copy)
_ASM_EXTABLE(5b,.L_fixup_4x8b_copy)
_ASM_EXTABLE(6b,.L_fixup_4x8b_copy)
_ASM_EXTABLE(7b,.L_fixup_4x8b_copy)
_ASM_EXTABLE(8b,.L_fixup_4x8b_copy)
_ASM_EXTABLE(9b,.L_fixup_4x8b_copy)
_ASM_EXTABLE(10b,.L_fixup_4x8b_copy)
_ASM_EXTABLE(11b,.L_fixup_4x8b_copy)
_ASM_EXTABLE(12b,.L_fixup_4x8b_copy)
_ASM_EXTABLE(13b,.L_fixup_4x8b_copy)
_ASM_EXTABLE(14b,.L_fixup_4x8b_copy)
_ASM_EXTABLE(15b,.L_fixup_4x8b_copy)
_ASM_EXTABLE(16b,.L_fixup_4x8b_copy)
_ASM_EXTABLE(20b,.L_fixup_8b_copy)
_ASM_EXTABLE(21b,.L_fixup_8b_copy)
_ASM_EXTABLE(30b,.L_fixup_4b_copy)
_ASM_EXTABLE(31b,.L_fixup_4b_copy)
_ASM_EXTABLE(40b,.L_fixup_1b_copy)
_ASM_EXTABLE(41b,.L_fixup_1b_copy)
ENDPROC(__copy_user_nocache)
EXPORT_SYMBOL(__copy_user_nocache)