diff options
| author | Linus Torvalds <torvalds@linux-foundation.org> | 2023-04-15 23:22:31 +0300 |
|---|---|---|
| committer | Linus Torvalds <torvalds@linux-foundation.org> | 2023-04-19 03:05:28 +0300 |
| commit | d2c95f9d6802cc518d71d9795f4d9da54fb4e24d (patch) | |
| tree | ca5498de4284cf09409c69cbc0c1ddc869d2a747 /arch/x86/lib | |
| parent | adfcf4231b8cbc2d9c1e7bfaa965b907e60639eb (diff) | |
| download | linux-d2c95f9d6802cc518d71d9795f4d9da54fb4e24d.tar.xz | |
x86: don't use REP_GOOD or ERMS for user memory clearing
The modern target to use is FSRS (Fast Short REP STOS), and the other
cases should only be used for bigger areas (ie mainly things like page
clearing).
Note! This changes the conditional for the inlining from FSRM ("fast
short rep movs") to FSRS ("fast short rep stos").
We'll have a separate fixup for AMD microarchitectures that have a good
'rep stosb' yet do not set the new Intel-specific FSRS bit (because FSRM
was there first).
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'arch/x86/lib')
| -rw-r--r-- | arch/x86/lib/clear_page_64.S | 75 |
1 files changed, 0 insertions, 75 deletions
diff --git a/arch/x86/lib/clear_page_64.S b/arch/x86/lib/clear_page_64.S index ecbfb4dd3b01..fcd01b9f8d50 100644 --- a/arch/x86/lib/clear_page_64.S +++ b/arch/x86/lib/clear_page_64.S @@ -113,78 +113,3 @@ SYM_FUNC_START(clear_user_original) _ASM_EXTABLE_UA(.Lbytes, .Lbytes_exception) SYM_FUNC_END(clear_user_original) EXPORT_SYMBOL(clear_user_original) - -/* - * Alternative clear user-space when CPU feature X86_FEATURE_REP_GOOD is - * present. - * Input: - * rdi destination - * rcx count - * - * Output: - * rcx: uncleared bytes or 0 if successful. - */ -SYM_FUNC_START(clear_user_rep_good) - # call the original thing for less than a cacheline - cmp $64, %rcx - jb clear_user_original - -.Lprep: - # copy lower 32-bits for rest bytes - mov %ecx, %edx - shr $3, %rcx - jz .Lrep_good_rest_bytes - -.Lrep_good_qwords: - rep stosq - -.Lrep_good_rest_bytes: - and $7, %edx - jz .Lrep_good_exit - -.Lrep_good_bytes: - mov %edx, %ecx - rep stosb - -.Lrep_good_exit: - # see .Lexit comment above - xor %eax, %eax - RET - -.Lrep_good_qwords_exception: - # convert remaining qwords back into bytes to return to caller - shl $3, %rcx - and $7, %edx - add %rdx, %rcx - jmp .Lrep_good_exit - - _ASM_EXTABLE_UA(.Lrep_good_qwords, .Lrep_good_qwords_exception) - _ASM_EXTABLE_UA(.Lrep_good_bytes, .Lrep_good_exit) -SYM_FUNC_END(clear_user_rep_good) -EXPORT_SYMBOL(clear_user_rep_good) - -/* - * Alternative clear user-space when CPU feature X86_FEATURE_ERMS is present. - * Input: - * rdi destination - * rcx count - * - * Output: - * rcx: uncleared bytes or 0 if successful. - * - */ -SYM_FUNC_START(clear_user_erms) - # call the original thing for less than a cacheline - cmp $64, %rcx - jb clear_user_original - -.Lerms_bytes: - rep stosb - -.Lerms_exit: - xorl %eax,%eax - RET - - _ASM_EXTABLE_UA(.Lerms_bytes, .Lerms_exit) -SYM_FUNC_END(clear_user_erms) -EXPORT_SYMBOL(clear_user_erms) |