The patch below removes an unnecessary locked instruction from switch_to().
TIF_FORK is only ever set in copy_thread() on initial process creation, and
gets cleared during the first scheduling of the process. As such, it is
safe to use an unlocked test for the flag within switch_to()./
Signed-off-by: Benjamin LaHaise <[email protected]
diff --git a/arch/x86/include/asm/system.h b/arch/x86/include/asm/system.h
index 2ed3f0f..ee2e589 100644
--- a/arch/x86/include/asm/system.h
+++ b/arch/x86/include/asm/system.h
@@ -96,15 +96,15 @@ do { \
"thread_return:\n\t" \
"movq %%gs:%P[pda_pcurrent],%%rsi\n\t" \
"movq %P[thread_info](%%rsi),%%r8\n\t" \
- LOCK_PREFIX "btr %[tif_fork],%P[ti_flags](%%r8)\n\t" \
"movq %%rax,%%rdi\n\t" \
- "jc ret_from_fork\n\t" \
+ "testl %[_tif_fork],%P[ti_flags](%%r8)\n\t" \
+ "jnz ret_from_fork\n\t" \
RESTORE_CONTEXT \
: "=a" (last) \
: [next] "S" (next), [prev] "D" (prev), \
[threadrsp] "i" (offsetof(struct task_struct, thread.sp)), \
[ti_flags] "i" (offsetof(struct thread_info, flags)), \
- [tif_fork] "i" (TIF_FORK), \
+ [_tif_fork] "i" (_TIF_FORK), \
[thread_info] "i" (offsetof(struct task_struct, stack)), \
[pda_pcurrent] "i" (offsetof(struct x8664_pda, pcurrent)) \
: "memory", "cc" __EXTRA_CLOBBER)
diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S
index b86f332..16b4587 100644
--- a/arch/x86/kernel/entry_64.S
+++ b/arch/x86/kernel/entry_64.S
@@ -247,6 +247,7 @@ ENTRY(native_usergs_sysret64)
*/
/* rdi: prev */
ENTRY(ret_from_fork)
+ LOCK ; btr $TIF_FORK,TI_flags(%r8)
CFI_DEFAULT_STACK
push kernel_eflags(%rip)
CFI_ADJUST_CFA_OFFSET 8
* Benjamin LaHaise <[email protected]> wrote:
> The patch below removes an unnecessary locked instruction from
> switch_to(). TIF_FORK is only ever set in copy_thread() on initial
> process creation, and gets cleared during the first scheduling of the
> process. As such, it is safe to use an unlocked test for the flag
> within switch_to()./
>
> Signed-off-by: Benjamin LaHaise <[email protected]
>
> diff --git a/arch/x86/include/asm/system.h b/arch/x86/include/asm/system.h
> index 2ed3f0f..ee2e589 100644
> --- a/arch/x86/include/asm/system.h
> +++ b/arch/x86/include/asm/system.h
> @@ -96,15 +96,15 @@ do { \
> "thread_return:\n\t" \
> "movq %%gs:%P[pda_pcurrent],%%rsi\n\t" \
> "movq %P[thread_info](%%rsi),%%r8\n\t" \
> - LOCK_PREFIX "btr %[tif_fork],%P[ti_flags](%%r8)\n\t" \
> "movq %%rax,%%rdi\n\t" \
> - "jc ret_from_fork\n\t" \
> + "testl %[_tif_fork],%P[ti_flags](%%r8)\n\t" \
> + "jnz ret_from_fork\n\t" \
> RESTORE_CONTEXT \
> : "=a" (last) \
> : [next] "S" (next), [prev] "D" (prev), \
> [threadrsp] "i" (offsetof(struct task_struct, thread.sp)), \
> [ti_flags] "i" (offsetof(struct thread_info, flags)), \
> - [tif_fork] "i" (TIF_FORK), \
> + [_tif_fork] "i" (_TIF_FORK), \
> [thread_info] "i" (offsetof(struct task_struct, stack)), \
> [pda_pcurrent] "i" (offsetof(struct x8664_pda, pcurrent)) \
> : "memory", "cc" __EXTRA_CLOBBER)
> diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S
> index b86f332..16b4587 100644
> --- a/arch/x86/kernel/entry_64.S
> +++ b/arch/x86/kernel/entry_64.S
> @@ -247,6 +247,7 @@ ENTRY(native_usergs_sysret64)
> */
> /* rdi: prev */
> ENTRY(ret_from_fork)
> + LOCK ; btr $TIF_FORK,TI_flags(%r8)
> CFI_DEFAULT_STACK
nice!
Small detail, it does not apply to latest -git cleanly:
hunk #1 FAILED at 247.
1 out of 1 hunk FAILED -- rejects in file arch/x86/kernel/entry_64.S
Probably trivial - would you mind to repost?
Ingo
Here is a rediffed and test booted version of the patch against linus' git
tree as of a couple of hours ago.
The patch below removes an unnecessary locked instruction from
switch_to(). TIF_FORK is only ever set in copy_thread() on initial
process creation, and gets cleared during the first scheduling of the
process. As such, it is safe to use an unlocked test for the flag
within switch_to()./
Signed-off-by: Benjamin LaHaise <[email protected]>
diff --git a/arch/x86/include/asm/system.h b/arch/x86/include/asm/system.h
index 8e626ea..fa47b1e 100644
--- a/arch/x86/include/asm/system.h
+++ b/arch/x86/include/asm/system.h
@@ -96,15 +96,15 @@ do { \
"thread_return:\n\t" \
"movq %%gs:%P[pda_pcurrent],%%rsi\n\t" \
"movq %P[thread_info](%%rsi),%%r8\n\t" \
- LOCK_PREFIX "btr %[tif_fork],%P[ti_flags](%%r8)\n\t" \
"movq %%rax,%%rdi\n\t" \
- "jc ret_from_fork\n\t" \
+ "testl %[_tif_fork],%P[ti_flags](%%r8)\n\t" \
+ "jnz ret_from_fork\n\t" \
RESTORE_CONTEXT \
: "=a" (last) \
: [next] "S" (next), [prev] "D" (prev), \
[threadrsp] "i" (offsetof(struct task_struct, thread.sp)), \
[ti_flags] "i" (offsetof(struct thread_info, flags)), \
- [tif_fork] "i" (TIF_FORK), \
+ [_tif_fork] "i" (_TIF_FORK), \
[thread_info] "i" (offsetof(struct task_struct, stack)), \
[pda_pcurrent] "i" (offsetof(struct x8664_pda, pcurrent)) \
: "memory", "cc" __EXTRA_CLOBBER)
diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S
index e28c7a9..38dd374 100644
--- a/arch/x86/kernel/entry_64.S
+++ b/arch/x86/kernel/entry_64.S
@@ -408,6 +408,8 @@ END(save_paranoid)
ENTRY(ret_from_fork)
DEFAULT_FRAME
+ LOCK ; btr $TIF_FORK,TI_flags(%r8)
+
push kernel_eflags(%rip)
CFI_ADJUST_CFA_OFFSET 8
popf # reset kernel eflags
* Benjamin LaHaise <[email protected]> wrote:
> Here is a rediffed and test booted version of the patch against linus' git
> tree as of a couple of hours ago.
>
> The patch below removes an unnecessary locked instruction from
> switch_to(). TIF_FORK is only ever set in copy_thread() on initial
> process creation, and gets cleared during the first scheduling of the
> process. As such, it is safe to use an unlocked test for the flag
> within switch_to()./
>
> Signed-off-by: Benjamin LaHaise <[email protected]>
applied to tip/x86/asm, thanks Ben!
Ingo