2021-12-10 23:02:25

by Chang S. Bae

[permalink] [raw]
Subject: [PATCH v2] signal: Skip the altstack update when not needed

== Background ==

Support for large, "dynamic" fpstates was recently merged. This
included code to ensure that sigaltstacks are sufficiently sized for
these large states. A new lock was added to remove races between
enabling large features and setting up sigaltstacks.

== Problem ==

The new lock (sigaltstack_lock()) is acquired in the sigreturn path
before restoring the old sigaltstack. Unfortunately, contention on the
new lock causes a measurable signal handling performance regression [1].
However, the common case is that no *changes* are made to the
sigaltstack state at sigreturn.

== Solution ==

do_sigaltstack() acquires sigaltstack_lock() and is used for both
sys_sigaltstack() and restoring the sigaltstack in sys_sigreturn().
Check for changes to the sigaltstack before taking the lock. If no
changes were made, return before acquiring the lock.

This removes lock contention from the common-case sigreturn path.

Reported-by: kernel test robot <[email protected]>
[1] https://lore.kernel.org/lkml/20211207012128.GA16074@xsang-OptiPlex-9020/
Fixes: 3aac3ebea08f ("x86/signal: Implement sigaltstack size validation")
Signed-off-by: Chang S. Bae <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Dave Hansen <[email protected]>
---
Changes from v1:
* Rewrite the changelog. (Dave Hansen)
* Add the code comment. (Dave Hansen)
---
kernel/signal.c | 9 +++++++++
1 file changed, 9 insertions(+)

diff --git a/kernel/signal.c b/kernel/signal.c
index a629b11bf3e0..dfcee3888b00 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -4185,6 +4185,15 @@ do_sigaltstack (const stack_t *ss, stack_t *oss, unsigned long sp,
ss_mode != 0))
return -EINVAL;

+ /*
+ * Return before taking any locks if no actual
+ * sigaltstack changes were requested.
+ */
+ if (t->sas_ss_sp == (unsigned long)ss_sp &&
+ t->sas_ss_size == ss_size &&
+ t->sas_ss_flags == ss_flags)
+ return 0;
+
sigaltstack_lock();
if (ss_mode == SS_DISABLE) {
ss_size = 0;
--
2.17.1



2021-12-13 21:25:41

by Thomas Gleixner

[permalink] [raw]
Subject: Re: [PATCH v2] signal: Skip the altstack update when not needed

On Fri, Dec 10 2021 at 14:55, Chang S. Bae wrote:
> do_sigaltstack() acquires sigaltstack_lock() and is used for both
> sys_sigaltstack() and restoring the sigaltstack in sys_sigreturn().
> Check for changes to the sigaltstack before taking the lock. If no
> changes were made, return before acquiring the lock.
>
> This removes lock contention from the common-case sigreturn path.
>
> Reported-by: kernel test robot <[email protected]>
> [1] https://lore.kernel.org/lkml/20211207012128.GA16074@xsang-OptiPlex-9020/
> Fixes: 3aac3ebea08f ("x86/signal: Implement sigaltstack size validation")
> Signed-off-by: Chang S. Bae <[email protected]>
> Cc: Thomas Gleixner <[email protected]>

Reviewed-by: Thomas Gleixner <[email protected]>

2021-12-14 01:01:38

by Chang S. Bae

[permalink] [raw]
Subject: Re: [PATCH v2] signal: Skip the altstack update when not needed

On Dec 13, 2021, at 13:25, Thomas Gleixner <[email protected]> wrote:
>
> Reviewed-by: Thomas Gleixner <[email protected]>

Thanks!
Chang

Subject: [tip: sched/urgent] signal: Skip the altstack update when not needed

The following commit has been merged into the sched/urgent branch of tip:

Commit-ID: 6c3118c32129b4197999a8928ba776bcabd0f5c4
Gitweb: https://git.kernel.org/tip/6c3118c32129b4197999a8928ba776bcabd0f5c4
Author: Chang S. Bae <[email protected]>
AuthorDate: Fri, 10 Dec 2021 14:55:03 -08:00
Committer: Dave Hansen <[email protected]>
CommitterDate: Tue, 14 Dec 2021 13:08:36 -08:00

signal: Skip the altstack update when not needed

== Background ==

Support for large, "dynamic" fpstates was recently merged. This
included code to ensure that sigaltstacks are sufficiently sized for
these large states. A new lock was added to remove races between
enabling large features and setting up sigaltstacks.

== Problem ==

The new lock (sigaltstack_lock()) is acquired in the sigreturn path
before restoring the old sigaltstack. Unfortunately, contention on the
new lock causes a measurable signal handling performance regression [1].
However, the common case is that no *changes* are made to the
sigaltstack state at sigreturn.

== Solution ==

do_sigaltstack() acquires sigaltstack_lock() and is used for both
sys_sigaltstack() and restoring the sigaltstack in sys_sigreturn().
Check for changes to the sigaltstack before taking the lock. If no
changes were made, return before acquiring the lock.

This removes lock contention from the common-case sigreturn path.

[1] https://lore.kernel.org/lkml/20211207012128.GA16074@xsang-OptiPlex-9020/

Fixes: 3aac3ebea08f ("x86/signal: Implement sigaltstack size validation")
Reported-by: kernel test robot <[email protected]>
Signed-off-by: Chang S. Bae <[email protected]>
Signed-off-by: Dave Hansen <[email protected]>
Reviewed-by: Thomas Gleixner <[email protected]>
Link: https://lkml.kernel.org/r/[email protected]
---
kernel/signal.c | 9 +++++++++
1 file changed, 9 insertions(+)

diff --git a/kernel/signal.c b/kernel/signal.c
index a629b11..dfcee38 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -4185,6 +4185,15 @@ do_sigaltstack (const stack_t *ss, stack_t *oss, unsigned long sp,
ss_mode != 0))
return -EINVAL;

+ /*
+ * Return before taking any locks if no actual
+ * sigaltstack changes were requested.
+ */
+ if (t->sas_ss_sp == (unsigned long)ss_sp &&
+ t->sas_ss_size == ss_size &&
+ t->sas_ss_flags == ss_flags)
+ return 0;
+
sigaltstack_lock();
if (ss_mode == SS_DISABLE) {
ss_size = 0;