Re: [PATCH] kernel/entry: Remove some redundancy checks on syscall works

Re: [PATCH] kernel/entry: Remove some redundancy checks on syscall works

[Khalid Ali]

On Wed, Jun 11 2025 at 11:43, Khalid Ali wrote:
> > There is a redundant checks of thread syscall work.

> Not really.
>
> >  After we read thread syscall work we are checking the work bits using
>
> We are doing nothing. Please write your changelogs in imperative mood
> and do not try to impersonate code.

Sorry, i guess my english sucks.

> > SYSCALL_WORK_ENTER and SYSCALL_WORK_EXIT on syscall entry and exit
> > respectively, and at the same time syscall_trace_enter() and
> > syscall_exit_work() checking bits one by one, the bits we already checked.
> > This is redundancy. So either we need to check the work bits one by one as I
> > did, or check as whole. On my prespective, i think the way code is
> > implemented now checking work bits one by one is simpler and gives us
> > more granular control.

> That's just wrong and absolutely not redundant. Care to look at the
> definition of SYSCALL_WORK_ENTER:
>
> #define SYSCALL_WORK_ENTER	(SYSCALL_WORK_SECCOMP |			\
>				 SYSCALL_WORK_SYSCALL_TRACEPOINT |	\
>				 SYSCALL_WORK_SYSCALL_TRACE |		\
>				 SYSCALL_WORK_SYSCALL_EMU |		\
>				 SYSCALL_WORK_SYSCALL_AUDIT |		\
>				 SYSCALL_WORK_SYSCALL_USER_DISPATCH |	\
>				 ARCH_SYSCALL_WORK_ENTER)
>
> So this initial check avoids:
>
>    1) Doing an unconditional out of line call
>
>    2) Checking bit for bit to figure out that there is none set.
>
> Same applies for SYSCALL_WORK_EXIT.
>
> Your change neither makes anything simpler nor provides more granular
> control.
>
> All it does is adding overhead and therefore guaranteed to introduce a
> performance regression.
>
> Not going to happen.
>
> Thanks,
>
>        tglx
Thanks, for the response and noted all your points, however i spotted some minor details also:

First if we are talking about performance then we may need likely() on SYSCALL_WORK_ENTER since 
the probability of condition evaluating as true is very high.

Second syscall_enter_audit() missing SYSCALL_WORK_SYSCALL_AUDIT	evaluation, aren't we supposed to call
it only if SYSCALL_WORK_SYSCALL_AUDIT is set?

Should i create another patch fixing these two points, of course if i am right?

Thanks, Khalid Ali

Re: [PATCH] kernel/entry: Remove some redundancy checks on syscall works

[Thomas Gleixner]

Can you please reply to the mail you received, so that there are proper
In-Reply-To and References tags in the mail, which are required for mail
threading?

I almost missed your replies because they ended up as single mail
threads without reference somewhere in my endless mail pile.

On Fri, Jun 13 2025 at 20:28, Khalid Ali wrote:

> First if we are talking about performance then we may need likely() on
> SYSCALL_WORK_ENTER since the probability of condition evaluating as
> true is very high.

That depends on the system configuration scenario and the likely() has
been omitted on purpose.

> Second syscall_enter_audit() missing SYSCALL_WORK_SYSCALL_AUDIT
> evaluation, aren't we supposed to call it only if
> SYSCALL_WORK_SYSCALL_AUDIT is set?

That's redundant as syscall_enter_audit() checks for a valid audit
context already. Both are valid indicators and go in lockstep. So it
might be arguable that evaluating the work bit is cheaper than the
context check, but I doubt it's measurable.

Thanks,

        tglx

Re: [PATCH] include/linux: Fix outdated comment on entry-common.h

[Khalid Ali]

> That's redundant as syscall_enter_audit() checks for a valid audit
> context already. Both are valid indicators and go in lockstep. So it
> might be arguable that evaluating the work bit is cheaper than the
> context check, but I doubt it's measurable.

This will also increase readability and make things more consistent. So i think 
we should use  SYSCALL_WORK_SYSCALL_AUDIT bit to check if we need to do auditing,
instead of checking audit context, on both entry and exit.

So should i do with another patch, what i am verifying is if we really agree with one point.

Thanks
Khalid Ali

Re: [PATCH] kernel/entry: Remove some redundancy checks on syscall

[Khalid Ali]

Since there were no much objection here i think we agreed with that, so i sent v2 
for this one. Could you please redirect our conversasion there. This patch was wrong 
so i took your last suggestion and improved it.

Thanks for your patience,
Khalid Ali