linux-kernel - Re: [PATCH v2 8/8] powerpc: Enable Generic Entry/Exit for syscalls.

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <lv6iie4ykgxbwemnhrwsao3qc6oq52lugk7rkjcvh26cubcp55@rwudr7xpnlif>
Date: Tue, 16 Dec 2025 00:02:15 +0530
From: Mukesh Kumar Chaurasiya <mchauras@...ux.ibm.com>
To: Segher Boessenkool <segher@...nel.crashing.org>
Cc: Mukesh Kumar Chaurasiya <mkchauras@...ux.ibm.com>, maddy@...ux.ibm.com,
        mpe@...erman.id.au, npiggin@...il.com, christophe.leroy@...roup.eu,
        oleg@...hat.com, kees@...nel.org, luto@...capital.net,
        wad@...omium.org, thuth@...hat.com, sshegde@...ux.ibm.com,
        charlie@...osinc.com, macro@...am.me.uk, akpm@...ux-foundation.org,
        ldv@...ace.io, deller@....de, ankur.a.arora@...cle.com,
        tglx@...utronix.de, thomas.weissschuh@...utronix.de,
        peterz@...radead.org, menglong8.dong@...il.com, bigeasy@...utronix.de,
        namcao@...utronix.de, kan.liang@...ux.intel.com, mingo@...nel.org,
        atrajeev@...ux.vnet.ibm.com, mark.barnett@....com,
        linuxppc-dev@...ts.ozlabs.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 8/8] powerpc: Enable Generic Entry/Exit for syscalls.

On Sun, Dec 14, 2025 at 10:20:12AM -0600, Segher Boessenkool wrote:
> Hi!
> 
> On Sun, Dec 14, 2025 at 06:32:44PM +0530, Mukesh Kumar Chaurasiya wrote:
> > 
> > | Test            | With Patch | Without Patch | % Change |
> > | --------------- | ---------- | ------------- | -------- |
> > | fork usec/op    | 833.986    | 814.809       | +2.35%   |
> 
> What causes this regression, did you investigate?  Maybe there is
> something simple you can do to avoid this degradation :-)  All other
> numbers look just fine :-)
> 
Hey,

I ran this multiple times and took the average this time, below are the
results:
===========================================
Without patch
===========================================
╰─❯ perf bench syscall fork 
# Running 'syscall/fork' benchmark:
# Executed 10,000 fork() calls
     Total time: 8.514 [sec]

     851.415300 usecs/op
          1,174 ops/sec

╰─❯ perf bench syscall fork
# Running 'syscall/fork' benchmark:
# Executed 10,000 fork() calls
     Total time: 8.572 [sec]

     857.293600 usecs/op
          1,166 ops/sec

╰─❯ perf bench syscall fork 
# Running 'syscall/fork' benchmark:
# Executed 10,000 fork() calls
     Total time: 8.695 [sec]

     869.536500 usecs/op
          1,150 ops/sec
===========================================
With patch
===========================================
╰─❯ perf bench syscall fork 
# Running 'syscall/fork' benchmark:
# Executed 10,000 fork() calls
     Total time: 8.482 [sec]

     848.241300 usecs/op
          1,178 ops/sec

╰─❯ perf bench syscall fork 
# Running 'syscall/fork' benchmark:
# Executed 10,000 fork() calls
     Total time: 8.623 [sec]

     862.389000 usecs/op
          1,159 ops/sec

╰─❯ perf bench syscall fork 
# Running 'syscall/fork' benchmark:
# Executed 10,000 fork() calls
     Total time: 8.530 [sec]

     853.037200 usecs/op
          1,172 ops/sec
===========================================
Average:
===========================================
With Patch:
854.4964 usecs/op
1169 ops/sec

Without patch:
859.4151 usecs/op
1163 ops/sec

That's ~0.5% improvement when i take average through the runs.
This we can ignore as a standard deviation and consider that there are
no regression for these.

Regards,
Mukesh

> > | Test            | With Patch | Without Patch | % Change |
> > | --------------- | ---------- | ------------- | -------- |
> > | fork ops/sec    | 1,199      | 1,227         | -2.28%   |
> 
> (Same thing seen from another side)
> 
> 
> Segher