From: David Laight <david.laight.linux@gmail•com>
To: Mathias Stearn <mathias@mongodb•com>
Cc: Thomas Gleixner <tglx@linutronix•de>,
Dmitry Vyukov <dvyukov@google•com>,
Jinjie Ruan <ruanjinjie@huawei•com>,
linux-man@vger•kernel.org, Mark Rutland <mark.rutland@arm•com>,
Mathieu Desnoyers <mathieu.desnoyers@efficios•com>,
Catalin Marinas <catalin.marinas@arm•com>,
Will Deacon <will@kernel•org>, Boqun Feng <boqun.feng@gmail•com>,
"Paul E. McKenney" <paulmck@kernel•org>,
Chris Kennelly <ckennelly@google•com>,
regressions@lists•linux.dev, linux-kernel@vger•kernel.org,
linux-arm-kernel@lists•infradead.org,
Peter Zijlstra <peterz@infradead•org>,
Ingo Molnar <mingo@kernel•org>,
Blake Oler <blake.oler@mongodb•com>
Subject: Re: [REGRESSION] rseq: refactoring in v6.19 broke everyone on arm64 and tcmalloc everywhere
Date: Thu, 23 Apr 2026 13:24:46 +0100 [thread overview]
Message-ID: <20260423132446.70478a78@pumpkin> (raw)
In-Reply-To: <CAHnCjA0UBNXfjHw=Y34OrAyGRNUtVF+zWd3ugyX6pd_mCk8K9w@mail.gmail.com>
On Thu, 23 Apr 2026 12:51:22 +0200
Mathias Stearn <mathias@mongodb•com> wrote:
> On Thu, Apr 23, 2026 at 12:39 PM Thomas Gleixner <tglx@linutronix•de> wrote:
> > The kernel clears rseq_cs reliably when user space was interrupted and:
> >
> > the task was preempted
> > or
> > the return from interrupt delivers a signal
> >
> > If the task invoked a syscall then there is absolutely no reason to do
> > either of this because syscalls from within a critical section are a
> > bug and catched when enabling rseq debugging.
> >
> > The original code did this along with unconditionally updating CPU/MMCID
> > which resulted in ~15% performance regression on a syscall heavy
> > database benchmark once glibc started to register rseq.
>
> Just to be clear TCMalloc does not need either rseq_cs to be cleared
> or cpu_id_start to be written to on syscalls because it doesn't do
> syscalls from critical sections. It will actually benefit (slightly)
> from not updating cpu_id_start on syscalls.
>
> It is specifically in the cases where an rseq would need to be aborted
> (preemption, signals, migration, and membarrier IPI with the rseq
> flag) that TCMalloc relies on cpu_id_start being written. It does rely
> on that write even when not inside the critical section, because it
> effectively uses that to detect if there were any would-cause-abort
> events in between two critical sections. But since it leaves the
> rseq_cs pointer non-null between critical sections, so you dont need
> to add _any_ overhead for programs that never make use of rseq after
> registration, or add any overhead to syscalls even for those who do.
>
That sounds like one long rseq sequence where the 'restart' path
detects that some of the operations have already been done.
David
next prev parent reply other threads:[~2026-04-23 12:25 UTC|newest]
Thread overview: 56+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <CAHnCjA25b+nO2n5CeifknSKHssJpPrjnf+dtr7UgzRw4Zgu=oA@mail.gmail.com>
2026-04-22 12:56 ` [REGRESSION] rseq: refactoring in v6.19 broke everyone on arm64 and tcmalloc everywhere Peter Zijlstra
2026-04-22 13:13 ` Peter Zijlstra
2026-04-23 10:38 ` Mathias Stearn
[not found] ` <CAHnCjA2fa+dP1+yCYNQrTXQaW-JdtfMj7wMikwMeeCRg-3NhiA@mail.gmail.com>
2026-04-23 11:48 ` Thomas Gleixner
2026-04-23 12:11 ` Mathias Stearn
2026-04-23 17:19 ` Thomas Gleixner
2026-04-23 17:38 ` Chris Kennelly
2026-04-23 17:47 ` Mathieu Desnoyers
2026-04-23 19:39 ` Thomas Gleixner
2026-04-23 17:41 ` Linus Torvalds
2026-04-23 18:35 ` Mathias Stearn
2026-04-23 18:53 ` Mark Rutland
2026-04-23 21:03 ` Thomas Gleixner
2026-04-23 21:28 ` Linus Torvalds
2026-04-23 23:08 ` Linus Torvalds
2026-04-27 7:06 ` Florian Weimer
2026-04-27 16:12 ` Linus Torvalds
2026-04-22 13:09 ` Mark Rutland
2026-04-22 17:49 ` Thomas Gleixner
2026-04-22 18:11 ` Mark Rutland
2026-04-22 19:47 ` Thomas Gleixner
2026-04-23 1:48 ` Jinjie Ruan
2026-04-23 5:53 ` Dmitry Vyukov
2026-04-23 10:39 ` Thomas Gleixner
2026-04-23 10:51 ` Mathias Stearn
2026-04-23 12:24 ` David Laight [this message]
2026-04-23 19:31 ` Thomas Gleixner
2026-04-24 7:56 ` Dmitry Vyukov
2026-04-24 8:32 ` Mathias Stearn
2026-04-24 9:30 ` Dmitry Vyukov
2026-04-24 14:16 ` Thomas Gleixner
2026-04-24 15:03 ` Peter Zijlstra
2026-04-24 19:44 ` Thomas Gleixner
2026-04-26 22:04 ` Thomas Gleixner
2026-04-27 7:40 ` Florian Weimer
2026-04-27 11:03 ` Thomas Gleixner
2026-04-27 18:35 ` Mathieu Desnoyers
2026-04-27 21:06 ` Thomas Gleixner
2026-04-28 6:11 ` Dmitry Vyukov
2026-04-28 8:07 ` Thomas Gleixner
2026-04-28 8:18 ` Thomas Gleixner
[not found] ` <CACT4Y+b_RH2eZMuh1YUyqnoK-5KUpdWW4z1q2ZQWkY_GcBqmNw@mail.gmail.com>
[not found] ` <CAHnCjA2sCwOumOjWm=wW=Kj0C83KVW5zS+51=9=YSeAzuEaVQA@mail.gmail.com>
2026-04-28 15:46 ` Thomas Gleixner
2026-04-28 7:39 ` Peter Zijlstra
2026-04-28 8:13 ` Peter Zijlstra
2026-04-28 8:51 ` Thomas Gleixner
2026-04-28 8:03 ` Peter Zijlstra
2026-04-28 8:36 ` Thomas Gleixner
2026-04-23 12:11 ` Alejandro Colomar
2026-04-23 12:54 ` Mathieu Desnoyers
2026-04-23 12:29 ` Mathieu Desnoyers
2026-04-23 12:36 ` Dmitry Vyukov
2026-04-23 12:53 ` Mathieu Desnoyers
2026-04-23 12:58 ` Dmitry Vyukov
2026-04-24 16:45 ` [PATCH] arm64/entry: Fix arm64-specific rseq brokenness (was: Re: [REGRESSION] rseq: refactoring in v6.19 broke everyone on arm64) " Mark Rutland
2026-04-28 1:39 ` [PATCH] arm64/entry: Fix arm64-specific rseq brokenness Jinjie Ruan
2026-04-28 13:40 ` Mark Rutland
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20260423132446.70478a78@pumpkin \
--to=david.laight.linux@gmail$(echo .)com \
--cc=blake.oler@mongodb$(echo .)com \
--cc=boqun.feng@gmail$(echo .)com \
--cc=catalin.marinas@arm$(echo .)com \
--cc=ckennelly@google$(echo .)com \
--cc=dvyukov@google$(echo .)com \
--cc=linux-arm-kernel@lists$(echo .)infradead.org \
--cc=linux-kernel@vger$(echo .)kernel.org \
--cc=linux-man@vger$(echo .)kernel.org \
--cc=mark.rutland@arm$(echo .)com \
--cc=mathias@mongodb$(echo .)com \
--cc=mathieu.desnoyers@efficios$(echo .)com \
--cc=mingo@kernel$(echo .)org \
--cc=paulmck@kernel$(echo .)org \
--cc=peterz@infradead$(echo .)org \
--cc=regressions@lists$(echo .)linux.dev \
--cc=ruanjinjie@huawei$(echo .)com \
--cc=tglx@linutronix$(echo .)de \
--cc=will@kernel$(echo .)org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox