[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAA2zVHrsJTX2sY3ZcMY7eAY11go4Qf+WJKHSVqzpV0qB2-R0LA@mail.gmail.com>
Date: Fri, 24 Jul 2026 18:06:45 -0400
From: James Y Knight <jyknight@...gle.com>
To: musl@...ts.openwall.com
Cc: Szabolcs Nagy <nsz@...t70.net>, Jake Adamson <jakeadamson@...gle.com>,
Peter Oskolkov <posk@...gle.com>, Peter Zijlstra <peterz@...radead.org>, mathieu.desnoyers@...icios.com
Subject: Re: Re: Question: number of arguments of musl's membarrier
does not match Linux syscall
On Fri, Jul 24, 2026 at 7:46 AM Nick Desaulniers <ndesaulniers@...gle.com>
wrote:
> > google broke it (silently)
> >
> >
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=2a36ab717e8fe678d98f81c14a0b124712719840
>
> Oops, did it get reported, yet?
>
> I didn't see any follow ups one the original LKML thread.
> https://lore.kernel.org/all/20200923233618.2572849-1-posk@google.com/T/#u
>
> Patch from 2020...perhaps too late to do anything about it. Though shall
> not break userspace..
But it's not a breakage to add another arg to a syscall which is only
looked at when a newly-introduced flag is passed in the existing arg.
While it might be nice if the kernel expressed opinions as to how syscalls
ought to be wrapped with userspace C functions, there's no actual problem
with that _as a syscall change_.
E.g., libc implementations could, if they wished, simply add a brand-new
wrapper function, membarrier3(int, unsigned, int), which still invokes the
same syscall number as the 2-arg "membarrier(int, unsigned)". It's really
just a matter of documentation and agreement between the various userspace
libc implementations here as to what they wish a wrapper function to look
like, not any issue on the kernel side.
Content of type "text/html" skipped
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
