[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20190911200818.GB9017@brightrain.aerifal.cx>
Date: Wed, 11 Sep 2019 16:08:18 -0400
From: Rich Felker <dalias@...c.org>
To: Florian Weimer <fweimer@...hat.com>
Cc: Carlos O'Donell <carlos@...hat.com>,
Mathieu Desnoyers <mathieu.desnoyers@...icios.com>,
Joseph Myers <joseph@...esourcery.com>,
Szabolcs Nagy <szabolcs.nagy@....com>,
libc-alpha@...rceware.org, Thomas Gleixner <tglx@...utronix.de>,
Ben Maurer <bmaurer@...com>,
Peter Zijlstra <peterz@...radead.org>,
"Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>,
Boqun Feng <boqun.feng@...il.com>,
Will Deacon <will.deacon@....com>,
Dave Watson <davejwatson@...com>, Paul Turner <pjt@...gle.com>,
linux-kernel@...r.kernel.org, linux-api@...r.kernel.org
Subject: Re: [PATCH glibc 2.31 1/5] glibc: Perform rseq(2) registration at C
startup and thread creation (v12)
On Wed, Sep 11, 2019 at 09:54:23PM +0200, Florian Weimer wrote:
> * Carlos O'Donell:
>
> > On 9/11/19 3:08 PM, Florian Weimer wrote:
> >> * Carlos O'Donell:
> >>
> >>> It would be easier to merge the patch set if it were just an unconditional
> >>> registration like we do for set_robust_list().
> >>
> >> Note that this depends on the in-tree system call numbers list, which I
> >> still need to finish according to Joseph's specifications.
> >
> > Which work is this? Do you have a URL reference to WIP?
>
> <https://sourceware.org/ml/libc-alpha/2019-05/msg00630.html>
> <https://sourceware.org/ml/libc-alpha/2019-06/msg00015.html>
>
> I think realistically this is needed for the Y2038 work as well if we
> want to support building glibc with older kernel headers. “glibc 2..31
> will have Y2038 support and rseq support, but only if it runs on a
> current and it happens to have been built against sufficiently recent
> kernel headers” is a bit difficult to explain. The current kernel part
> is easy enough to understand, but the impact of the kernel headers on
> the feature set has always been tough to explain. Especially if you
> factor in vendor kernels with system call backports.
I'm in favor of in-tree syscall numbers list. If you don't want O(n)
per-arch work, though, you could just define the 'base number' for
each arch and use the fact that all the new syscalls share a common
numbering (i.e. base + constant depending only on syscall). I think
including the list with glibc is more robust though, and would
eliminate the need to check for definitions of older (pre-unification)
syscalls glibc wants to use.
Rich
Powered by blists - more mailing lists