[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <alpine.DEB.2.22.394.2004090945110.31491@tglase-nb.lan.tarent.de>
Date: Thu, 9 Apr 2020 09:47:51 +0200 (CEST)
From: Thorsten Glaser <t.glaser@...ent.de>
To: Andy Lutomirski <luto@...nel.org>
cc: X86 ML <x86@...nel.org>, LKML <linux-kernel@...r.kernel.org>,
Thomas Gleixner <tglx@...utronix.de>, 954294@...s.debian.org
Subject: Re: __X32_SYSCALL_BIT being defined as UL constant breaks
userspace
On Wed, 8 Apr 2020, Andy Lutomirski wrote:
> One might reasonably ask whether it makes sense for syscall nrs to be
> signed at all.
It doesn’t, but it’s probably this way for hysteric raisins.
> But regardless, this breaks userspace and we should fix it. I can
> whip up a patch to split it into X32_SYSCALL_BIT (unsigned long) and
> __X32_SYSCALL_BIT (uapi, int). Thomas, etc, does this seem
This would help with the issue, thanks.
> reasonable? (For those not following all the machinations, this
> change caused some userspace build failures in libseccomp and/or
> systemd for reasons that are vaguely silly.)
Not very silly:
printf("%d\n", __NR_mmap);
That with -Wformat and -Werror (or -Werror=format) rightfully warns,
as the compiler cannot, on x32 (where int=long), detect that, on
architectures where int≠long the constant is int.
Even worse, switching userspace to unsigned long globally would
completely hose this on LP64 architectures.
So this is, in my opinion, completely justified. (Disclaimer: I’m
not affiliated with systemd and not even running it except udev.)
bye,
//mirabilos
--
tarent solutions GmbH
Rochusstraße 2-4, D-53123 Bonn • http://www.tarent.de/
Tel: +49 228 54881-393 • Fax: +49 228 54881-235
HRB 5168 (AG Bonn) • USt-ID (VAT): DE122264941
Geschäftsführer: Dr. Stefan Barth, Kai Ebenrett, Boris Esser, Alexander Steeg
Powered by blists - more mailing lists