Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
From:   Thomas Gleixner <tglx@linutronix.de>
To:     Andy Lutomirski <luto@kernel.org>,
        Thorsten Glaser <t.glaser@tarent.de>, X86 ML <x86@kernel.org>,
        LKML <linux-kernel@vger.kernel.org>
Cc:     Andy Lutomirski <luto@kernel.org>, 954294@bugs.debian.org
Subject: Re: __X32_SYSCALL_BIT being defined as UL constant breaks userspace
In-Reply-To: <CALCETrUkC0mfbo44+Jb7p6f19S+nH5VAEZAwTaYApbNqy9Fj-w@mail.gmail.com>
References: <alpine.DEB.2.22.394.2004081607050.27197@tglase.lan.tarent.de> <CALCETrUkC0mfbo44+Jb7p6f19S+nH5VAEZAwTaYApbNqy9Fj-w@mail.gmail.com>
Date:   Thu, 09 Apr 2020 09:47:17 +0200
Message-ID: <87wo6pp0ca.fsf@nanos.tec.linutronix.de>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 8BIT
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Andy Lutomirski <luto@kernel.org> writes:
> On Wed, Apr 8, 2020 at 7:34 AM Thorsten Glaser <t.glaser@tarent.de> wrote:
>> asm/unistd_x32.h:#define __NR_mmap (__X32_SYSCALL_BIT + 9)
>>
>> This construct is, thankfully, still usable in something like
>>         #if (__NR_mmap > __NR_somethingelse)
>> but as __X32_SYSCALL_BIT is no longer int its type also isn’t.
>>
>> Therefore I ask you to revert this change, bringing x32 closer
>> to all other architectures.
>>
>
> One might reasonably ask whether it makes sense for syscall nrs to be
> signed at all.
>
> 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
> 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.)

Yes.