MIME-Version: 1.0
In-Reply-To: <CA+55aFw+WVjwPBPKpnKcZUkqX=mYWv7A6M44uJpmWuzGU0zpqw@mail.gmail.com>
References: <20170221231216.y56gb62vkn5ewgea@treble> <20170222210548.GC8467@amd>
 <20170222212103.tigzbw5sfrwd7uwh@treble> <20170222224755.GA4310@amd>
 <20170222225614.4z4z24uz6l2iz6qm@treble> <20170222231808.hmr6ulbvfnrg2at7@treble>
 <20170223201039.GB5177@amd> <20170225050439.7dplheb6nyne4nkm@treble>
 <20170302234514.3qcqdozibcltkdai@treble> <20170306163807.GA20689@amd>
 <20170307173821.yknj5htr7plgdwxv@treble> <CA+55aFw+WVjwPBPKpnKcZUkqX=mYWv7A6M44uJpmWuzGU0zpqw@mail.gmail.com>
From: Andy Lutomirski <luto@amacapital.net>
Date: Tue, 7 Mar 2017 09:59:44 -0800
Message-ID: <CALCETrWHdra-Po0rNrHn-cV2G=AM2pWR45rRf4t2RZfQR8Lqjg@mail.gmail.com>
Subject: Re: v4.10: kernel stack frame pointer .. has bad value (null)
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>, Pavel Machek <pavel@ucw.cz>,
        kernel list <linux-kernel@vger.kernel.org>,
        Ingo Molnar <mingo@kernel.org>, Andrew Lutomirski <luto@kernel.org>,
        Borislav Petkov <bp@alien8.de>, Brian Gerst <brgerst@gmail.com>,
        Denys Vlasenko <dvlasenk@redhat.com>, Peter Anvin <hpa@zytor.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Thomas Gleixner <tglx@linutronix.de>
Content-Type: text/plain; charset=UTF-8
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1310
Lines: 33

On Tue, Mar 7, 2017 at 9:52 AM, Linus Torvalds
<torvalds@linux-foundation.org> wrote:
> On Tue, Mar 7, 2017 at 9:38 AM, Josh Poimboeuf <jpoimboe@redhat.com> wrote:
>>
>> So I'm thinking we should have -maccumulate-outgoing-args always enabled
>> on x86_32 just like we already do on x86_64.
>
> Ugh. I realize we have workarounds for bugs, but I think
> -maccumulate-outgoing-args is nasty. It just generates worse code by
> avoiding the much nicer push/pop sequences, afaik.
>
> On x86-64 it's not such a big deal, because we pass the first six
> arguments in registers anyway, so the arguments on the stack is a
> fairly unusual special case.
>
> But on x86-32, we only have three argument registers, so this
> braindamage is potentially worse.
>
> I guess we already do this in most situations due to the gcc bugs, but
> I do think it's sad that we would do it for our _own_ bugs too.
>

Is it our bug or a gcc bug?  I would have thought
-fno-omit-frame-pointer meant that the call-frame-to-return-address
offset should be constant and -fomit-frame-pointer meant "do
whatever".

Also, maybe I'm missing something, but does gcc's code even allow the
function to return sensibly?  It could do it by a nasty calculation
involving backing out the old esp from edi, but that seems quite
overcomplicated.

--Andy