Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Wed, 5 Jul 2023 23:31:13 +0200
From:   Peter Zijlstra <peterz@infradead.org>
To:     Kees Cook <keescook@chromium.org>
Cc:     Jann Horn <jannh@google.com>, Kees Cook <kees@kernel.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Linux LLVM <llvm@lists.linux.dev>,
        the arch/x86 maintainers <x86@kernel.org>
Subject: Re: CONFIG_UBSAN_TRAP #UD error message on x86 [was: Re:
 [CRASH][BISECTED] 6.4.1 crash in boot]
Message-ID: <20230705213113.GC2813335@hirez.programming.kicks-ass.net>
References: <202307022230.6121D6896@keescook>
 <ed3e4315-b149-2f9e-70d4-45d7f83b9922@alu.unizg.hr>
 <202307031149.823F9A3@keescook>
 <202307031606.3FD9D0D@keescook>
 <fbf4103b-6471-1f06-34a8-4aa227f5e091@alu.unizg.hr>
 <3CDD1188-7B2A-4D53-9B8F-C07BC39844E9@kernel.org>
 <19d0e63c-db47-8f02-2f79-0923f675898c@alu.unizg.hr>
 <DA3FEB08-DF39-406B-89CC-9076CFCF597A@kernel.org>
 <CAG48ez0kmSqS_WJ1=neKpxKmMMkPYgNHfBD6ZSif_EWnso84wQ@mail.gmail.com>
 <202307051405.E746464@keescook>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <202307051405.E746464@keescook>
Precedence: bulk

On Wed, Jul 05, 2023 at 02:08:09PM -0700, Kees Cook wrote:

> > Even just a "WARNING: CONFIG_UBSAN_TRAP active, #UD might be caused by
> > that" on every #UD that does not come from a known BUG() location or
> > such might be better than nothing...
> 
> I've considered it, but usually CONFIG_UBSAN_TRAP isn't accidentally
> set. Also, the crash info is something we can get help from on the
> compiler side, to mark up where the traps are, similar to what we do
> with KCFI, but it hasn't happened yet for x86. For example, arm64
> already encodes the details in the trap instruction itself:
> https://git.kernel.org/linus/25b84002afb9dc9a91a7ea67166879c13ad82422

Right, so you could easily use a different #UD instruction that has an
immediate, something like:

  0f b9 40 ff          ud1    -0x1(%rax),%rax

or even:

  0f b9 80 00 ff ff ff         ud1    -0x100(%rax),%rax

if you need a 32bit value.

It shouldn't be hard to fix up the #UD handler to decode the instruction
and obtain the displacement for a clue.

Typically we use ud2 because it's the smallest #UD instruction (2 bytes)
and that's enough, but if you want to provide additional clues, there's
options...