linux-kernel - Re: Simplify load_unaligned_zeropad() (was Re: [GIT PULL] Ceph updates for 5.20-rc1)

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <YvycqWVs5JbBR1D+@worktop.programming.kicks-ass.net>
Date:   Wed, 17 Aug 2022 09:45:45 +0200
From:   Peter Zijlstra <peterz@...radead.org>
To:     Linus Torvalds <torvalds@...ux-foundation.org>
Cc:     Al Viro <viro@...iv.linux.org.uk>,
        Nathan Chancellor <nathan@...nel.org>,
        Nick Desaulniers <ndesaulniers@...gle.com>,
        Jeff Layton <jlayton@...nel.org>,
        Ilya Dryomov <idryomov@...il.com>, ceph-devel@...r.kernel.org,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Matthew Wilcox <willy@...radead.org>,
        clang-built-linux <llvm@...ts.linux.dev>
Subject: Re: Simplify load_unaligned_zeropad() (was Re: [GIT PULL] Ceph
 updates for 5.20-rc1)

On Tue, Aug 16, 2022 at 10:57:45AM -0700, Linus Torvalds wrote:

> > > +     if (insn_decode(&insn, (void *) regs->ip, len, INSN_MODE_KERN))
> > > +             return false;
> >
> > We have insn_decode_kernel() for exactly this (very) common case.
> 
> I did that originally, and then I undid it in disgust, because that
> interface is too simple.
> 
> In particular, it just uses MAX_INSN_SIZE blindly. Which I didn't want
> to do when I actually had the instruction size.
> 
> Yes, yes, I also check the decode size after-the-fact, but I didn't
> want the decoder to even look at the invalid bytes.
> 
> This exception case is about the data being at the end of the page, I
> wanted the fixup to be aware of code being at the end of a page too.

I don't want to argue this point too much; but I will anyway :-)

IMO if the decoder ends up out of bounds its a decoder bug either way
around. That is, we *know* there is a full instruction at the given IP
because we got into this exception path.

( it would be possible to add further constraints on trapnr )

Irrespective of the length constraint given to the decoder, it should
not decode/access things past this instruction (without being careful
about it).

Anyway, I'm fine with the patch as you have it.