netdev - Re: UBSAN: object-size-mismatch in wg

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]

Date:   Mon, 11 Jan 2021 13:14:55 -0500
From:   Jeffrey Walton <noloader@...il.com>
To:     Dmitry Vyukov <dvyukov@...gle.com>
Cc:     Netdev <netdev@...r.kernel.org>,
        syzkaller-bugs <syzkaller-bugs@...glegroups.com>,
        WireGuard mailing list <wireguard@...ts.zx2c4.com>
Subject: Re: UBSAN: object-size-mismatch in wg_xmit

On Mon, Jan 11, 2021 at 12:58 PM Dmitry Vyukov <dvyukov@...gle.com> wrote:
>
> On Mon, Jan 11, 2021 at 6:35 PM Jeffrey Walton <noloader@...il.com> wrote:
> >
> > On Mon, Jan 11, 2021 at 12:20 PM Dmitry Vyukov <dvyukov@...gle.com> wrote:
> > > ...
> > > FTR, I've disabled the following UBSAN configs:
> > > UBSAN_MISC
> > > UBSAN_DIV_ZERO
> > > UBSAN_BOOL
> > > UBSAN_OBJECT_SIZE
> > > UBSAN_SIGNED_OVERFLOW
> > > UBSAN_UNSIGNED_OVERFLOW
> > > UBSAN_ENUM
> > > UBSAN_ALIGNMENT
> > > UBSAN_UNREACHABLE
> > >
> > > Only these are enabled now:
> > > UBSAN_BOUNDS
> > > UBSAN_SHIFT
> > >
> > > This is commit:
> > > https://github.com/google/syzkaller/commit/2c1f2513486f21d26b1942ce77ffc782677fbf4e
> >
> > I think the commit cut too deep.
> >
> > The overflows are important if folks are building with compilers other than GCC.
> >
> > The aligned data accesses are important on platforms like MIPS64 and Sparc64.
> >
> > Object size is important because it catches destination buffer overflows.
> >
> > I don't know what's in miscellaneous. There may be something useful in there.
>
> Hi Jeff,
>
> See the commit for reasons why each of these is disabled.
> E.g. object size, somebody first needs to fix bugs like this one.
> While things like skbuff have these UBs on trivial workloads, there is
> no point in involving fuzzing and making it crash on this trivial bug
> all the time and stopping doing any other kernel testing as the
> result.

Going off-topic a bit, what would you suggest for UBSAN_OBJECT_SIZE?

It seems to me object size checking is being conflated with object
type. It seems to me they need to be split: UBSAN_OBJECT_SIZE for
actual object sizes, and UBSAN_OBJECT_TYPE for the casts.

I still have a bitter taste in my mouth from
https://www.cvedetails.com/bugtraq-bid/57602/libupnp-Multiple-Buffer-Overflow-Vulnerabilities.html.
I hate to see buffer checks go away. (And I realize the kernel folks
are more skilled than the guy who wrote libupnp).

Jeff