Bug#782561: please backport commit ccfe8c3f7e52 from upstream
Hi Ben,
On Sun, Apr 19, 2015 at 03:15:35PM +0100, Ben Hutchings wrote:
> However, the upstream developers tell me that in this particular case
> the buffers will always have some extra padding that makes the
> overflows harmless in practice.
>
> Have you actually tested that this commit fixes the crashes you've
> seen?
Yes, although not on 3.16. I had a reproducer involving specific DNS
traffic which crashed my laptop reliably on 3.17, 3.18 and 3.19, and
the only workaround was to blacklist the aesni module or use another
encryption mode.
When I noticed this commit on the crypto mailing list I applied it on
top of 3.19.1 (I think), and the machine no longer crashed when
receiving this traffic. It could be a coincidence of course, but I felt
confident because most of the traces indicated some kind of memory
corruption, and this patch fitted the profile pretty well.
For example, here's a photo I took of the crash on 3.18.4:
https://orebokech.com/tmp/IMG_20150129_181653.jpg
Cheers,
--
Romain Francoise <rfrancoise@debian.org>
http://people.debian.org/~rfrancoise/
Reply to: