Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Mon, 15 Dec 2014 15:13:42 +0100
From: Sebastian Krahmer <krahmer@...e.de>
To: oss-security@...ts.openwall.com
Subject: Re: blkid command injection

On Tue, Dec 02, 2014 at 12:22:04PM +0100, Sebastian Krahmer wrote:
> 
> On Fri, Nov 28, 2014 at 12:17:24AM +1100, Murray McAllister wrote:
> > On 11/27/2014 02:25 AM, Sebastian Krahmer wrote:
> >> Hi
> >>
> >> There is a command injection inside blkid. It uses caching
> >> files (/dev/.blkid.tab or /run/blkid/blkid.tab) to store info about the
> >> UUID, LABEL etc it finds on certain devices.
> >>
> >> However, it does not strip " character, so it can be confused to
> >> build variable names containing embedded shell metas, which it would usually
> >> encode inside the value.
> >>
> >> Given an USB stick with /dev/sdb1 you can:
> >>
> >> # mkfs.ext4 -L 'X"`/tmp/foo` "' /dev/sdb1
> >> # blkid -o udev /dev/sdb1
> >> ID_FS_LABEL=X__/tmp/foo___
> >> [...]
> >>
> >> Seems to be OK, but invoking blkid a second time, taking the cache in effect:
> >>
> >> # blkid -o udev /dev/sdb1
> >> ID_FS_LABEL=X
> >> ID_FS_LABEL_ENC=X
> >> ID_FS_`/tmp/foo` "" UUID=...
> >> [...]
> >>
> >>
> >> "blkid -o udev" is often used in root context via udev or in automounters
> >> (uam-pmount) to construct key=value environment variables inside shell scripts
> >> which are then evaluated.
> >> Might be possible to construct an embedded LD_PRELOAD= as well for the binary
> >> case.
> >>
> >> By injecting > character one can probably construct whole fake cache entries.
> >>
> >> Sebastian
> >>
> >>
> >>
> >>
> >
> > Karel Zak has committed a patch:
> >
> > https://github.com/karelzak/util-linux/commit/89e90ae7b2826110ea28c1c0eb8e7c56c3907bdc
> >
> 
> Thanks. Patch looks good to me. I contacted upstream about additional
> fixes which you might want to include as well, so we can release it alltogether. The
> severity of command injection is probably not that high that we need
> updates immediately.

FWIW, the issue in question can be found here:

https://bugzilla.suse.com/show_bug.cgi?id=907434

Interesting for 32 bit systems where size_t has 32bit and so a mult of two
32bit words does not fit into the result. This would allow poor-man's badUSB attacks
with lovely crafted GPT's because its probed via systemd/udev upon plugin.

Thanks to Karel Zak for fixing a wrong > in the initial patch.

Sebastian

-- 

~ perl self.pl
~ $_='print"\$_=\47$_\47;eval"';eval
~ krahmer@...e.de - SuSE Security Team

Powered by blists - more mailing lists

Please check out the Open Source Software Security Wiki, which is counterpart to this mailing list.

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.