LWN.net Logo

Lots more kernel releases

A new set of stable releases - 2.6.16.26 and 2.6.17.6 - is out. These releases improve the fix for the (serious) /proc vulnerability.

On the unstable side, Linus has released 2.6.18-rc2. This release consists mostly of fixes, but it also contains the per-task delay accounting patch set. See the long-format changelog for the details.


(Log in to post comments)

Lots more kernel releases

Posted Jul 16, 2006 0:05 UTC (Sun) by NightMonkey (subscriber, #23051) [Link]

Hmm... What kernel version introduced this /proc vulnerability?

Lots more kernel releases

Posted Jul 16, 2006 3:48 UTC (Sun) by afalko (subscriber, #37028) [Link]

2.6.16.25 and 2.6.17.5: http://lwn.net/Articles/191486/

Lots more kernel releases

Posted Jul 16, 2006 4:19 UTC (Sun) by NightMonkey (subscriber, #23051) [Link]

Hrm.... Did that version introduce it, or just fix it?

Lots more kernel releases

Posted Jul 16, 2006 4:24 UTC (Sun) by NightMonkey (subscriber, #23051) [Link]

Ah, sorry, answered my own question. From the CVE at http://www.frsirt.com/english/advisories/2006/2816:

'Technical Description

A vulnerability has been identified in Linux Kernel, which could be exploited by local attackers to obtain elevated privileges. This flaw is due to a race condition in "fs/proc/base.c", which could be exploited by malicious users to execute arbitrary commands with "root" privileges.

Note : A fully functional exploit has been released.

Affected Products

Linux Kernel version 2.6.17.4 and prior
Linux Kernel version 2.6.16.24 and prior'

Yikes! This is a big problem.

Workaround: remount /proc nosuid,noexec

Posted Jul 16, 2006 9:43 UTC (Sun) by Duncan (guest, #6647) [Link]

On the Gentoo-devel list, which I follow as I run Gentoo and want to get a
heads-up on things coming down the pike, one of the kernel guys said it
wouldn't affect anyone who has /proc mounted nosuid,noexec. A bit of
testing later and they hadn't found any reason /not/ to do so (there was
speculation about a couple things but it turned out they worked fine
anyway), and I've been running that way for several hours, now, tho I did
update the kernel and it happened to be convenient to reboot at the time
so I did.

For anyone depending on the local security, I'd still recommend
double-checking it as I don't know enough about it to verify it myself,
but doing a /proc remount nosuid,noexec might be a useful workaround for
those who find it inconvenient to reboot ATM. That's how I have fstab
configured to (re)mount it, now, as it seems to make sense in any case,
and Gentoo will probably be making that the default in the next
baselayout, as well.

Duncan

Workaround: remount /proc nosuid,noexec

Posted Jul 16, 2006 21:36 UTC (Sun) by NightMonkey (subscriber, #23051) [Link]

Perhaps not even a reboot is neeeded? I just did this:

mount -o remount,nosuid,noexec /proc

Then, "mount" shows:

proc on /proc type proc (rw,noexec,nosuid)

Any drawbacks?

Workaround: remount /proc nosuid,noexec

Posted Jul 16, 2006 23:36 UTC (Sun) by Los__D (subscriber, #15263) [Link]

That was exactly what he said: "but doing a /proc remount nosuid,noexec might be a useful workaround for those who find it inconvenient to reboot ATM" :)

Workaround: remount /proc nosuid,noexec

Posted Jul 16, 2006 23:43 UTC (Sun) by NightMonkey (subscriber, #23051) [Link]

Your right. And I HAD my coffee. Uh oh.

Workaround: remount /proc nosuid,noexec

Posted Jul 17, 2006 16:14 UTC (Mon) by djrom (subscriber, #26074) [Link]

As usual, when someone suggests "noexec" for solving a security problem, it's not really a solution. It can't harm to do it anyway, but it's pretty easy to bypass. Just replace "/proc/executable" by "/lib/ld-linux.so.2 /proc/executable" and the exploit will work without a glitch.

Nope

Posted Jul 17, 2006 21:05 UTC (Mon) by JoeBuck (subscriber, #2330) [Link]

The exploit evidently is blocked if /proc is mounted nosuid. noexec has no effect.

Nope

Posted Jul 17, 2006 23:10 UTC (Mon) by djrom (subscriber, #26074) [Link]

Oops, I didn't checked for this precise exploit. I still think it's good to have it reminded as a general information on the usefulness of noexec mount option, but I stand corrected about for the exploit we were talking about :)

Lots more kernel releases

Posted Jul 16, 2006 4:27 UTC (Sun) by Method (guest, #26150) [Link]

FYI any machine running the SELinux targeted policy (or strict obviously) was not affected.

see http://james-morris.livejournal.com/12599.html

Lots more kernel releases

Posted Jul 17, 2006 14:41 UTC (Mon) by hmh (subscriber, #3838) [Link]

2.6.16.27 is out... ouch.
http://thread.gmane.org/gmane.linux.kernel/427744/focus=4...

(the thread is misnamed, don't let it confuse you).

aukasz Stelmach:
IPV6: Fix source address selection.

Greg Kroah-Hartman:
Linux 2.6.16.27

Ian Abbott:
USB serial ftdi_sio: Prevent userspace DoS (CVE-2006-2936)

YOSHIFUJI Hideaki:
IPV6 ADDRCONF: Fix default source address selection without CONFIG_IPV6_PRIVACY

Copyright © 2006, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds