[prev in list] [next in list] [prev in thread] [next in thread] 

List:       oss-security
Subject:    Re: [oss-security] Linux 2.6.30+/SELinux/RHEL5 test kernel 0day,
From:       Julien Tinnes <jt () cr0 ! org>
Date:       2009-07-20 18:26:31
Message-ID: deb7a2310907201126q16707cdbv94634ea2be0d7b87 () mail ! gmail ! com
[Download RAW message or body]

On Mon, Jul 20, 2009 at 4:29 AM, Solar Designer<solar@openwall.com> wrote:
> Marcus,
>
> On Mon, Jul 20, 2009 at 12:01:47PM +0200, Marcus Meissner wrote:
>> - fixed the personality - PER_CLEAR_ON_SETTID inheritance issue (CVE-2009-1895)
>>   to work around mmap_min_addr protection.
>>   Affects 2.6.23-2.6.30.1
>
> What makes you think this does not affect earlier kernels?  This does
> not match my analysis, but maybe I am missing something, hence I ask.
>
> BTW, as you're aware, this fix is a hardening measure for/against
> SUID-root programs with a certain class of design errors in them; it is
> not exactly a fix for the kernel itself, although it should be in the
> kernel.  I do not mean to downplay the issue, but I think it is
> important that we distinguish the different types of changes that we are
> making in response to Brad's exploit.

Hey Alexander,

for a SUID-root binary, dropping privileges and giving control back to
the user without going through exec to re-set the address space is
probably not the best idea indeed. However I still believe this was a
kernel issue, if you (really) know what you're doing (bzero-ing your
secrets in the address space etc..), this should not be strictly
forbidden.
Without this patch, you could also disable address space layout
randomization in SUID binaries, and in some cases exploit NULL ptr
dereferences bugs in SUID-root binaries more easily. Granted, both
would be non issues on a sane SUID binary, but this is what security
in-depth is about. And being able to map a page in a SUID binary just
feels wrong anyway :)

Cheers,

Julien

[prev in list] [next in list] [prev in thread] [next in thread]