Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 22 Dec 2010 14:06:35 +0000
From: Tim Brown <tmb@...35.com>
To: oss-security@...ts.openwall.com
Subject: Re: Re: Breaking the links: Exploiting the linker

On Wednesday 22 December 2010 11:46:41 Jamie Nguyen wrote:
> Tim Brown <timb@...> writes:
> > In the interests of a thorough peer review I'd be curious what people
> > think of the following paper I've been working on Linux and POSIX
> > linkers:
> > 
> > http://www.nth-dimension.org.uk/downloads.php?id=77
> > 
> > A previous revision has already been reviewed but constructive criticism
> > is always useful.  There are some sections that I have removed whilst I
> > wait on vendors but I'm particularly interested in feedback on pertinent
> > references or threats that I may have missed.  As per the abstract, the
> > aim of the paper wasn't to claim everything as my own but rather to
> > document as much about the current state of art as possible.
> > 
> > Tim
> 
> Hi,
> 
> I am somewhat unknowledgeable about the whole linking process, but I was
> testing out the execution of a file using ld on a filesystem mounted with
> noexec. I followed the example you gave of copying the '/usr/bin/id'
> executable to a user writeable directory and removing the executable bit.
> 
> After removing the executable bit, I was still able to execute this on a
> normal filesystem using /lib/ld-linux-x86_64.so.2 but on a filesystem
> mounted with noexec this method did not work.
> 
> You suggest in the article:
> 
> "...if you're mounting devices with noexec the you should probably ensure
> that they [sic] the runtime linker can't be executed either."
> 
> Forgive me if I am being dim, because from what I can see, mounting with
> noexec seems to solve the issue of using ld-linux-x86-64.so.2 to execute
> non-executable files.

You're not being dim.  On Linux, mounting the file system with noexec prevents 
the kernel mmap()ing the pages with execute permissions.  Removing the execute 
bit on a binary doesn't cause the same behaviour.  In the paper I was 
describing the general case.  This is something taviso or stealth mentioned to 
me too so I will update the paper to make this distinction clear.  Thanks for 
the feedback.

Tim
-- 
Tim Brown
<mailto:tmb@...35.com>

Download attachment "signature.asc " of type "application/pgp-signature" (837 bytes)

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.