Re: pam_timestamp internals

[Sebastian Krahmer <krahmer () suse de>]

Hi


On Mon, Mar 31, 2014 at 02:32:09PM +0400, Dmitry V. Levin wrote:
> Hi,
>
> On Mon, Mar 24, 2014 at 01:46:43PM +0100, Sebastian Krahmer wrote:
> > When playing with some PAM modules for my own projects, I came
> > across some implications of pam_timestamp (which is part of
> > upstream linux-pam) that should probably be addressed.
> >
> > Most importantly, there seems to be a path traversal issue:
>
> Thanks, Sebastian!  The issue has been fixed in upstream linux-pam by commit
> https://git.fedorahosted.org/cgit/linux-pam.git/commit/?id=Linux-PAM-1_1_8-32-g9dcead8

Thanks for taking care. I was about to write a patch on my own, but seems
not necessary anymore.

However, I think that

+       if (!strlen(tty) || !strcmp(tty, ".") || !strcmp(tty, "..")) {

could be insufficient. Any occurence of "." inside tty name
should be evil. Above strcmp() matches exactly "." or "..",
but you also want "../../" etc which should pass above check.

For the ruser check, the strchr(ruser, '/') safes this, but
".." occurence may also be treatened appropriately.

Sebastian



-- 

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