Re: [PATCH 8/9] proc: protect /proc/<pid>/{environ,pagemap} acrossexecve

From: Djalal Harouni
Date: Sun Mar 11 2012 - 12:57:51 EST


On Sun, Mar 11, 2012 at 11:05:23AM +0300, Alexey Dobriyan wrote:
> On Sun, Mar 11, 2012 at 12:25:18AM +0100, Djalal Harouni wrote:
> > The /proc/<pid>/{environ,pagemap} are sensitive files which must be
> > protected across execve to avoid information leaks.
> >
> > These files are protected by attaching them to their task at open time by
> > saving the exec_id of the target task, this way in read we just compare
> > the target task's exec_id and the previously saved exec_id of the
> > proc_file_private struct, in other words we just bind these files to their
> > appropriate process image at open time. We do this since we are able to do
> > proper permission checks (ptrace) at each syscall, so we do not care about
> > the reader.
> >
> > Another important rule is to set the exec_id of the target task before the
> > permission checks at open, this way we do not race against target task
> > execve, and it will be more effective if the exec_id check at read/write
> > times are delayed as much as possible to be sure that the target task do
> > not change during execve.
> >
> > This patch adds the open file_operation to the
> > /proc/<pid>/{environ,pagemap} so we are able to set the exec_id of the
> > target task and to do the appropriate permission checks. The exec_id check
> > is done in the related read file_operation.
>
> ->open is duplicated.
Right, I'll unify the code in a generic open function that does:
* alloc and setup proc_file_private (which includes the exec_id)
* ptrace check using mm_for_maps()
unify only those who check PTRACE_MODE_READ
* save priv_file_proc.

This applies also to the release functions, I'll re-submit it.

Thanks Alexey.

--
tixxdz
http://opendz.org
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/