Re: [RFC] c/r: prctl: Add ability to set new mm_struct::exe_file

[Oleg Nesterov <oleg@redhat.com>]

On 03/05, Cyrill Gorcunov wrote:
>
> On Mon, Mar 05, 2012 at 03:26:55PM +0100, Oleg Nesterov wrote:
> > > OK, I won't argue, probably this makes sense to make sure that
> > > admin can't get a heart attack looking at /proc/pid/exe.
> > >
> > > But the O_RDONLY check looks strange. We are not going to write
> > > to this file, we only set the name (and that is why I think it
> > > should be mm->exe_path). What is the point to check that the file
> > > was opened without FMODE_WRITE? Even if there were any security
> > > risk the apllication can open this file again with the different
> > > flags.
> >
>
> Hi Oleg!
>
> Replying to both your email -- I wanted to be as close to open_exec
> as possible.

I see. But open_exec() is different, it returns the file we are going
to read/mmap. PR_SET_MM_EXE_FILE is different, I think O_RDONLY buys
nothing and looks confusing.

Anyway, as I said I won't argue.

> This prctl does cheat the kernel

Yep. Except, well, it cheats the user-space.

> but with this tests
> the cheating should be minimized (it's almost the same as open_exec
> does).

I don't reallt understand "minimized" ;) With this tests
proc/pid/exe can't look "obviously wrong", I agree. But that is all.

> > Seriously, I think we should cleanup this before c/r adds more
> > ugliness. I'll try to make the patch today.
>
> Cleanup what? If you mean this patch -- just point me what
> should I do.

I just sent the patch, "turn mm->exe_file into mm->exe_path"

> > And with all these checks I am no longer sure that fd is better
> > than filename ;)
>
> This security tests was a reason why I've used open_exec in
> first version of the patch

Yes, but me and Pavel forced you to use "int fd" ;)

> (and I still would prefer to
> have open_exec here instead of fd).

With the patch I sent "struct file *" is not needed at all.
I think prctl() can use user_path().

> As to allow-write-access -- it should be cleaned once process
> finished, no?

Exactly! And who will increment ->i_writecount? Nobody, that is
the problem.

Oleg.