Re: [PATCH 2/2] sys, seccomp: add PR_SECCOMP_EXT and SECCOMP_EXT_ACT_TSYNC

[Will Drewry]

On Tue, Jan 14, 2014 at 2:13 PM, Oleg Nesterov <oleg@xxxxxxxxxx> wrote:
> On 01/14, Oleg Nesterov wrote:
>>
>> On 01/14, Oleg Nesterov wrote:
>> >
>> > > +                 get_seccomp_filter(caller);
>> > > +                 /*
>> > > +                  * Drop the task reference to the shared ancestor since
>> > > +                  * current's path will hold a reference.  (This also
>> > > +                  * allows a put before the assignment.)
>> > > +                  */
>> > > +                 put_seccomp_filter(thread);
>> > > +                 thread->seccomp.filter = caller->seccomp.filter;
>> >
>> > As I said, I do not understand this patch yet, but this looks suspicious.
>> >
>> > Why we can't race with this thread doing clone(CLONE_THREAD) ? We do
>> > not the the new thread yet, but its ->seccomp can be already copied
>> > by copy_process(), no?

Ah - I thought the tasklist_lock would catch that, but of course that
happens before
the tasklist_lock is needed.

>>
>> And it seems that this can obviously race with seccomp_attach_filter()
>> called by this "thread".

And... I was hoping the task_lock would cover any attach cases, but
missing the copy_process() is a problem.
>
> Heh. I just noticed that this patch is not first in series, and I wasn't
> cc'ed. I found this one on marc.info,

Sorry!  I shouldn't have relied on cc-cmd, I usually mess it up.

>
>         http://marc.info/?l=linux-kernel&m=138964557211277
>
> this explains task_lock(). But this can't fix the race with copy_process,
> and the patch itself doesn't look right... if nothing else, we can't do
> copy_from_user() under task_lock().

Thanks -- I'll take a more critical look!
--
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/