Re: Kernel stack read with PTRACE_EVENT_EXIT and io_uring threads

From: ebiederm@xmission.com (Eric W. Biederman)
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>,
	Michael Schmitz <schmitzmic@gmail.com>,
	linux-arch <linux-arch@vger.kernel.org>,
	Jens Axboe <axboe@kernel.dk>, Oleg Nesterov <oleg@redhat.com>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	Richard Henderson <rth@twiddle.net>,
	Ivan Kokshaysky <ink@jurassic.park.msu.ru>,
	Matt Turner <mattst88@gmail.com>,
	alpha <linux-alpha@vger.kernel.org>,
	Geert Uytterhoeven <geert@linux-m68k.org>,
	linux-m68k <linux-m68k@lists.linux-m68k.org>,
	Arnd Bergmann <arnd@kernel.org>,
	Ley Foon Tan <ley.foon.tan@intel.com>, Tejun Heo <tj@kernel.org>,
	Kees Cook <keescook@chromium.org>,
	Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Subject: Re: Kernel stack read with PTRACE_EVENT_EXIT and io_uring threads
Date: Tue, 22 Jun 2021 16:02:46 -0500	[thread overview]
Message-ID: <87h7hpbojt.fsf@disp2133> (raw)
In-Reply-To: <CAHk-=wjtagi3g5thA-T8ooM8AXcy3brdHzugCPU0itdbpDYH_A@mail.gmail.com> (Linus Torvalds's message of "Mon, 21 Jun 2021 16:36:33 -0700")

Linus Torvalds <torvalds@linux-foundation.org> writes:

> On Mon, Jun 21, 2021 at 4:23 PM Al Viro <viro@zeniv.linux.org.uk> wrote:
>>
>>         How would it help e.g. oopsen on the way out of timer interrupts?
>> IMO we simply shouldn't allow ptrace access if the tracee is in that kind
>> of state, on any architecture...
>
> Yeah no, we can't do the "wait for ptrace" when the exit is due to an
> oops. Although honestly, we have other cases like that where do_exit()
> isn't 100% robust if you kill something in an interrupt. Like all the
> locks it leaves locked etc.
>
> So do_exit() from a timer interrupt is going to cause problems
> regardless. I agree it's probably a good idea to try to avoid causing
> even more with the odd ptrace thing, but I don't think ptrace_event is
> some really "fundamental" problem at that point - it's just one detail
> among many many.
>
> So I was more thinking of the debug patch for m68k to catch all the
> _regular_ cases, and all the other random cases of ptrace_event() or
> ptrace_notify().
>
> Although maybe we've really caught them all. The exit case was clearly
> missing, and the thread fork case was scrogged. There are patches for
> the known problems. The patches I really don't like are the
> verification ones to find any unknown ones..

We still have nios2 which copied the m68k logic at some point.  I think
that is a processor that is still ``shipping'' and that people might
still be using in new designs.

I haven't looked closely enough to see what the other architectures with
caller saved registers are doing.

The challenging ones are /proc/pid/syscall and seccomp which want to see
all of the system call arguments.  I think every architecture always
saves the system call arguments unconditionally, so those cases are
probably not as interesting.  But they certain look like they could be
trouble.

Eric