Re: [PATCH] x86/unwind/orc: Handle kretprobes_trampoline

From: Marios Pomonis
Date: Fri Oct 29 2021 - 14:20:07 EST


On Thu, Oct 21, 2021 at 8:13 AM Masami Hiramatsu <mhiramat@xxxxxxxxxx> wrote:
>
> On Thu, 14 Oct 2021 19:16:43 +0900
> Masami Hiramatsu <mhiramat@xxxxxxxxxx> wrote:
>
> > On Wed, 13 Oct 2021 21:52:36 -0700
> > Kees Cook <keescook@xxxxxxxxxxxx> wrote:
> >
> > > On Wed, Oct 13, 2021 at 06:41:01PM -0700, Josh Poimboeuf wrote:
> > > > On Mon, Oct 11, 2021 at 02:03:26PM -0700, Kees Cook wrote:
> > > > > On Thu, Sep 02, 2021 at 07:13:26PM -0700, Kees Cook wrote:
> > > > > > From: Marios Pomonis <pomonis@xxxxxxxxxx>
> > > > > >
> > > > > > Fix a bug in the ORC unwinder when kretprobes has replaced a return
> > > > > > address with the address of `kretprobes_trampoline'. ORC mistakenly
> > > > > > assumes that the address in the stack is a return address and decrements
> > > > > > it by 1 in order to find the proper depth of the next frame.
> >
> > Hmm, with my fixes[1], the kretprobe_trampoline address in the stack will be
> > replaced with the correct return address. In that case, that behavior
> > sounds correct.
> >
> > [1] https://lore.kernel.org/all/163163030719.489837.2236069935502195491.stgit@devnote2/
>
> Here is the code which I applied this on my series.
>
> /* Find IP, SP and possibly regs: */
> switch (orc->type) {
> case UNWIND_HINT_TYPE_CALL:
> ip_p = sp - sizeof(long);
>
> if (!deref_stack_reg(state, ip_p, &state->ip))
> goto err;
>
> state->ip = unwind_recover_ret_addr(state, state->ip,
> (unsigned long *)ip_p);
> state->sp = sp;
> state->regs = NULL;
> state->prev_regs = NULL;
> state->signal = is_kretprobe_trampoline(state->ip);
> break;
>
> Actually, this cause a build issue because I introduced more generic is_kretprobe_trampoline().
> Anyway, after calling unwind_recover_ret_addr(), the state->ip should be fixed.
> This means that the is_kretprobe_trampoline(state->ip) always be false, and
> that is correct because state->ip is recovered with the correct return address
> which is call instruction + 5.
>
> So this patch seems not needed, hmm...
>
> Thank you,
>
> --
> Masami Hiramatsu <mhiramat@xxxxxxxxxx>

You're right, I made a mistake when testing this code; this is what
happens when you create patches with debugging changes and then forget
to remove them. I re-checked and your patch does solve the issue, so
the cover mail fix is not needed (I had created it against the
then-linux-next branch which didn't include your patch).

Thanks for catching this and I apologize for the (very) late response!