Re: [PATCH 1/3] futex: do not pagefault_disable infutex_atomic_cmpxchg_inatomic()

From: Thomas Gleixner
Date: Mon Mar 14 2011 - 05:15:28 EST


On Sun, 13 Mar 2011, Linus Torvalds wrote:
> On Thu, Mar 10, 2011 at 6:47 PM, Michel Lespinasse <walken@xxxxxxxxxx> wrote:
> > kernel/futex.c disables page faults before calling
> > futex_atomic_cmpxchg_inatomic(), so there is no need to do it again
> > within that function.
>
> This seems totally bogus.
>
> Even the comment is crap.
>
> Sure, the callers may disable preemption, but that has NOTHING to do
> with "pagefault_disable()". Th epagefault_[en/dis]able functions will
> touch the preempt count EVEN IF PREEMPTION ISN'T EVEN ENABLED!
>
> So what the f*ck does that "Note that preemption is disabled.." crap even mean?
>
> The thing is made even worse by the fact that as far as I can tell,
> the comment simply isn't true at all (even if you were to ignore the
> fundamental confusion about preemption vs the pagefault
> disable/enable). Not all callers of futex_atomic_cmpxchg_inatomic() do
> anything of the sort, whether it's preemptibility _or_ the proper
> pagefault_disable/enable(). Just look at the exit_robust_list() ->
> handle_futex_death(), for example.
>
> This kind of patch is the kind that personally makes me want to put
> you on a spam-list. Misleading commit messages with bogus and
> fundamentally incorrect added comments in the code. WTF?
>
> Did I miss some patch that changed that, or is this really as horribly
> bad as I think it is? I see it already made it into -tip.

That's my fault.

I really checked the call sites of futex_atomic_cmpxchg_inatomic() and
totally failed to see the one in handle_futex_death() which does not
use the helper function cmpxchg_futex_value_locked(). That helper
function is safe and does the right thing:

pagefault_disable();
curval = futex_atomic_cmpxchg_inatomic(uaddr, uval, newval);
pagefault_enable();

So, that means we have all call sites covered except one, which needs
to be fixed _AND_ also pushed into stable as all arch implementations
except ARM rely on the caller doing the pagefault_disable().

And I missed the bogus comment as well. Sigh.

Thanks,

tglx
--
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/