Re: [PATCH v2 -mm -next] ipc,sem: fix lockdep false positive

[Rik van Riel <riel@surriel.com>]

On 03/29/2013 09:08 AM, Michel Lespinasse wrote:

>> That is where I had it initially. I may have gotten too clever
>> and worked on keeping more accesses read-only. If you want, I
>> can move it back here and re-submit the patch :)
>
> I think I would prefer that - I feel having it upper serves little
> purpose, as you still need to check it here, so we might as well be
> optimistic and check it only here. Or maybe I've missed the benefit of
> having it earlier - I just don't see it.

OK, will do.

> It still makes it possible (though less likely) that each simple semop
> taking this path might make the next one take it too, though. So, I'm
> not sure how to balance that against the starvation possibility. I'll
> leave it up to you then :)

Lets optimistically assume that userland is using the semaphores
for synchronization. That rather limits the potential for starvation.

-- 
All rights reversed.