Messages in this thread | | | Date | Mon, 9 Jun 2008 15:20:31 +0200 | From | Ingo Molnar <> | Subject | Re: [PATCH 2/2] __down_common: use signal_pending_state() |
| |
* Matthew Wilcox <matthew@wil.cx> wrote:
> On Wed, Jun 04, 2008 at 09:09:06PM +0400, Oleg Nesterov wrote: > > Cleanup. __down_common() can use the new signal_pending_state() helper. > > This is a bad optimisation. __down_common gets inlined and the > constant 'state' versions are optimised away for the versions which > don't apply. > > NAK this patch.
well but we had a schedule() bug around this whole topic of TASK_KILLABLE, so i'd say robustness trumps micro-optimizations. Is this really a big issue? The dual-ness of TASK_KILLABLE checks is a bit ugly, and it led to a bug as well.
Ingo
| |