lkml.org 
[lkml]   [2005]   [Jan]   [4]   [last100]   RSS Feed
Views: [more markup]  [less markup]  [headers]  [forward] 
 
Messages in this thread
/
DateTue, 4 Jan 2005 10:33:30 +0100
FromIngo Molnar <>
SubjectRe: 2.6.10-mm1
* Christoph Hellwig <hch@infradead.org> wrote:

> On Tue, Jan 04, 2005 at 10:04:08AM +0100, Ingo Molnar wrote:
> > or is it the addition of _smp_processor_id() as a way to signal 'this
> > smp_processor_id() call in a preemptible region is fine, trust me'?
> 
> Yes.
> 
> > We
> > could do smp_processor_id_preempt() or some other name - any better
> > suggestions?
> 
> I'd just kill the debug check and rely on the eye of the review to not
> let new users of smp_processor_id slip in.

relying on that is quite futile. E.g. in the block IO code it needed 3-4
iterations even after the first instance was found to get all the cases
right. There are functions that are always called from under a lock then
some unlocked call happens and we've got trouble. Often the bug is some
very rare and obscure corruption of a statistics value, nobody really
notices that.

by today i think we've identified most of the places that can safely do
smp_processor_id() in a preemptible section (in x86 and x64) - it's only
around 3% of the total smp_processor_id() use. I'd rather allow these
exceptions and flag new exceptions as they get added - they are added at
least an order of magnitude more rarely than smp_processor_id() gets
added.

	Ingo
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 14:09    [from the cache]
©2003-2008