Re: [RFC] Event counters [1/3]: Basic counter functionality

[Nick Piggin <nickpiggin@yahoo.com.au>]

Marcelo Tosatti wrote:

> 
> What about this addition to the documentation above, to make it a little more 
> verbose:
> 
> 	The possible race scenario is restricted to kernel preemption,
> 	and could happen as follows:
> 
> 	thread A				thread B
> a)	movl    xyz(%ebp), %eax			movl    xyz(%ebp), %eax
> b)	incl    %eax				incl    %eax
> c)	movl    %eax, xyz(%ebp)			movl    %eax, xyz(%ebp)
> 
> Thread A can be preempted in b), and thread B succesfully increments the
> counter, writing it back to memory. Now thread A resumes execution, with
> its stale copy of the counter, and overwrites the current counter.
> 
> Resulting in increments lost.
> 
> However that should be relatively rare condition.
> 

Hi Guys,

I've been waiting for some mm/ patches to clear from -mm before commenting
too much... however I see that this patch is actually against -mm itself,
with my __mod_page_state stuff in it... that makes the page state accounting
much lighter weight AND is not racy.

So I'm not exactly sure why such a patch as this is wanted now? Are there
any more xxx_page_state hotspots? (I admit to only looking at page faults,
page allocator, and page reclaim).

Thanks,
Nick

-- 
SUSE Labs, Novell Inc.
Send instant messages to your online friends http://au.messenger.yahoo.com