Re: 2.5.59-mm1

[Andrew Morton]

Anton Blanchard <[e-mail]> wrote:
>
>  
> > -cputimes_stat.patch
> > 
> >  Not to Linus's taste.
> 
> Pity, I just had another reason to use this today (Checking if a network
> app was locking itself down to a cpu)

You can query that with sched_getaffinity()

> > -lockless-current_kernel_time.patch
> > 
> >  Is ia32-only.
> 
> We can fix that. Ive been avoiding it because it will take some non
> trivial cleanup of our ppc64 time.c. Based on how often get_current_time
> is appearing in profiles

Have you some numbers handy?

> and also how gettimeofday has been known to
> cause problems on large SMP (due to read_lock on xtime starving
> write_lock in the timer irq) I think this should get merged.
> 

OK.  I've had basically zero success getting non-ia32 people to test these
patches, and breaking their build won't help that.

But yes, this code needs to go ahead.