Re: [PATCH] show per-process swap usage via procfs v3

[KAMEZAWA Hiroyuki]

On Thu, 12 Nov 2009 10:20:29 -0500 (EST)
Christoph Lameter <cl@xxxxxxxxxxxxxxxxxxxx> wrote:

> On Wed, 11 Nov 2009, KAMEZAWA Hiroyuki wrote:
> 
> >
> > Index: mm-test-kernel/include/linux/mm_types.h
> > ===================================================================
> > --- mm-test-kernel.orig/include/linux/mm_types.h
> > +++ mm-test-kernel/include/linux/mm_types.h
> > @@ -228,6 +228,7 @@ struct mm_struct {
> >  	 */
> >  	mm_counter_t _file_rss;
> >  	mm_counter_t _anon_rss;
> > +	mm_counter_t _swap_usage;
> 
> This is going to be another hit on vm performance if we get down this
> road.
> 
> At least put
> 
> #ifdef CONFIG_SWAP ?
> 
> around this so that we can switch it off?
> 
Hmm, okay. But I'm not sure I can do it in clean way.
(Or, I'll wait for you updates for mm_counters, or I do by myself.)

> > @@ -597,7 +600,9 @@ copy_one_pte(struct mm_struct *dst_mm, s
> >  						 &src_mm->mmlist);
> >  				spin_unlock(&mmlist_lock);
> >  			}
> > -			if (is_write_migration_entry(entry) &&
> > +			if (!non_swap_entry(entry))
> > +				rss[2]++;
> > +			else if (is_write_migration_entry(entry) &&
> >  					is_cow_mapping(vm_flags)) {
> >  				/*
> 
> What are the implications for fork performance?

This path is executed when page table entry contains a entry of
  !pte_none() && !pte_present().

There are not very big chance to reach here.(this path is under unlikely()).

Thanks,
-Kame

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