Re: [PATCH] 2/5 VM changes: skip-writepage.patch

[Nikita Danilov]

Andrew Morton writes:
 > Nikita Danilov <[e-mail]> wrote:
 > >
 > > Don't call ->writepage from VM scanner when page is met for the first time
 > >  during scan.
 > 
 > 
 > I don't think we need all the code reorganisation.  Something like this:

I separated call to ->writepage into special function because of some
other patch that I didn't send yet.

 > 
 > 
 >  include/linux/page-flags.h |    7 +++++++
 >  mm/page_alloc.c            |    1 +
 >  mm/truncate.c              |    2 ++

[...]

 > @@ -328,6 +328,8 @@ shrink_list(struct list_head *page_list,
 >  		 * See swapfile.c:page_queue_congested().
 >  		 */
 >  		if (PageDirty(page)) {
 > +			if (!TestSetPageSkipped(page))
 > +				goto keep_locked;
 >  			if (!is_page_cache_freeable(page))
 >  				goto keep_locked;
 >  			if (!mapping)
 > @@ -351,6 +353,7 @@ shrink_list(struct list_head *page_list,
 >  				list_move(&page->list, &mapping->locked_pages);
 >  				spin_unlock(&mapping->page_lock);
 >  
 > +				ClearPageSkipped(page);

Good idea.

By the way, have you noted that patch changes wbc.nonblocking to be 0,
if called from kswapd or pdflush, what do you think?

 >  				SetPageReclaim(page);
 >  				res = mapping->a_ops->writepage(page, &wbc);
 >  

Nikita.