Re: [PATCH V3 3/7] mm: reclaim MADV_FREE pages

[Shaohua Li]

On Thu, Feb 16, 2017 at 01:40:18PM -0500, Johannes Weiner wrote:
> On Tue, Feb 14, 2017 at 11:36:09AM -0800, Shaohua Li wrote:
> > @@ -1419,11 +1419,18 @@ static int try_to_unmap_one(struct page *page, struct vm_area_struct *vma,
> >  			VM_BUG_ON_PAGE(!PageSwapCache(page) && PageSwapBacked(page),
> >  				page);
> >  
> > -			if (!PageDirty(page) && (flags & TTU_LZFREE)) {
> > -				/* It's a freeable page by MADV_FREE */
> > -				dec_mm_counter(mm, MM_ANONPAGES);
> > -				rp->lazyfreed++;
> > -				goto discard;
> > +			if (flags & TTU_LZFREE) {
> > +				if (!PageDirty(page)) {
> > +					/* It's a freeable page by MADV_FREE */
> > +					dec_mm_counter(mm, MM_ANONPAGES);
> > +					rp->lazyfreed++;
> > +					goto discard;
> > +				} else {
> > +					set_pte_at(mm, address, pvmw.pte, pteval);
> > +					ret = SWAP_FAIL;
> > +					page_vma_mapped_walk_done(&pvmw);
> > +					break;
> > +				}
> 
> I don't understand why we need the TTU_LZFREE bit in general. More on
> that below at the callsite.

Sounds useless flag, don't see any reason we shouldn't free the MADV_FREE page
in places other than reclaim. Looks TTU_UNMAP is useless too..

> > @@ -911,7 +911,7 @@ static void page_check_dirty_writeback(struct page *page,
> >  	 * Anonymous pages are not handled by flushers and must be written
> >  	 * from reclaim context. Do not stall reclaim based on them
> >  	 */
> > -	if (!page_is_file_cache(page)) {
> > +	if (!page_is_file_cache(page) || page_is_lazyfree(page)) {
> 
> Do we need this? MADV_FREE clears the dirty bit off the page; we could
> just let them go through with the function without any special-casing.

this is just to zero dirty and writeback
> 
> > @@ -986,6 +986,8 @@ static unsigned long shrink_page_list(struct list_head *page_list,
> >  
> >  		sc->nr_scanned++;
> >  
> > +		lazyfree = page_is_lazyfree(page);
> > +
> >  		if (unlikely(!page_evictable(page)))
> >  			goto cull_mlocked;
> >  
> > @@ -993,7 +995,7 @@ static unsigned long shrink_page_list(struct list_head *page_list,
> >  			goto keep_locked;
> >  
> >  		/* Double the slab pressure for mapped and swapcache pages */
> > -		if (page_mapped(page) || PageSwapCache(page))
> > +		if ((page_mapped(page) || PageSwapCache(page)) && !lazyfree)
> >  			sc->nr_scanned++;
> >  
> >  		may_enter_fs = (sc->gfp_mask & __GFP_FS) ||
> > @@ -1119,13 +1121,13 @@ static unsigned long shrink_page_list(struct list_head *page_list,
> >  		/*
> >  		 * Anonymous process memory has backing store?
> >  		 * Try to allocate it some swap space here.
> > +		 * Lazyfree page could be freed directly
> >  		 */
> > -		if (PageAnon(page) && !PageSwapCache(page)) {
> > +		if (PageAnon(page) && !PageSwapCache(page) && !lazyfree) {
> 
> lazyfree duplicates the anon check. As per the previous email, IMO it
> would be much preferable to get rid of that "lazyfree" obscuring here.
> 
> This would simply be:
> 
> 		if (PageAnon(page) && PageSwapBacked && !PageSwapCache)

I'd agree if we only don't need the lazyfree variable, but I think we still
need to check it in other places. More in below.

> > @@ -1142,7 +1144,7 @@ static unsigned long shrink_page_list(struct list_head *page_list,
> >  		 * The page is mapped into the page tables of one or more
> >  		 * processes. Try to unmap it here.
> >  		 */
> > -		if (page_mapped(page) && mapping) {
> > +		if (page_mapped(page) && (mapping || lazyfree)) {
> 
> Do we actually need to filter for mapping || lazyfree? If we fail to
> allocate swap, we don't reach here. If the page is a truncated file
> page, ttu returns pretty much instantly with SWAP_AGAIN. We should be
> able to just check for page_mapped() alone, no?

checking the mapping is faster than running into try_to_unamp, right?

> >  			switch (ret = try_to_unmap(page, lazyfree ?
> >  				(ttu_flags | TTU_BATCH_FLUSH | TTU_LZFREE) :
> >  				(ttu_flags | TTU_BATCH_FLUSH))) {
> 
> That bit I don't understand. Why do we need to pass TTU_LZFREE? What
> information does that carry that cannot be gathered from inside ttu?
> 
> I.e. when ttu runs into PageAnon, can it simply check !PageSwapBacked?
> And if it's still clean, it can lazyfreed++; goto discard.
> 
> Am I overlooking something?
> 
> > @@ -1154,7 +1156,14 @@ static unsigned long shrink_page_list(struct list_head *page_list,
> >  			case SWAP_MLOCK:
> >  				goto cull_mlocked;
> >  			case SWAP_LZFREE:
> > -				goto lazyfree;
> > +				/* follow __remove_mapping for reference */
> > +				if (page_ref_freeze(page, 1)) {
> > +					if (!PageDirty(page))
> > +						goto lazyfree;
> > +					else
> > +						page_ref_unfreeze(page, 1);
> > +				}
> > +				goto keep_locked;
> >  			case SWAP_SUCCESS:
> >  				; /* try to free the page below */
> 
> This is a similar situation.
> 
> Can we let the page go through the regular __remove_mapping() process
> and simply have that function check for PageAnon && !PageSwapBacked?

That will make the code more complicated. We don't call __remove_mapping if
!mapping. And we need to do bypass in __remove_mapping, for example, avoid
taking mapping->lock.
 
> > @@ -1266,10 +1275,9 @@ static unsigned long shrink_page_list(struct list_head *page_list,
> >  			}
> >  		}
> >  
> > -lazyfree:
> >  		if (!mapping || !__remove_mapping(mapping, page, true))
> >  			goto keep_locked;
> > -
> > +lazyfree:
> 
> ... eliminating this special casing.
> 
> > @@ -1294,6 +1302,8 @@ static unsigned long shrink_page_list(struct list_head *page_list,
> >  cull_mlocked:
> >  		if (PageSwapCache(page))
> >  			try_to_free_swap(page);
> > +		if (lazyfree)
> > +			clear_page_lazyfree(page);
> 
> Why cancel the MADV_FREE state? The combination seems non-sensical,
> but we can simply retain the invalidated state while the page goes to
> the unevictable list; munlock should move it back to inactive_file.

This depends on the policy. If user locks the page, I think it's reasonable to
assume the page is hot, so it doesn't make sense to treat the page lazyfree.
 
> >  		unlock_page(page);
> >  		list_add(&page->lru, &ret_pages);
> >  		continue;
> > @@ -1303,6 +1313,8 @@ static unsigned long shrink_page_list(struct list_head *page_list,
> >  		if (PageSwapCache(page) && mem_cgroup_swap_full(page))
> >  			try_to_free_swap(page);
> >  		VM_BUG_ON_PAGE(PageActive(page), page);
> > +		if (lazyfree)
> > +			clear_page_lazyfree(page);
> 
> This is similar too.
> 
> Can we leave simply leave the page alone here? The only way we get to
> this point is if somebody is reading the invalidated page. It's weird
> for a lazyfreed page to become active, but it doesn't seem to warrant
> active intervention here.

So the unmap fails here probably because the page is dirty, which means the
page is written recently. It makes sense to assume the page is hot.

Thanks,
Shaohua