Re: [PATCH 4/5] mm: fix check_move_unevictable_pages() on THP

[Alex Shi]

在 2020/8/31 上午5:08, Hugh Dickins 写道:
> check_move_unevictable_pages() is used in making unevictable shmem pages
> evictable: by shmem_unlock_mapping(), drm_gem_check_release_pagevec() and
> i915/gem check_release_pagevec().  Those may pass down subpages of a huge
> page, when /sys/kernel/mm/transparent_hugepage/shmem_enabled is "force".
> 
> That does not crash or warn at present, but the accounting of vmstats
> unevictable_pgs_scanned and unevictable_pgs_rescued is inconsistent:
> scanned being incremented on each subpage, rescued only on the head
> (since tails already appear evictable once the head has been updated).
> 
> 5.8 commit 5d91f31faf8e ("mm: swap: fix vmstats for huge page") has
> established that vm_events in general (and unevictable_pgs_rescued in
> particular) should count every subpage: so follow that precedent here.
> 
> Do this in such a way that if mem_cgroup_page_lruvec() is made stricter
> (to check page->mem_cgroup is always set), no problem: skip the tails
> before calling it, and add thp_nr_pages() to vmstats on the head.
> 
> Signed-off-by: Hugh Dickins <hughd@xxxxxxxxxx>
> ---
> Nothing here worth going to stable, since it's just a testing config
> that is fixed, whose event numbers are not very important; but this
> will be needed before Alex Shi's warning, and might as well go in now.
> 
> The callers of check_move_unevictable_pages() could be optimized,
> to skip over tails: but Matthew Wilcox has other changes in flight
> there, so let's skip the optimization for now.
> 
>  mm/vmscan.c |   10 ++++++++--
>  1 file changed, 8 insertions(+), 2 deletions(-)
> 
> --- 5.9-rc2/mm/vmscan.c	2020-08-16 17:32:50.721507348 -0700
> +++ linux/mm/vmscan.c	2020-08-28 17:47:10.595580876 -0700
> @@ -4260,8 +4260,14 @@ void check_move_unevictable_pages(struct
>  	for (i = 0; i < pvec->nr; i++) {
>  		struct page *page = pvec->pages[i];
>  		struct pglist_data *pagepgdat = page_pgdat(page);
> +		int nr_pages;
> +
> +		if (PageTransTail(page))
> +			continue;
> +
> +		nr_pages = thp_nr_pages(page);
> +		pgscanned += nr_pages;
>  
> -		pgscanned++;
>  		if (pagepgdat != pgdat) {
>  			if (pgdat)
>  				spin_unlock_irq(&pgdat->lru_lock);
> @@ -4280,7 +4286,7 @@ void check_move_unevictable_pages(struct
>  			ClearPageUnevictable(page);
>  			del_page_from_lru_list(page, lruvec, LRU_UNEVICTABLE);
>  			add_page_to_lru_list(page, lruvec, lru);

So, we might randomly del or add a thp tail page into lru?
It's interesting to know here. :)

Thanks
Alex

> -			pgrescued++;
> +			pgrescued += nr_pages;
>  		}
>  	}
>  
>