From: Joonsoo Kim <iamjoonsoo.kim@xxxxxxx> Subject: mm/memory: fix IO cost for anonymous page With synchronous IO swap device, swap-in is directly handled in fault code. Since IO cost notation isn't added there, with synchronous IO swap device, LRU balancing could be wrongly biased. Fix it to count it in fault code. Link: http://lkml.kernel.org/r/1592288204-27734-4-git-send-email-iamjoonsoo.kim@xxxxxxx Fixes: 314b57fb0460001 ("mm: balance LRU lists based on relative thrashing cache sizing") Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@xxxxxxx> Acked-by: Johannes Weiner <hannes@xxxxxxxxxxx> Cc: Joonsoo Kim <js1304@xxxxxxxxx> Cc: Michal Hocko <mhocko@xxxxxxxx> Cc: Minchan Kim <minchan.kim@xxxxxxxxx> Cc: Rik van Riel <riel@xxxxxxxxxxx> Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> --- mm/memory.c | 8 ++++++++ 1 file changed, 8 insertions(+) --- a/mm/memory.c~mm-memory-fix-io-cost-for-anonymous-page +++ a/mm/memory.c @@ -3146,6 +3146,14 @@ vm_fault_t do_swap_page(struct vm_fault goto out_page; } + /* + * XXX: Move to lru_cache_add() when it + * supports new vs putback + */ + spin_lock_irq(&page_pgdat(page)->lru_lock); + lru_note_cost_page(page); + spin_unlock_irq(&page_pgdat(page)->lru_lock); + lru_cache_add(page); swap_readpage(page, true); } _
