| Message ID | 20230915083417.3190512-1-liushixin2@huawei.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | [v6] mm: vmscan: try to reclaim swapcache pages if no swap space | expand |
On Fri, 15 Sep 2023 16:34:17 +0800 Liu Shixin <liushixin2@huawei.com> wrote: > When spaces of swap devices are exhausted, only file pages can be > reclaimed. But there are still some swapcache pages in anon lru list. > This can lead to a premature out-of-memory. > > The problem is found with such step: > > Firstly, set a 9MB disk swap space, then create a cgroup with 10MB > memory limit, then runs an program to allocates about 15MB memory. > > The problem occurs occasionally, which may need about 100 times [1]. > > Fix it by checking number of swapcache pages in can_reclaim_anon_pages(). > If the number is not zero, return true and set swapcache_only to 1. > When scan anon lru list in swapcache_only mode, non-swapcache pages will > be skipped to isolate in order to accelerate reclaim efficiency. > > However, in swapcache_only mode, the scan count still increased when scan > non-swapcache pages because there are large number of non-swapcache pages > and rare swapcache pages in swapcache_only mode, and if the non-swapcache > is skipped and do not count, the scan of pages in isolate_lru_folios() can > eventually lead to hung task, just as Sachin reported [2]. > > By the way, since there are enough times of memory reclaim before OOM, it > is not need to isolate too much swapcache pages in one times. > mhocko earlier suspected this might impact global reclaim. Have you looked into that further?
On 2023/10/5 0:18, Andrew Morton wrote: > On Fri, 15 Sep 2023 16:34:17 +0800 Liu Shixin <liushixin2@huawei.com> wrote: > >> When spaces of swap devices are exhausted, only file pages can be >> reclaimed. But there are still some swapcache pages in anon lru list. >> This can lead to a premature out-of-memory. >> >> The problem is found with such step: >> >> Firstly, set a 9MB disk swap space, then create a cgroup with 10MB >> memory limit, then runs an program to allocates about 15MB memory. >> >> The problem occurs occasionally, which may need about 100 times [1]. >> >> Fix it by checking number of swapcache pages in can_reclaim_anon_pages(). >> If the number is not zero, return true and set swapcache_only to 1. >> When scan anon lru list in swapcache_only mode, non-swapcache pages will >> be skipped to isolate in order to accelerate reclaim efficiency. >> >> However, in swapcache_only mode, the scan count still increased when scan >> non-swapcache pages because there are large number of non-swapcache pages >> and rare swapcache pages in swapcache_only mode, and if the non-swapcache >> is skipped and do not count, the scan of pages in isolate_lru_folios() can >> eventually lead to hung task, just as Sachin reported [2]. >> >> By the way, since there are enough times of memory reclaim before OOM, it >> is not need to isolate too much swapcache pages in one times. >> > mhocko earlier suspected this might impact global reclaim. Have you > looked into that further? > . Alreadly test the case of global reclaim using ltp testcases. In previous version, the reclaim can stall in isolate_lru_folios() because I didn't count non-swapcache pages into scan so it will waste too much times scanning non-swapcache pages. In this version, I count non-swapcache pages into scan too and terminate when scan reaches the threshold. So it will not stall in reclaim now.
diff --git a/include/linux/swap.h b/include/linux/swap.h index 493487ed7c38..19f30a29e1f1 100644 --- a/include/linux/swap.h +++ b/include/linux/swap.h @@ -658,6 +658,7 @@ static inline void mem_cgroup_uncharge_swap(swp_entry_t entry, unsigned int nr_p } extern long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg); +extern long mem_cgroup_get_nr_swapcache_pages(struct mem_cgroup *memcg); extern bool mem_cgroup_swap_full(struct folio *folio); #else static inline void mem_cgroup_swapout(struct folio *folio, swp_entry_t entry) @@ -680,6 +681,11 @@ static inline long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg) return get_nr_swap_pages(); } +static inline long mem_cgroup_get_nr_swapcache_pages(struct mem_cgroup *memcg) +{ + return total_swapcache_pages(); +} + static inline bool mem_cgroup_swap_full(struct folio *folio) { return vm_swap_full(); diff --git a/mm/memcontrol.c b/mm/memcontrol.c index 34bae9e63ff2..26ae2e08fd9a 100644 --- a/mm/memcontrol.c +++ b/mm/memcontrol.c @@ -7573,6 +7573,14 @@ long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg) return nr_swap_pages; } +long mem_cgroup_get_nr_swapcache_pages(struct mem_cgroup *memcg) +{ + if (mem_cgroup_disabled()) + return total_swapcache_pages(); + + return memcg_page_state(memcg, NR_SWAPCACHE); +} + bool mem_cgroup_swap_full(struct folio *folio) { struct mem_cgroup *memcg; diff --git a/mm/vmscan.c b/mm/vmscan.c index 3df0e2a59052..48c092cc6088 100644 --- a/mm/vmscan.c +++ b/mm/vmscan.c @@ -136,6 +136,9 @@ struct scan_control { /* Always discard instead of demoting to lower tier memory */ unsigned int no_demotion:1; + /* Swap space is exhausted, only reclaim swapcache for anon LRU */ + unsigned int swapcache_only:1; + /* Allocation order */ s8 order; @@ -319,10 +322,22 @@ static inline bool can_reclaim_anon_pages(struct mem_cgroup *memcg, */ if (get_nr_swap_pages() > 0) return true; + /* Is there any swapcache pages to reclaim? */ + if (total_swapcache_pages() > 0) { + if (sc) + sc->swapcache_only = 1; + return true; + } } else { /* Is the memcg below its swap limit? */ if (mem_cgroup_get_nr_swap_pages(memcg) > 0) return true; + /* Is there any swapcache pages in memcg to reclaim? */ + if (mem_cgroup_get_nr_swapcache_pages(memcg) > 0) { + if (sc) + sc->swapcache_only = 1; + return true; + } } /* @@ -1642,6 +1657,15 @@ static unsigned long isolate_lru_folios(unsigned long nr_to_scan, */ scan += nr_pages; + /* + * Count non-swapcache too because the swapcache pages may + * be rare and it takes too much times here if not count + * the non-swapcache pages. + */ + if (unlikely(sc->swapcache_only && !is_file_lru(lru) && + !folio_test_swapcache(folio))) + goto move; + if (!folio_test_lru(folio)) goto move; if (!sc->may_unmap && folio_mapped(folio))