On Tue, Apr 21, 2020 at 09:06:37PM +0900, Jaewon Kim wrote: > > > On 2020년 04월 20일 15:19, Jaewon Kim wrote: > > > > On 2020년 04월 18일 08:45, Jaewon Kim wrote: > >> On 2020년 04월 18일 00:13, Minchan Kim wrote: > >>> On Thu, Apr 16, 2020 at 05:38:37PM -0700, Minchan Kim wrote: > >>>> Hi Jaewon, > >>>> > >>>> On Thu, Apr 16, 2020 at 12:35:14PM +0900, Jaewon Kim wrote: > >>>>> This patch fix nr_isolate_* mismatch problem between cma and dirty > >>>>> lazyfree page. > >>>>> > >>>>> If try_to_unmap_one is used for reclaim and it detects a dirty lazyfree > >>>>> page, then the lazyfree page is changed to a normal anon page having > >>>>> SwapBacked by commit 18863d3a3f59 ("mm: remove SWAP_DIRTY in ttu"). Even > >>>>> with the change, reclaim context correctly counts isolated files because > >>>>> it uses is_file_lru to distinguish file. And the change to anon is not > >>>>> happened if try_to_unmap_one is used for migration. So migration context > >>>>> like compaction also correctly counts isolated files even though it uses > >>>>> page_is_file_lru insted of is_file_lru. Recently page_is_file_cache was > >>>>> renamed to page_is_file_lru by commit 9de4f22a60f7 ("mm: code cleanup for > >>>>> MADV_FREE"). > >>>>> > >>>>> But the nr_isolate_* mismatch problem happens on cma alloc. There is > >>>>> reclaim_clean_pages_from_list which is being used only by cma. It was > >>>>> introduced by commit 02c6de8d757c ("mm: cma: discard clean pages during > >>>>> contiguous allocation instead of migration") to reclaim clean file pages > >>>>> without migration. The cma alloc uses both reclaim_clean_pages_from_list > >>>>> and migrate_pages, and it uses page_is_file_lru to count isolated > >>>>> files. If there are dirty lazyfree pages allocated from cma memory > >>>>> region, the pages are counted as isolated file at the beginging but are > >>>>> counted as isolated anon after finished. > >>>>> > >>>>> Mem-Info: > >>>>> Node 0 active_anon:3045904kB inactive_anon:611448kB active_file:14892kB inactive_file:205636kB unevictable:10416kB isolated(anon):0kB isolated(file):37664kB mapped:630216kB dirty:384kB writeback:0kB shmem:42576kB writeback_tmp:0kB unstable:0kB all_unreclaimable? no > >>>>> > >>>>> Like log above, there was too much isolated file, 37664kB, which > >>>>> triggers too_many_isolated in reclaim when there is no isolated file in > >>>>> system wide. It could be reproducible by running two programs, doing > >>>>> MADV_FREE, writing and doing cma alloc, respectively. Although isolated > >>>>> anon is 0, I found that the internal value of isolated anon was the > >>>>> negative value of isolated file. > >>>>> > >>>>> Fix this by skipping anon pages on reclaim_clean_pages_from_list. The > >>>>> lazyfree page can be checked by both PageAnon(page) and > >>>>> page_is_file_lru. But in this case, PageAnon is enough to skip all > >>>>> anon pages. > >>>>> > >>>>> Reported-by: Yong-Taek Lee <ytk.lee@xxxxxxxxxxx> > >>>>> Signed-off-by: Jaewon Kim <jaewon31.kim@xxxxxxxxxxx> > >>>> Thanks for the investigation! > >>>> The thing is MADV_FREEed page since supporting swapless could change > >>>> his LRU status between reclaim. > >>>> > >>>> I am worry about voiding the optimization we have kept in CMA but > >>>> also don't have good idea, either so I tend to agree with this. > >>>> > >>>> Let me Cc Johannes who might have better idea. > >>>> > >>>>> --- > >>>>> mm/vmscan.c | 3 +++ > >>>>> 1 file changed, 3 insertions(+) > >>>>> > >>>>> diff --git a/mm/vmscan.c b/mm/vmscan.c > >>>>> index b06868fc4926..9380a18eef5e 100644 > >>>>> --- a/mm/vmscan.c > >>>>> +++ b/mm/vmscan.c > >>>>> @@ -1497,6 +1497,9 @@ unsigned long reclaim_clean_pages_from_list(struct zone *zone, > >>>>> LIST_HEAD(clean_pages); > >>>>> > >>>>> list_for_each_entry_safe(page, next, page_list, lru) { > >>>>> + /* to avoid race with MADV_FREE anon page */ > >>>>> + if (PageAnon(page)) > >>>>> + continue; > >>>>> if (page_is_file_lru(page) && !PageDirty(page) && > >>>>> !__PageMovable(page) && !PageUnevictable(page)) { > >>>>> ClearPageActive(page); > >>>>> -- > >>>>> 2.13.7 > >>>>> > >>> Hi Jaewon, > >>> > >>> How about this idea? I think it could solve the issue with keeping > >>> CMA alloc latency optimization. > >> Hello Minchan > >> > >> It looks good to me except compilation error. > >> > >> And to apply this patch on other stable branches, we may need some other > >> dependent patches though. > >> > >> Thank you > >>> diff --git a/include/linux/vmstat.h b/include/linux/vmstat.h > >>> index 292485f3d24d..10cc932e209a 100644 > >>> --- a/include/linux/vmstat.h > >>> +++ b/include/linux/vmstat.h > >>> @@ -29,6 +29,7 @@ struct reclaim_stat { > >>> unsigned nr_activate[2]; > >>> unsigned nr_ref_keep; > >>> unsigned nr_unmap_fail; > >>> + unsigned nr_lazyfree_fail; > >>> }; > >>> > >>> enum writeback_stat_item { > >>> diff --git a/mm/vmscan.c b/mm/vmscan.c > >>> index 4c8a1cdccbba..b390f6094f2f 100644 > >>> --- a/mm/vmscan.c > >>> +++ b/mm/vmscan.c > >>> @@ -1296,11 +1296,15 @@ static unsigned long shrink_page_list(struct list_head *page_list, > >>> */ > >>> if (page_mapped(page)) { > >>> enum ttu_flags flags = ttu_flags | TTU_BATCH_FLUSH; > >>> + bool lazyfree = PageAnon(page) && !PageSwapBacked(page); > >>> > >>> if (unlikely(PageTransHuge(page))) > >>> flags |= TTU_SPLIT_HUGE_PMD; > >>> + > >>> if (!try_to_unmap(page, flags)) { > >>> stat->nr_unmap_fail += nr_pages; > >>> + if (lazyfree && PageSwapBacked(page)) > >>> + stat->nr_lazyfree_fail += nr_pages; > >>> goto activate_locked; > >>> } > >>> } > >>> @@ -1492,8 +1496,8 @@ unsigned long reclaim_clean_pages_from_list(struct zone *zone, > >>> .priority = DEF_PRIORITY, > >>> .may_unmap = 1, > >>> }; > >>> - struct reclaim_stat dummy_stat; > >>> - unsigned long ret; > >>> + struct reclaim_stat stat; > >>> + unsigned long reclaimed; > >>> struct page *page, *next; > >>> LIST_HEAD(clean_pages); > >>> > >>> @@ -1505,11 +1509,21 @@ unsigned long reclaim_clean_pages_from_list(struct zone *zone, > >>> } > >>> } > >>> > >>> - ret = shrink_page_list(&clean_pages, zone->zone_pgdat, &sc, > >>> - TTU_IGNORE_ACCESS, &dummy_stat, true); > >>> + reclaimed = shrink_page_list(&clean_pages, zone->zone_pgdat, &sc, > >>> + TTU_IGNORE_ACCESS, &stat, true); > >>> list_splice(&clean_pages, page_list); > >>> - mod_node_page_state(zone->zone_pgdat, NR_ISOLATED_FILE, -ret); > >>> - return ret; > >>> + mod_node_page_state(zone->zone_pgdat, NR_ISOLATED_FILE, -reclaimed); > >>> + /* > >>> + * Since lazyfree pages are isolated from file LRU from the beginning, > >>> + * they will rotate back to anonymous LRU in the end if it failed to > >>> + * discard so isolated count will be mismatched. > >>> + * Compensate the isolated count for both LRU lists. > >>> + */ > >>> + mod_node_page_state(zone->zone_pgdat, NR_ISOLATED_ANON, > >>> + stat->nr_lazyfree_fail); > >>> + mod_node_page_state(zone->zone_pgdat, NR_ISOLATED_FILE, > >>> + -stat->nr_lazyfree_fail); > >> should be stat.nr_lazyfree_fail and -stat.nr_lazyfree_fail instead of -> > >>> + return reclaimed; > >>> } > >>> > >>> /* > > Let me just CC Shaohua Li for commit 802a3a92ad7a ("mm: reclaim MADV_FREE pages") > > , because I missed him/her on other mail thread > > : Sorry, I think I pointed a wrong commit, the SwapBacked was recovered > > : by commit 802a3a92ad7a ("mm: reclaim MADV_FREE pages"). > > > > > > Hello Minchan > > > > I tested on my v4.19 based source tree and it seems to work. > > > > Prior to your patch I applied commit 060f005f0747 ( > > "mm/vmscan.c: do not allocate duplicate stack variables in shrink_page_list()" > > for struct reclaim_stat. > > > > I considered other dependent changes below to follow code changes, not really needed for this issue though. > > v5.3 98879b3b9edc mm: vmscan: correct some vmscan counters for THP swapout > > v5.2 886cf1901db9 mm: move recent_rotated pages calculation to shrink_inactive_list() > Hello Minchan > > Are you preparing a complete patch for this issue? > Sorry if I am bugging you. Hi Jaewon Sorry for the late. You catched the bug and gave good description with the solution. What I did was just suggestion for alternative so feel free to send the patch with your SoB. You could use my Suggested-by and Acked-by. It's totally your credit! Thanks! Thanks!