On Thu 05-01-17 11:46:11, Andrew Morton wrote: [...] > > Since upgrading from kernel-PAE 4.5.3 on fedora 24 to kernel-PAE 4.8.10 on > > fedora 25, I get OOM when I run my daily rsync for backup. I upgraded to > > kernel-PAE-4.9.0-1.fc26.i686 and the problem still occurs. The OOM occurs > > although the system doesn't use any swap and memory is not used up either. > > > > See https://bugzilla.redhat.com/show_bug.cgi?id=1401012 > > > > Here is the dmesg from today: > > > > [32863.748720] gpg-agent invoked oom-killer: > > gfp_mask=0x27000c0(GFP_KERNEL_ACCOUNT|__GFP_NOTRACK), nodemask=0, order=1, oom_score_adj=0 this is a lowmem request [...] > > [32863.748789] active_anon:122505 inactive_anon:129240 isolated_anon:0 > > active_file:174922 inactive_file:371696 isolated_file:64 > > unevictable:8 dirty:0 writeback:0 unstable:0 > > slab_reclaimable:186174 slab_unreclaimable:17717 > > mapped:69769 shmem:11168 pagetables:2174 bounce:0 > > free:13565 free_pcp:660 free_cma:0 there is a lot of page cache and anonymous memory but... > > [32863.748792] Node 0 active_anon:490020kB inactive_anon:516960kB > > active_file:699688kB inactive_file:1486784kB unevictable:32kB > > isolated(anon):0kB isolated(file):256kB mapped:279076kB dirty:0kB writeback:0kB > > shmem:44672kB writeback_tmp:0kB unstable:0kB pages_scanned:9963129 > > all_unreclaimable? yes > > [32863.748795] DMA free:3260kB min:68kB low:84kB high:100kB active_anon:0kB > > inactive_anon:0kB active_file:0kB inactive_file:0kB unevictable:0kB > > writepending:0kB present:15992kB managed:15916kB mlocked:0kB > > slab_reclaimable:12460kB slab_unreclaimable:132kB kernel_stack:64kB > > pagetables:0kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB > > [32863.748796] lowmem_reserve[]: 0 798 4005 4005 > > [32863.748801] Normal free:3880kB min:3580kB low:4472kB high:5364kB > > active_anon:0kB inactive_anon:0kB active_file:1220kB inactive_file:68kB > > unevictable:0kB writepending:0kB present:892920kB managed:830896kB mlocked:0kB no anonymous memory is from eligible zones. There is some pagecache but 1.2MB doesn't sound all that much. There is a known regression from 4.8 when the active list aging is broken with memcg enabled but I am not sure this would make much of a difference here. You can try http://lkml.kernel.org/r/20170104100825.3729-1-mhocko@xxxxxxxxxx but it seems that the problem you are seeing is really the lowmem depletion which is hard to come around with 32b kernels. > > slab_reclaimable:732236kB slab_unreclaimable:70736kB kernel_stack:2560kB slab consumption is really high. It has eaten a majority of the lowmem. I would focus on who is eating that memory. Try to watch /proc/slabinfo for anomalies. > > pagetables:0kB bounce:0kB free_pcp:1252kB local_pcp:624kB free_cma:0kB -- Michal Hocko SUSE Labs -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>