On Wed, Jul 16, 2014 at 02:10:07PM -0400, Naoya Horiguchi wrote: > Hi, > > It seems that when a process in some memcg tries to allocate more than > memcg.limit_in_bytes, oom happens instead of swaping out in > mmotm-2014-07-09-17-08 (memcg.memsw.limit_in_bytes is large enough). > It does work in v3.16-rc3, so I think latest patches changed something. > I'm not familiar with memcg internally, so no idea about what caused it. > Could you see the problem? There are a lot of changes in memory and swap accounting, but I can not reproduce what you are describing: I set up a cgroup with a 100MB memory limit and an unlimited memory+swap, then start a task in there that faults 200MB worth of anonymous pages. The result is 100MB in memory, 100MB in swap: cache 0 rss 104267776 rss_huge 0 mapped_file 0 writeback 0 swap 105545728 pgpgin 26367 pgpgout 25950 pgfault 26695 pgmajfault 32 inactive_anon 52285440 active_anon 51982336 inactive_file 0 active_file 0 unevictable 0 hierarchical_memory_limit 104857600 hierarchical_memsw_limit 18446744073709551615 Filename Type Size Used Priority /swapfile file 8388604 109800 -1 Could you provide more detail on your configuration and test case? Thanks! -- To unsubscribe from this list: send the line "unsubscribe cgroups" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
