On Tue, Sep 23, 2014 at 07:48:27AM -0400, Johannes Weiner wrote: > On Tue, Sep 23, 2014 at 12:29:27PM +0400, Vladimir Davydov wrote: > > On Mon, Sep 22, 2014 at 10:52:50PM -0700, Greg Thelen wrote: > > > In this condition, if res usage is at limit then there's no point in > > > swapping because memsw.usage is already maximal. Prior to this patch > > > I think the kernel did the right thing, but not afterwards. > > > > > > Before this patch: > > > if res.usage == res.limit, try_charge() indirectly calls > > > try_to_free_mem_cgroup_pages(noswap=true) > > > > But this is wrong. If we fail to charge res, we should try to do swap > > out along with page cache reclaim. Swap out won't affect memsw.usage, > > but will diminish res.usage so that the allocation may succeed. > > But we know that the memsw limit must be hit as well in that case, and > swapping only makes progress in the sense that we are then succeeding > the memory charge. But we still fail to charge memsw. Yeah, I admit I said nonsense. The problem Greg pointed out does exist. I think your second patch (charging memsw before res) should fix it. Thanks, Vladimir -- 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
