On Fri 20-10-17 14:43:06, Du, Changbin wrote: > On Thu, Oct 19, 2017 at 11:52:49PM +0500, Михаил Гаврилов wrote: > > On 19 October 2017 at 08:56, Du, Changbin <changbin.du@xxxxxxxxx> wrote: > > > On Thu, Oct 19, 2017 at 01:16:48AM +0500, Михаил Гаврилов wrote: > > > I am curious about this, how can slub try to alloc compound page but the order > > > is 0? This is wrong. > > > > Nobody seems to know how this could happen. Can any logs shed light on this? > > > After checking the code, kernel can handle such case. So please ignore my last > comment. > > The warning is reporting OOM, first you need confirm if you have enough free > memory? If that is true, then it is not a programmer error. The kernel is not OOM. It just failed to allocate for GFP_NOWAIT which means that no memory reclaim could be used to free up potentially unused page cache. This means that kswapd is not able to free up memory in the pace it is allocated. Such an allocation failure shouldn't be critical and the caller should have means to fall back to a regular allocation or retry later. You can play with min_free_kbytes and increase it to kick the background reclaim sooner. -- 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>
