On Fri, Oct 24, 2014 at 11:27:49AM +0900, Minchan Kim wrote: > Hi Joonsoo, > > I know you spend much effort for investigate/fix this subtle problem. > So, you should be hero. > > Thanks for really nice work! Hello, Thanks. :) > > > > Joonsoo Kim (4): > > mm/page_alloc: fix incorrect isolation behavior by rechecking > > migratetype > > mm/page_alloc: add freepage on isolate pageblock to correct buddy > > list > > mm/page_alloc: move migratetype recheck logic to __free_one_page() > > So, [1-3], > Acked-by: Minchan Kim <minchan@xxxxxxxxxx> Thanks, too. > > mm/page_alloc: restrict max order of merging on isolated pageblock > > As you noted in description, this patch has a side effect which doesn't > merge buddies. Most of all, I agree your assumptions but it's not true always. > > Who knows there is a driver which want a higher page above pageblock? > Who knows there is no allocation/free of the isolated range right before > highest allocation request? > Even, your patch introduces new exception rule for page allocator. > > "Hey, allocator, from now on, you could have unmerged buddies > in your list so please advertise it to your customer" > > So, all of users of the allocator should consider that exception so > it might hit us sometime. > > I want to fix that in isolation undo time. > Thanks, again! Okay. I will try it. The reason I implement as current is that it makes process of isolation/un-isolation asymetric and needs to copy and paste some code to handle this specialty. That would possibly result in maintainance overhead. But, yes, exception of buddy property is also bad situation. I will implement it and send it soon. Thanks. -- 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>