On Tue, Aug 09, 2022 at 06:15:37PM +0900, Sergey Senozhatsky wrote: > On (22/08/09 11:12), Lukas Czerner wrote: > > Hi Jiri, > > > > I've tried a quick xfstests run on ext2 on zram and I can't see any > > issues like this so far. I will run a full test and report back in case > > there is anything obvious. > > AFAICT this should be visible only when we are under memory pressure, > so that direct reclaim from zs_malloc handle allocation call makes a > difference. > True, I haven't seen the other email from Jiri, sorry about that. I can confirm that under memory pressure it is in fact reproducible with xfstests and also I can confirm that reverting e7be8d1dd983156bbdd22c0319b71119a8fbb697 makes it go away. But Jiri has a better repro already anyway. Thanks! -Lukas
