On Fri 15-07-16 13:02:17, Mikulas Patocka wrote: > > > On Fri, 15 Jul 2016, Michal Hocko wrote: > > > On Fri 15-07-16 08:11:22, Mikulas Patocka wrote: > > > > > > The stacktraces showed that the kcryptd process was throttled when it > > > tried to do mempool allocation. Mempool adds the __GFP_NORETRY flag to the > > > allocation, but unfortunatelly, this flag doesn't prevent the allocator > > > from throttling. > > > > Yes and in fact it shouldn't prevent any throttling. The flag merely > > says that the allocation should give up rather than retry > > reclaim/compaction again and again. > > > > > I say that the process doing mempool allocation shouldn't ever be > > > throttled. Maybe add __GFP_NOTHROTTLE? > > > > A specific gfp flag would be an option but we are slowly running out of > > bit space there and I am not yet convinced PF_LESS_THROTTLE is > > unsuitable. > > PF_LESS_THROTTLE will make it throttle less, but it doesn't eliminate > throttling entirely. So, maybe add PF_NO_THROTTLE? But PF_* flags are also > almost exhausted. I am not really sure we can make anybody so special to not throttle at all. Seeing a congested backig device sounds like a reasonable compromise. Besides that it seems that we do not really need to eliminate wait_iff_congested for dm to work properly again AFAIU. I plan to repost both patch today after some more internal review. If we need to do more changes I would suggest making them in separet patches. -- Michal Hocko SUSE Labs -- dm-devel mailing list dm-devel@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/dm-devel
