On Tue 18-02-20 20:41:12, Sachin Sant wrote: > > >> Yes, I can recreate the same problem with the patch applied on top of > >> 5.6.0-rc2. > > > > And just to make sure. This was with http://lkml.kernel.org/r/fff0e636-4c36-ed10-281c-8cdb0687c839@xxxxxxxxxxxxx > > right? > > > Yes, the same patch. > > > If yes, is it possible that the specific node is somehow crippled (e.g. > > some nodes don't have any memory and thus the allocator blows up)? In > > other words what is the numa topology? (numactl -H) > > > > Here is the o/p of numactl > > # numactl -H > available: 2 nodes (0-1) > node 0 cpus: > node 0 size: 0 MB > node 0 free: 0 MB OK, so what I expected. The node0 is memory less or simply not present at all. Fun! Anyway, I do not think it is expected that kmalloc_node just blows up on those nodes. The page allocator simply falls back to the closest node. Something for kmalloc maintainers I believe. A short summary. kmalloc_node blows up when trying to allocate from a memory less node. > node 1 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 > node 1 size: 35247 MB > node 1 free: 30907 MB > node distances: > node 0 1 > 0: 10 40 > 1: 40 10 > # > > Thanks > -Sachin -- Michal Hocko SUSE Labs
