On Tue 18-02-20 20:41:12, Sachin Sant wrote:

quoted

quoted

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@virtuozzo.com
right?

Yes, the same patch.

quoted

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