Re: [PATCH v2 8/8] mm: Centralize & improve oom reporting in show_mem.c

[Michal Hocko]

On Tue 26-04-22 03:26:12, Kent Overstreet wrote:
[...]
> Anyways, the reason I think this allocation is fine is it's GFP_NOWAIT and it's
> completely fine if it fails - all we lose is some diagnostics, and also it's
> released right away.

I think you are still missing the PF_MEMALLOC point. Please have a look
how this leads to no reclaim recursion so GFP_NOWAIT has no meaning when
you are allocating from PF_MEMALLOC context (which the oom killer and
any reclaim path is). Also have a look at how __gfp_pfmemalloc_flags
makes the allocation request from that context ALLOC_NO_WATERMARKS.
See?
-- 
Michal Hocko
SUSE Labs