Re: [patch] mm, oom: stop reclaiming if GFP_ATOMIC will start failing soon

[Tetsuo Handa]

On 2020/04/26 9:27, Andrew Morton wrote:
> Well, what's really going on here?
> 
> Is networking potentially consuming an unbounded amount of memory?  If
> so, then killing a process will just cause networking to consume more
> memory then hit against the same thing.  So presumably the answer is
> "no, the watermarks are inappropriately set for this workload".
Maybe somebody is abusing __GFP_MEMALLOC allocations. David, please include whole
context (e.g. which function from which process) instead of only memory stat lines.