On 2020/04/27 12:12, David Rientjes wrote: > Tetsuo: the specific allocation that triggers a page allocation failure is > not interesting; we have tens of thousands of examples. Each example is > simply the unlucky last GFP_ATOMIC allocation that fails; the interesting > point is the amount of free memory. In other words, when free memory is > below ALLOC_HIGH watermarks, we assume that we have depleted memory > reserves *faster* than when user allocations started to fail. In the > interest of userspace being responsive, we should oom kill here. My interest is, which function (and which process if process context) is [ab]using GFP_ATOMIC (or __GFP_MEMALLOC) allocations enough to hit memory allocation failure. GFP_NOWAIT (or __GFP_NOMEMALLOC) could be used if that allocation can't sleep and can't shortly recover free memory.
