On Tue, Aug 31, 2021 at 01:44:49PM +0100, Mel Gorman wrote: > That's your answer -- the PCP count has been corrupted or misaccounted. > Given this is a Fedora kernel, check for any patches affecting > mm/page_alloc.c that could be accounting related or that would affect > the IRQ disabling or zone lock acquisition for problems. Another > possibility is memory corruption -- either kernel or the hardware > itself. Hmm, I don't see any changes to mm/page_alloc.c from Fedora for this kernel. What about a memory allocation originating from inside an NMI? I think this could occur quite easily with an eBPF program registered to a tracepoint, as some of the eBPF helpers do allocate memory on the fly for certain map types. > > I tried to find some way that this could happen, but the only thing I could > > think of was that maybe an allocation had both __GFP_RECLAIMABLE and > > __GFP_MOVABLE set in its gfp mask, in which case the rmqueue() call in > > get_page_from_freelist() would pass in a migratetype equal to MIGRATE_PCPTYPES > > and then pages could be added to an out-of-bounds pcp list while still > > incrementing the overall pcp count. This seems pretty unlikely though. > > It's unlikely because it would be an outright bug to specify both flags. Perhaps that VM_WARN_ON should be changed to a VM_BUG_ON? Thanks, Sultan
