On 1/17/2023 3:39 PM, Vlastimil Babka wrote: > So either the assumption of refcount being already 0 is wrong (shouldn't be, > AFAIK?), or this atomic operation effectively prevents some very subtle race > (although IIRC atomic_set() has no barrier semantics defined, it could still > affect a specific CPU? Is there difference of atomic_set() for 32bit and 64bit? This issue is only triggered on 32bit configuration. > > I guess we could > - try to restore that set_page_count(p, 0); on current kernel to see if it > kills the bug I tried this. But didn't try too hard. 20 times reboot didn't trigger the issue with this line added to commit 0af8489b02. Without this line, hit the issue 4 times in 10 times reboot testing. Regards Yin, Fengwei
