Re: [PATCH 1/2] mm: memory_hotplug: add TTU_HWPOISON for poisoned folio during migrate

[David Hildenbrand]

> We have faced a different race here as follow:
>
> CPU#0			cpu #1
> 			offline_pages
> 			do_migrate_range
> memory_failure
> TestSetPageHWPoison
> 			// folio is mark poison here
> 			unmap_poisoned_folio // should kill task here
> ...
> hwpoison_user_mappings

Interesting; TestSetPageHWPoison is called before we grab the folio 
lock, so even grabbing that around unmap_poisoned_folio() cannot 
completely help.

--
Cheers,

David / dhildenb