Re: [PATCH] mm/memory-failure: fix infinite UCE for VM_PFNMAP pfn

[David Hildenbrand <david@redhat.com>]

On 07.08.25 13:06, Jinjiang Tu wrote:
> 
> 在 2025/8/6 20:39, David Hildenbrand 写道:
>> On 06.08.25 04:05, Jinjiang Tu wrote:
>>> When memory_failure() is called for a already hwpoisoned pfn,
>>> kill_accessing_process() will be called to kill current task.
>>> However, if
>>> the vma of the accessing vaddr is VM_PFNMAP, walk_page_range() will skip
>>> the vma in walk_page_test() and return 0.
>>
>> That check is dubious and it's on my todo list to revisit that,
>> because it doesn't make sense for COW VM_PFNMAP mappings.
>>
>> But I am curious, is this about having an anon page in a COW
>> VM_PFNMAP, or on "what kidn of page" do you get the fault?
> The pfn is RAM pfn, which is mapped with remap_pfn_range(). There is no
> COW.
> 

Ah, and that works because in check_hwpoisoned_entry() we are only using 
pte_pfn().

-- 
Cheers,

David / dhildenb