Re: [PATCH] secretmem: Conert page_is_secretmem() to folio_is_secretmem()

[David Hildenbrand <david@redhat.com>]

On 21.08.23 17:53, Matthew Wilcox wrote:
> On Mon, Aug 21, 2023 at 05:32:50PM +0200, David Hildenbrand wrote:
>>> -	if (PageCompound(page) || !PageLRU(page))
>>> +	if (folio_test_large(folio) || folio_test_lru(folio))
>>>    		return false;
>>
>> You converted a "!PageLRU(page)" into a "folio_test_lru(folio)".
>>
>> Shouldn't that be a "!folio_test_lru(folio)" ?
> 
> Well, that's embarrassing.  Obviously I don't have any secretmem
> hardware, so my smoke-test didn't catch it.  Thanks, will do a v2.
> 

We do have a selftest:

tools/testing/selftests/mm/memfd_secret.c

I doubt it would have caught it, because ptrace doesn't use GUP-fast 
IIRC, so we always fail early in GUP code when doing the 
vma_is_secretmem() check.

We'd explicitly have to call something that triggers GUP-fast 
(get_user_pages_fast/pin_user_pages_fast).

-- 
Cheers,

David / dhildenb