linux-kernel - Re: [PATCH 2/5] mm,hugetlb: Document the reason to lock the folio in the faulting path

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <a5e098d1-ee5a-447f-9e05-0187b22500e1@redhat.com>
Date: Mon, 16 Jun 2025 11:22:43 +0200
From: David Hildenbrand <david@...hat.com>
To: Oscar Salvador <osalvador@...e.de>
Cc: Andrew Morton <akpm@...ux-foundation.org>,
 Muchun Song <muchun.song@...ux.dev>, James Houghton <jthoughton@...gle.com>,
 Peter Xu <peterx@...hat.com>, Gavin Guo <gavinguo@...lia.com>,
 linux-mm@...ck.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 2/5] mm,hugetlb: Document the reason to lock the folio in
 the faulting path

On 14.06.25 11:07, Oscar Salvador wrote:
> On Fri, Jun 13, 2025 at 11:47:50PM +0200, Oscar Salvador wrote:
>> Maybe it's because it's Friday, but I'm confused as to why
>> do_pte_missing()->do_fault()->do_cow_fault() holds the lock while do_wp_page() doesn't
>> although it might the file's page we have to copy.
> 
> Scratch that, I see my confusion.
> The first time we map the file privately, the folio must remain stable.
> But if we already mapped it privately before (R/O), and we write fault on it,
> we don't need to be stable (e.g: uptodated).
> 
> But I think my comment on hugetlb_no_page() still holds, because
> 
> hugetlb_fault->hugetlb_no_page->hugetlb_wp
> 
> would be similar to do_pte_missing->do_cow, and in do_cow we hold both
> the reference and the lock.

Well, there is an important difference:

hugetlb_fault->hugetlb_no_page->hugetlb_wp

already *mapped* the pagecache page into the page table.

See
	if (anon_rmap)
		hugetlb_add_new_anon_rmap(folio, vma, vmf->address);
	else
		hugetlb_add_file_rmap(folio);

So at that point it would be "stable" unless I am missing something?

So once we are in hugetlb_wp(), that path much rather corresponds to 
do_wp_page()->wp_page_copy.

> Were we might not need the lock is in hugetlb_fault->hugetlb_wp, which
> would be similar to do_wp_page()->wp_page_copy.

Exactly.

> Of course we will need to take it if it is an anonymous folio because we need
> to check the re-use case.

Yes.

> 
> So, it gets complicated because hugetlb_no_page() needs to call
> hugetlb_wp() with the lock held in case it is a pagecache folio,

Per above discussion: why? After we mapped the pagecache folio, we can 
unlock the folio I think.

and
> and the same time hugetlb_wp() needs to take the lock if it us an anonymous
> one for the re-use case.

I think it hugetlb_wp() really only needs the lock for the anon folio.

-- 
Cheers,

David / dhildenb