linux-kernel - Re: [PATCH v1 1/2] mm: let pte_lockptr() consume a pte

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: <Zqk0mzD68lImF3y4@x1n>
Date: Tue, 30 Jul 2024 14:44:43 -0400
From: Peter Xu <peterx@...hat.com>
To: David Hildenbrand <david@...hat.com>
Cc: linux-kernel@...r.kernel.org, linux-mm@...ck.org,
	Andrew Morton <akpm@...ux-foundation.org>,
	Muchun Song <muchun.song@...ux.dev>,
	Oscar Salvador <osalvador@...e.de>,
	Qi Zheng <zhengqi.arch@...edance.com>,
	Hugh Dickins <hughd@...gle.com>
Subject: Re: [PATCH v1 1/2] mm: let pte_lockptr() consume a pte_t pointer

On Mon, Jul 29, 2024 at 07:46:26PM +0200, David Hildenbrand wrote:
> I see what you mean but this is a very similar pattern as used in
> collapse_pte_mapped_thp(), no? There we have
> 
> start_pte = pte_offset_map_nolock(mm, pmd, haddr, &ptl);
> ...
> if (!pml)
> 	spin_lock(ptl);
> ...
> pte_unmap(start_pte);
> if (!pml)
> 	spin_unlock(ptl);
> 
> 
> Again, I don't have a strong opinion on this, but doing it more similar to
> collapse_pte_mapped_thp() to obtain locks makes it clearer to me. But if I
> am missing something obvious please shout and I'll change it.

Right.. I don't think that path can change the pte pgtable either, and
there is even the line Hugh left showing it's impossible:

	if (!start_pte)		/* mmap_lock + page lock should prevent this */
		goto abort;

I was thinking maybe the page lock is the critical one, irrelevant of mmap
lock.

No strong opinion either.  Not sure whether Hugh has some thoughts.  But
maybe if we stick with pte_offset_map_nolock() and if there'll be a repost
anyway, we could add a similar comment like this one showing that the pte
pgtable should be actually as stable as the ptlock.

Thanks,

-- 
Peter Xu