linux-kernel - Re: [PATCH v2 08/10] mm/hugetlb: Make walk_hugetlb

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] [day] [month] [year] [list]

Message-ID: <Y5Nm5KXeb9Nb/oTX@x1n>
Date:   Fri, 9 Dec 2022 11:48:36 -0500
From:   Peter Xu <peterx@...hat.com>
To:     David Hildenbrand <david@...hat.com>
Cc:     linux-mm@...ck.org, linux-kernel@...r.kernel.org,
        Muchun Song <songmuchun@...edance.com>,
        John Hubbard <jhubbard@...dia.com>,
        Andrea Arcangeli <aarcange@...hat.com>,
        James Houghton <jthoughton@...gle.com>,
        Jann Horn <jannh@...gle.com>, Rik van Riel <riel@...riel.com>,
        Miaohe Lin <linmiaohe@...wei.com>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Mike Kravetz <mike.kravetz@...cle.com>,
        Nadav Amit <nadav.amit@...il.com>
Subject: Re: [PATCH v2 08/10] mm/hugetlb: Make walk_hugetlb_range() safe to
 pmd unshare

On Fri, Dec 09, 2022 at 04:18:11PM +0100, David Hildenbrand wrote:
> On 09.12.22 15:39, Peter Xu wrote:
> > On Fri, Dec 09, 2022 at 11:24:55AM +0100, David Hildenbrand wrote:
> > > For such cases, it would be good to have any evidence that it really helps.
> > 
> > I don't know much on the s390 path, but if a process has a large hugetlb
> > vma, even MADV_DONTNEED will be blocked for whatever long time if there's
> > another process or thread scanning pagemap for this vma.
> > 
> > Would this justify a bit?
> 
> I get your point. But that raises the question if we should voluntarily drop
> the VMA lock already in the caller every now and then on such large VMAs and
> maybe move even the cond_resched() into the common page walker, if you get
> what I mean?
> 
> On a preemtible kernel you could reschedule just before you drop the lock
> and call cond_resched() ... hmm

On full preempt the old cond_resched() doesn't work anyway.  IIUC most
distros now should be using dynamic preempt which makes voluntary mode by
default, which this change should help.

> 
> No strong opinion here, it just looked a bit weird to optimize for a
> cond_resched() if we might just reschedule either way even without the
> cond_resched().

It's really not the core of the patchset.  Let me drop it for now.

-- 
Peter Xu