[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <30d86a2d-4af2-d840-91be-2e68c73a07bd@redhat.com>
Date: Tue, 1 Aug 2023 18:15:48 +0200
From: David Hildenbrand <david@...hat.com>
To: Peter Xu <peterx@...hat.com>
Cc: linux-kernel@...r.kernel.org, linux-mm@...ck.org,
linux-fsdevel@...r.kernel.org, kvm@...r.kernel.org,
linux-kselftest@...r.kernel.org,
Andrew Morton <akpm@...ux-foundation.org>,
Linus Torvalds <torvalds@...ux-foundation.org>,
liubo <liubo254@...wei.com>,
Matthew Wilcox <willy@...radead.org>,
Hugh Dickins <hughd@...gle.com>,
Jason Gunthorpe <jgg@...pe.ca>,
John Hubbard <jhubbard@...dia.com>,
Mel Gorman <mgorman@...e.de>, Shuah Khan <shuah@...nel.org>,
Paolo Bonzini <pbonzini@...hat.com>, stable@...r.kernel.org
Subject: Re: [PATCH v2 1/8] mm/gup: reintroduce FOLL_NUMA as
FOLL_HONOR_NUMA_FAULT
On 01.08.23 17:48, Peter Xu wrote:
> On Tue, Aug 01, 2023 at 02:48:37PM +0200, David Hildenbrand wrote:
>> @@ -2240,6 +2244,12 @@ static bool is_valid_gup_args(struct page **pages, int *locked,
>> gup_flags |= FOLL_UNLOCKABLE;
>> }
>>
>> + /*
>> + * For now, always trigger NUMA hinting faults. Some GUP users like
>> + * KVM really require it to benefit from autonuma.
>> + */
>> + gup_flags |= FOLL_HONOR_NUMA_FAULT;
>
> Since at it, do we want to not set it for FOLL_REMOTE, which still sounds
> like a good thing to have?
I thought about that, but decided against making that patch here more
complicated to eventually rip it again all out in #4.
I fully agree that FOLL_REMOTE does not make too much sense, but let's
rather keep it simple for this patch.
Thanks!
>
> Other than that, looks good here.
>
> Side note: when I was looking at the flags again just to check the
> interactions over numa balancing, I found FOLL_NOFAULT and I highly suspect
> that's not needed, instead it just wants to use follow_page[_mask]() with
> some proper gup flags passed over.. but that's off topic.
Be prepared for my proposal of removing foll_flags from follow_page() ;)
(accompanied by a proper documentation)
Especially as we have FOLL_PIN users of FOLL_NOFAULT, follow_page() is a
bad fit.
--
Cheers,
David / dhildenb
Powered by blists - more mailing lists