Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Mon, 31 Jul 2023 12:05:35 -0400
From:   Peter Xu <peterx@redhat.com>
To:     John Hubbard <jhubbard@nvidia.com>
Cc:     David Hildenbrand <david@redhat.com>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        linux-kernel@vger.kernel.org, linux-mm@kvack.org,
        linux-fsdevel@vger.kernel.org,
        Andrew Morton <akpm@linux-foundation.org>,
        liubo <liubo254@huawei.com>,
        Matthew Wilcox <willy@infradead.org>,
        Hugh Dickins <hughd@google.com>, Jason Gunthorpe <jgg@ziepe.ca>
Subject: Re: [PATCH v1 0/4] smaps / mm/gup: fix gup_can_follow_protnone
 fallout
Message-ID: <ZMfbz2JFKvAkaKg8@x1n>
References: <CAHk-=wiig=N75AGP7UAG9scmghWAqsTB5NRO6RiWLOB5YWfcTQ@mail.gmail.com>
 <ZMQZfn/hUURmfqWN@x1n>
 <CAHk-=wgRiP_9X0rRdZKT8nhemZGNateMtb366t37d8-x7VRs=g@mail.gmail.com>
 <e74b735e-56c8-8e62-976f-f448f7d4370c@redhat.com>
 <CAHk-=wgG1kfPR6vtA2W8DMFOSSVMOhKz1_w5bwUn4_QxyYHnTA@mail.gmail.com>
 <69a5f457-63b6-2d4f-e5c0-4b3de1e6c9f1@redhat.com>
 <ZMQxNzDcYTQRjWNh@x1n>
 <edd9b468-2d60-1df7-a515-22475fd94fe2@nvidia.com>
 <ZMQ32RRJlW/aDYAE@x1n>
 <118e571c-a79f-5020-b9fd-4c0a3722236d@nvidia.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <118e571c-a79f-5020-b9fd-4c0a3722236d@nvidia.com>
Precedence: bulk

On Fri, Jul 28, 2023 at 03:00:04PM -0700, John Hubbard wrote:
> On 7/28/23 14:49, Peter Xu wrote:
> > > The story of how FOLL_NUMA and FOLL_FORCE became entangled was enlightening,
> > > by the way, and now that I've read it I don't want to go back. :)
> > 
> > Yeah I fully agree we should hopefully remove the NUMA / FORCE
> > tangling.. even if we want to revert back to the FOLL_NUMA flag we may want
> > to not revive that specific part.  I had a feeling that we're all on the
> > same page there.
> > 
> 
> Yes, I think so. :)
> 
> > It's more about the further step to make FOLL_NUMA opt-in for GUP.
> 
> Let's say "FOLL_HONOR_NUMA_FAULT" for this next discussion, but yes. So
> given that our API allows passing in FOLL_ flags, I don't understand the
> objection to letting different callers pass in, or not pass in, that
> flag.
> 
> It's the perfect way to clean up the whole thing. As Linus suggested
> slightly earlier here, there can be a comment at the call site,
> explaining why KVM needs FOLL_HONOR_NUMA_FAULT, and you're good, right?

I'm good even if we want to experiment anything, as long as (at least) kvm
is all covered then I'm not against it, not yet strongly.

But again, IMHO we're not only talking about "cleaning up" of any flag - if
that falls into "cleanup" first, which I'm not 100% sure yet - we're
takling about changing GUP abi.  What I wanted to point out to be careful
is we're literally changing the GUP abi for all kernel modules on numa
implications.

Thanks,

-- 
Peter Xu