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, 27 Jun 2022 11:29:47 -0400
From:   Peter Xu <peterx@redhat.com>
To:     Jason Gunthorpe <jgg@ziepe.ca>
Cc:     kvm@vger.kernel.org, linux-kernel@vger.kernel.org,
        Paolo Bonzini <pbonzini@redhat.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        David Hildenbrand <david@redhat.com>,
        "Dr . David Alan Gilbert" <dgilbert@redhat.com>,
        Andrea Arcangeli <aarcange@redhat.com>,
        Linux MM Mailing List <linux-mm@kvack.org>,
        Sean Christopherson <seanjc@google.com>
Subject: Re: [PATCH 1/4] mm/gup: Add FOLL_INTERRUPTIBLE
Message-ID: <YrnM6x7QWo6NmFqf@xz-m1.local>
References: <20220622213656.81546-1-peterx@redhat.com>
 <20220622213656.81546-2-peterx@redhat.com>
 <20220625003554.GJ23621@ziepe.ca>
 <YrZjeEv1Z2IDMwgy@xz-m1.local>
 <20220625235904.GK23621@ziepe.ca>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <20220625235904.GK23621@ziepe.ca>
Precedence: bulk

On Sat, Jun 25, 2022 at 08:59:04PM -0300, Jason Gunthorpe wrote:
> On Fri, Jun 24, 2022 at 09:23:04PM -0400, Peter Xu wrote:
> > If to go back to the original question with a shorter answer: if the ioctl
> > context that GUP upon a page that will never be with a uffd context, then
> > it's probably not gonna help at all.. at least not before we use
> > FAULT_FLAG_INTERRUPTIBLE outside uffd page fault handling.
> 
> I think I would be more interested in this if it could abort a swap
> in, for instance. Doesn't this happen if it flows the interruptible
> flag into the VMA's fault handler?

The idea makes sense, but it doesn't work like that right now, afaict. We
need to teach lock_page_or_retry() to be able to consume the flag as
discussed.

I don't see a major blocker for it if lock_page_or_retry() is the only one
we'd like to touch.  Say, the only caller currently is do_swap_page()
(there's also remove_device_exclusive_entry() but just deeper in the
stack). Looks doable so far but I'll need to think about it..  I can keep
you updated if I get something, but it'll be separate from this patchset.

Thanks,

-- 
Peter Xu