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;
MIME-Version: 1.0
References: <20230609005158.2421285-1-surenb@google.com> <20230609005158.2421285-6-surenb@google.com>
 <ZIM/O54Q0waFq/tx@casper.infradead.org> <CAJuCfpE4VYz-Z4_aS3d9-8FGtQ-F4f7adYcJqRk3P3Ks7WPgQA@mail.gmail.com>
In-Reply-To: <CAJuCfpE4VYz-Z4_aS3d9-8FGtQ-F4f7adYcJqRk3P3Ks7WPgQA@mail.gmail.com>
From:   Suren Baghdasaryan <surenb@google.com>
Date:   Fri, 9 Jun 2023 11:55:29 -0700
Message-ID: <CAJuCfpG9JeHBKF0fzqR7xpDufpm7HVwgfbQVDeKYW24TWkpckw@mail.gmail.com>
Subject: Re: [PATCH v2 5/6] mm: implement folio wait under VMA lock
To:     Matthew Wilcox <willy@infradead.org>
Cc:     akpm@linux-foundation.org, hannes@cmpxchg.org, mhocko@suse.com,
        josef@toxicpanda.com, jack@suse.cz, ldufour@linux.ibm.com,
        laurent.dufour@fr.ibm.com, michel@lespinasse.org,
        liam.howlett@oracle.com, jglisse@google.com, vbabka@suse.cz,
        minchan@google.com, dave@stgolabs.net, punit.agrawal@bytedance.com,
        lstoakes@gmail.com, hdanton@sina.com, apopple@nvidia.com,
        peterx@redhat.com, ying.huang@intel.com, david@redhat.com,
        yuzhao@google.com, dhowells@redhat.com, hughd@google.com,
        viro@zeniv.linux.org.uk, brauner@kernel.org,
        pasha.tatashin@soleen.com, linux-mm@kvack.org,
        linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org,
        kernel-team@android.com
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Fri, Jun 9, 2023 at 11:49=E2=80=AFAM Suren Baghdasaryan <surenb@google.c=
om> wrote:
>
> On Fri, Jun 9, 2023 at 8:03=E2=80=AFAM Matthew Wilcox <willy@infradead.or=
g> wrote:
> >
> > On Thu, Jun 08, 2023 at 05:51:57PM -0700, Suren Baghdasaryan wrote:
> > >  static inline bool folio_lock_or_retry(struct folio *folio,
> > > -             struct mm_struct *mm, unsigned int flags)
> > > +             struct vm_area_struct *vma, unsigned int flags,
> > > +             bool *lock_dropped)
> >
> > I hate these double-return-value functions.
> >
> > How about this for an API:
> >
> > vm_fault_t folio_lock_fault(struct folio *folio, struct vm_fault *vmf)
> > {
> >         might_sleep();
> >         if (folio_trylock(folio))
> >                 return 0;
> >         return __folio_lock_fault(folio, vmf);
> > }
> >
> > Then the users look like ...
> >
> > > @@ -3580,8 +3581,10 @@ static vm_fault_t remove_device_exclusive_entr=
y(struct vm_fault *vmf)
> > >       if (!folio_try_get(folio))
> > >               return 0;
> > >
> > > -     if (!folio_lock_or_retry(folio, vma->vm_mm, vmf->flags)) {
> > > +     if (!folio_lock_or_retry(folio, vma, vmf->flags, &lock_dropped)=
) {
> > >               folio_put(folio);
> > > +             if (lock_dropped && vmf->flags & FAULT_FLAG_VMA_LOCK)
> > > +                     return VM_FAULT_VMA_UNLOCKED | VM_FAULT_RETRY;
> > >               return VM_FAULT_RETRY;
> > >       }
> >
> >         ret =3D folio_lock_fault(folio, vmf);
> >         if (ret)
> >                 return ret;
> >
> > > @@ -3837,9 +3840,9 @@ vm_fault_t do_swap_page(struct vm_fault *vmf)
> > >               goto out_release;
> > >       }
> > >
> > > -     locked =3D folio_lock_or_retry(folio, vma->vm_mm, vmf->flags);
> > > -
> > > -     if (!locked) {
> > > +     if (!folio_lock_or_retry(folio, vma, vmf->flags, &lock_dropped)=
) {
> > > +             if (lock_dropped && vmf->flags & FAULT_FLAG_VMA_LOCK)
> > > +                     ret |=3D VM_FAULT_VMA_UNLOCKED;
> > >               ret |=3D VM_FAULT_RETRY;
> > >               goto out_release;
> > >       }
> >
> >         ret |=3D folio_lock_fault(folio, vmf);
> >         if (ret & VM_FAULT_RETRY)
> >                 goto out_release;
> >
> > ie instead of trying to reconstruct what __folio_lock_fault() did from
> > its outputs, we just let folio_lock_fault() tell us what it did.
>
> Thanks for taking a look!
> Ok, I think what you are suggesting is to have a new set of
> folio_lock_fault()/__folio_lock_fault() functions which return
> vm_fault_t directly, __folio_lock_fault() will use
> __folio_lock_or_retry() internally and will adjust its return value
> based on __folio_lock_or_retry()'s return and the lock releasing rules
> described in the comments for __folio_lock_or_retry(). Is my
> understanding correct?

Oh, after rereading I think you are suggesting to replace
folio_lock_or_retry()/__folio_lock_or_retry() with
folio_lock_fault()/__folio_lock_fault(), not to add them. Is that
right?