Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20200618222225.102337-1-axelrasmussen@google.com>
 <20200618222225.102337-2-axelrasmussen@google.com> <20200619083023.GS576905@hirez.programming.kicks-ass.net>
 <CAJHvVciGnK9Ve7NeZ=LUbWT3PGhG7hRWh+J80ncR6Mgwk316fw@mail.gmail.com>
In-Reply-To: <CAJHvVciGnK9Ve7NeZ=LUbWT3PGhG7hRWh+J80ncR6Mgwk316fw@mail.gmail.com>
From:   Axel Rasmussen <axelrasmussen@google.com>
Date:   Thu, 25 Jun 2020 13:25:56 -0700
Message-ID: <CAJHvVch0py68FBc4sDfc+7xN7=oyfHXvM1gwun4hZrBdzrHr5Q@mail.gmail.com>
Subject: Re: [RFC PATCH v3 1/1] mmap_lock: add tracepoints around mmap_lock acquisition
To:     Peter Zijlstra <peterz@infradead.org>
Cc:     Michel Lespinasse <walken@google.com>,
        David Howells <dhowells@redhat.com>,
        linux-kernel@vger.kernel.org, linux-mm@kvack.org,
        Jonathan Adams <jwadams@google.com>,
        David Rientjes <rientjes@google.com>,
        Ying Han <yinghan@google.com>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Peter, one additional question for you.

Would you accept this patch if:
- The rwsem.{h,c} and lockdep.h changes were reverted
- It was split up into 2-3 commits
?

Thanks!

On Fri, Jun 19, 2020 at 9:28 AM Axel Rasmussen <axelrasmussen@google.com> wrote:
>
> On Fri, Jun 19, 2020 at 1:30 AM Peter Zijlstra <peterz@infradead.org> wrote:
> >
> > On Thu, Jun 18, 2020 at 03:22:25PM -0700, Axel Rasmussen wrote:
> > > diff --git a/kernel/locking/rwsem.c b/kernel/locking/rwsem.c
> > > index f11b9bd3431d..6aabea1cbc5d 100644
> > > --- a/kernel/locking/rwsem.c
> > > +++ b/kernel/locking/rwsem.c
> > > @@ -1495,6 +1495,20 @@ void __sched down_read(struct rw_semaphore *sem)
> > >  }
> > >  EXPORT_SYMBOL(down_read);
> > >
> > > +/*
> > > + * lock for reading
> > > + */
> > > +void __sched down_read_contended_hook(struct rw_semaphore *sem,
> > > +                                   void (*pre)(void *),
> > > +                                   void (*post)(void *), void *arg)
> > > +{
> > > +     might_sleep();
> > > +     rwsem_acquire_read(&sem->dep_map, 0, 0, _RET_IP_);
> > > +     LOCK_CONTENDED_HOOK(sem, __down_read_trylock, __down_read, pre, post,
> > > +                         arg);
> > > +}
> > > +EXPORT_SYMBOL(down_read_contended_hook);
> > > +
> > >  int __sched down_read_killable(struct rw_semaphore *sem)
> > >  {
> > >       might_sleep();
> > > @@ -1509,6 +1523,24 @@ int __sched down_read_killable(struct rw_semaphore *sem)
> > >  }
> > >  EXPORT_SYMBOL(down_read_killable);
> > >
> > > +int __sched down_read_killable_contended_hook(struct rw_semaphore *sem,
> > > +                                           void (*pre)(void *),
> > > +                                           void (*post)(void *, int),
> > > +                                           void *arg)
> > > +{
> > > +     might_sleep();
> > > +     rwsem_acquire_read(&sem->dep_map, 0, 0, _RET_IP_);
> > > +
> > > +     if (LOCK_CONTENDED_HOOK_RETURN(sem, __down_read_trylock,
> > > +                                    __down_read_killable, pre, post, arg)) {
> > > +             rwsem_release(&sem->dep_map, _RET_IP_);
> > > +             return -EINTR;
> > > +     }
> > > +
> > > +     return 0;
> > > +}
> > > +EXPORT_SYMBOL(down_read_killable_contended_hook);
> > > +
> > >  /*
> > >   * trylock for reading -- returns 1 if successful, 0 if contention
> > >   */
> > > @@ -1533,6 +1565,20 @@ void __sched down_write(struct rw_semaphore *sem)
> > >  }
> > >  EXPORT_SYMBOL(down_write);
> > >
> > > +/*
> > > + * lock for writing
> > > + */
> > > +void __sched down_write_contended_hook(struct rw_semaphore *sem,
> > > +                                    void (*pre)(void *),
> > > +                                    void (*post)(void *), void *arg)
> > > +{
> > > +     might_sleep();
> > > +     rwsem_acquire(&sem->dep_map, 0, 0, _RET_IP_);
> > > +     LOCK_CONTENDED_HOOK(sem, __down_write_trylock, __down_write, pre, post,
> > > +                         arg);
> > > +}
> > > +EXPORT_SYMBOL(down_write_contended_hook);
> > > +
> > >  /*
> > >   * lock for writing
> > >   */
> > > @@ -1551,6 +1597,24 @@ int __sched down_write_killable(struct rw_semaphore *sem)
> > >  }
> > >  EXPORT_SYMBOL(down_write_killable);
> > >
> > > +int __sched down_write_killable_contended_hook(struct rw_semaphore *sem,
> > > +                                            void (*pre)(void *),
> > > +                                            void (*post)(void *, int),
> > > +                                            void *arg)
> > > +{
> > > +     might_sleep();
> > > +     rwsem_acquire_read(&sem->dep_map, 0, 0, _RET_IP_);
> > > +
> > > +     if (LOCK_CONTENDED_HOOK_RETURN(sem, __down_write_trylock,
> > > +                                    __down_write_killable, pre, post, arg)) {
> > > +             rwsem_release(&sem->dep_map, _RET_IP_);
> > > +             return -EINTR;
> > > +     }
> > > +
> > > +     return 0;
> > > +}
> > > +EXPORT_SYMBOL(down_write_killable_contended_hook);
> > > +
> > >  /*
> > >   * trylock for writing -- returns 1 if successful, 0 if contention
> > >   */
> >
> > NAK, absolutely not going to happen. This is an atrocious API to expose,
> > worse you're exporting.
>
> Ack about splitting this up.
>
> Thanks for taking a look. :)