Re: [PATCH v4] locking/rwsem: Make handoff bit handling more consistent

[Waiman Long]

On 11/15/21 11:01, Peter Zijlstra wrote:
> On Sun, Nov 14, 2021 at 10:38:57PM -0500, Waiman Long wrote:
> > On 11/12/21 07:10, Peter Zijlstra wrote:
> > > Argh, rwsem_mark_wake() doesn't clear HANDOFF when list_empty(), and
> > > write_slowpath() is *far* too clever about all of this.
> > rwsem_mark_wake() does clear the HANDOFF flag if it was set.
> Argh, yeah, I got confused by the whole !woken case, but that case won't
> ever hit list_empty() either. Perhaps that stuff could use a bit of a
> reflow too.
I think your modification already have included the rewrite for that part.
> > > > @@ -1098,7 +1110,7 @@ rwsem_down_write_slowpath(struct rw_semaphore *sem, int state)
> > > >    		 * In this case, we attempt to acquire the lock again
> > > >    		 * without sleeping.
> > > >    		 */
> > > > -		if (wstate == WRITER_HANDOFF) {
> > > > +		if (waiter.handoff_set) {
> > > I'm thinking this wants to be something like:
> > >
> > > 		if (rwsem_first_waiter(sem) == &waiter && waiter.handoff_set) {
> > handoff_set flag is only set when the waiter becomes the first.
> Yes, but a random waiter can wake up and see it be set and also start
> spinning.

The handoff_set flag can only be true for a first waiter. A random 
waiter in the middle of a wait queue will never has this flag set.

This flag is set in two places in rwsem_try_write_lock():

1)

               if (has_handoff && !first)
                        return false;
                new = count;

                if (count & RWSEM_LOCK_MASK) {
                        /*
                         * Only the first waiter can inherit a 
previously set
                         * handoff bit.
                         */
                        waiter->handoff_set = has_handoff;

handoff_set can only be set to true here if first is also true. In that 
case, it will also return false immediately afterward.

2)

        if (new & RWSEM_FLAG_HANDOFF) {
                waiter->handoff_set = true;
                lockevent_inc(rwsem_wlock_handoff);
                return false;
        }

Again, only first waiter will have a chance of setting the handoff bit 
and have handoff_set set to true.

> > > >    			enum owner_state owner_state;
> > > >    			preempt_disable();
> > > @@ -575,6 +610,11 @@ static inline bool rwsem_try_write_lock(
> > >    		return false;
> > >    	}
> > > +	/*
> > > +	 * Have rwsem_try_write_lock() fully imply rwsem_del_waiter() on
> > > +	 * success.
> > > +	 */
> > > +	list_del(&waiter->list);
> > >    	rwsem_set_owner(sem);
> > >    	return true;
> > >    }
> > > @@ -1128,16 +1153,14 @@ rwsem_down_write_slowpath(struct rw_sema
> > >    		raw_spin_lock_irq(&sem->wait_lock);
> > >    	}
> > >    	__set_current_state(TASK_RUNNING);
> > > -	list_del(&waiter.list);
> > +    rwsem_del_waiter(sem, &waiters); ?
> I tried that, but then we get an extra atomic in this path. As is I made
> try_write_lock() do the full del_waiter, see the hunk above.

You are right. I missed your change in rwsem_try_write_lock().

Thanks,
Longman