Re: [PATCH v3 2/5] locking/rwsem: Limit # of null owner retries for handoff writer

[Waiman Long]

On 10/24/22 09:31, Peter Zijlstra wrote:
> On Mon, Oct 17, 2022 at 05:13:53PM -0400, Waiman Long wrote:
> > Commit 91d2a812dfb9 ("locking/rwsem: Make handoff writer optimistically
> > spin on owner") assumes that when the owner field is changed to NULL,
> > the lock will become free soon.  That assumption may not be correct
> > especially if the handoff writer doing the spinning is a RT task which
> > may preempt another task from completing its action of either freeing
> > the rwsem or properly setting up owner.
> I'm confused again -- rwsem_*_owner() has
> lockdep_assert_preemption_disabled(). But more specifically; why can the
> RT task preempt a lock-op like that?

There is a special case raised by Mukesh that can happen. I quoted his 
text here:

---------------------------

Looks like, There is still a window for a race.

There is a chance when a reader who came first added it's BIAS and goes 
to slowpath and before it gets added to wait list it got preempted by RT 
task which  goes to slowpath as well and being the first waiter gets its 
hand-off bit set and not able to get the lock due to following condition 
in rwsem_try_write_lock()

 630                 if (count & RWSEM_LOCK_MASK) {  ==> reader has 
sets its bias
..
...

 634
 635                         new |= RWSEM_FLAG_HANDOFF;
 636                 } else {
 637                         new |= RWSEM_WRITER_LOCKED;


---------------------->----------------------->-------------------------

First reader (1)          writer(2) RT task             Lock holder(3)

It sets
RWSEM_READER_BIAS.
while it is going to
slowpath(as the lock
was held by (3)) and
before it got added
to the waiters list
it got preempted
by (2).
                        RT task also takes
                        the slowpath and add              release the
                        itself into waiting list          rwsem lock
            and since it is the first         clear the
                        it is the next one to get         owner.
                        the lock but it can not
                        get the lock as (count &
                        RWSEM_LOCK_MASK) is set
                        as (1) has added it but
                        not able to remove its
            adjustment.

----------------------

To fix that we either has to disable preemption in down_read() and 
reenable it in rwsem_down_read_slowpath after decrementing the 
RWSEM_READER_BIAS or to limit the number of trylock-spinning attempt 
like this patch. The latter approach seems a bit less messy and I am 
going to take it back out anyway in patch 4. I will put a summary of 
that special case in the patch description.

Cheers,
Longman