Re: [PATCH 3/3] kernel/workqueue: Suppress a false positive lockdep complaint

[Johannes Berg <johannes@sipsolutions.net>]

On Thu, 2018-10-25 at 16:21 -0400, Theodore Y. Ts'o wrote:

> We can guarantee it from someone who is looking at the code path.  In
> dio_set_defer_completion:

[snip]

Right, it's indeed pretty obvious. I shouldn't have tried to reply
before the kids went to bed, that made me cut some corners ;-)

> The race found in the syzbot reproducer has multiple threads all
> running DIO writes at the same time.  So we have multiple threads
> calling sb_init_dio_done_wq, but all but one will lose the race, and
> then call destry_workqueue on the freshly created (but never used)
> workqueue.

Right.

> We could replace the destroy_workqueue(wq) with a
> "I_solemnly_swear_this_workqueue_has_never_been_used_please_destroy(wq)".

:-)

> Or, as Tejun suggested, "destroy_workqueue_skip_drain(wq)", but there is
> no way for the workqueue code to know whether the caller was using the
> interface correctly.  So this basically becomes a philosophical
> question about whether or not we trust the caller to be correct or
> not.

Right. Same with the lockdep annotation I suggested over in my other
email, of course. I think that the set of APIs I wrote there
({drain,flush,destroy}_workqueue_nested()) might be more generally
useful in other cases, not just this one, and I suspect that this code
would basically be the only user of destroy_workqueue_skip_drain().

> I don't see an obvious way that we can test to make sure the workqueue
> is never used without actually taking a performance.  Am I correct
> that we would need to take the wq->mutex before we can mess with the
> wq->flags field?

I don't really know, sorry.

johannes