[PATCH v3 hmm 06/12] mm/hmm: Hold on to the mmget for the lifetime of the range

[Jason Gunthorpe]

On Sat, Jun 15, 2019 at 07:14:35AM -0700, Christoph Hellwig wrote:
> >  	mutex_lock(&hmm->lock);
> > -	list_for_each_entry(range, &hmm->ranges, list)
> > -		range->valid = false;
> > -	wake_up_all(&hmm->wq);
> > +	/*
> > +	 * Since hmm_range_register() holds the mmget() lock hmm_release() is
> > +	 * prevented as long as a range exists.
> > +	 */
> > +	WARN_ON(!list_empty(&hmm->ranges));
> >  	mutex_unlock(&hmm->lock);
> 
> This can just use list_empty_careful and avoid the lock entirely.

Sure, it is just a debugging helper and the mmput should serialize
thinigs enough to be reliable. I had to move the RCU patch ahead of
this. Thanks

diff --git a/mm/hmm.c b/mm/hmm.c
index a9ace28984ea42..1eddda45cefae7 100644
--- a/mm/hmm.c
+++ b/mm/hmm.c
@@ -124,13 +124,11 @@ static void hmm_release(struct mmu_notifier *mn, struct mm_struct *mm)
 	if (!kref_get_unless_zero(&hmm->kref))
 		return;
 
-	mutex_lock(&hmm->lock);
 	/*
 	 * Since hmm_range_register() holds the mmget() lock hmm_release() is
 	 * prevented as long as a range exists.
 	 */
-	WARN_ON(!list_empty(&hmm->ranges));
-	mutex_unlock(&hmm->lock);
+	WARN_ON(!list_empty_careful(&hmm->ranges));
 
 	down_write(&hmm->mirrors_sem);
 	mirror = list_first_entry_or_null(&hmm->mirrors, struct hmm_mirror,
@@ -938,7 +936,7 @@ void hmm_range_unregister(struct hmm_range *range)
 		return;
 
 	mutex_lock(&hmm->lock);
-	list_del(&range->list);
+	list_del_init(&range->list);
 	mutex_unlock(&hmm->lock);
 
 	/* Drop reference taken by hmm_range_register() */