Re: [PATCH 1/2] vmpressure: in-kernel notifications

[Glauber Costa <glommer@parallels.com>]

[-- Attachment #1: Type: text/plain, Size: 399 bytes --]

On 04/24/2013 11:42 PM, Greg Thelen wrote:
>> +	vmpr->notify_userspace = true;
> Should notify_userspace get cleared sometime?  Seems like we might need
> to clear or decrement notify_userspace in vmpressure_event() when
> calling eventfd_signal().
> 
I am folding the attached patch and keeping the acks unless the ackers
oppose.

Greg, any other problem you spot here? Thanks for the review BTW.


[-- Attachment #2: vmpressure.diff --]
[-- Type: text/x-patch, Size: 789 bytes --]

diff --git a/mm/vmpressure.c b/mm/vmpressure.c
index 1a082a0..e16256e 100644
--- a/mm/vmpressure.c
+++ b/mm/vmpressure.c
@@ -164,6 +164,7 @@ static bool vmpressure_event(struct vmpressure *vmpr,
 		}
 	}
 
+	vmpr->notify_userspace = false;
 	mutex_unlock(&vmpr->events_lock);
 
 	return signalled;
@@ -249,8 +250,13 @@ void vmpressure(gfp_t gfp, struct mem_cgroup *memcg,
 	mutex_lock(&vmpr->sr_lock);
 	vmpr->scanned += scanned;
 	vmpr->reclaimed += reclaimed;
-	vmpr->notify_userspace = true;
 	scanned = vmpr->scanned;
+	/*
+	 * If we didn't reach this point, only kernel events will be triggered.
+	 * It is the job of the worker thread to clean this up once the
+	 * notifications are all delivered.
+	 */
+	vmpr->notify_userspace = true;
 	mutex_unlock(&vmpr->sr_lock);
 
 schedule: