From mboxrd@z Thu Jan 1 00:00:00 1970 From: Olaf Hering Subject: Re: RFC: mem_event: use wait queue when ring is full Date: Wed, 23 Nov 2011 17:49:18 +0100 Message-ID: <20111123164918.GB6000@aepfle.de> References: <6c2682114a95a003123344f552a2c64c.squirrel@webmail.lagarcavilla.org> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: Content-Disposition: inline In-Reply-To: <6c2682114a95a003123344f552a2c64c.squirrel@webmail.lagarcavilla.org> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Andres Lagar-Cavilla Cc: xen-devel@lists.xensource.com, adin@gridncetric.ca List-Id: xen-devel@lists.xenproject.org On Tue, Nov 22, Andres Lagar-Cavilla wrote: > Olaf, two questions here > > - do you have any insight for events caused by foreign mappings? Those > will be lost with a full ring, with or without wait queues The callers of mem_event_check_ring() have to retry if the ring is full. Thats what happens with p2m_mem_paging_populate(), the callers return -ENOENT and expect a retry at some later point. > - we have posted a patch (twice) previously, with changes to ring > management, most importantly sending guest vcpus to sleep when space in > the ring is < d->max_vcpus. I see these two patches as complementary. What > is your take? I'm not proposing to include my patch as is, because it has one issue: wake_up will start all waiting vcpus even if there is just a single slot free in the ringbuffer. You patch is better in this respect because only a few will be started again. I will send comments for it later. Olaf