From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Vrabel Subject: [PATCH 0/3] xen: evtchn and gntdev device fixes and perf improvements Date: Fri, 19 Jul 2013 15:51:57 +0100 Message-ID: <1374245520-19270-1-git-send-email-david.vrabel@citrix.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: xen-devel@lists.xen.org Cc: David Vrabel List-Id: xen-devel@lists.xenproject.org This series includes a small collection of fixes improving the evtchn and gntdev devices. Patch 1 fixes a bug in the evtchn device that may cause deadlocks when unbinding events or closing the device. You may wish to consider it for stable. Patch 2 is a (very) minor performance improvement to m2p_remove_override() when used by the the gntdev device. The TLB flush that it did not have any significant performance cost (since it's a local CPU and single address only). Patch 3 improves the scalability of the evtchn device when it is used by multiple processes (e.g., multiple qemus). As you can see from the graph[1] it is a signficant improvement but still less than ideal. I suspect that this may be due to the per-domain event lock inside Xen rather than anything kernel-side. The graphed data was collected from dom0 with 8 VCPUs on a host with 8 CPUs. David [1] http://xenbits.xen.org/people/dvrabel/evtchn-device-scalability.png