From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from [140.186.70.92] (port=39091 helo=eggs.gnu.org) by lists.gnu.org with esmtp (Exim 4.43) id 1OrBS1-0000Hj-WF for qemu-devel@nongnu.org; Thu, 02 Sep 2010 11:11:46 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.69) (envelope-from ) id 1OrBHU-0001MQ-Ko for qemu-devel@nongnu.org; Thu, 02 Sep 2010 11:00:49 -0400 Received: from mx1.redhat.com ([209.132.183.28]:27382) by eggs.gnu.org with esmtp (Exim 4.69) (envelope-from ) id 1OrBHU-0001M9-Dy for qemu-devel@nongnu.org; Thu, 02 Sep 2010 11:00:48 -0400 From: Alex Williamson Date: Thu, 02 Sep 2010 09:00:42 -0600 Message-ID: <20100902150041.11862.65901.stgit@s20.home> MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Subject: [Qemu-devel] [PATCH v2 0/4] virtio-net: More configurability and bh handling for tx List-Id: qemu-devel.nongnu.org List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org Cc: chrisw@redhat.com, kvm@vger.kernel.org, quintela@redhat.com, jes.sorensen@redhat.com, mst@redhat.com, alex.williamson@redhat.com Incorporated feedback. txburst= and txtimer= are now "x-" prefixed developer options. I added a tx= option, because I do want there to be a supported way to switch between TX strategies. This also drops the magic value of txtimer= 1 or 0 (setting default timeout or switching modes). I also dropped the trickiness around only enabling the bottom half for the iothread since Anthony indicates we don't care about performance for the non-iothread case. New performance data against both qemu.git and qemu-kvm.git: https://spreadsheets.google.com/pub?key=0AoEm50Bac2U7dGdlREhrWUpPVEdMcTJaX0RjSEgtc3c&hl=en&single=true&gid=0&output=html This shows the base performance versus patched with new default for various packet sizes and netperf test loads. The guest is a 2-way SMP, connected directly to the host via a tap (no bridge). Thanks, Alex --- Alex Williamson (4): virtio-net: Introduce a new bottom half packet TX virtio-net: Rename tx_timer_active to tx_waiting virtio-net: Limit number of packets sent per TX flush virtio-net: Make tx_timer timeout configurable hw/s390-virtio-bus.c | 8 +++ hw/s390-virtio-bus.h | 1 hw/syborg_virtio.c | 8 +++ hw/virtio-net.c | 129 +++++++++++++++++++++++++++++++++++++++++--------- hw/virtio-net.h | 14 +++++ hw/virtio-pci.c | 8 +++ hw/virtio.h | 4 +- 7 files changed, 144 insertions(+), 28 deletions(-)