From mboxrd@z Thu Jan 1 00:00:00 1970 From: Wei Wang Subject: [PATCH 1/6] Vhost-pci RFC: Introduction Date: Sun, 29 May 2016 07:36:30 +0800 Message-ID: <1464478595-146533-2-git-send-email-wei.w.wang@intel.com> References: <1464478595-146533-1-git-send-email-wei.w.wang@intel.com> Cc: Wei Wang To: kvm@vger.kernel.org, qemu-devel@nongnu.org, virtio-comment@lists.oasis-open.org, virtio-dev@lists.oasis-open.org, mst@redhat.com, stefanha@redhat.com, pbonzini@redhat.com Return-path: Received: from mga09.intel.com ([134.134.136.24]:57044 "EHLO mga09.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752507AbcE1Pkb (ORCPT ); Sat, 28 May 2016 11:40:31 -0400 In-Reply-To: <1464478595-146533-1-git-send-email-wei.w.wang@intel.com> Sender: kvm-owner@vger.kernel.org List-ID: Signed-off-by: Wei Wang --- Introduction | 31 +++++++++++++++++++++++++++++++ 1 file changed, 31 insertions(+) create mode 100644 Introduction diff --git a/Introduction b/Introduction new file mode 100644 index 0000000..8774676 --- /dev/null +++ b/Introduction @@ -0,0 +1,31 @@ +Current vhost-user based backend designs for virtio-net devices present scaling +challenges, as communication intensive applications (e.g. virtual network +functions) running in VMs start to stress this centralized design and resources +assigned to it. + +Vhost-pci was initially proposed by Michael S. Tsirkin with vIOMMU support to +offer a protected and point-to-point based inter-VM communication. In many +cases, such as network function virtualization (NFV) and software defined +networking (SDN) usages, VMs in an isolated network trust each other and they +may be chained together to complete a task. In these use cases, people care +more about performance than security. In this RFC we present a comprehensive +design of vhost-pci without vIOMMU support. A VM with such a vhost-pci device +is able to copy data to another VM's memory directly. The advantages of using +vhost-pci over vhost-user are: 1) one less packet copy per packet transfer; and +2) better scalability. + +To follow the naming conventions in the virtio specification, we call the VM +who sends packets to the destination VM the device VM, and the VM who provides +the vring and receives packets the driver VM. The vhost-pci device/driver works +independently in the device VM. It can be considered as a simple device mapping +the entire memory of a driver VM. But a lot of times, it may be used to +communicate to a virtio device in the driver VM. That is, it usually plays the +role of a backend part of a virtio device of a driver VM. + +The vhost-pci design is not limited to networking usages. The design presented +in this RFC is quite fundamental, and it is potentially useful for other +inter-VM data moving usages. For the convenience of descriptions, we will +simply use "virtio device" to refer to the device on a driver VM that is backed +by a vhost-pci device. The figures of this RFC are shown in this link: +https://etherpad.opnfv.org/p/vhost-pci_RFC + -- 1.8.3.1 From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:55813) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1b6gLz-0000jM-C4 for qemu-devel@nongnu.org; Sat, 28 May 2016 11:40:44 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1b6gLt-0004DI-2i for qemu-devel@nongnu.org; Sat, 28 May 2016 11:40:42 -0400 Received: from mga09.intel.com ([134.134.136.24]:54329) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1b6gLs-0004Bx-Rm for qemu-devel@nongnu.org; Sat, 28 May 2016 11:40:36 -0400 From: Wei Wang Date: Sun, 29 May 2016 07:36:30 +0800 Message-Id: <1464478595-146533-2-git-send-email-wei.w.wang@intel.com> In-Reply-To: <1464478595-146533-1-git-send-email-wei.w.wang@intel.com> References: <1464478595-146533-1-git-send-email-wei.w.wang@intel.com> Subject: [Qemu-devel] [PATCH 1/6] Vhost-pci RFC: Introduction List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: kvm@vger.kernel.org, qemu-devel@nongnu.org, virtio-comment@lists.oasis-open.org, virtio-dev@lists.oasis-open.org, mst@redhat.com, stefanha@redhat.com, pbonzini@redhat.com Cc: Wei Wang Signed-off-by: Wei Wang --- Introduction | 31 +++++++++++++++++++++++++++++++ 1 file changed, 31 insertions(+) create mode 100644 Introduction diff --git a/Introduction b/Introduction new file mode 100644 index 0000000..8774676 --- /dev/null +++ b/Introduction @@ -0,0 +1,31 @@ +Current vhost-user based backend designs for virtio-net devices present scaling +challenges, as communication intensive applications (e.g. virtual network +functions) running in VMs start to stress this centralized design and resources +assigned to it. + +Vhost-pci was initially proposed by Michael S. Tsirkin with vIOMMU support to +offer a protected and point-to-point based inter-VM communication. In many +cases, such as network function virtualization (NFV) and software defined +networking (SDN) usages, VMs in an isolated network trust each other and they +may be chained together to complete a task. In these use cases, people care +more about performance than security. In this RFC we present a comprehensive +design of vhost-pci without vIOMMU support. A VM with such a vhost-pci device +is able to copy data to another VM's memory directly. The advantages of using +vhost-pci over vhost-user are: 1) one less packet copy per packet transfer; and +2) better scalability. + +To follow the naming conventions in the virtio specification, we call the VM +who sends packets to the destination VM the device VM, and the VM who provides +the vring and receives packets the driver VM. The vhost-pci device/driver works +independently in the device VM. It can be considered as a simple device mapping +the entire memory of a driver VM. But a lot of times, it may be used to +communicate to a virtio device in the driver VM. That is, it usually plays the +role of a backend part of a virtio device of a driver VM. + +The vhost-pci design is not limited to networking usages. The design presented +in this RFC is quite fundamental, and it is potentially useful for other +inter-VM data moving usages. For the convenience of descriptions, we will +simply use "virtio device" to refer to the device on a driver VM that is backed +by a vhost-pci device. The figures of this RFC are shown in this link: +https://etherpad.opnfv.org/p/vhost-pci_RFC + -- 1.8.3.1