From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([140.186.70.92]:34412) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1QMQBv-0002Ea-3N for qemu-devel@nongnu.org; Tue, 17 May 2011 15:44:29 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1QMQBt-000200-Sy for qemu-devel@nongnu.org; Tue, 17 May 2011 15:44:27 -0400 Received: from e32.co.us.ibm.com ([32.97.110.150]:53025) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1QMQBt-0001zk-Ji for qemu-devel@nongnu.org; Tue, 17 May 2011 15:44:25 -0400 Received: from d03relay05.boulder.ibm.com (d03relay05.boulder.ibm.com [9.17.195.107]) by e32.co.us.ibm.com (8.14.4/8.13.1) with ESMTP id p4HJWwR9016945 for ; Tue, 17 May 2011 13:32:58 -0600 Received: from d03av02.boulder.ibm.com (d03av02.boulder.ibm.com [9.17.195.168]) by d03relay05.boulder.ibm.com (8.13.8/8.13.8/NCO v10.0) with ESMTP id p4HJiEOU184782 for ; Tue, 17 May 2011 13:44:18 -0600 Received: from d03av02.boulder.ibm.com (loopback [127.0.0.1]) by d03av02.boulder.ibm.com (8.14.4/8.13.1/NCO v10.0 AVout) with ESMTP id p4HDhd0I016916 for ; Tue, 17 May 2011 07:43:39 -0600 From: "Venkateswararao Jujjuri (JV)" Date: Tue, 17 May 2011 12:43:26 -0700 Message-Id: <1305661431-21289-1-git-send-email-jvrao@linux.vnet.ibm.com> Subject: [Qemu-devel] [V2 0/25] Async threading for VirtFS using glib threads & coroutines. List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org Cc: aliguori@us.ibm.com, "Venkateswararao Jujjuri \"" , stefanha@linux.vnet.ibm.com VirtFS (fileserver base on 9P) performs many blocking system calls in the vCPU context. This effort is to move the blocking calls out of vCPU/IO thread context, into asynchronous threads. Anthony's " Add hard build dependency on glib" patch and Kevin/Stefan's coroutine effort is a prerequisite. Changes from V1 --------------- o Redesigned to use bh as per Stefan's suggestion. This made the code very simple but is little less performant compared to V1. Anthony suggested to go-in with cleaner code and design (This version) and deal with the performance later. Just to put in perspective: Sequential Writes of creating 1GB files using ffsb o Write size 8k With bh: 66.9MB/sec Without bh (marshalling routines): 74.9 MB/sec o Write size 128k With bh: 117MB/sec Without bh (marshalling routines): 122MB/sec o Addressed all review comments. o Using readdir_r is currently being worked on. Will add as a patch on top. This patch set contains: - Converting all 9pfs calls into coroutines. - Each 9P operation will be modified for: - Remove post* functions. These are our call back functions which makes the code very hard to read. Now with coroutines, we can achieve the same state machine model with nice sequential code flow. - Move errno access near to the local_syscall() - Introduce asynchronous threading This series has the basic infrastructure and few routines like mkdir,monod,unlink,readdir,xattr,lstat, etc converted. Currently we are working on converting and testing other 9P operations also into this model and those patches will follow shortly. Removing callback functions made some of the patches little lengthy. Here is the git tree for the reviewer convenience. http://repo.or.cz/w/qemu/aliguori/jvrao.git/shortlog/refs/heads/9p-coroutine-bh-round1 Signed-off-by: Venkateswararao Jujjuri "