From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([208.118.235.92]:42744) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1SWlBZ-0005sh-9i for qemu-devel@nongnu.org; Tue, 22 May 2012 05:15:26 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1SWlBS-00051w-O8 for qemu-devel@nongnu.org; Tue, 22 May 2012 05:15:20 -0400 Received: from mx1.redhat.com ([209.132.183.28]:46835) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1SWlBS-00051T-GD for qemu-devel@nongnu.org; Tue, 22 May 2012 05:15:14 -0400 Message-ID: <4FBB591A.8020506@redhat.com> Date: Tue, 22 May 2012 11:15:06 +0200 From: Paolo Bonzini MIME-Version: 1.0 References: <4FBA1CC4.7020009@redhat.com> In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] Weird iscsi/fd-event issue since recent merge of event system changes List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: ronnie sahlberg Cc: Kevin Wolf , Stefan Hajnoczi , qemu-devel , Stefan Hajnoczi Il 22/05/2012 11:07, ronnie sahlberg ha scritto: > Below iscsi_aio_readv is printed when we enter this function. > This function prepares a CDB and sets it up for queueing, but it does > not actually write it to the socket. > At the end of this function, we set up the events by calling > iscsi_set_events() and we set it up for writeable. > > iscsi_process_write() is when we enter the fd-is-writeable callback. > > iscsi_aio_readv 1337676989.429822 > iscsi_set_events write_event?Y 1337676989.429834 > iscsi_process_write 1337676989.484424 > iscsi_set_events write_event?N 1337676989.484507 > iscsi_process_read 1337676989.484628 > iscsi_aio_read10_cb 1337676989.484647 > iscsi_set_events write_event?N 1337676989.484660 > iscsi_readv_writev_bh_cb 1337676989.484665 > iscsi_aio_readv 1337676989.484751 Can you add to the debugging a "strace -e select" (and also note down what fd is used for the socket), and/or instrument qemu_iohandler_{fill,poll} in qemu's iohandler.c? Paolo