From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:40657) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1V5MM2-0006XZ-Dk for qemu-devel@nongnu.org; Fri, 02 Aug 2013 16:53:43 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1V5MM1-0005SP-5g for qemu-devel@nongnu.org; Fri, 02 Aug 2013 16:53:42 -0400 Received: from mail.avalus.com ([2001:41c8:10:1dd::10]:39440) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1V5Ble-00006T-8m for qemu-devel@nongnu.org; Fri, 02 Aug 2013 05:35:26 -0400 Date: Fri, 02 Aug 2013 10:35:10 +0100 From: Alex Bligh Message-ID: In-Reply-To: References: <1374819052-4292-1-git-send-email-morita.kazutaka@lab.ntt.co.jp> <1374819052-4292-7-git-send-email-morita.kazutaka@lab.ntt.co.jp> <20130730135858.GC7471@stefanha-thinkpad.redhat.com> <87mwp04nrb.wl%morita.kazutaka@lab.ntt.co.jp> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Content-Disposition: inline Subject: Re: [Qemu-devel] [sheepdog] [PATCH v4 06/10] coroutine: add co_aio_sleep_ns() to allow sleep in block drivers Reply-To: Alex Bligh List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: liu ping fan , MORITA Kazutaka Cc: Kevin Wolf , pingfank@linux.vnet.ibm.com, sheepdog@lists.wpkg.org, Alex Bligh , Stefan Hajnoczi , qemu-devel@nongnu.org, Stefan Hajnoczi , Paolo Bonzini Pingfan, --On 2 August 2013 16:19:31 +0800 liu ping fan wrote: >> Okay, I'll update this patch based on the AioContext timer. I'm also >> happy to help Alex and Pingfan to finish the implementation. >> > Alex, do you make a further step than V4? If you have, could you share > it with us?So we can rebase our patches onto yours. I've got no changes beyond v4 yet. The changes I am thinking of making are: 1. changing the AioContext * in QEMUTimerList to be a pointer to a callback instead. This is a tiny and self-contained change and may well be unnecessary. 2. Whatever comes out of Paolo's long email of yesterday, the contents of which I have not fully digested. However, I think what this boils down to is calling *_notify slightly less often. 3. Someone (Paolo/Kevin?) suggested renaming the *new* timer functions (i.e. the ones with timerlist in the name). That won't affect anything current. I suggest you go with v4 for the time being. If I end up doing a v5 and your code doesn't sit well on top of it, I can rebase to yours or something. I'd expect changes in v5 to be really small. -- Alex Bligh