From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:36597) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1YL8ke-0002O8-5c for qemu-devel@nongnu.org; Tue, 10 Feb 2015 06:13:09 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1YL8ka-0004gX-5h for qemu-devel@nongnu.org; Tue, 10 Feb 2015 06:13:08 -0500 Received: from mail-ie0-f178.google.com ([209.85.223.178]:41517) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1YL8ka-0004gQ-2D for qemu-devel@nongnu.org; Tue, 10 Feb 2015 06:13:04 -0500 Received: by iecrd18 with SMTP id rd18so11700165iec.8 for ; Tue, 10 Feb 2015 03:13:03 -0800 (PST) Date: Tue, 10 Feb 2015 19:12:56 +0800 From: Liu Yuan Message-ID: <20150210111256.GG18727@ubuntu-trusty> References: <1422347727-13006-1-git-send-email-ishizaki.teruaki@lab.ntt.co.jp> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1422347727-13006-1-git-send-email-ishizaki.teruaki@lab.ntt.co.jp> Subject: Re: [Qemu-devel] [PATCH v4] sheepdog: selectable object size support List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Teruaki Ishizaki Cc: kwolf@redhat.com, mitake.hitoshi@lab.ntt.co.jp, sheepdog@lists.wpkg.org, qemu-devel@nongnu.org, stefanha@redhat.com On Tue, Jan 27, 2015 at 05:35:27PM +0900, Teruaki Ishizaki wrote: > Previously, qemu block driver of sheepdog used hard-coded VDI object size. > This patch enables users to handle "block_size_shift" value for > calculating VDI object size. > > When you start qemu, you don't need to specify additional command option. > > But when you create the VDI which doesn't have default object size > with qemu-img command, you specify block_size_shift option. > > If you want to create a VDI of 8MB(1 << 23) object size, > you need to specify following command option. > > # qemu-img create -o block_size_shift=23 sheepdog:test1 100M > > In addition, when you don't specify qemu-img command option, > a default value of sheepdog cluster is used for creating VDI. > > # qemu-img create sheepdog:test2 100M > > Signed-off-by: Teruaki Ishizaki > --- > V4: > - Limit a read/write buffer size for creating a preallocated VDI. > - Replace a parse function for the block_size_shift option. > - Fix an error message. > > V3: > - Delete the needless operation of buffer. > - Delete the needless operations of request header. > for SD_OP_GET_CLUSTER_DEFAULT. > - Fix coding style problems. > > V2: > - Fix coding style problem (white space). > - Add members, store_policy and block_size_shift to struct SheepdogVdiReq. > - Initialize request header to use block_size_shift specified by user. > --- > block/sheepdog.c | 138 ++++++++++++++++++++++++++++++++++++++------- > include/block/block_int.h | 1 + > 2 files changed, 119 insertions(+), 20 deletions(-) > > diff --git a/block/sheepdog.c b/block/sheepdog.c > index be3176f..a43b947 100644 > --- a/block/sheepdog.c > +++ b/block/sheepdog.c > @@ -37,6 +37,7 @@ > #define SD_OP_READ_VDIS 0x15 > #define SD_OP_FLUSH_VDI 0x16 > #define SD_OP_DEL_VDI 0x17 > +#define SD_OP_GET_CLUSTER_DEFAULT 0x18 This might not be necessary. For old qemu or the qemu-img without setting option, the block_size_shift will be 0. If we make 0 to represent 4MB object, then we don't need to get the default cluster object size. We migth even get rid of the idea of cluster default size. The downsize is that, if we want to create a vdi with different size not the default 4MB, we have to write it every time for qemu-img or dog. If we choose to keep the idea of cluster default size, I think we'd also try to avoid call this request from QEMU to make backward compatibility easier. In this scenario, 0 might be used to ask new sheep to decide to use cluster default size. Both old qemu and new QEMU will send 0 to sheep and both old and new sheep can handle 0 though it has different meanings. Table for this bit as 0: Qe: qemu SD: Sheep daemon CDS: Cluster Default Size Ign: Ignored by the sheep daemon Qe/sd new old new CDS Ign old CDS NULL I think this approach is acceptable. The difference to your patch is that we don't send SD_OP_GET_CLUSTER_DEFAULT to sheep daemon and SD_OP_GET_CLUSTER_DEFAULT can be removed. Thanks Yuan