From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:35866) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1aBwXE-0005kG-E7 for qemu-devel@nongnu.org; Wed, 23 Dec 2015 22:25:49 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1aBwXD-0000nr-0D for qemu-devel@nongnu.org; Wed, 23 Dec 2015 22:25:48 -0500 Date: Thu, 24 Dec 2015 11:25:38 +0800 From: Fam Zheng Message-ID: <20151224032538.GC23372@ad.usersys.redhat.com> References: <1450850347-5291-1-git-send-email-famz@redhat.com> <1450850347-5291-5-git-send-email-famz@redhat.com> <567B4207.8030300@redhat.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <567B4207.8030300@redhat.com> Subject: Re: [Qemu-devel] [PATCH v3 4/5] qmp: Add blockdev-mirror command List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Max Reitz Cc: Kevin Wolf , qemu-block@nongnu.org, Jeff Cody , qemu-devel@nongnu.org, Markus Armbruster , Stefan Hajnoczi , pbonzini@redhat.com On Thu, 12/24 01:53, Max Reitz wrote: > On 23.12.2015 06:59, Fam Zheng wrote: > > This will start a mirror job from a named device to another named > > device, its relation with drive-mirror is similar with blockdev-backup > > to drive-backup. > > > > In blockdev-mirror, the target node should be prepared by blockdev-add, > > which will be responsible for assigning a name to the new node, so > > we don't have 'node-name' parameter. > > > > Signed-off-by: Fam Zheng > > --- > > blockdev.c | 62 ++++++++++++++++++++++++++++++++++++++++++++++++++++ > > qapi/block-core.json | 47 +++++++++++++++++++++++++++++++++++++++ > > qmp-commands.hx | 48 ++++++++++++++++++++++++++++++++++++++++ > > 3 files changed, 157 insertions(+) > > It appears you haven't addressed the comments for v2. I only had a > single one (regarding documentation), but Markus had a couple ones, so > those may be worth addressing. Will look into that. > > > > > diff --git a/blockdev.c b/blockdev.c > > index f42e171..2df0c6d 100644 > > --- a/blockdev.c > > +++ b/blockdev.c > > @@ -3345,6 +3345,10 @@ static void blockdev_mirror_common(BlockDriverState *bs, > > if (bdrv_op_is_blocked(target, BLOCK_OP_TYPE_MIRROR_TARGET, errp)) { > > return; > > } > > + if (target->blk) { > > + error_setg(errp, "Cannot mirror to an attached block device"); > > + return; > > + } > > > > if (!bs->backing && sync == MIRROR_SYNC_MODE_TOP) { > > sync = MIRROR_SYNC_MODE_FULL; > > @@ -3518,6 +3522,64 @@ out: > > aio_context_release(aio_context); > > } > > > > +void qmp_blockdev_mirror(const char *device, const char *target, > > + bool has_replaces, const char *replaces, > > + MirrorSyncMode sync, > > + bool has_speed, int64_t speed, > > + bool has_granularity, uint32_t granularity, > > + bool has_buf_size, int64_t buf_size, > > + bool has_on_source_error, > > + BlockdevOnError on_source_error, > > + bool has_on_target_error, > > + BlockdevOnError on_target_error, > > + Error **errp) > > +{ > > + BlockDriverState *bs; > > + BlockBackend *blk; > > + BlockDriverState *target_bs; > > + AioContext *aio_context; > > + Error *local_err = NULL; > > + > > + blk = blk_by_name(device); > > + if (!blk) { > > + error_setg(errp, "Device '%s' not found", device); > > + return; > > + } > > + bs = blk_bs(blk); > > + > > + if (!bs) { > > + error_setg(errp, "Device '%s' has no media", device); > > + return; > > + } > > + > > + target_bs = bdrv_lookup_bs(target, target, errp); > > + if (!target_bs) { > > + return; > > + } > > + > > + aio_context = bdrv_get_aio_context(bs); > > + aio_context_acquire(aio_context); > > + > > + bdrv_ref(target_bs); > > + bdrv_set_aio_context(target_bs, aio_context); > > + > > + blockdev_mirror_common(bs, target_bs, > > + has_replaces, replaces, sync, > > + has_speed, speed, > > + has_granularity, granularity, > > + has_buf_size, buf_size, > > + has_on_source_error, on_source_error, > > + has_on_target_error, on_target_error, > > + true, true, > > Shouldn't this be "false, false,", or, ideally, set by the user? I think true is correct here because then it will be effectively controlled by open flags of target. I.e. mirror.c always sets BDRV_REQ_MAY_UNMAP, and bdrv_co_write_zeroes has: if (!(bs->open_flags & BDRV_O_UNMAP)) { flags &= ~BDRV_REQ_MAY_UNMAP; } Fam > > > + &local_err); > > + if (local_err) { > > + error_propagate(errp, local_err); > > + bdrv_unref(target_bs); > > + } > > + > > + aio_context_release(aio_context); > > +} > > + > > Max >