From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:42560) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1bFIHM-0008UM-B0 for qemu-devel@nongnu.org; Tue, 21 Jun 2016 05:47:33 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1bFIHG-0005nT-D3 for qemu-devel@nongnu.org; Tue, 21 Jun 2016 05:47:31 -0400 Sender: Paolo Bonzini References: <1466500894-9710-1-git-send-email-kwolf@redhat.com> From: Paolo Bonzini Message-ID: <3f872c08-06d8-d755-9369-02ecd0d6d000@redhat.com> Date: Tue, 21 Jun 2016 11:47:21 +0200 MIME-Version: 1.0 In-Reply-To: <1466500894-9710-1-git-send-email-kwolf@redhat.com> Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] [PATCH 00/17] block: Convert common I/O path to BdrvChild List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Kevin Wolf , qemu-block@nongnu.org Cc: famz@redhat.com, jcody@redhat.com, qemu-devel@nongnu.org, mreitz@redhat.com, stefanha@redhat.com On 21/06/2016 11:21, Kevin Wolf wrote: > This series converts all I/O function in the core block layer up to > bdrv_co_preadv/pwritev() to taking a BdrvChild as their first parameter > instead of a BlockDriverState. > > The original motivation for this change were op blockers, where one of > the biggest problems is making sure that every user of block devices > actually registers correctly with the op blockers system. If the I/O > functions know which parent a request comes from (BdrvChild basically > corresponds to an edge in our block device graph), it can use assertions > to make sure that that parent has actually registered its activities and > thereby ensured that it doesn't conflict with other users. > > There are, however, more benefits we get from this change. The most > important one is probably that it enforces important aspects of the > block layer design like that external users go through a BlockBackend > and request are internally routed along the edges of the graph. Accesses > to random BDSes are no longer possible, you need to own an actual child > reference so you can make a request. I still fail to understand what is the rationale for this change. The API is weird; you read from a disk, not from an edge, and in fact the first thing all the APIs do is dereference the BdrvChild... The assertions are nice, but I'm not sure it's a good idea to design a whole API around them. Paolo > The work on this series already led to a few cleanups and BlockBackend > conversions in master, and this series contains a few more. > > As a bonus, all the block drivers using bs->file->bs everywhere can now > go back to bs->file, which is a little nicer to read.