[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [RFC PATCH v2 20/23] qcow2: Cancel COW when overwritten
From: |
Kevin Wolf |
Subject: |
Re: [Qemu-devel] [RFC PATCH v2 20/23] qcow2: Cancel COW when overwritten |
Date: |
Thu, 21 Feb 2013 10:47:05 +0100 |
User-agent: |
Mutt/1.5.21 (2010-09-15) |
On Mon, Feb 18, 2013 at 04:46:49PM +0100, Stefan Hajnoczi wrote:
> On Wed, Feb 13, 2013 at 02:22:10PM +0100, Kevin Wolf wrote:
> > @@ -707,6 +710,16 @@ int qcow2_alloc_cluster_link_l2(BlockDriverState *bs,
> > QCowL2Meta *m)
> > }
> >
> > /* Update L2 table. */
> > + qemu_co_mutex_unlock(&s->lock);
> > + qemu_co_rwlock_wrlock(&m->l2_writeback_lock);
> > + has_wr_lock = true;
> > + qemu_co_mutex_lock(&s->lock);
> > +
> > + if (m->no_l2_update) {
>
> QcowL2Meta now has a no_l2_update field. A sign that we're abusing the
> QcowL2Meta struct and really need a different abstraction?
I think the abstraction is the right one, even though maybe it could use
a rename. Maybe QcowL2Meta -> QcowCOWRequest or something. The comment
in qcow2.h correctly describes what it's meant for.
/**
* Describes an in-flight (part of a) write request that writes to clusters
* that are not referenced in their L2 table yet.
*/
typedef struct QCowL2Meta
However, I think m->no_l2_update is actually redundant: The goal is that
only one request that touches a cluster should be responsible for
updating the L2 table, and it makes sense to choose the first one to do
that. We already have this information in m->parent: The first one is
the root of the tree described by these parent links. So if I'm not
mistaken, m->no_l2_update == (m->parent != NULL).
Kevin
- Re: [Qemu-devel] [RFC PATCH v2 19/23] qcow2: Add error handling to the l2meta coroutine, (continued)
[Qemu-devel] [RFC PATCH v2 14/23] qcow2: Use byte granularity in qcow2_alloc_cluster_offset(), Kevin Wolf, 2013/02/13
[Qemu-devel] [RFC PATCH v2 17/23] qcow2: Move COW and L2 update into own coroutine, Kevin Wolf, 2013/02/13
[Qemu-devel] [RFC PATCH v2 22/23] qcow2: Move cluster gathering to a non-looping loop, Kevin Wolf, 2013/02/13
[Qemu-devel] [RFC PATCH v2 21/23] qemu-iotests: Another concurrent multicluster allocation case, Kevin Wolf, 2013/02/13
[Qemu-devel] [RFC PATCH v2 20/23] qcow2: Cancel COW when overwritten, Kevin Wolf, 2013/02/13
[Qemu-devel] [RFC PATCH v2 23/23] qcow2: Gather clusters in a looping loop, Kevin Wolf, 2013/02/13
Re: [Qemu-devel] [RFC PATCH v2 00/23] qcow2: Delayed COW, Stefan Hajnoczi, 2013/02/14
Re: [Qemu-devel] [RFC PATCH v2 00/23] qcow2: Delayed COW, Stefan Hajnoczi, 2013/02/18