Re: [tip:tools/kvm] kvm tools, qcow: Fix copy-on-write image corruption

Re: [tip:tools/kvm] kvm tools, qcow: Fix copy-on-write image corruption

[Kevin Wolf]

Am 20.07.2011 21:49, schrieb tip-bot for Pekka Enberg:
> Commit-ID:  4839bcd6b04a1b628f0f00d535ddca2ee872e9fa
> Gitweb:     http://git.kernel.org/tip/4839bcd6b04a1b628f0f00d535ddca2ee872e9fa
> Author:     Pekka Enberg <penberg@kernel.org>
> AuthorDate: Wed, 20 Jul 2011 16:28:31 +0300
> Committer:  Pekka Enberg <penberg@kernel.org>
> CommitDate: Wed, 20 Jul 2011 16:28:31 +0300
> 
> kvm tools, qcow: Fix copy-on-write image corruption
> 
> We don't handle refcount table properly so make sure we only write to clusters
> that have the "copied" flag set.
> 
> Cc: Kevin Wolf <kwolf@redhat.com>
> Cc: Sasha Levin <levinsasha928@gmail.com>
> Signed-off-by: Pekka Enberg <penberg@kernel.org>

Hm, I must have missed this patch on the KVM mailing list...

You should also error out before creating a new L2 table as its refcount
would have to be updated, too. On the other hand, the patch is a bit too
restrictive, because it applies the check to qcow1 which doesn't even
have a refcount table.

I'm not sure if everyone is aware of it, so another hint: Now that you
can't grow qcow2 any more, you might find it useful to use qemu-img
create -f qcow2 -o preallocation=metadata, so that all clusters are
already present and you don't get random failures.

Kevin

Re: [tip:tools/kvm] kvm tools, qcow: Fix copy-on-write image corruption

[Pekka Enberg]

On Thu, 21 Jul 2011, Kevin Wolf wrote:
> Hm, I must have missed this patch on the KVM mailing list...

It wasn't on the list, sorry.

> You should also error out before creating a new L2 table as its refcount
> would have to be updated, too. On the other hand, the patch is a bit too
> restrictive, because it applies the check to qcow1 which doesn't even
> have a refcount table.

Oh, right, I broken QCOW1.

> I'm not sure if everyone is aware of it, so another hint: Now that you
> can't grow qcow2 any more, you might find it useful to use qemu-img
> create -f qcow2 -o preallocation=metadata, so that all clusters are
> already present and you don't get random failures.

Yup, I'm aware that you can't grow images anymore. The patch was meant as 
a stop-gap measure while I'm hacking on refcount table support.

 			Pekka

Re: [tip:tools/kvm] kvm tools, qcow: Fix copy-on-write image corruption

[Kevin Wolf]

Am 22.07.2011 10:21, schrieb Pekka Enberg:
> On Thu, 21 Jul 2011, Kevin Wolf wrote:
>> Hm, I must have missed this patch on the KVM mailing list...
> 
> It wasn't on the list, sorry.
> 
>> You should also error out before creating a new L2 table as its refcount
>> would have to be updated, too. On the other hand, the patch is a bit too
>> restrictive, because it applies the check to qcow1 which doesn't even
>> have a refcount table.
> 
> Oh, right, I broken QCOW1.
> 
>> I'm not sure if everyone is aware of it, so another hint: Now that you
>> can't grow qcow2 any more, you might find it useful to use qemu-img
>> create -f qcow2 -o preallocation=metadata, so that all clusters are
>> already present and you don't get random failures.
> 
> Yup, I'm aware that you can't grow images anymore. The patch was meant as 
> a stop-gap measure while I'm hacking on refcount table support.

Yeah, I was pretty sure you were aware of _that_, but maybe not everyone
knows that metadata preallocation exists.

Kevin