On 08/21/2014 02:14 AM, Kevin Wolf wrote:
>>>> + if (offset_into_cluster(s, l2_entry & L2E_OFFSET_MASK)) {
>>>> + fprintf(stderr, "qcow2: Cannot free unaligned cluster
>>>> %#llx\n",
>>>> + l2_entry & L2E_OFFSET_MASK);
>>>> + } else {
>>>> + qcow2_free_clusters(bs, l2_entry & L2E_OFFSET_MASK,
>>>> + nb_clusters << s->cluster_bits, type);
>>>> + }
>>> Hm... Why isn't this a corruption like any other? Unconditional
>>> fprintf() is something I don't like a lot.
>>
>> We already do it in qcow2_free_clusters().
>>
>> I decided not to make it a corruption because we don't lose
>> anything. The entry is corrupted, but we don't need it anymore
>> anyway; it's overwritten with 0 and wherever the cluster might have
>> been meant to be located, it doesn't matter anymore.
>
> I can see your point. This is a tough one: On the one hand, it is
> undoubtedly corruption, and usually there is not just one corrupted
> entry, so you want the user to check the image. On the other hand, yes,
> this is merely a cluster leak and breaking the VM for that might be a
> bit too much.
>
> Still just printing a line on stderr and continuing otherwise doesn't
> feel quite right, the user will usually miss the message because it ends
> up in the log of a seemingly well working guest and if printed
> unconditionally, it may still flood the logs.
>
> Eric, would management be able to make something useful out of this if
> we sent a QMP event?Libvirt could certainly expose a QMP event to upper layers (oVirt, VDSM, OpenStack, ...) for them to be made aware that "your image had a non-fatal corruption, you may want to check if there are other problems going on". I don't think libvirt would do anything in particular with this knowledge beyond passthrough, but it does sound like a reasonable event to add. -- Eric Blake eblake redhat com +1-919-301-3266 Libvirt virtualization library http://libvirt.org
signature.asc
Description: OpenPGP digital signature
