[PATCH 2/2] virStorageSourceIsSameLocation: Special-case storage sources of type 'volume'

[Martin Kletzander]

On Mon, Nov 29, 2021 at 02:31:55PM +0100, Peter Krempa wrote:
>On Fri, Nov 26, 2021 at 14:50:37 +0100, Martin Kletzander wrote:
>> On Fri, Nov 26, 2021 at 02:07:54PM +0100, Peter Krempa wrote:
>> > The function is used also to compare virStorageSource which may not be
>> > resolved to the image at that point in which case the 'path' is not yet
>> > populated and the actual type is not yet set. This means that the
>> > function fails to consider two identical volume-based disks as pointing
>> > to the same thing.
>> >
>> > Add a special case for both images being type=volume in which case we
>> > compare only the pool/volume names.
>> >
>> > Closes: https://gitlab.com/libvirt/libvirt/-/issues/240
>> > Signed-off-by: Peter Krempa <pkrempa at redhat.com>
>> > ---
>> > src/conf/storage_source_conf.c | 7 +++++++
>> > 1 file changed, 7 insertions(+)
>> >
>> > diff --git a/src/conf/storage_source_conf.c b/src/conf/storage_source_conf.c
>> > index 44944e1dbd..c0acee189a 100644
>> > --- a/src/conf/storage_source_conf.c
>> > +++ b/src/conf/storage_source_conf.c
>> > @@ -920,6 +920,13 @@ virStorageSourceIsSameLocation(virStorageSource *a,
>> >         virStorageSourceIsEmpty(b))
>> >         return true;
>> >
>> > +    /* for disk type=volume we must check just pool/volume names as they might
>> > +     * not yet be resolved if e.g. we are comparing against the persistent def */
>> > +    if (a->type == VIR_STORAGE_TYPE_VOLUME && b->type == VIR_STORAGE_TYPE_VOLUME) {
>> > +        return STREQ(a->srcpool->pool, b->srcpool->pool) &&
>> > +               STREQ(a->srcpool->volume, b->srcpool->volume);
>> > +    }
>> > +
>>
>> Does this hold true even for disk sources that differ in the `mode`
>> attribute?  I'm just asking for clarity.  If that was taken into
>> consideration and the difference does not matter here, then this is
>
>The idea is that the function should return true if both are pointing to
>the same backing, in those terms the 'mode' is irrelevant as it's still
>the same image.
>
>This code is used in blockjob finalization, where it doesn't matter at
>all and in cdrom image changing code, where it very theoretically could
>be used to change the mode, but it's questionable whether that would
>even make sense in any real usage.
>

OK, I had to ask because I read the documentation around 10 times and
wasn't sure whether it can change where the source points to, but it
looks like it does no, just changes how to represent the path to the
device.  Sorry for the noise.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://listman.redhat.com/archives/libvir-list/attachments/20211129/156d4565/attachment-0001.sig>