Description:
Cannot shallow blockcopy a vm disk backing chain to a block device (whose format=raw)

Versions:
qemu-kvm-rhev-2.9.0-16.el7_4.6.x86_64
libvirt-3.7.0-2.el7.x86_64

REGRESSION: not reproduded with libvirt-3.2.0-14.el7_4.3.x86_64

How reproducible:
100%

Steps:
1. create a transient vm
## virsh create /tmp/avocado-vt-vm1
Domain avocado-vt-vm1 created from /tmp/avocado-vt-vm1

2. create a disk only snapshot for it
## virsh snapshot-create avocado-vt-vm1 --disk-only
Domain snapshot 1505209180 created


3. prepare a block device with format=raw
## dd if=/dev/urandom of=/dev/sde bs=1M count=1000
qemu1000+0 records in
1000+0 records out
1048576000 bytes (1.0 GB) copied, 4.52631 s, 232 MB/s

## qemu-img info /dev/sde
image: /dev/sde
file format: raw
virtual size: 20G (21474836480 bytes)
disk size: 0

4. do a shallow blockcopy to /dev/sde
## virsh blockcopy avocado-vt-vm1 vda /dev/sde  --blockdev --shallow
error: internal error: unable to execute QEMU command 'drive-mirror': Backing file not supported for file format 'raw'

Actual result:
blockcopy failed.

Expected result:
blockcopy can be executed as usual

Additional info:
The reason is libvirt commit 703abf1d7 has:
----------------------------------------
@@ -16692,36 +16692,49 @@ qemuDomainBlockCopyValidateMirror(virStorageSourcePtr mirror,
     int desttype = virStorageSourceGetActualType(mirror);
     struct stat st;
...
+        if (S_ISBLK(st.st_mode)) {
+            /* if the target is a block device, assume that we are reusing it,
+             * so there are no attempts to create it */
+            *reuse = true;
----------------------------------------

so for block device mirror, the **reuse** will always be set to **true** and later used in:
----------------------------------------
if (!mirror->format) {                                                      
    if (!reuse) {                                                           
        mirror->format = disk->src->format;                                 
    } else {                                                                
        /* If the user passed the REUSE_EXT flag, then either they          
         * can also pass the RAW flag or use XML to tell us the format.     
         * So if we get here, we assume it is safe for us to probe the      
         * format from the file that we will be using.  */                  
        mirror->format = virStorageFileProbeFormat(mirror->path, cfg->user,
                                                   cfg->group);             
    }                                                                       
}   
----------------------------------------


and when we do a shallow blockcopy, this will send a qmp cmd as:
{"execute":"drive-mirror","arguments":{"device":"drive-virtio-disk0","target":"/dev/sde","sync":"top","mode":"absolute-paths","format":"raw"},"id":"libvirt-12"} which trigger the error message.
But in previous libvirt versions the qmp command with have "format": "qcow2" instead, so this will be successful.


Additional info:
And format=raw doesn't affect full chain blockcopy, following virsh/qmp cmd won't trigger error:
----------------------------------------
## virsh blockcopy avocado-vt-vm1 vda /dev/sde  --blockdev
Block Copy started

{"device":"drive-virtio-disk0","target":"/dev/sde","sync":"full","mode":"absolute-paths","format":"raw"},"id":"libvirt-15"}'
----------------------------------------

So seems only shallow copy to a format=raw block device is not supported. And above libvirt commit just expose this problem. Pls check if we need to do some workarounds such as prepare the block device to qcow2 format as usual when we do shallow blockcopy.