gluster partial reads refusal conflicts with qcow2

[flumm]

there is an inconsistency in how qemu creates qcow2 files, which causes an error in the gluster (and possibly other block drivers)

the problem is that the gluster backend expects the filesize to be 512 byte aligned, which is not the case anymore since 2.7.0 when using the file backend for qcow2 files with a backing file

the error is then
Could not open 'gluster://gluster01/gv0/bar2.qcow2': Could not read L1 table: Input/output error

steps to reproduce:

 * create a.qcow2
 * create b.qcow2 with a.qcow2 as base via filesystem (without gluster)
   b.qcow2 filesize is not a multiple of 512 bytes
 * move both files to a gluster share
 * access to b.qcow2 via gluster block driver fails

example:

have a gluster server at 'gluster01' with a volume 'gv0' (gluster versions tested: 3.7.15,3.8.5,3.8.5)

root@pc:~# mount -t glusterfs gluster01:/gv0 /mnt/gluster
root@pc:~# qemu-img create -f qcow2 gluster://gluster01/gv0/foo.qcow2 100M
Formatting 'gluster://gluster01/gv0/foo.qcow2', fmt=qcow2 size=104857600 encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
root@pc:~# qemu-img info /mnt/gluster/foo.qcow2
image: /mnt/gluster/foo.qcow2
file format: qcow2
virtual size: 100M (104857600 bytes)
disk size: 193K
cluster_size: 65536
Format specific information:
    compat: 1.1
    lazy refcounts: false
    refcount bits: 16
    corrupt: false
root@pc:~# qemu-img info gluster://gluster01/gv0/foo.qcow2
image: gluster://gluster01/gv0/foo.qcow2
file format: qcow2
virtual size: 100M (104857600 bytes)
disk size: 193K
cluster_size: 65536
Format specific information:
    compat: 1.1
    lazy refcounts: false
    refcount bits: 16
    corrupt: false
root@pc:~# qemu-img create -f qcow2 -b foo.qcow2 gluster://gluster01/gv0/bar.qcow2
Formatting 'gluster://gluster01/gv0/bar.qcow2', fmt=qcow2 size=104857600 backing_file=foo.qcow2 encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
root@pc:~# qemu-img info /mnt/gluster/bar.qcow2
image: /mnt/gluster/bar.qcow2
file format: qcow2
virtual size: 100M (104857600 bytes)
disk size: 193K
cluster_size: 65536
backing file: foo.qcow2 (actual path: /mnt/gluster/foo.qcow2)
Format specific information:
    compat: 1.1
    lazy refcounts: false
    refcount bits: 16
    corrupt: false
root@pc:~# qemu-img info gluster://gluster01/gv0/bar.qcow2
image: gluster://gluster01/gv0/bar.qcow2
file format: qcow2
virtual size: 100M (104857600 bytes)
disk size: 193K
cluster_size: 65536
backing file: foo.qcow2 (actual path: gluster://gluster01/gv0/foo.qcow2)
Format specific information:
    compat: 1.1
    lazy refcounts: false
    refcount bits: 16
    corrupt: false
root@pc:~# qemu-img create -f qcow2 -b foo.qcow2 /mnt/gluster/bar2.qcow2
Formatting '/mnt/gluster/bar2.qcow2', fmt=qcow2 size=104857600 backing_file=foo.qcow2 encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
root@pc:~# qemu-img info /mnt/gluster/bar2.qcow2
image: /mnt/gluster/bar2.qcow2
file format: qcow2
virtual size: 100M (104857600 bytes)
disk size: 193K
cluster_size: 65536
backing file: foo.qcow2 (actual path: /mnt/gluster/foo.qcow2)
Format specific information:
    compat: 1.1
    lazy refcounts: false
    refcount bits: 16
    corrupt: false
root@pc:~# qemu-img info gluster://gluster01/gv0/bar2.qcow2
qemu-img: Could not open 'gluster://gluster01/gv0/bar2.qcow2': Could not read L1 table: Input/output error
root@pc:~# ls -l /mnt/gluster/
total 578
-rw-r--r-- 1 root root 196616 Nov 25 09:07 bar2.qcow2
-rw------- 1 root root 197120 Nov 25 09:07 bar.qcow2
-rw------- 1 root root 197120 Nov 25 09:06 foo.qcow2
drwxr-xr-x 6 root root 46 Nov 24 16:51 images

here you can see that the file created with directory path is not 512 byte aligned, while the one created through the gluster api is

also, when creating a qcow2 with the nfs block driver, the filesize is also a multiple of 512, but reading a non aligned file with nfs works however

[Thomas Huth (th-huth)]

The QEMU project is currently considering to move its bug tracking to another system. For this we need to know which bugs are still valid and which could be closed already. Thus we are setting all older bugs to
"Incomplete" now.
If you still think this bug report here is valid, then please switch the state back to "New" within the next 60 days, otherwise this report will be marked as "Expired". Thank you and sorry for the inconvenience.

[Launchpad Janitor (janitor)]

[Expired for QEMU because there has been no activity for 60 days.]