Re: [Qemu-devel] [PATCH 0/4] virtio: handle zoned backing devices

[Kevin Wolf <kwolf@redhat.com>]

Am 10.07.2019 um 12:09 hat Paolo Bonzini geschrieben:
> On 09/07/19 22:38, Dmitry Fomichev wrote:
> > Currently, attaching zoned block devices (i.e. storage devices
> > compliant to ZAC/ZBC standards) using several virtio methods doesn't
> > work - the zoned devices appear as regular block devices at the guest.
> > This may cause unexpected i/o errors and, potentially, some data
> > corruption.
> > 
> > To be more precise, attaching a zoned device via virtio-pci-blk,
> > virtio-scsi-pci/scsi-disk or virtio-scsi-pci/scsi-hd demonstrates the
> > above behavior. A simple fix is needed to make
> > virtio-scsi-pci/scsi-block work and this is covered by a different
> > patch. The virtio-scsi-pci/scsi-generic method appears to handle zoned
> > devices without problems.
> 
> The problem with this approach is that other devices (e.g. ide-hd or sd
> card) also break with zoned devices and the only way to fix it would be
> to add code denying zoned block devices to all of them.
> 
> The question then becomes how to define a whitelist.  One possiblity is
> to add a QOM interface (for example TYPE_ZONED_BLOCK_SUPPORT) to
> scsi-block and scsi-generic.  In do_parse_drive you can query the
> BlockBackend with bdrv_get_zoned_info, and return an error if the
> backend is a zoned block device and the device does not implement
> TYPE_ZONED_BLOCK_SUPPORT.  (Commit 6b1566c is an example of adding a new
> QOM interface; in your case, it would be simpler as the interface would
> not have any method).  Kevin, what do you think?

What about non-device users such as block jobs or (NBD) exports? Won't
they have to special-case such devices, too? In fact, what about image
format drivers or even filters?

I feel that this needs to be managed at the BDS level somehow. Not sure
which mechanism to use, though. Permissions would be suitable for a
blacklist approach, but I agree with you that we need a whitelist
instead.

Hm... Actually, file-posix implements .bdrv_check_perm and could just
refuse attaching a parent there if it doesn't request a specific
permission like BLK_PERM_SUPPORT_ZONED. That should give us the
whitelist semantics through existing infrastructure.

Kevin