On Mon, Jun 25, 2012 at 03:51:13AM -0400, mc@linux.vnet.ibm.com wrote:

Quoting Stefan Hajnoczi [off-list ref]:

quoted

On Wed, Jun 20, 2012 at 7:47 AM, Cong Meng [off-list ref] wrote:

quoted

Implement the hotplug() and hot_unplug() interfaces in virtio-scsi,
by signal
the virtio_scsi.ko in guest kernel via event virtual queue.

The counterpart patch of virtio_scsi.ko will be sent soon in another thread.

quoted

Signed-off-by: Cong Meng <redacted>
Signed-off-by: Sen Wang <redacted>
---
hw/virtio-scsi.c |   72
+++++++++++++++++++++++++++++++++++++++++++++++++++--
1 files changed, 69 insertions(+), 3 deletions(-)

I compared against the virtio-scsi specification and this looks good:
http://ozlabs.org/~rusty/virtio-spec/virtio-0.9.5.pdf

Dropped events and event throttling are not implemented by this patch.
This means that the guest can miss events if it runs out of event
queue elements.  A scenario that might be able to trigger this is if
multiple LUNs are hotplugged in a single QEMU monitor callback.

Implementing dropped events is easy in hw/virtio-scsi.c.  Keep a bool
or counter of dropped events and report them when the guest kicks us
with a free event element (virtio_scsi_handle_event).

Yes. It's easy to do this in qemu. But I'm not sure what should be done
in virtio-scsi.ko to respond the "VIRTIO_SCSI_T_EVENTS_MISSED" event.
The spec says "poll the logical units for unit attention conditions", or
just a whole bus rescan?

I'm not sure what the answer is either, maybe you can find an existing
SCSI LLD that does what you need.

Stefan