Re: [Qemu-devel] [PATCH] virtio-scsi: forward scsibus for virtio-scsi-pci.

["Michael S. Tsirkin" <mst@redhat.com>]

On Tue, Jun 11, 2013 at 09:32:05AM +0200, Frederic Konrad wrote:
> On 11/06/2013 09:21, Michael S. Tsirkin wrote:
> >On Tue, Jun 11, 2013 at 08:43:51AM +0200, Frederic Konrad wrote:
> >>On 10/06/2013 19:00, Michael S. Tsirkin wrote:
> >>>On Mon, Jun 10, 2013 at 11:53:04AM +0200, fred.konrad@greensocs.com wrote:
> >>>>From: KONRAD Frederic <fred.konrad@greensocs.com>
> >>>>
> >>>>This fix a bug with scsi hotplug on virtio-scsi-pci:
> >>>>
> >>>>As virtio-scsi-pci doesn't have any scsi bus, we need to forward scsi-hot-add
> >>>>to the virtio-scsi-device plugged on the virtio-bus.
> >>>>
> >>>>Reported-by: Alexey Kardashevskiy <aik@ozlabs.ru>
> >>>>Signed-off-by: KONRAD Frederic <fred.konrad@greensocs.com>
> >>>>---
> >>>>  hw/pci/pci-hotplug.c | 19 +++++++++++++++++--
> >>>>  1 file changed, 17 insertions(+), 2 deletions(-)
> >>>>
> >>>>diff --git a/hw/pci/pci-hotplug.c b/hw/pci/pci-hotplug.c
> >>>>index 12287d1..c708752 100644
> >>>>--- a/hw/pci/pci-hotplug.c
> >>>>+++ b/hw/pci/pci-hotplug.c
> >>>>@@ -30,6 +30,8 @@
> >>>>  #include "monitor/monitor.h"
> >>>>  #include "hw/scsi/scsi.h"
> >>>>  #include "hw/virtio/virtio-blk.h"
> >>>>+#include "hw/virtio/virtio-scsi.h"
> >>>>+#include "hw/virtio/virtio-pci.h"
> >>>>  #include "qemu/config-file.h"
> >>>>  #include "sysemu/blockdev.h"
> >>>>  #include "qapi/error.h"
> >>>>@@ -79,13 +81,26 @@ static int scsi_hot_add(Monitor *mon, DeviceState *adapter,
> >>>>  {
> >>>>      SCSIBus *scsibus;
> >>>>      SCSIDevice *scsidev;
> >>>>+    VirtIOPCIProxy *virtio_proxy;
> >>>>      scsibus = (SCSIBus *)
> >>>>          object_dynamic_cast(OBJECT(QLIST_FIRST(&adapter->child_bus)),
> >>>>                              TYPE_SCSI_BUS);
> >>>>      if (!scsibus) {
> >>>>-	error_report("Device is not a SCSI adapter");
> >>>>-	return -1;
> >>>>+        /*
> >>>>+         * Check if the adapter is a virtio-scsi-pci, and forward scsi_hot_add
> >>>>+         * to the virtio-scsi-device.
> >>>>+         */
> >>>>+        if (!object_dynamic_cast(OBJECT(adapter), TYPE_VIRTIO_SCSI_PCI)) {
> >>>>+            error_report("Device is not a SCSI adapter");
> >>>>+            return -1;
> >>>>+        }
> >>>>+        virtio_proxy = VIRTIO_PCI(adapter);
> >>>>+        adapter = DEVICE(virtio_proxy->bus.vdev);
> >>>>+        scsibus = (SCSIBus *)
> >>>>+                  object_dynamic_cast(OBJECT(QLIST_FIRST(&adapter->child_bus)),
> >>>>+                            TYPE_SCSI_BUS);
> >>>>+        assert(scsibus);
> >>>>      }
> >>>>      /*
> >>>By the way I really wonder. pci-hotplug.c was supposed to
> >>>be legacy interface.
> >>>Is this the only way to add scsi disks?
> >>>And are other ways broken, too?
> >>Here you can add scsi disks to a given device.
> >>I think the other ways add scsi disks to a given bus?
> >>
> >>Do you see any other?
> >>
> >>Thanks,
> >>Fred
> >I don't know, that's why I'm asking.
> >
> >If it is, that's crazy. There's no reason for
> >it to be tied to pci at all.
> >
> >The commands in pci-hotplug.c are legacy - they don't support things
> >like multi root systems and there's no sane way to make them do it.
> >
> >Your patch probably makes sense for stable, but for 1.6,
> >let's fix this properly.
> 
> Ok, what do you propose to fix that properly?

The problem is not your patch, it's that generally
we don't appear have a supported way to add drives to devices.

I think drive_hot_add should first of all stop
calling pci_drive_hot_add unless pci_addr
is specified. Teach it to get device id
instead. Add a QMP interface, not just HMP.
Rename the file from device-hotplug.c 


> 
> If we want to keep the behaviour of drive_add command,
> I
> think we have two solutions:
>     A/ Change the scsi_hot_add to check if it is a virtio-scsi-pci
> device and take
>         virtio-scsi-device's scsi bus.
>     B/ Make somehow virtio-scsi-pci having a scsi-bus
>         (and the same in virtio-scsi-device).
> 
> I chose A because it seems a lot easier. For the moment B seems not
> possible,
> but would it be a cleaner solution? What do you think?
> 
> Fred

It sounds cleaner though I'm not sure how to implement this.

> >
> >>>>-- 
> >>>>1.8.1.4