[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [libvirt] [PATCH 05/18] qemu: Support interface model=virtio-{non-}transitional

On 01/18/2019 09:20 AM, Andrea Bolognani wrote:
On Thu, 2019-01-17 at 12:52 -0500, Cole Robinson wrote:
Add new <interface> model handling for virtio transitional devices. Ex:

   <model type='virtio-transitional'/>

* "virtio-transitional" maps to qemu "virtio-net-pci-transitional"
* "virtio-non-transitional" maps to qemu "virtio-net-pci-non-transitional"

Signed-off-by: Cole Robinson <crobinso redhat com>
(cherry picked from commit b6698b81846e2010e0cc030bcd6c2c549cf04e97)

Guess this slipped in somehow :)

@@ -1112,6 +1116,8 @@ struct virQEMUCapsStringFlags virQEMUCapsObjectTypes[] = {
      { "memory-backend-memfd", QEMU_CAPS_OBJECT_MEMORY_MEMFD },
      {"virtio-blk-pci-transitional", QEMU_CAPS_DEVICE_VIRTIO_BLK_TRANSITIONAL},
      {"virtio-blk-pci-non-transitional", QEMU_CAPS_DEVICE_VIRTIO_BLK_NON_TRANSITIONAL},
+    {"virtio-net-pci-transitional", QEMU_CAPS_DEVICE_VIRTIO_NET_TRANSITIONAL},
+    {"virtio-net-pci-non-transitional", QEMU_CAPS_DEVICE_VIRTIO_NET_NON_TRANSITIONAL},

Same comments about spacing, naming the capabilities and introducing
them in their separate commit as for virtio-blk apply.

@@ -449,6 +449,7 @@ qemuBuildVirtioTransitional(virBufferPtr buf,
                              virQEMUCapsPtr qemuCaps,
                              virDomainDeviceAddressType type,
                              int model,
+                            const char *modelstr,


We really, really need to make the handling of model names for
network interface more sane.

In the meantime, can you add some sort of (ideally compile-time)
sanity check to ensure that the function is either called with
model == -1 && modelStr != NULL or model >= 0 && modelStr == NULL?
Or at the very least some documentation on how you're supposed to
be calling it.

I'll respond to the rest of the review in time but I just wanted to mention here: I'm working on a series to try and untangle the net model mess and it undoes this hack. It converts nearly all usage of net->model internally to use an enum, but preserves the string passthrough in xen+qemu to be dealt with separately since that will take some discussion and probably need to be phased out after a grace period. I plan to finish + post the series today unless I hit unexpected hurdles.

Thanks for the reviews!

- Cole

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]