Skip to contentSkip to conversation

virtio-iommu-pci: Advertise the device as modern-only

PATCH In review Eric Auger 2 participants 6 messages Sep 8, 2020 – Sep 18, 2020

lore
Share & tools
lore

Eric Auger

author
#0 lore raw

v2 cover virtio-iommu-pci: Advertise the device as modern-only

Since commit 9b3a35ec82 ("virtio: verify that legacy support is not accidentally on"), we are forced to use disable-legacy=on when instantiating the virtio-iommu-pci device. This also revealed that the unrealize() function is likely to call g_tree_destroy() on NULL gtrees, which triggers assertions.

Best Regards

Eric

History:
v1 -> v2:
- Fix patch #1 commit message typos
- Add Connie's R-b on both patches

Eric Auger (2):
virtio-iommu: Check gtrees are non null before destroying them
virtio-iommu-pci: force virtio version 1

 hw/virtio/virtio-iommu-pci.c | 2 +-
 hw/virtio/virtio-iommu.c     | 8 ++++++--
 2 files changed, 7 insertions(+), 3 deletions(-)
Signature · 1 line
2.21.3
author
#2 lore raw

PATCH v2 2/2 virtio-iommu-pci: force virtio version 1

Commit 9b3a35ec82 ("virtio: verify that legacy support is not accidentally on") added a safety check that requires to set 'disable-legacy=on' on virtio-iommu-pci:

qemu-system-aarch64: -device virtio-iommu-pci: device is modern-only, use disable-legacy=on

virtio-iommu was introduced after the release of VIRTIO 1.0 specifications, so it should be 'modern-only'.

This patch forces virtio version 1 and removes the 'transitional_name' property removing the need to specify 'disable-legacy=on' on virtio-iommu-pci device.

---
v1 -> v2:
- Added Connie's R-b

---
 hw/virtio/virtio-iommu-pci.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
hw/virtio/virtio-iommu-pci.c +1 -1
index ba62d60a0a..3b6f7a11c6 100644
@@ -68,6 +68,7 @@ static void virtio_iommu_pci_realize(VirtIOPCIProxy *vpci_dev, Error **errp)
     object_property_set_link(OBJECT(dev), "primary-bus",
                              OBJECT(pci_get_bus(&vpci_dev->pci_dev)),
                              &error_abort);
+    virtio_pci_force_virtio_1(vpci_dev);
     qdev_realize(vdev, BUS(&vpci_dev->bus), errp);
 }
 
@@ -97,7 +98,6 @@ static void virtio_iommu_pci_instance_init(Object *obj)
 static const VirtioPCIDeviceTypeInfo virtio_iommu_pci_info = {
     .base_name             = TYPE_VIRTIO_IOMMU_PCI,
     .generic_name          = "virtio-iommu-pci",
-    .transitional_name     = "virtio-iommu-pci-transitional",
     .non_transitional_name = "virtio-iommu-pci-non-transitional",
     .instance_size = sizeof(VirtIOIOMMUPCI),
     .instance_init = virtio_iommu_pci_instance_init,
Signature · 1 line
2.21.3
author
#1 lore raw

PATCH v2 1/2 virtio-iommu: Check gtrees are non null before destroying them

In reply to Eric Auger

If realize fails, domains and endpoints trees may be NULL. On unrealize(), this produces assertions:

"GLib: g_tree_destroy: assertion 'tree != NULL' failed"

Check that the trees are non NULL before destroying them.

---
v1 -> v2:
- fix the commit message
- Added Connie's R-b

---
 hw/virtio/virtio-iommu.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
hw/virtio/virtio-iommu.c +6 -2
index 5d56865e56..21ec63b108 100644
@@ -801,8 +801,12 @@ static void virtio_iommu_device_unrealize(DeviceState *dev)
     VirtIOIOMMU *s = VIRTIO_IOMMU(dev);
 
     g_hash_table_destroy(s->as_by_busptr);
-    g_tree_destroy(s->domains);
-    g_tree_destroy(s->endpoints);
+    if (s->domains) {
+        g_tree_destroy(s->domains);
+    }
+    if (s->endpoints) {
+        g_tree_destroy(s->endpoints);
+    }
 
     virtio_delete_queue(s->req_vq);
     virtio_delete_queue(s->event_vq);
Signature · 1 line
2.21.3
#3 lore raw

Re: [PATCH v2 2/2] virtio-iommu-pci: force virtio version 1

In reply to Eric Auger
Quoted 7 lines · On Tue, 8 Sep 2020 21:33:09 +0200 Eric Auger <[email protected]> wrote:

Commit 9b3a35ec82 ("virtio: verify that legacy support is not accidentally on") added a safety check that requires to set 'disable-legacy=on' on virtio-iommu-pci:

qemu-system-aarch64: -device virtio-iommu-pci: device is modern-only, use disable-legacy=on

Quoted 6 lines

virtio-iommu was introduced after the release of VIRTIO 1.0 specifications, so it should be 'modern-only'.

This patch forces virtio version 1 and removes the 'transitional_name' property removing the need to specify 'disable-legacy=on' on virtio-iommu-pci device.

Not sure whether this patch has been queued already, and how much we care about migration compatibility for virtio-iommu, but would it make sense to force modern on 5.1+ compat machines only? (see
https://lore.kernel.org/qemu-devel/[email protected]/)

Quoted 32 lines

---
v1 -> v2:
- Added Connie's R-b

---
 hw/virtio/virtio-iommu-pci.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
hw/virtio/virtio-iommu-pci.c +1 -1
index ba62d60a0a..3b6f7a11c6 100644
@@ -68,6 +68,7 @@ static void virtio_iommu_pci_realize(VirtIOPCIProxy *vpci_dev, Error **errp)
     object_property_set_link(OBJECT(dev), "primary-bus",
                              OBJECT(pci_get_bus(&vpci_dev->pci_dev)),
                              &error_abort);
+    virtio_pci_force_virtio_1(vpci_dev);
     qdev_realize(vdev, BUS(&vpci_dev->bus), errp);
 }
 
@@ -97,7 +98,6 @@ static void virtio_iommu_pci_instance_init(Object *obj)
 static const VirtioPCIDeviceTypeInfo virtio_iommu_pci_info = {
     .base_name             = TYPE_VIRTIO_IOMMU_PCI,
     .generic_name          = "virtio-iommu-pci",
-    .transitional_name     = "virtio-iommu-pci-transitional",
     .non_transitional_name = "virtio-iommu-pci-non-transitional",
     .instance_size = sizeof(VirtIOIOMMUPCI),
     .instance_init = virtio_iommu_pci_instance_init,
author
#4 lore raw

Re: [PATCH v2 2/2] virtio-iommu-pci: force virtio version 1

Hi Connie,

Quoted 13 lines · Cornelia Huck
Quoted 10 lines · On Tue, 8 Sep 2020 21:33:09 +0200 Eric Auger <[email protected]> wrote:

Commit 9b3a35ec82 ("virtio: verify that legacy support is not accidentally on") added a safety check that requires to set 'disable-legacy=on' on virtio-iommu-pci:

qemu-system-aarch64: -device virtio-iommu-pci: device is modern-only, use disable-legacy=on

virtio-iommu was introduced after the release of VIRTIO 1.0 specifications, so it should be 'modern-only'.

Quoted 8 lines
Quoted 3 lines

This patch forces virtio version 1 and removes the 'transitional_name' property removing the need to specify 'disable-legacy=on' on virtio-iommu-pci device.

Not sure whether this patch has been queued already, and how much we care about migration compatibility for virtio-iommu, but would it make sense to force modern on 5.1+ compat machines only? (see
https://lore.kernel.org/qemu-devel/[email protected]/)

I don't think it was pulled yet.

Quoted 3 lines
Quoted 3 lines

The virtio-iommu-pci device only is usable on ARM in dt mode so I don't think it has production users at the moment.

Thanks

Eric

Quoted 32 lines
Quoted 30 lines

---
v1 -> v2:
- Added Connie's R-b

---
 hw/virtio/virtio-iommu-pci.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
hw/virtio/virtio-iommu-pci.c +1 -1
index ba62d60a0a..3b6f7a11c6 100644
@@ -68,6 +68,7 @@ static void virtio_iommu_pci_realize(VirtIOPCIProxy *vpci_dev, Error **errp)
     object_property_set_link(OBJECT(dev), "primary-bus",
                              OBJECT(pci_get_bus(&vpci_dev->pci_dev)),
                              &error_abort);
+    virtio_pci_force_virtio_1(vpci_dev);
     qdev_realize(vdev, BUS(&vpci_dev->bus), errp);
 }
 
@@ -97,7 +98,6 @@ static void virtio_iommu_pci_instance_init(Object *obj)
 static const VirtioPCIDeviceTypeInfo virtio_iommu_pci_info = {
     .base_name             = TYPE_VIRTIO_IOMMU_PCI,
     .generic_name          = "virtio-iommu-pci",
-    .transitional_name     = "virtio-iommu-pci-transitional",
     .non_transitional_name = "virtio-iommu-pci-non-transitional",
     .instance_size = sizeof(VirtIOIOMMUPCI),
     .instance_init = virtio_iommu_pci_instance_init,
#5 lore raw

Re: [PATCH v2 2/2] virtio-iommu-pci: force virtio version 1

Quoted 25 lines · On Fri, 18 Sep 2020 12:24:02 +0200 Auger Eric <[email protected]> wrote:

Hi Connie,

Quoted 13 lines · Cornelia Huck
Quoted 10 lines · On Tue, 8 Sep 2020 21:33:09 +0200 Eric Auger <[email protected]> wrote:

Commit 9b3a35ec82 ("virtio: verify that legacy support is not accidentally on") added a safety check that requires to set 'disable-legacy=on' on virtio-iommu-pci:

qemu-system-aarch64: -device virtio-iommu-pci: device is modern-only, use disable-legacy=on

virtio-iommu was introduced after the release of VIRTIO 1.0 specifications, so it should be 'modern-only'.

Quoted 8 lines
Quoted 3 lines

This patch forces virtio version 1 and removes the 'transitional_name' property removing the need to specify 'disable-legacy=on' on virtio-iommu-pci device.

Not sure whether this patch has been queued already, and how much we care about migration compatibility for virtio-iommu, but would it make sense to force modern on 5.1+ compat machines only? (see
https://lore.kernel.org/qemu-devel/[email protected]/)

Quoted 7 lines

I don't think it was pulled yet.

Quoted 3 lines
Quoted 3 lines

The virtio-iommu-pci device only is usable on ARM in dt mode so I don't think it has production users at the moment.

OK, then we can keep this patch here, I guess.