Re: [Qemu-devel] [PATCH] virtio: Mark virtio-device as non-user-creatable

[Halil Pasic]

On 08/03/2017 04:45 PM, Eduardo Habkost wrote:
> On Wed, Aug 02, 2017 at 02:33:00PM +0200, Halil Pasic wrote:
>>
>>
>> On 08/02/2017 01:01 AM, Eduardo Habkost wrote:
>>> TYPE_VIRTIO_DEVICE devices are already not usable with -device
>>> and device_add, but they are reported as user-creatable on
>>> "-device help" and through monitor interfaces.
>>
>> I've tried -device virtio-rng on s390x and from what I see, it
>> seems we 'auto-magically' create the 'controlling' virtio-rng-ccw
>> device. So I have to ask what do you mean by 'already not usable'?
> 
> virtio-rng is just an alias for virtio-rng-, and it's
> not a TYPE_VIRTIO_DEVICE device.
> 
> virtio-rng-device is the TYPE_VIRTIO_DEVICE subclass, and you
> shouldn't be able to use "-device virtio-rng-device".  Except
> that it works with "qemu-system-arm -machine virt", as I noted on
> another message, and this patch breaks it.
> 

Sorry, I was in a hurry leaving for holiday (I'm in a holiday
for approx 2 weeks) and acted on assumptions -- obviously wrong
ones.

AFAIU you withdrew the patch because of this arm exception in
the other message.

Sorry for the noise.

Halil

Re: [Qemu-devel] [PATCH] virtio: Mark virtio-device as non-user-creatable

[Eduardo Habkost]

On Wed, Aug 02, 2017 at 02:33:00PM +0200, Halil Pasic wrote:
> 
> 
> On 08/02/2017 01:01 AM, Eduardo Habkost wrote:
> > TYPE_VIRTIO_DEVICE devices are already not usable with -device
> > and device_add, but they are reported as user-creatable on
> > "-device help" and through monitor interfaces.
> 
> I've tried -device virtio-rng on s390x and from what I see, it
> seems we 'auto-magically' create the 'controlling' virtio-rng-ccw
> device. So I have to ask what do you mean by 'already not usable'?

virtio-rng is just an alias for virtio-rng-, and it's
not a TYPE_VIRTIO_DEVICE device.

virtio-rng-device is the TYPE_VIRTIO_DEVICE subclass, and you
shouldn't be able to use "-device virtio-rng-device".  Except
that it works with "qemu-system-arm -machine virt", as I noted on
another message, and this patch breaks it.

-- 
Eduardo

Re: [Qemu-devel] [PATCH] virtio: Mark virtio-device as non-user-creatable

[Eduardo Habkost]

On Tue, Aug 01, 2017 at 08:01:55PM -0300, Eduardo Habkost wrote:
> TYPE_VIRTIO_DEVICE devices are already not usable with -device
> and device_add, but they are reported as user-creatable on
> "-device help" and through monitor interfaces.
> 
> Mark them as not user-creatable to avoid confusing users, and to
> allow automated testing (e.g. scripts/device-crash-test) to skip
> them.
> 
> Before this patch, device-crash-test will try to test
> virtio-device devices with all machine-types:
> 
>   $ time ./scripts/device-crash-test -D virtio-device -v 
> ./x86_64-softmmu/qemu-system-x86_64
>   [...]
>   INFO: Total: 1088 test cases
>   INFO: Skipped 408 test cases
> 
>   real0m49.775s
> 
> After this patch, the script won't try to test virtio-device
> devices:
> 
>   $ time ./scripts/device-crash-test -D virtio-device -v 
> ./x86_64-softmmu/qemu-system-x86_64
>   INFO: Total: 0 test cases
> 
>   real0m0.092s
> 
> Signed-off-by: Eduardo Habkost 
> ---
>  hw/virtio/virtio.c | 11 +++
>  1 file changed, 11 insertions(+)
> 
> diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
> index 464947f..c4bdb94 100644
> --- a/hw/virtio/virtio.c
> +++ b/hw/virtio/virtio.c
> @@ -2653,6 +2653,17 @@ static void virtio_device_class_init(ObjectClass 
> *klass, void *data)
>  dc->unrealize = virtio_device_unrealize;
>  dc->bus_type = TYPE_VIRTIO_BUS;
>  dc->props = virtio_properties;
> +/*
> + * Reason:
> + * - TYPE_VIRTIO_DEVICE devices are not visible to guests
> + *   unless they are created and controlled by transport-specific
> + *   devices (virtio-pci, virtio-mmio, and virtio-ccw).
> + * - A TYPE_VIRTIO_BUS bus is never available for plugging
> + *   using -device/device_add, as virtio-bus buses are
> + *   created on the fly and immediately populated by the
> + *   transport-specific devices' realize methods.
> + */

Oops, I just found out that this is not true on virtio-mmio:
unused virtio-mmio-bus buses are available for plugging when
virtio-mmio devices are created.

So at least on arm, there are virtio-bus buses where
virtio-device devices can be plugged by users, and this patch is
incorrect.

> +dc->user_creatable = false;
>  vdc->start_ioeventfd = virtio_device_start_ioeventfd_impl;
>  vdc->stop_ioeventfd = virtio_device_stop_ioeventfd_impl;
>  
> -- 
> 2.9.4
> 

-- 
Eduardo

Re: [Qemu-devel] [PATCH] virtio: Mark virtio-device as non-user-creatable

[Halil Pasic]

On 08/02/2017 01:01 AM, Eduardo Habkost wrote:
> TYPE_VIRTIO_DEVICE devices are already not usable with -device
> and device_add, but they are reported as user-creatable on
> "-device help" and through monitor interfaces.

I've tried -device virtio-rng on s390x and from what I see, it
seems we 'auto-magically' create the 'controlling' virtio-rng-ccw
device. So I have to ask what do you mean by 'already not usable'?

> 
> Mark them as not user-creatable to avoid confusing users, and to
> allow automated testing (e.g. scripts/device-crash-test) to skip
> them.
> 
> Before this patch, device-crash-test will try to test
> virtio-device devices with all machine-types:
> 
>   $ time ./scripts/device-crash-test -D virtio-device -v 
> ./x86_64-softmmu/qemu-system-x86_64
>   [...]
>   INFO: Total: 1088 test cases
>   INFO: Skipped 408 test cases
> 
>   real0m49.775s
> 
> After this patch, the script won't try to test virtio-device
> devices:
> 
>   $ time ./scripts/device-crash-test -D virtio-device -v 
> ./x86_64-softmmu/qemu-system-x86_64
>   INFO: Total: 0 test cases
> 
>   real0m0.092s
> 
> Signed-off-by: Eduardo Habkost 
> ---
>  hw/virtio/virtio.c | 11 +++
>  1 file changed, 11 insertions(+)
> 
> diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
> index 464947f..c4bdb94 100644
> --- a/hw/virtio/virtio.c
> +++ b/hw/virtio/virtio.c
> @@ -2653,6 +2653,17 @@ static void virtio_device_class_init(ObjectClass 
> *klass, void *data)
>  dc->unrealize = virtio_device_unrealize;
>  dc->bus_type = TYPE_VIRTIO_BUS;
>  dc->props = virtio_properties;
> +/*
> + * Reason:
> + * - TYPE_VIRTIO_DEVICE devices are not visible to guests
> + *   unless they are created and controlled by transport-specific
> + *   devices (virtio-pci, virtio-mmio, and virtio-ccw).
> + * - A TYPE_VIRTIO_BUS bus is never available for plugging
> + *   using -device/device_add, as virtio-bus buses are
> + *   created on the fly and immediately populated by the
> + *   transport-specific devices' realize methods.
> + */
> +dc->user_creatable = false;
>  vdc->start_ioeventfd = virtio_device_start_ioeventfd_impl;
>  vdc->stop_ioeventfd = virtio_device_stop_ioeventfd_impl;
> 

[Qemu-devel] [PATCH] virtio: Mark virtio-device as non-user-creatable

[Eduardo Habkost]

TYPE_VIRTIO_DEVICE devices are already not usable with -device
and device_add, but they are reported as user-creatable on
"-device help" and through monitor interfaces.

Mark them as not user-creatable to avoid confusing users, and to
allow automated testing (e.g. scripts/device-crash-test) to skip
them.

Before this patch, device-crash-test will try to test
virtio-device devices with all machine-types:

  $ time ./scripts/device-crash-test -D virtio-device -v 
./x86_64-softmmu/qemu-system-x86_64
  [...]
  INFO: Total: 1088 test cases
  INFO: Skipped 408 test cases

  real0m49.775s

After this patch, the script won't try to test virtio-device
devices:

  $ time ./scripts/device-crash-test -D virtio-device -v 
./x86_64-softmmu/qemu-system-x86_64
  INFO: Total: 0 test cases

  real0m0.092s

Signed-off-by: Eduardo Habkost 
---
 hw/virtio/virtio.c | 11 +++
 1 file changed, 11 insertions(+)

diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
index 464947f..c4bdb94 100644
--- a/hw/virtio/virtio.c
+++ b/hw/virtio/virtio.c
@@ -2653,6 +2653,17 @@ static void virtio_device_class_init(ObjectClass *klass, 
void *data)
 dc->unrealize = virtio_device_unrealize;
 dc->bus_type = TYPE_VIRTIO_BUS;
 dc->props = virtio_properties;
+/*
+ * Reason:
+ * - TYPE_VIRTIO_DEVICE devices are not visible to guests
+ *   unless they are created and controlled by transport-specific
+ *   devices (virtio-pci, virtio-mmio, and virtio-ccw).
+ * - A TYPE_VIRTIO_BUS bus is never available for plugging
+ *   using -device/device_add, as virtio-bus buses are
+ *   created on the fly and immediately populated by the
+ *   transport-specific devices' realize methods.
+ */
+dc->user_creatable = false;
 vdc->start_ioeventfd = virtio_device_start_ioeventfd_impl;
 vdc->stop_ioeventfd = virtio_device_stop_ioeventfd_impl;
 
-- 
2.9.4