Re: [PATCH v1] virtio-pci: Fix legacy device flag setting error in probe

[Feng Liu <feliu@nvidia.com>]

On 2023-07-19 p.m.10:27, Jason Wang wrote:
> External email: Use caution opening links or attachments
> 
> 
> On Wed, Jul 19, 2023 at 11:46 PM Feng Liu <feliu@nvidia.com> wrote:
>>
>> The 'is_legacy' flag is used to differentiate between legacy vs modern
>> device. Currently, it is based on the value of vp_dev->ldev.ioaddr.
>> However, due to the shared memory of the union between struct
>> virtio_pci_legacy_device and struct virtio_pci_modern_device, when
>> virtio_pci_modern_probe modifies the content of struct
>> virtio_pci_modern_device, it affects the content of struct
>> virtio_pci_legacy_device, and ldev.ioaddr is no longer zero, causing
>> the 'is_legacy' flag to be set as true. To resolve issue, when legacy
>> device is probed, mark 'is_legacy' as true, when modern device is
>> probed, keep 'is_legacy' as false.
>>
>> Fixes: 4f0fc22534e3 ("virtio_pci: Optimize virtio_pci_device structure size")
>> Signed-off-by: Feng Liu <feliu@nvidia.com>
>> Reviewed-by: Parav Pandit <parav@nvidia.com>
>> Reviewed-by: Jiri Pirko <jiri@nvidia.com>
>> ---
>>   drivers/virtio/virtio_pci_common.c | 2 --
>>   drivers/virtio/virtio_pci_legacy.c | 1 +
>>   2 files changed, 1 insertion(+), 2 deletions(-)
>>
>> diff --git a/drivers/virtio/virtio_pci_common.c b/drivers/virtio/virtio_pci_common.c
>> index a6c86f916dbd..c2524a7207cf 100644
>> --- a/drivers/virtio/virtio_pci_common.c
>> +++ b/drivers/virtio/virtio_pci_common.c
>> @@ -557,8 +557,6 @@ static int virtio_pci_probe(struct pci_dev *pci_dev,
>>
>>          pci_set_master(pci_dev);
>>
>> -       vp_dev->is_legacy = vp_dev->ldev.ioaddr ? true : false;
>> -
>>          rc = register_virtio_device(&vp_dev->vdev);
>>          reg_dev = vp_dev;
>>          if (rc)
>> diff --git a/drivers/virtio/virtio_pci_legacy.c b/drivers/virtio/virtio_pci_legacy.c
>> index 2257f1b3d8ae..d9cbb02b35a1 100644
>> --- a/drivers/virtio/virtio_pci_legacy.c
>> +++ b/drivers/virtio/virtio_pci_legacy.c
>> @@ -223,6 +223,7 @@ int virtio_pci_legacy_probe(struct virtio_pci_device *vp_dev)
>>          vp_dev->config_vector = vp_config_vector;
>>          vp_dev->setup_vq = setup_vq;
>>          vp_dev->del_vq = del_vq;
>> +       vp_dev->is_legacy = true;
> 
> This seems break force_legacy for modern device:
> 
>          if (force_legacy) {
>                  rc = virtio_pci_legacy_probe(vp_dev);
>                  /* Also try modern mode if we can't map BAR0 (no IO space). */
>                  if (rc == -ENODEV || rc == -ENOMEM)
>                          rc = virtio_pci_modern_probe(vp_dev);
> 
> Thanks
> 

Hi, Jason

In the case of force_legacy, if no IO space occurs, function will return 
directly after vp_legacy_probe, and will not run vp_dev->is_legacy = 
true; because vp_dev is allocated through kzalloc, the default 
vp_dev->is_legacy is false, which It is expected for modern device, so 
it will not break modern device.

What do you think?

int virtio_pci_legacy_probe(struct virtio_pci_device *vp_dev)
{
[...]

	rc = vp_legacy_probe(ldev);
	if (rc)
		return rc;  /* if no IO space, function will return from here */

[...]
	vp_dev->is_legacy = true;
}


>>
>>          return 0;
>>   }
>> --
>> 2.37.1 (Apple Git-137.1)
>>
>