Linux Virtualization - Nov 2016 - [PATCH] virtio_mmio: Set dev.release() to avoid warning

I think not. In virtio_pci_common, vp_dev is allocated by kzalloc so a
kfree is needed. Here vm_dev is allocated by devm_kmalloc which is
"automatically freed on driver detach" from the comment
(drivers/base/devres.c:769).

On Thu, Nov 24, 2016 at 6:37 PM, Jason Wang <jasowang at redhat.com>
wrote:
>
>
> On 2016?11?24? 08:31, Yuan Liu wrote:
>
>> From: Yuan Liu <liuyuan at google.com>
>>
>> Fix a warning thrown from virtio_mmio_remove():
>> Device 'virtio0' does not have a release() function
>>
>> The fix is according to virtio_pci_probe() of
>> drivers/virtio/virtio_pci_common.c
>>
>> Signed-off-by: Yuan Liu <liuyuan at google.com>
>> ---
>>   drivers/virtio/virtio_mmio.c | 2 ++
>>   1 file changed, 2 insertions(+)
>>
>> diff --git a/drivers/virtio/virtio_mmio.c
b/drivers/virtio/virtio_mmio.c
>> index 48bfea9..d47a2fc 100644
>> --- a/drivers/virtio/virtio_mmio.c
>> +++ b/drivers/virtio/virtio_mmio.c
>> @@ -489,6 +489,7 @@ static const struct virtio_config_ops
>> virtio_mmio_config_ops = {
>>   };
>>     +static void virtio_mmio_release_dev_empty(struct device *_d) {}
>>
>
> Do we need to free vm_dev here?
>
>
>     /* Platform device */
>>   @@ -511,6 +512,7 @@ static int virtio_mmio_probe(struct
platform_device
>> *pdev)
>>                 return  -ENOMEM;
>>         vm_dev->vdev.dev.parent = &pdev->dev;
>> +       vm_dev->vdev.dev.release = virtio_mmio_release_dev_empty;
>>         vm_dev->vdev.config = &virtio_mmio_config_ops;
>>         vm_dev->pdev = pdev;
>>         INIT_LIST_HEAD(&vm_dev->virtqueues);
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.linuxfoundation.org/pipermail/virtualization/attachments/20161124/a2c0c98b/attachment-0001.html>

On 2016?11?25? 10:47, Yuan Liu wrote:
> I think not. In virtio_pci_common, vp_dev is allocated by kzalloc so a 
> kfree is needed. Here vm_dev is allocated by devm_kmalloc which is 
> "automatically freed on driver detach" from the comment 
> (drivers/base/devres.c:769).
I see, thanks.

Reviewed-by: Jason Wang <jasowang at redhat.com>
>
> On Thu, Nov 24, 2016 at 6:37 PM, Jason Wang <jasowang at redhat.com 
> <mailto:jasowang at redhat.com>> wrote:
>
>
>
>     On 2016?11?24? 08:31, Yuan Liu wrote:
>
>         From: Yuan Liu <liuyuan at google.com <mailto:liuyuan at
google.com>>
>
>         Fix a warning thrown from virtio_mmio_remove():
>         Device 'virtio0' does not have a release() function
>
>         The fix is according to virtio_pci_probe() of
>         drivers/virtio/virtio_pci_common.c
>
>         Signed-off-by: Yuan Liu <liuyuan at google.com
>         <mailto:liuyuan at google.com>>
>         ---
>           drivers/virtio/virtio_mmio.c | 2 ++
>           1 file changed, 2 insertions(+)
>
>         diff --git a/drivers/virtio/virtio_mmio.c
>         b/drivers/virtio/virtio_mmio.c
>         index 48bfea9..d47a2fc 100644
>         --- a/drivers/virtio/virtio_mmio.c
>         +++ b/drivers/virtio/virtio_mmio.c
>         @@ -489,6 +489,7 @@ static const struct virtio_config_ops
>         virtio_mmio_config_ops = {
>           };
>             +static void virtio_mmio_release_dev_empty(struct device
>         *_d) {}
>
>
>     Do we need to free vm_dev here?
>
>
>             /* Platform device */
>           @@ -511,6 +512,7 @@ static int virtio_mmio_probe(struct
>         platform_device *pdev)
>                         return  -ENOMEM;
>                 vm_dev->vdev.dev.parent = &pdev->dev;
>         +       vm_dev->vdev.dev.release =
virtio_mmio_release_dev_empty;
>                 vm_dev->vdev.config = &virtio_mmio_config_ops;
>                 vm_dev->pdev = pdev;
>                 INIT_LIST_HEAD(&vm_dev->virtqueues);
>
>
>