[PATCH] vfio: Fix NULL pointer dereference in vfio_pci_bus_notifier

[PATCH] vfio: Fix NULL pointer dereference in vfio_pci_bus_notifier

[Kunwu Chan]

kasprintf() returns a pointer to dynamically allocated memory
which can be NULL upon failure. Ensure the allocation was successful
by checking the pointer validity.

Signed-off-by: Kunwu Chan <chentao@kylinos.cn>
---
 drivers/vfio/pci/vfio_pci_core.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/vfio/pci/vfio_pci_core.c b/drivers/vfio/pci/vfio_pci_core.c
index 1cbc990d42e0..74e5b89a3a0c 100644
--- a/drivers/vfio/pci/vfio_pci_core.c
+++ b/drivers/vfio/pci/vfio_pci_core.c
@@ -2047,6 +2047,8 @@ static int vfio_pci_bus_notifier(struct notifier_block *nb,
 			 pci_name(pdev));
 		pdev->driver_override = kasprintf(GFP_KERNEL, "%s",
 						  vdev->vdev.ops->name);
+		if (!pdev->driver_override)
+			return -ENOMEM;
 	} else if (action == BUS_NOTIFY_BOUND_DRIVER &&
 		   pdev->is_virtfn && physfn == vdev->pdev) {
 		struct pci_driver *drv = pci_dev_driver(pdev);
-- 
2.39.2

Re: [PATCH] vfio: Fix NULL pointer dereference in vfio_pci_bus_notifier

[Alex Williamson]

On Fri, 12 Jan 2024 14:22:21 +0800
Kunwu Chan <chentao@kylinos.cn> wrote:

> kasprintf() returns a pointer to dynamically allocated memory
> which can be NULL upon failure. Ensure the allocation was successful
> by checking the pointer validity.
> 
> Signed-off-by: Kunwu Chan <chentao@kylinos.cn>
> ---
>  drivers/vfio/pci/vfio_pci_core.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/drivers/vfio/pci/vfio_pci_core.c b/drivers/vfio/pci/vfio_pci_core.c
> index 1cbc990d42e0..74e5b89a3a0c 100644
> --- a/drivers/vfio/pci/vfio_pci_core.c
> +++ b/drivers/vfio/pci/vfio_pci_core.c
> @@ -2047,6 +2047,8 @@ static int vfio_pci_bus_notifier(struct notifier_block *nb,
>  			 pci_name(pdev));
>  		pdev->driver_override = kasprintf(GFP_KERNEL, "%s",
>  						  vdev->vdev.ops->name);
> +		if (!pdev->driver_override)
> +			return -ENOMEM;
>  	} else if (action == BUS_NOTIFY_BOUND_DRIVER &&
>  		   pdev->is_virtfn && physfn == vdev->pdev) {
>  		struct pci_driver *drv = pci_dev_driver(pdev);

This is a blocking notifier callback, so errno isn't a proper return
value, nor does it accomplish anything.  We're into the realm of
worrying about small allocation failures here, which I understand
essentially cannot happen, but about the best we could do at this
point would be to WARN_ON if we weren't able to allocate an override.
Thanks,

Alex

Re: [PATCH] vfio: Fix NULL pointer dereference in vfio_pci_bus_notifier

[Kunwu Chan]

On 2024/1/12 23:30, Alex Williamson wrote:
> On Fri, 12 Jan 2024 14:22:21 +0800
> Kunwu Chan <chentao@kylinos.cn> wrote:
> 
>> kasprintf() returns a pointer to dynamically allocated memory
>> which can be NULL upon failure. Ensure the allocation was successful
>> by checking the pointer validity.
>>
>> Signed-off-by: Kunwu Chan <chentao@kylinos.cn>
>> ---
>>   drivers/vfio/pci/vfio_pci_core.c | 2 ++
>>   1 file changed, 2 insertions(+)
>>
>> diff --git a/drivers/vfio/pci/vfio_pci_core.c b/drivers/vfio/pci/vfio_pci_core.c
>> index 1cbc990d42e0..74e5b89a3a0c 100644
>> --- a/drivers/vfio/pci/vfio_pci_core.c
>> +++ b/drivers/vfio/pci/vfio_pci_core.c
>> @@ -2047,6 +2047,8 @@ static int vfio_pci_bus_notifier(struct notifier_block *nb,
>>   			 pci_name(pdev));
>>   		pdev->driver_override = kasprintf(GFP_KERNEL, "%s",
>>   						  vdev->vdev.ops->name);
>> +		if (!pdev->driver_override)
>> +			return -ENOMEM;
>>   	} else if (action == BUS_NOTIFY_BOUND_DRIVER &&
>>   		   pdev->is_virtfn && physfn == vdev->pdev) {
>>   		struct pci_driver *drv = pci_dev_driver(pdev);
> 
> This is a blocking notifier callback, so errno isn't a proper return
> value, nor does it accomplish anything.  We're into the realm of
> worrying about small allocation failures here, which I understand
> essentially cannot happen, but about the best we could do at this
> point would be to WARN_ON if we weren't able to allocate an override.
Thanks for your reply.
I'll update v2 patch use WARN_ON to print some callstack msgs when we 
weren't able to allocate an override.

These msgs could reduce some of the worries and help us to find what happed.

> Thanks,
> 
> Alex
> 
-- 
Thanks,
   Kunwu