[PATCH v3] usb: core: Prevent null pointer dereference in update_port_device_state

[PATCH v3] usb: core: Prevent null pointer dereference in update_port_device_state

[Udipto Goswami]

Currently,the function update_port_device_state gets the usb_hub from
udev->parent by calling usb_hub_to_struct_hub.
However, in case the actconfig or the maxchild is 0, the usb_hub would
be NULL and upon further accessing to get port_dev would result in null
pointer dereference.

Fix this by introducing an if check after the usb_hub is populated.

Fixes: 83cb2604f641 ("usb: core: add sysfs entry for usb device state")
Cc: stable@vger.kernel.org
Signed-off-by: Udipto Goswami <quic_ugoswami@quicinc.com>
---
v3: Re-wrote the comment for better context.
v2: Introduced comment for the if check & CC'ed stable.

 drivers/usb/core/hub.c | 20 +++++++++++++++++---
 1 file changed, 17 insertions(+), 3 deletions(-)

diff --git a/drivers/usb/core/hub.c b/drivers/usb/core/hub.c
index ffd7c99e24a3..6b514546e59b 100644
--- a/drivers/usb/core/hub.c
+++ b/drivers/usb/core/hub.c
@@ -2053,9 +2053,23 @@ static void update_port_device_state(struct usb_device *udev)
 
 	if (udev->parent) {
 		hub = usb_hub_to_struct_hub(udev->parent);
-		port_dev = hub->ports[udev->portnum - 1];
-		WRITE_ONCE(port_dev->state, udev->state);
-		sysfs_notify_dirent(port_dev->state_kn);
+
+		/*
+		 * The Link Layer Validation System Driver (lvstest)
+		 * has procedure of unbinding the hub before running
+		 * the rest of the procedure. This triggers
+		 * hub_disconnect will set the hub's maxchild to 0.
+		 * This would result usb_hub_to_struct_hub in this
+		 * function to return NULL.
+		 *
+		 * Add if check to avoid running into NULL pointer
+		 * de-reference.
+		 */
+		if (hub) {
+			port_dev = hub->ports[udev->portnum - 1];
+			WRITE_ONCE(port_dev->state, udev->state);
+			sysfs_notify_dirent(port_dev->state_kn);
+		}
 	}
 }
 
-- 
2.17.1

Re: [PATCH v3] usb: core: Prevent null pointer dereference in update_port_device_state

[Sergei Shtylyov]

On 1/9/24 9:17 AM, Udipto Goswami wrote:

> Currently,the function update_port_device_state gets the usb_hub from
> udev->parent by calling usb_hub_to_struct_hub.
> However, in case the actconfig or the maxchild is 0, the usb_hub would
> be NULL and upon further accessing to get port_dev would result in null
> pointer dereference.
> 
> Fix this by introducing an if check after the usb_hub is populated.
> 
> Fixes: 83cb2604f641 ("usb: core: add sysfs entry for usb device state")
> Cc: stable@vger.kernel.org
> Signed-off-by: Udipto Goswami <quic_ugoswami@quicinc.com>
> ---
> v3: Re-wrote the comment for better context.
> v2: Introduced comment for the if check & CC'ed stable.
> 
>  drivers/usb/core/hub.c | 20 +++++++++++++++++---
>  1 file changed, 17 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/usb/core/hub.c b/drivers/usb/core/hub.c
> index ffd7c99e24a3..6b514546e59b 100644
> --- a/drivers/usb/core/hub.c
> +++ b/drivers/usb/core/hub.c
> @@ -2053,9 +2053,23 @@ static void update_port_device_state(struct usb_device *udev)
>  
>  	if (udev->parent) {
>  		hub = usb_hub_to_struct_hub(udev->parent);
> -		port_dev = hub->ports[udev->portnum - 1];
> -		WRITE_ONCE(port_dev->state, udev->state);
> -		sysfs_notify_dirent(port_dev->state_kn);
> +
> +		/*
> +		 * The Link Layer Validation System Driver (lvstest)
> +		 * has procedure of unbinding the hub before running
> +		 * the rest of the procedure. This triggers
> +		 * hub_disconnect will set the hub's maxchild to 0.

   I can't parse this sentence, s/th is missing...

> +		 * This would result usb_hub_to_struct_hub in this
> +		 * function to return NULL.

   "This would result in usb_hub_to_struct_hub in this function
returning NULL", perhaps?

> +		 *
> +		 * Add if check to avoid running into NULL pointer
> +		 * de-reference.
> +		 */
> +		if (hub) {
> +			port_dev = hub->ports[udev->portnum - 1];
> +			WRITE_ONCE(port_dev->state, udev->state);
> +			sysfs_notify_dirent(port_dev->state_kn);
> +		}
>  	}
>  }
>  

MBR, Sergey

Re: [PATCH v3] usb: core: Prevent null pointer dereference in update_port_device_state

[Udipto Goswami]

Hi Sergei,

On 1/9/2024 3:01 PM, Sergei Shtylyov wrote:
> On 1/9/24 9:17 AM, Udipto Goswami wrote:
> 
>> Currently,the function update_port_device_state gets the usb_hub from
>> udev->parent by calling usb_hub_to_struct_hub.
>> However, in case the actconfig or the maxchild is 0, the usb_hub would
>> be NULL and upon further accessing to get port_dev would result in null
>> pointer dereference.
>>
>> Fix this by introducing an if check after the usb_hub is populated.
>>
>> Fixes: 83cb2604f641 ("usb: core: add sysfs entry for usb device state")
>> Cc: stable@vger.kernel.org
>> Signed-off-by: Udipto Goswami <quic_ugoswami@quicinc.com>
>> ---
>> v3: Re-wrote the comment for better context.
>> v2: Introduced comment for the if check & CC'ed stable.
>>
>>   drivers/usb/core/hub.c | 20 +++++++++++++++++---
>>   1 file changed, 17 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/usb/core/hub.c b/drivers/usb/core/hub.c
>> index ffd7c99e24a3..6b514546e59b 100644
>> --- a/drivers/usb/core/hub.c
>> +++ b/drivers/usb/core/hub.c
>> @@ -2053,9 +2053,23 @@ static void update_port_device_state(struct usb_device *udev)
>>   
>>   	if (udev->parent) {
>>   		hub = usb_hub_to_struct_hub(udev->parent);
>> -		port_dev = hub->ports[udev->portnum - 1];
>> -		WRITE_ONCE(port_dev->state, udev->state);
>> -		sysfs_notify_dirent(port_dev->state_kn);
>> +
>> +		/*
>> +		 * The Link Layer Validation System Driver (lvstest)
>> +		 * has procedure of unbinding the hub before running
>> +		 * the rest of the procedure. This triggers
>> +		 * hub_disconnect will set the hub's maxchild to 0.
> 
>     I can't parse this sentence, s/th is missing...
Thanks for the review.
Maybe this would sound better?

"This triggers hub_disconnect which will set hub's maxchild to 0"
> 
>> +		 * This would result usb_hub_to_struct_hub in this
>> +		 * function to return NULL.
> 
>     "This would result in usb_hub_to_struct_hub in this function
> returning NULL", perhaps?

yah sound better. Will take care of it in next version.

Thanks,
-Udipto

Re: [PATCH v3] usb: core: Prevent null pointer dereference in update_port_device_state

[Sergei Shtylyov]

On 1/9/24 2:57 PM, Udipto Goswami wrote:
[...]

>>> Currently,the function update_port_device_state gets the usb_hub from
>>> udev->parent by calling usb_hub_to_struct_hub.
>>> However, in case the actconfig or the maxchild is 0, the usb_hub would
>>> be NULL and upon further accessing to get port_dev would result in null
>>> pointer dereference.
>>>
>>> Fix this by introducing an if check after the usb_hub is populated.
>>>
>>> Fixes: 83cb2604f641 ("usb: core: add sysfs entry for usb device state")
>>> Cc: stable@vger.kernel.org
>>> Signed-off-by: Udipto Goswami <quic_ugoswami@quicinc.com>
>>> ---
>>> v3: Re-wrote the comment for better context.
>>> v2: Introduced comment for the if check & CC'ed stable.
>>>
>>>   drivers/usb/core/hub.c | 20 +++++++++++++++++---
>>>   1 file changed, 17 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/drivers/usb/core/hub.c b/drivers/usb/core/hub.c
>>> index ffd7c99e24a3..6b514546e59b 100644
>>> --- a/drivers/usb/core/hub.c
>>> +++ b/drivers/usb/core/hub.c
>>> @@ -2053,9 +2053,23 @@ static void update_port_device_state(struct usb_device *udev)
>>>         if (udev->parent) {
>>>           hub = usb_hub_to_struct_hub(udev->parent);
>>> -        port_dev = hub->ports[udev->portnum - 1];
>>> -        WRITE_ONCE(port_dev->state, udev->state);
>>> -        sysfs_notify_dirent(port_dev->state_kn);
>>> +
>>> +        /*
>>> +         * The Link Layer Validation System Driver (lvstest)
>>> +         * has procedure of unbinding the hub before running
>>> +         * the rest of the procedure. This triggers
>>> +         * hub_disconnect will set the hub's maxchild to 0.
>>
>>     I can't parse this sentence, s/th is missing...
> Thanks for the review.
> Maybe this would sound better?
> 
> "This triggers hub_disconnect which will set hub's maxchild to 0"

   That seems parsable. :-)

>>> +         * This would result usb_hub_to_struct_hub in this
>>> +         * function to return NULL.
>>
>>     "This would result in usb_hub_to_struct_hub in this function
>> returning NULL", perhaps?
> 
> yah sound better. Will take care of it in next version.

   Probably "in this function" should be dropped altogether...

> Thanks,
> -Udipto

MBR, Sergey

Re: [PATCH v3] usb: core: Prevent null pointer dereference in update_port_device_state

[Udipto Goswami]

On 1/9/2024 8:10 PM, Sergei Shtylyov wrote:
> On 1/9/24 2:57 PM, Udipto Goswami wrote:
> [...]
> 
>>>> Currently,the function update_port_device_state gets the usb_hub from
>>>> udev->parent by calling usb_hub_to_struct_hub.
>>>> However, in case the actconfig or the maxchild is 0, the usb_hub would
>>>> be NULL and upon further accessing to get port_dev would result in null
>>>> pointer dereference.
>>>>
>>>> Fix this by introducing an if check after the usb_hub is populated.
>>>>
>>>> Fixes: 83cb2604f641 ("usb: core: add sysfs entry for usb device state")
>>>> Cc: stable@vger.kernel.org
>>>> Signed-off-by: Udipto Goswami <quic_ugoswami@quicinc.com>
>>>> ---
>>>> v3: Re-wrote the comment for better context.
>>>> v2: Introduced comment for the if check & CC'ed stable.
>>>>
>>>>    drivers/usb/core/hub.c | 20 +++++++++++++++++---
>>>>    1 file changed, 17 insertions(+), 3 deletions(-)
>>>>
>>>> diff --git a/drivers/usb/core/hub.c b/drivers/usb/core/hub.c
>>>> index ffd7c99e24a3..6b514546e59b 100644
>>>> --- a/drivers/usb/core/hub.c
>>>> +++ b/drivers/usb/core/hub.c
>>>> @@ -2053,9 +2053,23 @@ static void update_port_device_state(struct usb_device *udev)
>>>>          if (udev->parent) {
>>>>            hub = usb_hub_to_struct_hub(udev->parent);
>>>> -        port_dev = hub->ports[udev->portnum - 1];
>>>> -        WRITE_ONCE(port_dev->state, udev->state);
>>>> -        sysfs_notify_dirent(port_dev->state_kn);
>>>> +
>>>> +        /*
>>>> +         * The Link Layer Validation System Driver (lvstest)
>>>> +         * has procedure of unbinding the hub before running
>>>> +         * the rest of the procedure. This triggers
>>>> +         * hub_disconnect will set the hub's maxchild to 0.
>>>
>>>      I can't parse this sentence, s/th is missing...
>> Thanks for the review.
>> Maybe this would sound better?
>>
>> "This triggers hub_disconnect which will set hub's maxchild to 0"
> 
>     That seems parsable. :-)
> 
>>>> +         * This would result usb_hub_to_struct_hub in this
>>>> +         * function to return NULL.
>>>
>>>      "This would result in usb_hub_to_struct_hub in this function
>>> returning NULL", perhaps?
>>
>> yah sound better. Will take care of it in next version.
> 
>     Probably "in this function" should be dropped altogether...

sure, i'll remove in v4.

Thanks,
-Udipto