Re: A potential broken at platform driver?

[Richard Gong <richard.gong@linux.intel.com>]

Hi Greg,

On 6/4/19 12:03 PM, Greg KH wrote:
> On Tue, Jun 04, 2019 at 11:13:02AM -0500, Richard Gong wrote:
>>
>> Hi Greg,
>>
>> On 6/4/19 9:28 AM, Greg KH wrote:
>>> On Tue, Jun 04, 2019 at 12:33:03PM +0200, Romain Izard wrote:
>>>> On Mon, Jun 03, 2019 at 08:02:55PM +0200, Greg KH wrote:
>>>>>> @@ -394,7 +432,7 @@ static struct platform_driver stratix10_rsu_driver = {
>>>>>>    	.remove = stratix10_rsu_remove,
>>>>>>    	.driver = {
>>>>>>    		.name = "stratix10-rsu",
>>>>>> -		.groups = rsu_groups,
>>>>>> +//		.groups = rsu_groups,
>>>>>
>>>>> Are you sure this is the correct pointer?  I think that might be
>>>>> pointing to the driver's attributes, not the device's attributes.
>>>>>
>>>>> If platform drivers do not have a way to register groups properly, then
>>>>> that really needs to be fixed, as trying to register it by yourself as
>>>>> you are doing, is ripe for racing with userspace.
>>>> This is a very common issue with platform drivers, and it seems to me that
>>>> it is not possible to add device attributes when binding a device to a
>>>> driver without entering the race condition.
>>>>
>>>> My understanding is the following one:
>>>>
>>>> The root cause is that the device has already been created and reported
>>>> to the userspace with a KOBJ_ADD uevent before the device and the driver
>>>> are bound together. On receiving this event, userspace will react, and
>>>> it will try to read the device's attributes. In parallel the kernel will
>>>> try to find a matching driver. If a driver is found, the kernel will
>>>> call the probe function from the driver with the device as a parameter,
>>>> and if successful a KOBJ_BIND uevent will be sent to userspace, but this
>>>> is a recent addition.
>>>>
>>>> Unfortunately, not all created devices will be bound to a driver, and the
>>>> existing udev code relies on KOBJ_ADD uevents rather than KOBJ_BIND uevents.
>>>> If new per-device attributes have been added to the device during the
>>>> binding stage userspace may or may not see them, depending on when userspace
>>>> tries to read the device's attributes.
>>>>
>>>> I have this possible workaround, but I do not know if it is a good solution:
>>>>
>>>> When binding the device and the driver together, create a new device as a
>>>> child to the current device, and fill its "groups" member to point to the
>>>> per-device attributes' group. As the device will be created with all the
>>>> attributes, it will not be affected by the race issues. The functions
>>>> handling the attributes will need to be modified to use the parents of their
>>>> "device" parameter, instead of the device itself. Additionnaly, the sysfs
>>>> location of the attributes will be different, as the child device will show
>>>> up in the sysfs path. But for a newly introduced device this will not be
>>>> a problem.
>>>>
>>>> Is this a good compromise ?
>>>
>>> Not really.  You just want the attributes on the platform device itself.
>>>
>>> Given the horrible hack that platform devices are today, what's one more
>>> hack!
>>>
>>> Here's a patch below of what should probably be done here.  Richard, can
>>> you change your code to use the new dev_groups pointer in the struct
>>> platform_driver and this patch and let me know if that works or not?
>>>
>>> Note, I've only compiled this code, not tested it...
>>>
>>
>> Your patch works.
>>
>> Many thanks for your help!
> 
> Nice!
> 
> I guess I need to turn it into a real patch now.  Let me do that tonight
> and see if I can convert some existing drivers to use it as well...
> 

Sorry for asking.

I haven't seen your patch, did you release that?

Regards,
Richard

> thanks,
> 
> greg k-h
>