Re: [PATCH] 2.5.31 driverfs: patch for your consideration

[Adam Belay]

[e-mail] wrote:

>>Also if You're interested here's the write support for "driver".
>>
>
>Suppose we did support this. If you write the name of a driver to a file, 
>we search the bus's list of drivers for a match. We then let the bus 
>compare the hardware IDs and call probe if it matches. 
>
That's exactly how my code works.

>
>
>One big problem is that the IDs in the driver are marked __devinitdata, so
>they're thrown away after init (if hotplugging is not enabled). So, we 
>would have to change every driver.
>
Found that out when I tested binding the agpgart driver.

> 
>
>Besides, it just doesn't make sense. If $user wants to use a different 
>or third party driver, let them rmmod and insmod. 
>
Ok, I guess that makes sense.  My interface was primarily for special 
cases anyway.  What does need to be done is a user level program that 
finds and loads the proper modules automatically.  Maybe we could use 
the existing hotplug scripts or we could even start from scratch.  Maybe 
we should make a file in the source tree where driver developers can 
list their supported hardware IDs but more importantly documentation on 
the attributes registered into driverfs.

>
>
>>PS:  Would you be interested in a patch that would port the pnpbios
>>driver to the driver model?
>>
>
>Yes. 
>
great!

Thanks,
Adam