Re: [PATCH] request_firmware(): fixes and polishing.

[Dmitry Torokhov <dtor_core@ameritech.net>]

On Tuesday 24 February 2004 08:34 pm, Manuel Estrada Sainz wrote:
> 
>  Hi,
> 
>  Please apply.
> 
>  Dmitry Torokhov has been criticizing my code for some days (Thanks Dmitry),
>  and here is the result. It should be ready for -mm tree.
>  
>  Simon Kelly tested the patch series and reported improvement with some
>  problems he was having.
> 

I have couple more fixes to the firmware loader class:

firmware-pin_module.patch:
  - we need to pin the firmware module if we successfully registered firmware
    class device and "put" it in device re4lease function otherwise the
    module could be unloaded too early. Consider:
    - some module requests firmware
    - firmware loader registers class device and calls hotplug
    - userspace hangs keeping 
    - firmware loader times out
    - the calls module is unloaded. Now firmware loader has 0 refcount and
      cazn be unloaded as well leaving device class behind.
    Am I seeing things?

firmware-hotplug.patch
  - I stll think that we sould not call userspace until we registered all
    necessary attributes. Now that Greg put my changes to kobject in mainline
    ist very easy -  just inhibit hotplug handler until we are ready and call
    kobject_hitplug by ourselves.

Please comment.

-- 
Dmitry