Re: [PATCH 1/2] input: make serio_register_driver() return error code

[Dmitry Torokhov]

On 11/7/06, Akinobu Mita <akinobu.mita@xxxxxxxxx> wrote:
> serio_register_driver() may fail under memory shortage.
>
> When serio_register_driver() called, it queues SERIO_REGISTER_DRIVER
> event into global serio_event_list, and then kseriod kernel thread
> handles that event and do driver_register().
>
> But event allocation by serio_register_driver() may fail.
> Because it is GFP_ATOMIC allocation. It will cause the problem
> by serio_unregister_driver() with not being registered driver
> at module_exit() time
>
> This patch makes serio_register_driver() call driver_register()
> directly instead of kseriod so that it can check whether
> driver_register() is succeeded or not.

This slows down boot process because probing for mice and keyboards
takes too long (for some touchpads it takes about 4 seconds to do
reset). We could change allocation from GFP_ATOMIC to GFP_KERNEL for
SERIO_REGISTER_DRIVER events to make it more robust but otherwise I'd
leave serio_register_driver return void. You could also add a flag to
serio driver indicating whether registration is complete and check
that flag in serio_unregister_driver so it does not do stupid things.

--
Dmitry
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/