Re: [PATCH] efivars: prevent oops on unload when efi is not enabled

[Mike Waychison <mikew@google.com>]

On Fri, May 6, 2011 at 3:04 PM, Matt Domsch <Matt_Domsch@dell.com> wrote:
> On Fri, May 06, 2011 at 02:57:02PM -0500, Randy Dunlap wrote:
>> From: Randy Dunlap <randy.dunlap@oracle.com>
>>
>> unregister_efivars() should check for efi_enabled and not undo
>> allocations when efi is not enabled.  Otherwise there is an Oops
>> during module unload:
>
> Thanks Randy.
>
> I'm not sure why the driver is still loaded at all if efi_enabled==0.
>
> static int __init
> efivars_init(void)
> {
> ...
>        if (!efi_enabled)
>                return 0;
>
>
>
> That return 0 should be return -ENODEV I think.  Mike?

Ya, I changed efivars so that others could still use the
register_efivars/unregister_efivars interfaces when !efi_enabled.
They are used by the drivers/firmware/google/gsmi.c driver currently
pending in linux-next.

I think the right approach in this case would be to do

        if (efi_enabled) {
               unregister_efivars(&__efivars);
               kobject_put(efi_kobj);
        }

in efivars_exit().

>
> Thanks,
> Matt
>
> --
> Matt Domsch
> Technology Strategist
> Dell | Office of the CTO
>