Re: Oops on ibmasm

[Greg KH <greg@kroah.com>]

On Wed, Mar 08, 2006 at 05:59:25PM -0500, Dave Jones wrote:
> On Thu, Mar 09, 2006 at 09:41:45AM +1100, Srihari Vijayaraghavan wrote:
>  > command count: 1
>  > input: ibmasm RSA I remote mouse as
>  > /class/input/input2
>  > input: ibmasm RSA I remote keyboard as
>  > /class/input/input3
>  > ibmasm remote responding to events on RSA card 0
>  > command count: 2
>  > ibmasm_exec_command:130 at 1141819512.780778
>  > do_exec_command:107 at 1141819512.780787
>  > respond to interrupt at 1141819512.782055
>  > exec_next_command:150 at 1141819512.782094
>  > finished interrupt at   1141819512.782103
>  > command count: 1
>  > Unable to handle kernel paging request at virtual
>  > address 6b6b6b6b
> 
> the problem is we do this..
> 
>     command_put(sp->current_command);
> 	exec_next_command(sp);
> 
> command_put drops the refcount of the kobject in sp,
> which then gets freed in the .release (free_command)
> which ends up kfree'ing 'sp'.
> 
> unsurprisingly, it goes bang the next time something
> tries to access it.  (Ie the 2nd line above)
> 
> It's totally busted.

I think there's a patch floating around somewhere to actually fix up
this driver so it works.  Max, any ideas about this?

thanks,

greg k-h