Re: [Bug #13780] NULL pointer dereference loading powernowk8

[Kurt Roeckx <kurt-burXGKnpAKGzQB+pC5nmwQ@public.gmane.org>]

On Tue, Sep 15, 2009 at 04:45:37PM -0700, David Rientjes wrote:
> On Thu, 10 Sep 2009, Rafael J. Wysocki wrote:
> 
> > > So I did a biset, and this is the result:
> > > $ git bisect good
> > > 0e64a0c982c06a6b8f5e2a7f29eb108fdf257b2f is first bad commit
> 
> 0e64a0c has this cleanup, which isn't equivalent to the previous code:

Please look at the patch I attached to the bug report.  There is
other code that is not equivalent.  It's in comment #17:
http://bugzilla.kernel.org/show_bug.cgi?id=13780#c17

And the patch itself:
http://bugzilla.kernel.org/attachment.cgi?id=23048

This patch fixes my problem.

> @@ -779,11 +834,12 @@ static int powernow_k8_cpu_init_acpi(struct powernow_k8_data *data)
>  		goto err_out;
>  	}
>  
> -	if ((data->acpi_data.control_register.space_id != ACPI_ADR_SPACE_FIXED_HARDWARE) ||
> -		(data->acpi_data.status_register.space_id != ACPI_ADR_SPACE_FIXED_HARDWARE)) {
> +	space_id = data->acpi_data.control_register.space_id;
> +	if ((space_id != ACPI_ADR_SPACE_FIXED_HARDWARE) ||
> +		(space_id != ACPI_ADR_SPACE_FIXED_HARDWARE)) {
>  		dprintk("Invalid control/status registers (%x - %x)\n",
>  			data->acpi_data.control_register.space_id,
> -			data->acpi_data.status_register.space_id);
> +			space_id);
>  		goto err_out;
>  	}
>  
> This got fixed in 2c701b1 in 2.6.31-rc8, so this is most likely the 
> problem that Kurt was encountering.
> 
> Kurt, could you try 2.6.31 and see if you still experience the issue?  If 
> not, this can be closed.

I also noticed this difference (comment #18), patched it, and
tried it.  This is not the issue.  Also note that I tested the
2.6.31-rc9 kernel which still had the problem.

The bug report might be a little confusing since after comment #18
there are older mails pasted in the bug report.


Kurt