via82xx.c vs. sonypi.c i/o region conflict on vaio

via82xx.c vs. sonypi.c i/o region conflict on vaio

[Greg Ingram]

Howdy,

(I sent a similar version of this message to Jaroslav Kysela
<perex@suse.cz> but haven't heard back from him.)

I'm working with a Sony Vaio PCG-FXA32.  The "Sony Programmable I/O
Controller Driver v1.22"  (drivers/char/sonypi.c). finds a controller and
grabs ports 0x1080-0x109f.  The sound module (sound/pci/via82xx.c) finds a
VIA686A and grabs ports 0x1000-0x10ff.  The ranges overlap. I can't load
both modules.

I modified the sound driver to grab only 128 ports instead of 256 and the
driver works fine on this hardware.  In 2.6.7, it's line 2049 or so of
sound/pci/via82xx.c:

old:	if ((chip->res_port = request_region(chip->port, 256, card->driver)) == NULL) {
new:	if ((chip->res_port = request_region(chip->port, 256, card->driver)) == NULL) {

Does anyone know specifically that some chipsets need so many ports?  If
so, what info for this hardware can I supply to fix the sound driver
permanently?

Regards,

- Greg

Re: via82xx.c vs. sonypi.c i/o region conflict on vaio

[Jeff Garzik]

Greg Ingram wrote:
> I modified the sound driver to grab only 128 ports instead of 256 and the
> driver works fine on this hardware.  In 2.6.7, it's line 2049 or so of
> sound/pci/via82xx.c:
> 
> old:	if ((chip->res_port = request_region(chip->port, 256, card->driver)) == NULL) {
> new:	if ((chip->res_port = request_region(chip->port, 256, card->driver)) == NULL) {


I don't see any difference between these two lines.

Regardless, I see two bugs:

1) Hardcoding 256 for resource size.  Should be using pci_resource_len()

2) via82xx sound driver should be using pci_request_regions() and 
pci_release_regions(), not request_region.  Doing this eliminates issue #1.

	Jeff

Re: via82xx.c vs. sonypi.c i/o region conflict on vaio

[Greg Ingram]

On Wed, 14 Jul 2004, Jeff Garzik wrote:

> Greg Ingram wrote:
> > I modified the sound driver to grab only 128 ports instead of 256 and the
> > driver works fine on this hardware.  In 2.6.7, it's line 2049 or so of
> > sound/pci/via82xx.c:
> >
> > old:	if ((chip->res_port = request_region(chip->port, 256, card->driver)) == NULL) {
> > new:	if ((chip->res_port = request_region(chip->port, 256, card->driver)) == NULL) {
>
>
> I don't see any difference between these two lines.

Doh!  In new I replaced '256' with '128'.

> Regardless, I see two bugs:
>
> 1) Hardcoding 256 for resource size.  Should be using pci_resource_len()
>
> 2) via82xx sound driver should be using pci_request_regions() and
> pci_release_regions(), not request_region.  Doing this eliminates issue #1.

Is the fix in?  Or is there something I need to do?

Incidentally, on the notebook, cutting the region grabbed by the audio
driver also eliminated a nifty hang-the-machine issue if I pressed Fn-F7
trying to shove video out the external display port.

Regards,

- Greg

Re: via82xx.c vs. sonypi.c i/o region conflict on vaio

[Takashi Iwai]

At Wed, 14 Jul 2004 15:31:44 -0400,
Jeff Garzik wrote:
> 
> Greg Ingram wrote:
> > I modified the sound driver to grab only 128 ports instead of 256 and the
> > driver works fine on this hardware.  In 2.6.7, it's line 2049 or so of
> > sound/pci/via82xx.c:
> > 
> > old:	if ((chip->res_port = request_region(chip->port, 256, card->driver)) == NULL) {
> > new:	if ((chip->res_port = request_region(chip->port, 256, card->driver)) == NULL) {
> 
> 
> I don't see any difference between these two lines.
> 
> Regardless, I see two bugs:
> 
> 1) Hardcoding 256 for resource size.  Should be using pci_resource_len()
> 
> 2) via82xx sound driver should be using pci_request_regions() and 
> pci_release_regions(), not request_region.  Doing this eliminates issue #1.

Agreed, it's much simpler.
I'll fix it (over all ALSA PCI drivers).


Takashi