On 9/2/05, Daniel Drake [off-list ref] wrote:

Bartlomiej Zolnierkiewicz wrote:

quoted

Same thing as with VT6420 support:

I'm still concerned about VIA IDE chipset + VT6410 combo
(AFAIR I've also seen VT6410 on PCI add-on card but I can be wrong).

via82cxxx.c needs to be fixed to support multiple controllers first.

Hows this? I don't have any hardware with two VIA controllers, however I have
tested this on a pc which has a single vt8233a controller.

---

Support multiple controllers in the via82cxxx IDE driver

Signed-off-by: Daniel Drake <redacted>

--- linux/drivers/ide/pci/via82cxxx.c.orig	2005-08-31 01:32:05.000000000 +0100
+++ linux/drivers/ide/pci/via82cxxx.c	2005-09-02 01:16:59.000000000 +0100

@@ -101,11 +101,19 @@ static struct via_isa_bridge {
 	{ NULL }
 };

-static struct via_isa_bridge *via_config;
-static unsigned int via_80w;
-static unsigned int via_clock;
 static char *via_dma[] = { "MWDMA16", "UDMA33", "UDMA66", "UDMA100",

"UDMA133" };

I would really prefer not to add per host struct via82xxx_dev,
(making it per hwif and doing extra match in ->init_hwif() is acceptable).

+struct via82cxxx_dev
+{
+	struct pci_dev *pci_dev, *isa_dev;

pci_dev is needed only for /proc/via and I would prefer /proc/via
to vanish because it complicates driver needlessly (could you do
this in separate patch?).

isa_dev has no relevance for vt6410 and won't be needed if
/proc/via goes away

+	struct via_isa_bridge *via_config;

Please instead add via_config_find() which would
find proper via_config given PCI ID.

+	unsigned int via_clock;

Global via_clock is OK as IDE core doesn't
support per bus PCI clocks anyway.

+	unsigned int via_80w;

Cable detection code should be moved to separate function
and be called from ->init_hwif() (required for future hotplug support).

Otherwise patch looks fine.

Thanks and sorry for the delay,
Bartlomiej