Get superio serial to work with 2.6

This patch is needed to get serial console to work on SuckyIO based systems on 2.6. It was mostly done by Grant (and willy, I think)

Unfortunately it requires disabling some bits of code in inventory. What do you guys think?

Can someone try this on a suckyio system that has USB running?

randolph

Index: arch/parisc/kernel/inventory.c
===================================================================
RCS file: /var/cvs/linux-2.6/arch/parisc/kernel/inventory.c,v retrieving revision 1.2

diff -u -p -r1.2 inventory.c
@@ -532,7 +532,12 @@ static void __init system_map_inventory(
 	 * Otherwise the machine might crash during iommu setup.
 	 */
 	pdc_io_reset();
-	pdc_io_reset_devices();
+
+	/*
+	 * Unfortunately if we reset devices here, serial console
+	 * stops working :-(
+	 */
+	/* pdc_io_reset_devices(); */
 #endif
 
 	for (i = 0; status != PDC_BAD_PROC && status != PDC_NE_MOD; i++) {

Index: drivers/parisc/superio.c
===================================================================
RCS file: /var/cvs/linux-2.6/drivers/parisc/superio.c,v retrieving revision 1.3

My preference right now is to move those PDC calls into sba_cdio.c. Then add hueristics that call PDC_IO_RESET_AND_CLEAR if suckyio/USB is active or the LAN active. If suckyio serial console, then add code to save/restore suckyio config space around the reset.

So quite a bit of hackery that's really only needed when the IO MMU is actually used.

hrm...I don't have the serial console setup at the moment. I suggest plugging in a USB mouse and moving the mouse around during boot. Not sure that's sufficient but maybe.

The USB/DMA problem occurs when booting with USB Keyboard/mouse and not a serial console. Ie, until 2.6.0 boots with STICON enabled, I can't really test the USB/DMA stuff.

grant

hmmm.. that looks like a hpux name ;-) sba_iommu.c?

ok, so should i just keep this patch in my private tree for now? or commit it and let you fix the rest of it? ;-)

i doubt that will make a difference. if the usb driver is not yet loaded (i.e. during inventory) i don't thing any usb events will be happening, right? anyway, i guess i can try this tomorrow.

randolph

Randolph Chung
Debian GNU/Linux Developer, hppa/ia64 ports
http://www.tausq.org/

yeah...brainfart. I was up too late last night again...

commit and I'll fix the rest later.

It does make a difference. The IODC left HW in an enabled and operational state when it handed off to the OS. That's harmless for a serial console. It's not harmless when the device can master DMA transactions and we reprogram the IO MMU.

grant