IO ports assign

[YhLu]

Eric,

I have tested s2880 with all addon cards, and found interesting problem.

When I plug Adaptec 29320 PCI X scsi card into slot 3 and 4. That share
BUS_A of 8131 with LSI SCSI. Linux bios got IO port allocation problem on
bus 1.

BUS 1: 1000-2fff
LSI will use 3000 later...

Regards

Yinghai Lu

Allocating resources...
ASSIGN RESOURCES, bus 0
PCI: 00:01.0 1c <- [0x00001000 - 0x00002fff] bus 1 io
PCI: 00:01.0 24 <- [0xfe300000 - 0xfe2fffff] bus 1 prefmem
PCI: 00:01.0 20 <- [0xfe100000 - 0xfe1fffff] bus 1 mem
ASSIGN RESOURCES, bus 1
PCI: 01:07.0 10 <- [0x00001000 - 0x000010ff] io
PCI: 01:07.0 14 <- [0xfe180000 - 0xfe181fff] mem
PCI: 01:07.0 1c <- [0x00001400 - 0x000014ff] io
PCI: 01:07.1 10 <- [0x00001800 - 0x000018ff] io
PCI: 01:07.1 14 <- [0xfe182000 - 0xfe183fff] mem
PCI: 01:07.1 1c <- [0x00001c00 - 0x00001cff] io
PCI: 01:08.0 10 <- [0x00002000 - 0x000020ff] io
PCI: 01:08.0 14 <- [0xfe184000 - 0xfe185fff] mem
PCI: 01:08.0 1c <- [0x00002400 - 0x000024ff] io
PCI: 01:08.1 10 <- [0x00002800 - 0x000028ff] io
PCI: 01:08.1 14 <- [0xfe186000 - 0xfe187fff] mem
PCI: 01:08.1 1c <- [0x00002c00 - 0x00002cff] io
PCI: 01:09.0 10 <- [0xfe100000 - 0xfe10ffff] mem
PCI: 01:09.0 18 <- [0xfe110000 - 0xfe11ffff] mem
PCI: 01:09.1 10 <- [0xfe120000 - 0xfe12ffff] mem
PCI: 01:09.1 18 <- [0xfe130000 - 0xfe13ffff] mem
PCI: 01:0a.0 10 <- [0x00003000 - 0x000030ff] io
PCI: 01:0a.0 14 <- [0xfe140000 - 0xfe14ffff] mem
PCI: 01:0a.0 1c <- [0xfe150000 - 0xfe15ffff] mem
PCI: 01:0a.1 10 <- [0x00003400 - 0x000034ff] io
PCI: 01:0a.1 14 <- [0xfe160000 - 0xfe16ffff] mem
PCI: 01:0a.1 1c <- [0xfe170000 - 0xfe17ffff] mem
ASSIGNED RESOURCES, bus 1

YhLu  writes:
> Eric,
>
> I have tested s2880 with all addon cards, and found interesting problem.
>
> When I plug Adaptec 29320 PCI X scsi card into slot 3 and 4. That share
> BUS_A of 8131 with LSI SCSI. Linux bios got IO port allocation problem on
> bus 1.
>
> BUS 1: 1000-2fff
> LSI will use 3000 later...

I believe this is a bug that failed to get pushed forward
from freebios tree.  I have just committed this fix and
a patch is attached.

The problem is because it was performing it's size calculations
at address 0.  It figured the I/O space registers could be safely
packed more tightly then they actually can be.

Eric

bash-2.05a$ cvs diff -u src/devices/device.c
[e-mail]'s password: 
Index: src/devices/device.c
===================================================================
RCS file: /cvsroot/freebios/freebios2/src/devices/device.c,v
retrieving revision 1.4
diff -u -r1.4 device.c
--- src/devices/device.c	21 Jul 2003 20:13:42 -0000	1.4
+++ src/devices/device.c	4 Aug 2003 19:52:30 -0000
@@ -217,8 +217,12 @@
    	if (resource->flags & IORESOURCE_IO) {
    		/* Don't allow potential aliases over the
    		 * legacy pci expansion card addresses.
+			 * The legacy pci decodes only 10 bits,
+			 * uses 100h - 3ffh. Therefor, only 0 - ff
+			 * can be used out of each 400h block of io
+			 * space.
    		 */
-			if ((base > 0x3ff) && ((base & 0x300) != 0)) {
+			if ((base & 0x300) != 0) {
    			base = (base & ~0x3ff) + 0x400;
    		}
    		/* Don't allow allocations in the VGA IO range.