fixup for pci config_access on alchemy au1x000

Hi all,

I've encountered a serious problem with PCI config space access on Au1x000 platforms with recent 2.6.x-kernel. With 2.4.31 the same hardware works fine. So I was looking for the differences:

Symptoms:
- no PCI-device is seen on bootup though two or three cards are present
- lspci output is empty
- OR: lspci shows 20 times the same device
(- OR: in some slot-configurations it worked anyhow)

System(s):
1. platform with Au1500 and three PCI-devices (actually a mycable XXS1500 with backplane for three PCI-devices)
2. platform with Au1550 and two PCI-devices (custom board)

Debugging:
I digged down to the config_access() of the au1xxx-processors in arch/mips/pci/ops-au1000.c and switched on DEBUG.

The code of config_access() seems to be almost the same as of the 2.4.x-kernel. But the "pci_cfg_vm->addr" returned by get_vm_area(0x2000, 0) once on booting is different. That's of course not forbidden. But the alignment seems to be wrong. In my case, I received:

2.4.31: pci_cfg_vm->addr = c0000000
2.6.18-rc5: pci_cfg_vm->addr = c0101000

To make it short: With 2.6.x it fails on the first config-access with:
"PCI ERR detected: status 83a00356".

Fixup:
My fix is now, to use the VM_IOREMAP-flag in the get_vm_area call. This flag seems to be introduced in mm/vmalloc.c a long time ago (in 2.6.7-bk13, I found in gitweb).
Now, the returned address is pci_cfg_vm->addr = c0104000 and everything works fine.

What do you think about my fixup-patch?
Nobody's using the get_vm_area()-call without any flag ("0"). Was it only forgotten in the arch/mips/pci/ops-au1000.c?

Or am I completely wrong?

Thanks a lot for your comments!

Best regard,

Alexander

--- linux-2.6.18-rc5/arch/mips/pci/ops-au1000.c 2006-08-28 12:09:15.000000000
+0200
+++ linux-2.6.18-rc5-dev/arch/mips/pci/ops-au1000.c 2006-08-29
13:08:59.000000000 +0200
@@ -110,7 +110,7 @@
if (first_cfg) {
/* reserve a wired entry for pci config accesses */ first_cfg = 0;
- pci_cfg_vm = get_vm_area(0x2000, 0);
+ pci_cfg_vm = get_vm_area(0x2000, VM_IOREMAP);
if (!pci_cfg_vm)
panic (KERN_ERR "PCI unable to get vm area\n");
pci_cfg_wired_entry = read_c0_wired();

Alexander Bigga     Tel: +49 4873 90 10 866
mycable GmbH        Fax: +49 4873 901 976
Boeker Stieg 43
D-24613 Aukrug      eMail: [email protected]

The patch looks ok, so just the usual technicalities:

- Please include a Signed-off-by: line.
- Don't use a bloody crapmailer that garbles patches into some sort of ASCII spinach.

Ralf

Thanks for applying.

Ok. Will do this in future.

I'm really sorry. I don't like spinach too ;-) For future mailings and patches, I will take mutt again.

Alexander

Alexander Bigga     Tel: +49 4873 90 10 866
mycable GmbH        Fax: +49 4873 901 976
Boeker Stieg 43
D-24613 Aukrug      eMail: [email protected]

Hello.

You're correct -- this code was only working by some chance. Once you get
a virtual address not aligned to 8K, it breaks completely since the pages can't constitute a valid pair for wired entry anymore.
Actually, in 2.4 the situation seems to be even worse as get_vm_area()
there has no provision for the address alignment at all!

WBR, Sergei