[commit] r5581 - trunk/src/southbridge/amd/cs5536

[repository service]

Author: stuge
Date: Sun May 23 06:50:41 2010
New Revision: 5581
URL: https://tracker.coreboot.org/trac/coreboot/changeset/5581

Log:
Fix bug from r5476 re CS5536 device search during GeodeLX PCI domain enable

cs5536.c:chipsetinit() is called during northbridge pci_domain_enable()
which happens before scan_bus() so the device tree does not have PCI
vendor/device ids yet. Let's use dev_find_slot() for now. This works
only as long as the CS5536 has PCI device id 0xf in all mainboards,
and a better solution is needed in case that ever changes!

Signed-off-by: Peter Stuge 
Acked-by: Nathan Williams

Modified:
   trunk/src/southbridge/amd/cs5536/cs5536.c

Modified: trunk/src/southbridge/amd/cs5536/cs5536.c
==============================================================================
--- trunk/src/southbridge/amd/cs5536/cs5536.c	Sat May 22 17:07:15 2010	(r5580)
+++ trunk/src/southbridge/amd/cs5536/cs5536.c	Sun May 23 06:50:41 2010	(r5581)
@@ -530,8 +530,7 @@
    struct southbridge_amd_cs5536_config *sb;
    struct msrinit *csi;

-	dev = dev_find_device(PCI_VENDOR_ID_AMD,
-			PCI_DEVICE_ID_AMD_CS5536_ISA, 0);
+	dev = dev_find_slot(0, PCI_DEVFN(0xf, 0));

if (!dev) {
    	printk(BIOS_ERR, "CS5536 not found.\n");

[Myles Watson]

> -----Original Message-----
> From: [e-mail] [mailto:[e-mail]]
> On Behalf Of repository service
> Sent: Saturday, May 22, 2010 10:51 PM
> To: [e-mail]
> Subject: [coreboot] [commit] r5581 - trunk/src/southbridge/amd/cs5536
>
> Author: stuge
> Date: Sun May 23 06:50:41 2010
> New Revision: 5581
> URL: https://tracker.coreboot.org/trac/coreboot/changeset/5581
>
> Log:
> Fix bug from r5476 re CS5536 device search during GeodeLX PCI domain
> enable
>
> cs5536.c:chipsetinit() is called during northbridge pci_domain_enable()
> which happens before scan_bus() so the device tree does not have PCI
> vendor/device ids yet. Let's use dev_find_slot() for now. This works
> only as long as the CS5536 has PCI device id 0xf in all mainboards,
> and a better solution is needed in case that ever changes!

It seems like you could check the device ID of 0xf once you get it, so that
someone with a board like that would know why it's failing.
> dev = dev_find_device(PCI_VENDOR_ID_AMD,
> 	PCI_DEVICE_ID_AMD_CS5536_ISA, 0);
> dev = dev_find_slot(0, PCI_DEVFN(0xf, 0));

if (pci_read_32(PCI_DEV(0,0xf,0), PCI_VENDOR_ID !=
        ((PCI_DEVICE_ID_AMD_CS5536_ISA << 16) |
         (PCI_VENDOR_ID_AMD))) {
    	printk(BIOS_ERR, "CS5536 not at 0xf.\n");
...

Thanks,
Myles