[coreboot] pci_read_config8 crash in mainboard.c

[Carl-Daniel Hailfinger]

On 03.03.2009 17:37, Carl-Daniel Hailfinger wrote:
> On 02.03.2009 16:40, Carl-Daniel Hailfinger wrote:
>   
>> On 02.03.2009 16:33, ron minnich wrote:
>>   
>>     
>>> On Mon, Mar 2, 2009 at 7:30 AM, Carl-Daniel Hailfinger
>>> <c-d.hailfinger.devel.2006 at gmx.net> wrote:
>>>
>>>   
>>>     
>>>       
>>>> AFAIK it has been the case at least since AMD published their 690G/SB600
>>>> port, maybe even before that. For a fun time, grep over the tree for
>>>> pci_cf8_conf1 and you'll see it is not constrained to mainboard code.
>>>>     
>>>>       
>>>>         
>>> ok, has anyone checked to see if it's the code that determines type
>>> 1/type 2 access going wrong somehow?
>>>   
>>>     
>>>       
>> Maybe. pci_set_method() is not called on any K8 platform. AFAICS type 1
>> is hardcoded there, so pci_set_method should not be necessary in theory.
>> I don't know whether setting the access method happens early enough for
>> the mainboard code.
>>
>>
>>   
>>     
>>> If so, there's an easy fix :-)
>>>   
>>>     
>>>       
>> Hopefully. Once Ward boots tomorrow with my debug patch, we'll know
>> where exactly it hangs and probably also why.
>>   
>>     
>
> http://ward.vandewege.net/coreboot/m2a-vm/m2a-vm-with-fix-printk-format-warning-patch.log
>
> Ouch. We hang inside get_pbus.
>   


Ward, can you revert src/devices/pci_ops.c again and apply this patch?

Index: src/devices/pci_ops.c
===================================================================
--- src/devices/pci_ops.c	(Revision 3967)
+++ src/devices/pci_ops.c	(Arbeitskopie)
@@ -24,11 +24,19 @@
 #include <device/pci_ids.h>
 #include <device/pci_ops.h>
 
-static struct bus *get_pbus(device_t dev)
+struct bus *get_pbus(device_t dev)
 {
+	printk_spew("%s entered\n", __func__);
 	struct bus *pbus = dev->bus;
+	printk_spew("%s before loop, pbus=%p, pbus->dev=%p, pbus->dev->ops=%p, pbus->children=%p\n", __func__, pbus, pbus ? pbus->dev : 0xdeadbeef, (pbus && pbus->dev) ? pbus->dev->ops : 0xdeadbeef, pbus ? pbus->children: 0xdeadbeef);
 	while(pbus && pbus->dev && !ops_pci_bus(pbus)) {
+		printk_spew("%s inside loop begin, pbus=%p, pbus->dev=%p, pbus->dev->ops=%p, pbus->children=%p\n", __func__, pbus, pbus ? pbus->dev : 0xdeadbeef, (pbus && pbus->dev) ? pbus->dev->ops : 0xdeadbeef, pbus ? pbus->children: 0xdeadbeef);
+		if (pbus == pbus->dev->bus) {
+			printk_alert("%s stuck in endless loop for %s, breaking out\n", __func__, dev_path(dev));
+			break;
+		}
 		pbus = pbus->dev->bus;
+		printk_spew("%s inside loop end, pbus=%p, pbus->dev=%p, pbus->dev->ops=%p, pbus->children=%p\n", __func__, pbus, pbus ? pbus->dev : 0xdeadbeef, (pbus && pbus->dev) ? pbus->dev->ops : 0xdeadbeef, pbus ? pbus->children: 0xdeadbeef);
 	}
 	if (!pbus || !pbus->dev || !pbus->dev->ops || !pbus->dev->ops->ops_pci_bus) {
 		printk_alert("%s Cannot find pci bus operations", dev_path(dev));



Regards,
Carl-Daniel

-- 
http://www.hailfinger.org/