On Mon, 2007-12-10 at 17:32 -0700, Marc Jones wrote:
When I have seen a soft reset at this point it has been due to a memory problem. Try simpler configurations and try putting some basic memory test just before CAR is disabled. I would try on the interesting boundaries. For example, check where LinuxBIOS gets copied at 2MB and across a dimm boundary like 1GB+2MB.
ram_check(0x00200000, 0x00200000 + (640 * 1024)); ram_check(0x40200000, 0x40200000 + (640 * 1024));
Also, most of the memory registers are located in Function 2 if you decide to dump and compare them.
Thanks, so far I'm not seeing any anomalies with RAM. I've peeked and poked RAM at a number of boundaries with no apparent problems.
A clue I've encountered is that the device 00:01.1 is there in the PCI bus scan during CAR and absent after CAR at the time of the pci bus scan (pci_scan_bus). There is a "feature" in the southbridge regarding this I don't yet understand. I have to put my forensic glasses back on...
btw: based upon this information I've deduced that if a device is listed in the Config.lb but does not appear in the bus scan it is identified as a "Left over device." Is this correct?
I'm thinking that an experiment which might prove interesting is to temporarily change pci_scan_bus to not die when this device is "Left over..." and see where that gets me.
Steve