I've stumbled upon a problem trying to get a NVIDIA 7300 SE together with my K9VGM mobo, a problem "pci=nommconf" or other grub parameter could not solve with a 2.6.27-11 kernel.
The symptom was a flow of "+------ PCI-Express Device Error ------+" at boot time when plugging the video card in:
+------ PCI-Express Device Error ------+ Error Severity^I^I: Uncorrected (Fatal) PCIE Bus Error type^I: Transaction Layer Malformed TLP ^I: First Receiver ID^I^I: 0010 VendorID=1106h, DeviceID=a238h, Bus=00h, Device=02h, Function=00h TLB Header: 4a000001 02000004 00100f00 00000000 pcieport-driver 0000:00:02.0: broadcast error_detected message nvidia 0000:02:00.0: device has no AER-aware driver pcieport-driver 0000:00:02.0: Root Port link has been reset pcieport-driver 0000:00:02.0: broadcast mmio_enabled message pcieport-driver 0000:00:02.0: broadcast resume message pcieport-driver 0000:00:02.0: AER driver successfully recovered
That was not a NVIDIA driver issue. Despite the +------ PCI-Express Device Error ------+, some of which were "fatal", my machine was working fine for an hour or two, and then, would hopelessly freeze.
Digging, I found this thread on kerneltrap.org Why do so many machines need "noapic"?. The kernel developers said something was bloated within the kernel, something that seemed to remain broken in my 2.6.27-11 kernel (?).
I also meekly admit my lack of knowledge. I guess that it is a hardware incompatibility problem. That could be the meaning of the "device has no AER-aware driver" message. A pothole in the PCI bus in which my NVIDIA 7300 SE has fallen? May be it is solved with the very latest kernel version?