PCI: use ACPI to verify extended config space on x86
In-Reply-To: <[email protected]>
Quoted 11 lines · Arjan van de Ven
Quoted 3 lines
Extend the verification for PCI-X/PCI-Express extended config space pointer. Checks whether the MCFG address range is listed as a motherboard resource, per the PCI firmware spec.
I'm still not quite happy about this; the entire point of the check is that we CAN'T trust the ACPI implementation, and want a second opinion. This patch basically asks the ACPI implementation if we can trust the ACPI implementation. I'm not sure that's a good idea.
And I understood that most issues went away with the more relaxed check that is in gregkh's tree already (if not in mainline, I should check that).
The more-relaxed check is in mainline. I wrote it, but it didn't even fix the problem on my own machine. This did.
According to Rajesh, the spec doesn't require the MCFG space to be e820-reserved, so that's not really a valid check.
Signature · 2 lines
Chuck "You can't read a newspaper if you can't read." --George W. Bush
Quoted 16 lines · Chuck Ebbert
In-Reply-To: <[email protected]>
Quoted 11 lines · Arjan van de Ven
Quoted 3 lines
Extend the verification for PCI-X/PCI-Express extended config space pointer. Checks whether the MCFG address range is listed as a motherboard resource, per the PCI firmware spec.
I'm still not quite happy about this; the entire point of the check is that we CAN'T trust the ACPI implementation, and want a second opinion. This patch basically asks the ACPI implementation if we can trust the ACPI implementation. I'm not sure that's a good idea.
And I understood that most issues went away with the more relaxed check that is in gregkh's tree already (if not in mainline, I should check that).
Quoted 2 lines
The more-relaxed check is in mainline. I wrote it, but it didn't even fix the problem on my own machine.
Why did you submit it then when it didn't work?
Quoted 4 lines
This did.
According to Rajesh, the spec doesn't require the MCFG space to be e820-reserved, so that's not really a valid check.
Anyways Rajesh's patch is probably the way to go. If the ACPI implementatin is self consistent it can be probably trusted.
The e820 check was just a heuristic and it clearly wasn't a good one.
-Andi