Re: [RFC] i386: per-cpu mmconfig map

From: OGAWA Hirofumi
Date: Sun Dec 24 2006 - 12:47:07 EST


Arjan van de Ven <arjan@xxxxxxxxxxxxx> writes:

>> [ Alternatively, we could make a per-cpu mapping area or something. Not
>> that it's probably worth it, but if we wanted to avoid all locking and
>> instead just disable preemption, that would be the way to go. --Linus ]
>> This patch is a draft of Linus's suggestion. This seems work for me.
>> And this removes pci_config_lock in mmconfig access path, I think we
>> don't need lock on this path, although we need to disable IRQ.
>> Comment?
> I like the idea and the implementation, I have just one concern:
> Does your schema still work if the Non-Maskable-Interrupt code uses
> config space? It may do that I suspect to deal with ECC memory errors ;(

I didn't notice the problem of NMI at all. However if NMI path use it,
pci_config_lock would be deadlock already. So I think NMI can't access
to PCI config... Hm...
OGAWA Hirofumi <hirofumi@xxxxxxxxxxxxxxxxxx>
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at
Please read the FAQ at