Issue with the emulated PCI bridge implementation

[arnd@arndb.de (Arnd Bergmann)]

On Friday 03 January 2014 11:44:24 Jason Gunthorpe wrote:
> On Fri, Jan 03, 2014 at 01:22:31PM +0100, Arnd Bergmann wrote:
> 
> > However the part that made me wonder is that an e1000e with a PCI bridge
> > actually /should/not/ need to allocate an I/O space window with a precious
> > mbus resource, since AFAIK this adapter does not have an I/O space BARs.
> 
> IIRC the e1000 still has a legacy IO port BAR.. (guessing it is used
> for PXE boot on x86?)

Ok, that would certainly explain it. I actually found a reference to the
I/O BAR in the e1000 driver now.

> My patch set to allow the DT to turn off IO port allocation is already
> in mainline - drop the IO ranges from the DT and no IO resources are
> consumed at all.

Ah, nice! I haven't followed that, but it sounds like a very good
solution.

> Though, that is an interesting point, a small refinement would be to
> not allocate the pci io map if IO is turned off as well..

Do you mean not call pci_ioremap_io() and pci_add_resource_offset(), or
something else? Either way, it sounds like the correct thing to do.
I wonder if the PCI core has any problems when there is no IO resource,
but it's probably fine.

	Arnd