Re: [PATCH 1/2] PCI: Split pci_assign_unassigned_resources to per root bus

[Benjamin Herrenschmidt <benh@kernel.crashing.org>]

On Mon, 2013-05-06 at 16:15 -0700, Yinghai Lu wrote:
> BenH reported that there is some assign unassigned resource problem
> in powerpc.
> 
> It turns out after
> | commit 0c5be0cb0edfe3b5c4b62eac68aa2aa15ec681af
> | Date:   Thu Feb 23 19:23:29 2012 -0800
> |
> |    PCI: Retry on IORESOURCE_IO type allocations
> 
> even the root bus does not have io port range, it will keep retrying
> to realloc with mmio.
> 
> Current retry logic is : try with must+optional at first, and if
> it fails will try must then try to extend must with optional.
> That will fail as mmio-non-pref and mmio-pref for bridge will
> be next to each other. So we have no chance to extend mmio-non-pref.

Shouldn't you do completely separate passes for IO and MMIO through the
whole lot instead since they are completely separate address spaces ?

IE. Even a legit setup with IOs, in case of failures on the IO side
(resource exhaustion, FW bugs, ...) we shouldn't retry the memory side
and vice-versa.

Note that I'm a bit worried by how invasive your proposed patch is. I
will let Gavin test it and if it works I'm happy but keep in mind that I
really need that fixed in 3.10 ... (especially since there are noises
around about a certain enterprise distro using that version...)

Cheers,
Ben.