Re: Bug report: the extended PCI config space is missed with 6.2-rc2

[Bjorn Helgaas]

On Thu, Jan 05, 2023 at 06:47:44PM -0600, Bjorn Helgaas wrote:
> On Fri, Jan 06, 2023 at 12:22:09AM +0000, Luck, Tony wrote:
> > > ...and Dave, who reported that CXL enumeration was busted in -rc2, says
> > > this patch fixes that. So you can also add:
> > >
> > > Tested-by: Dave Jiang <dave.jiang@xxxxxxxxx>
> > 
> > Also seems good for my Broadwell/EDAC system.
> > 
> > Boot messages mentioning MMCONFIG are:
> > 
> > $ dmesg | grep MMCONFIG
> > PCI: MMCONFIG for domain 0000 [bus 00-ff] at [mem 0x80000000-0x8fffffff] (base 0x80000000)
> > PCI: not using MMCONFIG
> > PCI: MMCONFIG for domain 0000 [bus 00-ff] at [mem 0x80000000-0x8fffffff] (base 0x80000000)
> > PCI: MMCONFIG at [mem 0x80000000-0x8fffffff] reserved in ACPI motherboard resources or EFI
> 
> This part looks ok.
> 
> > PCI: MMCONFIG for 0000 [bus00-7f] at [mem 0x80000000-0x87ffffff] (base 0x80000000) (size reduced!)
> > acpi PNP0A03:00: fail to add MMCONFIG information, can't access extended configuration space under this bridge
> > acpi PNP0A03:01: fail to add MMCONFIG information, can't access extended configuration space under this bridge
> > acpi PNP0A08:02: fail to add MMCONFIG information, can't access extended configuration space under this bridge
> > acpi PNP0A08:03: fail to add MMCONFIG information, can't access extended configuration space under this bridge
> 
> But the rest of this still looks like a regression.  From your
> previous dmesg log:
> 
>   PCI: MMCONFIG for domain 0000 [bus 00-ff] at [mem 0x80000000-0x8fffffff] (base 0x80000000)
>   PNP0A03:00: host bridge to domain 0000 [bus ff]
>   PNP0A03:01: host bridge to domain 0000 [bus bf]
>   PNP0A03:02: host bridge to domain 0000 [bus 7f]
>   PNP0A03:03: host bridge to domain 0000 [bus 3f]
>   PNP0A08:00: host bridge to domain 0000 [bus 00-3e]
>   PNP0A08:01: host bridge to domain 0000 [bus 40-7e]
>   PNP0A08:02: host bridge to domain 0000 [bus 80-be]
>   PNP0A08:03: host bridge to domain 0000 [bus c0-fe]
> 
> That MMCONFIG space should cover all those buses, but something is
> going wrong.

Tony, would you mind collecting a dmesg log with "efi=debug"?  I want
to see the EFI_MEMORY_MAPPED_IO size and what we remove from E820.

Bjorn