Re: SPARC64: getting "no compatible bridge window" errors :/

[Bjorn Helgaas]

On Mon, Sep 26, 2022 at 12:11:06PM -0500, Bjorn Helgaas wrote:
> [+cc Alex, David, sparclinux, LKML]
> 
> On Sun, Sep 25, 2022 at 06:59:23PM +0200, Richard Rogalski wrote:
> > I hope this is the right place for this.
> 
> This is great, thanks a lot for your report!  Is this a regression?
> If so, what's the most recent kernel that worked?  
> 
> > In my dmesg output, I get things like:
> > 
> > pci 0000:04:00.0: can't claim VGA legacy [mem 0x000a0000-0x000bffff]: no compatible bridge window
> > pci 0000:06:00.0: can't claim VGA legacy [mem 0x000a0000-0x000bffff]: no compatible bridge window
> > pci 0000:06:00.1: can't claim BAR 0 [mem 0x84110200000-0x84110203fff 64bit]: no compatible bridge window
> > 
> > I opened a bug for amdgpu [here](https://gitlab.freedesktop.org/drm/amd/-/issues/2169) but looking further into it I think it is caused by deeper PCIe problems :\
> > 
> > https://gitlab.freedesktop.org/drm/amd/uploads/cbf47807972c8a990bb2a8cdbb39ad9e/8C7CA9QNG dmesg log
> > https://gitlab.freedesktop.org/drm/amd/uploads/6a799425dea50febd82f8bc11e54433a/ll.txt lspci -vv
> > https://gitlab.freedesktop.org/drm/amd/uploads/7d4a794b1f7d67a1ffcdee5dfdec3ad6/config.txt kernel .config
> 
> Your error output attachment [1] contains an address that looks like
> it's in 06:00.0 BAR 5:
> 
>   pci 0000:06:00.0: reg 0x24: [mem 0x84001200000-0x8400123ffff]
>   NON-RESUMABLE ERROR: insn effective address [0x0000084001201410]
> 
> This looks like an amdgpu issue.  There have been recent changes like
> c1c39032a074 ("drm/amdgpu: make sure to init common IP before gmc")
> and dd6aeb4e5f59 ("drm/amdgpu: Don't enable LTR if not supported")
> that could be related.
> 
> The PCI "no compatible bridge window" warnings are definitely an
> issue, but I don't think they're related to the amdgpu crash:
> 
>   pci@400: PCI MEM64 [mem 0x84100000000-0x84dffffffff] offset 80000000000
>   pci_bus 0000:00: root bus resource [mem 0x84100000000-0x84dffffffff] (bus address [0x4100000000-0x4dffffffff])
>   pci 0000:09:00.0: can't claim BAR 0 [mem 0x84120000000-0x8412007ffff 64bit]: no compatible bridge window
> 
> Those and this from lspci:
> 
>   0000:01:00.0 bridge to [bus 02-09] window [mem 0x4100000000-0x412fffffff pref]
>   0000:02:0c.0 bridge to [bus 09]    window [mem 0x4120000000-0x412fffffff pref]
>   0000:09:00.0 Intel 82599ES NIC Region 0: Memory at 0x84120000000
> 
> are telling us there's something wrong with how the resource-to-bus
> offset is being applied.  It looks like the offset was applied to the
> NIC BAR, but didn't get applied to the bridge windows.
> 
> Could you start a new thread here (linux-kernel@xxxxxxxxxxxxxxx,
> linux-pci@xxxxxxxxxxxxxxx, and sparclinux@xxxxxxxxxxxxxxx) for this
> issue and attach the dmesg log when booting with "ofpci_debug=1"?

Any chance you could collect a dmesg log with "ofpci_debug=1"?

I'd like to look at the resource-to-bus offset issue.

> Do the devices we complain about (NICs and storage HBAs 09:00.0,
> 09:00.1, 0d:00.0, 0d:00.1, 0e:00.0, 0f:00.0, 0001:03:00.0,
> 0001:03:00.1, 0001:0:00.0, 0001:0a:00.1) work?
> 
> Bjorn
> 
> [1] https://gitlab.freedesktop.org/drm/amd/uploads/b51f4d6783eeebf90de9a400525d07d6/qq