Panic in quirk_usb_early_handoff

[slash.tmp@free.fr (Mason)]

On 03/03/2017 20:02, Robin Murphy wrote:

> On 03/03/17 17:15, Mason wrote:
> 
>> [    1.264893] Unable to handle kernel paging request at virtual address d08664f4
> 
> Note that that's a reasonable approximation of a vmalloc address...
> 
> ...and that specifically it's r0 + r3...
> 
> ...and again. And always at the same PC, too. Looking at
> quirk_usb_early_handoff(), it mostly seems to go off poking bridge
> resources, so I'd hazard a guess that it's down to your host driver,
> with something uninitialised (or already freed) being used as an offset
> into some ioremapped resource (which given the consistency of r0 is
> probably allocated pretty early on).

When you say "host driver", do you mean the USB driver,
or the PCIe controller driver?

I am currently writing the PCIe controller driver, so I do
expect a large number of bugs there; but the USB driver is
just the generic XHCI driver. Although I now realize that
I wrote no DT node for the USB HW... Would that explain
the random weirdness?

> "addr2line -i -e vmlinux c039fe44", and work backwards from there ;) In
> particular I'd follow the provenance of r3.

I'll definitely take a closer look. Thanks for the disassembly.

Regards.