Re: [PATCH qemu v23] spapr: Fix implementation of Open Firmware client interface

[Alexey Kardashevskiy <aik@ozlabs.ru>]

On 09/07/2021 23:28, BALATON Zoltan wrote:
> On Fri, 9 Jul 2021, Alexey Kardashevskiy wrote:
>> On 09/07/2021 08:34, BALATON Zoltan wrote:
>>> MorphOS still boots but this breaks Linux which changes a few things 
>>> in the device tree to fix it up to make it look the way it thinks is 
>>> better.
>>
>>
>> What are those things? What does the change break precisely? Does the 
>> kernel stop booting?
>> Can you please send output with the trace_vof_setprop tracepoint enabled?
> 
> It's fixing up some props that on Pegasos2 firmware are not how Linux 
> expects them.

Why does it need to fix them then? You are building the FDT in QEMU, 
built it in the way Linux like and then you do not depend on the kernel 
fixing them up. What do I miss?

 From traces I see that (besides PCI) it mostly sets props for 
linux-initrd/bootargs which you rather need to handle to keep the 
machine's properties and the FDT in sync.


> Without this it's not booting. Attached is the trace 
> output with VOF v23 as it is now (nosetprop) and another one after the 
> patch that adds setprop callback to pegasos2 I'm sending separately. 
> That patch restores Linux boot but I still think all this boilerplate 
> would not be needed if we kept the default to allow setprop and that 
> results in overall simpler code. If something breaks becuase of enabling 
> setprop by default (which normally works on real Open Firmware) it's 
> easy enough to debug by enabling vof_setprop trace points so I don't see 
> this adds any value other than making board code more complex.



-- 
Alexey