RPi4B: modern firmware vs. Device tree loaded to 0x4000 (size 0xbe0c) [fails] vs. to 0x1f0000 (size 0xbd90) [works]?

[Klaus Cucinauomo]

> Am 11.10.2020 um 08:31 schrieb Mark Millard <marklmi at yahoo.com>:
> 
> starting USB...
> Bus xhci_pci: probe failed, error -110
> No working controllers found
> 
> So (B) would seem to be a u-boot 2020.10 issue, having
> nothing specifically to do with FreeBSD or with
> rpi3-psci-monitor use.


It’s an xhci-issue in the DeviceTree in the 8GB hardware-design(there’s no more dedicated chip for it).
The VL805-firmware has to be loaded explicitly ; requires  an early xhci-reset :
Again :
https://patchwork.ozlabs.org/project/linux-pci/patch/20200629161845.6021-4-nsaenzjulienne@suse.de/
https://patchwork.ozlabs.org/project/linux-pci/patch/20200629161845.6021-5-nsaenzjulienne@suse.de/
afaik  since these patches are not yet upstreamed we have to patch by ourselves.


> Am 11.10.2020 um 11:13 schrieb Robert Crowston <crowston at protonmail.com>:
> 
> We have a patched version of armstub8-gic. We use it to install some code in a highly privileged context that we later call from supervisor mode to start the secondary CPUs. Linux uses a different approach to spin up the other CPUs.

That’s the problem, armstub8-gic should have been removed when using modern firmware because it no longer depends on it.

> Am 11.10.2020 um 11:13 schrieb Robert Crowston <crowston at protonmail.com>:
> 
> I haven’t followed this thread in huge detail, but it seems like the raspberry pi people changed everything again


Yes, the RPI was built to Rob ( I mean to Rob your weekend free time :-) Ha Ha



> Am 11.10.2020 um 11:39 schrieb Robert Crowston <crowston at protonmail.com>:
> 
> In our patched version of armstub8-gic, we attempt to modify the device tree at runtime to add PSCI nodes……...
> 
> — RHC.

… which results in panic after removing/unloading   armstub8-gic  :

https://dmesgd.nycbug.org/index.cgi?do=view&id=5702


Regards
K.