rpi4-uefi.dev Re: RaspberryPi 4B 8G model not boot

[Klaus Küchemann]

> Am 30.05.2020 um 22:58 schrieb myfreeweb <greg at unrelenting.technology>:
> 
> 
> 
> On May 30, 2020 7:46:37 PM UTC, "Klaus Küchemann via freebsd-arm" <freebsd-arm at freebsd.org> wrote:
>> 
>> rpi4-uefi.dev has a stunning support-team(including at jmcneill of NetBSD and OpenBSD-hacker @kettenis )
>> and they are absolutely interested in FreeBSD and willing to help,
>> but 1st we have to repair generic_xhci_acpi …
>> looking closely into that file we will see that quite „nothing" is inside it :-)
> 
> Yeah, there shouldn't be anything else in generic_xhci_acpi and there's nothing wrong with it.

I was sure you will answer,, Hi and thanks for stepping in this discussion  :-), 
.. O.K., why do we do an acpi-split if there’s nothing else than PNP0D10 in it ? 
NetBSD & OpenBSD don’t do  that file-split , afaik

> 
> Looks like the problem is that the PCIe controller isn't getting initialized,
pcie isn`t exposed ( to the OS) in RPI4UEFI-dev , that’s what @AndrejWarkentin told us …
And he told us what to do (in general, while not the easiest to understand ;-)


> which is why xhci device memory is all 0xdead. The _INI or whatever ACPI method on that does the initialization probably isn't working correctly under FreeBSD. It does run, I have seen its debug message when I built an acpi_debug kernel. But it's not accomplishing its goal?? Maybe the memory regions it writes to aren't mapped correctly or something.
Yep, look here(you’ve seen that already) :
https://github.com/tianocore/edk2-platforms/blob/master/Platform/RaspberryPi/AcpiTables/Xhci.asl#L118
( attention : QWord )

> 
> Silly debug idea: log all memory access initiated by acpi in FreeBSD and NetBSD and compare :)
> 
> Less silly idea: can someone who's really really familiar with FreeBSD's acpica integration take a look at the RPi4 dsdt already?? Please??
see:
https://github.com/openbsd/src/blob/master/sys/dev/acpi/dsdt.c

--

cutoff of an OpenBSD-file :


xhci_acpi_parse_resources(int crsidx, union acpi_resource *crs, void *arg)
{
	struct xhci_acpi_softc *sc = arg;
	int type = AML_CRSTYPE(crs);

	switch (type) {
	case LR_MEM32FIXED:
		/* XHCI registers are specified by the first resource. */
		if (sc->sc_size == 0) {
			sc->sc_addr = crs->lr_m32fixed._bas;
			sc->sc_size = crs->lr_m32fixed._len;
		}
		break;
	case LR_QWORD:
		/* XHCI registers are specified by the first resource. */
		if (sc->sc_size == 0) {
			sc->sc_addr = crs->lr_qword._min;
			sc->sc_size = crs->lr_qword._len;
		}
		break;
	case LR_EXTIRQ:
		sc->sc_irq = crs->lr_extirq.irq[0];
		sc->sc_irq_flags = crs->lr_extirq.flags;
		break;
	}

	return 0;
}


!! the QWord-thing belongs to the RPI4 !!
—
to anticipate it: if we have solved this problem, we will probably end up on vfs_mountroot because the uSD driver is not recognized under acpi, afaik …

Regards