Re: Deprecate the ppc405 boards in QEMU?

[Thomas Huth]

On 20/10/2021 16.31, BALATON Zoltan wrote:
> On Wed, 20 Oct 2021, LEROY Christophe wrote:
> > Le 20/10/2021 à 12:10, Cédric Le Goater a écrit :
> > > I dont understand how
> > >
> > >    static bd_t bd;
> > >
> > > can be updated in the kernel.
> >
> > It's not updated in the kernel.
> >
> > It is supposed to be provided by UBoot to Linux Kernel. But modern
> > kernels don't take that anymore, they take a device tree. For this
> > reason cuboot takes the content of bd to build/update the device tree.
> >
> > Looks like QEMU also provides the bd, see ref405ep_init()
> >
> > I managed to get a kernel booting with the following change (and with
> > CONFIG_ETHERNET removed)
> >
> > diff --git a/arch/powerpc/boot/cuboot-hotfoot.c
> > b/arch/powerpc/boot/cuboot-hotfoot.c
> > index 888a6b9bfead..63a9545ff55d 100644
> > --- a/arch/powerpc/boot/cuboot-hotfoot.c
> > +++ b/arch/powerpc/boot/cuboot-hotfoot.c
> > @@ -132,6 +132,12 @@ void platform_init(unsigned long r3, unsigned long
> > r4, unsigned long r5,
> >                    unsigned long r6, unsigned long r7)
> >  {
> >         CUBOOT_INIT();
> > +        bd.bi_intfreq = 133333333;
> > +        bd.bi_busfreq = 33333333;
> > +        bd.bi_procfreq = 133333333;
> > +        bd.bi_plb_busfreq = 33333333;
> > +        bd.bi_pci_busfreq = 33333333;
> > +        bd.bi_opbfreq = 33333333;
> >         platform_ops.fixups = hotfoot_fixups;
> >          platform_ops.exit = ibm40x_dbcr_reset;
> >         fdt_init(_dtb_start);
>
> So maybe taihu should also provide this boot info when linux_boot is true 
> (i.e. using -kernel) like the ref405ep does? Usually when using -kernel 
> without -bios then QEMU has to also emulate enough of what the firmware 
> would otherwise do like setting up devices and setting boot environment. Or 
> if we have both -bios and -kernel then maybe -kernel should tell the 
> firmware to boot a kernel but that may need a way to do that like setting 
> variables in nvram but we don't have models of that in taihu. This taihu 
> machine seems to be an early skeleton that wasn't finished, the ref405ep 
> seems to be more advanced.

I agree, looking code, the ref405ep board seems to be in a better shape than 
the taihu board. My u-boot image seems to run fine with both machines, so 
I'd suggest that we deprecate (and later remove) the taihu board, and keep 
the ref405ep board in QEMU if it is still helpful for Christophe (or anybody 
else).

 Thomas