Trouble starting AP cpu's on all winner R40/V40 chip, help needed?

[Søren Schmidt]

> On 15 Oct 2019, at 14.27, Søren Schmidt <soren.schmidt at gmail.com> wrote:
> 
> Hi Manu
> 
>> 
>>> Hi all
>>> 
>>> I found some time to wrestle the allwinner R40/v40 I?ve sitting around, but I?m bashing my head against getting the AP?s booted properly.
>>> 
>>> I don?t seem to be able to get the CPU to run at all, or (more likely) I don?t have a way to get them startet at ?mpentry? to get them going correctly.
>>> 
>>> I?m not sure I understand how those AP?s gets their starting address, the only reference to ?mpentry? is aw_mp.c:
>>> 
>>> 
>>> bus_space_write_4(fdtbus_bs_tag, cpucfg, CPUCFG_P_REG0,
>>>           pmap_kextract((vm_offset_t)mpentry));
>>> 
>>> This is used on all all winner cpu?s (more or less) but according til the R40/v40 docs that register doesn?t exist which would explain the failure?
> 
>> 
>> R40 seems to use different registers for cpu bringup, so you will need
>> to define a new platform for it in aw_machdep.c and defines some new
>> function for ap bringup in aw_mp.c
>> The User Manual section 3.2 of the R40 have all the needed info.
>> 
> 
> Yeps, already did that, and I can reset the cores etc, that seems to work (register bits flips), but there are no mention of regs to put the starting address into.
> 
> Chap 3.2 doesn’t mention anything in that regard (at least in the docs I’ve found online) its moved to chap 4.2 in the latest R40/V40/T3 docs I’ve found bu still no mention of starting address.
> 
> I looks somewhat like the A20 ie it has control for 2 cpu’s the last two are missing from the register ref, which suggests to me that the docs might be inaccurate at best.
> 
> I looked into “Tina-linux” that they made for it, but I havn’t found out how they start it there either..
> 
> I might also just have stared me blind at it :)

Found that the start adresse is somewhere else:

CPU software entry register offset 00bc in sysctl instead of cpuctl…

I’ll try that when I get home later today,,,

-Søren