rpi2: cpufreq(4) support lost ?

[Claude Buisson]

On 11/13/2017 13:20, Emmanuel Vadot wrote:
> On Mon, 13 Nov 2017 03:18:06 -0800
> Mark Millard <markmi at dsl-only.net> wrote:
> 
>> [sys/arm/broadcom/bcm2835/bcm2835_cpufreq.c does not
>> list "brcm,bcm2836": That is what I would have found
>> if I had looked at the details before babbling
>> incoherently.]
> 
>   That seems to be the problem yes.
>   I'll confirm tonight and commit a fix if nobody does it before me.
> 
>> On 2017-Nov-13, at 1:54 AM, Emmanuel Vadot <manu at bidouilliste.com> wrote:
>>
>>> On Sun, 12 Nov 2017 15:23:39 -0800
>>> Mark Millard <markmi at dsl-only.net> wrote:
>>>
>>>> [This is a Linux *.dt* source issue, not specific to
>>>> amrmv6 vs. armv7: FreeBSD has switched to Linux *.dt*
>>>> source files, 4.13 most recently if I remember right.]
>>>
>>> This we do not use the DTS from Linux for RPI1/2 I doubt that.
>>
>> My apologies to all for the mistake for the rpi* context.
>>
>>>> On 2017-Nov-12, at 2:30 PM, Claude Buisson <clbuisson at orange.fr> wrote:
>>>>
>>>>> On 11/12/2017 20:03, Herbert J. Skuhra wrote:
>>>>> . . .
>>>>>
>>>>> So, something musst be wrong with the dtb file. I'm using now a file from
>>>>> r323309, where cpufreq is available again.
>>>> . . .
>>>
>>> The way to get functionality back is to find the real problem.
>>
>> Looking around, if I gather right, the only *.dt*
>> source change after -r323309 was in -r324184
>> in rpi2.dts :
>>
>> compatible = "brcm,bcm2709";
>>
>> was replaced by:
>>
>> compatible = "brcm,bcm2836";
>>
>> Folks are reporting that the new code works with
>> the old .dtb from before this change. This suggests
>> that may be the differences between the lines:
>>
>> FDT_PLATFORM_DEF2(bcm2836, bcm2836_legacy, "bcm2836 (legacy)", 0, "brcm,bcm2709", 100);
>> FDT_PLATFORM_DEF2(bcm2836, bcm2836, "bcm2836", 0, "brcm,bcm2836", 100);
>>
>> in sys/arm/broadcom/bcm2835/bcm2835_machdep.c might
>> be involved in some way. In fact:
>>
>> /usr/src/sys/arm/broadcom/bcm2835/bcm2835_cpufreq.c
>>
>> has:
>>
>> static struct ofw_compat_data compat_data[] = {
>>          { "broadcom,bcm2835-vc",        1 },
>>          { "broadcom,bcm2708-vc",        1 },
>>          { "brcm,bcm2709",       1 },
>>          { NULL, 0 }
>> };
>>
>> which does not list "brcm,bcm2836" as an alternative
>> but does list the "brcm,bcm2709" alternative.
>>
>> My guess is this explains the cpufreq code working
>> for the older *.dtb files but not the newer ones.
>>
>>>> . . .
>>>>
>>>> ===
>>>> Mark Millard
>>>> markmi at dsl-only.net
>>>
>>> All this is mostly wrong and doesn't have anything to do with the
>>> cpufreq problem on RPI2.
>>
>> Again: sorry for the misapplied context.
>>
>>> Please stay focus.
>>
>> I clearly did not do appropriate investigation
>> and incorrectly presumed what context would apply.
>> My mistake for sure.
>>
>>
>> Side note relative to BPI-M3 . . .
>>
>> FYI: The beginning of:
>>
>> https://lists.freebsd.org/pipermail/freebsd-arm/2017-November/017039.html
>>
>> lists what I've done locally to keep the BPI-M3 going as
>> of head -r325700 .
>>
>> ===

>> Mark Millard
>> markmi at dsl-only.net
>  

> 

Just to confirm that the patch @ r325768 solves the cpufreq problem

Claude Buisson