Re: [PATCH v6 00/41] ARM: davinci: convert to common clock framework​

[David Lechner]

On 01/23/2018 01:53 PM, Bartosz Golaszewski wrote:
> 2018-01-23 20:24 GMT+01:00 David Lechner <david at lechnology.com>:
>> On 01/23/2018 12:34 PM, Bartosz Golaszewski wrote:
>>>
>>> 2018-01-23 19:26 GMT+01:00 David Lechner <david at lechnology.com>:
>>>>
>>>> On 01/23/2018 12:10 PM, Bartosz Golaszewski wrote:
>>>>>
>>>>>
>>>>> 2018-01-23 18:03 GMT+01:00 Adam Ford <aford173 at gmail.com>:
>>>>>>
>>>>>>
>>>>>> On Tue, Jan 23, 2018 at 10:06 AM, David Lechner <david at lechnology.com>
>>>>>> wrote:
>>>>>>>
>>>>>>>
>>>>>>> On 01/23/2018 10:03 AM, David Lechner wrote:
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> You can see if the clock is enabled by running:
>>>>>>>>
>>>>>>>>         cat /sys/kernel/debug/clk/clk_summary
>>>>>>>>
>>>>>>>
>>>>>>> I just realized if you can't boot, you can't do this. :-/
>>>>>>
>>>>>>
>>>>>>
>>>>>> I can boot with the latest set in your git repo, but the Ethernet
>>>>>> doesn't apparently fully operate.  I don't get errors, but I cannot
>>>>>> get a dhcp address.
>>>>>>
>>>>>> I'll try to do a more exhaustive test later today to get an idea of
>>>>>> what works and what doesn't.  When I ran my basic tests, I just did a
>>>>>> feel-good boot test (and reboot test)
>>>>>>
>>>>>> adam
>>>>>
>>>>>
>>>>>
>>>>> FYI: manually calling clk_prepare_enable() in the davinci_mdio driver
>>>>> seems to at least fix the ethernet. In master branch it's done by
>>>>> pm_runtime_get_sync() (in davinci_mdio_reset()). However I'm still
>>>>> getting several oopses and WARNs so there's some more work to do.
>>>>>
>>>>
>>>> Hmm... I'm wondering if we need to also add #power-domain-cells to the
>>>> PSC clocks and power-domains properties to the consumers.
>>>>
>>>> For this specific case though, it seems strange to me that the drivers
>>>> to clk_get() and clk_get_rate() but never enable the clocks.
>>>>
>>>>
>>>> Also, are the oopses and WARNs the same as before?
>>>
>>>
>>> No, the ones before were all related to the ethernet failing, now I
>>> get several stack traces from drm. Posted them on pastebin[1].
>>>
>>
>> It looks like the LCDC driver is the same way. It does clk_get() but
>> not clk_prepare_enable().
>>
>>
> 
> In the mdio case - the problem is that devm_clk_get() doesn't fail,
> but somehow the clock doesn't end up in the list of the device's
> clocks - which is why it's not enabled by pm_runtime_get_sync().
> 


Right. This is because devm_clk_get() now finds the clock via device
tree instead of a clkdev lookup entry. However, I think that the PM
notifier registered in arch/arm/mach-davinci/pm_domain.c only uses
the clkdev lookup to match the con_id and does not use device tree.
The same thing is happing in mdio, emac and lcdc.