Re: [PATCH v6 2/2] clocksource: Ingenic: Add support for the Ingenic X1000 OST.

[Paul Cercueil <paul@crapouillou.net>]

Hi Zhou,

Le sam. 18 juil. 2020 à 21:42, Zhou Yanjie <zhouyanjie@wanyeetech.com> 
a écrit :
> Hello Paul and Daniel,
> 
> 在 2020/7/18 下午9:12, Paul Cercueil 写道:
>> Hi Daniel,
>> 
>> Le ven. 17 juil. 2020 à 10:02, Daniel Lezcano 
>> \x7f<daniel.lezcano@linaro.org> a écrit :
>>> On 17/07/2020 08:13, Zhou Yanjie wrote:
>>>>  Hi Daniel,
>>>> 
>>>>  在 2020/7/17 下午12:20, Daniel Lezcano 写道:
>>>>>  On 10/07/2020 19:02, 周琰杰 (Zhou Yanjie) wrote:
>>>>>>  X1000 and SoCs after X1000 (such as X1500 and X1830) had a 
>>>>>> separate
>>>>>>  OST, it no longer belongs to TCU. This driver will register 
>>>>>> both a
>>>>>>  clocksource and a sched_clock to the system.
>>>>>> 
>>>>>>  Tested-by: 周正 (Zhou Zheng) <sernia.zhou@foxmail.com>
>>>>>>  Co-developed-by: 漆鹏振 (Qi Pengzhen) <aric.pzqi@ingenic.com>
>>>>>>  Signed-off-by: 漆鹏振 (Qi Pengzhen) <aric.pzqi@ingenic.com>
>>>>>>  Signed-off-by: 周琰杰 (Zhou Yanjie) 
>>>>>> <zhouyanjie@wanyeetech.com>
>>>>>>  Reviewed-by: Paul Cercueil <paul@crapouillou.net>
>>>>>>  ---
>>>>>> 
>>>>>>  Notes:
>>>>>>       v1->v2:
>>>>>>       Fix compile warnings.
>>>>>>       Reported-by: kernel test robot <lkp@intel.com>
>>>>>>            v2->v3:
>>>>>>       No change.
>>>>>>            v3->v4:
>>>>>>       1.Rename "ost" to "sysost"
>>>>>>       1.Remove unrelated changes.
>>>>>>       2.Remove ost_clock_parent enum.
>>>>>>       3.Remove 
>>>>>> ost->percpu_timer_channel/ost->global_timer_channel.
>>>>>>       4.Set up independent .recalc_rate/.set_rate for 
>>>>>> percpu/global
>>>>>>  timer.
>>>>>>       5.No longer call functions in variable declarations.
>>>>>>            v4->v5:
>>>>>>       Use "of_io_request_and_map()" instead "of_iomap()".
>>>>>>       Suggested-by: Paul Cercueil <paul@crapouillou.net>
>>>>>>            v5->v6:
>>>>>>       No change.
>>>>>> 
>>>>>>    drivers/clocksource/Kconfig          |  11 +
>>>>>>    drivers/clocksource/Makefile         |   1 +
>>>>>>    drivers/clocksource/ingenic-sysost.c | 539
>>>>>>  +++++++++++++++++++++++++++++++++++
>>>>>>    3 files changed, 551 insertions(+)
>>>>>>    create mode 100644 drivers/clocksource/ingenic-sysost.c
>>>>>> 
>>>>>>  diff --git a/drivers/clocksource/Kconfig 
>>>>>> \x7f\x7f\x7f\x7f\x7fb/drivers/clocksource/Kconfig
>>>>>>  index 91418381fcd4..1bca8b8fb30f 100644
>>>>>>  --- a/drivers/clocksource/Kconfig
>>>>>>  +++ b/drivers/clocksource/Kconfig
>>>>>>  @@ -696,6 +696,17 @@ config INGENIC_TIMER
>>>>>>        help
>>>>>>          Support for the timer/counter unit of the Ingenic JZ 
>>>>>> SoCs.
>>>>>>    +config INGENIC_SYSOST
>>>>>>  +    bool "Clocksource/timer using the SYSOST in Ingenic X SoCs"
>>>>>  We usually use silent options and let the platform's Kconfig 
>>>>> \x7f\x7f\x7f\x7fenable it.
>>>>>  We show up the option only when COMPILE_TEST is enabled.
>>>>> 
>>>>>  Is there a reason to do it differently?
>>>> 
>>>> 
>>>>  Do you mean
>>>> 
>>>>  bool "Clocksource/timer using the SYSOST in Ingenic X SoCs"
>>>> 
>>>>  or
>>>> 
>>>>  default MACH_INGENIC ?
>>> 
>>> Both, no default here.
>>> 
>>> eg.
>>> 
>>> bool "Clocksource/timer using the SYSOST in Ingenic X SoCs" if 
>>> \x7f\x7fCOMPILE_TEST
>>> 
>>> and
>>> 
>>> in arch/mips/Kconfig in the config MACH_INGENIC section :
>>> 
>>> ...
>>> select INGENIC_SYSOST
>>> ...
>> 
>> Disagreed. That's not how we do things on MIPS. Selecting 
>> MACH_INGENIC \x7fmeans "this kernel will support Ingenic SoCs", but not 
>> that it will \x7fonly support these. Hence the depends on MIPS / 
>> default MACH_INGENIC.
>> 
>> As for the select INGENIC_SYSOST, this driver only applies to a few 
>> \x7fSoCs, I certainly don't want it to be force-enabled. I don't even 
>> wait \x7fit to be force-enabled on X1000, since it is optional there 
>> too.
>> 
>> Cheers,
>> -Paul
> 
> 
> If we still need to keep the "default MACH_INGENIC", then Daniel can 
> directly apply the v6 version.
> 
> If we need to use the silent options, maybe we can enable them 
> separately according to 
> MACH_JZ4740/MACH_JZ4770/MACH_JZ4780/MACH_X1000/MACH_X1830.
> 
> In fact, I think X1000 and X1830 need to enable this driver in most 
> cases, because the current test has found that use TCU to provide 
> clocksource and clockevent will cause data loss/error when 
> transmitting data through spi or ethernet. And these errors no longer 
> appear after using OST.

This is likely because the clock is too fast, try to reduce it by a 
factor 2 or 4, it should behave better.

If it turns out OST is really needed, then it should be selected from 
MACH_X1000/X1830, not MACH_INGENIC.

Cheers,
-Paul

> Thanks and best regards!
> 
> 
>> 
>>> 
>>>>  This driver has some origins from "INGENIC_TIMER" driver and
>>>>  "INGENIC_OST" driver.
>>>>  Early Ingenic processors used TCU (timer/counter unit, has 6 or 8
>>>>  generic timer channels) to provide clocksource and clockevent 
>>>> (both \x7f\x7f\x7fwith
>>>>  only 16bit precision). This part of the processor can only use
>>>>  "INGENIC_TIMER" driver.
>>>> 
>>>>  Later processors provide an independent 32bit or 64bit timer 
>>>> channel
>>>>  (still under TCU, known as ost channel, this channel can not 
>>>> generate
>>>>  interrupt) to provid higher precision clocksource. The 
>>>> "INGENIC_OST"
>>>>  driver is for this channel. These processors can use 
>>>> "INGENIC_TIMER"
>>>>  driver, but using "INGENIC_OST" driver to provide higher precision
>>>>  clocksource would be a better choice (clockevent still needs to be
>>>>  provided by generic timer channel of TCU, and still 16bit 
>>>> precision).
>>>> 
>>>>  And the recent processors provide a SYSOST components, it is 
>>>> \x7f\x7f\x7findependent
>>>>  from TCU, including a 64bit timer channel for clocksource and a 
>>>> 32bit
>>>>  timer channel for clockevent. Although these processors can also 
>>>> use
>>>>  "INGENIC_TIMER" driver, but the better choice is completely 
>>>> \x7f\x7f\x7findependent
>>>>  use of "INGENIC_SYSOST" driver to provide higher precision 
>>>> clocksource
>>>>  and clockevent.
>>> 
>>> Ok, the rating should do the job then.
>>> 
>>> Thanks for the explanation.
>>> 
>>>>  You may have already noticed that this independent SYSOST 
>>>> component is
>>>>  like an upgraded and streamlined TCU, which only retains one 
>>>> generic
>>>>  timer channel that can generate interrupts, upgrade it from 16bit 
>>>> to
>>>>  32bit, and then retain the 64bit ost channel. so the driver code 
>>>> and
>>>>  Kconfig code of this patch is largely referenced
>>>>  "INGENIC_TIMER" driver and "INGENIC_OST" driver.
>>>> 
>>>>  Thanks and best regards!
>>>> 
>>>>>>  +    default MACH_INGENIC
>>>>>>  +    depends on MIPS || COMPILE_TEST
>>>>>>  +    depends on COMMON_CLK
>>>>>>  +    select MFD_SYSCON
>>>>>>  +    select TIMER_OF
>>>>>>  +    select IRQ_DOMAIN
>>>>>>  +    help
>>>>>>  +      Support for the SYSOST of the Ingenic X Series SoCs.
>>>>>>  +
>>>>>  [ ... ]
>>>>> 
>>>>> 
>>> 
>>> 
>>> --
>>> <http://www.linaro.org/> Linaro.org │ Open source software for 
>>> ARM SoCs
>>> 
>>> Follow Linaro:  <http://www.facebook.com/pages/Linaro> Facebook |
>>> <http://twitter.com/#!/linaroorg> Twitter |
>>> <http://www.linaro.org/linaro-blog/> Blog
>>