linux-kernel - Re: [PATCH v4 7/8] clocksource: Add a new timer-ingenic driver

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <73091b00-205b-3133-6650-21b6e3fda70f@linaro.org>
Date:   Tue, 3 Apr 2018 11:59:47 +0200
From:   Daniel Lezcano <daniel.lezcano@...aro.org>
To:     Paul Cercueil <paul@...pouillou.net>
Cc:     Thomas Gleixner <tglx@...utronix.de>,
        Jason Cooper <jason@...edaemon.net>,
        Marc Zyngier <marc.zyngier@....com>,
        Lee Jones <lee.jones@...aro.org>,
        Ralf Baechle <ralf@...ux-mips.org>,
        Rob Herring <robh+dt@...nel.org>,
        Jonathan Corbet <corbet@....net>,
        Mark Rutland <mark.rutland@....com>,
        James Hogan <jhogan@...nel.org>,
        Maarten ter Huurne <maarten@...ewalker.org>,
        linux-clk@...r.kernel.org, devicetree@...r.kernel.org,
        linux-kernel@...r.kernel.org, linux-mips@...ux-mips.org,
        linux-doc@...r.kernel.org
Subject: Re: [PATCH v4 7/8] clocksource: Add a new timer-ingenic driver

On 31/03/2018 19:46, Paul Cercueil wrote:
> Le 2018-03-31 10:10, Daniel Lezcano a écrit :
>> On 29/03/2018 16:52, Paul Cercueil wrote:
>>>
>>>
>>> Le mer. 28 mars 2018 à 18:25, Daniel Lezcano <daniel.lezcano@...aro.org>
>>> a écrit :
>>>> On 28/03/2018 17:15, Paul Cercueil wrote:
>>>>>  Le 2018-03-24 07:26, Daniel Lezcano a écrit :
>>>>>>  On 18/03/2018 00:29, Paul Cercueil wrote:
>>>>>>>  This driver will use the TCU (Timer Counter Unit) present on the
>>>>>>> Ingenic
>>>>>>>  JZ47xx SoCs to provide the kernel with a clocksource and timers.
>>>>>>
>>>>>>  Please provide a more detailed description about the timer.
>>>>>
>>>>>  There's a doc file for that :)
>>>>
>>>> Usually, when there is a new driver I ask for a description in the
>>>> changelog for reference.
>>>>
>>>>>>  Where is the clocksource ?
>>>>>
>>>>>  Right, there is no clocksource, just timers.
>>>>>
>>>>>>  I don't see the point of using channel idx and pwm checking here.
>>>>>>
>>>>>>  There is one clockevent, why create multiple channels ? Can't you
>>>>>> stick
>>>>>>  to the usual init routine for a timer.
>>>>>
>>>>>  So the idea is that we use all the TCU channels that won't be used
>>>>> for PWM
>>>>>  as timers. Hence the PWM checking. Why is this bad?
>>>>
>>>> It is not bad but arguable. By checking the channels used by the pwm in
>>>> the code, you introduce an adherence between two subsystems even if it
>>>> is just related to the DT parsing part.
>>>>
>>>> As it is not needed to have more than one timer in the time framework
>>>> (at least with the same characteristics), the pwm channels check is
>>>> pointless. We can assume the author of the DT file is smart enough to
>>>> prevent conflicts and define a pwm and a timer properly instead of
>>>> adding more code complexity.
>>>>
>>>> In addition, simplifying the code will allow you to use the timer-of
>>>> code and reduce very significantly the init function.
>>>
>>> That's what I had in my V1 and V2, my DT node for the timer-ingenic
>>> driver
>>> had a "timers" property (e.g. "timers = <4 5>;") to select the channels
>>> that
>>> should be used as timers. Then Rob told me I shouldn't do that, and
>>> instead
>>> detect the channels that will be used for PWM.
>>>
>>
>> [ ... ]
>>
>> How do you specify the channels used for PWM ?
> 
> To detect the channels that will be used as PWM I parse the whole
> devicetree
> searching for "pwms" properties; check that the PWM handle is for our
> TCU PWM
> driver; then read the PWM number from there.
> 
> Of course it's hackish, and it only works for devicetree. I preferred the
> method with the "timers" property.

Do you have a DT portion describing that? Eg somewhere in the kernel's
git tree ?

>From what I understood, we can specify the channel for a pwm but not for
a timer, there is certainly something I'm missing.

>>>>>>>
>>>>>>>  +config INGENIC_TIMER
>>>>>>>  +    bool "Clocksource/timer using the TCU in Ingenic JZ SoCs"
>>>>>>>  +    depends on MACH_INGENIC || COMPILE_TEST
>>>>>>
>>>>>>  bool "Clocksource/timer using the TCU in Ingenic JZ SoCs" if
>>>>>> COMPILE_TEST
>>>>>>
>>>>>>  Remove the depends MACH_INGENIC.
>>>>>
>>>>>  This driver is not useful on anything else than Ingenic SoCs, why
>>>>> should I
>>>>>  remove MACH_INGENIC then?
>>>>
>>>> For COMPILE_TEST on x86.
>>>
>>> Well that's a logical OR right here, so it will work...
>>
>> Right, I missed the second part of the condition. For consistency
>> reason, we don't add a dependency on the platform. The platform will
>> select it. Look the other timer options and you will see there is no
>> MACH deps. I'm trying consolidating all these options to have same
>> format and hopefully factor them out.
> 
> I'm all for factorisation, but what I dislike with not depending on
> MACH_INGENIC, is that the driver now appears in the menuconfig for
> every arch, even if it only applies to one MIPS SoC.

Can you do the following change?

bool "Clocksource/timer using the TCU in Ingenic JZ SoCs" if COMPILE_TEST

so it will appear only when the COMPILE_TEST option is set whatever the
platform which is the purpose of this option to increase compile test
coverage.


-- 
 <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