linux-kernel - Re: [PATCH v2 4/6] pwm: omap-dmtimer: Fix pwm disabling sequence

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] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 11 Mar 2020 09:42:39 +0530
From:   Lokesh Vutla <lokeshvutla@...com>
To:     Tony Lindgren <tony@...mide.com>
CC:     Thierry Reding <thierry.reding@...il.com>,
        Uwe Kleine-König <u.kleine-koenig@...gutronix.de>,
        Linux OMAP Mailing List <linux-omap@...r.kernel.org>,
        <linux-kernel@...r.kernel.org>, <linux-pwm@...r.kernel.org>,
        Sekhar Nori <nsekhar@...com>, Vignesh R <vigneshr@...com>,
        Sebastian Reichel <sre@...nel.org>
Subject: Re: [PATCH v2 4/6] pwm: omap-dmtimer: Fix pwm disabling sequence



On 10/03/20 9:22 PM, Tony Lindgren wrote:
> * Lokesh Vutla <lokeshvutla@...com> [200310 07:06]:
>> Hi Tony,
>>
>> [...snip...]
>>
>>>>>>  
>>>>>> +	/*
>>>>>> +	 * Disable auto reload so that the current cycle gets completed and
>>>>>> +	 * then the counter stops.
>>>>>> +	 */
>>>>>>  	mutex_lock(&omap->mutex);
>>>>>> -	omap->pdata->stop(omap->dm_timer);
>>>>>> +	omap->pdata->set_pwm(omap->dm_timer,
>>>>>> +			     pwm_get_polarity(pwm) == PWM_POLARITY_INVERSED,
>>>>>> +			     true, OMAP_TIMER_TRIGGER_OVERFLOW_AND_COMPARE,
>>>>>> +			     false);
>>>>>> +
>>>>>>  	mutex_unlock(&omap->mutex);
>>>>>>  }
>>>>>
>>>>> I'm seeing an issue with this patch where after use something is
>>>>> left on and power consumption stays higher by about 30 mW after
>>>>> use.
>>>>
>>>> Interesting...What is the PWM period and duty cycle in the test case?
>>>> Can you dump the following registers before and after disabling:
>>>> - TLDR
>>>> - TMAR
>>>> - TCLR
>>>
>>> Here's the state dumped before and after in omap_dm_timer_set_pwm():
>>>
>>> omap_timer 4803e000.timer: XXX set_pwm before: tldr: fffffeb8 tmar: fffffffe tclr: 00000040
>>> omap_timer 4803e000.timer: XXX set_pwm after: tldr: fffffeb8 tmar: fffffffe tclr: 00001842
>>> omap_timer 4013e000.timer: XXX set_pwm before: tldr: fffffeb8 tmar: fffffffe tclr: 00000040
>>> omap_timer 4013e000.timer: XXX set_pwm after: tldr: fffffeb8 tmar: fffffffe tclr: 00001842
>>> omap_timer 4013e000.timer: XXX set_pwm before: tldr: fffffeb8 tmar: fffffffe tclr: 00001843
>>> omap_timer 4013e000.timer: XXX set_pwm after: tldr: fffffeb8 tmar: fffffffe tclr: 00001841
>>> omap_timer 4803e000.timer: XXX set_pwm before: tldr: fffffeb8 tmar: fffffffe tclr: 00001843
>>> omap_timer 4803e000.timer: XXX set_pwm after: tldr: fffffeb8 tmar: fffffffe tclr: 00001841
>>>
>>
>> Looking at the registers:
>> period = 327 *(1000/clk_freq in MHz) ns
>> duty_cycle =  perioid.
>>
>> I did simulate this behavior on BeagleBoneBlack on timer7. PWM is going down
>> after disabling.
>>
>>> So looks like the start bit is still enabled after use?
>>
>> Right, that is expected. The start bit gets disabled automatically once the pwm
>> period completes. This is because auto reload bit is off. That's the main idea
>> of this patch so that PWM period is completed after disabling, else PWM is
>> stopped abruptly.
> 
> OK
> 
>> Not sure why it is not happening in your case. If you think it is not needed, I
>> can drop this patch and add a limitation saying that PWM gets disabled
>> immediately without completing the current cycle.
> 
> Could it be that we now have the cpu_pm notifier restore something
> invalid after exiting idle that restarts the counter?

If that's the case, mis behavior should have happened without this patch as well.

Is it possible for you to dump the registers when you are observing higher power
consumption after the use?

However, I see an issue with the patch itself as pm_runtime is not disabled
after the pwm is stopped. Not sure how that could be nullified with this approach.


Thanks and regards,
Lokesh