[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <d5506b03-e035-aca9-5e01-e642c2bc5111@samsung.com>
Date: Mon, 28 Jan 2019 12:57:25 +0100
From: Marek Szyprowski <m.szyprowski@...sung.com>
To: myungjoo.ham@...sung.com,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
"linux-pm@...r.kernel.org" <linux-pm@...r.kernel.org>,
"linux-samsung-soc@...r.kernel.org"
<linux-samsung-soc@...r.kernel.org>
Cc: Chanwoo Choi <cw00.choi@...sung.com>,
Krzysztof Kozlowski <krzk@...nel.org>,
Bartlomiej Zolnierkiewicz <b.zolnierkie@...sung.com>,
Lukasz Luba <l.luba@...tner.samsung.com>,
Markus Reichl <m.reichl@...etechno.de>
Subject: Re: [PATCH] devfreq: Suspend all devices on system shutdown
Hi MyungJoo,
On 2019-01-28 09:05, MyungJoo Ham wrote:
>> This way devfreq core ensures that all its devices will be set to safe
>> operation points before reboot operation. There are board on which some
>> aggressive power saving operation points are behind the capabilities of
>> the bootloader to properly reset the hardware and boot the board. This
>> way one can avoid board crash early after reboot.
>>
>> Similar pattern is used in CPUfreq subsystem.
>>
>> Reported-by: Markus Reichl <m.reichl@...etechno.de>
>> Signed-off-by: Marek Szyprowski <m.szyprowski@...sung.com>
>> ---
> You are invoking ALL devfreq suspend callbacks at shutdown
> with this commit.
>
> Can you make it invoke only devices explicitly saying their needs
> to handle "SHUTDOWN" event?
>
> For example, we can add a flag at struct devfreq_dev_profile:
> "uint32_t requirement"
> , where
> 0x1: need to operate at the initial frequency for suspend
> 0x2: need to operate at the initial frequency for shutdown
> 0x4: it forgets its status at resume, reconfigure frequency at resume.
> (or reverse 0x1's semantics for the backward compatibility)
Frankly speaking this looks like an over-engineering. Switching to safe
frequency during reboot shouldn't have any negative side-effects. IMHO
such flags can be always added later, once there will be a real use case
for them.
Best regards
--
Marek Szyprowski, PhD
Samsung R&D Institute Poland
Powered by blists - more mailing lists