linux-kernel - Re: [PATCH 1/2] watchdog: rti_wdt: Use managed APIs to handle runtime PM

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: <0f3277cc-0e4a-42c9-920c-87e1e6d6c0ca@ti.com>
Date:   Tue, 21 Nov 2023 09:30:34 +0530
From:   Vignesh Raghavendra <vigneshr@...com>
To:     Guenter Roeck <linux@...ck-us.net>,
        Wim Van Sebroeck <wim@...ux-watchdog.org>
CC:     Tero Kristo <t-kristo@...nel.org>,
        <linux-watchdog@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
        <linux-arm-kernel@...ts.infradead.org>, <afd@...com>,
        <n-francis@...com>
Subject: Re: [PATCH 1/2] watchdog: rti_wdt: Use managed APIs to handle runtime
 PM

Hi,

On 10/11/23 20:33, Guenter Roeck wrote:
> On 11/10/23 02:07, Vignesh Raghavendra wrote:
>> Switch to devm_pm_runtime_enable() to simplify error handling in driver
>> probe.
>>
> 
> This also replaces the call to pm_runtime_resume_and_get() without
> explanation.
> Worse, the next patch conditionally re-introduces pm_runtime_put_sync()
> on the probe function.
> 
>> Signed-off-by: Vignesh Raghavendra <vigneshr@...com>
>> ---
>>   drivers/watchdog/rti_wdt.c | 30 ++++++++----------------------
>>   1 file changed, 8 insertions(+), 22 deletions(-)
>>
>> diff --git a/drivers/watchdog/rti_wdt.c b/drivers/watchdog/rti_wdt.c
>> index 8e1be7ba0103..163bdeb6929a 100644
>> --- a/drivers/watchdog/rti_wdt.c
>> +++ b/drivers/watchdog/rti_wdt.c
>> @@ -236,12 +236,8 @@ static int rti_wdt_probe(struct platform_device
>> *pdev)
>>       if (wdt->freq < 32768)
>>           wdt->freq = wdt->freq * 9 / 10;
>>   -    pm_runtime_enable(dev);
>> -    ret = pm_runtime_resume_and_get(dev);
>> -    if (ret < 0) {
>> -        pm_runtime_disable(&pdev->dev);
>> -        return dev_err_probe(dev, ret, "runtime pm failed\n");
>> -    }
>> +    devm_pm_runtime_enable(dev);
> 
> devm_pm_runtime_enable() returns an error code. I don't think ignoring it
> is a good idea.
> 


Oops, yes...

>> +    pm_runtime_get_noresume(dev);
> 
> Is this functionally identical to pm_runtime_resume_and_get() ?
> That would require further explanation. Why is it not necessary
> to resume here ?

include/linux/pm_runtime.h ::
 pm_runtime_resume_and_get - Bump up usage counter of a device and
resume it.

vs

pm_runtime_get_noresume - Bump up runtime PM usage counter of a device.

During probe, device is already active. Hence, there is really no need
to call driver level runtime_resume() callback as there is really no
context to resume from. Driver currently doesnt have runtime_pm calls
which I intend to add as a later patch. I guess, its probably better to
move this patch to same series.

Also, missed a call to pm_runtime_put_noidle() in the error path.


So for now I will respin 2/2 as standalone fix and repost along with
runtime_pm support.

Apologies, for the delayed response!

-- 
Regards
Vignesh