Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <e9d113fb-5cd1-d93d-3d8f-fa9c1e55a8e2@redhat.com>
Date:   Fri, 9 Dec 2022 13:15:21 +0100
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.5.1
Subject: Re: [PATCH] i2c: designware: Fix unbalanced suspended flag
Content-Language: en-US, nl
To:     Richard Fitzgerald <rf@opensource.cirrus.com>, wsa@kernel.org,
        jarkko.nikula@linux.intel.com, andriy.shevchenko@linux.intel.com,
        mika.westerberg@linux.intel.com, jsd@semihalf.com
Cc:     linux-i2c@vger.kernel.org, linux-kernel@vger.kernel.org,
        patches@opensource.cirrus.com
References: <20221209114034.18025-1-rf@opensource.cirrus.com>
From:   Hans de Goede <hdegoede@redhat.com>
In-Reply-To: <20221209114034.18025-1-rf@opensource.cirrus.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

Hi Richard,

On 12/9/22 12:40, Richard Fitzgerald wrote:
> Ensure that i2c_mark_adapter_suspended() is always balanced by a call to
> i2c_mark_adapter_resumed().
> 
> Don't set DPM_FLAG_MAY_SKIP_RESUME to skip system early_resume stage if the
> driver was runtime-suspended. Instead, always call dw_i2c_plat_resume() and
> use pm_runtime_suspended() to determine whether we need to power up the
> hardware.
> 
> The unbalanced suspended flag was introduced by
> commit c57813b8b288 ("i2c: designware: Lock the adapter while setting the
> suspended flag")
> 
> Before that commit, the system and runtime PM used the same functions. The
> DPM_FLAG_MAY_SKIP_RESUME was used to skip the system resume if the driver
> had been in runtime-suspend. If system resume was skipped, the suspended
> flag would be cleared by the next runtime resume. The check of the
> suspended flag was _after_ the call to pm_runtime_get_sync() in
> i2c_dw_xfer(). So either a system resume or a runtime resume would clear
> the flag before it was checked.
> 
> Having introduced the unbalanced suspended flag with that commit, a further
> commit 80704a84a9f8
> ("i2c: designware: Use the i2c_mark_adapter_suspended/resumed() helpers")
> 
> changed from using a local suspended flag to using the
> i2c_mark_adapter_suspended/resumed() functions. These use a flag that is
> checked by I2C core code before issuing the transfer to the bus driver, so
> there was no opportunity for the bus driver to runtime resume itself before
> the flag check.
> 
> Signed-off-by: Richard Fitzgerald <rf@opensource.cirrus.com>
> Fixes: c57813b8b288 ("i2c: designware: Lock the adapter while setting the suspended flag")

It is not entirely clear to me where the unbalance you claim to see comes
from? When runtime-suspended SMART_SUSPEND should keep it suspended at which point
the system suspend callback will never run ?

Are you sure that you are not maybe seeing a suspend/resume ordering issue?

Did you add printk messages to the suspend/resume callbacks of
i2c-designware-platdrv.c which show the system suspend callback
being called but not the system resume one ?

I guess that is possible with DPM_FLAG_MAY_SKIP_RESUME, but
since we also use SMART_SUSPEND I would expect the system-suspend
callback to also always be skipped for runtime-suspended controllers.


> ---
> Apologies if you get this message multiple times. I'm having trouble
> with my SMTP server.
> ---
>  drivers/i2c/busses/i2c-designware-platdrv.c | 10 +++++-----
>  1 file changed, 5 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/i2c/busses/i2c-designware-platdrv.c b/drivers/i2c/busses/i2c-designware-platdrv.c
> index ba043b547393..d805b8c7e797 100644
> --- a/drivers/i2c/busses/i2c-designware-platdrv.c
> +++ b/drivers/i2c/busses/i2c-designware-platdrv.c
> @@ -351,13 +351,11 @@ static int dw_i2c_plat_probe(struct platform_device *pdev)
>  
>  	if (dev->flags & ACCESS_NO_IRQ_SUSPEND) {
>  		dev_pm_set_driver_flags(&pdev->dev,
> -					DPM_FLAG_SMART_PREPARE |
> -					DPM_FLAG_MAY_SKIP_RESUME);
> +					DPM_FLAG_SMART_PREPARE);
>  	} else {
>  		dev_pm_set_driver_flags(&pdev->dev,
>  					DPM_FLAG_SMART_PREPARE |
> -					DPM_FLAG_SMART_SUSPEND |
> -					DPM_FLAG_MAY_SKIP_RESUME);
> +					DPM_FLAG_SMART_SUSPEND);
>  	}
>  
>  	device_enable_async_suspend(&pdev->dev);
> @@ -475,7 +473,9 @@ static int __maybe_unused dw_i2c_plat_resume(struct device *dev)
>  {
>  	struct dw_i2c_dev *i_dev = dev_get_drvdata(dev);
>  
> -	dw_i2c_plat_runtime_resume(dev);
> +	if (!pm_runtime_suspended(dev))
> +		dw_i2c_plat_runtime_resume(dev);
> +

I'm afraid that this is always going to run now, before this callback gets
called drivers/base/power/main.c: device_resume_noirq() does:

        skip_resume = dev_pm_skip_resume(dev);

        if (skip_resume)
                pm_runtime_set_suspended(dev);
        else if (dev_pm_skip_suspend(dev))
                pm_runtime_set_active(dev);

Where skip_resume now is false since you dropped the
DPM_FLAG_MAY_SKIP_RESUME flag and dev_pm_skip_suspend(dev)
will return true (see below) for runtime-suspended controllers,
so they will be marked active here. and then your
!pm_runtime_suspended(dev) will always be false.

Did you add a printk to both the if + else paths
and have you ever seen the controller not get
resumed with this test added ?

Regards,

Hans


bool dev_pm_skip_suspend(struct device *dev)
{
        return dev_pm_test_driver_flags(dev, DPM_FLAG_SMART_SUSPEND) &&
                pm_runtime_status_suspended(dev);
}


>  	i2c_mark_adapter_resumed(&i_dev->adapter);
>  
>  	return 0;