Date: Thu, 21 Apr 2016 16:15:50 +0800
From: Jisheng Zhang <jszhang@marvell.com>
To: Jarkko Nikula <jarkko.nikula@linux.intel.com>
CC: <andriy.shevchenko@linux.intel.com>, <mika.westerberg@linux.intel.com>,
        <wsa@the-dreams.de>, <linux-i2c@vger.kernel.org>,
        <linux-kernel@vger.kernel.org>, <linux-arm-kernel@lists.infradead.org>
Subject: Re: [PATCH 1/4] i2c: designware-platdrv: Fix runtime PM
 initialization
Message-ID: <20160421161550.023d503f@xhacker>
In-Reply-To: <571885E8.4000301@linux.intel.com>
References: <1460638414-5987-1-git-send-email-jszhang@marvell.com>
	<1460638414-5987-2-git-send-email-jszhang@marvell.com>
	<571789C4.5020904@linux.intel.com>	<20160421110859.1509ad2c@xhacker>
	<571885E8.4000301@linux.intel.com>
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 2757
Lines: 70

On Thu, 21 Apr 2016 10:48:56 +0300
Jarkko Nikula <jarkko.nikula@linux.intel.com> wrote:

> Hi
> 
> On 04/21/2016 06:08 AM, Jisheng Zhang wrote:
> > Dear Jarkko,
> >
> > On Wed, 20 Apr 2016 16:53:08 +0300 Jarkko Nikula wrote:
> >  
> >> On 04/14/2016 03:53 PM, Jisheng Zhang wrote:  
> >>> When pm_runtime_enable() was being called, the device's usage counter
> >>> was 0, causing the PM layer to runtime-suspend the device.  We then
> >>> went on to call i2c_dw_probe() on a suspended device, which could hung.
> >>>
> >>> Fix this by incrementing the usage counter before pm_runtime_enable().
> >>>
> >>> Signed-off-by: Jisheng Zhang <jszhang@marvell.com>
> >>> ---
> >>>    drivers/i2c/busses/i2c-designware-platdrv.c | 14 +++++++++++++-
> >>>    1 file changed, 13 insertions(+), 1 deletion(-)
> >>>
> >>> diff --git a/drivers/i2c/busses/i2c-designware-platdrv.c b/drivers/i2c/busses/i2c-designware-platdrv.c
> >>> index d656657..00f9e99 100644
> >>> --- a/drivers/i2c/busses/i2c-designware-platdrv.c
> >>> +++ b/drivers/i2c/busses/i2c-designware-platdrv.c
> >>> @@ -246,6 +246,7 @@ static int dw_i2c_plat_probe(struct platform_device *pdev)
> >>>    	if (dev->pm_runtime_disabled) {
> >>>    		pm_runtime_forbid(&pdev->dev);
> >>>    	} else {
> >>> +		pm_runtime_get_noresume(&pdev->dev);
> >>>    		pm_runtime_set_autosuspend_delay(&pdev->dev, 1000);
> >>>    		pm_runtime_use_autosuspend(&pdev->dev);
> >>>    		pm_runtime_set_active(&pdev->dev);  
> >>
> >> pm_runtime_enable() here after pm_runtime_set_active() shouldn't suspend
> >> as far as I understand which made me thinking if there is some other  
> >
> > FWICT, on arm DT platform, the device usage counter is zero at the beginning,
> > once we enable rpm, the i2c have chance to runtime suspend.
> >  
> Yes it is same also for ACPI platforms. What I like to understand what 
> is actually causing the runtime suspend in your case.

OK, got your point.

> 
> If I add delay longer than 1 second between pm_runtime_enable() and 
> i2c_dw_probe() suspends still happens only after probe finishes. It will 
> be triggered by the drivers/base/dd.c: driver_probe_device() when code 
> calls the pm_request_idle().
> 
> I'm wondering is it possible there is a race with deferred probe etc. 
> that is causing the issue you are seeing.

Hmm, I do see there's deferred probe for the i2c hosts. something like:

[    0.177338] platform f7fcc000.i2c: Driver i2c_designware requests probe deferral

Is it possible as this:

probe => driver return EPROBE_DEFER => pm_request_idle() => probe again
=> since the device usage count is zero, once rpm is enabled, we have
chance to suspend?

And from another side, is it reasonable to always increase the usage count
until probe is done?

Thanks,
Jisheng