Re: [RFC PATCH] mmc: Enable wakeup_sources for mmc core

From: Zoran Markovic <zoran.markovic@linaro.org>
To: Ulf Hansson <ulf.hansson@linaro.org>
Cc: Colin Cross <ccross@android.com>,
	lkml <linux-kernel@vger.kernel.org>,
	linux-mmc@vger.kernel.org, San Mehat <san@google.com>,
	John Stultz <john.stultz@linaro.org>, Chris Ball <cjb@laptop.org>,
	Johan Rudholm <johan.rudholm@stericsson.com>,
	Jaehoon Chung <jh80.chung@samsung.com>,
	Konstantin Dorfman <kdorfman@codeaurora.org>,
	Guennadi Liakhovetski <g.liakhovetski@gmx.de>,
	Tejun Heo <tj@kernel.org>,
	Andrew Morton <akpm@linux-foundation.org>
Subject: Re: [RFC PATCH] mmc: Enable wakeup_sources for mmc core
Date: Thu, 1 Aug 2013 00:42:28 -0700	[thread overview]
Message-ID: <CAME+o4kpsPkoTpS7_0STDpMQpZr0KyfD1prbYYJs147-5DiMkg@mail.gmail.com> (raw)
In-Reply-To: <CAPDyKFpKYeqoFn=S6P2tu1s2C28daCh4vcHTUCCRUOx2_MHShg@mail.gmail.com>

Ulf,
I got confirmation from Broadcom that all cell phone reference designs
have card insert/removal configured as a wakeup IRQ. Unless our
customers change that - which I doubt - this results in a considerable
number of products implementing this feature.

Please let me know how you wish to proceed.

Cheers,
Zoran

On 26 June 2013 13:57, Ulf Hansson <ulf.hansson@linaro.org> wrote:
> On 24 June 2013 21:58, Zoran Markovic <zoran.markovic@linaro.org> wrote:
>>>> This patch is ported from the Android common tree, so you've probably
>>>> been using it.
>>>
>>> We removed more or less all Android code in the mmc subsystem, since
>>> it just didn't work. :-)
>>>
>>> The "deferred resume" was very useful though, so after some rework we
>>> kept it and could then improve the system resume time significantly.
>>
>> For what it's worth, I fixed one bug I noticed in the Android kernel:
>> if a system has a non-removable MMC device, a suspend/resume cycle on
>> this device would hold a wake lock forever. This was a visible issue
>> on the panda board I am using.
>>
>> If there are doubts on whether or not the system should stay awake
>> during a MMC mount, we have the option to make the calls to
>> wakeup_source_register/unregister configurable. Skipping these calls
>> would leave the .ws field NULL, in which case
>> __pm_stay_awake/__pm_relax/__pm_wakeup_event would do nothing.
>
> Even if we make this feature configurable, I can't see any host driver
> that would benefit from it as of today. The reason is simply that host
> drivers do not configure it's card detect irq as a wakeup irq. Myself
> is also having quite hard to see the benefit of doing that, but I
> don't know all the use cases.
>
> Let's see if we can get someone else to provide input...
>
>>
>> Thoughts?
>>
>> - Zoran
>
>
> Kind regards
> Ulf Hansson