Re: [PATCH 1/3] SLIMbus: Device management on SLIMbus

From: Sagar Dharia <sdharia@codeaurora.org>
To: Mark Brown <broonie@kernel.org>
Cc: gregkh@linuxfoundation.org, linux-kernel@vger.kernel.org,
	bp@suse.de, poeschel@lemonage.de, treding@nvidia.com,
	gong.chen@linux.intel.com, andreas.noever@gmail.com,
	alan@linux.intel.com, mathieu.poirier@linaro.org,
	daniel@ffwll.ch, oded.gabbay@amd.com, jkosina@suse.cz,
	sharon.dvir1@mail.huji.ac.il, joe@perches.com,
	davem@davemloft.net, james.hogan@imgtec.com,
	michael.opdenacker@free-electrons.com,
	daniel.thompson@linaro.org, nkaje@codeaurora.org,
	kheitke@audience.com, mlocke@codeaurora.org,
	agross@codeaurora.org, linux-arm-msm@vger.kernel.org
Subject: Re: [PATCH 1/3] SLIMbus: Device management on SLIMbus
Date: Wed, 17 Jun 2015 11:10:09 -0600	[thread overview]
Message-ID: <5581A9F1.6010301@codeaurora.org> (raw)
In-Reply-To: <20150617114501.GB3214@sirena.org.uk>

On 6/17/2015 5:45 AM, Mark Brown wrote:
> On Tue, Jun 16, 2015 at 09:22:31AM -0600, Sagar Dharia wrote:
>> On 6/15/2015 4:54 AM, Mark Brown wrote:
>>> On Sat, Jun 13, 2015 at 11:49:16PM -0600, Sagar Dharia wrote:
>>>> +void slim_ctrl_add_boarddevs(struct slim_controller *ctrl)
>>>> +{
>>> Why are these operations split?
>> Some slaves may choose to do expensive operations in their probes (or wait
>> for logical address assignment). That will delay/block controller
>> registration and further HW initialization of the controller.
>> I did not see any downside/side-effect in splitting them. I am open however
>> to have them combined if that's the preferred way.
> How does this resolve the issue?  As far as I can see it just shuffles
> it around so any delay happens later (and possibly not at a convenient
> time), possibly not helping if there's multiple controllers.  If devices
> are doing excessively expensive things in their probe that seems like
> something that we should fix in the drivers.  For LA assignment
> shouldn't we be addressing that by having a callback when the LA is
> assigned?  That will avoid single threading effects and is more like the
> pattern for other buses.  Slimbus isn't particularly unique in this
> regard.
I agree that using the 'device_up' callback (it's called when the device 
has LA assignment) will make sure slaves can keep their probes non-blocking.
I was just trying to accommodate that blocking call by slaves and did 
not realize the downside you now mentioned about multiple-controllers
I will combine these in next patchset.
Thanks
Sagar
>
> Greg, the LA assignment thing here is an example of the issue is an
> example of the pattern I raised a while ago (but never got round to
> coding up handling of) where we have devices on an enumerable bus with
> static registrations.  LA assignment is the end of the hotplug process
> for a Slimbus device but most devices will need something doing to power
> them up so they enumerate.


-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
hosted by The Linux Foundation