[PATCH 2/5] drivercore: Add driver probe deferral mechanism

[Ming Lei]

CC Rafael and linux-pm

On Thu, Oct 13, 2011 at 12:09 PM, Grant Likely
<grant.likely at secretlab.ca> wrote:
> On Tue, Oct 11, 2011 at 08:29:18PM +0800, Ming Lei wrote:
>> On Tue, Oct 11, 2011 at 1:37 AM, Andrei Warkentin <awarkentin at vmware.com> wrote:
>> > Hi,
>> >
>> > ----- Original Message -----
>> >> From: "Greg KH" <greg at kroah.com>
>> >> To: "Josh Triplett" <josh at joshtriplett.org>
>> >> Cc: "G, Manjunath Kondaiah" <manjugk at ti.com>, linux-arm-kernel at lists.infradead.org, "Grant Likely"
>> >> <grant.likely at secretlab.ca>, linux-omap at vger.kernel.org, linux-mmc at vger.kernel.org, linux-kernel at vger.kernel.org,
>> >> "Dilan Lee" <dilee at nvidia.com>, "Mark Brown" <broonie at opensource.wolfsonmicro.com>, Manjunath at jasper.es
>> >> Sent: Saturday, October 8, 2011 11:55:02 AM
>> >> Subject: Re: [PATCH 2/5] drivercore: Add driver probe deferral mechanism
>> >>
>> >
>> > I'm a bit of a fly on the wall here, but I'm curious how this impacts suspend/resume.
>> > device_initialize->device_pm_init are called from device_register, so certainly this
>> > patch doesn't also ensure that the PM ordering matches probe ordering, which is bound
>> > to break suspend, right? Was this ever tested with the OMAP target? Shouldn't the
>>
>> Inside device_add(), device_pm_add is called before bus_probe_device,
>> so the patch can't change the device order in pm list, and just change
>> the driver probe order.
>
> That's the way it works now, but can it be reworked?  It would be

IMO, it depends on what shape you plan to rework.  Currently, the
deferred probe may found a resource dependency, but I am not sure
that pm dependency is same with the resource dependency found
during probe.

> possible to adjust the list order after successful probe.  However,
> I'm not clear on the ordering rules for the dpm_list.  Right now it is
> explicitly ordered to have parents before children, but as already
> expressed, that doesn't accurately represent ordering constraints for
> multiple device dependancies.

Maybe we should understand the correct model of the ordering constraints
for the multiple device dependancies first, could you give a description or
some examples about it?

>
> So, reordering the list would probably require maintaining the
> existing parent-child ordering constraint, but to also shift
> devices (and any possible children?) to be after drivers that are
> already probed.  That alone will be difficult to implement and get
> right, but maybe the constraints can be simplified.  It needs some
> further thought.
>
> g.
>
>


thanks,
-- 
Ming Lei