Re: How to handle devices sitting on multiple busses ?

[Anssi Hannula <anssi.hannula@gmail.com>]

Laurent Pinchart wrote:
[...]
> As the bridge and I2C master live their own life in the Linux device tree, 
> they are initialized, suspended, resumed and destroyed independently. The 
> sensor being an I2C slave device, Linux initializes it after the I2C master 
> device is initialized, but doesn't ensure that the bridge is initialized first 
> as well. A similar problem occurs during suspend/resume, as the I2C slave 
> needs to be suspended before and resumed after the video bridge.
> 
> Have you ever encountered such a situation before ? Is there a clean way for a 
> device to have multiple parents, or do you have plans for such a possibility 
> in the future ? I would be willing to give an implementation a try if you can 
> provide me with some guidelines.
> 

It looks to me like this patch is related to your problem:
http://article.gmane.org/gmane.linux.power-management.general/15651

With a quick glance it doesn't seem to involve changes in device
initialization or destroying, though, but this is not really my area of
expertise.

-- 
Anssi Hannula