Re: [PATCH RFC] omap3isp: prevent releasing MC too early

[Mauro Carvalho Chehab <mchehab@s-opensource.com>]

Em Thu, 15 Dec 2016 16:04:51 +0200
Laurent Pinchart <laurent.pinchart@ideasonboard.com> escreveu:

We have now two threads discussing the same subject, which is bad, as
we'll end repeating the same arguments on different threads...

Let's use the "[PATCH RFC 00/21]" for those discussions, as it seems we're
reaching to somewhere there.

> Even if you're not entirely convinced by the reasons 
> explained in this mail thread, remember that we will need sooner or later to 
> implement support for media graph update at runtime. Refcounting will be 
> needed, let's design it in the cleanest possible way.

As I said, I'm not against using some other approach and even
adding refcounting to each graph object.

What I am against is on a patchset that starts by breaking 
the USB drivers that use the media controller.

Btw, I'm starting to suspect that getting rid of devm_*alloc()
on OMAP3, as proposed by the 00/21 thread is addressing a symptom of
the problem and not a cause, and that using get_device()/put_device()
may help fixing such issues. See Hans comments on that thread.

Thanks,
Mauro