On Wed, 22 September 2010 James Simmons [off-list ref] wrote:

quoted

quoted

I have a tree at

http://git.infradead.org/users/jsimmons/linuxconsole-2.6.git

but currently fbcon is broken so I'm tracing down the problem.

Thanks for the reference to your tree!

What's you opinion regarding my changes to fbcon in my RFC patch?
Are they ok or would you prefer having fbcon changed to stop peeking
at registered_fb list and just operate directly on fb_info everywhere
it needs it? (that is let con2fb_map[] point to fb_info instead of
indexes into registered_fb? (I have a preference for the second one
and will try it out)

I'm aiming to kill off registered_fb. I agree that fb_info should be used 
directly and we can avoid the ref-count. As for con2fb_map that is a 
little more complex. 

Refcounting can't be fully avoided on fb side but certainly can be on
(most of) fbcon side (except around fbcon's calls to fbops.fb_open and
fbops.fb_release).

Will attempt the fbcon change as a separate patch my ref-counted fb_info
patch will depend on. (will probably happen during week-end)

Thanks,
Bruno