Fw: [framebuffer] broken again :/

Fw: [framebuffer] broken again :/

[Andrew Morton]

Begin forwarded message:

Date: Wed, 21 Jul 2004 16:00:44 +0200
From: Pawe³ Sikora <pluto@pld-linux.org>
To: Andrew Morton <akpm@osdl.org>
Subject: [framebuffer] broken again :/


Hi,

# vanilla 2.8.8rc2...

I switch *all* consoles to 1024x768(75Hz,32bpp) (rivafb) mode.
(http://149.156.124.14/~pluto/tmp/kernel/step1.png)
Even without launching a bootsplash while switching
to another console I observe below effect:
(strange resolution and horrible aspect ratio)
http://149.156.124.14/~pluto/tmp/kernel/step2.png
I switch consoles again to 1024x768(75Hz,32bpp) mode
and get an oops:
http://149.156.124.14/~pluto/tmp/kernel/step3.png

# vanilla 2.6.7...

I switch *all* consoles to 1024x768(75Hz,32bpp) (rivafb) mode.
While switching to another console everything is all right.
(no resolution/aspect changes, no oops).

# summary

Framebuffer in 2.6.8-xx is broken (again) :/
Writing to lkml == Writing to /dev/null :/
Please, forward this email to fb-related developers.
At now, PLD-Linux rolls back to 2.6.7(+security fixes).

2.6.8 is unusable :(

-- 
/* Copyright (C) 2003, SCO, Inc. This is valuable Intellectual Property. */

                           #define say(x) lie(x)


-------------------------------------------------------
This SF.Net email is sponsored by BEA Weblogic Workshop
FREE Java Enterprise J2EE developer tools!
Get your free copy of BEA WebLogic Workshop 8.1 today.
http://ads.osdn.com/?ad_idG21&alloc_id\x10040&op=click

Re: Fw: [framebuffer] broken again :/

[Antonino A. Daplas]

On Thursday 22 July 2004 13:45, Andrew Morton wrote:
> Begin forwarded message:
>
> Date: Wed, 21 Jul 2004 16:00:44 +0200
> From: Pawe³ Sikora <pluto@pld-linux.org>
> To: Andrew Morton <akpm@osdl.org>
> Subject: [framebuffer] broken again :/
>
>
> Hi,
>
> # vanilla 2.8.8rc2...
>
> I switch *all* consoles to 1024x768(75Hz,32bpp) (rivafb) mode.
> (http://149.156.124.14/~pluto/tmp/kernel/step1.png)
> Even without launching a bootsplash while switching
> to another console I observe below effect:
> (strange resolution and horrible aspect ratio)

The problem is that the bootsplash code uses vc_resize() to create a 'box' 
effect (small viewport within the display).  Calling vc_resize() has no 
effect in 2.4 as fbcon does not have a con_resize hook, but in 2.6, calling 
vc_resize() is equivalent to doing an fbset.  This produces the unusual 
aspect ratio and resolution.
 
> http://149.156.124.14/~pluto/tmp/kernel/step2.png
> I switch consoles again to 1024x768(75Hz,32bpp) mode
> and get an oops:
> http://149.156.124.14/~pluto/tmp/kernel/step3.png

The oops here is from the bootsplash code, perhaps brought about  by the 
unexpected mode change.

>
> # vanilla 2.6.7...
>
> I switch *all* consoles to 1024x768(75Hz,32bpp) (rivafb) mode.
> While switching to another console everything is all right.
> (no resolution/aspect changes, no oops).

Yes, fbset calls are ignored in pre 2.6.7-mm1 kernels, so this is expected.  
Latest kernel snapshots already handle fbset requests.

One solution for the bootsplash code in 2.6 is for it to disable the 
con_resize/fbcon_resize hook in fbcon.c. You should let the bootsplash 
developers know about this complication. 

For rivafb, enable DDC/I2C support in your kernel config and try booting with 
the 'strictmode' option. If you do not have a GTF capable display and 
assuming your monitor can do DDC and has a valid EDID block, the 'strictmode' 
option will force rivafb to restrict itself to using video modes creaed from 
the EDID block. This, hopefully, avoids the unusual aspect ratio and 
resolution change.

> Framebuffer in 2.6.8-xx is broken (again) :/
> Writing to lkml == Writing to /dev/null :/

Next time, write to fbdev-devel list for any fbdev/fbcon related problems.

Tony




-------------------------------------------------------
This SF.Net email is sponsored by BEA Weblogic Workshop
FREE Java Enterprise J2EE developer tools!
Get your free copy of BEA WebLogic Workshop 8.1 today.
http://ads.osdn.com/?ad_idG21&alloc_id\x10040&op=click