-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

On Dec 12, 2006, at 12:30, Johan Wall=E9n wrote:

> the new accelerated Radeon console driver for Radeon video cards seems
> to work nicely, but there are two strange things going on with Radeon
> Mobility 9700:
>
> Every time a recently current kernel boots (at least on a PowerBook
> G4), a lot of noise that seems related to the last boot (?) is shown
> before the latter part of the kernel dmesg output is shown (the noise
> is quite difficult to describe),

Yeah, radeonfb needs to clear the screen when switching modes. It's on=20=

my todo-list, along with streamlining the switching process itself,=20
disabling hardware-accelerated drawing on R3x00 etc.

> and sometimes (this indeed only happens sometimes; something like=20
> every (50-100)th boot), the console gets messed up completely.  When=20=

> this noise is really bad, the text
> shown on the screen is fuzzy (fluctuates a lot), and there seems to be=20=

> some phantom text scrolling in the background.

Ouch, this sounds like radeon somehow managed to mess up the video mode=20=

( as in output doesn't match your display ). I've never seen this on my=20=

iBook but I'm not sure I booted it more than 50 times with radeonfb.

> When X starts, it is in these cases completely messed up: the=20
> scrolling of random text
> continues, and the stuff X is supposed to show is fluctuating rapidly.

Hmm, this is something that needs to be resolved by someone with access=20=

to ATI documentation.

> So the two strange things are:  (1) why is the radeon driver=20
> correlating
> so strongly with the last reboot,

Because nobody bothered to clean out video memory. Trivial to fix, I=20
just need to get around to do it.

>  and (2) what could be done to really make it work nicely?

If you can get access to ATI's docs find out what exactly they changed=20=

in the colour expansion hardware with the R3xx.

> As an average user, it seems like radeonfb is left in a strange state=20=

> in some cases, so some initialisation might be missing?

Likely. It was committed for some embedded MIPS machine and I only made=20=

it work on macppc ( or rather - made it work on my iBook ).=20
Unfortunately the person who wrote radeonfb is no longer with the=20
NetBSD project.

> (P.S.  I do have a nonstandard version of
> xsrc/xfree/xc/programs/Xserver/hw/xfree86/ati/radeon_driver.c,
> so I'll better check that radeonfb* includes all the relevant changes
> before I start looking for errors.)

That shouldn't really matter to the kernel driver.

have fun
Michael
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (Darwin)

iQEVAwUBRX77ZcpnzkX8Yg2nAQI7qgf/XtlHWgGg4vB9Tkds6FGFRcD92X6WIMT2
WTvG3A8+3E2fpZjCYYmXWSmITB+5yYDWtrV7Kvo2UlfWYEO/Q143YBr6hbv1bO3d
DYYnIg+SpnyZExz8BPPefmzoP5zPkuXITt7Vq8LvnebvmTqQ7v/VwJlXresnpULu
fZTQbcFr53HvF+UABPLUeZ2Ur4DTfnpZBTcb+BUfTjfSBUqUYnKqDIF+Whmhv6wr
7IkiI2xO7eZ0pfmZ8apsIV9YesDIqthKLKfIoON325qO0PFGqK1GNJqIRJKyJdIR
UsYcIQgEzkElW/G82B/vaCZ6PElUu8eS6EgslMi2UmK741C7hmAT8g=3D=3D
=3DUOJv
-----END PGP SIGNATURE-----