TC1.1 Live CD - ATI Rage Mobility chipset missing in R128 driver

Adam Jackson ajax at redhat.com
Mon Oct 18 17:08:31 UTC 2010


On Thu, 2010-10-14 at 15:09 -0500, Robert G. (Doc) Savage wrote:

> In /usr/share/hwdata/pci.ids I found:
> 
>         4c45  Rage Mobility M3 AGP
>         4c46  Rage Mobility M3 AGP 2x
>                 1002 0155  IBM Thinkpad A22p
>                 1014 0155  IBM Thinkpad A22p
>                 1028 00b1  Latitude C600
> 
> Do you suppose the aty128fb.ko kernel driver contains the correct
> parameter info for one of the device numbers but not the other (e.g.
> 1002 but not 1014)? The display tearing problem I'm seeing goes back a
> long time (prior to F7 for sure). Could we have been missing half a loaf
> all this time?

Well, let's ask the driver:

% modinfo aty128fb | grep -i 4c4
alias:          pci:v00001002d00004C46sv*sd*bc*sc*i*
alias:          pci:v00001002d00004C45sv*sd*bc*sc*i*

sv* means it will bind to any subvendor id (1002 or 1014 in this case).
So, no, I don't think that has any relevance.

> James MacKenzie has experienced the same problem on his A22p. The only
> solution I've found is the xorg.conf file I attached to my first msg.
> Before I hacked that together I had to use the VESA driver, which is now
> abominably slow.
> 
> A key number in my xorg.conf file is the hsync value found by the gtf
> utility: 74.52 kHz. That's 501 Hz higher than the default value listed
> in Xorg.0.log. The difference is blocking a 1600x1200 display.

We're carrying a patch to the r128 driver that, if we don't get EDID
from the display, will attempt to use the preprogrammed panel size to
compute appropriate sync ranges for mode validation.  The old version of
it used the CVT reduced blanking formula since that's common for laptop
panels, but:

% cvt -r 1600 1200
# 1600x1200 59.92 Hz (CVT 1.92M3-R) hsync: 74.01 kHz; pclk: 130.25 MHz
Modeline "1600x1200R"  130.25  1600 1648 1680 1760  1200 1203 1207 1235 +hsync -vsync

Which is where that 74.01 is coming from.  The patched version of the
driver I gave you switches to the normal CVT formula, which seems
reasonable since the r128 hardware predates CVT-R by quite a bit:

% cvt 1600 1200 
# 1600x1200 59.87 Hz (CVT 1.92M3) hsync: 74.54 kHz; pclk: 161.00 MHz
Modeline "1600x1200_60.00"  161.00  1600 1712 1880 2160  1200 1203 1207 1245 -hsync +vsync

Which should be large enough to fit the 1600x1200 mode in the default
mode list.

However the X log you emailed me still shows the driver using the old
74.01kHz max hsync, so I suspect you weren't testing what you thought
you were testing.

- ajax



More information about the test mailing list