Re: odd xorg-video-intel problem



On Sun, Aug 05, 2007 at 11:24:46PM +0900, Takehiko Abe wrote:
Andrew J. Barr wrote:

The file /var/log/Xorg.0.log would help with diagnostics.

only if I know how to interpret it...

There are in total 10 entries marked 'EE', all identical:

(EE) intel(0): Unable to write to SDVOCTRL_E for SDVOB Slave 0x70.

Is this normal? ... OK, silly question. Errors are not normal.

Also, consider posting to the xorg@xxxxxxxxxxxxxxxxxxxxx list, a
number of Intel employees are active there.

I think I need to do my homework before posting there. And that
would be quite a task for me. I am very new to XWindows.

Additionally, consult the driver documentation. I believe you can
turn on a debug mode in the modesetting code by specifying an option
in xorg.conf. This might help you or someone else figure out what's
wrong.

Doesn't help me. For instance, with debugging enabled, the above error
entry became:

(EE) intel(0): Unable to write to SDVOCTRL_E for SDVOB Slave 0x70.
(II) intel(0): SDVO: W: 7A 02 (SDVO_CMD_SET_CONTROL_BUS_SWITCH)
(II) intel(0): SDVO: W: 7A 02 (SDVO_CMD_SET_CONTROL_BUS_SWITCH)


My own experience with the 965:

I have a 965 connected to an widescreen 1440x900 LCD monitor via a VGA
connector. When X starts, it starts in a non-native resolution, but
after I log in the GNOME session software changes the native
resolution which I specified in the Screen Resolution control panel.

I wonder how modlines and the control panel interacts. If the setting
in a control panel overrides modeline setting, why do I need modelines?


the control panel doesn't override the modelines. It presents a list
of all available modes as determined by xorg. This includes probed
modes that may not be specified in the xorg.conf.

btw, I tried xvidtune and it appears no matter what values I specify it
says:

Sorry: You have requested a mode-line that is not possible
or not supported by your hardware configuration.


yes. don't bother with xvidtune.

using xrandr you can list and change to whatever available modes you
like. If you find the right mode in that list, then use that modes
information instead of the custom modes I'm about to talk about.


Weird.

sigh... I thought I could avoid this kind of hassle by choosing the
intel 965G.

heh. heh. well...

okay, her's what I did and it finally works beautifully on my wife's
machien with 1440x900. The problem I had is that xorg was always
choosing a mode at 59.9 hz which the monitor couldn't display properly
and it would also cause the monitor blankout sometimes. I used the
online modeline generator to get a modeline that fit the specs of the
monitor.

This custom modeline I entered into xrg.conf with a names like
1440x900_75. I think the _ was critical as it put that modeline at the
top of the list which makes it the one xorg chooses. Other attempts by
me used a more normal mode name and that put the mode further down the
list causing xorg to not choose it by default (the list I'm referring
to is the one produced by xrandr which is what xorg uses now, I
believe).

pertinent parts of xorg.conf


Section "Device"
Identifier "Intel Corporation 82845G/GL[Brookdale-G]/GE
Chipset Integrated Graphics Device"
Driver "intel"
BusID "PCI:0:2:0"
# Option "UseFBDev" "true"
Option "ModeDebug" "YES"
Option "Monitor-VGA" "19MVWK"
EndSection

note the "Monitor-VGA" option which specifies whichmonitor to
use. Though we only have one, I believe this is important.


Section "Monitor"
Identifier "19MVWk"
Option "DPMS"
HorizSync 31-81
VertRefresh 56-76
Modeline "1440x900_75" 136.49 1440 1536 1688 1936 900 901
904 940 -HSync +Vsync
EndSection

That monitor-vga option *makes* xorg look to this section for the
modes. note that the modeline should be all on one line. Also, I
speced the vert refresh just slightly over the max because its
possible for a modeline to calculate out to something other than the
exact hz setting and if that's the case, then xorg will reject it even
though its perfectly fine. for example, the 59.9 hz I mentioned
above. xorg calls it a 60hz setting, but its not, and if the probed
vertrefresh rate was 60hz then it would be rejected because the
calculated rate wouldn't fit.

I hope this ramble helps a bit.

A

Attachment: signature.asc
Description: Digital signature