[lm-sensors] sensors-detect changed screen's vertical rate

[Jean Delvare]

Hi Numan,

On Mon, 2 Apr 2012 00:01:51 +0300, Numan DEMİRDÖĞEN wrote:
> After using sensors-detect for the first time, screen's vertical rate
> turned to 1023 from 800. The original resolution of screen is
> 1280x800 at 60Hzt but now it is 1280x1023 and I can not change it
> whatever I did. I tried xrandr to revert it to original but if I
> choose a different rate other than 1280x1023, resolution looks
> horrible. It looks like its EDID was also changed.

I'm very sorry about that. Needless to say sensors-detect isn't
supposed to do that. This is the first report ever of EDID corruption,
if that's what it is. This could be caused by a specific uncommon chip
on one of your graphics chip I2C buses, or by a bug in the graphics chip
driver.

> Could you help me to solve this issue?

I'll do my best.

> uname -a
> Linux else 3.0.0-17-generic #30-Ubuntu SMP Thu Mar 8 17:34:21 UTC 2012
> i686 i686 i386 GNU/Linux

For the records, please also tell us:
* Which version of sensors-detect you used?
* Did you leave the default answer to every question, or did you ask
  for probing which was disabled by default?

I see you're using the radeon driver on a RV380 chip. Do I understand
correctly that this is a laptop and the screen you're talking about is
the laptop's own screen? Fujitsu-Siemens Amilo Pro V2045, right?

It would be great if you could dig your logs for a dump of your EDID
_before_ the corruption. The radeon driver logs this
to /var/log/Xorg.0.log. If you can't find that, it would be great to
find another user with the same laptop who could provide this dump.

> sudo i2cdetect -l
> i2c-0   i2c             Radeon i2c bit bus DVI_DDC              I2C adapter
> i2c-1   i2c             Radeon i2c bit bus VGA_DDC     I2C adapter
> i2c-2   i2c             Radeon i2c hw bus MM_I2C        I2C adapter
> i2c-3   i2c             Radeon i2c bit bus MONID                I2C adapter
> 
> lspci output: http://pastebin.com/hmREjFgu
> xrandr output: http://pastebin.com/abGqSApj

First modeline looks pretty wrong, and physical dimensions too. 

You could try xrandr's --newmode option. First run xrandr --verbose to
have the detailed timings for 1280x1023, and then use --newmode with
1023 replaced by 800 (and all other relevant values shifted
accordingly.) This might at least make the machine usable until we find
how fix your EDID.

> i2cdump output: http://pastebin.com/bCtKEFMq

I'd be more interested in the output of i2cdetect for each but for now.
Also I'm not sure what you were trying to achieve with:

> i2cdump 2 0x60

I suspect you meant 0x50, not 0x60. Please try that and report. I would
also need the (bogus) EDID dump from /var/log/Xorg.0.log.

> get-edid | read | edid output: http://pastebin.com/Z5dM2WFW
> dmesg | grep drm output: http://pastebin.com/S9fF8CuA
> dmesg output: http://pastebin.com/h3UVmxVs

Meanwhile I'll refresh my knowledge of how timings are encoded in the
EDID. Hopefully we can write the right contents back to it.

Meanwhile I'll see if I can change sensors-detect to skip probing
address 0x50 on graphics chip I2C buses. It serves no purpose, so if
it's going to cause huge trouble on some systems, we should just not do
it.

-- 
Jean Delvare