Too many bugs to isolate a bug when swithching from Alt-F1 to Alt-F7

dufresnep · 2011-11-21 23:31:58

First, I consider Linux to be the less buggy OS out there.
Well, Windows would be less buggy, but it does not really exist, since you have to pay to use it.
Still, I use to consider Linux very buggy. This opinion I hold true more than ever.

Here is my box:
A P5GZ-MX motherboard (intel 945GZ), with 1 GB Memory, and an Celeron D @ 2.66 GHz.
Now, the integrated VGA connector is badly soldered, so that I used to put something to hold the cable up to have an image.
This is not the monitor Cable that is broken (I tried an other one and had the same problem).
When this was not enough anymore, I put my box upside-down (reversing side of DVD and HD too), that way gravity would hold the cable up (eh... down).
When this was not enough anymore, I put an old Radeon 7200 (an R100 generation ATI PCI card).

Then, my problems began (eh... I means continue as usual).

Let's begin the story with when I installed Arch Linux (again, well the first time with the Radeon 7200 card), about three weeks ago.

Well, I set the BIOS to prefer the PCI card, to internal one, for what should be now obvious reason.
I installed Xorg... and things were not so bad. Except, one in about 3 times, when switching back from (Ctrl-Alt-F1 console) to (Alt-F7 Xorg).
I would have THE bug. Well, basically, one or many window would be empty. Mouse would move, but not interact with anything on screen. Sometimes mouse is jailed inside a window (cannot move out of a window). Recently (maybe just because Firefox not occupying full screen), I would see that some of the text from (Ctrl-Alt-F1 text console) would appears (very brokenly) under the windows (about 4 times on the screen, in green on black).
I still can swith to text mode with Ctrl-Alt-F1 after bug, and then can use the console.
Nothing interesting would show in Xorg.0.log after the bug happened.

I thought maybe I should try a different window manager. I tried icewm rather than OpenBox (or twm).
At first, I thought OpenBox was the bug, because icewn seems not exhibit THE bug.
But then, I came to the conclusion, that icewm just had made the bug happen one in about twenty (rathen than in three).

So I thought that I should install fbcon and vesa driver too (there were some warning they were not installed).
No change.
Then I thought, maybe I should try with the vesa driver rather than the radeon one.
Surprise:
(EE) VESA(0): V_BIOS address 0x3d810 out of range
(EE) screen found, but none have a usable configuration

Then I thought, I should try with latest kernel. Downloaded it.
Still the bug... with 3.1.1-1-Arch (32 bits BTW).

Ok, so let's stick with radeon. Reading on possible Kernel Mode Setting with Radeon, I created /etc/modprobe.d/radeon.conf with:
options radeon modeset=1

Well, just restarting with startx, show in Xorg.0.log, that KMS was still active.
So, I thought, let's reboot and see.
But I did not see.
Well, my GRUB menu, have NO vga= on the kernel line. So that I boot in plain text (as far as I know).
At first, Linux seems to boot fine, but then not much before when login prompt would appears... screen became totally black.
I realized I was still able to log in without seeing what I do, and 'reboot'.
Ok, try the fallback... not working either.
So basically, my computer still was working, except I would not see anything.

Then I began to think... well, maybe I could try to connect the monitor in the 'badly soldered' internal VGA connector.
Surprise, surprise: the login prompt! Ouf!
Well, strangely I now need to hold the cable up, so that I thought: maybe this is time to put my box down side up (remember, it is already up side down). But, was too lazy for it. I just edited back /etc/modprobe.d/radeon.conf with modese=0, put back the monitor in the Radeon 7200 card, and rebooted.

Things were 'fine' again.
That's when, I thought I should write a message for maybe asking help, but frankly more to publicly complain.

roygbiv · 2011-11-21 23:35:37

TLDR, but it seems this belongs elsewhere on the forum..

jasonwryan · 2011-11-21 23:39:42

dufresnep wrote:

That's when, I thought I should write a message for maybe asking help, but frankly more to publicly complain.

Yes, the world needs more of that...

Have you even entertained the notion that you may be dealing with a hardware problem?

This should be TGN'ed.

karol · 2011-11-22 00:05:55

jasonwryan wrote:

Have you even entertained the notion that you may be dealing with a hardware problem?

He posted in 'Kernel & Hardware' subforum :-)

dufresnep · 2011-11-22 02:05:22

If I considered it is an hardware problem?
No, except for the badly soldered VGA connector.

My hypothesis goes like this for now:
-Maybe the VESA driver not working is caused by the new Version 12 ABI changes that happens some days ago http://cgit.freedesktop.org/xorg/driver … -vesa/log/
-I suspect that the fact that when the radeon driver have the option to disable KernelModeSetting my Box choose to switch to internal VGA is due to a bug in DRM Direct Rendering Manager, that is now in the kernel since 2.6.39 (according to http://www.phoronix.com/scan.php?page=n … &px=OTI0OQ ) [which explains why it happens even before I start X]
-I suspect that the occasionnal bug from Ctrl-Alt-F1 console to Alt-F7 Xorg, is due to KernelModeSetting bug

So that is why I am writing in the 'Kernel' section. ;-)

dufresnep · 2011-11-23 06:23:09

Revised hypothesis:

I now believe that the bug while switching from console back to Xorg is not hardware related, and should probably happens to anyone. Also I have a new procedure that seems to always trigger it:
1) on console (Ctrl-Alt-F1) run startx (without ending & )
2) You should be in Xorg, go back to console with Ctrl-Alt-F1
3) type Enter
4) Type Alt-F7 to go back in Xorg... hopefully for me in bogus state
Now, the message that seems to be written in some titlebars are the same, I think, that I would get when I go back to console and do: Ctrl-C (2 or three times)... then after the bug I would see some:
xterm: Ressource temporarily not available (or something very similar)

About V_BIOS address 0x3d810 out of range I am now convinced it is while loading (initializing) int10 module,
The sad part is that if I add Option "NoInt10" "True" to vesa driver section (doc seems to wrongly suggest to put it in Server section)... then yes, the out of range does not appears, but vesa will continue to unload itself... and finally no usable screen found message would continue to appears. So I suppose Int10 module have become mandatory to vesa rather than optional.

Arch Linux

#1 2011-11-21 23:31:58

Too many bugs to isolate a bug when swithching from Alt-F1 to Alt-F7

#2 2011-11-21 23:35:37

Re: Too many bugs to isolate a bug when swithching from Alt-F1 to Alt-F7

#3 2011-11-21 23:39:42

Re: Too many bugs to isolate a bug when swithching from Alt-F1 to Alt-F7

#4 2011-11-22 00:05:55

Re: Too many bugs to isolate a bug when swithching from Alt-F1 to Alt-F7

#5 2011-11-22 02:05:22

Re: Too many bugs to isolate a bug when swithching from Alt-F1 to Alt-F7

#6 2011-11-23 06:23:09

Re: Too many bugs to isolate a bug when swithching from Alt-F1 to Alt-F7

Board footer