xenial: xorg problems

[Jason W]

Of course you can compile a kernel with dCore, that is one of the tests I do to make sure the toolchain related packages are working.  But it is not how the official dCore kernel is made nor can one simply sce-import a new kernel.

[sm8ps]

I believe that I am bitten by this bug, too. The machine in question has a Intel 855GM graphics card.
. When I have "xorg-intel" as well as "graphics-4.2.9-tinycore" in 'sceboot.lst' then the system will show a totally black screen.
. When I only load xorg-intel during booting then the X-server crashes. The following error message remains on the screen:
"
modprobe: module i915 not found in modules.dep
modprobe: module fbcon not found in modules.dep
intel: waited 2020 ms for i915.ko driver to load
modprobe: module i915 not found in modules.dep
error setting MTRR (base = ..., size= ..., type = 1) Invalid argument (22)
xinit: connection to X server lost
(...)
"
Either way the system can be reached via SSH.

Does this seem to fit in here? Is there any hope for finding a solution? Is the situation maybe different on Debian? Trusty does not seem to work any better, does it?

[nitram]

This fixed my black screen boot on older Intel graphic chip, lost most support in more recent Debian/Ubuntu releases but now runs Xenial and Stretch fine. For me disabling all acceleration (DRI false) overcomes black screen, 2D performance remains decent. From your error it looks like you still need to load the graphics extension, as Xorg is seeking the module. This should allow better resolutions than falling back to xorg-vesa:
xorg-intel
graphics-4.2.9-tinycore

Then try this /etc/X11/xorg.conf file. Don't just exit WM and  startx  , save the xorg.conf file followed by full reboot:

Code: [Select]

Section "Monitor"
Identifier   "Monitor0"
EndSection

Section "Device"
Option      "DRI"    "false"
Identifier  "Card0"
Driver      "intel"
BusID       "PCI:0:1:0"
EndSection

Section "Screen"
Identifier "Screen0"
Device     "Card0"
Monitor    "Monitor0"
SubSection "Display"
Depth 16
#Modes "1024x768"
Modes "1280x800"
EndSubSection
EndSection


[sm8ps]

Thanks, Nitram, but I have to give up. Nothing whatsoever can get this *!^'- card to cooperate with Xorg. I tried your xorg.conf but it did not improve matters. I could get the graphics extension to not blank my screen but indeed add a higher screen resolution but that is all I have achieved. Xvesa does work but I want to use Wacom digitizer and some compositing capabilities for this project.
This is really frustrating! I have read somewhere that newer kernel version (4.5+) have less bugs in the i915 module so I shall give it another try when Core upgrades its kernels.

OK, I am stupid. I used to issue ''startx'' only and that gave me all the errors described here. For my project, I want to use the tablet as an overlay screen bridging other computers to a beamer so as to allow writing over the laptop screen. For that I intend to run a "naked" X server without window manager. That is why I did not go any further than the bare minimum.

After the waves of quantom randomness had faded a bit, I installed Flwm and launched it and sure enough, my despair turned to joy. Isn't that a neat story for the actual season? 

Peace, everybody!

[sm8ps]

Some further remarks for fellow i915 users. The machine in question is quite dated and so I thought that setting DRI to false was appropriate. After quite some bisecting it turned out that this is not the problem at all. Instead, it is mandatory to have the i915 module use the option "invert_brightness=1" in '/etc/modprobe.d/i915.conf'. With this option, not even an 'xorg.conf' is necessary.

Upon loading graphics-4.2.9-tinycore (yes, back on dCore-xenial again), it spits out four DRM-errors about pipe underrun but that does not seem to be harmful -- at least so far it is not

All these nice discoveries have one tiny issue: they do only work when loading graphics-4.2.9-tinycore manually but not when having it loaded automatically through 'sceboot.lst'. I may well add "/usr/bin/sce-load graphics-4.2.9-tinycore" to '~/.profile' but that seems quite quirky. Plus it needs an additional "sleep 5" in order to function properly. -- Anybody got any better idea how to overcome this or why this is even necessary?

Cheers!

[nitram]

Congratulations, now we know how you spent your Christmas!

Guess it's a race condition, as per the documentation, have you tried adding  i915.invert_brightness=1  as a kernel parameter instead?
https://wiki.archlinux.org/index.php/backlight#Inverted_Brightness_.28Intel_i915_only.29

Also found this, you've probably read through all this already today:
http://people.skolelinux.org/pere/blog/Fixing_the_Linux_black_screen_of_death_on_machines_with_Intel_HD_video.html

Once you've finalized your solution i will update the wiki or feel free to update yourself so next user doesn't need to reinvent the wheel:
http://wiki.tinycorelinux.net/dcore:xorg-video_driver_packages#dcore_graphic_notes
http://wiki.tinycorelinux.net/dcore:installations#dcore_hardware_installation_notes

Thanks.

[Jason W]

JLS - I misunderstood the question of why the dCore kernel is not compiled with dCore.  The aim was to be able to share kernel modules between Core and dCore.  Since there is a config change between Core and dCore in regards to DEVTMPFS since dCore now has to accomodate systemd, the build is not exactly the same.  Perhaps one day if the kernel config is again the same, the ability to share kernel modules from Core to dCore and vice versa can be realized.

[sm8ps]

Thanks for the pointers, Nitram! Adding the kernel parameter to the boot stanza works well. In fact, as mentioned in the second of your links at the very end, it is enough to add "acpi_backlight=vendor". This also eliminates the need for loading the graphics extension manually. Yay!

I have updated both sections in the wiki.