[SOLVED] Illegible text + narrow horizontal distortion with i915 Intel driver.

edbarx · 09-27-2016, 01:05 AM

The Problem:
a) Immediately after kernel boot, text is unreadable with successive pixel lines shifted to the left by a constant number of pixels. This continues until the text is reformatted by custom font settings.

b) In Firefox smooth scrolling causes narrow horizontal distortion. Videos can be played smoothly but I haven't done any testing for long as I want to avoid to cause overheating in case the graphics load is being passed to the CPU.

c) Kernel complaints requesting a UEFI/BIOS upgrade.

Code:

[    7.471266] [drm] Wrong MCH_SSKPD value: 0x16040307
[    7.471269] [drm] This can cause pipe underruns and display issues.
[    7.471270] [drm] Please upgrade your BIOS to fix this.
[    7.489912] fbcon: inteldrmfb (fb0) is primary device
[    7.598128] EFI Variables Facility v0.08 2004-May-17
[    7.724203] hp_accel: laptop model unknown, using default axes configuration
[    7.729494] jmb38x_ms 0000:02:00.3: enabling device (0000 -> 0002)
[    7.757025] [drm:ilk_display_irq_handler] *ERROR* Pipe A FIFO underrun
[    7.759849] [drm:cpt_set_fifo_underrun_reporting] *ERROR* uncleared pch fifo underrun on pch transcoder A
[    7.759851] [drm:cpt_serr_int_handler] *ERROR* PCH transcoder A FIFO underrun
[    7.768795] input: PC Speaker as /devices/platform/pcspkr/input/input14
[    7.856025] lis3lv02d: 8 bits 3DC sensor found
[    7.919352] cfg80211: Calling CRDA to update world regulatory domain
[    8.144477] Console: switching to colour frame buffer device 170x48
[    8.152154] i915 0000:00:02.0: fb0: inteldrmfb frame buffer device
[    8.152159] i915 0000:00:02.0: registered panic notifier
[    8.265292] pstore: Registered efi as persistent store backend
[    8.294325] ath9k 0000:03:00.0: enabling device (0000 -> 0002)
[    8.302137] ath: EEPROM regdomain: 0x60

My graphics hardware:

Code:

00:02.0 VGA compatible controller: Intel Corporation 2nd Generation Core Processor Family Integrated Graphics Controller (rev 09)

Another name for the above "may be" Intel HD Graphics 2000.

Processor:

Code:

Architecture:          x86_64
CPU op-mode(s):        32-bit, 64-bit
Byte Order:            Little Endian
CPU(s):                4
On-line CPU(s) list:   0-3
Thread(s) per core:    2
Core(s) per socket:    2
Socket(s):             1
NUMA node(s):          1
Vendor ID:             GenuineIntel
CPU family:            6
Model:                 42
Model name:            Intel(R) Core(TM) i5-2450M CPU @ 2.50GHz
Stepping:              7
CPU MHz:               800.000
CPU max MHz:           3100.0000
CPU min MHz:           800.0000
BogoMIPS:              4988.74
Virtualization:        VT-x

Linux Distribution:
Debian/Devuan Jessie

Kernel: 3.16.0-4-amd64 (is the best performing kernel with respect to graphics)

mrmazda · 09-27-2016, 02:11 AM

http://ark.intel.com/products/53452/...up-to-3_10-GHz shows "Intel® HD Graphics 3000" may be more likely than 2000.

Which DM/DE are you running?

If perchance Devuan is using Plymouth, I'd try disabling or uninstalling it.

http://www.phoronix.com/scan.php?pag...ndon-Intel-DDX suggests it may be worth giving the modesetting driver a try. With the 1.16.4 server in Jessie the modesetting driver remains a separate package that will normally be used automatically if the intel driver is uninstalled. In server 1.17.0 the modesetting driver was merged directly into the server. All my Jessie installations on Intel gfx are Core2Duo (older than yours around 3-4 years), pure Debian and working acceptably full time both on the vttys and running Firefox.

edbarx · 09-27-2016, 06:17 AM

I am using XFCE 4.10. Plymouth is not installed according to "apt-cache policy". I did not understand what you mean by modesetting driver. I remember I tried nomodeset as a kernel parameter with horrible results. The display was never readable with nomodeset forcing me to turn off the machine.

mrmazda · 09-27-2016, 07:12 PM

# grep RETT /etc/os-release
PRETTY_NAME="Debian GNU/Linux 8 (jessie)"
# apt-cache search r-xorg-video | egrep -v 'free|debug' | egrep 'ntel|veau|ati|amd|modes'
xserver-xorg-video-ati - X.Org X server -- AMD/ATI display driver wrapper
xserver-xorg-video-intel - X.Org X server -- Intel i8xx, i9xx display driver
xserver-xorg-video-modesetting - X.Org X server -- Generic modesetting driver
xserver-xorg-video-nouveau - X.Org X server -- Nouveau display driver
#
Above you see listed the four mainstream gfxchip drivers for Xorg provided by standard Jessie sources.

In Stretch and Sid there are only three, because the generic modesetting driver was merged into the server.

KMS is short for Kernel Mode Setting. KMS was implemented in Linux around eight years ago. Since then it has become default, with Xorg video drivers rewritten to accommodate it, replacing Xorg video drivers that had previously operated entirely in userland.

The nomodeset cmdline parameter exists primarily as a troubleshooting enabler, but is required for normal functionality by non-mainstream Xorg video drivers (e.g. MGA) and by certain proprietary drivers (e.g. NVidia). Nomodeset on cmdline will disable KMS, thus preventing the four mainstream gfxchip Xorg video drivers from functioning.

The URL provided in my previous thread post was meant to enlighten the reader that xserver-xorg-video-intel, as are xserver-xorg-video-ati and xserver-xorg-video-nouveau, is in the process of being deprecated in favor of xserver-xorg-video-modesetting. The newer's suitability is in the process of maturing, so with some gfxchips it works as well or better, and with others not. Only by testing on your own hardware can its suitability be determined.

edbarx · 09-28-2016, 01:07 AM

I posted my question here in the hope of receiving information that can resolve my issue. I run the following as a normal non-privileged user and got exactly the same results:

Code:

edbarx@edbarx-pc3:~$ apt-cache search r-xorg-video | egrep -v 'free|debug' | egrep 'ntel|veau|ati|amd|modes'
xserver-xorg-video-ati - X.Org X server -- AMD/ATI display driver wrapper
xserver-xorg-video-intel - X.Org X server -- Intel i8xx, i9xx display driver
xserver-xorg-video-modesetting - X.Org X server -- Generic modesetting driver
xserver-xorg-video-nouveau - X.Org X server -- Nouveau display driver

I am interested in a workaround that allows my graphics hardware use its built in 3D acceleration to eliminate visual distortion.

If there is no solution, it means, only MS Windows will run well on this machine although this machine, can still be used as a server for a thin client like the Pi, but that incurs more expense which I want to avoid.

It seems these days GNU/SystemD/Linux is slowly ceasing to be an alternative for those who want to save on hardware. The fashion is to merge diverse parts together adding more to bloat just like what MS Windows used to do.

The irony of all this is, that I purchase a laptop with integrated Intel graphics for exactly the reason that Intel is said to play nice on Linux! Sadly, this is the SECOND TIME I have been bitten by Intel hardware. The first one was the INFAMOUS BACKLIGHT ISSUE that unashamedly lasted for a WHOLE YEAR! But, there was a workaround which I used and the problem was like non-existent. However, this time, it seems, the problem will not be resolved. Filing a bug report with freedesktop.org resulted in a request for me to upgrade to kernel 4.7 or better. In the hope of getting a properly working display, I compiled a vanilla kernel on this same machine, but the result was even more graphics issues!

The first strategy at a solution that I tried was to attempt to upgrade the UEFI. I tried to do it from three operating systems failing every time. I used MS Windows XP, MS Windows 7 and Debian Jessie. If anyone knows of a solution path that I can follow, please post.

mrmazda · 09-28-2016, 02:40 AM

Quote:

Originally Posted by edbarx

I am interested in a workaround that allows my graphics hardware use its built in 3D acceleration to eliminate visual distortion.

Exactly what graphics hardware is that? AFAICT, i5-2450M, Intel Corporation 2nd Generation Core Processor Family Integrated Graphics Controller (rev 09) and laptop model unknown don't tell a complete graphics hardware story.

Are you sure eliminating visual distortion is not possible without xserver-xorg-video-intel installed? I'm using kernels older than 4.7 and servers older than 1.18.4 on Haswell and older Intel gfx (and with a maximum of 2 CPU cores). I can't recall seeing any kind of unexpected distortion scrolling in any Mozilla browser or on the vttys on any of various LCD displays using any of the common modes from 1280x1024 through 1440x900, 1680x1050, 1920x1080 and 1920x1200.

edbarx · 09-28-2016, 06:07 AM

Laptop Model: HP Probook 4540s
CPU: Intel(R) Core(TM) i5-2450M CPU @ 2.50GHz
RAM:

Code:

root@edbarx-pc3:/home/edbarx# free -m
             total       used       free     shared    buffers     cached
Mem:          3845        875       2969         64         36        419
-/+ buffers/cache:        419       3425
Swap:         3811          0       3811

GPU:

Code:

root@edbarx-pc3:/home/edbarx# lspci | grep -i vga
00:02.0 VGA compatible controller: Intel Corporation 2nd Generation Core Processor Family Integrated Graphics Controller (rev 09)

The above is what I get with "lspci".

Distribution: Debian/Devuan 64 bit Jessie running with sysvinit. SystemD exhibited exactly the same symptoms.

The kernel complains as follows:

Code:

root@edbarx-pc3:/home/edbarx# dmesg | grep -A 10 Wrong
[   15.029947] [drm] Wrong MCH_SSKPD value: 0x16040307
[   15.029951] [drm] This can cause pipe underruns and display issues.
[   15.029952] [drm] Please upgrade your BIOS to fix this.

I struggled to upgrade the UEFI for several days without success. For some weird reason both virtualised Windows installations and this Jessie (devuanised) installation failed.

ADDED LATER to add more information:
I removed xserver-xorg-video-intel without the least of effects. I am still experiencing distortion while scrolling in Firefox and garbage text immediately upon kernel boot.

EDIT 2:
This behaviour is a registered kernel bug as shown here. It was said some users succeeded to get rid of the bug using kernel version 4.3. Further towards the beginning of the thread, it was suggested to downgrade the kernel as versions prior 3.16 did not exhibit the bug.

Now, the headache is to downgrade the kernel.

mrmazda · 09-28-2016, 02:52 PM

Now that we know the laptop model we know the chipset is Mobile Intel HM76 Express (Panther Point) and the approximate hardware age is 4 years, rather than hardware newer than should be supported in your choice of distro.

If it was mine, I wouldn't try to downgrade the kernel. Instead, I'd try the LTS kernel used by Stretch, 4.4, and if that was not enough to solve the problem, I'd try Stretch's Xorg and Intel driver. I'd probably also ask on the intel-gfx mailing list, where people read who deal with Intel video issues on a daily basis. Barring a solution based there, I'd next try booting live media of a distro newer than Jessie/Devuan, such as Knoppix 7.6.1 (kernel 4.2.6), Fedora 24 (kernel 4.5.5) or Xubuntu (kernel 4.4). Video trouble shouldn't have anything to do with the init system used. At least you'd know if the problem seems limited to your choice of distro if any or all fail to exhibit the same trouble.

ondoho · 09-29-2016, 12:53 AM

just a thought - it could also be accel related.
try this.

edbarx · 09-30-2016, 01:57 AM

EDIT: Sorry for the "wild edit" but this is more relevant after I dist-upgraded.

I upgraded my Devuan Jessie installation to Devuan Testing (ASCII) with kernel 4.6.0-1-rt-amd64. I will try installing Intel proprietary driver for my current kernel.

Hopefully, the problem is resolved.

edbarx · 10-12-2016, 06:47 AM

Upgrading Debian Jessie to Devuan ASCII solved the issue. Glxinfo now lists screenfuls of information which means hardware acceleration is working. However, there is still a minor problem with the display being unreadable as soon as the kernel takes control of the machine.