Notebook reboots on intel KMS and refuse to load intel Xorg driver.

sari · 2015-09-06 14:04:58

Hello,

I have a fresh installation of Arch on a ThinkPad X220 computer.
This model has an LCD display with maximum resolution of 1366x768.
It has integrated intel graphics chip, see the output of lspci -vvkqqnn here.
First problem is that I am unable to boot but only with `nomodeset` parameter. If I don't give any, or I set `i915.modeset=1` then the computer reboots immediately when loading initramfs, otherwise while loading the system and trying to switch to framebuffer. Module i915 is compiled in the initramfs. I experienced the same with Linux Mint installer, a fresh Linux Mint install, the Arch installer, and now with a fresh Arch system.

My boot parameters are currently the following:

root=/dev/mapper/vgroup-root cryptdevice=/dev/sda2:vgroup:allow-discards init=/usr/lib/systemd/systemd ro acpi_osi="!Windows 2012" nomodeset

The module i915 is loaded later, see here the `lsmod | grep i915`.

When I start Xorg, it loads automatically the `vesa` driver, and sets the resoultion to `1024x768`, which is distorted on this screen.
Other resolutions don't appear in `xrandr` output.
See also the Xorg.log, without any custom config except keyboard layouts.

After I tried to set up an Xorg config to load the `intel` driver. See the `/etc/X11/xorg.conf.d/09-monitor.conf` here.

Using this configuration resulted a `No devices detected` error, see the log here.

I have xorg-server-utils, xorg-xinit, xf86-video-intel, intel-ucode and many other packages installed (ask if you think some other might be relevant).

Here I found a similar issue, with no solution.

Lone_Wolf · 2015-09-07 11:33:08

First problem is that I am unable to boot but only with `nomodeset` parameter.

That is the root cause of your problems, and we'll need to figure out how to solve it.

Do you have latest firmware installed ?
are you using uefi or legacy (MBR) boot mode ?

sari · 2015-09-07 16:14:04

Thanks Lone_Wolf,

Yes, now I realized that the main problem is with i915 with KMS, because Xorg will only work if KMS works. As soon as I load the i915 with KMS enabled, the system reboots, does not matter if I load it early or late.

I use MBR boot, EFI disabled.

Which firmware do you mean, the bios? I haven't flashed the bios, but I will check the version.

In the meantime I checked the EDID block, it seems ok, see here.

Also I started the kernel with late modeset and with insane debug options as it is described in Arch Wiki. After I read the journal with `journalctl -b -1 -a`. I don't see anything related, maybe because I don't know what to look for. Here is the debug log. The message "AGP: No AGP bridge found" do you think might be relevant? As I understand, the system is determined to reboot from near line #6091, is that right?

sari · 2015-09-07 16:27:00

The BIOS is not the recent, it is version UEFI 1.17 (8DET47WW) / EPC 1.08 (8ZHT24WW), while the recent is UEFI 1.40 (8DET70WW) / EPC 1.24 (8DHT34WW).

Last edited by sari (2015-09-07 16:27:17)

ewaller · 2015-09-07 18:00:12

Allow me to reaffirm what Sari asked. Do you have the latest Intel microcode installed and configured properly?

sari · 2015-09-07 18:30:01

Hi ewaller,

Yes I have installed the recent intel-ucode package, and have `../intel-ucode.img` before the initramfs. I tried everything with and without loading the microcode, just to see if this is related. In the kernel debug log you can see the following lines:

microcode: CPU0 microcode updated early to revision 0x29, date = 2013-06-12
microcode: CPU2 microcode updated early to revision 0x29, date = 2013-06-12
microcode: CPU0 sig=0x206a7, pf=0x10, revision=0x29
microcode: CPU1 sig=0x206a7, pf=0x10, revision=0x29
microcode: CPU2 sig=0x206a7, pf=0x10, revision=0x29
microcode: CPU3 sig=0x206a7, pf=0x10, revision=0x29
microcode: Microcode Update Driver: v2.00 <tigran@aivazian.fsnet.co.uk>, Peter Oruba
systemd-tmpfiles[280]: Entry "/sys/devices/system/cpu/microcode/reload" does not match any include prefix, skipping.
systemd-udevd[295]: No module matches 'platform:microcode'
systemd-tmpfiles[368]: Running create action for entry w /sys/devices/system/cpu/microcode/reload
Writing to "/sys/devices/system/cpu/microcode/reload".
Running remove action for entry w /sys/devices/system/cpu/microcode/reload

Other news: I successfully enabled i915 KMS both early and late, with the `nolapic` parameter. I am able to use now the intel Xorg driver with all the possible resolutions. However, this can not be a final solution, as now I have only one CPU core on a 4 core machine. In the BIOS all CPU hyperthreading related features are enabled.

I don't understand: why nolapic solved the problem with KMS? How I could achieve this without loosing 3 CPU cores?

I also tried parameters `noapic` and `acpi_osi=` (the latter is nothing, i.e. no os reported to the bios). These do not affect the problem.

Last edited by sari (2015-09-07 18:42:47)

Lone_Wolf · 2015-09-07 20:22:03

"AGP: No AGP bridge found"

AGP = Accelerated Graphics Port, the forerunner of PCIe . It hasn't been used in pcs for 5 years or longer.
You can ignore messages about agp .

nolapic = no l(ocal) Advanced Programmable Interrupt Controller .
Disabling it does most of the time prevent smp / multicore use.

However, the fact that nolapic makes your card usable with modesetting indicates the problem is with accessing hardware.
A common error on x86_64 processors running an x86_64 OS is that IOMMU is disabled in firmware .
(for 32-bit OS , IOMMU does nothing useful and may cause problems, so many manufacturers disable it by default).

Check your bios settings and enable IOMMU, then try booting without nolapic again.

Last edited by Lone_Wolf (2015-09-07 20:24:07)

sari · 2015-09-07 22:06:36

Thanks!

I don't find such parameter in the BIOS setup, actually there are very few parameters.
These are:
Core Multi-Processing: Enabled
Intel Hyper-Threading Technology: Enabled
Intel AMT Control: Enabled
CIRA Timeout: 0
Console Type: VT100+

If I check in dmesg for iommu (dmesg | grep -i iommu) I get this:

[    0.357148] dmar: IOMMU 0: reg_base_addr fed90000 ver 1:0 cap c0000020e60262 ecap f0101a
[    0.357159] dmar: IOMMU 1: reg_base_addr fed91000 ver 1:0 cap c9008020660262 ecap f0105a

Does it confirm if IOMMU is enabled?

Lone_Wolf · 2015-09-08 11:54:15

It should be present.

Sofar everything we found points at a low-level problem very close to the hardware, I think it's time for you to update the bios firmware to latest version.

sari · 2015-09-08 21:30:01

I updated the BIOS to the recent version (UEFI 1.40 (8DET70WW) / EPC 1.24 (8DHT34WW)). The problem still persists.

Lone_Wolf · 2015-09-09 21:47:07

Ok, let's try getting more info from linux .

boot with the "insane debug options" , but remove "nomodeset i915.modeset=0" from the command line and post the log.
Comparing the 2 logs may give some clues.

some more general questions :
full lspci -k output please
Is archlinux the only OS on this laptop ?
Please try to verify if this laptop has a 32-but uefi or 64-bit uefi .

Last edited by Lone_Wolf (2015-09-09 21:47:25)

sari · 2015-09-11 18:47:18

Hey Lone_Wolf,

Thanks for your message.

Here is an insane debug log from a boot with `nolapic` and with early KMS (when kms works, have good resolution, but only one CPU core). And here is the version when `nolapic` does not present, so I have all the 4 cores, but the KMS enabled late, and then the computer reboots. In this latter version nomodeset option was present, and the i915 module was not included in the initramfs, otherwise the reboot happens in initramfs stage, and no log could be saved to the disk (only printed on the screen). So KMS enabled late, when the i915 module loads, and the line `options i915 modeset=1` in `/etc/modprobe.d/i915.conf` ensure KMS is enabled. This could be suppressed only with `i915.modeset=0` parameter.

And here you can see the output of `lspci -k`.

Yes, Arch Linux is the only operating system installed. How to check if it is 32 or 64 bit UEFI? I think it should be 64 bit. Only legacy mode is enabled in BIOS, and boots with syslinux.

Last edited by sari (2015-09-11 18:54:47)

dragonn · 2015-09-12 08:39:41

You are booting Arch in legacy mode on your system? Did you try to boot normal in UEFI mode?

Lone_Wolf · 2015-09-13 21:03:15

You accidentally posted the same log twice, both are with nolapic and have same timestamp .

The lspci -k output is weird, it lists no vga cards .

Some systems with 64-bit processors have a 32-bit UEFI , and this can lead to problems under linux. No clue how to check it, except maybe asking the manfacturer of the motherboard.

I think it's time to try out UEFI-boot .
change to uefi boot in bios, then try booting the arch install CD (it supports both bios & UEFI-boot afaik).

Arch Linux

#1 2015-09-06 14:04:58

Notebook reboots on intel KMS and refuse to load intel Xorg driver.

#2 2015-09-07 11:33:08

Re: Notebook reboots on intel KMS and refuse to load intel Xorg driver.

#3 2015-09-07 16:14:04

Re: Notebook reboots on intel KMS and refuse to load intel Xorg driver.

#4 2015-09-07 16:27:00

Re: Notebook reboots on intel KMS and refuse to load intel Xorg driver.

#5 2015-09-07 18:00:12

Re: Notebook reboots on intel KMS and refuse to load intel Xorg driver.

#6 2015-09-07 18:30:01

Re: Notebook reboots on intel KMS and refuse to load intel Xorg driver.

#7 2015-09-07 20:22:03

Re: Notebook reboots on intel KMS and refuse to load intel Xorg driver.

#8 2015-09-07 22:06:36

Re: Notebook reboots on intel KMS and refuse to load intel Xorg driver.

#9 2015-09-08 11:54:15

Re: Notebook reboots on intel KMS and refuse to load intel Xorg driver.

#10 2015-09-08 21:30:01

Re: Notebook reboots on intel KMS and refuse to load intel Xorg driver.

#11 2015-09-09 21:47:07

Re: Notebook reboots on intel KMS and refuse to load intel Xorg driver.

#12 2015-09-11 18:47:18

Re: Notebook reboots on intel KMS and refuse to load intel Xorg driver.

#13 2015-09-12 08:39:41

Re: Notebook reboots on intel KMS and refuse to load intel Xorg driver.

#14 2015-09-13 21:03:15

Re: Notebook reboots on intel KMS and refuse to load intel Xorg driver.

Board footer