Xorg with Skylake graphics hangs

deathangle908 · 2015-12-13 15:05:20

I've bought msi gs40 that has intel i7 6700HQ which means integrated graphics hd 530. After a day of struggling I was able to start xorg with HD530. For this I addes i915 to initramfs and i915.preliminary_hw_support=1 to kernel params. But xorg hangs in the end of the day, sometime right after start, sometimes after a few hours of using. I tried archlinux with:

xorg 1.6, 1.7, 1.8
kernels 4.2.5 ( from arch repository), 4.3 (aur linux-pf), 4.4 (aur linux-git), 4.1-13 lts, 4.1.15 lts
cinnamon, plasma5
xf86-video-intel (from arch repo), xf86-video-intel-git (aur)
And another os (Xubuntu, even Linux Mint failed)

Everything posted above failed. Windows runs just fine so it's not hardware issue. Everything just freezes after some time of using and I get glib stacktrace in journalctl (meanwhile xorg.log and dmesg are fine) Sometimes I get "couldn't open dis display :0" in dmesg and applications refuse to start. Kde shows "Bad windows", cinnamon show another errors etc. Sometimes I get triangles in windows. Sure I can kill xorg and start it again and so on. If I live linux "alive" for 6 hours for example it will die (tty switching won't work, everything is dead just fans are working).

It seems to me like Skylake support issue. I'm just curious have anyone been able to run Archlinux on Skylake integrated graphics? If anyone, please post your configuration here (kernel version, boot parameters, xorg and intel driver version)

EDIT
Seems like related bug

Last edited by deathangle908 (2015-12-24 21:20:51)

mouseman · 2015-12-14 12:31:28

deathangle908 wrote:

Windows runs just fine so it's not hardware issue.

deathangle908 wrote:

Sometimes I get triangles in windows.

What are you trying to say? Display corruption is acceptable but crashes or hanging isn't?

Neither is acceptable and if you get display corruption in Windows there is something wrong.

Monitor your CPU temperature is the first thing I would do. And test your memory with memtest86+ (FYI, integrated graphics uses a portion of your system RAM).

I own a Skylake with HD530 and it works flawlessly, in Linux and in Windows.

deathangle908 · 2015-12-14 14:46:41

deathangle908 wrote:

Windows runs just fine so it's not hardware issue.

OS windows runs fine.

deathangle908 wrote:

Sometimes I get triangles in windows.

I get triangles in X windows system, for example yakuake anymated with white triangle

I own a Skylake with HD530 and it works flawlessly, in Linux ...

Oh man, you have no idea... It's been for two weeks already...
Could you please post the output of:

uname -a
pacman -Qs video-intel
pacman -Qs xorg-server
kernel boot parameters
grep MODULES /etc/mkinitcpio.conf
Exact laptop model you're using
/etc/X11/xorg.conf.d/*, /etc/X11/xorg.conf

Thanks in advance!

mouseman · 2015-12-14 19:54:20

In order:
4.2.5-1-ARCH
local/xf86-video-intel 1:2.99.917+478+gdf72bc5-3
local/xorg-server 1.18.0-3
local/xorg-server-common 1.18.0-3
options root=PARTUUID=0b017073-c443-4a7b-9dcf-f3b5e82f7d31 rw i915.preliminary_hw_support=1 elevator=cfq quiet

Section "Device"
	Identifier	"Intel Graphics"
	Driver		"intel"
	Option		"AccelMethod"	"glamor"
	Option		"DRI"		"3"
	Option		"TearFree"	"true"
EndSection

Not a laptop, though. Gigabyte GA-Z170X-Gaming 7 with i5 6600K.

I'm stil thinking hardware/temperature related. Your Windows might control the fans better. Try lm_sensors in Linux to monitor cpu temps or set your fans to a higher speed in the BIOS if that's possible.

Last edited by mouseman (2015-12-14 19:58:54)

deathangle908 · 2015-12-15 08:13:57

mouseman wrote:

.I'm stil thinking hardware/temperature related.

Temperature is fine: 56C (132F). Moreover it could hang right after start.

TheSgtBilko · 2015-12-15 21:00:37

Had same issue with random complete freezes on i5-6200U.

Fixed temporarily as suggested at end of "freeze section" in wiki: https://wiki.archlinux.org/index.php/In … tel_driver

Add "intel_pstate=disable" in /etc/default/grub and then rebuild grub.

I think there's a thread here in forums too.

Last edited by TheSgtBilko (2015-12-15 21:01:40)

deathangle908 · 2015-12-16 20:09:58

TheSgtBilko wrote:

kernel param "intel_pstate=disable"

Disabling pstate didn't work for me.With `i915.semaphores=1` kernel param xorg doesn't start at all. ` Option "NoAccel" "True"` and ` Option "DRI" "False"` - the same
Skylake graphics crashes a few times faster if I connect external monitor (meaning multiple monitors configuration)
Here's the dmesg after crash with external monitors. When I'm sitting only on laptop's display dmesg stacktrace is a rare case.
Here's journalctl -xb after crash with only laptop display.

Last edited by deathangle908 (2015-12-17 08:28:36)

jono · 2015-12-19 20:38:12

I think I might be having a similar problem which I have been trying to narrow down for weeks. I just built a desktop machine with similar specs and seems to be crashing within a couple hours of use. Specifically, it is a hard freeze. No mouse movement. It can stay that way for hours. Cant ping the machine. No RESIUB. I have to hard power down the machine.

Reproducing the crash has been the biggest problem. What I do is load 20 copies of the same video in VLC and wait 10 minutes to 2 hours and then it freezes. I tried 'stress' with various parameters but could not get it to crash.

There is nothing in the logs related to this in dmesg or journalctl which is what has been making me think this is a hardware problem. But after seeing a couple posts like yours, I wondered if it was kernel related. I have tried many boot parameters - none have helped much. But i915.preliminary_hw_support=1 seems required.

I'm using Antergos (Arch) with Kernel 4.2.5-1.

I've reseated the RAM, ran memory tests, tried MBR vs UEFI disks, tried SSD and HDD disks. Same problem.

Though I was able to run Ubuntu 15.10 (Kernel 4.2.0-19) and it did not crash (I ran it for 15 hours). Now in Arch I have downgraded to the LTS Kernel (4.1.14-1-lts) and it has not crashed - so far running 17 hours with videos.

Side question. I tried Ubuntu 14 LTS and it uses kernel 3.13 or something. Video is smooth and does not require i915.preliminary_hw_support=1 . I am wondering why I have to use this boot param at all.

Here are some specs:

motherboard MSI-Z170M, Intel i7 6700K, 16GB DDR4-3200 single dim, 350W power supply
local/xf86-video-intel 1:2.99.917+478+gdf72bc5-3
local/xorg-server 1.18.0-3

grub entry (you can see I am using Intel microcode as well):

menuentry 'Antergos Linux (LTS Kernel)' --class antergos --class arch --class gnu-linux --class gnu --class os $menuentry_id_option 'gnulinux-simple-9b... {
load_video
set gfxpayload=keep
insmod gzio
insmod part_msdos
insmod ext2
cryptomount -u
set root='hd0,msdos1'
if [ x$feature_platform_search_hint = xy ]; then
search --no-floppy --fs-uuid --set=root --hint-bios=hd0,msdos1 --hint-efi=hd0,msdos1 --hint-baremetal=ahci0,msdos1 e6.....
else
search --no-floppy --fs-uuid --set=root e6....
fi
echo 'Loading linux kernel ...'
linux /vmlinuz-linux-lts root=UUID=9b.... rw cryptdevice=/dev/disk/by-uuid/6b...:cryptAntergos resume=UUID=b3... quiet i915.preliminary_hw_support=1
echo 'Loading initial ramdisk ...'
initrd /intel-ucode.img /initramfs-linux-lts.img
}

Last edited by jono (2015-12-19 20:57:18)

deathangle908 · 2015-12-19 22:23:13

jono wrote:

Specifically, it is a hard freeze. No mouse movement. It can stay that way for hours. Cant ping the machine. No RESIUB. I have to hard power down the machine.

I have exactly the same! I tested with other configuration (than described below) it hangs faster but I was able to switch to another tty sometimes after freeze. I also asked a question in unixstackexchange to draw attention, maybe I meet somebody else with similar issue. The best result I get is 2-3 hours with configuration below. If I change anything it will be even less. Another thing I note that starting gparted (gnome disk manager) gives a change of ~25% to hang the xorg immediately (works like a magic from video games).

#kernel 4.4-rc5
i915.preliminary_hw_support=1 nouveau.modeset=0  intel_idle.max_cstate=1
#xorg
Section "Device"
   Identifier  "Intel Graphics"
   Driver      "intel"
   Option      "DRI"  "False"
EndSection

jono wrote:

I am using Intel microcode as well

UnFortunately intel haven't released microcode for skylake, this means they didn't invented any improvements for lame architecture or CPU bios is up to date.

jono wrote:

Now in Arch I have downgraded to the LTS Kernel (4.1.14-1-lts) and it has not crashed - so far running 17 hours with videos.

Hm... It seems like I know another guy that succeeded with 4.1 kernel but not above.. I'mma give it try!

I also tried Ubuntu and mint. I was able to run mint from liveusb (editing grub's options) and install it with mint's gui installer. But when I boot into Mint it just freezes right after start. Ubuntu liveusb xorg doesn't start at all (the same freeze), so I even wasn't able to install it. I tried a hell of a lot configurations and parameter options. I try again and again it seems like the only option is wait until kernel (i915) will officially support Skylake ( as it should since 4.3)

Last edited by deathangle908 (2015-12-19 23:06:08)

mrunion · 2015-12-21 15:35:50

@jono: Just so you know Antegros is not Arch, and is not supported in these forums. Thanks for your contribution, but you need to report this on the Antegros forums.

graysky · 2015-12-21 16:05:46

mouseman wrote:

I own a Skylake with HD530 and it works flawlessly, in Linux and in Windows.

Would you mind emailing me? Considering the purchase of a skylake chip and would like ask you a few questions about your setup. You seem to have disabled forum initiated emails. My email address is the first line of my PKGBUILD I maintain, for example. Thank you.

mouseman · 2015-12-21 17:45:30

graysky wrote:

mouseman wrote:
I own a Skylake with HD530 and it works flawlessly, in Linux and in Windows.
Would you mind emailing me? Considering the purchase of a skylake chip and would like ask you a few questions about your setup. You seem to have disabled forum initiated emails. My email address is the first line of my PKGBUILD I maintain, for example. Thank you.

Done!

Also adjusted my profile to hide my address but allow emails.

Last edited by mouseman (2015-12-21 17:46:54)

Optinuss · 2015-12-21 21:39:10

I'm facing exactly the same issues as you on my new laptop. (Schenker P506)
Specs: i7-6700HQ, GTX 970M, Chipset Mobile Intel HM170 Express
I'm on a fresh Arch installation and already tried Linux Kernel 4.2.5-1 and LTS 4.1.15-1.
Xorg Version : 1.18 + SDDM + Plasma Desktop
First I guessed it was caused of Bumblebee but after removing it completly I'm pretty sure its a fault in the intel driver. (I could run programs with optirun and nvidia just fine though!)
I already tried these kernel parameters in combination with i915.preliminary_hw_support=1 :
intel_idle.max_cstate=1, i915.enable_execlists=0, i915.semaphores=1
With semaphores screen just freezed directly after login. The other apparantly don't change anything at all. After login I can use the system for few seconds or minutes and before it freezes completly.
@mouseman: Great to hear that you got the HD530 working. It would be awesome if you could share your setup or maybe send me a mail?

jono · 2015-12-22 05:11:16

Reporting back again.
After the downgrate to the LTS kernel I was able to run my machine for 2+ days. No crash. To be sure, I disabled the intel microcode as you mentioned and it seemed to have no affect (running 14+ hours now). So LTS was what seemed to work for me.

Also, sidenote: I do not have a graphics card. So no Nvidia or anything else to murk this up.

PS. Yes, I am aware of the fact that I'm using Antergos and this is and Arch forum. Also Mint, Ubuntu and Windows are not Arch. As you can see we are all grasping at straws here trying to find any correlation possible. My guesses have gone from SSD corruption, to broken motherboard, to kernel compatabiliy.

mouseman · 2015-12-22 09:51:09

[2015-12-21 11:09] [ALPM] upgraded virtualbox-host-modules (5.0.10-2.1 -> 5.0.12-1)
[2015-12-21 11:09] [ALPM-SCRIPTLET] In order to use the new version, reload all virtualbox modules manually.
[2015-12-21 11:09] [ALPM] upgraded virtualbox (5.0.10-2 -> 5.0.12-1)
[2015-12-21 11:09] [ALPM] upgraded virtualbox-guest-iso (5.0.10-1 -> 5.0.12-1)
[2015-12-21 11:09] [ALPM] upgraded wine (1.8rc4-1 -> 1.8-1)
[2015-12-21 11:09] [ALPM] upgraded xf86-video-intel (1:2.99.917+478+gdf72bc5-3 -> 1:2.99.917+519+g8229390-1)
[2015-12-21 11:09] [ALPM] transaction completed

Since this upgrade a Windows VM in Virtualbox was consistently crashing my machine. Complete hang, sound in a loop and frozen mouse. I can't find anything in the logs (journalctl or from Virtualbox itself).

I downgraded the Intel driver back to the previous version and it's been stable again.

[2015-12-22 10:23] [PACMAN] Running 'pacman -U xf86-video-intel-1:2.99.917+478+gdf72bc5-3-x86_64.pkg.tar.xz'
[2015-12-22 10:24] [ALPM] transaction started
[2015-12-22 10:24] [ALPM] downgraded xf86-video-intel (1:2.99.917+519+g8229390-1 -> 1:2.99.917+478+gdf72bc5-3)
[2015-12-22 10:24] [ALPM] transaction completed

I just wish there was something in a log somewhere to show the crash. I'm guessing a kernel panic?

deathangle908 · 2015-12-23 14:52:31

Optinuss wrote:

nvidia just fine though

What a bummer I can't run neither nvidia neither intel. Would you mind just to take a look at mine nvidia topic. In a few words: xorg with nvidia starts, applications start (I hear skype sound), but I get no backlight (blackscreen) . I've spent like 10+ hours trying to make nvidia work.
Could you please share all your configurations with nvidia, to be specific

uname -a
pacman -Qs nvidia
pacman -Qs xorg-server
kernel boot parameters
grep MODULES /etc/mkinitcpio.conf
/etc/X11/xorg.conf.d/*, /etc/X11/xorg.conf
Maybe you did some additional setup

Optinuss wrote:

With semaphores screen just freezed directly after login

Same same, all the same.

Last edited by deathangle908 (2015-12-23 15:03:16)

deathangle908 · 2015-12-24 00:59:11

Well, first error: can't run applications on 4.1.13lts after 1 hour of using, I can't open new applications they all fail with error

can't open Display :0

: cat /proc/cmdline 
initrd=\initramfs-linux-lts.img root=/dev/nvme0n1p4 rw i915.preliminary_hw_support=1 nouveau.modeset=0  intel_idle.max_cstate=1
: cat /etc/X11/xorg.conf.d/20-intel.conf 
Section "Device"
   Identifier  "Intel Graphics"
   Driver      "intel"
   Option      "DRI"  "False"
EndSection

I also tried without intel.conf and kernel params below

initrd=\initramfs-linux-lts.img root=/dev/nvme0n1p4 rw i915.preliminary_hw_support=1 nouveau.modeset=0

I also tried 4.1.15-1 lts - it freezes too

jono wrote:

No crash

I have no success with lts either:( . Seems like everyone else succeeded which means I'm doing something wrong. Did you use this package? Do you have any kernel params, xorg conf. Could you please post the whole stack of package that you're using.

UPDATE
I downgraded intel driver to 478v apart with lts kernel. This combinations works better than others but still hangs in the end of the day. The best uptime I got is 7 hours, then kernel just freezed. Sometimes it freezes after 1 hour. Another thing that tty switching (alt+ctrl+f2) on this configuration doesn't work. After switching back to xorg it shows broken artifacts (triangles, below mouse is all white etc, which means xorg is completely unusable after switching. I just give up Here's current configuration.

: pacman -Qs xf86-video-intel
local/xf86-video-intel 1:2.99.917+478+gdf72bc5-3 (xorg-drivers xorg)
    X.org Intel i810/i830/i915/945G/G965+ video drivers
: uname -r
4.1.15-1-lts
: cat /proc/cmdline 
initrd=\intel-ucode.img initrd=\initramfs-linux-lts.img root=/dev/nvme0n1p4 rw i915.preliminary_hw_support=1 nouveau.modeset=0
: uptime 
 16:33:13 up  5:47,

Note that I don't have cstate in kernel params since it makes it even worse. I also turned on cstate in bios, without it xorg hangs really fast. I would like to switch to optimus (my laptop has gtx 970m) but I can't make it work, it just shows black screen.

Last edited by deathangle908 (2015-12-25 21:13:43)

deathangle908 · 2015-12-24 19:46:17

EDIT
Hm It's seems to me that xorg freezing is actually a kernel freezing, Looks like related to this bug.

Last edited by deathangle908 (2015-12-25 23:34:01)

jono · 2015-12-30 22:21:42

Another update since I have been running 4.1 kernel for a couple weeks now. Unfortunately I am occasionally getting very random crashes. Sometimes after running for several days straight, it will hard freeze. So my workaround is not a complete solution for me, but it is closer to one.

Note that we are talking about different crashes here. Mine does not seem to be X related. These are hard freezes. No key command gets me out of it, I cant quit X. There is nothing in the logs.

My real question now is could this be a hardware failure? Since I have no logs I dont know this is not a hardware issue.

wildfowl · 2016-01-01 22:50:51

Can you try the following string added to your kernel parameters.

intel_iommu=igfx_off

deathangle908 · 2016-01-03 02:17:14

jono wrote:

Mine does not seem to be X related.

Well it's hard to say if it's related to X. It seems to me it's i915 is failing. I worked a lot of time without xorg and kernel didn't freeze. Meanwhile running xorg produces i915 kernel stacktraces, sometime nothing works after xorg freeze (no hdd sound, nod leds blinking, not tty switching, the picture is freezed), sometimes I'm able to switch to tty and work in the terminal... Very mysterious...

jono wrote:

There is nothing in the logs.

How do you detect that there's nothing it the logs? Usually kernel at least says "because kernel failure" when it dies. You can set dmesg to save logs to file so you check it after reboot. Furthermore journalctl by default logs kernel messages.

headkase · 2016-01-03 06:16:19

jono wrote:

PS. Yes, I am aware of the fact that I'm using Antergos and this is and Arch forum. Also Mint, Ubuntu and Windows are not Arch. As you can see we are all grasping at straws here trying to find any correlation possible. My guesses have gone from SSD corruption, to broken motherboard, to kernel compatabiliy.

jono, the problem with you using Antergos is that it is not Arch. Whatever the Antergos people did to their distribution may not be the same as what Arch did with theirs. Which means that something that works for you may not work for an Arch user or vice-versa. Posting on this forum while not being an Arch user is a form of help-vampirism. You are receiving help with your issues and your help and the associated discussion does not necessarily benefit Arch - as the information may, as said, not actually work on Arch. It can be frustrating when other communities are not as active as Arch but the fact remains that if you would like to benefit from the good-will of these forums then you really should just install Arch and then everything would be just peachy discussion-wise. That is my opinion.

jono · 2016-01-03 19:48:19

Now that 4.3 stable has been released, I have been using that with success so far (3 days). Of course I am able to remove the i915.preliminary_hw_support=1 flag as well. No crashes yet.

deathangle908 wrote:

jono wrote:
There is nothing in the logs.
How do you detect that there's nothing it the logs? Usually kernel at least says "because kernel failure" when it dies. You can set dmesg to save logs to file so you check it after reboot. Furthermore journalctl by default logs kernel messages.

journalctl --boot=1

I have also watched dmesg both on the running machine and from a shelled in connection. But since this crash is hard, my network connection terminates so there is nothing to see.

headkase wrote:

jono, the problem with you using Antergos is that it is not Arch.

I hear what you are saying but I will have to just disagree on that one. Antergos is not much more then Arch with a few added packages. Here are all the ones I have installed from their repo:

> paclist antergos
antergos-gnome-defaults-list 1.0-1
antergos-keyring 20150806-1
antergos-mirrorlist 20150715-1
antergos-wallpapers 0.6-1
antergos-welcome 0.0.2-2
gnome-shell-extension-dash-to-dock 1:v49-1
grub2-theme-antergos 0.1-1
lightdm-webkit2-greeter 0.2.3-3
numix-frost-themes 3.0.4-1
numix-icon-theme 0.r1631.be82dae-1
numix-icon-theme-square 1:0.r82.bb9313e-2
package-query 1.7-2
pamac 2.4.3-4

The kernel for example, is not changed. The Antergos community would be right in pointing me upstream. And maybe even more upstream then the distro. I can look to kernel.org which I have done as well. But since the kernel is configured in the arch repos, I am discussion this here.

In no way am I blaming any party for this or necessarilly asking for a solution. I am just sharing the problems I am having, because I know I have benefitted from others sharing the issues they have with similar hardware and software.

jasonwryan · 2016-01-04 07:08:56

jono wrote:

headkase wrote:
jono, the problem with you using Antergos is that it is not Arch.
I hear what you are saying but I will have to just disagree on that one.

Disagree all you like, it doesn't change the fact that Antegros is not supported here: https://wiki.archlinux.org/index.php/Fo … pport_ONLY

deathangle908 · 2016-01-08 01:17:55

headkase wrote:

jono, the problem with you using Antergos is that it is not Arch.

I would also disagree with you on this one, despite you're "Forum & Wiki Admin" and I'm just a member. I think the main purpose of this forum for people to help each other. That's what exactly we are trying to do. Sometimes you don't need to follow the rules that strictly. Meanwhile I'm 110% sure that antegros bug is still the same as archlinux's, furthermore it's not even related to archlinux, to be specific it's kernel only. But here we can together find a solution like a combination of archlinux's packages (that still present in antegros) + their config. I don't want to start a hollywar, this is mine opinion, but rules are rules...

I updated to archlinux 4.3.3-2, with enabled cstate in bios and "i915.preliminary_hw_support=1 intel_idle.max_cstate=1" issue still persists. On this time:

working time is about 2m-10m, after it xorg hangs
If I restart xorg after it hangs, kernel will hang with it
dmesg doesn't have any messages when xorg crashes
Journalctl has BadWindows qt errors
disabling cstate in bios - xorg crashes right after start

Last edited by deathangle908 (2016-01-08 01:24:32)

Arch Linux

#1 2015-12-13 15:05:20

Xorg with Skylake graphics hangs

#2 2015-12-14 12:31:28

Re: Xorg with Skylake graphics hangs

#3 2015-12-14 14:46:41

Re: Xorg with Skylake graphics hangs

#4 2015-12-14 19:54:20

Re: Xorg with Skylake graphics hangs

#5 2015-12-15 08:13:57

Re: Xorg with Skylake graphics hangs

#6 2015-12-15 21:00:37

Re: Xorg with Skylake graphics hangs

#7 2015-12-16 20:09:58

Re: Xorg with Skylake graphics hangs

#8 2015-12-19 20:38:12

Re: Xorg with Skylake graphics hangs

#9 2015-12-19 22:23:13

Re: Xorg with Skylake graphics hangs

#10 2015-12-21 15:35:50

Re: Xorg with Skylake graphics hangs

#11 2015-12-21 16:05:46

Re: Xorg with Skylake graphics hangs

#12 2015-12-21 17:45:30

Re: Xorg with Skylake graphics hangs

#13 2015-12-21 21:39:10

Re: Xorg with Skylake graphics hangs

#14 2015-12-22 05:11:16

Re: Xorg with Skylake graphics hangs

#15 2015-12-22 09:51:09

Re: Xorg with Skylake graphics hangs

#16 2015-12-23 14:52:31

Re: Xorg with Skylake graphics hangs

#17 2015-12-24 00:59:11

Re: Xorg with Skylake graphics hangs

#18 2015-12-24 19:46:17

Re: Xorg with Skylake graphics hangs

#19 2015-12-30 22:21:42

Re: Xorg with Skylake graphics hangs

#20 2016-01-01 22:50:51

Re: Xorg with Skylake graphics hangs

#21 2016-01-03 02:17:14

Re: Xorg with Skylake graphics hangs

#22 2016-01-03 06:16:19

Re: Xorg with Skylake graphics hangs

#23 2016-01-03 19:48:19

Re: Xorg with Skylake graphics hangs

#24 2016-01-04 07:08:56

Re: Xorg with Skylake graphics hangs

#25 2016-01-08 01:17:55

Re: Xorg with Skylake graphics hangs

Board footer