Arch Linux

knewyousir

Member

Registered: 2018-02-22

Posts: 20

Arch Crashing Periodically - AMDGPU and/or Mobo/Processor Issues??

Hi all!
My arch build is coming along nicely but I am having trouble getting to be stable.
I'd like to eventually use it host virtual servers, and have it safely up for weeks/months at a time,
but for now am having problems getting it to be stable for more than 12hrs or a day at a time.
I'm starting to wonder if the hardware I purchased is suitable for running Arch Linux stably.

Been examining

journalctl -p err

and am seeing a couple main errors affecting me at this point:
Would someone be able to point me in the right direction towards resolving these issues?:

Mar 17 19:43:25 knewyousir kernel: [drm:amdgpu_atombios_dp_link_train [amdgpu]] *ERROR* displayport link status failed
Mar 17 19:43:25 knewyousir kernel: [drm:amdgpu_atombios_dp_link_train [amdgpu]] *ERROR* clock recovery failed
Mar 17 19:43:25 knewyousir kernel: [drm:amdgpu_atombios_dp_link_train [amdgpu]] *ERROR* displayport link status failed
Mar 17 19:43:25 knewyousir kernel: [drm:amdgpu_atombios_dp_link_train [amdgpu]] *ERROR* clock recovery failed
Mar 17 23:36:20 knewyousir kcheckpass[17263]: pam_tally(kde:auth): Error opening /var/log/faillog for update
Mar 17 23:36:20 knewyousir kcheckpass[17263]: pam_tally(kde:auth): Error opening /var/log/faillog for read
Mar 17 23:36:29 knewyousir kcheckpass[17263]: pam_tally(kde:auth): Error opening /var/log/faillog for update
Mar 17 23:36:29 knewyousir kcheckpass[17263]: pam_tally(kde:auth): Error opening /var/log/faillog for read
Mar 17 23:36:29 knewyousir kcheckpass[17263]: pam_tally(kde:setcred): Error opening /var/log/faillog for update
Mar 17 23:36:29 knewyousir kcheckpass[17263]: pam_tally(kde:setcred): Error opening /var/log/faillog for update
Mar 17 23:39:10 knewyousir kdeinit5[776]: QThread: Destroyed while thread is still running
-- Reboot --

-Seems like the displayport connection with the monitor is lost, which starts a death spiral of sorts. Could this be an issue with my amdgpu driver?

Feb 22 17:09:15 knewyousir systemd[676]: at-spi-dbus-bus.service: Failed to add inotify watch descriptor for control group /user.slice/user-1000.slice/user@1000.service/at-spi-dbus-bus.service: No space left on device
-- Reboot --

-No idea about this one. I looked it up and seemingly had something to do with GNOME, but I use KDE plasma right now...

Mar 21 04:00:24 knewyousir kernel: xhci_hcd 0000:11:00.3: AMD-Vi: Event logged [IO_PAGE_FAULT domain=0x0008 address=0x08000000fffdb180 flags=0x0030]
-- Reboot --

-Have been seeing this error periodically ever since I got Arch up and running. Something to do with my mobo/processor (Taichi X370/Ryzen 7 1700x) and the Kernel? I read elsewhere that AMD had given refunds to Linux users that had purchase certain models of the Ryzen line, including mine. Is there any hope for fixing this one?

Thanks for your help in advance

Last edited by knewyousir (2018-03-21 23:14:40)

Morn

Member

Registered: 2012-09-02

Posts: 886

Re: Arch Crashing Periodically - AMDGPU and/or Mobo/Processor Issues??

If you think DisplayPort or the GPU are the problem, maybe you should unplug the monitor and see if the system is stable then. Just ssh in from another machine and see what happens. Then maybe later use "ssh -X" and start KDE. Just circumvent the graphics card for a while via X's remote features. And try the virtual machines in that way too.

The Linux Ryzen errors were gcc core dumps and MCE errors, so if you get neither of those, your Ryzen should be fine. My Ryzen 1700 is stable now, but I have not tried virtualization (it's not even enabled in the BIOS), so there could still be an issue with AMD-Vi I suppose.

seth

Member

From: Won't reply 2 private help req

Registered: 2012-09-03

Posts: 78,857

Re: Arch Crashing Periodically - AMDGPU and/or Mobo/Processor Issues??

Did you notice that the link training issues predate the reboot by ~4h?

That said:
https://bbs.archlinux.org/viewtopic.php?id=233304
https://bbs.archlinux.org/viewtopic.php?id=226668

---

How to upload text · How to boot w/o GUI · Disable Windows Fast-Start! · Fix your xinitrc

knewyousir

Member

Registered: 2018-02-22

Posts: 20

Re: Arch Crashing Periodically - AMDGPU and/or Mobo/Processor Issues??

Thanks for the suggestions Seth and Morn- I started by checking out the Ryzen issues threads, went into the bios and turned off some of the suggested settings, namely I disabled c-states and coolnquiet but left ASLR on. I also tweaked my ram speed down to something slow but safe (2166 vs 34XX (forgot)). I am on v3.20 bios which may not support max speed on my RAM yet apparently. Its only been one day, but I haven't had any more display-related crashes since I made the bios changes.
Fingers crossed that these do the trick!

Last edited by knewyousir (2018-03-24 02:52:21)

Morn

Member

Registered: 2012-09-02

Posts: 886

Re: Arch Crashing Periodically - AMDGPU and/or Mobo/Processor Issues??

journalctl | grep "Hardware Error"

is a command you should run on Ryzen systems from time to time to make sure everything is OK. Ideally it prints nothing of course...

knewyousir

Member

Registered: 2018-02-22

Posts: 20

Re: Arch Crashing Periodically - AMDGPU and/or Mobo/Processor Issues??

Ok- so had been working well but had two seemingly random crashes in the last 16 hours.

This crash instance below caused a red screen and loss of audio earlier today while firefox and clementine ran behind a kde plasma lock screen. Mobo codes seemed to indicate hardware was working fine, but as soon as I attempted to change the console (with a mere keyboard press), it started flashing '00' to indicate a full-blown crash.

Mar 25 15:10:25 knewyousir kernel: [drm:drm_atomic_helper_wait_for_flip_done [drm_kms_helper]] *ERROR* [CRTC:42:crtc-0] flip_done timed out
Mar 25 15:13:41 knewyousir kernel: INFO: task Xorg:732 blocked for more than 120 seconds.
Mar 25 15:13:41 knewyousir kernel:       Tainted: G        W        4.15.10-1-ARCH #1
Mar 25 15:13:41 knewyousir kernel: "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
Mar 25 15:15:44 knewyousir kernel: INFO: task Xorg:732 blocked for more than 120 seconds.
Mar 25 15:15:44 knewyousir kernel:       Tainted: G        W        4.15.10-1-ARCH #1
Mar 25 15:15:44 knewyousir kernel: "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
Mar 25 15:17:47 knewyousir kernel: INFO: task Xorg:732 blocked for more than 120 seconds.
Mar 25 15:17:47 knewyousir kernel:       Tainted: G        W        4.15.10-1-ARCH #1
Mar 25 15:17:47 knewyousir kernel: "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
Mar 25 15:19:50 knewyousir kernel: INFO: task Xorg:732 blocked for more than 120 seconds.
Mar 25 15:19:50 knewyousir kernel:       Tainted: G        W        4.15.10-1-ARCH #1
Mar 25 15:19:50 knewyousir kernel: "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
Mar 25 15:21:53 knewyousir kernel: INFO: task Xorg:732 blocked for more than 120 seconds.
Mar 25 15:21:53 knewyousir kernel:       Tainted: G        W        4.15.10-1-ARCH #1
Mar 25 15:21:53 knewyousir kernel: "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.

I'm thinking that this is the same cause as the prior crashes, although I don't know why it wouldn't have shown up in the log before, other than maybe the bios settings I changed allowed log entries to be made?

I did a bit of research (thread here)and it appears to have something to do with the kernel and the display. Xorg of course is a display server. It seems like Xorg's process has been blocked by the kernel for some reason and then it terminates?

knewyousir

Member

Registered: 2018-02-22

Posts: 20

Re: Arch Crashing Periodically - AMDGPU and/or Mobo/Processor Issues??

Another crash today:

Mar 26 17:35:55 knewyousir kernel: mce: [Hardware Error]: CPU 8: Machine Check: 0 Bank 5: bea0000000000108
Mar 26 17:35:55 knewyousir kernel: mce: [Hardware Error]: TSC 0 ADDR 1ffff9c109ec6 MISC d012000101000000 SYND 4d000000 IPID 500b000000000 
Mar 26 17:35:55 knewyousir kernel: mce: [Hardware Error]: PROCESSOR 2:800f11 TIME 1522088852 SOCKET 0 APIC 8 microcode 8001129

Context was:
primary monitor :0 was running behind lock screen with monitor power off (through power mgmt not hw switch).
I ssh'd in (sha keys + OTP) and was running vnc session on a virtual desktop.
Left it idle for a few minutes while I chatted with an employee.
When I came back, the vnc session had a spinning timer on it and was unresponsive.
Closed the session and looked at ssh logs. Don't have the exact message on me, but basically said
"pipe was broken". Further ssh attempts were met with "network unresponsive".
Got home and CPU had rebooted itself and was asking for my encryption password to unlock LUKS...

Last edited by knewyousir (2018-03-26 21:52:36)

Morn

Member

Registered: 2012-09-02

Posts: 886

Re: Arch Crashing Periodically - AMDGPU and/or Mobo/Processor Issues??

Yes, that last one is the infamous Ryzen MCE error. You could try the "processor.max_cstate=1" kernel parameter, as these MCE crashes seem to be caused by CPU power saving features.