[Nix-dev] Disappearing X session on some systems

[Evan Rowley]

The issue on the ThinkPad E465 seems to have been fixed. No longer have I
experienced a locked X session. The problem may have been related to GDM.
It was some error messages received during nixos-rebuild switch which led
me down this path. The fix was achieved after applying a workaround
supplied by digs: https://github.com/NixOS/nixpkgs/issues/24172#
issuecomment-304540789

Unfortunately I was not able to test the fix on the SuperMicro Intel with
Nvidia card. Before I could test, I was receiving garbled images on my
monitor from boot. I assumed the 10-year old Nvidia 8800 GT had given up
the ghost, so swapped it with a spare Radeon 7750. I verified with the
Radeon 7750 that GDM-related error messages were no longer being produced
by running nixos-rebuild switch - the intended fix of the Issue 24172
workaround.

The latest issue is the SuperMicro Intel machine cannot keep an X session
running - even when I'm actively using a session. Sessions freeze up within
1 - 5 minutes after starting. The mouse cursor still works, but nothing on
the screen responds to it. I have used journalctl to review a few logs of
what happens when X freezes. Around the time stuck to the clock in the X
session, these things happen:

Up to one minute before, org.freedesktop.PackageKit and org.bluez both
fail. I'm not sure if this is related or not.
.gnome-shell-wr[1699]: JS LOG: Error calling StartServiceByName for
org.freedesktop.PackageKit:
GDBus.Error:org.freedesktop.DBus.Error.TimedOut: Failed to activate service
'org.freedesktop.PackageKit': timed out
...
dbus[1378]: [system] Failed to activate service 'org.bluez': timed out
...

At the exact moment when X gets stuck, these kernel messages are logged:
kernel: DMAR: DRHD: handling fault status reg 2
kernel: DMAR: [INTR-REMAP] Request device [00:00.0] fault index 13 [fault
reason 38] Blocked an interrupt request due to source-id verification
failure
...

After the above kervel messages, these gdm and gnome messages follow:
/nix/store/cslvy4s12r9y0ixi33x79m3906zbsimb-gdm-3.22.0/libexec/gdm-x-session[2088]:
(II) AIGLX: Suspending AIGLX clients for VT switch
xsession[2030]: gnome-session-binary[2055]: DEBUG(+): emitting
SessionIsActive
gnome-session-binary[2055]: DEBUG(+): emitting SessionIsActive
...

After the above messages, gdm-related "systemd-logind: got pause" messages
appear:
/nix/store/cslvy4s12r9y0ixi33x79m3906zbsimb-gdm-3.22.0/libexec/gdm-x-session[2016]:
(II) systemd-logind: got pause for 13:70
/nix/store/cslvy4s12r9y0ixi33x79m3906zbsimb-gdm-3.22.0/libexec/gdm-x-session[2016]:
(II) systemd-logind: got pause for 13:65
/nix/store/cslvy4s12r9y0ixi33x79m3906zbsimb-gdm-3.22.0/libexec/gdm-x-session[2016]:
(II) systemd-logind: got pause for 226:0
/nix/store/cslvy4s12r9y0ixi33x79m3906zbsimb-gdm-3.22.0/libexec/gdm-x-session[2016]:
(II) systemd-logind: got pause for 13:64
/nix/store/cslvy4s12r9y0ixi33x79m3906zbsimb-gdm-3.22.0/libexec/gdm-x-session[2016]:
(II) systemd-logind: got pause for 13:68
/nix/store/cslvy4s12r9y0ixi33x79m3906zbsimb-gdm-3.22.0/libexec/gdm-x-session[2016]:
(II) systemd-logind: got pause for 13:69
...

After the above messages, there is then a repeat of these 'GSD_IS_WACOM_DEVICE
(device)' failed & invalid pointer messages:
gnome-settings-[2177]: gsd_wacom_device_get_device_type: assertion
'GSD_IS_WACOM_DEVICE (device)' failed
gnome-settings-[2177]: gsd_wacom_device_get_settings: assertion
'GSD_IS_WACOM_DEVICE (device)' failed
gnome-settings-[2248]: invalid (NULL) pointer instance
...

Except for the org.freedesktop.PackageKit and org.bluez messages, all of
the messages here are invariably displayed in this order, within the span
of a minute, every time X freezes on this machine. The exact meaning of
this I'm unsure of. I am considering opening up a separate issue on GitHub.





On Sun, May 28, 2017 at 1:31 PM, Evan Rowley <rowley.evan at gmail.com> wrote:

> Upon further inspection, it appears bug 24172, bug 22470, and pull request
> 25621 may be related to the X session problem I mentioned.
>
> Gdm on nvidia
> https://github.com/NixOS/nixpkgs/pull/25621
>
> nixos-rebuild switch fails in X terminal
> https://github.com/NixOS/nixpkgs/issues/22470
>
> systemd-vconsole-setup.service seems to get started after every switch
> https://github.com/NixOS/nixpkgs/issues/23118
>
>
> On Sat, May 27, 2017 at 11:57 AM, Evan Rowley <rowley.evan at gmail.com>
> wrote:
>
>> Has anyone left their NixOS machine running for awhile, come back to it,
>> and saw a console on tty7 where X used to be running? I experience this
>> when I leave two of my NixOS systems alone for awhile. Both are using Gnome
>> 3. It affects my ThinkPad E465 which is running an AMD A10-8700P & Radeon
>> R6 M340DX. It also affects my SuperMicro SYS-7038A-i with Intel Xeon CPUs
>> and Nvidia 8800 GT graphics. It never happens while I'm using these, just
>> only when I've left them alone for a while - like overnight, for example. I
>> have not seen this behavior on my NixOS ThinkPad T420 with Intel i5-2520M
>> and integrated graphics. I realize that the T420 is relatively mature in
>> Linux support & things working well is somewhat expected - but a crashing X
>> session seems to be abnormal, even for hardware that's less tested. One
>> major difference between the T420 and the other systems is the T420 is
>> using integrated graphics while the other systems are using dedicated
>> graphics cards - which is why I listed out the cards here. If anyone knows
>> of ongoing issues with NixOS/X/graphics which could be related, how it
>> further investigate the issue, ideas on a fix, or is experiencing the same
>> kind of issue - please reply. :)
>>
>> --
>>  - EJR
>>
>
>
>
> --
>  - EJR
>



-- 
 - EJR
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.science.uu.nl/pipermail/nix-dev/attachments/20170529/8c7d2984/attachment.html>