[solved] nemo crashes + chrome freezes + systemctl suspend hangs

dr1fter · 2024-02-25 09:29:57

Some months ago, I had a similar issue (hanging applications the eventually loaded after some tens of seconds), which I could also fix (after finding this thread some months ago) by removing `xdg-desktop-portal-gnome` as found here: https://bbs.archlinux.org/viewtopic.php … 6#p2153026.

Since yesterday (I rebooted my machine yesterday after ~4w w/o reboot), I have a similar (but different) new issue:

Almost all applications seem to startup with usual loading performance; however:

- `nemo` (cinnamon's file-manager) will not start at all (it will hang for some few tens of seconds, and then terminate without ever showing up); *update*: after some additional runs of `nemo`, it eventually _did_ startup some of the times I tried
- `google-chrome` will sometimes freeze (only one window at a time) for some tens of seconds
- `systemctl suspend` will hang "forever" and not even react to CTRL-C

I uploaded the output of `strace -o nemo.strace.txt nemo .` to https://0x0.st/H5FL.txt.gz (use

curl https://0x0.st/H5FL.txt.gz | gunzip

to retrieve.

I did not do any recent configuration changes, and also did not install or remove any packages in the past few weeks. I did however install an additional 32 GiB of main memory (upgrading my machine from 32 to 64 GiB in total; however, I would actually not have assumed this might be related to the issues I see). I let BIOS (/EFI) decide on memory configurations + did not see any crashes, esp. from memory-bound workloads, like some golang-builds.

I uploaded the output of `loginctl session-status` to

https://0x0.st/H5F9.txt.gz

In addition:

$ echo $DBUS_SESSION_BUS_ADDRESS 
unix:path=/run/user/1000/bus

@Seth: you asked me to also provide a dump of my system-journal. Could you hint me as to how to do that?

Last edited by dr1fter (2024-02-26 18:24:08)

seth · 2024-02-25 14:34:07

sudo journalctl -b | curl -F 'file=@-' 0x0.st

The session looks unsuspicious.
You're getting timeouts on the AT-SPI bus and the X11 connection.
If you don't rely on on the a11y features, you can disable the AT-SPI w/ "export NO_AT_BRIDGE=1" but it would hint at dbus issues.

What does

write(2, "Could not register nemo as a rem"..., 57) = 57

actually write as eror before nemo fails?

dr1fter · 2024-02-25 20:44:44

@seth: I think this is the output:

[x] ocm]$ strace -o strace.nemo.txt nemo . 
Could not register nemo as a remote: Timeout was reached

Note that I did not see this output before or after the one time I ran strace as quoted above (in the other cases, `nemo` did not print anything to stdout/stderr, I think).

The timeout is somewhat consistent w/ the strace-dump, where the process seems to wait for a resource and eventually seems to give up waiting (I am not too familiar w/ reading through strace-dumps)

update: exporting NO_AT_BRIDGE=1 does fix the issue for starting `nemo` and also for running `systemctl suspend` (as for google-chrome, it will be harder to tell, as this was a sporadic issue). So thanks a lot for this workaround.

However, interestingly, after exporting NO_AT_BRIDGE=1 in one terminal (I always have six terminals open - two per monitor (three monitors)) and starting nemo successfully + putting my machine to sleep successfully, and resuming from sleep, I could successfully run `nemo` from another terminal (where I did not export said env-var). Did this one-time starting of a nemo process with the env-var exported cause a persistent effect for the running X11 session?

Last edited by dr1fter (2024-02-25 20:53:54)

seth · 2024-02-25 21:09:02

I rather suspect that nemo will run a single process and subsequent invocations will just spawn new windows.

Please post your complete system journal for the (previous, assuming you're blocking the AT bridge globally now) boot:

sudo journalctl -b -1 | curl -F 'file=@-' 0x0.st

so we can see what's wrong w/ your dbus config at large or whether there's something else broken w/ at-spi
If you don't need that, it's perfectly fine to disable it and avoid some useless process and an extra dbus session, but it should™ work and if it doesn't, you better ensure that the cause for the failure won't cause more problems.

dr1fter · 2024-02-25 21:30:09

I uploaded the output:

http://0x0.st/HR-N.txt

. Not sure whether I am currently blocking AT bridge globally (I exported the env-var only in one terminal, and did not reboot yet).

Generally speaking, I would prefer to find out what is broken. However, until today, I did not even know at-api exists. From a quick google-search (also for the NO_AT_BRIDGE-env-var), I learnt it seems to be related to accessibility-functionality. If this is true, then maybe the best thing for me would be to fully disable/remove all related packages (I do not use any accessibility features).

seth · 2024-02-25 21:46:54

Yes is and yes you can just deactivate it (I've disabled it since ever), but you won't be able to remove it w/o messing with the package system (eitehr ignoring deps or adding a lot of NoExtract's)
Journal is a relatively short run (1m)

Feb 24 19:14:50 arch systemd[1]: Started Session 1 of User dr1fter.
…
Feb 24 19:14:50 arch systemd[924]: Starting Accessibility services bus...
Feb 24 19:14:50 arch systemd[924]: Started Accessibility services bus.
Feb 24 19:14:50 arch at-spi-bus-launcher[1145]: Policy to allow eavesdropping in /usr/share/defaults/at-spi2/accessibility.conf +15: Eavesdropping is deprecated and ignored
Feb 24 19:14:50 arch at-spi-bus-launcher[1145]: Policy to allow eavesdropping in /usr/share/defaults/at-spi2/accessibility.conf +17: Eavesdropping is deprecated and ignored
Feb 24 19:14:50 arch dbus-broker-launch[1145]: Ready
Feb 24 19:14:50 arch systemd[924]: Created slice Slice /app/dbus-:1.36-org.a11y.atspi.Registry.
Feb 24 19:14:50 arch systemd[924]: Started dbus-:1.36-org.a11y.atspi.Registry@0.service.
Feb 24 19:14:50 arch udisksd[1164]: udisks daemon version 2.10.1 starting
Feb 24 19:14:50 arch at-spi2-registryd[1156]: SpiRegistry daemon is running with well-known name - org.a11y.atspi.Registry
…
Feb 24 19:15:09 arch systemd[1]: Stopping Session 1 of User dr1fter...

You get lots of problems when logging out

Feb 24 19:15:10 arch at-spi2-registr[2632]: Could not open X display
Feb 24 19:15:10 arch at-spi2-registryd[2632]: SpiRegistry daemon is running with well-known name - org.a11y.atspi.Registry
Feb 24 19:15:10 arch at-spi2-registr[2632]: AT-SPI: Cannot open default display
Feb 24 19:15:10 arch systemd[924]: Starting flatpak document portal service...
Feb 24 19:15:10 arch systemd[924]: dbus-:1.36-org.a11y.atspi.Registry@1.service: Main process exited, code=exited, status=1/FAILURE
Feb 24 19:15:10 arch systemd[924]: dbus-:1.36-org.a11y.atspi.Registry@1.service: Failed with result 'exit-code'.
…
Feb 24 19:15:10 arch xdg-desktop-por[2661]: cannot open display: :0
Feb 24 19:15:10 arch systemd[924]: xdg-desktop-portal-gtk.service: Main process exited, code=exited, status=1/FAILURE
Feb 24 19:15:10 arch systemd[924]: xdg-desktop-portal-gtk.service: Failed with result 'exit-code'.
Feb 24 19:15:10 arch systemd[924]: Failed to start Portal service (GTK/GNOME implementation).
Feb 24 19:15:10 arch systemd[924]: Starting Portal service (GTK/GNOME implementation)...
Feb 24 19:15:10 arch xdg-desktop-por[2663]: cannot open display: :0
…

but I don't think that's critical.
Did you ever attempt to run nemo during that boot?

dr1fter · 2024-02-26 07:49:32

seth wrote:

Did you ever attempt to run nemo during that boot?

Well, technically, I never rebooted since before I created this thread (I did put my machine to sleep + resumed a few times). So yes, I started `nemo` a couple of times (actually, multiple times before and after posting / reading a reply from you (seth) to this thread), and ever since I exported env-var `NO_AT_BRIDGE=1`, I no longer observed the crashes any more.

I sometimes observe slight glitch when resuming from standby: Typically, I can wake-up my machine by pressing any key on my keyboard (which is nice). Sometimes (e.g. just a couple of minutes ago), however, any keystrokes are ignored when my machine is in standby. In this case, I can wake it up by pressing the power-button; in most cases, I need to press it for 1..2s (as opposed to just a very short "click"). Maybe this is caused from those shutdown-errors you saw (truth be told, this glitch never bugged me that much, so I just accepted it thus-far).

As for disabling at-api: is all I need to do exporting this magic env-var, or is there more to do for full disabling?

Last edited by dr1fter (2024-02-26 07:52:27)

seth · 2024-02-26 08:25:20

You posted the journal of the previous boot (as suggested because I wasn't sure whether you had rebooted) - it stops w/ a shutdown.

sudo journalctl -b | curl -F 'file=@-' 0x0.st

Wakeup triggers as ACPI driven, they're not related to at-spi and you don't quit the session for an S3
Either the keyboard bus powers down or gets disabled in /proc/acpi/wakeup or the your BIOS is buggy/confused (the latter in case there's a parallel, hibernating windows)

is there more to do for full disabling?

Nope.

dr1fter · 2024-02-26 08:41:45

gotcha. I dumped the log from current boot:

http://0x0.st/HRXH.txt

No parallel Windows (or actually any other OS) on my machine :-)

So I guess disabling at-spi did the trick for me (still: strange that this started to be broken all of the sudden). Thank you so much! Should I add a `[Solved]` token to the thread's title?

seth · 2024-02-26 09:13:52

There're no obvious errors in that journal, nemo/at-spi might object to the missing portal implementation

Feb 25 09:33:49 arch sudo[237377]:  dr1fter : TTY=pts/1 ; PWD=/mnt/shared_profile/src/sap/ocm ; USER=root ; COMMAND=/usr/bin/pacman -Qs portal
Feb 25 09:34:26 arch sudo[237417]:  dr1fter : TTY=pts/1 ; PWD=/mnt/shared_profile/src/sap/ocm ; USER=root ; COMMAND=/usr/bin/pacman -Qi xdg-desktop-portal-xapp
Feb 25 09:34:39 arch sudo[237592]:  dr1fter : TTY=pts/1 ; PWD=/mnt/shared_profile/src/sap/ocm ; USER=root ; COMMAND=/usr/bin/pacman -Qi xdg-desktop-portal-gtk
Feb 25 09:37:36 arch sudo[238634]:  dr1fter : TTY=pts/1 ; PWD=/mnt/shared_profile/src/sap/ocm ; USER=root ; COMMAND=/usr/bin/pacman -R xdg-desktop-portal-gtk
Feb 25 09:38:00 arch sudo[238689]:  dr1fter : TTY=pts/1 ; PWD=/mnt/shared_profile/src/sap/ocm ; USER=root ; COMMAND=/usr/bin/pacman -R xdg-desktop-portal-xapp
Feb 25 09:38:12 arch sudo[238694]:  dr1fter : TTY=pts/1 ; PWD=/mnt/shared_profile/src/sap/ocm ; USER=root ; COMMAND=/usr/bin/pacman -S xdg-desktop-portal-gtk

But it got never started.

You could re-login/reboot w/o NO_AT_BRIDGE=1 to see whether that was the cause (for posterity), then export the variable and yes, please always remember to mark resolved threads by editing your initial posts subject - so others will know that there's no task left, but maybe a solution to find.
Thanks.

dr1fter · 2024-02-26 18:23:02

So I just rebooted (using `reboot now`) twice, first with `NO_AT_BRIDGE=1` exported in my user's $HOME/.bashrc, then without NO_AT_BRIDGE being exported (I decided against adding into /etc/profile or similar as it seemed not to be necessary from previous observation).

In both cases, I started my usual set of terminals, and started `nemo` as a (detached; using `( setsid "$@" &> /dev/null & )`) (sub)process from one of those terminals (my preferred way of starting (sub)processes). This time, in both cases, I did not observe any issues starting `nemo`, nor any other deviations from expected behaviour.

So in summary:

- my issues are all resolved (thanks a lot your your help seth!)
- if first time of exporting NO_AT_BRIDGE=1 + starting nemo with this env-var set did _not_ have a permanent side-effect, then I am at a loss considering what actually _did_ fix the issues
- it seems generally like a sane choice to export this env-var if not requiring accessibility-features

Arch Linux

#1 2024-02-25 09:29:57

[solved] nemo crashes + chrome freezes + systemctl suspend hangs

#2 2024-02-25 14:34:07

Re: [solved] nemo crashes + chrome freezes + systemctl suspend hangs

#3 2024-02-25 20:44:44

Re: [solved] nemo crashes + chrome freezes + systemctl suspend hangs

#4 2024-02-25 21:09:02

Re: [solved] nemo crashes + chrome freezes + systemctl suspend hangs

#5 2024-02-25 21:30:09

Re: [solved] nemo crashes + chrome freezes + systemctl suspend hangs

#6 2024-02-25 21:46:54

Re: [solved] nemo crashes + chrome freezes + systemctl suspend hangs

#7 2024-02-26 07:49:32

Re: [solved] nemo crashes + chrome freezes + systemctl suspend hangs

#8 2024-02-26 08:25:20

Re: [solved] nemo crashes + chrome freezes + systemctl suspend hangs

#9 2024-02-26 08:41:45

Re: [solved] nemo crashes + chrome freezes + systemctl suspend hangs

#10 2024-02-26 09:13:52

Re: [solved] nemo crashes + chrome freezes + systemctl suspend hangs

#11 2024-02-26 18:23:02

Re: [solved] nemo crashes + chrome freezes + systemctl suspend hangs

Board footer