[Bug 1063060] New: [Build 20171010] gnome fails to load cirrus module when restart x11 using systemctl isolate
http://bugzilla.opensuse.org/show_bug.cgi?id=1063060 Bug ID: 1063060 Summary: [Build 20171010] gnome fails to load cirrus module when restart x11 using systemctl isolate Classification: openSUSE Product: openSUSE Tumbleweed Version: Current Hardware: Other URL: http://openqa.opensuse.org/tests/503069/modules/multi_ users_dm/steps/49 OS: Other Status: NEW Severity: Normal Priority: P5 - None Component: GNOME Assignee: bnc-team-gnome@forge.provo.novell.com Reporter: riafarov@suse.com QA Contact: qa-bugs@suse.de Found By: --- Blocker: --- Created attachment 744138 --> http://bugzilla.opensuse.org/attachment.cgi?id=744138&action=edit Set of logs ## Observation In our test we verify how X behaves with multiple users. To do that we need to diable autologin and consequently restart x11, which we do by issuing systemctl isolate multi-user.target and systemctl isolate graphical.target commands. Steps to reproduce: 1) Create 500 users in the system. Use following script: #!/bin/bash n_users="$1" encrypted_password="$2" for i in `seq 1 $n_users` ; do num=`printf "%02d" $i` useradd -m -d /home/user${i} -c "User #${num}" -p "$encrypted_password" "user${i}" done which accepts number of users to create and password 2) Unload and restart dm: systemctl isolate multi-user.target and systemctl isolate graphical.target 3) Restart x11 second time: systemctl isolate multi-user.target and systemctl isolate graphical.target After that results differ. With previous build system got frozen, probably because bug in cirrus driver. With current builds it doesn't start, here we get stuck in loading screen: [multi_users_dm](http://openqa.opensuse.org/tests/503069/modules/multi_users_dm/steps/49) After that it's not possible to start gdm normally, and it's not possible to switch to another tty, which may be qemu issue. I also was not able to reproduce issue when using virtio vga. Please find full set of logs in the attachment. In the logs I can see following error: Oct 12 11:24:02 susetest /usr/lib/gdm/gdm-x-session[3970]: (WW) CIRRUS: More than one matching Device section found: Autoconfigured Video Device cirrus Oct 12 11:24:02 susetest /usr/lib/gdm/gdm-x-session[3970]: (--) Chipset CLGD5446 found Oct 12 11:24:02 susetest /usr/lib/gdm/gdm-x-session[3970]: (EE) cirrus: The PCI device 0xb8 at 00@00:02:0 has a kernel module claiming it. Oct 12 11:24:02 susetest /usr/lib/gdm/gdm-x-session[3970]: (EE) cirrus: This driver cannot operate until it has been unloaded. [ 606.718] (WW) CIRRUS: More than one matching Device section found: Autoconfigured Video Device cirrus [ 606.718] (--) Chipset CLGD5446 found [ 606.718] (EE) cirrus: The PCI device 0xb8 at 00@00:02:0 has a kernel module claiming it. [ 606.718] (EE) cirrus: This driver cannot operate until it has been unloaded. [ 606.718] (II) modeset(0): using drv /dev/dri/card0 [ 606.718] (WW) Falling back to old probe method for fbdev [ 606.718] (II) Loading sub module "fbdevhw" [ 606.718] (II) LoadModule: "fbdevhw" [ 606.718] (II) Loading /usr/lib64/xorg/modules/libfbdevhw.so [ 606.719] (II) Module fbdevhw: vendor="X.Org Foundation" [ 606.719] compiled for 1.19.4, module version = 0.0.2 [ 606.719] ABI class: X.Org Video Driver, version 23.0 [ 606.719] (EE) open /dev/fb0: Permission denied [ 606.719] (WW) Falling back to old probe method for vesa [ 606.719] (WW) VGA arbiter: cannot open kernel arbiter, no multi-card support [ 606.719] (II) modeset(0): Using 24bpp hw front buffer with 32bpp shadow openQA test in scenario opensuse-Tumbleweed-DVD-x86_64-extra_tests_on_gnome@64bit fails in [multi_users_dm](http://openqa.opensuse.org/tests/503069/modules/multi_users_dm/steps/49) ## Reproducible Fails since (at least) Build [20170531](http://openqa.opensuse.org/tests/412233) ## Expected result Last good: (unknown) (or more recent) ## Further details Always latest result in this scenario: [latest](http://openqa.opensuse.org/tests/latest?distri=opensuse&arch=x86_64&machine=64bit&flavor=DVD&test=extra_tests_on_gnome&version=Tumbleweed) -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1063060 Rodion Iafarov <riafarov@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- See Also| |https://bugzilla.opensuse.o | |rg/show_bug.cgi?id=1020363 -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1063060 http://bugzilla.opensuse.org/show_bug.cgi?id=1063060#c1 Simon Lees <simonf.lees@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |simonf.lees@suse.com --- Comment #1 from Simon Lees <simonf.lees@suse.com> --- "systemctl restart display-manager" is likely a better way to restart x11, not sure if it will help you in this case though. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1063060 http://bugzilla.opensuse.org/show_bug.cgi?id=1063060#c2 --- Comment #2 from Rodion Iafarov <riafarov@suse.com> --- (In reply to Simon Lees from comment #1)
"systemctl restart display-manager" is likely a better way to restart x11, not sure if it will help you in this case though.
Hi Simon. Yes, that's true. Whereas it's still common way used for unloading gui to switch target levels. We will consider to replace this part in the test though. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1063060 http://bugzilla.opensuse.org/show_bug.cgi?id=1063060#c7 Ludwig Nussel <lnussel@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Priority|P5 - None |P2 - High Component|GNOME |X.Org Version|Current |Leap 15.0 Assignee|bnc-team-gnome@forge.provo. |xorg-maintainer-bugs@forge. |novell.com |provo.novell.com Product|openSUSE Tumbleweed |openSUSE Distribution Target Milestone|--- |Leap 15.0 QA Contact|qa-bugs@suse.de |xorg-maintainer-bugs@forge. | |provo.novell.com --- Comment #7 from Ludwig Nussel <lnussel@suse.com> --- Is this really a gnome issue? Sounds more like something related to X and the cirrus driver itself. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1063060 http://bugzilla.opensuse.org/show_bug.cgi?id=1063060#c8 --- Comment #8 from Stefan Dirsch <sndirsch@suse.com> --- The X log is probably misleading. cirrus UMS driver is being tried first (although I'm not sure why it still exists, it has been dropped with sle15), but fails since KMS is active (no issue at all), then also fbdev is being tried, but gdm now starts the Xserver for the session as regular user, which cannot access /dev/fb0 (permissions are 660 for root.video). Then modesetting. Here the logfile got truncated! My guess is that it is successful. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1063060 http://bugzilla.opensuse.org/show_bug.cgi?id=1063060#c9 --- Comment #9 from Stefan Dirsch <sndirsch@suse.com> --- Ok. According to the logfiles I could find at the references in openqa and attached to this bugreport modesetting driver appears to work fine in 1024x768 mode on cirrus VGA emulation of qemu. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1063060 Stefan Dirsch <sndirsch@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |riafarov@suse.com Flags| |needinfo?(riafarov@suse.com | |) -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1063060 Max Staudt <mstaudt@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |sndirsch@suse.com Component|X.Org |GNOME Assignee|xorg-maintainer-bugs@forge. |bnc-team-gnome@forge.provo. |provo.novell.com |novell.com QA Contact|xorg-maintainer-bugs@forge. |qa-bugs@suse.de |provo.novell.com | -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1063060 Oliver Kurz <okurz@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |okurz@suse.com -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1063060 http://bugzilla.opensuse.org/show_bug.cgi?id=1063060#c16 --- Comment #16 from Ludwig Nussel <lnussel@suse.com> --- ping! -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1063060 http://bugzilla.opensuse.org/show_bug.cgi?id=1063060#c17 patrick shanahan <paka@opensuse.org> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |paka@opensuse.org --- Comment #17 from patrick shanahan <paka@opensuse.org> --- sounds list the same as https://bugzilla.opensuse.org/show_bug.cgi?id=1020363 which still has not been addressed. removal of /usr/lib/systemd/user/sockets.target.wants/dbus.socket before entering graphical.session appears to allow plasma session access. note, *before* attempting access to runlevel 5 after booting. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1063060 http://bugzilla.opensuse.org/show_bug.cgi?id=1063060#c18 --- Comment #18 from Simon Lees <simonf.lees@suse.com> --- (In reply to patrick shanahan from comment #17)
sounds list the same as https://bugzilla.opensuse.org/show_bug.cgi?id=1020363 which still has not been addressed.
removal of /usr/lib/systemd/user/sockets.target.wants/dbus.socket before entering graphical.session appears to allow plasma session access. note, *before* attempting access to runlevel 5 after booting.
It seems reasonable user session would require a dbus socket, I don't think removing this is the correct answer long term. But I am not the worlds biggest systemd expert maybe there is another way it could be done. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1063060 http://bugzilla.opensuse.org/show_bug.cgi?id=1063060#c19 --- Comment #19 from patrick shanahan <paka@opensuse.org> --- (In reply to Simon Lees from comment #18)
(In reply to patrick shanahan from comment #17)
sounds list the same as https://bugzilla.opensuse.org/show_bug.cgi?id=1020363 which still has not been addressed.
removal of /usr/lib/systemd/user/sockets.target.wants/dbus.socket before entering graphical.session appears to allow plasma session access. note, *before* attempting access to runlevel 5 after booting.
It seems reasonable user session would require a dbus socket, I don't think removing this is the correct answer long term. But I am not the worlds biggest systemd expert maybe there is another way it could be done.
dbus will start and create it's sockets w/o this re your comment #14 to but #1020363 Simon Lees 2017-05-09 01:13:57 UTC (In reply to patrick shanahan from comment #13)
(In reply to Simon Lees from comment #12)
(In reply to patrick shanahan from comment #11)
I would be happy to provide information. But I only just found the request for information by checking status of the report. Seems even though I an listed on the cc: list, I have not received any notice.(In reply to Simon Lees from comment #8)
Patrick I guess we need to debug your issue further to work out whats preventing dbus from closing in this case could you provide the output of "systemctl status user-XXX.slice" where XXX is your uid?
Thanks for the info, i'll try and replicate this tomorrow, its not what I was expecting to see. dbus isn't running in the multi-user.target (which is expected if nothing is using it) so nothing should be preventing it from running when the graphical target is started.
Just confirming you are running tumbleweed?
yes, 20170505
info may be skewed by having remote ssh access. I can alway trigger the failure that way. I can open a graphical session from an clean boot w/o any remote ssh sessions active.
did you want to see the output w/o the remote session?
ps: there is a problem with mail, I am not receiving and my addr is listed in cc
I can replicate the problem here so i'll chase it up with upstream to see what we can do about it. [reply] [−] Comment 15 Simon Lees 2017-05-12 03:00:33 UTC
From the following, it seems that the whole user session is closed when isolate is run which may lead to bigger issues then just with dbus
test@linux-gruo:~> dbus-monitor signal time=1494556806.294641 sender=org.freedesktop.DBus -> destination=:1.1 serial=2 path=/org/freedesktop/DBus; interface=org.freedesktop.DBus; member=NameAcquired string ":1.1" ..... test@linux-gruo:~> date Fri May 12 12:10:10 ACST 2017 test@linux-gruo:~> date Fri May 12 12:10:32 ACST 2017 < systemctl isolate graphical.target run from a different session> test@linux-gruo:~> dbus-monitor Failed to open connection to session bus: Failed to connect to socket /run/user/1001/bus: Connection refused test@linux-gruo:~> date Fri May 12 12:10:43 ACST 2017 test@linux-gruo:~> journalctl Hint: You are currently not seeing messages from other users and the system. Users in the 'systemd-journal' group can see all messages. Pass -q to turn off this notice. ...... May 12 12:10:01 linux-gruo systemd[1554]: Reached target Timers. May 12 12:10:01 linux-gruo systemd[1554]: Starting D-Bus User Message Bus Socket. May 12 12:10:01 linux-gruo systemd[1554]: Reached target Paths. May 12 12:10:01 linux-gruo systemd[1554]: Listening on D-Bus User Message Bus Socket. May 12 12:10:01 linux-gruo systemd[1554]: Reached target Sockets. May 12 12:10:01 linux-gruo systemd[1554]: Reached target Basic System. May 12 12:10:01 linux-gruo systemd[1554]: Reached target Default. May 12 12:10:01 linux-gruo systemd[1554]: Startup finished in 66ms. May 12 12:10:06 linux-gruo systemd[1554]: Started D-Bus User Message Bus. May 12 12:10:06 linux-gruo dbus-daemon[1603]: Connection :1.1 (uid=1001 pid=1602 comm="dbus-monitor ") became a monitor. May 12 12:10:08 linux-gruo dbus-daemon[1603]: Monitoring connection :1.1 closed. May 12 12:10:33 linux-gruo systemd[1554]: Stopping D-Bus User Message Bus... May 12 12:10:33 linux-gruo systemd[1554]: Stopped target Default. May 12 12:10:33 linux-gruo systemd[1554]: Stopped D-Bus User Message Bus. May 12 12:10:33 linux-gruo systemd[1554]: Stopped target Basic System. May 12 12:10:33 linux-gruo systemd[1554]: Stopped target Timers. May 12 12:10:33 linux-gruo systemd[1554]: Stopped target Paths. May 12 12:10:33 linux-gruo systemd[1554]: Stopped target Sockets. May 12 12:10:33 linux-gruo systemd[1554]: Closed D-Bus User Message Bus Socket. May 12 12:10:33 linux-gruo systemd[1554]: Reached target Shutdown. May 12 12:10:33 linux-gruo systemd[1554]: Starting Exit the Session... May 12 12:10:33 linux-gruo systemd[1554]: Received SIGRTMIN+24 from PID 1661 (kill). May 12 12:10:33 linux-gruo systemd[1555]: pam_unix(systemd-user:session): session closed for user test -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1063060 http://bugzilla.opensuse.org/show_bug.cgi?id=1063060#c21 Dominik Heidler <dheidler@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |dheidler@suse.com --- Comment #21 from Dominik Heidler <dheidler@suse.com> --- Recent failure with odd graphics effects: https://openqa.opensuse.org/tests/656915#step/multi_users_dm/15 -- You are receiving this mail because: You are on the CC list for the bug.
participants (1)
-
bugzilla_noreply＠novell.com