[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Nate Graham]

https://bugs.kde.org/show_bug.cgi?id=354250

Nate Graham  changed:

   What|Removed |Added

 CC||n...@kde.org
 Resolution|--- |FIXED
 Status|REOPENED|RESOLVED

--- Comment #38 from Nate Graham  ---
Let's close it. If anyone an still reproduce this in Plasma 5.13 or later,
please re-open and leave a comment.

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Alexander Mentyu]

https://bugs.kde.org/show_bug.cgi?id=354250

Alexander Mentyu  changed:

   What|Removed |Added

 CC||notux...@gmail.com

--- Comment #37 from Alexander Mentyu  ---
Is the issue can be closed?

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Kai Uwe Broulik via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

Kai Uwe Broulik  changed:

   What|Removed |Added

 CC||akarsh.si...@kdemail.net

--- Comment #36 from Kai Uwe Broulik  ---
*** Bug 369584 has been marked as a duplicate of this bug. ***

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Fabian Köster via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #35 from Fabian Köster  ---
I can confirm that this bug is fixed in 5.7.95 or at least the probability for
it has been reduced.

@Kai
So I guess your patch did help, thank you very much for your persistent effort!

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Fabian Köster via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #34 from Fabian Köster  ---
(In reply to Kai Uwe Broulik from comment #33)
> I think so. That's why I pushed it this morning, tagging is usually around
> 10 to 12

Great, will test your patch as soon as I updated my system.

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Kai Uwe Broulik via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #33 from Kai Uwe Broulik  ---
I think so. That's why I pushed it this morning, tagging is usually around 10
to 12

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Fabian Köster via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #32 from Fabian Köster  ---
(In reply to Kai Uwe Broulik from comment #31)
> When the session becomes inactive, clear all idle timeouts, so we don't get
> them delivered
> while we're inactive or just when we're about to become active again.
> 
> Depending on your setup, PowerDevil might be blocked on X before it gets the
> "session is now
> inactive" signal, so it still suspends when becoming active again :/ It
> should still make it
> a bit less likely (and works for me :D)

That would be great! Will this patch be included in 5.7.95?

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Kai Uwe Broulik via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #31 from Kai Uwe Broulik  ---
Git commit ece8adacb35b8897407d92d995f362a6aea644be by Kai Uwe Broulik.
Committed on 15/09/2016 at 07:15.
Pushed by broulik into branch 'master'.

Clear idle timeouts when session becomes inactive

When the session becomes inactive, clear all idle timeouts, so we don't get
them delivered
while we're inactive or just when we're about to become active again.

Depending on your setup, PowerDevil might be blocked on X before it gets the
"session is now
inactive" signal, so it still suspends when becoming active again :/ It should
still make it
a bit less likely (and works for me :D)

Differential Revision: https://phabricator.kde.org/D2033

M  +15   -0daemon/powerdevilcore.cpp
M  +8-5daemon/powerdevilpolicyagent.cpp
M  +1-0daemon/powerdevilpolicyagent.h

http://commits.kde.org/powerdevil/ece8adacb35b8897407d92d995f362a6aea644be

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Oliver Henshaw via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #30 from Oliver Henshaw  ---
Created attachment 99846
  --> https://bugs.kde.org/attachment.cgi?id=99846=edit
extra logging to help understand the problem

(In reply to Fabian Köster from comment #14)
> (In reply to Kai Uwe Broulik from comment #13)
> > If you can, try https://phabricator.kde.org/D2033 as I don't switch sessions
> > that often :)
> 
> I applied your patch and tested again, but the issue still happens :-(

Kai's first patch (Diff 1, i.e. https://phabricator.kde.org/D2033?id=4799)
seems to work for me, could you try it again with the attached patch on top and
post the logs?

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Rex Dieter via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

Rex Dieter  changed:

   What|Removed |Added

 CC||rdie...@math.unl.edu

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Fabian Köster via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #29 from Fabian Köster  ---
Please note that I do not use Fedora but Gentoo Linux which still executes X as
root.

"ps aux | grep /usr/bin/X" confirms that: 

root  2229  2.2  0.5 407300 89944 tty1 Ssl+ 15:39   1:32 /usr/bin/X
-nolisten tcp -auth /var/run/sddm/{c31eee08-f7e1-45f3-974b-14c6c8998aad}
-background none -noreset -displayfd 18 vt1

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Oliver Henshaw via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #28 from Oliver Henshaw  ---
Note that this wasn't a problem on F21, so this is a regression somewhere, not
necessarily a misdesign feature. Could be an X regression, could be a kde 4 ->
5 regression, could be a qt 4 -> 5 regression.

Note that I see a similar backtrace for Xorg for the session I'm actually in -
that might just be business as usual.

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Martin Gräßlin via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #27 from Martin Gräßlin  ---
both threads are waiting (X can do multi-threading, that's like wow I'm
impressed). Though the one is a graphics thread and is somewhere in radeon.
That could be an indication for being blocked and waiting for something to
happen when resumed. But overall I'm not familiar enough with the code to make
a guess.

So I just assume that non-root X is not doing anything when the session is
switched away. AFAIU their architecture is very similar to what kwin_wayland
does for that situation.

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Oliver Henshaw via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #26 from Oliver Henshaw  ---
(In reply to Martin Gräßlin from comment #24)
> From comment @18 it looks like X freezes while being on the inactive
> session. As Fedora is mentioned I assume it's an X not run as root? If X
> freezes all xcb requests are going to block. From client side there is
> nothing which can be done about that. There is no timeout and no way to
> cancel.

Yep, it's xorg-x11-server-Xorg-1.17.4-1.fc22.x86_64 with the server running as
non-root.

> 
> I would be interested in seeing a backtrace of X at that time as I assume
> it's also frozen.

# pstack 19966
Thread 2 (Thread 0x7f133d647700 (LWP 19995)):
#0  pthread_cond_wait@@GLIBC_2.3.2 () at
../sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S:185
#1  0x7f13401db6da in cnd_wait (mtx=0x27ce4b0, cond=0x27ce4d8) at
../../../../../include/c11/threads_posix.h:154
#2  pipe_semaphore_wait (sema=0x27ce4b0) at
../../../../../src/gallium/auxiliary/os/os_thread.h:259
#3  radeon_drm_cs_emit_ioctl (param=param@entry=0x27ce180) at
radeon_drm_winsys.c:636
#4  0x7f13401dae37 in impl_thrd_routine (p=) at
../../../../../include/c11/threads_posix.h:87
#5  0x7f134963d555 in start_thread (arg=0x7f133d647700) at
pthread_create.c:333
#6  0x7f1349955ded in clone () at
../sysdeps/unix/sysv/linux/x86_64/clone.S:109
Thread 1 (Thread 0x7f134be499c0 (LWP 19966)):
#0  0x7f134994c2e3 in select () at ../sysdeps/unix/syscall-template.S:81
#1  0x005935af in WaitForSomething
(pClientsReady=pClientsReady@entry=0x2aa4530) at WaitFor.c:227
#2  0x0043a3e1 in Dispatch () at dispatch.c:358
#3  0x0043e6eb in dix_main (argc=14, argv=0x7fff89842b38,
envp=) at main.c:298
#4  0x7f1349873700 in __libc_start_main (main=0x428ae0 , argc=14,
argv=0x7fff89842b38, init=, fini=,
rtld_fini=, stack_end=0x7fff89842b28) at libc-start.c:289
#5  0x00428b19 in _start ()

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Fabian Köster via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #25 from Fabian Köster  ---
(In reply to Kai Uwe Broulik from comment #23)
> I just updated the patch at [1] to clear idle timeouts when the session
> becomes inactive.
> 
> [1] https://phabricator.kde.org/D2033

This patch does not solve the issue for me :-/

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Martin Gräßlin via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #24 from Martin Gräßlin  ---
>From comment @18 it looks like X freezes while being on the inactive session.
As Fedora is mentioned I assume it's an X not run as root? If X freezes all xcb
requests are going to block. From client side there is nothing which can be
done about that. There is no timeout and no way to cancel.

I would be interested in seeing a backtrace of X at that time as I assume it's
also frozen.

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Kai Uwe Broulik via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #23 from Kai Uwe Broulik  ---
I just updated the patch at [1] to clear idle timeouts when the session becomes
inactive.

[1] https://phabricator.kde.org/D2033

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Kai Uwe Broulik via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #22 from Kai Uwe Broulik  ---
Thanks. 
What I get looks ok, safe for the idle event mid-air:

2016-06-28T22:09:55 ACTIVE SESSION PATH CHANGED:
"/org/freedesktop/login1/session/_32"
2016-06-28T22:09:55 Current session is now inactive
2016-06-28T22:10:25 Idle timeout reached but we're not the active session, not
triggering actions
2016-06-28T22:12:24 Created alarm 18874377
2016-06-28T22:12:24 Idle timeout reached but we're not the active session, not
triggering actions
[...]
2016-06-28T22:12:25 ACTIVE SESSION PATH CHANGED:
"/org/freedesktop/login1/session/_320"
2016-06-28T22:12:25 Current session is now active

However indeed it should probably unregister all idle events when the session
becomes inactive.

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Oliver Henshaw via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #21 from Oliver Henshaw  ---

> @Oliver: Thanks for the investigation. Indeed, this looks like the cause for
> the issue. When PowerDevil gets stuck because X blocks it never notices it's
> no longer in the active session and shit hits the fan.
> 
Actually if you set QT_MESSAGE_PATTERN so you see timestamps, e.g. following
https://woboq.com/blog/nice-debug-output-with-qt.html , and compare it to your
watch: then you see that the suspend session calls only happen after you've
switched to the other user. The "Current session is now active" lines only
happen later because (a) that's how the timing of dbus-vs-xcb events works out,
and (b) powerdevil does a roundtrip back to the logind dbus service to see what
the currently active session is.

So it gets events fired when it wasn't active, but not until it's active again.
Then the policy code checks whether it's active and of course it is, so the
suspend goes ahead. Then the active-tracking parts of powerdevil realise it's
active just after it resumes.

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Fabian Köster via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #20 from Fabian Köster  ---
Created attachment 99752
  --> https://bugs.kde.org/attachment.cgi?id=99752=edit
gstack log

pstack was not available for my platform so I used gstack. Hope it is useful
nevertheless.

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Kai Uwe Broulik via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

Kai Uwe Broulik  changed:

   What|Removed |Added

 CC||mgraess...@kde.org

--- Comment #19 from Kai Uwe Broulik  ---
@Fabian: Thanks a lot. It seems PowerDevil never actually notices that its no
longer the current session, ie. above the gap you can see "Session is now
active", and a while down you see it triggering suspend and only thereafter it
says again it's now active.

@Oliver: Thanks for the investigation. Indeed, this looks like the cause for
the issue. When PowerDevil gets stuck because X blocks it never notices it's no
longer in the active session and shit hits the fan.

CC'ing Martin as maintainer of KIdleTime.

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Oliver Henshaw via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

Oliver Henshaw  changed:

   What|Removed |Added

 CC||oliver.hens...@gmail.com

--- Comment #18 from Oliver Henshaw  ---
Created attachment 99751
  --> https://bugs.kde.org/attachment.cgi?id=99751=edit
stacktrace showing that main thread of other user's kded5 is blocked in
xcb/xlib

Can you identify the kded5 process PID belonging to the user you're about to
switch to with
$ ps aux | grep kded5

Then do
$ pstack PID
(as root or in logged in as the other user via your terminal) and paste the
results here?

On Fedora 22 I get the attached result. It looks like powerdevil gets stuck in
xcb while waiting for a reply from an Xlib call (one that should return
straight away). And I see QXcbConnection::processXcbEvents further up the
stack. Maybe there's a problem with nesting the xlib-on-xcb eventloop inside an
xcb event?

Whatever's happening, it seems to get unstuck when you actually switch to the
other user and all the kidletime events come flooding in. I also see floods of
other events in kded, e.g. the software update applet spams me with a load of
identical messages telling me how many packages are due for update.

If this is indeed the problem, completing the port of kidletime to xcb looks
like part of the long-term solution. I think there were problems when that was
tried before but hopefully everything that is called from the xcb event can be
ported.

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Fabian Köster via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #17 from Fabian Köster  ---
Created attachment 99750
  --> https://bugs.kde.org/attachment.cgi?id=99750=edit
kded5 log

So, thats the requested log of kded5. Hope it helps!

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Fabian Köster via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #16 from Fabian Köster  ---
(In reply to Kai Uwe Broulik from comment #15)
> Can you run kded5 from console on the other session and switch to it after a
> while? Maybe debug output there can give us a clue what's going on. Look for
> "powerdevil: Current session is now (in)active" and what's happening around
> there.

OK, I will do that and report!

> Do you have logind?

Yes, I am using systemd with logind.

loginctl --version
systemd 226
+PAM -AUDIT -SELINUX +IMA -APPARMOR +SMACK -SYSVINIT +UTMP +LIBCRYPTSETUP
-GCRYPT -GNUTLS +ACL -XZ +LZ4 +SECCOMP +BLKID -ELFUTILS +KMOD +IDN

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Kai Uwe Broulik via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #15 from Kai Uwe Broulik  ---
Meh. I cannot reproduce anymore here. All I get is a bunch of "Idle timeout
reached but we're not the active session, not triggering actions" but no
suspend/screen dimming.

Can you run kded5 from console on the other session and switch to it after a
while? Maybe debug output there can give us a clue what's going on. Look for
"powerdevil: Current session is now (in)active" and what's happening around
there. Do you have logind?

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Fabian Köster via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #14 from Fabian Köster  ---
(In reply to Kai Uwe Broulik from comment #13)
> If you can, try https://phabricator.kde.org/D2033 as I don't switch sessions
> that often :)

I applied your patch and tested again, but the issue still happens :-(

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Kai Uwe Broulik via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #13 from Kai Uwe Broulik  ---
If you can, try https://phabricator.kde.org/D2033 as I don't switch sessions
that often :)

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Fabian Köster via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #12 from Fabian Köster  ---
Still happens in 5.7 Beta (5.6.95)

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Gregor Mi via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #11 from Gregor Mi  ---
Reproduced on 5.5.2

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Kai Uwe Broulik via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

Kai Uwe Broulik  changed:

   What|Removed |Added

 Resolution|FIXED   |---
 Status|RESOLVED|REOPENED

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Kai Uwe Broulik via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #9 from Kai Uwe Broulik  ---
*** Bug 356941 has been marked as a duplicate of this bug. ***

-- 
You are receiving this mail because:
You are watching all bug changes.

[Powerdevil] [Bug 354250] "Switch User" sometimes sends computer to Standby

[Fabian Köster via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=354250

--- Comment #8 from Fabian Köster  ---
The problem still occurs with Plasma 5.5.0 (and powerdevil 5.5.0) on Gentoo.
Maybe the fix was not sufficient?

-- 
You are receiving this mail because:
You are watching all bug changes.