kwin 100% of cpu, second session

[DimanNe]

Steps to reproduce
Login as user1
Login as user2
Press Ctrl+Alt+F7 (to switch to first user)

kwin from second user starts to eat 100% of cpu.

Here is my investigation of what is going on

1) ====== GDB ======
I attached to running (and eating 100% of cpu) kwin, paused ("ctrl+c" command) and run ("continue" command) kwin several times and watched output of info threads command to figure out what thread is eating 100%.
What I got:

(gdb) info threads
  Id Target Id Frame
  3 Thread 0x7fa21ecb6700 (LWP 25128) "QProcessManager" 0x00007fa23f3c8de3 in select () at ../sysdeps/unix/syscall-template.S:81
  2 Thread 0x7fa21c9c6700 (LWP 25131) "kwin" pthread_cond_wait@@GLIBC_2.3.2 ()
    at ../nptl/sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S:185
* 1 Thread 0x7fa23fbfe7c0 (LWP 25123) "kwin" 0x00007fa236ea7f84 in __GI___pthread_mutex_lock (mutex=0x2520918) at pthread_mutex_lock.c:54
(gdb) continue

// continue, wait some time, ctrl+c

(gdb) info threads
  Id Target Id Frame
  3 Thread 0x7fa21ecb6700 (LWP 25128) "QProcessManager" 0x00007fa23f3c8de3 in select () at ../sysdeps/unix/syscall-template.S:81
  2 Thread 0x7fa21c9c6700 (LWP 25131) "kwin" pthread_cond_wait@@GLIBC_2.3.2 ()
    at ../nptl/sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S:185
* 1 Thread 0x7fa23fbfe7c0 (LWP 25123) "kwin" 0x00007fa23f3d1957 in __libc_recv (fd=9, buf=0x2520974, n=4096, flags=-1, flags@entry=0)
    at ../sysdeps/unix/sysv/linux/x86_64/recv.c:33

// continue, wait some time, ctrl+c

(gdb) info threads
  Id Target Id Frame
  3 Thread 0x7fa21ecb6700 (LWP 25128) "QProcessManager" 0x00007fa23f3c8de3 in select () at ../sysdeps/unix/syscall-template.S:81
  2 Thread 0x7fa21c9c6700 (LWP 25131) "kwin" pthread_cond_wait@@GLIBC_2.3.2 ()
    at ../nptl/sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S:185
* 1 Thread 0x7fa23fbfe7c0 (LWP 25123) "kwin" 0x00007fa23f3c8de3 in select () at ../sysdeps/unix/syscall-template.S:81

and so on...

As you can see first thread is really doing something.

Now, when I know thread number I can see backtrace:
First backtrace
[Switching to thread 1 (Thread 0x7fa23fbfe7c0 (LWP 25123))]
#0 0x00007fa238b11303 in ?? () from /usr/lib/x86_64-linux-gnu/libQtGui.so.4
(gdb) backtrace
#0 0x00007fa238b11303 in ?? () from /usr/lib/x86_64-linux-gnu/libQtGui.so.4
#1 0x00007fa2396d05ef in QEventLoop::processEvents(QFlags<QEventLoop::ProcessEventsFlag>) () from /usr/lib/x86_64-linux-gnu/libQtCore.so.4
#2 0x00007fa2396d08e5 in QEventLoop::exec(QFlags<QEventLoop::ProcessEventsFlag>) () from /usr/lib/x86_64-linux-gnu/libQtCore.so.4
#3 0x00007fa2396d5e5b in QCoreApplication::exec() () from /usr/lib/x86_64-linux-gnu/libQtCore.so.4
#4 0x00007fa23f719956 in kdemain () from /usr/lib/kde4/libkdeinit/libkdeinit4_kwin.so
#5 0x00007fa23f2f7de5 in __libc_start_main (main=0x4006d0, argc=3, ubp_av=0x7ffffe010848, init=<optimized out>, fini=<optimized out>,
    rtld_fini=<optimized out>, stack_end=0x7ffffe010838) at libc-start.c:260
#6 0x00000000004006fe in _start ()

// continue, wait some time, ctrl+c

And second one (after "continue" and "ctrl+c")
(gdb) backtrace
#0 0x00007fa23f3c8de3 in select () at ../sysdeps/unix/syscall-template.S:81
#1 0x00007fa2396fc33b in qt_safe_select(int, fd_set*, fd_set*, fd_set*, timeval const*) () from /usr/lib/x86_64-linux-gnu/libQtCore.so.4
#2 0x00007fa239701904 in QEventDispatcherUNIXPrivate::doSelect(QFlags<QEventLoop::ProcessEventsFlag>, timeval*) ()
   from /usr/lib/x86_64-linux-gnu/libQtCore.so.4
#3 0x00007fa239701d02 in QEventDispatcherUNIX::processEvents(QFlags<QEventLoop::ProcessEventsFlag>) ()
   from /usr/lib/x86_64-linux-gnu/libQtCore.so.4
#4 0x00007fa238b112f6 in ?? () from /usr/lib/x86_64-linux-gnu/libQtGui.so.4
#5 0x00007fa2396d05ef in QEventLoop::processEvents(QFlags<QEventLoop::ProcessEventsFlag>) () from /usr/lib/x86_64-linux-gnu/libQtCore.so.4
#6 0x00007fa2396d08e5 in QEventLoop::exec(QFlags<QEventLoop::ProcessEventsFlag>) () from /usr/lib/x86_64-linux-gnu/libQtCore.so.4
#7 0x00007fa2396d5e5b in QCoreApplication::exec() () from /usr/lib/x86_64-linux-gnu/libQtCore.so.4
#8 0x00007fa23f719956 in kdemain () from /usr/lib/kde4/libkdeinit/libkdeinit4_kwin.so
#9 0x00007fa23f2f7de5 in __libc_start_main (main=0x4006d0, argc=3, ubp_av=0x7ffffe010848, init=<optimized out>, fini=<optimized out>,
    rtld_fini=<optimized out>, stack_end=0x7ffffe010838) at libc-start.c:260
#10 0x00000000004006fe in _start ()
(gdb)

2) ====== strace ======
Now lets see which syscalls are called, I ran this command
strace -p 25123 -s 1000 -o strace1.txt
And in 1-2 seconds I got file (with 50000 lines) which contains ONLY these messages:

select(29, [4 5 9 10 14 28], [], [], {0, 0}) = 0 (Timeout)
recvfrom(9, 0x2520974, 4096, 0, 0, 0) = -1 EAGAIN (Resource temporarily unavailable)
select(29, [4 5 9 10 14 28], [], [], {0, 0}) = 0 (Timeout)
recvfrom(9, 0x2520974, 4096, 0, 0, 0) = -1 EAGAIN (Resource temporarily unavailable)
select(29, [4 5 9 10 14 28], [], [], {0, 0}) = 0 (Timeout)
recvfrom(9, 0x2520974, 4096, 0, 0, 0) = -1 EAGAIN (Resource temporarily unavailable)
select(29, [4 5 9 10 14 28], [], [], {0, 0}) = 0 (Timeout)
recvfrom(9, 0x2520974, 4096, 0, 0, 0) = -1 EAGAIN (Resource temporarily unavailable)
select(29, [4 5 9 10 14 28], [], [], {0, 0}) = 0 (Timeout)

Notice first argument of recvfrom - 9, this fd we already saw at gdb "info threads":
* 1 Thread 0x7fa23fbfe7c0 (LWP 25123) "kwin" 0x00007fa23f3d1957 in __libc_recv (fd=9, buf=0x2520974, n=4096, flags=-1,

ProblemType: Bug
DistroRelease: Ubuntu 13.10
Package: kde-window-manager 4:4.11.3-0ubuntu0.1
ProcVersionSignature: Ubuntu 3.11.0-15.23-generic 3.11.10
Uname: Linux 3.11.0-15-generic x86_64
NonfreeKernelModules: nvidia
ApportVersion: 2.12.5-0ubuntu2.2
Architecture: amd64
Date: Fri Jan 24 21:20:31 2014
InstallationDate: Installed on 2013-11-11 (74 days ago)
InstallationMedia: Kubuntu 13.10 "Saucy Salamander" - Release amd64 (20131016.1)
MarkForUpload: True
SourcePackage: kde-workspace
UpgradeStatus: No upgrade log present (probably fresh install)

Arch forum posts:
https://bbs.archlinux.org/viewtopic.php?id=186904
https://bbs.archlinux.org/viewtopic.php?id=173030

[DimanNe (dimanne)]

I am not sure, but it seems, that function denoted in gdb backtrace as
#0 0x00007fa238b11303 in ?? () from /usr/lib/x86_64-linux-gnu/libQtGui.so.4
is
bool QEventDispatcherX11::processEvents(QEventLoop::ProcessEventsFlags flags)
from
qt-src/src/gui/kernel/qeventdispatcher_x11.cpp

[Rohan Garg (rohangarg)]

Hi
Could you please install the Qt4 debug symbols (libqt4-dbg) and then submit a bug report on https://bugs.kde.org/enter_bug.cgi?format=guided against the 'kwin' product.

If you're running KDE SC < 4.12 . I would recommend upgrading to KDE SC 4.12.1 via the Kubuntu Backports PPA ( https://launchpad.net/~kubuntu-ppa/+archive/backports ) , reproducing the bug and reporting the bug on the KDE Bug tracker with appropriate backtraces.

[John Hubbard (ender8282)]

Was this ever reported upstream? If so would someone kindly piont me to the upstream bug. I am currently seeing this on kubuntu-14.04 using kde 4.13.

[Paul Andreassen (paulone)]

I've seen the lockup from kwin using 100% cpu on two different Kubuntu 14.04 machines, without needing to switch users. These machines were ungrades from previous versions, 12.04 and 13.10.

I think its fixed by resetting the kwin configure file by:

run "System Settings"
select "Configure desktop effects"
then clicking "Defaults" (at the bottom) and then "Apply" (at the bottom right).

If you can get to term with Alt-Ctrl-F1, you can use "DISPLAY=:0 kwin --replace" to fix a broken kwin using 100% cpu. This also works if you can "ssh" in.

[Dima Ryazanov (dima-gmail)]

I keep running into this every time I switch away from X. Here's a stack trace with Qt debug symbols, in KDE 4.13.3:

#0 0x00007f7514c4f8e3 in select () at ../sysdeps/unix/syscall-template.S:81
#1 0x00007f750f0401c7 in qt_safe_select (nfds=23, fdread=0xcf7f88, fdwrite=0xcf8220,
    fdexcept=0xcf84b8, orig_timeout=0x7fff4a848190) at kernel/qcore_unix.cpp:89
#2 0x00007f750f045824 in QEventDispatcherUNIXPrivate::doSelect (this=this@entry=0xcf7dd0,
    flags=..., timeout=0x7fff4a848190) at kernel/qeventdispatcher_unix.cpp:223
#3 0x00007f750f045c92 in QEventDispatcherUNIX::processEvents (this=this@entry=0xc98af0,
    flags=...) at kernel/qeventdispatcher_unix.cpp:936
#4 0x00007f750e4514d6 in QEventDispatcherX11::processEvents (this=0xc98af0, flags=...)
    at kernel/qeventdispatcher_x11.cpp:152
#5 0x00007f750f0140af in QEventLoop::processEvents (this=this@entry=0x7fff4a848460,
    flags=...) at kernel/qeventloop.cpp:149
#6 0x00007f750f0143a5 in QEventLoop::exec (this=this@entry=0x7fff4a848460, flags=...)
    at kernel/qeventloop.cpp:204
#7 0x00007f750f019b79 in QCoreApplication::exec () at kernel/qcoreapplication.cpp:1225
#8 0x00007f750e3ad37c in QApplication::exec () at kernel/qapplication.cpp:3828
#9 0x00007f7514f9e9d6 in kdemain (argc=3, argv=0x7fff4a848738) at ../../kwin/main.cpp:597
#10 0x00007f7514b7eec5 in __libc_start_main (main=0x4006d0 <main(int, char**)>, argc=3,
    argv=0x7fff4a848738, init=<optimized out>, fini=<optimized out>,
    rtld_fini=<optimized out>, stack_end=0x7fff4a848728) at libc-start.c:287
#11 0x00000000004006fe in _start ()

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.