Reply to topic

KDE 4.5.4 excruciatingly slow in Fedora 14

User avatar MickO
Registered Member
Posts
8
Karma
0
OS
Greetings - I did scour the board for this issue, but have not come up with anything that matches my situation (I reserve the right to have bad search-fu :< ).

Basically - I have a very nice laptop (Core2Duo T9300, 8GB RAM, Quadro FX 1600M) I run Fedora x86_64 on, and KDE has _always_ been my preferred window manager. Recently I did a Fedora 13 -> 14 preupgrade, and all seemed well. I replaced the harddrive with a larger one and in the process installed F14 fresh, and ever since, KDE is mind-numbingly slow; starting KDE takes 10-15 minutes. Even in GNOME, starting a KDE app takes several minutes. An strace for konsole shows (among other things) that:
Code: Select all
...
sendmsg(3, {msg_name(0)=NULL, msg_iov(2)=[{"l\1\0\1.\0\0\0\v\0\0\0\212\0\0\0\1\1o\0\20\0\0\0/MainApp"..., 160}, {"\0\0\0\0&\0\0\0\0\0\0\22/home/mick.ohrberg\0\0"..., 46}], msg_controllen=0, msg_flags=0}, MSG_NOSIGNAL) = 206
poll([{fd=3, events=POLLIN}], 1, 25000) = 1 ([{fd=3, revents=POLLIN}])
recvmsg(3, {msg_name(0)=NULL, msg_iov(1)=[{"l\2\1\1\302\v\0\0\t\0\0\0/\0\0\0\6\1s\0\6\0\0\0:1.226\0\0"..., 2048}], msg_controllen=0, msg_flags=MSG_CMSG_CLOEXEC}, MSG_CMSG_CLOEXEC) = 2048
recvmsg(3, {msg_name(0)=NULL, msg_iov(1)=[{"ection=\"in\"/>\n    </method>\n    "..., 2048}], msg_controllen=0, msg_flags=MSG_CMSG_CLOEXEC}, MSG_CMSG_CLOEXEC) = 1026
poll([{fd=3, events=POLLIN}], 1, 10621QMetaObject::invokeMethod: No such method Konsole::Application::loadCommandLineOptionsForNewInstance()
) = 1 ([{fd=3, revents=POLLIN}])
recvmsg(3, {msg_name(0)=NULL, msg_iov(1)=[{"l\2\1\1\4\0\0\0\f\0\0\0/\0\0\0\6\1s\0\6\0\0\0:1.226\0\0"..., 2048}], msg_controllen=0, msg_flags=MSG_CMSG_CLOEXEC}, MSG_CMSG_CLOEXEC) = 68
recvmsg(3, 0x7fff1cf05850, MSG_CMSG_CLOEXEC) = -1 EAGAIN (Resource temporarily unavailable)
...

When I look up file descriptor 3 for that PID, it turns out it's a symlink to a socket that doesn't exist. I am not sure yet whether this is relevant to my issue or not. CPU load (and HD activity) during the startup process is negligible.

I gambled that it had to do with dbus (I have since learned that GNOME uses dbus as well) and started ,y system with init=/bin/systemd instead of init=/sbin/startup, but to no avail.

If anyone has seen and resolved an issue like this, please help me get my KDE back :)

Thank you in advance!


--
A: Because it's counterintuitive to the way we read.
Q: Why is top-posting bad?

See http://www.gnu.org/philosophy/no-word-attachments.html
User avatar google01103
Manager
Posts
6668
Karma
25
you can look at this thread, similar problem but op never states if fixed, this specific post may help determine issue viewtopic.php?f=66&t=89804#p169590 states "You may wish to try switching to a virtual terminal and watching ~/.xsession-errors. It may be getting hung on a particular component for some reason ( perhaps the network state has changed the computer's hostname? )"

things to try - 4.6beta, 4.5.3, login as new user (or rename your kde4 dir), dl or run a live cd (Fedora 13, opensuse maybe) to see if it's a Fedora thing (as 14 is a preupgrade)

ps nice laptop


OpenSuse Leap 42.1 x64, Plasma 5.x

User avatar MickO
Registered Member
Posts
8
Karma
0
OS
Thanks! I love this lump - it's huge (17" screen), but she's got it where it counts :) The main reason for the memory is that I'm doing a lot of VMware stuffs.

I did a few things in the thread you suggested, alas! .xsession-errors says nothing. Renaming .kde and .kderc makes no difference, other than losing settings. :)

I want to make sure it's understood that it seems to apply to *all* KDE applications - konsole, amarok, kmahjongg - and worth mentioning is also that once the application is up and running, it seems to be fine.

I _am_ *ahem* running a tainted nVidia kernel, but when I was running with nouveau I got the same result.


--
A: Because it's counterintuitive to the way we read.
Q: Why is top-posting bad?

See http://www.gnu.org/philosophy/no-word-attachments.html
User avatar google01103
Manager
Posts
6668
Karma
25
you could try private messaging the original poster of the thread I suggested and ask if and how he resolved his problem, though he doesn't appear to be an active forum user he may reply

in a terminal session (alt+ctrl+f1, alt+ctrl+f2, etc) you can monitor disk usage with iotop and cpu usage with htop


OpenSuse Leap 42.1 x64, Plasma 5.x

User avatar MickO
Registered Member
Posts
8
Karma
0
OS
I don't have iotop/htop, but I might pull those down. What I did do was utilize nmon (_very_ handy tool, open source contribution by IBM, see http://nmon.sourceforge.net/pmwiki.php?n=Site.Download ) and see that neither disk, memory, cpu nor network is really reacting in any noticeable way when I start konsole, amarok etc. It's almost as though there's a sleep in there for any k* application - very odd indeed.


--
A: Because it's counterintuitive to the way we read.
Q: Why is top-posting bad?

See http://www.gnu.org/philosophy/no-word-attachments.html
User avatar bcooksley
Administrator
Posts
19765
Karma
87
OS
This could possibly be an issue with it being unable to reach D-Bus (which stores sockets in ~/.dbus) or KDEInit (uses sockets via symlinks in ~/.kde4/, actual sockets located at /tmp/ksocket-$USER)

Can you please verify that "dbus-daemon" and "kdeinit4" are both running under your current user after starting a KDE application?


KDE Sysadmin
[img]http://forum.kde.org/content/bcooksley_sig.png[/img]
User avatar MickO
Registered Member
Posts
8
Karma
0
OS
bcooksley wrote:This could possibly be an issue with it being unable to reach D-Bus (which stores sockets in ~/.dbus) or KDEInit (uses sockets via symlinks in ~/.kde4/, actual sockets located at /tmp/ksocket-$USER)

I started konsole, but sockets (or symlinks) do not seem to be created in ~/.dbus; however there is a socket symlink in ~/.kde (there's not ~/.kde4 - I'm assuming that's a Fedora customization):
Code: Select all
$ ls -al ~/.kde
...
lrwxrwxrwx.  1 <owner> <group>   25 Dec 27 07:58 socket-<hostname> -> /tmp/ksocket-<username>
...
$ ls -al /tmp/ksocket-<username>
...
srw-------.  1 <owner> <group>    0 Dec 27 15:47 kdeinit4__0
srwxrwxr-x.  1 <owner> <group>    0 Dec 27 15:47 klauncherMT4559.slave-socket
...

In the above example, <username> = <owner> and <group>, which are UID and GID 500, respectively.

bcooksley wrote:Can you please verify that "dbus-daemon" and "kdeinit4" are both running under your current user after starting a KDE application?

They are indeed both running:
Code: Select all
$ ps -ef |grep kdeinit
500       4557     1  0 Dec27 ?        00:00:00 kdeinit4: kdeinit4 Running...
and
Code: Select all
$ ps -ef | grep dbus
dbus      1382     1  0 Dec27 ?        00:00:22 dbus-daemon --system
500       3495     1  0 Dec27 ?        00:00:00 dbus-launch --sh-syntax --exit-with-session
500       3496     1  0 Dec27 ?        00:00:49 /bin/dbus-daemon --fork --print-pid 5 --print-address 7 --session


This is, as a point of clarification, the konsole application launched from within GNOME.


--
A: Because it's counterintuitive to the way we read.
Q: Why is top-posting bad?

See http://www.gnu.org/philosophy/no-word-attachments.html
User avatar bcooksley
Administrator
Posts
19765
Karma
87
OS
Note that Konsole itself won't create the socket(s) in ~/.dbus, it merely connects to them.

Just interested, does "qdbus" take a long time to run?


KDE Sysadmin
[img]http://forum.kde.org/content/bcooksley_sig.png[/img]
david04
Registered Member
Posts
3
Karma
0
OS
Hy,

I seem to have the same problem.

Another piece of information:
When I start a KDE app, for example kile, with strace:

Code: Select all
brk(0x2697000)                          = 0x2697000
brk(0x2693000)                          = 0x2693000
poll([{fd=8, events=POLLIN|POLLOUT}], 1, -1) = 1 ([{fd=8, revents=POLLOUT}])
writev(8, [{"\22\0\17\0\t\0\240\6#\0\0\0#\0\0\0 \3\260\1\t\0\0\0C\0\0\0\1\0\0\0"..., 1280}, {NULL, 0}, {"", 0}], 3) = 1280
poll([{fd=8, events=POLLIN}], 1, -1)    = 1 ([{fd=8, revents=POLLIN}])
read(8, "\34\0\342\0\t\0\240\6#\0\0\0\265Cf\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 1184
read(8, 0x24846d4, 4096)                = -1 EAGAIN (Resource temporarily unavailable)
futex(0x7f40e2aa0008, FUTEX_WAIT_BITSET|FUTEX_CLOCK_REALTIME, 2, {1293721719, 0}, ffffffff
<It just hangs in calls like this for a lot of time>
User avatar MickO
Registered Member
Posts
8
Karma
0
OS
bcooksley wrote:Just interested, does "qdbus" take a long time to run?

In fact, it doesn't:
Code: Select all
$ time qdbus
:1.0
 org.gnome.SessionManager
:1.10
 org.gtk.vfs.Daemon
:1.100
 im.pidgin.purple.PurpleService

[snip]

:1.91
:1.93
:1.97
org.freedesktop.DBus

real   0m0.058s
user   0m0.018s
sys   0m0.013s


--
A: Because it's counterintuitive to the way we read.
Q: Why is top-posting bad?

See http://www.gnu.org/philosophy/no-word-attachments.html
User avatar bcooksley
Administrator
Posts
19765
Karma
87
OS
Does this occur only with certain applications (Konsole, Dolphin) or with all applications (ie. KWrite is also affected). Also, have you tried attaching gdb once it has hung to try and backtrace where it has hung? You will need debugging symbols for that (probably for libdbus, qt and kdelibs at least)


KDE Sysadmin
[img]http://forum.kde.org/content/bcooksley_sig.png[/img]
david04
Registered Member
Posts
3
Karma
0
OS
bcooksley wrote:Does this occur only with certain applications (Konsole, Dolphin) or with all applications (ie. KWrite is also affected). Also, have you tried attaching gdb once it has hung to try and backtrace where it has hung? You will need debugging symbols for that (probably for libdbus, qt and kdelibs at least)


For me this occurs with every KDE app I tried (including kwrite).
User avatar bcooksley
Administrator
Posts
19765
Karma
87
OS
Can you please try attaching gdb once it has hung and generating a full backtrace in this case?


KDE Sysadmin
[img]http://forum.kde.org/content/bcooksley_sig.png[/img]
User avatar MickO
Registered Member
Posts
8
Karma
0
OS
bcooksley wrote:Can you please try attaching gdb once it has hung and generating a full backtrace in this case?

I noticed something alarming when I tried it. By mistake I ran gdb as root (ouchie wa-wa) and lo and behold - kwrite, konsole and kbounce started up quickly and without fail. So - I may have lead y'all down the proverbial primrose path of complexity when in reality it's a simple permissions issue. WHAT exactly this permissions issue is, however, I cannot yet tell, so if there are any suggestions on what to check (since I'm not a developer, and CERTAINLY not as KDE savvy as you good folks are), please let me know.


--
A: Because it's counterintuitive to the way we read.
Q: Why is top-posting bad?

See http://www.gnu.org/philosophy/no-word-attachments.html
User avatar MickO
Registered Member
Posts
8
Karma
0
OS
Additional information - once I start Kwrite as my non-root user, the /proc/<pid>/fd directory is rather ugly with MANY red symlinks, for example:
Code: Select all
l-wx------. 1 <uid> <gid> 64 Jan  3 10:21 10 -> pipe:[295904]
l-wx------. 1 <uid> <gid> 64 Jan  3 10:21 11 -> pipe:[295913]
lr-x------. 1 <uid> <gid> 64 Jan  3 10:21 12 -> pipe:[295914]
l-wx------. 1 <uid> <gid> 64 Jan  3 10:21 13 -> pipe:[295914]
lrwx------. 1 <uid> <gid> 64 Jan  3 10:21 14 -> socket:[295915]
lrwx------. 1 <uid> <gid> 64 Jan  3 10:21 15 -> socket:[295917]
lrwx------. 1 <uid> <gid> 64 Jan  3 10:21 16 -> socket:[295920]
lrwx------. 1 <uid> <gid> 64 Jan  3 10:21 17 -> socket:[295921]
lrwx------. 1 <uid> <gid> 64 Jan  3 10:21 18 -> socket:[295923]
lrwx------. 1 <uid> <gid> 64 Jan  3 10:21 19 -> socket:[295925]

When I do the same as the root user, I get the same result, albeit a lot fewer of them - unsure if this is relevant or not.


--
A: Because it's counterintuitive to the way we read.
Q: Why is top-posting bad?

See http://www.gnu.org/philosophy/no-word-attachments.html

 
Reply to topic

Bookmarks



Who is online

Registered users: andrewm, Baidu [Spider], Bing [Bot], Google [Bot], Merlimau, Sogou [Bot], ssanderson, Yahoo [Bot]