[kde-freebsd] KDE Slows down X

[Kris Moore]

On 09/18/2012 10:54, Kris Moore wrote:
> On 09/18/2012 10:02, Andriy Gapon wrote:
>> on 18/09/2012 16:20 Kris Moore said the following:
>>> KDE Team,
>>>
>>> I've gotten a number of reports from our users, that when using KDE
>>> 4.8.4 and leaving the system up for more than 24-48 hours eventually it
>>> causes Xorg to use 100% CPU and the desktop slows down to a crawl. This
>>> is *only* occurring when KDE is the running desktop, Gnome, LXDE and
>>> others don't seem to have this issue. I've confirmed it here as well on
>>> an ATI system, so its doing it on both that and NVIDIA hardware. Logging
>>> out and back in again fixes the problem.
>>>
>>> Has anybody else run into this? Its easy to duplicate, just leave your
>>> desktop session up for a day or two. I tried disabling all the KDE
>>> Energy Savings in their control panel, but it still occurs. Can anybody
>>> else think of any knobs to tweak?
>>>
>> What HW?  Anything interesting in Xorg.log or system messages when the slowdown
>> happens?
>>
> Its happening on all the systems we've tested at the iX office. Various
> nvidia cards, Intel i5 / i7 chipsets. On the laptop I tested here it is
> running an ATI radeon 5400 series, i5 Proc. When this occurs, I don't
> see any particular messages in dmesg or Xorg.0.log that would indicate
> the problem. We've tried it with 3D enabled and disabled to try and rule
> that out.
>

Further information on this issue. I left two laptops running for about
6 days, both using KDE 4.8.4 and xorg-server-1.1.0.6,1.

One laptop was using the NVIDIA driver, another using ATI with the older
libdrm without KMS. On both systems, X began to use up more and more CPU
with the only thing running being a Konsole session with "top" being
displayed. On each system, no errors are shown in Xorg.0.log, or dmesg,
or anyplace else I could think of to check.

Here's the output of "top" after it had been running for a while:

---------------------------------------------------------
last pid:  7337;  load averages:  0.63,  0.88, 
1.03                               up 6+23:06:15  12:04:32
113 processes: 2 running, 111 sleeping
CPU:  9.6% user,  0.0% nice,  0.1% system,  0.0% interrupt, 90.3% idle
Mem: 1918M Active, 458M Inact, 655M Wired, 21M Cache, 402M Buf, 726M Free
Swap: 2000M Total, 4636K Used, 1995M Free

PID USERNAME   THR PRI NICE   SIZE    RES STATE    C   TIME   WCPU COMMAND
2095 root            1     -21  r31  1684M  1590M CPU1   1 151.2H 38.96%
Xorg
30770 root          1     20   0      14232K  1508K select 2   0:18 
0.98% moused
2931 jamie2        4     24   0      890M     158M select   3  23:19 
0.00% kdeinit4
----------------------------------------------------------

After I logout / login again, it goes back down to a more expected level:

-----------------------------------------------------------
last pid:  9270;  load averages:  0.51,  0.72, 
0.92                               up 6+23:09:44  12:08:01
131 processes: 1 running, 130 sleeping
CPU:  3.7% user,  0.0% nice,  1.7% system,  0.1% interrupt, 94.5% idle
Mem: 645M Active, 604M Inact, 645M Wired, 15M Cache, 402M Buf, 1868M Free
Swap: 2000M Total, 660K Used, 1999M Free

PID USERNAME    THR PRI NICE   SIZE    RES STATE   C   TIME   WCPU COMMAND
7471 root          1  23    0   394M   299M select  3   0:05  2.98% Xorg
9207 jamie2        2  40    0   408M 73716K select  2   0:01  0.98% kdeinit4
2141 root          2  20    0 51784K  4920K select  0   3:06  0.00% upowerd
------------------------------------------------------------

Looking at the memory usage, do you think it could be a memory leak
somewhere?

With 9.1 taking so long to get wrapped up, do you think we may get
4.9.x  into ports before it hits? I'm hopeful it has already been fixed
upstream :)

-- 
Kris Moore
PC-BSD Software
iXsystems