lkml.org 
[lkml]   [2007]   [Jun]   [13]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectCFS-v16: top shows incorrect CPU% for multi-threaded application
Hi,

When I run a multithreaded application, consisting of a "main thread"
that is mostly idle, and
3 "worker threads" (using as much CPU as they can get), 'top' and 'ps'
show that
the application uses 0% CPU.
If I turn off thread details in top, and ps it correctly shows the CPU
usage of each thread.

This happens only on 2.6.22-rc4-cfs-v16, on 2.6.21-1 everything is ok
(top shows sum-of-all-threads-CPU-usage, i.e. ~98,99%).

Can I help debug/track down the cause of this issue? (how?)

uname -a: Linux lightspeed2 2.6.22-rc4-cfs-v16-g845a2fdc-dirty #2 SMP
Sun Jun 10 14:23:08 EEST 2007 x86_64 GNU/Linux

Running top with 'show threads off':
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
5932 clamav 20 0 277m 53m 1608 S 0.0 2.7 0:00.81 clam

Running top with 'show threads on':
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
5953 clamav 20 0 277m 53m 1608 R 99.9 2.7 1:50.33 clamd
5952 clamav 20 0 277m 53m 1608 R 99.9 2.7 1:49.18 clamd
5956 clamav 20 0 277m 53m 1608 R 99.9 2.7 1:47.72 clamd
5932 clamav 20 0 277m 53m 1608 S 0.0 2.7 0:00.81 clamd
5951 clamav 20 0 277m 53m 1608 S 0.0 2.7 0:01.78 clamd

^First time, after second refresh the 99% is down to 30%

top - 22:25:06 up 16 min, 1 user, load average: 3.51, 3.45, 2.23
Tasks: 161 total, 4 running, 157 sleeping, 0 stopped, 0 zombie
Cpu(s): 96.3%us, 2.7%sy, 0.0%ni, 0.0%id, 0.0%wa, 0.3%hi, 0.7%si, 0.0%st
Mem: 2060820k total, 2041144k used, 19676k free, 33948k buffers
Swap: 0k total, 0k used, 0k free, 1502632k cached

Somewhat later:

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
5953 clamav 20 0 279m 52m 1632 R 30.9 2.6 4:14.37 clamd
5956 clamav 20 0 279m 52m 1632 R 30.9 2.6 4:11.67 clamd
5952 clamav 20 0 279m 52m 1632 R 28.9 2.6 4:14.20 clamd
5951 clamav 20 0 279m 52m 1632 S 0.3 2.6 0:04.55 clamd
5932 clamav 20 0 279m 52m 1632 S 0.0 2.6 0:00.81 clamd

Also same results from ps:

edwin@lightspeed2:~$ ps -eLf|grep clamd
clamav 5932 1 5932 0 5 22:10 ? 00:00:00 clamd
clamav 5932 1 5951 0 5 22:11 ? 00:00:02 clamd
clamav 5932 1 5952 31 5 22:11 ? 00:02:27 clamd
clamav 5932 1 5953 31 5 22:11 ? 00:02:28 clamd
clamav 5932 1 5956 31 5 22:11 ? 00:02:26 clamd
edwin 5950 5731 5950 0 1 22:11 pts/3 00:00:00 clamdscan
--quiet -m .
edwin 6746 5729 6746 0 1 22:18 pts/1 00:00:00 grep clamd
edwin@lightspeed2:~$ ps -e|grep clamd
5932 ? 00:00:00 clamd
5950 pts/3 00:00:00 clamdscan

Output of /proc/sched_debug:
Sched Debug Version: v0.02
now at 449457590200 nsecs

cpu: 0
.nr_running : 5
.raw_weighted_load : 4206
.nr_switches : 716023
.nr_load_updates : 448815
.nr_uninterruptible : 0
.jiffies : 4295116153
.next_balance : 4295116155
.curr->pid : 6599
.clock : 356291196511
.prev_clock_raw : 459148341095
.clock_warps : 0
.clock_overflows : 8686
.clock_unstable_events : 0
.clock_max_delta : 1998116
.fair_clock : 192139480240
.delta_fair_clock : 200255631498
.exec_clock : 350544921775
.delta_exec_clock : 350544921775
.wait_runtime : -48647436
.wait_runtime_overruns : 280656
.wait_runtime_underruns: 4285
.cpu_load[0] : 4206
.cpu_load[1] : 4206
.cpu_load[2] : 4190
.cpu_load[3] : 4024
.cpu_load[4] : 3736
.wait_runtime_rq_sum : 3955762

runnable tasks:
task PID tree-key delta waiting
switches prio sum-exec sum-wait sum-sleep
wait-overrun wait-underrun
------------------------------------------------------------------------------------------------------------------------------------------------------------------
clamd 5952 192137574072 -1906168 449816
98914 120 102684256359 2613992521 8984991859
1006 245
clamd 5953 192139027677 -452563 448071
97449 120 103668648209 1738317592 5672775365
542 323
clamd 5956 192137649522 -1830718 1097630
97852 120 101209019733 2430217564 7866850540
957 220
munin-graph 6561 192137708550 -1771690 -275731
3885 130 2792201054 -761359 0
0 0
R cat 6599 192137244264 -2235976 2235976
2 120 61234 2235976 84374

If you need further info, please ask.

Best regards,
--Edwin
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2007-06-13 21:39    [W:0.049 / U:1.316 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site