| View previous topic :: View next topic |
| Author |
Message |
zzaappp
Apprentice
Joined: 06 Jun 2003
Posts: 176
|
 Posted: Sat Jul 19, 2008 2:45 pm Post subject: top(1) reporting bad cpu percentage [SOLVED] |
 |
|
I was doing a trivial compile build today on the spanking new box I just built, and was a bit surprised to see top(1) reporting the C++ compiler using more than 1000% of available cpu. See the first line of the process report, "cc1plus", below, and look under the "CPU%" column:
| Code: | top - 10:35:55 up 14 min, 4 users, load average: 1.66, 0.59, 0.28
Tasks: 131 total, 4 running, 127 sleeping, 0 stopped, 0 zombie
Cpu(s): 4.2%us, 0.4%sy, 0.0%ni, 94.5%id, 0.8%wa, 0.0%hi, 0.0%si, 0.0%st
Mem: 8310568k total, 346292k used, 7964276k free, 19936k buffers
Swap: 0k total, 0k used, 0k free, 151792k cached
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
7623 root 20 0 48956 41m 4596 R 1805 0.5 0:09.11 cc1plus
7659 root 20 0 19072 14m 4396 R 22 0.2 0:00.11 cc1plus
5384 root 20 0 268m 64m 25m S 4 0.8 1:09.67 java
4330 root 20 0 10516 8536 1956 R 2 0.1 0:55.29 Xvnc
1 root 20 0 1596 548 476 S 0 0.0 0:14.63 init
2 root 15 -5 0 0 0 S 0 0.0 0:00.00 kthreadd
3 root RT -5 0 0 0 S 0 0.0 0:00.00 migration/0
4 root 15 -5 0 0 0 S 0 0.0 0:00.00 ksoftirqd/0
5 root RT -5 0 0 0 S 0 0.0 0:00.00 migration/1
6 root 15 -5 0 0 0 S 0 0.0 0:04.29 ksoftirqd/1
7 root RT -5 0 0 0 S 0 0.0 0:00.00 migration/2
8 root 15 -5 0 0 0 S 0 0.0 0:00.00 ksoftirqd/2
9 root RT -5 0 0 0 S 0 0.0 0:00.00 migration/3
10 root 15 -5 0 0 0 S 0 0.0 0:12.88 ksoftirqd/3
11 root 15 -5 0 0 0 S 0 0.0 0:00.00 events/0
12 root 15 -5 0 0 0 S 0 0.0 0:00.08 events/1
13 root 15 -5 0 0 0 S 0 0.0 0:00.00 events/2
14 root 15 -5 0 0 0 S 0 0.0 0:00.00 events/3
15 root 15 -5 0 0 0 S 0 0.0 0:25.76 khelper
87 root 15 -5 0 0 0 S 0 0.0 0:00.02 kblockd/0
88 root 15 -5 0 0 0 S 0 0.0 0:00.00 kblockd/1
89 root 15 -5 0 0 0 S 0 0.0 0:00.00 kblockd/2
90 root 15 -5 0 0 0 S 0 0.0 0:00.00 kblockd/3
92 root 15 -5 0 0 0 S 0 0.0 0:00.00 kacpid
93 root 15 -5 0 0 0 S 0 0.0 0:04.28 kacpi_notify
176 root 15 -5 0 0 0 S 0 0.0 0:00.00 cqueue
182 root 15 -5 0 0 0 S 0 0.0 0:00.00 ata/0
183 root 15 -5 0 0 0 S 0 0.0 0:00.00 ata/1
184 root 15 -5 0 0 0 S 0 0.0 0:00.00 ata/2
185 root 15 -5 0 0 0 S 0 0.0 0:00.00 ata/3
186 root 15 -5 0 0 0 S 0 0.0 0:00.00 ata_aux
187 root 15 -5 0 0 0 S 0 0.0 0:00.00 ksuspend_usbd
192 root 15 -5 0 0 0 S 0 0.0 0:00.71 khubd
195 root 15 -5 0 0 0 S 0 0.0 0:00.07 kseriod
198 root 15 -5 0 0 0 S 0 0.0 0:00.00 kmmcd
261 root 20 0 0 0 0 S 0 0.0 0:00.00 pdflush
262 root 20 0 0 0 0 S 0 0.0 0:00.00 pdflush
263 root 15 -5 0 0 0 S 0 0.0 0:00.00 kswapd0 |
I originally compiled the kernel as Tickless. When I saw this report, I rebuilt it with Tickless disabled. That made no difference. Just an FYI, this is using an intel Quad Core cpu for a total of 4 cpu cores.
I recall an article a few months ago in LINUX JOURNAL that spoke of a bug in top(1)'s way of calculating cpu usage, but I'm fuzzy on the details, the context that causes it. Has anyone seen this behavior before?
Many thanks!
-z
Last edited by zzaappp on Sun Jul 20, 2008 5:31 pm; edited 1 time in total |
|
| Back to top |
|
 |
alex.blackbit
Advocate
Joined: 26 Jul 2005
Posts: 2397
|
| Posted: Sat Jul 19, 2008 6:26 pm Post subject: |
|
|
very interesting.
the summary on the top says that your system is almost idle. very strange.
when you hit "1", you see the load distribution over the cpus.
you can also try using "htop" maybe it gives different results. |
|
| Back to top |
|
|
zzaappp
Apprentice
Joined: 06 Jun 2003
Posts: 176
|
| Posted: Sun Jul 20, 2008 5:31 am Post subject: |
|
|
I tried htop(1) concurrently with top(1) so I could see both sets of numbers. I then ran the build test so that the CPU's were ramped up fairly high. Straight away there was a clear difference: htop's numbers stayed realistic, and top's numbers spiked impossibly high.
Evidently there is a bug in top. Has something in /proc changed lately? I would really like to take a peek at the code and see what the differences are between the two utilities.
Either way, thanks for the info!
-z |
|
| Back to top |
|
|
Hu
Administrator
Joined: 06 Mar 2007
Posts: 23089
|
| Posted: Sun Jul 20, 2008 4:02 pm Post subject: |
|
|
| You could use dev-util/strace to monitor the system calls of the two processes. It will produce quite a lot of output, but it should show you how each of them is getting their respective inputs, which may give you a clue about what part of /proc to investigate. |
|
| Back to top |
|
|
zzaappp
Apprentice
Joined: 06 Jun 2003
Posts: 176
|
| Posted: Sun Jul 20, 2008 5:30 pm Post subject: |
|
|
I did a quick analysis of the code in top. Top lives in the procps project along with a number of other utilities that need access to /proc. In the case of top(1), a call is made into the procps library to a function called readproc(); readproc parses the contents of files in /proc/%id%/task/*. Evidently the structure of one or more files in /proc/%id%/task/ have changed, and readproc is unaware, or perhaps not compiled correctly to handle it.
htop(1), on the other hand, has its own parser and doesn't use the readproc call that procps provides to top(1).
I could dig deeper, but I think I'll use htop instead.
-z |
|
| Back to top |
|
|
alex.blackbit
Advocate
Joined: 26 Jul 2005
Posts: 2397
|
| Posted: Sun Jul 20, 2008 6:13 pm Post subject: |
|
|
htop isn't a standard unix tool, but i use it myself a lot, because it just works very well and is feature rich.
good choice. |
|
| Back to top |
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
Kernel & Hardware
