Bug 438900
| Summary: | PC card daemon overhead too high | ||
|---|---|---|---|
| Product: | Red Hat Enterprise MRG | Reporter: | IBM Bug Proxy <bugproxy> |
| Component: | realtime-kernel | Assignee: | Red Hat Real Time Maintenance <rt-maint> |
| Status: | CLOSED NOTABUG | QA Contact: | |
| Severity: | medium | Docs Contact: | |
| Priority: | low | ||
| Version: | beta | CC: | bhu |
| Target Milestone: | --- | ||
| Target Release: | --- | ||
| Hardware: | x86_64 | ||
| OS: | All | ||
| Whiteboard: | |||
| Fixed In Version: | Doc Type: | Bug Fix | |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | 2008-05-27 22:07:20 UTC | Type: | --- |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
| Bug Depends On: | |||
| Bug Blocks: | 448636 | ||
------- Comment From jstultz.com 2008-04-02 11:18 EDT------- Disabling via chkconfig is likely the solution here. Needs to be documented, then I think this is ok to mark resolved. closing ------- Comment From sripathi.com 2008-05-28 05:47 EDT------- Clark, you have closed this bug as NOTABUG. Did you decide so because you concluded that the PC card daemon does not affect latencies? |
=Comment: #0================================================= Alan P. Stevens <alan_stevens.com> - 2008-03-13 07:17 EDT Problem description: The pcscd daemon is using more CPU than any other daemon ( such as softirq-net- rx ), and although it is a low priority task, any additional background noise could lead to higher preemption costs to RT task, cache pollution and other undersirable impacts on determinism. It appeared with RHEL 5.1 and so in SR3 kernels onwards. System tap shows the activity (from strace) as file opens every second or so - pcscd(3928) open ("/dev/bus/usb", O_RDONLY|O_DIRECTORY|O_NONBLOCK) pcscd(3928) open ("/dev/bus/usb/001", O_RDONLY|O_DIRECTORY|O_NONBLOCK) pcscd(3928) open ("/dev/bus/usb/001/001", O_RDWR) pcscd(3928) open ("/dev/bus/usb/001/001", O_RDWR) And 'top' output sorted by cumulative CPU time confirms pcscd is the worst offender... Cpu(s): 0.0%us, 0.0%sy, 0.0%ni,100.0%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st Mem: 4010632k total, 3002848k used, 1007784k free, 224912k buffers Swap: 1959920k total, 88k used, 1959832k free, 2469824k cached PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ TIME COMMAND 3928 root 25 0 18948 1392 592 S 0 0.0 0:40.36 0:40 pcscd 46 root -91 0 0 0 0 S 0 0.0 0:23.34 0:23 softirq-net- rx/ 10035 root 11 -5 0 0 0 S 0 0.0 0:11.53 0:11 kjournald 7 root -91 0 0 0 0 S 0 0.0 0:10.73 0:10 softirq-net- rx/ 20 root -91 0 0 0 0 S 0 0.0 0:10.49 0:10 softirq-net- rx/ 2060 root 10 -5 0 0 0 S 0 0.0 0:10.48 0:10 IRQ-24 33 root -91 0 0 0 0 S 0 0.0 0:07.84 0:07 softirq-net- rx/ 47 root -51 0 0 0 0 S 0 0.0 0:05.55 0:05 softirq- block/3 246 root 10 -5 0 0 0 S 0 0.0 0:04.79 0:04 kswapd0 2251 root 15 0 6708 904 388 S 0 0.0 0:03.97 0:03 dhclient 3955 root 25 0 95468 1200 904 S 0 0.0 0:02.44 0:02 automount 635 root 11 -5 0 0 0 S 0 0.0 0:02.14 0:02 kjournald 476 root 10 -5 0 0 0 S 0 0.0 0:02.07 0:02 IRQ-26 44 root -51 0 0 0 0 S 0 0.0 0:02.00 0:02 softirq- timer/3 4228 avahi 15 0 23216 1472 1176 S 0 0.0 0:01.91 0:01 avahi-daemon 4239 haldaemo 15 0 30832 3948 1656 S 0 0.1 0:01.76 0:01 hald 21993 root 15 0 0 0 0 S 0 0.0 0:01.62 0:01 pdflush 1 root 18 0 10312 632 540 S 0 0.0 0:01.17 0:01 init 10 root -51 0 0 0 0 S 0 0.0 0:01.08 0:01 softirq- sched/0 18 root -51 0 0 0 0 S 0 0.0 0:00.83 0:00 softirq- timer/1 37 root -93 0 0 0 0 S 0 0.0 0:00.79 0:00 softirq- hrtimer 23 root -51 0 0 0 0 S 0 0.0 0:00.77 0:00 softirq- sched/1 5 root -51 0 0 0 0 S 0 0.0 0:00.71 0:00 softirq- timer/0 4217 root 15 0 244m 17m 3608 S 0 0.4 0:00.55 0:00 yum-updatesd 7506 stevens 15 0 85124 1876 1148 S 0 0.0 0:00.34 0:00 sshd 31 root -51 0 0 0 0 S 0 0.0 0:00.32 0:00 softirq- timer/2 686 root 15 -4 12616 828 392 S 0 0.0 0:00.31 0:00 udevd 36 root -51 0 0 0 0 S 0 0.0 0:00.29 0:00 softirq- sched/2 2312 root 15 0 5876 616 492 S 0 0.0 0:00.27 0:00 syslogd 7507 stevens 16 0 66044 1508 1184 S 0 0.0 0:00.27 0:00 bash 40 root 5 -10 0 0 0 S 0 0.0 0:00.24 0:00 desched/2 53 root 5 -10 0 0 0 S 0 0.0 0:00.24 0:00 desched/3 27 root 5 -10 0 0 0 S 0 0.0 0:00.17 0:00 desched/1 14 root 5 -10 0 0 0 S 0 0.0 0:00.16 0:00 desched/0 8958 stevens 15 0 12712 1088 804 R 0 0.0 0:00.12 0:00 top 41 root RT 0 0 0 0 S 0 0.0 0:00.11 0:00 migration/3 54 root 0 -20 0 0 0 S 0 0.0 0:00.11 0:00 events/0 Next step - I will see if running without pcscd makes any measureable difference to Calibrate / Recalibrate. =Comment: #2================================================= Sripathi Kodi <sripathi.com> - 2008-03-13 07:47 EDT pcscd is the smart card daemon. On my LS20 pcscd has not consumed much CPU yet. However, the system has only been running for around 3 hours. [root@llm50 ~]# ps -ef |grep pcscd root 2750 1 0 14:23 ? 00:00:02 pcscd strace shows that it keeps opening and closing /dev/bus/usb/001/001 all the time. We can turn off this service if we can confirm that we don't need it. =Comment: #3================================================= Sripathi Kodi <sripathi.com> - 2008-03-13 07:49 EDT Google says pcscd has given problems to others in the past: https://fcp.surfsite.org/modules/newbb/viewtopic.php?viewmode=flat&topic_id=46565&forum=12 http://www.nabble.com/pcscd,-sun-java,-100--CPU-usage-td14574113.html http://fcp.surfsite.org/modules/newbb/viewtopic.php?viewmode=flat&order=DESC&topic_id=34080&forum=12&move=prev&topic_time=1171910543 =Comment: #4================================================= John G. Stultz <jstultz.com> - 2008-03-24 17:08 EDT Pretty easy to turn it off via chkconfig. Probably worth bringing up w/ RH so its on their radar as well.