[OpenWrt-Devel] NOHZ: local_softirq_pending 08 -- IPQ4019 on master

[Jeff Kletsky]

Just flashed a build off `master` and am seeing "new" error messages
starting after the network has started, a couple times,
then every 20 seconds, seemingly like clockwork.

root at test:/# dmesg | fgrep NOHZ
[   36.955401] NOHZ: local_softirq_pending 08
[   57.439420] NOHZ: local_softirq_pending 08
[   98.397107] NOHZ: local_softirq_pending 08
[  118.881226] NOHZ: local_softirq_pending 08
[  139.356879] NOHZ: local_softirq_pending 08
[  159.840218] NOHZ: local_softirq_pending 08
[  180.315962] NOHZ: local_softirq_pending 08
[  200.799979] NOHZ: local_softirq_pending 08
[  221.275747] NOHZ: local_softirq_pending 08
[  241.756644] NOHZ: local_softirq_pending 08

Still looking into it, but curious if anyone else is seeing this.

Linux 4.19.53

commit 629e6538a1 (openwrt/master, openwrt/HEAD, master)
Author: Christian Lamparter <redacted>
Date:   Sat Jun 22 12:23:10 2019 +0200


Similar behavior on the alternate firmware on the device

commit eaad2211db
Author: Jo-Philipp Wich <redacted>
Date:   Fri Jun 21 17:41:28 2019 +0200

root at test:/# dmesg | fgrep NOHZ
[   56.800024] NOHZ: local_softirq_pending 08
[   77.281594] NOHZ: local_softirq_pending 08
[   97.763395] NOHZ: local_softirq_pending 08
[  138.718065] NOHZ: local_softirq_pending 08
[  138.720722] NOHZ: local_softirq_pending 08
[  159.197431] NOHZ: local_softirq_pending 08
[  200.157115] NOHZ: local_softirq_pending 08
[  220.636004] NOHZ: local_softirq_pending 08
[  241.116778] NOHZ: local_softirq_pending 08
[  241.119413] NOHZ: local_softirq_pending 08


Some mention of `NOHZ: local_softirq_pending 08` back in 2012
https://groups.google.com/forum/#!topic/linux.kernel/YeH1CTNT9vU

Other mention at
https://askubuntu.com/questions/878030/occasional-dmesg-log-nohz-local-softirq-pending-08

> Add nohz=off to the kernel parameters during boot to disable it.
> 
>> This option causes RCU to attempt to accelerate grace periods
>> in order to allow CPUs to enter dynticks-idle state more quickly.
>> On the other hand, this option increases the overhead
>> of the dynticks-idle checking, particularly on systems
>> with large numbers of CPUs.

I'm not sure I'd consider an IPQ4019 to have "large numbers of CPUs"
but I figured I'd lay out what I have found so far.


Jeff

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.infradead.org/pipermail/openwrt-devel/attachments/20190622/f447a983/attachment.htm>
-------------- next part --------------
_______________________________________________
openwrt-devel mailing list
openwrt-devel at lists.openwrt.org
https://lists.openwrt.org/mailman/listinfo/openwrt-devel