linux: Messages about clocksource and timekeeping watchdog

[pmenzel]

@thomas noticed new Linux informational messages on phaedra with Linux 5.10.103:

[    0.000000] Linux version 5.10.103.mx64.429 (root@theinternet.molgen.mpg.de) (gcc (GCC) 7.5.0, GNU ld (GNU Binutils) 2.37) #1 SMP Mon Mar 7 16:41:58 CET 2022
[    0.000000] Command line: BOOT_IMAGE=/boot/bzImage-5.10.103.mx64.429 root=LABEL=root ro crashkernel=64G-:256M console=ttyS0,115200n8 console=tty0 init=/bin/systemd audit=0 random.trust_cpu=on systemd.unified_cgroup_hierarchy
[…]
[ 2324.018824] clocksource: timekeeping watchdog on CPU39: hpet wd-wd read-back delay of 96031ns
[ 2324.027704] clocksource: wd-tsc-wd read-back delay of 132977ns, clock-skew test skipped!
[ 2832.515481] clocksource: timekeeping watchdog on CPU32: hpet wd-wd read-back delay of 57200ns
[ 2832.524390] clocksource: wd-tsc-wd read-back delay of 109301ns, clock-skew test skipped!
[ 2994.515649] clocksource: timekeeping watchdog on CPU100: hpet wd-wd read-back delay of 60063ns
[ 2994.524619] clocksource: wd-tsc-wd read-back delay of 108882ns, clock-skew test skipped!
[…]

I did $ pssh -H "$(hostconfig --list mx64)" -p 100 -i grep \"timekeeping\ watchdog\ on\" /var/log/messages\* and got the hits below:

• amaru
• bohemianrhapsody
• dose
• elcattivo
• holidayincambodia
• koprastasophobie
• nieselregen
• phaedra
• simsalabimbambasaladusaladim
• steel
• sympathyforthedevil
• talyn

[donald]

Looks like some of these are rather old (e.g. "amaru" from 2021-07-13) so we had this before.

Only a few systems had this during the current boot (dmesg) and - aside phaedra - only a few times:

elcattivo: 4
koprastasophobie:1
phaedra: 215
talyn: 1
dose: 17
bohemianrhapsody: 5

There is a revert for an decreased tolerance: https://lore.kernel.org/lkml/20211122030223.GG34844@shbuild999.sh.intel.com/ , maybe this is related.

[thomas]

Hmm,

Can you recheck your results?

Here I get one old hit

[root@nieselregen:~]
#grep -i 'timekeeping watchdog ' /var/log/messages*
/var/log/messages.2021-10:2021-10-28T18:45:12.669944+02:00 nieselregen kernel: [12001177.137417] clocksource: timekeeping watchdog on CPU0: Marking clocksource 'tsc' as unstable because the skew is too large:
[root@nieselregen:~]

[donald]

This message is from October 2021. I was referring to the current boot.

root:nieselregen:~/# dmesg|grep 'timekeeping watchdog '
root:nieselregen:~/# 

[thomas]

And I was referring to Pauls post not yours :)

[donald]

Got it :-)