Bug ID 569072: kernel: Fast TSC calibration failed

Last Modified: Nov 22, 2021

Bug Tracker

Affected Product:  See more info
BIG-IP All(all modules)

Known Affected Versions:
11.4.1

Opened: Jan 21, 2016
Severity: 3-Major

Symptoms

Occasionally when booting bigip 800/1660/3600/3950 platforms, the normal attempt to calibrate the TSC's 'quickly' fails. By itself this does not mean the TSC's are not calibrated properly, but without seeing more dmesg output, it is hard to determine if by the time user programs run whether or not the TSC's across the cpus are actually out of sync.

Impact

Generally this messages does indicate that the TSC for each cpu is not in sync with the other cpus, but only if slower methods of syncronization also fail.

Conditions

No particular fingerprints on this - it just happens intermittently. It is a very old linux problem on older Intel cpus. More modern cpus have 'fixed' or 'constant' TSC frequency features in the hardware so the TSC's are always in sync. The way to know if this actually failed is to look at the following dmesg output: # dmesg | grep clocksource Switching to clocksource hpet Refined TSC clocksource calibration: 2294.254 MHz. Switching to clocksource tsc The 2nd and 3rd lines mean that the TSC has properly syncronized. If all you see is the hpet being used as the clock source, then the tsc's could not be used because they did not properly syncronize.

Workaround

Power-cycling the machine (if actually needed) will typically resolve this issue.

Fix Information

None

Behavior Change