Bug 197203

Summary: Migration process shows huge CPU time
Product: Timers Reporter: Chris Mindas (chris.mindas)
Component: OtherAssignee: john stultz (john.stultz)
Status: NEW ---    
Severity: normal    
Priority: P1    
Hardware: Intel   
OS: Linux   
Kernel Version: 4.4.0-97-generic #120-Ubuntu SMP Tue Sep 19 17:28:18 UTC 2017 x86_64 x86_64 x86_64 GNU/Linux Subsystem:
Regression: No Bisected commit-id:

Description Chris Mindas 2017-10-11 23:27:42 UTC 
CPU: Intel i9-7900x 10 cores.
Distribution: Ubuntu 16.04.3

On boot, the processes migration/1 and migration/2 indicate that they have used a huge amount of CPU time (the times are meaningless, typically between 2-100 hours, so it seems unlikely that migration is actually using this amount of CPU time). The other migration processes are always zero.  This occurs at some point during the boot process before I can log into the system.

ps -ealdf | grep migration
1 S root          9      2  0 -40   - -     0 -      19:20 ?        00:00:00 [migration/0]
1 S root         12      2 99 -40   - -     0 -      19:20 ?        02:38:00 [migration/1]
1 S root         18      2 99 -40   - -     0 -      19:20 ?        02:38:00 [migration/2]
1 S root         23      2  0 -40   - -     0 -      19:20 ?        00:00:00 [migration/3]
1 S root         28      2  0 -40   - -     0 -      19:20 ?        00:00:00 [migration/4]

After logging in, the migration processes do not consume a significant amount of CPU time, no matter what is running.  I checked _SC_CLK_TCK after logging in thinking this might be screwed up:
sysconf(_SC_CLK_TCK) = 100

Interestingly enough, if I power off the system, migration/1 and migration/2 show reasonable results after the boot process is finished:
1 S root          9      2  0 -40   - -     0 -      16:42 ?        00:00:00 [migration/0]
1 S root         12      2  0 -40   - -     0 -      16:42 ?        00:00:01 [migration/1]
1 S root         18      2  0 -40   - -     0 -      16:42 ?        00:00:01 [migration/2]
1 S root         23      2  0 -40   - -     0 -      16:42 ?        00:00:00 [migration/3]
1 S root         28      2  0 -40   - -     0 -      16:42 ?        00:00:00 [migration/4]
(truncated...)

Warm booting the system results in the original problem.

Note that this doesn't seem to cause any problems with the system (except messing up some of the CPU usage stats in webmin).