[PATCH v2 0/2] timers/nohz: Optimize /proc/stat idle/iowait fluctuation

Xin Zhao posted 2 patches 3 days, 17 hours ago
fs/proc/stat.c           |  4 +++
include/linux/tick.h     |  4 +++
kernel/time/tick-sched.c | 58 ++++++++++++++++++++++++++++++++++++++--
3 files changed, 64 insertions(+), 2 deletions(-)
[PATCH v2 0/2] timers/nohz: Optimize /proc/stat idle/iowait fluctuation
Posted by Xin Zhao 3 days, 17 hours ago
The idle and iowait statistics in /proc/stat are obtained through
get_idle_time() and get_iowait_time(). Assuming CONFIG_NO_HZ_COMMON is
enabled, when CPU is online, the idle and iowait values use the
idle_sleeptime and iowait_sleeptime statistics from tick_cpu_sched, but
use CPUTIME_IDLE and CPUTIME_IOWAIT items from kernel_cpustat when CPU
is offline. Although /proc/stat do not print statistics of offline CPU,
it still print aggregated statistics of all possible CPUs.

ick_cpu_sched and kernel_cpustat are maintained by different logic,
leading to a significant gap. The first line of the data below shows the
/proc/stat output when only one CPU remains after CPU offline, the second
line shows the /proc/stat output after all CPUs are brought back online:

cpu  2408558 2 916619 4275883 5403 123758 64685 0 0 0
cpu  2408588 2 916693 4200737 4184 123762 64686 0 0 0

Obviously, other values do not experience significant fluctuations, while
idle/iowait statistics show a substantial decrease, which make system CPU
monitoring troublesome.

get_cpu_idle_time_us() calculates the latest cpu idle time based on
idle_entrytime and current time. When CPU is idle when offline, the value
return by get_cpu_idle_time_us() will continue to increase, which is
unexpected. get_cpu_iowait_time_us() has the similar calculation logic.
When CPU is in the iowait state when offline, the value return by
get_cpu_iowait_time_us() will continue to increase.

Introduce get_cpu_idle_time_us_offline() as the _offline variants of
get_cpu_idle_time_us(). get_cpu_idle_time_us_offline() just return the
same value of idle_sleeptime without any calculation. In this way,
/proc/stat logic can use it to get a correct CPU idle time, which remains
unchanged during CPU offline period. Also, the aggregated statistics of
all possible CPUs printed by /proc/stat will not experience significant
fluctuation when CPU hotplug.
So as the newly added get_cpu_iowait_time_us_offline().

In addition, revise a comment about broken iowait counter update race
related to the topic.

Xin Zhao (2):
  timers/nohz: Revise a comment about broken iowait counter update race
  timers/nohz: Avoid /proc/stat idle/iowait fluctuation when cpu hotplug

 fs/proc/stat.c           |  4 +++
 include/linux/tick.h     |  4 +++
 kernel/time/tick-sched.c | 58 ++++++++++++++++++++++++++++++++++++++--
 3 files changed, 64 insertions(+), 2 deletions(-)

-- 
2.34.1