On Thu, 4 Dec 2025 18:24:57 +0100
"Rafael J. Wysocki" <[email protected]> wrote:
> > I'm not exactly sure what you mean here. There is an "onchange" trigger you
> > can use to trigger a synthetic event whenever a change happens. But I think
> > the data here wants to know which CPU had its policy change. Hence the CPU
> > mask.
>
> IIUC he wants to trace frequency changes per policy, not per CPU
> (because there are cases in which multiple CPUs belong to one policy
> and arguably the frequency doesn't need to be traced for all of them),
> but tooling should know which CPUs belong to the same policy, so it
> should be straightforward to use that knowledge when processing the
> traces.
In case you only care about frequency changes, you could do this:
# echo 'freq_change u32 state;' > /sys/kernel/tracing/synthetic_events
# echo 'hist:keys=common_type:s=state:onchange($s).trace(freq_change,$s)' >
/sys/kernel/tracing/events/power/cpu_frequency/trigger
# echo 1 > /sys/kernel/tracing/events/synthetic/freq_change/enable
# cat /sys/kernel/tracing/trace
# tracer: nop
#
# entries-in-buffer/entries-written: 2833/2833 #P:56
#
# _-----=> irqs-off/BH-disabled
# / _----=> need-resched
# | / _---=> hardirq/softirq
# || / _--=> preempt-depth
# ||| / _-=> migrate-disable
# |||| / delay
# TASK-PID CPU# ||||| TIMESTAMP FUNCTION
# | | | ||||| | |
sed-596089 [034] d..5. 2687140.288806: freq_change: state=2000000
bash-596090 [020] d.s4. 2687140.290407: freq_change: state=1900000
<idle>-0 [028] d.s7. 2687140.290425: freq_change: state=3000000
bash-596090 [020] d..5. 2687140.291152: freq_change: state=1900000
<idle>-0 [000] dNs5. 2687140.326526: freq_change: state=1200000
CPU 3/KVM-10724 [019] d.s5. 2687140.358418: freq_change: state=2100000
CPU 6/KVM-10727 [021] d.h5. 2687140.394403: freq_change: state=1300000
CPU 6/KVM-10727 [021] d.h5. 2687140.398403: freq_change: state=1400000
CPU 6/KVM-10727 [021] d.h5. 2687140.402402: freq_change: state=1500000
CPU 6/KVM-10727 [021] d.h5. 2687140.406400: freq_change: state=1600000
CPU 6/KVM-10727 [021] d.h5. 2687140.410404: freq_change: state=1700000
[..]
Which BTW, I'll be giving a talk about synthetic events at OSS in Tokyo ;-)
(cheap plug!)
https://ossjapan2025.sched.com/event/29FoB/synthetic-events-and-tracing-latency-within-the-kernel-steven-rostedt-google?iframe=yes&w=100%&sidebar=yes&bg=no
-- Steve
