drivers/cpuidle/governors/haltpoll.c | 15 ++++++++----- include/trace/events/power.h | 33 ++++++++++++++++++++++++++++ 2 files changed, 43 insertions(+), 5 deletions(-)
Add trace points as are implemented in KVM host halt polling.
This helps tune guest halt polling params.
Signed-off-by: Eiichi Tsukata <eiichi.tsukata@nutanix.com>
---
drivers/cpuidle/governors/haltpoll.c | 15 ++++++++-----
include/trace/events/power.h | 33 ++++++++++++++++++++++++++++
2 files changed, 43 insertions(+), 5 deletions(-)
diff --git a/drivers/cpuidle/governors/haltpoll.c b/drivers/cpuidle/governors/haltpoll.c
index cb2a96eafc02..a5b6ad32956c 100644
--- a/drivers/cpuidle/governors/haltpoll.c
+++ b/drivers/cpuidle/governors/haltpoll.c
@@ -19,6 +19,7 @@
#include <linux/sched.h>
#include <linux/module.h>
#include <linux/kvm_para.h>
+#include <trace/events/power.h>
static unsigned int guest_halt_poll_ns __read_mostly = 200000;
module_param(guest_halt_poll_ns, uint, 0644);
@@ -77,13 +78,16 @@ static int haltpoll_select(struct cpuidle_driver *drv,
static void adjust_poll_limit(struct cpuidle_device *dev, u64 block_ns)
{
- unsigned int val;
+ unsigned int val, old;
- /* Grow cpu_halt_poll_us if
- * cpu_halt_poll_us < block_ns < guest_halt_poll_us
+ val = dev->poll_limit_ns;
+ old = val;
+
+ /* Grow poll_limit_ns if
+ * poll_limit_ns < block_ns < guest_halt_poll_ns
*/
if (block_ns > dev->poll_limit_ns && block_ns <= guest_halt_poll_ns) {
- val = dev->poll_limit_ns * guest_halt_poll_grow;
+ val *= guest_halt_poll_grow;
if (val < guest_halt_poll_grow_start)
val = guest_halt_poll_grow_start;
@@ -91,16 +95,17 @@ static void adjust_poll_limit(struct cpuidle_device *dev, u64 block_ns)
val = guest_halt_poll_ns;
dev->poll_limit_ns = val;
+ trace_guest_halt_poll_ns_grow(smp_processor_id(), val, old);
} else if (block_ns > guest_halt_poll_ns &&
guest_halt_poll_allow_shrink) {
unsigned int shrink = guest_halt_poll_shrink;
- val = dev->poll_limit_ns;
if (shrink == 0)
val = 0;
else
val /= shrink;
dev->poll_limit_ns = val;
+ trace_guest_halt_poll_ns_shrink(smp_processor_id(), val, old);
}
}
diff --git a/include/trace/events/power.h b/include/trace/events/power.h
index af5018aa9517..db065af9c3c0 100644
--- a/include/trace/events/power.h
+++ b/include/trace/events/power.h
@@ -500,6 +500,39 @@ DEFINE_EVENT(dev_pm_qos_request, dev_pm_qos_remove_request,
TP_ARGS(name, type, new_value)
);
+
+TRACE_EVENT(guest_halt_poll_ns,
+
+ TP_PROTO(bool grow, unsigned int cpu_id,
+ unsigned int new, unsigned int old),
+
+ TP_ARGS(grow, cpu_id, new, old),
+
+ TP_STRUCT__entry(
+ __field(bool, grow)
+ __field(unsigned int, cpu_id)
+ __field(unsigned int, new)
+ __field(unsigned int, old)
+ ),
+
+ TP_fast_assign(
+ __entry->grow = grow;
+ __entry->cpu_id = cpu_id;
+ __entry->new = new;
+ __entry->old = old;
+ ),
+
+ TP_printk("cpu %u: halt_poll_ns %u (%s %u)",
+ __entry->cpu_id,
+ __entry->new,
+ __entry->grow ? "grow" : "shrink",
+ __entry->old)
+);
+
+#define trace_guest_halt_poll_ns_grow(cpu_id, new, old) \
+ trace_guest_halt_poll_ns(true, cpu_id, new, old)
+#define trace_guest_halt_poll_ns_shrink(cpu_id, new, old) \
+ trace_guest_halt_poll_ns(false, cpu_id, new, old)
#endif /* _TRACE_POWER_H */
/* This part must be outside protection */
--
2.36.1
On Mon, May 23, 2022 at 11:53:32PM +0000, Eiichi Tsukata wrote: > Add trace points as are implemented in KVM host halt polling. > This helps tune guest halt polling params. > > Signed-off-by: Eiichi Tsukata <eiichi.tsukata@nutanix.com> > --- > drivers/cpuidle/governors/haltpoll.c | 15 ++++++++----- > include/trace/events/power.h | 33 ++++++++++++++++++++++++++++ > 2 files changed, 43 insertions(+), 5 deletions(-) > > diff --git a/drivers/cpuidle/governors/haltpoll.c b/drivers/cpuidle/governors/haltpoll.c > index cb2a96eafc02..a5b6ad32956c 100644 > --- a/drivers/cpuidle/governors/haltpoll.c > +++ b/drivers/cpuidle/governors/haltpoll.c > @@ -19,6 +19,7 @@ > #include <linux/sched.h> > #include <linux/module.h> > #include <linux/kvm_para.h> > +#include <trace/events/power.h> > > static unsigned int guest_halt_poll_ns __read_mostly = 200000; > module_param(guest_halt_poll_ns, uint, 0644); > @@ -77,13 +78,16 @@ static int haltpoll_select(struct cpuidle_driver *drv, > > static void adjust_poll_limit(struct cpuidle_device *dev, u64 block_ns) > { > - unsigned int val; > + unsigned int val, old; > > - /* Grow cpu_halt_poll_us if > - * cpu_halt_poll_us < block_ns < guest_halt_poll_us > + val = dev->poll_limit_ns; > + old = val; > + > + /* Grow poll_limit_ns if > + * poll_limit_ns < block_ns < guest_halt_poll_ns > */ > if (block_ns > dev->poll_limit_ns && block_ns <= guest_halt_poll_ns) { > - val = dev->poll_limit_ns * guest_halt_poll_grow; > + val *= guest_halt_poll_grow; > > if (val < guest_halt_poll_grow_start) > val = guest_halt_poll_grow_start; > @@ -91,16 +95,17 @@ static void adjust_poll_limit(struct cpuidle_device *dev, u64 block_ns) > val = guest_halt_poll_ns; > Can do it before the assignment: trace_guest_halt_poll_ns_grow(val, dev->poll_limit_ns); > dev->poll_limit_ns = val; > + trace_guest_halt_poll_ns_grow(smp_processor_id(), val, old); > } else if (block_ns > guest_halt_poll_ns && > guest_halt_poll_allow_shrink) { > unsigned int shrink = guest_halt_poll_shrink; > > - val = dev->poll_limit_ns; > if (shrink == 0) > val = 0; > else > val /= shrink; > dev->poll_limit_ns = val; > + trace_guest_halt_poll_ns_shrink(smp_processor_id(), val, old); > } > } > > diff --git a/include/trace/events/power.h b/include/trace/events/power.h > index af5018aa9517..db065af9c3c0 100644 > --- a/include/trace/events/power.h > +++ b/include/trace/events/power.h > @@ -500,6 +500,39 @@ DEFINE_EVENT(dev_pm_qos_request, dev_pm_qos_remove_request, > > TP_ARGS(name, type, new_value) > ); > + > +TRACE_EVENT(guest_halt_poll_ns, > + > + TP_PROTO(bool grow, unsigned int cpu_id, > + unsigned int new, unsigned int old), > + > + TP_ARGS(grow, cpu_id, new, old), > + > + TP_STRUCT__entry( > + __field(bool, grow) > + __field(unsigned int, cpu_id) > + __field(unsigned int, new) > + __field(unsigned int, old) > + ), > + > + TP_fast_assign( > + __entry->grow = grow; > + __entry->cpu_id = cpu_id; > + __entry->new = new; > + __entry->old = old; > + ), > + > + TP_printk("cpu %u: halt_poll_ns %u (%s %u)", > + __entry->cpu_id, > + __entry->new, > + __entry->grow ? "grow" : "shrink", > + __entry->old) > +); > + > +#define trace_guest_halt_poll_ns_grow(cpu_id, new, old) \ > + trace_guest_halt_poll_ns(true, cpu_id, new, old) > +#define trace_guest_halt_poll_ns_shrink(cpu_id, new, old) \ > + trace_guest_halt_poll_ns(false, cpu_id, new, old) > #endif /* _TRACE_POWER_H */ > > /* This part must be outside protection */ > -- > 2.36.1 > >
Hi Marcelo > On May 26, 2022, at 2:31, Marcelo Tosatti <mtosatti@redhat.com> wrote: > > On Mon, May 23, 2022 at 11:53:32PM +0000, Eiichi Tsukata wrote: >> Add trace points as are implemented in KVM host halt polling. >> This helps tune guest halt polling params. >> >> Signed-off-by: Eiichi Tsukata <eiichi.tsukata@nutanix.com> >> --- >> drivers/cpuidle/governors/haltpoll.c | 15 ++++++++----- >> include/trace/events/power.h | 33 ++++++++++++++++++++++++++++ >> 2 files changed, 43 insertions(+), 5 deletions(-) >> >> diff --git a/drivers/cpuidle/governors/haltpoll.c b/drivers/cpuidle/governors/haltpoll.c >> index cb2a96eafc02..a5b6ad32956c 100644 >> --- a/drivers/cpuidle/governors/haltpoll.c >> +++ b/drivers/cpuidle/governors/haltpoll.c >> @@ -19,6 +19,7 @@ >> #include <linux/sched.h> >> #include <linux/module.h> >> #include <linux/kvm_para.h> >> +#include <trace/events/power.h> >> >> static unsigned int guest_halt_poll_ns __read_mostly = 200000; >> module_param(guest_halt_poll_ns, uint, 0644); >> @@ -77,13 +78,16 @@ static int haltpoll_select(struct cpuidle_driver *drv, >> >> static void adjust_poll_limit(struct cpuidle_device *dev, u64 block_ns) >> { >> - unsigned int val; >> + unsigned int val, old; >> >> - /* Grow cpu_halt_poll_us if >> - * cpu_halt_poll_us < block_ns < guest_halt_poll_us >> + val = dev->poll_limit_ns; >> + old = val; >> + >> + /* Grow poll_limit_ns if >> + * poll_limit_ns < block_ns < guest_halt_poll_ns >> */ >> if (block_ns > dev->poll_limit_ns && block_ns <= guest_halt_poll_ns) { >> - val = dev->poll_limit_ns * guest_halt_poll_grow; >> + val *= guest_halt_poll_grow; >> >> if (val < guest_halt_poll_grow_start) >> val = guest_halt_poll_grow_start; >> @@ -91,16 +95,17 @@ static void adjust_poll_limit(struct cpuidle_device *dev, u64 block_ns) >> val = guest_halt_poll_ns; >> > > Can do it before the assignment: > > trace_guest_halt_poll_ns_grow(val, dev->poll_limit_ns); Sure. Will fix it in v4. Thanks Eiichi
On Mon, 23 May 2022 23:53:32 +0000 Eiichi Tsukata <eiichi.tsukata@nutanix.com> wrote: > @@ -91,16 +95,17 @@ static void adjust_poll_limit(struct cpuidle_device *dev, u64 block_ns) > val = guest_halt_poll_ns; > > dev->poll_limit_ns = val; > + trace_guest_halt_poll_ns_grow(smp_processor_id(), val, old); Why are you passing in smp_processor_id()? > } else if (block_ns > guest_halt_poll_ns && > guest_halt_poll_allow_shrink) { > unsigned int shrink = guest_halt_poll_shrink; > > - val = dev->poll_limit_ns; > if (shrink == 0) > val = 0; > else > val /= shrink; > dev->poll_limit_ns = val; > + trace_guest_halt_poll_ns_shrink(smp_processor_id(), val, old); > } > } > > diff --git a/include/trace/events/power.h b/include/trace/events/power.h > index af5018aa9517..db065af9c3c0 100644 > --- a/include/trace/events/power.h > +++ b/include/trace/events/power.h > @@ -500,6 +500,39 @@ DEFINE_EVENT(dev_pm_qos_request, dev_pm_qos_remove_request, > > TP_ARGS(name, type, new_value) > ); > + > +TRACE_EVENT(guest_halt_poll_ns, > + > + TP_PROTO(bool grow, unsigned int cpu_id, > + unsigned int new, unsigned int old), > + > + TP_ARGS(grow, cpu_id, new, old), > + > + TP_STRUCT__entry( > + __field(bool, grow) > + __field(unsigned int, cpu_id) > + __field(unsigned int, new) > + __field(unsigned int, old) > + ), > + > + TP_fast_assign( > + __entry->grow = grow; > + __entry->cpu_id = cpu_id; You are wasting space to save the cpu_id, as the trace event already knows what CPU it occurred on. # echo 1 > events/sched/enable # cat trace # TASK-PID CPU# ||||| TIMESTAMP FUNCTION # | | | ||||| | | systemd-1 [004] ..... 15.872715: ftrace_boot_snapshot: ** Boot snapshot taken ** systemd-1 [001] ..... 22.555418: initcall_start: func=fuse_len_args+0x0/0x30 [fuse] systemd-1 [001] ..... 22.555425: initcall_finish: func=fuse_len_args+0x0/0x30 [fuse] ret=0 modprobe-643 [006] ..... 26.737355: initcall_start: func=wmidev_evaluate_method+0x46/0x100 [wmi] modprobe-643 [006] ..... 26.742491: initcall_finish: func=wmidev_evaluate_method+0x46/0x100 [wmi] ret=0 -- Steve > + __entry->new = new; > + __entry->old = old; > + ), > + > + TP_printk("cpu %u: halt_poll_ns %u (%s %u)", > + __entry->cpu_id, > + __entry->new, > + __entry->grow ? "grow" : "shrink", > + __entry->old) > +); > +
Hi Steven > On May 26, 2022, at 1:02, Steven Rostedt <rostedt@goodmis.org> wrote: > > On Mon, 23 May 2022 23:53:32 +0000 > Eiichi Tsukata <eiichi.tsukata@nutanix.com> wrote: > >> @@ -91,16 +95,17 @@ static void adjust_poll_limit(struct cpuidle_device *dev, u64 block_ns) >> val = guest_halt_poll_ns; >> >> dev->poll_limit_ns = val; >> + trace_guest_halt_poll_ns_grow(smp_processor_id(), val, old); > > Why are you passing in smp_processor_id()? > >> } else if (block_ns > guest_halt_poll_ns && >> guest_halt_poll_allow_shrink) { >> unsigned int shrink = guest_halt_poll_shrink; >> >> - val = dev->poll_limit_ns; >> if (shrink == 0) >> val = 0; >> else >> val /= shrink; >> dev->poll_limit_ns = val; >> + trace_guest_halt_poll_ns_shrink(smp_processor_id(), val, old); >> } >> } >> >> diff --git a/include/trace/events/power.h b/include/trace/events/power.h >> index af5018aa9517..db065af9c3c0 100644 >> --- a/include/trace/events/power.h >> +++ b/include/trace/events/power.h >> @@ -500,6 +500,39 @@ DEFINE_EVENT(dev_pm_qos_request, dev_pm_qos_remove_request, >> >> TP_ARGS(name, type, new_value) >> ); >> + >> +TRACE_EVENT(guest_halt_poll_ns, >> + >> + TP_PROTO(bool grow, unsigned int cpu_id, >> + unsigned int new, unsigned int old), >> + >> + TP_ARGS(grow, cpu_id, new, old), >> + >> + TP_STRUCT__entry( >> + __field(bool, grow) >> + __field(unsigned int, cpu_id) >> + __field(unsigned int, new) >> + __field(unsigned int, old) >> + ), >> + >> + TP_fast_assign( >> + __entry->grow = grow; >> + __entry->cpu_id = cpu_id; > > You are wasting space to save the cpu_id, as the trace event already knows > what CPU it occurred on. > > # echo 1 > events/sched/enable > # cat trace > # TASK-PID CPU# ||||| TIMESTAMP FUNCTION > # | | | ||||| | | > systemd-1 [004] ..... 15.872715: ftrace_boot_snapshot: ** Boot snapshot taken ** > systemd-1 [001] ..... 22.555418: initcall_start: func=fuse_len_args+0x0/0x30 [fuse] > systemd-1 [001] ..... 22.555425: initcall_finish: func=fuse_len_args+0x0/0x30 [fuse] ret=0 > modprobe-643 [006] ..... 26.737355: initcall_start: func=wmidev_evaluate_method+0x46/0x100 [wmi] > modprobe-643 [006] ..... 26.742491: initcall_finish: func=wmidev_evaluate_method+0x46/0x100 [wmi] ret=0 > > -- Steve Thanks for your suggestion. I added cpu_id as there is a similar precedent “trace_cpu_idle” but I think we can remove cpu_id. Will fix it in v3. Eiichi
© 2016 - 2024 Red Hat, Inc.