Re: [Patch v3 2/7] sched: Add infrastructure to store and update instantaneous thermal pressure

[Vincent Guittot]

Hi Thara,

On Wed, 16 Oct 2019 at 23:22, Thara Gopinath <thara.gopinath@xxxxxxxxxx> wrote:
>
> Hi Vincent,
>
> Thanks for the review
> On 10/14/2019 11:50 AM, Vincent Guittot wrote:
> > Hi Thara,
> >
> > On Mon, 14 Oct 2019 at 02:58, Thara Gopinath <thara.gopinath@xxxxxxxxxx> wrote:
> >>
> >> Add thermal.c and thermal.h files that provides interface
> >> APIs to initialize, update/average, track, accumulate and decay
> >> thermal pressure per cpu basis. A per cpu structure max_capacity_info is
> >> introduced to keep track of instantaneous per cpu thermal pressure.
> >> Thermal pressure is the delta between max_capacity and cap_capacity.
> >> API update_periodic_maxcap is called for periodic accumulate and decay
> >> of the thermal pressure. It is to to be called from a periodic tick
> >> function. This API calculates the delta between max_capacity and
> >> cap_capacity and passes on the delta to update_thermal_avg to do the
> >> necessary accumulate, decay and average. API update_maxcap_capacity is for
> >> the system to update the thermal pressure by updating cap_capacity.
> >> Considering, update_periodic_maxcap reads cap_capacity and
> >> update_maxcap_capacity writes into cap_capacity, one can argue for
> >> some sort of locking mechanism to avoid a stale value.
> >> But considering update_periodic_maxcap can be called from a system
> >> critical path like scheduler tick function, a locking mechanism is not
> >> ideal. This means that it is possible the value used to
> >> calculate average thermal pressure for a cpu can be stale for upto 1
> >> tick period.
> >>
> >> Signed-off-by: Thara Gopinath <thara.gopinath@xxxxxxxxxx>
> >> ---
> >>  include/linux/sched.h  | 14 +++++++++++
> >>  kernel/sched/Makefile  |  2 +-
> >>  kernel/sched/thermal.c | 66 ++++++++++++++++++++++++++++++++++++++++++++++++++
> >>  kernel/sched/thermal.h | 13 ++++++++++
> >>  4 files changed, 94 insertions(+), 1 deletion(-)
> >>  create mode 100644 kernel/sched/thermal.c
> >>  create mode 100644 kernel/sched/thermal.h
> >>
> >> diff --git a/include/linux/sched.h b/include/linux/sched.h
> >> index 2c2e56b..875ce2b 100644
> >> --- a/include/linux/sched.h
> >> +++ b/include/linux/sched.h
> >> @@ -1983,6 +1983,20 @@ static inline void rseq_syscall(struct pt_regs *regs)
> >>
> >>  #endif
> >>
> >> +#ifdef CONFIG_SMP
> >> +void update_maxcap_capacity(int cpu, u64 capacity);
> >> +
> >> +void populate_max_capacity_info(void);
> >> +#else
> >> +static inline void update_maxcap_capacity(int cpu, u64 capacity)
> >> +{
> >> +}
> >> +
> >> +static inline void populate_max_capacity_info(void)
> >> +{
> >> +}
> >> +#endif
> >> +
> >>  const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
> >>  char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
> >>  int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
> >> diff --git a/kernel/sched/Makefile b/kernel/sched/Makefile
> >> index 21fb5a5..4d3b820 100644
> >> --- a/kernel/sched/Makefile
> >> +++ b/kernel/sched/Makefile
> >> @@ -20,7 +20,7 @@ obj-y += core.o loadavg.o clock.o cputime.o
> >>  obj-y += idle.o fair.o rt.o deadline.o
> >>  obj-y += wait.o wait_bit.o swait.o completion.o
> >>
> >> -obj-$(CONFIG_SMP) += cpupri.o cpudeadline.o topology.o stop_task.o pelt.o
> >> +obj-$(CONFIG_SMP) += cpupri.o cpudeadline.o topology.o stop_task.o pelt.o thermal.o
> >>  obj-$(CONFIG_SCHED_AUTOGROUP) += autogroup.o
> >>  obj-$(CONFIG_SCHEDSTATS) += stats.o
> >>  obj-$(CONFIG_SCHED_DEBUG) += debug.o
> >> diff --git a/kernel/sched/thermal.c b/kernel/sched/thermal.c
> >> new file mode 100644
> >> index 0000000..5f0b2d4
> >> --- /dev/null
> >> +++ b/kernel/sched/thermal.c
> >> @@ -0,0 +1,66 @@
> >> +// SPDX-License-Identifier: GPL-2.0
> >> +/*
> >> + * Sceduler Thermal Interactions
> >> + *
> >> + *  Copyright (C) 2018 Linaro, Inc., Thara Gopinath <thara.gopinath@xxxxxxxxxx>
> >> + */
> >> +
> >> +#include <linux/sched.h>
> >> +#include "sched.h"
> >> +#include "pelt.h"
> >> +#include "thermal.h"
> >> +
> >> +struct max_capacity_info {
> >> +       unsigned long max_capacity;
> >> +       unsigned long cap_capacity;
> >> +};
> >> +
> >> +static DEFINE_PER_CPU(struct max_capacity_info, max_cap);
> >> +
> >> +void update_maxcap_capacity(int cpu, u64 capacity)
> >> +{
> >> +       struct max_capacity_info *__max_cap;
> >> +       unsigned long __capacity;
> >> +
> >> +       __max_cap = (&per_cpu(max_cap, cpu));
> >> +       if (!__max_cap) {
> >> +               pr_err("no max_capacity_info structure for cpu %d\n", cpu);
> >> +               return;
> >> +       }
> >> +
> >> +       /* Normalize the capacity */
> >> +       __capacity = (capacity * arch_scale_cpu_capacity(cpu)) >>
> >> +                                                       SCHED_CAPACITY_SHIFT;
> >> +       pr_debug("updating cpu%d capped capacity from %lu to %lu\n", cpu, __max_cap->cap_capacity, __capacity);
> >> +
> >> +       __max_cap->cap_capacity = __capacity;
> >> +}
> >> +
> >> +void populate_max_capacity_info(void)
> >> +{
> >> +       struct max_capacity_info *__max_cap;
> >> +       u64 capacity;
> >> +       int cpu;
> >> +
> >> +       for_each_possible_cpu(cpu) {
> >> +               __max_cap = (&per_cpu(max_cap, cpu));
> >> +               if (!__max_cap)
> >> +                       continue;
> >> +               capacity = arch_scale_cpu_capacity(cpu);
> >> +               __max_cap->max_capacity = capacity;
> >> +               __max_cap->cap_capacity = capacity;
> >> +               pr_debug("cpu %d max capacity set to %ld\n", cpu, __max_cap->max_capacity);
> >> +       }
> >> +}
> >
> > everything above seems to be there for the cpu cooling device and
> > should be included in it instead. The scheduler only need the capacity
> > capping
> > The cpu cooling device should just set the delta capacity in the
> > per-cpu variable (see my comment below)
> It can be a firmware  updating the thermal pressure instead of cpu
> cooling device. Or may be some other entity .So instead of replicating
> this code, isnt't it better to reside in one place ?

But you have now idea which kind of metrics will be provided by
firmware. It might not be a capped frequency and a max frequency but
other metrics.
Also, __max_cap->max_capacity just save in a new per cpu struct the
return of arch_scale_cpu_capacity but this doesn't give us real
benefit




> >
> >> +
> >> +void update_periodic_maxcap(struct rq *rq)
> >> +{
> >> +       struct max_capacity_info *__max_cap = (&per_cpu(max_cap, cpu_of(rq)));
> >> +       unsigned long delta;
> >> +
> >> +       if (!__max_cap)
> >> +               return;
> >> +
> >> +       delta = __max_cap->max_capacity - __max_cap->cap_capacity;
> >
> > Why don't you just save the delta in the per_cpu variable instead of
> > the struct max_capacity_info ? You have to compute the delta every
> > tick whereas we can expect it to not change that much compared to the
> > number of tick.
>
> Again I think thermal pressure can be applied from other entities like
> firmware as well. But  I agree with your point that calculating delta
> every tick is not a good idea. How about I move it to
> update_maxcap_capacity so that delta is calculate every time an update
> comes from cpu cooling device or anybody else ?
>
> Warm Regards
> Thara
>