2023-06-13 04:23:23

by Ricardo Neri

[permalink] [raw]
Subject: [PATCH v4 02/24] sched: Add interfaces for IPC classes

Add the interfaces that architectures shall implement to convey the data
to support IPC classes.

arch_update_ipcc() updates the IPC classification of the current task as
given by hardware.

arch_get_ipcc_score() provides a performance score for a given IPC class
when placed on a specific CPU. Higher scores indicate higher performance.

When a driver or equivalent enablement code has configured the necessary
hardware to support IPC classes, it should call sched_enable_ipc_classes()
to notify the scheduler that it can start using IPC classes data.

The number of classes and the score of each class of task are determined
by hardware.

Cc: Ben Segall <[email protected]>
Cc: Daniel Bristot de Oliveira <[email protected]>
Cc: Dietmar Eggemann <[email protected]>
Cc: Ionela Voinescu <[email protected]>
Cc: Joel Fernandes (Google) <[email protected]>
Cc: Len Brown <[email protected]>
Cc: Lukasz Luba <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Perry Yuan <[email protected]>
Cc: Rafael J. Wysocki <[email protected]>
Cc: Srinivas Pandruvada <[email protected]>
Cc: Steven Rostedt <[email protected]>
Cc: Tim C. Chen <[email protected]>
Cc: Valentin Schneider <[email protected]>
Cc: Zhao Liu <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Signed-off-by: Ricardo Neri <[email protected]>
---
Changes since v3:
* None

Changes since v2:
* Clarified the properties of the IPC score: abstract, and linear. It can
normalized when needed. (Ionela)
* Selected a better default IPC score. (Ionela)
* Removed arch_has_ipc_classes(). It is not suitable for hardware that is
not ready to support IPC classes after boot. (Lukasz)
* Added a new sched_enable_ipc_classes() interface that drivers or
enablement code can call when ready to support IPC classes. (Lukasz)

Changes since v1:
* Shortened the names of the IPCC interfaces (PeterZ):
sched_task_classes_enabled >> sched_ipcc_enabled
arch_has_task_classes >> arch_has_ipc_classes
arch_update_task_class >> arch_update_ipcc
arch_get_task_class_score >> arch_get_ipcc_score
* Removed smt_siblings_idle argument from arch_update_ipcc(). (PeterZ)
---
include/linux/sched/topology.h | 6 ++++
kernel/sched/sched.h | 66 ++++++++++++++++++++++++++++++++++
kernel/sched/topology.c | 9 +++++
3 files changed, 81 insertions(+)

diff --git a/include/linux/sched/topology.h b/include/linux/sched/topology.h
index 816df6cc444e..5b084d3c9ad1 100644
--- a/include/linux/sched/topology.h
+++ b/include/linux/sched/topology.h
@@ -280,4 +280,10 @@ static inline int task_node(const struct task_struct *p)
return cpu_to_node(task_cpu(p));
}

+#ifdef CONFIG_IPC_CLASSES
+extern void sched_enable_ipc_classes(void);
+#else
+static inline void sched_enable_ipc_classes(void) { }
+#endif
+
#endif /* _LINUX_SCHED_TOPOLOGY_H */
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 556496c77dc2..03fb53e9f340 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -2544,6 +2544,72 @@ void arch_scale_freq_tick(void)
}
#endif

+#ifdef CONFIG_IPC_CLASSES
+DECLARE_STATIC_KEY_FALSE(sched_ipcc);
+
+static inline bool sched_ipcc_enabled(void)
+{
+ return static_branch_unlikely(&sched_ipcc);
+}
+
+#ifndef arch_update_ipcc
+/**
+ * arch_update_ipcc() - Update the IPC class of the current task
+ * @curr: The current task
+ *
+ * Request that the IPC classification of @curr is updated.
+ *
+ * Returns: none
+ */
+static __always_inline
+void arch_update_ipcc(struct task_struct *curr)
+{
+}
+#endif
+
+#ifndef arch_get_ipcc_score
+
+#define SCHED_IPCC_SCORE_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
+/**
+ * arch_get_ipcc_score() - Get the IPC score of a class of task
+ * @ipcc: The IPC class
+ * @cpu: A CPU number
+ *
+ * The IPC performance scores reflects (but it is not identical to) the number
+ * of instructions retired per cycle for a given IPC class. It is a linear and
+ * abstract metric. Higher scores reflect better performance.
+ *
+ * The IPC score can be normalized with respect to the class, i, with the
+ * highest IPC score on the CPU, c, with highest performance:
+ *
+ * IPC(i, c)
+ * ------------------------------------ * SCHED_IPCC_SCORE_SCALE
+ * max(IPC(i, c) : (i, c))
+ *
+ * Scheduling schemes that want to use the IPC score along with other
+ * normalized metrics for scheduling (e.g., CPU capacity) may need to normalize
+ * it.
+ *
+ * Other scheduling schemes (e.g., asym_packing) do not need normalization.
+ *
+ * Returns the performance score of an IPC class, @ipcc, when running on @cpu.
+ * Error when either @ipcc or @cpu are invalid.
+ */
+static __always_inline
+unsigned long arch_get_ipcc_score(unsigned short ipcc, int cpu)
+{
+ return SCHED_IPCC_SCORE_SCALE;
+}
+#endif
+#else /* CONFIG_IPC_CLASSES */
+
+#define arch_get_ipcc_score(ipcc, cpu) (-EINVAL)
+#define arch_update_ipcc(curr)
+
+static inline bool sched_ipcc_enabled(void) { return false; }
+
+#endif /* CONFIG_IPC_CLASSES */
+
#ifndef arch_scale_freq_capacity
/**
* arch_scale_freq_capacity - get the frequency scale factor of a given CPU.
diff --git a/kernel/sched/topology.c b/kernel/sched/topology.c
index ca4472281c28..c3a633a4b474 100644
--- a/kernel/sched/topology.c
+++ b/kernel/sched/topology.c
@@ -672,6 +672,15 @@ DEFINE_PER_CPU(struct sched_domain __rcu *, sd_asym_packing);
DEFINE_PER_CPU(struct sched_domain __rcu *, sd_asym_cpucapacity);
DEFINE_STATIC_KEY_FALSE(sched_asym_cpucapacity);

+#ifdef CONFIG_IPC_CLASSES
+DEFINE_STATIC_KEY_FALSE(sched_ipcc);
+
+void sched_enable_ipc_classes(void)
+{
+ static_branch_enable_cpuslocked(&sched_ipcc);
+}
+#endif
+
static void update_top_cache_domain(int cpu)
{
struct sched_domain_shared *sds = NULL;
--
2.25.1