libtest: Add scheduler test support

Add support to record scheduler operations.  This support is especially
important for tests in SMP configurations since the thread switch extension is
quite difficult to use due to the asynchronous nature of thread dispatching.
In contrast, the scheduler operations occur normally in a deterministic order.

Update #3716.

1 files changed, 658 insertions, 0 deletions

diff --git a/cpukit/libtest/t-test-scheduler.c b/cpukit/libtest/t-test-scheduler.c
new file mode 100644
index 0000000000..3f4a59c674
--- /dev/null
+++ b/cpukit/libtest/t-test-scheduler.c
@@ -0,0 +1,658 @@
+/* SPDX-License-Identifier: BSD-2-Clause */
+
+/**
+ * @file
+ *
+ * @ingroup RTEMSTestFrameworkImpl
+ *
+ * @brief This source file contains the implementation of the scheduler test
+ *   support API.
+ */
+
+/*
+ * Copyright (C) 2021 embedded brains GmbH (http://www.embedded-brains.de)
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifdef HAVE_CONFIG_H
+#include "config.h"
+#endif
+
+#include <rtems/test-scheduler.h>
+
+#include <rtems.h>
+#include <rtems/score/percpu.h>
+#include <rtems/score/schedulerimpl.h>
+
+typedef struct {
+	RTEMS_INTERRUPT_LOCK_MEMBER(lock)
+	T_scheduler_log *active;
+	T_scheduler_event_handler handler;
+	void *arg;
+} T_scheduler_context;
+
+static T_scheduler_context T_scheduler_instance = {
+#ifdef RTEMS_SMP
+	.lock = RTEMS_INTERRUPT_LOCK_INITIALIZER("Test Scheduler"),
+#endif
+	.active = NULL
+};
+
+void
+T_scheduler_set_event_handler(T_scheduler_event_handler handler, void *arg)
+{
+	T_scheduler_context *ctx;
+	rtems_interrupt_lock_context lock_context;
+
+	ctx = &T_scheduler_instance;
+
+	rtems_interrupt_lock_acquire(&ctx->lock, &lock_context);
+	ctx->handler = handler;
+	ctx->arg = arg;
+	rtems_interrupt_lock_release(&ctx->lock, &lock_context);
+}
+
+static void
+T_scheduler_before_operation(T_scheduler_event *event)
+{
+	T_scheduler_context *ctx;
+	rtems_interrupt_lock_context lock_context;
+	T_scheduler_log *log;
+	T_scheduler_event_handler handler;
+	void *arg;
+	Per_CPU_Control *cpu_self;
+
+	ctx = &T_scheduler_instance;
+	log = ctx->active;
+	handler = ctx->handler;
+
+	if (log == NULL && handler == NULL) {
+		return;
+	}
+
+	rtems_interrupt_lock_interrupt_disable(&lock_context);
+	cpu_self = _Per_CPU_Get();
+	event->cpu = _Per_CPU_Get_index( cpu_self );
+
+	if (_Per_CPU_Is_ISR_in_progress(cpu_self)) {
+		event->executing = NULL;
+	} else {
+		event->executing = _Per_CPU_Get_executing(cpu_self);
+	}
+
+	event->instant = T_now();
+
+	rtems_interrupt_lock_acquire_isr(&ctx->lock, &lock_context);
+	handler = ctx->handler;
+	arg = ctx->arg;
+	rtems_interrupt_lock_release(&ctx->lock, &lock_context);
+
+	if (handler != NULL) {
+		(*handler)(arg, event, T_SCHEDULER_BEFORE);
+	}
+}
+
+static void
+T_scheduler_record_event(T_scheduler_event *event)
+{
+	T_scheduler_context *ctx;
+	rtems_interrupt_lock_context lock_context;
+	T_scheduler_log *log;
+	T_scheduler_event_handler handler;
+
+	ctx = &T_scheduler_instance;
+	log = ctx->active;
+	handler = ctx->handler;
+
+	if (log == NULL && handler == NULL) {
+		return;
+	}
+
+	rtems_interrupt_lock_acquire(&ctx->lock, &lock_context);
+
+#ifdef RTEMS_SMP
+	handler = ctx->handler;
+#endif
+
+	if (handler != NULL) {
+		void *arg;
+
+		arg = ctx->arg;
+
+		rtems_interrupt_lock_release(&ctx->lock, &lock_context);
+		(*handler)(arg, event, T_SCHEDULER_AFTER);
+		rtems_interrupt_lock_acquire(&ctx->lock, &lock_context);
+	}
+
+#ifdef RTEMS_SMP
+	log = ctx->active;
+#endif
+
+	if (log != NULL) {
+		size_t recorded;
+
+		++log->header.operations;
+		recorded = log->header.recorded;
+
+		if (recorded < log->header.capacity) {
+			log->header.recorded = recorded + 1;
+			log->events[recorded] = *event;
+		}
+	}
+
+	rtems_interrupt_lock_release(&ctx->lock, &lock_context);
+}
+
+T_scheduler_log *
+T_scheduler_record(T_scheduler_log *log)
+{
+	rtems_interrupt_lock_context lock_context;
+	T_scheduler_log *previous;
+	T_scheduler_context *ctx;
+
+	if (log != NULL) {
+		log->header.recorded = 0;
+		log->header.operations = 0;
+	}
+
+	ctx = &T_scheduler_instance;
+
+	rtems_interrupt_lock_acquire(&ctx->lock, &lock_context);
+	previous = ctx->active;
+	ctx->active = log;
+	rtems_interrupt_lock_release(&ctx->lock, &lock_context);
+
+	return previous;
+}
+
+T_scheduler_log *
+T_scheduler_record_2(T_scheduler_log_2 *log)
+{
+	log->header.capacity = T_ARRAY_SIZE(log->events);
+	return T_scheduler_record((T_scheduler_log *)log);
+}
+
+T_scheduler_log *
+T_scheduler_record_4(T_scheduler_log_4 *log)
+{
+	log->header.capacity = T_ARRAY_SIZE(log->events);
+	return T_scheduler_record((T_scheduler_log *)log);
+}
+
+T_scheduler_log *
+T_scheduler_record_10(T_scheduler_log_10 *log)
+{
+	log->header.capacity = T_ARRAY_SIZE(log->events);
+	return T_scheduler_record((T_scheduler_log *)log);
+}
+
+T_scheduler_log *
+T_scheduler_record_20(T_scheduler_log_20 *log)
+{
+	log->header.capacity = T_ARRAY_SIZE(log->events);
+	return T_scheduler_record((T_scheduler_log *)log);
+}
+
+T_scheduler_log *
+T_scheduler_record_40(T_scheduler_log_40 *log)
+{
+	log->header.capacity = T_ARRAY_SIZE(log->events);
+	return T_scheduler_record((T_scheduler_log *)log);
+}
+
+const T_scheduler_event T_scheduler_event_null;
+
+const T_scheduler_event *
+T_scheduler_next(T_scheduler_header *header, T_scheduler_operation operation,
+    size_t *index)
+{
+	T_scheduler_log *log;
+	size_t i;
+
+	log = (T_scheduler_log *)header;
+
+	if (log == NULL) {
+		return &T_scheduler_event_null;
+	}
+
+	i = *index;
+
+	while (i < log->header.recorded) {
+		if (operation == T_SCHEDULER_ANY ||
+		    operation == log->events[i].operation) {
+			*index = i + 1;
+			return &log->events[i];
+		}
+
+		++i;
+	}
+
+	return &T_scheduler_event_null;
+}
+
+const T_scheduler_event *
+T_scheduler_next_any(T_scheduler_header *header, size_t *index)
+{
+	return T_scheduler_next(header, T_SCHEDULER_ANY, index);
+}
+
+void
+T_scheduler_initialize(const Scheduler_Control *scheduler)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_INITIALIZE
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	T_scheduler_before_operation(&event);
+	(*operations->initialize)(scheduler);
+	T_scheduler_record_event(&event);
+}
+
+void
+T_scheduler_schedule(const Scheduler_Control *scheduler,
+   Thread_Control *thread)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_SCHEDULE,
+		.thread = thread
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	T_scheduler_before_operation(&event);
+	(*operations->schedule)(scheduler, thread);
+	T_scheduler_record_event(&event);
+}
+
+void
+T_scheduler_yield(const Scheduler_Control *scheduler, Thread_Control *thread,
+    Scheduler_Node *node)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_YIELD,
+		.thread = thread,
+		.node = node
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	T_scheduler_before_operation(&event);
+	(*operations->yield)(scheduler, thread, node);
+	T_scheduler_record_event(&event);
+}
+
+void
+T_scheduler_block(const Scheduler_Control *scheduler, Thread_Control *thread,
+    Scheduler_Node *node)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_BLOCK,
+		.thread = thread,
+		.node = node
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	T_scheduler_before_operation(&event);
+	(*operations->block)(scheduler, thread, node);
+	T_scheduler_record_event(&event);
+}
+
+void
+T_scheduler_unblock(const Scheduler_Control *scheduler, Thread_Control *thread,
+    Scheduler_Node *node)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_UNBLOCK,
+		.thread = thread,
+		.node = node
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	T_scheduler_before_operation(&event);
+	(*operations->unblock)(scheduler, thread, node);
+	T_scheduler_record_event(&event);
+}
+
+void
+T_scheduler_update_priority(const Scheduler_Control *scheduler,
+    Thread_Control *thread, Scheduler_Node *node)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_UPDATE_PRIORITY,
+		.thread = thread,
+		.node = node
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	T_scheduler_before_operation(&event);
+	(*operations->update_priority)(scheduler, thread, node);
+	T_scheduler_record_event(&event);
+}
+
+Priority_Control T_scheduler_map_priority(const Scheduler_Control *scheduler,
+    Priority_Control priority)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_MAP_PRIORITY
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	event.map_unmap_priority.in = priority;
+	T_scheduler_before_operation(&event);
+	event.map_unmap_priority.out = (*operations->map_priority)(scheduler,
+	    priority);
+	T_scheduler_record_event(&event);
+	return event.map_unmap_priority.out;
+}
+
+Priority_Control T_scheduler_unmap_priority(const Scheduler_Control *scheduler,
+    Priority_Control priority)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_UNMAP_PRIORITY
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	event.map_unmap_priority.in = priority;
+	T_scheduler_before_operation(&event);
+	event.map_unmap_priority.out = (*operations->unmap_priority)(scheduler,
+	    priority);
+	T_scheduler_record_event(&event);
+	return event.map_unmap_priority.out;
+}
+
+void
+T_scheduler_node_initialize(const Scheduler_Control *scheduler,
+    Scheduler_Node *node, Thread_Control *thread, Priority_Control priority)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_NODE_INITIALIZE,
+		.thread = thread,
+		.node = node
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	T_scheduler_before_operation(&event);
+	(*operations->node_initialize)(scheduler, node, thread, priority);
+	T_scheduler_record_event(&event);
+}
+
+void
+T_scheduler_node_destroy(const Scheduler_Control *scheduler,
+    Scheduler_Node *node)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_NODE_DESTROY,
+		.node = node
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	T_scheduler_before_operation(&event);
+	(*operations->node_destroy)(scheduler, node);
+	T_scheduler_record_event(&event);
+}
+
+void
+T_scheduler_release_job(const Scheduler_Control *scheduler,
+    Thread_Control *thread, Priority_Node *priority, uint64_t deadline,
+    Thread_queue_Context *queue_context)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_RELEASE_JOB,
+		.thread = thread
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	event.release_job.priority = priority;
+	event.release_job.deadline = deadline;
+	T_scheduler_before_operation(&event);
+	(*operations->release_job)(scheduler, thread, priority, deadline, queue_context);
+	T_scheduler_record_event(&event);
+}
+
+void
+T_scheduler_cancel_job(const Scheduler_Control *scheduler,
+    Thread_Control *thread, Priority_Node *priority,
+    Thread_queue_Context *queue_context)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_CANCEL_JOB,
+		.thread = thread
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	event.cancel_job.priority = priority;
+	T_scheduler_before_operation(&event);
+	(*operations->cancel_job)(scheduler, thread, priority, queue_context);
+	T_scheduler_record_event(&event);
+}
+
+void
+T_scheduler_start_idle(const Scheduler_Control *scheduler,
+    Thread_Control *thread, Per_CPU_Control *cpu)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_START_IDLE,
+		.thread = thread
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	T_scheduler_before_operation(&event);
+	(*operations->start_idle)(scheduler, thread, cpu);
+	T_scheduler_record_event(&event);
+}
+
+#ifdef RTEMS_SMP
+bool
+T_scheduler_ask_for_help(const Scheduler_Control *scheduler,
+    Thread_Control *thread, Scheduler_Node *node)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_ASK_FOR_HELP,
+		.thread = thread,
+		.node = node
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	T_scheduler_before_operation(&event);
+	event.ask_for_help.success = (*operations->ask_for_help)(scheduler,
+	    thread, node);
+	T_scheduler_record_event(&event);
+	return event.ask_for_help.success;
+}
+
+void
+T_scheduler_reconsider_help_request(const Scheduler_Control *scheduler,
+    Thread_Control *thread, Scheduler_Node *node)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_RECONSIDER_HELP_REQUEST,
+		.thread = thread,
+		.node = node
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	T_scheduler_before_operation(&event);
+	(*operations->reconsider_help_request)(scheduler, thread, node);
+	T_scheduler_record_event(&event);
+}
+
+void
+T_scheduler_withdraw_node(const Scheduler_Control *scheduler,
+    Thread_Control *thread, Scheduler_Node *node,
+    Thread_Scheduler_state next_state)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_WITHDRAW_NODE,
+		.thread = thread,
+		.node = node
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	event.withdraw_node.next_state = next_state;
+	T_scheduler_before_operation(&event);
+	(*operations->withdraw_node)(scheduler, thread, node, next_state);
+	T_scheduler_record_event(&event);
+}
+
+void
+T_scheduler_make_sticky(const Scheduler_Control *scheduler, Thread_Control *thread,
+    Scheduler_Node *node)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_MAKE_STICKY,
+		.thread = thread,
+		.node = node
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	T_scheduler_before_operation(&event);
+	(*operations->make_sticky)(scheduler, thread, node);
+	T_scheduler_record_event(&event);
+}
+
+void
+T_scheduler_clean_sticky(const Scheduler_Control *scheduler, Thread_Control *thread,
+    Scheduler_Node *node)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_CLEAN_STICKY,
+		.thread = thread,
+		.node = node
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	T_scheduler_before_operation(&event);
+	(*operations->clean_sticky)(scheduler, thread, node);
+	T_scheduler_record_event(&event);
+}
+
+void
+T_scheduler_pin(const Scheduler_Control *scheduler, Thread_Control *thread,
+    Scheduler_Node *node, Per_CPU_Control *cpu)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_PIN,
+		.thread = thread,
+		.node = node
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	event.pin_unpin.cpu = cpu;
+	T_scheduler_before_operation(&event);
+	(*operations->pin)(scheduler, thread, node, cpu);
+	T_scheduler_record_event(&event);
+}
+
+void
+T_scheduler_unpin(const Scheduler_Control *scheduler, Thread_Control *thread,
+    Scheduler_Node *node, Per_CPU_Control *cpu)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_UNPIN,
+		.thread = thread,
+		.node = node
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	event.pin_unpin.cpu = cpu;
+	T_scheduler_before_operation(&event);
+	(*operations->unpin)(scheduler, thread, node, cpu);
+	T_scheduler_record_event(&event);
+}
+
+void
+T_scheduler_add_processor(const Scheduler_Control *scheduler,
+    Thread_Control *idle)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_ADD_PROCESSOR,
+		.thread = idle
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	event.add_processor.idle = idle;
+	T_scheduler_before_operation(&event);
+	(*operations->add_processor)(scheduler, idle);
+	T_scheduler_record_event(&event);
+}
+
+Thread_Control *
+T_scheduler_remove_processor(const Scheduler_Control *scheduler,
+    Per_CPU_Control *cpu)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_REMOVE_PROCESSOR
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	event.remove_processor.cpu = cpu;
+	T_scheduler_before_operation(&event);
+	event.remove_processor.idle =
+	    (*operations->remove_processor)(scheduler, cpu);
+	T_scheduler_record_event(&event);
+	return event.remove_processor.idle;
+}
+
+Status_Control
+T_scheduler_set_affinity(const Scheduler_Control *scheduler,
+    Thread_Control *thread, Scheduler_Node *node,
+    const Processor_mask *affinity)
+{
+	T_scheduler_event event = {
+		.operation = T_SCHEDULER_SET_AFFINITY,
+		.thread = thread,
+		.node = node
+	};
+	const Scheduler_Operations *operations;
+
+	operations = &T_scheduler_operations[_Scheduler_Get_index(scheduler)];
+	event.set_affinity.affinity = *affinity;
+	T_scheduler_before_operation(&event);
+	event.set_affinity.status = (*operations->set_affinity)(scheduler,
+	    thread, node, affinity);
+	T_scheduler_record_event(&event);
+	return event.set_affinity.status;
+}
+#endif