Add ability to request a shared event queue

To allow components with a simple need to schedule a few events to not
have to create their own threads, with all the associated memory
overhead, add 2 central calls to get shared normal and an
interrupt-deferral event queues, each dispatched on their own shared
threads.

For non-RTOS systems, just the normal event queue is provided, and the
application would have to dispatch this itself. This
application-dispatch is also available via a config option, to
potentially save memory by reusing the main thread.

Possible future improvement: the ability for separate components to
request a minimum stack size, and have the JSON combine these requests.
(Analogous tooling has already been mooted for mbed TLS config options
like key size).

events/mbed_events.h

@@ -28,6 +28,8 @@
 #include "events/EventQueue.h"
 #include "events/Event.h"
 
+#include "events/mbed_shared_queues.h"
+
 using namespace events;
 
 #endif

events/mbed_lib.json

@@ -1,6 +1,26 @@
 {
     "name": "events",
     "config": {
-        "present": 1
+        "present": 1,
+        "shared-stacksize": {
+            "help": "Stack size (bytes) for shared event queue thread",
+            "value": 1024
+        },
+        "shared-eventsize": {
+            "help": "Event buffer size (bytes) for shared event queue",
+            "value": 256
+        },
+        "shared-dispatch-from-application": {
+            "help": "No thread created for shared event queue - application will call dispatch from another thread (eg dispatch_forever at end of main)",
+            "value": false
+        },
+        "shared-highprio-stacksize": {
+            "help": "Stack size (bytes) for shared high-priority event queue thread",
+            "value": 1024
+        },
+        "shared-highprio-eventsize": {
+            "help": "Event buffer size (bytes) for shared high-priority event queue",
+            "value": 256
+        }
     }
 }

events/mbed_shared_queues.cpp

@@ -0,0 +1,72 @@
+/* events
+ * Copyright (c) 2017 ARM Limited
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "events/mbed_shared_queues.h"
+#include "mbed.h"
+
+using namespace events;
+
+namespace mbed {
+
+#ifdef MBED_CONF_RTOS_PRESENT
+/* Create an event queue, and start the thread that dispatches it. Static
+ * variables mean this happens once the first time each template instantiation
+ * is called. This is currently instantiated no more than twice.
+ */
+template
+<osPriority Priority, size_t QueueSize, size_t StackSize>
+EventQueue *do_shared_event_queue_with_thread()
+{
+    static uint64_t queue_buffer[QueueSize / sizeof(uint64_t)];
+    static EventQueue queue(sizeof queue_buffer, (unsigned char *) queue_buffer);
+
+    static uint64_t stack[StackSize / sizeof(uint64_t)];
+    static Thread thread(Priority, StackSize, (unsigned char *) stack);
+
+    Thread::State state = thread.get_state();
+    if (state == Thread::Inactive || state == Thread::Deleted) {
+        osStatus status = thread.start(callback(&queue, &EventQueue::dispatch_forever));
+        MBED_ASSERT(status == osOK);
+        if (status != osOK) {
+            return NULL;
+        }
+    }
+
+    return &queue;
+}
+#endif
+
+EventQueue *mbed_event_queue()
+{
+#if MBED_CONF_EVENTS_SHARED_DISPATCH_FROM_APPLICATION || !defined MBED_CONF_RTOS_PRESENT
+    /* Only create the EventQueue, but no dispatching thread */
+    static unsigned char queue_buffer[MBED_CONF_EVENTS_SHARED_EVENTSIZE];
+    static EventQueue queue(sizeof queue_buffer, queue_buffer);
+
+    return &queue;
+#else
+    return do_shared_event_queue_with_thread<osPriorityNormal, MBED_CONF_EVENTS_SHARED_EVENTSIZE, MBED_CONF_EVENTS_SHARED_STACKSIZE>();
+#endif
+}
+
+#ifdef MBED_CONF_RTOS_PRESENT
+EventQueue *mbed_highprio_event_queue()
+{
+    return do_shared_event_queue_with_thread<osPriorityHigh, MBED_CONF_EVENTS_SHARED_HIGHPRIO_EVENTSIZE, MBED_CONF_EVENTS_SHARED_HIGHPRIO_STACKSIZE>();
+}
+#endif
+
+}

events/mbed_shared_queues.h

@@ -0,0 +1,81 @@
+
+/** \addtogroup events */
+/** @{*/
+/* events
+ * Copyright (c) 2017 ARM Limited
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#ifndef MBED_SHARED_QUEUES_H
+#define MBED_SHARED_QUEUES_H
+
+#include "events/EventQueue.h"
+
+namespace mbed {
+
+/**
+ * Return a pointer to an EventQueue, on which normal tasks can be queued.
+ *
+ * All calls to this return the same EventQueue - it and its dispatch thread
+ * are created on the first call to this function. The dispatch thread
+ * runs at default priority (currently osPriorityNormal).
+ *
+ * The EventQueue returned may be used to call() Events, or to chain() other
+ * EventQueues so that they are run in the same context.
+ *
+ * Events (or chained EventQueues) executing on the normal event queue should
+ * normally take less than 10ms to execute, to avoid starving other users. As
+ * such, users can expect that event latency will typically be 10ms or less,
+ * but could occasionally be significantly higher if many events are queued.
+ *
+ * If an RTOS is not present or the configuration option
+ * `events.shared-dispatch-from-application` is set to true, then this
+ * does not create a dedicated dispatch thread - instead the application is
+ * expected to run the EventQueue's dispatch, eg from main. This is necessary
+ * for the event loop to work without an RTOS, or an RTOS system can can save
+ * memory by reusing the main stack.
+ *
+ * @return pointer to event queue
+ */
+events::EventQueue *mbed_event_queue();
+
+#ifdef MBED_CONF_RTOS_PRESENT
+/**
+ * Return a pointer to an EventQueue, on which small high-priority tasks can
+ * be queues, such as simple deferrals from interrupt.
+ *
+ * All calls to this return the same EventQueue - it and its thread are
+ * created on the first call to this function. The dispatch thread
+ * runs at a high priority (currently osPriorityHigh).
+ *
+ * The EventQueue returned may be used to call() Events, or to chain() other
+ * EventQueues so that they are run in the same context.
+ *
+ * Events (or chained EventQueues) executing on the high-priority event queue
+ * should normally take less than 100us to execute, to avoid starving other
+ * users. As such, users can expect that event latency will typically be 100us
+ * or less, but could occasionally be significantly higher if many events are
+ * queued.
+ *
+ * @return pointer to high-priority event queue
+ */
+
+events::EventQueue *mbed_highprio_event_queue();
+
+#endif // MBED_CONF_RTOS_PRESENT
+
+};
+
+#endif
+
+/** @}*/