sched: integrate into kent hierarchy

This is an initial, rudimentary attempt at integrating the scheduler
into the kent hierarchy.  There are likely gonna be drastic changes
in the future, and i haven't even tried running the whole thing.

But hey, the code compiles again now!

include/ardix/atomic.h

@@ -2,7 +2,13 @@
 
 #pragma once
 
-/** Enter atomic context. */
+/**
+ * Enter atomic context.
+ *
+ * Since Ardix does not support SMP, this method will simply increment a
+ * reference counter that is checked in the scheduler interrupt routine before
+ * performing the context switch.
+ */
 void atomic_enter(void);
 
 /** Leave atomic context. */

include/ardix/sched.h

@@ -2,6 +2,7 @@
 
 #pragma once
 
+#include <ardix/kent.h>
 #include <ardix/list.h>
 #include <ardix/types.h>
 
@@ -39,6 +40,7 @@ enum task_state {
 
 /** Stores an entire process image. */
 struct task {
+	struct kent kent;
 	/** current stack pointer (only gets updated for task switching) */
 	void *sp;
 	/** first address of the stack (highest if the stack grows downwards) */
@@ -61,7 +63,7 @@ int sched_init(void);
 /**
  * Switch to the next task (interrupt context only).
  * Must be called directly from within an interrupt routine.
- * This selects a new task to be  run and updates the old and new task's `state`
+ * This selects a new task to be run and updates the old and new task's `state`
  * field to the appropriate value.
  *
  * @param curr_sp: stack pointer of the current task
@@ -75,7 +77,7 @@ void *sched_switch(void *curr_sp);
  * @param task: the task to make a copy of
  * @returns the new (child) task copy, or `NULL` on failure
  */
-struct task *sched_task_clone(struct task *dest);
+struct task *sched_task_clone(struct task *task);
 
 /**
  * Request the scheduler be invoked early, resulting in the current task to

kernel/kent.c

@@ -8,31 +8,18 @@
 #include <errno.h>
 #include <stddef.h>
 
-struct kent *kent_root = NULL;
-
-static void kent_root_destroy(struct kent *kent)
-{
-	/*
-	 * this callback should never actually be executed in the first place
-	 * because the kent root lives as long as the kernel is running but hey,
-	 * it's not like our flash memory has a size limit or anything :)
-	 */
-	free(kent);
-	kent_root = NULL;
-}
-
 static struct kent_ops kent_root_ops = {
-	.destroy = &kent_root_destroy,
+	.destroy = NULL,
 };
 
+struct kent _kent_root;
+struct kent *kent_root = NULL;
+
 int kent_root_init(void)
 {
 	if (kent_root != NULL)
 		return -EEXIST;
-
-	kent_root = malloc(sizeof(*kent_root));
-	if (kent_root == NULL)
-		return -ENOMEM;
+	kent_root = &_kent_root;
 
 	kent_root->parent = NULL;
 	kent_root->operations = &kent_root_ops;

kernel/sched.c

@@ -18,6 +18,17 @@ extern uint32_t _estack;
 static struct task *_sched_tasktab[CONFIG_SCHED_MAXTASK];
 struct task *_sched_current_task;
 
+static void sched_kent_destroy(struct kent *kent)
+{
+	struct task *task = container_of(kent, struct task, kent);
+	_sched_tasktab[task->pid] = NULL;
+	free(task);
+}
+
+static struct kent_ops sched_kent_ops = {
+	.destroy = sched_kent_destroy,
+};
+
 int sched_init(void)
 {
 	int i;
@@ -26,19 +37,29 @@ int sched_init(void)
 	if (_sched_current_task == NULL)
 		return -ENOMEM;
 
-	_sched_current_task->sp = &_sstack;
-	_sched_current_task->stack_bottom = &_estack;
-	_sched_current_task->pid = 0;
-	_sched_current_task->state = TASK_READY;
-	_sched_tasktab[0] = _sched_current_task;
+	_sched_current_task->kent.parent = kent_root;
+	_sched_current_task->kent.operations = &sched_kent_ops;
+	i = kent_init(&_sched_current_task->kent);
+	if (i == 0) {
+		_sched_current_task->sp = &_sstack;
+		_sched_current_task->stack_bottom = &_estack;
+		_sched_current_task->pid = 0;
+		_sched_current_task->state = TASK_READY;
+		_sched_tasktab[0] = _sched_current_task;
 
-	for (i = 1; i < CONFIG_SCHED_MAXTASK; i++)
-		_sched_tasktab[i] = NULL;
+		for (i = 1; i < CONFIG_SCHED_MAXTASK; i++)
+			_sched_tasktab[i] = NULL;
 
-	i = arch_watchdog_init();
+		i = arch_watchdog_init();
 
-	if (i == 0)
-		i = arch_sched_hwtimer_init(CONFIG_SCHED_MAXTASK);
+		if (i == 0)
+			i = arch_sched_hwtimer_init(CONFIG_SCHED_MAXTASK);
+	}
+
+	/*
+	 * we don't really need to deallocate resources on error because we
+	 * are going to panic anyways if the scheduler fails to initialize.
+	 */
 
 	return i;
 }
@@ -71,6 +92,7 @@ void *sched_process_switch(void *curr_sp)
 	while (1) {
 		nextpid++;
 		nextpid %= CONFIG_SCHED_MAXTASK;
+
 		tmp = _sched_tasktab[nextpid];
 		if (tmp != NULL && sched_task_should_run(tmp)) {
 			_sched_current_task = tmp;
@@ -98,10 +120,15 @@ struct task *sched_fork(struct task *parent)
 	if (pid == CONFIG_SCHED_MAXTASK)
 		goto err_maxtask;
 
-	child->pid = pid;
-	list_init(&child->children);
-	list_insert(&parent->children, &child->siblings);
+	child->kent.parent = &parent->kent;
+	child->kent.operations = &sched_kent_ops;
+	if (kent_init(&child->kent) != 0)
+		goto err_kent;
 
+	child->pid = pid;
+	return child;
+
+err_kent:
 err_maxtask:
 	free(child);
 err_alloc: