kernel/arch: optimize memory use of some thread fields

Some thread fields were 32-bit wide, when they are not even close to
using that full range of values. They are instead changed to 8-bit fields.

- prio can fit in one byte, limiting the priorities range to -128 to 127

- recursive scheduler locking can be limited to 255; a rollover results
  most probably from a logic error

- flags are split into execution flags and thread states; 8 bits is
  enough for each of them currently, with at worst two states and four
  flags to spare (on x86, on other archs, there are six flags to spare)

Doing this saves 8 bytes per stack. It also sets up an incoming
enhancement when checking if the current thread is preemptible on
interrupt exit.

Change-Id: Ieb5321a5b99f99173b0605dd4a193c3bc7ddabf4
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>

10 files changed, 59 insertions, 39 deletions

diff --git a/kernel/Kconfig b/kernel/Kconfig
index a4930b9b4..d987f335d 100644
--- a/kernel/Kconfig
+++ b/kernel/Kconfig
@@ -37,6 +37,7 @@ config NUM_COOP_PRIORITIES
 	prompt "Number of coop priorities" if MULTITHREADING
 	default 16
 	default 1 if !MULTITHREADING
+	range 0 128
 	help
 	Number of cooperative priorities configured in the system. Gives access
 	to priorities:
@@ -66,6 +67,7 @@ config NUM_PREEMPT_PRIORITIES
 	prompt "Number of preemptible priorities" if MULTITHREADING
 	default 15
 	default 0 if !MULTITHREADING
+	range 0 128
 	help
 	Number of preemptible priorities available in the system. Gives access
 	to priorities 0 to CONFIG_NUM_PREEMPT_PRIORITIES - 1.
diff --git a/kernel/include/kernel_offsets.h b/kernel/include/kernel_offsets.h
index 73553c007..63ac010ac 100644
--- a/kernel/include/kernel_offsets.h
+++ b/kernel/include/kernel_offsets.h
@@ -48,7 +48,8 @@ GEN_OFFSET_SYM(_kernel_t, current_fp);
 
 GEN_ABSOLUTE_SYM(_STRUCT_KERNEL_SIZE, sizeof(struct _kernel));
 
-GEN_OFFSET_SYM(_thread_base_t, flags);
+GEN_OFFSET_SYM(_thread_base_t, execution_flags);
+GEN_OFFSET_SYM(_thread_base_t, thread_state);
 GEN_OFFSET_SYM(_thread_base_t, prio);
 GEN_OFFSET_SYM(_thread_base_t, sched_locked);
 GEN_OFFSET_SYM(_thread_base_t, swap_data);
diff --git a/kernel/include/kernel_structs.h b/kernel/include/kernel_structs.h
index 234ea7770..570a6398d 100644
--- a/kernel/include/kernel_structs.h
+++ b/kernel/include/kernel_structs.h
@@ -24,14 +24,17 @@
 #endif
 
 /*
- * Common bitmask definitions for the struct tcs->flags bit field.
+ * bitmask definitions for the execution_flags and state
  *
  * Must be before kerneL_arch_data.h because it might need them to be already
  * defined.
  */
 
-/* thread is defined statically */
-#define K_STATIC (1 << 0)
+
+/* states: common uses low bits, arch-specific use high bits */
+
+/* system thread that must not abort */
+#define K_ESSENTIAL (1 << 0)
 
 /* Thread is waiting on an object */
 #define K_PENDING (1 << 1)
@@ -48,13 +51,20 @@
 /* Not a real thread */
 #define K_DUMMY (1 << 5)
 
-/* system thread that must not abort */
-#define K_ESSENTIAL (1 << 6)
+/* end - states */
+
+
+/* execution flags: common uses low bits, arch-specific use high bits */
+
+/* thread is defined statically */
+#define K_STATIC (1 << 0)
 
 #if defined(CONFIG_FP_SHARING)
 /* thread uses floating point registers */
-#define K_FP_REGS (1 << 7)
+#define K_FP_REGS (1 << 1)
 #endif
+/* end - execution flags */
+
 
 #include <kernel_arch_data.h>
 
@@ -76,13 +86,16 @@ struct _thread_base {
 	sys_dnode_t k_q_node;
 
 	/* execution flags */
-	uint32_t flags;
+	uint8_t execution_flags;
 
-	/* thread priority used to sort linked list */
-	int prio;
+	/* thread state */
+	uint8_t thread_state;
 
 	/* scheduler lock count */
-	volatile uint32_t sched_locked;
+	volatile uint8_t sched_locked;
+
+	/* thread priority used to sort linked list */
+	int8_t prio;
 
 	/* data returned by APIs */
 	void *swap_data;
diff --git a/kernel/include/ksched.h b/kernel/include/ksched.h
index b9c919c45..f83613ba6 100644
--- a/kernel/include/ksched.h
+++ b/kernel/include/ksched.h
@@ -271,25 +271,25 @@ static ALWAYS_INLINE void _sched_unlock_no_reschedule(void)
 
 static inline void _set_thread_states(struct k_thread *thread, uint32_t states)
 {
-	thread->base.flags |= states;
+	thread->base.thread_state |= states;
 }
 
 static inline void _reset_thread_states(struct k_thread *thread,
 					uint32_t states)
 {
-	thread->base.flags &= ~states;
+	thread->base.thread_state &= ~states;
 }
 
 /* mark a thread as being suspended */
 static inline void _mark_thread_as_suspended(struct k_thread *thread)
 {
-	thread->base.flags |= K_SUSPENDED;
+	thread->base.thread_state |= K_SUSPENDED;
 }
 
 /* mark a thread as not being suspended */
 static inline void _mark_thread_as_not_suspended(struct k_thread *thread)
 {
-	thread->base.flags &= ~K_SUSPENDED;
+	thread->base.thread_state &= ~K_SUSPENDED;
 }
 
 static ALWAYS_INLINE int _is_thread_timeout_expired(struct k_thread *thread)
@@ -313,14 +313,14 @@ static inline int _is_thread_timeout_active(struct k_thread *thread)
 
 static inline int _has_thread_started(struct k_thread *thread)
 {
-	return !(thread->base.flags & K_PRESTART);
+	return !(thread->base.thread_state & K_PRESTART);
 }
 
 static inline int _is_thread_prevented_from_running(struct k_thread *thread)
 {
-	return thread->base.flags & (K_PENDING   | K_PRESTART |
-				     K_DEAD      | K_DUMMY    |
-				     K_SUSPENDED);
+	return thread->base.thread_state & (K_PENDING   | K_PRESTART |
+					    K_DEAD      | K_DUMMY    |
+					    K_SUSPENDED);
 
 }
 
@@ -334,19 +334,19 @@ static inline int _is_thread_ready(struct k_thread *thread)
 /* mark a thread as pending in its TCS */
 static inline void _mark_thread_as_pending(struct k_thread *thread)
 {
-	thread->base.flags |= K_PENDING;
+	thread->base.thread_state |= K_PENDING;
 }
 
 /* mark a thread as not pending in its TCS */
 static inline void _mark_thread_as_not_pending(struct k_thread *thread)
 {
-	thread->base.flags &= ~K_PENDING;
+	thread->base.thread_state &= ~K_PENDING;
 }
 
 /* check if a thread is pending */
 static inline int _is_thread_pending(struct k_thread *thread)
 {
-	return !!(thread->base.flags & K_PENDING);
+	return !!(thread->base.thread_state & K_PENDING);
 }
 
 /**
@@ -356,7 +356,7 @@ static inline int _is_thread_pending(struct k_thread *thread)
  */
 static inline void _mark_thread_as_started(struct k_thread *thread)
 {
-	thread->base.flags &= ~K_PRESTART;
+	thread->base.thread_state &= ~K_PRESTART;
 }
 
 /*
@@ -394,7 +394,7 @@ static inline void _ready_thread(struct k_thread *thread)
  */
 static inline void _mark_thread_as_dead(struct k_thread *thread)
 {
-	thread->base.flags |= K_DEAD;
+	thread->base.thread_state |= K_DEAD;
 }
 
 /*
@@ -463,7 +463,7 @@ static inline struct k_thread *_unpend_first_thread(_wait_q_t *wait_q)
 /* must be called with interrupts locked */
 static inline void _unpend_thread(struct k_thread *thread)
 {
-	__ASSERT(thread->base.flags & K_PENDING, "");
+	__ASSERT(thread->base.thread_state & K_PENDING, "");
 
 	sys_dlist_remove(&thread->base.k_q_node);
 	_mark_thread_as_not_pending(thread);
diff --git a/kernel/include/offsets_short.h b/kernel/include/offsets_short.h
index 32918648f..1b7c83159 100644
--- a/kernel/include/offsets_short.h
+++ b/kernel/include/offsets_short.h
@@ -53,8 +53,11 @@
 
 /* base */
 
-#define _thread_offset_to_flags \
-	(___thread_t_base_OFFSET + ___thread_base_t_flags_OFFSET)
+#define _thread_offset_to_thread_state \
+	(___thread_t_base_OFFSET + ___thread_base_t_thread_state_OFFSET)
+
+#define _thread_offset_to_execution_flags \
+	(___thread_t_base_OFFSET + ___thread_base_t_execution_flags_OFFSET)
 
 #define _thread_offset_to_prio \
 	(___thread_t_base_OFFSET + ___thread_base_t_prio_OFFSET)
diff --git a/kernel/init.c b/kernel/init.c
index 01fa27ae9..6f78ab34a 100644
--- a/kernel/init.c
+++ b/kernel/init.c
@@ -203,7 +203,7 @@ static void _main(void *unused1, void *unused2, void *unused3)
 	main();
 
 	/* Terminate thread normally since it has no more work to do */
-	_main_thread->base.flags &= ~K_ESSENTIAL;
+	_main_thread->base.thread_state &= ~K_ESSENTIAL;
 }
 
 void __weak main(void)
@@ -238,7 +238,7 @@ static void prepare_multithreading(struct k_thread *dummy_thread)
 
 	_current = dummy_thread;
 
-	dummy_thread->base.flags = K_ESSENTIAL;
+	dummy_thread->base.thread_state = K_ESSENTIAL;
 #endif
 
 	/* _kernel.ready_q is all zeroes */
diff --git a/kernel/mailbox.c b/kernel/mailbox.c
index 40a70b812..9483b2e2a 100644
--- a/kernel/mailbox.c
+++ b/kernel/mailbox.c
@@ -211,7 +211,7 @@ static void _mbox_message_dispose(struct k_mbox_msg *rx_msg)
 	 * asynchronous send: free asynchronous message descriptor +
 	 * dummy thread pair, then give semaphore (if needed)
 	 */
-	if (sending_thread->base.flags & K_DUMMY) {
+	if (sending_thread->base.thread_state & K_DUMMY) {
 		struct k_sem *async_sem = tx_msg->_async_sem;
 
 		_mbox_async_free((struct k_mbox_async *)sending_thread);
@@ -286,7 +286,7 @@ static int _mbox_message_put(struct k_mbox *mbox, struct k_mbox_msg *tx_msg,
 			 * note: dummy sending thread sits (unqueued)
 			 * until the receiver consumes the message
 			 */
-			if (sending_thread->base.flags & K_DUMMY) {
+			if (sending_thread->base.thread_state & K_DUMMY) {
 				_reschedule_threads(key);
 				return 0;
 			}
@@ -310,7 +310,7 @@ static int _mbox_message_put(struct k_mbox *mbox, struct k_mbox_msg *tx_msg,
 
 #if (CONFIG_NUM_MBOX_ASYNC_MSGS > 0)
 	/* asynchronous send: dummy thread waits on tx queue for receiver */
-	if (sending_thread->base.flags & K_DUMMY) {
+	if (sending_thread->base.thread_state & K_DUMMY) {
 		_pend_thread(sending_thread, &mbox->tx_msg_queue, K_FOREVER);
 		irq_unlock(key);
 		return 0;
diff --git a/kernel/pipes.c b/kernel/pipes.c
index f1a4c1ec6..f44f841ac 100644
--- a/kernel/pipes.c
+++ b/kernel/pipes.c
@@ -111,7 +111,7 @@ static int init_pipes_module(struct device *dev)
 	 */
 
 	for (int i = 0; i < CONFIG_NUM_PIPE_ASYNC_MSGS; i++) {
-		async_msg[i].thread.flags = K_DUMMY;
+		async_msg[i].thread.thread_state = K_DUMMY;
 		async_msg[i].thread.swap_data = &async_msg[i].desc;
 		k_stack_push(&pipe_async_msgs, (uint32_t)&async_msg[i]);
 	}
@@ -377,7 +377,7 @@ static void _pipe_thread_ready(struct k_thread *thread)
 	unsigned int  key;
 
 #if (CONFIG_NUM_PIPE_ASYNC_MSGS > 0)
-	if (thread->base.flags & K_DUMMY) {
+	if (thread->base.thread_state & K_DUMMY) {
 		_pipe_async_finish((struct k_pipe_async *)thread);
 		return;
 	}
diff --git a/kernel/sem.c b/kernel/sem.c
index 56e4eba2b..ca420e741 100644
--- a/kernel/sem.c
+++ b/kernel/sem.c
@@ -170,7 +170,7 @@ static int handle_sem_group(struct k_sem *sem, struct k_thread *thread)
 	sys_dnode_t  *node;
 	sys_dnode_t  *next;
 
-	if (!(thread->base.flags & K_DUMMY)) {
+	if (!(thread->base.thread_state & K_DUMMY)) {
 		/*
 		 * The awakened thread is a real thread and thus was not
 		 * involved in a semaphore group operation.
diff --git a/kernel/thread.c b/kernel/thread.c
index 030f415ac..925b97f0e 100644
--- a/kernel/thread.c
+++ b/kernel/thread.c
@@ -86,7 +86,7 @@ int k_is_in_isr(void)
  */
 void _thread_essential_set(void)
 {
-	_current->base.flags |= K_ESSENTIAL;
+	_current->base.thread_state |= K_ESSENTIAL;
 }
 
 /*
@@ -96,7 +96,7 @@ void _thread_essential_set(void)
  */
 void _thread_essential_clear(void)
 {
-	_current->base.flags &= ~K_ESSENTIAL;
+	_current->base.thread_state &= ~K_ESSENTIAL;
 }
 
 /*
@@ -106,7 +106,7 @@ void _thread_essential_clear(void)
  */
 int _is_thread_essential(void)
 {
-	return _current->base.flags & K_ESSENTIAL;
+	return _current->base.thread_state & K_ESSENTIAL;
 }
 
 void k_busy_wait(uint32_t usec_to_wait)
@@ -437,7 +437,8 @@ void _init_thread_base(struct _thread_base *thread_base, int priority,
 {
 	/* k_q_node is initialized upon first insertion in a list */
 
-	thread_base->flags = options | initial_state;
+	thread_base->execution_flags = (uint8_t)options;
+	thread_base->thread_state = (uint8_t)initial_state;
 
 	thread_base->prio = priority;