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 <[email protected]>

arch/arc/core/fast_irq.S

@@ -150,11 +150,11 @@ SECTION_FUNC(TEXT, _firq_exit)
 .balign 4
 _firq_check_for_swap:
 	/* coop thread ? do not schedule */
-	ld_s r0, [r2, _thread_offset_to_prio]
+	ldb.x r0, [r2, _thread_offset_to_prio]
 	brlt r0, 0, _firq_no_reschedule
 
 	/* scheduler locked ? do not schedule */
-	ld_s r0, [r2, _thread_offset_to_sched_locked]
+	ldb_s r0, [r2, _thread_offset_to_sched_locked]
 	brne r0, 0, _firq_no_reschedule
 
 	/* Check if the current thread (in r2) is the cached thread */

arch/arc/core/fault_s.S

@@ -168,11 +168,11 @@ _trap_return:
 .balign 4
 _trap_check_for_swap:
 	/* coop thread ? do not schedule */
-	ld_s r0, [r2, _thread_offset_to_prio]
+	ldb.x r0, [r2, _thread_offset_to_prio]
 	brlt r0, 0, _trap_return
 
 	/* scheduler locked ? do not schedule */
-	ld_s r0, [r2, _thread_offset_to_sched_locked]
+	ldb_s r0, [r2, _thread_offset_to_sched_locked]
 	brne r0, 0, _trap_return
 
 	/* check if the current thread needs to be rescheduled */

arch/arc/core/regular_irq.S

@@ -138,13 +138,13 @@ SECTION_FUNC(TEXT, _rirq_exit)
 	 */
 
 	/* coop thread ? do not schedule */
-	ld_s r0, [r2, _thread_offset_to_prio]
+	ldb.x r0, [r2, _thread_offset_to_prio]
 	cmp_s r0, 0
 	blt.d _rirq_no_reschedule
 	ld sp, [r2, _thread_offset_to_sp]
 
 	/* scheduler locked ? do not schedule */
-	ld_s r0, [r2, _thread_offset_to_sched_locked]
+	ldb_s r0, [r2, _thread_offset_to_sched_locked]
 #ifdef CONFIG_ARC_STACK_CHECKING
 	breq.d r0, 0,  _rirq_reschedule_check
 	ld sp, [r2, _thread_offset_to_sp]

arch/arm/core/exc_exit.S

@@ -84,8 +84,13 @@ SECTION_SUBSEC_FUNC(TEXT, _HandlerModeExit, _ExcExit)
     ldr r0, =_kernel
 
     ldr r1, [r0, #_kernel_offset_to_current]
-    ldr r2, [r1, #_thread_offset_to_prio]
-    ldr r3, [r1, #_thread_offset_to_sched_locked]
+#ifdef CONFIG_CPU_CORTEX_M0_M0PLUS
+    movs r3, #_thread_offset_to_prio
+    ldrsb r2, [r1, r3]
+#else
+    ldrsb r2, [r1, #_thread_offset_to_prio]
+#endif
+    ldrb r3, [r1, #_thread_offset_to_sched_locked]
 
     /* coop thread ? do not schedule */
     cmp r2, #0

arch/nios2/core/exception.S

@@ -132,11 +132,11 @@ on_irq_stack:
 	 */
 
 	/* Do not reschedule coop threads (threads that have negative prio) */
-	ldw  r12, _thread_offset_to_prio(r11)
+	ldb  r12, _thread_offset_to_prio(r11)
 	blt  r12, zero, no_reschedule
 
 	/* Do not reschedule if scheduler is locked */
-	ldw  r12, _thread_offset_to_sched_locked(r11)
+	ldb  r12, _thread_offset_to_sched_locked(r11)
 	bne  r12, zero, no_reschedule
 
 	/* Call into the kernel to see if a scheduling decision is necessary */

arch/x86/core/excstub.S

@@ -135,7 +135,7 @@ SECTION_FUNC(TEXT, _exception_enter)
      * registers and the stack of the preempted thread.
      */
 
-	testl	$EXC_ACTIVE, _thread_offset_to_flags(%edx)
+	testb	$EXC_ACTIVE, _thread_offset_to_execution_flags(%edx)
 	jne	alreadyInException
 	movl	%esp, _thread_offset_to_esf(%edx)
 
@@ -151,7 +151,7 @@ alreadyInException:
 	 * handled in the event of a context switch.
 	 */
 
-	orl	$EXC_ACTIVE, _thread_offset_to_flags(%edx)
+	orb	$EXC_ACTIVE, _thread_offset_to_execution_flags(%edx)
 
 #endif /* CONFIG_FP_SHARING || CONFIG_GDB_INFO */
 
@@ -210,7 +210,7 @@ allDone:
 	 * outermost exception).
 	 */
 
-	andl	$~EXC_ACTIVE, _thread_offset_to_flags (%ecx)
+	andb	$~EXC_ACTIVE, _thread_offset_to_execution_flags (%ecx)
 
 nestedException:
 #endif /* CONFIG_FP_SHARING || CONFIG_GDB_INFO */

arch/x86/core/float.c

@@ -70,7 +70,7 @@ extern uint32_t _sse_mxcsr_default_value;
 static void _FpCtxSave(struct tcs *tcs)
 {
 #ifdef CONFIG_SSE
-	if (tcs->base.flags & K_SSE_REGS) {
+	if (tcs->base.execution_flags & K_SSE_REGS) {
 		_do_fp_and_sse_regs_save(&tcs->arch.preempFloatReg);
 		return;
 	}
@@ -88,7 +88,7 @@ static inline void _FpCtxInit(struct tcs *tcs)
 {
 	_do_fp_regs_init();
 #ifdef CONFIG_SSE
-	if (tcs->base.flags & K_SSE_REGS) {
+	if (tcs->base.execution_flags & K_SSE_REGS) {
 		_do_sse_regs_init();
 	}
 #endif
@@ -114,7 +114,7 @@ void k_float_enable(struct tcs *tcs, unsigned int options)
 
 	/* Indicate thread requires floating point context saving */
 
-	tcs->base.flags |= options;
+	tcs->base.execution_flags |= (uint8_t)options;
 
 	/*
 	 * The current thread might not allow FP instructions, so clear CR0[TS]
@@ -132,7 +132,7 @@ void k_float_enable(struct tcs *tcs, unsigned int options)
 
 	fp_owner = _kernel.current_fp;
 	if (fp_owner) {
-		if (fp_owner->base.flags & INT_OR_EXC_MASK) {
+		if (fp_owner->base.execution_flags & INT_OR_EXC_MASK) {
 			_FpCtxSave(fp_owner);
 		}
 	}
@@ -158,7 +158,7 @@ void k_float_enable(struct tcs *tcs, unsigned int options)
 		 * of the FPU to them (unless we need it ourselves).
 		 */
 
-		if ((_current->base.flags & _FP_USER_MASK) == 0) {
+		if ((_current->base.execution_flags & _FP_USER_MASK) == 0) {
 			/*
 			 * We are not FP-capable, so mark FPU as owned by the
 			 * thread we've just enabled FP support for, then
@@ -212,7 +212,7 @@ void k_float_disable(struct tcs *tcs)
 
 	/* Disable all floating point capabilities for the thread */
 
-	tcs->base.flags &= ~_FP_USER_MASK;
+	tcs->base.execution_flags &= ~_FP_USER_MASK;
 
 	if (tcs == _current) {
 		_FpAccessDisable();

arch/x86/core/intstub.S

@@ -294,11 +294,11 @@ alreadyOnIntStack:
 	 */
 
 	/* do not reschedule coop threads (negative priority) */
-	cmpl	$0, _thread_offset_to_prio(%edx)
+	cmpb	$0, _thread_offset_to_prio(%edx)
 	jl	noReschedule
 
 	/* do not reschedule if scheduler is locked */
-	cmpl	$0, _thread_offset_to_sched_locked(%edx)
+	cmpb	$0, _thread_offset_to_sched_locked(%edx)
 	jne	noReschedule
 
 
@@ -314,7 +314,7 @@ alreadyOnIntStack:
 	 */
 
 #if defined(CONFIG_FP_SHARING) ||  defined(CONFIG_GDB_INFO)
-	orl	$INT_ACTIVE, _thread_offset_to_flags(%edx)
+	orb	$INT_ACTIVE, _thread_offset_to_execution_flags(%edx)
 #endif
 
 	/*
@@ -362,7 +362,7 @@ alreadyOnIntStack:
 	 */
 
 	movl	_kernel + _kernel_offset_to_current, %eax
-	andl	$~INT_ACTIVE, _thread_offset_to_flags (%eax)
+	andb	$~INT_ACTIVE, _thread_offset_to_execution_flags(%eax)
 #endif /* CONFIG_FP_SHARING || CONFIG_GDB_INFO */
 
 	/* Restore volatile registers and return to the interrupted thread */

arch/x86/core/swap.S

@@ -54,9 +54,9 @@
  * the non-volatile integer registers need to be saved in the TCS of the
  * outgoing thread.  The restoration of the integer registers of the incoming
  * thread depends on whether that thread was preemptively context switched out.
- * The INT_ACTIVE and EXC_ACTIVE bits in the k_thread->flags field will signify
- * that the thread was preemptively context switched out, and thus both the
- * volatile and non-volatile integer registers need to be restored.
+ * The INT_ACTIVE and EXC_ACTIVE bits in the k_thread->execution_flags field
+ * will signify that the thread was preemptively context switched out, and thus
+ * both the volatile and non-volatile integer registers need to be restored.
  *
  * The non-volatile registers need to be scrubbed to ensure they contain no
  * sensitive information that could compromise system security.  This is to
@@ -161,7 +161,7 @@ SECTION_FUNC(TEXT, _Swap)
 	 * _and_ whether the thread was context switched out preemptively.
 	 */
 
-	testl	$_FP_USER_MASK, _thread_offset_to_flags(%eax)
+	testb	$_FP_USER_MASK, _thread_offset_to_execution_flags(%eax)
 	je 	restoreContext_NoFloatSwap
 
 
@@ -197,12 +197,12 @@ SECTION_FUNC(TEXT, _Swap)
 	 * was preemptively context switched.
 	 */
 
-	testl	$INT_OR_EXC_MASK, _thread_offset_to_flags(%ebx)
+	testb	$INT_OR_EXC_MASK, _thread_offset_to_execution_flags(%ebx)
 	je	restoreContext_NoFloatSave
 
 
 #ifdef CONFIG_SSE
-	testl	$K_SSE_REGS, _thread_offset_to_flags(%ebx)
+	testb	$K_SSE_REGS, _thread_offset_to_execution_flags(%ebx)
 	je	x87FloatSave
 
 	/*
@@ -237,11 +237,11 @@ restoreContext_NoFloatSave:
 	 * was previously preemptively context switched out.
 	 */
 
-	testl   $INT_OR_EXC_MASK, _thread_offset_to_flags(%eax)
+	testb   $INT_OR_EXC_MASK, _thread_offset_to_execution_flags(%eax)
 	je 	restoreContext_NoFloatRestore
 
 #ifdef CONFIG_SSE
-	testl	$K_SSE_REGS, _thread_offset_to_flags(%eax)
+	testb	$K_SSE_REGS, _thread_offset_to_execution_flags(%eax)
 	je	x87FloatRestore
 
 	fxrstor	_thread_offset_to_preempFloatReg(%eax)
@@ -276,7 +276,7 @@ restoreContext_NoFloatSwap:
 	 * registers
 	 */
 
-	testl	$_FP_USER_MASK, _thread_offset_to_flags(%eax)
+	testb	$_FP_USER_MASK, _thread_offset_to_execution_flags(%eax)
 	jne	CROHandlingDone
 
 	/*

arch/x86/include/kernel_arch_data.h

@@ -52,19 +52,19 @@
 
 #define STACK_ALIGN_SIZE 4
 
-/* x86 Bitmask definitions for the struct k_thread->flags bit field */
+/* x86 Bitmask definitions for struct k_thread->execution_flags */
 
 /* executing context is interrupt handler */
-#define INT_ACTIVE (1 << 31)
+#define INT_ACTIVE (1 << 7)
 
 /* executing context is exception handler */
-#define EXC_ACTIVE (1 << 30)
+#define EXC_ACTIVE (1 << 6)
 
 #define INT_OR_EXC_MASK (INT_ACTIVE | EXC_ACTIVE)
 
 #if defined(CONFIG_FP_SHARING) && defined(CONFIG_SSE)
 /* thread uses SSEx (and also FP) registers */
-#define K_SSE_REGS (1 << 29)
+#define K_SSE_REGS (1 << 5)
 #endif
 
 #if defined(CONFIG_FP_SHARING) && defined(CONFIG_SSE)

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.

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);

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;