Preempt support and fix ARM build for thread_info changes

arch/arm/config.in

@@ -464,6 +464,7 @@ tristate 'Kernel support for a.out binaries' CONFIG_BINFMT_AOUT
 tristate 'Kernel support for ELF binaries' CONFIG_BINFMT_ELF
 tristate 'Kernel support for MISC binaries' CONFIG_BINFMT_MISC
 bool 'Power Management support' CONFIG_PM
+dep_bool 'Preemptible Kernel (experimental)' CONFIG_PREEMPT $CONFIG_CPU_32 $CONFIG_EXPERIMENTAL
 dep_tristate 'Advanced Power Management Emulation' CONFIG_APM $CONFIG_PM
 dep_tristate 'RISC OS personality' CONFIG_ARTHUR $CONFIG_CPU_32
 

arch/arm/kernel/armksyms.c

@@ -21,6 +21,7 @@
 #include <linux/pm.h>
 #include <linux/tty.h>
 #include <linux/vt_kern.h>
+#include <linux/smp_lock.h>
 
 #include <asm/byteorder.h>
 #include <asm/elf.h>
@@ -273,3 +274,7 @@ EXPORT_SYMBOL_NOVERS(__down_trylock_failed);
 EXPORT_SYMBOL_NOVERS(__up_wakeup);
 
 EXPORT_SYMBOL(get_wchan);
+
+#ifdef CONFIG_PREEMPT
+EXPORT_SYMBOL(kernel_flag);
+#endif

arch/arm/kernel/entry-armv.S

@@ -15,6 +15,7 @@
  */
 #include <linux/config.h>
 #include "entry-header.S"
+#include <asm/thread_info.h>
 
 
 #ifdef IOC_BASE
@@ -690,8 +691,7 @@ __dabt_svc:	sub	sp, sp, #S_FRAME_SIZE
 		msr	cpsr_c, r9
 		mov	r2, sp
 		bl	SYMBOL_NAME(do_DataAbort)
-		mov	r0, #PSR_I_BIT | MODE_SVC
-		msr	cpsr_c, r0
+		set_cpsr_c r0, #PSR_I_BIT | MODE_SVC
 		ldr	r0, [sp, #S_PSR]
 		msr	spsr, r0
 		ldmia	sp, {r0 - pc}^			@ load r0 - pc, cpsr
@@ -705,17 +705,50 @@ __irq_svc:	sub	sp, sp, #S_FRAME_SIZE
 		add	r4, sp, #S_SP
 		mov	r6, lr
 		stmia	r4, {r5, r6, r7, r8, r9}	@ save sp_SVC, lr_SVC, pc, cpsr, old_ro
+#ifdef CONFIG_PREEMPT
+		get_thread_info r8
+		ldr	r9, [r8, #TI_PREEMPT]		@ get preempt count
+		add	r7, r9, #1			@ increment it
+		str	r7, [r8, #TI_PREEMPT]
+#endif
 1:		get_irqnr_and_base r0, r6, r5, lr
 		movne	r1, sp
 		@
 		@ routine called with r0 = irq number, r1 = struct pt_regs *
 		@
 		adrsvc	ne, lr, 1b
-		bne	do_IRQ
+		bne	asm_do_IRQ
+#ifdef CONFIG_PREEMPT
+		ldr	r0, [r8, #TI_FLAGS]		@ get flags
+		tst	r0, #_TIF_NEED_RESCHED
+		ldrne	r6, .LCirq_stat
+		blne	svc_preempt
+preempt_return:
+		ldr	r0, [r8, #TI_PREEMPT]		@ read preempt value
+		teq	r0, r7
+		strne	r0, [r0, -r0]			@ bug()
+		str	r9, [r8, #TI_PREEMPT]		@ restore preempt count
+#endif
 		ldr	r0, [sp, #S_PSR]		@ irqs are already disabled
 		msr	spsr, r0
 		ldmia	sp, {r0 - pc}^			@ load r0 - pc, cpsr
 
+#ifdef CONFIG_PREEMPT
+svc_preempt:	teq	r9, #0
+		movne	pc, lr
+		ldr	r0, [r6, #4]			@ local_irq_count
+		ldr	r1, [r6, #8]			@ local_b_count
+		adds	r0, r0, r1
+		movne	pc, lr
+1:		set_cpsr_c r0, #MODE_SVC		@ enable IRQs
+		bl	SYMBOL_NAME(preempt_schedule)
+		set_cpsr_c r0, #PSR_I_BIT | MODE_SVC	@ disable IRQs
+		ldr	r0, [r8, #TI_FLAGS]
+		tst	r0, #_TIF_NEED_RESCHED
+		bne	1b
+		b	preempt_return
+#endif
+
 		.align	5
 __und_svc:	sub	sp, sp, #S_FRAME_SIZE
 		stmia	sp, {r0 - r12}			@ save r0 - r12
@@ -733,8 +766,7 @@ __und_svc:	sub	sp, sp, #S_FRAME_SIZE
 		mov	r1, sp				@ struct pt_regs *regs
 		bl	SYMBOL_NAME(do_undefinstr)
 
-1:		mov	r0, #PSR_I_BIT | MODE_SVC
-		msr	cpsr_c, r0
+1:		set_cpsr_c r0, #PSR_I_BIT | MODE_SVC
 		ldr	lr, [sp, #S_PSR]		@ Get SVC cpsr
 		msr	spsr, lr
 		ldmia	sp, {r0 - pc}^			@ Restore SVC registers
@@ -755,8 +787,7 @@ __pabt_svc:	sub	sp, sp, #S_FRAME_SIZE
 		mov	r0, r2				@ address (pc)
 		mov	r1, sp				@ regs
 		bl	SYMBOL_NAME(do_PrefetchAbort)	@ call abort handler
-		mov	r0, #PSR_I_BIT | MODE_SVC
-		msr	cpsr_c, r0
+		set_cpsr_c r0, #PSR_I_BIT | MODE_SVC
 		ldr	r0, [sp, #S_PSR]
 		msr	spsr, r0
 		ldmia	sp, {r0 - pc}^			@ load r0 - pc, cpsr
@@ -769,6 +800,9 @@ __pabt_svc:	sub	sp, sp, #S_FRAME_SIZE
 .LCprocfns:	.word	SYMBOL_NAME(processor)
 #endif
 .LCfp:		.word	SYMBOL_NAME(fp_enter)
+#ifdef CONFIG_PREEMPT
+.LCirq_stat:	.word	SYMBOL_NAME(irq_stat)
+#endif
 
 		irq_prio_table
 
@@ -793,8 +827,7 @@ __dabt_usr:	sub	sp, sp, #S_FRAME_SIZE		@ Allocate frame size in one go
 #else
 		bl	cpu_data_abort
 #endif
-		mov	r2, #MODE_SVC
-		msr	cpsr_c, r2			@ Enable interrupts
+		set_cpsr_c r2, #MODE_SVC		@ Enable interrupts
 		mov	r2, sp
 		adrsvc	al, lr, ret_from_exception
 		b	SYMBOL_NAME(do_DataAbort)
@@ -809,15 +842,29 @@ __irq_usr:	sub	sp, sp, #S_FRAME_SIZE
 		stmdb	r8, {sp, lr}^
 		alignment_trap r4, r7, __temp_irq
 		zero_fp
+#ifdef CONFIG_PREEMPT
+		get_thread_info r8
+		ldr	r9, [r8, #TI_PREEMPT]		@ get preempt count
+		add	r7, r9, #1			@ increment it
+		str	r7, [r8, #TI_PREEMPT]
+#endif
 1:		get_irqnr_and_base r0, r6, r5, lr
 		movne	r1, sp
 		adrsvc	ne, lr, 1b
 		@
 		@ routine called with r0 = irq number, r1 = struct pt_regs *
 		@
-		bne	do_IRQ
+		bne	asm_do_IRQ
+#ifdef CONFIG_PREEMPT
+		ldr	r0, [r8, #TI_PREEMPT]
+		teq	r0, r7
+		strne	r0, [r0, -r0]
+		str	r9, [r8, #TI_PREEMPT]
+		mov	tsk, r8
+#else
+		get_thread_info tsk
+#endif
 		mov	why, #0
-		get_current_task tsk
 		b	ret_to_user
 
 		.align	5
@@ -833,15 +880,15 @@ __und_usr:	sub	sp, sp, #S_FRAME_SIZE		@ Allocate frame size in one go
 		adrsvc	al, r9, ret_from_exception	@ r9  = normal FP return
 		adrsvc	al, lr, fpundefinstr		@ lr  = undefined instr return
 
-call_fpe:	get_current_task r10
+call_fpe:	get_thread_info r10			@ get current thread
+		ldr	r4, [r10, #TI_TASK]		@ get current task
 		mov	r8, #1
-		strb	r8, [r10, #TSK_USED_MATH]	@ set current->used_math
+		strb	r8, [r4, #TSK_USED_MATH]	@ set current->used_math
 		ldr	r4, .LCfp
-		add	r10, r10, #TSS_FPESAVE		@ r10 = workspace
+		add	r10, r10, #TI_FPSTATE		@ r10 = workspace
 		ldr	pc, [r4]			@ Call FP module USR entry point
 
-fpundefinstr:	mov	r0, #MODE_SVC
-		msr	cpsr_c, r0			@ Enable interrupts
+fpundefinstr:	set_cpsr_c r0, #MODE_SVC		@ Enable interrupts
 		mov	r0, lr
 		mov	r1, sp
 		adrsvc	al, lr, ret_from_exception
@@ -857,8 +904,7 @@ __pabt_usr:	sub	sp, sp, #S_FRAME_SIZE		@ Allocate frame size in one go
 		stmdb	r8, {sp, lr}^			@ Save sp_usr lr_usr
 		alignment_trap r4, r7, __temp_abt
 		zero_fp
-		mov	r0, #MODE_SVC
-		msr	cpsr_c, r0			@ Enable interrupts
+		set_cpsr_c r0, #MODE_SVC		@ Enable interrupts
 		mov	r0, r5				@ address (pc)
 		mov	r1, sp				@ regs
 		bl	SYMBOL_NAME(do_PrefetchAbort)	@ call abort handler
@@ -867,7 +913,7 @@ __pabt_usr:	sub	sp, sp, #S_FRAME_SIZE		@ Allocate frame size in one go
  * This is the return code to user mode for abort handlers
  */
 ENTRY(ret_from_exception)
-		get_current_task tsk
+		get_thread_info tsk
 		mov	why, #0
 		b	ret_to_user
 
@@ -877,16 +923,16 @@ ENTRY(fp_enter)
 		.text
 /*
  * Register switch for ARMv3 and ARMv4 processors
- * r0 = previous, r1 = next, return previous.
+ * r0 = previous thread_info, r1 = next thread_info, return previous.
  * previous and next are guaranteed not to be the same.
  */
 ENTRY(__switch_to)
 		stmfd	sp!, {r4 - sl, fp, lr}		@ Store most regs on stack
 		mrs	ip, cpsr
 		str	ip, [sp, #-4]!			@ Save cpsr_SVC
-		str	sp, [r0, #TSS_SAVE]		@ Save sp_SVC
-		ldr	sp, [r1, #TSS_SAVE]		@ Get saved sp_SVC
-		ldr	r2, [r1, #TSS_DOMAIN]
+		str	sp, [r0, #TI_CPU_SAVE]		@ Save sp_SVC
+		ldr	sp, [r1, #TI_CPU_SAVE]		@ Get saved sp_SVC
+		ldr	r2, [r1, #TI_CPU_DOMAIN]
 		ldr	ip, [sp], #4
 		mcr	p15, 0, r2, c3, c0		@ Set domain register
 		msr	spsr, ip			@ Save tasks CPSR into SPSR for this return

arch/arm/kernel/entry-common.S

@@ -9,6 +9,7 @@
  */
 #include <linux/config.h>
 #include "entry-header.S"
+#include <asm/thread_info.h>
 
 /* 
  * We rely on the fact that R0 is at the bottom of the stack (due to
@@ -34,39 +35,42 @@ ENTRY(__do_softirq)
  * stack.
  */
 ret_fast_syscall:
-#error	ldr	r1, [tsk, #TSK_NEED_RESCHED]
-#error	ldr	r2, [tsk, #TSK_SIGPENDING]
-	teq	r1, #0				@ need_resched || sigpending
-	teqeq	r2, #0
-	bne	slow
+	set_cpsr_c r1, #PSR_I_BIT | MODE_SVC	@ disable interrupts
+	ldr	r1, [tsk, #TI_FLAGS]
+	tst	r1, #_TIF_WORK_MASK
+	bne	ret_fast_work
 	fast_restore_user_regs
 
 /*
  * Ok, we need to do extra processing, enter the slow path.
  */
-slow:	str	r0, [sp, #S_R0+S_OFF]!	@ returned r0
-	b	1f
+ret_fast_work:
+	str	r0, [sp, #S_R0+S_OFF]!	@ returned r0
+	b	work_pending
 
+work_resched:
+	bl	SYMBOL_NAME(schedule)
 /*
  * "slow" syscall return path.  "why" tells us if this was a real syscall.
  */
-reschedule:
-	bl	SYMBOL_NAME(schedule)
 ENTRY(ret_to_user)
 ret_slow_syscall:
-#error	ldr	r1, [tsk, #TSK_NEED_RESCHED]
-#error	ldr	r2, [tsk, #TSK_SIGPENDING]
-1:	teq	r1, #0				@ need_resched => schedule()
-	bne	reschedule
-	teq	r2, #0				@ sigpending => do_signal()
-	blne	__do_signal
+	set_cpsr_c r1, #PSR_I_BIT | MODE_SVC
+	ldr	r1, [tsk, #TI_FLAGS]
+	tst	r1, #_TIF_WORK_MASK
+	beq	no_work_pending
+work_pending:
+	tst	r1, #_TIF_NEED_RESCHED
+	bne	work_resched
+	tst	r1, #_TIF_NOTIFY_RESUME | _TIF_SIGPENDING
+	blne	__do_notify_resume
+no_work_pending:
 	restore_user_regs
 
-__do_signal:
-	mov	r0, #0				@ NULL 'oldset'
-	mov	r1, sp				@ 'regs'
+__do_notify_resume:
+	mov	r0, sp				@ 'regs'
 	mov	r2, why				@ 'syscall'
-#error	b	SYMBOL_NAME(do_signal)		@ note the bl above sets lr
+	b	SYMBOL_NAME(do_notify_resume)	@ note the bl above sets lr
 
 /*
  * This is how we return from a fork.  __switch_to will be calling us
@@ -75,14 +79,14 @@ __do_signal:
  */
 ENTRY(ret_from_fork)
 	bl	SYMBOL_NAME(schedule_tail)
-	get_current_task tsk
-	ldr	ip, [tsk, #TSK_PTRACE]		@ check for syscall tracing
+	get_thread_info tsk
+	ldr	ip, [tsk, #TI_FLAGS]		@ check for syscall tracing
 	mov	why, #1
-	tst	ip, #PT_TRACESYS		@ are we tracing syscalls?
+	tst	ip, #_TIF_SYSCALL_TRACE		@ are we tracing syscalls?
 	beq	ret_slow_syscall
 	mov	r1, sp
 	mov	r0, #1				@ trace exit [IP = 1]
-#error	bl	SYMBOL_NAME(syscall_trace)
+	bl	SYMBOL_NAME(syscall_trace)
 	b	ret_slow_syscall
 	
 
@@ -134,12 +138,12 @@ ENTRY(vector_swi)
 
 	str	r4, [sp, #-S_OFF]!		@ push fifth arg
 
-	get_current_task tsk
-	ldr	ip, [tsk, #TSK_PTRACE]		@ check for syscall tracing
+	get_thread_info tsk
+	ldr	ip, [tsk, #TI_FLAGS]		@ check for syscall tracing
 	bic	scno, scno, #0xff000000		@ mask off SWI op-code
 	eor	scno, scno, #OS_NUMBER << 20	@ check OS number
 	adr	tbl, sys_call_table		@ load syscall table pointer
-	tst	ip, #PT_TRACESYS		@ are we tracing syscalls?
+	tst	ip, #_TIF_SYSCALL_TRACE		@ are we tracing syscalls?
 	bne	__sys_trace
 
 	adrsvc	al, lr, ret_fast_syscall	@ return address
@@ -160,7 +164,7 @@ ENTRY(vector_swi)
 __sys_trace:
 	add	r1, sp, #S_OFF
 	mov	r0, #0				@ trace entry [IP = 0]
-#error	bl	SYMBOL_NAME(syscall_trace)
+	bl	SYMBOL_NAME(syscall_trace)
 
 	adrsvc	al, lr, __sys_trace_return	@ return address
 	add	r1, sp, #S_R0 + S_OFF		@ pointer to regs
@@ -173,7 +177,7 @@ __sys_trace_return:
 	str	r0, [sp, #S_R0 + S_OFF]!	@ save returned r0
 	mov	r1, sp
 	mov	r0, #1				@ trace exit [IP = 1]
-#error	bl	SYMBOL_NAME(syscall_trace)
+	bl	SYMBOL_NAME(syscall_trace)
 	b	ret_slow_syscall
 
 	.align	5

arch/arm/kernel/entry-header.S

@@ -113,7 +113,7 @@
 		msr	cpsr_c, \temp
 		.endm
 
-		.macro	get_current_task, rd
+		.macro	get_thread_info, rd
 		mov	\rd, sp, lsr #13
 		mov	\rd, \rd, lsl #13
 		.endm
@@ -171,7 +171,7 @@
 		.macro	initialise_traps_extra
 		.endm
 
-		.macro	get_current_task, rd
+		.macro	get_thread_info, rd
 		mov	\rd, sp, lsr #13
 		mov	\rd, \rd, lsl #13
 		.endm
@@ -200,7 +200,7 @@
 scno	.req	r7		@ syscall number
 tbl	.req	r8		@ syscall table pointer
 why	.req	r8		@ Linux syscall (!= 0)
-tsk	.req	r9				@ current task
+tsk	.req	r9		@ current thread_info
 
 /*
  * Get the system call number.
@@ -217,3 +217,12 @@ tsk	.req	r9				@ current task
 #endif
 	.endm
 
+	.macro	set_cpsr_c, reg, mode
+#if 1
+	mov	\reg, \mode
+	msr	cpsr_c, \reg
+#else
+	msr	cpsr_c, \mode
+#endif
+	.endm
+

arch/arm/kernel/head.S

@@ -149,7 +149,7 @@ __switch_data:	.long	__mmap_switched
 		.long	SYMBOL_NAME(processor_id)
 		.long	SYMBOL_NAME(__machine_arch_type)
 		.long	SYMBOL_NAME(cr_alignment)
-		.long	SYMBOL_NAME(init_task_union)+8192
+		.long	SYMBOL_NAME(init_thread_union)+8192
 
 		.type	__ret, %function
 __ret:		ldr	lr, __switch_data

arch/arm/kernel/init_task.c

@@ -16,7 +16,7 @@ static struct signal_struct init_signals = INIT_SIGNALS;
 struct mm_struct init_mm = INIT_MM(init_mm);
 
 /*
- * Initial task structure.
+ * Initial thread structure.
  *
  * We need to make sure that this is 8192-byte aligned due to the
  * way process stacks are handled. This is done by making sure
@@ -25,5 +25,13 @@ struct mm_struct init_mm = INIT_MM(init_mm);
  *
  * The things we do for performance..
  */
-union task_union init_task_union __attribute__((__section__(".init.task"))) =
-		{ INIT_TASK(init_task_union.task) };
+union thread_union init_thread_union
+	__attribute__((__section__(".init.task"))) =
+		{ INIT_THREAD_INFO(init_task) };
+
+/*
+ * Initial task structure.
+ *
+ * All other task structs will be allocated on slabs in fork.c
+ */
+struct task_struct init_task = INIT_TASK(init_task);

arch/arm/kernel/process.c

@@ -19,6 +19,7 @@
 #include <linux/ptrace.h>
 #include <linux/slab.h>
 #include <linux/user.h>
+#include <linux/a.out.h>
 #include <linux/delay.h>
 #include <linux/reboot.h>
 #include <linux/interrupt.h>
@@ -27,6 +28,7 @@
 #include <asm/system.h>
 #include <asm/io.h>
 #include <asm/leds.h>
+#include <asm/processor.h>
 #include <asm/uaccess.h>
 
 /*
@@ -83,9 +85,7 @@ void (*pm_power_off)(void);
 void cpu_idle(void)
 {
 	/* endless idle loop with no priority at all */
-	init_idle();
-	current->nice = 20;
-
+	preempt_disable();
 	while (1) {
 		void (*idle)(void) = pm_idle;
 		if (!idle)
@@ -228,51 +228,61 @@ void show_fpregs(struct user_fp *regs)
 /*
  * Task structure and kernel stack allocation.
  */
-static struct task_struct *task_struct_head;
-static unsigned int nr_task_struct;
+static unsigned long *thread_info_head;
+static unsigned int nr_thread_info;
 
 #ifdef CONFIG_CPU_32
 #define EXTRA_TASK_STRUCT	4
+#define ll_alloc_task_struct() ((struct thread_info *) __get_free_pages(GFP_KERNEL,1))
+#define ll_free_task_struct(p) free_pages((unsigned long)(p),1)
 #else
+extern unsigned long get_page_8k(int priority);
+extern void free_page_8k(unsigned long page);
+
 #define EXTRA_TASK_STRUCT	0
+#define ll_alloc_task_struct()	((struct task_struct *)get_page_8k(GFP_KERNEL))
+#define ll_free_task_struct(p)  free_page_8k((unsigned long)(p))
 #endif
 
-struct task_struct *alloc_task_struct(void)
+struct thread_info *alloc_thread_info(void)
 {
-	struct task_struct *tsk;
+	struct thread_info *thread = NULL;
 
-	if (EXTRA_TASK_STRUCT)
-		tsk = task_struct_head;
-	else
-		tsk = NULL;
+	if (EXTRA_TASK_STRUCT) {
+		unsigned long *p = thread_info_head;
 
-	if (tsk) {
-		task_struct_head = tsk->next_task;
-		nr_task_struct -= 1;
-	} else
-		tsk = ll_alloc_task_struct();
+		if (p) {
+			thread_info_head = (unsigned long *)p[0];
+			nr_thread_info -= 1;
+		}
+		thread = (struct thread_info *)p;
+	}
+
+	if (!thread)
+		thread = ll_alloc_task_struct();
 
 #ifdef CONFIG_SYSRQ
 	/*
 	 * The stack must be cleared if you want SYSRQ-T to
 	 * give sensible stack usage information
 	 */
-	if (tsk) {
-		char *p = (char *)tsk;
+	if (thread) {
+		char *p = (char *)thread;
 		memzero(p+KERNEL_STACK_SIZE, KERNEL_STACK_SIZE);
 	}
 #endif
-	return tsk;
+	return thread;
 }
 
-void __free_task_struct(struct task_struct *p)
+void free_thread_info(struct thread_info *thread)
 {
-	if (EXTRA_TASK_STRUCT && nr_task_struct < EXTRA_TASK_STRUCT) {
-		p->next_task = task_struct_head;
-		task_struct_head = p;
-		nr_task_struct += 1;
+	if (EXTRA_TASK_STRUCT && nr_thread_info < EXTRA_TASK_STRUCT) {
+		unsigned long *p = (unsigned long *)thread;
+		p[0] = (unsigned long)thread_info_head;
+		thread_info_head = p;
+		nr_thread_info += 1;
 	} else
-		ll_free_task_struct(p);
+		ll_free_task_struct(thread);
 }
 
 /*
@@ -284,10 +294,13 @@ void exit_thread(void)
 
 void flush_thread(void)
 {
-	memset(&current->thread.debug, 0, sizeof(struct debug_info));
-	memset(&current->thread.fpstate, 0, sizeof(union fp_state));
+	struct thread_info *thread = current_thread_info();
+	struct task_struct *tsk = current;
+
+	memset(&tsk->thread.debug, 0, sizeof(struct debug_info));
+	memset(&thread->fpstate, 0, sizeof(union fp_state));
+
 	current->used_math = 0;
-	current->flags &= ~PF_USEDFPU;
 }
 
 void release_thread(struct task_struct *dead_task)
@@ -300,21 +313,19 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long esp,
 	unsigned long unused,
 	struct task_struct * p, struct pt_regs * regs)
 {
-	struct pt_regs * childregs;
-	struct context_save_struct * save;
+	struct pt_regs *childregs;
+	struct cpu_context_save *save;
 
-	atomic_set(&p->thread.refcount, 1);
-
-	childregs = ((struct pt_regs *)((unsigned long)p + 8192)) - 1;
+	childregs = ((struct pt_regs *)((unsigned long)p->thread_info + THREAD_SIZE)) - 1;
 	*childregs = *regs;
 	childregs->ARM_r0 = 0;
 	childregs->ARM_sp = esp;
 
-	save = ((struct context_save_struct *)(childregs)) - 1;
+	save = ((struct cpu_context_save *)(childregs)) - 1;
 	*save = INIT_CSS;
 	save->pc |= (unsigned long)ret_from_fork;
 
-	p->thread.save = save;
+	p->thread_info->cpu_context = save;
 
 	return 0;
 }
@@ -324,10 +335,13 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long esp,
  */
 int dump_fpu (struct pt_regs *regs, struct user_fp *fp)
 {
-	if (current->used_math)
-		memcpy(fp, &current->thread.fpstate.soft, sizeof (*fp));
+	struct thread_info *thread = current_thread_info();
+	int used_math = current->used_math;
+
+	if (used_math)
+		memcpy(fp, &thread->fpstate.soft, sizeof (*fp));
 
-	return current->used_math;
+	return used_math;
 }
 
 /*
@@ -405,7 +419,7 @@ unsigned long get_wchan(struct task_struct *p)
 		return 0;
 
 	stack_page = 4096 + (unsigned long)p;
-	fp = get_css_fp(&p->thread);
+	fp = thread_saved_fp(p);
 	do {
 		if (fp < stack_page || fp > 4092+stack_page)
 			return 0;

arch/arm/kernel/ptrace.c

@@ -456,9 +456,9 @@ static int do_ptrace(int request, struct task_struct *child, long addr, long dat
 			if ((unsigned long) data > _NSIG)
 				break;
 			if (request == PTRACE_SYSCALL)
-				child->ptrace |= PT_TRACESYS;
+				set_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
 			else
-				child->ptrace &= ~PT_TRACESYS;
+				clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
 			child->exit_code = data;
 			/* make sure single-step breakpoint is gone. */
 			__ptrace_cancel_bpt(child);
@@ -491,7 +491,7 @@ static int do_ptrace(int request, struct task_struct *child, long addr, long dat
 			if ((unsigned long) data > _NSIG)
 				break;
 			child->thread.debug.nsaved = -1;
-			child->ptrace &= ~PT_TRACESYS;
+			clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
 			child->exit_code = data;
 			/* give it a chance to run. */
 			wake_up_process(child);
@@ -548,7 +548,7 @@ static int do_ptrace(int request, struct task_struct *child, long addr, long dat
 				break;
 
 			/* we should check child->used_math here */
-			ret = __copy_to_user((void *)data, &child->thread.fpstate,
+			ret = __copy_to_user((void *)data, &child->thread_info->fpstate,
 					     sizeof(struct user_fp)) ? -EFAULT : 0;
 			break;
 		
@@ -561,7 +561,7 @@ static int do_ptrace(int request, struct task_struct *child, long addr, long dat
 				break;
 
 			child->used_math = 1;
-			ret = __copy_from_user(&child->thread.fpstate, (void *)data,
+			ret = __copy_from_user(&child->thread_info->fpstate, (void *)data,
 					   sizeof(struct user_fp)) ? -EFAULT : 0;
 			break;
 
@@ -617,7 +617,7 @@ asmlinkage int sys_ptrace(long request, long pid, long addr, long data)
 	ret = do_ptrace(request, child, addr, data);
 
 out_tsk:
-	free_task_struct(child);
+	put_task_struct(child);
 out:
 	unlock_kernel();
 	return ret;
@@ -627,8 +627,9 @@ asmlinkage void syscall_trace(int why, struct pt_regs *regs)
 {
 	unsigned long ip;
 
-	if ((current->ptrace & (PT_PTRACED|PT_TRACESYS))
-			!= (PT_PTRACED|PT_TRACESYS))
+	if (!test_thread_flag(TIF_SYSCALL_TRACE))
+		return;
+	if (!(current->ptrace & PT_PTRACED))
 		return;
 
 	/*

arch/arm/kernel/setup.c

@@ -38,6 +38,10 @@
 #define CONFIG_CMDLINE ""
 #endif
 
+#ifdef CONFIG_PREEMPT
+spinlock_t kernel_flag __cacheline_aligned_in_smp = SPIN_LOCK_UNLOCKED;
+#endif
+
 #if defined(CONFIG_FPE_NWFPE) || defined(CONFIG_FPE_FASTFPE)
 char fpe_type[8];
 

arch/arm/kernel/signal.c

@@ -480,6 +480,7 @@ static void
 handle_signal(unsigned long sig, struct k_sigaction *ka,
 	      siginfo_t *info, sigset_t *oldset, struct pt_regs * regs)
 {
+	struct thread_info *thread = current_thread_info();
 	struct task_struct *tsk = current;
 	int usig = sig;
 	int ret;
@@ -487,8 +488,8 @@ handle_signal(unsigned long sig, struct k_sigaction *ka,
 	/*
 	 * translate the signal
 	 */
-	if (usig < 32 && tsk->exec_domain && tsk->exec_domain->signal_invmap)
-		usig = tsk->exec_domain->signal_invmap[usig];
+	if (usig < 32 && thread->exec_domain && thread->exec_domain->signal_invmap)
+		usig = thread->exec_domain->signal_invmap[usig];
 
 	/*
 	 * Set up the stack frame
@@ -532,7 +533,7 @@ handle_signal(unsigned long sig, struct k_sigaction *ka,
  * the kernel can handle, and then we build all the user-level signal handling
  * stack-frames in one go after that.
  */
-asmlinkage int do_signal(sigset_t *oldset, struct pt_regs *regs, int syscall)
+int do_signal(sigset_t *oldset, struct pt_regs *regs, int syscall)
 {
 	struct k_sigaction *ka;
 	siginfo_t info;
@@ -678,3 +679,10 @@ asmlinkage int do_signal(sigset_t *oldset, struct pt_regs *regs, int syscall)
 		ptrace_set_bpt(current);
 	return 0;
 }
+
+asmlinkage void
+do_notify_resume(struct pt_regs *regs, unsigned int thread_flags, int syscall)
+{
+	if (thread_flags & _TIF_SIGPENDING)
+		do_signal(NULL, regs, syscall);
+}

arch/arm/kernel/traps.c

@@ -115,7 +115,7 @@ static void dump_instr(struct pt_regs *regs)
 
 static void dump_stack(struct task_struct *tsk, unsigned long sp)
 {
-	dump_mem("Stack: ", sp - 16, 8192+(unsigned long)tsk);
+	dump_mem("Stack: ", sp - 16, 8192+(unsigned long)tsk->thread_info);
 }
 
 static void dump_backtrace(struct pt_regs *regs, struct task_struct *tsk)
@@ -146,7 +146,7 @@ static void dump_backtrace(struct pt_regs *regs, struct task_struct *tsk)
 void show_trace_task(struct task_struct *tsk)
 {
 	if (tsk != current) {
-		unsigned int fp = tsk->thread.save->fp;
+		unsigned int fp = thread_saved_fp(tsk);
 		c_backtrace(fp, 0x10);
 	}
 }
@@ -304,16 +304,17 @@ asmlinkage void bad_mode(struct pt_regs *regs, int reason, int proc_mode)
 
 static int bad_syscall(int n, struct pt_regs *regs)
 {
+	struct thread_info *thread = current_thread_info();
 	siginfo_t info;
 
 	/* You might think just testing `handler' would be enough, but PER_LINUX
 	 * points it to no_lcall7 to catch undercover SVr4 binaries.  Gutted.
 	 */
-	if (current->personality != PER_LINUX && current->exec_domain->handler) {
+	if (current->personality != PER_LINUX && thread->exec_domain->handler) {
 		/* Hand it off to iBCS.  The extra parameter and consequent type 
 		 * forcing is necessary because of the weird ARM calling convention.
 		 */
-		current->exec_domain->handler(n, regs);
+		thread->exec_domain->handler(n, regs);
 		return regs->ARM_r0;
 	}
 

arch/arm/lib/getuser.S

@@ -27,12 +27,13 @@
  * Note also that it is intended that __get_user_bad is not global.
  */
 #include <asm/constants.h>
+#include <asm/thread_info.h>
 
 	.global	__get_user_1
 __get_user_1:
 	bic	r1, sp, #0x1f00
 	bic	r1, r1, #0x00ff
-	ldr	r1, [r1, #TSK_ADDR_LIMIT]
+	ldr	r1, [r1, #TI_ADDR_LIMIT]
 	sub	r1, r1, #1
 	cmp	r0, r1
 1:	ldrlsbt	r1, [r0]
@@ -44,7 +45,7 @@ __get_user_1:
 __get_user_2:
 	bic	r2, sp, #0x1f00
 	bic	r2, r2, #0x00ff
-	ldr	r2, [r2, #TSK_ADDR_LIMIT]
+	ldr	r2, [r2, #TI_ADDR_LIMIT]
 	sub	r2, r2, #2
 	cmp	r0, r2
 2:	ldrlsbt	r1, [r0], #1
@@ -62,7 +63,7 @@ __get_user_2:
 __get_user_4:
 	bic	r1, sp, #0x1f00
 	bic	r1, r1, #0x00ff
-	ldr	r1, [r1, #TSK_ADDR_LIMIT]
+	ldr	r1, [r1, #TI_ADDR_LIMIT]
 	sub	r1, r1, #4
 	cmp	r0, r1
 4:	ldrlst	r1, [r0]
@@ -74,7 +75,7 @@ __get_user_4:
 __get_user_8:
 	bic	r2, sp, #0x1f00
 	bic	r2, r2, #0x00ff
-	ldr	r2, [r2, #TSK_ADDR_LIMIT]
+	ldr	r2, [r2, #TI_ADDR_LIMIT]
 	sub	r2, r2, #8
 	cmp	r0, r2
 5:	ldrlst	r1, [r0], #4

arch/arm/lib/putuser.S

@@ -27,12 +27,13 @@
  * Note also that it is intended that __put_user_bad is not global.
  */
 #include <asm/constants.h>
+#include <asm/thread_info.h>
 
 	.global	__put_user_1
 __put_user_1:
 	bic	r2, sp, #0x1f00
 	bic	r2, r2, #0x00ff
-	ldr	r2, [r2, #TSK_ADDR_LIMIT]
+	ldr	r2, [r2, #TI_ADDR_LIMIT]
 	sub	r2, r2, #1
 	cmp	r0, r2
 1:	strlsbt	r1, [r0]
@@ -44,7 +45,7 @@ __put_user_1:
 __put_user_2:
 	bic	r2, sp, #0x1f00
 	bic	r2, r2, #0x00ff
-	ldr	r2, [r2, #TSK_ADDR_LIMIT]
+	ldr	r2, [r2, #TI_ADDR_LIMIT]
 	sub	r2, r2, #2
 	cmp	r0, r2
 	movls	r2, r1, lsr #8
@@ -63,7 +64,7 @@ __put_user_2:
 __put_user_4:
 	bic	r2, sp, #0x1f00
 	bic	r2, r2, #0x00ff
-	ldr	r2, [r2, #TSK_ADDR_LIMIT]
+	ldr	r2, [r2, #TI_ADDR_LIMIT]
 	sub	r2, r2, #4
 	cmp	r0, r2
 4:	strlst	r1, [r0]
@@ -75,7 +76,7 @@ __put_user_4:
 __put_user_8:
 	bic	ip, sp, #0x1f00
 	bic	ip, ip, #0x00ff
-	ldr	ip, [ip, #TSK_ADDR_LIMIT]
+	ldr	ip, [ip, #TI_ADDR_LIMIT]
 	sub	ip, ip, #8
 	cmp	r0, ip
 5:	strlst	r1, [r0], #4

arch/arm/tools/getconstants.c

@@ -32,23 +32,21 @@
 #endif
 
 #define OFF_TSK(n) (unsigned long)&(((struct task_struct *)0)->n)
+#define OFF_VMA(n) (unsigned long)&(((struct vm_area_struct *)0)->n)
 
 #define DEFN(name,off) asm("\n#define "name" %0" :: "I" (off))
 
 void func(void)
 {
-#error DEFN("TSK_SIGPENDING",		OFF_TSK(sigpending));
-DEFN("TSK_ADDR_LIMIT",		OFF_TSK(addr_limit));
-#error DEFN("TSK_NEED_RESCHED",	OFF_TSK(need_resched));
-#error DEFN("TSK_PTRACE",		OFF_TSK(ptrace));
 DEFN("TSK_USED_MATH",		OFF_TSK(used_math));
+DEFN("TSK_ACTIVE_MM",		OFF_TSK(active_mm));
 
-DEFN("TSS_SAVE",		OFF_TSK(thread.save));
-DEFN("TSS_FPESAVE",		OFF_TSK(thread.fpstate.soft.save));
+DEFN("VMA_VM_MM",		OFF_VMA(vm_mm));
+DEFN("VMA_VM_FLAGS",		OFF_VMA(vm_flags));
 
-#ifdef CONFIG_CPU_32
-DEFN("TSS_DOMAIN",		OFF_TSK(thread.domain));
+DEFN("VM_EXEC",			VM_EXEC);
 
+#ifdef CONFIG_CPU_32
 DEFN("HPTE_TYPE_SMALL",		PTE_TYPE_SMALL);
 DEFN("HPTE_AP_READ",		PTE_AP_READ);
 DEFN("HPTE_AP_WRITE",		PTE_AP_WRITE);

include/asm-arm/bitops.h

@@ -301,6 +301,23 @@ static inline unsigned long ffz(unsigned long word)
         return k;
 }
 
+/*
+ * ffz = Find First Zero in word. Undefined if no zero exists,
+ * so code should check against ~0UL first..
+ */
+static inline unsigned long __ffs(unsigned long word)
+{
+	int k;
+
+	k = 31;
+	if (word & 0x0000ffff) { k -= 16; word <<= 16; }
+	if (word & 0x00ff0000) { k -= 8;  word <<= 8;  }
+	if (word & 0x0f000000) { k -= 4;  word <<= 4;  }
+	if (word & 0x30000000) { k -= 2;  word <<= 2;  }
+	if (word & 0x40000000) { k -= 1; }
+        return k;
+}
+
 /*
  * ffs: find first bit set. This is defined the same way as
  * the libc and compiler builtin ffs routines, therefore

include/asm-arm/current.h

 #ifndef _ASMARM_CURRENT_H
 #define _ASMARM_CURRENT_H
 
+#include <asm/thread_info.h>
+
 static inline struct task_struct *get_current(void) __attribute__ (( __const__ ));
 
 static inline struct task_struct *get_current(void)
 {
-	register unsigned long sp asm ("sp");
-	return (struct task_struct *)(sp & ~0x1fff);
+	return current_thread_info()->task;
 }
 
 #define current (get_current())

include/asm-arm/fpstate.h

+/*
+ *  linux/include/asm-arm/fpstate.h
+ *
+ *  Copyright (C) 1995 Russell King
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#ifndef __ASM_ARM_FPSTATE_H
+#define __ASM_ARM_FPSTATE_H
+
+#define FP_SIZE 35
+
+struct fp_hard_struct {
+	unsigned int save[FP_SIZE];		/* as yet undefined */
+};
+
+struct fp_soft_struct {
+	unsigned int save[FP_SIZE];		/* undefined information */
+};
+
+union fp_state {
+	struct fp_hard_struct	hard;
+	struct fp_soft_struct	soft;
+};
+
+#endif

include/asm-arm/hardirq.h

@@ -34,6 +34,7 @@ typedef struct {
 #define irq_exit(cpu,irq)	(local_irq_count(cpu)--)
 
 #define synchronize_irq()	do { } while (0)
+#define release_irqlock(cpu)	do { } while (0)
 
 #else
 #error SMP not supported

include/asm-arm/mmu_context.h

@@ -42,14 +42,34 @@ static inline void
 switch_mm(struct mm_struct *prev, struct mm_struct *next,
 	  struct task_struct *tsk, unsigned int cpu)
 {
-	if (prev != next) {
+	if (prev != next)
 		cpu_switch_mm(next->pgd, tsk);
-		clear_bit(cpu, &prev->cpu_vm_mask);
-	}
-	set_bit(cpu, &next->cpu_vm_mask);
 }
 
 #define activate_mm(prev, next) \
 	switch_mm((prev),(next),NULL,smp_processor_id())
 
+/*
+ * Find first bit set in a 168-bit bitmap, where the first
+ * 128 bits are unlikely to be set.
+ */
+static inline int sched_find_first_bit(unsigned long *b)
+{
+#if MAX_RT_PRIO != 128 || MAX_PRIO != 168
+#error update this function
+#endif
+
+	if (unlikely(b[0]))
+		return __ffs(b[0]);
+	if (unlikely(b[1]))
+		return __ffs(b[1]) + 32;
+	if (unlikely(b[2]))
+		return __ffs(b[2]) + 64;
+	if (unlikely(b[3]))
+		return __ffs(b[3]) + 96;
+	if (b[4])
+		return __ffs(b[4]) + MAX_RT_PRIO;
+	return __ffs(b[5]) + 32 + MAX_RT_PRIO;
+}
+
 #endif

include/asm-arm/pgalloc.h

@@ -57,40 +57,48 @@ static inline pgd_t *get_pgd_fast(void)
 {
 	unsigned long *ret;
 
+	preempt_disable();
 	if ((ret = pgd_quicklist) != NULL) {
 		pgd_quicklist = (unsigned long *)__pgd_next(ret);
 		ret[1] = ret[2];
 		clean_dcache_entry(ret + 1);
 		pgtable_cache_size--;
 	}
+	preempt_enable();
 	return (pgd_t *)ret;
 }
 
 static inline void free_pgd_fast(pgd_t *pgd)
 {
+	preempt_disable();
 	__pgd_next(pgd) = (unsigned long) pgd_quicklist;
 	pgd_quicklist = (unsigned long *) pgd;
 	pgtable_cache_size++;
+	preempt_enable();
 }
 
 static inline pte_t *pte_alloc_one_fast(struct mm_struct *mm, unsigned long address)
 {
 	unsigned long *ret;
 
+	preempt_disable();
 	if((ret = pte_quicklist) != NULL) {
 		pte_quicklist = (unsigned long *)__pte_next(ret);
 		ret[0] = 0;
 		clean_dcache_entry(ret);
 		pgtable_cache_size--;
 	}
+	preempt_enable();
 	return (pte_t *)ret;
 }
 
 static inline void free_pte_fast(pte_t *pte)
 {
+	preempt_disable();
 	__pte_next(pte) = (unsigned long) pte_quicklist;
 	pte_quicklist = (unsigned long *) pte;
 	pgtable_cache_size++;
+	preempt_enable();
 }
 
 #else	/* CONFIG_NO_PGT_CACHE */

include/asm-arm/proc-armo/processor.h

@@ -23,7 +23,7 @@
 
 #define KERNEL_STACK_SIZE 4096
 
-struct context_save_struct {
+struct cpu_context_save {
 	unsigned long r4;
 	unsigned long r5;
 	unsigned long r6;
@@ -35,7 +35,7 @@ struct context_save_struct {
 	unsigned long pc;
 };
 
-#define INIT_CSS (struct context_save_struct){ 0, 0, 0, 0, 0, 0, 0, 0, SVC26_MODE }
+#define INIT_CSS (struct cpu_context_save){ 0, 0, 0, 0, 0, 0, 0, 0, SVC26_MODE }
 
 typedef struct {
 	void (*put_byte)(void);			/* Special calling convention */
@@ -74,14 +74,4 @@ extern uaccess_t uaccess_user, uaccess_kernel;
 #define KSTK_EIP(tsk)	(((unsigned long *)(4096+(unsigned long)(tsk)))[1022])
 #define KSTK_ESP(tsk)	(((unsigned long *)(4096+(unsigned long)(tsk)))[1020])
 
-/* Allocation and freeing of basic task resources. */
-/*
- * NOTE! The task struct and the stack go together
- */
-extern unsigned long get_page_8k(int priority);
-extern void free_page_8k(unsigned long page);
-
-#define ll_alloc_task_struct()	((struct task_struct *)get_page_8k(GFP_KERNEL))
-#define ll_free_task_struct(p)  free_page_8k((unsigned long)(p))
-
 #endif

include/asm-arm/proc-armv/processor.h

@@ -22,7 +22,7 @@
 
 #define KERNEL_STACK_SIZE	PAGE_SIZE
 
-struct context_save_struct {
+struct cpu_context_save {
 	unsigned long cpsr;
 	unsigned long r4;
 	unsigned long r5;
@@ -35,13 +35,10 @@ struct context_save_struct {
 	unsigned long pc;
 };
 
-#define INIT_CSS (struct context_save_struct){ SVC_MODE, 0, 0, 0, 0, 0, 0, 0, 0, 0 }
+#define INIT_CSS (struct cpu_context_save){ SVC_MODE, 0, 0, 0, 0, 0, 0, 0, 0, 0 }
 
-#define EXTRA_THREAD_STRUCT						\
-	unsigned int	domain;
-
-#define EXTRA_THREAD_STRUCT_INIT					\
-	domain:	  domain_val(DOMAIN_USER, DOMAIN_CLIENT) |		\
+#define INIT_EXTRA_THREAD_INFO						\
+	cpu_domain:	  domain_val(DOMAIN_USER, DOMAIN_CLIENT) |	\
 			  domain_val(DOMAIN_KERNEL, DOMAIN_MANAGER) |	\
 			  domain_val(DOMAIN_IO, DOMAIN_CLIENT)
 
@@ -64,11 +61,4 @@ struct context_save_struct {
 #define KSTK_EIP(tsk)	(((unsigned long *)(4096+(unsigned long)(tsk)))[1021])
 #define KSTK_ESP(tsk)	(((unsigned long *)(4096+(unsigned long)(tsk)))[1019])
 
-/* Allocation and freeing of basic task resources. */
-/*
- * NOTE! The task struct and the stack go together
- */
-#define ll_alloc_task_struct() ((struct task_struct *) __get_free_pages(GFP_KERNEL,1))
-#define ll_free_task_struct(p) free_pages((unsigned long)(p),1)
-
 #endif

include/asm-arm/proc-armv/uaccess.h

@@ -16,8 +16,7 @@
 
 static inline void set_fs (mm_segment_t fs)
 {
-	current->addr_limit = fs;
-
+	current_thread_info()->addr_limit = fs;
 	modify_domain(DOMAIN_KERNEL, fs ? DOMAIN_CLIENT : DOMAIN_MANAGER);
 }
 
@@ -26,7 +25,7 @@ static inline void set_fs (mm_segment_t fs)
 	unsigned long flag, sum; \
 	__asm__("adds %1, %2, %3; sbcccs %1, %1, %0; movcc %0, #0" \
 		: "=&r" (flag), "=&r" (sum) \
-		: "r" (addr), "Ir" (size), "0" (current->addr_limit) \
+		: "r" (addr), "Ir" (size), "0" (current_thread_info()->addr_limit) \
 		: "cc"); \
 	flag; })
 
@@ -34,7 +33,7 @@ static inline void set_fs (mm_segment_t fs)
 	unsigned long flag; \
 	__asm__("cmp %2, %0; movlo %0, #0" \
 		: "=&r" (flag) \
-		: "0" (current->addr_limit), "r" (addr) \
+		: "0" (current_thread_info()->addr_limit), "r" (addr) \
 		: "cc"); \
 	(flag == 0); })
 

include/asm-arm/processor.h

@@ -17,23 +17,6 @@
  */
 #define current_text_addr() ({ __label__ _l; _l: &&_l;})
 
-#define FP_SIZE 35
-
-struct fp_hard_struct {
-	unsigned int save[FP_SIZE];		/* as yet undefined */
-};
-
-struct fp_soft_struct {
-	unsigned int save[FP_SIZE];		/* undefined information */
-};
-
-union fp_state {
-	struct fp_hard_struct	hard;
-	struct fp_soft_struct	soft;
-};
-
-typedef unsigned long mm_segment_t;		/* domain register	*/
-
 #ifdef __KERNEL__
 
 #define EISA_bus 0
@@ -54,37 +37,15 @@ struct debug_info {
 };
 
 struct thread_struct {
-	atomic_t			refcount;
 							/* fault info	  */
 	unsigned long			address;
 	unsigned long			trap_no;
 	unsigned long			error_code;
-							/* floating point */
-	union fp_state			fpstate;
 							/* debugging	  */
 	struct debug_info		debug;
-							/* context info	  */
-	struct context_save_struct	*save;
-	EXTRA_THREAD_STRUCT
 };
 
-#define INIT_THREAD  {					\
-	refcount:	ATOMIC_INIT(1),			\
-	EXTRA_THREAD_STRUCT_INIT			\
-}
-
-/*
- * Return saved PC of a blocked thread.
- */
-static inline unsigned long thread_saved_pc(struct thread_struct *t)
-{
-	return t->save ? pc_pointer(t->save->pc) : 0;
-}
-
-static inline unsigned long get_css_fp(struct thread_struct *t)
-{
-	return t->save ? t->save->fp : 0;
-}
+#define INIT_THREAD  {	}
 
 /* Forward declaration, a strange C thing */
 struct task_struct;
@@ -98,20 +59,6 @@ extern void release_thread(struct task_struct *);
 
 unsigned long get_wchan(struct task_struct *p);
 
-#define THREAD_SIZE	(8192)
-
-extern struct task_struct *alloc_task_struct(void);
-extern void __free_task_struct(struct task_struct *);
-#define get_task_struct(p)	atomic_inc(&(p)->thread.refcount)
-#define free_task_struct(p)					\
- do {								\
-	if (atomic_dec_and_test(&(p)->thread.refcount))		\
-		__free_task_struct((p));			\
- } while (0)
-
-#define init_task	(init_task_union.task)
-#define init_stack	(init_task_union.stack)
-
 #define cpu_relax()			do { } while (0)
 
 /*

include/asm-arm/smplock.h

@@ -3,22 +3,25 @@
  *
  * Default SMP lock implementation
  */
+#include <linux/config.h>
 #include <linux/interrupt.h>
 #include <linux/spinlock.h>
 
 extern spinlock_t kernel_flag;
 
+#ifdef CONFIG_PREEMPT
+#define kernel_locked()		preempt_get_count()
+#else
 #define kernel_locked()		spin_is_locked(&kernel_flag)
+#endif
 
 /*
  * Release global kernel lock and global interrupt lock
  */
 #define release_kernel_lock(task, cpu)		\
 do {						\
-	if (task->lock_depth >= 0) \
+	if (unlikely(task->lock_depth >= 0))	\
 		spin_unlock(&kernel_flag);	\
-	release_irqlock(cpu); \
-	__sti(); \
 } while (0)
 
 /*
@@ -26,7 +29,7 @@ do { \
  */
 #define reacquire_kernel_lock(task)		\
 do {						\
-	if (task->lock_depth >= 0) \
+	if (unlikely(task->lock_depth >= 0))	\
 		spin_lock(&kernel_flag);	\
 } while (0)
 
@@ -40,8 +43,14 @@ do { \
  */
 static inline void lock_kernel(void)
 {
+#ifdef CONFIG_PREEMPT
+	if (current->lock_depth == -1)
+		spin_lock(&kernel_flag);
+	++current->lock_depth;
+#else
 	if (!++current->lock_depth)
 		spin_lock(&kernel_flag);
+#endif
 }
 
 static inline void unlock_kernel(void)

include/asm-arm/softirq.h

@@ -5,20 +5,22 @@
 #include <asm/hardirq.h>
 
 #define __cpu_bh_enable(cpu) \
-		do { barrier(); local_bh_count(cpu)--; } while (0)
+		do { barrier(); local_bh_count(cpu)--; preempt_enable(); } while (0)
 #define cpu_bh_disable(cpu) \
-		do { local_bh_count(cpu)++; barrier(); } while (0)
+		do { preempt_disable(); local_bh_count(cpu)++; barrier(); } while (0)
 
 #define local_bh_disable()	cpu_bh_disable(smp_processor_id())
 #define __local_bh_enable()	__cpu_bh_enable(smp_processor_id())
 
 #define in_softirq()		(local_bh_count(smp_processor_id()) != 0)
 
-#define local_bh_enable()						\
+#define _local_bh_enable()						\
 do {									\
 	unsigned int *ptr = &local_bh_count(smp_processor_id());	\
 	if (!--*ptr && ptr[-2])						\
 		__asm__("bl%? __do_softirq": : : "lr");/* out of line */\
 } while (0)
 
+#define local_bh_enable() do { _local_bh_enable(); preempt_enable(); } while (0)
+
 #endif	/* __ASM_SOFTIRQ_H */

include/asm-arm/system.h

@@ -6,6 +6,8 @@
 #include <linux/config.h>
 #include <linux/kernel.h>
 
+struct thread_info;
+
 /* information about the system we're running on */
 extern unsigned int system_rev;
 extern unsigned int system_serial_low;
@@ -48,11 +50,12 @@ extern asmlinkage void __backtrace(void);
  * `prev' will never be the same as `next'.
  * The `mb' is to tell GCC not to cache `current' across this call.
  */
-extern struct task_struct *__switch_to(struct task_struct *prev, struct task_struct *next);
+struct thread_info;
+extern struct task_struct *__switch_to(struct thread_info *, struct thread_info *);
 
 #define switch_to(prev,next,last)					\
 	do {			 					\
-		last = __switch_to(prev,next);	\
+		last = __switch_to(prev->thread_info,next->thread_info); \
 		mb();							\
 	} while (0)
 

include/asm-arm/thread_info.h

+/*
+ *  linux/include/asm-arm/thread_info.h
+ *
+ *  Copyright (C) 2002 Russell King.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#ifndef __ASM_ARM_THREAD_INFO_H
+#define __ASM_ARM_THREAD_INFO_H
+
+#ifdef __KERNEL__
+
+#ifndef __ASSEMBLY__
+
+struct task_struct;
+struct exec_domain;
+
+#include <asm/fpstate.h>
+#include <asm/proc/processor.h>
+#include <asm/ptrace.h>
+#include <asm/types.h>
+
+typedef unsigned long mm_segment_t;		/* domain register	*/
+
+/*
+ * low level task data that entry.S needs immediate access to.
+ */
+struct thread_info {
+	__u32			flags;		/* low level flags */
+	__s32			preempt_count;	/* 0 => preemptable, <0 => bug */
+	mm_segment_t		addr_limit;	/* address limit */
+	__u32			cpu;		/* cpu */
+	struct cpu_context_save	*cpu_context;	/* cpu context */
+	__u32			cpu_domain;	/* cpu domain */
+	struct task_struct	*task;		/* main task structure */
+	struct exec_domain	*exec_domain;	/* execution domain */
+	union fp_state		fpstate;
+};
+
+#define INIT_THREAD_INFO(tsk)			\
+{						\
+	task:		&tsk,			\
+	exec_domain:	&default_exec_domain,	\
+	flags:		0,			\
+	preempt_count:	0,			\
+	addr_limit:	KERNEL_DS,		\
+	INIT_EXTRA_THREAD_INFO,			\
+}
+
+#define init_thread_info	(init_thread_union.thread_info)
+#define init_stack		(init_thread_union.stack)
+
+/*
+ * how to get the thread information struct from C
+ */
+static inline struct thread_info *current_thread_info(void) __attribute__ (( __const__ ));
+
+static inline struct thread_info *current_thread_info(void)
+{
+	register unsigned long sp asm ("sp");
+	return (struct thread_info *)(sp & ~0x1fff);
+}
+
+#define THREAD_SIZE		(8192)
+
+extern struct thread_info *alloc_thread_info(void);
+extern void free_thread_info(struct thread_info *);
+
+#define get_thread_info(ti)	get_task_struct((ti)->task)
+#define put_thread_info(ti)	put_task_struct((ti)->task)
+
+static inline unsigned long __thread_saved_pc(struct thread_info *thread)
+{
+	struct cpu_context_save *context = thread->cpu_context;
+
+	return context ? pc_pointer(context->pc) : 0;
+}
+
+static inline unsigned long __thread_saved_fp(struct thread_info *thread)
+{
+	struct cpu_context_save *context = thread->cpu_context;
+
+	return context ? context->fp : 0;
+}
+
+#define thread_saved_pc(tsk)	__thread_saved_pc((tsk)->thread_info)
+#define thread_saved_fp(tsk)	__thread_saved_fp((tsk)->thread_info)
+
+#else /* !__ASSEMBLY__ */
+
+#define TI_FLAGS	0
+#define TI_PREEMPT	4
+#define TI_ADDR_LIMIT	8
+#define TI_CPU		12
+#define TI_CPU_SAVE	16
+#define TI_CPU_DOMAIN	20
+#define TI_TASK		24
+#define TI_EXEC_DOMAIN	28
+#define TI_FPSTATE	32
+
+#endif
+
+/*
+ * thread information flags:
+ *  TIF_SYSCALL_TRACE	- syscall trace active
+ *  TIF_NOTIFY_RESUME	- resumption notification requested
+ *  TIF_SIGPENDING	- signal pending
+ *  TIF_NEED_RESCHED	- rescheduling necessary
+ *  TIF_USEDFPU		- FPU was used by this task this quantum (SMP)
+ *  TIF_POLLING_NRFLAG	- true if poll_idle() is polling TIF_NEED_RESCHED
+ */
+#define TIF_NOTIFY_RESUME	0
+#define TIF_SIGPENDING		1
+#define TIF_NEED_RESCHED	2
+#define TIF_SYSCALL_TRACE	8
+#define TIF_USED_FPU		16
+#define TIF_POLLING_NRFLAG	17
+
+#define _TIF_NOTIFY_RESUME	(1 << TIF_NOTIFY_RESUME)
+#define _TIF_SIGPENDING		(1 << TIF_SIGPENDING)
+#define _TIF_NEED_RESCHED	(1 << TIF_NEED_RESCHED)
+#define _TIF_SYSCALL_TRACE	(1 << TIF_SYSCALL_TRACE)
+#define _TIF_USED_FPU		(1 << TIF_USED_FPU)
+#define _TIF_POLLING_NRFLAG	(1 << TIF_POLLING_NRFLAG)
+
+/*
+ * Change these and you break ASM code in entry-common.S
+ */
+#define _TIF_WORK_MASK		0x000000ff
+
+#endif /* __KERNEL__ */
+#endif /* __ASM_ARM_THREAD_INFO_H */

include/asm-arm/uaccess.h

@@ -32,7 +32,7 @@ struct exception_table_entry
 extern unsigned long search_exception_table(unsigned long);
 
 #define get_ds()	(KERNEL_DS)
-#define get_fs()	(current->addr_limit)
+#define get_fs()	(current_thread_info()->addr_limit)
 #define segment_eq(a,b)	((a) == (b))
 
 #include <asm/proc/uaccess.h>