O(1) scheduler K3+ for IA64

diff -urN 2.4.17-ia64-kdbv2.1-K3/arch/i386/kernel/i8259.c 2.4.17-ia64-kdbv2.1-k3y_al2/arch/i386/kernel/i8259.c
@@ -79,7 +79,6 @@
  * through the ICC by us (IPIs)
  */
 #ifdef CONFIG_SMP
-BUILD_SMP_INTERRUPT(task_migration_interrupt,TASK_MIGRATION_VECTOR)
 BUILD_SMP_INTERRUPT(reschedule_interrupt,RESCHEDULE_VECTOR)
 BUILD_SMP_INTERRUPT(invalidate_interrupt,INVALIDATE_TLB_VECTOR)
 BUILD_SMP_INTERRUPT(call_function_interrupt,CALL_FUNCTION_VECTOR)

@@ -474,9 +473,6 @@
 	 */
 	set_intr_gate(RESCHEDULE_VECTOR, reschedule_interrupt);
 
-	/* IPI for task migration */
-	set_intr_gate(TASK_MIGRATION_VECTOR, task_migration_interrupt);
-
 	/* IPI for invalidation */
 	set_intr_gate(INVALIDATE_TLB_VECTOR, invalidate_interrupt);
 

diff -urN 2.4.17-ia64-kdbv2.1-K3/arch/i386/kernel/smp.c 2.4.17-ia64-kdbv2.1-k3y_al2/arch/i386/kernel/smp.c
@@ -485,35 +485,6 @@
 	do_flush_tlb_all_local();
 }
 
-static spinlock_t migration_lock = SPIN_LOCK_UNLOCKED;
-static task_t *new_task;
-
-/*
- * This function sends a 'task migration' IPI to another CPU.
- * Must be called from syscall contexts, with interrupts *enabled*.
- */
-void smp_migrate_task(int cpu, task_t *p)
-{
-	/*
-	 * The target CPU will unlock the migration spinlock:
-	 */
-	spin_lock(&migration_lock);
-	new_task = p;
-	send_IPI_mask(1 << cpu, TASK_MIGRATION_VECTOR);
-}
-
-/*
- * Task migration callback.
- */
-asmlinkage void smp_task_migration_interrupt(void)
-{
-	task_t *p;
-
-	ack_APIC_irq();
-	p = new_task;
-	spin_unlock(&migration_lock);
-	sched_task_migrated(p);
-}
 /*
  * this function sends a 'reschedule' IPI to another CPU.
  * it goes straight through and wastes no time serializing

diff -urN 2.4.17-ia64-kdbv2.1-K3/arch/ia64/kdb/kdba_bt.c 2.4.17-ia64-kdbv2.1-k3y_al2/arch/ia64/kdb/kdba_bt.c
@@ -197,7 +197,7 @@
 	}
 #ifdef CONFIG_SMP
 	else if (task_has_cpu(p)) {
-		sw = kdb_sw[p->processor];
+		sw = kdb_sw[p->cpu];
 	}
 #endif
 	else {

diff -urN 2.4.17-ia64-kdbv2.1-K3/arch/ia64/kernel/irq_ia64.c 2.4.17-ia64-kdbv2.1-k3y_al2/arch/ia64/kernel/irq_ia64.c
@@ -148,6 +148,7 @@
 	flags:		SA_INTERRUPT,
 	name:		"IPI"
 };
+
 #endif
 
 void

diff -urN 2.4.17-ia64-kdbv2.1-K3/arch/ia64/kernel/process.c 2.4.17-ia64-kdbv2.1-k3y_al2/arch/ia64/kernel/process.c
@@ -125,9 +125,6 @@
 cpu_idle (void *unused)
 {
 	/* endless idle loop with no priority at all */
-	init_idle();
-	current->nice = 20;
-	current->counter = -100;
 
 
 	while (1) {

@@ -136,11 +133,10 @@
 			min_xtp();
 #endif
 
-		while (!current->need_resched) {
+		if (!current->need_resched) {
 #ifdef CONFIG_IA64_SGI_SN
 			snidle();
 #endif
-			continue;
 		}
 
 #ifdef CONFIG_IA64_SGI_SN

@@ -258,7 +254,7 @@
 
 	if (user_mode(child_ptregs)) {
 		if (user_stack_base) {
-			child_ptregs->r12 = user_stack_base + user_stack_size;
+			child_ptregs->r12 = user_stack_base + user_stack_size - 16;
 			child_ptregs->ar_bspstore = user_stack_base;
 			child_ptregs->ar_rnat = 0;
 			child_ptregs->loadrs = 0;

diff -urN 2.4.17-ia64-kdbv2.1-K3/arch/ia64/kernel/setup.c 2.4.17-ia64-kdbv2.1-k3y_al2/arch/ia64/kernel/setup.c
@@ -375,10 +375,10 @@
 {
 #ifdef CONFIG_SMP
 #	define lpj	c->loops_per_jiffy
-#	define cpu	c->processor
+#	define cpum	c->processor
 #else
 #	define lpj	loops_per_jiffy
-#	define cpu	0
+#	define cpum	0
 #endif
 	char family[32], features[128], *cp;
 	struct cpuinfo_ia64 *c = v;

@@ -417,7 +417,7 @@
 		   "cpu MHz    : %lu.%06lu\n"
 		   "itc MHz    : %lu.%06lu\n"
 		   "BogoMIPS   : %lu.%02lu\n\n",
-		   cpu, c->vendor, family, c->model, c->revision, c->archrev,
+		   cpum, c->vendor, family, c->model, c->revision, c->archrev,
 		   features, c->ppn, c->number,
 		   c->proc_freq / 1000000, c->proc_freq % 1000000,
 		   c->itc_freq / 1000000, c->itc_freq % 1000000,

diff -urN 2.4.17-ia64-kdbv2.1-K3/arch/ia64/kernel/smp.c 2.4.17-ia64-kdbv2.1-k3y_al2/arch/ia64/kernel/smp.c
@@ -200,6 +200,12 @@
 }
 
 void
+smp_send_reschedule_all(void)
+{
+	send_IPI_all(IA64_IPI_RESCHEDULE);
+}
+
+void
 smp_flush_tlb_all (void)
 {
 	smp_call_function ((void (*)(void *))__flush_tlb_all,0,1,1);

diff -urN 2.4.17-ia64-kdbv2.1-K3/arch/ia64/kernel/smpboot.c 2.4.17-ia64-kdbv2.1-k3y_al2/arch/ia64/kernel/smpboot.c
@@ -356,6 +356,7 @@
 	local_irq_enable();
 	calibrate_delay();
 	local_cpu_data->loops_per_jiffy = loops_per_jiffy;
+	ia64_disable_timer();
 	/*
 	 * Allow the master to continue.
 	 */

@@ -379,7 +380,8 @@
 	Dprintk("CPU %d is set to go.\n", smp_processor_id());
 	while (!atomic_read(&smp_commenced))
 		;
-
+	/* reenable timer interrupts */
+	ia64_cpu_local_tick();
 	Dprintk("CPU %d is starting idle.\n", smp_processor_id());
 	return cpu_idle();
 }

@@ -416,11 +418,10 @@
 	if (!idle)
 		panic("No idle process for CPU %d", cpu);
 
-	task_set_cpu(idle, cpu);	/* we schedule the first task manually */
+	init_idle(idle, cpu);
 
 	ia64_cpu_to_sapicid[cpu] = sapicid;
 
-	del_from_runqueue(idle);
 	unhash_process(idle);
 	init_tasks[cpu] = idle;
 

@@ -481,8 +482,7 @@
 	printk("Boot processor id 0x%x/0x%x\n", 0, boot_cpu_id);
 
 	global_irq_holder = 0;
-	current->processor = 0;
-	init_idle();
+	current->cpu = 0;
 
 	/*
 	 * If SMP should be disabled, then really disable it!

@@ -569,3 +569,9 @@
 		smp_num_cpus = 1;
 	}
 }
+
+/* Number of ticks we consider an idle tasks still cache-hot.
+ * For Itanium: with 1GB/s bandwidth we need 4ms to fill up 4MB L3 cache...
+ * So let's try 10 ticks.
+ */
+unsigned long cache_decay_ticks;

diff -urN 2.4.17-ia64-kdbv2.1-K3/arch/ia64/kernel/time.c 2.4.17-ia64-kdbv2.1-k3y_al2/arch/ia64/kernel/time.c
@@ -209,7 +209,7 @@
 /*
  * Encapsulate access to the itm structure for SMP.
  */
-void __init
+void
 ia64_cpu_local_tick (void)
 {
 	int cpu = smp_processor_id();

@@ -298,3 +298,9 @@
 	efi_gettimeofday((struct timeval *) &xtime);
 	ia64_init_itm();
 }
+
+void __init ia64_disable_timer(void)
+{
+	ia64_set_itv(IA64_TIMER_VECTOR | IA64_TIMER_MASK);
+}
+

diff -urN 2.4.17-ia64-kdbv2.1-K3/arch/ia64/mm/fault.c 2.4.17-ia64-kdbv2.1-k3y_al2/arch/ia64/mm/fault.c
@@ -194,8 +194,7 @@
   out_of_memory:
 	up_read(&mm->mmap_sem);
 	if (current->pid = 1) {
-		current->policy |= SCHED_YIELD;
-		schedule();
+		yield();
 		down_read(&mm->mmap_sem);
 		goto survive;
 	}

diff -urN 2.4.17-ia64-kdbv2.1-K3/arch/ia64/mm/tlb.c 2.4.17-ia64-kdbv2.1-k3y_al2/arch/ia64/mm/tlb.c
@@ -48,6 +48,7 @@
 {
 	unsigned long tsk_context, max_ctx = ia64_ctx.max_ctx;
 	struct task_struct *tsk;
+	int i;
 
 	if (ia64_ctx.next > max_ctx)
 		ia64_ctx.next = 300;	/* skip daemons */

@@ -76,7 +77,11 @@
 			ia64_ctx.limit = tsk_context;
 	}
 	read_unlock(&tasklist_lock);
-	flush_tlb_all();
+	//flush_tlb_all(); /* potential race condition with O(1) scheduler [EF] */
+	for (i=0; i<smp_num_cpus; i++)
+		cpu_data(i)->tlb_flush = 1;
+	__flush_tlb_all();
+	local_cpu_data->tlb_flush = 0;
 }
 
 static inline void

diff -urN 2.4.17-ia64-kdbv2.1-K3/arch/ia64/tools/print_offsets.c 2.4.17-ia64-kdbv2.1-k3y_al2/arch/ia64/tools/print_offsets.c
@@ -54,7 +54,7 @@
     { "IA64_TASK_PTRACE_OFFSET",	offsetof (struct task_struct, ptrace) },
     { "IA64_TASK_SIGPENDING_OFFSET",	offsetof (struct task_struct, sigpending) },
     { "IA64_TASK_NEED_RESCHED_OFFSET",	offsetof (struct task_struct, need_resched) },
-    { "IA64_TASK_PROCESSOR_OFFSET",	offsetof (struct task_struct, processor) },
+    { "IA64_TASK_PROCESSOR_OFFSET",	offsetof (struct task_struct, cpu) },
     { "IA64_TASK_THREAD_OFFSET",	offsetof (struct task_struct, thread) },
     { "IA64_TASK_THREAD_KSP_OFFSET",	offsetof (struct task_struct, thread.ksp) },
 #ifdef CONFIG_PERFMON

diff -urN 2.4.17-ia64-kdbv2.1-K3/fs/pipe.c 2.4.17-ia64-kdbv2.1-k3y_al2/fs/pipe.c
@@ -115,7 +115,7 @@
 		 * writers synchronously that there is more
 		 * room.
 		 */
-		wake_up_interruptible_sync(PIPE_WAIT(*inode));
+		wake_up_interruptible(PIPE_WAIT(*inode));
 		if (!PIPE_EMPTY(*inode))
 			BUG();
 		goto do_more_read;

@@ -213,7 +213,7 @@
 			 * is going to give up this CPU, so it doesnt have
 			 * to do idle reschedules.
 			 */
-			wake_up_interruptible_sync(PIPE_WAIT(*inode));
+			wake_up_interruptible(PIPE_WAIT(*inode));
 			PIPE_WAITING_WRITERS(*inode)++;
 			pipe_wait(inode);
 			PIPE_WAITING_WRITERS(*inode)--;

diff -urN 2.4.17-ia64-kdbv2.1-K3/include/asm-i386/hw_irq.h 2.4.17-ia64-kdbv2.1-k3y_al2/include/asm-i386/hw_irq.h
@@ -41,8 +41,7 @@
 #define ERROR_APIC_VECTOR	0xfe
 #define INVALIDATE_TLB_VECTOR	0xfd
 #define RESCHEDULE_VECTOR	0xfc
-#define TASK_MIGRATION_VECTOR	0xfb
-#define CALL_FUNCTION_VECTOR	0xfa
+#define CALL_FUNCTION_VECTOR	0xfb
 
 /*
  * Local APIC timer IRQ vector is on a different priority level,

80 lines not on this page · raw message

diff -urN 2.4.17-ia64-kdbv2.1-K3/include/asm-ia64/hw_irq.h 2.4.17-ia64-kdbv2.1-k3y_al2/include/asm-ia64/hw_irq.h
@@ -65,6 +65,9 @@
         IA64_IPI_DM_EXTINT =    0x7,    /* pend an 8259-compatible interrupt. */
 };
 
+/* bit for masking and discarding timer interrupts on IA64 */
+#define IA64_TIMER_MASK         (1<<16)
+
 extern __u8 isa_irq_to_vector_map[16];
 #define isa_irq_to_vector(x)	isa_irq_to_vector_map[(x)]
 

diff -urN 2.4.17-ia64-kdbv2.1-K3/include/asm-ia64/mmu_context.h 2.4.17-ia64-kdbv2.1-k3y_al2/include/asm-ia64/mmu_context.h
@@ -44,16 +44,34 @@
 {
 }
 
+/* 
+ * When the context counter wraps around all TLBs need to be flushed because
+ * an old context number might have been reused. This is signalled by a bit
+ * set in ia64_ctx.flush, which is checked in the routine below. Called by
+ * activate_mm().                                        <[email protected]>
+ */
+static inline void
+delayed_tlb_flush (void)
+{
+	extern void __flush_tlb_all (void);
+
+	if (unlikely(local_cpu_data->tlb_flush)) {
+		__flush_tlb_all();
+		local_cpu_data->tlb_flush = 0;
+	}
+}
+
 static inline void
 get_new_mmu_context (struct mm_struct *mm)
 {
-	spin_lock(&ia64_ctx.lock);
+	unsigned long flags;
+	spin_lock_irqsave(&ia64_ctx.lock,flags);
 	{
 		if (ia64_ctx.next >= ia64_ctx.limit)
 			wrap_mmu_context(mm);
 		mm->context = ia64_ctx.next++;
 	}
-	spin_unlock(&ia64_ctx.lock);
+	spin_unlock_irqrestore(&ia64_ctx.lock,flags);
 
 }
 

@@ -113,11 +131,28 @@
 	 * We may get interrupts here, but that's OK because interrupt
 	 * handlers cannot touch user-space.
 	 */
+	delayed_tlb_flush();
 	ia64_set_kr(IA64_KR_PT_BASE, __pa(next->pgd));
 	get_mmu_context(next);
 	reload_context(next);
 }
 
+/*
+ * Needed for the O(1) MQ scheduler.
+ */
+#if MAX_PRIO >= 192
+# error update this function. */
+#endif
+
+static inline int sched_find_first_bit(unsigned long *b)
+{
+	if (unlikely(b[0]))
+		return __ffs(b[0]);
+	if (b[1])
+		return 64 + __ffs(b[1]);
+	return __ffs(b[2]) + 128;
+}
+
 #define switch_mm(prev_mm,next_mm,next_task,cpu)	activate_mm(prev_mm, next_mm)
 
 # endif /* ! __ASSEMBLY__ */

diff -urN 2.4.17-ia64-kdbv2.1-K3/include/asm-ia64/pgalloc.h 2.4.17-ia64-kdbv2.1-k3y_al2/include/asm-ia64/pgalloc.h
@@ -160,9 +160,12 @@
 
 #ifdef CONFIG_SMP
   extern void smp_flush_tlb_all (void);
+  extern void smp_flush_tlb_all_nowait (void);
 # define flush_tlb_all()	smp_flush_tlb_all()
+# define flush_tlb_all_nowait()	smp_flush_tlb_all_nowait()
 #else
 # define flush_tlb_all()	__flush_tlb_all()
+# define flush_tlb_all_nowait()	__flush_tlb_all()
 #endif
 
 /*

diff -urN 2.4.17-ia64-kdbv2.1-K3/include/asm-ia64/processor.h 2.4.17-ia64-kdbv2.1-k3y_al2/include/asm-ia64/processor.h
@@ -258,6 +258,7 @@
 	/* CPUID-derived information: */
 	__u64 ppn;
 	__u64 features;
+	__u8 tlb_flush;		/* flush TLB before next context switch if non-zero */
 	__u8 number;
 	__u8 revision;
 	__u8 model;

diff -urN 2.4.17-ia64-kdbv2.1-K3/include/asm-ia64/smp.h 2.4.17-ia64-kdbv2.1-k3y_al2/include/asm-ia64/smp.h
@@ -27,7 +27,7 @@
 #define SMP_IRQ_REDIRECTION	(1 << 0)
 #define SMP_IPI_REDIRECTION	(1 << 1)
 
-#define smp_processor_id()	(current->processor)
+#define smp_processor_id()	(current->cpu)
 
 extern struct smp_boot_data {
 	int cpu_count;

@@ -48,6 +48,9 @@
 
 extern unsigned long ap_wakeup_vector;
 
+extern void smp_send_reschedule(int cpu);
+extern void smp_send_reschedule_all(void);
+
 /*
  * Function to map hard smp processor id to logical id.  Slow, so
  * don't use this in performance-critical code.

@@ -110,12 +113,6 @@
 
 #define NO_PROC_ID		0xffffffff	/* no processor magic marker */
 
-/*
- * Extra overhead to move a task from one cpu to another (due to TLB and cache misses).
- * Expressed in "negative nice value" units (larger number means higher priority/penalty).
- */
-#define PROC_CHANGE_PENALTY	20
-
 extern void __init init_smp_config (void);
 extern void smp_do_timer (struct pt_regs *regs);
 

diff -urN 2.4.17-ia64-kdbv2.1-K3/include/asm-ia64/spinlock.h 2.4.17-ia64-kdbv2.1-k3y_al2/include/asm-ia64/spinlock.h
@@ -84,7 +84,7 @@
 	"mov r29 = 1\n"						\
 	";;\n"							\
 	"1:\n"							\
-	"ld4.bias r2 = [%0]\n"					\
+	"ld4 r2 = [%0]\n"					\
 	";;\n"							\
 	"cmp4.eq p0,p7 = r0,r2\n"				\
 	"(p7) br.cond.spnt.few 1b \n"				\

diff -urN 2.4.17-ia64-kdbv2.1-K3/include/linux/sched.h 2.4.17-ia64-kdbv2.1-k3y_al2/include/linux/sched.h
@@ -149,8 +149,7 @@
 extern void update_one_process(task_t *p, unsigned long user,
 			       unsigned long system, int cpu);
 extern void scheduler_tick(int user_tick, int system);
-extern void sched_task_migrated(task_t *p);
-extern void smp_migrate_task(int cpu, task_t *task);
+extern void migration_init(void);
 extern unsigned long cache_decay_ticks;
 
 #define	MAX_SCHEDULE_TIMEOUT	LONG_MAX

@@ -350,6 +349,10 @@
 
 	wait_queue_head_t wait_chldexit;	/* for wait4() */
 	struct completion *vfork_done;		/* for vfork() */
+
+	list_t migration_list;
+	struct semaphore migration_sem;
+
 	unsigned long rt_priority;
 	unsigned long it_real_value, it_prof_value, it_virt_value;
 	unsigned long it_real_incr, it_prof_incr, it_virt_incr;

@@ -444,7 +447,12 @@
  */
 #define _STK_LIM	(8*1024*1024)
 
+#ifdef CONFIG_SMP
 extern void set_cpus_allowed(task_t *p, unsigned long new_mask);
+#else
+# define set_cpus_allowed(p, new_mask) do { } while (0)
+#endif
+
 extern void set_user_nice(task_t *p, long nice);
 extern int task_prio(task_t *p);
 extern int task_nice(task_t *p);

@@ -476,6 +484,8 @@
     mm:			NULL,						\
     active_mm:		&init_mm,					\
     run_list:		LIST_HEAD_INIT(tsk.run_list),			\
+    migration_list:	LIST_HEAD_INIT(tsk.migration_list),		\
+    migration_sem:	__MUTEX_INITIALIZER(tsk.migration_sem),		\
     time_slice:		HZ,						\
     next_task:		&tsk,						\
     prev_task:		&tsk,						\

@@ -571,7 +581,6 @@
 #define CURRENT_TIME (xtime.tv_sec)
 
 extern void FASTCALL(__wake_up(wait_queue_head_t *q, unsigned int mode, int nr));
-extern void FASTCALL(__wake_up_sync(wait_queue_head_t *q, unsigned int mode, int nr));
 extern void FASTCALL(sleep_on(wait_queue_head_t *q));
 extern long FASTCALL(sleep_on_timeout(wait_queue_head_t *q,
 				      signed long timeout));

@@ -585,13 +594,9 @@
 #define wake_up(x)			__wake_up((x),TASK_UNINTERRUPTIBLE | TASK_INTERRUPTIBLE, 1)
 #define wake_up_nr(x, nr)		__wake_up((x),TASK_UNINTERRUPTIBLE | TASK_INTERRUPTIBLE, nr)
 #define wake_up_all(x)			__wake_up((x),TASK_UNINTERRUPTIBLE | TASK_INTERRUPTIBLE, 0)
-#define wake_up_sync(x)			__wake_up_sync((x),TASK_UNINTERRUPTIBLE | TASK_INTERRUPTIBLE, 1)
-#define wake_up_sync_nr(x, nr)		__wake_up_sync((x),TASK_UNINTERRUPTIBLE | TASK_INTERRUPTIBLE, nr)
 #define wake_up_interruptible(x)	__wake_up((x),TASK_INTERRUPTIBLE, 1)
 #define wake_up_interruptible_nr(x, nr)	__wake_up((x),TASK_INTERRUPTIBLE, nr)
 #define wake_up_interruptible_all(x)	__wake_up((x),TASK_INTERRUPTIBLE, 0)
-#define wake_up_interruptible_sync(x)	__wake_up_sync((x),TASK_INTERRUPTIBLE, 1)
-#define wake_up_interruptible_sync_nr(x) __wake_up_sync((x),TASK_INTERRUPTIBLE,  nr)
 asmlinkage long sys_wait4(pid_t pid,unsigned int * stat_addr, int options, struct rusage * ru);
 
 extern int in_group_p(gid_t);

diff -urN 2.4.17-ia64-kdbv2.1-K3/include/linux/smp.h 2.4.17-ia64-kdbv2.1-k3y_al2/include/linux/smp.h
@@ -24,12 +24,6 @@
 extern void smp_send_stop(void);
 
 /*
- * sends a 'reschedule' event to another CPU:
- */
-extern void FASTCALL(smp_send_reschedule(int cpu));
-
-
-/*
  * Boot processor call to load the other CPU's
  */
 extern void smp_boot_cpus(void);

diff -urN 2.4.17-ia64-kdbv2.1-K3/init/main.c 2.4.17-ia64-kdbv2.1-k3y_al2/init/main.c
@@ -698,7 +698,12 @@
  */
 static void __init do_basic_setup(void)
 {
-
+	/*
+	 * Let the per-CPU migration threads start up:
+	 */
+#if CONFIG_SMP
+	migration_init();
+#endif
 	/*
 	 * Tell the world that we're going to be the grim
 	 * reaper of innocent orphaned children.

diff -urN 2.4.17-ia64-kdbv2.1-K3/kdb/kdbmain.c 2.4.17-ia64-kdbv2.1-k3y_al2/kdb/kdbmain.c
@@ -2344,7 +2344,7 @@
 	for_each_task(p) {
 		kdb_printf("0x%p %08d %08d  %1.1d  %3.3d  %s  0x%p%c%s\n",
 			   (void *)p, p->pid, p->p_pptr->pid,
-			   task_has_cpu(p), p->processor,
+			   task_has_cpu(p), p->cpu,
 			   (p->state = 0)?"run ":(p->state>0)?"stop":"unrn",
 			   (void *)(&p->thread),
 			   (p = current) ? '*': ' ',

diff -urN 2.4.17-ia64-kdbv2.1-K3/kernel/fork.c 2.4.17-ia64-kdbv2.1-k3y_al2/kernel/fork.c
@@ -640,6 +640,10 @@
 	{
 		int i;
 
+		if (likely(p->cpus_allowed & (1UL<<smp_processor_id())))
+			p->cpu = smp_processor_id();
+		else
+			p->cpu = __ffs(p->cpus_allowed);
 		/* ?? should we just memset this ?? */
 		for(i = 0; i < smp_num_cpus; i++)
 			p->per_cpu_utime[cpu_logical_map(i)] diff -urN 2.4.17-ia64-kdbv2.1-K3/kernel/ksyms.c 2.4.17-ia64-kdbv2.1-k3y_al2/kernel/ksyms.c

@@ -441,7 +441,6 @@
 /* process management */
 EXPORT_SYMBOL(complete_and_exit);
 EXPORT_SYMBOL(__wake_up);
-EXPORT_SYMBOL(__wake_up_sync);
 EXPORT_SYMBOL(wake_up_process);
 EXPORT_SYMBOL(sleep_on);
 EXPORT_SYMBOL(sleep_on_timeout);

@@ -451,7 +450,9 @@
 EXPORT_SYMBOL(schedule_timeout);
 EXPORT_SYMBOL(sys_sched_yield);
 EXPORT_SYMBOL(set_user_nice);
+#ifdef CONFIG_SMP
 EXPORT_SYMBOL(set_cpus_allowed);
+#endif
 EXPORT_SYMBOL(jiffies);
 EXPORT_SYMBOL(xtime);
 EXPORT_SYMBOL(do_gettimeofday);

diff -urN 2.4.17-ia64-kdbv2.1-K3/kernel/printk.c 2.4.17-ia64-kdbv2.1-k3y_al2/kernel/printk.c
@@ -25,6 +25,8 @@
 #include <linux/init.h>
 #include <linux/module.h>
 #include <linux/interrupt.h>			/* For in_interrupt() */
+#include <linux/config.h>
+#include <linux/delay.h>
 
 #include <asm/uaccess.h>
 

577 lines not on this page · raw message

diff -urN 2.4.17-ia64-kdbv2.1-K3/kernel/timer.c 2.4.17-ia64-kdbv2.1-k3y_al2/kernel/timer.c
@@ -584,18 +584,7 @@
 	int cpu = smp_processor_id(), system = user_tick ^ 1;
 
 	update_one_process(p, user_tick, system, cpu);
-	if (p->pid) {
-		if (--p->counter <= 0) {
-			p->counter = 0;
-			p->need_resched = 1;
-		}
-		if (p->nice > 0)
-			kstat.per_cpu_nice[cpu] += user_tick;
-		else
-			kstat.per_cpu_user[cpu] += user_tick;
-		kstat.per_cpu_system[cpu] += system;
-	} else if (really_local_bh_count() || really_local_irq_count() > 1)
-		kstat.per_cpu_system[cpu] += system;
+	scheduler_tick(user_tick, system);
 }
 
 /*

@@ -1539,7 +1539,8 @@
 
 	for (;;) {
 		if (test_and_clear_bit(smp_processor_id(), &migration_mask))
-			current->cpus_allowed = 1 << smp_processor_id();
+                        printk("migration_task on cpu=%d mask=%lx\n",
+                               cpu(),current->cpus_allowed);
 		if (current->need_resched)
 			schedule();
 		if (!migration_mask)

@@ -1509,10 +1509,10 @@
 	down(&req.sem);
 }
 
-static volatile unsigned long migration_mask;
-
 static int migration_thread(void * unused)
 {
+	int bind_cpu = (int) (long) unused;
+	int cpu = cpu_logical_map(bind_cpu);
 	struct sched_param param = { sched_priority: 99 };
 	runqueue_t *rq;
 	int ret;

@@ -1520,31 +1520,19 @@
 	daemonize();
 	sigfillset(&current->blocked);
 	set_fs(KERNEL_DS);
-	ret = setscheduler(0, SCHED_FIFO, &param);
 
 	/*
-	 * We have to migrate manually - there is no migration thread
-	 * to do this for us yet :-)
-	 *
-	 * We use the following property of the Linux scheduler. At
-	 * this point no other task is running, so by keeping all
-	 * migration threads running, the load-balancer will distribute
-	 * them between all CPUs equally. At that point every migration
-	 * task binds itself to the current CPU.
+	 * The first migration task is started on CPU #0. This one can migrate
+	 * the tasks to their destination CPUs.
 	 */
-
-	/* wait for all migration threads to start up. */
-	while (!migration_mask)
-		yield();
-
-	for (;;) {
-		if (test_and_clear_bit(smp_processor_id(), &migration_mask))
-			current->cpus_allowed = 1 << smp_processor_id();
-		if (current->need_resched)
-			schedule();
-		if (!migration_mask)
-			break;
+	if (cpu != 0) {
+		while (!cpu_rq(cpu_logical_map(0))->migration_thread)
+			yield();
+		set_cpus_allowed(current, 1UL << cpu);
 	}
+	printk("migration_task %d on cpu=%d\n",cpu,smp_processor_id());
+	ret = setscheduler(0, SCHED_FIFO, &param);
+
 	rq = this_rq();
 	rq->migration_thread = current;
 

@@ -1602,21 +1590,16 @@
 {
 	int cpu;
 
+	current->cpus_allowed = 1UL << cpu_logical_map(0);
 	for (cpu = 0; cpu < smp_num_cpus; cpu++) {
-		current->cpus_allowed = 1UL << cpu_logical_map(cpu);
-		if (kernel_thread(migration_thread, NULL,
+		if (kernel_thread(migration_thread, (void *) (long) cpu,
 				CLONE_FS | CLONE_FILES | CLONE_SIGNAL) < 0)
 			BUG();
-		else
-			current->cpus_allowed = -1L;
 	}
-
-	migration_mask = (1 << smp_num_cpus) - 1;
+	current->cpus_allowed = -1L;
 
 	for (cpu = 0; cpu < smp_num_cpus; cpu++)
 		while (!cpu_rq(cpu)->migration_thread)
 			schedule_timeout(2);
-	if (migration_mask)
-		BUG();
 }
 #endif