1 files changed, 86 insertions, 63 deletions

diff --git a/arch/arc/kernel/smp.c b/arch/arc/kernel/smp.c
index bca3052c956..c802bb50060 100644
--- a/arch/arc/kernel/smp.c
+++ b/arch/arc/kernel/smp.c
@@ -95,7 +95,7 @@ void __init smp_cpus_done(unsigned int max_cpus)
  *        If it turns out to be elaborate, it's better to code it in assembly
  *
  */
-void __attribute__((weak)) arc_platform_smp_wait_to_boot(int cpu)
+void __weak arc_platform_smp_wait_to_boot(int cpu)
 {
 	/*
 	 * As a hack for debugging - since debugger will single-step over the
@@ -128,6 +128,7 @@ void start_kernel_secondary(void)
 	atomic_inc(&mm->mm_users);
 	atomic_inc(&mm->mm_count);
 	current->active_mm = mm;
+	cpumask_set_cpu(cpu, mm_cpumask(mm));
 
 	notify_cpu_starting(cpu);
 	set_cpu_online(cpu, true);
@@ -137,7 +138,7 @@ void start_kernel_secondary(void)
 	if (machine_desc->init_smp)
 		machine_desc->init_smp(smp_processor_id());
 
-	arc_local_timer_setup(cpu);
+	arc_local_timer_setup();
 
 	local_irq_enable();
 	preempt_disable();
@@ -196,52 +197,65 @@ int __init setup_profiling_timer(unsigned int multiplier)
 /*              Inter Processor Interrupt Handling                           */
 /*****************************************************************************/
 
-/*
- * structures for inter-processor calls
- * A Collection of single bit ipi messages
- *
- */
-
-/*
- * TODO_rajesh investigate tlb message types.
- * IPI Timer not needed because each ARC has an individual Interrupting Timer
- */
 enum ipi_msg_type {
-	IPI_NOP = 0,
+	IPI_EMPTY = 0,
 	IPI_RESCHEDULE = 1,
 	IPI_CALL_FUNC,
-	IPI_CALL_FUNC_SINGLE,
-	IPI_CPU_STOP
+	IPI_CPU_STOP,
 };
 
-struct ipi_data {
-	unsigned long bits;
-};
+/*
+ * In arches with IRQ for each msg type (above), receiver can use IRQ-id  to
+ * figure out what msg was sent. For those which don't (ARC has dedicated IPI
+ * IRQ), the msg-type needs to be conveyed via per-cpu data
+ */
 
-static DEFINE_PER_CPU(struct ipi_data, ipi_data);
+static DEFINE_PER_CPU(unsigned long, ipi_data);
 
-static void ipi_send_msg(const struct cpumask *callmap, enum ipi_msg_type msg)
+static void ipi_send_msg_one(int cpu, enum ipi_msg_type msg)
 {
+	unsigned long __percpu *ipi_data_ptr = per_cpu_ptr(&ipi_data, cpu);
+	unsigned long old, new;
 	unsigned long flags;
-	unsigned int cpu;
+
+	pr_debug("%d Sending msg [%d] to %d\n", smp_processor_id(), msg, cpu);
 
 	local_irq_save(flags);
 
-	for_each_cpu(cpu, callmap) {
-		struct ipi_data *ipi = &per_cpu(ipi_data, cpu);
-		set_bit(msg, &ipi->bits);
-	}
+	/*
+	 * Atomically write new msg bit (in case others are writing too),
+	 * and read back old value
+	 */
+	do {
+		new = old = *ipi_data_ptr;
+		new |= 1U << msg;
+	} while (cmpxchg(ipi_data_ptr, old, new) != old);
 
-	/* Call the platform specific cross-CPU call function  */
-	if (plat_smp_ops.ipi_send)
-		plat_smp_ops.ipi_send((void *)callmap);
+	/*
+	 * Call the platform specific IPI kick function, but avoid if possible:
+	 * Only do so if there's no pending msg from other concurrent sender(s).
+	 * Otherwise, recevier will see this msg as well when it takes the
+	 * IPI corresponding to that msg. This is true, even if it is already in
+	 * IPI handler, because !@old means it has not yet dequeued the msg(s)
+	 * so @new msg can be a free-loader
+	 */
+	if (plat_smp_ops.ipi_send && !old)
+		plat_smp_ops.ipi_send(cpu);
 
 	local_irq_restore(flags);
 }
 
+static void ipi_send_msg(const struct cpumask *callmap, enum ipi_msg_type msg)
+{
+	unsigned int cpu;
+
+	for_each_cpu(cpu, callmap)
+		ipi_send_msg_one(cpu, msg);
+}
+
 void smp_send_reschedule(int cpu)
 {
-	ipi_send_msg(cpumask_of(cpu), IPI_RESCHEDULE);
+	ipi_send_msg_one(cpu, IPI_RESCHEDULE);
 }
 
 void smp_send_stop(void)
@@ -254,7 +268,7 @@ void smp_send_stop(void)
 
 void arch_send_call_function_single_ipi(int cpu)
 {
-	ipi_send_msg(cpumask_of(cpu), IPI_CALL_FUNC_SINGLE);
+	ipi_send_msg_one(cpu, IPI_CALL_FUNC);
 }
 
 void arch_send_call_function_ipi_mask(const struct cpumask *mask)
@@ -265,37 +279,29 @@ void arch_send_call_function_ipi_mask(const struct cpumask *mask)
 /*
  * ipi_cpu_stop - handle IPI from smp_send_stop()
  */
-static void ipi_cpu_stop(unsigned int cpu)
+static void ipi_cpu_stop(void)
 {
 	machine_halt();
 }
 
-static inline void __do_IPI(unsigned long *ops, struct ipi_data *ipi, int cpu)
+static inline void __do_IPI(unsigned long msg)
 {
-	unsigned long msg = 0;
-
-	do {
-		msg = find_next_bit(ops, BITS_PER_LONG, msg+1);
+	switch (msg) {
+	case IPI_RESCHEDULE:
+		scheduler_ipi();
+		break;
 
-		switch (msg) {
-		case IPI_RESCHEDULE:
-			scheduler_ipi();
-			break;
-
-		case IPI_CALL_FUNC:
-			generic_smp_call_function_interrupt();
-			break;
-
-		case IPI_CALL_FUNC_SINGLE:
-			generic_smp_call_function_single_interrupt();
-			break;
+	case IPI_CALL_FUNC:
+		generic_smp_call_function_interrupt();
+		break;
 
-		case IPI_CPU_STOP:
-			ipi_cpu_stop(cpu);
-			break;
-		}
-	} while (msg < BITS_PER_LONG);
+	case IPI_CPU_STOP:
+		ipi_cpu_stop();
+		break;
 
+	default:
+		pr_warn("IPI with unexpected msg %ld\n", msg);
+	}
 }
 
 /*
@@ -304,19 +310,25 @@ static inline void __do_IPI(unsigned long *ops, struct ipi_data *ipi, int cpu)
  */
 irqreturn_t do_IPI(int irq, void *dev_id)
 {
-	int cpu = smp_processor_id();
-	struct ipi_data *ipi = &per_cpu(ipi_data, cpu);
-	unsigned long ops;
+	unsigned long pending;
+
+	pr_debug("IPI [%ld] received on cpu %d\n",
+		 *this_cpu_ptr(&ipi_data), smp_processor_id());
 
 	if (plat_smp_ops.ipi_clear)
-		plat_smp_ops.ipi_clear(cpu, irq);
+		plat_smp_ops.ipi_clear(irq);
 
 	/*
-	 * XXX: is this loop really needed
-	 * And do we need to move ipi_clean inside
+	 * "dequeue" the msg corresponding to this IPI (and possibly other
+	 * piggybacked msg from elided IPIs: see ipi_send_msg_one() above)
 	 */
-	while ((ops = xchg(&ipi->bits, 0)) != 0)
-		__do_IPI(&ops, ipi, cpu);
+	pending = xchg(this_cpu_ptr(&ipi_data), 0);
+
+	do {
+		unsigned long msg = __ffs(pending);
+		__do_IPI(msg);
+		pending &= ~(1U << msg);
+	} while (pending);
 
 	return IRQ_HANDLED;
 }
@@ -325,8 +337,19 @@ irqreturn_t do_IPI(int irq, void *dev_id)
  * API called by platform code to hookup arch-common ISR to their IPI IRQ
  */
 static DEFINE_PER_CPU(int, ipi_dev);
+
+static struct irqaction arc_ipi_irq = {
+        .name    = "IPI Interrupt",
+        .flags   = IRQF_PERCPU,
+        .handler = do_IPI,
+};
+
 int smp_ipi_irq_setup(int cpu, int irq)
 {
-	int *dev_id = &per_cpu(ipi_dev, smp_processor_id());
-	return request_percpu_irq(irq, do_IPI, "IPI Interrupt", dev_id);
+	if (!cpu)
+		return setup_irq(irq, &arc_ipi_irq);
+	else
+		arch_unmask_irq(irq);
+
+	return 0;
 }