1 files changed, 26 insertions, 24 deletions

diff --git a/arch/sparc64/kernel/smp.c b/arch/sparc64/kernel/smp.c
index 7dc28a484268..90eaca3ec9a6 100644
--- a/arch/sparc64/kernel/smp.c
+++ b/arch/sparc64/kernel/smp.c
@@ -745,12 +745,21 @@ struct call_data_struct {
         int wait;
 };
 
-static DEFINE_SPINLOCK(call_lock);
+static __cacheline_aligned_in_smp DEFINE_SPINLOCK(call_lock);
 static struct call_data_struct *call_data;
 
 extern unsigned long xcall_call_function;
 
-/*
+/**
+ * smp_call_function(): Run a function on all other CPUs.
+ * @func: The function to run. This must be fast and non-blocking.
+ * @info: An arbitrary pointer to pass to the function.
+ * @nonatomic: currently unused.
+ * @wait: If true, wait (atomically) until function has completed on other CPUs.
+ *
+ * Returns 0 on success, else a negative status code. Does not return until
+ * remote CPUs are nearly ready to execute <<func>> or are or have executed.
+ *
  * You must not call this function with disabled interrupts or from a
  * hardware interrupt handler or from a bottom half handler.
  */
@@ -759,7 +768,6 @@ static int smp_call_function_mask(void (*func)(void *info), void *info,
 {
         struct call_data_struct data;
         int cpus;
-        long timeout;
 
         /* Can deadlock when called with interrupts disabled */
         WARN_ON(irqs_disabled());
@@ -777,31 +785,18 @@ static int smp_call_function_mask(void (*func)(void *info), void *info,
                 goto out_unlock;
 
         call_data = &data;
+        mb();
 
         smp_cross_call_masked(&xcall_call_function, 0, 0, 0, mask);
 
-        /* 
-         * Wait for other cpus to complete function or at
-         * least snap the call data.
-         */
-        timeout = 1000000;
-        while (atomic_read(&data.finished) != cpus) {
-                if (--timeout <= 0)
-                        goto out_timeout;
-                barrier();
-                udelay(1);
-        }
+        /* Wait for response */
+        while (atomic_read(&data.finished) != cpus)
+                cpu_relax();
 
 out_unlock:
         spin_unlock(&call_lock);
 
         return 0;
-
-out_timeout:
-        spin_unlock(&call_lock);
-        printk("XCALL: Remote cpus not responding, ncpus=%d finished=%d\n",
-               cpus, atomic_read(&data.finished));
-        return 0;
 }
 
 int smp_call_function(void (*func)(void *info), void *info,
@@ -830,9 +825,16 @@ void smp_call_function_client(int irq, struct pt_regs *regs)
 
 static void tsb_sync(void *info)
 {
+        struct trap_per_cpu *tp = &trap_block[raw_smp_processor_id()];
         struct mm_struct *mm = info;
 
-        if (current->active_mm == mm)
+        /* It is not valid to test "currrent->active_mm == mm" here.
+         *
+         * The value of "current" is not changed atomically with
+         * switch_mm().  But that's OK, we just need to check the
+         * current cpu's trap block PGD physical address.
+         */
+        if (tp->pgd_paddr == __pa(mm->pgd))
                 tsb_context_switch(mm);
 }
 
@@ -1278,7 +1280,7 @@ int setup_profiling_timer(unsigned int multiplier)
                 return -EINVAL;
 
         spin_lock_irqsave(&prof_setup_lock, flags);
-        for_each_cpu(i)
+        for_each_possible_cpu(i)
                 prof_multiplier(i) = multiplier;
         current_tick_offset = (timer_tick_offset / multiplier);
         spin_unlock_irqrestore(&prof_setup_lock, flags);
@@ -1306,12 +1308,12 @@ void __init smp_prepare_cpus(unsigned int max_cpus)
                 }
         }
 
-        for_each_cpu(i) {
+        for_each_possible_cpu(i) {
                 if (tlb_type == hypervisor) {
                         int j;
 
                         /* XXX get this mapping from machine description */
-                        for_each_cpu(j) {
+                        for_each_possible_cpu(j) {
                                 if ((j >> 2) == (i >> 2))
                                         cpu_set(j, cpu_sibling_map[i]);
                         }