On Thu, Mar 15, 2012 at 5:04 PM, Kevin Hilman [off-list ref] wrote:

Colin Cross [off-list ref] writes:

quoted

+/**
+ * cpuidle_coupled_cpu_set_alive - adjust alive_count during hotplug transitions
+ * @cpu: target cpu number
+ * @alive: whether the target cpu is going up or down
+ *
+ * Run on the cpu that is bringing up the target cpu, before the target cpu
+ * has been booted, or after the target cpu is completely dead.
+ */
+static void cpuidle_coupled_cpu_set_alive(int cpu, bool alive)
+{
+     struct cpuidle_device *dev;
+     struct cpuidle_coupled *coupled;
+
+     mutex_lock(&cpuidle_lock);
+
+     dev = per_cpu(cpuidle_devices, cpu);
+     if (!dev->coupled)
+             goto out;
+
+     coupled = dev->coupled;
+
+     /*
+      * waiting_count must be at least 1 less than alive_count, because
+      * this cpu is not waiting.  Spin until all cpus have noticed this cpu
+      * is not idle and exited the ready loop before changing alive_count.
+      */
+     while (atomic_read(&coupled->ready_count))
+             cpu_relax();
+
+     smp_mb__before_atomic_inc();
+     atomic_inc(&coupled->alive_count);

This doesn't look quite right.  alive_count is incrmented whether the
CPU is going up or down?

Maybe I misunderstood something, but I don't see anywhere where
alive_count is decrmemented after a CPU is removed.

Oops, dropped the atomic_dec when I merged from two separate functions
for up and down to a single function that takes a bool.

quoted

+     smp_mb__after_atomic_inc();
+
+     if (alive)
+             coupled->requested_state[dev->cpu] = CPUIDLE_COUPLED_NOT_IDLE;
+     else
+             coupled->requested_state[dev->cpu] = CPUIDLE_COUPLED_DEAD;
+
+out:
+     mutex_unlock(&cpuidle_lock);
+}
+