rcu-tasks: Fix computation of CPU-to-list shift counts

The ->percpu_enqueue_shift field is used to map from the running CPU
number to the index of the corresponding callback list.  This mapping
can change at runtime in response to varying callback load, resulting
in varying levels of contention on the callback-list locks.

Unfortunately, the initial value of this field is correct only if the
system happens to have a power-of-two number of CPUs, otherwise the
callbacks from the high-numbered CPUs can be placed into the callback list
indexed by 1 (rather than 0), and those index-1 callbacks will be ignored.
This can result in soft lockups and hangs.

This commit therefore corrects this mapping, adding one to this shift
count as needed for systems having odd numbers of CPUs.

Fixes: 7a30871b ("rcu-tasks: Introduce ->percpu_enqueue_shift for dynamic queue selection")
Reported-by: Andrii Nakryiko <andrii.nakryiko@gmail.com>
Cc: Reported-by: Martin Lau <kafai@fb.com>
Cc: Neeraj Upadhyay <neeraj.iitr10@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>

kernel/rcu/tasks.h

@@ -123,7 +123,7 @@ static struct rcu_tasks rt_name =							\
 	.call_func = call,								\
 	.rtpcpu = &rt_name ## __percpu,							\
 	.name = n,									\
-	.percpu_enqueue_shift = ilog2(CONFIG_NR_CPUS),					\
+	.percpu_enqueue_shift = ilog2(CONFIG_NR_CPUS) + 1,				\
 	.percpu_enqueue_lim = 1,							\
 	.percpu_dequeue_lim = 1,							\
 	.barrier_q_mutex = __MUTEX_INITIALIZER(rt_name.barrier_q_mutex),		\
@@ -216,6 +216,7 @@ static void cblist_init_generic(struct rcu_tasks *rtp)
 	int cpu;
 	unsigned long flags;
 	int lim;
+	int shift;
 
 	raw_spin_lock_irqsave(&rtp->cbs_gbl_lock, flags);
 	if (rcu_task_enqueue_lim < 0) {
@@ -229,7 +230,10 @@ static void cblist_init_generic(struct rcu_tasks *rtp)
 
 	if (lim > nr_cpu_ids)
 		lim = nr_cpu_ids;
-	WRITE_ONCE(rtp->percpu_enqueue_shift, ilog2(nr_cpu_ids / lim));
+	shift = ilog2(nr_cpu_ids / lim);
+	if (((nr_cpu_ids - 1) >> shift) >= lim)
+		shift++;
+	WRITE_ONCE(rtp->percpu_enqueue_shift, shift);
 	WRITE_ONCE(rtp->percpu_dequeue_lim, lim);
 	smp_store_release(&rtp->percpu_enqueue_lim, lim);
 	for_each_possible_cpu(cpu) {
@@ -298,7 +302,7 @@ static void call_rcu_tasks_generic(struct rcu_head *rhp, rcu_callback_t func,
 	if (unlikely(needadjust)) {
 		raw_spin_lock_irqsave(&rtp->cbs_gbl_lock, flags);
 		if (rtp->percpu_enqueue_lim != nr_cpu_ids) {
-			WRITE_ONCE(rtp->percpu_enqueue_shift, ilog2(nr_cpu_ids));
+			WRITE_ONCE(rtp->percpu_enqueue_shift, ilog2(nr_cpu_ids) + 1);
 			WRITE_ONCE(rtp->percpu_dequeue_lim, nr_cpu_ids);
 			smp_store_release(&rtp->percpu_enqueue_lim, nr_cpu_ids);
 			pr_info("Switching %s to per-CPU callback queuing.\n", rtp->name);
@@ -413,7 +417,7 @@ static int rcu_tasks_need_gpcb(struct rcu_tasks *rtp)
 	if (rcu_task_cb_adjust && ncbs <= rcu_task_collapse_lim) {
 		raw_spin_lock_irqsave(&rtp->cbs_gbl_lock, flags);
 		if (rtp->percpu_enqueue_lim > 1) {
-			WRITE_ONCE(rtp->percpu_enqueue_shift, ilog2(nr_cpu_ids));
+			WRITE_ONCE(rtp->percpu_enqueue_shift, ilog2(nr_cpu_ids) + 1);
 			smp_store_release(&rtp->percpu_enqueue_lim, 1);
 			rtp->percpu_dequeue_gpseq = get_state_synchronize_rcu();
 			pr_info("Starting switch %s to CPU-0 callback queuing.\n", rtp->name);