@@ -3559,7 +3559,9 @@ EXPORT_SYMBOL_GPL(start_poll_synchronize_rcu_full);
* If @false is returned, it is the caller's responsibility to invoke this
* function later on until it does return @true. Alternatively, the caller
* can explicitly wait for a grace period, for example, by passing @oldstate
- * to cond_synchronize_rcu() or by directly invoking synchronize_rcu().
+ * to either cond_synchronize_rcu() or cond_synchronize_rcu_expedited()
+ * on the one hand or by directly invoking either synchronize_rcu() or
+ * synchronize_rcu_expedited() on the other.
*
* Yes, this function does not take counter wrap into account.
* But counter wrap is harmless. If the counter wraps, we have waited for
@@ -3570,6 +3572,12 @@ EXPORT_SYMBOL_GPL(start_poll_synchronize_rcu_full);
* completed. Alternatively, they can use get_completed_synchronize_rcu()
* to get a guaranteed-completed grace-period state.
*
+ * In addition, because oldstate compresses the grace-period state for
+ * both normal and expedited grace periods into a single unsigned long,
+ * it can miss a grace period when synchronize_rcu() runs concurrently
+ * with synchronize_rcu_expedited(). If this is unacceptable, please
+ * instead use the _full() variant of these polling APIs.
+ *
* This function provides the same memory-ordering guarantees that
* would be provided by a synchronize_rcu() that was invoked at the call
* to the function that provided @oldstate, and that returned at the end
This commit emphasizes the possibility of concurrent calls to synchronize_rcu() and synchronize_rcu_expedited() causing one or the other of the two grace periods being lost from the viewpoint of poll_state_synchronize_rcu(). If you cannot afford to lose grace periods this way, you should instead use the _full() variants of the polled RCU API, for example, poll_state_synchronize_rcu_full(). Signed-off-by: Paul E. McKenney <paulmck@kernel.org> --- kernel/rcu/tree.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-)