Merge tag 'rcu.2025.09.26a' of git://git.kernel.org/pub/scm/linux/kernel/git/rcu/linux

Pull RCU updates from Paul McKenney:
 "Documentation updates:

   - Update whatisRCU.rst and checklist.rst for recent RCU API additions

   - Fix RCU documentation formatting and typos

   - Replace dead Ottawa Linux Symposium links in RTFP.txt

  Miscellaneous RCU updates:

   - Document that rcu_barrier() hurries RCU_LAZY callbacks

   - Remove redundant interrupt disabling from
     rcu_preempt_deferred_qs_handler()

   - Move list_for_each_rcu from list.h to rculist.h, and adjust the
     include directive in kernel/cgroup/dmem.c accordingly

   - Make initial set of changes to accommodate upcoming
     system_percpu_wq changes

  SRCU updates:

   - Create an srcu_read_lock_fast_notrace() for eventual use in
     tracing, including adding guards

   - Document the reliance on per-CPU operations as implicit RCU readers
     in __srcu_read_{,un}lock_fast()

   - Document the srcu_flip() function's memory-barrier D's relationship
     to SRCU-fast readers

   - Remove a redundant preempt_disable() and preempt_enable() pair from
     srcu_gp_start_if_needed()

  Torture-test updates:

   - Fix jitter.sh spin time so that it actually varies as advertised.
     It is still quite coarse-grained, but at least it does now vary

   - Update torture.sh help text to include the not-so-new --do-normal
     parameter, which permits (for example) testing KCSAN kernels
     without doing non-debug kernels

   - Fix a number of false-positive diagnostics that were being
     triggered by rcutorture starting before boot completed. Running
     multiple near-CPU-bound rcutorture processes when there is only the
     boot CPU is after all a bit excessive

   - Substitute kcalloc() for kzalloc()

   - Remove a redundant kfree() and NULL out kfree()ed objects"

* tag 'rcu.2025.09.26a' of git://git.kernel.org/pub/scm/linux/kernel/git/rcu/linux: (31 commits)
  rcu: WQ_UNBOUND added to sync_wq workqueue
  rcu: WQ_PERCPU added to alloc_workqueue users
  rcu: replace use of system_wq with system_percpu_wq
  refperf: Set reader_tasks to NULL after kfree()
  refperf: Remove redundant kfree() after torture_stop_kthread()
  srcu/tiny: Remove preempt_disable/enable() in srcu_gp_start_if_needed()
  srcu: Document srcu_flip() memory-barrier D relation to SRCU-fast
  srcu: Document __srcu_read_{,un}lock_fast() implicit RCU readers
  rculist: move list_for_each_rcu() to where it belongs
  refscale: Use kcalloc() instead of kzalloc()
  rcutorture: Use kcalloc() instead of kzalloc()
  docs: rcu: Replace multiple dead OLS links in RTFP.txt
  doc: Fix typo in RCU's torture.rst documentation
  Documentation: RCU: Retitle toctree index
  Documentation: RCU: Reduce toctree depth
  Documentation: RCU: Wrap kvm-remote.sh rerun snippet in literal code block
  rcu: docs: Requirements.rst: Abide by conventions of kernel documentation
  doc: Add RCU guards to checklist.rst
  doc: Update whatisRCU.rst for recent RCU API additions
  rcutorture: Delay forward-progress testing until boot completes
  ...

1 files changed, 7 insertions, 2 deletions

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 8eff357b0436..31690ffa452a 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -3800,6 +3800,11 @@ static void rcu_barrier_handler(void *cpu_in)
  * to complete.  For example, if there are no RCU callbacks queued anywhere
  * in the system, then rcu_barrier() is within its rights to return
  * immediately, without waiting for anything, much less an RCU grace period.
+ * In fact, rcu_barrier() will normally not result in any RCU grace periods
+ * beyond those that were already destined to be executed.
+ *
+ * In kernels built with CONFIG_RCU_LAZY=y, this function also hurries all
+ * pending lazy RCU callbacks.
  */
 void rcu_barrier(void)
 {
@@ -4885,10 +4890,10 @@ void __init rcu_init(void)
 	rcutree_online_cpu(cpu);
 
 	/* Create workqueue for Tree SRCU and for expedited GPs. */
-	rcu_gp_wq = alloc_workqueue("rcu_gp", WQ_MEM_RECLAIM, 0);
+	rcu_gp_wq = alloc_workqueue("rcu_gp", WQ_MEM_RECLAIM | WQ_PERCPU, 0);
 	WARN_ON(!rcu_gp_wq);
 
-	sync_wq = alloc_workqueue("sync_wq", WQ_MEM_RECLAIM, 0);
+	sync_wq = alloc_workqueue("sync_wq", WQ_MEM_RECLAIM | WQ_UNBOUND, 0);
 	WARN_ON(!sync_wq);
 
 	/* Respect if explicitly disabled via a boot parameter. */