[PATCH v2 01/21] srcu: Permit Tiny SRCU srcu_read_unlock() with interrupts disabled

[PATCH v2 01/21] srcu: Permit Tiny SRCU srcu_read_unlock() with interrupts disabled

[Paul E. McKenney]

The current Tiny SRCU implementation of srcu_read_unlock() awakens
the grace-period processing when exiting the outermost SRCU read-side
critical section.  However, not all Linux-kernel configurations and
contexts permit swake_up_one() to be invoked while interrupts are
disabled, and this can result in indefinitely extended SRCU grace periods.
This commit therefore only invokes swake_up_one() when interrupts are
enabled, and introduces polling to the grace-period workqueue handler.

Reported-by: kernel test robot <oliver.sang@intel.com>
Reported-by: Zqiang <qiang.zhang@linux.dev>
Closes: https://lore.kernel.org/oe-lkp/202508261642.b15eefbb-lkp@intel.com
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
---
 kernel/rcu/srcutiny.c | 13 +++++++++----
 1 file changed, 9 insertions(+), 4 deletions(-)

diff --git a/kernel/rcu/srcutiny.c b/kernel/rcu/srcutiny.c
index e3b64a5e0ec7e1..3450c3751ef7ad 100644
--- a/kernel/rcu/srcutiny.c
+++ b/kernel/rcu/srcutiny.c
@@ -106,15 +106,15 @@ void __srcu_read_unlock(struct srcu_struct *ssp, int idx)
 	newval = READ_ONCE(ssp->srcu_lock_nesting[idx]) - 1;
 	WRITE_ONCE(ssp->srcu_lock_nesting[idx], newval);
 	preempt_enable();
-	if (!newval && READ_ONCE(ssp->srcu_gp_waiting) && in_task())
+	if (!newval && READ_ONCE(ssp->srcu_gp_waiting) && in_task() && !irqs_disabled())
 		swake_up_one(&ssp->srcu_wq);
 }
 EXPORT_SYMBOL_GPL(__srcu_read_unlock);
 
 /*
  * Workqueue handler to drive one grace period and invoke any callbacks
- * that become ready as a result.  Single-CPU and !PREEMPTION operation
- * means that we get away with murder on synchronization.  ;-)
+ * that become ready as a result.  Single-CPU operation and preemption
+ * disabling mean that we get away with murder on synchronization.  ;-)
  */
 void srcu_drive_gp(struct work_struct *wp)
 {
@@ -141,7 +141,12 @@ void srcu_drive_gp(struct work_struct *wp)
 	WRITE_ONCE(ssp->srcu_idx, ssp->srcu_idx + 1);
 	WRITE_ONCE(ssp->srcu_gp_waiting, true);  /* srcu_read_unlock() wakes! */
 	preempt_enable();
-	swait_event_exclusive(ssp->srcu_wq, !READ_ONCE(ssp->srcu_lock_nesting[idx]));
+	do {
+		// Deadlock issues prevent __srcu_read_unlock() from
+		// doing an unconditional wakeup, so polling is required.
+		swait_event_timeout_exclusive(ssp->srcu_wq,
+					      !READ_ONCE(ssp->srcu_lock_nesting[idx]), HZ / 10);
+	} while (READ_ONCE(ssp->srcu_lock_nesting[idx]));
 	preempt_disable();  // Needed for PREEMPT_LAZY
 	WRITE_ONCE(ssp->srcu_gp_waiting, false); /* srcu_read_unlock() cheap. */
 	WRITE_ONCE(ssp->srcu_idx, ssp->srcu_idx + 1);
-- 
2.40.1

[PATCH v2 02/21] rcu: Re-implement RCU Tasks Trace in terms of SRCU-fast

[Paul E. McKenney]

This commit saves more than 500 lines of RCU code by re-implementing
RCU Tasks Trace in terms of SRCU-fast.  Follow-up work will remove
more code that does not cause problems by its presence, but that is no
longer required.

This variant places smp_mb() in rcu_read_{,un}lock_trace(), which will
be removed on common-case architectures in a later commit.

[ paulmck: Apply kernel test robot, Boqun Feng, and Zqiang feedback. ]
[ paulmck: Split out Tiny SRCU fixes per Andrii Nakryiko feedback. ]

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Tested-by: kernel test robot <oliver.sang@intel.com>
Cc: Andrii Nakryiko <andrii@kernel.org>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: <bpf@vger.kernel.org>
---
 include/linux/rcupdate_trace.h | 107 ++++--
 include/linux/sched.h          |   1 +
 kernel/rcu/tasks.h             | 617 +--------------------------------
 3 files changed, 95 insertions(+), 630 deletions(-)

diff --git a/include/linux/rcupdate_trace.h b/include/linux/rcupdate_trace.h
index e6c44eb428ab63..3f46cbe6700038 100644
--- a/include/linux/rcupdate_trace.h
+++ b/include/linux/rcupdate_trace.h
@@ -12,28 +12,28 @@
 #include <linux/rcupdate.h>
 #include <linux/cleanup.h>
 
-extern struct lockdep_map rcu_trace_lock_map;
+#ifdef CONFIG_TASKS_TRACE_RCU
+extern struct srcu_struct rcu_tasks_trace_srcu_struct;
+#endif // #ifdef CONFIG_TASKS_TRACE_RCU
 
-#ifdef CONFIG_DEBUG_LOCK_ALLOC
+#if defined(CONFIG_DEBUG_LOCK_ALLOC) && defined(CONFIG_TASKS_TRACE_RCU)
 
 static inline int rcu_read_lock_trace_held(void)
 {
-	return lock_is_held(&rcu_trace_lock_map);
+	return srcu_read_lock_held(&rcu_tasks_trace_srcu_struct);
 }
 
-#else /* #ifdef CONFIG_DEBUG_LOCK_ALLOC */
+#else // #if defined(CONFIG_DEBUG_LOCK_ALLOC) && defined(CONFIG_TASKS_TRACE_RCU)
 
 static inline int rcu_read_lock_trace_held(void)
 {
 	return 1;
 }
 
-#endif /* #else #ifdef CONFIG_DEBUG_LOCK_ALLOC */
+#endif // #else // #if defined(CONFIG_DEBUG_LOCK_ALLOC) && defined(CONFIG_TASKS_TRACE_RCU)
 
 #ifdef CONFIG_TASKS_TRACE_RCU
 
-void rcu_read_unlock_trace_special(struct task_struct *t);
-
 /**
  * rcu_read_lock_trace - mark beginning of RCU-trace read-side critical section
  *
@@ -50,12 +50,14 @@ static inline void rcu_read_lock_trace(void)
 {
 	struct task_struct *t = current;
 
-	WRITE_ONCE(t->trc_reader_nesting, READ_ONCE(t->trc_reader_nesting) + 1);
-	barrier();
-	if (IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB) &&
-	    t->trc_reader_special.b.need_mb)
-		smp_mb(); // Pairs with update-side barriers
-	rcu_lock_acquire(&rcu_trace_lock_map);
+	if (t->trc_reader_nesting++) {
+		// In case we interrupted a Tasks Trace RCU reader.
+		rcu_try_lock_acquire(&rcu_tasks_trace_srcu_struct.dep_map);
+		return;
+	}
+	barrier();  // nesting before scp to protect against interrupt handler.
+	t->trc_reader_scp = srcu_read_lock_fast(&rcu_tasks_trace_srcu_struct);
+	smp_mb(); // Placeholder for more selective ordering
 }
 
 /**
@@ -69,26 +71,75 @@ static inline void rcu_read_lock_trace(void)
  */
 static inline void rcu_read_unlock_trace(void)
 {
-	int nesting;
+	struct srcu_ctr __percpu *scp;
 	struct task_struct *t = current;
 
-	rcu_lock_release(&rcu_trace_lock_map);
-	nesting = READ_ONCE(t->trc_reader_nesting) - 1;
-	barrier(); // Critical section before disabling.
-	// Disable IPI-based setting of .need_qs.
-	WRITE_ONCE(t->trc_reader_nesting, INT_MIN + nesting);
-	if (likely(!READ_ONCE(t->trc_reader_special.s)) || nesting) {
-		WRITE_ONCE(t->trc_reader_nesting, nesting);
-		return;  // We assume shallow reader nesting.
-	}
-	WARN_ON_ONCE(nesting != 0);
-	rcu_read_unlock_trace_special(t);
+	smp_mb(); // Placeholder for more selective ordering
+	scp = t->trc_reader_scp;
+	barrier();  // scp before nesting to protect against interrupt handler.
+	if (!--t->trc_reader_nesting)
+		srcu_read_unlock_fast(&rcu_tasks_trace_srcu_struct, scp);
+	else
+		srcu_lock_release(&rcu_tasks_trace_srcu_struct.dep_map);
+}
+
+/**
+ * call_rcu_tasks_trace() - Queue a callback trace task-based grace period
+ * @rhp: structure to be used for queueing the RCU updates.
+ * @func: actual callback function to be invoked after the grace period
+ *
+ * The callback function will be invoked some time after a trace rcu-tasks
+ * grace period elapses, in other words after all currently executing
+ * trace rcu-tasks read-side critical sections have completed. These
+ * read-side critical sections are delimited by calls to rcu_read_lock_trace()
+ * and rcu_read_unlock_trace().
+ *
+ * See the description of call_rcu() for more detailed information on
+ * memory ordering guarantees.
+ */
+static inline void call_rcu_tasks_trace(struct rcu_head *rhp, rcu_callback_t func)
+{
+	call_srcu(&rcu_tasks_trace_srcu_struct, rhp, func);
+}
+
+/**
+ * synchronize_rcu_tasks_trace - wait for a trace rcu-tasks grace period
+ *
+ * Control will return to the caller some time after a trace rcu-tasks
+ * grace period has elapsed, in other words after all currently executing
+ * trace rcu-tasks read-side critical sections have elapsed. These read-side
+ * critical sections are delimited by calls to rcu_read_lock_trace()
+ * and rcu_read_unlock_trace().
+ *
+ * This is a very specialized primitive, intended only for a few uses in
+ * tracing and other situations requiring manipulation of function preambles
+ * and profiling hooks.  The synchronize_rcu_tasks_trace() function is not
+ * (yet) intended for heavy use from multiple CPUs.
+ *
+ * See the description of synchronize_rcu() for more detailed information
+ * on memory ordering guarantees.
+ */
+static inline void synchronize_rcu_tasks_trace(void)
+{
+	synchronize_srcu(&rcu_tasks_trace_srcu_struct);
 }
 
-void call_rcu_tasks_trace(struct rcu_head *rhp, rcu_callback_t func);
-void synchronize_rcu_tasks_trace(void);
-void rcu_barrier_tasks_trace(void);
+/**
+ * rcu_barrier_tasks_trace - Wait for in-flight call_rcu_tasks_trace() callbacks.
+ *
+ * Note that rcu_barrier_tasks_trace() is not obligated to actually wait,
+ * for example, if there are no pending callbacks.
+ */
+static inline void rcu_barrier_tasks_trace(void)
+{
+	srcu_barrier(&rcu_tasks_trace_srcu_struct);
+}
+
+// Placeholders to enable stepwise transition.
+void rcu_tasks_trace_get_gp_data(int *flags, unsigned long *gp_seq);
+void __init rcu_tasks_trace_suppress_unused(void);
 struct task_struct *get_rcu_tasks_trace_gp_kthread(void);
+
 #else
 /*
  * The BPF JIT forms these addresses even when it doesn't call these
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 2b272382673d62..89d3646155525f 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -939,6 +939,7 @@ struct task_struct {
 
 #ifdef CONFIG_TASKS_TRACE_RCU
 	int				trc_reader_nesting;
+	struct srcu_ctr __percpu	*trc_reader_scp;
 	int				trc_ipi_to_cpu;
 	union rcu_special		trc_reader_special;
 	struct list_head		trc_holdout_list;
diff --git a/kernel/rcu/tasks.h b/kernel/rcu/tasks.h
index 2dc044fd126eb0..418fa242cf0288 100644
--- a/kernel/rcu/tasks.h
+++ b/kernel/rcu/tasks.h
@@ -718,7 +718,6 @@ static void __init rcu_tasks_bootup_oddness(void)
 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
 }
 
-
 /* Dump out rcutorture-relevant state common to all RCU-tasks flavors. */
 static void show_rcu_tasks_generic_gp_kthread(struct rcu_tasks *rtp, char *s)
 {
@@ -803,7 +802,7 @@ static void rcu_tasks_torture_stats_print_generic(struct rcu_tasks *rtp, char *t
 
 static void exit_tasks_rcu_finish_trace(struct task_struct *t);
 
-#if defined(CONFIG_TASKS_RCU) || defined(CONFIG_TASKS_TRACE_RCU)
+#if defined(CONFIG_TASKS_RCU)
 
 ////////////////////////////////////////////////////////////////////////
 //
@@ -898,7 +897,7 @@ static void rcu_tasks_wait_gp(struct rcu_tasks *rtp)
 	rtp->postgp_func(rtp);
 }
 
-#endif /* #if defined(CONFIG_TASKS_RCU) || defined(CONFIG_TASKS_TRACE_RCU) */
+#endif /* #if defined(CONFIG_TASKS_RCU) */
 
 #ifdef CONFIG_TASKS_RCU
 
@@ -1453,94 +1452,23 @@ EXPORT_SYMBOL_GPL(rcu_tasks_rude_get_gp_data);
 //
 // Tracing variant of Tasks RCU.  This variant is designed to be used
 // to protect tracing hooks, including those of BPF.  This variant
-// therefore:
-//
-// 1.	Has explicit read-side markers to allow finite grace periods
-//	in the face of in-kernel loops for PREEMPT=n builds.
-//
-// 2.	Protects code in the idle loop, exception entry/exit, and
-//	CPU-hotplug code paths, similar to the capabilities of SRCU.
-//
-// 3.	Avoids expensive read-side instructions, having overhead similar
-//	to that of Preemptible RCU.
-//
-// There are of course downsides.  For example, the grace-period code
-// can send IPIs to CPUs, even when those CPUs are in the idle loop or
-// in nohz_full userspace.  If needed, these downsides can be at least
-// partially remedied.
-//
-// Perhaps most important, this variant of RCU does not affect the vanilla
-// flavors, rcu_preempt and rcu_sched.  The fact that RCU Tasks Trace
-// readers can operate from idle, offline, and exception entry/exit in no
-// way allows rcu_preempt and rcu_sched readers to also do so.
-//
-// The implementation uses rcu_tasks_wait_gp(), which relies on function
-// pointers in the rcu_tasks structure.  The rcu_spawn_tasks_trace_kthread()
-// function sets these function pointers up so that rcu_tasks_wait_gp()
-// invokes these functions in this order:
-//
-// rcu_tasks_trace_pregp_step():
-//	Disables CPU hotplug, adds all currently executing tasks to the
-//	holdout list, then checks the state of all tasks that blocked
-//	or were preempted within their current RCU Tasks Trace read-side
-//	critical section, adding them to the holdout list if appropriate.
-//	Finally, this function re-enables CPU hotplug.
-// The ->pertask_func() pointer is NULL, so there is no per-task processing.
-// rcu_tasks_trace_postscan():
-//	Invokes synchronize_rcu() to wait for late-stage exiting tasks
-//	to finish exiting.
-// check_all_holdout_tasks_trace(), repeatedly until holdout list is empty:
-//	Scans the holdout list, attempting to identify a quiescent state
-//	for each task on the list.  If there is a quiescent state, the
-//	corresponding task is removed from the holdout list.  Once this
-//	list is empty, the grace period has completed.
-// rcu_tasks_trace_postgp():
-//	Provides the needed full memory barrier and does debug checks.
-//
-// The exit_tasks_rcu_finish_trace() synchronizes with exiting tasks.
-//
-// Pre-grace-period update-side code is ordered before the grace period
-// via the ->cbs_lock and barriers in rcu_tasks_kthread().  Pre-grace-period
-// read-side code is ordered before the grace period by atomic operations
-// on .b.need_qs flag of each task involved in this process, or by scheduler
-// context-switch ordering (for locked-down non-running readers).
-
-// The lockdep state must be outside of #ifdef to be useful.
-#ifdef CONFIG_DEBUG_LOCK_ALLOC
-static struct lock_class_key rcu_lock_trace_key;
-struct lockdep_map rcu_trace_lock_map =
-	STATIC_LOCKDEP_MAP_INIT("rcu_read_lock_trace", &rcu_lock_trace_key);
-EXPORT_SYMBOL_GPL(rcu_trace_lock_map);
-#endif /* #ifdef CONFIG_DEBUG_LOCK_ALLOC */
+// is implemented via a straightforward mapping onto SRCU-fast.
 
 #ifdef CONFIG_TASKS_TRACE_RCU
 
-// Record outstanding IPIs to each CPU.  No point in sending two...
-static DEFINE_PER_CPU(bool, trc_ipi_to_cpu);
-
-// The number of detections of task quiescent state relying on
-// heavyweight readers executing explicit memory barriers.
-static unsigned long n_heavy_reader_attempts;
-static unsigned long n_heavy_reader_updates;
-static unsigned long n_heavy_reader_ofl_updates;
-static unsigned long n_trc_holdouts;
+DEFINE_SRCU(rcu_tasks_trace_srcu_struct);
+EXPORT_SYMBOL_GPL(rcu_tasks_trace_srcu_struct);
 
-void call_rcu_tasks_trace(struct rcu_head *rhp, rcu_callback_t func);
-DEFINE_RCU_TASKS(rcu_tasks_trace, rcu_tasks_wait_gp, call_rcu_tasks_trace,
-		 "RCU Tasks Trace");
-
-/* Load from ->trc_reader_special.b.need_qs with proper ordering. */
-static u8 rcu_ld_need_qs(struct task_struct *t)
+// Placeholder to suppress build errors through transition period.
+void __init rcu_tasks_trace_suppress_unused(void)
 {
-	smp_mb(); // Enforce full grace-period ordering.
-	return smp_load_acquire(&t->trc_reader_special.b.need_qs);
-}
-
-/* Store to ->trc_reader_special.b.need_qs with proper ordering. */
-static void rcu_st_need_qs(struct task_struct *t, u8 v)
-{
-	smp_store_release(&t->trc_reader_special.b.need_qs, v);
-	smp_mb(); // Enforce full grace-period ordering.
+	show_rcu_tasks_generic_gp_kthread(NULL, NULL);
+	rcu_spawn_tasks_kthread_generic(NULL);
+	synchronize_rcu_tasks_generic(NULL);
+	call_rcu_tasks_generic(NULL, NULL, NULL);
+	call_rcu_tasks_iw_wakeup(NULL);
+	cblist_init_generic(NULL);
+	rcu_tasks_torture_stats_print_generic(NULL, NULL, NULL, NULL);
 }
 
 /*
@@ -1555,321 +1483,12 @@ u8 rcu_trc_cmpxchg_need_qs(struct task_struct *t, u8 old, u8 new)
 }
 EXPORT_SYMBOL_GPL(rcu_trc_cmpxchg_need_qs);
 
-/*
- * If we are the last reader, signal the grace-period kthread.
- * Also remove from the per-CPU list of blocked tasks.
- */
-void rcu_read_unlock_trace_special(struct task_struct *t)
-{
-	unsigned long flags;
-	struct rcu_tasks_percpu *rtpcp;
-	union rcu_special trs;
-
-	// Open-coded full-word version of rcu_ld_need_qs().
-	smp_mb(); // Enforce full grace-period ordering.
-	trs = smp_load_acquire(&t->trc_reader_special);
-
-	if (IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB) && t->trc_reader_special.b.need_mb)
-		smp_mb(); // Pairs with update-side barriers.
-	// Update .need_qs before ->trc_reader_nesting for irq/NMI handlers.
-	if (trs.b.need_qs == (TRC_NEED_QS_CHECKED | TRC_NEED_QS)) {
-		u8 result = rcu_trc_cmpxchg_need_qs(t, TRC_NEED_QS_CHECKED | TRC_NEED_QS,
-						       TRC_NEED_QS_CHECKED);
-
-		WARN_ONCE(result != trs.b.need_qs, "%s: result = %d", __func__, result);
-	}
-	if (trs.b.blocked) {
-		rtpcp = per_cpu_ptr(rcu_tasks_trace.rtpcpu, t->trc_blkd_cpu);
-		raw_spin_lock_irqsave_rcu_node(rtpcp, flags);
-		list_del_init(&t->trc_blkd_node);
-		WRITE_ONCE(t->trc_reader_special.b.blocked, false);
-		raw_spin_unlock_irqrestore_rcu_node(rtpcp, flags);
-	}
-	WRITE_ONCE(t->trc_reader_nesting, 0);
-}
-EXPORT_SYMBOL_GPL(rcu_read_unlock_trace_special);
-
 /* Add a newly blocked reader task to its CPU's list. */
 void rcu_tasks_trace_qs_blkd(struct task_struct *t)
 {
-	unsigned long flags;
-	struct rcu_tasks_percpu *rtpcp;
-
-	local_irq_save(flags);
-	rtpcp = this_cpu_ptr(rcu_tasks_trace.rtpcpu);
-	raw_spin_lock_rcu_node(rtpcp); // irqs already disabled
-	t->trc_blkd_cpu = smp_processor_id();
-	if (!rtpcp->rtp_blkd_tasks.next)
-		INIT_LIST_HEAD(&rtpcp->rtp_blkd_tasks);
-	list_add(&t->trc_blkd_node, &rtpcp->rtp_blkd_tasks);
-	WRITE_ONCE(t->trc_reader_special.b.blocked, true);
-	raw_spin_unlock_irqrestore_rcu_node(rtpcp, flags);
 }
 EXPORT_SYMBOL_GPL(rcu_tasks_trace_qs_blkd);
 
-/* Add a task to the holdout list, if it is not already on the list. */
-static void trc_add_holdout(struct task_struct *t, struct list_head *bhp)
-{
-	if (list_empty(&t->trc_holdout_list)) {
-		get_task_struct(t);
-		list_add(&t->trc_holdout_list, bhp);
-		n_trc_holdouts++;
-	}
-}
-
-/* Remove a task from the holdout list, if it is in fact present. */
-static void trc_del_holdout(struct task_struct *t)
-{
-	if (!list_empty(&t->trc_holdout_list)) {
-		list_del_init(&t->trc_holdout_list);
-		put_task_struct(t);
-		n_trc_holdouts--;
-	}
-}
-
-/* IPI handler to check task state. */
-static void trc_read_check_handler(void *t_in)
-{
-	int nesting;
-	struct task_struct *t = current;
-	struct task_struct *texp = t_in;
-
-	// If the task is no longer running on this CPU, leave.
-	if (unlikely(texp != t))
-		goto reset_ipi; // Already on holdout list, so will check later.
-
-	// If the task is not in a read-side critical section, and
-	// if this is the last reader, awaken the grace-period kthread.
-	nesting = READ_ONCE(t->trc_reader_nesting);
-	if (likely(!nesting)) {
-		rcu_trc_cmpxchg_need_qs(t, 0, TRC_NEED_QS_CHECKED);
-		goto reset_ipi;
-	}
-	// If we are racing with an rcu_read_unlock_trace(), try again later.
-	if (unlikely(nesting < 0))
-		goto reset_ipi;
-
-	// Get here if the task is in a read-side critical section.
-	// Set its state so that it will update state for the grace-period
-	// kthread upon exit from that critical section.
-	rcu_trc_cmpxchg_need_qs(t, 0, TRC_NEED_QS | TRC_NEED_QS_CHECKED);
-
-reset_ipi:
-	// Allow future IPIs to be sent on CPU and for task.
-	// Also order this IPI handler against any later manipulations of
-	// the intended task.
-	smp_store_release(per_cpu_ptr(&trc_ipi_to_cpu, smp_processor_id()), false); // ^^^
-	smp_store_release(&texp->trc_ipi_to_cpu, -1); // ^^^
-}
-
-/* Callback function for scheduler to check locked-down task.  */
-static int trc_inspect_reader(struct task_struct *t, void *bhp_in)
-{
-	struct list_head *bhp = bhp_in;
-	int cpu = task_cpu(t);
-	int nesting;
-	bool ofl = cpu_is_offline(cpu);
-
-	if (task_curr(t) && !ofl) {
-		// If no chance of heavyweight readers, do it the hard way.
-		if (!IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB))
-			return -EINVAL;
-
-		// If heavyweight readers are enabled on the remote task,
-		// we can inspect its state despite its currently running.
-		// However, we cannot safely change its state.
-		n_heavy_reader_attempts++;
-		// Check for "running" idle tasks on offline CPUs.
-		if (!rcu_watching_zero_in_eqs(cpu, &t->trc_reader_nesting))
-			return -EINVAL; // No quiescent state, do it the hard way.
-		n_heavy_reader_updates++;
-		nesting = 0;
-	} else {
-		// The task is not running, so C-language access is safe.
-		nesting = t->trc_reader_nesting;
-		WARN_ON_ONCE(ofl && task_curr(t) && (t != idle_task(task_cpu(t))));
-		if (IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB) && ofl)
-			n_heavy_reader_ofl_updates++;
-	}
-
-	// If not exiting a read-side critical section, mark as checked
-	// so that the grace-period kthread will remove it from the
-	// holdout list.
-	if (!nesting) {
-		rcu_trc_cmpxchg_need_qs(t, 0, TRC_NEED_QS_CHECKED);
-		return 0;  // In QS, so done.
-	}
-	if (nesting < 0)
-		return -EINVAL; // Reader transitioning, try again later.
-
-	// The task is in a read-side critical section, so set up its
-	// state so that it will update state upon exit from that critical
-	// section.
-	if (!rcu_trc_cmpxchg_need_qs(t, 0, TRC_NEED_QS | TRC_NEED_QS_CHECKED))
-		trc_add_holdout(t, bhp);
-	return 0;
-}
-
-/* Attempt to extract the state for the specified task. */
-static void trc_wait_for_one_reader(struct task_struct *t,
-				    struct list_head *bhp)
-{
-	int cpu;
-
-	// If a previous IPI is still in flight, let it complete.
-	if (smp_load_acquire(&t->trc_ipi_to_cpu) != -1) // Order IPI
-		return;
-
-	// The current task had better be in a quiescent state.
-	if (t == current) {
-		rcu_trc_cmpxchg_need_qs(t, 0, TRC_NEED_QS_CHECKED);
-		WARN_ON_ONCE(READ_ONCE(t->trc_reader_nesting));
-		return;
-	}
-
-	// Attempt to nail down the task for inspection.
-	get_task_struct(t);
-	if (!task_call_func(t, trc_inspect_reader, bhp)) {
-		put_task_struct(t);
-		return;
-	}
-	put_task_struct(t);
-
-	// If this task is not yet on the holdout list, then we are in
-	// an RCU read-side critical section.  Otherwise, the invocation of
-	// trc_add_holdout() that added it to the list did the necessary
-	// get_task_struct().  Either way, the task cannot be freed out
-	// from under this code.
-
-	// If currently running, send an IPI, either way, add to list.
-	trc_add_holdout(t, bhp);
-	if (task_curr(t) &&
-	    time_after(jiffies + 1, rcu_tasks_trace.gp_start + rcu_task_ipi_delay)) {
-		// The task is currently running, so try IPIing it.
-		cpu = task_cpu(t);
-
-		// If there is already an IPI outstanding, let it happen.
-		if (per_cpu(trc_ipi_to_cpu, cpu) || t->trc_ipi_to_cpu >= 0)
-			return;
-
-		per_cpu(trc_ipi_to_cpu, cpu) = true;
-		t->trc_ipi_to_cpu = cpu;
-		rcu_tasks_trace.n_ipis++;
-		if (smp_call_function_single(cpu, trc_read_check_handler, t, 0)) {
-			// Just in case there is some other reason for
-			// failure than the target CPU being offline.
-			WARN_ONCE(1, "%s():  smp_call_function_single() failed for CPU: %d\n",
-				  __func__, cpu);
-			rcu_tasks_trace.n_ipis_fails++;
-			per_cpu(trc_ipi_to_cpu, cpu) = false;
-			t->trc_ipi_to_cpu = -1;
-		}
-	}
-}
-
-/*
- * Initialize for first-round processing for the specified task.
- * Return false if task is NULL or already taken care of, true otherwise.
- */
-static bool rcu_tasks_trace_pertask_prep(struct task_struct *t, bool notself)
-{
-	// During early boot when there is only the one boot CPU, there
-	// is no idle task for the other CPUs.	Also, the grace-period
-	// kthread is always in a quiescent state.  In addition, just return
-	// if this task is already on the list.
-	if (unlikely(t == NULL) || (t == current && notself) || !list_empty(&t->trc_holdout_list))
-		return false;
-
-	rcu_st_need_qs(t, 0);
-	t->trc_ipi_to_cpu = -1;
-	return true;
-}
-
-/* Do first-round processing for the specified task. */
-static void rcu_tasks_trace_pertask(struct task_struct *t, struct list_head *hop)
-{
-	if (rcu_tasks_trace_pertask_prep(t, true))
-		trc_wait_for_one_reader(t, hop);
-}
-
-/* Initialize for a new RCU-tasks-trace grace period. */
-static void rcu_tasks_trace_pregp_step(struct list_head *hop)
-{
-	LIST_HEAD(blkd_tasks);
-	int cpu;
-	unsigned long flags;
-	struct rcu_tasks_percpu *rtpcp;
-	struct task_struct *t;
-
-	// There shouldn't be any old IPIs, but...
-	for_each_possible_cpu(cpu)
-		WARN_ON_ONCE(per_cpu(trc_ipi_to_cpu, cpu));
-
-	// Disable CPU hotplug across the CPU scan for the benefit of
-	// any IPIs that might be needed.  This also waits for all readers
-	// in CPU-hotplug code paths.
-	cpus_read_lock();
-
-	// These rcu_tasks_trace_pertask_prep() calls are serialized to
-	// allow safe access to the hop list.
-	for_each_online_cpu(cpu) {
-		rcu_read_lock();
-		// Note that cpu_curr_snapshot() picks up the target
-		// CPU's current task while its runqueue is locked with
-		// an smp_mb__after_spinlock().  This ensures that either
-		// the grace-period kthread will see that task's read-side
-		// critical section or the task will see the updater's pre-GP
-		// accesses.  The trailing smp_mb() in cpu_curr_snapshot()
-		// does not currently play a role other than simplify
-		// that function's ordering semantics.  If these simplified
-		// ordering semantics continue to be redundant, that smp_mb()
-		// might be removed.
-		t = cpu_curr_snapshot(cpu);
-		if (rcu_tasks_trace_pertask_prep(t, true))
-			trc_add_holdout(t, hop);
-		rcu_read_unlock();
-		cond_resched_tasks_rcu_qs();
-	}
-
-	// Only after all running tasks have been accounted for is it
-	// safe to take care of the tasks that have blocked within their
-	// current RCU tasks trace read-side critical section.
-	for_each_possible_cpu(cpu) {
-		rtpcp = per_cpu_ptr(rcu_tasks_trace.rtpcpu, cpu);
-		raw_spin_lock_irqsave_rcu_node(rtpcp, flags);
-		list_splice_init(&rtpcp->rtp_blkd_tasks, &blkd_tasks);
-		while (!list_empty(&blkd_tasks)) {
-			rcu_read_lock();
-			t = list_first_entry(&blkd_tasks, struct task_struct, trc_blkd_node);
-			list_del_init(&t->trc_blkd_node);
-			list_add(&t->trc_blkd_node, &rtpcp->rtp_blkd_tasks);
-			raw_spin_unlock_irqrestore_rcu_node(rtpcp, flags);
-			rcu_tasks_trace_pertask(t, hop);
-			rcu_read_unlock();
-			raw_spin_lock_irqsave_rcu_node(rtpcp, flags);
-		}
-		raw_spin_unlock_irqrestore_rcu_node(rtpcp, flags);
-		cond_resched_tasks_rcu_qs();
-	}
-
-	// Re-enable CPU hotplug now that the holdout list is populated.
-	cpus_read_unlock();
-}
-
-/*
- * Do intermediate processing between task and holdout scans.
- */
-static void rcu_tasks_trace_postscan(struct list_head *hop)
-{
-	// Wait for late-stage exiting tasks to finish exiting.
-	// These might have passed the call to exit_tasks_rcu_finish().
-
-	// If you remove the following line, update rcu_trace_implies_rcu_gp()!!!
-	synchronize_rcu();
-	// Any tasks that exit after this point will set
-	// TRC_NEED_QS_CHECKED in ->trc_reader_special.b.need_qs.
-}
-
 /* Communicate task state back to the RCU tasks trace stall warning request. */
 struct trc_stall_chk_rdr {
 	int nesting;
@@ -1877,241 +1496,39 @@ struct trc_stall_chk_rdr {
 	u8 needqs;
 };
 
-static int trc_check_slow_task(struct task_struct *t, void *arg)
-{
-	struct trc_stall_chk_rdr *trc_rdrp = arg;
-
-	if (task_curr(t) && cpu_online(task_cpu(t)))
-		return false; // It is running, so decline to inspect it.
-	trc_rdrp->nesting = READ_ONCE(t->trc_reader_nesting);
-	trc_rdrp->ipi_to_cpu = READ_ONCE(t->trc_ipi_to_cpu);
-	trc_rdrp->needqs = rcu_ld_need_qs(t);
-	return true;
-}
-
-/* Show the state of a task stalling the current RCU tasks trace GP. */
-static void show_stalled_task_trace(struct task_struct *t, bool *firstreport)
-{
-	int cpu;
-	struct trc_stall_chk_rdr trc_rdr;
-	bool is_idle_tsk = is_idle_task(t);
-
-	if (*firstreport) {
-		pr_err("INFO: rcu_tasks_trace detected stalls on tasks:\n");
-		*firstreport = false;
-	}
-	cpu = task_cpu(t);
-	if (!task_call_func(t, trc_check_slow_task, &trc_rdr))
-		pr_alert("P%d: %c%c\n",
-			 t->pid,
-			 ".I"[t->trc_ipi_to_cpu >= 0],
-			 ".i"[is_idle_tsk]);
-	else
-		pr_alert("P%d: %c%c%c%c nesting: %d%c%c cpu: %d%s\n",
-			 t->pid,
-			 ".I"[trc_rdr.ipi_to_cpu >= 0],
-			 ".i"[is_idle_tsk],
-			 ".N"[cpu >= 0 && tick_nohz_full_cpu(cpu)],
-			 ".B"[!!data_race(t->trc_reader_special.b.blocked)],
-			 trc_rdr.nesting,
-			 " !CN"[trc_rdr.needqs & 0x3],
-			 " ?"[trc_rdr.needqs > 0x3],
-			 cpu, cpu_online(cpu) ? "" : "(offline)");
-	sched_show_task(t);
-}
-
-/* List stalled IPIs for RCU tasks trace. */
-static void show_stalled_ipi_trace(void)
-{
-	int cpu;
-
-	for_each_possible_cpu(cpu)
-		if (per_cpu(trc_ipi_to_cpu, cpu))
-			pr_alert("\tIPI outstanding to CPU %d\n", cpu);
-}
-
-/* Do one scan of the holdout list. */
-static void check_all_holdout_tasks_trace(struct list_head *hop,
-					  bool needreport, bool *firstreport)
-{
-	struct task_struct *g, *t;
-
-	// Disable CPU hotplug across the holdout list scan for IPIs.
-	cpus_read_lock();
-
-	list_for_each_entry_safe(t, g, hop, trc_holdout_list) {
-		// If safe and needed, try to check the current task.
-		if (READ_ONCE(t->trc_ipi_to_cpu) == -1 &&
-		    !(rcu_ld_need_qs(t) & TRC_NEED_QS_CHECKED))
-			trc_wait_for_one_reader(t, hop);
-
-		// If check succeeded, remove this task from the list.
-		if (smp_load_acquire(&t->trc_ipi_to_cpu) == -1 &&
-		    rcu_ld_need_qs(t) == TRC_NEED_QS_CHECKED)
-			trc_del_holdout(t);
-		else if (needreport)
-			show_stalled_task_trace(t, firstreport);
-		cond_resched_tasks_rcu_qs();
-	}
-
-	// Re-enable CPU hotplug now that the holdout list scan has completed.
-	cpus_read_unlock();
-
-	if (needreport) {
-		if (*firstreport)
-			pr_err("INFO: rcu_tasks_trace detected stalls? (Late IPI?)\n");
-		show_stalled_ipi_trace();
-	}
-}
-
-static void rcu_tasks_trace_empty_fn(void *unused)
-{
-}
-
-/* Wait for grace period to complete and provide ordering. */
-static void rcu_tasks_trace_postgp(struct rcu_tasks *rtp)
-{
-	int cpu;
-
-	// Wait for any lingering IPI handlers to complete.  Note that
-	// if a CPU has gone offline or transitioned to userspace in the
-	// meantime, all IPI handlers should have been drained beforehand.
-	// Yes, this assumes that CPUs process IPIs in order.  If that ever
-	// changes, there will need to be a recheck and/or timed wait.
-	for_each_online_cpu(cpu)
-		if (WARN_ON_ONCE(smp_load_acquire(per_cpu_ptr(&trc_ipi_to_cpu, cpu))))
-			smp_call_function_single(cpu, rcu_tasks_trace_empty_fn, NULL, 1);
-
-	smp_mb(); // Caller's code must be ordered after wakeup.
-		  // Pairs with pretty much every ordering primitive.
-}
-
 /* Report any needed quiescent state for this exiting task. */
 static void exit_tasks_rcu_finish_trace(struct task_struct *t)
 {
-	union rcu_special trs = READ_ONCE(t->trc_reader_special);
-
-	rcu_trc_cmpxchg_need_qs(t, 0, TRC_NEED_QS_CHECKED);
-	WARN_ON_ONCE(READ_ONCE(t->trc_reader_nesting));
-	if (WARN_ON_ONCE(rcu_ld_need_qs(t) & TRC_NEED_QS || trs.b.blocked))
-		rcu_read_unlock_trace_special(t);
-	else
-		WRITE_ONCE(t->trc_reader_nesting, 0);
-}
-
-/**
- * call_rcu_tasks_trace() - Queue a callback trace task-based grace period
- * @rhp: structure to be used for queueing the RCU updates.
- * @func: actual callback function to be invoked after the grace period
- *
- * The callback function will be invoked some time after a trace rcu-tasks
- * grace period elapses, in other words after all currently executing
- * trace rcu-tasks read-side critical sections have completed. These
- * read-side critical sections are delimited by calls to rcu_read_lock_trace()
- * and rcu_read_unlock_trace().
- *
- * See the description of call_rcu() for more detailed information on
- * memory ordering guarantees.
- */
-void call_rcu_tasks_trace(struct rcu_head *rhp, rcu_callback_t func)
-{
-	call_rcu_tasks_generic(rhp, func, &rcu_tasks_trace);
-}
-EXPORT_SYMBOL_GPL(call_rcu_tasks_trace);
-
-/**
- * synchronize_rcu_tasks_trace - wait for a trace rcu-tasks grace period
- *
- * Control will return to the caller some time after a trace rcu-tasks
- * grace period has elapsed, in other words after all currently executing
- * trace rcu-tasks read-side critical sections have elapsed. These read-side
- * critical sections are delimited by calls to rcu_read_lock_trace()
- * and rcu_read_unlock_trace().
- *
- * This is a very specialized primitive, intended only for a few uses in
- * tracing and other situations requiring manipulation of function preambles
- * and profiling hooks.  The synchronize_rcu_tasks_trace() function is not
- * (yet) intended for heavy use from multiple CPUs.
- *
- * See the description of synchronize_rcu() for more detailed information
- * on memory ordering guarantees.
- */
-void synchronize_rcu_tasks_trace(void)
-{
-	RCU_LOCKDEP_WARN(lock_is_held(&rcu_trace_lock_map), "Illegal synchronize_rcu_tasks_trace() in RCU Tasks Trace read-side critical section");
-	synchronize_rcu_tasks_generic(&rcu_tasks_trace);
 }
-EXPORT_SYMBOL_GPL(synchronize_rcu_tasks_trace);
-
-/**
- * rcu_barrier_tasks_trace - Wait for in-flight call_rcu_tasks_trace() callbacks.
- *
- * Although the current implementation is guaranteed to wait, it is not
- * obligated to, for example, if there are no pending callbacks.
- */
-void rcu_barrier_tasks_trace(void)
-{
-	rcu_barrier_tasks_generic(&rcu_tasks_trace);
-}
-EXPORT_SYMBOL_GPL(rcu_barrier_tasks_trace);
 
 int rcu_tasks_trace_lazy_ms = -1;
 module_param(rcu_tasks_trace_lazy_ms, int, 0444);
 
 static int __init rcu_spawn_tasks_trace_kthread(void)
 {
-	if (IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB)) {
-		rcu_tasks_trace.gp_sleep = HZ / 10;
-		rcu_tasks_trace.init_fract = HZ / 10;
-	} else {
-		rcu_tasks_trace.gp_sleep = HZ / 200;
-		if (rcu_tasks_trace.gp_sleep <= 0)
-			rcu_tasks_trace.gp_sleep = 1;
-		rcu_tasks_trace.init_fract = HZ / 200;
-		if (rcu_tasks_trace.init_fract <= 0)
-			rcu_tasks_trace.init_fract = 1;
-	}
-	if (rcu_tasks_trace_lazy_ms >= 0)
-		rcu_tasks_trace.lazy_jiffies = msecs_to_jiffies(rcu_tasks_trace_lazy_ms);
-	rcu_tasks_trace.pregp_func = rcu_tasks_trace_pregp_step;
-	rcu_tasks_trace.postscan_func = rcu_tasks_trace_postscan;
-	rcu_tasks_trace.holdouts_func = check_all_holdout_tasks_trace;
-	rcu_tasks_trace.postgp_func = rcu_tasks_trace_postgp;
-	rcu_spawn_tasks_kthread_generic(&rcu_tasks_trace);
 	return 0;
 }
 
 #if !defined(CONFIG_TINY_RCU)
 void show_rcu_tasks_trace_gp_kthread(void)
 {
-	char buf[64];
-
-	snprintf(buf, sizeof(buf), "N%lu h:%lu/%lu/%lu",
-		data_race(n_trc_holdouts),
-		data_race(n_heavy_reader_ofl_updates),
-		data_race(n_heavy_reader_updates),
-		data_race(n_heavy_reader_attempts));
-	show_rcu_tasks_generic_gp_kthread(&rcu_tasks_trace, buf);
 }
 EXPORT_SYMBOL_GPL(show_rcu_tasks_trace_gp_kthread);
 
 void rcu_tasks_trace_torture_stats_print(char *tt, char *tf)
 {
-	rcu_tasks_torture_stats_print_generic(&rcu_tasks_trace, tt, tf, "");
 }
 EXPORT_SYMBOL_GPL(rcu_tasks_trace_torture_stats_print);
 #endif // !defined(CONFIG_TINY_RCU)
 
 struct task_struct *get_rcu_tasks_trace_gp_kthread(void)
 {
-	return rcu_tasks_trace.kthread_ptr;
+	return NULL;
 }
 EXPORT_SYMBOL_GPL(get_rcu_tasks_trace_gp_kthread);
 
 void rcu_tasks_trace_get_gp_data(int *flags, unsigned long *gp_seq)
 {
-	*flags = 0;
-	*gp_seq = rcu_seq_current(&rcu_tasks_trace.tasks_gp_seq);
 }
 EXPORT_SYMBOL_GPL(rcu_tasks_trace_get_gp_data);
 
@@ -2251,10 +1668,6 @@ void __init tasks_cblist_init_generic(void)
 #ifdef CONFIG_TASKS_RUDE_RCU
 	cblist_init_generic(&rcu_tasks_rude);
 #endif
-
-#ifdef CONFIG_TASKS_TRACE_RCU
-	cblist_init_generic(&rcu_tasks_trace);
-#endif
 }
 
 static int __init rcu_init_tasks_generic(void)
-- 
2.40.1

[PATCH v2 03/21] context_tracking: Remove rcu_task_trace_heavyweight_{enter,exit}()

[Paul E. McKenney]

Because SRCU-fast does not use IPIs for its grace periods, there is
no need for real-time workloads to switch to an IPI-free mode, and
there is in turn no need for either rcu_task_trace_heavyweight_enter()
or rcu_task_trace_heavyweight_exit().  This commit therefore removes them.

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Cc: Frederic Weisbecker <frederic@kernel.org>
Cc: Andrii Nakryiko <andrii@kernel.org>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: <bpf@vger.kernel.org>
---
 kernel/context_tracking.c | 20 --------------------
 1 file changed, 20 deletions(-)

diff --git a/kernel/context_tracking.c b/kernel/context_tracking.c
index fb5be6e9b423f2..a743e7ffa6c00f 100644
--- a/kernel/context_tracking.c
+++ b/kernel/context_tracking.c
@@ -54,24 +54,6 @@ static __always_inline void rcu_task_enter(void)
 #endif /* #if defined(CONFIG_TASKS_RCU) && defined(CONFIG_NO_HZ_FULL) */
 }
 
-/* Turn on heavyweight RCU tasks trace readers on kernel exit. */
-static __always_inline void rcu_task_trace_heavyweight_enter(void)
-{
-#ifdef CONFIG_TASKS_TRACE_RCU
-	if (IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB))
-		current->trc_reader_special.b.need_mb = true;
-#endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
-}
-
-/* Turn off heavyweight RCU tasks trace readers on kernel entry. */
-static __always_inline void rcu_task_trace_heavyweight_exit(void)
-{
-#ifdef CONFIG_TASKS_TRACE_RCU
-	if (IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB))
-		current->trc_reader_special.b.need_mb = false;
-#endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
-}
-
 /*
  * Record entry into an extended quiescent state.  This is only to be
  * called when not already in an extended quiescent state, that is,
@@ -85,7 +67,6 @@ static noinstr void ct_kernel_exit_state(int offset)
 	 * critical sections, and we also must force ordering with the
 	 * next idle sojourn.
 	 */
-	rcu_task_trace_heavyweight_enter();  // Before CT state update!
 	// RCU is still watching.  Better not be in extended quiescent state!
 	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) && !rcu_is_watching_curr_cpu());
 	(void)ct_state_inc(offset);
@@ -108,7 +89,6 @@ static noinstr void ct_kernel_enter_state(int offset)
 	 */
 	seq = ct_state_inc(offset);
 	// RCU is now watching.  Better not be in an extended quiescent state!
-	rcu_task_trace_heavyweight_exit();  // After CT state update!
 	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) && !(seq & CT_RCU_WATCHING));
 }
 
-- 
2.40.1

[PATCH v2 04/21] rcu: Clean up after the SRCU-fastification of RCU Tasks Trace

[Paul E. McKenney]

Now that RCU Tasks Trace has been re-implemented in terms of SRCU-fast,
the ->trc_ipi_to_cpu, ->trc_blkd_cpu, ->trc_blkd_node, ->trc_holdout_list,
and ->trc_reader_special task_struct fields are no longer used.

In addition, the rcu_tasks_trace_qs(), rcu_tasks_trace_qs_blkd(),
exit_tasks_rcu_finish_trace(), and rcu_spawn_tasks_trace_kthread(),
show_rcu_tasks_trace_gp_kthread(), rcu_tasks_trace_get_gp_data(),
rcu_tasks_trace_torture_stats_print(), and get_rcu_tasks_trace_gp_kthread()
functions and all the other functions that they invoke are no longer used.

Also, the TRC_NEED_QS and TRC_NEED_QS_CHECKED CPP macros are no longer used.
Neither are the rcu_tasks_trace_lazy_ms and rcu_task_ipi_delay rcupdate
module parameters and the TASKS_TRACE_RCU_READ_MB Kconfig option.

This commit therefore removes all of them.

[ paulmck: Apply Alexei Starovoitov feedback. ]

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Cc: Andrii Nakryiko <andrii@kernel.org>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: <bpf@vger.kernel.org>
---
 .../admin-guide/kernel-parameters.txt         | 15 ----
 include/linux/rcupdate.h                      | 31 +-------
 include/linux/rcupdate_trace.h                |  2 -
 include/linux/sched.h                         |  5 --
 init/init_task.c                              |  3 -
 kernel/fork.c                                 |  3 -
 kernel/rcu/Kconfig                            | 18 -----
 kernel/rcu/rcu.h                              |  9 ---
 kernel/rcu/rcuscale.c                         |  7 --
 kernel/rcu/rcutorture.c                       |  2 -
 kernel/rcu/tasks.h                            | 79 +------------------
 .../selftests/rcutorture/configs/rcu/TRACE01  |  1 -
 .../selftests/rcutorture/configs/rcu/TRACE02  |  1 -
 13 files changed, 2 insertions(+), 174 deletions(-)

diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index 747a55abf4946b..40fc198bed4db9 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -6078,13 +6078,6 @@
 			dynamically) adjusted.	This parameter is intended
 			for use in testing.
 
-	rcupdate.rcu_task_ipi_delay= [KNL]
-			Set time in jiffies during which RCU tasks will
-			avoid sending IPIs, starting with the beginning
-			of a given grace period.  Setting a large
-			number avoids disturbing real-time workloads,
-			but lengthens grace periods.
-
 	rcupdate.rcu_task_lazy_lim= [KNL]
 			Number of callbacks on a given CPU that will
 			cancel laziness on that CPU.  Use -1 to disable
@@ -6128,14 +6121,6 @@
 			of zero will disable batching.	Batching is
 			always disabled for synchronize_rcu_tasks().
 
-	rcupdate.rcu_tasks_trace_lazy_ms= [KNL]
-			Set timeout in milliseconds RCU Tasks
-			Trace asynchronous callback batching for
-			call_rcu_tasks_trace().  A negative value
-			will take the default.	A value of zero will
-			disable batching.  Batching is always disabled
-			for synchronize_rcu_tasks_trace().
-
 	rcupdate.rcu_self_test= [KNL]
 			Run the RCU early boot self tests
 
diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
index 120536f4c6eb1d..2237c56d32417b 100644
--- a/include/linux/rcupdate.h
+++ b/include/linux/rcupdate.h
@@ -175,36 +175,7 @@ void rcu_tasks_torture_stats_print(char *tt, char *tf);
 # define synchronize_rcu_tasks synchronize_rcu
 # endif
 
-# ifdef CONFIG_TASKS_TRACE_RCU
-// Bits for ->trc_reader_special.b.need_qs field.
-#define TRC_NEED_QS		0x1  // Task needs a quiescent state.
-#define TRC_NEED_QS_CHECKED	0x2  // Task has been checked for needing quiescent state.
-
-u8 rcu_trc_cmpxchg_need_qs(struct task_struct *t, u8 old, u8 new);
-void rcu_tasks_trace_qs_blkd(struct task_struct *t);
-
-# define rcu_tasks_trace_qs(t)							\
-	do {									\
-		int ___rttq_nesting = READ_ONCE((t)->trc_reader_nesting);	\
-										\
-		if (unlikely(READ_ONCE((t)->trc_reader_special.b.need_qs) == TRC_NEED_QS) &&	\
-		    likely(!___rttq_nesting)) {					\
-			rcu_trc_cmpxchg_need_qs((t), TRC_NEED_QS, TRC_NEED_QS_CHECKED);	\
-		} else if (___rttq_nesting && ___rttq_nesting != INT_MIN &&	\
-			   !READ_ONCE((t)->trc_reader_special.b.blocked)) {	\
-			rcu_tasks_trace_qs_blkd(t);				\
-		}								\
-	} while (0)
-void rcu_tasks_trace_torture_stats_print(char *tt, char *tf);
-# else
-# define rcu_tasks_trace_qs(t) do { } while (0)
-# endif
-
-#define rcu_tasks_qs(t, preempt)					\
-do {									\
-	rcu_tasks_classic_qs((t), (preempt));				\
-	rcu_tasks_trace_qs(t);						\
-} while (0)
+#define rcu_tasks_qs(t, preempt) rcu_tasks_classic_qs((t), (preempt))
 
 # ifdef CONFIG_TASKS_RUDE_RCU
 void synchronize_rcu_tasks_rude(void);
diff --git a/include/linux/rcupdate_trace.h b/include/linux/rcupdate_trace.h
index 3f46cbe6700038..0bd47f12ecd17b 100644
--- a/include/linux/rcupdate_trace.h
+++ b/include/linux/rcupdate_trace.h
@@ -136,9 +136,7 @@ static inline void rcu_barrier_tasks_trace(void)
 }
 
 // Placeholders to enable stepwise transition.
-void rcu_tasks_trace_get_gp_data(int *flags, unsigned long *gp_seq);
 void __init rcu_tasks_trace_suppress_unused(void);
-struct task_struct *get_rcu_tasks_trace_gp_kthread(void);
 
 #else
 /*
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 89d3646155525f..577fafd22a0e6f 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -940,11 +940,6 @@ struct task_struct {
 #ifdef CONFIG_TASKS_TRACE_RCU
 	int				trc_reader_nesting;
 	struct srcu_ctr __percpu	*trc_reader_scp;
-	int				trc_ipi_to_cpu;
-	union rcu_special		trc_reader_special;
-	struct list_head		trc_holdout_list;
-	struct list_head		trc_blkd_node;
-	int				trc_blkd_cpu;
 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
 
 	struct sched_info		sched_info;
diff --git a/init/init_task.c b/init/init_task.c
index e557f622bd9061..0c075f3cc8fc5a 100644
--- a/init/init_task.c
+++ b/init/init_task.c
@@ -165,9 +165,6 @@ struct task_struct init_task __aligned(L1_CACHE_BYTES) = {
 #endif
 #ifdef CONFIG_TASKS_TRACE_RCU
 	.trc_reader_nesting = 0,
-	.trc_reader_special.s = 0,
-	.trc_holdout_list = LIST_HEAD_INIT(init_task.trc_holdout_list),
-	.trc_blkd_node = LIST_HEAD_INIT(init_task.trc_blkd_node),
 #endif
 #ifdef CONFIG_CPUSETS
 	.mems_allowed_seq = SEQCNT_SPINLOCK_ZERO(init_task.mems_allowed_seq,
diff --git a/kernel/fork.c b/kernel/fork.c
index af673856499dca..5686d50b62cfaf 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -1780,9 +1780,6 @@ static inline void rcu_copy_process(struct task_struct *p)
 #endif /* #ifdef CONFIG_TASKS_RCU */
 #ifdef CONFIG_TASKS_TRACE_RCU
 	p->trc_reader_nesting = 0;
-	p->trc_reader_special.s = 0;
-	INIT_LIST_HEAD(&p->trc_holdout_list);
-	INIT_LIST_HEAD(&p->trc_blkd_node);
 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
 }
 
diff --git a/kernel/rcu/Kconfig b/kernel/rcu/Kconfig
index ceaf6594f634cd..54b4c4aa553a4a 100644
--- a/kernel/rcu/Kconfig
+++ b/kernel/rcu/Kconfig
@@ -313,24 +313,6 @@ config RCU_NOCB_CPU_CB_BOOST
 	  Say Y here if you want to set RT priority for offloading kthreads.
 	  Say N here if you are building a !PREEMPT_RT kernel and are unsure.
 
-config TASKS_TRACE_RCU_READ_MB
-	bool "Tasks Trace RCU readers use memory barriers in user and idle"
-	depends on RCU_EXPERT && TASKS_TRACE_RCU
-	default PREEMPT_RT || NR_CPUS < 8
-	help
-	  Use this option to further reduce the number of IPIs sent
-	  to CPUs executing in userspace or idle during tasks trace
-	  RCU grace periods.  Given that a reasonable setting of
-	  the rcupdate.rcu_task_ipi_delay kernel boot parameter
-	  eliminates such IPIs for many workloads, proper setting
-	  of this Kconfig option is important mostly for aggressive
-	  real-time installations and for battery-powered devices,
-	  hence the default chosen above.
-
-	  Say Y here if you hate IPIs.
-	  Say N here if you hate read-side memory barriers.
-	  Take the default if you are unsure.
-
 config RCU_LAZY
 	bool "RCU callback lazy invocation functionality"
 	depends on RCU_NOCB_CPU
diff --git a/kernel/rcu/rcu.h b/kernel/rcu/rcu.h
index 9cf01832a6c3d1..dc5d614b372c1e 100644
--- a/kernel/rcu/rcu.h
+++ b/kernel/rcu/rcu.h
@@ -544,10 +544,6 @@ struct task_struct *get_rcu_tasks_rude_gp_kthread(void);
 void rcu_tasks_rude_get_gp_data(int *flags, unsigned long *gp_seq);
 #endif // # ifdef CONFIG_TASKS_RUDE_RCU
 
-#ifdef CONFIG_TASKS_TRACE_RCU
-void rcu_tasks_trace_get_gp_data(int *flags, unsigned long *gp_seq);
-#endif
-
 #ifdef CONFIG_TASKS_RCU_GENERIC
 void tasks_cblist_init_generic(void);
 #else /* #ifdef CONFIG_TASKS_RCU_GENERIC */
@@ -673,11 +669,6 @@ void show_rcu_tasks_rude_gp_kthread(void);
 #else
 static inline void show_rcu_tasks_rude_gp_kthread(void) {}
 #endif
-#if !defined(CONFIG_TINY_RCU) && defined(CONFIG_TASKS_TRACE_RCU)
-void show_rcu_tasks_trace_gp_kthread(void);
-#else
-static inline void show_rcu_tasks_trace_gp_kthread(void) {}
-#endif
 
 #ifdef CONFIG_TINY_RCU
 static inline bool rcu_cpu_beenfullyonline(int cpu) { return true; }
diff --git a/kernel/rcu/rcuscale.c b/kernel/rcu/rcuscale.c
index b521d04559927a..17d038e26b65de 100644
--- a/kernel/rcu/rcuscale.c
+++ b/kernel/rcu/rcuscale.c
@@ -400,11 +400,6 @@ static void tasks_trace_scale_read_unlock(int idx)
 	rcu_read_unlock_trace();
 }
 
-static void rcu_tasks_trace_scale_stats(void)
-{
-	rcu_tasks_trace_torture_stats_print(scale_type, SCALE_FLAG);
-}
-
 static struct rcu_scale_ops tasks_tracing_ops = {
 	.ptype		= RCU_TASKS_FLAVOR,
 	.init		= rcu_sync_scale_init,
@@ -416,8 +411,6 @@ static struct rcu_scale_ops tasks_tracing_ops = {
 	.gp_barrier	= rcu_barrier_tasks_trace,
 	.sync		= synchronize_rcu_tasks_trace,
 	.exp_sync	= synchronize_rcu_tasks_trace,
-	.rso_gp_kthread	= get_rcu_tasks_trace_gp_kthread,
-	.stats		= IS_ENABLED(CONFIG_TINY_RCU) ? NULL : rcu_tasks_trace_scale_stats,
 	.name		= "tasks-tracing"
 };
 
diff --git a/kernel/rcu/rcutorture.c b/kernel/rcu/rcutorture.c
index 72619e5e8549dc..485fa822b6a753 100644
--- a/kernel/rcu/rcutorture.c
+++ b/kernel/rcu/rcutorture.c
@@ -1130,8 +1130,6 @@ static struct rcu_torture_ops tasks_tracing_ops = {
 	.exp_sync	= synchronize_rcu_tasks_trace,
 	.call		= call_rcu_tasks_trace,
 	.cb_barrier	= rcu_barrier_tasks_trace,
-	.gp_kthread_dbg	= show_rcu_tasks_trace_gp_kthread,
-	.get_gp_data    = rcu_tasks_trace_get_gp_data,
 	.cbflood_max	= 50000,
 	.irq_capable	= 1,
 	.slow_gps	= 1,
diff --git a/kernel/rcu/tasks.h b/kernel/rcu/tasks.h
index 418fa242cf0288..fd1fe80ddde484 100644
--- a/kernel/rcu/tasks.h
+++ b/kernel/rcu/tasks.h
@@ -161,11 +161,6 @@ static void tasks_rcu_exit_srcu_stall(struct timer_list *unused);
 static DEFINE_TIMER(tasks_rcu_exit_srcu_stall_timer, tasks_rcu_exit_srcu_stall);
 #endif
 
-/* Avoid IPIing CPUs early in the grace period. */
-#define RCU_TASK_IPI_DELAY (IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB) ? HZ / 2 : 0)
-static int rcu_task_ipi_delay __read_mostly = RCU_TASK_IPI_DELAY;
-module_param(rcu_task_ipi_delay, int, 0644);
-
 /* Control stall timeouts.  Disable with <= 0, otherwise jiffies till stall. */
 #define RCU_TASK_BOOT_STALL_TIMEOUT (HZ * 30)
 #define RCU_TASK_STALL_TIMEOUT (HZ * 60 * 10)
@@ -800,8 +795,6 @@ static void rcu_tasks_torture_stats_print_generic(struct rcu_tasks *rtp, char *t
 
 #endif // #ifndef CONFIG_TINY_RCU
 
-static void exit_tasks_rcu_finish_trace(struct task_struct *t);
-
 #if defined(CONFIG_TASKS_RCU)
 
 ////////////////////////////////////////////////////////////////////////
@@ -1321,13 +1314,11 @@ void exit_tasks_rcu_finish(void)
 	raw_spin_lock_irqsave_rcu_node(rtpcp, flags);
 	list_del_init(&t->rcu_tasks_exit_list);
 	raw_spin_unlock_irqrestore_rcu_node(rtpcp, flags);
-
-	exit_tasks_rcu_finish_trace(t);
 }
 
 #else /* #ifdef CONFIG_TASKS_RCU */
 void exit_tasks_rcu_start(void) { }
-void exit_tasks_rcu_finish(void) { exit_tasks_rcu_finish_trace(current); }
+void exit_tasks_rcu_finish(void) { }
 #endif /* #else #ifdef CONFIG_TASKS_RCU */
 
 #ifdef CONFIG_TASKS_RUDE_RCU
@@ -1471,69 +1462,6 @@ void __init rcu_tasks_trace_suppress_unused(void)
 	rcu_tasks_torture_stats_print_generic(NULL, NULL, NULL, NULL);
 }
 
-/*
- * Do a cmpxchg() on ->trc_reader_special.b.need_qs, allowing for
- * the four-byte operand-size restriction of some platforms.
- *
- * Returns the old value, which is often ignored.
- */
-u8 rcu_trc_cmpxchg_need_qs(struct task_struct *t, u8 old, u8 new)
-{
-	return cmpxchg(&t->trc_reader_special.b.need_qs, old, new);
-}
-EXPORT_SYMBOL_GPL(rcu_trc_cmpxchg_need_qs);
-
-/* Add a newly blocked reader task to its CPU's list. */
-void rcu_tasks_trace_qs_blkd(struct task_struct *t)
-{
-}
-EXPORT_SYMBOL_GPL(rcu_tasks_trace_qs_blkd);
-
-/* Communicate task state back to the RCU tasks trace stall warning request. */
-struct trc_stall_chk_rdr {
-	int nesting;
-	int ipi_to_cpu;
-	u8 needqs;
-};
-
-/* Report any needed quiescent state for this exiting task. */
-static void exit_tasks_rcu_finish_trace(struct task_struct *t)
-{
-}
-
-int rcu_tasks_trace_lazy_ms = -1;
-module_param(rcu_tasks_trace_lazy_ms, int, 0444);
-
-static int __init rcu_spawn_tasks_trace_kthread(void)
-{
-	return 0;
-}
-
-#if !defined(CONFIG_TINY_RCU)
-void show_rcu_tasks_trace_gp_kthread(void)
-{
-}
-EXPORT_SYMBOL_GPL(show_rcu_tasks_trace_gp_kthread);
-
-void rcu_tasks_trace_torture_stats_print(char *tt, char *tf)
-{
-}
-EXPORT_SYMBOL_GPL(rcu_tasks_trace_torture_stats_print);
-#endif // !defined(CONFIG_TINY_RCU)
-
-struct task_struct *get_rcu_tasks_trace_gp_kthread(void)
-{
-	return NULL;
-}
-EXPORT_SYMBOL_GPL(get_rcu_tasks_trace_gp_kthread);
-
-void rcu_tasks_trace_get_gp_data(int *flags, unsigned long *gp_seq)
-{
-}
-EXPORT_SYMBOL_GPL(rcu_tasks_trace_get_gp_data);
-
-#else /* #ifdef CONFIG_TASKS_TRACE_RCU */
-static void exit_tasks_rcu_finish_trace(struct task_struct *t) { }
 #endif /* #else #ifdef CONFIG_TASKS_TRACE_RCU */
 
 #ifndef CONFIG_TINY_RCU
@@ -1541,7 +1469,6 @@ void show_rcu_tasks_gp_kthreads(void)
 {
 	show_rcu_tasks_classic_gp_kthread();
 	show_rcu_tasks_rude_gp_kthread();
-	show_rcu_tasks_trace_gp_kthread();
 }
 #endif /* #ifndef CONFIG_TINY_RCU */
 
@@ -1680,10 +1607,6 @@ static int __init rcu_init_tasks_generic(void)
 	rcu_spawn_tasks_rude_kthread();
 #endif
 
-#ifdef CONFIG_TASKS_TRACE_RCU
-	rcu_spawn_tasks_trace_kthread();
-#endif
-
 	// Run the self-tests.
 	rcu_tasks_initiate_self_tests();
 
diff --git a/tools/testing/selftests/rcutorture/configs/rcu/TRACE01 b/tools/testing/selftests/rcutorture/configs/rcu/TRACE01
index 85b407467454a2..18efab346381a4 100644
--- a/tools/testing/selftests/rcutorture/configs/rcu/TRACE01
+++ b/tools/testing/selftests/rcutorture/configs/rcu/TRACE01
@@ -10,5 +10,4 @@ CONFIG_PROVE_LOCKING=n
 #CHECK#CONFIG_PROVE_RCU=n
 CONFIG_FORCE_TASKS_TRACE_RCU=y
 #CHECK#CONFIG_TASKS_TRACE_RCU=y
-CONFIG_TASKS_TRACE_RCU_READ_MB=y
 CONFIG_RCU_EXPERT=y
diff --git a/tools/testing/selftests/rcutorture/configs/rcu/TRACE02 b/tools/testing/selftests/rcutorture/configs/rcu/TRACE02
index 9003c56cd76484..8da390e8282977 100644
--- a/tools/testing/selftests/rcutorture/configs/rcu/TRACE02
+++ b/tools/testing/selftests/rcutorture/configs/rcu/TRACE02
@@ -9,6 +9,5 @@ CONFIG_PROVE_LOCKING=y
 #CHECK#CONFIG_PROVE_RCU=y
 CONFIG_FORCE_TASKS_TRACE_RCU=y
 #CHECK#CONFIG_TASKS_TRACE_RCU=y
-CONFIG_TASKS_TRACE_RCU_READ_MB=n
 CONFIG_RCU_EXPERT=y
 CONFIG_DEBUG_OBJECTS=y
-- 
2.40.1

[PATCH v2 05/21] rcu: Move rcu_tasks_trace_srcu_struct out of #ifdef CONFIG_TASKS_RCU_GENERIC

[Paul E. McKenney]

Moving the rcu_tasks_trace_srcu_struct structure instance out
from under the CONFIG_TASKS_RCU_GENERIC Kconfig option permits
the CONFIG_TASKS_TRACE_RCU Kconfig option to stop enabling this
CONFIG_TASKS_RCU_GENERIC Kconfig option.  This commit also therefore
makes it so.

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Cc: Andrii Nakryiko <andrii@kernel.org>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: <bpf@vger.kernel.org>
---
 kernel/rcu/Kconfig |  2 +-
 kernel/rcu/tasks.h | 38 +++++++++++++-------------------------
 2 files changed, 14 insertions(+), 26 deletions(-)

diff --git a/kernel/rcu/Kconfig b/kernel/rcu/Kconfig
index 54b4c4aa553a4a..73a6cc364628b5 100644
--- a/kernel/rcu/Kconfig
+++ b/kernel/rcu/Kconfig
@@ -82,7 +82,7 @@ config NEED_SRCU_NMI_SAFE
 	def_bool HAVE_NMI && !ARCH_HAS_NMI_SAFE_THIS_CPU_OPS && !TINY_SRCU
 
 config TASKS_RCU_GENERIC
-	def_bool TASKS_RCU || TASKS_RUDE_RCU || TASKS_TRACE_RCU
+	def_bool TASKS_RCU || TASKS_RUDE_RCU
 	help
 	  This option enables generic infrastructure code supporting
 	  task-based RCU implementations.  Not for manual selection.
diff --git a/kernel/rcu/tasks.h b/kernel/rcu/tasks.h
index fd1fe80ddde484..833e180db744f2 100644
--- a/kernel/rcu/tasks.h
+++ b/kernel/rcu/tasks.h
@@ -1439,31 +1439,6 @@ EXPORT_SYMBOL_GPL(rcu_tasks_rude_get_gp_data);
 
 #endif /* #ifdef CONFIG_TASKS_RUDE_RCU */
 
-////////////////////////////////////////////////////////////////////////
-//
-// Tracing variant of Tasks RCU.  This variant is designed to be used
-// to protect tracing hooks, including those of BPF.  This variant
-// is implemented via a straightforward mapping onto SRCU-fast.
-
-#ifdef CONFIG_TASKS_TRACE_RCU
-
-DEFINE_SRCU(rcu_tasks_trace_srcu_struct);
-EXPORT_SYMBOL_GPL(rcu_tasks_trace_srcu_struct);
-
-// Placeholder to suppress build errors through transition period.
-void __init rcu_tasks_trace_suppress_unused(void)
-{
-	show_rcu_tasks_generic_gp_kthread(NULL, NULL);
-	rcu_spawn_tasks_kthread_generic(NULL);
-	synchronize_rcu_tasks_generic(NULL);
-	call_rcu_tasks_generic(NULL, NULL, NULL);
-	call_rcu_tasks_iw_wakeup(NULL);
-	cblist_init_generic(NULL);
-	rcu_tasks_torture_stats_print_generic(NULL, NULL, NULL, NULL);
-}
-
-#endif /* #else #ifdef CONFIG_TASKS_TRACE_RCU */
-
 #ifndef CONFIG_TINY_RCU
 void show_rcu_tasks_gp_kthreads(void)
 {
@@ -1617,3 +1592,16 @@ core_initcall(rcu_init_tasks_generic);
 #else /* #ifdef CONFIG_TASKS_RCU_GENERIC */
 static inline void rcu_tasks_bootup_oddness(void) {}
 #endif /* #else #ifdef CONFIG_TASKS_RCU_GENERIC */
+
+#ifdef CONFIG_TASKS_TRACE_RCU
+
+////////////////////////////////////////////////////////////////////////
+//
+// Tracing variant of Tasks RCU.  This variant is designed to be used
+// to protect tracing hooks, including those of BPF.  This variant
+// is implemented via a straightforward mapping onto SRCU-fast.
+
+DEFINE_SRCU(rcu_tasks_trace_srcu_struct);
+EXPORT_SYMBOL_GPL(rcu_tasks_trace_srcu_struct);
+
+#endif /* #else #ifdef CONFIG_TASKS_TRACE_RCU */
-- 
2.40.1

[PATCH v2 0/21] Implement RCU Tasks Trace in terms of SRCU-fast and optimize

[Paul E. McKenney]

Hello!

This v2 series re-implements RCU Tasks Trace in terms of SRCU-fast,
reducing the size of the Linux-kernel RCU implementation by several
hundred lines of code.  It also removes a conditional branch from the
srcu_read_lock_fast() implementation in order to make SRCU-fast a bit
more fastpath-friendly.  The patches are as follows:

1.	Permit Tiny SRCU srcu_read_unlock() with interrupts disabled.

2.	Re-implement RCU Tasks Trace in terms of SRCU-fast.

3.	context_tracking: Remove
	rcu_task_trace_heavyweight_{enter,exit}().

4.	Clean up after the SRCU-fastification of RCU Tasks Trace.

5.	Move rcu_tasks_trace_srcu_struct out of #ifdef
	CONFIG_TASKS_RCU_GENERIC.

6.	Create an srcu_expedite_current() function.

7.	Test srcu_expedite_current().

8.	Add noinstr-fast rcu_read_{,un}lock_tasks_trace() APIs.

9.	Update Requirements.rst for RCU Tasks Trace.

10.	Deprecate rcu_read_{,un}lock_trace().

11.	Create a DEFINE_SRCU_FAST().

12.	Create an rcu_tasks_trace_expedite_current() function.

13.	Test rcu_tasks_trace_expedite_current().

14.	Make grace-period determination use ssp->srcu_reader_flavor.

15.	Exercise DEFINE_STATIC_SRCU_FAST() and init_srcu_struct_fast().

16.	Exercise DEFINE_STATIC_SRCU_FAST() and init_srcu_struct_fast().

17.	Require special srcu_struct define/init for SRCU-fast readers.

18.	Make SRCU-fast readers enforce use of SRCU-fast definition/init.

19.	Update for SRCU-fast definitions and initialization.

20.	Guard __DECLARE_TRACE() use of __DO_TRACE_CALL() with SRCU-fast.

21.	Mark diagnostic functions as notrace.

Changes since v1:

o	Consolidate RCU Tasks Trace cleanup patches per Alexei Starovoitov
	feedback.  This explains the decrease from 34 patches in v1 to
	only 21 in this v2 series.

o	While consolidating, consolidate the noinstr-fast patches and the
	DEFINE_SRCU_FAST()/init_srcu_struct_fast() patches.

o	Upgrade comments per Peter Zijlstra feedback.

o	Extract Tiny SRCU bug fix into its own commit per Andrii Nakryiko
	feedback.

o	Switch srcu_expedite_current() from preempt_disable() to
	migrate_disable() for PREEMPT_RT kernels per Zqiang feedback.

o	Apply tags.

https://lore.kernel.org/all/580ea2de-799a-4ddc-bde9-c16f3fb1e6e7@paulmck-laptop/

						Thanx, Paul

------------------------------------------------------------------------

 Documentation/RCU/Design/Requirements/Requirements.rst   |   33 
 b/Documentation/RCU/Design/Requirements/Requirements.rst |   12 
 b/Documentation/RCU/checklist.rst                        |   12 
 b/Documentation/RCU/whatisRCU.rst                        |    3 
 b/Documentation/admin-guide/kernel-parameters.txt        |   15 
 b/include/linux/notifier.h                               |    2 
 b/include/linux/rcupdate.h                               |   31 
 b/include/linux/rcupdate_trace.h                         |  107 +-
 b/include/linux/sched.h                                  |    1 
 b/include/linux/srcu.h                                   |   16 
 b/include/linux/srcutiny.h                               |    1 
 b/include/linux/srcutree.h                               |    8 
 b/include/linux/tracepoint.h                             |   45 -
 b/include/trace/perf.h                                   |    4 
 b/include/trace/trace_events.h                           |    4 
 b/init/init_task.c                                       |    3 
 b/kernel/context_tracking.c                              |   20 
 b/kernel/fork.c                                          |    3 
 b/kernel/rcu/Kconfig                                     |   18 
 b/kernel/rcu/rcu.h                                       |    9 
 b/kernel/rcu/rcuscale.c                                  |    7 
 b/kernel/rcu/rcutorture.c                                |    2 
 b/kernel/rcu/refscale.c                                  |    9 
 b/kernel/rcu/srcutiny.c                                  |   13 
 b/kernel/rcu/srcutree.c                                  |   58 +
 b/kernel/rcu/tasks.h                                     |  617 ---------------
 b/kernel/rcu/tree.c                                      |    2 
 b/kernel/rcu/update.c                                    |    8 
 b/kernel/tracepoint.c                                    |   21 
 b/scripts/checkpatch.pl                                  |    2 
 b/tools/testing/selftests/rcutorture/configs/rcu/TRACE01 |    1 
 b/tools/testing/selftests/rcutorture/configs/rcu/TRACE02 |    1 
 include/linux/rcupdate_trace.h                           |   81 +
 include/linux/sched.h                                    |    5 
 include/linux/srcu.h                                     |   40 
 include/linux/srcutiny.h                                 |   14 
 include/linux/srcutree.h                                 |   50 -
 kernel/rcu/Kconfig                                       |   25 
 kernel/rcu/rcutorture.c                                  |   32 
 kernel/rcu/srcutree.c                                    |   39 
 kernel/rcu/tasks.h                                       |  129 ---
 41 files changed, 525 insertions(+), 978 deletions(-)

[PATCH v2 06/21] srcu: Create an srcu_expedite_current() function

[Paul E. McKenney]

This commit creates an srcu_expedite_current() function that expedites
the current (and possibly the next) SRCU grace period for the specified
srcu_struct structure.  This functionality will be inherited by RCU
Tasks Trace courtesy of its mapping to SRCU fast.

If the current SRCU grace period is already waiting, that wait will
complete before the expediting takes effect.  If there is no SRCU grace
period in flight, this function might well create one.

[ paulmck: Apply Zqiang feedback for PREEMPT_RT use. ]

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Cc: Andrii Nakryiko <andrii@kernel.org>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: <bpf@vger.kernel.org>
---
 include/linux/srcutiny.h |  1 +
 include/linux/srcutree.h |  8 ++++++
 kernel/rcu/srcutree.c    | 58 ++++++++++++++++++++++++++++++++++++++++
 3 files changed, 67 insertions(+)

diff --git a/include/linux/srcutiny.h b/include/linux/srcutiny.h
index 51ce25f07930ee..3bfbd44cb1b37d 100644
--- a/include/linux/srcutiny.h
+++ b/include/linux/srcutiny.h
@@ -103,6 +103,7 @@ static inline void srcu_barrier(struct srcu_struct *ssp)
 	synchronize_srcu(ssp);
 }
 
+static inline void srcu_expedite_current(struct srcu_struct *ssp) { }
 #define srcu_check_read_flavor(ssp, read_flavor) do { } while (0)
 #define srcu_check_read_flavor_force(ssp, read_flavor) do { } while (0)
 
diff --git a/include/linux/srcutree.h b/include/linux/srcutree.h
index 42098e0fa0b7dd..93ad18acd6d02d 100644
--- a/include/linux/srcutree.h
+++ b/include/linux/srcutree.h
@@ -42,6 +42,8 @@ struct srcu_data {
 	struct timer_list delay_work;		/* Delay for CB invoking */
 	struct work_struct work;		/* Context for CB invoking. */
 	struct rcu_head srcu_barrier_head;	/* For srcu_barrier() use. */
+	struct rcu_head srcu_ec_head;		/* For srcu_expedite_current() use. */
+	int srcu_ec_state;			/*  State for srcu_expedite_current(). */
 	struct srcu_node *mynode;		/* Leaf srcu_node. */
 	unsigned long grpmask;			/* Mask for leaf srcu_node */
 						/*  ->srcu_data_have_cbs[]. */
@@ -135,6 +137,11 @@ struct srcu_struct {
 #define SRCU_STATE_SCAN1	1
 #define SRCU_STATE_SCAN2	2
 
+/* Values for srcu_expedite_current() state (->srcu_ec_state). */
+#define SRCU_EC_IDLE		0
+#define SRCU_EC_PENDING		1
+#define SRCU_EC_REPOST		2
+
 /*
  * Values for initializing gp sequence fields. Higher values allow wrap arounds to
  * occur earlier.
@@ -210,6 +217,7 @@ struct srcu_struct {
 int __srcu_read_lock(struct srcu_struct *ssp) __acquires(ssp);
 void synchronize_srcu_expedited(struct srcu_struct *ssp);
 void srcu_barrier(struct srcu_struct *ssp);
+void srcu_expedite_current(struct srcu_struct *ssp);
 void srcu_torture_stats_print(struct srcu_struct *ssp, char *tt, char *tf);
 
 // Converts a per-CPU pointer to an ->srcu_ctrs[] array element to that
diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
index 1ff94b76d91f15..38b440b0b0c80b 100644
--- a/kernel/rcu/srcutree.c
+++ b/kernel/rcu/srcutree.c
@@ -1688,6 +1688,64 @@ void srcu_barrier(struct srcu_struct *ssp)
 }
 EXPORT_SYMBOL_GPL(srcu_barrier);
 
+/* Callback for srcu_expedite_current() usage. */
+static void srcu_expedite_current_cb(struct rcu_head *rhp)
+{
+	unsigned long flags;
+	bool needcb = false;
+	struct srcu_data *sdp = container_of(rhp, struct srcu_data, srcu_ec_head);
+
+	spin_lock_irqsave_sdp_contention(sdp, &flags);
+	if (sdp->srcu_ec_state == SRCU_EC_IDLE) {
+		WARN_ON_ONCE(1);
+	} else if (sdp->srcu_ec_state == SRCU_EC_PENDING) {
+		sdp->srcu_ec_state = SRCU_EC_IDLE;
+	} else {
+		WARN_ON_ONCE(sdp->srcu_ec_state != SRCU_EC_REPOST);
+		sdp->srcu_ec_state = SRCU_EC_PENDING;
+		needcb = true;
+	}
+	spin_unlock_irqrestore_rcu_node(sdp, flags);
+	// If needed, requeue ourselves as an expedited SRCU callback.
+	if (needcb)
+		__call_srcu(sdp->ssp, &sdp->srcu_ec_head, srcu_expedite_current_cb, false);
+}
+
+/**
+ * srcu_expedite_current - Expedite the current SRCU grace period
+ * @ssp: srcu_struct to expedite.
+ *
+ * Cause the current SRCU grace period to become expedited.  The grace
+ * period following the current one might also be expedited.  If there is
+ * no current grace period, one might be created.  If the current grace
+ * period is currently sleeping, that sleep will complete before expediting
+ * will take effect.
+ */
+void srcu_expedite_current(struct srcu_struct *ssp)
+{
+	unsigned long flags;
+	bool needcb = false;
+	struct srcu_data *sdp;
+
+	migrate_disable();
+	sdp = this_cpu_ptr(ssp->sda);
+	spin_lock_irqsave_sdp_contention(sdp, &flags);
+	if (sdp->srcu_ec_state == SRCU_EC_IDLE) {
+		sdp->srcu_ec_state = SRCU_EC_PENDING;
+		needcb = true;
+	} else if (sdp->srcu_ec_state == SRCU_EC_PENDING) {
+		sdp->srcu_ec_state = SRCU_EC_REPOST;
+	} else {
+		WARN_ON_ONCE(sdp->srcu_ec_state != SRCU_EC_REPOST);
+	}
+	spin_unlock_irqrestore_rcu_node(sdp, flags);
+	// If needed, queue an expedited SRCU callback.
+	if (needcb)
+		__call_srcu(ssp, &sdp->srcu_ec_head, srcu_expedite_current_cb, false);
+	migrate_enable();
+}
+EXPORT_SYMBOL_GPL(srcu_expedite_current);
+
 /**
  * srcu_batches_completed - return batches completed.
  * @ssp: srcu_struct on which to report batch completion.
-- 
2.40.1

[PATCH v2 07/21] rcutorture: Test srcu_expedite_current()

[Paul E. McKenney]

This commit adds a ->exp_current member to the rcu_torture_ops structure
to test the srcu_expedite_current() function.

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Cc: Andrii Nakryiko <andrii@kernel.org>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: <bpf@vger.kernel.org>
---
 kernel/rcu/rcutorture.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/kernel/rcu/rcutorture.c b/kernel/rcu/rcutorture.c
index 485fa822b6a753..64803d09fc733a 100644
--- a/kernel/rcu/rcutorture.c
+++ b/kernel/rcu/rcutorture.c
@@ -389,6 +389,7 @@ struct rcu_torture_ops {
 	void (*deferred_free)(struct rcu_torture *p);
 	void (*sync)(void);
 	void (*exp_sync)(void);
+	void (*exp_current)(void);
 	unsigned long (*get_gp_state_exp)(void);
 	unsigned long (*start_gp_poll_exp)(void);
 	void (*start_gp_poll_exp_full)(struct rcu_gp_oldstate *rgosp);
@@ -857,6 +858,11 @@ static void srcu_torture_synchronize_expedited(void)
 	synchronize_srcu_expedited(srcu_ctlp);
 }
 
+static void srcu_torture_expedite_current(void)
+{
+	srcu_expedite_current(srcu_ctlp);
+}
+
 static struct rcu_torture_ops srcu_ops = {
 	.ttype		= SRCU_FLAVOR,
 	.init		= rcu_sync_torture_init,
@@ -871,6 +877,7 @@ static struct rcu_torture_ops srcu_ops = {
 	.deferred_free	= srcu_torture_deferred_free,
 	.sync		= srcu_torture_synchronize,
 	.exp_sync	= srcu_torture_synchronize_expedited,
+	.exp_current	= srcu_torture_expedite_current,
 	.same_gp_state	= same_state_synchronize_srcu,
 	.get_comp_state = get_completed_synchronize_srcu,
 	.get_gp_state	= srcu_torture_get_gp_state,
@@ -919,6 +926,7 @@ static struct rcu_torture_ops srcud_ops = {
 	.deferred_free	= srcu_torture_deferred_free,
 	.sync		= srcu_torture_synchronize,
 	.exp_sync	= srcu_torture_synchronize_expedited,
+	.exp_current	= srcu_torture_expedite_current,
 	.same_gp_state	= same_state_synchronize_srcu,
 	.get_comp_state = get_completed_synchronize_srcu,
 	.get_gp_state	= srcu_torture_get_gp_state,
@@ -1698,6 +1706,8 @@ rcu_torture_writer(void *arg)
 					ulo[i] = cur_ops->get_comp_state();
 				gp_snap = cur_ops->start_gp_poll();
 				rcu_torture_writer_state = RTWS_POLL_WAIT;
+				if (cur_ops->exp_current && !torture_random(&rand) % 0xff)
+					cur_ops->exp_current();
 				while (!cur_ops->poll_gp_state(gp_snap)) {
 					gp_snap1 = cur_ops->get_gp_state();
 					for (i = 0; i < ulo_size; i++)
@@ -1718,6 +1728,8 @@ rcu_torture_writer(void *arg)
 					cur_ops->get_comp_state_full(&rgo[i]);
 				cur_ops->start_gp_poll_full(&gp_snap_full);
 				rcu_torture_writer_state = RTWS_POLL_WAIT_FULL;
+				if (cur_ops->exp_current && !torture_random(&rand) % 0xff)
+					cur_ops->exp_current();
 				while (!cur_ops->poll_gp_state_full(&gp_snap_full)) {
 					cur_ops->get_gp_state_full(&gp_snap1_full);
 					for (i = 0; i < rgo_size; i++)
-- 
2.40.1

[PATCH v2 08/21] rcu: Add noinstr-fast rcu_read_{,un}lock_tasks_trace() APIs

[Paul E. McKenney]

When expressing RCU Tasks Trace in terms of SRCU-fast, it was
necessary to keep a nesting count and per-CPU srcu_ctr structure
pointer in the task_struct structure, which is slow to access.
But an alternative is to instead make rcu_read_lock_tasks_trace() and
rcu_read_unlock_tasks_trace(), which match the underlying SRCU-fast
semantics, avoiding the task_struct accesses.

When all callers have switched to the new API, the previous
rcu_read_lock_trace() and rcu_read_unlock_trace() APIs will be removed.

The rcu_read_{,un}lock_{,tasks_}trace() functions need to use smp_mb()
only if invoked where RCU is not watching, that is, from locations where
a call to rcu_is_watching() would return false.  In architectures that
define the ARCH_WANTS_NO_INSTR Kconfig option, use of noinstr and friends
ensures that tracing happens only where RCU is watching, so those
architectures can dispense entirely with the read-side calls to smp_mb().

Other architectures include these read-side calls by default, but in many
installations there might be either larger than average tolerance for
risk, prohibition of removing tracing on a running system, or careful
review and approval of removing of tracing.  Such installations can
build their kernels with CONFIG_TASKS_TRACE_RCU_NO_MB=y to avoid those
read-side calls to smp_mb(), thus accepting responsibility for run-time
removal of tracing from code regions that RCU is not watching.

Those wishing to disable read-side memory barriers for an entire
architecture can select this TASKS_TRACE_RCU_NO_MB Kconfig option,
hence the polarity.

[ paulmck: Apply Peter Zijlstra feedback. ]

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Cc: Andrii Nakryiko <andrii@kernel.org>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: <bpf@vger.kernel.org>
---
 include/linux/rcupdate_trace.h | 65 +++++++++++++++++++++++++++++-----
 kernel/rcu/Kconfig             | 23 ++++++++++++
 kernel/rcu/tasks.h             |  7 +++-
 3 files changed, 86 insertions(+), 9 deletions(-)

diff --git a/include/linux/rcupdate_trace.h b/include/linux/rcupdate_trace.h
index 0bd47f12ecd17b..f47ba9c074601c 100644
--- a/include/linux/rcupdate_trace.h
+++ b/include/linux/rcupdate_trace.h
@@ -34,6 +34,53 @@ static inline int rcu_read_lock_trace_held(void)
 
 #ifdef CONFIG_TASKS_TRACE_RCU
 
+/**
+ * rcu_read_lock_tasks_trace - mark beginning of RCU-trace read-side critical section
+ *
+ * When synchronize_rcu_tasks_trace() is invoked by one task, then that
+ * task is guaranteed to block until all other tasks exit their read-side
+ * critical sections.  Similarly, if call_rcu_trace() is invoked on one
+ * task while other tasks are within RCU read-side critical sections,
+ * invocation of the corresponding RCU callback is deferred until after
+ * the all the other tasks exit their critical sections.
+ *
+ * For more details, please see the documentation for
+ * srcu_read_lock_fast().  For a description of how implicit RCU
+ * readers provide the needed ordering for architectures defining the
+ * ARCH_WANTS_NO_INSTR Kconfig option (and thus promising never to trace
+ * code where RCU is not watching), please see the __srcu_read_lock_fast()
+ * (non-kerneldoc) header comment.  Otherwise, the smp_mb() below provided
+ * the needed ordering.
+ */
+static inline struct srcu_ctr __percpu *rcu_read_lock_tasks_trace(void)
+{
+	struct srcu_ctr __percpu *ret = __srcu_read_lock_fast(&rcu_tasks_trace_srcu_struct);
+
+	rcu_try_lock_acquire(&rcu_tasks_trace_srcu_struct.dep_map);
+	if (!IS_ENABLED(CONFIG_TASKS_TRACE_RCU_NO_MB))
+		smp_mb(); // Provide ordering on noinstr-incomplete architectures.
+	return ret;
+}
+
+/**
+ * rcu_read_unlock_tasks_trace - mark end of RCU-trace read-side critical section
+ * @scp: return value from corresponding rcu_read_lock_tasks_trace().
+ *
+ * Pairs with the preceding call to rcu_read_lock_tasks_trace() that
+ * returned the value passed in via scp.
+ *
+ * For more details, please see the documentation for rcu_read_unlock().
+ * For memory-ordering information, please see the header comment for the
+ * rcu_read_lock_tasks_trace() function.
+ */
+static inline void rcu_read_unlock_tasks_trace(struct srcu_ctr __percpu *scp)
+{
+	if (!IS_ENABLED(CONFIG_TASKS_TRACE_RCU_NO_MB))
+		smp_mb(); // Provide ordering on noinstr-incomplete architectures.
+	__srcu_read_unlock_fast(&rcu_tasks_trace_srcu_struct, scp);
+	srcu_lock_release(&rcu_tasks_trace_srcu_struct.dep_map);
+}
+
 /**
  * rcu_read_lock_trace - mark beginning of RCU-trace read-side critical section
  *
@@ -50,14 +97,15 @@ static inline void rcu_read_lock_trace(void)
 {
 	struct task_struct *t = current;
 
+	rcu_try_lock_acquire(&rcu_tasks_trace_srcu_struct.dep_map);
 	if (t->trc_reader_nesting++) {
 		// In case we interrupted a Tasks Trace RCU reader.
-		rcu_try_lock_acquire(&rcu_tasks_trace_srcu_struct.dep_map);
 		return;
 	}
 	barrier();  // nesting before scp to protect against interrupt handler.
-	t->trc_reader_scp = srcu_read_lock_fast(&rcu_tasks_trace_srcu_struct);
-	smp_mb(); // Placeholder for more selective ordering
+	t->trc_reader_scp = __srcu_read_lock_fast(&rcu_tasks_trace_srcu_struct);
+	if (!IS_ENABLED(CONFIG_TASKS_TRACE_RCU_NO_MB))
+		smp_mb(); // Placeholder for more selective ordering
 }
 
 /**
@@ -74,13 +122,14 @@ static inline void rcu_read_unlock_trace(void)
 	struct srcu_ctr __percpu *scp;
 	struct task_struct *t = current;
 
-	smp_mb(); // Placeholder for more selective ordering
 	scp = t->trc_reader_scp;
 	barrier();  // scp before nesting to protect against interrupt handler.
-	if (!--t->trc_reader_nesting)
-		srcu_read_unlock_fast(&rcu_tasks_trace_srcu_struct, scp);
-	else
-		srcu_lock_release(&rcu_tasks_trace_srcu_struct.dep_map);
+	if (!--t->trc_reader_nesting) {
+		if (!IS_ENABLED(CONFIG_TASKS_TRACE_RCU_NO_MB))
+			smp_mb(); // Placeholder for more selective ordering
+		__srcu_read_unlock_fast(&rcu_tasks_trace_srcu_struct, scp);
+	}
+	srcu_lock_release(&rcu_tasks_trace_srcu_struct.dep_map);
 }
 
 /**
diff --git a/kernel/rcu/Kconfig b/kernel/rcu/Kconfig
index 73a6cc364628b5..6a319e2926589f 100644
--- a/kernel/rcu/Kconfig
+++ b/kernel/rcu/Kconfig
@@ -142,6 +142,29 @@ config TASKS_TRACE_RCU
 	default n
 	select IRQ_WORK
 
+config TASKS_TRACE_RCU_NO_MB
+	bool "Override RCU Tasks Trace inclusion of read-side memory barriers"
+	depends on RCU_EXPERT && TASKS_TRACE_RCU
+	default ARCH_WANTS_NO_INSTR
+	help
+	  This option prevents the use of read-side memory barriers in
+	  rcu_read_lock_tasks_trace() and rcu_read_unlock_tasks_trace()
+	  even in kernels built with CONFIG_ARCH_WANTS_NO_INSTR=n, that is,
+	  in kernels that do not have noinstr set up in entry/exit code.
+	  By setting this option, you are promising to carefully review
+	  use of ftrace, BPF, and friends to ensure that no tracing
+	  operation is attached to a function that runs in that portion
+	  of the entry/exit code that RCU does not watch, that is,
+	  where rcu_is_watching() returns false.  Alternatively, you
+	  might choose to never remove traces except by rebooting.
+
+	  Those wishing to disable read-side memory barriers for an entire
+	  architecture can select this Kconfig option, hence the polarity.
+
+	  Say Y here if you need speed and will review use of tracing.
+	  Say N here for certain esoteric testing of RCU itself.
+	  Take the default if you are unsure.
+
 config RCU_STALL_COMMON
 	def_bool TREE_RCU
 	help
diff --git a/kernel/rcu/tasks.h b/kernel/rcu/tasks.h
index 833e180db744f2..bf1226834c9423 100644
--- a/kernel/rcu/tasks.h
+++ b/kernel/rcu/tasks.h
@@ -1600,8 +1600,13 @@ static inline void rcu_tasks_bootup_oddness(void) {}
 // Tracing variant of Tasks RCU.  This variant is designed to be used
 // to protect tracing hooks, including those of BPF.  This variant
 // is implemented via a straightforward mapping onto SRCU-fast.
+// DEFINE_SRCU_FAST() is required because rcu_read_lock_trace() must
+// use __srcu_read_lock_fast() in order to bypass the rcu_is_watching()
+// checks in kernels built with CONFIG_TASKS_TRACE_RCU_NO_MB=n, which also
+// bypasses the srcu_check_read_flavor_force() that would otherwise mark
+// rcu_tasks_trace_srcu_struct as needing SRCU-fast readers.
 
-DEFINE_SRCU(rcu_tasks_trace_srcu_struct);
+DEFINE_SRCU_FAST(rcu_tasks_trace_srcu_struct);
 EXPORT_SYMBOL_GPL(rcu_tasks_trace_srcu_struct);
 
 #endif /* #else #ifdef CONFIG_TASKS_TRACE_RCU */
-- 
2.40.1

[PATCH v2 09/21] rcu: Update Requirements.rst for RCU Tasks Trace

[Paul E. McKenney]

This commit updates the documentation to declare that RCU Tasks Trace
is implemented as a thin wrapper around SRCU-fast.

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Cc: Andrii Nakryiko <andrii@kernel.org>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: <bpf@vger.kernel.org>
---
 .../RCU/Design/Requirements/Requirements.rst         | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/Documentation/RCU/Design/Requirements/Requirements.rst b/Documentation/RCU/Design/Requirements/Requirements.rst
index f24b3c0b9b0dc6..4a116d7a564edc 100644
--- a/Documentation/RCU/Design/Requirements/Requirements.rst
+++ b/Documentation/RCU/Design/Requirements/Requirements.rst
@@ -2779,12 +2779,12 @@ Tasks Trace RCU
 ~~~~~~~~~~~~~~~
 
 Some forms of tracing need to sleep in readers, but cannot tolerate
-SRCU's read-side overhead, which includes a full memory barrier in both
-srcu_read_lock() and srcu_read_unlock().  This need is handled by a
-Tasks Trace RCU that uses scheduler locking and IPIs to synchronize with
-readers.  Real-time systems that cannot tolerate IPIs may build their
-kernels with ``CONFIG_TASKS_TRACE_RCU_READ_MB=y``, which avoids the IPIs at
-the expense of adding full memory barriers to the read-side primitives.
+SRCU's read-side overhead, which includes a full memory barrier in
+both srcu_read_lock() and srcu_read_unlock().  This need is handled by
+a Tasks Trace RCU API implemented as thin wrappers around SRCU-fast,
+which avoids the read-side memory barriers, at least for architectures
+that apply noinstr to kernel entry/exit code (or that build with
+``CONFIG_TASKS_TRACE_RCU_NO_MB=y``.
 
 The tasks-trace-RCU API is also reasonably compact,
 consisting of rcu_read_lock_trace(), rcu_read_unlock_trace(),
-- 
2.40.1

[PATCH v2 10/21] checkpatch: Deprecate rcu_read_{,un}lock_trace()

[Paul E. McKenney]

Uses of rcu_read_lock_trace() and rcu_read_unlock_trace()
are better served by the new rcu_read_lock_tasks_trace() and
rcu_read_unlock_tasks_trace() APIs.  Therefore, mark the old APIs as
deprecated.

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Acked-by: Joe Perches <joe@perches.com>
Cc: Andy Whitcroft <apw@canonical.com>
Cc: Dwaipayan Ray <dwaipayanray1@gmail.com>
Cc: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Cc: Andrii Nakryiko <andrii@kernel.org>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: <bpf@vger.kernel.org>
---
 scripts/checkpatch.pl | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
index e722dd6fa8ef3d..3bb7d35a5cfcba 100755
--- a/scripts/checkpatch.pl
+++ b/scripts/checkpatch.pl
@@ -860,6 +860,8 @@ our %deprecated_apis = (
 	"kunmap"				=> "kunmap_local",
 	"kmap_atomic"				=> "kmap_local_page",
 	"kunmap_atomic"				=> "kunmap_local",
+	"rcu_read_lock_trace"			=> "rcu_read_lock_tasks_trace",
+	"rcu_read_unlock_trace"			=> "rcu_read_unlock_tasks_trace",
 );
 
 #Create a search pattern for all these strings to speed up a loop below
-- 
2.40.1

[PATCH v2 11/21] srcu: Create a DEFINE_SRCU_FAST()

[Paul E. McKenney]

This commit creates DEFINE_SRCU_FAST() and DEFINE_STATIC_SRCU_FAST()
macros that are similar to DEFINE_SRCU() and DEFINE_STATIC_SRCU(),
but which create srcu_struct structures that are usable only by readers
initiated by srcu_read_lock_fast() and friends.

This commit does make DEFINE_SRCU_FAST() available to modules, in which
case the per-CPU srcu_data structures are not created at compile time, but
rather at module-load time.  This means that the >srcu_reader_flavor field
of the srcu_data structure is not available.  Therefore,
this commit instead creates an ->srcu_reader_flavor field in the
srcu_struct structure, adds arguments to the DEFINE_SRCU()-related
macros to initialize this new field, and extends the checks in the
__srcu_check_read_flavor() function to include this new field.

This commit also allows dynamically allocated srcu_struct structure
to be marked for SRCU-fast readers.  It does so by defining a new
init_srcu_struct_fast() function that marks the specified srcu_struct
structure for use by srcu_read_lock_fast() and friends.

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: <bpf@vger.kernel.org>
---
 include/linux/notifier.h |  2 +-
 include/linux/srcu.h     | 16 ++++++++++++++--
 include/linux/srcutiny.h | 13 ++++++++++---
 include/linux/srcutree.h | 34 +++++++++++++++++++++++-----------
 kernel/rcu/srcutree.c    | 36 ++++++++++++++++++++++++++++++++++--
 5 files changed, 82 insertions(+), 19 deletions(-)

diff --git a/include/linux/notifier.h b/include/linux/notifier.h
index b42e6473496882..01b6c9d9956f90 100644
--- a/include/linux/notifier.h
+++ b/include/linux/notifier.h
@@ -109,7 +109,7 @@ extern void srcu_init_notifier_head(struct srcu_notifier_head *nh);
 		.mutex = __MUTEX_INITIALIZER(name.mutex),	\
 		.head = NULL,					\
 		.srcuu = __SRCU_USAGE_INIT(name.srcuu),		\
-		.srcu = __SRCU_STRUCT_INIT(name.srcu, name.srcuu, pcpu), \
+		.srcu = __SRCU_STRUCT_INIT(name.srcu, name.srcuu, pcpu, 0), \
 	}
 
 #define ATOMIC_NOTIFIER_HEAD(name)				\
diff --git a/include/linux/srcu.h b/include/linux/srcu.h
index ada65b58bc4c5f..26de47820c58cd 100644
--- a/include/linux/srcu.h
+++ b/include/linux/srcu.h
@@ -25,8 +25,10 @@ struct srcu_struct;
 
 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 
-int __init_srcu_struct(struct srcu_struct *ssp, const char *name,
-		       struct lock_class_key *key);
+int __init_srcu_struct(struct srcu_struct *ssp, const char *name, struct lock_class_key *key);
+#ifndef CONFIG_TINY_SRCU
+int __init_srcu_struct_fast(struct srcu_struct *ssp, const char *name, struct lock_class_key *key);
+#endif // #ifndef CONFIG_TINY_SRCU
 
 #define init_srcu_struct(ssp) \
 ({ \
@@ -35,10 +37,20 @@ int __init_srcu_struct(struct srcu_struct *ssp, const char *name,
 	__init_srcu_struct((ssp), #ssp, &__srcu_key); \
 })
 
+#define init_srcu_struct_fast(ssp) \
+({ \
+	static struct lock_class_key __srcu_key; \
+	\
+	__init_srcu_struct_fast((ssp), #ssp, &__srcu_key); \
+})
+
 #define __SRCU_DEP_MAP_INIT(srcu_name)	.dep_map = { .name = #srcu_name },
 #else /* #ifdef CONFIG_DEBUG_LOCK_ALLOC */
 
 int init_srcu_struct(struct srcu_struct *ssp);
+#ifndef CONFIG_TINY_SRCU
+int init_srcu_struct_fast(struct srcu_struct *ssp);
+#endif // #ifndef CONFIG_TINY_SRCU
 
 #define __SRCU_DEP_MAP_INIT(srcu_name)
 #endif /* #else #ifdef CONFIG_DEBUG_LOCK_ALLOC */
diff --git a/include/linux/srcutiny.h b/include/linux/srcutiny.h
index 3bfbd44cb1b37d..92e6ab53398fc0 100644
--- a/include/linux/srcutiny.h
+++ b/include/linux/srcutiny.h
@@ -31,7 +31,7 @@ struct srcu_struct {
 
 void srcu_drive_gp(struct work_struct *wp);
 
-#define __SRCU_STRUCT_INIT(name, __ignored, ___ignored)			\
+#define __SRCU_STRUCT_INIT(name, __ignored, ___ignored, ____ignored)	\
 {									\
 	.srcu_wq = __SWAIT_QUEUE_HEAD_INITIALIZER(name.srcu_wq),	\
 	.srcu_cb_tail = &name.srcu_cb_head,				\
@@ -44,13 +44,20 @@ void srcu_drive_gp(struct work_struct *wp);
  * Tree SRCU, which needs some per-CPU data.
  */
 #define DEFINE_SRCU(name) \
-	struct srcu_struct name = __SRCU_STRUCT_INIT(name, name, name)
+	struct srcu_struct name = __SRCU_STRUCT_INIT(name, name, name, name)
 #define DEFINE_STATIC_SRCU(name) \
-	static struct srcu_struct name = __SRCU_STRUCT_INIT(name, name, name)
+	static struct srcu_struct name = __SRCU_STRUCT_INIT(name, name, name, name)
+#define DEFINE_SRCU_FAST(name) DEFINE_SRCU(name)
+#define DEFINE_STATIC_SRCU_FAST(name) \
+	static struct srcu_struct name = __SRCU_STRUCT_INIT(name, name, name, name)
 
 // Dummy structure for srcu_notifier_head.
 struct srcu_usage { };
 #define __SRCU_USAGE_INIT(name) { }
+#define __init_srcu_struct_fast __init_srcu_struct
+#ifndef CONFIG_DEBUG_LOCK_ALLOC
+#define init_srcu_struct_fast init_srcu_struct
+#endif // #ifndef CONFIG_DEBUG_LOCK_ALLOC
 
 void synchronize_srcu(struct srcu_struct *ssp);
 
diff --git a/include/linux/srcutree.h b/include/linux/srcutree.h
index 93ad18acd6d02d..05400f70baa40a 100644
--- a/include/linux/srcutree.h
+++ b/include/linux/srcutree.h
@@ -104,6 +104,7 @@ struct srcu_usage {
 struct srcu_struct {
 	struct srcu_ctr __percpu *srcu_ctrp;
 	struct srcu_data __percpu *sda;		/* Per-CPU srcu_data array. */
+	u8 srcu_reader_flavor;
 	struct lockdep_map dep_map;
 	struct srcu_usage *srcu_sup;		/* Update-side data. */
 };
@@ -162,20 +163,21 @@ struct srcu_struct {
 	.work = __DELAYED_WORK_INITIALIZER(name.work, NULL, 0),					\
 }
 
-#define __SRCU_STRUCT_INIT_COMMON(name, usage_name)						\
+#define __SRCU_STRUCT_INIT_COMMON(name, usage_name, fast)					\
 	.srcu_sup = &usage_name,								\
+	.srcu_reader_flavor = fast,								\
 	__SRCU_DEP_MAP_INIT(name)
 
-#define __SRCU_STRUCT_INIT_MODULE(name, usage_name)						\
+#define __SRCU_STRUCT_INIT_MODULE(name, usage_name, fast)					\
 {												\
-	__SRCU_STRUCT_INIT_COMMON(name, usage_name)						\
+	__SRCU_STRUCT_INIT_COMMON(name, usage_name, fast)					\
 }
 
-#define __SRCU_STRUCT_INIT(name, usage_name, pcpu_name)						\
+#define __SRCU_STRUCT_INIT(name, usage_name, pcpu_name, fast)					\
 {												\
 	.sda = &pcpu_name,									\
 	.srcu_ctrp = &pcpu_name.srcu_ctrs[0],							\
-	__SRCU_STRUCT_INIT_COMMON(name, usage_name)						\
+	__SRCU_STRUCT_INIT_COMMON(name, usage_name, fast)						\
 }
 
 /*
@@ -196,23 +198,33 @@ struct srcu_struct {
  *	init_srcu_struct(&my_srcu);
  *
  * See include/linux/percpu-defs.h for the rules on per-CPU variables.
+ *
+ * DEFINE_SRCU_FAST() creates an srcu_struct and associated structures
+ * whose readers must be of the SRCU-fast variety.  DEFINE_SRCU_FAST()
+ * cannot be used within modules due to there being no place to
+ * put the desired ->srcu_reader_flavor value.  If in-module use of
+ * DEFINE_SRCU_FAST() becomes necessary, the srcu_struct structure will
+ * need to grow in order to store this value.
  */
 #ifdef MODULE
-# define __DEFINE_SRCU(name, is_static)								\
+# define __DEFINE_SRCU(name, fast, is_static)							\
 	static struct srcu_usage name##_srcu_usage = __SRCU_USAGE_INIT(name##_srcu_usage);	\
-	is_static struct srcu_struct name = __SRCU_STRUCT_INIT_MODULE(name, name##_srcu_usage);	\
+	is_static struct srcu_struct name = __SRCU_STRUCT_INIT_MODULE(name, name##_srcu_usage,	\
+								      fast);			\
 	extern struct srcu_struct * const __srcu_struct_##name;					\
 	struct srcu_struct * const __srcu_struct_##name						\
 		__section("___srcu_struct_ptrs") = &name
 #else
-# define __DEFINE_SRCU(name, is_static)								\
+# define __DEFINE_SRCU(name, fast, is_static)							\
 	static DEFINE_PER_CPU(struct srcu_data, name##_srcu_data);				\
 	static struct srcu_usage name##_srcu_usage = __SRCU_USAGE_INIT(name##_srcu_usage);	\
 	is_static struct srcu_struct name =							\
-		__SRCU_STRUCT_INIT(name, name##_srcu_usage, name##_srcu_data)
+		__SRCU_STRUCT_INIT(name, name##_srcu_usage, name##_srcu_data, fast)
 #endif
-#define DEFINE_SRCU(name)		__DEFINE_SRCU(name, /* not static */)
-#define DEFINE_STATIC_SRCU(name)	__DEFINE_SRCU(name, static)
+#define DEFINE_SRCU(name)		__DEFINE_SRCU(name, 0, /* not static */)
+#define DEFINE_STATIC_SRCU(name)	__DEFINE_SRCU(name, 0, static)
+#define DEFINE_SRCU_FAST(name)		__DEFINE_SRCU(name, SRCU_READ_FLAVOR_FAST, /* not static */)
+#define DEFINE_STATIC_SRCU_FAST(name)	__DEFINE_SRCU(name, SRCU_READ_FLAVOR_FAST, static)
 
 int __srcu_read_lock(struct srcu_struct *ssp) __acquires(ssp);
 void synchronize_srcu_expedited(struct srcu_struct *ssp);
diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
index 38b440b0b0c80b..9869a13b876342 100644
--- a/kernel/rcu/srcutree.c
+++ b/kernel/rcu/srcutree.c
@@ -286,16 +286,29 @@ static int init_srcu_struct_fields(struct srcu_struct *ssp, bool is_static)
 
 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 
-int __init_srcu_struct(struct srcu_struct *ssp, const char *name,
-		       struct lock_class_key *key)
+static int
+__init_srcu_struct_common(struct srcu_struct *ssp, const char *name, struct lock_class_key *key)
 {
 	/* Don't re-initialize a lock while it is held. */
 	debug_check_no_locks_freed((void *)ssp, sizeof(*ssp));
 	lockdep_init_map(&ssp->dep_map, name, key, 0);
 	return init_srcu_struct_fields(ssp, false);
 }
+
+int __init_srcu_struct(struct srcu_struct *ssp, const char *name, struct lock_class_key *key)
+{
+	ssp->srcu_reader_flavor = 0;
+	return __init_srcu_struct_common(ssp, name, key);
+}
 EXPORT_SYMBOL_GPL(__init_srcu_struct);
 
+int __init_srcu_struct_fast(struct srcu_struct *ssp, const char *name, struct lock_class_key *key)
+{
+	ssp->srcu_reader_flavor = SRCU_READ_FLAVOR_FAST;
+	return __init_srcu_struct_common(ssp, name, key);
+}
+EXPORT_SYMBOL_GPL(__init_srcu_struct_fast);
+
 #else /* #ifdef CONFIG_DEBUG_LOCK_ALLOC */
 
 /**
@@ -308,10 +321,26 @@ EXPORT_SYMBOL_GPL(__init_srcu_struct);
  */
 int init_srcu_struct(struct srcu_struct *ssp)
 {
+	ssp->srcu_reader_flavor = 0;
 	return init_srcu_struct_fields(ssp, false);
 }
 EXPORT_SYMBOL_GPL(init_srcu_struct);
 
+/**
+ * init_srcu_struct_fast - initialize a fast-reader sleep-RCU structure
+ * @ssp: structure to initialize.
+ *
+ * Must invoke this on a given srcu_struct before passing that srcu_struct
+ * to any other function.  Each srcu_struct represents a separate domain
+ * of SRCU protection.
+ */
+int init_srcu_struct_fast(struct srcu_struct *ssp)
+{
+	ssp->srcu_reader_flavor = SRCU_READ_FLAVOR_FAST;
+	return init_srcu_struct_fields(ssp, false);
+}
+EXPORT_SYMBOL_GPL(init_srcu_struct_fast);
+
 #endif /* #else #ifdef CONFIG_DEBUG_LOCK_ALLOC */
 
 /*
@@ -734,6 +763,9 @@ void __srcu_check_read_flavor(struct srcu_struct *ssp, int read_flavor)
 
 	sdp = raw_cpu_ptr(ssp->sda);
 	old_read_flavor = READ_ONCE(sdp->srcu_reader_flavor);
+	WARN_ON_ONCE(ssp->srcu_reader_flavor && read_flavor != ssp->srcu_reader_flavor);
+	WARN_ON_ONCE(old_read_flavor && ssp->srcu_reader_flavor &&
+		     old_read_flavor != ssp->srcu_reader_flavor);
 	if (!old_read_flavor) {
 		old_read_flavor = cmpxchg(&sdp->srcu_reader_flavor, 0, read_flavor);
 		if (!old_read_flavor)
-- 
2.40.1

[PATCH v2 12/21] srcu: Create an rcu_tasks_trace_expedite_current() function

[Paul E. McKenney]

This commit creates an rcu_tasks_trace_expedite_current() function
that expedites the current (and possibly the next) RCU Tasks Trace
grace period.

If the current RCU Tasks Trace grace period is already waiting, that wait
will complete before the expediting takes effect.  If there is no RCU
Tasks Trace  grace period in flight, this function might well create one.

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Cc: Andrii Nakryiko <andrii@kernel.org>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: <bpf@vger.kernel.org>
---
 include/linux/rcupdate_trace.h | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/include/linux/rcupdate_trace.h b/include/linux/rcupdate_trace.h
index f47ba9c074601c..cee89e51e45cbb 100644
--- a/include/linux/rcupdate_trace.h
+++ b/include/linux/rcupdate_trace.h
@@ -184,6 +184,20 @@ static inline void rcu_barrier_tasks_trace(void)
 	srcu_barrier(&rcu_tasks_trace_srcu_struct);
 }
 
+/**
+ * rcu_tasks_trace_expedite_current - Expedite the current Tasks Trace RCU grace period
+ *
+ * Cause the current Tasks Trace RCU grace period to become expedited.
+ * The grace period following the current one might also be expedited.
+ * If there is no current grace period, one might be created.  If the
+ * current grace period is currently sleeping, that sleep will complete
+ * before expediting will take effect.
+ */
+static inline void rcu_tasks_trace_expedite_current(void)
+{
+	srcu_expedite_current(&rcu_tasks_trace_srcu_struct);
+}
+
 // Placeholders to enable stepwise transition.
 void __init rcu_tasks_trace_suppress_unused(void);
 
-- 
2.40.1

[PATCH v2 13/21] rcutorture: Test rcu_tasks_trace_expedite_current()

[Paul E. McKenney]

This commit adds a ->exp_current member to the tasks_tracing_ops structure
to test the rcu_tasks_trace_expedite_current() function.

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Cc: Andrii Nakryiko <andrii@kernel.org>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: <bpf@vger.kernel.org>
---
 kernel/rcu/rcutorture.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/kernel/rcu/rcutorture.c b/kernel/rcu/rcutorture.c
index 64803d09fc733a..2e3806b996a80a 100644
--- a/kernel/rcu/rcutorture.c
+++ b/kernel/rcu/rcutorture.c
@@ -1136,6 +1136,7 @@ static struct rcu_torture_ops tasks_tracing_ops = {
 	.deferred_free	= rcu_tasks_tracing_torture_deferred_free,
 	.sync		= synchronize_rcu_tasks_trace,
 	.exp_sync	= synchronize_rcu_tasks_trace,
+	.exp_current	= rcu_tasks_trace_expedite_current,
 	.call		= call_rcu_tasks_trace,
 	.cb_barrier	= rcu_barrier_tasks_trace,
 	.cbflood_max	= 50000,
-- 
2.40.1

[PATCH v2 14/21] srcu: Make grace-period determination use ssp->srcu_reader_flavor

[Paul E. McKenney]

This commit causes the srcu_readers_unlock_idx() function to take the
srcu_struct structure's ->srcu_reader_flavor field into account.  This
ensures that structures defined via DEFINE_SRCU_FAST( or initialized via
init_srcu_struct_fast() have their grace periods use synchronize_srcu()
or synchronize_srcu_expedited() instead of smp_mb(), even before the
first SRCU reader has been entered.

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: <bpf@vger.kernel.org>
---
 kernel/rcu/srcutree.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
index 9869a13b876342..c29203b23d1ad7 100644
--- a/kernel/rcu/srcutree.c
+++ b/kernel/rcu/srcutree.c
@@ -490,7 +490,7 @@ static bool srcu_readers_lock_idx(struct srcu_struct *ssp, int idx, bool gp, uns
 static unsigned long srcu_readers_unlock_idx(struct srcu_struct *ssp, int idx, unsigned long *rdm)
 {
 	int cpu;
-	unsigned long mask = 0;
+	unsigned long mask = ssp->srcu_reader_flavor;
 	unsigned long sum = 0;
 
 	for_each_possible_cpu(cpu) {
-- 
2.40.1

[PATCH v2 15/21] rcutorture: Exercise DEFINE_STATIC_SRCU_FAST() and init_srcu_struct_fast()

[Paul E. McKenney]

This commit updates the initialization for the "srcu" and "srcud" torture
types to use DEFINE_STATIC_SRCU_FAST() and init_srcu_struct_fast(),
respectively, when reader_flavor is equal to SRCU_READ_FLAVOR_FAST.

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: <bpf@vger.kernel.org>
---
 kernel/rcu/rcutorture.c | 19 +++++++++++++++----
 1 file changed, 15 insertions(+), 4 deletions(-)

diff --git a/kernel/rcu/rcutorture.c b/kernel/rcu/rcutorture.c
index 2e3806b996a80a..e859a1ca7dc841 100644
--- a/kernel/rcu/rcutorture.c
+++ b/kernel/rcu/rcutorture.c
@@ -692,10 +692,18 @@ static struct rcu_torture_ops rcu_busted_ops = {
  */
 
 DEFINE_STATIC_SRCU(srcu_ctl);
+DEFINE_STATIC_SRCU_FAST(srcu_ctlf);
 static struct srcu_struct srcu_ctld;
 static struct srcu_struct *srcu_ctlp = &srcu_ctl;
 static struct rcu_torture_ops srcud_ops;
 
+static void srcu_torture_init(void)
+{
+	rcu_sync_torture_init();
+	if (reader_flavor & SRCU_READ_FLAVOR_FAST)
+		srcu_ctlp = &srcu_ctlf;
+}
+
 static void srcu_get_gp_data(int *flags, unsigned long *gp_seq)
 {
 	srcutorture_get_gp_data(srcu_ctlp, flags, gp_seq);
@@ -865,7 +873,7 @@ static void srcu_torture_expedite_current(void)
 
 static struct rcu_torture_ops srcu_ops = {
 	.ttype		= SRCU_FLAVOR,
-	.init		= rcu_sync_torture_init,
+	.init		= srcu_torture_init,
 	.readlock	= srcu_torture_read_lock,
 	.read_delay	= srcu_read_delay,
 	.readunlock	= srcu_torture_read_unlock,
@@ -897,10 +905,13 @@ static struct rcu_torture_ops srcu_ops = {
 	.name		= "srcu"
 };
 
-static void srcu_torture_init(void)
+static void srcud_torture_init(void)
 {
 	rcu_sync_torture_init();
-	WARN_ON(init_srcu_struct(&srcu_ctld));
+	if (reader_flavor & SRCU_READ_FLAVOR_FAST)
+		WARN_ON(init_srcu_struct_fast(&srcu_ctld));
+	else
+		WARN_ON(init_srcu_struct(&srcu_ctld));
 	srcu_ctlp = &srcu_ctld;
 }
 
@@ -913,7 +924,7 @@ static void srcu_torture_cleanup(void)
 /* As above, but dynamically allocated. */
 static struct rcu_torture_ops srcud_ops = {
 	.ttype		= SRCU_FLAVOR,
-	.init		= srcu_torture_init,
+	.init		= srcud_torture_init,
 	.cleanup	= srcu_torture_cleanup,
 	.readlock	= srcu_torture_read_lock,
 	.read_delay	= srcu_read_delay,
-- 
2.40.1

[PATCH v2 16/21] refscale: Exercise DEFINE_STATIC_SRCU_FAST() and init_srcu_struct_fast()

[Paul E. McKenney]

This commit updates the initialization for the "srcu-fast" scale
type to use DEFINE_STATIC_SRCU_FAST() when reader_flavor is equal to
SRCU_READ_FLAVOR_FAST.

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: <bpf@vger.kernel.org>
---
 kernel/rcu/refscale.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/kernel/rcu/refscale.c b/kernel/rcu/refscale.c
index 19841704d8f579..ece77f6d055b85 100644
--- a/kernel/rcu/refscale.c
+++ b/kernel/rcu/refscale.c
@@ -184,6 +184,7 @@ static const struct ref_scale_ops rcu_ops = {
 
 // Definitions for SRCU ref scale testing.
 DEFINE_STATIC_SRCU(srcu_refctl_scale);
+DEFINE_STATIC_SRCU_FAST(srcu_fast_refctl_scale);
 static struct srcu_struct *srcu_ctlp = &srcu_refctl_scale;
 
 static void srcu_ref_scale_read_section(const int nloops)
@@ -216,6 +217,12 @@ static const struct ref_scale_ops srcu_ops = {
 	.name		= "srcu"
 };
 
+static bool srcu_fast_sync_scale_init(void)
+{
+	srcu_ctlp = &srcu_fast_refctl_scale;
+	return true;
+}
+
 static void srcu_fast_ref_scale_read_section(const int nloops)
 {
 	int i;
@@ -240,7 +247,7 @@ static void srcu_fast_ref_scale_delay_section(const int nloops, const int udl, c
 }
 
 static const struct ref_scale_ops srcu_fast_ops = {
-	.init		= rcu_sync_scale_init,
+	.init		= srcu_fast_sync_scale_init,
 	.readsection	= srcu_fast_ref_scale_read_section,
 	.delaysection	= srcu_fast_ref_scale_delay_section,
 	.name		= "srcu-fast"
-- 
2.40.1

[PATCH v2 17/21] srcu: Require special srcu_struct define/init for SRCU-fast readers

[Paul E. McKenney]

This commit adds CONFIG_PROVE_RCU=y checking to enforce the new rule that
srcu_struct structures passed to srcu_read_lock_fast() and other SRCU-fast
read-side markers be either initialized with init_srcu_struct_fast()
on the one hand or defined using either DEFINE_SRCU_FAST() or
DEFINE_STATIC_SRCU_FAST().  This will enable removal of the non-debug
read-side checks from srcu_read_lock_fast() and friends, which on my
laptop provides a 25% speedup (which admittedly amounts to about half
a nanosecond, but when tracing fastpaths...)

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: <bpf@vger.kernel.org>
---
 include/linux/srcu.h  | 34 ++++++++++++++++++++++------------
 kernel/rcu/srcutree.c |  1 +
 2 files changed, 23 insertions(+), 12 deletions(-)

diff --git a/include/linux/srcu.h b/include/linux/srcu.h
index 26de47820c58cd..2982b5a6930fa6 100644
--- a/include/linux/srcu.h
+++ b/include/linux/srcu.h
@@ -271,17 +271,26 @@ static inline int srcu_read_lock(struct srcu_struct *ssp) __acquires(ssp)
  * @ssp: srcu_struct in which to register the new reader.
  *
  * Enter an SRCU read-side critical section, but for a light-weight
- * smp_mb()-free reader.  See srcu_read_lock() for more information.
- *
- * If srcu_read_lock_fast() is ever used on an srcu_struct structure,
- * then none of the other flavors may be used, whether before, during,
- * or after.  Note that grace-period auto-expediting is disabled for _fast
- * srcu_struct structures because auto-expedited grace periods invoke
- * synchronize_rcu_expedited(), IPIs and all.
- *
- * Note that srcu_read_lock_fast() can be invoked only from those contexts
- * where RCU is watching, that is, from contexts where it would be legal
- * to invoke rcu_read_lock().  Otherwise, lockdep will complain.
+ * smp_mb()-free reader.  See srcu_read_lock() for more information.  This
+ * function is NMI-safe, in a manner similar to srcu_read_lock_nmisafe().
+ *
+ * For srcu_read_lock_fast() to be used on an srcu_struct structure,
+ * that structure must have been defined using either DEFINE_SRCU_FAST()
+ * or DEFINE_STATIC_SRCU_FAST() on the one hand or initialized with
+ * init_srcu_struct_fast() on the other.  Such an srcu_struct structure
+ * cannot be passed to any non-fast variant of srcu_read_{,un}lock() or
+ * srcu_{down,up}_read().  In kernels built with CONFIG_PROVE_RCU=y,
+ * __srcu_check_read_flavor() will complain bitterly if you ignore this
+ * restriction.
+ *
+ * Grace-period auto-expediting is disabled for SRCU-fast srcu_struct
+ * structures because SRCU-fast expedited grace periods invoke
+ * synchronize_rcu_expedited(), IPIs and all.  If you need expedited
+ * SRCU-fast grace periods, use synchronize_srcu_expedited().
+ *
+ * The srcu_read_lock_fast() function can be invoked only from those
+ * contexts where RCU is watching, that is, from contexts where it would
+ * be legal to invoke rcu_read_lock().  Otherwise, lockdep will complain.
  */
 static inline struct srcu_ctr __percpu *srcu_read_lock_fast(struct srcu_struct *ssp) __acquires(ssp)
 {
@@ -317,7 +326,8 @@ static inline struct srcu_ctr __percpu *srcu_read_lock_fast_notrace(struct srcu_
  * srcu_down_read() for more information.
  *
  * The same srcu_struct may be used concurrently by srcu_down_read_fast()
- * and srcu_read_lock_fast().
+ * and srcu_read_lock_fast().  However, the same definition/initialization
+ * requirements called out for srcu_read_lock_safe() apply.
  */
 static inline struct srcu_ctr __percpu *srcu_down_read_fast(struct srcu_struct *ssp) __acquires(ssp)
 {
diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
index c29203b23d1ad7..2f8aa280911e19 100644
--- a/kernel/rcu/srcutree.c
+++ b/kernel/rcu/srcutree.c
@@ -766,6 +766,7 @@ void __srcu_check_read_flavor(struct srcu_struct *ssp, int read_flavor)
 	WARN_ON_ONCE(ssp->srcu_reader_flavor && read_flavor != ssp->srcu_reader_flavor);
 	WARN_ON_ONCE(old_read_flavor && ssp->srcu_reader_flavor &&
 		     old_read_flavor != ssp->srcu_reader_flavor);
+	WARN_ON_ONCE(read_flavor == SRCU_READ_FLAVOR_FAST && !ssp->srcu_reader_flavor);
 	if (!old_read_flavor) {
 		old_read_flavor = cmpxchg(&sdp->srcu_reader_flavor, 0, read_flavor);
 		if (!old_read_flavor)
-- 
2.40.1

[PATCH v2 18/21] srcu: Make SRCU-fast readers enforce use of SRCU-fast definition/init

[Paul E. McKenney]

This commit makes CONFIG_PROVE_RCU=y kernels enforce the new rule
that srcu_struct structures that are passed to srcu_read_lock_fast()
and other SRCU-fast read-side markers be either initialized with
init_srcu_struct_fast() on the one hand or defined with DEFINE_SRCU_FAST()
or DEFINE_STATIC_SRCU_FAST() on the other.

This eliminates the read-side test that was formerly included in
srcu_read_lock_fast() and friends, speeding these primitives up by
about 25% (admittedly only about half of a nanosecond, but when tracing
on fastpaths...)

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: <bpf@vger.kernel.org>
---
 include/linux/srcu.h     |  6 +++---
 include/linux/srcutiny.h |  1 -
 include/linux/srcutree.h | 16 +---------------
 kernel/rcu/tasks.h       |  5 -----
 4 files changed, 4 insertions(+), 24 deletions(-)

diff --git a/include/linux/srcu.h b/include/linux/srcu.h
index 2982b5a6930fa6..41e27c1d917d3e 100644
--- a/include/linux/srcu.h
+++ b/include/linux/srcu.h
@@ -297,7 +297,7 @@ static inline struct srcu_ctr __percpu *srcu_read_lock_fast(struct srcu_struct *
 	struct srcu_ctr __percpu *retval;
 
 	RCU_LOCKDEP_WARN(!rcu_is_watching(), "RCU must be watching srcu_read_lock_fast().");
-	srcu_check_read_flavor_force(ssp, SRCU_READ_FLAVOR_FAST);
+	srcu_check_read_flavor(ssp, SRCU_READ_FLAVOR_FAST);
 	retval = __srcu_read_lock_fast(ssp);
 	rcu_try_lock_acquire(&ssp->dep_map);
 	return retval;
@@ -312,7 +312,7 @@ static inline struct srcu_ctr __percpu *srcu_read_lock_fast_notrace(struct srcu_
 {
 	struct srcu_ctr __percpu *retval;
 
-	srcu_check_read_flavor_force(ssp, SRCU_READ_FLAVOR_FAST);
+	srcu_check_read_flavor(ssp, SRCU_READ_FLAVOR_FAST);
 	retval = __srcu_read_lock_fast(ssp);
 	return retval;
 }
@@ -333,7 +333,7 @@ static inline struct srcu_ctr __percpu *srcu_down_read_fast(struct srcu_struct *
 {
 	WARN_ON_ONCE(IS_ENABLED(CONFIG_PROVE_RCU) && in_nmi());
 	RCU_LOCKDEP_WARN(!rcu_is_watching(), "RCU must be watching srcu_down_read_fast().");
-	srcu_check_read_flavor_force(ssp, SRCU_READ_FLAVOR_FAST);
+	srcu_check_read_flavor(ssp, SRCU_READ_FLAVOR_FAST);
 	return __srcu_read_lock_fast(ssp);
 }
 
diff --git a/include/linux/srcutiny.h b/include/linux/srcutiny.h
index 92e6ab53398fc0..1ecc3393fb26be 100644
--- a/include/linux/srcutiny.h
+++ b/include/linux/srcutiny.h
@@ -112,7 +112,6 @@ static inline void srcu_barrier(struct srcu_struct *ssp)
 
 static inline void srcu_expedite_current(struct srcu_struct *ssp) { }
 #define srcu_check_read_flavor(ssp, read_flavor) do { } while (0)
-#define srcu_check_read_flavor_force(ssp, read_flavor) do { } while (0)
 
 /* Defined here to avoid size increase for non-torture kernels. */
 static inline void srcu_torture_stats_print(struct srcu_struct *ssp,
diff --git a/include/linux/srcutree.h b/include/linux/srcutree.h
index 05400f70baa40a..fd24ec146af614 100644
--- a/include/linux/srcutree.h
+++ b/include/linux/srcutree.h
@@ -311,21 +311,7 @@ __srcu_read_unlock_fast(struct srcu_struct *ssp, struct srcu_ctr __percpu *scp)
 
 void __srcu_check_read_flavor(struct srcu_struct *ssp, int read_flavor);
 
-// Record reader usage even for CONFIG_PROVE_RCU=n kernels.  This is
-// needed only for flavors that require grace-period smp_mb() calls to be
-// promoted to synchronize_rcu().
-static inline void srcu_check_read_flavor_force(struct srcu_struct *ssp, int read_flavor)
-{
-	struct srcu_data *sdp = raw_cpu_ptr(ssp->sda);
-
-	if (likely(READ_ONCE(sdp->srcu_reader_flavor) & read_flavor))
-		return;
-
-	// Note that the cmpxchg() in __srcu_check_read_flavor() is fully ordered.
-	__srcu_check_read_flavor(ssp, read_flavor);
-}
-
-// Record non-_lite() usage only for CONFIG_PROVE_RCU=y kernels.
+// Record SRCU-reader usage type only for CONFIG_PROVE_RCU=y kernels.
 static inline void srcu_check_read_flavor(struct srcu_struct *ssp, int read_flavor)
 {
 	if (IS_ENABLED(CONFIG_PROVE_RCU))
diff --git a/kernel/rcu/tasks.h b/kernel/rcu/tasks.h
index bf1226834c9423..76f952196a2921 100644
--- a/kernel/rcu/tasks.h
+++ b/kernel/rcu/tasks.h
@@ -1600,11 +1600,6 @@ static inline void rcu_tasks_bootup_oddness(void) {}
 // Tracing variant of Tasks RCU.  This variant is designed to be used
 // to protect tracing hooks, including those of BPF.  This variant
 // is implemented via a straightforward mapping onto SRCU-fast.
-// DEFINE_SRCU_FAST() is required because rcu_read_lock_trace() must
-// use __srcu_read_lock_fast() in order to bypass the rcu_is_watching()
-// checks in kernels built with CONFIG_TASKS_TRACE_RCU_NO_MB=n, which also
-// bypasses the srcu_check_read_flavor_force() that would otherwise mark
-// rcu_tasks_trace_srcu_struct as needing SRCU-fast readers.
 
 DEFINE_SRCU_FAST(rcu_tasks_trace_srcu_struct);
 EXPORT_SYMBOL_GPL(rcu_tasks_trace_srcu_struct);
-- 
2.40.1

[PATCH v2 19/21] doc: Update for SRCU-fast definitions and initialization

[Paul E. McKenney]

This commit documents the DEFINE_SRCU_FAST(), DEFINE_STATIC_SRCU_FAST(),
and init_srcu_struct_fast() API members.

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: <bpf@vger.kernel.org>
---
 .../RCU/Design/Requirements/Requirements.rst  | 33 ++++++++++---------
 Documentation/RCU/checklist.rst               | 12 ++++---
 Documentation/RCU/whatisRCU.rst               |  3 ++
 3 files changed, 27 insertions(+), 21 deletions(-)

diff --git a/Documentation/RCU/Design/Requirements/Requirements.rst b/Documentation/RCU/Design/Requirements/Requirements.rst
index 4a116d7a564edc..b5cdbba3ec2e71 100644
--- a/Documentation/RCU/Design/Requirements/Requirements.rst
+++ b/Documentation/RCU/Design/Requirements/Requirements.rst
@@ -2637,15 +2637,16 @@ synchronize_srcu() for some other domain ``ss1``, and if an
 that was held across as ``ss``-domain synchronize_srcu(), deadlock
 would again be possible. Such a deadlock cycle could extend across an
 arbitrarily large number of different SRCU domains. Again, with great
-power comes great responsibility.
+power comes great responsibility, though lockdep is now able to detect
+this sort of deadlock.
 
-Unlike the other RCU flavors, SRCU read-side critical sections can run
-on idle and even offline CPUs. This ability requires that
-srcu_read_lock() and srcu_read_unlock() contain memory barriers,
-which means that SRCU readers will run a bit slower than would RCU
-readers. It also motivates the smp_mb__after_srcu_read_unlock() API,
-which, in combination with srcu_read_unlock(), guarantees a full
-memory barrier.
+Unlike the other RCU flavors, SRCU read-side critical sections can run on
+idle and even offline CPUs, with the exception of srcu_read_lock_fast()
+and friends.  This ability requires that srcu_read_lock() and
+srcu_read_unlock() contain memory barriers, which means that SRCU
+readers will run a bit slower than would RCU readers. It also motivates
+the smp_mb__after_srcu_read_unlock() API, which, in combination with
+srcu_read_unlock(), guarantees a full memory barrier.
 
 Also unlike other RCU flavors, synchronize_srcu() may **not** be
 invoked from CPU-hotplug notifiers, due to the fact that SRCU grace
@@ -2681,15 +2682,15 @@ run some tests first. SRCU just might need a few adjustment to deal with
 that sort of load. Of course, your mileage may vary based on the speed
 of your CPUs and the size of your memory.
 
-The `SRCU
-API <https://lwn.net/Articles/609973/#RCU%20Per-Flavor%20API%20Table>`__
+The `SRCU API
+<https://lwn.net/Articles/609973/#RCU%20Per-Flavor%20API%20Table>`__
 includes srcu_read_lock(), srcu_read_unlock(),
-srcu_dereference(), srcu_dereference_check(),
-synchronize_srcu(), synchronize_srcu_expedited(),
-call_srcu(), srcu_barrier(), and srcu_read_lock_held(). It
-also includes DEFINE_SRCU(), DEFINE_STATIC_SRCU(), and
-init_srcu_struct() APIs for defining and initializing
-``srcu_struct`` structures.
+srcu_dereference(), srcu_dereference_check(), synchronize_srcu(),
+synchronize_srcu_expedited(), call_srcu(), srcu_barrier(),
+and srcu_read_lock_held(). It also includes DEFINE_SRCU(),
+DEFINE_STATIC_SRCU(), DEFINE_SRCU_FAST(), DEFINE_STATIC_SRCU_FAST(),
+init_srcu_struct(), and init_srcu_struct_fast() APIs for defining and
+initializing ``srcu_struct`` structures.
 
 More recently, the SRCU API has added polling interfaces:
 
diff --git a/Documentation/RCU/checklist.rst b/Documentation/RCU/checklist.rst
index c9bfb2b218e525..4b30f701225fdb 100644
--- a/Documentation/RCU/checklist.rst
+++ b/Documentation/RCU/checklist.rst
@@ -417,11 +417,13 @@ over a rather long period of time, but improvements are always welcome!
 	you should be using RCU rather than SRCU, because RCU is almost
 	always faster and easier to use than is SRCU.
 
-	Also unlike other forms of RCU, explicit initialization and
-	cleanup is required either at build time via DEFINE_SRCU()
-	or DEFINE_STATIC_SRCU() or at runtime via init_srcu_struct()
-	and cleanup_srcu_struct().  These last two are passed a
-	"struct srcu_struct" that defines the scope of a given
+	Also unlike other forms of RCU, explicit initialization
+	and cleanup is required either at build time via
+	DEFINE_SRCU(), DEFINE_STATIC_SRCU(), DEFINE_SRCU_FAST(),
+	or DEFINE_STATIC_SRCU_FAST() or at runtime via either
+	init_srcu_struct() or init_srcu_struct_fast() and
+	cleanup_srcu_struct().	These last three are passed a
+	`struct srcu_struct` that defines the scope of a given
 	SRCU domain.  Once initialized, the srcu_struct is passed
 	to srcu_read_lock(), srcu_read_unlock() synchronize_srcu(),
 	synchronize_srcu_expedited(), and call_srcu().	A given
diff --git a/Documentation/RCU/whatisRCU.rst b/Documentation/RCU/whatisRCU.rst
index cf0b0ac9f4636a..a1582bd653d115 100644
--- a/Documentation/RCU/whatisRCU.rst
+++ b/Documentation/RCU/whatisRCU.rst
@@ -1227,7 +1227,10 @@ SRCU: Initialization/cleanup/ordering::
 
 	DEFINE_SRCU
 	DEFINE_STATIC_SRCU
+	DEFINE_SRCU_FAST        // for srcu_read_lock_fast() and friends
+	DEFINE_STATIC_SRCU_FAST // for srcu_read_lock_fast() and friends
 	init_srcu_struct
+	init_srcu_struct_fast
 	cleanup_srcu_struct
 	smp_mb__after_srcu_read_unlock
 
-- 
2.40.1

[PATCH v2 20/21] tracing: Guard __DECLARE_TRACE() use of __DO_TRACE_CALL() with SRCU-fast

[Paul E. McKenney]

The current use of guard(preempt_notrace)() within __DECLARE_TRACE()
to protect invocation of __DO_TRACE_CALL() means that BPF programs
attached to tracepoints are non-preemptible.  This is unhelpful in
real-time systems, whose users apparently wish to use BPF while also
achieving low latencies.  (Who knew?)

One option would be to use preemptible RCU, but this introduces
many opportunities for infinite recursion, which many consider to
be counterproductive, especially given the relatively small stacks
provided by the Linux kernel.  These opportunities could be shut down
by sufficiently energetic duplication of code, but this sort of thing
is considered impolite in some circles.

Therefore, use the shiny new SRCU-fast API, which provides somewhat faster
readers than those of preemptible RCU, at least on my laptop, where
task_struct access is more expensive than access to per-CPU variables.
And SRCU fast provides way faster readers than does SRCU, courtesy of
being able to avoid the read-side use of smp_mb().  Also, it is quite
straightforward to create srcu_read_{,un}lock_fast_notrace() functions.

While in the area, SRCU now supports early boot call_srcu().  Therefore,
remove the checks that used to avoid such use from rcu_free_old_probes()
before this commit was applied:

e53244e2c893 ("tracepoint: Remove SRCU protection")

The current commit can be thought of as an approximate revert of that
commit, with some compensating additions of preemption disabling pointed
out by Steven Rostedt (thank you, Steven!).  This preemption disabling
uses guard(preempt_notrace)(), and while in the area a couple of other
use cases were also converted to guards.

However, Yonghong Song points out that BPF expects non-sleepable BPF
programs to remain on the same CPU, which means that migration must
be disabled whenever preemption remains enabled.  In addition, non-RT
kernels have performance expectations on BPF that would be violated
by allowing the BPF programs to be preempted.

Therefore, continue to disable preemption in non-RT kernels, and protect
the BPF program with both SRCU and migration disabling for RT kernels,
and even then only if preemption is not already disabled.

[ paulmck: Apply kernel test robot and Yonghong Song feedback. ]

Link: https://lore.kernel.org/all/20250613152218.1924093-1-bigeasy@linutronix.de/
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: <bpf@vger.kernel.org>
---
 include/linux/tracepoint.h   | 45 ++++++++++++++++++++++--------------
 include/trace/perf.h         |  4 ++--
 include/trace/trace_events.h |  4 ++--
 kernel/tracepoint.c          | 21 ++++++++++++++++-
 4 files changed, 52 insertions(+), 22 deletions(-)

diff --git a/include/linux/tracepoint.h b/include/linux/tracepoint.h
index 826ce3f8e1f851..9f8b19cd303acc 100644
--- a/include/linux/tracepoint.h
+++ b/include/linux/tracepoint.h
@@ -33,6 +33,8 @@ struct trace_eval_map {
 
 #define TRACEPOINT_DEFAULT_PRIO	10
 
+extern struct srcu_struct tracepoint_srcu;
+
 extern int
 tracepoint_probe_register(struct tracepoint *tp, void *probe, void *data);
 extern int
@@ -115,7 +117,10 @@ void for_each_tracepoint_in_module(struct module *mod,
 static inline void tracepoint_synchronize_unregister(void)
 {
 	synchronize_rcu_tasks_trace();
-	synchronize_rcu();
+	if (IS_ENABLED(CONFIG_PREEMPT_RT))
+		synchronize_srcu(&tracepoint_srcu);
+	else
+		synchronize_rcu();
 }
 static inline bool tracepoint_is_faultable(struct tracepoint *tp)
 {
@@ -266,23 +271,29 @@ static inline struct tracepoint *tracepoint_ptr_deref(tracepoint_ptr_t *p)
 		return static_branch_unlikely(&__tracepoint_##name.key);\
 	}
 
-#define __DECLARE_TRACE(name, proto, args, cond, data_proto)		\
+#define __DECLARE_TRACE(name, proto, args, cond, data_proto)			\
 	__DECLARE_TRACE_COMMON(name, PARAMS(proto), PARAMS(args), PARAMS(data_proto)) \
-	static inline void __do_trace_##name(proto)			\
-	{								\
-		if (cond) {						\
-			guard(preempt_notrace)();			\
-			__DO_TRACE_CALL(name, TP_ARGS(args));		\
-		}							\
-	}								\
-	static inline void trace_##name(proto)				\
-	{								\
-		if (static_branch_unlikely(&__tracepoint_##name.key))	\
-			__do_trace_##name(args);			\
-		if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) {		\
-			WARN_ONCE(!rcu_is_watching(),			\
-				  "RCU not watching for tracepoint");	\
-		}							\
+	static inline void __do_trace_##name(proto)				\
+	{									\
+		if (cond) {							\
+			if (IS_ENABLED(CONFIG_PREEMPT_RT) && preemptible()) {	\
+				guard(srcu_fast_notrace)(&tracepoint_srcu);	\
+				guard(migrate)();				\
+				__DO_TRACE_CALL(name, TP_ARGS(args));		\
+			} else {						\
+				guard(preempt_notrace)();			\
+				__DO_TRACE_CALL(name, TP_ARGS(args));		\
+			}							\
+		}								\
+	}									\
+	static inline void trace_##name(proto)					\
+	{									\
+		if (static_branch_unlikely(&__tracepoint_##name.key))		\
+			__do_trace_##name(args);				\
+		if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) {			\
+			WARN_ONCE(!rcu_is_watching(),				\
+				  "RCU not watching for tracepoint");		\
+		}								\
 	}
 
 #define __DECLARE_TRACE_SYSCALL(name, proto, args, data_proto)		\
diff --git a/include/trace/perf.h b/include/trace/perf.h
index a1754b73a8f55b..348ad1d9b5566e 100644
--- a/include/trace/perf.h
+++ b/include/trace/perf.h
@@ -71,6 +71,7 @@ perf_trace_##call(void *__data, proto)					\
 	u64 __count __attribute__((unused));				\
 	struct task_struct *__task __attribute__((unused));		\
 									\
+	guard(preempt_notrace)();					\
 	do_perf_trace_##call(__data, args);				\
 }
 
@@ -85,9 +86,8 @@ perf_trace_##call(void *__data, proto)					\
 	struct task_struct *__task __attribute__((unused));		\
 									\
 	might_fault();							\
-	preempt_disable_notrace();					\
+	guard(preempt_notrace)();					\
 	do_perf_trace_##call(__data, args);				\
-	preempt_enable_notrace();					\
 }
 
 /*
diff --git a/include/trace/trace_events.h b/include/trace/trace_events.h
index 4f22136fd4656c..fbc07d353be6b6 100644
--- a/include/trace/trace_events.h
+++ b/include/trace/trace_events.h
@@ -436,6 +436,7 @@ __DECLARE_EVENT_CLASS(call, PARAMS(proto), PARAMS(args), PARAMS(tstruct), \
 static notrace void							\
 trace_event_raw_event_##call(void *__data, proto)			\
 {									\
+	guard(preempt_notrace)();					\
 	do_trace_event_raw_event_##call(__data, args);			\
 }
 
@@ -447,9 +448,8 @@ static notrace void							\
 trace_event_raw_event_##call(void *__data, proto)			\
 {									\
 	might_fault();							\
-	preempt_disable_notrace();					\
+	guard(preempt_notrace)();					\
 	do_trace_event_raw_event_##call(__data, args);			\
-	preempt_enable_notrace();					\
 }
 
 /*
diff --git a/kernel/tracepoint.c b/kernel/tracepoint.c
index 62719d2941c900..21bb6779821476 100644
--- a/kernel/tracepoint.c
+++ b/kernel/tracepoint.c
@@ -25,6 +25,9 @@ enum tp_func_state {
 extern tracepoint_ptr_t __start___tracepoints_ptrs[];
 extern tracepoint_ptr_t __stop___tracepoints_ptrs[];
 
+DEFINE_SRCU_FAST(tracepoint_srcu);
+EXPORT_SYMBOL_GPL(tracepoint_srcu);
+
 enum tp_transition_sync {
 	TP_TRANSITION_SYNC_1_0_1,
 	TP_TRANSITION_SYNC_N_2_1,
@@ -34,6 +37,7 @@ enum tp_transition_sync {
 
 struct tp_transition_snapshot {
 	unsigned long rcu;
+	unsigned long srcu_gp;
 	bool ongoing;
 };
 
@@ -46,6 +50,7 @@ static void tp_rcu_get_state(enum tp_transition_sync sync)
 
 	/* Keep the latest get_state snapshot. */
 	snapshot->rcu = get_state_synchronize_rcu();
+	snapshot->srcu_gp = start_poll_synchronize_srcu(&tracepoint_srcu);
 	snapshot->ongoing = true;
 }
 
@@ -56,6 +61,8 @@ static void tp_rcu_cond_sync(enum tp_transition_sync sync)
 	if (!snapshot->ongoing)
 		return;
 	cond_synchronize_rcu(snapshot->rcu);
+	if (!poll_state_synchronize_srcu(&tracepoint_srcu, snapshot->srcu_gp))
+		synchronize_srcu(&tracepoint_srcu);
 	snapshot->ongoing = false;
 }
 
@@ -101,17 +108,29 @@ static inline void *allocate_probes(int count)
 	return p == NULL ? NULL : p->probes;
 }
 
-static void rcu_free_old_probes(struct rcu_head *head)
+static void srcu_free_old_probes(struct rcu_head *head)
 {
 	kfree(container_of(head, struct tp_probes, rcu));
 }
 
+static void rcu_free_old_probes(struct rcu_head *head)
+{
+	call_srcu(&tracepoint_srcu, head, srcu_free_old_probes);
+}
+
 static inline void release_probes(struct tracepoint *tp, struct tracepoint_func *old)
 {
 	if (old) {
 		struct tp_probes *tp_probes = container_of(old,
 			struct tp_probes, probes[0]);
 
+		/*
+		 * Tracepoint probes are protected by either RCU or
+		 * Tasks Trace RCU and also by SRCU.  By calling the SRCU
+		 * callback in the [Tasks Trace] RCU callback we cover
+		 * both cases. So let us chain the SRCU and [Tasks Trace]
+		 * RCU callbacks to wait for both grace periods.
+		 */
 		if (tracepoint_is_faultable(tp))
 			call_rcu_tasks_trace(&tp_probes->rcu, rcu_free_old_probes);
 		else
-- 
2.40.1

[PATCH v2 21/21] rcu: Mark diagnostic functions as notrace

[Paul E. McKenney]

The rcu_lockdep_current_cpu_online(), rcu_read_lock_sched_held(),
rcu_read_lock_held(), rcu_read_lock_bh_held(), rcu_read_lock_any_held()
are used by tracing-related code paths, so putting traces on them is
unlikely to make anyone happy.  This commit therefore marks them all
"notrace".

Reported-by: Leon Hwang <leon.hwang@linux.dev>
Reported-by: Alexei Starovoitov <alexei.starovoitov@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
---
 kernel/rcu/tree.c   | 2 +-
 kernel/rcu/update.c | 8 ++++----
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 31690ffa452afe..8ddd07fed36334 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -4021,7 +4021,7 @@ bool rcu_cpu_online(int cpu)
  * RCU on an offline processor during initial boot, hence the check for
  * rcu_scheduler_fully_active.
  */
-bool rcu_lockdep_current_cpu_online(void)
+bool notrace rcu_lockdep_current_cpu_online(void)
 {
 	struct rcu_data *rdp;
 	bool ret = false;
diff --git a/kernel/rcu/update.c b/kernel/rcu/update.c
index c912b594ba987f..dfeba9b3539508 100644
--- a/kernel/rcu/update.c
+++ b/kernel/rcu/update.c
@@ -117,7 +117,7 @@ static bool rcu_read_lock_held_common(bool *ret)
 	return false;
 }
 
-int rcu_read_lock_sched_held(void)
+int notrace rcu_read_lock_sched_held(void)
 {
 	bool ret;
 
@@ -342,7 +342,7 @@ EXPORT_SYMBOL_GPL(debug_lockdep_rcu_enabled);
  * Note that rcu_read_lock() is disallowed if the CPU is either idle or
  * offline from an RCU perspective, so check for those as well.
  */
-int rcu_read_lock_held(void)
+int notrace rcu_read_lock_held(void)
 {
 	bool ret;
 
@@ -367,7 +367,7 @@ EXPORT_SYMBOL_GPL(rcu_read_lock_held);
  * Note that rcu_read_lock_bh() is disallowed if the CPU is either idle or
  * offline from an RCU perspective, so check for those as well.
  */
-int rcu_read_lock_bh_held(void)
+int notrace rcu_read_lock_bh_held(void)
 {
 	bool ret;
 
@@ -377,7 +377,7 @@ int rcu_read_lock_bh_held(void)
 }
 EXPORT_SYMBOL_GPL(rcu_read_lock_bh_held);
 
-int rcu_read_lock_any_held(void)
+int notrace rcu_read_lock_any_held(void)
 {
 	bool ret;
 
-- 
2.40.1

Re: [PATCH v2 08/21] rcu: Add noinstr-fast rcu_read_{,un}lock_tasks_trace() APIs

[Alexei Starovoitov]

On Wed, Oct 1, 2025 at 7:48 AM Paul E. McKenney <paulmck@kernel.org> wrote:
>
> +static inline struct srcu_ctr __percpu *rcu_read_lock_tasks_trace(void)
> +{
> +       struct srcu_ctr __percpu *ret = __srcu_read_lock_fast(&rcu_tasks_trace_srcu_struct);
> +
> +       rcu_try_lock_acquire(&rcu_tasks_trace_srcu_struct.dep_map);
> +       if (!IS_ENABLED(CONFIG_TASKS_TRACE_RCU_NO_MB))
> +               smp_mb(); // Provide ordering on noinstr-incomplete architectures.
> +       return ret;
> +}

...

> @@ -50,14 +97,15 @@ static inline void rcu_read_lock_trace(void)
>  {
>         struct task_struct *t = current;
>
> +       rcu_try_lock_acquire(&rcu_tasks_trace_srcu_struct.dep_map);
>         if (t->trc_reader_nesting++) {
>                 // In case we interrupted a Tasks Trace RCU reader.
> -               rcu_try_lock_acquire(&rcu_tasks_trace_srcu_struct.dep_map);
>                 return;
>         }
>         barrier();  // nesting before scp to protect against interrupt handler.
> -       t->trc_reader_scp = srcu_read_lock_fast(&rcu_tasks_trace_srcu_struct);
> -       smp_mb(); // Placeholder for more selective ordering
> +       t->trc_reader_scp = __srcu_read_lock_fast(&rcu_tasks_trace_srcu_struct);
> +       if (!IS_ENABLED(CONFIG_TASKS_TRACE_RCU_NO_MB))
> +               smp_mb(); // Placeholder for more selective ordering
>  }

Since srcu_fast() __percpu pointers must be incremented/decremented
within the same task, should we expose "raw" rcu_read_lock_tasks_trace()
at all?
rcu_read_lock_trace() stashes that pointer within a task,
so implementation guarantees that unlock will happen within the same task,
while _tasks_trace() requires the user not to do stupid things.

I guess it's fine to have both versions and the amount of copy paste
seems justified, but I keep wondering.
Especially since _tasks_trace() needs more work on bpf trampoline
side to pass this pointer around from lock to unlock.
We can add extra 8 bytes to struct bpf_tramp_run_ctx and save it there,
but set/reset run_ctx operates on current anyway, so it's not clear
which version will be faster. I suspect _trace() will be good enough.
Especially since trc_reader_nesting is kinda an optimization.

Re: [PATCH v2 08/21] rcu: Add noinstr-fast rcu_read_{,un}lock_tasks_trace() APIs

[Paul E. McKenney]

On Wed, Oct 01, 2025 at 06:37:33PM -0700, Alexei Starovoitov wrote:
> On Wed, Oct 1, 2025 at 7:48 AM Paul E. McKenney <paulmck@kernel.org> wrote:
> >
> > +static inline struct srcu_ctr __percpu *rcu_read_lock_tasks_trace(void)
> > +{
> > +       struct srcu_ctr __percpu *ret = __srcu_read_lock_fast(&rcu_tasks_trace_srcu_struct);
> > +
> > +       rcu_try_lock_acquire(&rcu_tasks_trace_srcu_struct.dep_map);
> > +       if (!IS_ENABLED(CONFIG_TASKS_TRACE_RCU_NO_MB))
> > +               smp_mb(); // Provide ordering on noinstr-incomplete architectures.
> > +       return ret;
> > +}
> 
> ...
> 
> > @@ -50,14 +97,15 @@ static inline void rcu_read_lock_trace(void)
> >  {
> >         struct task_struct *t = current;
> >
> > +       rcu_try_lock_acquire(&rcu_tasks_trace_srcu_struct.dep_map);
> >         if (t->trc_reader_nesting++) {
> >                 // In case we interrupted a Tasks Trace RCU reader.
> > -               rcu_try_lock_acquire(&rcu_tasks_trace_srcu_struct.dep_map);
> >                 return;
> >         }
> >         barrier();  // nesting before scp to protect against interrupt handler.
> > -       t->trc_reader_scp = srcu_read_lock_fast(&rcu_tasks_trace_srcu_struct);
> > -       smp_mb(); // Placeholder for more selective ordering
> > +       t->trc_reader_scp = __srcu_read_lock_fast(&rcu_tasks_trace_srcu_struct);
> > +       if (!IS_ENABLED(CONFIG_TASKS_TRACE_RCU_NO_MB))
> > +               smp_mb(); // Placeholder for more selective ordering
> >  }
> 
> Since srcu_fast() __percpu pointers must be incremented/decremented
> within the same task, should we expose "raw" rcu_read_lock_tasks_trace()
> at all?
> rcu_read_lock_trace() stashes that pointer within a task,
> so implementation guarantees that unlock will happen within the same task,
> while _tasks_trace() requires the user not to do stupid things.
> 
> I guess it's fine to have both versions and the amount of copy paste
> seems justified, but I keep wondering.
> Especially since _tasks_trace() needs more work on bpf trampoline
> side to pass this pointer around from lock to unlock.
> We can add extra 8 bytes to struct bpf_tramp_run_ctx and save it there,
> but set/reset run_ctx operates on current anyway, so it's not clear
> which version will be faster. I suspect _trace() will be good enough.
> Especially since trc_reader_nesting is kinda an optimization.

The idea is to convert callers and get rid of rcu_read_lock_trace()
in favor of rcu_read_lock_tasks_trace(), the reason being the slow
task_struct access on x86.  But if the extra storage is an issue for
some use cases, we can keep both.  In that case, I would of course reduce
the copy-pasta in a future patch.

							Thanx, Paul

Re: [PATCH v2 02/21] rcu: Re-implement RCU Tasks Trace in terms of SRCU-fast

[Frederic Weisbecker]

Le Wed, Oct 01, 2025 at 07:48:13AM -0700, Paul E. McKenney a écrit :
> This commit saves more than 500 lines of RCU code by re-implementing
> RCU Tasks Trace in terms of SRCU-fast.  Follow-up work will remove
> more code that does not cause problems by its presence, but that is no
> longer required.
> 
> This variant places smp_mb() in rcu_read_{,un}lock_trace(), which will
> be removed on common-case architectures in a later commit.

The changelog doesn't mention what this is ordering :-)

> 
> [ paulmck: Apply kernel test robot, Boqun Feng, and Zqiang feedback. ]
> [ paulmck: Split out Tiny SRCU fixes per Andrii Nakryiko feedback. ]
> 
> Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
> Tested-by: kernel test robot <oliver.sang@intel.com>
> Cc: Andrii Nakryiko <andrii@kernel.org>
> Cc: Alexei Starovoitov <ast@kernel.org>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Cc: <bpf@vger.kernel.org>
> ---
[...]
> @@ -50,12 +50,14 @@ static inline void rcu_read_lock_trace(void)
>  {
>  	struct task_struct *t = current;
>  
> -	WRITE_ONCE(t->trc_reader_nesting, READ_ONCE(t->trc_reader_nesting) + 1);
> -	barrier();
> -	if (IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB) &&
> -	    t->trc_reader_special.b.need_mb)
> -		smp_mb(); // Pairs with update-side barriers
> -	rcu_lock_acquire(&rcu_trace_lock_map);
> +	if (t->trc_reader_nesting++) {
> +		// In case we interrupted a Tasks Trace RCU reader.
> +		rcu_try_lock_acquire(&rcu_tasks_trace_srcu_struct.dep_map);
> +		return;
> +	}
> +	barrier();  // nesting before scp to protect against interrupt handler.
> +	t->trc_reader_scp = srcu_read_lock_fast(&rcu_tasks_trace_srcu_struct);
> +	smp_mb(); // Placeholder for more selective ordering

Mysterious :-)

>  }
>  
>  /**
> @@ -69,26 +71,75 @@ static inline void rcu_read_lock_trace(void)
>   */
>  static inline void rcu_read_unlock_trace(void)
>  {
> -	int nesting;
> +	struct srcu_ctr __percpu *scp;
>  	struct task_struct *t = current;
>  
> -	rcu_lock_release(&rcu_trace_lock_map);
> -	nesting = READ_ONCE(t->trc_reader_nesting) - 1;
> -	barrier(); // Critical section before disabling.
> -	// Disable IPI-based setting of .need_qs.
> -	WRITE_ONCE(t->trc_reader_nesting, INT_MIN + nesting);
> -	if (likely(!READ_ONCE(t->trc_reader_special.s)) || nesting) {
> -		WRITE_ONCE(t->trc_reader_nesting, nesting);
> -		return;  // We assume shallow reader nesting.
> -	}
> -	WARN_ON_ONCE(nesting != 0);
> -	rcu_read_unlock_trace_special(t);
> +	smp_mb(); // Placeholder for more selective ordering

Bizarre :-)

> +	scp = t->trc_reader_scp;
> +	barrier();  // scp before nesting to protect against interrupt handler.

What is it protecting against interrupt?

> +	if (!--t->trc_reader_nesting)
> +		srcu_read_unlock_fast(&rcu_tasks_trace_srcu_struct, scp);
> +	else
> +		srcu_lock_release(&rcu_tasks_trace_srcu_struct.dep_map);
> +}

Thanks (very happy to see all the rest of the code going away!)

-- 
Frederic Weisbecker
SUSE Labs

Re: [PATCH v2 08/21] rcu: Add noinstr-fast rcu_read_{,un}lock_tasks_trace() APIs

[Alexei Starovoitov]

On Thu, Oct 2, 2025 at 6:38 AM Paul E. McKenney <paulmck@kernel.org> wrote:
>
> On Wed, Oct 01, 2025 at 06:37:33PM -0700, Alexei Starovoitov wrote:
> > On Wed, Oct 1, 2025 at 7:48 AM Paul E. McKenney <paulmck@kernel.org> wrote:
> > >
> > > +static inline struct srcu_ctr __percpu *rcu_read_lock_tasks_trace(void)
> > > +{
> > > +       struct srcu_ctr __percpu *ret = __srcu_read_lock_fast(&rcu_tasks_trace_srcu_struct);
> > > +
> > > +       rcu_try_lock_acquire(&rcu_tasks_trace_srcu_struct.dep_map);
> > > +       if (!IS_ENABLED(CONFIG_TASKS_TRACE_RCU_NO_MB))
> > > +               smp_mb(); // Provide ordering on noinstr-incomplete architectures.
> > > +       return ret;
> > > +}
> >
> > ...
> >
> > > @@ -50,14 +97,15 @@ static inline void rcu_read_lock_trace(void)
> > >  {
> > >         struct task_struct *t = current;
> > >
> > > +       rcu_try_lock_acquire(&rcu_tasks_trace_srcu_struct.dep_map);
> > >         if (t->trc_reader_nesting++) {
> > >                 // In case we interrupted a Tasks Trace RCU reader.
> > > -               rcu_try_lock_acquire(&rcu_tasks_trace_srcu_struct.dep_map);
> > >                 return;
> > >         }
> > >         barrier();  // nesting before scp to protect against interrupt handler.
> > > -       t->trc_reader_scp = srcu_read_lock_fast(&rcu_tasks_trace_srcu_struct);
> > > -       smp_mb(); // Placeholder for more selective ordering
> > > +       t->trc_reader_scp = __srcu_read_lock_fast(&rcu_tasks_trace_srcu_struct);
> > > +       if (!IS_ENABLED(CONFIG_TASKS_TRACE_RCU_NO_MB))
> > > +               smp_mb(); // Placeholder for more selective ordering
> > >  }
> >
> > Since srcu_fast() __percpu pointers must be incremented/decremented
> > within the same task, should we expose "raw" rcu_read_lock_tasks_trace()
> > at all?
> > rcu_read_lock_trace() stashes that pointer within a task,
> > so implementation guarantees that unlock will happen within the same task,
> > while _tasks_trace() requires the user not to do stupid things.
> >
> > I guess it's fine to have both versions and the amount of copy paste
> > seems justified, but I keep wondering.
> > Especially since _tasks_trace() needs more work on bpf trampoline
> > side to pass this pointer around from lock to unlock.
> > We can add extra 8 bytes to struct bpf_tramp_run_ctx and save it there,
> > but set/reset run_ctx operates on current anyway, so it's not clear
> > which version will be faster. I suspect _trace() will be good enough.
> > Especially since trc_reader_nesting is kinda an optimization.
>
> The idea is to convert callers and get rid of rcu_read_lock_trace()
> in favor of rcu_read_lock_tasks_trace(), the reason being the slow
> task_struct access on x86.  But if the extra storage is an issue for
> some use cases, we can keep both.  In that case, I would of course reduce
> the copy-pasta in a future patch.

slow task_struct access on x86? That's news to me.
Why is it slow?
static __always_inline struct task_struct *get_current(void)
{
        if (IS_ENABLED(CONFIG_USE_X86_SEG_SUPPORT))
                return this_cpu_read_const(const_current_task);

        return this_cpu_read_stable(current_task);
}


The former is used with gcc 14+ while later is with clang.
I don't understand the difference between the two.
I'm guessing gcc14+ can be optimized better within the function,
but both look plenty fast.

We need current access anyway for run_ctx.

Re: [PATCH v2 08/21] rcu: Add noinstr-fast rcu_read_{,un}lock_tasks_trace() APIs

[Paul E. McKenney]

On Thu, Oct 02, 2025 at 08:56:01AM -0700, Alexei Starovoitov wrote:
> On Thu, Oct 2, 2025 at 6:38 AM Paul E. McKenney <paulmck@kernel.org> wrote:
> >
> > On Wed, Oct 01, 2025 at 06:37:33PM -0700, Alexei Starovoitov wrote:
> > > On Wed, Oct 1, 2025 at 7:48 AM Paul E. McKenney <paulmck@kernel.org> wrote:
> > > >
> > > > +static inline struct srcu_ctr __percpu *rcu_read_lock_tasks_trace(void)
> > > > +{
> > > > +       struct srcu_ctr __percpu *ret = __srcu_read_lock_fast(&rcu_tasks_trace_srcu_struct);
> > > > +
> > > > +       rcu_try_lock_acquire(&rcu_tasks_trace_srcu_struct.dep_map);
> > > > +       if (!IS_ENABLED(CONFIG_TASKS_TRACE_RCU_NO_MB))
> > > > +               smp_mb(); // Provide ordering on noinstr-incomplete architectures.
> > > > +       return ret;
> > > > +}
> > >
> > > ...
> > >
> > > > @@ -50,14 +97,15 @@ static inline void rcu_read_lock_trace(void)
> > > >  {
> > > >         struct task_struct *t = current;
> > > >
> > > > +       rcu_try_lock_acquire(&rcu_tasks_trace_srcu_struct.dep_map);
> > > >         if (t->trc_reader_nesting++) {
> > > >                 // In case we interrupted a Tasks Trace RCU reader.
> > > > -               rcu_try_lock_acquire(&rcu_tasks_trace_srcu_struct.dep_map);
> > > >                 return;
> > > >         }
> > > >         barrier();  // nesting before scp to protect against interrupt handler.
> > > > -       t->trc_reader_scp = srcu_read_lock_fast(&rcu_tasks_trace_srcu_struct);
> > > > -       smp_mb(); // Placeholder for more selective ordering
> > > > +       t->trc_reader_scp = __srcu_read_lock_fast(&rcu_tasks_trace_srcu_struct);
> > > > +       if (!IS_ENABLED(CONFIG_TASKS_TRACE_RCU_NO_MB))
> > > > +               smp_mb(); // Placeholder for more selective ordering
> > > >  }
> > >
> > > Since srcu_fast() __percpu pointers must be incremented/decremented
> > > within the same task, should we expose "raw" rcu_read_lock_tasks_trace()
> > > at all?
> > > rcu_read_lock_trace() stashes that pointer within a task,
> > > so implementation guarantees that unlock will happen within the same task,
> > > while _tasks_trace() requires the user not to do stupid things.
> > >
> > > I guess it's fine to have both versions and the amount of copy paste
> > > seems justified, but I keep wondering.
> > > Especially since _tasks_trace() needs more work on bpf trampoline
> > > side to pass this pointer around from lock to unlock.
> > > We can add extra 8 bytes to struct bpf_tramp_run_ctx and save it there,
> > > but set/reset run_ctx operates on current anyway, so it's not clear
> > > which version will be faster. I suspect _trace() will be good enough.
> > > Especially since trc_reader_nesting is kinda an optimization.
> >
> > The idea is to convert callers and get rid of rcu_read_lock_trace()
> > in favor of rcu_read_lock_tasks_trace(), the reason being the slow
> > task_struct access on x86.  But if the extra storage is an issue for
> > some use cases, we can keep both.  In that case, I would of course reduce
> > the copy-pasta in a future patch.
> 
> slow task_struct access on x86? That's news to me.
> Why is it slow?
> static __always_inline struct task_struct *get_current(void)
> {
>         if (IS_ENABLED(CONFIG_USE_X86_SEG_SUPPORT))
>                 return this_cpu_read_const(const_current_task);
> 
>         return this_cpu_read_stable(current_task);
> }
> 
> 
> The former is used with gcc 14+ while later is with clang.
> I don't understand the difference between the two.
> I'm guessing gcc14+ can be optimized better within the function,
> but both look plenty fast.
> 
> We need current access anyway for run_ctx.

Last I measured it, task_struct access was quite a bit slower than was
access to per-CPU variables.  The assembly language was such that this
was unsurprising.

But maybe things have changed, and it certainly would be a good thing
if task_struct access had improved.  Once I get done hammering it with
functional tests, I will of course do benchmarking and adjust as needed.

							Thanx, Paul

Re: [PATCH v2 02/21] rcu: Re-implement RCU Tasks Trace in terms of SRCU-fast

[Paul E. McKenney]

On Thu, Oct 02, 2025 at 05:46:10PM +0200, Frederic Weisbecker wrote:
> Le Wed, Oct 01, 2025 at 07:48:13AM -0700, Paul E. McKenney a écrit :
> > This commit saves more than 500 lines of RCU code by re-implementing
> > RCU Tasks Trace in terms of SRCU-fast.  Follow-up work will remove
> > more code that does not cause problems by its presence, but that is no
> > longer required.
> > 
> > This variant places smp_mb() in rcu_read_{,un}lock_trace(), which will
> > be removed on common-case architectures in a later commit.
> 
> The changelog doesn't mention what this is ordering :-)

"The ordering that dare not be named"?  ;-)

How about like this for that second paragraph?

	This variant places smp_mb() in rcu_read_{,un}lock_trace(),
	which will be removed on common-case architectures in a
	later commit.  In the meantime, it serves to enforce ordering
	between the underlying srcu_read_{,un}lock_fast() markers and
	the intervening critical section, even on architectures that
	permit attaching tracepoints on regions of code not watched
	by RCU.  Such architectures defeat SRCU-fast's use of implicit
	single-instruction, interrupts-disabled, and atomic-operation
	RCU read-side critical sections, which have no effect when RCU is
	not watching.  The aforementioned later commit will insert these
	smp_mb() calls only on architectures that have not used noinstr to
	prevent attaching tracepoints to code where RCU is not watching.

> > [ paulmck: Apply kernel test robot, Boqun Feng, and Zqiang feedback. ]
> > [ paulmck: Split out Tiny SRCU fixes per Andrii Nakryiko feedback. ]
> > 
> > Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
> > Tested-by: kernel test robot <oliver.sang@intel.com>
> > Cc: Andrii Nakryiko <andrii@kernel.org>
> > Cc: Alexei Starovoitov <ast@kernel.org>
> > Cc: Peter Zijlstra <peterz@infradead.org>
> > Cc: <bpf@vger.kernel.org>
> > ---
> [...]
> > @@ -50,12 +50,14 @@ static inline void rcu_read_lock_trace(void)
> >  {
> >  	struct task_struct *t = current;
> >  
> > -	WRITE_ONCE(t->trc_reader_nesting, READ_ONCE(t->trc_reader_nesting) + 1);
> > -	barrier();
> > -	if (IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB) &&
> > -	    t->trc_reader_special.b.need_mb)
> > -		smp_mb(); // Pairs with update-side barriers
> > -	rcu_lock_acquire(&rcu_trace_lock_map);
> > +	if (t->trc_reader_nesting++) {
> > +		// In case we interrupted a Tasks Trace RCU reader.
> > +		rcu_try_lock_acquire(&rcu_tasks_trace_srcu_struct.dep_map);
> > +		return;
> > +	}
> > +	barrier();  // nesting before scp to protect against interrupt handler.
> > +	t->trc_reader_scp = srcu_read_lock_fast(&rcu_tasks_trace_srcu_struct);
> > +	smp_mb(); // Placeholder for more selective ordering
> 
> Mysterious :-)

Does the reworked commit-log paragraph help clear up this mystery?

> >  }
> >  
> >  /**
> > @@ -69,26 +71,75 @@ static inline void rcu_read_lock_trace(void)
> >   */
> >  static inline void rcu_read_unlock_trace(void)
> >  {
> > -	int nesting;
> > +	struct srcu_ctr __percpu *scp;
> >  	struct task_struct *t = current;
> >  
> > -	rcu_lock_release(&rcu_trace_lock_map);
> > -	nesting = READ_ONCE(t->trc_reader_nesting) - 1;
> > -	barrier(); // Critical section before disabling.
> > -	// Disable IPI-based setting of .need_qs.
> > -	WRITE_ONCE(t->trc_reader_nesting, INT_MIN + nesting);
> > -	if (likely(!READ_ONCE(t->trc_reader_special.s)) || nesting) {
> > -		WRITE_ONCE(t->trc_reader_nesting, nesting);
> > -		return;  // We assume shallow reader nesting.
> > -	}
> > -	WARN_ON_ONCE(nesting != 0);
> > -	rcu_read_unlock_trace_special(t);
> > +	smp_mb(); // Placeholder for more selective ordering
> 
> Bizarre :-)

And this bizarreness?  ;-)

> > +	scp = t->trc_reader_scp;
> > +	barrier();  // scp before nesting to protect against interrupt handler.
> 
> What is it protecting against interrupt?

The incrementing of ->trc_reader_nesting vs the fetch of ->trc_reader_scp.

							Thanx, Paul

> > +	if (!--t->trc_reader_nesting)
> > +		srcu_read_unlock_fast(&rcu_tasks_trace_srcu_struct, scp);
> > +	else
> > +		srcu_lock_release(&rcu_tasks_trace_srcu_struct.dep_map);
> > +}
> 
> Thanks (very happy to see all the rest of the code going away!)
> 
> -- 
> Frederic Weisbecker
> SUSE Labs

Re: [PATCH v2 02/21] rcu: Re-implement RCU Tasks Trace in terms of SRCU-fast

[Frederic Weisbecker]

Le Sat, Oct 04, 2025 at 02:47:08AM -0700, Paul E. McKenney a écrit :
> On Thu, Oct 02, 2025 at 05:46:10PM +0200, Frederic Weisbecker wrote:
> > Le Wed, Oct 01, 2025 at 07:48:13AM -0700, Paul E. McKenney a écrit :
> > > This commit saves more than 500 lines of RCU code by re-implementing
> > > RCU Tasks Trace in terms of SRCU-fast.  Follow-up work will remove
> > > more code that does not cause problems by its presence, but that is no
> > > longer required.
> > > 
> > > This variant places smp_mb() in rcu_read_{,un}lock_trace(), which will
> > > be removed on common-case architectures in a later commit.
> > 
> > The changelog doesn't mention what this is ordering :-)
> 
> "The ordering that dare not be named"?  ;-)
> 
> How about like this for that second paragraph?
> 
> 	This variant places smp_mb() in rcu_read_{,un}lock_trace(),
> 	which will be removed on common-case architectures in a
> 	later commit.  In the meantime, it serves to enforce ordering
> 	between the underlying srcu_read_{,un}lock_fast() markers and
> 	the intervening critical section, even on architectures that
> 	permit attaching tracepoints on regions of code not watched
> 	by RCU.  Such architectures defeat SRCU-fast's use of implicit
> 	single-instruction, interrupts-disabled, and atomic-operation
> 	RCU read-side critical sections, which have no effect when RCU is
> 	not watching.  The aforementioned later commit will insert these
> 	smp_mb() calls only on architectures that have not used noinstr to
> 	prevent attaching tracepoints to code where RCU is not watching.

Oh I see now. So basically this forces the SRCU-slow behaviour by
restoring the full barriers that are within SRCU-slow's srcu_read_[un]lock()
(can we add a word about that?) for those architectures due to unwatched
RCU sections that can escape the vigilance of the synchronize_rcu() on
the write side.

Thanks.

-- 
Frederic Weisbecker
SUSE Labs

Re: [PATCH v2 02/21] rcu: Re-implement RCU Tasks Trace in terms of SRCU-fast

[Paul E. McKenney]

On Tue, Oct 07, 2025 at 02:10:41PM +0200, Frederic Weisbecker wrote:
> Le Sat, Oct 04, 2025 at 02:47:08AM -0700, Paul E. McKenney a écrit :
> > On Thu, Oct 02, 2025 at 05:46:10PM +0200, Frederic Weisbecker wrote:
> > > Le Wed, Oct 01, 2025 at 07:48:13AM -0700, Paul E. McKenney a écrit :
> > > > This commit saves more than 500 lines of RCU code by re-implementing
> > > > RCU Tasks Trace in terms of SRCU-fast.  Follow-up work will remove
> > > > more code that does not cause problems by its presence, but that is no
> > > > longer required.
> > > > 
> > > > This variant places smp_mb() in rcu_read_{,un}lock_trace(), which will
> > > > be removed on common-case architectures in a later commit.
> > > 
> > > The changelog doesn't mention what this is ordering :-)
> > 
> > "The ordering that dare not be named"?  ;-)
> > 
> > How about like this for that second paragraph?
> > 
> > 	This variant places smp_mb() in rcu_read_{,un}lock_trace(),
> > 	which will be removed on common-case architectures in a
> > 	later commit.  In the meantime, it serves to enforce ordering
> > 	between the underlying srcu_read_{,un}lock_fast() markers and
> > 	the intervening critical section, even on architectures that
> > 	permit attaching tracepoints on regions of code not watched
> > 	by RCU.  Such architectures defeat SRCU-fast's use of implicit
> > 	single-instruction, interrupts-disabled, and atomic-operation
> > 	RCU read-side critical sections, which have no effect when RCU is
> > 	not watching.  The aforementioned later commit will insert these
> > 	smp_mb() calls only on architectures that have not used noinstr to
> > 	prevent attaching tracepoints to code where RCU is not watching.
> 
> Oh I see now. So basically this forces the SRCU-slow behaviour by
> restoring the full barriers that are within SRCU-slow's srcu_read_[un]lock()
> (can we add a word about that?) for those architectures due to unwatched
> RCU sections that can escape the vigilance of the synchronize_rcu() on
> the write side.

You got it!  I will add the connection to old-school srcu_read_[un]lock()
on my next rebase.

							Thanx, Paul

> Thanks.
> 
> -- 
> Frederic Weisbecker
> SUSE Labs