sched_ext, rcu: Upgrade RCU stall paths to report cpumask of stalled CPUs

scx_rcu_cpu_stall() previously recorded the detector CPU rather than the
stalled one, and the expedited grace period path had no stalled CPU to
report at all.

Thread a cpumask through panic_on_rcu_stall() and scx_rcu_cpu_stall()
to capture all stalled CPUs. Report cpumask_first() as exit_cpu and the
full CPU list in the exit message. Task-only stalls yield exit_cpu = -1.

Store the stall mask in scx_sched rather than scx_exit_info, keeping the
BPF-visible struct unchanged. scx_dump_state() reads sch->stall_cpus
directly and dumps all stalled CPUs first to avoid losing them to
truncation.

Signed-off-by: Cheng-Yang Chou <yphbchou0911@gmail.com>
Reviewed-by: Paul E. McKenney <paulmck@kernel.org>
Reviewed-by: Andrea Righi <arighi@nvidia.com>
Signed-off-by: Tejun Heo <tj@kernel.org>

6 files changed, 83 insertions, 16 deletions

diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index 20b2343aa344..75cb8b119fb7 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -263,7 +263,7 @@ void sched_ext_dead(struct task_struct *p);
 void print_scx_info(const char *log_lvl, struct task_struct *p);
 void scx_softlockup(u32 dur_s);
 bool scx_hardlockup(int cpu);
-bool scx_rcu_cpu_stall(void);
+bool scx_rcu_cpu_stall(const struct cpumask *stalled_mask);
 
 #else	/* !CONFIG_SCHED_CLASS_EXT */
 
@@ -271,7 +271,7 @@ static inline void sched_ext_dead(struct task_struct *p) {}
 static inline void print_scx_info(const char *log_lvl, struct task_struct *p) {}
 static inline void scx_softlockup(u32 dur_s) {}
 static inline bool scx_hardlockup(int cpu) { return false; }
-static inline bool scx_rcu_cpu_stall(void) { return false; }
+static inline bool scx_rcu_cpu_stall(const struct cpumask *stalled_mask) { return false; }
 
 #endif	/* CONFIG_SCHED_CLASS_EXT */
 
diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 03a43d3d2616..415583c35f8c 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -4900,6 +4900,9 @@ static void __init rcu_dump_rcu_node_tree(void)
 
 struct workqueue_struct *rcu_gp_wq;
 
+static struct cpumask rcu_stall_cpumask;
+static struct cpumask rcu_exp_stall_cpumask;
+
 void __init rcu_init(void)
 {
 	int cpu = smp_processor_id();
diff --git a/kernel/rcu/tree_exp.h b/kernel/rcu/tree_exp.h
index 82cada459e5d..46b6907f1b09 100644
--- a/kernel/rcu/tree_exp.h
+++ b/kernel/rcu/tree_exp.h
@@ -578,6 +578,7 @@ static void synchronize_rcu_expedited_stall(unsigned long jiffies_start, unsigne
 			if (!(READ_ONCE(rnp->expmask) & mask))
 				continue;
 			ndetected++;
+			cpumask_set_cpu(cpu, &rcu_exp_stall_cpumask);
 			rdp = per_cpu_ptr(&rcu_data, cpu);
 			pr_cont(" %d-%c%c%c%c", cpu,
 				"O."[!!cpu_online(cpu)],
@@ -665,6 +666,8 @@ static void synchronize_rcu_expedited_wait(void)
 		if (rcu_stall_is_suppressed())
 			continue;
 
+		cpumask_clear(&rcu_exp_stall_cpumask);
+
 		nbcon_cpu_emergency_enter();
 
 		j = jiffies;
@@ -675,7 +678,7 @@ static void synchronize_rcu_expedited_wait(void)
 
 		nbcon_cpu_emergency_exit();
 
-		panic_on_rcu_stall();
+		panic_on_rcu_stall(&rcu_exp_stall_cpumask);
 	}
 }
 
diff --git a/kernel/rcu/tree_stall.h b/kernel/rcu/tree_stall.h
index cf7ae51cba40..ebf381936eb1 100644
--- a/kernel/rcu/tree_stall.h
+++ b/kernel/rcu/tree_stall.h
@@ -159,7 +159,7 @@ static int __init check_cpu_stall_init(void)
 early_initcall(check_cpu_stall_init);
 
 /* If so specified via sysctl, panic, yielding cleaner stall-warning output. */
-static void panic_on_rcu_stall(void)
+static void panic_on_rcu_stall(const struct cpumask *stalled_mask)
 {
 	static int cpu_stall;
 
@@ -167,7 +167,7 @@ static void panic_on_rcu_stall(void)
 	 * Attempt to kick out the BPF scheduler if it's installed and defer
 	 * the panic to give the system a chance to recover.
 	 */
-	if (scx_rcu_cpu_stall())
+	if (scx_rcu_cpu_stall(stalled_mask))
 		return;
 
 	if (++cpu_stall < sysctl_max_rcu_stall_to_panic)
@@ -644,6 +644,8 @@ static void print_other_cpu_stall(unsigned long gp_seq, unsigned long gps)
 	if (rcu_stall_is_suppressed())
 		return;
 
+	cpumask_clear(&rcu_stall_cpumask);
+
 	nbcon_cpu_emergency_enter();
 
 	/*
@@ -659,6 +661,7 @@ static void print_other_cpu_stall(unsigned long gp_seq, unsigned long gps)
 			for_each_leaf_node_possible_cpu(rnp, cpu)
 				if (rnp->qsmask & leaf_node_cpu_bit(rnp, cpu)) {
 					print_cpu_stall_info(cpu);
+					cpumask_set_cpu(cpu, &rcu_stall_cpumask);
 					ndetected++;
 				}
 		}
@@ -700,7 +703,7 @@ static void print_other_cpu_stall(unsigned long gp_seq, unsigned long gps)
 
 	nbcon_cpu_emergency_exit();
 
-	panic_on_rcu_stall();
+	panic_on_rcu_stall(&rcu_stall_cpumask);
 
 	rcu_force_quiescent_state();  /* Kick them all. */
 }
@@ -753,7 +756,9 @@ static void print_cpu_stall(unsigned long gp_seq, unsigned long gps)
 
 	nbcon_cpu_emergency_exit();
 
-	panic_on_rcu_stall();
+	cpumask_clear(&rcu_stall_cpumask);
+	cpumask_set_cpu(smp_processor_id(), &rcu_stall_cpumask);
+	panic_on_rcu_stall(&rcu_stall_cpumask);
 
 	/*
 	 * Attempt to revive the RCU machinery by forcing a context switch.
diff --git a/kernel/sched/ext/ext.c b/kernel/sched/ext/ext.c
index 7044a591e4c5..9c9cb9d08bca 100644
--- a/kernel/sched/ext/ext.c
+++ b/kernel/sched/ext/ext.c
@@ -4853,6 +4853,8 @@ static const struct attribute_group scx_global_attr_group = {
 
 static void free_pnode(struct scx_sched_pnode *pnode);
 static void free_exit_info(struct scx_exit_info *ei);
+static const char *scx_exit_reason(enum scx_exit_kind kind);
+static bool scx_claim_exit(struct scx_sched *sch, enum scx_exit_kind kind);
 
 static s32 scx_set_cmask_scratch_alloc(struct scx_sched *sch)
 {
@@ -4909,6 +4911,7 @@ static void scx_sched_free_rcu_work(struct work_struct *work)
 	timer_shutdown_sync(&sch->bypass_lb_timer);
 	free_cpumask_var(sch->bypass_lb_donee_cpumask);
 	free_cpumask_var(sch->bypass_lb_resched_cpumask);
+	free_cpumask_var(sch->stall_cpus);
 
 #ifdef CONFIG_EXT_SUB_SCHED
 	kfree(sch->cgrp_path);
@@ -5138,9 +5141,46 @@ static __printf(2, 3) bool handle_lockup(int exit_cpu, const char *fmt, ...)
  * resolve the reported RCU stall. %false if sched_ext is not enabled or someone
  * else already initiated abort.
  */
-bool scx_rcu_cpu_stall(void)
+bool scx_rcu_cpu_stall(const struct cpumask *stalled_mask)
 {
-	return handle_lockup(-1, "RCU CPU stall detected!");
+	struct scx_sched *sch;
+	struct scx_exit_info *ei;
+	int exit_cpu;
+
+	guard(rcu)();
+
+	sch = rcu_dereference(scx_root);
+	if (unlikely(!sch))
+		return false;
+
+	switch (scx_enable_state()) {
+	case SCX_ENABLING:
+	case SCX_ENABLED:
+		break;
+	default:
+		return false;
+	}
+
+	exit_cpu = cpumask_empty(stalled_mask) ? -1 : (int)cpumask_first(stalled_mask);
+	ei = sch->exit_info;
+
+	guard(preempt)();
+
+	if (!scx_claim_exit(sch, SCX_EXIT_ERROR))
+		return false;
+
+#ifdef CONFIG_STACKTRACE
+	ei->bt_len = stack_trace_save(ei->bt, SCX_EXIT_BT_LEN, 1);
+#endif
+	scnprintf(ei->msg, SCX_EXIT_MSG_LEN, "RCU CPU stall on CPUs (%*pbl)",
+		  cpumask_pr_args(stalled_mask));
+	ei->kind = SCX_EXIT_ERROR;
+	ei->reason = scx_exit_reason(SCX_EXIT_ERROR);
+	ei->exit_cpu = exit_cpu;
+	cpumask_copy(sch->stall_cpus, stalled_mask);
+
+	irq_work_queue(&sch->disable_irq_work);
+	return true;
 }
 
 /**
@@ -6587,14 +6627,23 @@ static void scx_dump_state(struct scx_sched *sch, struct scx_exit_info *ei,
 	dump_line(&s, "----------");
 
 	/*
-	 * Dump the exit CPU first so it isn't lost to dump truncation, then
-	 * walk the rest in order, skipping the one already dumped.
+	 * Dump stalled CPUs first so they aren't lost to dump truncation, then
+	 * walk the rest in order. Fall back to exit_cpu if no stall mask set.
 	 */
-	if (ei->exit_cpu >= 0)
-		scx_dump_cpu(sch, &s, &dctx, ei->exit_cpu, dump_all_tasks);
-	for_each_possible_cpu(cpu) {
-		if (cpu != ei->exit_cpu)
+	if (!cpumask_empty(sch->stall_cpus)) {
+		for_each_cpu(cpu, sch->stall_cpus)
 			scx_dump_cpu(sch, &s, &dctx, cpu, dump_all_tasks);
+		for_each_possible_cpu(cpu) {
+			if (!cpumask_test_cpu(cpu, sch->stall_cpus))
+				scx_dump_cpu(sch, &s, &dctx, cpu, dump_all_tasks);
+		}
+	} else {
+		if (ei->exit_cpu >= 0)
+			scx_dump_cpu(sch, &s, &dctx, ei->exit_cpu, dump_all_tasks);
+		for_each_possible_cpu(cpu) {
+			if (cpu != ei->exit_cpu)
+				scx_dump_cpu(sch, &s, &dctx, cpu, dump_all_tasks);
+		}
 	}
 
 	dump_newline(&s);
@@ -6831,6 +6880,10 @@ static struct scx_sched *scx_alloc_and_add_sched(struct scx_enable_cmd *cmd,
 		ret = -ENOMEM;
 		goto err_free_lb_cpumask;
 	}
+	if (!zalloc_cpumask_var(&sch->stall_cpus, GFP_KERNEL)) {
+		ret = -ENOMEM;
+		goto err_free_lb_resched_cpumask;
+	}
 	/*
 	 * Copy ops through the right union view. For cid-form the source is
 	 * struct sched_ext_ops_cid which lacks the trailing cpu_acquire/
@@ -6914,8 +6967,10 @@ static struct scx_sched *scx_alloc_and_add_sched(struct scx_enable_cmd *cmd,
 #ifdef CONFIG_EXT_SUB_SCHED
 err_free_lb_resched:
 	RCU_INIT_POINTER(ops->priv, NULL);
-	free_cpumask_var(sch->bypass_lb_resched_cpumask);
+	free_cpumask_var(sch->stall_cpus);
 #endif
+err_free_lb_resched_cpumask:
+	free_cpumask_var(sch->bypass_lb_resched_cpumask);
 err_free_lb_cpumask:
 	free_cpumask_var(sch->bypass_lb_donee_cpumask);
 err_stop_helper:
diff --git a/kernel/sched/ext/internal.h b/kernel/sched/ext/internal.h
index 65eceefcf5e2..75522a5f28f4 100644
--- a/kernel/sched/ext/internal.h
+++ b/kernel/sched/ext/internal.h
@@ -1205,6 +1205,7 @@ struct scx_sched {
 	struct timer_list	bypass_lb_timer;
 	cpumask_var_t		bypass_lb_donee_cpumask;
 	cpumask_var_t		bypass_lb_resched_cpumask;
+	cpumask_var_t		stall_cpus;
 	struct rcu_work		rcu_work;
 
 	/* all ancestors including self */