RE: rcu_preempt caused oom

[He, Bo]

I don't see the rcutree.sysrq_rcu parameter in v4.19 kernel, I also checked the latest kernel and the latest tag v4.20-rc6, not see the sysrq_rcu.
Please correct me if I have something wrong.

-----Original Message-----
From: Paul E. McKenney <paulmck@xxxxxxxxxxxxx> 
Sent: Thursday, December 13, 2018 5:03 AM
To: He, Bo <bo.he@xxxxxxxxx>
Cc: Steven Rostedt <rostedt@xxxxxxxxxxx>; linux-kernel@xxxxxxxxxxxxxxx; josh@xxxxxxxxxxxxxxxx; mathieu.desnoyers@xxxxxxxxxxxx; jiangshanlai@xxxxxxxxx; Zhang, Jun <jun.zhang@xxxxxxxxx>; Xiao, Jin <jin.xiao@xxxxxxxxx>; Zhang, Yanmin <yanmin.zhang@xxxxxxxxx>; Bai, Jie A <jie.a.bai@xxxxxxxxx>
Subject: Re: rcu_preempt caused oom

On Wed, Dec 12, 2018 at 07:42:24AM -0800, Paul E. McKenney wrote:
> On Wed, Dec 12, 2018 at 01:21:33PM +0000, He, Bo wrote:
> > we reproduce on two boards, but I still not see the show_rcu_gp_kthreads() dump logs, it seems the patch can't catch the scenario.
> > I double confirmed the CONFIG_PROVE_RCU=y is enabled in the config as it's extracted from the /proc/config.gz.
> 
> Strange.
> 
> Are the systems responsive to sysrq keys once failure occurs?  If so, 
> I will provide you a sysrq-R or some such to dump out the RCU state.

Or, as it turns out, sysrq-y if booting with rcutree.sysrq_rcu=1 using the patch below.  Only lightly tested.

							Thanx, Paul

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

commit adfc7dff659495a3433d5084256be59eee0ac6df
Author: Paul E. McKenney <paulmck@xxxxxxxxxxxxx>
Date:   Mon Dec 10 16:33:59 2018 -0800

    rcu: Improve diagnostics for failed RCU grace-period start
    
    Backported from v4.21/v5.0
    
    If a grace period fails to start (for example, because you commented
    out the last two lines of rcu_accelerate_cbs_unlocked()), rcu_core()
    will invoke rcu_check_gp_start_stall(), which will notice and complain.
    However, this complaint is lacking crucial debugging information such
    as when the last wakeup executed and what the value of ->gp_seq was at
    that time.  This commit therefore removes the current pr_alert() from
    rcu_check_gp_start_stall(), instead invoking show_rcu_gp_kthreads(),
    which has been updated to print the needed information, which is collected
    by rcu_gp_kthread_wake().
    
    Signed-off-by: Paul E. McKenney <paulmck@xxxxxxxxxxxxx>

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index 0b760c1369f7..4bcd8753e293 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -626,25 +626,57 @@ void rcu_sched_force_quiescent_state(void)
 }
 EXPORT_SYMBOL_GPL(rcu_sched_force_quiescent_state);
 
+/*
+ * Convert a ->gp_state value to a character string.
+ */
+static const char *gp_state_getname(short gs) {
+	if (gs < 0 || gs >= ARRAY_SIZE(gp_state_names))
+		return "???";
+	return gp_state_names[gs];
+}
+
+/*
+ * Return the root node of the specified rcu_state structure.
+ */
+static struct rcu_node *rcu_get_root(struct rcu_state *rsp) {
+	return &rsp->node[0];
+}
+
 /*
  * Show the state of the grace-period kthreads.
  */
 void show_rcu_gp_kthreads(void)
 {
 	int cpu;
+	unsigned long j;
+	unsigned long ja;
+	unsigned long jr;
+	unsigned long jw;
 	struct rcu_data *rdp;
 	struct rcu_node *rnp;
 	struct rcu_state *rsp;
 
+	j = jiffies;
 	for_each_rcu_flavor(rsp) {
-		pr_info("%s: wait state: %d ->state: %#lx\n",
-			rsp->name, rsp->gp_state, rsp->gp_kthread->state);
+		ja = j - READ_ONCE(rsp->gp_activity);
+		jr = j - READ_ONCE(rsp->gp_req_activity);
+		jw = j - READ_ONCE(rsp->gp_wake_time);
+		pr_info("%s: wait state: %s(%d) ->state: %#lx delta ->gp_activity %lu ->gp_req_activity %lu ->gp_wake_time %lu ->gp_wake_seq %ld ->gp_seq %ld ->gp_seq_needed %ld ->gp_flags %#x\n",
+			rsp->name, gp_state_getname(rsp->gp_state),
+			rsp->gp_state,
+			rsp->gp_kthread ? rsp->gp_kthread->state : 0x1ffffL,
+			ja, jr, jw, (long)READ_ONCE(rsp->gp_wake_seq),
+			(long)READ_ONCE(rsp->gp_seq),
+			(long)READ_ONCE(rcu_get_root(rsp)->gp_seq_needed),
+			READ_ONCE(rsp->gp_flags));
 		rcu_for_each_node_breadth_first(rsp, rnp) {
 			if (ULONG_CMP_GE(rsp->gp_seq, rnp->gp_seq_needed))
 				continue;
-			pr_info("\trcu_node %d:%d ->gp_seq %lu ->gp_seq_needed %lu\n",
-				rnp->grplo, rnp->grphi, rnp->gp_seq,
-				rnp->gp_seq_needed);
+			pr_info("\trcu_node %d:%d ->gp_seq %ld ->gp_seq_needed %ld\n",
+				rnp->grplo, rnp->grphi, (long)rnp->gp_seq,
+				(long)rnp->gp_seq_needed);
 			if (!rcu_is_leaf_node(rnp))
 				continue;
 			for_each_leaf_node_possible_cpu(rnp, cpu) { @@ -653,8 +685,8 @@ void show_rcu_gp_kthreads(void)
 				    ULONG_CMP_GE(rsp->gp_seq,
 						 rdp->gp_seq_needed))
 					continue;
-				pr_info("\tcpu %d ->gp_seq_needed %lu\n",
-					cpu, rdp->gp_seq_needed);
+				pr_info("\tcpu %d ->gp_seq_needed %ld\n",
+					cpu, (long)rdp->gp_seq_needed);
 			}
 		}
 		/* sched_show_task(rsp->gp_kthread); */ @@ -690,14 +722,6 @@ void rcutorture_get_gp_data(enum rcutorture_type test_type, int *flags,  }  EXPORT_SYMBOL_GPL(rcutorture_get_gp_data);
 
-/*
- * Return the root node of the specified rcu_state structure.
- */
-static struct rcu_node *rcu_get_root(struct rcu_state *rsp) -{
-	return &rsp->node[0];
-}
-
 /*
  * Enter an RCU extended quiescent state, which can be either the
  * idle loop or adaptive-tickless usermode execution.
@@ -1285,16 +1309,6 @@ static void record_gp_stall_check_time(struct rcu_state *rsp)
 	rsp->n_force_qs_gpstart = READ_ONCE(rsp->n_force_qs);  }
 
-/*
- * Convert a ->gp_state value to a character string.
- */
-static const char *gp_state_getname(short gs) -{
-	if (gs < 0 || gs >= ARRAY_SIZE(gp_state_names))
-		return "???";
-	return gp_state_names[gs];
-}
-
 /*
  * Complain about starvation of grace-period kthread.
  */
@@ -1693,7 +1707,8 @@ static bool rcu_future_gp_cleanup(struct rcu_state *rsp, struct rcu_node *rnp)
  * Don't do a self-awaken, and don't bother awakening when there is
  * nothing for the grace-period kthread to do (as in several CPUs
  * raced to awaken, and we lost), and finally don't try to awaken
- * a kthread that has not yet been created.
+ * a kthread that has not yet been created.  If all those checks are
+ * passed, track some debug information and awaken.
  */
 static void rcu_gp_kthread_wake(struct rcu_state *rsp)  { @@ -1701,6 +1716,8 @@ static void rcu_gp_kthread_wake(struct rcu_state *rsp)
 	    !READ_ONCE(rsp->gp_flags) ||
 	    !rsp->gp_kthread)
 		return;
+	WRITE_ONCE(rsp->gp_wake_time, jiffies);
+	WRITE_ONCE(rsp->gp_wake_seq, READ_ONCE(rsp->gp_seq));
 	swake_up_one(&rsp->gp_wq);
 }
 
@@ -2802,16 +2819,11 @@ rcu_check_gp_start_stall(struct rcu_state *rsp, struct rcu_node *rnp,
 		raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
 		return;
 	}
-	pr_alert("%s: g%ld->%ld gar:%lu ga:%lu f%#x gs:%d %s->state:%#lx\n",
-		 __func__, (long)READ_ONCE(rsp->gp_seq),
-		 (long)READ_ONCE(rnp_root->gp_seq_needed),
-		 j - rsp->gp_req_activity, j - rsp->gp_activity,
-		 rsp->gp_flags, rsp->gp_state, rsp->name,
-		 rsp->gp_kthread ? rsp->gp_kthread->state : 0x1ffffL);
 	WARN_ON(1);
 	if (rnp_root != rnp)
 		raw_spin_unlock_rcu_node(rnp_root);
 	raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
+	show_rcu_gp_kthreads();
 }
 
 /*
diff --git a/kernel/rcu/tree.h b/kernel/rcu/tree.h index 4e74df768c57..0e051d9b5f1a 100644
--- a/kernel/rcu/tree.h
+++ b/kernel/rcu/tree.h
@@ -327,6 +327,8 @@ struct rcu_state {
 	struct swait_queue_head gp_wq;		/* Where GP task waits. */
 	short gp_flags;				/* Commands for GP task. */
 	short gp_state;				/* GP kthread sleep state. */
+	unsigned long gp_wake_time;		/* Last GP kthread wake. */
+	unsigned long gp_wake_seq;		/* ->gp_seq at ^^^. */
 
 	/* End of fields guarded by root rcu_node's lock. */