[PATCH rcu 01/23] rcu: Dump rcuc kthread status for CPUs not reporting quiescent state

From: Paul E. McKenney
Date: Mon Jun 20 2022 - 19:12:36 EST


From: Zqiang <qiang1.zhang@xxxxxxxxx>

If the rcutree.use_softirq kernel boot parameter is disabled, then it is
possible that a RCU CPU stall is due to the rcuc kthreads being starved of
CPU time. There is currently no easy way to infer this from the RCU CPU
stall warning output. This commit therefore adds a string of the form "
rcuc=%ld jiffies(starved)" to a given CPU's output if the corresponding
rcuc kthread has been starved for more than two seconds.

[ paulmck: Eliminate extraneous space characters. ]

Signed-off-by: Zqiang <qiang1.zhang@xxxxxxxxx>
Signed-off-by: Paul E. McKenney <paulmck@xxxxxxxxxx>
---
kernel/rcu/tree_stall.h | 49 ++++++++++++++++++-----------------------
1 file changed, 21 insertions(+), 28 deletions(-)

diff --git a/kernel/rcu/tree_stall.h b/kernel/rcu/tree_stall.h
index 4995c078cff98..3556637768fd5 100644
--- a/kernel/rcu/tree_stall.h
+++ b/kernel/rcu/tree_stall.h
@@ -409,7 +409,19 @@ static bool rcu_is_gp_kthread_starving(unsigned long *jp)

static bool rcu_is_rcuc_kthread_starving(struct rcu_data *rdp, unsigned long *jp)
{
- unsigned long j = jiffies - READ_ONCE(rdp->rcuc_activity);
+ int cpu;
+ struct task_struct *rcuc;
+ unsigned long j;
+
+ rcuc = rdp->rcu_cpu_kthread_task;
+ if (!rcuc)
+ return false;
+
+ cpu = task_cpu(rcuc);
+ if (cpu_is_offline(cpu) || idle_cpu(cpu))
+ return false;
+
+ j = jiffies - READ_ONCE(rdp->rcuc_activity);

if (jp)
*jp = j;
@@ -434,6 +446,9 @@ static void print_cpu_stall_info(int cpu)
struct rcu_data *rdp = per_cpu_ptr(&rcu_data, cpu);
char *ticks_title;
unsigned long ticks_value;
+ bool rcuc_starved;
+ unsigned long j;
+ char buf[32];

/*
* We could be printing a lot while holding a spinlock. Avoid
@@ -451,7 +466,10 @@ static void print_cpu_stall_info(int cpu)
delta = rcu_seq_ctr(rdp->mynode->gp_seq - rdp->rcu_iw_gp_seq);
falsepositive = rcu_is_gp_kthread_starving(NULL) &&
rcu_dynticks_in_eqs(rcu_dynticks_snap(rdp));
- pr_err("\t%d-%c%c%c%c: (%lu %s) idle=%03x/%ld/%#lx softirq=%u/%u fqs=%ld %s\n",
+ rcuc_starved = rcu_is_rcuc_kthread_starving(rdp, &j);
+ if (rcuc_starved)
+ sprintf(buf, " rcuc=%ld jiffies(starved)", j);
+ pr_err("\t%d-%c%c%c%c: (%lu %s) idle=%03x/%ld/%#lx softirq=%u/%u fqs=%ld%s%s\n",
cpu,
"O."[!!cpu_online(cpu)],
"o."[!!(rdp->grpmask & rdp->mynode->qsmaskinit)],
@@ -464,32 +482,10 @@ static void print_cpu_stall_info(int cpu)
rdp->dynticks_nesting, rdp->dynticks_nmi_nesting,
rdp->softirq_snap, kstat_softirqs_cpu(RCU_SOFTIRQ, cpu),
data_race(rcu_state.n_force_qs) - rcu_state.n_force_qs_gpstart,
+ rcuc_starved ? buf : "",
falsepositive ? " (false positive?)" : "");
}

-static void rcuc_kthread_dump(struct rcu_data *rdp)
-{
- int cpu;
- unsigned long j;
- struct task_struct *rcuc;
-
- rcuc = rdp->rcu_cpu_kthread_task;
- if (!rcuc)
- return;
-
- cpu = task_cpu(rcuc);
- if (cpu_is_offline(cpu) || idle_cpu(cpu))
- return;
-
- if (!rcu_is_rcuc_kthread_starving(rdp, &j))
- return;
-
- pr_err("%s kthread starved for %ld jiffies\n", rcuc->comm, j);
- sched_show_task(rcuc);
- if (!trigger_single_cpu_backtrace(cpu))
- dump_cpu_task(cpu);
-}
-
/* Complain about starvation of grace-period kthread. */
static void rcu_check_gp_kthread_starvation(void)
{
@@ -662,9 +658,6 @@ static void print_cpu_stall(unsigned long gps)
rcu_check_gp_kthread_expired_fqs_timer();
rcu_check_gp_kthread_starvation();

- if (!use_softirq)
- rcuc_kthread_dump(rdp);
-
rcu_dump_cpu_stacks();

raw_spin_lock_irqsave_rcu_node(rnp, flags);
--
2.31.1.189.g2e36527f23