[PATCH] locking/lockdep: Dump state of percpu_rwsem upon hung up.

[Tetsuo Handa]

This is a temporary patch which should not go to linux.git.

syzbot is hitting hung task problems at __sb_start_write() [1]. While hung
tasks at __sb_start_write() suggest that filesystem was frozen, syzbot is
not doing ioctl(FIFREEZE) requests. Therefore, the root cause of hung tasks
is currently unknown. As a first step for debugging this problem, let's
check what atomic_long_read(&sem->rw_sem.count) says when hung task is
reported. Since it is impossible to reproduce this problem locally, this
patch was made in order to test linux-next.git using syzbot infrastructure.

[1] https://syzkaller.appspot.com/bug?id=287aa8708bc940d0ca1645223c53dd4c2d203be6

Signed-off-by: Tetsuo Handa <penguin-kernel@xxxxxxxxxxxxxxxxxxx>
Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Cc: Ingo Molnar <mingo@xxxxxxxxxx>
Cc: Will Deacon <will.deacon@xxxxxxx>
Cc: Dmitry Vyukov <dvyukov@xxxxxxxxxx>
---
 include/linux/sched.h         |  1 +
 kernel/hung_task.c            | 25 +++++++++++++++++++++++++
 kernel/locking/percpu-rwsem.c |  2 ++
 3 files changed, 28 insertions(+)

diff --git a/include/linux/sched.h b/include/linux/sched.h
index cf1edc0..2aaaf49 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -1183,6 +1183,7 @@ struct task_struct {
 	/* Used by LSM modules for access restriction: */
 	void				*security;
 #endif
+	struct percpu_rw_semaphore	*pcpu_rwsem_read_waiting;
 
 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
 	unsigned long			lowest_stack;
diff --git a/kernel/hung_task.c b/kernel/hung_task.c
index b9132d1..03d7d9b 100644
--- a/kernel/hung_task.c
+++ b/kernel/hung_task.c
@@ -18,6 +18,7 @@
 #include <linux/utsname.h>
 #include <linux/sched/signal.h>
 #include <linux/sched/debug.h>
+#include <linux/percpu-rwsem.h>
 
 #include <trace/events/sched.h>
 
@@ -82,6 +83,29 @@ static int __init hung_task_panic_setup(char *str)
 	.notifier_call = hung_task_panic,
 };
 
+static void dump_percpu_rwsem_state(struct percpu_rw_semaphore *sem)
+{
+	unsigned int sum = 0;
+	int cpu;
+
+	if (!sem)
+		return;
+	pr_info("percpu_rw_semaphore(%p)\n", sem);
+	pr_info("  ->rw_sem.count=0x%lx\n",
+		atomic_long_read(&sem->rw_sem.count));
+	pr_info("  ->rss.gp_state=%d\n", sem->rss.gp_state);
+	pr_info("  ->rss.gp_count=%d\n", sem->rss.gp_count);
+	pr_info("  ->rss.cb_state=%d\n", sem->rss.cb_state);
+	pr_info("  ->rss.gp_type=%d\n", sem->rss.gp_type);
+	pr_info("  ->readers_block=%d\n", sem->readers_block);
+	for_each_possible_cpu(cpu)
+		sum += per_cpu(*sem->read_count, cpu);
+	pr_info("  ->read_count=%d\n", sum);
+	pr_info("  ->list_empty(rw_sem.wait_list)=%d\n",
+	       list_empty(&sem->rw_sem.wait_list));
+	pr_info("  ->writer.task=%p\n", sem->writer.task);
+}
+
 static void check_hung_task(struct task_struct *t, unsigned long timeout)
 {
 	unsigned long switch_count = t->nvcsw + t->nivcsw;
@@ -130,6 +154,7 @@ static void check_hung_task(struct task_struct *t, unsigned long timeout)
 		pr_err("\"echo 0 > /proc/sys/kernel/hung_task_timeout_secs\""
 			" disables this message.\n");
 		sched_show_task(t);
+		dump_percpu_rwsem_state(t->pcpu_rwsem_read_waiting);
 		hung_task_show_lock = true;
 	}
 
diff --git a/kernel/locking/percpu-rwsem.c b/kernel/locking/percpu-rwsem.c
index 883cf1b..b3654ab 100644
--- a/kernel/locking/percpu-rwsem.c
+++ b/kernel/locking/percpu-rwsem.c
@@ -82,7 +82,9 @@ int __percpu_down_read(struct percpu_rw_semaphore *sem, int try)
 	/*
 	 * Avoid lockdep for the down/up_read() we already have them.
 	 */
+	current->pcpu_rwsem_read_waiting = sem;
 	__down_read(&sem->rw_sem);
+	current->pcpu_rwsem_read_waiting = NULL;
 	this_cpu_inc(*sem->read_count);
 	__up_read(&sem->rw_sem);
 
-- 
1.8.3.1