Re: [circular locking bug] Re: [patch 00/15] clocksource / timekeeping rework V4 (resend V3 + bug fix)

[Martin Schwidefsky <schwidefsky@de.ibm.com>]

On Mon, 17 Aug 2009 11:28:07 +0200
Ingo Molnar <mingo@elte.hu> wrote:

> 
> -tip testing still triggers lockdep troubles with latest 
> timers/core. I think timekeeping_notify() might be involved.
> 
> Bootlog below, config attached.
> [   43.081726] =======================================================
> [   43.089512] [ INFO: possible circular locking dependency detected ]
> [   43.090020] 2.6.31-rc6-tip-01039-ge54d9cb-dirty #4937
> [   43.090020] -------------------------------------------------------
> [   43.090020] S99local/5002 is trying to acquire lock:
> [   43.090020]  (events){+.+.+.}, at: [<ffffffff8106ff4c>] cleanup_workqueue_thread+0x2d/0xdd
> [   43.090020] 
> [   43.090020] but task is already holding lock:
> [   43.090020]  (cpu_add_remove_lock){+.+.+.}, at: [<ffffffff818387aa>] cpu_down+0x3b/0xb4
> [   43.090020] 
> [   43.090020] which lock already depends on the new lock.
> [   43.090020] 
> [   43.090020] 
> [   43.090020] the existing dependency chain (in reverse order) is:
> [   43.090020] 
> [   43.090020] -> #4 (cpu_add_remove_lock){+.+.+.}:
> [   43.090020]        [<ffffffff8108925d>] check_prev_add+0x18e/0x225
> [   43.090020]        [<ffffffff8108976a>] validate_chain+0x476/0x556
> [   43.090020]        [<ffffffff81089bfa>] __lock_acquire+0x3b0/0x43b
> [   43.090020]        [<ffffffff81089d5a>] lock_acquire+0xd5/0x10e
> [   43.090020]        [<ffffffff81872229>] __mutex_lock_common+0x5b/0x3b0
> [   43.090020]        [<ffffffff81872683>] mutex_lock_nested+0x45/0x62
> [   43.090020]        [<ffffffff8105de75>] cpu_maps_update_begin+0x25/0x3b
> [   43.090020]        [<ffffffff81070348>] __create_workqueue_key+0x137/0x200
> [   43.090020]        [<ffffffff810a2397>] stop_machine_create+0x4d/0xc2
> [   43.090020]        [<ffffffff810a2438>] stop_machine+0x2c/0x74
> [   43.090020]        [<ffffffff8107f163>] timekeeping_notify+0x31/0x4c
> [   43.090020]        [<ffffffff81080258>] clocksource_select+0xc9/0xe8
> [   43.090020]        [<ffffffff81080b8c>] clocksource_register+0x70/0xa1
> [   43.090020]        [<ffffffff81d4cc6a>] init_acpi_pm_clocksource+0xf3/0x114
> [   43.090020]        [<ffffffff8100a092>] do_one_initcall+0x65/0x15b
> [   43.090020]        [<ffffffff81d0d992>] do_basic_setup+0x59/0x82
> [   43.090020]        [<ffffffff81d0da3e>] kernel_init+0x83/0xdd
> [   43.090020]        [<ffffffff8101520a>] child_rip+0xa/0x20
> [   43.090020] 
> [   43.090020] -> #3 (setup_lock){+.+.+.}:
> [   43.090020]        [<ffffffff8108925d>] check_prev_add+0x18e/0x225
> [   43.090020]        [<ffffffff8108976a>] validate_chain+0x476/0x556
> [   43.090020]        [<ffffffff81089bfa>] __lock_acquire+0x3b0/0x43b
> [   43.090020]        [<ffffffff81089d5a>] lock_acquire+0xd5/0x10e
> [   43.090020]        [<ffffffff81872229>] __mutex_lock_common+0x5b/0x3b0
> [   43.090020]        [<ffffffff81872683>] mutex_lock_nested+0x45/0x62
> [   43.090020]        [<ffffffff810a236f>] stop_machine_create+0x25/0xc2
> [   43.090020]        [<ffffffff810a2438>] stop_machine+0x2c/0x74
> [   43.090020]        [<ffffffff8107f163>] timekeeping_notify+0x31/0x4c
> [   43.090020]        [<ffffffff81080258>] clocksource_select+0xc9/0xe8
> [   43.090020]        [<ffffffff81080b8c>] clocksource_register+0x70/0xa1
> [   43.090020]        [<ffffffff81d4cc6a>] init_acpi_pm_clocksource+0xf3/0x114
> [   43.090020]        [<ffffffff8100a092>] do_one_initcall+0x65/0x15b
> [   43.090020]        [<ffffffff81d0d992>] do_basic_setup+0x59/0x82
> [   43.090020]        [<ffffffff81d0da3e>] kernel_init+0x83/0xdd
> [   43.090020]        [<ffffffff8101520a>] child_rip+0xa/0x20
> [   43.090020] 
> [   43.090020] -> #2 (clocksource_mutex){+.+.+.}:
> [   43.090020]        [<ffffffff8108925d>] check_prev_add+0x18e/0x225
> [   43.090020]        [<ffffffff8108976a>] validate_chain+0x476/0x556
> [   43.090020]        [<ffffffff81089bfa>] __lock_acquire+0x3b0/0x43b
> [   43.090020]        [<ffffffff81089d5a>] lock_acquire+0xd5/0x10e
> [   43.090020]        [<ffffffff81872229>] __mutex_lock_common+0x5b/0x3b0
> [   43.090020]        [<ffffffff81872683>] mutex_lock_nested+0x45/0x62
> [   43.090020]        [<ffffffff8108076c>] clocksource_change_rating+0x34/0xaf
> [   43.090020]        [<ffffffff8108091a>] clocksource_watchdog_work+0x133/0x16f
> [   43.090020]        [<ffffffff8106f9ca>] run_workqueue+0x161/0x265
> [   43.090020]        [<ffffffff8106fbb3>] worker_thread+0xe5/0x10c
> [   43.090020]        [<ffffffff81074b64>] kthread+0x98/0xa0
> [   43.090020]        [<ffffffff8101520a>] child_rip+0xa/0x20
> [   43.090020] 
> [   43.090020] -> #1 (&watchdog_work){+.+...}:
> [   43.090020]        [<ffffffff8108925d>] check_prev_add+0x18e/0x225
> [   43.090020]        [<ffffffff8108976a>] validate_chain+0x476/0x556
> [   43.090020]        [<ffffffff81089bfa>] __lock_acquire+0x3b0/0x43b
> [   43.090020]        [<ffffffff81089d5a>] lock_acquire+0xd5/0x10e
> [   43.090020]        [<ffffffff8106f9c4>] run_workqueue+0x15b/0x265
> [   43.090020]        [<ffffffff8106fbb3>] worker_thread+0xe5/0x10c
> [   43.090020]        [<ffffffff81074b64>] kthread+0x98/0xa0
> [   43.090020]        [<ffffffff8101520a>] child_rip+0xa/0x20
> [   43.090020] 
> [   43.090020] -> #0 (events){+.+.+.}:
> [   43.090020]        [<ffffffff81089146>] check_prev_add+0x77/0x225
> [   43.090020]        [<ffffffff8108976a>] validate_chain+0x476/0x556
> [   43.090020]        [<ffffffff81089bfa>] __lock_acquire+0x3b0/0x43b
> [   43.090020]        [<ffffffff81089d5a>] lock_acquire+0xd5/0x10e
> [   43.090020]        [<ffffffff8106ff73>] cleanup_workqueue_thread+0x54/0xdd
> [   43.090020]        [<ffffffff81839c41>] workqueue_cpu_callback+0xdb/0x13f
> [   43.090020]        [<ffffffff8187714e>] notifier_call_chain+0x6d/0xb5
> [   43.090020]        [<ffffffff8107a19e>] raw_notifier_call_chain+0x22/0x38
> [   43.090020]        [<ffffffff81838735>] _cpu_down+0x2c2/0x2fc
> [   43.090020]        [<ffffffff818387cc>] cpu_down+0x5d/0xb4
> [   43.090020]        [<ffffffff8183988e>] store_online+0x3f/0x9a
> [   43.090020]        [<ffffffff813c2e41>] sysdev_store+0x2e/0x44
> [   43.090020]        [<ffffffff811559a5>] sysfs_write_file+0xf3/0x13c
> [   43.090020]        [<ffffffff810f6f7a>] vfs_write+0xbe/0x130
> [   43.090020]        [<ffffffff810f70e2>] sys_write+0x56/0x93
> [   43.090020]        [<ffffffff81014042>] system_call_fastpath+0x16/0x1b
> [   43.090020] 
> [   43.090020] other info that might help us debug this:
> [   43.090020] 
> [   43.090020] 2 locks held by S99local/5002:
> [   43.090020]  #0:  (&buffer->mutex){+.+.+.}, at: [<ffffffff811558f9>] sysfs_write_file+0x47/0x13c
> [   43.090020]  #1:  (cpu_add_remove_lock){+.+.+.}, at: [<ffffffff818387aa>] cpu_down+0x3b/0xb4
> [   43.090020] 
> [   43.090020] stack backtrace:
> [   43.090020] Pid: 5002, comm: S99local Not tainted 2.6.31-rc6-tip-01039-ge54d9cb-dirty #4937
> [   43.090020] Call Trace:
> [   43.090020]  [<ffffffff81088bb0>] print_circular_bug+0xcf/0xf4
> [   43.090020]  [<ffffffff81089146>] check_prev_add+0x77/0x225
> [   43.090020]  [<ffffffff818710e5>] ? wait_for_common+0xeb/0x150
> [   43.090020]  [<ffffffff8108976a>] validate_chain+0x476/0x556
> [   43.090020]  [<ffffffff81087d17>] ? mark_lock+0x31/0x17d
> [   43.090020]  [<ffffffff81089bfa>] __lock_acquire+0x3b0/0x43b
> [   43.090020]  [<ffffffff8106ff4c>] ? cleanup_workqueue_thread+0x2d/0xdd
> [   43.090020]  [<ffffffff8106ff4c>] ? cleanup_workqueue_thread+0x2d/0xdd
> [   43.090020]  [<ffffffff81089d5a>] lock_acquire+0xd5/0x10e
> [   43.090020]  [<ffffffff8106ff4c>] ? cleanup_workqueue_thread+0x2d/0xdd
> [   43.090020]  [<ffffffff81871283>] ? wait_for_completion+0x2b/0x41
> [   43.090020]  [<ffffffff81074be5>] ? kthread_stop+0x79/0xd9
> [   43.090020]  [<ffffffff8106ff73>] cleanup_workqueue_thread+0x54/0xdd
> [   43.090020]  [<ffffffff8106ff4c>] ? cleanup_workqueue_thread+0x2d/0xdd
> [   43.090020]  [<ffffffff81839c41>] workqueue_cpu_callback+0xdb/0x13f
> [   43.090020]  [<ffffffff8187714e>] notifier_call_chain+0x6d/0xb5
> [   43.090020]  [<ffffffff8107a19e>] raw_notifier_call_chain+0x22/0x38
> [   43.090020]  [<ffffffff81838735>] _cpu_down+0x2c2/0x2fc
> [   43.090020]  [<ffffffff818710e5>] ? wait_for_common+0xeb/0x150
> [   43.090020]  [<ffffffff818387cc>] cpu_down+0x5d/0xb4
> [   43.090020]  [<ffffffff8183988e>] store_online+0x3f/0x9a
> [   43.090020]  [<ffffffff813c2e41>] sysdev_store+0x2e/0x44
> [   43.090020]  [<ffffffff811559a5>] sysfs_write_file+0xf3/0x13c
> [   43.090020]  [<ffffffff810f65fa>] ? rw_verify_area+0x97/0xd1
> [   43.090020]  [<ffffffff810f6f7a>] vfs_write+0xbe/0x130
> [   43.090020]  [<ffffffff810f70e2>] sys_write+0x56/0x93
> [   43.090020]  [<ffffffff81014042>] system_call_fastpath+0x16/0x1b
> [   43.829633] lockdep: fixing up alternatives.

I though about this circular dependency for a while and came to the
conclusion that stop_machine from a multithreaded workqueue is not
allowed. It deadlocks with cpu_down. We either use a single-threaded
workqueue or an explicitly created kernel thread for the clocksource
downgrade. The kthread_run solution seems to be less painful than
to create a single-threaded workqueue. Untested patch below.

--
Subject: [PATCH] clocksource watchdog circular locking dependency

From: Martin Schwidefsky <schwidefsky@de.ibm.com>

stop_machine from a multithreaded workqueue is not allowed because
of a circular locking dependency between cpu_down and the workqueue
execution. Use a kernel thread to do the clocksource downgrade.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: john stultz <johnstul@us.ibm.com>
Cc: Daniel Walker <dwalker@fifo99.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
---

---
 kernel/time/clocksource.c |   19 ++++++++++++++++---
 1 file changed, 16 insertions(+), 3 deletions(-)

Index: linux-2.6/kernel/time/clocksource.c
===================================================================
--- linux-2.6.orig/kernel/time/clocksource.c
+++ linux-2.6/kernel/time/clocksource.c
@@ -29,6 +29,7 @@
 #include <linux/module.h>
 #include <linux/sched.h> /* for spin_unlock_irq() using preempt_count() m68k */
 #include <linux/tick.h>
+#include <linux/kthread.h>
 
 void timecounter_init(struct timecounter *tc,
 		      const struct cyclecounter *cc,
@@ -130,7 +131,7 @@ static DEFINE_SPINLOCK(watchdog_lock);
 static cycle_t watchdog_last;
 static int watchdog_running;
 
-static void clocksource_watchdog_work(struct work_struct *work);
+static int clocksource_watchdog_kthread(void *data);
 
 /*
  * Interval: 0.5sec Threshold: 0.0625s
@@ -138,6 +139,15 @@ static void clocksource_watchdog_work(st
 #define WATCHDOG_INTERVAL (HZ >> 1)
 #define WATCHDOG_THRESHOLD (NSEC_PER_SEC >> 4)
 
+static void clocksource_watchdog_work(struct work_struct *work)
+{
+	/*
+	 * If kthread_run fails the next watchdog scan over the
+	 * watchdog_list will find the unstable clock again.
+	 */
+	kthread_run(clocksource_watchdog_kthread, NULL, "kwatchdog");
+}
+
 static void clocksource_unstable(struct clocksource *cs, int64_t delta)
 {
 	printk(KERN_WARNING "Clocksource %s unstable (delta = %Ld ns)\n",
@@ -166,8 +176,10 @@ static void clocksource_watchdog(unsigne
 	list_for_each_entry(cs, &watchdog_list, wd_list) {
 
 		/* Clocksource already marked unstable? */
-		if (cs->flags & CLOCK_SOURCE_UNSTABLE)
+		if (cs->flags & CLOCK_SOURCE_UNSTABLE) {
+			schedule_work(&watchdog_work);
 			continue;
+		}
 
 		csnow = cs->read(cs);
 
@@ -303,7 +315,7 @@ static void clocksource_dequeue_watchdog
 	spin_unlock_irqrestore(&watchdog_lock, flags);
 }
 
-static void clocksource_watchdog_work(struct work_struct *work)
+static int clocksource_watchdog_kthread(void *data)
 {
 	struct clocksource *cs, *tmp;
 	unsigned long flags;
@@ -324,6 +336,7 @@ static void clocksource_watchdog_work(st
 		list_del_init(&cs->wd_list);
 		clocksource_change_rating(cs, 0);
 	}
+	return 0;
 }
 
 #else /* CONFIG_CLOCKSOURCE_WATCHDOG */

-- 
blue skies,
   Martin.

"Reality continues to ruin my life." - Calvin.