Re: [PATCH] sched/core: make "Preemption disabled at" message more useful

[Vegard Nossum]

On 27 July 2016 at 11:15, Ingo Molnar <mingo@xxxxxxxxxx> wrote:
>
> * Vegard Nossum <vegard.nossum@xxxxxxxxxx> wrote:
[...]

> These two blocks could be merged trivially, avoiding an #ifdef pair ...

>> @@ -7541,6 +7550,9 @@ EXPORT_SYMBOL(__might_sleep);
>>  void ___might_sleep(const char *file, int line, int preempt_offset)
>>  {
>>       static unsigned long prev_jiffy;        /* ratelimiting */
>> +#ifdef CONFIG_DEBUG_PREEMPT
>> +     unsigned long preempt_disable_ip;
>> +#endif
>>
>>       rcu_sleep_check(); /* WARN_ON_ONCE() by default, no rate limit reqd. */
>>       if ((preempt_count_equals(preempt_offset) && !irqs_disabled() &&
>> @@ -7551,6 +7563,11 @@ void ___might_sleep(const char *file, int line, int preempt_offset)
>>               return;
>>       prev_jiffy = jiffies;
>>
>> +#ifdef CONFIG_DEBUG_PREEMPT
>> +     /* Save this before calling printk(), since that will clobber it */
>> +     preempt_disable_ip = current->preempt_disable_ip;
>> +#endif
>
> Ditto.

I'm assuming you want to declare and initialise preempt_disable_ip at
once here, but it generates slightly worse code since it dereferences
current->preempt_disable_ip in the "fast path" (i.e. a sleeping
function is NOT called from an invalid context).

If you really prefer that I've attached an updated patch.

Let me know if I misunderstood :-)


Vegard
From 332d49d4da90f00304de73c7412311576e2e7f78 Mon Sep 17 00:00:00 2001
From: Vegard Nossum <vegard.nossum@xxxxxxxxxx>
Date: Sat, 23 Jul 2016 09:46:39 +0200
Subject: [PATCH] sched/core: make "Preemption disabled at" message more useful

This message is currently really useless since it always prints a value
that comes from the printk() we just did, e.g.:

    BUG: sleeping function called from invalid context at mm/slab.h:388
    in_atomic(): 0, irqs_disabled(): 0, pid: 31996, name: trinity-c1
    Preemption disabled at:[<ffffffff8119db33>] down_trylock+0x13/0x80

    BUG: sleeping function called from invalid context at include/linux/freezer.h:56
    in_atomic(): 0, irqs_disabled(): 0, pid: 31996, name: trinity-c1
    Preemption disabled at:[<ffffffff811aaa37>] console_unlock+0x2f7/0x930

Here, both down_trylock() and console_unlock() is somewhere in the
printk() path.

We should save the value before calling printk() and use the saved value
instead. That immediately reveals the offending callsite:

    BUG: sleeping function called from invalid context at mm/slab.h:388
    in_atomic(): 0, irqs_disabled(): 0, pid: 14971, name: trinity-c2
    Preemption disabled at:[<ffffffff819bcd46>] rhashtable_walk_start+0x46/0x150

(Bug report: http://marc.info/?l=linux-netdev&m=146925979821849&w=2)

Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Cc: Paul E. McKenney <paulmck@xxxxxxxxxxxxxxxxxx>
Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Cc: Ingo Molnar <mingo@xxxxxxxxxx>
Cc: Rusty Russel <rusty@xxxxxxxxxxxxxxx>
Signed-off-by: Vegard Nossum <vegard.nossum@xxxxxxxxxx>
---
 kernel/sched/core.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 7171cf9..e06354e 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -3144,6 +3144,11 @@ static inline void preempt_latency_stop(int val) { }
  */
 static noinline void __schedule_bug(struct task_struct *prev)
 {
+#ifdef CONFIG_DEBUG_PREEMPT
+	/* Save this before calling printk(), since that will clobber it */
+	unsigned long preempt_disable_ip = current->preempt_disable_ip;
+#endif
+
 	if (oops_in_progress)
 		return;
 
@@ -3157,7 +3162,7 @@ static noinline void __schedule_bug(struct task_struct *prev)
 #ifdef CONFIG_DEBUG_PREEMPT
 	if (in_atomic_preempt_off()) {
 		pr_err("Preemption disabled at:");
-		print_ip_sym(current->preempt_disable_ip);
+		print_ip_sym(preempt_disable_ip);
 		pr_cont("\n");
 	}
 #endif
@@ -7541,6 +7546,10 @@ EXPORT_SYMBOL(__might_sleep);
 void ___might_sleep(const char *file, int line, int preempt_offset)
 {
 	static unsigned long prev_jiffy;	/* ratelimiting */
+#ifdef CONFIG_DEBUG_PREEMPT
+	/* Save this before calling printk(), since that will clobber it */
+	unsigned long preempt_disable_ip = current->preempt_disable_ip;
+#endif
 
 	rcu_sleep_check(); /* WARN_ON_ONCE() by default, no rate limit reqd. */
 	if ((preempt_count_equals(preempt_offset) && !irqs_disabled() &&
@@ -7568,7 +7577,7 @@ void ___might_sleep(const char *file, int line, int preempt_offset)
 #ifdef CONFIG_DEBUG_PREEMPT
 	if (!preempt_count_equals(preempt_offset)) {
 		pr_err("Preemption disabled at:");
-		print_ip_sym(current->preempt_disable_ip);
+		print_ip_sym(preempt_disable_ip);
 		pr_cont("\n");
 	}
 #endif
-- 
1.9.1