Skip to contentSkip to conversation

lockdep: call trace_hardirqs_off after clearing hardirqs_enabled

PATCH In review Qi Hou 3 participants 4 messages Mar 13, 2017 – Mar 14, 2017 Locking primitives

lore
Share & tools
lore

Qi Hou

author
#0 lore raw

PATCH 1/1 lockdep: call trace_hardirqs_off after clearing hardirqs_enabled

The current order of calls within trace_hardirqs_off() would provoke an "unannotated irqs-off" WARNING.

This warning was reported by check_flags() when it found that the hardirqs has been disabled but the irq-flags state, hardirqs_enabled, has not been cleared.

Check_flags() is called via trace_hardirqs_off(), and trace_hardirqs_off() will be called by functions kind of local_irq_disable() if CONFIG_TRACE_IRQFLAGS has been enabled.

The calltrace from functions kind of local_irq_diasble() to check_flags() simplied as below:

local_irq_disable()
|- trace_hardirqs_off()

     |- trace_hardirqs_off_caller()
         |- preemptirqsoff_hist()
            |- __DO_TRACE()
               |- rcu_dereference_sched()
                   |- rcu_read_lock_sched_held()
                       |- lock_is_held()
                           |- check_flags()

The order of calls within trace_hardirqs_off() can be simplified as below:

local_irq_disable()
|- raw_local_irq_disable()
|- trace_hardirqs_off()
|- trace_hardirqs_off() --> check_flags()
|- hardirqs_enabled = 0

We could find that the call to check_flags() has been made between disabling hardirqs and clearing hardirqs_enabled. That will definitely cause an "unannotated irqs-off" warning.

To fix this "unannotated irqs-off" warning, we must call trace_hardirqs_off() after that we clear hardirqs_enabled.

---
 kernel/locking/lockdep.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
kernel/locking/lockdep.c +2 -2
index 577f026..e8b35e4 100644
@@ -2627,8 +2627,6 @@ __visible void trace_hardirqs_off_caller(unsigned long ip)
 {
 	struct task_struct *curr = current;
 
-	time_hardirqs_off(CALLER_ADDR0, ip);
-
 	if (unlikely(!debug_locks || current->lockdep_recursion))
 		return;
 
@@ -2649,6 +2647,8 @@ __visible void trace_hardirqs_off_caller(unsigned long ip)
 		debug_atomic_inc(hardirqs_off_events);
 	} else
 		debug_atomic_inc(redundant_hardirqs_off);
+
+	time_hardirqs_off(CALLER_ADDR0, ip);
 }
 EXPORT_SYMBOL(trace_hardirqs_off_caller);
 
Signature · 1 line
1.9.1
#1 lore raw
Quoted 10 lines · On Mon, 13 Mar 2017 18:08:48 +0800 Qi Hou <[email protected]> wrote:

The current order of calls within trace_hardirqs_off() would provoke an "unannotated irqs-off" WARNING.

This warning was reported by check_flags() when it found that the hardirqs has been disabled but the irq-flags state, hardirqs_enabled, has not been cleared.

Check_flags() is called via trace_hardirqs_off(), and trace_hardirqs_off() will be called by functions kind of local_irq_disable() if CONFIG_TRACE_IRQFLAGS has been enabled.

Quoted 7 lines

The calltrace from functions kind of local_irq_diasble() to check_flags() simplied as below:

local_irq_disable()
|- trace_hardirqs_off()
|- trace_hardirqs_off_caller()
|- preemptirqsoff_hist()

There is no preemptirqsoff_hist() function in the kernel.

-- Steve

Quoted 48 lines
            |- __DO_TRACE()
               |- rcu_dereference_sched()
                   |- rcu_read_lock_sched_held()
                       |- lock_is_held()
                           |- check_flags()

The order of calls within trace_hardirqs_off() can be simplified as below:

local_irq_disable()
|- raw_local_irq_disable()
|- trace_hardirqs_off()
|- trace_hardirqs_off() --> check_flags()
|- hardirqs_enabled = 0

We could find that the call to check_flags() has been made between disabling hardirqs and clearing hardirqs_enabled. That will definitely cause an "unannotated irqs-off" warning.

To fix this "unannotated irqs-off" warning, we must call trace_hardirqs_off() after that we clear hardirqs_enabled.

---
 kernel/locking/lockdep.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
kernel/locking/lockdep.c +2 -2
index 577f026..e8b35e4 100644
@@ -2627,8 +2627,6 @@ __visible void trace_hardirqs_off_caller(unsigned long ip)
 {
 	struct task_struct *curr = current;
 
-	time_hardirqs_off(CALLER_ADDR0, ip);
-
 	if (unlikely(!debug_locks || current->lockdep_recursion))
 		return;
 
@@ -2649,6 +2647,8 @@ __visible void trace_hardirqs_off_caller(unsigned long ip)
 		debug_atomic_inc(hardirqs_off_events);
 	} else
 		debug_atomic_inc(redundant_hardirqs_off);
+
+	time_hardirqs_off(CALLER_ADDR0, ip);
 }
 EXPORT_SYMBOL(trace_hardirqs_off_caller);
 
author
#2 lore raw
Quoted 13 lines · Steven Rostedt
Quoted 10 lines · On Mon, 13 Mar 2017 18:08:48 +0800 Qi Hou <[email protected]> wrote:

The current order of calls within trace_hardirqs_off() would provoke an "unannotated irqs-off" WARNING.

This warning was reported by check_flags() when it found that the hardirqs has been disabled but the irq-flags state, hardirqs_enabled, has not been cleared.

Check_flags() is called via trace_hardirqs_off(), and trace_hardirqs_off() will be called by functions kind of local_irq_disable() if CONFIG_TRACE_IRQFLAGS has been enabled.

Quoted 8 lines
Quoted 7 lines

The calltrace from functions kind of local_irq_diasble() to check_flags() simplied as below:

local_irq_disable()
|- trace_hardirqs_off()
|- trace_hardirqs_off_caller()
|- preemptirqsoff_hist()

There is no preemptirqsoff_hist() function in the kernel.

Sorry, it's a mistake. It should have been "trace_hardirqs_off_caller()", defined in file kernel/locking/lockdep.c.

And it will call trace_preemptirqsoff_hist_rcuidle(), trace_preemptirqsoff_hist_rcuidle will() call check_flags().

I will correct it. And after carefully checking the patch, I will re-send it.

Very very sorry for my mistake. That wasted your time.

---
Best regards,
Qi Hou

Quoted 51 lines

-- Steve

Quoted 48 lines
             |- __DO_TRACE()
                |- rcu_dereference_sched()
                    |- rcu_read_lock_sched_held()
                        |- lock_is_held()
                            |- check_flags()

The order of calls within trace_hardirqs_off() can be simplified as below:

local_irq_disable()
|- raw_local_irq_disable()
|- trace_hardirqs_off()
|- trace_hardirqs_off() --> check_flags()
|- hardirqs_enabled = 0

We could find that the call to check_flags() has been made between disabling hardirqs and clearing hardirqs_enabled. That will definitely cause an "unannotated irqs-off" warning.

To fix this "unannotated irqs-off" warning, we must call trace_hardirqs_off() after that we clear hardirqs_enabled.

---
  kernel/locking/lockdep.c | 4 ++--
  1 file changed, 2 insertions(+), 2 deletions(-)
kernel/locking/lockdep.c +2 -2
index 577f026..e8b35e4 100644
@@ -2627,8 +2627,6 @@ __visible void trace_hardirqs_off_caller(unsigned long ip)
  {
  	struct task_struct *curr = current;
 
-	time_hardirqs_off(CALLER_ADDR0, ip);
-
  	if (unlikely(!debug_locks || current->lockdep_recursion))
  		return;
 
@@ -2649,6 +2647,8 @@ __visible void trace_hardirqs_off_caller(unsigned long ip)
  		debug_atomic_inc(hardirqs_off_events);
  	} else
  		debug_atomic_inc(redundant_hardirqs_off);
+
+	time_hardirqs_off(CALLER_ADDR0, ip);
  }
  EXPORT_SYMBOL(trace_hardirqs_off_caller);
 
Signature · 4 lines
Best regards,
Qi Hou
Phone number: +86-10-8477-8608
Address: Floor 15, Building B, Wangjing Plaza, No.9 Zhong-Huan Nanlu, Chaoyang District
maintainer
#3 lore raw
In reply to Qi Hou

Thanks for Cc'ing me.

Quoted 31 lines · Qi Hou

The current order of calls within trace_hardirqs_off() would provoke an "unannotated irqs-off" WARNING.

This warning was reported by check_flags() when it found that the hardirqs has been disabled but the irq-flags state, hardirqs_enabled, has not been cleared.

Check_flags() is called via trace_hardirqs_off(), and trace_hardirqs_off() will be called by functions kind of local_irq_disable() if CONFIG_TRACE_IRQFLAGS has been enabled.

The calltrace from functions kind of local_irq_diasble() to check_flags() simplied as below:

local_irq_disable()
|- trace_hardirqs_off()

     |- trace_hardirqs_off_caller()
         |- preemptirqsoff_hist()
            |- __DO_TRACE()
               |- rcu_dereference_sched()
                   |- rcu_read_lock_sched_held()
                       |- lock_is_held()
                           |- check_flags()

The order of calls within trace_hardirqs_off() can be simplified as below:

local_irq_disable()
|- raw_local_irq_disable()
|- trace_hardirqs_off()
|- trace_hardirqs_off() --> check_flags()
|- hardirqs_enabled = 0

Quoted 6 lines

We could find that the call to check_flags() has been made between disabling hardirqs and clearing hardirqs_enabled. That will definitely cause an "unannotated irqs-off" warning.

To fix this "unannotated irqs-off" warning, we must call trace_hardirqs_off() after that we clear hardirqs_enabled.

Here you say: "trace_hardirqs_off()"

Quoted 10 lines
---
 kernel/locking/lockdep.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
kernel/locking/lockdep.c +0 -0
index 577f026..e8b35e4 100644

kernel/locking/lockdep.c __visible void trace_hardirqs_off_caller(unsigned long ip) · line 2630

Quoted 4 lines
kernel/locking/lockdep.c +0 -1
@@ -2627,8 +2627,6 @@ __visible void trace_hardirqs_off_caller(unsigned long ip)
 {
 	struct task_struct *curr = current;
 
-	time_hardirqs_off(CALLER_ADDR0, ip);

This is: "TIME_hardirqs_off()", notice the difference?

Care to try to write a changelog that makes sense?