lockdep: Correctly annotate hardirq context in irq_exit()

There was a reported deadlock on -rt which lockdep didn't report.

It turns out that in irq_exit() we tell lockdep that the hardirq
context ends and then do all kinds of locking afterwards.

To fix it, move trace_hardirq_exit() to the very end of irq_exit(), this
ensures all locking in tick_irq_exit() and rcu_irq_exit() are properly
recorded as happening from hardirq context.

This however leads to the 'fun' little problem of running softirqs
while in hardirq context. To cure this make the softirq code a little
more complex (in the CONFIG_TRACE_IRQFLAGS case).

Due to stack swizzling arch dependent trickery we cannot pass an
argument to __do_softirq() to tell it if it was done from hardirq
context or not; so use a side-band argument.

When we do __do_softirq() from hardirq context, 'atomically' flip to
softirq context and back, so that no locking goes without being in
either hard- or soft-irq context.

I didn't find any new problems in mainline using this patch, but it
did show the -rt problem.

Cc: stable-rt@vger.kernel.org
Reported-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/n/tip-dgwc5cdksbn0jk09vbmcc9sa@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>

1 files changed, 42 insertions, 3 deletions

diff --git a/kernel/softirq.c b/kernel/softirq.c
index 34fe1db0d5e0..c2c6bc828dfb 100644
--- a/kernel/softirq.c
+++ b/kernel/softirq.c
@@ -297,6 +297,44 @@ EXPORT_SYMBOL(local_bh_enable_ip);
  */
 #define MAX_SOFTIRQ_RESTART 10
 
+#ifdef CONFIG_TRACE_IRQFLAGS
+/*
+ * Convoluted means of passing __do_softirq() a message through the various
+ * architecture execute_on_stack() bits.
+ *
+ * When we run softirqs from irq_exit() and thus on the hardirq stack we need
+ * to keep the lockdep irq context tracking as tight as possible in order to
+ * not miss-qualify lock contexts and miss possible deadlocks.
+ */
+static DEFINE_PER_CPU(int, softirq_from_hardirq);
+
+static inline void lockdep_softirq_from_hardirq(void)
+{
+	this_cpu_write(softirq_from_hardirq, 1);
+}
+
+static inline void lockdep_softirq_start(void)
+{
+	if (this_cpu_read(softirq_from_hardirq))
+		trace_hardirq_exit();
+	lockdep_softirq_enter();
+}
+
+static inline void lockdep_softirq_end(void)
+{
+	lockdep_softirq_exit();
+	if (this_cpu_read(softirq_from_hardirq)) {
+		this_cpu_write(softirq_from_hardirq, 0);
+		trace_hardirq_enter();
+	}
+}
+
+#else
+static inline void lockdep_softirq_from_hardirq(void) { }
+static inline void lockdep_softirq_start(void) { }
+static inline void lockdep_softirq_end(void) { }
+#endif
+
 asmlinkage void __do_softirq(void)
 {
 	__u32 pending;
@@ -308,7 +346,7 @@ asmlinkage void __do_softirq(void)
 
 	__local_bh_disable((unsigned long)__builtin_return_address(0),
 			   SOFTIRQ_OFFSET);
-	lockdep_softirq_enter();
+	lockdep_softirq_start();
 
 	cpu = smp_processor_id();
 restart:
@@ -324,7 +362,7 @@ restart:
 	if (pending)
 		wakeup_softirqd();
 
-	lockdep_softirq_exit();
+	lockdep_softirq_end();
 
 	account_system_vtime(current);
 	__local_bh_enable(SOFTIRQ_OFFSET);
@@ -582,6 +620,7 @@ static inline void invoke_softirq(void)
 {
 #ifndef CONFIG_PREEMPT_RT_FULL
 	if (!force_irqthreads) {
+		lockdep_softirq_from_hardirq();
 #ifdef __ARCH_IRQ_EXIT_IRQS_DISABLED
 		__do_softirq();
 #else
@@ -604,7 +643,6 @@ static inline void invoke_softirq(void)
 void irq_exit(void)
 {
 	account_system_vtime(current);
-	trace_hardirq_exit();
 	sub_preempt_count(IRQ_EXIT_OFFSET);
 	if (!in_interrupt() && local_softirq_pending())
 		invoke_softirq();
@@ -615,6 +653,7 @@ void irq_exit(void)
 		tick_nohz_irq_exit();
 #endif
 	rcu_irq_exit();
+	trace_hardirq_exit(); /* must be last! */
 	sched_preempt_enable_no_resched();
 }