rcu: Dump local stack if cannot dump all CPUs' stacks
authorPaul E. McKenney <paulmck@linux.vnet.ibm.com>
Wed, 3 Aug 2011 10:34:24 +0000 (03:34 -0700)
committerPaul E. McKenney <paulmck@linux.vnet.ibm.com>
Thu, 29 Sep 2011 04:38:36 +0000 (21:38 -0700)
The trigger_all_cpu_backtrace() function is a no-op in architectures that
do not define arch_trigger_all_cpu_backtrace.  On such architectures, RCU
CPU stall warning messages contain no stack trace information, which makes
debugging quite difficult.  This commit therefore substitutes dump_stack()
for architectures that do not define arch_trigger_all_cpu_backtrace,
so that at least the local CPU's stack is dumped as part of the RCU CPU
stall warning message.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
kernel/rcutree.c

index ebd18e5..a07bf55 100644 (file)
@@ -583,7 +583,8 @@ static void print_other_cpu_stall(struct rcu_state *rsp)
        }
        printk("} (detected by %d, t=%ld jiffies)\n",
               smp_processor_id(), (long)(jiffies - rsp->gp_start));
-       trigger_all_cpu_backtrace();
+       if (!trigger_all_cpu_backtrace())
+               dump_stack();
 
        /* If so configured, complain about tasks blocking the grace period. */
 
@@ -604,7 +605,8 @@ static void print_cpu_stall(struct rcu_state *rsp)
         */
        printk(KERN_ERR "INFO: %s detected stall on CPU %d (t=%lu jiffies)\n",
               rsp->name, smp_processor_id(), jiffies - rsp->gp_start);
-       trigger_all_cpu_backtrace();
+       if (!trigger_all_cpu_backtrace())
+               dump_stack();
 
        raw_spin_lock_irqsave(&rnp->lock, flags);
        if (ULONG_CMP_GE(jiffies, rsp->jiffies_stall))