rcu: Don't use NMIs to dump other CPUs' stacks
authorPaul E. McKenney <paulmck@linux.vnet.ibm.com>
Wed, 18 Jun 2014 16:18:31 +0000 (09:18 -0700)
committerPaul E. McKenney <paulmck@linux.vnet.ibm.com>
Wed, 9 Jul 2014 16:15:04 +0000 (09:15 -0700)
commitbc1dce514e9b29b64df28a533015885862f47814
tree72977f4949a34fa20736dfdd69c6f740ae086d37
parentc0f489d2c6fec8994c642c2ec925eb858727dc7b
rcu: Don't use NMIs to dump other CPUs' stacks

Although NMI-based stack dumps are in principle more accurate, they are
also more likely to trigger deadlocks.  This commit therefore replaces
all uses of trigger_all_cpu_backtrace() with rcu_dump_cpu_stacks(), so
that the CPU detecting an RCU CPU stall does the stack dumping.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
kernel/rcu/tree.c