rcu: Fix stall-warning deadlock due to non-release of rcu_node ->lock
authorYanfei Xu <yanfei.xu@windriver.com>
Sun, 16 May 2021 09:50:10 +0000 (17:50 +0800)
committerPaul E. McKenney <paulmck@kernel.org>
Fri, 6 Aug 2021 20:41:42 +0000 (13:41 -0700)
commitdc87740c8a6806bd2162bfb441770e4e53be5601
tree13a3871468ac9b4845cbb9acad7322acf0cc0585
parente6a901a44f76878ed1653626c9ff4cfc5a3f58f8
rcu: Fix stall-warning deadlock due to non-release of rcu_node ->lock

If rcu_print_task_stall() is invoked on an rcu_node structure that does
not contain any tasks blocking the current grace period, it takes an
early exit that fails to release that rcu_node structure's lock.  This
results in a self-deadlock, which is detected by lockdep.

To reproduce this bug:

tools/testing/selftests/rcutorture/bin/kvm.sh --allcpus --duration 3 --trust-make --configs "TREE03" --kconfig "CONFIG_PROVE_LOCKING=y" --bootargs "rcutorture.stall_cpu=30 rcutorture.stall_cpu_block=1 rcutorture.fwd_progress=0 rcutorture.test_boost=0"

This will also result in other complaints, including RCU's scheduler
hook complaining about blocking rather than preemption and an rcutorture
writer stall.

Only a partial RCU CPU stall warning message will be printed because of
the self-deadlock.

This commit therefore releases the lock on the rcu_print_task_stall()
function's early exit path.

Fixes: c583bcb8f5ed ("rcu: Don't invoke try_invoke_on_locked_down_task() with irqs disabled")
Tested-by: Qais Yousef <qais.yousef@arm.com>
Signed-off-by: Yanfei Xu <yanfei.xu@windriver.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
kernel/rcu/tree_stall.h