]> git.baikalelectronics.ru Git - kernel.git/commit
rcu: Don't acquire lock in NMI handler in rcu_nmi_enter_common()
authorPaul E. McKenney <paulmck@kernel.org>
Thu, 12 Mar 2020 23:55:07 +0000 (16:55 -0700)
committerPaul E. McKenney <paulmck@kernel.org>
Sun, 5 Apr 2020 21:22:15 +0000 (14:22 -0700)
commita698dd7585492d724c8f3db0bcaeb44e58f64e0d
treebce8b75a3caa6a7f7feeadb20187c7c5f2851539
parent3dbfc463848f672b7e5a855cf6a1baae55ca4699
rcu: Don't acquire lock in NMI handler in rcu_nmi_enter_common()

The rcu_nmi_enter_common() function can be invoked both in interrupt
and NMI handlers.  If it is invoked from process context (as opposed
to userspace or idle context) on a nohz_full CPU, it might acquire the
CPU's leaf rcu_node structure's ->lock.  Because this lock is held only
with interrupts disabled, this is safe from an interrupt handler, but
doing so from an NMI handler can result in self-deadlock.

This commit therefore adds "irq" to the "if" condition so as to only
acquire the ->lock from irq handlers or process context, never from
an NMI handler.

Fixes: 3fe13bff6962 ("rcu: Avoid tick_dep_set_cpu() misordering")
Reported-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Cc: <stable@vger.kernel.org> # 5.5.x
kernel/rcu/tree.c