Commit 0f41c0dd authored by Paul E. McKenney's avatar Paul E. McKenney
Browse files

rcu: Provide diagnostic option to slow down grace-period scans 



Grace-period scans of the rcu_node combining tree normally
proceed quite quickly, so that it is very difficult to reproduce
races against them.  This commit therefore allows grace-period
pre-initialization and cleanup to be artificially slowed down,
increasing race-reproduction probability.  A pair of pairs of new
Kconfig parameters are provided, RCU_TORTURE_TEST_SLOW_PREINIT to
enable the slowing down of propagating CPU-hotplug changes up the
combining tree along with RCU_TORTURE_TEST_SLOW_PREINIT_DELAY to
specify the delay in jiffies, and RCU_TORTURE_TEST_SLOW_CLEANUP
to enable the slowing down of the end-of-grace-period cleanup scan
along with RCU_TORTURE_TEST_SLOW_CLEANUP_DELAY to specify the delay
in jiffies.  Boot-time parameters named rcutree.gp_preinit_delay and
rcutree.gp_cleanup_delay allow these delays to be specified at boot time.

Signed-off-by: default avatarPaul E. McKenney <paulmck@linux.vnet.ibm.com>
parent 3eaaaf6c

Documentation/kernel-parameters.txt

+14 −2
Original line number Diff line number Diff line
@@ -2992,11 +2992,23 @@ bytes respectively. Such letter suffixes can also be entirely omitted. 
			Set maximum number of finished RCU callbacks to

			process in one batch.



	rcutree.gp_cleanup_delay=	[KNL]

			Set the number of jiffies to delay each step of

			RCU grace-period cleanup.  This only has effect

			when CONFIG_RCU_TORTURE_TEST_SLOW_CLEANUP is set.



	rcutree.gp_init_delay=	[KNL]

			Set the number of jiffies to delay each step of

			RCU grace-period initialization.  This only has

			effect when CONFIG_RCU_TORTURE_TEST_SLOW_INIT is

			set.

			effect when CONFIG_RCU_TORTURE_TEST_SLOW_INIT

			is set.



	rcutree.gp_preinit_delay=	[KNL]

			Set the number of jiffies to delay each step of

			RCU grace-period pre-initialization, that is,

			the propagation of recent CPU-hotplug changes up

			the rcu_node combining tree.  This only has effect

			when CONFIG_RCU_TORTURE_TEST_SLOW_PREINIT is set.



	rcutree.rcu_fanout_leaf= [KNL]

			Increase the number of CPUs assigned to each

kernel/rcu/tree.c

+25 −4
Original line number Diff line number Diff line
@@ -163,6 +163,14 @@ static int kthread_prio = CONFIG_RCU_KTHREAD_PRIO; 
module_param(kthread_prio, int, 0644);



/* Delay in jiffies for grace-period initialization delays, debug only. */



#ifdef CONFIG_RCU_TORTURE_TEST_SLOW_PREINIT

static int gp_preinit_delay = CONFIG_RCU_TORTURE_TEST_SLOW_PREINIT_DELAY;

module_param(gp_preinit_delay, int, 0644);

#else /* #ifdef CONFIG_RCU_TORTURE_TEST_SLOW_PREINIT */

static const int gp_preinit_delay;

#endif /* #else #ifdef CONFIG_RCU_TORTURE_TEST_SLOW_PREINIT */



#ifdef CONFIG_RCU_TORTURE_TEST_SLOW_INIT

static int gp_init_delay = CONFIG_RCU_TORTURE_TEST_SLOW_INIT_DELAY;

module_param(gp_init_delay, int, 0644);
@@ -170,6 +178,13 @@ module_param(gp_init_delay, int, 0644); 
static const int gp_init_delay;

#endif /* #else #ifdef CONFIG_RCU_TORTURE_TEST_SLOW_INIT */



#ifdef CONFIG_RCU_TORTURE_TEST_SLOW_CLEANUP

static int gp_cleanup_delay = CONFIG_RCU_TORTURE_TEST_SLOW_CLEANUP_DELAY;

module_param(gp_cleanup_delay, int, 0644);

#else /* #ifdef CONFIG_RCU_TORTURE_TEST_SLOW_CLEANUP */

static const int gp_cleanup_delay;

#endif /* #else #ifdef CONFIG_RCU_TORTURE_TEST_SLOW_CLEANUP */



/*

 * Number of grace periods between delays, normalized by the duration of

 * the delay.  The longer the the delay, the more the grace periods between
@@ -1742,6 +1757,13 @@ static void note_gp_changes(struct rcu_state *rsp, struct rcu_data *rdp) 
		rcu_gp_kthread_wake(rsp);

}



static void rcu_gp_slow(struct rcu_state *rsp, int delay)

{

	if (delay > 0 &&

	    !(rsp->gpnum % (rcu_num_nodes * PER_RCU_NODE_PERIOD * delay)))

		schedule_timeout_uninterruptible(delay);

}



/*

 * Initialize a new grace period.  Return 0 if no grace period required.

 */
@@ -1784,6 +1806,7 @@ static int rcu_gp_init(struct rcu_state *rsp) 
	 * will handle subsequent offline CPUs.

	 */

	rcu_for_each_leaf_node(rsp, rnp) {

		rcu_gp_slow(rsp, gp_preinit_delay);

		raw_spin_lock_irq(&rnp->lock);

		smp_mb__after_unlock_lock();

		if (rnp->qsmaskinit == rnp->qsmaskinitnext &&
@@ -1840,6 +1863,7 @@ static int rcu_gp_init(struct rcu_state *rsp) 
	 * process finishes, because this kthread handles both.

	 */

	rcu_for_each_node_breadth_first(rsp, rnp) {

		rcu_gp_slow(rsp, gp_init_delay);

		raw_spin_lock_irq(&rnp->lock);

		smp_mb__after_unlock_lock();

		rdp = this_cpu_ptr(rsp->rda);
@@ -1857,10 +1881,6 @@ static int rcu_gp_init(struct rcu_state *rsp) 
		raw_spin_unlock_irq(&rnp->lock);

		cond_resched_rcu_qs();

		WRITE_ONCE(rsp->gp_activity, jiffies);

		if (gp_init_delay > 0 &&

		    !(rsp->gpnum %

		      (rcu_num_nodes * PER_RCU_NODE_PERIOD * gp_init_delay)))

			schedule_timeout_uninterruptible(gp_init_delay);

	}



	return 1;
@@ -1955,6 +1975,7 @@ static void rcu_gp_cleanup(struct rcu_state *rsp) 
		raw_spin_unlock_irq(&rnp->lock);

		cond_resched_rcu_qs();

		WRITE_ONCE(rsp->gp_activity, jiffies);

		rcu_gp_slow(rsp, gp_cleanup_delay);

	}

	rnp = rcu_get_root(rsp);

	raw_spin_lock_irq(&rnp->lock);

lib/Kconfig.debug

+52 −2
Original line number Diff line number Diff line
@@ -1261,12 +1261,38 @@ config RCU_TORTURE_TEST_RUNNABLE 
	  Say N here if you want the RCU torture tests to start only

	  after being manually enabled via /proc.



config RCU_TORTURE_TEST_SLOW_PREINIT

	bool "Slow down RCU grace-period pre-initialization to expose races"

	depends on RCU_TORTURE_TEST

	help

	  This option delays grace-period pre-initialization (the

	  propagation of CPU-hotplug changes up the rcu_node combining

	  tree) for a few jiffies between initializing each pair of

	  consecutive rcu_node structures.  This helps to expose races

	  involving grace-period pre-initialization, in other words, it

	  makes your kernel less stable.  It can also greatly increase

	  grace-period latency, especially on systems with large numbers

	  of CPUs.  This is useful when torture-testing RCU, but in

	  almost no other circumstance.



	  Say Y here if you want your system to crash and hang more often.

	  Say N if you want a sane system.



config RCU_TORTURE_TEST_SLOW_PREINIT_DELAY

	int "How much to slow down RCU grace-period pre-initialization"

	range 0 5

	default 3

	depends on RCU_TORTURE_TEST_SLOW_PREINIT

	help

	  This option specifies the number of jiffies to wait between

	  each rcu_node structure pre-initialization step.



config RCU_TORTURE_TEST_SLOW_INIT

	bool "Slow down RCU grace-period initialization to expose races"

	depends on RCU_TORTURE_TEST

	help

	  This option makes grace-period initialization block for a

	  few jiffies between initializing each pair of consecutive

	  This option delays grace-period initialization for a few

	  jiffies between initializing each pair of consecutive

	  rcu_node structures.	This helps to expose races involving

	  grace-period initialization, in other words, it makes your

	  kernel less stable.  It can also greatly increase grace-period
@@ -1286,6 +1312,30 @@ config RCU_TORTURE_TEST_SLOW_INIT_DELAY 
	  This option specifies the number of jiffies to wait between

	  each rcu_node structure initialization.



config RCU_TORTURE_TEST_SLOW_CLEANUP

	bool "Slow down RCU grace-period cleanup to expose races"

	depends on RCU_TORTURE_TEST

	help

	  This option delays grace-period cleanup for a few jiffies

	  between cleaning up each pair of consecutive rcu_node

	  structures.  This helps to expose races involving grace-period

	  cleanup, in other words, it makes your kernel less stable.

	  It can also greatly increase grace-period latency, especially

	  on systems with large numbers of CPUs.  This is useful when

	  torture-testing RCU, but in almost no other circumstance.



	  Say Y here if you want your system to crash and hang more often.

	  Say N if you want a sane system.



config RCU_TORTURE_TEST_SLOW_CLEANUP_DELAY

	int "How much to slow down RCU grace-period cleanup"

	range 0 5

	default 3

	depends on RCU_TORTURE_TEST_SLOW_CLEANUP

	help

	  This option specifies the number of jiffies to wait between

	  each rcu_node structure cleanup operation.



config RCU_CPU_STALL_TIMEOUT

	int "RCU CPU stall timeout in seconds"

	depends on RCU_STALL_COMMON

tools/testing/selftests/rcutorture/configs/rcu/CFcommon

+2 −0
Original line number Diff line number Diff line 
CONFIG_RCU_TORTURE_TEST=y

CONFIG_PRINTK_TIME=y

CONFIG_RCU_TORTURE_TEST_SLOW_CLEANUP=y

CONFIG_RCU_TORTURE_TEST_SLOW_INIT=y

CONFIG_RCU_TORTURE_TEST_SLOW_PREINIT=y

CRA Git | Maintained and supported by SUSTech CRA and CCSE