Skip to contentSkip to conversation

RCU priority boosting that survives vicious testing

RFC PATCH In review Paul E. McKenney 3 participants 10 messages Feb 1, 2007 – Feb 2, 2007

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

cover RCU priority boosting that survives vicious testing

Hello!

This is yet another update of the RCU priority-boosting patch (see http://lkml.org/lkml/2007/1/24/294 for the previous version). This series contains the RCU-boost patch itself and rcutorture modifications to test it more thoroughly. This version incorporates feedback from Josh Triplett and applies cleanly to 2.6.20-rc6-rt6.

Thanx, Paul

#1 lore raw

RFC PATCH 1/2 RCU priority boosting that survives vicious testing

Here is the RCU priority-boosting patch. Pretty close to the
http://lkml.org/lkml/2007/1/24/295 version. This patch prevents preempted or blocked low-priority RCU readers from indefinitely stalling RCU grace periods.

---

 include/linux/init_task.h  |   12 +
 include/linux/rcupdate.h   |   12 +
 include/linux/rcupreempt.h |   20 +
 include/linux/sched.h      |   16 +
 init/main.c                |    1 
 kernel/Kconfig.preempt     |   32 ++
 kernel/fork.c              |    6 
 kernel/rcupreempt.c        |  528 +++++++++++++++++++++++++++++++++++++++++++++
 kernel/rtmutex.c           |    7 
 kernel/sched.c             |    5 
 10 files changed, 636 insertions(+), 3 deletions(-)
include/linux/init_task.h +12 -0
diff -urpNa -X dontdiff linux-2.6.20-rc4-rt1/include/linux/init_task.h linux-2.6.20-rc4-rt1-rcub/include/linux/init_task.h
@@ -87,6 +87,17 @@ extern struct nsproxy init_nsproxy;
 	.siglock	= __SPIN_LOCK_UNLOCKED(sighand.siglock),	\
 }
 
+#ifdef CONFIG_PREEMPT_RCU_BOOST
+#define INIT_RCU_BOOST_PRIO .rcu_prio	= MAX_PRIO,
+#define INIT_PREEMPT_RCU_BOOST(tsk)					\
+	.rcub_rbdp	= NULL,						\
+	.rcub_state	= RCU_BOOST_IDLE,				\
+	.rcub_entry	= LIST_HEAD_INIT(tsk.rcub_entry),
+#else /* #ifdef CONFIG_PREEMPT_RCU_BOOST */
+#define INIT_RCU_BOOST_PRIO
+#define INIT_PREEMPT_RCU_BOOST(tsk)
+#endif /* #else #ifdef CONFIG_PREEMPT_RCU_BOOST */
+
 extern struct group_info init_groups;
 
 /*
@@ -143,6 +154,7 @@ extern struct group_info init_groups;
 	.pi_lock	= RAW_SPIN_LOCK_UNLOCKED(tsk.pi_lock),		\
 	INIT_TRACE_IRQFLAGS						\
 	INIT_LOCKDEP							\
+	INIT_PREEMPT_RCU_BOOST(tsk)					\
 }
 
 
include/linux/rcupdate.h +12 -0
diff -urpNa -X dontdiff linux-2.6.20-rc4-rt1/include/linux/rcupdate.h linux-2.6.20-rc4-rt1-rcub/include/linux/rcupdate.h
@@ -227,6 +227,18 @@ extern void rcu_barrier(void);
 extern void rcu_init(void);
 extern void rcu_advance_callbacks(int cpu, int user);
 extern void rcu_check_callbacks(int cpu, int user);
+#ifdef CONFIG_PREEMPT_RCU_BOOST
+extern void init_rcu_boost_late(void);
+extern void __rcu_preempt_boost(void);
+#define rcu_preempt_boost() \
+	do { \
+		if (unlikely(current->rcu_read_lock_nesting > 0)) \
+			__rcu_preempt_boost(); \
+	} while (0)
+#else /* #ifdef CONFIG_PREEMPT_RCU_BOOST */
+#define init_rcu_boost_late()
+#define rcu_preempt_boost()
+#endif /* #else #ifdef CONFIG_PREEMPT_RCU_BOOST */
 
 #endif /* __KERNEL__ */
 #endif /* __LINUX_RCUPDATE_H */
include/linux/rcupreempt.h +20 -0
diff -urpNa -X dontdiff linux-2.6.20-rc4-rt1/include/linux/rcupreempt.h linux-2.6.20-rc4-rt1-rcub/include/linux/rcupreempt.h
@@ -42,6 +42,26 @@
 #include <linux/cpumask.h>
 #include <linux/seqlock.h>
 
+#ifdef CONFIG_PREEMPT_RCU_BOOST
+/*
+ * Task state with respect to being RCU-boosted.  This state is changed
+ * by the task itself in response to the following three events:
+ * 1. Preemption (or block on lock) while in RCU read-side critical section.
+ * 2. Outermost rcu_read_unlock() for blocked RCU read-side critical section.
+ *
+ * The RCU-boost task also updates the state when boosting priority.
+ */
+enum rcu_boost_state {
+	RCU_BOOST_IDLE = 0,	   /* Not yet blocked if in RCU read-side. */
+	RCU_BOOST_BLOCKED = 1,	   /* Blocked from RCU read-side. */
+	RCU_BOOSTED = 2,	   /* Boosting complete. */
+	RCU_BOOST_INVALID = 3,	   /* For bogus state sightings. */
+};
+
+#define N_RCU_BOOST_STATE (RCU_BOOST_INVALID + 1)
+
+#endif /* #ifdef CONFIG_PREEMPT_RCU_BOOST */
+
 #define rcu_qsctr_inc(cpu)
 #define rcu_bh_qsctr_inc(cpu)
 #define call_rcu_bh(head, rcu) call_rcu(head, rcu)
include/linux/sched.h +16 -0
diff -urpNa -X dontdiff linux-2.6.20-rc4-rt1/include/linux/sched.h linux-2.6.20-rc4-rt1-rcub/include/linux/sched.h
@@ -699,6 +699,14 @@ struct signal_struct {
 #define is_rt_policy(p)		((p) != SCHED_NORMAL && (p) != SCHED_BATCH)
 #define has_rt_policy(p)	unlikely(is_rt_policy((p)->policy))
 
+#ifdef CONFIG_PREEMPT_RCU_BOOST
+#define set_rcu_prio(p, prio)  ((p)->rcu_prio = (prio))
+#define get_rcu_prio(p) ((p)->rcu_prio)
+#else /* #ifdef CONFIG_PREEMPT_RCU_BOOST */
+#define set_rcu_prio(p, prio)  do { } while (0)
+#define get_rcu_prio(p) MAX_PRIO
+#endif /* #else #ifdef CONFIG_PREEMPT_RCU_BOOST */
+
 /*
  * Some day this will be a full-fledged user tracking system..
  */
@@ -982,6 +990,9 @@ struct task_struct {
 #endif
 	int load_weight;	/* for niceness load balancing purposes */
 	int prio, static_prio, normal_prio;
+#ifdef CONFIG_PREEMPT_RCU_BOOST
+	int rcu_prio;
+#endif /* #ifdef CONFIG_PREEMPT_RCU_BOOST */
 	struct list_head run_list;
 	struct prio_array *array;
 
@@ -1003,6 +1014,11 @@ struct task_struct {
         atomic_t *rcu_flipctr1;
         atomic_t *rcu_flipctr2;
 #endif
+#ifdef CONFIG_PREEMPT_RCU_BOOST
+	struct rcu_boost_dat *rcub_rbdp;
+	enum rcu_boost_state rcub_state;
+	struct list_head rcub_entry;
+#endif /* #ifdef CONFIG_PREEMPT_RCU_BOOST */
 
 #if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT)
 	struct sched_info sched_info;
init/main.c +1 -0
diff -urpNa -X dontdiff linux-2.6.20-rc4-rt1/init/main.c linux-2.6.20-rc4-rt1-rcub/init/main.c
@@ -712,6 +712,7 @@ static void __init do_basic_setup(void)
 	init_workqueues();
 	usermodehelper_init();
 	driver_init();
+	init_rcu_boost_late();
 
 #ifdef CONFIG_SYSCTL
 	sysctl_init();
kernel/Kconfig.preempt +32 -0
diff -urpNa -X dontdiff linux-2.6.20-rc4-rt1/kernel/Kconfig.preempt linux-2.6.20-rc4-rt1-rcub/kernel/Kconfig.preempt
@@ -176,3 +176,35 @@ config RCU_TRACE
 
 	  Say Y here if you want to enable RCU tracing
 	  Say N if you are unsure.
+
+config PREEMPT_RCU_BOOST
+	bool "Enable priority boosting of RCU read-side critical sections"
+	depends on PREEMPT_RCU
+	default n
+	help
+	  This option permits priority boosting of RCU read-side critical
+	  sections that have been preempted in order to prevent indefinite
+	  delay of grace periods in face of runaway non-realtime processes.
+
+	  Say N if you are unsure.
+
+config PREEMPT_RCU_BOOST_STATS
+	bool "Enable RCU priority-boosting statistic printing"
+	depends on PREEMPT_RCU_BOOST
+	default n
+	help
+	  This option enables debug printk()s of RCU boost statistics,
+	  which are normally only used to debug RCU priority boost
+	  implementations.
+
+	  Say N if you are unsure.
+
+config PREEMPT_RCU_BOOST_STATS_INTERVAL
+	int "RCU priority-boosting statistic printing interval (seconds)"
+	depends on PREEMPT_RCU_BOOST_STATS
+	default 100
+	range 10 86400
+	help
+	  This option controls the timing of debug printk()s of RCU boost
+	  statistics, which are normally only used to debug RCU priority
+	  boost implementations.
kernel/fork.c +6 -0
diff -urpNa -X dontdiff linux-2.6.20-rc4-rt1/kernel/fork.c linux-2.6.20-rc4-rt1-rcub/kernel/fork.c
@@ -1129,6 +1129,12 @@ static struct task_struct *copy_process(
 	p->softirq_context = 0;
 #endif
 	p->pagefault_disabled = 0;
+#ifdef CONFIG_PREEMPT_RCU_BOOST
+	p->rcu_prio = MAX_PRIO;
+	p->rcub_rbdp = NULL;
+	p->rcub_state = RCU_BOOST_IDLE;
+	INIT_LIST_HEAD(&p->rcub_entry);
+#endif
 #ifdef CONFIG_LOCKDEP
 	p->lockdep_depth = 0; /* no locks held yet */
 	p->curr_chain_key = 0;
kernel/rcupreempt.c +528 -0

558 lines not on this page · raw message

kernel/rtmutex.c +4 -3
diff -urpNa -X dontdiff linux-2.6.20-rc4-rt1/kernel/rtmutex.c linux-2.6.20-rc4-rt1-rcub/kernel/rtmutex.c
@@ -105,11 +105,12 @@ static inline void init_lists(struct rt_
  */
 int rt_mutex_getprio(struct task_struct *task)
 {
+	int prio = min(task->normal_prio, get_rcu_prio(task));
+
 	if (likely(!task_has_pi_waiters(task)))
-		return task->normal_prio;
+		return prio;
 
-	return min(task_top_pi_waiter(task)->pi_list_entry.prio,
-		   task->normal_prio);
+	return min(task_top_pi_waiter(task)->pi_list_entry.prio, prio);
 }
 
 /*
kernel/sched.c +5 -0
diff -urpNa -X dontdiff linux-2.6.20-rc4-rt1/kernel/sched.c linux-2.6.20-rc4-rt1-rcub/kernel/sched.c
@@ -1962,6 +1962,7 @@ void fastcall sched_fork(struct task_str
 	 * Make sure we do not leak PI boosting priority to the child:
 	 */
 	p->prio = current->normal_prio;
+	set_rcu_prio(p, MAX_PRIO);
 
 	INIT_LIST_HEAD(&p->run_list);
 	p->array = NULL;
@@ -2044,6 +2045,7 @@ void fastcall wake_up_new_task(struct ta
 			else {
 				p->prio = current->prio;
 				p->normal_prio = current->normal_prio;
+				set_rcu_prio(p, MAX_PRIO);
 				__activate_task_after(p, current, rq);
 			}
 			set_need_resched();
@@ -3986,6 +3988,8 @@ void __sched __schedule(void)
 	}
 	profile_hit(SCHED_PROFILING, __builtin_return_address(0));
 
+	rcu_preempt_boost();
+
 	preempt_disable(); // FIXME: disable irqs here
 	prev = current;
 	release_kernel_lock(prev);
@@ -5725,6 +5729,7 @@ void __cpuinit init_idle(struct task_str
 	idle->sleep_avg = 0;
 	idle->array = NULL;
 	idle->prio = idle->normal_prio = MAX_PRIO;
+	set_rcu_prio(idle, MAX_PRIO);
 	idle->state = TASK_RUNNING;
 	idle->cpus_allowed = cpumask_of_cpu(cpu);
 	set_task_cpu(idle, cpu);
#4 lore raw

RFC PATCH 2/2 RCU priority boosting additions to rcutorture

In reply to Paul E. McKenney

This patch adds an optional preemption kernel thread to the rcutorture tests. This thread sets itself to a low RT priority and chews up CPU in 10-second bursts, verifying that grace periods progress during this 10-second interval. Passes RCU torture testing on a 4-CPU (a pair of 2-CPU dies) 64-bit Xeon system.

---

 rcutorture.c |   90 +++++++++++++++++++++++++++++++++++++++++++++++++----------
 1 files changed, 76 insertions(+), 14 deletions(-)
kernel/rcutorture.c +76 -14
diff -urpNa -X dontdiff linux-2.6.20-rc6-rt4/kernel/rcutorture.c linux-2.6.20-rc6-rt4-rcubtorture/kernel/rcutorture.c
@@ -58,6 +58,7 @@ static int stat_interval;	/* Interval be
 static int verbose;		/* Print more debug info. */
 static int test_no_idle_hz;	/* Test RCU's support for tickless idle CPUs. */
 static int shuffle_interval = 5; /* Interval between shuffles (in sec)*/
+static int preempt_torture = 0;	/* Realtime task preempts torture readers. */
 static char *torture_type = "rcu"; /* What RCU implementation to torture. */
 
 module_param(nreaders, int, 0);
@@ -72,6 +73,8 @@ module_param(test_no_idle_hz, bool, 0);
 MODULE_PARM_DESC(test_no_idle_hz, "Test support for tickless idle CPUs");
 module_param(shuffle_interval, int, 0);
 MODULE_PARM_DESC(shuffle_interval, "Number of seconds between shuffles");
+module_param(preempt_torture, bool, 0);
+MODULE_PARM_DESC(preempt_torture, "Enable realtime preemption torture");
 module_param(torture_type, charp, 0);
 MODULE_PARM_DESC(torture_type, "Type of RCU to torture (rcu, rcu_bh, srcu)");
 
@@ -194,6 +197,8 @@ struct rcu_torture_ops {
 	int (*completed)(void);
 	void (*deferredfree)(struct rcu_torture *p);
 	void (*sync)(void);
+	long (*preemptstart)(void);
+	void (*preemptend)(void);
 	int (*stats)(char *page);
 	char *name;
 };
@@ -258,16 +263,74 @@ static void rcu_torture_deferred_free(st
 	call_rcu(&p->rtort_rcu, rcu_torture_cb);
 }
 
+static struct task_struct *rcu_preeempt_task;
+static unsigned long rcu_torture_preempt_errors = 0;
+
+static int rcu_torture_preempt(void *arg)
+{
+	int completedstart;
+	int err;
+	time_t gcstart;
+	struct sched_param sp;
+
+	sp.sched_priority = MAX_RT_PRIO - 1;
+	err = sched_setscheduler(current, SCHED_RR, &sp);
+	if (err != 0)
+		printk(KERN_ALERT "rcu_torture_preempt() priority err: %d\n",
+		       err);
+	current->flags |= PF_NOFREEZE;
+
+	do {
+		completedstart = rcu_torture_completed();
+		gcstart = xtime.tv_sec;
+		while ((xtime.tv_sec - gcstart < 10) &&
+		       (rcu_torture_completed() == completedstart))
+			cond_resched();
+		if (rcu_torture_completed() == completedstart)
+			rcu_torture_preempt_errors++;
+		schedule_timeout_interruptible(HZ);
+	} while (!kthread_should_stop());
+	return 0;
+}
+
+static long rcu_preempt_start(void)
+{
+	long retval = 0;
+
+	rcu_preeempt_task = kthread_run(rcu_torture_preempt, NULL,
+					"rcu_torture_preempt");
+	if (IS_ERR(rcu_preeempt_task)) {
+		VERBOSE_PRINTK_ERRSTRING("Failed to create preempter");
+		retval = PTR_ERR(rcu_preeempt_task);
+		rcu_preeempt_task = NULL;
+	}
+	return retval;
+}
+
+static void rcu_preempt_end(void)
+{
+	if (rcu_preeempt_task != NULL) {
+		VERBOSE_PRINTK_STRING("Stopping rcu_preempt task");
+		kthread_stop(rcu_preeempt_task);
+	}
+	rcu_preeempt_task = NULL;
+}
+
+static int rcu_preempt_stats(char *page) {
+	return sprintf(page,
+		       "Preemption stalls: %lu\n", rcu_torture_preempt_errors);
+}
+
 static struct rcu_torture_ops rcu_ops = {
-	.init = NULL,
-	.cleanup = NULL,
 	.readlock = rcu_torture_read_lock,
 	.readdelay = rcu_read_delay,
 	.readunlock = rcu_torture_read_unlock,
 	.completed = rcu_torture_completed,
 	.deferredfree = rcu_torture_deferred_free,
 	.sync = synchronize_rcu,
-	.stats = NULL,
+	.preemptstart = rcu_preempt_start,
+	.preemptend = rcu_preempt_end,
+	.stats = rcu_preempt_stats,
 	.name = "rcu"
 };
 
@@ -299,14 +362,12 @@ static void rcu_sync_torture_init(void)
 
 static struct rcu_torture_ops rcu_sync_ops = {
 	.init = rcu_sync_torture_init,
-	.cleanup = NULL,
 	.readlock = rcu_torture_read_lock,
 	.readdelay = rcu_read_delay,
 	.readunlock = rcu_torture_read_unlock,
 	.completed = rcu_torture_completed,
 	.deferredfree = rcu_sync_torture_deferred_free,
 	.sync = synchronize_rcu,
-	.stats = NULL,
 	.name = "rcu_sync"
 };
 
@@ -362,28 +423,23 @@ static void rcu_bh_torture_synchronize(v
 }
 
 static struct rcu_torture_ops rcu_bh_ops = {
-	.init = NULL,
-	.cleanup = NULL,
 	.readlock = rcu_bh_torture_read_lock,
 	.readdelay = rcu_read_delay,  /* just reuse rcu's version. */
 	.readunlock = rcu_bh_torture_read_unlock,
 	.completed = rcu_bh_torture_completed,
 	.deferredfree = rcu_bh_torture_deferred_free,
 	.sync = rcu_bh_torture_synchronize,
-	.stats = NULL,
 	.name = "rcu_bh"
 };
 
 static struct rcu_torture_ops rcu_bh_sync_ops = {
 	.init = rcu_sync_torture_init,
-	.cleanup = NULL,
 	.readlock = rcu_bh_torture_read_lock,
 	.readdelay = rcu_read_delay,  /* just reuse rcu's version. */
 	.readunlock = rcu_bh_torture_read_unlock,
 	.completed = rcu_bh_torture_completed,
 	.deferredfree = rcu_sync_torture_deferred_free,
 	.sync = rcu_bh_torture_synchronize,
-	.stats = NULL,
 	.name = "rcu_bh_sync"
 };
 
@@ -495,14 +551,12 @@ static void sched_torture_synchronize(vo
 
 static struct rcu_torture_ops sched_ops = {
 	.init = rcu_sync_torture_init,
-	.cleanup = NULL,
 	.readlock = sched_torture_read_lock,
 	.readdelay = rcu_read_delay,  /* just reuse rcu's version. */
 	.readunlock = sched_torture_read_unlock,
 	.completed = sched_torture_completed,
 	.deferredfree = rcu_sync_torture_deferred_free,
 	.sync = sched_torture_synchronize,
-	.stats = NULL,
 	.name = "sched"
 };
 
@@ -801,9 +855,10 @@ rcu_torture_print_module_parms(char *tag
 	printk(KERN_ALERT "%s" TORTURE_FLAG
 		"--- %s: nreaders=%d nfakewriters=%d "
 		"stat_interval=%d verbose=%d test_no_idle_hz=%d "
-		"shuffle_interval = %d\n",
+		"shuffle_interval=%d preempt_torture=%d\n",
 		torture_type, tag, nrealreaders, nfakewriters,
-		stat_interval, verbose, test_no_idle_hz, shuffle_interval);
+		stat_interval, verbose, test_no_idle_hz, shuffle_interval,
+		preempt_torture);
 }
 
 static void
@@ -856,6 +911,8 @@ rcu_torture_cleanup(void)
 		kthread_stop(stats_task);
 	}
 	stats_task = NULL;
+	if (preempt_torture && (cur_ops->preemptend != NULL))
+		cur_ops->preemptend();
 
 	/* Wait for all RCU callbacks to fire.  */
 	rcu_barrier();
@@ -997,6 +1054,11 @@ rcu_torture_init(void)
 			goto unwind;
 		}
 	}
+	if (preempt_torture && (cur_ops->preemptstart != NULL)) {
+		firsterr = cur_ops->preemptstart();
+		if (firsterr != 0)
+			goto unwind;
+	}
 	return 0;
 
 unwind:
#5 lore raw

Re: [RFC PATCH -rt 2/2] RCU priority boosting additions to rcutorture

Hi Paul.

Quoted 5 lines · Paul E. McKenney

This patch adds an optional preemption kernel thread to the rcutorture tests. This thread sets itself to a low RT priority and chews up CPU in 10-second bursts, verifying that grace periods progress during this 10-second interval. Passes RCU torture testing on a 4-CPU (a pair of 2-CPU dies) 64-bit Xeon system.

[...]

Quoted 17 lines
kernel/rcutorture.c +16 -0
+static int rcu_torture_preempt(void *arg)
+{
+	int completedstart;
+	int err;
+	time_t gcstart;
+	struct sched_param sp;
+
+	sp.sched_priority = MAX_RT_PRIO - 1;
+	err = sched_setscheduler(current, SCHED_RR, &sp);
+	if (err != 0)
+		printk(KERN_ALERT "rcu_torture_preempt() priority err: %d\n",
+		       err);
+	current->flags |= PF_NOFREEZE;
+
+	do {
+		completedstart = rcu_torture_completed();

kernel/rcutorture.c static void rcu_torture_deferred_free(st · line 294

Quoted 10 lines
kernel/rcutorture.c +10 -0
+		gcstart = xtime.tv_sec;
+		while ((xtime.tv_sec - gcstart < 10) &&
+		       (rcu_torture_completed() == completedstart))
+			cond_resched();
+		if (rcu_torture_completed() == completedstart)
+			rcu_torture_preempt_errors++;
+		schedule_timeout_interruptible(HZ);
+	} while (!kthread_should_stop());
+	return 0;
+}

Does it need to be NOFREEZE? I would think that it should be frozen during a suspend/hibernate.

Regards,

Nigel

#6 lore raw

Re: [RFC PATCH -rt 2/2] RCU priority boosting additions to rcutorture

Quoted 33 lines · Nigel Cunningham

Hi Paul.

Quoted 5 lines · Paul E. McKenney

This patch adds an optional preemption kernel thread to the rcutorture tests. This thread sets itself to a low RT priority and chews up CPU in 10-second bursts, verifying that grace periods progress during this 10-second interval. Passes RCU torture testing on a 4-CPU (a pair of 2-CPU dies) 64-bit Xeon system.

[...]

Quoted 22 lines
kernel/rcutorture.c +21 -0
+static int rcu_torture_preempt(void *arg)
+{
+	int completedstart;
+	int err;
+	time_t gcstart;
+	struct sched_param sp;
+
+	sp.sched_priority = MAX_RT_PRIO - 1;
+	err = sched_setscheduler(current, SCHED_RR, &sp);
+	if (err != 0)
+		printk(KERN_ALERT "rcu_torture_preempt() priority err: %d\n",
+		       err);
+	current->flags |= PF_NOFREEZE;
+
+	do {
+		completedstart = rcu_torture_completed();
+		gcstart = xtime.tv_sec;
+		while ((xtime.tv_sec - gcstart < 10) &&
+		       (rcu_torture_completed() == completedstart))
+			cond_resched();
+		if (rcu_torture_completed() == completedstart)
Quoted 8 lines

kernel/rcutorture.c static void rcu_torture_deferred_free(st · line 294

Quoted 5 lines
kernel/rcutorture.c +5 -0
+			rcu_torture_preempt_errors++;
+		schedule_timeout_interruptible(HZ);
+	} while (!kthread_should_stop());
+	return 0;
+}

Does it need to be NOFREEZE? I would think that it should be frozen during a suspend/hibernate.

Good to hear from you, Nigel!

Should indeed be OK to freeze during suspend/hibernate. Will my schedule_timeout_interruptible() be sufficient to allow the freeze to happen, or do I need to add an explicit try_to_freeze()?

Ah, and I probably need to use the same trick that mtd_blktrans_thread() does to avoid having all my sleeps killed of by an errant signal:

	spin_lock_irq(&current->sighand->siglock);
	sigfillset(&current->blocked);
	recalc_sigpending();
	spin_unlock_irq(&current->sighand->siglock);

Or is such paranoia unnecessary?

Thanx, Paul

#7 lore raw

Re: [RFC PATCH -rt 2/2] RCU priority boosting additions to rcutorture

Hi Paul.

Quoted 1 line · Paul E. McKenney

Good to hear from you, Nigel!

Thanks :)

Quoted 3 lines

Should indeed be OK to freeze during suspend/hibernate. Will my schedule_timeout_interruptible() be sufficient to allow the freeze to happen, or do I need to add an explicit try_to_freeze()?

You need a try_to_freeze() - the process has to enter the refrigerator() function to be counted as frozen.

Quoted 9 lines

Ah, and I probably need to use the same trick that mtd_blktrans_thread() does to avoid having all my sleeps killed of by an errant signal:

	spin_lock_irq(&current->sighand->siglock);
	sigfillset(&current->blocked);
	recalc_sigpending();
	spin_unlock_irq(&current->sighand->siglock);

Or is such paranoia unnecessary?

Yeah. try_to_freeze() is a function now, so you can do something if (try_to_freeze()) goto sleep_again if you so desire.

Regards,

Nigel

#2 lore raw
In reply to Paul E. McKenney
Quoted 4 lines · Paul E. McKenney

Here is the RCU priority-boosting patch. Pretty close to the
http://lkml.org/lkml/2007/1/24/295 version. This patch prevents preempted or blocked low-priority RCU readers from indefinitely stalling RCU grace periods.

thanks - i've applied both patches to -rt and it's looking good so far! Find a small cosmetic fix below.

Ingo

kernel/rcupreempt.c +1 -1
Index: linux/kernel/rcupreempt.c
===================================================================
@@ -489,7 +489,7 @@ void init_rcu_boost_late(void)
 	rcu_boost_task = kthread_run(rcu_booster, NULL, "RCU Prio Booster");
 	if (IS_ERR(rcu_boost_task)) {
 		printk(KERN_ALERT
-		       "Unable to create RCU Priority Booster, errno %d\n",
+		       "Unable to create RCU Priority Booster, errno %ld\n",
 		       -PTR_ERR(rcu_boost_task));
 
 		/*
#8 lore raw

Re: [RFC PATCH -rt 2/2] RCU priority boosting additions to rcutorture

In reply to Nigel Cunningham
Quoted 5 lines · Nigel Cunningham

Hi Paul.

Quoted 1 line · Paul E. McKenney

Good to hear from you, Nigel!

Quoted 8 lines

Thanks :)

Quoted 3 lines

Should indeed be OK to freeze during suspend/hibernate. Will my schedule_timeout_interruptible() be sufficient to allow the freeze to happen, or do I need to add an explicit try_to_freeze()?

You need a try_to_freeze() - the process has to enter the refrigerator() function to be counted as frozen.

Even though it explicitly sleeps each time through the loop? Hmmm...

Quoted 8 lines
Quoted 8 lines

Ah, and I probably need to use the same trick that mtd_blktrans_thread() does to avoid having all my sleeps killed of by an errant signal:

	spin_lock_irq(&current->sighand->siglock);
	sigfillset(&current->blocked);
	recalc_sigpending();
	spin_unlock_irq(&current->sighand->siglock);
Quoted 4 lines
Quoted 1 line

Or is such paranoia unnecessary?

Yeah. try_to_freeze() is a function now, so you can do something if (try_to_freeze()) goto sleep_again if you so desire.

If try_to_freeze() succeeds, do I need to clean up signal state? It didn't look like it to me, but thought I should ask the expert!

My guess is that I can simply do:

	try_to_freeze();
	schedule_timeout_interruptible(HZ);

The schedule_timeout_interruptible() might return early, but if I don't care about getting a shorter than expected sleep, I am OK, right? Besides, one would have to get a couple of very closely spaced freeze_processes() calls for this to happen. ;-)

Thanx, Paul

#9 lore raw

Re: [RFC PATCH -rt 2/2] RCU priority boosting additions to rcutorture

Hi.

Quoted 8 lines · Paul E. McKenney
Quoted 7 lines · Nigel Cunningham

Hi Paul.

Quoted 1 line · Paul E. McKenney

Good to hear from you, Nigel!

Thanks :)

Quoted 8 lines
Quoted 6 lines
Quoted 3 lines

Should indeed be OK to freeze during suspend/hibernate. Will my schedule_timeout_interruptible() be sufficient to allow the freeze to happen, or do I need to add an explicit try_to_freeze()?

You need a try_to_freeze() - the process has to enter the refrigerator() function to be counted as frozen.

Even though it explicitly sleeps each time through the loop? Hmmm...

Yes. Sleeping isn't enough - we have to be sure it won't wake up and perform work at inappropriate times (we don't know what process X might do if it did wake; the result could be an inconsistent image). It therefore needs to enter the refrigerator function so that the freezer code can ensure it remains inactive until the suspend-to-whatever cycle is complete.

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

Ah, and I probably need to use the same trick that mtd_blktrans_thread() does to avoid having all my sleeps killed of by an errant signal:

	spin_lock_irq(&current->sighand->siglock);
	sigfillset(&current->blocked);
	recalc_sigpending();
	spin_unlock_irq(&current->sighand->siglock);
Quoted 7 lines
Quoted 4 lines
Quoted 1 line

Or is such paranoia unnecessary?

Yeah. try_to_freeze() is a function now, so you can do something if (try_to_freeze()) goto sleep_again if you so desire.

If try_to_freeze() succeeds, do I need to clean up signal state? It didn't look like it to me, but thought I should ask the expert!

No, you don't need to. We have recalc_sigpending() in the refrigerator function.

Quoted 9 lines

My guess is that I can simply do:

	try_to_freeze();
	schedule_timeout_interruptible(HZ);

The schedule_timeout_interruptible() might return early, but if I don't care about getting a shorter than expected sleep, I am OK, right? Besides, one would have to get a couple of very closely spaced freeze_processes() calls for this to happen. ;-)

Yes, that looks good to me.

Regards,

Nigel

#3 lore raw
In reply to Ingo Molnar
Quoted 10 lines · Ingo Molnar
Quoted 4 lines · Paul E. McKenney

Here is the RCU priority-boosting patch. Pretty close to the
http://lkml.org/lkml/2007/1/24/295 version. This patch prevents preempted or blocked low-priority RCU readers from indefinitely stalling RCU grace periods.

thanks - i've applied both patches to -rt and it's looking good so far! Find a small cosmetic fix below.

Sorry for the %d noise! :-/

I will send another patch addressing Nigel's freeze_processes() review comments.

Thanx, Paul

Quoted 15 lines

Ingo

Index: linux/kernel/rcupreempt.c
===================================================================

kernel/rcupreempt.c +1 -1
@@ -489,7 +489,7 @@ void init_rcu_boost_late(void)
 	rcu_boost_task = kthread_run(rcu_booster, NULL, "RCU Prio Booster");
 	if (IS_ERR(rcu_boost_task)) {
 		printk(KERN_ALERT
-		       "Unable to create RCU Priority Booster, errno %d\n",
+		       "Unable to create RCU Priority Booster, errno %ld\n",
 		       -PTR_ERR(rcu_boost_task));
 
 		/*