context_tracking: Convert state to atomic_t

Context tracking's state and dynticks counter are going to be merged
in a single field so that both updates can happen atomically and at the
same time. Prepare for that with converting the state into an atomic_t.

[ paulmck: Apply kernel test robot feedback. ]

Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Neeraj Upadhyay <quic_neeraju@quicinc.com>
Cc: Uladzislau Rezki <uladzislau.rezki@sony.com>
Cc: Joel Fernandes <joel@joelfernandes.org>
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: Nicolas Saenz Julienne <nsaenz@kernel.org>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Xiongfeng Wang <wangxiongfeng2@huawei.com>
Cc: Yu Liao <liaoyu15@huawei.com>
Cc: Phil Auld <pauld@redhat.com>
Cc: Paul Gortmaker<paul.gortmaker@windriver.com>
Cc: Alex Belits <abelits@marvell.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Reviewed-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Tested-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>

1 files changed, 94 insertions, 49 deletions

diff --git a/kernel/context_tracking.c b/kernel/context_tracking.c
index e485b6b01537..ca78ff27dc53 100644
--- a/kernel/context_tracking.c
+++ b/kernel/context_tracking.c
@@ -28,8 +28,8 @@ DEFINE_PER_CPU(struct context_tracking, context_tracking) = {
 #ifdef CONFIG_CONTEXT_TRACKING_IDLE
 	.dynticks_nesting = 1,
 	.dynticks_nmi_nesting = DYNTICK_IRQ_NONIDLE,
-	.dynticks = ATOMIC_INIT(1),
 #endif
+	.state = ATOMIC_INIT(RCU_DYNTICKS_IDX),
 };
 EXPORT_SYMBOL_GPL(context_tracking);
 
@@ -76,7 +76,7 @@ static __always_inline void rcu_dynticks_task_trace_exit(void)
  * RCU is watching prior to the call to this function and is no longer
  * watching upon return.
  */
-static noinstr void rcu_dynticks_eqs_enter(void)
+static noinstr void ct_kernel_exit_state(int offset)
 {
 	int seq;
 
@@ -86,9 +86,9 @@ static noinstr void rcu_dynticks_eqs_enter(void)
 	 * next idle sojourn.
 	 */
 	rcu_dynticks_task_trace_enter();  // Before ->dynticks update!
-	seq = rcu_dynticks_inc(1);
+	seq = ct_state_inc(offset);
 	// RCU is no longer watching.  Better be in extended quiescent state!
-	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) && (seq & 0x1));
+	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) && (seq & RCU_DYNTICKS_IDX));
 }
 
 /*
@@ -96,7 +96,7 @@ static noinstr void rcu_dynticks_eqs_enter(void)
  * called from an extended quiescent state, that is, RCU is not watching
  * prior to the call to this function and is watching upon return.
  */
-static noinstr void rcu_dynticks_eqs_exit(void)
+static noinstr void ct_kernel_enter_state(int offset)
 {
 	int seq;
 
@@ -105,10 +105,10 @@ static noinstr void rcu_dynticks_eqs_exit(void)
 	 * and we also must force ordering with the next RCU read-side
 	 * critical section.
 	 */
-	seq = rcu_dynticks_inc(1);
+	seq = ct_state_inc(offset);
 	// RCU is now watching.  Better not be in an extended quiescent state!
 	rcu_dynticks_task_trace_exit();  // After ->dynticks update!
-	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) && !(seq & 0x1));
+	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) && !(seq & RCU_DYNTICKS_IDX));
 }
 
 /*
@@ -119,7 +119,7 @@ static noinstr void rcu_dynticks_eqs_exit(void)
  * the possibility of usermode upcalls having messed up our count
  * of interrupt nesting level during the prior busy period.
  */
-static void noinstr rcu_eqs_enter(bool user)
+static void noinstr ct_kernel_exit(bool user, int offset)
 {
 	struct context_tracking *ct = this_cpu_ptr(&context_tracking);
 
@@ -139,13 +139,13 @@ static void noinstr rcu_eqs_enter(bool user)
 	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) && !user && !is_idle_task(current));
 	rcu_preempt_deferred_qs(current);
 
-	// instrumentation for the noinstr rcu_dynticks_eqs_enter()
-	instrument_atomic_write(&ct->dynticks, sizeof(ct->dynticks));
+	// instrumentation for the noinstr ct_kernel_exit_state()
+	instrument_atomic_write(&ct->state, sizeof(ct->state));
 
 	instrumentation_end();
 	WRITE_ONCE(ct->dynticks_nesting, 0); /* Avoid irq-access tearing. */
 	// RCU is watching here ...
-	rcu_dynticks_eqs_enter();
+	ct_kernel_exit_state(offset);
 	// ... but is no longer watching here.
 	rcu_dynticks_task_enter();
 }
@@ -158,7 +158,7 @@ static void noinstr rcu_eqs_enter(bool user)
  * allow for the possibility of usermode upcalls messing up our count of
  * interrupt nesting level during the busy period that is just now starting.
  */
-static void noinstr rcu_eqs_exit(bool user)
+static void noinstr ct_kernel_enter(bool user, int offset)
 {
 	struct context_tracking *ct = this_cpu_ptr(&context_tracking);
 	long oldval;
@@ -173,12 +173,12 @@ static void noinstr rcu_eqs_exit(bool user)
 	}
 	rcu_dynticks_task_exit();
 	// RCU is not watching here ...
-	rcu_dynticks_eqs_exit();
+	ct_kernel_enter_state(offset);
 	// ... but is watching here.
 	instrumentation_begin();
 
-	// instrumentation for the noinstr rcu_dynticks_eqs_exit()
-	instrument_atomic_write(&ct->dynticks, sizeof(ct->dynticks));
+	// instrumentation for the noinstr ct_kernel_enter_state()
+	instrument_atomic_write(&ct->state, sizeof(ct->state));
 
 	trace_rcu_dyntick(TPS("End"), ct_dynticks_nesting(), 1, ct_dynticks());
 	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) && !user && !is_idle_task(current));
@@ -192,7 +192,7 @@ static void noinstr rcu_eqs_exit(bool user)
  * ct_nmi_exit - inform RCU of exit from NMI context
  *
  * If we are returning from the outermost NMI handler that interrupted an
- * RCU-idle period, update ct->dynticks and ct->dynticks_nmi_nesting
+ * RCU-idle period, update ct->state and ct->dynticks_nmi_nesting
  * to let the RCU grace-period handling know that the CPU is back to
  * being RCU-idle.
  *
@@ -229,12 +229,12 @@ void noinstr ct_nmi_exit(void)
 	trace_rcu_dyntick(TPS("Startirq"), ct_dynticks_nmi_nesting(), 0, ct_dynticks());
 	WRITE_ONCE(ct->dynticks_nmi_nesting, 0); /* Avoid store tearing. */
 
-	// instrumentation for the noinstr rcu_dynticks_eqs_enter()
-	instrument_atomic_write(&ct->dynticks, sizeof(ct->dynticks));
+	// instrumentation for the noinstr ct_kernel_exit_state()
+	instrument_atomic_write(&ct->state, sizeof(ct->state));
 	instrumentation_end();
 
 	// RCU is watching here ...
-	rcu_dynticks_eqs_enter();
+	ct_kernel_exit_state(RCU_DYNTICKS_IDX);
 	// ... but is no longer watching here.
 
 	if (!in_nmi())
@@ -244,7 +244,7 @@ void noinstr ct_nmi_exit(void)
 /**
  * ct_nmi_enter - inform RCU of entry to NMI context
  *
- * If the CPU was idle from RCU's viewpoint, update ct->dynticks and
+ * If the CPU was idle from RCU's viewpoint, update ct->state and
  * ct->dynticks_nmi_nesting to let the RCU grace-period handling know
  * that the CPU is active.  This implementation permits nested NMIs, as
  * long as the nesting level does not overflow an int.  (You will probably
@@ -275,14 +275,14 @@ void noinstr ct_nmi_enter(void)
 			rcu_dynticks_task_exit();
 
 		// RCU is not watching here ...
-		rcu_dynticks_eqs_exit();
+		ct_kernel_enter_state(RCU_DYNTICKS_IDX);
 		// ... but is watching here.
 
 		instrumentation_begin();
 		// instrumentation for the noinstr rcu_dynticks_curr_cpu_in_eqs()
-		instrument_atomic_read(&ct->dynticks, sizeof(ct->dynticks));
-		// instrumentation for the noinstr rcu_dynticks_eqs_exit()
-		instrument_atomic_write(&ct->dynticks, sizeof(ct->dynticks));
+		instrument_atomic_read(&ct->state, sizeof(ct->state));
+		// instrumentation for the noinstr ct_kernel_enter_state()
+		instrument_atomic_write(&ct->state, sizeof(ct->state));
 
 		incby = 1;
 	} else if (!in_nmi()) {
@@ -315,7 +315,7 @@ void noinstr ct_nmi_enter(void)
 void noinstr ct_idle_enter(void)
 {
 	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) && !raw_irqs_disabled());
-	rcu_eqs_enter(false);
+	ct_kernel_exit(false, RCU_DYNTICKS_IDX + CONTEXT_IDLE);
 }
 EXPORT_SYMBOL_GPL(ct_idle_enter);
 
@@ -333,7 +333,7 @@ void noinstr ct_idle_exit(void)
 	unsigned long flags;
 
 	raw_local_irq_save(flags);
-	rcu_eqs_exit(false);
+	ct_kernel_enter(false, RCU_DYNTICKS_IDX - CONTEXT_IDLE);
 	raw_local_irq_restore(flags);
 }
 EXPORT_SYMBOL_GPL(ct_idle_exit);
@@ -421,8 +421,8 @@ void ct_irq_exit_irqson(void)
 	local_irq_restore(flags);
 }
 #else
-static __always_inline void rcu_eqs_enter(bool user) { }
-static __always_inline void rcu_eqs_exit(bool user) { }
+static __always_inline void ct_kernel_exit(bool user, int offset) { }
+static __always_inline void ct_kernel_enter(bool user, int offset) { }
 #endif /* #ifdef CONFIG_CONTEXT_TRACKING_IDLE */
 
 #ifdef CONFIG_CONTEXT_TRACKING_USER
@@ -463,6 +463,7 @@ static __always_inline void context_tracking_recursion_exit(void)
  */
 void noinstr __ct_user_enter(enum ctx_state state)
 {
+	struct context_tracking *ct = this_cpu_ptr(&context_tracking);
 	lockdep_assert_irqs_disabled();
 
 	/* Kernel threads aren't supposed to go to userspace */
@@ -471,8 +472,8 @@ void noinstr __ct_user_enter(enum ctx_state state)
 	if (!context_tracking_recursion_enter())
 		return;
 
-	if ( __this_cpu_read(context_tracking.state) != state) {
-		if (__this_cpu_read(context_tracking.active)) {
+	if (__ct_state() != state) {
+		if (ct->active) {
 			/*
 			 * At this stage, only low level arch entry code remains and
 			 * then we'll run in userspace. We can assume there won't be
@@ -492,28 +493,49 @@ void noinstr __ct_user_enter(enum ctx_state state)
 			 * that will fire and reschedule once we resume in user/guest mode.
 			 */
 			rcu_irq_work_resched();
+
 			/*
 			 * Enter RCU idle mode right before resuming userspace.  No use of RCU
 			 * is permitted between this call and rcu_eqs_exit(). This way the
 			 * CPU doesn't need to maintain the tick for RCU maintenance purposes
 			 * when the CPU runs in userspace.
 			 */
-			rcu_eqs_enter(true);
+			ct_kernel_exit(true, RCU_DYNTICKS_IDX + state);
+
+			/*
+			 * Special case if we only track user <-> kernel transitions for tickless
+			 * cputime accounting but we don't support RCU extended quiescent state.
+			 * In this we case we don't care about any concurrency/ordering.
+			 */
+			if (!IS_ENABLED(CONFIG_CONTEXT_TRACKING_IDLE))
+				atomic_set(&ct->state, state);
+		} else {
+			/*
+			 * Even if context tracking is disabled on this CPU, because it's outside
+			 * the full dynticks mask for example, we still have to keep track of the
+			 * context transitions and states to prevent inconsistency on those of
+			 * other CPUs.
+			 * If a task triggers an exception in userspace, sleep on the exception
+			 * handler and then migrate to another CPU, that new CPU must know where
+			 * the exception returns by the time we call exception_exit().
+			 * This information can only be provided by the previous CPU when it called
+			 * exception_enter().
+			 * OTOH we can spare the calls to vtime and RCU when context_tracking.active
+			 * is false because we know that CPU is not tickless.
+			 */
+			if (!IS_ENABLED(CONFIG_CONTEXT_TRACKING_IDLE)) {
+				/* Tracking for vtime only, no concurrent RCU EQS accounting */
+				atomic_set(&ct->state, state);
+			} else {
+				/*
+				 * Tracking for vtime and RCU EQS. Make sure we don't race
+				 * with NMIs. OTOH we don't care about ordering here since
+				 * RCU only requires RCU_DYNTICKS_IDX increments to be fully
+				 * ordered.
+				 */
+				atomic_add(state, &ct->state);
+			}
 		}
-		/*
-		 * Even if context tracking is disabled on this CPU, because it's outside
-		 * the full dynticks mask for example, we still have to keep track of the
-		 * context transitions and states to prevent inconsistency on those of
-		 * other CPUs.
-		 * If a task triggers an exception in userspace, sleep on the exception
-		 * handler and then migrate to another CPU, that new CPU must know where
-		 * the exception returns by the time we call exception_exit().
-		 * This information can only be provided by the previous CPU when it called
-		 * exception_enter().
-		 * OTOH we can spare the calls to vtime and RCU when context_tracking.active
-		 * is false because we know that CPU is not tickless.
-		 */
-		__this_cpu_write(context_tracking.state, state);
 	}
 	context_tracking_recursion_exit();
 }
@@ -581,24 +603,47 @@ NOKPROBE_SYMBOL(user_enter_callable);
  */
 void noinstr __ct_user_exit(enum ctx_state state)
 {
+	struct context_tracking *ct = this_cpu_ptr(&context_tracking);
+
 	if (!context_tracking_recursion_enter())
 		return;
 
-	if (__this_cpu_read(context_tracking.state) == state) {
-		if (__this_cpu_read(context_tracking.active)) {
+	if (__ct_state() == state) {
+		if (ct->active) {
 			/*
 			 * Exit RCU idle mode while entering the kernel because it can
 			 * run a RCU read side critical section anytime.
 			 */
-			rcu_eqs_exit(true);
+			ct_kernel_enter(true, RCU_DYNTICKS_IDX - state);
 			if (state == CONTEXT_USER) {
 				instrumentation_begin();
 				vtime_user_exit(current);
 				trace_user_exit(0);
 				instrumentation_end();
 			}
+
+			/*
+			 * Special case if we only track user <-> kernel transitions for tickless
+			 * cputime accounting but we don't support RCU extended quiescent state.
+			 * In this we case we don't care about any concurrency/ordering.
+			 */
+			if (!IS_ENABLED(CONFIG_CONTEXT_TRACKING_IDLE))
+				atomic_set(&ct->state, CONTEXT_KERNEL);
+
+		} else {
+			if (!IS_ENABLED(CONFIG_CONTEXT_TRACKING_IDLE)) {
+				/* Tracking for vtime only, no concurrent RCU EQS accounting */
+				atomic_set(&ct->state, CONTEXT_KERNEL);
+			} else {
+				/*
+				 * Tracking for vtime and RCU EQS. Make sure we don't race
+				 * with NMIs. OTOH we don't care about ordering here since
+				 * RCU only requires RCU_DYNTICKS_IDX increments to be fully
+				 * ordered.
+				 */
+				atomic_sub(state, &ct->state);
+			}
 		}
-		__this_cpu_write(context_tracking.state, CONTEXT_KERNEL);
 	}
 	context_tracking_recursion_exit();
 }