Merge branch 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull scheduler updates from Ingo Molnar:
 "The main changes in this cycle were:

   - another round of rq-clock handling debugging, robustization and
     fixes

   - PELT accounting improvements

   - CPU hotplug related ->cpus_allowed affinity handling fixes all
     around the tree

   - ... plus misc fixes, cleanups and updates"

* 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (35 commits)
  sched/x86: Update reschedule warning text
  crypto: N2 - Replace racy task affinity logic
  cpufreq/sparc-us2e: Replace racy task affinity logic
  cpufreq/sparc-us3: Replace racy task affinity logic
  cpufreq/sh: Replace racy task affinity logic
  cpufreq/ia64: Replace racy task affinity logic
  ACPI/processor: Replace racy task affinity logic
  ACPI/processor: Fix error handling in __acpi_processor_start()
  sparc/sysfs: Replace racy task affinity logic
  powerpc/smp: Replace open coded task affinity logic
  ia64/sn/hwperf: Replace racy task affinity logic
  ia64/salinfo: Replace racy task affinity logic
  workqueue: Provide work_on_cpu_safe()
  ia64/topology: Remove cpus_allowed manipulation
  sched/fair: Move the PELT constants into a generated header
  sched/fair: Increase PELT accuracy for small tasks
  sched/fair: Fix comments
  sched/Documentation: Add 'sched-pelt' tool
  sched/fair: Fix corner case in __accumulate_sum()
  sched/core: Remove 'task' parameter and rename tsk_restore_flags() to current_restore_flags()
  ...

Documentation/scheduler/sched-pelt.c

@@ -0,0 +1,108 @@
+/*
+ * The following program is used to generate the constants for
+ * computing sched averages.
+ *
+ * ==============================================================
+ *		C program (compile with -lm)
+ * ==============================================================
+ */
+
+#include <math.h>
+#include <stdio.h>
+
+#define HALFLIFE 32
+#define SHIFT 32
+
+double y;
+
+void calc_runnable_avg_yN_inv(void)
+{
+	int i;
+	unsigned int x;
+
+	printf("static const u32 runnable_avg_yN_inv[] = {");
+	for (i = 0; i < HALFLIFE; i++) {
+		x = ((1UL<<32)-1)*pow(y, i);
+
+		if (i % 6 == 0) printf("\n\t");
+		printf("0x%8x, ", x);
+	}
+	printf("\n};\n\n");
+}
+
+int sum = 1024;
+
+void calc_runnable_avg_yN_sum(void)
+{
+	int i;
+
+	printf("static const u32 runnable_avg_yN_sum[] = {\n\t    0,");
+	for (i = 1; i <= HALFLIFE; i++) {
+		if (i == 1)
+			sum *= y;
+		else
+			sum = sum*y + 1024*y;
+
+		if (i % 11 == 0)
+			printf("\n\t");
+
+		printf("%5d,", sum);
+	}
+	printf("\n};\n\n");
+}
+
+int n = -1;
+/* first period */
+long max = 1024;
+
+void calc_converged_max(void)
+{
+	long last = 0, y_inv = ((1UL<<32)-1)*y;
+
+	for (; ; n++) {
+		if (n > -1)
+			max = ((max*y_inv)>>SHIFT) + 1024;
+			/*
+			 * This is the same as:
+			 * max = max*y + 1024;
+			 */
+
+		if (last == max)
+			break;
+
+		last = max;
+	}
+	n--;
+	printf("#define LOAD_AVG_PERIOD %d\n", HALFLIFE);
+	printf("#define LOAD_AVG_MAX %ld\n", max);
+//	printf("#define LOAD_AVG_MAX_N %d\n\n", n);
+}
+
+void calc_accumulated_sum_32(void)
+{
+	int i, x = sum;
+
+	printf("static const u32 __accumulated_sum_N32[] = {\n\t     0,");
+	for (i = 1; i <= n/HALFLIFE+1; i++) {
+		if (i > 1)
+			x = x/2 + sum;
+
+		if (i % 6 == 0)
+			printf("\n\t");
+
+		printf("%6d,", x);
+	}
+	printf("\n};\n\n");
+}
+
+void main(void)
+{
+	printf("/* Generated by Documentation/scheduler/sched-pelt; do not modify. */\n\n");
+
+	y = pow(0.5, 1/(double)HALFLIFE);
+
+	calc_runnable_avg_yN_inv();
+//	calc_runnable_avg_yN_sum();
+	calc_converged_max();
+//	calc_accumulated_sum_32();
+}

arch/ia64/kernel/salinfo.c

@@ -179,14 +179,14 @@ struct salinfo_platform_oemdata_parms {
 	const u8 *efi_guid;
 	u8 **oemdata;
 	u64 *oemdata_size;
-	int ret;
 };
 
-static void
+static long
 salinfo_platform_oemdata_cpu(void *context)
 {
 	struct salinfo_platform_oemdata_parms *parms = context;
-	parms->ret = salinfo_platform_oemdata(parms->efi_guid, parms->oemdata, parms->oemdata_size);
+
+	return salinfo_platform_oemdata(parms->efi_guid, parms->oemdata, parms->oemdata_size);
 }
 
 static void
@@ -380,16 +380,7 @@ salinfo_log_release(struct inode *inode, struct file *file)
 	return 0;
 }
 
-static void
-call_on_cpu(int cpu, void (*fn)(void *), void *arg)
-{
-	cpumask_t save_cpus_allowed = current->cpus_allowed;
-	set_cpus_allowed_ptr(current, cpumask_of(cpu));
-	(*fn)(arg);
-	set_cpus_allowed_ptr(current, &save_cpus_allowed);
-}
-
-static void
+static long
 salinfo_log_read_cpu(void *context)
 {
 	struct salinfo_data *data = context;
@@ -399,6 +390,7 @@ salinfo_log_read_cpu(void *context)
 	/* Clear corrected errors as they are read from SAL */
 	if (rh->severity == sal_log_severity_corrected)
 		ia64_sal_clear_state_info(data->type);
+	return 0;
 }
 
 static void
@@ -430,7 +422,7 @@ retry:
 	spin_unlock_irqrestore(&data_saved_lock, flags);
 
 	if (!data->saved_num)
-		call_on_cpu(cpu, salinfo_log_read_cpu, data);
+		work_on_cpu_safe(cpu, salinfo_log_read_cpu, data);
 	if (!data->log_size) {
 		data->state = STATE_NO_DATA;
 		cpumask_clear_cpu(cpu, &data->cpu_event);
@@ -459,11 +451,13 @@ salinfo_log_read(struct file *file, char __user *buffer, size_t count, loff_t *p
 	return simple_read_from_buffer(buffer, count, ppos, buf, bufsize);
 }
 
-static void
+static long
 salinfo_log_clear_cpu(void *context)
 {
 	struct salinfo_data *data = context;
+
 	ia64_sal_clear_state_info(data->type);
+	return 0;
 }
 
 static int
@@ -486,7 +480,7 @@ salinfo_log_clear(struct salinfo_data *data, int cpu)
 	rh = (sal_log_record_header_t *)(data->log_buffer);
 	/* Corrected errors have already been cleared from SAL */
 	if (rh->severity != sal_log_severity_corrected)
-		call_on_cpu(cpu, salinfo_log_clear_cpu, data);
+		work_on_cpu_safe(cpu, salinfo_log_clear_cpu, data);
 	/* clearing a record may make a new record visible */
 	salinfo_log_new_read(cpu, data);
 	if (data->state == STATE_LOG_RECORD) {
@@ -531,9 +525,8 @@ salinfo_log_write(struct file *file, const char __user *buffer, size_t count, lo
 				.oemdata = &data->oemdata,
 				.oemdata_size = &data->oemdata_size
 			};
-			call_on_cpu(cpu, salinfo_platform_oemdata_cpu, &parms);
-			if (parms.ret)
-				count = parms.ret;
+			count = work_on_cpu_safe(cpu, salinfo_platform_oemdata_cpu,
+						 &parms);
 		} else
 			data->oemdata_size = 0;
 	} else

arch/ia64/kernel/topology.c

@@ -355,18 +355,12 @@ static int cache_add_dev(unsigned int cpu)
 	unsigned long i, j;
 	struct cache_info *this_object;
 	int retval = 0;
-	cpumask_t oldmask;
 
 	if (all_cpu_cache_info[cpu].kobj.parent)
 		return 0;
 
-	oldmask = current->cpus_allowed;
-	retval = set_cpus_allowed_ptr(current, cpumask_of(cpu));
-	if (unlikely(retval))
-		return retval;
 
 	retval = cpu_cache_sysfs_init(cpu);
-	set_cpus_allowed_ptr(current, &oldmask);
 	if (unlikely(retval < 0))
 		return retval;
 

arch/ia64/sn/kernel/sn2/sn_hwperf.c

@@ -598,12 +598,17 @@ static void sn_hwperf_call_sal(void *info)
 	op_info->ret = r;
 }
 
+static long sn_hwperf_call_sal_work(void *info)
+{
+	sn_hwperf_call_sal(info);
+	return 0;
+}
+
 static int sn_hwperf_op_cpu(struct sn_hwperf_op_info *op_info)
 {
 	u32 cpu;
 	u32 use_ipi;
 	int r = 0;
-	cpumask_t save_allowed;
 	
 	cpu = (op_info->a->arg & SN_HWPERF_ARG_CPU_MASK) >> 32;
 	use_ipi = op_info->a->arg & SN_HWPERF_ARG_USE_IPI_MASK;
@@ -629,13 +634,9 @@ static int sn_hwperf_op_cpu(struct sn_hwperf_op_info *op_info)
 			/* use an interprocessor interrupt to call SAL */
 			smp_call_function_single(cpu, sn_hwperf_call_sal,
 				op_info, 1);
-		}
-		else {
-			/* migrate the task before calling SAL */ 
-			save_allowed = current->cpus_allowed;
-			set_cpus_allowed_ptr(current, cpumask_of(cpu));
-			sn_hwperf_call_sal(op_info);
-			set_cpus_allowed_ptr(current, &save_allowed);
+		} else {
+			/* Call on the target CPU */
+			work_on_cpu_safe(cpu, sn_hwperf_call_sal_work, op_info);
 		}
 	}
 	r = op_info->ret;

arch/powerpc/kernel/smp.c

@@ -787,24 +787,21 @@ static struct sched_domain_topology_level powerpc_topology[] = {
 	{ NULL, },
 };
 
-void __init smp_cpus_done(unsigned int max_cpus)
+static __init long smp_setup_cpu_workfn(void *data __always_unused)
 {
-	cpumask_var_t old_mask;
+	smp_ops->setup_cpu(boot_cpuid);
+	return 0;
+}
 
-	/* We want the setup_cpu() here to be called from CPU 0, but our
-	 * init thread may have been "borrowed" by another CPU in the meantime
-	 * se we pin us down to CPU 0 for a short while
+void __init smp_cpus_done(unsigned int max_cpus)
+{
+	/*
+	 * We want the setup_cpu() here to be called on the boot CPU, but
+	 * init might run on any CPU, so make sure it's invoked on the boot
+	 * CPU.
 	 */
-	alloc_cpumask_var(&old_mask, GFP_NOWAIT);
-	cpumask_copy(old_mask, &current->cpus_allowed);
-	set_cpus_allowed_ptr(current, cpumask_of(boot_cpuid));
-	
 	if (smp_ops && smp_ops->setup_cpu)
-		smp_ops->setup_cpu(boot_cpuid);
-
-	set_cpus_allowed_ptr(current, old_mask);
-
-	free_cpumask_var(old_mask);
+		work_on_cpu_safe(boot_cpuid, smp_setup_cpu_workfn, NULL);
 
 	if (smp_ops && smp_ops->bringup_done)
 		smp_ops->bringup_done();
@@ -812,7 +809,6 @@ void __init smp_cpus_done(unsigned int max_cpus)
 	dump_numa_cpu_topology();
 
 	set_sched_topology(powerpc_topology);
-
 }
 
 #ifdef CONFIG_HOTPLUG_CPU

arch/sparc/kernel/sysfs.c

@@ -98,27 +98,7 @@ static struct attribute_group mmu_stat_group = {
 	.name = "mmu_stats",
 };
 
-/* XXX convert to rusty's on_one_cpu */
-static unsigned long run_on_cpu(unsigned long cpu,
-			        unsigned long (*func)(unsigned long),
-				unsigned long arg)
-{
-	cpumask_t old_affinity;
-	unsigned long ret;
-
-	cpumask_copy(&old_affinity, &current->cpus_allowed);
-	/* should return -EINVAL to userspace */
-	if (set_cpus_allowed_ptr(current, cpumask_of(cpu)))
-		return 0;
-
-	ret = func(arg);
-
-	set_cpus_allowed_ptr(current, &old_affinity);
-
-	return ret;
-}
-
-static unsigned long read_mmustat_enable(unsigned long junk)
+static long read_mmustat_enable(void *data __maybe_unused)
 {
 	unsigned long ra = 0;
 
@@ -127,11 +107,11 @@ static unsigned long read_mmustat_enable(unsigned long junk)
 	return ra != 0;
 }
 
-static unsigned long write_mmustat_enable(unsigned long val)
+static long write_mmustat_enable(void *data)
 {
-	unsigned long ra, orig_ra;
+	unsigned long ra, orig_ra, *val = data;
 
-	if (val)
+	if (*val)
 		ra = __pa(&per_cpu(mmu_stats, smp_processor_id()));
 	else
 		ra = 0UL;
@@ -142,7 +122,8 @@ static unsigned long write_mmustat_enable(unsigned long val)
 static ssize_t show_mmustat_enable(struct device *s,
 				struct device_attribute *attr, char *buf)
 {
-	unsigned long val = run_on_cpu(s->id, read_mmustat_enable, 0);
+	long val = work_on_cpu(s->id, read_mmustat_enable, NULL);
+
 	return sprintf(buf, "%lx\n", val);
 }
 
@@ -150,13 +131,15 @@ static ssize_t store_mmustat_enable(struct device *s,
 			struct device_attribute *attr, const char *buf,
 			size_t count)
 {
-	unsigned long val, err;
-	int ret = sscanf(buf, "%lu", &val);
+	unsigned long val;
+	long err;
+	int ret;
 
+	ret = sscanf(buf, "%lu", &val);
 	if (ret != 1)
 		return -EINVAL;
 
-	err = run_on_cpu(s->id, write_mmustat_enable, val);
+	err = work_on_cpu(s->id, write_mmustat_enable, &val);
 	if (err)
 		return -EIO;
 

arch/x86/kernel/smp.c

@@ -124,7 +124,7 @@ static bool smp_no_nmi_ipi = false;
 static void native_smp_send_reschedule(int cpu)
 {
 	if (unlikely(cpu_is_offline(cpu))) {
-		WARN_ON(1);
+		WARN(1, "sched: Unexpected reschedule of offline CPU#%d!\n", cpu);
 		return;
 	}
 	apic->send_IPI(cpu, RESCHEDULE_VECTOR);

drivers/acpi/processor_driver.c

@@ -251,6 +251,9 @@ static int __acpi_processor_start(struct acpi_device *device)
 	if (ACPI_SUCCESS(status))
 		return 0;
 
+	result = -ENODEV;
+	acpi_pss_perf_exit(pr, device);
+
 err_power_exit:
 	acpi_processor_power_exit(pr);
 	return result;
@@ -259,11 +262,16 @@ err_power_exit:
 static int acpi_processor_start(struct device *dev)
 {
 	struct acpi_device *device = ACPI_COMPANION(dev);
+	int ret;
 
 	if (!device)
 		return -ENODEV;
 
-	return __acpi_processor_start(device);
+	/* Protect against concurrent CPU hotplug operations */
+	get_online_cpus();
+	ret = __acpi_processor_start(device);
+	put_online_cpus();
+	return ret;
 }
 
 static int acpi_processor_stop(struct device *dev)

drivers/acpi/processor_throttling.c

@@ -62,8 +62,8 @@ struct acpi_processor_throttling_arg {
 #define THROTTLING_POSTCHANGE      (2)
 
 static int acpi_processor_get_throttling(struct acpi_processor *pr);
-int acpi_processor_set_throttling(struct acpi_processor *pr,
-						int state, bool force);
+static int __acpi_processor_set_throttling(struct acpi_processor *pr,
+					   int state, bool force, bool direct);
 
 static int acpi_processor_update_tsd_coord(void)
 {
@@ -891,7 +891,8 @@ static int acpi_processor_get_throttling_ptc(struct acpi_processor *pr)
 			ACPI_DEBUG_PRINT((ACPI_DB_INFO,
 				"Invalid throttling state, reset\n"));
 			state = 0;
-			ret = acpi_processor_set_throttling(pr, state, true);
+			ret = __acpi_processor_set_throttling(pr, state, true,
+							      true);
 			if (ret)
 				return ret;
 		}
@@ -901,36 +902,31 @@ static int acpi_processor_get_throttling_ptc(struct acpi_processor *pr)
 	return 0;
 }
 
-static int acpi_processor_get_throttling(struct acpi_processor *pr)
+static long __acpi_processor_get_throttling(void *data)
 {
-	cpumask_var_t saved_mask;
-	int ret;
+	struct acpi_processor *pr = data;
+
+	return pr->throttling.acpi_processor_get_throttling(pr);
+}
 
+static int acpi_processor_get_throttling(struct acpi_processor *pr)
+{
 	if (!pr)
 		return -EINVAL;
 
 	if (!pr->flags.throttling)
 		return -ENODEV;
 
-	if (!alloc_cpumask_var(&saved_mask, GFP_KERNEL))
-		return -ENOMEM;
-
 	/*
-	 * Migrate task to the cpu pointed by pr.
+	 * This is either called from the CPU hotplug callback of
+	 * processor_driver or via the ACPI probe function. In the latter
+	 * case the CPU is not guaranteed to be online. Both call sites are
+	 * protected against CPU hotplug.
 	 */
-	cpumask_copy(saved_mask, &current->cpus_allowed);
-	/* FIXME: use work_on_cpu() */
-	if (set_cpus_allowed_ptr(current, cpumask_of(pr->id))) {
-		/* Can't migrate to the target pr->id CPU. Exit */
-		free_cpumask_var(saved_mask);
+	if (!cpu_online(pr->id))
 		return -ENODEV;
-	}
-	ret = pr->throttling.acpi_processor_get_throttling(pr);
-	/* restore the previous state */
-	set_cpus_allowed_ptr(current, saved_mask);
-	free_cpumask_var(saved_mask);
 
-	return ret;
+	return work_on_cpu(pr->id, __acpi_processor_get_throttling, pr);
 }
 
 static int acpi_processor_get_fadt_info(struct acpi_processor *pr)
@@ -1080,8 +1076,15 @@ static long acpi_processor_throttling_fn(void *data)
 			arg->target_state, arg->force);
 }
 
-int acpi_processor_set_throttling(struct acpi_processor *pr,
-						int state, bool force)
+static int call_on_cpu(int cpu, long (*fn)(void *), void *arg, bool direct)
+{
+	if (direct)
+		return fn(arg);
+	return work_on_cpu(cpu, fn, arg);
+}
+
+static int __acpi_processor_set_throttling(struct acpi_processor *pr,
+					   int state, bool force, bool direct)
 {
 	int ret = 0;
 	unsigned int i;
@@ -1130,7 +1133,8 @@ int acpi_processor_set_throttling(struct acpi_processor *pr,
 		arg.pr = pr;
 		arg.target_state = state;
 		arg.force = force;
-		ret = work_on_cpu(pr->id, acpi_processor_throttling_fn, &arg);
+		ret = call_on_cpu(pr->id, acpi_processor_throttling_fn, &arg,
+				  direct);
 	} else {
 		/*
 		 * When the T-state coordination is SW_ALL or HW_ALL,
@@ -1163,8 +1167,8 @@ int acpi_processor_set_throttling(struct acpi_processor *pr,
 			arg.pr = match_pr;
 			arg.target_state = state;
 			arg.force = force;
-			ret = work_on_cpu(pr->id, acpi_processor_throttling_fn,
-				&arg);
+			ret = call_on_cpu(pr->id, acpi_processor_throttling_fn,
+					  &arg, direct);
 		}
 	}
 	/*
@@ -1182,6 +1186,12 @@ int acpi_processor_set_throttling(struct acpi_processor *pr,
 	return ret;
 }
 
+int acpi_processor_set_throttling(struct acpi_processor *pr, int state,
+				  bool force)
+{
+	return __acpi_processor_set_throttling(pr, state, force, false);
+}
+
 int acpi_processor_get_throttling_info(struct acpi_processor *pr)
 {
 	int result = 0;

drivers/block/nbd.c

@@ -381,7 +381,7 @@ static int sock_xmit(struct nbd_device *nbd, int index, int send,
 			*sent += result;
 	} while (msg_data_left(&msg));
 
-	tsk_restore_flags(current, pflags, PF_MEMALLOC);
+	current_restore_flags(pflags, PF_MEMALLOC);
 
 	return result;
 }

drivers/cpufreq/ia64-acpi-cpufreq.c

@@ -34,6 +34,11 @@ struct cpufreq_acpi_io {
 	unsigned int				resume;
 };
 
+struct cpufreq_acpi_req {
+	unsigned int		cpu;
+	unsigned int		state;
+};
+
 static struct cpufreq_acpi_io	*acpi_io_data[NR_CPUS];
 
 static struct cpufreq_driver acpi_cpufreq_driver;
@@ -83,8 +88,7 @@ processor_get_pstate (
 static unsigned
 extract_clock (
 	struct cpufreq_acpi_io *data,
-	unsigned value,
-	unsigned int cpu)
+	unsigned value)
 {
 	unsigned long i;
 
@@ -98,60 +102,43 @@ extract_clock (
 }
 
 
-static unsigned int
+static long
 processor_get_freq (
-	struct cpufreq_acpi_io	*data,
-	unsigned int		cpu)
+	void *arg)
 {
-	int			ret = 0;
-	u32			value = 0;
-	cpumask_t		saved_mask;
-	unsigned long 		clock_freq;
+	struct cpufreq_acpi_req *req = arg;
+	unsigned int		cpu = req->cpu;
+	struct cpufreq_acpi_io	*data = acpi_io_data[cpu];
+	u32			value;
+	int			ret;
 
 	pr_debug("processor_get_freq\n");
-
-	saved_mask = current->cpus_allowed;
-	set_cpus_allowed_ptr(current, cpumask_of(cpu));
 	if (smp_processor_id() != cpu)
-		goto migrate_end;
+		return -EAGAIN;
 
 	/* processor_get_pstate gets the instantaneous frequency */
 	ret = processor_get_pstate(&value);
-
 	if (ret) {
-		set_cpus_allowed_ptr(current, &saved_mask);
 		pr_warn("get performance failed with error %d\n", ret);
-		ret = 0;
-		goto migrate_end;
+		return ret;
 	}
-	clock_freq = extract_clock(data, value, cpu);
-	ret = (clock_freq*1000);
-
-migrate_end:
-	set_cpus_allowed_ptr(current, &saved_mask);
-	return ret;
+	return 1000 * extract_clock(data, value);
 }
 
 
-static int
+static long
 processor_set_freq (
-	struct cpufreq_acpi_io	*data,
-	struct cpufreq_policy   *policy,
-	int			state)
+	void *arg)
 {
-	int			ret = 0;
-	u32			value = 0;
-	cpumask_t		saved_mask;
-	int			retval;
+	struct cpufreq_acpi_req *req = arg;
+	unsigned int		cpu = req->cpu;
+	struct cpufreq_acpi_io	*data = acpi_io_data[cpu];
+	int			ret, state = req->state;
+	u32			value;
 
 	pr_debug("processor_set_freq\n");
-
-	saved_mask = current->cpus_allowed;
-	set_cpus_allowed_ptr(current, cpumask_of(policy->cpu));
-	if (smp_processor_id() != policy->cpu) {
-		retval = -EAGAIN;
-		goto migrate_end;
-	}
+	if (smp_processor_id() != cpu)
+		return -EAGAIN;
 
 	if (state == data->acpi_data.state) {
 		if (unlikely(data->resume)) {
@@ -159,8 +146,7 @@ processor_set_freq (
 			data->resume = 0;
 		} else {
 			pr_debug("Already at target state (P%d)\n", state);
-			retval = 0;
-			goto migrate_end;
+			return 0;
 		}
 	}
 
@@ -171,7 +157,6 @@ processor_set_freq (
 	 * First we write the target state's 'control' value to the
 	 * control_register.
 	 */
-
 	value = (u32) data->acpi_data.states[state].control;
 
 	pr_debug("Transitioning to state: 0x%08x\n", value);
@@ -179,17 +164,11 @@ processor_set_freq (
 	ret = processor_set_pstate(value);
 	if (ret) {
 		pr_warn("Transition failed with error %d\n", ret);
-		retval = -ENODEV;
-		goto migrate_end;
+		return -ENODEV;
 	}
 
 	data->acpi_data.state = state;
-
-	retval = 0;
-
-migrate_end:
-	set_cpus_allowed_ptr(current, &saved_mask);
-	return (retval);
+	return 0;
 }
 
 
@@ -197,11 +176,13 @@ static unsigned int
 acpi_cpufreq_get (
 	unsigned int		cpu)
 {
-	struct cpufreq_acpi_io *data = acpi_io_data[cpu];
+	struct cpufreq_acpi_req req;
+	long ret;
 
-	pr_debug("acpi_cpufreq_get\n");
+	req.cpu = cpu;
+	ret = work_on_cpu(cpu, processor_get_freq, &req);
 
-	return processor_get_freq(data, cpu);
+	return ret > 0 ? (unsigned int) ret : 0;
 }
 
 
@@ -210,7 +191,12 @@ acpi_cpufreq_target (
 	struct cpufreq_policy   *policy,
 	unsigned int index)
 {
-	return processor_set_freq(acpi_io_data[policy->cpu], policy, index);
+	struct cpufreq_acpi_req req;
+
+	req.cpu = policy->cpu;
+	req.state = index;
+
+	return work_on_cpu(req.cpu, processor_set_freq, &req);
 }
 
 static int

drivers/cpufreq/sh-cpufreq.c

@@ -30,54 +30,63 @@
 
 static DEFINE_PER_CPU(struct clk, sh_cpuclk);
 
+struct cpufreq_target {
+	struct cpufreq_policy	*policy;
+	unsigned int		freq;
+};
+
 static unsigned int sh_cpufreq_get(unsigned int cpu)
 {
 	return (clk_get_rate(&per_cpu(sh_cpuclk, cpu)) + 500) / 1000;
 }
 
-/*
- * Here we notify other drivers of the proposed change and the final change.
- */
-static int sh_cpufreq_target(struct cpufreq_policy *policy,
-			     unsigned int target_freq,
-			     unsigned int relation)
+static long __sh_cpufreq_target(void *arg)
 {
-	unsigned int cpu = policy->cpu;
+	struct cpufreq_target *target = arg;
+	struct cpufreq_policy *policy = target->policy;
+	int cpu = policy->cpu;
 	struct clk *cpuclk = &per_cpu(sh_cpuclk, cpu);
-	cpumask_t cpus_allowed;
 	struct cpufreq_freqs freqs;
 	struct device *dev;
 	long freq;
 
-	cpus_allowed = current->cpus_allowed;
-	set_cpus_allowed_ptr(current, cpumask_of(cpu));
-
-	BUG_ON(smp_processor_id() != cpu);
+	if (smp_processor_id() != cpu)
+		return -ENODEV;
 
 	dev = get_cpu_device(cpu);
 
 	/* Convert target_freq from kHz to Hz */
-	freq = clk_round_rate(cpuclk, target_freq * 1000);
+	freq = clk_round_rate(cpuclk, target->freq * 1000);
 
 	if (freq < (policy->min * 1000) || freq > (policy->max * 1000))
 		return -EINVAL;
 
-	dev_dbg(dev, "requested frequency %u Hz\n", target_freq * 1000);
+	dev_dbg(dev, "requested frequency %u Hz\n", target->freq * 1000);
 
 	freqs.old	= sh_cpufreq_get(cpu);
 	freqs.new	= (freq + 500) / 1000;
 	freqs.flags	= 0;
 
-	cpufreq_freq_transition_begin(policy, &freqs);
-	set_cpus_allowed_ptr(current, &cpus_allowed);
+	cpufreq_freq_transition_begin(target->policy, &freqs);
 	clk_set_rate(cpuclk, freq);
-	cpufreq_freq_transition_end(policy, &freqs, 0);
+	cpufreq_freq_transition_end(target->policy, &freqs, 0);
 
 	dev_dbg(dev, "set frequency %lu Hz\n", freq);
-
 	return 0;
 }
 
+/*
+ * Here we notify other drivers of the proposed change and the final change.
+ */
+static int sh_cpufreq_target(struct cpufreq_policy *policy,
+			     unsigned int target_freq,
+			     unsigned int relation)
+{
+	struct cpufreq_target data = { .policy = policy, .freq = target_freq };
+
+	return work_on_cpu(policy->cpu, __sh_cpufreq_target, &data);
+}
+
 static int sh_cpufreq_verify(struct cpufreq_policy *policy)
 {
 	struct clk *cpuclk = &per_cpu(sh_cpuclk, policy->cpu);

drivers/cpufreq/sparc-us2e-cpufreq.c

@@ -118,10 +118,6 @@ static void us2e_transition(unsigned long estar, unsigned long new_bits,
 			    unsigned long clock_tick,
 			    unsigned long old_divisor, unsigned long divisor)
 {
-	unsigned long flags;
-
-	local_irq_save(flags);
-
 	estar &= ~ESTAR_MODE_DIV_MASK;
 
 	/* This is based upon the state transition diagram in the IIe manual.  */
@@ -152,8 +148,6 @@ static void us2e_transition(unsigned long estar, unsigned long new_bits,
 	} else {
 		BUG();
 	}
-
-	local_irq_restore(flags);
 }
 
 static unsigned long index_to_estar_mode(unsigned int index)
@@ -229,48 +223,51 @@ static unsigned long estar_to_divisor(unsigned long estar)
 	return ret;
 }
 
+static void __us2e_freq_get(void *arg)
+{
+	unsigned long *estar = arg;
+
+	*estar = read_hbreg(HBIRD_ESTAR_MODE_ADDR);
+}
+
 static unsigned int us2e_freq_get(unsigned int cpu)
 {
-	cpumask_t cpus_allowed;
 	unsigned long clock_tick, estar;
 
-	cpumask_copy(&cpus_allowed, &current->cpus_allowed);
-	set_cpus_allowed_ptr(current, cpumask_of(cpu));
-
 	clock_tick = sparc64_get_clock_tick(cpu) / 1000;
-	estar = read_hbreg(HBIRD_ESTAR_MODE_ADDR);
-
-	set_cpus_allowed_ptr(current, &cpus_allowed);
+	if (smp_call_function_single(cpu, __us2e_freq_get, &estar, 1))
+		return 0;
 
 	return clock_tick / estar_to_divisor(estar);
 }
 
-static int us2e_freq_target(struct cpufreq_policy *policy, unsigned int index)
+static void __us2e_freq_target(void *arg)
 {
-	unsigned int cpu = policy->cpu;
+	unsigned int cpu = smp_processor_id();
+	unsigned int *index = arg;
 	unsigned long new_bits, new_freq;
 	unsigned long clock_tick, divisor, old_divisor, estar;
-	cpumask_t cpus_allowed;
-
-	cpumask_copy(&cpus_allowed, &current->cpus_allowed);
-	set_cpus_allowed_ptr(current, cpumask_of(cpu));
 
 	new_freq = clock_tick = sparc64_get_clock_tick(cpu) / 1000;
-	new_bits = index_to_estar_mode(index);
-	divisor = index_to_divisor(index);
+	new_bits = index_to_estar_mode(*index);
+	divisor = index_to_divisor(*index);
 	new_freq /= divisor;
 
 	estar = read_hbreg(HBIRD_ESTAR_MODE_ADDR);
 
 	old_divisor = estar_to_divisor(estar);
 
-	if (old_divisor != divisor)
+	if (old_divisor != divisor) {
 		us2e_transition(estar, new_bits, clock_tick * 1000,
 				old_divisor, divisor);
+	}
+}
 
-	set_cpus_allowed_ptr(current, &cpus_allowed);
+static int us2e_freq_target(struct cpufreq_policy *policy, unsigned int index)
+{
+	unsigned int cpu = policy->cpu;
 
-	return 0;
+	return smp_call_function_single(cpu, __us2e_freq_target, &index, 1);
 }
 
 static int __init us2e_freq_cpu_init(struct cpufreq_policy *policy)

drivers/cpufreq/sparc-us3-cpufreq.c

@@ -35,22 +35,28 @@ static struct us3_freq_percpu_info *us3_freq_table;
 #define SAFARI_CFG_DIV_32	0x0000000080000000UL
 #define SAFARI_CFG_DIV_MASK	0x00000000C0000000UL
 
-static unsigned long read_safari_cfg(void)
+static void read_safari_cfg(void *arg)
 {
-	unsigned long ret;
+	unsigned long ret, *val = arg;
 
 	__asm__ __volatile__("ldxa	[%%g0] %1, %0"
 			     : "=&r" (ret)
 			     : "i" (ASI_SAFARI_CONFIG));
-	return ret;
+	*val = ret;
 }
 
-static void write_safari_cfg(unsigned long val)
+static void update_safari_cfg(void *arg)
 {
+	unsigned long reg, *new_bits = arg;
+
+	read_safari_cfg(&reg);
+	reg &= ~SAFARI_CFG_DIV_MASK;
+	reg |= *new_bits;
+
 	__asm__ __volatile__("stxa	%0, [%%g0] %1\n\t"
 			     "membar	#Sync"
 			     : /* no outputs */
-			     : "r" (val), "i" (ASI_SAFARI_CONFIG)
+			     : "r" (reg), "i" (ASI_SAFARI_CONFIG)
 			     : "memory");
 }
 
@@ -78,29 +84,17 @@ static unsigned long get_current_freq(unsigned int cpu, unsigned long safari_cfg
 
 static unsigned int us3_freq_get(unsigned int cpu)
 {
-	cpumask_t cpus_allowed;
 	unsigned long reg;
-	unsigned int ret;
-
-	cpumask_copy(&cpus_allowed, &current->cpus_allowed);
-	set_cpus_allowed_ptr(current, cpumask_of(cpu));
-
-	reg = read_safari_cfg();
-	ret = get_current_freq(cpu, reg);
-
-	set_cpus_allowed_ptr(current, &cpus_allowed);
 
-	return ret;
+	if (smp_call_function_single(cpu, read_safari_cfg, &reg, 1))
+		return 0;
+	return get_current_freq(cpu, reg);
 }
 
 static int us3_freq_target(struct cpufreq_policy *policy, unsigned int index)
 {
 	unsigned int cpu = policy->cpu;
-	unsigned long new_bits, new_freq, reg;
-	cpumask_t cpus_allowed;
-
-	cpumask_copy(&cpus_allowed, &current->cpus_allowed);
-	set_cpus_allowed_ptr(current, cpumask_of(cpu));
+	unsigned long new_bits, new_freq;
 
 	new_freq = sparc64_get_clock_tick(cpu) / 1000;
 	switch (index) {
@@ -121,15 +115,7 @@ static int us3_freq_target(struct cpufreq_policy *policy, unsigned int index)
 		BUG();
 	}
 
-	reg = read_safari_cfg();
-
-	reg &= ~SAFARI_CFG_DIV_MASK;
-	reg |= new_bits;
-	write_safari_cfg(reg);
-
-	set_cpus_allowed_ptr(current, &cpus_allowed);
-
-	return 0;
+	return smp_call_function_single(cpu, update_safari_cfg, &new_bits, 1);
 }
 
 static int __init us3_freq_cpu_init(struct cpufreq_policy *policy)

drivers/crypto/n2_core.c

@@ -65,6 +65,11 @@ struct spu_queue {
 	struct list_head	list;
 };
 
+struct spu_qreg {
+	struct spu_queue	*queue;
+	unsigned long		type;
+};
+
 static struct spu_queue **cpu_to_cwq;
 static struct spu_queue **cpu_to_mau;
 
@@ -1631,31 +1636,27 @@ static void queue_cache_destroy(void)
 	kmem_cache_destroy(queue_cache[HV_NCS_QTYPE_CWQ - 1]);
 }
 
-static int spu_queue_register(struct spu_queue *p, unsigned long q_type)
+static long spu_queue_register_workfn(void *arg)
 {
-	cpumask_var_t old_allowed;
+	struct spu_qreg *qr = arg;
+	struct spu_queue *p = qr->queue;
+	unsigned long q_type = qr->type;
 	unsigned long hv_ret;
 
-	if (cpumask_empty(&p->sharing))
-		return -EINVAL;
-
-	if (!alloc_cpumask_var(&old_allowed, GFP_KERNEL))
-		return -ENOMEM;
-
-	cpumask_copy(old_allowed, &current->cpus_allowed);
-
-	set_cpus_allowed_ptr(current, &p->sharing);
-
 	hv_ret = sun4v_ncs_qconf(q_type, __pa(p->q),
 				 CWQ_NUM_ENTRIES, &p->qhandle);
 	if (!hv_ret)
 		sun4v_ncs_sethead_marker(p->qhandle, 0);
 
-	set_cpus_allowed_ptr(current, old_allowed);
+	return hv_ret ? -EINVAL : 0;
+}
 
-	free_cpumask_var(old_allowed);
+static int spu_queue_register(struct spu_queue *p, unsigned long q_type)
+{
+	int cpu = cpumask_any_and(&p->sharing, cpu_online_mask);
+	struct spu_qreg qr = { .queue = p, .type = q_type };
 
-	return (hv_ret ? -EINVAL : 0);
+	return work_on_cpu_safe(cpu, spu_queue_register_workfn, &qr);
 }
 
 static int spu_queue_setup(struct spu_queue *p)

drivers/scsi/iscsi_tcp.c

@@ -387,7 +387,7 @@ static int iscsi_sw_tcp_pdu_xmit(struct iscsi_task *task)
 		rc = 0;
 	}
 
-	tsk_restore_flags(current, pflags, PF_MEMALLOC);
+	current_restore_flags(pflags, PF_MEMALLOC);
 	return rc;
 }
 

fs/nfsd/vfs.c

@@ -1004,7 +1004,7 @@ out_nfserr:
 	else
 		err = nfserrno(host_err);
 	if (test_bit(RQ_LOCAL, &rqstp->rq_flags))
-		tsk_restore_flags(current, pflags, PF_LESS_THROTTLE);
+		current_restore_flags(pflags, PF_LESS_THROTTLE);
 	return err;
 }
 

include/linux/sched.h

@@ -1290,10 +1290,10 @@ TASK_PFA_TEST(LMK_WAITING, lmk_waiting)
 TASK_PFA_SET(LMK_WAITING, lmk_waiting)
 
 static inline void
-tsk_restore_flags(struct task_struct *task, unsigned long orig_flags, unsigned long flags)
+current_restore_flags(unsigned long orig_flags, unsigned long flags)
 {
-	task->flags &= ~flags;
-	task->flags |= orig_flags & flags;
+	current->flags &= ~flags;
+	current->flags |= orig_flags & flags;
 }
 
 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);

include/linux/workqueue.h

@@ -608,8 +608,13 @@ static inline long work_on_cpu(int cpu, long (*fn)(void *), void *arg)
 {
 	return fn(arg);
 }
+static inline long work_on_cpu_safe(int cpu, long (*fn)(void *), void *arg)
+{
+	return fn(arg);
+}
 #else
 long work_on_cpu(int cpu, long (*fn)(void *), void *arg);
+long work_on_cpu_safe(int cpu, long (*fn)(void *), void *arg);
 #endif /* CONFIG_SMP */
 
 #ifdef CONFIG_FREEZER

kernel/sched/core.c

@@ -85,21 +85,6 @@ int sysctl_sched_rt_runtime = 950000;
 /* CPUs with isolated domains */
 cpumask_var_t cpu_isolated_map;
 
-/*
- * this_rq_lock - lock this runqueue and disable interrupts.
- */
-static struct rq *this_rq_lock(void)
-	__acquires(rq->lock)
-{
-	struct rq *rq;
-
-	local_irq_disable();
-	rq = this_rq();
-	raw_spin_lock(&rq->lock);
-
-	return rq;
-}
-
 /*
  * __task_rq_lock - lock the rq @p resides on.
  */
@@ -233,8 +218,11 @@ void update_rq_clock(struct rq *rq)
 		return;
 
 #ifdef CONFIG_SCHED_DEBUG
+	if (sched_feat(WARN_DOUBLE_CLOCK))
+		SCHED_WARN_ON(rq->clock_update_flags & RQCF_UPDATED);
 	rq->clock_update_flags |= RQCF_UPDATED;
 #endif
+
 	delta = sched_clock_cpu(cpu_of(rq)) - rq->clock;
 	if (delta < 0)
 		return;
@@ -261,13 +249,14 @@ static void hrtick_clear(struct rq *rq)
 static enum hrtimer_restart hrtick(struct hrtimer *timer)
 {
 	struct rq *rq = container_of(timer, struct rq, hrtick_timer);
+	struct rq_flags rf;
 
 	WARN_ON_ONCE(cpu_of(rq) != smp_processor_id());
 
-	raw_spin_lock(&rq->lock);
+	rq_lock(rq, &rf);
 	update_rq_clock(rq);
 	rq->curr->sched_class->task_tick(rq, rq->curr, 1);
-	raw_spin_unlock(&rq->lock);
+	rq_unlock(rq, &rf);
 
 	return HRTIMER_NORESTART;
 }
@@ -287,11 +276,12 @@ static void __hrtick_restart(struct rq *rq)
 static void __hrtick_start(void *arg)
 {
 	struct rq *rq = arg;
+	struct rq_flags rf;
 
-	raw_spin_lock(&rq->lock);
+	rq_lock(rq, &rf);
 	__hrtick_restart(rq);
 	rq->hrtick_csd_pending = 0;
-	raw_spin_unlock(&rq->lock);
+	rq_unlock(rq, &rf);
 }
 
 /*
@@ -762,17 +752,23 @@ static void set_load_weight(struct task_struct *p)
 
 static inline void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
 {
-	update_rq_clock(rq);
+	if (!(flags & ENQUEUE_NOCLOCK))
+		update_rq_clock(rq);
+
 	if (!(flags & ENQUEUE_RESTORE))
 		sched_info_queued(rq, p);
+
 	p->sched_class->enqueue_task(rq, p, flags);
 }
 
 static inline void dequeue_task(struct rq *rq, struct task_struct *p, int flags)
 {
-	update_rq_clock(rq);
+	if (!(flags & DEQUEUE_NOCLOCK))
+		update_rq_clock(rq);
+
 	if (!(flags & DEQUEUE_SAVE))
 		sched_info_dequeued(rq, p);
+
 	p->sched_class->dequeue_task(rq, p, flags);
 }
 
@@ -946,18 +942,19 @@ void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
  *
  * Returns (locked) new rq. Old rq's lock is released.
  */
-static struct rq *move_queued_task(struct rq *rq, struct task_struct *p, int new_cpu)
+static struct rq *move_queued_task(struct rq *rq, struct rq_flags *rf,
+				   struct task_struct *p, int new_cpu)
 {
 	lockdep_assert_held(&rq->lock);
 
 	p->on_rq = TASK_ON_RQ_MIGRATING;
-	dequeue_task(rq, p, 0);
+	dequeue_task(rq, p, DEQUEUE_NOCLOCK);
 	set_task_cpu(p, new_cpu);
-	raw_spin_unlock(&rq->lock);
+	rq_unlock(rq, rf);
 
 	rq = cpu_rq(new_cpu);
 
-	raw_spin_lock(&rq->lock);
+	rq_lock(rq, rf);
 	BUG_ON(task_cpu(p) != new_cpu);
 	enqueue_task(rq, p, 0);
 	p->on_rq = TASK_ON_RQ_QUEUED;
@@ -980,7 +977,8 @@ struct migration_arg {
  * So we race with normal scheduler movements, but that's OK, as long
  * as the task is no longer on this CPU.
  */
-static struct rq *__migrate_task(struct rq *rq, struct task_struct *p, int dest_cpu)
+static struct rq *__migrate_task(struct rq *rq, struct rq_flags *rf,
+				 struct task_struct *p, int dest_cpu)
 {
 	if (unlikely(!cpu_active(dest_cpu)))
 		return rq;
@@ -989,7 +987,8 @@ static struct rq *__migrate_task(struct rq *rq, struct task_struct *p, int dest_
 	if (!cpumask_test_cpu(dest_cpu, &p->cpus_allowed))
 		return rq;
 
-	rq = move_queued_task(rq, p, dest_cpu);
+	update_rq_clock(rq);
+	rq = move_queued_task(rq, rf, p, dest_cpu);
 
 	return rq;
 }
@@ -1004,6 +1003,7 @@ static int migration_cpu_stop(void *data)
 	struct migration_arg *arg = data;
 	struct task_struct *p = arg->task;
 	struct rq *rq = this_rq();
+	struct rq_flags rf;
 
 	/*
 	 * The original target CPU might have gone down and we might
@@ -1018,7 +1018,7 @@ static int migration_cpu_stop(void *data)
 	sched_ttwu_pending();
 
 	raw_spin_lock(&p->pi_lock);
-	raw_spin_lock(&rq->lock);
+	rq_lock(rq, &rf);
 	/*
 	 * If task_rq(p) != rq, it cannot be migrated here, because we're
 	 * holding rq->lock, if p->on_rq == 0 it cannot get enqueued because
@@ -1026,11 +1026,11 @@ static int migration_cpu_stop(void *data)
 	 */
 	if (task_rq(p) == rq) {
 		if (task_on_rq_queued(p))
-			rq = __migrate_task(rq, p, arg->dest_cpu);
+			rq = __migrate_task(rq, &rf, p, arg->dest_cpu);
 		else
 			p->wake_cpu = arg->dest_cpu;
 	}
-	raw_spin_unlock(&rq->lock);
+	rq_unlock(rq, &rf);
 	raw_spin_unlock(&p->pi_lock);
 
 	local_irq_enable();
@@ -1063,7 +1063,7 @@ void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
 		 * holding rq->lock.
 		 */
 		lockdep_assert_held(&rq->lock);
-		dequeue_task(rq, p, DEQUEUE_SAVE);
+		dequeue_task(rq, p, DEQUEUE_SAVE | DEQUEUE_NOCLOCK);
 	}
 	if (running)
 		put_prev_task(rq, p);
@@ -1071,7 +1071,7 @@ void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
 	p->sched_class->set_cpus_allowed(p, new_mask);
 
 	if (queued)
-		enqueue_task(rq, p, ENQUEUE_RESTORE);
+		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
 	if (running)
 		set_curr_task(rq, p);
 }
@@ -1150,9 +1150,7 @@ static int __set_cpus_allowed_ptr(struct task_struct *p,
 		 * OK, since we're going to drop the lock immediately
 		 * afterwards anyway.
 		 */
-		rq_unpin_lock(rq, &rf);
-		rq = move_queued_task(rq, p, dest_cpu);
-		rq_repin_lock(rq, &rf);
+		rq = move_queued_task(rq, &rf, p, dest_cpu);
 	}
 out:
 	task_rq_unlock(rq, p, &rf);
@@ -1217,16 +1215,24 @@ static void __migrate_swap_task(struct task_struct *p, int cpu)
 {
 	if (task_on_rq_queued(p)) {
 		struct rq *src_rq, *dst_rq;
+		struct rq_flags srf, drf;
 
 		src_rq = task_rq(p);
 		dst_rq = cpu_rq(cpu);
 
+		rq_pin_lock(src_rq, &srf);
+		rq_pin_lock(dst_rq, &drf);
+
 		p->on_rq = TASK_ON_RQ_MIGRATING;
 		deactivate_task(src_rq, p, 0);
 		set_task_cpu(p, cpu);
 		activate_task(dst_rq, p, 0);
 		p->on_rq = TASK_ON_RQ_QUEUED;
 		check_preempt_curr(dst_rq, p, 0);
+
+		rq_unpin_lock(dst_rq, &drf);
+		rq_unpin_lock(src_rq, &srf);
+
 	} else {
 		/*
 		 * Task isn't running anymore; make it appear like we migrated
@@ -1680,7 +1686,7 @@ static void
 ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags,
 		 struct rq_flags *rf)
 {
-	int en_flags = ENQUEUE_WAKEUP;
+	int en_flags = ENQUEUE_WAKEUP | ENQUEUE_NOCLOCK;
 
 	lockdep_assert_held(&rq->lock);
 
@@ -1726,14 +1732,13 @@ void sched_ttwu_pending(void)
 	struct rq *rq = this_rq();
 	struct llist_node *llist = llist_del_all(&rq->wake_list);
 	struct task_struct *p;
-	unsigned long flags;
 	struct rq_flags rf;
 
 	if (!llist)
 		return;
 
-	raw_spin_lock_irqsave(&rq->lock, flags);
-	rq_pin_lock(rq, &rf);
+	rq_lock_irqsave(rq, &rf);
+	update_rq_clock(rq);
 
 	while (llist) {
 		int wake_flags = 0;
@@ -1747,8 +1752,7 @@ void sched_ttwu_pending(void)
 		ttwu_do_activate(rq, p, wake_flags, &rf);
 	}
 
-	rq_unpin_lock(rq, &rf);
-	raw_spin_unlock_irqrestore(&rq->lock, flags);
+	rq_unlock_irqrestore(rq, &rf);
 }
 
 void scheduler_ipi(void)
@@ -1806,7 +1810,7 @@ static void ttwu_queue_remote(struct task_struct *p, int cpu, int wake_flags)
 void wake_up_if_idle(int cpu)
 {
 	struct rq *rq = cpu_rq(cpu);
-	unsigned long flags;
+	struct rq_flags rf;
 
 	rcu_read_lock();
 
@@ -1816,11 +1820,11 @@ void wake_up_if_idle(int cpu)
 	if (set_nr_if_polling(rq->idle)) {
 		trace_sched_wake_idle_without_ipi(cpu);
 	} else {
-		raw_spin_lock_irqsave(&rq->lock, flags);
+		rq_lock_irqsave(rq, &rf);
 		if (is_idle_task(rq->curr))
 			smp_send_reschedule(cpu);
 		/* Else CPU is not idle, do nothing here: */
-		raw_spin_unlock_irqrestore(&rq->lock, flags);
+		rq_unlock_irqrestore(rq, &rf);
 	}
 
 out:
@@ -1846,11 +1850,10 @@ static void ttwu_queue(struct task_struct *p, int cpu, int wake_flags)
 	}
 #endif
 
-	raw_spin_lock(&rq->lock);
-	rq_pin_lock(rq, &rf);
+	rq_lock(rq, &rf);
+	update_rq_clock(rq);
 	ttwu_do_activate(rq, p, wake_flags, &rf);
-	rq_unpin_lock(rq, &rf);
-	raw_spin_unlock(&rq->lock);
+	rq_unlock(rq, &rf);
 }
 
 /*
@@ -2097,11 +2100,9 @@ static void try_to_wake_up_local(struct task_struct *p, struct rq_flags *rf)
 		 * disabled avoiding further scheduler activity on it and we've
 		 * not yet picked a replacement task.
 		 */
-		rq_unpin_lock(rq, rf);
-		raw_spin_unlock(&rq->lock);
+		rq_unlock(rq, rf);
 		raw_spin_lock(&p->pi_lock);
-		raw_spin_lock(&rq->lock);
-		rq_repin_lock(rq, rf);
+		rq_relock(rq, rf);
 	}
 
 	if (!(p->state & TASK_NORMAL))
@@ -2114,7 +2115,7 @@ static void try_to_wake_up_local(struct task_struct *p, struct rq_flags *rf)
 			delayacct_blkio_end();
 			atomic_dec(&rq->nr_iowait);
 		}
-		ttwu_activate(rq, p, ENQUEUE_WAKEUP);
+		ttwu_activate(rq, p, ENQUEUE_WAKEUP | ENQUEUE_NOCLOCK);
 	}
 
 	ttwu_do_wakeup(rq, p, 0, rf);
@@ -2555,7 +2556,7 @@ void wake_up_new_task(struct task_struct *p)
 	update_rq_clock(rq);
 	post_init_entity_util_avg(&p->se);
 
-	activate_task(rq, p, 0);
+	activate_task(rq, p, ENQUEUE_NOCLOCK);
 	p->on_rq = TASK_ON_RQ_QUEUED;
 	trace_sched_wakeup_new(p);
 	check_preempt_curr(rq, p, WF_FORK);
@@ -3093,15 +3094,18 @@ void scheduler_tick(void)
 	int cpu = smp_processor_id();
 	struct rq *rq = cpu_rq(cpu);
 	struct task_struct *curr = rq->curr;
+	struct rq_flags rf;
 
 	sched_clock_tick();
 
-	raw_spin_lock(&rq->lock);
+	rq_lock(rq, &rf);
+
 	update_rq_clock(rq);
 	curr->sched_class->task_tick(rq, curr, 0);
 	cpu_load_update_active(rq);
 	calc_global_load_tick(rq);
-	raw_spin_unlock(&rq->lock);
+
+	rq_unlock(rq, &rf);
 
 	perf_event_task_tick();
 
@@ -3386,18 +3390,18 @@ static void __sched notrace __schedule(bool preempt)
 	 * done by the caller to avoid the race with signal_wake_up().
 	 */
 	smp_mb__before_spinlock();
-	raw_spin_lock(&rq->lock);
-	rq_pin_lock(rq, &rf);
+	rq_lock(rq, &rf);
 
 	/* Promote REQ to ACT */
 	rq->clock_update_flags <<= 1;
+	update_rq_clock(rq);
 
 	switch_count = &prev->nivcsw;
 	if (!preempt && prev->state) {
 		if (unlikely(signal_pending_state(prev->state, prev))) {
 			prev->state = TASK_RUNNING;
 		} else {
-			deactivate_task(rq, prev, DEQUEUE_SLEEP);
+			deactivate_task(rq, prev, DEQUEUE_SLEEP | DEQUEUE_NOCLOCK);
 			prev->on_rq = 0;
 
 			if (prev->in_iowait) {
@@ -3421,9 +3425,6 @@ static void __sched notrace __schedule(bool preempt)
 		switch_count = &prev->nvcsw;
 	}
 
-	if (task_on_rq_queued(prev))
-		update_rq_clock(rq);
-
 	next = pick_next_task(rq, prev, &rf);
 	clear_tsk_need_resched(prev);
 	clear_preempt_need_resched();
@@ -3439,8 +3440,7 @@ static void __sched notrace __schedule(bool preempt)
 		rq = context_switch(rq, prev, next, &rf);
 	} else {
 		rq->clock_update_flags &= ~(RQCF_ACT_SKIP|RQCF_REQ_SKIP);
-		rq_unpin_lock(rq, &rf);
-		raw_spin_unlock_irq(&rq->lock);
+		rq_unlock_irq(rq, &rf);
 	}
 
 	balance_callback(rq);
@@ -3684,7 +3684,8 @@ EXPORT_SYMBOL(default_wake_function);
  */
 void rt_mutex_setprio(struct task_struct *p, int prio)
 {
-	int oldprio, queued, running, queue_flag = DEQUEUE_SAVE | DEQUEUE_MOVE;
+	int oldprio, queued, running, queue_flag =
+		DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
 	const struct sched_class *prev_class;
 	struct rq_flags rf;
 	struct rq *rq;
@@ -3805,7 +3806,7 @@ void set_user_nice(struct task_struct *p, long nice)
 	queued = task_on_rq_queued(p);
 	running = task_current(rq, p);
 	if (queued)
-		dequeue_task(rq, p, DEQUEUE_SAVE);
+		dequeue_task(rq, p, DEQUEUE_SAVE | DEQUEUE_NOCLOCK);
 	if (running)
 		put_prev_task(rq, p);
 
@@ -3816,7 +3817,7 @@ void set_user_nice(struct task_struct *p, long nice)
 	delta = p->prio - old_prio;
 
 	if (queued) {
-		enqueue_task(rq, p, ENQUEUE_RESTORE);
+		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
 		/*
 		 * If the task increased its priority or is running and
 		 * lowered its priority, then reschedule its CPU:
@@ -4126,7 +4127,7 @@ static int __sched_setscheduler(struct task_struct *p,
 	const struct sched_class *prev_class;
 	struct rq_flags rf;
 	int reset_on_fork;
-	int queue_flags = DEQUEUE_SAVE | DEQUEUE_MOVE;
+	int queue_flags = DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
 	struct rq *rq;
 
 	/* May grab non-irq protected spin_locks: */
@@ -4923,7 +4924,12 @@ SYSCALL_DEFINE3(sched_getaffinity, pid_t, pid, unsigned int, len,
  */
 SYSCALL_DEFINE0(sched_yield)
 {
-	struct rq *rq = this_rq_lock();
+	struct rq_flags rf;
+	struct rq *rq;
+
+	local_irq_disable();
+	rq = this_rq();
+	rq_lock(rq, &rf);
 
 	schedstat_inc(rq->yld_count);
 	current->sched_class->yield_task(rq);
@@ -4932,9 +4938,8 @@ SYSCALL_DEFINE0(sched_yield)
 	 * Since we are going to call schedule() anyway, there's
 	 * no need to preempt or enable interrupts:
 	 */
-	__release(rq->lock);
-	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
-	do_raw_spin_unlock(&rq->lock);
+	preempt_disable();
+	rq_unlock(rq, &rf);
 	sched_preempt_enable_no_resched();
 
 	schedule();
@@ -5514,7 +5519,7 @@ void sched_setnuma(struct task_struct *p, int nid)
 	p->numa_preferred_nid = nid;
 
 	if (queued)
-		enqueue_task(rq, p, ENQUEUE_RESTORE);
+		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
 	if (running)
 		set_curr_task(rq, p);
 	task_rq_unlock(rq, p, &rf);
@@ -5579,11 +5584,11 @@ static struct task_struct fake_task = {
  * there's no concurrency possible, we hold the required locks anyway
  * because of lock validation efforts.
  */
-static void migrate_tasks(struct rq *dead_rq)
+static void migrate_tasks(struct rq *dead_rq, struct rq_flags *rf)
 {
 	struct rq *rq = dead_rq;
 	struct task_struct *next, *stop = rq->stop;
-	struct rq_flags rf;
+	struct rq_flags orf = *rf;
 	int dest_cpu;
 
 	/*
@@ -5602,9 +5607,7 @@ static void migrate_tasks(struct rq *dead_rq)
 	 * class method both need to have an up-to-date
 	 * value of rq->clock[_task]
 	 */
-	rq_pin_lock(rq, &rf);
 	update_rq_clock(rq);
-	rq_unpin_lock(rq, &rf);
 
 	for (;;) {
 		/*
@@ -5617,8 +5620,7 @@ static void migrate_tasks(struct rq *dead_rq)
 		/*
 		 * pick_next_task() assumes pinned rq->lock:
 		 */
-		rq_repin_lock(rq, &rf);
-		next = pick_next_task(rq, &fake_task, &rf);
+		next = pick_next_task(rq, &fake_task, rf);
 		BUG_ON(!next);
 		next->sched_class->put_prev_task(rq, next);
 
@@ -5631,10 +5633,9 @@ static void migrate_tasks(struct rq *dead_rq)
 		 * because !cpu_active at this point, which means load-balance
 		 * will not interfere. Also, stop-machine.
 		 */
-		rq_unpin_lock(rq, &rf);
-		raw_spin_unlock(&rq->lock);
+		rq_unlock(rq, rf);
 		raw_spin_lock(&next->pi_lock);
-		raw_spin_lock(&rq->lock);
+		rq_relock(rq, rf);
 
 		/*
 		 * Since we're inside stop-machine, _nothing_ should have
@@ -5648,12 +5649,12 @@ static void migrate_tasks(struct rq *dead_rq)
 
 		/* Find suitable destination for @next, with force if needed. */
 		dest_cpu = select_fallback_rq(dead_rq->cpu, next);
-
-		rq = __migrate_task(rq, next, dest_cpu);
+		rq = __migrate_task(rq, rf, next, dest_cpu);
 		if (rq != dead_rq) {
-			raw_spin_unlock(&rq->lock);
+			rq_unlock(rq, rf);
 			rq = dead_rq;
-			raw_spin_lock(&rq->lock);
+			*rf = orf;
+			rq_relock(rq, rf);
 		}
 		raw_spin_unlock(&next->pi_lock);
 	}
@@ -5766,7 +5767,7 @@ static int cpuset_cpu_inactive(unsigned int cpu)
 int sched_cpu_activate(unsigned int cpu)
 {
 	struct rq *rq = cpu_rq(cpu);
-	unsigned long flags;
+	struct rq_flags rf;
 
 	set_cpu_active(cpu, true);
 
@@ -5784,12 +5785,12 @@ int sched_cpu_activate(unsigned int cpu)
 	 * 2) At runtime, if cpuset_cpu_active() fails to rebuild the
 	 *    domains.
 	 */
-	raw_spin_lock_irqsave(&rq->lock, flags);
+	rq_lock_irqsave(rq, &rf);
 	if (rq->rd) {
 		BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
 		set_rq_online(rq);
 	}
-	raw_spin_unlock_irqrestore(&rq->lock, flags);
+	rq_unlock_irqrestore(rq, &rf);
 
 	update_max_interval();
 
@@ -5847,18 +5848,20 @@ int sched_cpu_starting(unsigned int cpu)
 int sched_cpu_dying(unsigned int cpu)
 {
 	struct rq *rq = cpu_rq(cpu);
-	unsigned long flags;
+	struct rq_flags rf;
 
 	/* Handle pending wakeups and then migrate everything off */
 	sched_ttwu_pending();
-	raw_spin_lock_irqsave(&rq->lock, flags);
+
+	rq_lock_irqsave(rq, &rf);
 	if (rq->rd) {
 		BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
 		set_rq_offline(rq);
 	}
-	migrate_tasks(rq);
+	migrate_tasks(rq, &rf);
 	BUG_ON(rq->nr_running != 1);
-	raw_spin_unlock_irqrestore(&rq->lock, flags);
+	rq_unlock_irqrestore(rq, &rf);
+
 	calc_load_migrate(rq);
 	update_max_interval();
 	nohz_balance_exit_idle(cpu);
@@ -6412,7 +6415,8 @@ static void sched_change_group(struct task_struct *tsk, int type)
  */
 void sched_move_task(struct task_struct *tsk)
 {
-	int queued, running;
+	int queued, running, queue_flags =
+		DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
 	struct rq_flags rf;
 	struct rq *rq;
 
@@ -6423,14 +6427,14 @@ void sched_move_task(struct task_struct *tsk)
 	queued = task_on_rq_queued(tsk);
 
 	if (queued)
-		dequeue_task(rq, tsk, DEQUEUE_SAVE | DEQUEUE_MOVE);
+		dequeue_task(rq, tsk, queue_flags);
 	if (running)
 		put_prev_task(rq, tsk);
 
 	sched_change_group(tsk, TASK_MOVE_GROUP);
 
 	if (queued)
-		enqueue_task(rq, tsk, ENQUEUE_RESTORE | ENQUEUE_MOVE);
+		enqueue_task(rq, tsk, queue_flags);
 	if (running)
 		set_curr_task(rq, tsk);
 
@@ -7008,14 +7012,15 @@ static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
 	for_each_online_cpu(i) {
 		struct cfs_rq *cfs_rq = tg->cfs_rq[i];
 		struct rq *rq = cfs_rq->rq;
+		struct rq_flags rf;
 
-		raw_spin_lock_irq(&rq->lock);
+		rq_lock_irq(rq, &rf);
 		cfs_rq->runtime_enabled = runtime_enabled;
 		cfs_rq->runtime_remaining = 0;
 
 		if (cfs_rq->throttled)
 			unthrottle_cfs_rq(cfs_rq);
-		raw_spin_unlock_irq(&rq->lock);
+		rq_unlock_irq(rq, &rf);
 	}
 	if (runtime_was_enabled && !runtime_enabled)
 		cfs_bandwidth_usage_dec();

kernel/sched/fair.c

@@ -717,18 +717,12 @@ static u64 sched_vslice(struct cfs_rq *cfs_rq, struct sched_entity *se)
 }
 
 #ifdef CONFIG_SMP
+
+#include "sched-pelt.h"
+
 static int select_idle_sibling(struct task_struct *p, int prev_cpu, int cpu);
 static unsigned long task_h_load(struct task_struct *p);
 
-/*
- * We choose a half-life close to 1 scheduling period.
- * Note: The tables runnable_avg_yN_inv and runnable_avg_yN_sum are
- * dependent on this value.
- */
-#define LOAD_AVG_PERIOD 32
-#define LOAD_AVG_MAX 47742 /* maximum possible load avg */
-#define LOAD_AVG_MAX_N 345 /* number of full periods to produce LOAD_AVG_MAX */
-
 /* Give new sched_entity start runnable values to heavy its load in infant time */
 void init_entity_runnable_average(struct sched_entity *se)
 {
@@ -2733,47 +2727,15 @@ static inline void update_cfs_shares(struct sched_entity *se)
 #endif /* CONFIG_FAIR_GROUP_SCHED */
 
 #ifdef CONFIG_SMP
-/* Precomputed fixed inverse multiplies for multiplication by y^n */
-static const u32 runnable_avg_yN_inv[] = {
-	0xffffffff, 0xfa83b2da, 0xf5257d14, 0xefe4b99a, 0xeac0c6e6, 0xe5b906e6,
-	0xe0ccdeeb, 0xdbfbb796, 0xd744fcc9, 0xd2a81d91, 0xce248c14, 0xc9b9bd85,
-	0xc5672a10, 0xc12c4cc9, 0xbd08a39e, 0xb8fbaf46, 0xb504f333, 0xb123f581,
-	0xad583ee9, 0xa9a15ab4, 0xa5fed6a9, 0xa2704302, 0x9ef5325f, 0x9b8d39b9,
-	0x9837f050, 0x94f4efa8, 0x91c3d373, 0x8ea4398a, 0x8b95c1e3, 0x88980e80,
-	0x85aac367, 0x82cd8698,
-};
-
-/*
- * Precomputed \Sum y^k { 1<=k<=n }.  These are floor(true_value) to prevent
- * over-estimates when re-combining.
- */
-static const u32 runnable_avg_yN_sum[] = {
-	    0, 1002, 1982, 2941, 3880, 4798, 5697, 6576, 7437, 8279, 9103,
-	 9909,10698,11470,12226,12966,13690,14398,15091,15769,16433,17082,
-	17718,18340,18949,19545,20128,20698,21256,21802,22336,22859,23371,
-};
-
-/*
- * Precomputed \Sum y^k { 1<=k<=n, where n%32=0). Values are rolled down to
- * lower integers. See Documentation/scheduler/sched-avg.txt how these
- * were generated:
- */
-static const u32 __accumulated_sum_N32[] = {
-	    0, 23371, 35056, 40899, 43820, 45281,
-	46011, 46376, 46559, 46650, 46696, 46719,
-};
-
 /*
  * Approximate:
  *   val * y^n,    where y^32 ~= 0.5 (~1 scheduling period)
  */
-static __always_inline u64 decay_load(u64 val, u64 n)
+static u64 decay_load(u64 val, u64 n)
 {
 	unsigned int local_n;
 
-	if (!n)
-		return val;
-	else if (unlikely(n > LOAD_AVG_PERIOD * 63))
+	if (unlikely(n > LOAD_AVG_PERIOD * 63))
 		return 0;
 
 	/* after bounds checking we can collapse to 32-bit */
@@ -2795,30 +2757,97 @@ static __always_inline u64 decay_load(u64 val, u64 n)
 	return val;
 }
 
+static u32 __accumulate_pelt_segments(u64 periods, u32 d1, u32 d3)
+{
+	u32 c1, c2, c3 = d3; /* y^0 == 1 */
+
+	/*
+	 * c1 = d1 y^p
+	 */
+	c1 = decay_load((u64)d1, periods);
+
+	/*
+	 *            p-1
+	 * c2 = 1024 \Sum y^n
+	 *            n=1
+	 *
+	 *              inf        inf
+	 *    = 1024 ( \Sum y^n - \Sum y^n - y^0 )
+	 *              n=0        n=p
+	 */
+	c2 = LOAD_AVG_MAX - decay_load(LOAD_AVG_MAX, periods) - 1024;
+
+	return c1 + c2 + c3;
+}
+
+#define cap_scale(v, s) ((v)*(s) >> SCHED_CAPACITY_SHIFT)
+
 /*
- * For updates fully spanning n periods, the contribution to runnable
- * average will be: \Sum 1024*y^n
+ * Accumulate the three separate parts of the sum; d1 the remainder
+ * of the last (incomplete) period, d2 the span of full periods and d3
+ * the remainder of the (incomplete) current period.
+ *
+ *           d1          d2           d3
+ *           ^           ^            ^
+ *           |           |            |
+ *         |<->|<----------------->|<--->|
+ * ... |---x---|------| ... |------|-----x (now)
+ *
+ *                           p-1
+ * u' = (u + d1) y^p + 1024 \Sum y^n + d3 y^0
+ *                           n=1
  *
- * We can compute this reasonably efficiently by combining:
- *   y^PERIOD = 1/2 with precomputed \Sum 1024*y^n {for  n <PERIOD}
+ *    = u y^p +					(Step 1)
+ *
+ *                     p-1
+ *      d1 y^p + 1024 \Sum y^n + d3 y^0		(Step 2)
+ *                     n=1
  */
-static u32 __compute_runnable_contrib(u64 n)
+static __always_inline u32
+accumulate_sum(u64 delta, int cpu, struct sched_avg *sa,
+	       unsigned long weight, int running, struct cfs_rq *cfs_rq)
 {
-	u32 contrib = 0;
+	unsigned long scale_freq, scale_cpu;
+	u32 contrib = (u32)delta; /* p == 0 -> delta < 1024 */
+	u64 periods;
 
-	if (likely(n <= LOAD_AVG_PERIOD))
-		return runnable_avg_yN_sum[n];
-	else if (unlikely(n >= LOAD_AVG_MAX_N))
-		return LOAD_AVG_MAX;
+	scale_freq = arch_scale_freq_capacity(NULL, cpu);
+	scale_cpu = arch_scale_cpu_capacity(NULL, cpu);
 
-	/* Since n < LOAD_AVG_MAX_N, n/LOAD_AVG_PERIOD < 11 */
-	contrib = __accumulated_sum_N32[n/LOAD_AVG_PERIOD];
-	n %= LOAD_AVG_PERIOD;
-	contrib = decay_load(contrib, n);
-	return contrib + runnable_avg_yN_sum[n];
-}
+	delta += sa->period_contrib;
+	periods = delta / 1024; /* A period is 1024us (~1ms) */
 
-#define cap_scale(v, s) ((v)*(s) >> SCHED_CAPACITY_SHIFT)
+	/*
+	 * Step 1: decay old *_sum if we crossed period boundaries.
+	 */
+	if (periods) {
+		sa->load_sum = decay_load(sa->load_sum, periods);
+		if (cfs_rq) {
+			cfs_rq->runnable_load_sum =
+				decay_load(cfs_rq->runnable_load_sum, periods);
+		}
+		sa->util_sum = decay_load((u64)(sa->util_sum), periods);
+
+		/*
+		 * Step 2
+		 */
+		delta %= 1024;
+		contrib = __accumulate_pelt_segments(periods,
+				1024 - sa->period_contrib, delta);
+	}
+	sa->period_contrib = delta;
+
+	contrib = cap_scale(contrib, scale_freq);
+	if (weight) {
+		sa->load_sum += weight * contrib;
+		if (cfs_rq)
+			cfs_rq->runnable_load_sum += weight * contrib;
+	}
+	if (running)
+		sa->util_sum += contrib * scale_cpu;
+
+	return periods;
+}
 
 /*
  * We can represent the historical contribution to runnable average as the
@@ -2849,13 +2878,10 @@ static u32 __compute_runnable_contrib(u64 n)
  *            = u_0 + u_1*y + u_2*y^2 + ... [re-labeling u_i --> u_{i+1}]
  */
 static __always_inline int
-__update_load_avg(u64 now, int cpu, struct sched_avg *sa,
+___update_load_avg(u64 now, int cpu, struct sched_avg *sa,
 		  unsigned long weight, int running, struct cfs_rq *cfs_rq)
 {
-	u64 delta, scaled_delta, periods;
-	u32 contrib;
-	unsigned int delta_w, scaled_delta_w, decayed = 0;
-	unsigned long scale_freq, scale_cpu;
+	u64 delta;
 
 	delta = now - sa->last_update_time;
 	/*
@@ -2874,83 +2900,52 @@ __update_load_avg(u64 now, int cpu, struct sched_avg *sa,
 	delta >>= 10;
 	if (!delta)
 		return 0;
-	sa->last_update_time = now;
-
-	scale_freq = arch_scale_freq_capacity(NULL, cpu);
-	scale_cpu = arch_scale_cpu_capacity(NULL, cpu);
-
-	/* delta_w is the amount already accumulated against our next period */
-	delta_w = sa->period_contrib;
-	if (delta + delta_w >= 1024) {
-		decayed = 1;
 
-		/* how much left for next period will start over, we don't know yet */
-		sa->period_contrib = 0;
+	sa->last_update_time += delta << 10;
 
-		/*
-		 * Now that we know we're crossing a period boundary, figure
-		 * out how much from delta we need to complete the current
-		 * period and accrue it.
-		 */
-		delta_w = 1024 - delta_w;
-		scaled_delta_w = cap_scale(delta_w, scale_freq);
-		if (weight) {
-			sa->load_sum += weight * scaled_delta_w;
-			if (cfs_rq) {
-				cfs_rq->runnable_load_sum +=
-						weight * scaled_delta_w;
-			}
-		}
-		if (running)
-			sa->util_sum += scaled_delta_w * scale_cpu;
-
-		delta -= delta_w;
-
-		/* Figure out how many additional periods this update spans */
-		periods = delta / 1024;
-		delta %= 1024;
+	/*
+	 * Now we know we crossed measurement unit boundaries. The *_avg
+	 * accrues by two steps:
+	 *
+	 * Step 1: accumulate *_sum since last_update_time. If we haven't
+	 * crossed period boundaries, finish.
+	 */
+	if (!accumulate_sum(delta, cpu, sa, weight, running, cfs_rq))
+		return 0;
 
-		sa->load_sum = decay_load(sa->load_sum, periods + 1);
-		if (cfs_rq) {
-			cfs_rq->runnable_load_sum =
-				decay_load(cfs_rq->runnable_load_sum, periods + 1);
-		}
-		sa->util_sum = decay_load((u64)(sa->util_sum), periods + 1);
-
-		/* Efficiently calculate \sum (1..n_period) 1024*y^i */
-		contrib = __compute_runnable_contrib(periods);
-		contrib = cap_scale(contrib, scale_freq);
-		if (weight) {
-			sa->load_sum += weight * contrib;
-			if (cfs_rq)
-				cfs_rq->runnable_load_sum += weight * contrib;
-		}
-		if (running)
-			sa->util_sum += contrib * scale_cpu;
+	/*
+	 * Step 2: update *_avg.
+	 */
+	sa->load_avg = div_u64(sa->load_sum, LOAD_AVG_MAX);
+	if (cfs_rq) {
+		cfs_rq->runnable_load_avg =
+			div_u64(cfs_rq->runnable_load_sum, LOAD_AVG_MAX);
 	}
+	sa->util_avg = sa->util_sum / LOAD_AVG_MAX;
 
-	/* Remainder of delta accrued against u_0` */
-	scaled_delta = cap_scale(delta, scale_freq);
-	if (weight) {
-		sa->load_sum += weight * scaled_delta;
-		if (cfs_rq)
-			cfs_rq->runnable_load_sum += weight * scaled_delta;
-	}
-	if (running)
-		sa->util_sum += scaled_delta * scale_cpu;
+	return 1;
+}
 
-	sa->period_contrib += delta;
+static int
+__update_load_avg_blocked_se(u64 now, int cpu, struct sched_entity *se)
+{
+	return ___update_load_avg(now, cpu, &se->avg, 0, 0, NULL);
+}
 
-	if (decayed) {
-		sa->load_avg = div_u64(sa->load_sum, LOAD_AVG_MAX);
-		if (cfs_rq) {
-			cfs_rq->runnable_load_avg =
-				div_u64(cfs_rq->runnable_load_sum, LOAD_AVG_MAX);
-		}
-		sa->util_avg = sa->util_sum / LOAD_AVG_MAX;
-	}
+static int
+__update_load_avg_se(u64 now, int cpu, struct cfs_rq *cfs_rq, struct sched_entity *se)
+{
+	return ___update_load_avg(now, cpu, &se->avg,
+				  se->on_rq * scale_load_down(se->load.weight),
+				  cfs_rq->curr == se, NULL);
+}
 
-	return decayed;
+static int
+__update_load_avg_cfs_rq(u64 now, int cpu, struct cfs_rq *cfs_rq)
+{
+	return ___update_load_avg(now, cpu, &cfs_rq->avg,
+			scale_load_down(cfs_rq->load.weight),
+			cfs_rq->curr != NULL, cfs_rq);
 }
 
 /*
@@ -3014,6 +3009,9 @@ static inline void update_tg_load_avg(struct cfs_rq *cfs_rq, int force)
 void set_task_rq_fair(struct sched_entity *se,
 		      struct cfs_rq *prev, struct cfs_rq *next)
 {
+	u64 p_last_update_time;
+	u64 n_last_update_time;
+
 	if (!sched_feat(ATTACH_AGE_LOAD))
 		return;
 
@@ -3024,11 +3022,11 @@ void set_task_rq_fair(struct sched_entity *se,
 	 * time. This will result in the wakee task is less decayed, but giving
 	 * the wakee more load sounds not bad.
 	 */
-	if (se->avg.last_update_time && prev) {
-		u64 p_last_update_time;
-		u64 n_last_update_time;
+	if (!(se->avg.last_update_time && prev))
+		return;
 
 #ifndef CONFIG_64BIT
+	{
 		u64 p_last_update_time_copy;
 		u64 n_last_update_time_copy;
 
@@ -3043,14 +3041,13 @@ void set_task_rq_fair(struct sched_entity *se,
 
 		} while (p_last_update_time != p_last_update_time_copy ||
 			 n_last_update_time != n_last_update_time_copy);
+	}
 #else
-		p_last_update_time = prev->avg.last_update_time;
-		n_last_update_time = next->avg.last_update_time;
+	p_last_update_time = prev->avg.last_update_time;
+	n_last_update_time = next->avg.last_update_time;
 #endif
-		__update_load_avg(p_last_update_time, cpu_of(rq_of(prev)),
-				  &se->avg, 0, 0, NULL);
-		se->avg.last_update_time = n_last_update_time;
-	}
+	__update_load_avg_blocked_se(p_last_update_time, cpu_of(rq_of(prev)), se);
+	se->avg.last_update_time = n_last_update_time;
 }
 
 /* Take into account change of utilization of a child task group */
@@ -3173,6 +3170,36 @@ static inline int propagate_entity_load_avg(struct sched_entity *se)
 	return 1;
 }
 
+/*
+ * Check if we need to update the load and the utilization of a blocked
+ * group_entity:
+ */
+static inline bool skip_blocked_update(struct sched_entity *se)
+{
+	struct cfs_rq *gcfs_rq = group_cfs_rq(se);
+
+	/*
+	 * If sched_entity still have not zero load or utilization, we have to
+	 * decay it:
+	 */
+	if (se->avg.load_avg || se->avg.util_avg)
+		return false;
+
+	/*
+	 * If there is a pending propagation, we have to update the load and
+	 * the utilization of the sched_entity:
+	 */
+	if (gcfs_rq->propagate_avg)
+		return false;
+
+	/*
+	 * Otherwise, the load and the utilization of the sched_entity is
+	 * already zero and there is no pending propagation, so it will be a
+	 * waste of time to try to decay it:
+	 */
+	return true;
+}
+
 #else /* CONFIG_FAIR_GROUP_SCHED */
 
 static inline void update_tg_load_avg(struct cfs_rq *cfs_rq, int force) {}
@@ -3265,8 +3292,7 @@ update_cfs_rq_load_avg(u64 now, struct cfs_rq *cfs_rq, bool update_freq)
 		set_tg_cfs_propagate(cfs_rq);
 	}
 
-	decayed = __update_load_avg(now, cpu_of(rq_of(cfs_rq)), sa,
-		scale_load_down(cfs_rq->load.weight), cfs_rq->curr != NULL, cfs_rq);
+	decayed = __update_load_avg_cfs_rq(now, cpu_of(rq_of(cfs_rq)), cfs_rq);
 
 #ifndef CONFIG_64BIT
 	smp_wmb();
@@ -3298,11 +3324,8 @@ static inline void update_load_avg(struct sched_entity *se, int flags)
 	 * Track task load average for carrying it to new CPU after migrated, and
 	 * track group sched_entity load average for task_h_load calc in migration
 	 */
-	if (se->avg.last_update_time && !(flags & SKIP_AGE_LOAD)) {
-		__update_load_avg(now, cpu, &se->avg,
-			  se->on_rq * scale_load_down(se->load.weight),
-			  cfs_rq->curr == se, NULL);
-	}
+	if (se->avg.last_update_time && !(flags & SKIP_AGE_LOAD))
+		__update_load_avg_se(now, cpu, cfs_rq, se);
 
 	decayed  = update_cfs_rq_load_avg(now, cfs_rq, true);
 	decayed |= propagate_entity_load_avg(se);
@@ -3407,7 +3430,7 @@ void sync_entity_load_avg(struct sched_entity *se)
 	u64 last_update_time;
 
 	last_update_time = cfs_rq_last_update_time(cfs_rq);
-	__update_load_avg(last_update_time, cpu_of(rq_of(cfs_rq)), &se->avg, 0, 0, NULL);
+	__update_load_avg_blocked_se(last_update_time, cpu_of(rq_of(cfs_rq)), se);
 }
 
 /*
@@ -4271,8 +4294,9 @@ static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b,
 	list_for_each_entry_rcu(cfs_rq, &cfs_b->throttled_cfs_rq,
 				throttled_list) {
 		struct rq *rq = rq_of(cfs_rq);
+		struct rq_flags rf;
 
-		raw_spin_lock(&rq->lock);
+		rq_lock(rq, &rf);
 		if (!cfs_rq_throttled(cfs_rq))
 			goto next;
 
@@ -4289,7 +4313,7 @@ static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b,
 			unthrottle_cfs_rq(cfs_rq);
 
 next:
-		raw_spin_unlock(&rq->lock);
+		rq_unlock(rq, &rf);
 
 		if (!remaining)
 			break;
@@ -5097,15 +5121,16 @@ void cpu_load_update_nohz_stop(void)
 	unsigned long curr_jiffies = READ_ONCE(jiffies);
 	struct rq *this_rq = this_rq();
 	unsigned long load;
+	struct rq_flags rf;
 
 	if (curr_jiffies == this_rq->last_load_update_tick)
 		return;
 
 	load = weighted_cpuload(cpu_of(this_rq));
-	raw_spin_lock(&this_rq->lock);
+	rq_lock(this_rq, &rf);
 	update_rq_clock(this_rq);
 	cpu_load_update_nohz(this_rq, curr_jiffies, load);
-	raw_spin_unlock(&this_rq->lock);
+	rq_unlock(this_rq, &rf);
 }
 #else /* !CONFIG_NO_HZ_COMMON */
 static inline void cpu_load_update_nohz(struct rq *this_rq,
@@ -6769,7 +6794,7 @@ static void detach_task(struct task_struct *p, struct lb_env *env)
 	lockdep_assert_held(&env->src_rq->lock);
 
 	p->on_rq = TASK_ON_RQ_MIGRATING;
-	deactivate_task(env->src_rq, p, 0);
+	deactivate_task(env->src_rq, p, DEQUEUE_NOCLOCK);
 	set_task_cpu(p, env->dst_cpu);
 }
 
@@ -6902,7 +6927,7 @@ static void attach_task(struct rq *rq, struct task_struct *p)
 	lockdep_assert_held(&rq->lock);
 
 	BUG_ON(task_rq(p) != rq);
-	activate_task(rq, p, 0);
+	activate_task(rq, p, ENQUEUE_NOCLOCK);
 	p->on_rq = TASK_ON_RQ_QUEUED;
 	check_preempt_curr(rq, p, 0);
 }
@@ -6913,9 +6938,12 @@ static void attach_task(struct rq *rq, struct task_struct *p)
  */
 static void attach_one_task(struct rq *rq, struct task_struct *p)
 {
-	raw_spin_lock(&rq->lock);
+	struct rq_flags rf;
+
+	rq_lock(rq, &rf);
+	update_rq_clock(rq);
 	attach_task(rq, p);
-	raw_spin_unlock(&rq->lock);
+	rq_unlock(rq, &rf);
 }
 
 /*
@@ -6926,8 +6954,10 @@ static void attach_tasks(struct lb_env *env)
 {
 	struct list_head *tasks = &env->tasks;
 	struct task_struct *p;
+	struct rq_flags rf;
 
-	raw_spin_lock(&env->dst_rq->lock);
+	rq_lock(env->dst_rq, &rf);
+	update_rq_clock(env->dst_rq);
 
 	while (!list_empty(tasks)) {
 		p = list_first_entry(tasks, struct task_struct, se.group_node);
@@ -6936,7 +6966,7 @@ static void attach_tasks(struct lb_env *env)
 		attach_task(env->dst_rq, p);
 	}
 
-	raw_spin_unlock(&env->dst_rq->lock);
+	rq_unlock(env->dst_rq, &rf);
 }
 
 #ifdef CONFIG_FAIR_GROUP_SCHED
@@ -6944,9 +6974,9 @@ static void update_blocked_averages(int cpu)
 {
 	struct rq *rq = cpu_rq(cpu);
 	struct cfs_rq *cfs_rq;
-	unsigned long flags;
+	struct rq_flags rf;
 
-	raw_spin_lock_irqsave(&rq->lock, flags);
+	rq_lock_irqsave(rq, &rf);
 	update_rq_clock(rq);
 
 	/*
@@ -6954,6 +6984,8 @@ static void update_blocked_averages(int cpu)
 	 * list_add_leaf_cfs_rq() for details.
 	 */
 	for_each_leaf_cfs_rq(rq, cfs_rq) {
+		struct sched_entity *se;
+
 		/* throttled entities do not contribute to load */
 		if (throttled_hierarchy(cfs_rq))
 			continue;
@@ -6961,11 +6993,12 @@ static void update_blocked_averages(int cpu)
 		if (update_cfs_rq_load_avg(cfs_rq_clock_task(cfs_rq), cfs_rq, true))
 			update_tg_load_avg(cfs_rq, 0);
 
-		/* Propagate pending load changes to the parent */
-		if (cfs_rq->tg->se[cpu])
-			update_load_avg(cfs_rq->tg->se[cpu], 0);
+		/* Propagate pending load changes to the parent, if any: */
+		se = cfs_rq->tg->se[cpu];
+		if (se && !skip_blocked_update(se))
+			update_load_avg(se, 0);
 	}
-	raw_spin_unlock_irqrestore(&rq->lock, flags);
+	rq_unlock_irqrestore(rq, &rf);
 }
 
 /*
@@ -7019,12 +7052,12 @@ static inline void update_blocked_averages(int cpu)
 {
 	struct rq *rq = cpu_rq(cpu);
 	struct cfs_rq *cfs_rq = &rq->cfs;
-	unsigned long flags;
+	struct rq_flags rf;
 
-	raw_spin_lock_irqsave(&rq->lock, flags);
+	rq_lock_irqsave(rq, &rf);
 	update_rq_clock(rq);
 	update_cfs_rq_load_avg(cfs_rq_clock_task(cfs_rq), cfs_rq, true);
-	raw_spin_unlock_irqrestore(&rq->lock, flags);
+	rq_unlock_irqrestore(rq, &rf);
 }
 
 static unsigned long task_h_load(struct task_struct *p)
@@ -7525,6 +7558,7 @@ static inline void update_sd_lb_stats(struct lb_env *env, struct sd_lb_stats *sd
 {
 	struct sched_domain *child = env->sd->child;
 	struct sched_group *sg = env->sd->groups;
+	struct sg_lb_stats *local = &sds->local_stat;
 	struct sg_lb_stats tmp_sgs;
 	int load_idx, prefer_sibling = 0;
 	bool overload = false;
@@ -7541,7 +7575,7 @@ static inline void update_sd_lb_stats(struct lb_env *env, struct sd_lb_stats *sd
 		local_group = cpumask_test_cpu(env->dst_cpu, sched_group_cpus(sg));
 		if (local_group) {
 			sds->local = sg;
-			sgs = &sds->local_stat;
+			sgs = local;
 
 			if (env->idle != CPU_NEWLY_IDLE ||
 			    time_after_eq(jiffies, sg->sgc->next_update))
@@ -7565,8 +7599,8 @@ static inline void update_sd_lb_stats(struct lb_env *env, struct sd_lb_stats *sd
 		 * the tasks on the system).
 		 */
 		if (prefer_sibling && sds->local &&
-		    group_has_capacity(env, &sds->local_stat) &&
-		    (sgs->sum_nr_running > 1)) {
+		    group_has_capacity(env, local) &&
+		    (sgs->sum_nr_running > local->sum_nr_running + 1)) {
 			sgs->group_no_capacity = 1;
 			sgs->group_type = group_classify(sg, sgs);
 		}
@@ -8042,7 +8076,7 @@ static int load_balance(int this_cpu, struct rq *this_rq,
 	struct sched_domain *sd_parent = sd->parent;
 	struct sched_group *group;
 	struct rq *busiest;
-	unsigned long flags;
+	struct rq_flags rf;
 	struct cpumask *cpus = this_cpu_cpumask_var_ptr(load_balance_mask);
 
 	struct lb_env env = {
@@ -8105,7 +8139,7 @@ redo:
 		env.loop_max  = min(sysctl_sched_nr_migrate, busiest->nr_running);
 
 more_balance:
-		raw_spin_lock_irqsave(&busiest->lock, flags);
+		rq_lock_irqsave(busiest, &rf);
 		update_rq_clock(busiest);
 
 		/*
@@ -8122,14 +8156,14 @@ more_balance:
 		 * See task_rq_lock() family for the details.
 		 */
 
-		raw_spin_unlock(&busiest->lock);
+		rq_unlock(busiest, &rf);
 
 		if (cur_ld_moved) {
 			attach_tasks(&env);
 			ld_moved += cur_ld_moved;
 		}
 
-		local_irq_restore(flags);
+		local_irq_restore(rf.flags);
 
 		if (env.flags & LBF_NEED_BREAK) {
 			env.flags &= ~LBF_NEED_BREAK;
@@ -8207,6 +8241,8 @@ more_balance:
 			sd->nr_balance_failed++;
 
 		if (need_active_balance(&env)) {
+			unsigned long flags;
+
 			raw_spin_lock_irqsave(&busiest->lock, flags);
 
 			/* don't kick the active_load_balance_cpu_stop,
@@ -8444,8 +8480,9 @@ static int active_load_balance_cpu_stop(void *data)
 	struct rq *target_rq = cpu_rq(target_cpu);
 	struct sched_domain *sd;
 	struct task_struct *p = NULL;
+	struct rq_flags rf;
 
-	raw_spin_lock_irq(&busiest_rq->lock);
+	rq_lock_irq(busiest_rq, &rf);
 
 	/* make sure the requested cpu hasn't gone down in the meantime */
 	if (unlikely(busiest_cpu != smp_processor_id() ||
@@ -8496,7 +8533,7 @@ static int active_load_balance_cpu_stop(void *data)
 	rcu_read_unlock();
 out_unlock:
 	busiest_rq->active_balance = 0;
-	raw_spin_unlock(&busiest_rq->lock);
+	rq_unlock(busiest_rq, &rf);
 
 	if (p)
 		attach_one_task(target_rq, p);
@@ -8794,10 +8831,13 @@ static void nohz_idle_balance(struct rq *this_rq, enum cpu_idle_type idle)
 		 * do the balance.
 		 */
 		if (time_after_eq(jiffies, rq->next_balance)) {
-			raw_spin_lock_irq(&rq->lock);
+			struct rq_flags rf;
+
+			rq_lock_irq(rq, &rf);
 			update_rq_clock(rq);
 			cpu_load_update_idle(rq);
-			raw_spin_unlock_irq(&rq->lock);
+			rq_unlock_irq(rq, &rf);
+
 			rebalance_domains(rq, CPU_IDLE);
 		}
 
@@ -8988,8 +9028,9 @@ static void task_fork_fair(struct task_struct *p)
 	struct cfs_rq *cfs_rq;
 	struct sched_entity *se = &p->se, *curr;
 	struct rq *rq = this_rq();
+	struct rq_flags rf;
 
-	raw_spin_lock(&rq->lock);
+	rq_lock(rq, &rf);
 	update_rq_clock(rq);
 
 	cfs_rq = task_cfs_rq(current);
@@ -9010,7 +9051,7 @@ static void task_fork_fair(struct task_struct *p)
 	}
 
 	se->vruntime -= cfs_rq->min_vruntime;
-	raw_spin_unlock(&rq->lock);
+	rq_unlock(rq, &rf);
 }
 
 /*
@@ -9372,7 +9413,6 @@ static DEFINE_MUTEX(shares_mutex);
 int sched_group_set_shares(struct task_group *tg, unsigned long shares)
 {
 	int i;
-	unsigned long flags;
 
 	/*
 	 * We can't change the weight of the root cgroup.
@@ -9389,19 +9429,17 @@ int sched_group_set_shares(struct task_group *tg, unsigned long shares)
 	tg->shares = shares;
 	for_each_possible_cpu(i) {
 		struct rq *rq = cpu_rq(i);
-		struct sched_entity *se;
+		struct sched_entity *se = tg->se[i];
+		struct rq_flags rf;
 
-		se = tg->se[i];
 		/* Propagate contribution to hierarchy */
-		raw_spin_lock_irqsave(&rq->lock, flags);
-
-		/* Possible calls to update_curr() need rq clock */
+		rq_lock_irqsave(rq, &rf);
 		update_rq_clock(rq);
 		for_each_sched_entity(se) {
 			update_load_avg(se, UPDATE_TG);
 			update_cfs_shares(se);
 		}
-		raw_spin_unlock_irqrestore(&rq->lock, flags);
+		rq_unlock_irqrestore(rq, &rf);
 	}
 
 done:

kernel/sched/features.h

@@ -56,6 +56,13 @@ SCHED_FEAT(TTWU_QUEUE, true)
  */
 SCHED_FEAT(SIS_AVG_CPU, false)
 
+/*
+ * Issue a WARN when we do multiple update_rq_clock() calls
+ * in a single rq->lock section. Default disabled because the
+ * annotations are not complete.
+ */
+SCHED_FEAT(WARN_DOUBLE_CLOCK, false)
+
 #ifdef HAVE_RT_PUSH_IPI
 /*
  * In order to avoid a thundering herd attack of CPUs that are

kernel/sched/rt.c

@@ -1927,6 +1927,87 @@ static int find_next_push_cpu(struct rq *rq)
 #define RT_PUSH_IPI_EXECUTING		1
 #define RT_PUSH_IPI_RESTART		2
 
+/*
+ * When a high priority task schedules out from a CPU and a lower priority
+ * task is scheduled in, a check is made to see if there's any RT tasks
+ * on other CPUs that are waiting to run because a higher priority RT task
+ * is currently running on its CPU. In this case, the CPU with multiple RT
+ * tasks queued on it (overloaded) needs to be notified that a CPU has opened
+ * up that may be able to run one of its non-running queued RT tasks.
+ *
+ * On large CPU boxes, there's the case that several CPUs could schedule
+ * a lower priority task at the same time, in which case it will look for
+ * any overloaded CPUs that it could pull a task from. To do this, the runqueue
+ * lock must be taken from that overloaded CPU. Having 10s of CPUs all fighting
+ * for a single overloaded CPU's runqueue lock can produce a large latency.
+ * (This has actually been observed on large boxes running cyclictest).
+ * Instead of taking the runqueue lock of the overloaded CPU, each of the
+ * CPUs that scheduled a lower priority task simply sends an IPI to the
+ * overloaded CPU. An IPI is much cheaper than taking an runqueue lock with
+ * lots of contention. The overloaded CPU will look to push its non-running
+ * RT task off, and if it does, it can then ignore the other IPIs coming
+ * in, and just pass those IPIs off to any other overloaded CPU.
+ *
+ * When a CPU schedules a lower priority task, it only sends an IPI to
+ * the "next" CPU that has overloaded RT tasks. This prevents IPI storms,
+ * as having 10 CPUs scheduling lower priority tasks and 10 CPUs with
+ * RT overloaded tasks, would cause 100 IPIs to go out at once.
+ *
+ * The overloaded RT CPU, when receiving an IPI, will try to push off its
+ * overloaded RT tasks and then send an IPI to the next CPU that has
+ * overloaded RT tasks. This stops when all CPUs with overloaded RT tasks
+ * have completed. Just because a CPU may have pushed off its own overloaded
+ * RT task does not mean it should stop sending the IPI around to other
+ * overloaded CPUs. There may be another RT task waiting to run on one of
+ * those CPUs that are of higher priority than the one that was just
+ * pushed.
+ *
+ * An optimization that could possibly be made is to make a CPU array similar
+ * to the cpupri array mask of all running RT tasks, but for the overloaded
+ * case, then the IPI could be sent to only the CPU with the highest priority
+ * RT task waiting, and that CPU could send off further IPIs to the CPU with
+ * the next highest waiting task. Since the overloaded case is much less likely
+ * to happen, the complexity of this implementation may not be worth it.
+ * Instead, just send an IPI around to all overloaded CPUs.
+ *
+ * The rq->rt.push_flags holds the status of the IPI that is going around.
+ * A run queue can only send out a single IPI at a time. The possible flags
+ * for rq->rt.push_flags are:
+ *
+ *    (None or zero):		No IPI is going around for the current rq
+ *    RT_PUSH_IPI_EXECUTING:	An IPI for the rq is being passed around
+ *    RT_PUSH_IPI_RESTART:	The priority of the running task for the rq
+ *				has changed, and the IPI should restart
+ *				circulating the overloaded CPUs again.
+ *
+ * rq->rt.push_cpu contains the CPU that is being sent the IPI. It is updated
+ * before sending to the next CPU.
+ *
+ * Instead of having all CPUs that schedule a lower priority task send
+ * an IPI to the same "first" CPU in the RT overload mask, they send it
+ * to the next overloaded CPU after their own CPU. This helps distribute
+ * the work when there's more than one overloaded CPU and multiple CPUs
+ * scheduling in lower priority tasks.
+ *
+ * When a rq schedules a lower priority task than what was currently
+ * running, the next CPU with overloaded RT tasks is examined first.
+ * That is, if CPU 1 and 5 are overloaded, and CPU 3 schedules a lower
+ * priority task, it will send an IPI first to CPU 5, then CPU 5 will
+ * send to CPU 1 if it is still overloaded. CPU 1 will clear the
+ * rq->rt.push_flags if RT_PUSH_IPI_RESTART is not set.
+ *
+ * The first CPU to notice IPI_RESTART is set, will clear that flag and then
+ * send an IPI to the next overloaded CPU after the rq->cpu and not the next
+ * CPU after push_cpu. That is, if CPU 1, 4 and 5 are overloaded when CPU 3
+ * schedules a lower priority task, and the IPI_RESTART gets set while the
+ * handling is being done on CPU 5, it will clear the flag and send it back to
+ * CPU 4 instead of CPU 1.
+ *
+ * Note, the above logic can be disabled by turning off the sched_feature
+ * RT_PUSH_IPI. Then the rq lock of the overloaded CPU will simply be
+ * taken by the CPU requesting a pull and the waiting RT task will be pulled
+ * by that CPU. This may be fine for machines with few CPUs.
+ */
 static void tell_cpu_to_push(struct rq *rq)
 {
 	int cpu;

kernel/sched/sched-pelt.h

@@ -0,0 +1,13 @@
+/* Generated by Documentation/scheduler/sched-pelt; do not modify. */
+
+static const u32 runnable_avg_yN_inv[] = {
+	0xffffffff, 0xfa83b2da, 0xf5257d14, 0xefe4b99a, 0xeac0c6e6, 0xe5b906e6,
+	0xe0ccdeeb, 0xdbfbb796, 0xd744fcc9, 0xd2a81d91, 0xce248c14, 0xc9b9bd85,
+	0xc5672a10, 0xc12c4cc9, 0xbd08a39e, 0xb8fbaf46, 0xb504f333, 0xb123f581,
+	0xad583ee9, 0xa9a15ab4, 0xa5fed6a9, 0xa2704302, 0x9ef5325f, 0x9b8d39b9,
+	0x9837f050, 0x94f4efa8, 0x91c3d373, 0x8ea4398a, 0x8b95c1e3, 0x88980e80,
+	0x85aac367, 0x82cd8698,
+};
+
+#define LOAD_AVG_PERIOD 32
+#define LOAD_AVG_MAX 47742

kernel/sched/sched.h

@@ -1331,15 +1331,17 @@ extern const u32 sched_prio_to_wmult[40];
 #define DEQUEUE_SLEEP		0x01
 #define DEQUEUE_SAVE		0x02 /* matches ENQUEUE_RESTORE */
 #define DEQUEUE_MOVE		0x04 /* matches ENQUEUE_MOVE */
+#define DEQUEUE_NOCLOCK		0x08 /* matches ENQUEUE_NOCLOCK */
 
 #define ENQUEUE_WAKEUP		0x01
 #define ENQUEUE_RESTORE		0x02
 #define ENQUEUE_MOVE		0x04
+#define ENQUEUE_NOCLOCK		0x08
 
-#define ENQUEUE_HEAD		0x08
-#define ENQUEUE_REPLENISH	0x10
+#define ENQUEUE_HEAD		0x10
+#define ENQUEUE_REPLENISH	0x20
 #ifdef CONFIG_SMP
-#define ENQUEUE_MIGRATED	0x20
+#define ENQUEUE_MIGRATED	0x40
 #else
 #define ENQUEUE_MIGRATED	0x00
 #endif
@@ -1624,6 +1626,7 @@ static inline void sched_avg_update(struct rq *rq) { }
 
 struct rq *__task_rq_lock(struct task_struct *p, struct rq_flags *rf)
 	__acquires(rq->lock);
+
 struct rq *task_rq_lock(struct task_struct *p, struct rq_flags *rf)
 	__acquires(p->pi_lock)
 	__acquires(rq->lock);
@@ -1645,6 +1648,62 @@ task_rq_unlock(struct rq *rq, struct task_struct *p, struct rq_flags *rf)
 	raw_spin_unlock_irqrestore(&p->pi_lock, rf->flags);
 }
 
+static inline void
+rq_lock_irqsave(struct rq *rq, struct rq_flags *rf)
+	__acquires(rq->lock)
+{
+	raw_spin_lock_irqsave(&rq->lock, rf->flags);
+	rq_pin_lock(rq, rf);
+}
+
+static inline void
+rq_lock_irq(struct rq *rq, struct rq_flags *rf)
+	__acquires(rq->lock)
+{
+	raw_spin_lock_irq(&rq->lock);
+	rq_pin_lock(rq, rf);
+}
+
+static inline void
+rq_lock(struct rq *rq, struct rq_flags *rf)
+	__acquires(rq->lock)
+{
+	raw_spin_lock(&rq->lock);
+	rq_pin_lock(rq, rf);
+}
+
+static inline void
+rq_relock(struct rq *rq, struct rq_flags *rf)
+	__acquires(rq->lock)
+{
+	raw_spin_lock(&rq->lock);
+	rq_repin_lock(rq, rf);
+}
+
+static inline void
+rq_unlock_irqrestore(struct rq *rq, struct rq_flags *rf)
+	__releases(rq->lock)
+{
+	rq_unpin_lock(rq, rf);
+	raw_spin_unlock_irqrestore(&rq->lock, rf->flags);
+}
+
+static inline void
+rq_unlock_irq(struct rq *rq, struct rq_flags *rf)
+	__releases(rq->lock)
+{
+	rq_unpin_lock(rq, rf);
+	raw_spin_unlock_irq(&rq->lock);
+}
+
+static inline void
+rq_unlock(struct rq *rq, struct rq_flags *rf)
+	__releases(rq->lock)
+{
+	rq_unpin_lock(rq, rf);
+	raw_spin_unlock(&rq->lock);
+}
+
 #ifdef CONFIG_SMP
 #ifdef CONFIG_PREEMPT
 

kernel/softirq.c

@@ -309,7 +309,7 @@ restart:
 	account_irq_exit_time(current);
 	__local_bh_enable(SOFTIRQ_OFFSET);
 	WARN_ON_ONCE(in_interrupt());
-	tsk_restore_flags(current, old_flags, PF_MEMALLOC);
+	current_restore_flags(old_flags, PF_MEMALLOC);
 }
 
 asmlinkage __visible void do_softirq(void)

kernel/workqueue.c

@@ -4734,6 +4734,29 @@ long work_on_cpu(int cpu, long (*fn)(void *), void *arg)
 	return wfc.ret;
 }
 EXPORT_SYMBOL_GPL(work_on_cpu);
+
+/**
+ * work_on_cpu_safe - run a function in thread context on a particular cpu
+ * @cpu: the cpu to run on
+ * @fn:  the function to run
+ * @arg: the function argument
+ *
+ * Disables CPU hotplug and calls work_on_cpu(). The caller must not hold
+ * any locks which would prevent @fn from completing.
+ *
+ * Return: The value @fn returns.
+ */
+long work_on_cpu_safe(int cpu, long (*fn)(void *), void *arg)
+{
+	long ret = -ENODEV;
+
+	get_online_cpus();
+	if (cpu_online(cpu))
+		ret = work_on_cpu(cpu, fn, arg);
+	put_online_cpus();
+	return ret;
+}
+EXPORT_SYMBOL_GPL(work_on_cpu_safe);
 #endif /* CONFIG_SMP */
 
 #ifdef CONFIG_FREEZER

net/core/dev.c

@@ -4243,7 +4243,7 @@ static int __netif_receive_skb(struct sk_buff *skb)
 		 */
 		current->flags |= PF_MEMALLOC;
 		ret = __netif_receive_skb_core(skb, true);
-		tsk_restore_flags(current, pflags, PF_MEMALLOC);
+		current_restore_flags(pflags, PF_MEMALLOC);
 	} else
 		ret = __netif_receive_skb_core(skb, false);
 

net/core/sock.c

@@ -325,7 +325,7 @@ int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
 
 	current->flags |= PF_MEMALLOC;
 	ret = sk->sk_backlog_rcv(sk, skb);
-	tsk_restore_flags(current, pflags, PF_MEMALLOC);
+	current_restore_flags(pflags, PF_MEMALLOC);
 
 	return ret;
 }