4 files changed, 612 insertions, 135 deletions

diff --git a/drivers/bus/Kconfig b/drivers/bus/Kconfig
index 603eb1be4f6a..b99729e36860 100644
--- a/drivers/bus/Kconfig
+++ b/drivers/bus/Kconfig
@@ -6,7 +6,7 @@ menu "Bus devices"
 
 config BRCMSTB_GISB_ARB
 	bool "Broadcom STB GISB bus arbiter"
-	depends on ARM
+	depends on ARM || MIPS
 	help
 	  Driver for the Broadcom Set Top Box System-on-a-chip internal bus
 	  arbiter. This driver provides timeout and target abort error handling
diff --git a/drivers/bus/arm-cci.c b/drivers/bus/arm-cci.c
index 7af78df241f2..860da40b78ef 100644
--- a/drivers/bus/arm-cci.c
+++ b/drivers/bus/arm-cci.c
@@ -16,17 +16,17 @@
 
 #include <linux/arm-cci.h>
 #include <linux/io.h>
+#include <linux/interrupt.h>
 #include <linux/module.h>
 #include <linux/of_address.h>
 #include <linux/of_irq.h>
 #include <linux/of_platform.h>
+#include <linux/perf_event.h>
 #include <linux/platform_device.h>
 #include <linux/slab.h>
 #include <linux/spinlock.h>
 
 #include <asm/cacheflush.h>
-#include <asm/irq_regs.h>
-#include <asm/pmu.h>
 #include <asm/smp_plat.h>
 
 #define DRIVER_NAME		"CCI-400"
@@ -98,6 +98,8 @@ static unsigned long cci_ctrl_phys;
 
 #define CCI_PMU_CNTR_BASE(idx)	((idx) * SZ_4K)
 
+#define CCI_PMU_CNTR_MASK	((1ULL << 32) -1)
+
 /*
  * Instead of an event id to monitor CCI cycles, a dedicated counter is
  * provided. Use 0xff to represent CCI cycles and hope that no future revisions
@@ -170,18 +172,29 @@ static char *const pmu_names[] = {
 	[CCI_REV_R1] = "CCI_400_r1",
 };
 
-struct cci_pmu_drv_data {
+struct cci_pmu_hw_events {
+	struct perf_event *events[CCI_PMU_MAX_HW_EVENTS];
+	unsigned long used_mask[BITS_TO_LONGS(CCI_PMU_MAX_HW_EVENTS)];
+	raw_spinlock_t pmu_lock;
+};
+
+struct cci_pmu {
 	void __iomem *base;
-	struct arm_pmu *cci_pmu;
+	struct pmu pmu;
 	int nr_irqs;
 	int irqs[CCI_PMU_MAX_HW_EVENTS];
 	unsigned long active_irqs;
-	struct perf_event *events[CCI_PMU_MAX_HW_EVENTS];
-	unsigned long used_mask[BITS_TO_LONGS(CCI_PMU_MAX_HW_EVENTS)];
 	struct pmu_port_event_ranges *port_ranges;
-	struct pmu_hw_events hw_events;
+	struct cci_pmu_hw_events hw_events;
+	struct platform_device *plat_device;
+	int num_events;
+	atomic_t active_events;
+	struct mutex reserve_mutex;
+	cpumask_t cpus;
 };
-static struct cci_pmu_drv_data *pmu;
+static struct cci_pmu *pmu;
+
+#define to_cci_pmu(c)	(container_of(c, struct cci_pmu, pmu))
 
 static bool is_duplicate_irq(int irq, int *irqs, int nr_irqs)
 {
@@ -252,7 +265,7 @@ static int pmu_validate_hw_event(u8 hw_event)
 	return -ENOENT;
 }
 
-static int pmu_is_valid_counter(struct arm_pmu *cci_pmu, int idx)
+static int pmu_is_valid_counter(struct cci_pmu *cci_pmu, int idx)
 {
 	return CCI_PMU_CYCLE_CNTR_IDX <= idx &&
 		idx <= CCI_PMU_CNTR_LAST(cci_pmu);
@@ -293,14 +306,9 @@ static u32 pmu_get_max_counters(void)
 	return n_cnts + 1;
 }
 
-static struct pmu_hw_events *pmu_get_hw_events(void)
-{
-	return &pmu->hw_events;
-}
-
-static int pmu_get_event_idx(struct pmu_hw_events *hw, struct perf_event *event)
+static int pmu_get_event_idx(struct cci_pmu_hw_events *hw, struct perf_event *event)
 {
-	struct arm_pmu *cci_pmu = to_arm_pmu(event->pmu);
+	struct cci_pmu *cci_pmu = to_cci_pmu(event->pmu);
 	struct hw_perf_event *hw_event = &event->hw;
 	unsigned long cci_event = hw_event->config_base & CCI_PMU_EVENT_MASK;
 	int idx;
@@ -336,7 +344,7 @@ static int pmu_map_event(struct perf_event *event)
 	return mapping;
 }
 
-static int pmu_request_irq(struct arm_pmu *cci_pmu, irq_handler_t handler)
+static int pmu_request_irq(struct cci_pmu *cci_pmu, irq_handler_t handler)
 {
 	int i;
 	struct platform_device *pmu_device = cci_pmu->plat_device;
@@ -371,17 +379,91 @@ static int pmu_request_irq(struct arm_pmu *cci_pmu, irq_handler_t handler)
 	return 0;
 }
 
+static void pmu_free_irq(struct cci_pmu *cci_pmu)
+{
+	int i;
+
+	for (i = 0; i < pmu->nr_irqs; i++) {
+		if (!test_and_clear_bit(i, &pmu->active_irqs))
+			continue;
+
+		free_irq(pmu->irqs[i], cci_pmu);
+	}
+}
+
+static u32 pmu_read_counter(struct perf_event *event)
+{
+	struct cci_pmu *cci_pmu = to_cci_pmu(event->pmu);
+	struct hw_perf_event *hw_counter = &event->hw;
+	int idx = hw_counter->idx;
+	u32 value;
+
+	if (unlikely(!pmu_is_valid_counter(cci_pmu, idx))) {
+		dev_err(&cci_pmu->plat_device->dev, "Invalid CCI PMU counter %d\n", idx);
+		return 0;
+	}
+	value = pmu_read_register(idx, CCI_PMU_CNTR);
+
+	return value;
+}
+
+static void pmu_write_counter(struct perf_event *event, u32 value)
+{
+	struct cci_pmu *cci_pmu = to_cci_pmu(event->pmu);
+	struct hw_perf_event *hw_counter = &event->hw;
+	int idx = hw_counter->idx;
+
+	if (unlikely(!pmu_is_valid_counter(cci_pmu, idx)))
+		dev_err(&cci_pmu->plat_device->dev, "Invalid CCI PMU counter %d\n", idx);
+	else
+		pmu_write_register(value, idx, CCI_PMU_CNTR);
+}
+
+static u64 pmu_event_update(struct perf_event *event)
+{
+	struct hw_perf_event *hwc = &event->hw;
+	u64 delta, prev_raw_count, new_raw_count;
+
+	do {
+		prev_raw_count = local64_read(&hwc->prev_count);
+		new_raw_count = pmu_read_counter(event);
+	} while (local64_cmpxchg(&hwc->prev_count, prev_raw_count,
+		 new_raw_count) != prev_raw_count);
+
+	delta = (new_raw_count - prev_raw_count) & CCI_PMU_CNTR_MASK;
+
+	local64_add(delta, &event->count);
+
+	return new_raw_count;
+}
+
+static void pmu_read(struct perf_event *event)
+{
+	pmu_event_update(event);
+}
+
+void pmu_event_set_period(struct perf_event *event)
+{
+	struct hw_perf_event *hwc = &event->hw;
+	/*
+	 * The CCI PMU counters have a period of 2^32. To account for the
+	 * possiblity of extreme interrupt latency we program for a period of
+	 * half that. Hopefully we can handle the interrupt before another 2^31
+	 * events occur and the counter overtakes its previous value.
+	 */
+	u64 val = 1ULL << 31;
+	local64_set(&hwc->prev_count, val);
+	pmu_write_counter(event, val);
+}
+
 static irqreturn_t pmu_handle_irq(int irq_num, void *dev)
 {
 	unsigned long flags;
-	struct arm_pmu *cci_pmu = (struct arm_pmu *)dev;
-	struct pmu_hw_events *events = cci_pmu->get_hw_events();
-	struct perf_sample_data data;
-	struct pt_regs *regs;
+	struct cci_pmu *cci_pmu = dev;
+	struct cci_pmu_hw_events *events = &pmu->hw_events;
 	int idx, handled = IRQ_NONE;
 
 	raw_spin_lock_irqsave(&events->pmu_lock, flags);
-	regs = get_irq_regs();
 	/*
 	 * Iterate over counters and update the corresponding perf events.
 	 * This should work regardless of whether we have per-counter overflow
@@ -403,154 +485,407 @@ static irqreturn_t pmu_handle_irq(int irq_num, void *dev)
 
 		pmu_write_register(CCI_PMU_OVRFLW_FLAG, idx, CCI_PMU_OVRFLW);
 
+		pmu_event_update(event);
+		pmu_event_set_period(event);
 		handled = IRQ_HANDLED;
-
-		armpmu_event_update(event);
-		perf_sample_data_init(&data, 0, hw_counter->last_period);
-		if (!armpmu_event_set_period(event))
-			continue;
-
-		if (perf_event_overflow(event, &data, regs))
-			cci_pmu->disable(event);
 	}
 	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
 
 	return IRQ_RETVAL(handled);
 }
 
-static void pmu_free_irq(struct arm_pmu *cci_pmu)
+static int cci_pmu_get_hw(struct cci_pmu *cci_pmu)
 {
-	int i;
+	int ret = pmu_request_irq(cci_pmu, pmu_handle_irq);
+	if (ret) {
+		pmu_free_irq(cci_pmu);
+		return ret;
+	}
+	return 0;
+}
 
-	for (i = 0; i < pmu->nr_irqs; i++) {
-		if (!test_and_clear_bit(i, &pmu->active_irqs))
-			continue;
+static void cci_pmu_put_hw(struct cci_pmu *cci_pmu)
+{
+	pmu_free_irq(cci_pmu);
+}
 
-		free_irq(pmu->irqs[i], cci_pmu);
+static void hw_perf_event_destroy(struct perf_event *event)
+{
+	struct cci_pmu *cci_pmu = to_cci_pmu(event->pmu);
+	atomic_t *active_events = &cci_pmu->active_events;
+	struct mutex *reserve_mutex = &cci_pmu->reserve_mutex;
+
+	if (atomic_dec_and_mutex_lock(active_events, reserve_mutex)) {
+		cci_pmu_put_hw(cci_pmu);
+		mutex_unlock(reserve_mutex);
 	}
 }
 
-static void pmu_enable_event(struct perf_event *event)
+static void cci_pmu_enable(struct pmu *pmu)
 {
+	struct cci_pmu *cci_pmu = to_cci_pmu(pmu);
+	struct cci_pmu_hw_events *hw_events = &cci_pmu->hw_events;
+	int enabled = bitmap_weight(hw_events->used_mask, cci_pmu->num_events);
 	unsigned long flags;
-	struct arm_pmu *cci_pmu = to_arm_pmu(event->pmu);
-	struct pmu_hw_events *events = cci_pmu->get_hw_events();
-	struct hw_perf_event *hw_counter = &event->hw;
-	int idx = hw_counter->idx;
+	u32 val;
+
+	if (!enabled)
+		return;
+
+	raw_spin_lock_irqsave(&hw_events->pmu_lock, flags);
+
+	/* Enable all the PMU counters. */
+	val = readl_relaxed(cci_ctrl_base + CCI_PMCR) | CCI_PMCR_CEN;
+	writel(val, cci_ctrl_base + CCI_PMCR);
+	raw_spin_unlock_irqrestore(&hw_events->pmu_lock, flags);
+
+}
+
+static void cci_pmu_disable(struct pmu *pmu)
+{
+	struct cci_pmu *cci_pmu = to_cci_pmu(pmu);
+	struct cci_pmu_hw_events *hw_events = &cci_pmu->hw_events;
+	unsigned long flags;
+	u32 val;
+
+	raw_spin_lock_irqsave(&hw_events->pmu_lock, flags);
+
+	/* Disable all the PMU counters. */
+	val = readl_relaxed(cci_ctrl_base + CCI_PMCR) & ~CCI_PMCR_CEN;
+	writel(val, cci_ctrl_base + CCI_PMCR);
+	raw_spin_unlock_irqrestore(&hw_events->pmu_lock, flags);
+}
+
+static void cci_pmu_start(struct perf_event *event, int pmu_flags)
+{
+	struct cci_pmu *cci_pmu = to_cci_pmu(event->pmu);
+	struct cci_pmu_hw_events *hw_events = &cci_pmu->hw_events;
+	struct hw_perf_event *hwc = &event->hw;
+	int idx = hwc->idx;
+	unsigned long flags;
+
+	/*
+	 * To handle interrupt latency, we always reprogram the period
+	 * regardlesss of PERF_EF_RELOAD.
+	 */
+	if (pmu_flags & PERF_EF_RELOAD)
+		WARN_ON_ONCE(!(hwc->state & PERF_HES_UPTODATE));
+
+	hwc->state = 0;
 
 	if (unlikely(!pmu_is_valid_counter(cci_pmu, idx))) {
 		dev_err(&cci_pmu->plat_device->dev, "Invalid CCI PMU counter %d\n", idx);
 		return;
 	}
 
-	raw_spin_lock_irqsave(&events->pmu_lock, flags);
+	raw_spin_lock_irqsave(&hw_events->pmu_lock, flags);
 
 	/* Configure the event to count, unless you are counting cycles */
 	if (idx != CCI_PMU_CYCLE_CNTR_IDX)
-		pmu_set_event(idx, hw_counter->config_base);
+		pmu_set_event(idx, hwc->config_base);
 
+	pmu_event_set_period(event);
 	pmu_enable_counter(idx);
 
-	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
+	raw_spin_unlock_irqrestore(&hw_events->pmu_lock, flags);
 }
 
-static void pmu_disable_event(struct perf_event *event)
+static void cci_pmu_stop(struct perf_event *event, int pmu_flags)
 {
-	struct arm_pmu *cci_pmu = to_arm_pmu(event->pmu);
-	struct hw_perf_event *hw_counter = &event->hw;
-	int idx = hw_counter->idx;
+	struct cci_pmu *cci_pmu = to_cci_pmu(event->pmu);
+	struct hw_perf_event *hwc = &event->hw;
+	int idx = hwc->idx;
+
+	if (hwc->state & PERF_HES_STOPPED)
+		return;
 
 	if (unlikely(!pmu_is_valid_counter(cci_pmu, idx))) {
 		dev_err(&cci_pmu->plat_device->dev, "Invalid CCI PMU counter %d\n", idx);
 		return;
 	}
 
+	/*
+	 * We always reprogram the counter, so ignore PERF_EF_UPDATE. See
+	 * cci_pmu_start()
+	 */
 	pmu_disable_counter(idx);
+	pmu_event_update(event);
+	hwc->state |= PERF_HES_STOPPED | PERF_HES_UPTODATE;
 }
 
-static void pmu_start(struct arm_pmu *cci_pmu)
+static int cci_pmu_add(struct perf_event *event, int flags)
 {
-	u32 val;
-	unsigned long flags;
-	struct pmu_hw_events *events = cci_pmu->get_hw_events();
+	struct cci_pmu *cci_pmu = to_cci_pmu(event->pmu);
+	struct cci_pmu_hw_events *hw_events = &cci_pmu->hw_events;
+	struct hw_perf_event *hwc = &event->hw;
+	int idx;
+	int err = 0;
 
-	raw_spin_lock_irqsave(&events->pmu_lock, flags);
+	perf_pmu_disable(event->pmu);
 
-	/* Enable all the PMU counters. */
-	val = readl_relaxed(cci_ctrl_base + CCI_PMCR) | CCI_PMCR_CEN;
-	writel(val, cci_ctrl_base + CCI_PMCR);
+	/* If we don't have a space for the counter then finish early. */
+	idx = pmu_get_event_idx(hw_events, event);
+	if (idx < 0) {
+		err = idx;
+		goto out;
+	}
 
-	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
+	event->hw.idx = idx;
+	hw_events->events[idx] = event;
+
+	hwc->state = PERF_HES_STOPPED | PERF_HES_UPTODATE;
+	if (flags & PERF_EF_START)
+		cci_pmu_start(event, PERF_EF_RELOAD);
+
+	/* Propagate our changes to the userspace mapping. */
+	perf_event_update_userpage(event);
+
+out:
+	perf_pmu_enable(event->pmu);
+	return err;
 }
 
-static void pmu_stop(struct arm_pmu *cci_pmu)
+static void cci_pmu_del(struct perf_event *event, int flags)
 {
-	u32 val;
-	unsigned long flags;
-	struct pmu_hw_events *events = cci_pmu->get_hw_events();
+	struct cci_pmu *cci_pmu = to_cci_pmu(event->pmu);
+	struct cci_pmu_hw_events *hw_events = &cci_pmu->hw_events;
+	struct hw_perf_event *hwc = &event->hw;
+	int idx = hwc->idx;
 
-	raw_spin_lock_irqsave(&events->pmu_lock, flags);
+	cci_pmu_stop(event, PERF_EF_UPDATE);
+	hw_events->events[idx] = NULL;
+	clear_bit(idx, hw_events->used_mask);
 
-	/* Disable all the PMU counters. */
-	val = readl_relaxed(cci_ctrl_base + CCI_PMCR) & ~CCI_PMCR_CEN;
-	writel(val, cci_ctrl_base + CCI_PMCR);
+	perf_event_update_userpage(event);
+}
 
-	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
+static int
+validate_event(struct cci_pmu_hw_events *hw_events,
+	       struct perf_event *event)
+{
+	if (is_software_event(event))
+		return 1;
+
+	if (event->state < PERF_EVENT_STATE_OFF)
+		return 1;
+
+	if (event->state == PERF_EVENT_STATE_OFF && !event->attr.enable_on_exec)
+		return 1;
+
+	return pmu_get_event_idx(hw_events, event) >= 0;
 }
 
-static u32 pmu_read_counter(struct perf_event *event)
+static int
+validate_group(struct perf_event *event)
 {
-	struct arm_pmu *cci_pmu = to_arm_pmu(event->pmu);
-	struct hw_perf_event *hw_counter = &event->hw;
-	int idx = hw_counter->idx;
-	u32 value;
+	struct perf_event *sibling, *leader = event->group_leader;
+	struct cci_pmu_hw_events fake_pmu = {
+		/*
+		 * Initialise the fake PMU. We only need to populate the
+		 * used_mask for the purposes of validation.
+		 */
+		.used_mask = CPU_BITS_NONE,
+	};
 
-	if (unlikely(!pmu_is_valid_counter(cci_pmu, idx))) {
-		dev_err(&cci_pmu->plat_device->dev, "Invalid CCI PMU counter %d\n", idx);
-		return 0;
+	if (!validate_event(&fake_pmu, leader))
+		return -EINVAL;
+
+	list_for_each_entry(sibling, &leader->sibling_list, group_entry) {
+		if (!validate_event(&fake_pmu, sibling))
+			return -EINVAL;
 	}
-	value = pmu_read_register(idx, CCI_PMU_CNTR);
 
-	return value;
+	if (!validate_event(&fake_pmu, event))
+		return -EINVAL;
+
+	return 0;
 }
 
-static void pmu_write_counter(struct perf_event *event, u32 value)
+static int
+__hw_perf_event_init(struct perf_event *event)
 {
-	struct arm_pmu *cci_pmu = to_arm_pmu(event->pmu);
-	struct hw_perf_event *hw_counter = &event->hw;
-	int idx = hw_counter->idx;
+	struct hw_perf_event *hwc = &event->hw;
+	int mapping;
 
-	if (unlikely(!pmu_is_valid_counter(cci_pmu, idx)))
-		dev_err(&cci_pmu->plat_device->dev, "Invalid CCI PMU counter %d\n", idx);
-	else
-		pmu_write_register(value, idx, CCI_PMU_CNTR);
+	mapping = pmu_map_event(event);
+
+	if (mapping < 0) {
+		pr_debug("event %x:%llx not supported\n", event->attr.type,
+			 event->attr.config);
+		return mapping;
+	}
+
+	/*
+	 * We don't assign an index until we actually place the event onto
+	 * hardware. Use -1 to signify that we haven't decided where to put it
+	 * yet.
+	 */
+	hwc->idx		= -1;
+	hwc->config_base	= 0;
+	hwc->config		= 0;
+	hwc->event_base		= 0;
+
+	/*
+	 * Store the event encoding into the config_base field.
+	 */
+	hwc->config_base	    |= (unsigned long)mapping;
+
+	/*
+	 * Limit the sample_period to half of the counter width. That way, the
+	 * new counter value is far less likely to overtake the previous one
+	 * unless you have some serious IRQ latency issues.
+	 */
+	hwc->sample_period  = CCI_PMU_CNTR_MASK >> 1;
+	hwc->last_period    = hwc->sample_period;
+	local64_set(&hwc->period_left, hwc->sample_period);
+
+	if (event->group_leader != event) {
+		if (validate_group(event) != 0)
+			return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int cci_pmu_event_init(struct perf_event *event)
+{
+	struct cci_pmu *cci_pmu = to_cci_pmu(event->pmu);
+	atomic_t *active_events = &cci_pmu->active_events;
+	int err = 0;
+	int cpu;
+
+	if (event->attr.type != event->pmu->type)
+		return -ENOENT;
+
+	/* Shared by all CPUs, no meaningful state to sample */
+	if (is_sampling_event(event) || event->attach_state & PERF_ATTACH_TASK)
+		return -EOPNOTSUPP;
+
+	/* We have no filtering of any kind */
+	if (event->attr.exclude_user	||
+	    event->attr.exclude_kernel	||
+	    event->attr.exclude_hv	||
+	    event->attr.exclude_idle	||
+	    event->attr.exclude_host	||
+	    event->attr.exclude_guest)
+		return -EINVAL;
+
+	/*
+	 * Following the example set by other "uncore" PMUs, we accept any CPU
+	 * and rewrite its affinity dynamically rather than having perf core
+	 * handle cpu == -1 and pid == -1 for this case.
+	 *
+	 * The perf core will pin online CPUs for the duration of this call and
+	 * the event being installed into its context, so the PMU's CPU can't
+	 * change under our feet.
+	 */
+	cpu = cpumask_first(&cci_pmu->cpus);
+	if (event->cpu < 0 || cpu < 0)
+		return -EINVAL;
+	event->cpu = cpu;
+
+	event->destroy = hw_perf_event_destroy;
+	if (!atomic_inc_not_zero(active_events)) {
+		mutex_lock(&cci_pmu->reserve_mutex);
+		if (atomic_read(active_events) == 0)
+			err = cci_pmu_get_hw(cci_pmu);
+		if (!err)
+			atomic_inc(active_events);
+		mutex_unlock(&cci_pmu->reserve_mutex);
+	}
+	if (err)
+		return err;
+
+	err = __hw_perf_event_init(event);
+	if (err)
+		hw_perf_event_destroy(event);
+
+	return err;
 }
 
-static int cci_pmu_init(struct arm_pmu *cci_pmu, struct platform_device *pdev)
+static ssize_t pmu_attr_cpumask_show(struct device *dev,
+				     struct device_attribute *attr, char *buf)
 {
-	*cci_pmu = (struct arm_pmu){
-		.name		  = pmu_names[probe_cci_revision()],
-		.max_period       = (1LLU << 32) - 1,
-		.get_hw_events    = pmu_get_hw_events,
-		.get_event_idx    = pmu_get_event_idx,
-		.map_event        = pmu_map_event,
-		.request_irq      = pmu_request_irq,
-		.handle_irq       = pmu_handle_irq,
-		.free_irq         = pmu_free_irq,
-		.enable           = pmu_enable_event,
-		.disable          = pmu_disable_event,
-		.start            = pmu_start,
-		.stop             = pmu_stop,
-		.read_counter     = pmu_read_counter,
-		.write_counter    = pmu_write_counter,
+	int n = cpulist_scnprintf(buf, PAGE_SIZE - 2, &pmu->cpus);
+
+	buf[n++] = '\n';
+	buf[n] = '\0';
+	return n;
+}
+
+static DEVICE_ATTR(cpumask, S_IRUGO, pmu_attr_cpumask_show, NULL);
+
+static struct attribute *pmu_attrs[] = {
+	&dev_attr_cpumask.attr,
+	NULL,
+};
+
+static struct attribute_group pmu_attr_group = {
+	.attrs = pmu_attrs,
+};
+
+static const struct attribute_group *pmu_attr_groups[] = {
+	&pmu_attr_group,
+	NULL
+};
+
+static int cci_pmu_init(struct cci_pmu *cci_pmu, struct platform_device *pdev)
+{
+	char *name = pmu_names[probe_cci_revision()];
+	cci_pmu->pmu = (struct pmu) {
+		.name		= pmu_names[probe_cci_revision()],
+		.task_ctx_nr	= perf_invalid_context,
+		.pmu_enable	= cci_pmu_enable,
+		.pmu_disable	= cci_pmu_disable,
+		.event_init	= cci_pmu_event_init,
+		.add		= cci_pmu_add,
+		.del		= cci_pmu_del,
+		.start		= cci_pmu_start,
+		.stop		= cci_pmu_stop,
+		.read		= pmu_read,
+		.attr_groups	= pmu_attr_groups,
 	};
 
 	cci_pmu->plat_device = pdev;
 	cci_pmu->num_events = pmu_get_max_counters();
 
-	return armpmu_register(cci_pmu, -1);
+	return perf_pmu_register(&cci_pmu->pmu, name, -1);
 }
 
+static int cci_pmu_cpu_notifier(struct notifier_block *self,
+				unsigned long action, void *hcpu)
+{
+	unsigned int cpu = (long)hcpu;
+	unsigned int target;
+
+	switch (action & ~CPU_TASKS_FROZEN) {
+	case CPU_DOWN_PREPARE:
+		if (!cpumask_test_and_clear_cpu(cpu, &pmu->cpus))
+			break;
+		target = cpumask_any_but(cpu_online_mask, cpu);
+		if (target < 0) // UP, last CPU
+			break;
+		/*
+		 * TODO: migrate context once core races on event->ctx have
+		 * been fixed.
+		 */
+		cpumask_set_cpu(target, &pmu->cpus);
+	default:
+		break;
+	}
+
+	return NOTIFY_OK;
+}
+
+static struct notifier_block cci_pmu_cpu_nb = {
+	.notifier_call	= cci_pmu_cpu_notifier,
+	/*
+	 * to migrate uncore events, our notifier should be executed
+	 * before perf core's notifier.
+	 */
+	.priority	= CPU_PRI_PERF + 1,
+};
+
 static const struct of_device_id arm_cci_pmu_matches[] = {
 	{
 		.compatible = "arm,cci-400-pmu",
@@ -604,15 +939,16 @@ static int cci_pmu_probe(struct platform_device *pdev)
 		return -EINVAL;
 	}
 
-	pmu->cci_pmu = devm_kzalloc(&pdev->dev, sizeof(*(pmu->cci_pmu)), GFP_KERNEL);
-	if (!pmu->cci_pmu)
-		return -ENOMEM;
-
-	pmu->hw_events.events = pmu->events;
-	pmu->hw_events.used_mask = pmu->used_mask;
 	raw_spin_lock_init(&pmu->hw_events.pmu_lock);
+	mutex_init(&pmu->reserve_mutex);
+	atomic_set(&pmu->active_events, 0);
+	cpumask_set_cpu(smp_processor_id(), &pmu->cpus);
+
+	ret = register_cpu_notifier(&cci_pmu_cpu_nb);
+	if (ret)
+		return ret;
 
-	ret = cci_pmu_init(pmu->cci_pmu, pdev);
+	ret = cci_pmu_init(pmu, pdev);
 	if (ret)
 		return ret;
 
diff --git a/drivers/bus/brcmstb_gisb.c b/drivers/bus/brcmstb_gisb.c
index e7ccd21a45c9..46de8dc39eb4 100644
--- a/drivers/bus/brcmstb_gisb.c
+++ b/drivers/bus/brcmstb_gisb.c
@@ -25,26 +25,72 @@
 #include <linux/bitops.h>
 #include <linux/pm.h>
 
+#ifdef CONFIG_ARM
 #include <asm/bug.h>
 #include <asm/signal.h>
+#endif
 
-#define ARB_TIMER			0x008
-#define ARB_ERR_CAP_CLR			0x7e4
 #define  ARB_ERR_CAP_CLEAR		(1 << 0)
-#define ARB_ERR_CAP_HI_ADDR		0x7e8
-#define ARB_ERR_CAP_ADDR		0x7ec
-#define ARB_ERR_CAP_DATA		0x7f0
-#define ARB_ERR_CAP_STATUS		0x7f4
 #define  ARB_ERR_CAP_STATUS_TIMEOUT	(1 << 12)
 #define  ARB_ERR_CAP_STATUS_TEA		(1 << 11)
 #define  ARB_ERR_CAP_STATUS_BS_SHIFT	(1 << 2)
 #define  ARB_ERR_CAP_STATUS_BS_MASK	0x3c
 #define  ARB_ERR_CAP_STATUS_WRITE	(1 << 1)
 #define  ARB_ERR_CAP_STATUS_VALID	(1 << 0)
-#define ARB_ERR_CAP_MASTER		0x7f8
+
+enum {
+	ARB_TIMER,
+	ARB_ERR_CAP_CLR,
+	ARB_ERR_CAP_HI_ADDR,
+	ARB_ERR_CAP_ADDR,
+	ARB_ERR_CAP_DATA,
+	ARB_ERR_CAP_STATUS,
+	ARB_ERR_CAP_MASTER,
+};
+
+static const int gisb_offsets_bcm7038[] = {
+	[ARB_TIMER]		= 0x00c,
+	[ARB_ERR_CAP_CLR]	= 0x0c4,
+	[ARB_ERR_CAP_HI_ADDR]	= -1,
+	[ARB_ERR_CAP_ADDR]	= 0x0c8,
+	[ARB_ERR_CAP_DATA]	= 0x0cc,
+	[ARB_ERR_CAP_STATUS]	= 0x0d0,
+	[ARB_ERR_CAP_MASTER]	= -1,
+};
+
+static const int gisb_offsets_bcm7400[] = {
+	[ARB_TIMER]		= 0x00c,
+	[ARB_ERR_CAP_CLR]	= 0x0c8,
+	[ARB_ERR_CAP_HI_ADDR]	= -1,
+	[ARB_ERR_CAP_ADDR]	= 0x0cc,
+	[ARB_ERR_CAP_DATA]	= 0x0d0,
+	[ARB_ERR_CAP_STATUS]	= 0x0d4,
+	[ARB_ERR_CAP_MASTER]	= 0x0d8,
+};
+
+static const int gisb_offsets_bcm7435[] = {
+	[ARB_TIMER]		= 0x00c,
+	[ARB_ERR_CAP_CLR]	= 0x168,
+	[ARB_ERR_CAP_HI_ADDR]	= -1,
+	[ARB_ERR_CAP_ADDR]	= 0x16c,
+	[ARB_ERR_CAP_DATA]	= 0x170,
+	[ARB_ERR_CAP_STATUS]	= 0x174,
+	[ARB_ERR_CAP_MASTER]	= 0x178,
+};
+
+static const int gisb_offsets_bcm7445[] = {
+	[ARB_TIMER]		= 0x008,
+	[ARB_ERR_CAP_CLR]	= 0x7e4,
+	[ARB_ERR_CAP_HI_ADDR]	= 0x7e8,
+	[ARB_ERR_CAP_ADDR]	= 0x7ec,
+	[ARB_ERR_CAP_DATA]	= 0x7f0,
+	[ARB_ERR_CAP_STATUS]	= 0x7f4,
+	[ARB_ERR_CAP_MASTER]	= 0x7f8,
+};
 
 struct brcmstb_gisb_arb_device {
 	void __iomem	*base;
+	const int	*gisb_offsets;
 	struct mutex	lock;
 	struct list_head next;
 	u32 valid_mask;
@@ -54,6 +100,26 @@ struct brcmstb_gisb_arb_device {
 
 static LIST_HEAD(brcmstb_gisb_arb_device_list);
 
+static u32 gisb_read(struct brcmstb_gisb_arb_device *gdev, int reg)
+{
+	int offset = gdev->gisb_offsets[reg];
+
+	/* return 1 if the hardware doesn't have ARB_ERR_CAP_MASTER */
+	if (offset == -1)
+		return 1;
+
+	return ioread32(gdev->base + offset);
+}
+
+static void gisb_write(struct brcmstb_gisb_arb_device *gdev, u32 val, int reg)
+{
+	int offset = gdev->gisb_offsets[reg];
+
+	if (offset == -1)
+		return;
+	iowrite32(val, gdev->base + reg);
+}
+
 static ssize_t gisb_arb_get_timeout(struct device *dev,
 				    struct device_attribute *attr,
 				    char *buf)
@@ -63,7 +129,7 @@ static ssize_t gisb_arb_get_timeout(struct device *dev,
 	u32 timeout;
 
 	mutex_lock(&gdev->lock);
-	timeout = ioread32(gdev->base + ARB_TIMER);
+	timeout = gisb_read(gdev, ARB_TIMER);
 	mutex_unlock(&gdev->lock);
 
 	return sprintf(buf, "%d", timeout);
@@ -85,7 +151,7 @@ static ssize_t gisb_arb_set_timeout(struct device *dev,
 		return -EINVAL;
 
 	mutex_lock(&gdev->lock);
-	iowrite32(val, gdev->base + ARB_TIMER);
+	gisb_write(gdev, val, ARB_TIMER);
 	mutex_unlock(&gdev->lock);
 
 	return count;
@@ -112,18 +178,18 @@ static int brcmstb_gisb_arb_decode_addr(struct brcmstb_gisb_arb_device *gdev,
 	const char *m_name;
 	char m_fmt[11];
 
-	cap_status = ioread32(gdev->base + ARB_ERR_CAP_STATUS);
+	cap_status = gisb_read(gdev, ARB_ERR_CAP_STATUS);
 
 	/* Invalid captured address, bail out */
 	if (!(cap_status & ARB_ERR_CAP_STATUS_VALID))
 		return 1;
 
 	/* Read the address and master */
-	arb_addr = ioread32(gdev->base + ARB_ERR_CAP_ADDR) & 0xffffffff;
+	arb_addr = gisb_read(gdev, ARB_ERR_CAP_ADDR) & 0xffffffff;
 #if (IS_ENABLED(CONFIG_PHYS_ADDR_T_64BIT))
-	arb_addr |= (u64)ioread32(gdev->base + ARB_ERR_CAP_HI_ADDR) << 32;
+	arb_addr |= (u64)gisb_read(gdev, ARB_ERR_CAP_HI_ADDR) << 32;
 #endif
-	master = ioread32(gdev->base + ARB_ERR_CAP_MASTER);
+	master = gisb_read(gdev, ARB_ERR_CAP_MASTER);
 
 	m_name = brcmstb_gisb_master_to_str(gdev, master);
 	if (!m_name) {
@@ -138,11 +204,12 @@ static int brcmstb_gisb_arb_decode_addr(struct brcmstb_gisb_arb_device *gdev,
 		m_name);
 
 	/* clear the GISB error */
-	iowrite32(ARB_ERR_CAP_CLEAR, gdev->base + ARB_ERR_CAP_CLR);
+	gisb_write(gdev, ARB_ERR_CAP_CLEAR, ARB_ERR_CAP_CLR);
 
 	return 0;
 }
 
+#ifdef CONFIG_ARM
 static int brcmstb_bus_error_handler(unsigned long addr, unsigned int fsr,
 				     struct pt_regs *regs)
 {
@@ -161,6 +228,7 @@ static int brcmstb_bus_error_handler(unsigned long addr, unsigned int fsr,
 
 	return ret;
 }
+#endif
 
 static irqreturn_t brcmstb_gisb_timeout_handler(int irq, void *dev_id)
 {
@@ -188,10 +256,20 @@ static struct attribute_group gisb_arb_sysfs_attr_group = {
 	.attrs = gisb_arb_sysfs_attrs,
 };
 
-static int brcmstb_gisb_arb_probe(struct platform_device *pdev)
+static const struct of_device_id brcmstb_gisb_arb_of_match[] = {
+	{ .compatible = "brcm,gisb-arb",         .data = gisb_offsets_bcm7445 },
+	{ .compatible = "brcm,bcm7445-gisb-arb", .data = gisb_offsets_bcm7445 },
+	{ .compatible = "brcm,bcm7435-gisb-arb", .data = gisb_offsets_bcm7435 },
+	{ .compatible = "brcm,bcm7400-gisb-arb", .data = gisb_offsets_bcm7400 },
+	{ .compatible = "brcm,bcm7038-gisb-arb", .data = gisb_offsets_bcm7038 },
+	{ },
+};
+
+static int __init brcmstb_gisb_arb_probe(struct platform_device *pdev)
 {
 	struct device_node *dn = pdev->dev.of_node;
 	struct brcmstb_gisb_arb_device *gdev;
+	const struct of_device_id *of_id;
 	struct resource *r;
 	int err, timeout_irq, tea_irq;
 	unsigned int num_masters, j = 0;
@@ -212,6 +290,13 @@ static int brcmstb_gisb_arb_probe(struct platform_device *pdev)
 	if (IS_ERR(gdev->base))
 		return PTR_ERR(gdev->base);
 
+	of_id = of_match_node(brcmstb_gisb_arb_of_match, dn);
+	if (!of_id) {
+		pr_err("failed to look up compatible string\n");
+		return -EINVAL;
+	}
+	gdev->gisb_offsets = of_id->data;
+
 	err = devm_request_irq(&pdev->dev, timeout_irq,
 				brcmstb_gisb_timeout_handler, 0, pdev->name,
 				gdev);
@@ -257,8 +342,10 @@ static int brcmstb_gisb_arb_probe(struct platform_device *pdev)
 
 	list_add_tail(&gdev->next, &brcmstb_gisb_arb_device_list);
 
+#ifdef CONFIG_ARM
 	hook_fault_code(22, brcmstb_bus_error_handler, SIGBUS, 0,
 			"imprecise external abort");
+#endif
 
 	dev_info(&pdev->dev, "registered mem: %p, irqs: %d, %d\n",
 			gdev->base, timeout_irq, tea_irq);
@@ -272,7 +359,7 @@ static int brcmstb_gisb_arb_suspend(struct device *dev)
 	struct platform_device *pdev = to_platform_device(dev);
 	struct brcmstb_gisb_arb_device *gdev = platform_get_drvdata(pdev);
 
-	gdev->saved_timeout = ioread32(gdev->base + ARB_TIMER);
+	gdev->saved_timeout = gisb_read(gdev, ARB_TIMER);
 
 	return 0;
 }
@@ -285,7 +372,7 @@ static int brcmstb_gisb_arb_resume_noirq(struct device *dev)
 	struct platform_device *pdev = to_platform_device(dev);
 	struct brcmstb_gisb_arb_device *gdev = platform_get_drvdata(pdev);
 
-	iowrite32(gdev->saved_timeout, gdev->base + ARB_TIMER);
+	gisb_write(gdev, gdev->saved_timeout, ARB_TIMER);
 
 	return 0;
 }
@@ -299,13 +386,7 @@ static const struct dev_pm_ops brcmstb_gisb_arb_pm_ops = {
 	.resume_noirq	= brcmstb_gisb_arb_resume_noirq,
 };
 
-static const struct of_device_id brcmstb_gisb_arb_of_match[] = {
-	{ .compatible = "brcm,gisb-arb" },
-	{ },
-};
-
 static struct platform_driver brcmstb_gisb_arb_driver = {
-	.probe	= brcmstb_gisb_arb_probe,
 	.driver = {
 		.name	= "brcm-gisb-arb",
 		.owner	= THIS_MODULE,
@@ -316,7 +397,8 @@ static struct platform_driver brcmstb_gisb_arb_driver = {
 
 static int __init brcm_gisb_driver_init(void)
 {
-	return platform_driver_register(&brcmstb_gisb_arb_driver);
+	return platform_driver_probe(&brcmstb_gisb_arb_driver,
+				     brcmstb_gisb_arb_probe);
 }
 
 module_init(brcm_gisb_driver_init);
diff --git a/drivers/bus/omap_l3_noc.c b/drivers/bus/omap_l3_noc.c
index 531ae591783b..17d86595951c 100644
--- a/drivers/bus/omap_l3_noc.c
+++ b/drivers/bus/omap_l3_noc.c
@@ -222,10 +222,14 @@ static irqreturn_t l3_interrupt_handler(int irq, void *_l3)
 			}
 
 			/* Error found so break the for loop */
-			break;
+			return IRQ_HANDLED;
 		}
 	}
-	return IRQ_HANDLED;
+
+	dev_err(l3->dev, "L3 %s IRQ not handled!!\n",
+		inttype ? "debug" : "application");
+
+	return IRQ_NONE;
 }
 
 static const struct of_device_id l3_noc_match[] = {
@@ -296,11 +300,66 @@ static int omap_l3_probe(struct platform_device *pdev)
 	return ret;
 }
 
+#ifdef	CONFIG_PM
+
+/**
+ * l3_resume_noirq() - resume function for l3_noc
+ * @dev:	pointer to l3_noc device structure
+ *
+ * We only have the resume handler only since we
+ * have already maintained the delta register
+ * configuration as part of configuring the system
+ */
+static int l3_resume_noirq(struct device *dev)
+{
+	struct omap_l3 *l3 = dev_get_drvdata(dev);
+	int i;
+	struct l3_flagmux_data *flag_mux;
+	void __iomem *base, *mask_regx = NULL;
+	u32 mask_val;
+
+	for (i = 0; i < l3->num_modules; i++) {
+		base = l3->l3_base[i];
+		flag_mux = l3->l3_flagmux[i];
+		if (!flag_mux->mask_app_bits && !flag_mux->mask_dbg_bits)
+			continue;
+
+		mask_regx = base + flag_mux->offset + L3_FLAGMUX_MASK0 +
+			   (L3_APPLICATION_ERROR << 3);
+		mask_val = readl_relaxed(mask_regx);
+		mask_val &= ~(flag_mux->mask_app_bits);
+
+		writel_relaxed(mask_val, mask_regx);
+		mask_regx = base + flag_mux->offset + L3_FLAGMUX_MASK0 +
+			   (L3_DEBUG_ERROR << 3);
+		mask_val = readl_relaxed(mask_regx);
+		mask_val &= ~(flag_mux->mask_dbg_bits);
+
+		writel_relaxed(mask_val, mask_regx);
+	}
+
+	/* Dummy read to force OCP barrier */
+	if (mask_regx)
+		(void)readl(mask_regx);
+
+	return 0;
+}
+
+static const struct dev_pm_ops l3_dev_pm_ops = {
+	.resume_noirq		= l3_resume_noirq,
+};
+
+#define L3_DEV_PM_OPS (&l3_dev_pm_ops)
+#else
+#define L3_DEV_PM_OPS NULL
+#endif
+
 static struct platform_driver omap_l3_driver = {
 	.probe		= omap_l3_probe,
 	.driver		= {
 		.name		= "omap_l3_noc",
 		.owner		= THIS_MODULE,
+		.pm		= L3_DEV_PM_OPS,
 		.of_match_table = of_match_ptr(l3_noc_match),
 	},
 };