Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / kernel / irq / chip.c
diff --git a/kernel/kernel/irq/chip.c b/kernel/kernel/irq/chip.c
new file mode 100644 (file)
index 0000000..eb9a4ea
--- /dev/null
@@ -0,0 +1,993 @@
+/*
+ * linux/kernel/irq/chip.c
+ *
+ * Copyright (C) 1992, 1998-2006 Linus Torvalds, Ingo Molnar
+ * Copyright (C) 2005-2006, Thomas Gleixner, Russell King
+ *
+ * This file contains the core interrupt handling code, for irq-chip
+ * based architectures.
+ *
+ * Detailed information is available in Documentation/DocBook/genericirq
+ */
+
+#include <linux/irq.h>
+#include <linux/msi.h>
+#include <linux/module.h>
+#include <linux/interrupt.h>
+#include <linux/kernel_stat.h>
+#include <linux/irqdomain.h>
+
+#include <trace/events/irq.h>
+
+#include "internals.h"
+
+/**
+ *     irq_set_chip - set the irq chip for an irq
+ *     @irq:   irq number
+ *     @chip:  pointer to irq chip description structure
+ */
+int irq_set_chip(unsigned int irq, struct irq_chip *chip)
+{
+       unsigned long flags;
+       struct irq_desc *desc = irq_get_desc_lock(irq, &flags, 0);
+
+       if (!desc)
+               return -EINVAL;
+
+       if (!chip)
+               chip = &no_irq_chip;
+
+       desc->irq_data.chip = chip;
+       irq_put_desc_unlock(desc, flags);
+       /*
+        * For !CONFIG_SPARSE_IRQ make the irq show up in
+        * allocated_irqs.
+        */
+       irq_mark_irq(irq);
+       return 0;
+}
+EXPORT_SYMBOL(irq_set_chip);
+
+/**
+ *     irq_set_type - set the irq trigger type for an irq
+ *     @irq:   irq number
+ *     @type:  IRQ_TYPE_{LEVEL,EDGE}_* value - see include/linux/irq.h
+ */
+int irq_set_irq_type(unsigned int irq, unsigned int type)
+{
+       unsigned long flags;
+       struct irq_desc *desc = irq_get_desc_buslock(irq, &flags, IRQ_GET_DESC_CHECK_GLOBAL);
+       int ret = 0;
+
+       if (!desc)
+               return -EINVAL;
+
+       type &= IRQ_TYPE_SENSE_MASK;
+       ret = __irq_set_trigger(desc, irq, type);
+       irq_put_desc_busunlock(desc, flags);
+       return ret;
+}
+EXPORT_SYMBOL(irq_set_irq_type);
+
+/**
+ *     irq_set_handler_data - set irq handler data for an irq
+ *     @irq:   Interrupt number
+ *     @data:  Pointer to interrupt specific data
+ *
+ *     Set the hardware irq controller data for an irq
+ */
+int irq_set_handler_data(unsigned int irq, void *data)
+{
+       unsigned long flags;
+       struct irq_desc *desc = irq_get_desc_lock(irq, &flags, 0);
+
+       if (!desc)
+               return -EINVAL;
+       desc->irq_data.handler_data = data;
+       irq_put_desc_unlock(desc, flags);
+       return 0;
+}
+EXPORT_SYMBOL(irq_set_handler_data);
+
+/**
+ *     irq_set_msi_desc_off - set MSI descriptor data for an irq at offset
+ *     @irq_base:      Interrupt number base
+ *     @irq_offset:    Interrupt number offset
+ *     @entry:         Pointer to MSI descriptor data
+ *
+ *     Set the MSI descriptor entry for an irq at offset
+ */
+int irq_set_msi_desc_off(unsigned int irq_base, unsigned int irq_offset,
+                        struct msi_desc *entry)
+{
+       unsigned long flags;
+       struct irq_desc *desc = irq_get_desc_lock(irq_base + irq_offset, &flags, IRQ_GET_DESC_CHECK_GLOBAL);
+
+       if (!desc)
+               return -EINVAL;
+       desc->irq_data.msi_desc = entry;
+       if (entry && !irq_offset)
+               entry->irq = irq_base;
+       irq_put_desc_unlock(desc, flags);
+       return 0;
+}
+
+/**
+ *     irq_set_msi_desc - set MSI descriptor data for an irq
+ *     @irq:   Interrupt number
+ *     @entry: Pointer to MSI descriptor data
+ *
+ *     Set the MSI descriptor entry for an irq
+ */
+int irq_set_msi_desc(unsigned int irq, struct msi_desc *entry)
+{
+       return irq_set_msi_desc_off(irq, 0, entry);
+}
+
+/**
+ *     irq_set_chip_data - set irq chip data for an irq
+ *     @irq:   Interrupt number
+ *     @data:  Pointer to chip specific data
+ *
+ *     Set the hardware irq chip data for an irq
+ */
+int irq_set_chip_data(unsigned int irq, void *data)
+{
+       unsigned long flags;
+       struct irq_desc *desc = irq_get_desc_lock(irq, &flags, 0);
+
+       if (!desc)
+               return -EINVAL;
+       desc->irq_data.chip_data = data;
+       irq_put_desc_unlock(desc, flags);
+       return 0;
+}
+EXPORT_SYMBOL(irq_set_chip_data);
+
+struct irq_data *irq_get_irq_data(unsigned int irq)
+{
+       struct irq_desc *desc = irq_to_desc(irq);
+
+       return desc ? &desc->irq_data : NULL;
+}
+EXPORT_SYMBOL_GPL(irq_get_irq_data);
+
+static void irq_state_clr_disabled(struct irq_desc *desc)
+{
+       irqd_clear(&desc->irq_data, IRQD_IRQ_DISABLED);
+}
+
+static void irq_state_set_disabled(struct irq_desc *desc)
+{
+       irqd_set(&desc->irq_data, IRQD_IRQ_DISABLED);
+}
+
+static void irq_state_clr_masked(struct irq_desc *desc)
+{
+       irqd_clear(&desc->irq_data, IRQD_IRQ_MASKED);
+}
+
+static void irq_state_set_masked(struct irq_desc *desc)
+{
+       irqd_set(&desc->irq_data, IRQD_IRQ_MASKED);
+}
+
+int irq_startup(struct irq_desc *desc, bool resend)
+{
+       int ret = 0;
+
+       irq_state_clr_disabled(desc);
+       desc->depth = 0;
+
+       irq_domain_activate_irq(&desc->irq_data);
+       if (desc->irq_data.chip->irq_startup) {
+               ret = desc->irq_data.chip->irq_startup(&desc->irq_data);
+               irq_state_clr_masked(desc);
+       } else {
+               irq_enable(desc);
+       }
+       if (resend)
+               check_irq_resend(desc, desc->irq_data.irq);
+       return ret;
+}
+
+void irq_shutdown(struct irq_desc *desc)
+{
+       irq_state_set_disabled(desc);
+       desc->depth = 1;
+       if (desc->irq_data.chip->irq_shutdown)
+               desc->irq_data.chip->irq_shutdown(&desc->irq_data);
+       else if (desc->irq_data.chip->irq_disable)
+               desc->irq_data.chip->irq_disable(&desc->irq_data);
+       else
+               desc->irq_data.chip->irq_mask(&desc->irq_data);
+       irq_domain_deactivate_irq(&desc->irq_data);
+       irq_state_set_masked(desc);
+}
+
+void irq_enable(struct irq_desc *desc)
+{
+       irq_state_clr_disabled(desc);
+       if (desc->irq_data.chip->irq_enable)
+               desc->irq_data.chip->irq_enable(&desc->irq_data);
+       else
+               desc->irq_data.chip->irq_unmask(&desc->irq_data);
+       irq_state_clr_masked(desc);
+}
+
+/**
+ * irq_disable - Mark interrupt disabled
+ * @desc:      irq descriptor which should be disabled
+ *
+ * If the chip does not implement the irq_disable callback, we
+ * use a lazy disable approach. That means we mark the interrupt
+ * disabled, but leave the hardware unmasked. That's an
+ * optimization because we avoid the hardware access for the
+ * common case where no interrupt happens after we marked it
+ * disabled. If an interrupt happens, then the interrupt flow
+ * handler masks the line at the hardware level and marks it
+ * pending.
+ */
+void irq_disable(struct irq_desc *desc)
+{
+       irq_state_set_disabled(desc);
+       if (desc->irq_data.chip->irq_disable) {
+               desc->irq_data.chip->irq_disable(&desc->irq_data);
+               irq_state_set_masked(desc);
+       }
+}
+
+void irq_percpu_enable(struct irq_desc *desc, unsigned int cpu)
+{
+       if (desc->irq_data.chip->irq_enable)
+               desc->irq_data.chip->irq_enable(&desc->irq_data);
+       else
+               desc->irq_data.chip->irq_unmask(&desc->irq_data);
+       cpumask_set_cpu(cpu, desc->percpu_enabled);
+}
+
+void irq_percpu_disable(struct irq_desc *desc, unsigned int cpu)
+{
+       if (desc->irq_data.chip->irq_disable)
+               desc->irq_data.chip->irq_disable(&desc->irq_data);
+       else
+               desc->irq_data.chip->irq_mask(&desc->irq_data);
+       cpumask_clear_cpu(cpu, desc->percpu_enabled);
+}
+
+static inline void mask_ack_irq(struct irq_desc *desc)
+{
+       if (desc->irq_data.chip->irq_mask_ack)
+               desc->irq_data.chip->irq_mask_ack(&desc->irq_data);
+       else {
+               desc->irq_data.chip->irq_mask(&desc->irq_data);
+               if (desc->irq_data.chip->irq_ack)
+                       desc->irq_data.chip->irq_ack(&desc->irq_data);
+       }
+       irq_state_set_masked(desc);
+}
+
+void mask_irq(struct irq_desc *desc)
+{
+       if (desc->irq_data.chip->irq_mask) {
+               desc->irq_data.chip->irq_mask(&desc->irq_data);
+               irq_state_set_masked(desc);
+       }
+}
+
+void unmask_irq(struct irq_desc *desc)
+{
+       if (desc->irq_data.chip->irq_unmask) {
+               desc->irq_data.chip->irq_unmask(&desc->irq_data);
+               irq_state_clr_masked(desc);
+       }
+}
+
+void unmask_threaded_irq(struct irq_desc *desc)
+{
+       struct irq_chip *chip = desc->irq_data.chip;
+
+       if (chip->flags & IRQCHIP_EOI_THREADED)
+               chip->irq_eoi(&desc->irq_data);
+
+       if (chip->irq_unmask) {
+               chip->irq_unmask(&desc->irq_data);
+               irq_state_clr_masked(desc);
+       }
+}
+
+/*
+ *     handle_nested_irq - Handle a nested irq from a irq thread
+ *     @irq:   the interrupt number
+ *
+ *     Handle interrupts which are nested into a threaded interrupt
+ *     handler. The handler function is called inside the calling
+ *     threads context.
+ */
+void handle_nested_irq(unsigned int irq)
+{
+       struct irq_desc *desc = irq_to_desc(irq);
+       struct irqaction *action;
+       irqreturn_t action_ret;
+
+       might_sleep();
+
+       raw_spin_lock_irq(&desc->lock);
+
+       desc->istate &= ~(IRQS_REPLAY | IRQS_WAITING);
+       kstat_incr_irqs_this_cpu(irq, desc);
+
+       action = desc->action;
+       if (unlikely(!action || irqd_irq_disabled(&desc->irq_data))) {
+               desc->istate |= IRQS_PENDING;
+               goto out_unlock;
+       }
+
+       irqd_set(&desc->irq_data, IRQD_IRQ_INPROGRESS);
+       raw_spin_unlock_irq(&desc->lock);
+
+       action_ret = action->thread_fn(action->irq, action->dev_id);
+       if (!noirqdebug)
+               note_interrupt(irq, desc, action_ret);
+
+       raw_spin_lock_irq(&desc->lock);
+       irqd_clear(&desc->irq_data, IRQD_IRQ_INPROGRESS);
+
+out_unlock:
+       raw_spin_unlock_irq(&desc->lock);
+}
+EXPORT_SYMBOL_GPL(handle_nested_irq);
+
+static bool irq_check_poll(struct irq_desc *desc)
+{
+       if (!(desc->istate & IRQS_POLL_INPROGRESS))
+               return false;
+       return irq_wait_for_poll(desc);
+}
+
+static bool irq_may_run(struct irq_desc *desc)
+{
+       unsigned int mask = IRQD_IRQ_INPROGRESS | IRQD_WAKEUP_ARMED;
+
+       /*
+        * If the interrupt is not in progress and is not an armed
+        * wakeup interrupt, proceed.
+        */
+       if (!irqd_has_set(&desc->irq_data, mask))
+               return true;
+
+       /*
+        * If the interrupt is an armed wakeup source, mark it pending
+        * and suspended, disable it and notify the pm core about the
+        * event.
+        */
+       if (irq_pm_check_wakeup(desc))
+               return false;
+
+       /*
+        * Handle a potential concurrent poll on a different core.
+        */
+       return irq_check_poll(desc);
+}
+
+/**
+ *     handle_simple_irq - Simple and software-decoded IRQs.
+ *     @irq:   the interrupt number
+ *     @desc:  the interrupt description structure for this irq
+ *
+ *     Simple interrupts are either sent from a demultiplexing interrupt
+ *     handler or come from hardware, where no interrupt hardware control
+ *     is necessary.
+ *
+ *     Note: The caller is expected to handle the ack, clear, mask and
+ *     unmask issues if necessary.
+ */
+void
+handle_simple_irq(unsigned int irq, struct irq_desc *desc)
+{
+       raw_spin_lock(&desc->lock);
+
+       if (!irq_may_run(desc))
+               goto out_unlock;
+
+       desc->istate &= ~(IRQS_REPLAY | IRQS_WAITING);
+       kstat_incr_irqs_this_cpu(irq, desc);
+
+       if (unlikely(!desc->action || irqd_irq_disabled(&desc->irq_data))) {
+               desc->istate |= IRQS_PENDING;
+               goto out_unlock;
+       }
+
+       handle_irq_event(desc);
+
+out_unlock:
+       raw_spin_unlock(&desc->lock);
+}
+EXPORT_SYMBOL_GPL(handle_simple_irq);
+
+/*
+ * Called unconditionally from handle_level_irq() and only for oneshot
+ * interrupts from handle_fasteoi_irq()
+ */
+static void cond_unmask_irq(struct irq_desc *desc)
+{
+       /*
+        * We need to unmask in the following cases:
+        * - Standard level irq (IRQF_ONESHOT is not set)
+        * - Oneshot irq which did not wake the thread (caused by a
+        *   spurious interrupt or a primary handler handling it
+        *   completely).
+        */
+       if (!irqd_irq_disabled(&desc->irq_data) &&
+           irqd_irq_masked(&desc->irq_data) && !desc->threads_oneshot)
+               unmask_irq(desc);
+}
+
+/**
+ *     handle_level_irq - Level type irq handler
+ *     @irq:   the interrupt number
+ *     @desc:  the interrupt description structure for this irq
+ *
+ *     Level type interrupts are active as long as the hardware line has
+ *     the active level. This may require to mask the interrupt and unmask
+ *     it after the associated handler has acknowledged the device, so the
+ *     interrupt line is back to inactive.
+ */
+void
+handle_level_irq(unsigned int irq, struct irq_desc *desc)
+{
+       raw_spin_lock(&desc->lock);
+       mask_ack_irq(desc);
+
+       if (!irq_may_run(desc))
+               goto out_unlock;
+
+       desc->istate &= ~(IRQS_REPLAY | IRQS_WAITING);
+       kstat_incr_irqs_this_cpu(irq, desc);
+
+       /*
+        * If its disabled or no action available
+        * keep it masked and get out of here
+        */
+       if (unlikely(!desc->action || irqd_irq_disabled(&desc->irq_data))) {
+               desc->istate |= IRQS_PENDING;
+               goto out_unlock;
+       }
+
+       handle_irq_event(desc);
+
+       cond_unmask_irq(desc);
+
+out_unlock:
+       raw_spin_unlock(&desc->lock);
+}
+EXPORT_SYMBOL_GPL(handle_level_irq);
+
+#ifdef CONFIG_IRQ_PREFLOW_FASTEOI
+static inline void preflow_handler(struct irq_desc *desc)
+{
+       if (desc->preflow_handler)
+               desc->preflow_handler(&desc->irq_data);
+}
+#else
+static inline void preflow_handler(struct irq_desc *desc) { }
+#endif
+
+static void cond_unmask_eoi_irq(struct irq_desc *desc, struct irq_chip *chip)
+{
+       if (!(desc->istate & IRQS_ONESHOT)) {
+               chip->irq_eoi(&desc->irq_data);
+               return;
+       }
+       /*
+        * We need to unmask in the following cases:
+        * - Oneshot irq which did not wake the thread (caused by a
+        *   spurious interrupt or a primary handler handling it
+        *   completely).
+        */
+       if (!irqd_irq_disabled(&desc->irq_data) &&
+           irqd_irq_masked(&desc->irq_data) && !desc->threads_oneshot) {
+               chip->irq_eoi(&desc->irq_data);
+               unmask_irq(desc);
+       } else if (!(chip->flags & IRQCHIP_EOI_THREADED)) {
+               chip->irq_eoi(&desc->irq_data);
+       }
+}
+
+/**
+ *     handle_fasteoi_irq - irq handler for transparent controllers
+ *     @irq:   the interrupt number
+ *     @desc:  the interrupt description structure for this irq
+ *
+ *     Only a single callback will be issued to the chip: an ->eoi()
+ *     call when the interrupt has been serviced. This enables support
+ *     for modern forms of interrupt handlers, which handle the flow
+ *     details in hardware, transparently.
+ */
+void
+handle_fasteoi_irq(unsigned int irq, struct irq_desc *desc)
+{
+       struct irq_chip *chip = desc->irq_data.chip;
+
+       raw_spin_lock(&desc->lock);
+
+       if (!irq_may_run(desc))
+               goto out;
+
+       desc->istate &= ~(IRQS_REPLAY | IRQS_WAITING);
+       kstat_incr_irqs_this_cpu(irq, desc);
+
+       /*
+        * If its disabled or no action available
+        * then mask it and get out of here:
+        */
+       if (unlikely(!desc->action || irqd_irq_disabled(&desc->irq_data))) {
+               desc->istate |= IRQS_PENDING;
+               mask_irq(desc);
+               goto out;
+       }
+
+       if (desc->istate & IRQS_ONESHOT)
+               mask_irq(desc);
+
+       preflow_handler(desc);
+       handle_irq_event(desc);
+
+       cond_unmask_eoi_irq(desc, chip);
+
+       raw_spin_unlock(&desc->lock);
+       return;
+out:
+       if (!(chip->flags & IRQCHIP_EOI_IF_HANDLED))
+               chip->irq_eoi(&desc->irq_data);
+       raw_spin_unlock(&desc->lock);
+}
+EXPORT_SYMBOL_GPL(handle_fasteoi_irq);
+
+/**
+ *     handle_edge_irq - edge type IRQ handler
+ *     @irq:   the interrupt number
+ *     @desc:  the interrupt description structure for this irq
+ *
+ *     Interrupt occures on the falling and/or rising edge of a hardware
+ *     signal. The occurrence is latched into the irq controller hardware
+ *     and must be acked in order to be reenabled. After the ack another
+ *     interrupt can happen on the same source even before the first one
+ *     is handled by the associated event handler. If this happens it
+ *     might be necessary to disable (mask) the interrupt depending on the
+ *     controller hardware. This requires to reenable the interrupt inside
+ *     of the loop which handles the interrupts which have arrived while
+ *     the handler was running. If all pending interrupts are handled, the
+ *     loop is left.
+ */
+void
+handle_edge_irq(unsigned int irq, struct irq_desc *desc)
+{
+       raw_spin_lock(&desc->lock);
+
+       desc->istate &= ~(IRQS_REPLAY | IRQS_WAITING);
+
+       if (!irq_may_run(desc)) {
+               desc->istate |= IRQS_PENDING;
+               mask_ack_irq(desc);
+               goto out_unlock;
+       }
+
+       /*
+        * If its disabled or no action available then mask it and get
+        * out of here.
+        */
+       if (irqd_irq_disabled(&desc->irq_data) || !desc->action) {
+               desc->istate |= IRQS_PENDING;
+               mask_ack_irq(desc);
+               goto out_unlock;
+       }
+
+       kstat_incr_irqs_this_cpu(irq, desc);
+
+       /* Start handling the irq */
+       desc->irq_data.chip->irq_ack(&desc->irq_data);
+
+       do {
+               if (unlikely(!desc->action)) {
+                       mask_irq(desc);
+                       goto out_unlock;
+               }
+
+               /*
+                * When another irq arrived while we were handling
+                * one, we could have masked the irq.
+                * Renable it, if it was not disabled in meantime.
+                */
+               if (unlikely(desc->istate & IRQS_PENDING)) {
+                       if (!irqd_irq_disabled(&desc->irq_data) &&
+                           irqd_irq_masked(&desc->irq_data))
+                               unmask_irq(desc);
+               }
+
+               handle_irq_event(desc);
+
+       } while ((desc->istate & IRQS_PENDING) &&
+                !irqd_irq_disabled(&desc->irq_data));
+
+out_unlock:
+       raw_spin_unlock(&desc->lock);
+}
+EXPORT_SYMBOL(handle_edge_irq);
+
+#ifdef CONFIG_IRQ_EDGE_EOI_HANDLER
+/**
+ *     handle_edge_eoi_irq - edge eoi type IRQ handler
+ *     @irq:   the interrupt number
+ *     @desc:  the interrupt description structure for this irq
+ *
+ * Similar as the above handle_edge_irq, but using eoi and w/o the
+ * mask/unmask logic.
+ */
+void handle_edge_eoi_irq(unsigned int irq, struct irq_desc *desc)
+{
+       struct irq_chip *chip = irq_desc_get_chip(desc);
+
+       raw_spin_lock(&desc->lock);
+
+       desc->istate &= ~(IRQS_REPLAY | IRQS_WAITING);
+
+       if (!irq_may_run(desc)) {
+               desc->istate |= IRQS_PENDING;
+               goto out_eoi;
+       }
+
+       /*
+        * If its disabled or no action available then mask it and get
+        * out of here.
+        */
+       if (irqd_irq_disabled(&desc->irq_data) || !desc->action) {
+               desc->istate |= IRQS_PENDING;
+               goto out_eoi;
+       }
+
+       kstat_incr_irqs_this_cpu(irq, desc);
+
+       do {
+               if (unlikely(!desc->action))
+                       goto out_eoi;
+
+               handle_irq_event(desc);
+
+       } while ((desc->istate & IRQS_PENDING) &&
+                !irqd_irq_disabled(&desc->irq_data));
+
+out_eoi:
+       chip->irq_eoi(&desc->irq_data);
+       raw_spin_unlock(&desc->lock);
+}
+#endif
+
+/**
+ *     handle_percpu_irq - Per CPU local irq handler
+ *     @irq:   the interrupt number
+ *     @desc:  the interrupt description structure for this irq
+ *
+ *     Per CPU interrupts on SMP machines without locking requirements
+ */
+void
+handle_percpu_irq(unsigned int irq, struct irq_desc *desc)
+{
+       struct irq_chip *chip = irq_desc_get_chip(desc);
+
+       kstat_incr_irqs_this_cpu(irq, desc);
+
+       if (chip->irq_ack)
+               chip->irq_ack(&desc->irq_data);
+
+       handle_irq_event_percpu(desc, desc->action);
+
+       if (chip->irq_eoi)
+               chip->irq_eoi(&desc->irq_data);
+}
+
+/**
+ * handle_percpu_devid_irq - Per CPU local irq handler with per cpu dev ids
+ * @irq:       the interrupt number
+ * @desc:      the interrupt description structure for this irq
+ *
+ * Per CPU interrupts on SMP machines without locking requirements. Same as
+ * handle_percpu_irq() above but with the following extras:
+ *
+ * action->percpu_dev_id is a pointer to percpu variables which
+ * contain the real device id for the cpu on which this handler is
+ * called
+ */
+void handle_percpu_devid_irq(unsigned int irq, struct irq_desc *desc)
+{
+       struct irq_chip *chip = irq_desc_get_chip(desc);
+       struct irqaction *action = desc->action;
+       void *dev_id = raw_cpu_ptr(action->percpu_dev_id);
+       irqreturn_t res;
+
+       kstat_incr_irqs_this_cpu(irq, desc);
+
+       if (chip->irq_ack)
+               chip->irq_ack(&desc->irq_data);
+
+       trace_irq_handler_entry(irq, action);
+       res = action->handler(irq, dev_id);
+       trace_irq_handler_exit(irq, action, res);
+
+       if (chip->irq_eoi)
+               chip->irq_eoi(&desc->irq_data);
+}
+
+void
+__irq_set_handler(unsigned int irq, irq_flow_handler_t handle, int is_chained,
+                 const char *name)
+{
+       unsigned long flags;
+       struct irq_desc *desc = irq_get_desc_buslock(irq, &flags, 0);
+
+       if (!desc)
+               return;
+
+       if (!handle) {
+               handle = handle_bad_irq;
+       } else {
+               struct irq_data *irq_data = &desc->irq_data;
+#ifdef CONFIG_IRQ_DOMAIN_HIERARCHY
+               /*
+                * With hierarchical domains we might run into a
+                * situation where the outermost chip is not yet set
+                * up, but the inner chips are there.  Instead of
+                * bailing we install the handler, but obviously we
+                * cannot enable/startup the interrupt at this point.
+                */
+               while (irq_data) {
+                       if (irq_data->chip != &no_irq_chip)
+                               break;
+                       /*
+                        * Bail out if the outer chip is not set up
+                        * and the interrrupt supposed to be started
+                        * right away.
+                        */
+                       if (WARN_ON(is_chained))
+                               goto out;
+                       /* Try the parent */
+                       irq_data = irq_data->parent_data;
+               }
+#endif
+               if (WARN_ON(!irq_data || irq_data->chip == &no_irq_chip))
+                       goto out;
+       }
+
+       /* Uninstall? */
+       if (handle == handle_bad_irq) {
+               if (desc->irq_data.chip != &no_irq_chip)
+                       mask_ack_irq(desc);
+               irq_state_set_disabled(desc);
+               desc->depth = 1;
+       }
+       desc->handle_irq = handle;
+       desc->name = name;
+
+       if (handle != handle_bad_irq && is_chained) {
+               irq_settings_set_noprobe(desc);
+               irq_settings_set_norequest(desc);
+               irq_settings_set_nothread(desc);
+               irq_startup(desc, true);
+       }
+out:
+       irq_put_desc_busunlock(desc, flags);
+}
+EXPORT_SYMBOL_GPL(__irq_set_handler);
+
+void
+irq_set_chip_and_handler_name(unsigned int irq, struct irq_chip *chip,
+                             irq_flow_handler_t handle, const char *name)
+{
+       irq_set_chip(irq, chip);
+       __irq_set_handler(irq, handle, 0, name);
+}
+EXPORT_SYMBOL_GPL(irq_set_chip_and_handler_name);
+
+void irq_modify_status(unsigned int irq, unsigned long clr, unsigned long set)
+{
+       unsigned long flags;
+       struct irq_desc *desc = irq_get_desc_lock(irq, &flags, 0);
+
+       if (!desc)
+               return;
+       irq_settings_clr_and_set(desc, clr, set);
+
+       irqd_clear(&desc->irq_data, IRQD_NO_BALANCING | IRQD_PER_CPU |
+                  IRQD_TRIGGER_MASK | IRQD_LEVEL | IRQD_MOVE_PCNTXT);
+       if (irq_settings_has_no_balance_set(desc))
+               irqd_set(&desc->irq_data, IRQD_NO_BALANCING);
+       if (irq_settings_is_per_cpu(desc))
+               irqd_set(&desc->irq_data, IRQD_PER_CPU);
+       if (irq_settings_can_move_pcntxt(desc))
+               irqd_set(&desc->irq_data, IRQD_MOVE_PCNTXT);
+       if (irq_settings_is_level(desc))
+               irqd_set(&desc->irq_data, IRQD_LEVEL);
+
+       irqd_set(&desc->irq_data, irq_settings_get_trigger_mask(desc));
+
+       irq_put_desc_unlock(desc, flags);
+}
+EXPORT_SYMBOL_GPL(irq_modify_status);
+
+/**
+ *     irq_cpu_online - Invoke all irq_cpu_online functions.
+ *
+ *     Iterate through all irqs and invoke the chip.irq_cpu_online()
+ *     for each.
+ */
+void irq_cpu_online(void)
+{
+       struct irq_desc *desc;
+       struct irq_chip *chip;
+       unsigned long flags;
+       unsigned int irq;
+
+       for_each_active_irq(irq) {
+               desc = irq_to_desc(irq);
+               if (!desc)
+                       continue;
+
+               raw_spin_lock_irqsave(&desc->lock, flags);
+
+               chip = irq_data_get_irq_chip(&desc->irq_data);
+               if (chip && chip->irq_cpu_online &&
+                   (!(chip->flags & IRQCHIP_ONOFFLINE_ENABLED) ||
+                    !irqd_irq_disabled(&desc->irq_data)))
+                       chip->irq_cpu_online(&desc->irq_data);
+
+               raw_spin_unlock_irqrestore(&desc->lock, flags);
+       }
+}
+
+/**
+ *     irq_cpu_offline - Invoke all irq_cpu_offline functions.
+ *
+ *     Iterate through all irqs and invoke the chip.irq_cpu_offline()
+ *     for each.
+ */
+void irq_cpu_offline(void)
+{
+       struct irq_desc *desc;
+       struct irq_chip *chip;
+       unsigned long flags;
+       unsigned int irq;
+
+       for_each_active_irq(irq) {
+               desc = irq_to_desc(irq);
+               if (!desc)
+                       continue;
+
+               raw_spin_lock_irqsave(&desc->lock, flags);
+
+               chip = irq_data_get_irq_chip(&desc->irq_data);
+               if (chip && chip->irq_cpu_offline &&
+                   (!(chip->flags & IRQCHIP_ONOFFLINE_ENABLED) ||
+                    !irqd_irq_disabled(&desc->irq_data)))
+                       chip->irq_cpu_offline(&desc->irq_data);
+
+               raw_spin_unlock_irqrestore(&desc->lock, flags);
+       }
+}
+
+#ifdef CONFIG_IRQ_DOMAIN_HIERARCHY
+/**
+ * irq_chip_ack_parent - Acknowledge the parent interrupt
+ * @data:      Pointer to interrupt specific data
+ */
+void irq_chip_ack_parent(struct irq_data *data)
+{
+       data = data->parent_data;
+       data->chip->irq_ack(data);
+}
+
+/**
+ * irq_chip_mask_parent - Mask the parent interrupt
+ * @data:      Pointer to interrupt specific data
+ */
+void irq_chip_mask_parent(struct irq_data *data)
+{
+       data = data->parent_data;
+       data->chip->irq_mask(data);
+}
+
+/**
+ * irq_chip_unmask_parent - Unmask the parent interrupt
+ * @data:      Pointer to interrupt specific data
+ */
+void irq_chip_unmask_parent(struct irq_data *data)
+{
+       data = data->parent_data;
+       data->chip->irq_unmask(data);
+}
+
+/**
+ * irq_chip_eoi_parent - Invoke EOI on the parent interrupt
+ * @data:      Pointer to interrupt specific data
+ */
+void irq_chip_eoi_parent(struct irq_data *data)
+{
+       data = data->parent_data;
+       data->chip->irq_eoi(data);
+}
+
+/**
+ * irq_chip_set_affinity_parent - Set affinity on the parent interrupt
+ * @data:      Pointer to interrupt specific data
+ * @dest:      The affinity mask to set
+ * @force:     Flag to enforce setting (disable online checks)
+ *
+ * Conditinal, as the underlying parent chip might not implement it.
+ */
+int irq_chip_set_affinity_parent(struct irq_data *data,
+                                const struct cpumask *dest, bool force)
+{
+       data = data->parent_data;
+       if (data->chip->irq_set_affinity)
+               return data->chip->irq_set_affinity(data, dest, force);
+
+       return -ENOSYS;
+}
+
+/**
+ * irq_chip_retrigger_hierarchy - Retrigger an interrupt in hardware
+ * @data:      Pointer to interrupt specific data
+ *
+ * Iterate through the domain hierarchy of the interrupt and check
+ * whether a hw retrigger function exists. If yes, invoke it.
+ */
+int irq_chip_retrigger_hierarchy(struct irq_data *data)
+{
+       for (data = data->parent_data; data; data = data->parent_data)
+               if (data->chip && data->chip->irq_retrigger)
+                       return data->chip->irq_retrigger(data);
+
+       return -ENOSYS;
+}
+
+/**
+ * irq_chip_set_wake_parent - Set/reset wake-up on the parent interrupt
+ * @data:      Pointer to interrupt specific data
+ * @on:                Whether to set or reset the wake-up capability of this irq
+ *
+ * Conditional, as the underlying parent chip might not implement it.
+ */
+int irq_chip_set_wake_parent(struct irq_data *data, unsigned int on)
+{
+       data = data->parent_data;
+       if (data->chip->irq_set_wake)
+               return data->chip->irq_set_wake(data, on);
+
+       return -ENOSYS;
+}
+#endif
+
+/**
+ * irq_chip_compose_msi_msg - Componse msi message for a irq chip
+ * @data:      Pointer to interrupt specific data
+ * @msg:       Pointer to the MSI message
+ *
+ * For hierarchical domains we find the first chip in the hierarchy
+ * which implements the irq_compose_msi_msg callback. For non
+ * hierarchical we use the top level chip.
+ */
+int irq_chip_compose_msi_msg(struct irq_data *data, struct msi_msg *msg)
+{
+       struct irq_data *pos = NULL;
+
+#ifdef CONFIG_IRQ_DOMAIN_HIERARCHY
+       for (; data; data = data->parent_data)
+#endif
+               if (data->chip && data->chip->irq_compose_msi_msg)
+                       pos = data;
+       if (!pos)
+               return -ENOSYS;
+
+       pos->chip->irq_compose_msi_msg(pos, msg);
+
+       return 0;
+}