Merge branch 'irqchip/gic-vic-move' into next/virt
authorOlof Johansson <olof@lixom.net>
Mon, 11 Feb 2013 17:05:29 +0000 (09:05 -0800)
committerOlof Johansson <olof@lixom.net>
Mon, 11 Feb 2013 17:05:29 +0000 (09:05 -0800)
1  2 
MAINTAINERS
arch/arm/kernel/smp.c
arch/arm/mach-imx/common.h
arch/arm/mach-imx/platsmp.c
arch/arm/mach-omap2/board-omap4panda.c
drivers/irqchip/irq-gic.c
drivers/mfd/db8500-prcmu.c

diff --cc MAINTAINERS
Simple merge
Simple merge
Simple merge
Simple merge
index 0000000000000000000000000000000000000000,69d9a395d54c671e0e1068c710cca76b92a2a5f5..644d724684232d9b389fcc056f78aaab1e155817
mode 000000,100644..100644
--- /dev/null
@@@ -1,0 -1,824 +1,845 @@@
 -      cpumask = readl_relaxed(base + GIC_DIST_TARGET + 0);
+ /*
+  *  linux/arch/arm/common/gic.c
+  *
+  *  Copyright (C) 2002 ARM Limited, All Rights Reserved.
+  *
+  * This program is free software; you can redistribute it and/or modify
+  * it under the terms of the GNU General Public License version 2 as
+  * published by the Free Software Foundation.
+  *
+  * Interrupt architecture for the GIC:
+  *
+  * o There is one Interrupt Distributor, which receives interrupts
+  *   from system devices and sends them to the Interrupt Controllers.
+  *
+  * o There is one CPU Interface per CPU, which sends interrupts sent
+  *   by the Distributor, and interrupts generated locally, to the
+  *   associated CPU. The base address of the CPU interface is usually
+  *   aliased so that the same address points to different chips depending
+  *   on the CPU it is accessed from.
+  *
+  * Note that IRQs 0-31 are special - they are local to each CPU.
+  * As such, the enable set/clear, pending set/clear and active bit
+  * registers are banked per-cpu for these sources.
+  */
+ #include <linux/init.h>
+ #include <linux/kernel.h>
+ #include <linux/err.h>
+ #include <linux/module.h>
+ #include <linux/list.h>
+ #include <linux/smp.h>
+ #include <linux/cpu_pm.h>
+ #include <linux/cpumask.h>
+ #include <linux/io.h>
+ #include <linux/of.h>
+ #include <linux/of_address.h>
+ #include <linux/of_irq.h>
+ #include <linux/irqdomain.h>
+ #include <linux/interrupt.h>
+ #include <linux/percpu.h>
+ #include <linux/slab.h>
+ #include <linux/irqchip/arm-gic.h>
+ #include <asm/irq.h>
+ #include <asm/exception.h>
+ #include <asm/smp_plat.h>
+ #include <asm/mach/irq.h>
+ #include "irqchip.h"
+ union gic_base {
+       void __iomem *common_base;
+       void __percpu __iomem **percpu_base;
+ };
+ struct gic_chip_data {
+       union gic_base dist_base;
+       union gic_base cpu_base;
+ #ifdef CONFIG_CPU_PM
+       u32 saved_spi_enable[DIV_ROUND_UP(1020, 32)];
+       u32 saved_spi_conf[DIV_ROUND_UP(1020, 16)];
+       u32 saved_spi_target[DIV_ROUND_UP(1020, 4)];
+       u32 __percpu *saved_ppi_enable;
+       u32 __percpu *saved_ppi_conf;
+ #endif
+       struct irq_domain *domain;
+       unsigned int gic_irqs;
+ #ifdef CONFIG_GIC_NON_BANKED
+       void __iomem *(*get_base)(union gic_base *);
+ #endif
+ };
+ static DEFINE_RAW_SPINLOCK(irq_controller_lock);
+ /*
+  * The GIC mapping of CPU interfaces does not necessarily match
+  * the logical CPU numbering.  Let's use a mapping as returned
+  * by the GIC itself.
+  */
+ #define NR_GIC_CPU_IF 8
+ static u8 gic_cpu_map[NR_GIC_CPU_IF] __read_mostly;
+ /*
+  * Supported arch specific GIC irq extension.
+  * Default make them NULL.
+  */
+ struct irq_chip gic_arch_extn = {
+       .irq_eoi        = NULL,
+       .irq_mask       = NULL,
+       .irq_unmask     = NULL,
+       .irq_retrigger  = NULL,
+       .irq_set_type   = NULL,
+       .irq_set_wake   = NULL,
+ };
+ #ifndef MAX_GIC_NR
+ #define MAX_GIC_NR    1
+ #endif
+ static struct gic_chip_data gic_data[MAX_GIC_NR] __read_mostly;
+ #ifdef CONFIG_GIC_NON_BANKED
+ static void __iomem *gic_get_percpu_base(union gic_base *base)
+ {
+       return *__this_cpu_ptr(base->percpu_base);
+ }
+ static void __iomem *gic_get_common_base(union gic_base *base)
+ {
+       return base->common_base;
+ }
+ static inline void __iomem *gic_data_dist_base(struct gic_chip_data *data)
+ {
+       return data->get_base(&data->dist_base);
+ }
+ static inline void __iomem *gic_data_cpu_base(struct gic_chip_data *data)
+ {
+       return data->get_base(&data->cpu_base);
+ }
+ static inline void gic_set_base_accessor(struct gic_chip_data *data,
+                                        void __iomem *(*f)(union gic_base *))
+ {
+       data->get_base = f;
+ }
+ #else
+ #define gic_data_dist_base(d) ((d)->dist_base.common_base)
+ #define gic_data_cpu_base(d)  ((d)->cpu_base.common_base)
+ #define gic_set_base_accessor(d,f)
+ #endif
+ static inline void __iomem *gic_dist_base(struct irq_data *d)
+ {
+       struct gic_chip_data *gic_data = irq_data_get_irq_chip_data(d);
+       return gic_data_dist_base(gic_data);
+ }
+ static inline void __iomem *gic_cpu_base(struct irq_data *d)
+ {
+       struct gic_chip_data *gic_data = irq_data_get_irq_chip_data(d);
+       return gic_data_cpu_base(gic_data);
+ }
+ static inline unsigned int gic_irq(struct irq_data *d)
+ {
+       return d->hwirq;
+ }
+ /*
+  * Routines to acknowledge, disable and enable interrupts
+  */
+ static void gic_mask_irq(struct irq_data *d)
+ {
+       u32 mask = 1 << (gic_irq(d) % 32);
+       raw_spin_lock(&irq_controller_lock);
+       writel_relaxed(mask, gic_dist_base(d) + GIC_DIST_ENABLE_CLEAR + (gic_irq(d) / 32) * 4);
+       if (gic_arch_extn.irq_mask)
+               gic_arch_extn.irq_mask(d);
+       raw_spin_unlock(&irq_controller_lock);
+ }
+ static void gic_unmask_irq(struct irq_data *d)
+ {
+       u32 mask = 1 << (gic_irq(d) % 32);
+       raw_spin_lock(&irq_controller_lock);
+       if (gic_arch_extn.irq_unmask)
+               gic_arch_extn.irq_unmask(d);
+       writel_relaxed(mask, gic_dist_base(d) + GIC_DIST_ENABLE_SET + (gic_irq(d) / 32) * 4);
+       raw_spin_unlock(&irq_controller_lock);
+ }
+ static void gic_eoi_irq(struct irq_data *d)
+ {
+       if (gic_arch_extn.irq_eoi) {
+               raw_spin_lock(&irq_controller_lock);
+               gic_arch_extn.irq_eoi(d);
+               raw_spin_unlock(&irq_controller_lock);
+       }
+       writel_relaxed(gic_irq(d), gic_cpu_base(d) + GIC_CPU_EOI);
+ }
+ static int gic_set_type(struct irq_data *d, unsigned int type)
+ {
+       void __iomem *base = gic_dist_base(d);
+       unsigned int gicirq = gic_irq(d);
+       u32 enablemask = 1 << (gicirq % 32);
+       u32 enableoff = (gicirq / 32) * 4;
+       u32 confmask = 0x2 << ((gicirq % 16) * 2);
+       u32 confoff = (gicirq / 16) * 4;
+       bool enabled = false;
+       u32 val;
+       /* Interrupt configuration for SGIs can't be changed */
+       if (gicirq < 16)
+               return -EINVAL;
+       if (type != IRQ_TYPE_LEVEL_HIGH && type != IRQ_TYPE_EDGE_RISING)
+               return -EINVAL;
+       raw_spin_lock(&irq_controller_lock);
+       if (gic_arch_extn.irq_set_type)
+               gic_arch_extn.irq_set_type(d, type);
+       val = readl_relaxed(base + GIC_DIST_CONFIG + confoff);
+       if (type == IRQ_TYPE_LEVEL_HIGH)
+               val &= ~confmask;
+       else if (type == IRQ_TYPE_EDGE_RISING)
+               val |= confmask;
+       /*
+        * As recommended by the spec, disable the interrupt before changing
+        * the configuration
+        */
+       if (readl_relaxed(base + GIC_DIST_ENABLE_SET + enableoff) & enablemask) {
+               writel_relaxed(enablemask, base + GIC_DIST_ENABLE_CLEAR + enableoff);
+               enabled = true;
+       }
+       writel_relaxed(val, base + GIC_DIST_CONFIG + confoff);
+       if (enabled)
+               writel_relaxed(enablemask, base + GIC_DIST_ENABLE_SET + enableoff);
+       raw_spin_unlock(&irq_controller_lock);
+       return 0;
+ }
+ static int gic_retrigger(struct irq_data *d)
+ {
+       if (gic_arch_extn.irq_retrigger)
+               return gic_arch_extn.irq_retrigger(d);
+       return -ENXIO;
+ }
+ #ifdef CONFIG_SMP
+ static int gic_set_affinity(struct irq_data *d, const struct cpumask *mask_val,
+                           bool force)
+ {
+       void __iomem *reg = gic_dist_base(d) + GIC_DIST_TARGET + (gic_irq(d) & ~3);
+       unsigned int shift = (gic_irq(d) % 4) * 8;
+       unsigned int cpu = cpumask_any_and(mask_val, cpu_online_mask);
+       u32 val, mask, bit;
+       if (cpu >= NR_GIC_CPU_IF || cpu >= nr_cpu_ids)
+               return -EINVAL;
+       mask = 0xff << shift;
+       bit = gic_cpu_map[cpu] << shift;
+       raw_spin_lock(&irq_controller_lock);
+       val = readl_relaxed(reg) & ~mask;
+       writel_relaxed(val | bit, reg);
+       raw_spin_unlock(&irq_controller_lock);
+       return IRQ_SET_MASK_OK;
+ }
+ #endif
+ #ifdef CONFIG_PM
+ static int gic_set_wake(struct irq_data *d, unsigned int on)
+ {
+       int ret = -ENXIO;
+       if (gic_arch_extn.irq_set_wake)
+               ret = gic_arch_extn.irq_set_wake(d, on);
+       return ret;
+ }
+ #else
+ #define gic_set_wake  NULL
+ #endif
+ static asmlinkage void __exception_irq_entry gic_handle_irq(struct pt_regs *regs)
+ {
+       u32 irqstat, irqnr;
+       struct gic_chip_data *gic = &gic_data[0];
+       void __iomem *cpu_base = gic_data_cpu_base(gic);
+       do {
+               irqstat = readl_relaxed(cpu_base + GIC_CPU_INTACK);
+               irqnr = irqstat & ~0x1c00;
+               if (likely(irqnr > 15 && irqnr < 1021)) {
+                       irqnr = irq_find_mapping(gic->domain, irqnr);
+                       handle_IRQ(irqnr, regs);
+                       continue;
+               }
+               if (irqnr < 16) {
+                       writel_relaxed(irqstat, cpu_base + GIC_CPU_EOI);
+ #ifdef CONFIG_SMP
+                       handle_IPI(irqnr, regs);
+ #endif
+                       continue;
+               }
+               break;
+       } while (1);
+ }
+ static void gic_handle_cascade_irq(unsigned int irq, struct irq_desc *desc)
+ {
+       struct gic_chip_data *chip_data = irq_get_handler_data(irq);
+       struct irq_chip *chip = irq_get_chip(irq);
+       unsigned int cascade_irq, gic_irq;
+       unsigned long status;
+       chained_irq_enter(chip, desc);
+       raw_spin_lock(&irq_controller_lock);
+       status = readl_relaxed(gic_data_cpu_base(chip_data) + GIC_CPU_INTACK);
+       raw_spin_unlock(&irq_controller_lock);
+       gic_irq = (status & 0x3ff);
+       if (gic_irq == 1023)
+               goto out;
+       cascade_irq = irq_find_mapping(chip_data->domain, gic_irq);
+       if (unlikely(gic_irq < 32 || gic_irq > 1020))
+               do_bad_IRQ(cascade_irq, desc);
+       else
+               generic_handle_irq(cascade_irq);
+  out:
+       chained_irq_exit(chip, desc);
+ }
+ static struct irq_chip gic_chip = {
+       .name                   = "GIC",
+       .irq_mask               = gic_mask_irq,
+       .irq_unmask             = gic_unmask_irq,
+       .irq_eoi                = gic_eoi_irq,
+       .irq_set_type           = gic_set_type,
+       .irq_retrigger          = gic_retrigger,
+ #ifdef CONFIG_SMP
+       .irq_set_affinity       = gic_set_affinity,
+ #endif
+       .irq_set_wake           = gic_set_wake,
+ };
+ void __init gic_cascade_irq(unsigned int gic_nr, unsigned int irq)
+ {
+       if (gic_nr >= MAX_GIC_NR)
+               BUG();
+       if (irq_set_handler_data(irq, &gic_data[gic_nr]) != 0)
+               BUG();
+       irq_set_chained_handler(irq, gic_handle_cascade_irq);
+ }
++static u8 gic_get_cpumask(struct gic_chip_data *gic)
++{
++      void __iomem *base = gic_data_dist_base(gic);
++      u32 mask, i;
++
++      for (i = mask = 0; i < 32; i += 4) {
++              mask = readl_relaxed(base + GIC_DIST_TARGET + i);
++              mask |= mask >> 16;
++              mask |= mask >> 8;
++              if (mask)
++                      break;
++      }
++
++      if (!mask)
++              pr_crit("GIC CPU mask not found - kernel will fail to boot.\n");
++
++      return mask;
++}
++
+ static void __init gic_dist_init(struct gic_chip_data *gic)
+ {
+       unsigned int i;
+       u32 cpumask;
+       unsigned int gic_irqs = gic->gic_irqs;
+       void __iomem *base = gic_data_dist_base(gic);
+       writel_relaxed(0, base + GIC_DIST_CTRL);
+       /*
+        * Set all global interrupts to be level triggered, active low.
+        */
+       for (i = 32; i < gic_irqs; i += 16)
+               writel_relaxed(0, base + GIC_DIST_CONFIG + i * 4 / 16);
+       /*
+        * Set all global interrupts to this CPU only.
+        */
 -      cpu_mask = readl_relaxed(dist_base + GIC_DIST_TARGET + 0);
++      cpumask = gic_get_cpumask(gic);
++      cpumask |= cpumask << 8;
++      cpumask |= cpumask << 16;
+       for (i = 32; i < gic_irqs; i += 4)
+               writel_relaxed(cpumask, base + GIC_DIST_TARGET + i * 4 / 4);
+       /*
+        * Set priority on all global interrupts.
+        */
+       for (i = 32; i < gic_irqs; i += 4)
+               writel_relaxed(0xa0a0a0a0, base + GIC_DIST_PRI + i * 4 / 4);
+       /*
+        * Disable all interrupts.  Leave the PPI and SGIs alone
+        * as these enables are banked registers.
+        */
+       for (i = 32; i < gic_irqs; i += 32)
+               writel_relaxed(0xffffffff, base + GIC_DIST_ENABLE_CLEAR + i * 4 / 32);
+       writel_relaxed(1, base + GIC_DIST_CTRL);
+ }
+ static void __cpuinit gic_cpu_init(struct gic_chip_data *gic)
+ {
+       void __iomem *dist_base = gic_data_dist_base(gic);
+       void __iomem *base = gic_data_cpu_base(gic);
+       unsigned int cpu_mask, cpu = smp_processor_id();
+       int i;
+       /*
+        * Get what the GIC says our CPU mask is.
+        */
+       BUG_ON(cpu >= NR_GIC_CPU_IF);
++      cpu_mask = gic_get_cpumask(gic);
+       gic_cpu_map[cpu] = cpu_mask;
+       /*
+        * Clear our mask from the other map entries in case they're
+        * still undefined.
+        */
+       for (i = 0; i < NR_GIC_CPU_IF; i++)
+               if (i != cpu)
+                       gic_cpu_map[i] &= ~cpu_mask;
+       /*
+        * Deal with the banked PPI and SGI interrupts - disable all
+        * PPI interrupts, ensure all SGI interrupts are enabled.
+        */
+       writel_relaxed(0xffff0000, dist_base + GIC_DIST_ENABLE_CLEAR);
+       writel_relaxed(0x0000ffff, dist_base + GIC_DIST_ENABLE_SET);
+       /*
+        * Set priority on PPI and SGI interrupts
+        */
+       for (i = 0; i < 32; i += 4)
+               writel_relaxed(0xa0a0a0a0, dist_base + GIC_DIST_PRI + i * 4 / 4);
+       writel_relaxed(0xf0, base + GIC_CPU_PRIMASK);
+       writel_relaxed(1, base + GIC_CPU_CTRL);
+ }
+ #ifdef CONFIG_CPU_PM
+ /*
+  * Saves the GIC distributor registers during suspend or idle.  Must be called
+  * with interrupts disabled but before powering down the GIC.  After calling
+  * this function, no interrupts will be delivered by the GIC, and another
+  * platform-specific wakeup source must be enabled.
+  */
+ static void gic_dist_save(unsigned int gic_nr)
+ {
+       unsigned int gic_irqs;
+       void __iomem *dist_base;
+       int i;
+       if (gic_nr >= MAX_GIC_NR)
+               BUG();
+       gic_irqs = gic_data[gic_nr].gic_irqs;
+       dist_base = gic_data_dist_base(&gic_data[gic_nr]);
+       if (!dist_base)
+               return;
+       for (i = 0; i < DIV_ROUND_UP(gic_irqs, 16); i++)
+               gic_data[gic_nr].saved_spi_conf[i] =
+                       readl_relaxed(dist_base + GIC_DIST_CONFIG + i * 4);
+       for (i = 0; i < DIV_ROUND_UP(gic_irqs, 4); i++)
+               gic_data[gic_nr].saved_spi_target[i] =
+                       readl_relaxed(dist_base + GIC_DIST_TARGET + i * 4);
+       for (i = 0; i < DIV_ROUND_UP(gic_irqs, 32); i++)
+               gic_data[gic_nr].saved_spi_enable[i] =
+                       readl_relaxed(dist_base + GIC_DIST_ENABLE_SET + i * 4);
+ }
+ /*
+  * Restores the GIC distributor registers during resume or when coming out of
+  * idle.  Must be called before enabling interrupts.  If a level interrupt
+  * that occured while the GIC was suspended is still present, it will be
+  * handled normally, but any edge interrupts that occured will not be seen by
+  * the GIC and need to be handled by the platform-specific wakeup source.
+  */
+ static void gic_dist_restore(unsigned int gic_nr)
+ {
+       unsigned int gic_irqs;
+       unsigned int i;
+       void __iomem *dist_base;
+       if (gic_nr >= MAX_GIC_NR)
+               BUG();
+       gic_irqs = gic_data[gic_nr].gic_irqs;
+       dist_base = gic_data_dist_base(&gic_data[gic_nr]);
+       if (!dist_base)
+               return;
+       writel_relaxed(0, dist_base + GIC_DIST_CTRL);
+       for (i = 0; i < DIV_ROUND_UP(gic_irqs, 16); i++)
+               writel_relaxed(gic_data[gic_nr].saved_spi_conf[i],
+                       dist_base + GIC_DIST_CONFIG + i * 4);
+       for (i = 0; i < DIV_ROUND_UP(gic_irqs, 4); i++)
+               writel_relaxed(0xa0a0a0a0,
+                       dist_base + GIC_DIST_PRI + i * 4);
+       for (i = 0; i < DIV_ROUND_UP(gic_irqs, 4); i++)
+               writel_relaxed(gic_data[gic_nr].saved_spi_target[i],
+                       dist_base + GIC_DIST_TARGET + i * 4);
+       for (i = 0; i < DIV_ROUND_UP(gic_irqs, 32); i++)
+               writel_relaxed(gic_data[gic_nr].saved_spi_enable[i],
+                       dist_base + GIC_DIST_ENABLE_SET + i * 4);
+       writel_relaxed(1, dist_base + GIC_DIST_CTRL);
+ }
+ static void gic_cpu_save(unsigned int gic_nr)
+ {
+       int i;
+       u32 *ptr;
+       void __iomem *dist_base;
+       void __iomem *cpu_base;
+       if (gic_nr >= MAX_GIC_NR)
+               BUG();
+       dist_base = gic_data_dist_base(&gic_data[gic_nr]);
+       cpu_base = gic_data_cpu_base(&gic_data[gic_nr]);
+       if (!dist_base || !cpu_base)
+               return;
+       ptr = __this_cpu_ptr(gic_data[gic_nr].saved_ppi_enable);
+       for (i = 0; i < DIV_ROUND_UP(32, 32); i++)
+               ptr[i] = readl_relaxed(dist_base + GIC_DIST_ENABLE_SET + i * 4);
+       ptr = __this_cpu_ptr(gic_data[gic_nr].saved_ppi_conf);
+       for (i = 0; i < DIV_ROUND_UP(32, 16); i++)
+               ptr[i] = readl_relaxed(dist_base + GIC_DIST_CONFIG + i * 4);
+ }
+ static void gic_cpu_restore(unsigned int gic_nr)
+ {
+       int i;
+       u32 *ptr;
+       void __iomem *dist_base;
+       void __iomem *cpu_base;
+       if (gic_nr >= MAX_GIC_NR)
+               BUG();
+       dist_base = gic_data_dist_base(&gic_data[gic_nr]);
+       cpu_base = gic_data_cpu_base(&gic_data[gic_nr]);
+       if (!dist_base || !cpu_base)
+               return;
+       ptr = __this_cpu_ptr(gic_data[gic_nr].saved_ppi_enable);
+       for (i = 0; i < DIV_ROUND_UP(32, 32); i++)
+               writel_relaxed(ptr[i], dist_base + GIC_DIST_ENABLE_SET + i * 4);
+       ptr = __this_cpu_ptr(gic_data[gic_nr].saved_ppi_conf);
+       for (i = 0; i < DIV_ROUND_UP(32, 16); i++)
+               writel_relaxed(ptr[i], dist_base + GIC_DIST_CONFIG + i * 4);
+       for (i = 0; i < DIV_ROUND_UP(32, 4); i++)
+               writel_relaxed(0xa0a0a0a0, dist_base + GIC_DIST_PRI + i * 4);
+       writel_relaxed(0xf0, cpu_base + GIC_CPU_PRIMASK);
+       writel_relaxed(1, cpu_base + GIC_CPU_CTRL);
+ }
+ static int gic_notifier(struct notifier_block *self, unsigned long cmd,       void *v)
+ {
+       int i;
+       for (i = 0; i < MAX_GIC_NR; i++) {
+ #ifdef CONFIG_GIC_NON_BANKED
+               /* Skip over unused GICs */
+               if (!gic_data[i].get_base)
+                       continue;
+ #endif
+               switch (cmd) {
+               case CPU_PM_ENTER:
+                       gic_cpu_save(i);
+                       break;
+               case CPU_PM_ENTER_FAILED:
+               case CPU_PM_EXIT:
+                       gic_cpu_restore(i);
+                       break;
+               case CPU_CLUSTER_PM_ENTER:
+                       gic_dist_save(i);
+                       break;
+               case CPU_CLUSTER_PM_ENTER_FAILED:
+               case CPU_CLUSTER_PM_EXIT:
+                       gic_dist_restore(i);
+                       break;
+               }
+       }
+       return NOTIFY_OK;
+ }
+ static struct notifier_block gic_notifier_block = {
+       .notifier_call = gic_notifier,
+ };
+ static void __init gic_pm_init(struct gic_chip_data *gic)
+ {
+       gic->saved_ppi_enable = __alloc_percpu(DIV_ROUND_UP(32, 32) * 4,
+               sizeof(u32));
+       BUG_ON(!gic->saved_ppi_enable);
+       gic->saved_ppi_conf = __alloc_percpu(DIV_ROUND_UP(32, 16) * 4,
+               sizeof(u32));
+       BUG_ON(!gic->saved_ppi_conf);
+       if (gic == &gic_data[0])
+               cpu_pm_register_notifier(&gic_notifier_block);
+ }
+ #else
+ static void __init gic_pm_init(struct gic_chip_data *gic)
+ {
+ }
+ #endif
+ #ifdef CONFIG_SMP
+ void gic_raise_softirq(const struct cpumask *mask, unsigned int irq)
+ {
+       int cpu;
+       unsigned long map = 0;
+       /* Convert our logical CPU mask into a physical one. */
+       for_each_cpu(cpu, mask)
+               map |= 1 << cpu_logical_map(cpu);
+       /*
+        * Ensure that stores to Normal memory are visible to the
+        * other CPUs before issuing the IPI.
+        */
+       dsb();
+       /* this always happens on GIC0 */
+       writel_relaxed(map << 16 | irq, gic_data_dist_base(&gic_data[0]) + GIC_DIST_SOFTINT);
+ }
+ #endif
+ static int gic_irq_domain_map(struct irq_domain *d, unsigned int irq,
+                               irq_hw_number_t hw)
+ {
+       if (hw < 32) {
+               irq_set_percpu_devid(irq);
+               irq_set_chip_and_handler(irq, &gic_chip,
+                                        handle_percpu_devid_irq);
+               set_irq_flags(irq, IRQF_VALID | IRQF_NOAUTOEN);
+       } else {
+               irq_set_chip_and_handler(irq, &gic_chip,
+                                        handle_fasteoi_irq);
+               set_irq_flags(irq, IRQF_VALID | IRQF_PROBE);
+       }
+       irq_set_chip_data(irq, d->host_data);
+       return 0;
+ }
+ static int gic_irq_domain_xlate(struct irq_domain *d,
+                               struct device_node *controller,
+                               const u32 *intspec, unsigned int intsize,
+                               unsigned long *out_hwirq, unsigned int *out_type)
+ {
+       if (d->of_node != controller)
+               return -EINVAL;
+       if (intsize < 3)
+               return -EINVAL;
+       /* Get the interrupt number and add 16 to skip over SGIs */
+       *out_hwirq = intspec[1] + 16;
+       /* For SPIs, we need to add 16 more to get the GIC irq ID number */
+       if (!intspec[0])
+               *out_hwirq += 16;
+       *out_type = intspec[2] & IRQ_TYPE_SENSE_MASK;
+       return 0;
+ }
+ const struct irq_domain_ops gic_irq_domain_ops = {
+       .map = gic_irq_domain_map,
+       .xlate = gic_irq_domain_xlate,
+ };
+ void __init gic_init_bases(unsigned int gic_nr, int irq_start,
+                          void __iomem *dist_base, void __iomem *cpu_base,
+                          u32 percpu_offset, struct device_node *node)
+ {
+       irq_hw_number_t hwirq_base;
+       struct gic_chip_data *gic;
+       int gic_irqs, irq_base, i;
+       BUG_ON(gic_nr >= MAX_GIC_NR);
+       gic = &gic_data[gic_nr];
+ #ifdef CONFIG_GIC_NON_BANKED
+       if (percpu_offset) { /* Frankein-GIC without banked registers... */
+               unsigned int cpu;
+               gic->dist_base.percpu_base = alloc_percpu(void __iomem *);
+               gic->cpu_base.percpu_base = alloc_percpu(void __iomem *);
+               if (WARN_ON(!gic->dist_base.percpu_base ||
+                           !gic->cpu_base.percpu_base)) {
+                       free_percpu(gic->dist_base.percpu_base);
+                       free_percpu(gic->cpu_base.percpu_base);
+                       return;
+               }
+               for_each_possible_cpu(cpu) {
+                       unsigned long offset = percpu_offset * cpu_logical_map(cpu);
+                       *per_cpu_ptr(gic->dist_base.percpu_base, cpu) = dist_base + offset;
+                       *per_cpu_ptr(gic->cpu_base.percpu_base, cpu) = cpu_base + offset;
+               }
+               gic_set_base_accessor(gic, gic_get_percpu_base);
+       } else
+ #endif
+       {                       /* Normal, sane GIC... */
+               WARN(percpu_offset,
+                    "GIC_NON_BANKED not enabled, ignoring %08x offset!",
+                    percpu_offset);
+               gic->dist_base.common_base = dist_base;
+               gic->cpu_base.common_base = cpu_base;
+               gic_set_base_accessor(gic, gic_get_common_base);
+       }
+       /*
+        * Initialize the CPU interface map to all CPUs.
+        * It will be refined as each CPU probes its ID.
+        */
+       for (i = 0; i < NR_GIC_CPU_IF; i++)
+               gic_cpu_map[i] = 0xff;
+       /*
+        * For primary GICs, skip over SGIs.
+        * For secondary GICs, skip over PPIs, too.
+        */
+       if (gic_nr == 0 && (irq_start & 31) > 0) {
+               hwirq_base = 16;
+               if (irq_start != -1)
+                       irq_start = (irq_start & ~31) + 16;
+       } else {
+               hwirq_base = 32;
+       }
+       /*
+        * Find out how many interrupts are supported.
+        * The GIC only supports up to 1020 interrupt sources.
+        */
+       gic_irqs = readl_relaxed(gic_data_dist_base(gic) + GIC_DIST_CTR) & 0x1f;
+       gic_irqs = (gic_irqs + 1) * 32;
+       if (gic_irqs > 1020)
+               gic_irqs = 1020;
+       gic->gic_irqs = gic_irqs;
+       gic_irqs -= hwirq_base; /* calculate # of irqs to allocate */
+       irq_base = irq_alloc_descs(irq_start, 16, gic_irqs, numa_node_id());
+       if (IS_ERR_VALUE(irq_base)) {
+               WARN(1, "Cannot allocate irq_descs @ IRQ%d, assuming pre-allocated\n",
+                    irq_start);
+               irq_base = irq_start;
+       }
+       gic->domain = irq_domain_add_legacy(node, gic_irqs, irq_base,
+                                   hwirq_base, &gic_irq_domain_ops, gic);
+       if (WARN_ON(!gic->domain))
+               return;
+ #ifdef CONFIG_SMP
+       set_smp_cross_call(gic_raise_softirq);
+ #endif
+       set_handle_irq(gic_handle_irq);
+       gic_chip.flags |= gic_arch_extn.flags;
+       gic_dist_init(gic);
+       gic_cpu_init(gic);
+       gic_pm_init(gic);
+ }
+ void __cpuinit gic_secondary_init(unsigned int gic_nr)
+ {
+       BUG_ON(gic_nr >= MAX_GIC_NR);
+       gic_cpu_init(&gic_data[gic_nr]);
+ }
+ #ifdef CONFIG_OF
+ static int gic_cnt __initdata = 0;
+ int __init gic_of_init(struct device_node *node, struct device_node *parent)
+ {
+       void __iomem *cpu_base;
+       void __iomem *dist_base;
+       u32 percpu_offset;
+       int irq;
+       if (WARN_ON(!node))
+               return -ENODEV;
+       dist_base = of_iomap(node, 0);
+       WARN(!dist_base, "unable to map gic dist registers\n");
+       cpu_base = of_iomap(node, 1);
+       WARN(!cpu_base, "unable to map gic cpu registers\n");
+       if (of_property_read_u32(node, "cpu-offset", &percpu_offset))
+               percpu_offset = 0;
+       gic_init_bases(gic_cnt, -1, dist_base, cpu_base, percpu_offset, node);
+       if (parent) {
+               irq = irq_of_parse_and_map(node, 0);
+               gic_cascade_irq(gic_cnt, irq);
+       }
+       gic_cnt++;
+       return 0;
+ }
+ IRQCHIP_DECLARE(cortex_a15_gic, "arm,cortex-a15-gic", gic_of_init);
+ IRQCHIP_DECLARE(cortex_a9_gic, "arm,cortex-a9-gic", gic_of_init);
+ IRQCHIP_DECLARE(msm_8660_qgic, "qcom,msm-8660-qgic", gic_of_init);
+ IRQCHIP_DECLARE(msm_qgic2, "qcom,msm-qgic2", gic_of_init);
+ #endif
Simple merge