[PATCH] x86-64: typo in __assign_irq_vector when updating pos for vector and offset
[sfrench/cifs-2.6.git] / arch / x86_64 / kernel / io_apic.c
index 97e90f6abcc45bc3e313f69b5877b7f4f704eef8..b848f48085100075cc275070a5dbc233139de366 100644 (file)
 #include <linux/delay.h>
 #include <linux/sched.h>
 #include <linux/smp_lock.h>
+#include <linux/pci.h>
 #include <linux/mc146818rtc.h>
 #include <linux/acpi.h>
 #include <linux/sysdev.h>
+#include <linux/msi.h>
+#include <linux/htirq.h>
 #ifdef CONFIG_ACPI
 #include <acpi/acpi_bus.h>
 #endif
 #include <asm/acpi.h>
 #include <asm/dma.h>
 #include <asm/nmi.h>
+#include <asm/msidef.h>
+#include <asm/hypertransport.h>
+
+static int assign_irq_vector(int irq, cpumask_t mask, cpumask_t *result);
 
 #define __apicdebuginit  __init
 
@@ -48,7 +55,7 @@ int sis_apic_bug; /* not actually supported, dummy for compile */
 
 static int no_timer_check;
 
-int disable_timer_pin_1 __initdata;
+static int disable_timer_pin_1 __initdata;
 
 int timer_over_8254 __initdata = 0;
 
@@ -81,14 +88,6 @@ static struct irq_pin_list {
        short apic, pin, next;
 } irq_2_pin[PIN_MAP_SIZE];
 
-int vector_irq[NR_VECTORS] __read_mostly = { [0 ... NR_VECTORS - 1] = -1};
-#ifdef CONFIG_PCI_MSI
-#define vector_to_irq(vector)  \
-       (platform_legacy_irq(vector) ? vector : vector_irq[vector])
-#else
-#define vector_to_irq(vector)  (vector)
-#endif
-
 #define __DO_ACTION(R, ACTION, FINAL)                                  \
                                                                        \
 {                                                                      \
@@ -111,12 +110,63 @@ int vector_irq[NR_VECTORS] __read_mostly = { [0 ... NR_VECTORS - 1] = -1};
        FINAL;                                                          \
 }
 
+union entry_union {
+       struct { u32 w1, w2; };
+       struct IO_APIC_route_entry entry;
+};
+
+static struct IO_APIC_route_entry ioapic_read_entry(int apic, int pin)
+{
+       union entry_union eu;
+       unsigned long flags;
+       spin_lock_irqsave(&ioapic_lock, flags);
+       eu.w1 = io_apic_read(apic, 0x10 + 2 * pin);
+       eu.w2 = io_apic_read(apic, 0x11 + 2 * pin);
+       spin_unlock_irqrestore(&ioapic_lock, flags);
+       return eu.entry;
+}
+
+static void ioapic_write_entry(int apic, int pin, struct IO_APIC_route_entry e)
+{
+       unsigned long flags;
+       union entry_union eu;
+       eu.entry = e;
+       spin_lock_irqsave(&ioapic_lock, flags);
+       io_apic_write(apic, 0x10 + 2*pin, eu.w1);
+       io_apic_write(apic, 0x11 + 2*pin, eu.w2);
+       spin_unlock_irqrestore(&ioapic_lock, flags);
+}
+
 #ifdef CONFIG_SMP
+static void __target_IO_APIC_irq(unsigned int irq, unsigned int dest, u8 vector)
+{
+       int apic, pin;
+       struct irq_pin_list *entry = irq_2_pin + irq;
+
+       BUG_ON(irq >= NR_IRQS);
+       for (;;) {
+               unsigned int reg;
+               apic = entry->apic;
+               pin = entry->pin;
+               if (pin == -1)
+                       break;
+               io_apic_write(apic, 0x11 + pin*2, dest);
+               reg = io_apic_read(apic, 0x10 + pin*2);
+               reg &= ~0x000000ff;
+               reg |= vector;
+               io_apic_modify(apic, reg);
+               if (!entry->next)
+                       break;
+               entry = irq_2_pin + entry->next;
+       }
+}
+
 static void set_ioapic_affinity_irq(unsigned int irq, cpumask_t mask)
 {
        unsigned long flags;
        unsigned int dest;
        cpumask_t tmp;
+       int vector;
 
        cpus_and(tmp, mask, cpu_online_map);
        if (cpus_empty(tmp))
@@ -124,7 +174,11 @@ static void set_ioapic_affinity_irq(unsigned int irq, cpumask_t mask)
 
        cpus_and(mask, tmp, CPU_MASK_ALL);
 
-       dest = cpu_mask_to_apicid(mask);
+       vector = assign_irq_vector(irq, mask, &tmp);
+       if (vector < 0)
+               return;
+
+       dest = cpu_mask_to_apicid(tmp);
 
        /*
         * Only the high 8 bits are valid.
@@ -132,14 +186,12 @@ static void set_ioapic_affinity_irq(unsigned int irq, cpumask_t mask)
        dest = SET_APIC_LOGICAL_ID(dest);
 
        spin_lock_irqsave(&ioapic_lock, flags);
-       __DO_ACTION(1, = dest, )
-       set_irq_info(irq, mask);
+       __target_IO_APIC_irq(irq, dest, vector);
+       set_native_irq_info(irq, mask);
        spin_unlock_irqrestore(&ioapic_lock, flags);
 }
 #endif
 
-static u8 gsi_2_irq[NR_IRQ_VECTORS] = { [0 ... NR_IRQ_VECTORS-1] = 0xFF };
-
 /*
  * The common case is 1:1 IRQ<->pin mappings. Sometimes there are
  * shared ISA-space IRQs, so we have to support them. We are super
@@ -196,13 +248,9 @@ static void unmask_IO_APIC_irq (unsigned int irq)
 static void clear_IO_APIC_pin(unsigned int apic, unsigned int pin)
 {
        struct IO_APIC_route_entry entry;
-       unsigned long flags;
 
        /* Check delivery_mode to be sure we're not clearing an SMI pin */
-       spin_lock_irqsave(&ioapic_lock, flags);
-       *(((int*)&entry) + 0) = io_apic_read(apic, 0x10 + 2 * pin);
-       *(((int*)&entry) + 1) = io_apic_read(apic, 0x11 + 2 * pin);
-       spin_unlock_irqrestore(&ioapic_lock, flags);
+       entry = ioapic_read_entry(apic, pin);
        if (entry.delivery_mode == dest_SMI)
                return;
        /*
@@ -210,10 +258,7 @@ static void clear_IO_APIC_pin(unsigned int apic, unsigned int pin)
         */
        memset(&entry, 0, sizeof(entry));
        entry.mask = 1;
-       spin_lock_irqsave(&ioapic_lock, flags);
-       io_apic_write(apic, 0x10 + 2 * pin, *(((int *)&entry) + 0));
-       io_apic_write(apic, 0x11 + 2 * pin, *(((int *)&entry) + 1));
-       spin_unlock_irqrestore(&ioapic_lock, flags);
+       ioapic_write_entry(apic, pin, entry);
 }
 
 static void clear_IO_APIC (void)
@@ -233,18 +278,17 @@ int ioapic_force;
 static int __init disable_ioapic_setup(char *str)
 {
        skip_ioapic_setup = 1;
-       return 1;
+       return 0;
 }
+early_param("noapic", disable_ioapic_setup);
 
-static int __init enable_ioapic_setup(char *str)
+/* Actually the next is obsolete, but keep it for paranoid reasons -AK */
+static int __init disable_timer_pin_setup(char *arg)
 {
-       ioapic_force = 1;
-       skip_ioapic_setup = 0;
+       disable_timer_pin_1 = 1;
        return 1;
 }
-
-__setup("noapic", disable_ioapic_setup);
-__setup("apic", enable_ioapic_setup);
+__setup("disable_timer_pin_1", disable_timer_pin_setup);
 
 static int __init setup_disable_8254_timer(char *s)
 {
@@ -260,107 +304,6 @@ static int __init setup_enable_8254_timer(char *s)
 __setup("disable_8254_timer", setup_disable_8254_timer);
 __setup("enable_8254_timer", setup_enable_8254_timer);
 
-#include <asm/pci-direct.h>
-#include <linux/pci_ids.h>
-#include <linux/pci.h>
-
-
-#ifdef CONFIG_ACPI
-
-static int nvidia_hpet_detected __initdata;
-
-static int __init nvidia_hpet_check(unsigned long phys, unsigned long size)
-{
-       nvidia_hpet_detected = 1;
-       return 0;
-}
-#endif
-
-/* Temporary Hack. Nvidia and VIA boards currently only work with IO-APIC
-   off. Check for an Nvidia or VIA PCI bridge and turn it off.
-   Use pci direct infrastructure because this runs before the PCI subsystem. 
-
-   Can be overwritten with "apic"
-
-   And another hack to disable the IOMMU on VIA chipsets.
-
-   ... and others. Really should move this somewhere else.
-
-   Kludge-O-Rama. */
-void __init check_ioapic(void) 
-{ 
-       int num,slot,func; 
-       /* Poor man's PCI discovery */
-       for (num = 0; num < 32; num++) { 
-               for (slot = 0; slot < 32; slot++) { 
-                       for (func = 0; func < 8; func++) { 
-                               u32 class;
-                               u32 vendor;
-                               u8 type;
-                               class = read_pci_config(num,slot,func,
-                                                       PCI_CLASS_REVISION);
-                               if (class == 0xffffffff)
-                                       break; 
-
-                               if ((class >> 16) != PCI_CLASS_BRIDGE_PCI)
-                                       continue; 
-
-                               vendor = read_pci_config(num, slot, func, 
-                                                        PCI_VENDOR_ID);
-                               vendor &= 0xffff;
-                               switch (vendor) { 
-                               case PCI_VENDOR_ID_VIA:
-#ifdef CONFIG_IOMMU
-                                       if ((end_pfn > MAX_DMA32_PFN ||
-                                            force_iommu) &&
-                                           !iommu_aperture_allowed) {
-                                               printk(KERN_INFO
-    "Looks like a VIA chipset. Disabling IOMMU. Override with \"iommu=allowed\"\n");
-                                               iommu_aperture_disabled = 1;
-                                       }
-#endif
-                                       return;
-                               case PCI_VENDOR_ID_NVIDIA:
-#ifdef CONFIG_ACPI
-                                       /*
-                                        * All timer overrides on Nvidia are
-                                        * wrong unless HPET is enabled.
-                                        */
-                                       nvidia_hpet_detected = 0;
-                                       acpi_table_parse(ACPI_HPET,
-                                                       nvidia_hpet_check);
-                                       if (nvidia_hpet_detected == 0) {
-                                               acpi_skip_timer_override = 1;
-                                               printk(KERN_INFO "Nvidia board "
-                                                   "detected. Ignoring ACPI "
-                                                   "timer override.\n");
-                                       }
-#endif
-                                       /* RED-PEN skip them on mptables too? */
-                                       return;
-
-                               /* This should be actually default, but
-                                  for 2.6.16 let's do it for ATI only where
-                                  it's really needed. */
-                               case PCI_VENDOR_ID_ATI:
-                                       if (timer_over_8254 == 1) {     
-                                               timer_over_8254 = 0;    
-                                       printk(KERN_INFO
-               "ATI board detected. Disabling timer routing over 8254.\n");
-                                       }       
-                                       return;
-                               } 
-
-
-                               /* No multi-function device? */
-                               type = read_pci_config_byte(num,slot,func,
-                                                           PCI_HEADER_TYPE);
-                               if (!(type & 0x80))
-                                       break;
-                       } 
-               }
-       }
-} 
 
 /*
  * Find the IRQ entry number of a certain pin.
@@ -389,7 +332,7 @@ static int __init find_isa_irq_pin(int irq, int type)
        for (i = 0; i < mp_irq_entries; i++) {
                int lbus = mp_irqs[i].mpc_srcbus;
 
-               if (mp_bus_id_to_type[lbus] == MP_BUS_ISA &&
+               if (test_bit(lbus, mp_bus_not_pci) &&
                    (mp_irqs[i].mpc_irqtype == type) &&
                    (mp_irqs[i].mpc_srcbusirq == irq))
 
@@ -405,7 +348,7 @@ static int __init find_isa_irq_apic(int irq, int type)
        for (i = 0; i < mp_irq_entries; i++) {
                int lbus = mp_irqs[i].mpc_srcbus;
 
-               if ((mp_bus_id_to_type[lbus] == MP_BUS_ISA) &&
+               if (test_bit(lbus, mp_bus_not_pci) &&
                    (mp_irqs[i].mpc_irqtype == type) &&
                    (mp_irqs[i].mpc_srcbusirq == irq))
                        break;
@@ -445,7 +388,7 @@ int IO_APIC_get_PCI_irq_vector(int bus, int slot, int pin)
                            mp_irqs[i].mpc_dstapic == MP_APIC_ALL)
                                break;
 
-               if ((mp_bus_id_to_type[lbus] == MP_BUS_PCI) &&
+               if (!test_bit(lbus, mp_bus_not_pci) &&
                    !mp_irqs[i].mpc_irqtype &&
                    (bus == lbus) &&
                    (slot == ((mp_irqs[i].mpc_srcbusirq >> 2) & 0x1f))) {
@@ -491,28 +434,11 @@ static int __init MPBIOS_polarity(int idx)
        switch (mp_irqs[idx].mpc_irqflag & 3)
        {
                case 0: /* conforms, ie. bus-type dependent polarity */
-               {
-                       switch (mp_bus_id_to_type[bus])
-                       {
-                               case MP_BUS_ISA: /* ISA pin */
-                               {
-                                       polarity = default_ISA_polarity(idx);
-                                       break;
-                               }
-                               case MP_BUS_PCI: /* PCI pin */
-                               {
-                                       polarity = default_PCI_polarity(idx);
-                                       break;
-                               }
-                               default:
-                               {
-                                       printk(KERN_WARNING "broken BIOS!!\n");
-                                       polarity = 1;
-                                       break;
-                               }
-                       }
+                       if (test_bit(bus, mp_bus_not_pci))
+                               polarity = default_ISA_polarity(idx);
+                       else
+                               polarity = default_PCI_polarity(idx);
                        break;
-               }
                case 1: /* high active */
                {
                        polarity = 0;
@@ -550,28 +476,11 @@ static int MPBIOS_trigger(int idx)
        switch ((mp_irqs[idx].mpc_irqflag>>2) & 3)
        {
                case 0: /* conforms, ie. bus-type dependent */
-               {
-                       switch (mp_bus_id_to_type[bus])
-                       {
-                               case MP_BUS_ISA: /* ISA pin */
-                               {
-                                       trigger = default_ISA_trigger(idx);
-                                       break;
-                               }
-                               case MP_BUS_PCI: /* PCI pin */
-                               {
-                                       trigger = default_PCI_trigger(idx);
-                                       break;
-                               }
-                               default:
-                               {
-                                       printk(KERN_WARNING "broken BIOS!!\n");
-                                       trigger = 1;
-                                       break;
-                               }
-                       }
+                       if (test_bit(bus, mp_bus_not_pci))
+                               trigger = default_ISA_trigger(idx);
+                       else
+                               trigger = default_PCI_trigger(idx);
                        break;
-               }
                case 1: /* edge */
                {
                        trigger = 0;
@@ -608,64 +517,6 @@ static inline int irq_trigger(int idx)
        return MPBIOS_trigger(idx);
 }
 
-static int next_irq = 16;
-
-/*
- * gsi_irq_sharing -- Name overload!  "irq" can be either a legacy IRQ
- * in the range 0-15, a linux IRQ in the range 0-223, or a GSI number
- * from ACPI, which can reach 800 in large boxen.
- *
- * Compact the sparse GSI space into a sequential IRQ series and reuse
- * vectors if possible.
- */
-int gsi_irq_sharing(int gsi)
-{
-       int i, tries, vector;
-
-       BUG_ON(gsi >= NR_IRQ_VECTORS);
-
-       if (platform_legacy_irq(gsi))
-               return gsi;
-
-       if (gsi_2_irq[gsi] != 0xFF)
-               return (int)gsi_2_irq[gsi];
-
-       tries = NR_IRQS;
-  try_again:
-       vector = assign_irq_vector(gsi);
-
-       /*
-        * Sharing vectors means sharing IRQs, so scan irq_vectors for previous
-        * use of vector and if found, return that IRQ.  However, we never want
-        * to share legacy IRQs, which usually have a different trigger mode
-        * than PCI.
-        */
-       for (i = 0; i < NR_IRQS; i++)
-               if (IO_APIC_VECTOR(i) == vector)
-                       break;
-       if (platform_legacy_irq(i)) {
-               if (--tries >= 0) {
-                       IO_APIC_VECTOR(i) = 0;
-                       goto try_again;
-               }
-               panic("gsi_irq_sharing: didn't find an IRQ using vector 0x%02X for GSI %d", vector, gsi);
-       }
-       if (i < NR_IRQS) {
-               gsi_2_irq[gsi] = i;
-               printk(KERN_INFO "GSI %d sharing vector 0x%02X and IRQ %d\n",
-                               gsi, vector, i);
-               return i;
-       }
-
-       i = next_irq++;
-       BUG_ON(i >= NR_IRQS);
-       gsi_2_irq[gsi] = i;
-       IO_APIC_VECTOR(i) = vector;
-       printk(KERN_INFO "GSI %d assigned vector 0x%02X and IRQ %d\n",
-                       gsi, vector, i);
-       return i;
-}
-
 static int pin_2_irq(int idx, int apic, int pin)
 {
        int irq, i;
@@ -677,31 +528,16 @@ static int pin_2_irq(int idx, int apic, int pin)
        if (mp_irqs[idx].mpc_dstirq != pin)
                printk(KERN_ERR "broken BIOS or MPTABLE parser, ayiee!!\n");
 
-       switch (mp_bus_id_to_type[bus])
-       {
-               case MP_BUS_ISA: /* ISA pin */
-               {
-                       irq = mp_irqs[idx].mpc_srcbusirq;
-                       break;
-               }
-               case MP_BUS_PCI: /* PCI pin */
-               {
-                       /*
-                        * PCI IRQs are mapped in order
-                        */
-                       i = irq = 0;
-                       while (i < apic)
-                               irq += nr_ioapic_registers[i++];
-                       irq += pin;
-                       irq = gsi_irq_sharing(irq);
-                       break;
-               }
-               default:
-               {
-                       printk(KERN_ERR "unknown bus type %d.\n",bus); 
-                       irq = 0;
-                       break;
-               }
+       if (test_bit(bus, mp_bus_not_pci)) {
+               irq = mp_irqs[idx].mpc_srcbusirq;
+       } else {
+               /*
+                * PCI IRQs are mapped in order
+                */
+               i = irq = 0;
+               while (i < apic)
+                       irq += nr_ioapic_registers[i++];
+               irq += pin;
        }
        BUG_ON(irq >= NR_IRQS);
        return irq;
@@ -725,46 +561,132 @@ static inline int IO_APIC_irq_trigger(int irq)
 }
 
 /* irq_vectors is indexed by the sum of all RTEs in all I/O APICs. */
-u8 irq_vector[NR_IRQ_VECTORS] __read_mostly = { FIRST_DEVICE_VECTOR , 0 };
+static u8 irq_vector[NR_IRQ_VECTORS] __read_mostly = {
+       [0] = FIRST_EXTERNAL_VECTOR + 0,
+       [1] = FIRST_EXTERNAL_VECTOR + 1,
+       [2] = FIRST_EXTERNAL_VECTOR + 2,
+       [3] = FIRST_EXTERNAL_VECTOR + 3,
+       [4] = FIRST_EXTERNAL_VECTOR + 4,
+       [5] = FIRST_EXTERNAL_VECTOR + 5,
+       [6] = FIRST_EXTERNAL_VECTOR + 6,
+       [7] = FIRST_EXTERNAL_VECTOR + 7,
+       [8] = FIRST_EXTERNAL_VECTOR + 8,
+       [9] = FIRST_EXTERNAL_VECTOR + 9,
+       [10] = FIRST_EXTERNAL_VECTOR + 10,
+       [11] = FIRST_EXTERNAL_VECTOR + 11,
+       [12] = FIRST_EXTERNAL_VECTOR + 12,
+       [13] = FIRST_EXTERNAL_VECTOR + 13,
+       [14] = FIRST_EXTERNAL_VECTOR + 14,
+       [15] = FIRST_EXTERNAL_VECTOR + 15,
+};
 
-int assign_irq_vector(int irq)
+static cpumask_t irq_domain[NR_IRQ_VECTORS] __read_mostly = {
+       [0] = CPU_MASK_ALL,
+       [1] = CPU_MASK_ALL,
+       [2] = CPU_MASK_ALL,
+       [3] = CPU_MASK_ALL,
+       [4] = CPU_MASK_ALL,
+       [5] = CPU_MASK_ALL,
+       [6] = CPU_MASK_ALL,
+       [7] = CPU_MASK_ALL,
+       [8] = CPU_MASK_ALL,
+       [9] = CPU_MASK_ALL,
+       [10] = CPU_MASK_ALL,
+       [11] = CPU_MASK_ALL,
+       [12] = CPU_MASK_ALL,
+       [13] = CPU_MASK_ALL,
+       [14] = CPU_MASK_ALL,
+       [15] = CPU_MASK_ALL,
+};
+
+static int __assign_irq_vector(int irq, cpumask_t mask, cpumask_t *result)
 {
-       static int current_vector = FIRST_DEVICE_VECTOR, offset = 0;
-       unsigned long flags;
-       int vector;
+       /*
+        * NOTE! The local APIC isn't very good at handling
+        * multiple interrupts at the same interrupt level.
+        * As the interrupt level is determined by taking the
+        * vector number and shifting that right by 4, we
+        * want to spread these out a bit so that they don't
+        * all fall in the same interrupt level.
+        *
+        * Also, we've got to be careful not to trash gate
+        * 0x80, because int 0x80 is hm, kind of importantish. ;)
+        */
+       static struct {
+               int vector;
+               int offset;
+       } pos[NR_CPUS] = { [ 0 ... NR_CPUS - 1] = {FIRST_DEVICE_VECTOR, 0} };
+       int old_vector = -1;
+       int cpu;
+
+       BUG_ON((unsigned)irq >= NR_IRQ_VECTORS);
+
+       if (irq_vector[irq] > 0)
+               old_vector = irq_vector[irq];
+       if (old_vector > 0) {
+               cpus_and(*result, irq_domain[irq], mask);
+               if (!cpus_empty(*result))
+                       return old_vector;
+       }
 
-       BUG_ON(irq != AUTO_ASSIGN && (unsigned)irq >= NR_IRQ_VECTORS);
+       for_each_cpu_mask(cpu, mask) {
+               cpumask_t domain;
+               int first, new_cpu;
+               int vector, offset;
 
-       spin_lock_irqsave(&vector_lock, flags);
+               domain = vector_allocation_domain(cpu);
+               first = first_cpu(domain);
 
-       if (irq != AUTO_ASSIGN && IO_APIC_VECTOR(irq) > 0) {
-               spin_unlock_irqrestore(&vector_lock, flags);
-               return IO_APIC_VECTOR(irq);
-       }
+               vector = pos[first].vector;
+               offset = pos[first].offset;
 next:
-       current_vector += 8;
-       if (current_vector == IA32_SYSCALL_VECTOR)
-               goto next;
-
-       if (current_vector >= FIRST_SYSTEM_VECTOR) {
-               /* If we run out of vectors on large boxen, must share them. */
-               offset = (offset + 1) % 8;
-               current_vector = FIRST_DEVICE_VECTOR + offset;
+               vector += 8;
+               if (vector >= FIRST_SYSTEM_VECTOR) {
+                       /* If we run out of vectors on large boxen, must share them. */
+                       offset = (offset + 1) % 8;
+                       vector = FIRST_DEVICE_VECTOR + offset;
+               }
+               if (unlikely(pos[first].vector == vector))
+                       continue;
+               if (vector == IA32_SYSCALL_VECTOR)
+                       goto next;
+               for_each_cpu_mask(new_cpu, domain)
+                       if (per_cpu(vector_irq, new_cpu)[vector] != -1)
+                               goto next;
+               /* Found one! */
+               for_each_cpu_mask(new_cpu, domain) {
+                       pos[new_cpu].vector = vector;
+                       pos[new_cpu].offset = offset;
+               }
+               if (old_vector >= 0) {
+                       int old_cpu;
+                       for_each_cpu_mask(old_cpu, irq_domain[irq])
+                               per_cpu(vector_irq, old_cpu)[old_vector] = -1;
+               }
+               for_each_cpu_mask(new_cpu, domain)
+                       per_cpu(vector_irq, new_cpu)[vector] = irq;
+               irq_vector[irq] = vector;
+               irq_domain[irq] = domain;
+               cpus_and(*result, domain, mask);
+               return vector;
        }
+       return -ENOSPC;
+}
 
-       vector = current_vector;
-       vector_irq[vector] = irq;
-       if (irq != AUTO_ASSIGN)
-               IO_APIC_VECTOR(irq) = vector;
+static int assign_irq_vector(int irq, cpumask_t mask, cpumask_t *result)
+{
+       int vector;
+       unsigned long flags;
 
+       spin_lock_irqsave(&vector_lock, flags);
+       vector = __assign_irq_vector(irq, mask, result);
        spin_unlock_irqrestore(&vector_lock, flags);
-
        return vector;
 }
 
 extern void (*interrupt[NR_IRQS])(void);
-static struct hw_interrupt_type ioapic_level_type;
-static struct hw_interrupt_type ioapic_edge_type;
+
+static struct irq_chip ioapic_chip;
 
 #define IOAPIC_AUTO    -1
 #define IOAPIC_EDGE    0
@@ -772,16 +694,13 @@ static struct hw_interrupt_type ioapic_edge_type;
 
 static void ioapic_register_intr(int irq, int vector, unsigned long trigger)
 {
-       unsigned idx;
-
-       idx = use_pci_vector() && !platform_legacy_irq(irq) ? vector : irq;
-
        if ((trigger == IOAPIC_AUTO && IO_APIC_irq_trigger(irq)) ||
                        trigger == IOAPIC_LEVEL)
-               irq_desc[idx].chip = &ioapic_level_type;
+               set_irq_chip_and_handler_name(irq, &ioapic_chip,
+                                             handle_fasteoi_irq, "fasteoi");
        else
-               irq_desc[idx].chip = &ioapic_edge_type;
-       set_intr_gate(vector, interrupt[idx]);
+               set_irq_chip_and_handler_name(irq, &ioapic_chip,
+                                             handle_edge_irq, "edge");
 }
 
 static void __init setup_IO_APIC_irqs(void)
@@ -831,16 +750,21 @@ static void __init setup_IO_APIC_irqs(void)
                        continue;
 
                if (IO_APIC_IRQ(irq)) {
-                       vector = assign_irq_vector(irq);
+                       cpumask_t mask;
+                       vector = assign_irq_vector(irq, TARGET_CPUS, &mask);
+                       if (vector < 0)
+                               continue;
+
+                       entry.dest.logical.logical_dest = cpu_mask_to_apicid(mask);
                        entry.vector = vector;
 
                        ioapic_register_intr(irq, vector, IOAPIC_AUTO);
                        if (!apic && (irq < 16))
                                disable_8259A_irq(irq);
                }
+               ioapic_write_entry(apic, pin, entry);
+
                spin_lock_irqsave(&ioapic_lock, flags);
-               io_apic_write(apic, 0x11+2*pin, *(((int *)&entry)+1));
-               io_apic_write(apic, 0x10+2*pin, *(((int *)&entry)+0));
                set_native_irq_info(irq, TARGET_CPUS);
                spin_unlock_irqrestore(&ioapic_lock, flags);
        }
@@ -882,7 +806,7 @@ static void __init setup_ExtINT_IRQ0_pin(unsigned int apic, unsigned int pin, in
         * The timer IRQ doesn't have to know that behind the
         * scene we have a 8259A-master in AEOI mode ...
         */
-       irq_desc[0].chip = &ioapic_edge_type;
+       set_irq_chip_and_handler_name(0, &ioapic_chip, handle_edge_irq, "edge");
 
        /*
         * Add it to the IO-APIC irq-routing table:
@@ -978,10 +902,7 @@ void __apicdebuginit print_IO_APIC(void)
        for (i = 0; i <= reg_01.bits.entries; i++) {
                struct IO_APIC_route_entry entry;
 
-               spin_lock_irqsave(&ioapic_lock, flags);
-               *(((int *)&entry)+0) = io_apic_read(apic, 0x10+i*2);
-               *(((int *)&entry)+1) = io_apic_read(apic, 0x11+i*2);
-               spin_unlock_irqrestore(&ioapic_lock, flags);
+               entry = ioapic_read_entry(apic, i);
 
                printk(KERN_DEBUG " %02x %03X %02X  ",
                        i,
@@ -1001,17 +922,12 @@ void __apicdebuginit print_IO_APIC(void)
                );
        }
        }
-       if (use_pci_vector())
-               printk(KERN_INFO "Using vector-based indexing\n");
        printk(KERN_DEBUG "IRQ to pin mappings:\n");
        for (i = 0; i < NR_IRQS; i++) {
                struct irq_pin_list *entry = irq_2_pin + i;
                if (entry->pin < 0)
                        continue;
-               if (use_pci_vector() && !platform_legacy_irq(i))
-                       printk(KERN_DEBUG "IRQ%d ", IO_APIC_VECTOR(i));
-               else
-                       printk(KERN_DEBUG "IRQ%d ", i);
+               printk(KERN_DEBUG "IRQ%d ", i);
                for (;;) {
                        printk("-> %d:%d", entry->apic, entry->pin);
                        if (!entry->next)
@@ -1191,11 +1107,7 @@ static void __init enable_IO_APIC(void)
                /* See if any of the pins is in ExtINT mode */
                for (pin = 0; pin < nr_ioapic_registers[apic]; pin++) {
                        struct IO_APIC_route_entry entry;
-                       spin_lock_irqsave(&ioapic_lock, flags);
-                       *(((int *)&entry) + 0) = io_apic_read(apic, 0x10 + 2 * pin);
-                       *(((int *)&entry) + 1) = io_apic_read(apic, 0x11 + 2 * pin);
-                       spin_unlock_irqrestore(&ioapic_lock, flags);
-
+                       entry = ioapic_read_entry(apic, pin);
 
                        /* If the interrupt line is enabled and in ExtInt mode
                         * I have found the pin where the i8259 is connected.
@@ -1247,7 +1159,6 @@ void disable_IO_APIC(void)
         */
        if (ioapic_i8259.pin != -1) {
                struct IO_APIC_route_entry entry;
-               unsigned long flags;
 
                memset(&entry, 0, sizeof(entry));
                entry.mask            = 0; /* Enabled */
@@ -1264,83 +1175,12 @@ void disable_IO_APIC(void)
                /*
                 * Add it to the IO-APIC irq-routing table:
                 */
-               spin_lock_irqsave(&ioapic_lock, flags);
-               io_apic_write(ioapic_i8259.apic, 0x11+2*ioapic_i8259.pin,
-                       *(((int *)&entry)+1));
-               io_apic_write(ioapic_i8259.apic, 0x10+2*ioapic_i8259.pin,
-                       *(((int *)&entry)+0));
-               spin_unlock_irqrestore(&ioapic_lock, flags);
+               ioapic_write_entry(ioapic_i8259.apic, ioapic_i8259.pin, entry);
        }
 
        disconnect_bsp_APIC(ioapic_i8259.pin != -1);
 }
 
-/*
- * function to set the IO-APIC physical IDs based on the
- * values stored in the MPC table.
- *
- * by Matt Domsch <Matt_Domsch@dell.com>  Tue Dec 21 12:25:05 CST 1999
- */
-
-static void __init setup_ioapic_ids_from_mpc (void)
-{
-       union IO_APIC_reg_00 reg_00;
-       int apic;
-       int i;
-       unsigned char old_id;
-       unsigned long flags;
-
-       /*
-        * Set the IOAPIC ID to the value stored in the MPC table.
-        */
-       for (apic = 0; apic < nr_ioapics; apic++) {
-
-               /* Read the register 0 value */
-               spin_lock_irqsave(&ioapic_lock, flags);
-               reg_00.raw = io_apic_read(apic, 0);
-               spin_unlock_irqrestore(&ioapic_lock, flags);
-               
-               old_id = mp_ioapics[apic].mpc_apicid;
-
-
-               printk(KERN_INFO "Using IO-APIC %d\n", mp_ioapics[apic].mpc_apicid);
-
-
-               /*
-                * We need to adjust the IRQ routing table
-                * if the ID changed.
-                */
-               if (old_id != mp_ioapics[apic].mpc_apicid)
-                       for (i = 0; i < mp_irq_entries; i++)
-                               if (mp_irqs[i].mpc_dstapic == old_id)
-                                       mp_irqs[i].mpc_dstapic
-                                               = mp_ioapics[apic].mpc_apicid;
-
-               /*
-                * Read the right value from the MPC table and
-                * write it into the ID register.
-                */
-               apic_printk(APIC_VERBOSE,KERN_INFO "...changing IO-APIC physical APIC ID to %d ...",
-                               mp_ioapics[apic].mpc_apicid);
-
-               reg_00.bits.ID = mp_ioapics[apic].mpc_apicid;
-               spin_lock_irqsave(&ioapic_lock, flags);
-               io_apic_write(apic, 0, reg_00.raw);
-               spin_unlock_irqrestore(&ioapic_lock, flags);
-
-               /*
-                * Sanity check
-                */
-               spin_lock_irqsave(&ioapic_lock, flags);
-               reg_00.raw = io_apic_read(apic, 0);
-               spin_unlock_irqrestore(&ioapic_lock, flags);
-               if (reg_00.bits.ID != mp_ioapics[apic].mpc_apicid)
-                       printk("could not set ID!\n");
-               else
-                       apic_printk(APIC_VERBOSE," ok.\n");
-       }
-}
-
 /*
  * There is a nasty bug in some older SMP boards, their mptable lies
  * about the timer IRQ. We do the following to work around the situation:
@@ -1394,7 +1234,7 @@ static int __init timer_irq_works(void)
  * an edge even if it isn't on the 8259A...
  */
 
-static unsigned int startup_edge_ioapic_irq(unsigned int irq)
+static unsigned int startup_ioapic_irq(unsigned int irq)
 {
        int was_pending = 0;
        unsigned long flags;
@@ -1411,107 +1251,16 @@ static unsigned int startup_edge_ioapic_irq(unsigned int irq)
        return was_pending;
 }
 
-/*
- * Once we have recorded IRQ_PENDING already, we can mask the
- * interrupt for real. This prevents IRQ storms from unhandled
- * devices.
- */
-static void ack_edge_ioapic_irq(unsigned int irq)
-{
-       move_irq(irq);
-       if ((irq_desc[irq].status & (IRQ_PENDING | IRQ_DISABLED))
-                                       == (IRQ_PENDING | IRQ_DISABLED))
-               mask_IO_APIC_irq(irq);
-       ack_APIC_irq();
-}
-
-/*
- * Level triggered interrupts can just be masked,
- * and shutting down and starting up the interrupt
- * is the same as enabling and disabling them -- except
- * with a startup need to return a "was pending" value.
- *
- * Level triggered interrupts are special because we
- * do not touch any IO-APIC register while handling
- * them. We ack the APIC in the end-IRQ handler, not
- * in the start-IRQ-handler. Protection against reentrance
- * from the same interrupt is still provided, both by the
- * generic IRQ layer and by the fact that an unacked local
- * APIC does not accept IRQs.
- */
-static unsigned int startup_level_ioapic_irq (unsigned int irq)
-{
-       unmask_IO_APIC_irq(irq);
-
-       return 0; /* don't check for pending */
-}
-
-static void end_level_ioapic_irq (unsigned int irq)
-{
-       move_irq(irq);
-       ack_APIC_irq();
-}
-
-#ifdef CONFIG_PCI_MSI
-static unsigned int startup_edge_ioapic_vector(unsigned int vector)
+static int ioapic_retrigger_irq(unsigned int irq)
 {
-       int irq = vector_to_irq(vector);
+       cpumask_t mask;
+       unsigned vector;
 
-       return startup_edge_ioapic_irq(irq);
-}
+       vector = irq_vector[irq];
+       cpus_clear(mask);
+       cpu_set(vector >> 8, mask);
 
-static void ack_edge_ioapic_vector(unsigned int vector)
-{
-       int irq = vector_to_irq(vector);
-
-       move_native_irq(vector);
-       ack_edge_ioapic_irq(irq);
-}
-
-static unsigned int startup_level_ioapic_vector (unsigned int vector)
-{
-       int irq = vector_to_irq(vector);
-
-       return startup_level_ioapic_irq (irq);
-}
-
-static void end_level_ioapic_vector (unsigned int vector)
-{
-       int irq = vector_to_irq(vector);
-
-       move_native_irq(vector);
-       end_level_ioapic_irq(irq);
-}
-
-static void mask_IO_APIC_vector (unsigned int vector)
-{
-       int irq = vector_to_irq(vector);
-
-       mask_IO_APIC_irq(irq);
-}
-
-static void unmask_IO_APIC_vector (unsigned int vector)
-{
-       int irq = vector_to_irq(vector);
-
-       unmask_IO_APIC_irq(irq);
-}
-
-#ifdef CONFIG_SMP
-static void set_ioapic_affinity_vector (unsigned int vector,
-                                       cpumask_t cpu_mask)
-{
-       int irq = vector_to_irq(vector);
-
-       set_native_irq_info(vector, cpu_mask);
-       set_ioapic_affinity_irq(irq, cpu_mask);
-}
-#endif // CONFIG_SMP
-#endif // CONFIG_PCI_MSI
-
-static int ioapic_retrigger(unsigned int irq)
-{
-       send_IPI_self(IO_APIC_VECTOR(irq));
+       send_IPI_mask(mask, vector & 0xff);
 
        return 1;
 }
@@ -1525,32 +1274,47 @@ static int ioapic_retrigger(unsigned int irq)
  * races.
  */
 
-static struct hw_interrupt_type ioapic_edge_type __read_mostly = {
-       .typename = "IO-APIC-edge",
-       .startup        = startup_edge_ioapic,
-       .shutdown       = shutdown_edge_ioapic,
-       .enable         = enable_edge_ioapic,
-       .disable        = disable_edge_ioapic,
-       .ack            = ack_edge_ioapic,
-       .end            = end_edge_ioapic,
-#ifdef CONFIG_SMP
-       .set_affinity = set_ioapic_affinity,
+static void ack_apic_edge(unsigned int irq)
+{
+       move_native_irq(irq);
+       ack_APIC_irq();
+}
+
+static void ack_apic_level(unsigned int irq)
+{
+       int do_unmask_irq = 0;
+
+#if defined(CONFIG_GENERIC_PENDING_IRQ) || defined(CONFIG_IRQBALANCE)
+       /* If we are moving the irq we need to mask it */
+       if (unlikely(irq_desc[irq].status & IRQ_MOVE_PENDING)) {
+               do_unmask_irq = 1;
+               mask_IO_APIC_irq(irq);
+       }
 #endif
-       .retrigger      = ioapic_retrigger,
-};
 
-static struct hw_interrupt_type ioapic_level_type __read_mostly = {
-       .typename = "IO-APIC-level",
-       .startup        = startup_level_ioapic,
-       .shutdown       = shutdown_level_ioapic,
-       .enable         = enable_level_ioapic,
-       .disable        = disable_level_ioapic,
-       .ack            = mask_and_ack_level_ioapic,
-       .end            = end_level_ioapic,
+       /*
+        * We must acknowledge the irq before we move it or the acknowledge will
+        * not propogate properly.
+        */
+       ack_APIC_irq();
+
+       /* Now we can move and renable the irq */
+       move_masked_irq(irq);
+       if (unlikely(do_unmask_irq))
+               unmask_IO_APIC_irq(irq);
+}
+
+static struct irq_chip ioapic_chip __read_mostly = {
+       .name           = "IO-APIC",
+       .startup        = startup_ioapic_irq,
+       .mask           = mask_IO_APIC_irq,
+       .unmask         = unmask_IO_APIC_irq,
+       .ack            = ack_apic_edge,
+       .eoi            = ack_apic_level,
 #ifdef CONFIG_SMP
-       .set_affinity = set_ioapic_affinity,
+       .set_affinity   = set_ioapic_affinity_irq,
 #endif
-       .retrigger      = ioapic_retrigger,
+       .retrigger      = ioapic_retrigger_irq,
 };
 
 static inline void init_IO_APIC_traps(void)
@@ -1570,12 +1334,7 @@ static inline void init_IO_APIC_traps(void)
         */
        for (irq = 0; irq < NR_IRQS ; irq++) {
                int tmp = irq;
-               if (use_pci_vector()) {
-                       if (!platform_legacy_irq(tmp))
-                               if ((tmp = vector_to_irq(tmp)) == -1)
-                                       continue;
-               }
-               if (IO_APIC_IRQ(tmp) && !IO_APIC_VECTOR(tmp)) {
+               if (IO_APIC_IRQ(tmp) && !irq_vector[tmp]) {
                        /*
                         * Hmm.. We don't have an entry for this,
                         * so default to an old-fashioned 8259
@@ -1585,7 +1344,7 @@ static inline void init_IO_APIC_traps(void)
                                make_8259A_irq(irq);
                        else
                                /* Strange. Oh, well.. */
-                               irq_desc[irq].chip = &no_irq_type;
+                               irq_desc[irq].chip = &no_irq_chip;
                }
        }
 }
@@ -1704,8 +1463,6 @@ static inline void unlock_ExtINT_logic(void)
        spin_unlock_irqrestore(&ioapic_lock, flags);
 }
 
-int timer_uses_ioapic_pin_0;
-
 /*
  * This code may look a bit paranoid, but it's supposed to cooperate with
  * a wide range of boards and BIOS bugs.  Fortunately only the timer IRQ
@@ -1718,13 +1475,13 @@ static inline void check_timer(void)
 {
        int apic1, pin1, apic2, pin2;
        int vector;
+       cpumask_t mask;
 
        /*
         * get/set the timer IRQ vector:
         */
        disable_8259A_irq(0);
-       vector = assign_irq_vector(0);
-       set_intr_gate(vector, interrupt[0]);
+       vector = assign_irq_vector(0, TARGET_CPUS, &mask);
 
        /*
         * Subtle, code in do_timer_interrupt() expects an AEOI
@@ -1743,9 +1500,6 @@ static inline void check_timer(void)
        pin2  = ioapic_i8259.pin;
        apic2 = ioapic_i8259.apic;
 
-       if (pin1 == 0)
-               timer_uses_ioapic_pin_0 = 1;
-
        apic_printk(APIC_VERBOSE,KERN_INFO "..TIMER: vector=0x%02X apic1=%d pin1=%d apic2=%d pin2=%d\n",
                vector, apic1, pin1, apic2, pin2);
 
@@ -1856,11 +1610,6 @@ void __init setup_IO_APIC(void)
 
        apic_printk(APIC_VERBOSE, "ENABLING IO-APIC IRQs\n");
 
-       /*
-        * Set up the IO-APIC IRQ routing table.
-        */
-       if (!acpi_ioapic)
-               setup_ioapic_ids_from_mpc();
        sync_Arb_IDs();
        setup_IO_APIC_irqs();
        init_IO_APIC_traps();
@@ -1879,17 +1628,12 @@ static int ioapic_suspend(struct sys_device *dev, pm_message_t state)
 {
        struct IO_APIC_route_entry *entry;
        struct sysfs_ioapic_data *data;
-       unsigned long flags;
        int i;
 
        data = container_of(dev, struct sysfs_ioapic_data, dev);
        entry = data->entry;
-       spin_lock_irqsave(&ioapic_lock, flags);
-       for (i = 0; i < nr_ioapic_registers[dev->id]; i ++, entry ++ ) {
-               *(((int *)entry) + 1) = io_apic_read(dev->id, 0x11 + 2 * i);
-               *(((int *)entry) + 0) = io_apic_read(dev->id, 0x10 + 2 * i);
-       }
-       spin_unlock_irqrestore(&ioapic_lock, flags);
+       for (i = 0; i < nr_ioapic_registers[dev->id]; i ++, entry ++ )
+               *entry = ioapic_read_entry(dev->id, i);
 
        return 0;
 }
@@ -1911,11 +1655,9 @@ static int ioapic_resume(struct sys_device *dev)
                reg_00.bits.ID = mp_ioapics[dev->id].mpc_apicid;
                io_apic_write(dev->id, 0, reg_00.raw);
        }
-       for (i = 0; i < nr_ioapic_registers[dev->id]; i ++, entry ++ ) {
-               io_apic_write(dev->id, 0x11+2*i, *(((int *)entry)+1));
-               io_apic_write(dev->id, 0x10+2*i, *(((int *)entry)+0));
-       }
        spin_unlock_irqrestore(&ioapic_lock, flags);
+       for (i = 0; i < nr_ioapic_registers[dev->id]; i++)
+               ioapic_write_entry(dev->id, i, entry[i]);
 
        return 0;
 }
@@ -1961,26 +1703,253 @@ static int __init ioapic_init_sysfs(void)
 
 device_initcall(ioapic_init_sysfs);
 
-/* --------------------------------------------------------------------------
-                          ACPI-based IOAPIC Configuration
-   -------------------------------------------------------------------------- */
+/*
+ * Dynamic irq allocate and deallocation
+ */
+int create_irq(void)
+{
+       /* Allocate an unused irq */
+       int irq;
+       int new;
+       int vector = 0;
+       unsigned long flags;
+       cpumask_t mask;
 
-#ifdef CONFIG_ACPI
+       irq = -ENOSPC;
+       spin_lock_irqsave(&vector_lock, flags);
+       for (new = (NR_IRQS - 1); new >= 0; new--) {
+               if (platform_legacy_irq(new))
+                       continue;
+               if (irq_vector[new] != 0)
+                       continue;
+               vector = __assign_irq_vector(new, TARGET_CPUS, &mask);
+               if (likely(vector > 0))
+                       irq = new;
+               break;
+       }
+       spin_unlock_irqrestore(&vector_lock, flags);
 
-#define IO_APIC_MAX_ID         0xFE
+       if (irq >= 0) {
+               dynamic_irq_init(irq);
+       }
+       return irq;
+}
 
-int __init io_apic_get_version (int ioapic)
+void destroy_irq(unsigned int irq)
 {
-       union IO_APIC_reg_01    reg_01;
        unsigned long flags;
 
-       spin_lock_irqsave(&ioapic_lock, flags);
-       reg_01.raw = io_apic_read(ioapic, 1);
-       spin_unlock_irqrestore(&ioapic_lock, flags);
+       dynamic_irq_cleanup(irq);
+
+       spin_lock_irqsave(&vector_lock, flags);
+       irq_vector[irq] = 0;
+       spin_unlock_irqrestore(&vector_lock, flags);
+}
+
+/*
+ * MSI mesage composition
+ */
+#ifdef CONFIG_PCI_MSI
+static int msi_compose_msg(struct pci_dev *pdev, unsigned int irq, struct msi_msg *msg)
+{
+       int vector;
+       unsigned dest;
+       cpumask_t tmp;
+
+       vector = assign_irq_vector(irq, TARGET_CPUS, &tmp);
+       if (vector >= 0) {
+               dest = cpu_mask_to_apicid(tmp);
+
+               msg->address_hi = MSI_ADDR_BASE_HI;
+               msg->address_lo =
+                       MSI_ADDR_BASE_LO |
+                       ((INT_DEST_MODE == 0) ?
+                               MSI_ADDR_DEST_MODE_PHYSICAL:
+                               MSI_ADDR_DEST_MODE_LOGICAL) |
+                       ((INT_DELIVERY_MODE != dest_LowestPrio) ?
+                               MSI_ADDR_REDIRECTION_CPU:
+                               MSI_ADDR_REDIRECTION_LOWPRI) |
+                       MSI_ADDR_DEST_ID(dest);
+
+               msg->data =
+                       MSI_DATA_TRIGGER_EDGE |
+                       MSI_DATA_LEVEL_ASSERT |
+                       ((INT_DELIVERY_MODE != dest_LowestPrio) ?
+                               MSI_DATA_DELIVERY_FIXED:
+                               MSI_DATA_DELIVERY_LOWPRI) |
+                       MSI_DATA_VECTOR(vector);
+       }
+       return vector;
+}
+
+#ifdef CONFIG_SMP
+static void set_msi_irq_affinity(unsigned int irq, cpumask_t mask)
+{
+       struct msi_msg msg;
+       unsigned int dest;
+       cpumask_t tmp;
+       int vector;
+
+       cpus_and(tmp, mask, cpu_online_map);
+       if (cpus_empty(tmp))
+               tmp = TARGET_CPUS;
+
+       cpus_and(mask, tmp, CPU_MASK_ALL);
+
+       vector = assign_irq_vector(irq, mask, &tmp);
+       if (vector < 0)
+               return;
+
+       dest = cpu_mask_to_apicid(tmp);
 
-       return reg_01.bits.version;
+       read_msi_msg(irq, &msg);
+
+       msg.data &= ~MSI_DATA_VECTOR_MASK;
+       msg.data |= MSI_DATA_VECTOR(vector);
+       msg.address_lo &= ~MSI_ADDR_DEST_ID_MASK;
+       msg.address_lo |= MSI_ADDR_DEST_ID(dest);
+
+       write_msi_msg(irq, &msg);
+       set_native_irq_info(irq, mask);
 }
+#endif /* CONFIG_SMP */
 
+/*
+ * IRQ Chip for MSI PCI/PCI-X/PCI-Express Devices,
+ * which implement the MSI or MSI-X Capability Structure.
+ */
+static struct irq_chip msi_chip = {
+       .name           = "PCI-MSI",
+       .unmask         = unmask_msi_irq,
+       .mask           = mask_msi_irq,
+       .ack            = ack_apic_edge,
+#ifdef CONFIG_SMP
+       .set_affinity   = set_msi_irq_affinity,
+#endif
+       .retrigger      = ioapic_retrigger_irq,
+};
+
+int arch_setup_msi_irq(unsigned int irq, struct pci_dev *dev)
+{
+       struct msi_msg msg;
+       int ret;
+       ret = msi_compose_msg(dev, irq, &msg);
+       if (ret < 0)
+               return ret;
+
+       write_msi_msg(irq, &msg);
+
+       set_irq_chip_and_handler_name(irq, &msi_chip, handle_edge_irq, "edge");
+
+       return 0;
+}
+
+void arch_teardown_msi_irq(unsigned int irq)
+{
+       return;
+}
+
+#endif /* CONFIG_PCI_MSI */
+
+/*
+ * Hypertransport interrupt support
+ */
+#ifdef CONFIG_HT_IRQ
+
+#ifdef CONFIG_SMP
+
+static void target_ht_irq(unsigned int irq, unsigned int dest, u8 vector)
+{
+       u32 low, high;
+       low  = read_ht_irq_low(irq);
+       high = read_ht_irq_high(irq);
+
+       low  &= ~(HT_IRQ_LOW_VECTOR_MASK | HT_IRQ_LOW_DEST_ID_MASK);
+       high &= ~(HT_IRQ_HIGH_DEST_ID_MASK);
+
+       low  |= HT_IRQ_LOW_VECTOR(vector) | HT_IRQ_LOW_DEST_ID(dest);
+       high |= HT_IRQ_HIGH_DEST_ID(dest);
+
+       write_ht_irq_low(irq, low);
+       write_ht_irq_high(irq, high);
+}
+
+static void set_ht_irq_affinity(unsigned int irq, cpumask_t mask)
+{
+       unsigned int dest;
+       cpumask_t tmp;
+       int vector;
+
+       cpus_and(tmp, mask, cpu_online_map);
+       if (cpus_empty(tmp))
+               tmp = TARGET_CPUS;
+
+       cpus_and(mask, tmp, CPU_MASK_ALL);
+
+       vector = assign_irq_vector(irq, mask, &tmp);
+       if (vector < 0)
+               return;
+
+       dest = cpu_mask_to_apicid(tmp);
+
+       target_ht_irq(irq, dest, vector & 0xff);
+       set_native_irq_info(irq, mask);
+}
+#endif
+
+static struct irq_chip ht_irq_chip = {
+       .name           = "PCI-HT",
+       .mask           = mask_ht_irq,
+       .unmask         = unmask_ht_irq,
+       .ack            = ack_apic_edge,
+#ifdef CONFIG_SMP
+       .set_affinity   = set_ht_irq_affinity,
+#endif
+       .retrigger      = ioapic_retrigger_irq,
+};
+
+int arch_setup_ht_irq(unsigned int irq, struct pci_dev *dev)
+{
+       int vector;
+       cpumask_t tmp;
+
+       vector = assign_irq_vector(irq, TARGET_CPUS, &tmp);
+       if (vector >= 0) {
+               u32 low, high;
+               unsigned dest;
+
+               dest = cpu_mask_to_apicid(tmp);
+
+               high =  HT_IRQ_HIGH_DEST_ID(dest);
+
+               low =   HT_IRQ_LOW_BASE |
+                       HT_IRQ_LOW_DEST_ID(dest) |
+                       HT_IRQ_LOW_VECTOR(vector) |
+                       ((INT_DEST_MODE == 0) ?
+                               HT_IRQ_LOW_DM_PHYSICAL :
+                               HT_IRQ_LOW_DM_LOGICAL) |
+                       HT_IRQ_LOW_RQEOI_EDGE |
+                       ((INT_DELIVERY_MODE != dest_LowestPrio) ?
+                               HT_IRQ_LOW_MT_FIXED :
+                               HT_IRQ_LOW_MT_ARBITRATED);
+
+               write_ht_irq_low(irq, low);
+               write_ht_irq_high(irq, high);
+
+               set_irq_chip_and_handler_name(irq, &ht_irq_chip,
+                                             handle_edge_irq, "edge");
+       }
+       return vector;
+}
+#endif /* CONFIG_HT_IRQ */
+
+/* --------------------------------------------------------------------------
+                          ACPI-based IOAPIC Configuration
+   -------------------------------------------------------------------------- */
+
+#ifdef CONFIG_ACPI
+
+#define IO_APIC_MAX_ID         0xFE
 
 int __init io_apic_get_redir_entries (int ioapic)
 {
@@ -1999,6 +1968,8 @@ int io_apic_set_pci_routing (int ioapic, int pin, int irq, int triggering, int p
 {
        struct IO_APIC_route_entry entry;
        unsigned long flags;
+       int vector;
+       cpumask_t mask;
 
        if (!IO_APIC_IRQ(irq)) {
                apic_printk(APIC_QUIET,KERN_ERR "IOAPIC[%d]: Invalid reference to IRQ 0\n",
@@ -2006,6 +1977,17 @@ int io_apic_set_pci_routing (int ioapic, int pin, int irq, int triggering, int p
                return -EINVAL;
        }
 
+       /*
+        * IRQs < 16 are already in the irq_2_pin[] map
+        */
+       if (irq >= 16)
+               add_pin_to_irq(irq, ioapic, pin);
+
+
+       vector = assign_irq_vector(irq, TARGET_CPUS, &mask);
+       if (vector < 0)
+               return vector;
+
        /*
         * Generate a PCI IRQ routing entry and program the IOAPIC accordingly.
         * Note that we mask (disable) IRQs now -- these get enabled when the
@@ -2016,19 +1998,11 @@ int io_apic_set_pci_routing (int ioapic, int pin, int irq, int triggering, int p
 
        entry.delivery_mode = INT_DELIVERY_MODE;
        entry.dest_mode = INT_DEST_MODE;
-       entry.dest.logical.logical_dest = cpu_mask_to_apicid(TARGET_CPUS);
+       entry.dest.logical.logical_dest = cpu_mask_to_apicid(mask);
        entry.trigger = triggering;
        entry.polarity = polarity;
        entry.mask = 1;                                  /* Disabled (masked) */
-
-       irq = gsi_irq_sharing(irq);
-       /*
-        * IRQs < 16 are already in the irq_2_pin[] map
-        */
-       if (irq >= 16)
-               add_pin_to_irq(irq, ioapic, pin);
-
-       entry.vector = assign_irq_vector(irq);
+       entry.vector = vector & 0xff;
 
        apic_printk(APIC_VERBOSE,KERN_DEBUG "IOAPIC[%d]: Set PCI routing entry (%d-%d -> 0x%x -> "
                "IRQ %d Mode:%i Active:%i)\n", ioapic, 
@@ -2040,10 +2014,10 @@ int io_apic_set_pci_routing (int ioapic, int pin, int irq, int triggering, int p
        if (!ioapic && (irq < 16))
                disable_8259A_irq(irq);
 
+       ioapic_write_entry(ioapic, pin, entry);
+
        spin_lock_irqsave(&ioapic_lock, flags);
-       io_apic_write(ioapic, 0x11+2*pin, *(((int *)&entry)+1));
-       io_apic_write(ioapic, 0x10+2*pin, *(((int *)&entry)+0));
-       set_native_irq_info(use_pci_vector() ?  entry.vector : irq, TARGET_CPUS);
+       set_native_irq_info(irq, TARGET_CPUS);
        spin_unlock_irqrestore(&ioapic_lock, flags);
 
        return 0;