MSI 2/6: change the pirq to be per-domain

Signed-off-by: Jiang Yunhong <yunhong.jiang@intel.com>
Signed-off-by: Shan Haitao <haitao.shan@intel.com>

18 files changed, 529 insertions, 25 deletions

diff --git a/tools/ioemu/hw/pass-through.c b/tools/ioemu/hw/pass-through.c
index 29ffda451e..0da7176d44 100644
--- a/tools/ioemu/hw/pass-through.c
+++ b/tools/ioemu/hw/pass-through.c
@@ -519,7 +519,21 @@ struct pt_dev * register_real_device(PCIBus *e_bus,
     e_intx = assigned_device->dev.config[0x3d]-1;
 
     if ( PT_MACHINE_IRQ_AUTO == machine_irq )
+    {
+        int pirq = pci_dev->irq;
+
         machine_irq = pci_dev->irq;
+        rc = xc_physdev_map_pirq(xc_handle, domid, MAP_PIRQ_TYPE_GSI,
+                                machine_irq, &pirq);
+
+        if ( rc )
+        {
+            /* TBD: unregister device in case of an error */
+            PT_LOG("Error: Mapping irq failed, rc = %d\n", rc);
+        }
+        else
+            machine_irq = pirq;
+    }
 
     /* bind machine_irq to device */
     if ( 0 != machine_irq )
diff --git a/tools/libxc/xc_physdev.c b/tools/libxc/xc_physdev.c
index 43ea7ff53f..1c5e8cb5da 100644
--- a/tools/libxc/xc_physdev.c
+++ b/tools/libxc/xc_physdev.c
@@ -19,3 +19,43 @@ int xc_physdev_pci_access_modify(int xc_handle,
     errno = ENOSYS;
     return -1;
 }
+
+int xc_physdev_map_pirq(int xc_handle,
+                        int domid,
+                        int type,
+                        int index,
+                        int *pirq)
+{
+    int rc;
+    struct physdev_map_pirq map;
+
+    if ( !pirq )
+        return -EINVAL;
+
+    map.domid = domid;
+    map.type = type;
+    map.index = index;
+    map.pirq = *pirq;
+
+    rc = do_physdev_op(xc_handle, PHYSDEVOP_map_pirq, &map);
+
+    if ( !rc )
+        *pirq = map.pirq;
+
+    return rc;
+}
+
+int xc_physdev_unmap_pirq(int xc_handle,
+                          int domid,
+                          int pirq)
+{
+    int rc;
+    struct physdev_unmap_pirq unmap;
+
+    unmap.domid = domid;
+    unmap.pirq = pirq;
+
+    rc = do_physdev_op(xc_handle, PHYSDEVOP_unmap_pirq, &unmap);
+
+    return rc;
+}
diff --git a/tools/libxc/xc_private.h b/tools/libxc/xc_private.h
index b068e45dbe..0570e62dfb 100644
--- a/tools/libxc/xc_private.h
+++ b/tools/libxc/xc_private.h
@@ -24,10 +24,12 @@
 #define DECLARE_HYPERCALL privcmd_hypercall_t hypercall = { 0 }
 #define DECLARE_DOMCTL struct xen_domctl domctl = { 0 }
 #define DECLARE_SYSCTL struct xen_sysctl sysctl = { 0 }
+#define DECLARE_PHYSDEV_OP struct physdev_op physdev_op = { 0 }
 #else
 #define DECLARE_HYPERCALL privcmd_hypercall_t hypercall
 #define DECLARE_DOMCTL struct xen_domctl domctl
 #define DECLARE_SYSCTL struct xen_sysctl sysctl
+#define DECLARE_PHYSDEV_OP struct physdev_op physdev_op
 #endif
 
 #undef PAGE_SHIFT
@@ -96,6 +98,34 @@ static inline int do_xen_version(int xc_handle, int cmd, void *dest)
     return do_xen_hypercall(xc_handle, &hypercall);
 }
 
+static inline int do_physdev_op(int xc_handle, int cmd, void *op)
+{
+    int ret = -1;
+
+    DECLARE_HYPERCALL;
+    hypercall.op = __HYPERVISOR_physdev_op;
+    hypercall.arg[0] = (unsigned long) cmd;
+    hypercall.arg[1] = (unsigned long) op;
+
+    if ( lock_pages(op, sizeof(*op)) != 0 )
+    {
+        PERROR("Could not lock memory for Xen hypercall");
+        goto out1;
+    }
+
+    if ( (ret = do_xen_hypercall(xc_handle, &hypercall)) < 0 )
+    {
+        if ( errno == EACCES )
+            DPRINTF("physdev operation failed -- need to"
+                    " rebuild the user-space tool set?\n");
+    }
+
+    unlock_pages(op, sizeof(*op));
+
+out1:
+    return ret;
+}
+
 static inline int do_domctl(int xc_handle, struct xen_domctl *domctl)
 {
     int ret = -1;
diff --git a/tools/libxc/xenctrl.h b/tools/libxc/xenctrl.h
index e05df73162..09e7becea3 100644
--- a/tools/libxc/xenctrl.h
+++ b/tools/libxc/xenctrl.h
@@ -21,6 +21,7 @@
 #include <stdint.h>
 #include <xen/xen.h>
 #include <xen/domctl.h>
+#include <xen/physdev.h>
 #include <xen/sysctl.h>
 #include <xen/version.h>
 #include <xen/event_channel.h>
@@ -849,6 +850,16 @@ int xc_gnttab_munmap(int xcg_handle,
 int xc_gnttab_set_max_grants(int xcg_handle,
 			     uint32_t count);
 
+int xc_physdev_map_pirq(int xc_handle,
+                        int domid,
+                        int type,
+                        int index,
+                        int *pirq);
+
+int xc_physdev_unmap_pirq(int xc_handle,
+                          int domid,
+                          int pirq);
+
 int xc_hvm_set_pci_intx_level(
     int xc_handle, domid_t dom,
     uint8_t domain, uint8_t bus, uint8_t device, uint8_t intx,
diff --git a/tools/python/xen/lowlevel/xc/xc.c b/tools/python/xen/lowlevel/xc/xc.c
index 53a26eed6b..709bbba62b 100644
--- a/tools/python/xen/lowlevel/xc/xc.c
+++ b/tools/python/xen/lowlevel/xc/xc.c
@@ -799,6 +799,26 @@ static PyObject *pyxc_evtchn_reset(XcObject *self,
     return zero;
 }
 
+static PyObject *pyxc_physdev_map_pirq(PyObject *self,
+                                       PyObject *args,
+                                       PyObject *kwds)
+{
+    XcObject *xc = (XcObject *)self;
+    uint32_t dom;
+    int index, pirq, ret;
+
+    static char *kwd_list[] = {"domid", "index", "pirq", NULL};
+
+    if ( !PyArg_ParseTupleAndKeywords(args, kwds, "iii", kwd_list,
+                                      &dom, &index, &pirq) )
+        return NULL;
+    ret = xc_physdev_map_pirq(xc->xc_handle, dom, MAP_PIRQ_TYPE_GSI,
+                             index, &pirq);
+    if ( ret != 0 )
+          return pyxc_error_to_exception();
+    return PyLong_FromUnsignedLong(pirq);
+}
+
 static PyObject *pyxc_physdev_pci_access_modify(XcObject *self,
                                                 PyObject *args,
                                                 PyObject *kwds)
@@ -1589,6 +1609,15 @@ static PyMethodDef pyxc_methods[] = {
       "Reset all connections.\n"
       " dom [int]: Domain to reset.\n" },
 
+    { "physdev_map_pirq",
+      (PyCFunction)pyxc_physdev_map_pirq,
+      METH_VARARGS | METH_KEYWORDS, "\n"
+      "map physical irq to guest pirq.\n"
+      " dom     [int]:      Identifier of domain to map for.\n"
+      " index   [int]:      physical irq.\n"
+      " pirq    [int]:      guest pirq.\n"
+      "Returns: [long] value of the param.\n" },
+
     { "physdev_pci_access_modify",
       (PyCFunction)pyxc_physdev_pci_access_modify,
       METH_VARARGS | METH_KEYWORDS, "\n"
diff --git a/tools/python/xen/xend/server/irqif.py b/tools/python/xen/xend/server/irqif.py
index 7e96726834..db4b1deedd 100644
--- a/tools/python/xen/xend/server/irqif.py
+++ b/tools/python/xen/xend/server/irqif.py
@@ -69,5 +69,10 @@ class IRQController(DevController):
             #todo non-fatal
             raise VmError(
                 'irq: Failed to configure irq: %d' % (pirq))
-
+        rc = xc.physdev_map_pirq(domid = self.getDomid(),
+                                index = pirq,
+                                pirq  = pirq)
+        if rc < 0:
+            raise VmError(
+                'irq: Failed to map irq %x' % (pirq))
         return (None, {}, {})
diff --git a/tools/python/xen/xend/server/pciif.py b/tools/python/xen/xend/server/pciif.py
index e616ebfe11..6f08e77e7c 100644
--- a/tools/python/xen/xend/server/pciif.py
+++ b/tools/python/xen/xend/server/pciif.py
@@ -271,6 +271,12 @@ class PciController(DevController):
             if rc<0:
                 raise VmError(('pci: failed to configure I/O memory on device '+
                             '%s - errno=%d')%(dev.name,rc))
+            rc = xc.physdev_map_pirq(domid = fe_domid,
+                                   index = dev.irq,
+                                   pirq  = dev.irq)
+            if rc < 0:
+                raise VmError(('pci: failed to map irq on device '+
+                            '%s - errno=%d')%(dev.name,rc))
 
         if dev.irq>0:
             log.debug('pci: enabling irq %d'%dev.irq)
diff --git a/xen/arch/x86/domain.c b/xen/arch/x86/domain.c
index 4fd1ac1f51..85ff65eadc 100644
--- a/xen/arch/x86/domain.c
+++ b/xen/arch/x86/domain.c
@@ -525,6 +525,8 @@ int arch_domain_create(struct domain *d, unsigned int domcr_flags)
             goto fail;
     }
 
+    spin_lock_init(&d->arch.irq_lock);
+
     if ( is_hvm_domain(d) )
     {
         if ( (rc = hvm_domain_initialise(d)) != 0 )
diff --git a/xen/arch/x86/hvm/vmx/intr.c b/xen/arch/x86/hvm/vmx/intr.c
index 3fcfc52175..72dfd1d6ac 100644
--- a/xen/arch/x86/hvm/vmx/intr.c
+++ b/xen/arch/x86/hvm/vmx/intr.c
@@ -121,7 +121,7 @@ static void vmx_dirq_assist(struct vcpu *v)
         if ( !test_and_clear_bit(irq, &hvm_irq_dpci->dirq_mask) )
             continue;
 
-        stop_timer(&hvm_irq_dpci->hvm_timer[irq_to_vector(irq)]);
+        stop_timer(&hvm_irq_dpci->hvm_timer[domain_irq_to_vector(d, irq)]);
 
         list_for_each_entry ( digl, &hvm_irq_dpci->mirq[irq].digl_list, list )
         {
@@ -140,7 +140,7 @@ static void vmx_dirq_assist(struct vcpu *v)
          * guest will never deal with the irq, then the physical interrupt line
          * will never be deasserted.
          */
-        set_timer(&hvm_irq_dpci->hvm_timer[irq_to_vector(irq)],
+        set_timer(&hvm_irq_dpci->hvm_timer[domain_irq_to_vector(d, irq)],
                   NOW() + PT_IRQ_TIME_OUT);
     }
 }
diff --git a/xen/arch/x86/io_apic.c b/xen/arch/x86/io_apic.c
index b7e50ae8f1..67cfed78b6 100644
--- a/xen/arch/x86/io_apic.c
+++ b/xen/arch/x86/io_apic.c
@@ -41,6 +41,25 @@
 int (*ioapic_renumber_irq)(int ioapic, int irq);
 atomic_t irq_mis_count;
 
+int msi_irq_enable = 0;
+boolean_param("msi_irq_enable", msi_irq_enable);
+
+int domain_irq_to_vector(struct domain *d, int irq)
+{
+    if ( !msi_irq_enable )
+        return irq_to_vector(irq);
+    else
+        return d->arch.pirq_vector[irq];
+}
+
+int domain_vector_to_irq(struct domain *d, int vector)
+{
+    if ( !msi_irq_enable )
+        return vector_to_irq(vector);
+    else
+        return d->arch.vector_pirq[vector];
+}
+
 /* Where if anywhere is the i8259 connect in external int mode */
 static struct { int pin, apic; } ioapic_i8259 = { -1, -1 };
 
diff --git a/xen/arch/x86/irq.c b/xen/arch/x86/irq.c
index fc050d7d60..ea840dfbdd 100644
--- a/xen/arch/x86/irq.c
+++ b/xen/arch/x86/irq.c
@@ -203,7 +203,6 @@ static DEFINE_PER_CPU(struct pending_eoi, pending_eoi[NR_VECTORS]);
 
 static void __do_IRQ_guest(int vector)
 {
-    unsigned int        irq = vector_to_irq(vector);
     irq_desc_t         *desc = &irq_desc[vector];
     irq_guest_action_t *action = (irq_guest_action_t *)desc->action;
     struct domain      *d;
@@ -232,7 +231,9 @@ static void __do_IRQ_guest(int vector)
 
     for ( i = 0; i < action->nr_guests; i++ )
     {
+        unsigned int irq;
         d = action->guest[i];
+        irq = domain_vector_to_irq(d, vector);
         if ( (action->ack_type != ACKTYPE_NONE) &&
              !test_and_set_bit(irq, d->pirq_mask) )
             action->in_flight++;
@@ -305,8 +306,10 @@ static void __pirq_guest_eoi(struct domain *d, int irq)
     irq_desc_t         *desc;
     irq_guest_action_t *action;
     cpumask_t           cpu_eoi_map;
+    int                 vector;
 
-    desc   = &irq_desc[irq_to_vector(irq)];
+    vector = domain_irq_to_vector(d, irq);
+    desc   = &irq_desc[vector];
     action = (irq_guest_action_t *)desc->action;
 
     spin_lock_irq(&desc->lock);
@@ -324,7 +327,7 @@ static void __pirq_guest_eoi(struct domain *d, int irq)
     if ( action->ack_type == ACKTYPE_UNMASK )
     {
         ASSERT(cpus_empty(action->cpu_eoi_map));
-        desc->handler->end(irq_to_vector(irq));
+        desc->handler->end(vector);
         spin_unlock_irq(&desc->lock);
         return;
     }
@@ -375,12 +378,12 @@ int pirq_guest_unmask(struct domain *d)
 }
 
 extern int ioapic_ack_new;
-int pirq_acktype(int irq)
+int pirq_acktype(struct domain *d, int irq)
 {
     irq_desc_t  *desc;
     unsigned int vector;
 
-    vector = irq_to_vector(irq);
+    vector = domain_irq_to_vector(d, irq);
     if ( vector == 0 )
         return ACKTYPE_NONE;
 
@@ -421,7 +424,7 @@ int pirq_acktype(int irq)
     return 0;
 }
 
-int pirq_shared(int irq)
+int pirq_shared(struct domain *d, int irq)
 {
     unsigned int        vector;
     irq_desc_t         *desc;
@@ -429,7 +432,7 @@ int pirq_shared(int irq)
     unsigned long       flags;
     int                 shared;
 
-    vector = irq_to_vector(irq);
+    vector = domain_irq_to_vector(d, irq);
     if ( vector == 0 )
         return 0;
 
@@ -453,7 +456,7 @@ int pirq_guest_bind(struct vcpu *v, int irq, int will_share)
     cpumask_t           cpumask = CPU_MASK_NONE;
 
  retry:
-    vector = irq_to_vector(irq);
+    vector = domain_irq_to_vector(v->domain, irq);
     if ( vector == 0 )
         return -EINVAL;
 
@@ -487,7 +490,7 @@ int pirq_guest_bind(struct vcpu *v, int irq, int will_share)
         action->nr_guests   = 0;
         action->in_flight   = 0;
         action->shareable   = will_share;
-        action->ack_type    = pirq_acktype(irq);
+        action->ack_type    = pirq_acktype(v->domain, irq);
         cpus_clear(action->cpu_eoi_map);
 
         desc->depth = 0;
@@ -538,13 +541,15 @@ int pirq_guest_bind(struct vcpu *v, int irq, int will_share)
 
 int pirq_guest_unbind(struct domain *d, int irq)
 {
-    unsigned int        vector = irq_to_vector(irq);
-    irq_desc_t         *desc = &irq_desc[vector];
+    unsigned int        vector;
+    irq_desc_t         *desc;
     irq_guest_action_t *action;
     cpumask_t           cpu_eoi_map;
     unsigned long       flags;
     int                 i;
 
+    vector = domain_irq_to_vector(d, irq);
+    desc = &irq_desc[vector];
     BUG_ON(vector == 0);
 
     spin_lock_irqsave(&desc->lock, flags);
diff --git a/xen/arch/x86/physdev.c b/xen/arch/x86/physdev.c
index e42379d4ac..dd396fed07 100644
--- a/xen/arch/x86/physdev.c
+++ b/xen/arch/x86/physdev.c
@@ -7,6 +7,7 @@
 #include <xen/irq.h>
 #include <xen/event.h>
 #include <xen/guest_access.h>
+#include <xen/iocap.h>
 #include <asm/current.h>
 #include <asm/hypercall.h>
 #include <public/xen.h>
@@ -24,6 +25,263 @@ int
 ioapic_guest_write(
     unsigned long physbase, unsigned int reg, u32 pval);
 
+static int get_free_pirq(struct domain *d, int type, int index)
+{
+    int i;
+
+    if ( d == NULL )
+        return -EINVAL;
+
+    ASSERT(spin_is_locked(&d->arch.irq_lock));
+
+    if ( type == MAP_PIRQ_TYPE_GSI )
+    {
+        for ( i = 16; i < NR_PIRQS; i++ )
+            if ( !d->arch.pirq_vector[i] )
+                break;
+        if ( i == NR_PIRQS )
+            return -ENOSPC;
+    }
+    else
+    {
+        for ( i = NR_PIRQS - 1; i >= 16; i-- )
+            if ( !d->arch.pirq_vector[i] )
+                break;
+        if ( i == 16 )
+            return -ENOSPC;
+    }
+
+    return i;
+}
+
+/*
+ * Caller hold the irq_lock
+ */
+static int map_domain_pirq(struct domain *d, int pirq, int vector, int type)
+{
+    int ret = 0;
+    int old_vector, old_pirq;
+
+    if ( d == NULL )
+        return -EINVAL;
+
+    ASSERT(spin_is_locked(&d->arch.irq_lock));
+
+    if ( !IS_PRIV(current->domain) )
+        return -EPERM;
+
+    if ( pirq < 0 || pirq >= NR_PIRQS || vector < 0 || vector >= NR_VECTORS )
+    {
+        gdprintk(XENLOG_G_ERR,
+                 "invalid pirq %x or vector %x\n", pirq, vector);
+        return -EINVAL;
+    }
+
+    old_vector = d->arch.pirq_vector[pirq];
+    old_pirq = d->arch.vector_pirq[vector];
+
+    if ( (old_vector && (old_vector != vector) ) ||
+         (old_pirq && (old_pirq != pirq)) )
+    {
+        gdprintk(XENLOG_G_ERR, "remap pirq %x vector %x while not unmap\n",
+                 pirq, vector);
+        ret = -EINVAL;
+        goto done;
+    }
+
+    ret = irq_permit_access(d, pirq);
+    if ( ret )
+    {
+        gdprintk(XENLOG_G_ERR, "add irq permit access %x failed\n", pirq);
+        ret = -EINVAL;
+        goto done;
+    }
+
+    d->arch.pirq_vector[pirq] = vector;
+    d->arch.vector_pirq[vector] = pirq;
+
+done:
+    return ret;
+}
+
+/*
+ * The pirq should has been unbound before this call
+ */
+static int unmap_domain_pirq(struct domain *d, int pirq)
+{
+    int ret = 0;
+    int vector;
+
+    if ( d == NULL || pirq < 0 || pirq > NR_PIRQS )
+        return -EINVAL;
+
+    if ( !IS_PRIV(current->domain) )
+        return -EINVAL;
+
+    ASSERT(spin_is_locked(&d->arch.irq_lock));
+
+    vector = d->arch.pirq_vector[pirq];
+
+    if ( !vector )
+    {
+        gdprintk(XENLOG_G_ERR, "domain %X: pirq %x not mapped still\n",
+                 d->domain_id, pirq);
+        ret = -EINVAL;
+    }
+    else
+        d->arch.pirq_vector[pirq] = d->arch.vector_pirq[vector] = 0;
+    ret = irq_deny_access(d, pirq);
+
+    if ( ret )
+        gdprintk(XENLOG_G_ERR, "deny irq %x access failed\n", pirq);
+
+    return ret;
+}
+
+extern int msi_irq_enable;
+static int physdev_map_pirq(struct physdev_map_pirq *map)
+{
+    struct domain *d;
+    int vector, pirq, ret = 0;
+    unsigned long flags;
+
+    /* if msi_irq_enable is not enabled,map always success */
+    if ( !msi_irq_enable )
+        return 0;
+
+    if ( !IS_PRIV(current->domain) )
+        return -EPERM;
+
+    if ( !map )
+        return -EINVAL;
+
+    if ( map->domid == DOMID_SELF )
+        d = rcu_lock_domain(current->domain);
+    else
+        d = rcu_lock_domain_by_id(map->domid);
+
+    if ( d == NULL )
+    {
+        ret = -ESRCH;
+        goto free_domain;
+    }
+
+    switch ( map->type )
+    {
+        case MAP_PIRQ_TYPE_GSI:
+            if ( map->index >= NR_IRQS )
+            {
+                ret = -EINVAL;
+                gdprintk(XENLOG_G_ERR,
+                         "map invalid irq %x\n", map->index);
+                goto free_domain;
+            }
+            vector = IO_APIC_VECTOR(map->index);
+            if ( !vector )
+            {
+                ret = -EINVAL;
+                gdprintk(XENLOG_G_ERR,
+                         "map irq with no vector %x\n", map->index);
+                goto free_domain;
+            }
+            break;
+        case MAP_PIRQ_TYPE_MSI:
+            vector = map->index;
+            if ( vector < 0 || vector >= NR_VECTORS )
+            {
+                ret = -EINVAL;
+                gdprintk(XENLOG_G_ERR,
+                         "map_pirq with wrong vector %x\n", map->index);
+                goto free_domain;
+            }
+            break;
+        default:
+            ret = -EINVAL;
+            gdprintk(XENLOG_G_ERR, "wrong map_pirq type %x\n", map->type);
+            goto free_domain;
+            break;
+    }
+
+    spin_lock_irqsave(&d->arch.irq_lock, flags);
+    if ( map->pirq == -1 )
+    {
+        if ( d->arch.vector_pirq[vector] )
+        {
+            gdprintk(XENLOG_G_ERR, "%x %x mapped already%x\n",
+                                    map->index, map->pirq,
+                                    d->arch.vector_pirq[vector]);
+            pirq = d->arch.vector_pirq[vector];
+        }
+        else
+        {
+            pirq = get_free_pirq(d, map->type, map->index);
+            if ( pirq < 0 )
+            {
+                ret = pirq;
+                gdprintk(XENLOG_G_ERR, "No free pirq\n");
+                goto done;
+            }
+        }
+    }
+    else
+    {
+        if ( d->arch.vector_pirq[vector] &&
+             d->arch.vector_pirq[vector] != map->pirq )
+        {
+            gdprintk(XENLOG_G_ERR, "%x conflict with %x\n",
+              map->index, map->pirq);
+            ret = -EEXIST;
+            goto done;
+        }
+        else
+            pirq = map->pirq;
+    }
+
+    ret = map_domain_pirq(d, pirq, vector, map->type);
+
+    if ( !ret )
+        map->pirq = pirq;
+done:
+    spin_unlock_irqrestore(&d->arch.irq_lock, flags);
+free_domain:
+    rcu_unlock_domain(d);
+    return ret;
+}
+
+static int physdev_unmap_pirq(struct physdev_unmap_pirq *unmap)
+{
+    struct domain *d;
+    unsigned long flags;
+    int ret;
+
+    if ( !msi_irq_enable )
+        return 0;
+
+    if ( !IS_PRIV(current->domain) )
+        return -EPERM;
+
+    if ( !unmap )
+        return -EINVAL;
+
+    if ( unmap->domid == DOMID_SELF )
+        d = rcu_lock_domain(current->domain);
+    else
+        d = rcu_lock_domain_by_id(unmap->domid);
+
+    if ( d == NULL )
+    {
+        rcu_unlock_domain(d);
+        return -ESRCH;
+    }
+
+    spin_lock_irqsave(&d->arch.irq_lock, flags);
+    ret = unmap_domain_pirq(d, unmap->pirq);
+    spin_unlock_irqrestore(&d->arch.irq_lock, flags);
+    rcu_unlock_domain(d);
+
+    return ret;
+}
+
 ret_t do_physdev_op(int cmd, XEN_GUEST_HANDLE(void) arg)
 {
     int irq;
@@ -57,14 +315,38 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_HANDLE(void) arg)
         if ( (irq < 0) || (irq >= NR_IRQS) )
             break;
         irq_status_query.flags = 0;
-        if ( pirq_acktype(irq) != 0 )
+        if ( pirq_acktype(v->domain, irq) != 0 )
             irq_status_query.flags |= XENIRQSTAT_needs_eoi;
-        if ( pirq_shared(irq) )
+        if ( pirq_shared(v->domain, irq) )
             irq_status_query.flags |= XENIRQSTAT_shared;
         ret = copy_to_guest(arg, &irq_status_query, 1) ? -EFAULT : 0;
         break;
     }
 
+    case PHYSDEVOP_map_pirq: {
+        struct physdev_map_pirq map;
+
+        ret = -EFAULT;
+        if ( copy_from_guest(&map, arg, 1) != 0 )
+            break;
+
+        ret = physdev_map_pirq(&map);
+        if ( copy_to_guest(arg, &map, 1) != 0 )
+            ret = -EFAULT;
+        break;
+    }
+
+    case PHYSDEVOP_unmap_pirq: {
+        struct physdev_unmap_pirq unmap;
+
+        ret = -EFAULT;
+        if ( copy_from_guest(&unmap, arg, 1) != 0 )
+            break;
+
+        ret = physdev_unmap_pirq(&unmap);
+        break;
+    }
+
     case PHYSDEVOP_apic_read: {
         struct physdev_apic apic;
         ret = -EFAULT;
@@ -99,6 +381,7 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_HANDLE(void) arg)
 
     case PHYSDEVOP_alloc_irq_vector: {
         struct physdev_irq irq_op;
+        unsigned long flags;
 
         ret = -EFAULT;
         if ( copy_from_guest(&irq_op, arg, 1) != 0 )
@@ -118,7 +401,20 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_HANDLE(void) arg)
             break;
 
         irq_op.vector = assign_irq_vector(irq);
-        ret = copy_to_guest(arg, &irq_op, 1) ? -EFAULT : 0;
+
+        ret = 0;
+
+        if ( msi_irq_enable )
+        {
+            spin_lock_irqsave(&dom0->arch.irq_lock, flags);
+            if ( irq != AUTO_ASSIGN )
+                ret = map_domain_pirq(dom0, irq_op.irq, irq_op.vector,
+                                     MAP_PIRQ_TYPE_GSI);
+            spin_unlock_irqrestore(&dom0->arch.irq_lock, flags);
+        }
+
+        if ( copy_to_guest(arg, &irq_op, 1) != 0 )
+            ret = -EFAULT;
         break;
     }
 
diff --git a/xen/drivers/passthrough/io.c b/xen/drivers/passthrough/io.c
index 528839550b..b0f55bc802 100644
--- a/xen/drivers/passthrough/io.c
+++ b/xen/drivers/passthrough/io.c
@@ -25,6 +25,7 @@ static void pt_irq_time_out(void *data)
 {
     struct hvm_mirq_dpci_mapping *irq_map = data;
     unsigned int guest_gsi, machine_gsi = 0;
+    int vector;
     struct hvm_irq_dpci *dpci = domain_get_irq_dpci(irq_map->dom);
     struct dev_intx_gsi_link *digl;
     uint32_t device, intx;
@@ -39,7 +40,8 @@ static void pt_irq_time_out(void *data)
     }
 
     clear_bit(machine_gsi, dpci->dirq_mask);
-    stop_timer(&dpci->hvm_timer[irq_to_vector(machine_gsi)]);
+    vector = domain_irq_to_vector(irq_map->dom, machine_gsi);
+    stop_timer(&dpci->hvm_timer[vector]);
     spin_lock(&dpci->dirq_lock);
     dpci->mirq[machine_gsi].pending = 0;
     spin_unlock(&dpci->dirq_lock);
@@ -98,7 +100,7 @@ int pt_irq_create_bind_vtd(
         hvm_irq_dpci->mirq[machine_gsi].valid = 1;
         hvm_irq_dpci->mirq[machine_gsi].dom = d;
 
-        init_timer(&hvm_irq_dpci->hvm_timer[irq_to_vector(machine_gsi)],
+        init_timer(&hvm_irq_dpci->hvm_timer[domain_irq_to_vector(d, machine_gsi)],
                    pt_irq_time_out, &hvm_irq_dpci->mirq[machine_gsi], 0);
         /* Deal with gsi for legacy devices */
         pirq_guest_bind(d->vcpu[0], machine_gsi, BIND_PIRQ__WILL_SHARE);
@@ -157,7 +159,7 @@ int pt_irq_destroy_bind_vtd(
         if ( list_empty(&hvm_irq_dpci->mirq[machine_gsi].digl_list) )
         {
             pirq_guest_unbind(d, machine_gsi);
-            kill_timer(&hvm_irq_dpci->hvm_timer[irq_to_vector(machine_gsi)]);
+            kill_timer(&hvm_irq_dpci->hvm_timer[domain_irq_to_vector(d, machine_gsi)]);
             hvm_irq_dpci->mirq[machine_gsi].dom   = NULL;
             hvm_irq_dpci->mirq[machine_gsi].valid = 0;
         }
@@ -185,7 +187,7 @@ int hvm_do_IRQ_dpci(struct domain *d, unsigned int mirq)
      * PIC) and we need to detect that.
      */
     set_bit(mirq, dpci->dirq_mask);
-    set_timer(&dpci->hvm_timer[irq_to_vector(mirq)],
+    set_timer(&dpci->hvm_timer[domain_irq_to_vector(d, mirq)],
               NOW() + PT_IRQ_TIME_OUT);
     vcpu_kick(d->vcpu[0]);
 
@@ -221,7 +223,7 @@ void hvm_dpci_eoi(struct domain *d, unsigned int guest_gsi,
 
         gdprintk(XENLOG_INFO VTDPREFIX,
                  "hvm_dpci_eoi:: mirq = %x\n", machine_gsi);
-        stop_timer(&hvm_irq_dpci->hvm_timer[irq_to_vector(machine_gsi)]);
+        stop_timer(&hvm_irq_dpci->hvm_timer[domain_irq_to_vector(d, machine_gsi)]);
         if ( (ent == NULL) || !ent->fields.mask )
             pirq_guest_eoi(d, machine_gsi);
     }
diff --git a/xen/drivers/passthrough/vtd/x86/vtd.c b/xen/drivers/passthrough/vtd/x86/vtd.c
index b78bce80ed..098a1b949a 100644
--- a/xen/drivers/passthrough/vtd/x86/vtd.c
+++ b/xen/drivers/passthrough/vtd/x86/vtd.c
@@ -114,7 +114,7 @@ void hvm_dpci_isairq_eoi(struct domain *d, unsigned int isairq)
                 if ( --dpci->mirq[i].pending == 0 )
                 {
                     spin_unlock(&dpci->dirq_lock);
-                    stop_timer(&dpci->hvm_timer[irq_to_vector(i)]);
+                    stop_timer(&dpci->hvm_timer[domain_irq_to_vector(d, i)]);
                     pirq_guest_eoi(d, i);
                 }
                 else
diff --git a/xen/include/asm-x86/domain.h b/xen/include/asm-x86/domain.h
index 8c4ee649b4..106a90656e 100644
--- a/xen/include/asm-x86/domain.h
+++ b/xen/include/asm-x86/domain.h
@@ -6,6 +6,7 @@
 #include <asm/hvm/vcpu.h>
 #include <asm/hvm/domain.h>
 #include <asm/e820.h>
+#include <asm/pirq.h>
 
 #define has_32bit_shinfo(d)    ((d)->arch.has_32bit_shinfo)
 #define is_pv_32bit_domain(d)  ((d)->arch.is_32bit_pv)
@@ -222,6 +223,10 @@ struct arch_domain
     /* Shadow translated domain: P2M mapping */
     pagetable_t phys_table;
 
+    spinlock_t irq_lock;
+    int vector_pirq[NR_VECTORS];
+    int pirq_vector[NR_PIRQS];
+
     /* Pseudophysical e820 map (XENMEM_memory_map).  */
     struct e820entry e820[3];
     unsigned int nr_e820;
diff --git a/xen/include/asm-x86/irq.h b/xen/include/asm-x86/irq.h
index e92e775945..8858b4eb95 100644
--- a/xen/include/asm-x86/irq.h
+++ b/xen/include/asm-x86/irq.h
@@ -49,7 +49,9 @@ extern unsigned long io_apic_irqs;
 extern atomic_t irq_err_count;
 extern atomic_t irq_mis_count;
 
-int pirq_acktype(int irq);
-int pirq_shared(int irq);
+int pirq_acktype(struct domain *d, int irq);
+int pirq_shared(struct domain *d , int irq);
 
+extern int domain_irq_to_vector(struct domain *d, int irq);
+extern int domain_vector_to_irq(struct domain *d, int vector);
 #endif /* _ASM_HW_IRQ_H */
diff --git a/xen/include/asm-x86/pirq.h b/xen/include/asm-x86/pirq.h
new file mode 100644
index 0000000000..2041262134
--- /dev/null
+++ b/xen/include/asm-x86/pirq.h
@@ -0,0 +1,11 @@
+#ifndef __XEN_PIRQ_H
+#define __XEN_PIRQ_H
+
+#define PIRQ_BASE       0
+#define NR_PIRQS        256
+
+#define DYNIRQ_BASE     (PIRQ_BASE + NR_PIRQS)
+#define NR_DYNIRQS      256
+
+#endif /* __XEN_PIRQ_H */
+
diff --git a/xen/include/public/physdev.h b/xen/include/public/physdev.h
index 75990545b2..fdf6c2bf16 100644
--- a/xen/include/public/physdev.h
+++ b/xen/include/public/physdev.h
@@ -121,6 +121,33 @@ struct physdev_irq {
 };
 typedef struct physdev_irq physdev_irq_t;
 DEFINE_XEN_GUEST_HANDLE(physdev_irq_t);
+ 
+#define MAP_PIRQ_TYPE_MSI               0x0
+#define MAP_PIRQ_TYPE_GSI               0x1
+#define MAP_PIRQ_TYPE_UNKNOWN           0x2
+
+#define PHYSDEVOP_map_pirq               13
+struct physdev_map_pirq {
+    domid_t domid;
+    /* IN */
+    int type;
+    /* IN */
+    int index;
+    /* IN or OUT */
+    int pirq;
+};
+typedef struct physdev_map_pirq physdev_map_pirq_t;
+DEFINE_XEN_GUEST_HANDLE(physdev_map_pirq_t);
+
+#define PHYSDEVOP_unmap_pirq             14
+struct physdev_unmap_pirq {
+    domid_t domid;
+    /* IN */
+    int pirq;
+};
+
+typedef struct physdev_unmap_pirq physdev_unmap_pirq_t;
+DEFINE_XEN_GUEST_HANDLE(physdev_unmap_pirq_t);
 
 /*
  * Argument to physdev_op_compat() hypercall. Superceded by new physdev_op()