PCI multi-seg: add new physdevop-s

The new PHYSDEVOP_pci_device_add is intended to be extensible, with a
first extension (to pass the proximity domain of a device) added right
away.

A couple of directly related functions at once get adjusted to account
for the segment number.

Should we deprecate the PHYSDEVOP_manage_pci_* sub-hypercalls?

Signed-off-by: Jan Beulich <jbeulich@suse.com>

xen/arch/ia64/xen/hypercall.c

@@ -665,7 +665,7 @@ long do_physdev_op(int cmd, XEN_GUEST_HANDLE(void) arg)
         if ( copy_from_guest(&manage_pci, arg, 1) != 0 )
             break;
 
-        ret = pci_add_device(manage_pci.bus, manage_pci.devfn, NULL);
+        ret = pci_add_device(0, manage_pci.bus, manage_pci.devfn, NULL);
         break;
     }
 
@@ -678,7 +678,7 @@ long do_physdev_op(int cmd, XEN_GUEST_HANDLE(void) arg)
         if ( copy_from_guest(&manage_pci, arg, 1) != 0 )
             break;
 
-        ret = pci_remove_device(manage_pci.bus, manage_pci.devfn);
+        ret = pci_remove_device(0, manage_pci.bus, manage_pci.devfn);
             break;
     }
 
@@ -698,7 +698,7 @@ long do_physdev_op(int cmd, XEN_GUEST_HANDLE(void) arg)
         pdev_info.is_virtfn = manage_pci_ext.is_virtfn;
         pdev_info.physfn.bus = manage_pci_ext.physfn.bus;
         pdev_info.physfn.devfn = manage_pci_ext.physfn.devfn;
-        ret = pci_add_device(manage_pci_ext.bus,
+        ret = pci_add_device(0, manage_pci_ext.bus,
                              manage_pci_ext.devfn,
                              &pdev_info);
         break;

xen/arch/x86/irq.c

@@ -1716,7 +1716,7 @@ int map_domain_pirq(
         if ( !cpu_has_apic )
             goto done;
 
-        pdev = pci_get_pdev(msi->bus, msi->devfn);
+        pdev = pci_get_pdev(msi->seg, msi->bus, msi->devfn);
         ret = pci_enable_msi(msi, &msi_desc);
         if ( ret )
             goto done;

xen/arch/x86/msi.c

@@ -528,7 +528,7 @@ static u64 read_pci_mem_bar(u8 bus, u8 slot, u8 func, u8 bir, int vf)
 
     if ( vf >= 0 )
     {
-        struct pci_dev *pdev = pci_get_pdev(bus, PCI_DEVFN(slot, func));
+        struct pci_dev *pdev = pci_get_pdev(0, bus, PCI_DEVFN(slot, func));
         unsigned int pos = pci_find_ext_capability(0, bus,
                                                    PCI_DEVFN(slot, func),
                                                    PCI_EXT_CAP_ID_SRIOV);
@@ -767,7 +767,7 @@ static int __pci_enable_msi(struct msi_info *msi, struct msi_desc **desc)
     struct msi_desc *old_desc;
 
     ASSERT(spin_is_locked(&pcidevs_lock));
-    pdev = pci_get_pdev(msi->bus, msi->devfn);
+    pdev = pci_get_pdev(msi->seg, msi->bus, msi->devfn);
     if ( !pdev )
         return -ENODEV;
 
@@ -775,7 +775,8 @@ static int __pci_enable_msi(struct msi_info *msi, struct msi_desc **desc)
     if ( old_desc )
     {
         dprintk(XENLOG_WARNING, "irq %d has already mapped to MSI on "
-                "device %02x:%02x.%01x.\n", msi->irq, msi->bus,
+                "device %04x:%02x:%02x.%01x\n",
+                msi->irq, msi->seg, msi->bus,
                 PCI_SLOT(msi->devfn), PCI_FUNC(msi->devfn));
         *desc = old_desc;
         return 0;
@@ -785,7 +786,7 @@ static int __pci_enable_msi(struct msi_info *msi, struct msi_desc **desc)
     if ( old_desc )
     {
         dprintk(XENLOG_WARNING, "MSI-X is already in use on "
-                "device %02x:%02x.%01x\n", msi->bus,
+                "device %04x:%02x:%02x.%01x\n", msi->seg, msi->bus,
                 PCI_SLOT(msi->devfn), PCI_FUNC(msi->devfn));
         pci_disable_msi(old_desc);
     }
@@ -830,7 +831,7 @@ static int __pci_enable_msix(struct msi_info *msi, struct msi_desc **desc)
     struct msi_desc *old_desc;
 
     ASSERT(spin_is_locked(&pcidevs_lock));
-    pdev = pci_get_pdev(msi->bus, msi->devfn);
+    pdev = pci_get_pdev(msi->seg, msi->bus, msi->devfn);
     if ( !pdev )
         return -ENODEV;
 
@@ -844,7 +845,8 @@ static int __pci_enable_msix(struct msi_info *msi, struct msi_desc **desc)
     if ( old_desc )
     {
         dprintk(XENLOG_WARNING, "irq %d has already mapped to MSIX on "
-                "device %02x:%02x.%01x.\n", msi->irq, msi->bus,
+                "device %04x:%02x:%02x.%01x\n",
+                msi->irq, msi->seg, msi->bus,
                 PCI_SLOT(msi->devfn), PCI_FUNC(msi->devfn));
         *desc = old_desc;
         return 0;
@@ -854,7 +856,7 @@ static int __pci_enable_msix(struct msi_info *msi, struct msi_desc **desc)
     if ( old_desc )
     {
         dprintk(XENLOG_WARNING, "MSI is already in use on "
-                "device %02x:%02x.%01x\n", msi->bus,
+                "device %04x:%02x:%02x.%01x\n", msi->seg, msi->bus,
                 PCI_SLOT(msi->devfn), PCI_FUNC(msi->devfn));
         pci_disable_msi(old_desc);
 
@@ -962,8 +964,10 @@ int pci_restore_msi_state(struct pci_dev *pdev)
 
         if (desc->msi_desc != entry)
         {
-            dprintk(XENLOG_ERR, "Restore MSI for dev %x:%x not set before?\n",
+            dprintk(XENLOG_ERR,
-                                pdev->bus, pdev->devfn);
+                    "Restore MSI for dev %04x:%02x:%02x:%x not set before?\n",
+                    pdev->seg, pdev->bus, PCI_SLOT(pdev->devfn),
+                    PCI_FUNC(pdev->devfn));
             spin_unlock_irqrestore(&desc->lock, flags);
             return -EINVAL;
         }

xen/arch/x86/physdev.c

@@ -357,6 +357,15 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_HANDLE(void) arg)
         if ( copy_from_guest(&map, arg, 1) != 0 )
             break;
 
+        if ( map.type == MAP_PIRQ_TYPE_MSI_SEG )
+        {
+            map.type = MAP_PIRQ_TYPE_MSI;
+            msi.seg = map.bus >> 16;
+        }
+        else
+        {
+            msi.seg = 0;
+        }
         msi.bus = map.bus;
         msi.devfn = map.devfn;
         msi.entry_nr = map.entry_nr;
@@ -480,7 +489,7 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_HANDLE(void) arg)
         if ( copy_from_guest(&manage_pci, arg, 1) != 0 )
             break;
 
-        ret = pci_add_device(manage_pci.bus, manage_pci.devfn, NULL);
+        ret = pci_add_device(0, manage_pci.bus, manage_pci.devfn, NULL);
         break;
     }
 
@@ -493,7 +502,7 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_HANDLE(void) arg)
         if ( copy_from_guest(&manage_pci, arg, 1) != 0 )
             break;
 
-        ret = pci_remove_device(manage_pci.bus, manage_pci.devfn);
+        ret = pci_remove_device(0, manage_pci.bus, manage_pci.devfn);
         break;
     }
 
@@ -517,12 +526,52 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_HANDLE(void) arg)
         pdev_info.is_virtfn = manage_pci_ext.is_virtfn;
         pdev_info.physfn.bus = manage_pci_ext.physfn.bus;
         pdev_info.physfn.devfn = manage_pci_ext.physfn.devfn;
-        ret = pci_add_device(manage_pci_ext.bus,
+        ret = pci_add_device(0, manage_pci_ext.bus,
                              manage_pci_ext.devfn,
                              &pdev_info);
         break;
     }
 
+    case PHYSDEVOP_pci_device_add: {
+        struct physdev_pci_device_add add;
+        struct pci_dev_info pdev_info;
+
+        ret = -EPERM;
+        if ( !IS_PRIV(current->domain) )
+            break;
+
+        ret = -EFAULT;
+        if ( copy_from_guest(&add, arg, 1) != 0 )
+            break;
+
+        pdev_info.is_extfn = !!(add.flags & XEN_PCI_DEV_EXTFN);
+        if ( add.flags & XEN_PCI_DEV_VIRTFN )
+        {
+            pdev_info.is_virtfn = 1;
+            pdev_info.physfn.bus = add.physfn.bus;
+            pdev_info.physfn.devfn = add.physfn.devfn;
+        }
+        else
+            pdev_info.is_virtfn = 0;
+        ret = pci_add_device(add.seg, add.bus, add.devfn, &pdev_info);
+        break;
+    }
+
+    case PHYSDEVOP_pci_device_remove: {
+        struct physdev_pci_device dev;
+
+        ret = -EPERM;
+        if ( !IS_PRIV(v->domain) )
+            break;
+
+        ret = -EFAULT;
+        if ( copy_from_guest(&dev, arg, 1) != 0 )
+            break;
+
+        ret = pci_remove_device(dev.seg, dev.bus, dev.devfn);
+        break;
+    }
+
 #ifdef __x86_64__
     case PHYSDEVOP_pci_mmcfg_reserved: {
         struct physdev_pci_mmcfg_reserved info;
@@ -554,11 +603,31 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_HANDLE(void) arg)
             break;
 
         spin_lock(&pcidevs_lock);
-        pdev = pci_get_pdev(restore_msi.bus, restore_msi.devfn);
+        pdev = pci_get_pdev(0, restore_msi.bus, restore_msi.devfn);
         ret = pdev ? pci_restore_msi_state(pdev) : -ENODEV;
         spin_unlock(&pcidevs_lock);
         break;
     }
+
+    case PHYSDEVOP_restore_msi_ext: {
+        struct physdev_pci_device dev;
+        struct pci_dev *pdev;
+
+        ret = -EPERM;
+        if ( !IS_PRIV(v->domain) )
+            break;
+
+        ret = -EFAULT;
+        if ( copy_from_guest(&dev, arg, 1) != 0 )
+            break;
+
+        spin_lock(&pcidevs_lock);
+        pdev = pci_get_pdev(dev.seg, dev.bus, dev.devfn);
+        ret = pdev ? pci_restore_msi_state(pdev) : -ENODEV;
+        spin_unlock(&pcidevs_lock);
+        break;
+    }
+
     case PHYSDEVOP_setup_gsi: {
         struct physdev_setup_gsi setup_gsi;
 

xen/arch/x86/x86_64/physdev.c

@@ -67,6 +67,14 @@ CHECK_physdev_get_free_pirq;
 CHECK_physdev_pci_mmcfg_reserved;
 #undef xen_physdev_pci_mmcfg_reserved
 
+#define xen_physdev_pci_device_add physdev_pci_device_add
+CHECK_physdev_pci_device_add
+#undef xen_physdev_pci_device_add
+
+#define xen_physdev_pci_device physdev_pci_device
+CHECK_physdev_pci_device
+#undef xen_physdev_pci_device
+
 #define COMPAT
 #undef guest_handle_okay
 #define guest_handle_okay          compat_handle_okay

xen/drivers/passthrough/amd/pci_amd_iommu.c

@@ -131,7 +131,7 @@ static void __init amd_iommu_setup_dom0_devices(struct domain *d)
     {
         for ( devfn = 0; devfn < 256; devfn++ )
         {
-            pdev = pci_get_pdev(bus, devfn);
+            pdev = pci_get_pdev(0, bus, devfn);
             if ( !pdev )
                 continue;
 
@@ -313,7 +313,7 @@ static int reassign_device( struct domain *source, struct domain *target,
     struct hvm_iommu *t = domain_hvm_iommu(target);
 
     ASSERT(spin_is_locked(&pcidevs_lock));
-    pdev = pci_get_pdev_by_domain(source, bus, devfn);
+    pdev = pci_get_pdev_by_domain(source, 0, bus, devfn);
     if ( !pdev )
         return -ENODEV;
 

xen/drivers/passthrough/iommu.c

@@ -282,7 +282,7 @@ int deassign_device(struct domain *d, u8 bus, u8 devfn)
         return -EINVAL;
 
     ASSERT(spin_is_locked(&pcidevs_lock));
-    pdev = pci_get_pdev(bus, devfn);
+    pdev = pci_get_pdev(0, bus, devfn);
     if ( !pdev )
         return -ENODEV;
 

xen/drivers/passthrough/pci.c

@@ -121,6 +121,7 @@ static struct pci_dev *alloc_pdev(struct pci_seg *pseg, u8 bus, u8 devfn)
         return NULL;
     memset(pdev, 0, sizeof(struct pci_dev));
 
+    *(u16*) &pdev->seg = pseg->nr;
     *((u8*) &pdev->bus) = bus;
     *((u8*) &pdev->devfn) = devfn;
     pdev->domain = NULL;
@@ -137,41 +138,59 @@ static void free_pdev(struct pci_dev *pdev)
     xfree(pdev);
 }
 
-struct pci_dev *pci_get_pdev(int bus, int devfn)
+struct pci_dev *pci_get_pdev(int seg, int bus, int devfn)
 {
-    struct pci_seg *pseg = get_pseg(0);
+    struct pci_seg *pseg = get_pseg(seg);
     struct pci_dev *pdev = NULL;
 
     ASSERT(spin_is_locked(&pcidevs_lock));
+    ASSERT(seg != -1 || bus == -1);
+    ASSERT(bus != -1 || devfn == -1);
 
     if ( !pseg )
-        return NULL;
+    {
+        if ( seg == -1 )
+            radix_tree_gang_lookup(&pci_segments, (void **)&pseg, 0, 1);
+        if ( !pseg )
+            return NULL;
+    }
 
-    list_for_each_entry ( pdev, &pseg->alldevs_list, alldevs_list )
+    do {
-        if ( (pdev->bus == bus || bus == -1) &&
+        list_for_each_entry ( pdev, &pseg->alldevs_list, alldevs_list )
-             (pdev->devfn == devfn || devfn == -1) )
+            if ( (pdev->bus == bus || bus == -1) &&
-        {
+                 (pdev->devfn == devfn || devfn == -1) )
-            return pdev;
+                return pdev;
-        }
+    } while ( radix_tree_gang_lookup(&pci_segments, (void **)&pseg,
+                                     pseg->nr + 1, 1) );
 
     return NULL;
 }
 
-struct pci_dev *pci_get_pdev_by_domain(struct domain *d, int bus, int devfn)
+struct pci_dev *pci_get_pdev_by_domain(
+    struct domain *d, int seg, int bus, int devfn)
 {
-    struct pci_seg *pseg = get_pseg(0);
+    struct pci_seg *pseg = get_pseg(seg);
     struct pci_dev *pdev = NULL;
 
-    if ( !pseg )
+    ASSERT(seg != -1 || bus == -1);
-        return NULL;
+    ASSERT(bus != -1 || devfn == -1);
 
-    list_for_each_entry ( pdev, &pseg->alldevs_list, alldevs_list )
+    if ( !pseg )
-         if ( (pdev->bus == bus || bus == -1) &&
+    {
-              (pdev->devfn == devfn || devfn == -1) &&
+        if ( seg == -1 )
-              (pdev->domain == d) )
+            radix_tree_gang_lookup(&pci_segments, (void **)&pseg, 0, 1);
-         {
+        if ( !pseg )
-             return pdev;
+            return NULL;
-         }
+    }
+
+    do {
+        list_for_each_entry ( pdev, &pseg->alldevs_list, alldevs_list )
+            if ( (pdev->bus == bus || bus == -1) &&
+                 (pdev->devfn == devfn || devfn == -1) &&
+                 (pdev->domain == d) )
+                return pdev;
+    } while ( radix_tree_gang_lookup(&pci_segments, (void **)&pseg,
+                                     pseg->nr + 1, 1) );
 
     return NULL;
 }
@@ -215,7 +234,7 @@ void pci_enable_acs(struct pci_dev *pdev)
     pci_conf_write16(bus, dev, func, pos + PCI_ACS_CTRL, ctrl);
 }
 
-int pci_add_device(u8 bus, u8 devfn, const struct pci_dev_info *info)
+int pci_add_device(u16 seg, u8 bus, u8 devfn, const struct pci_dev_info *info)
 {
     struct pci_seg *pseg;
     struct pci_dev *pdev;
@@ -230,17 +249,20 @@ int pci_add_device(u8 bus, u8 devfn, const struct pci_dev_info *info)
     else if (info->is_virtfn)
     {
         spin_lock(&pcidevs_lock);
-        pdev = pci_get_pdev(info->physfn.bus, info->physfn.devfn);
+        pdev = pci_get_pdev(seg, info->physfn.bus, info->physfn.devfn);
         spin_unlock(&pcidevs_lock);
         if ( !pdev )
-            pci_add_device(info->physfn.bus, info->physfn.devfn, NULL);
+            pci_add_device(seg, info->physfn.bus, info->physfn.devfn, NULL);
         pdev_type = "virtual function";
     }
     else
-        return -EINVAL;
+    {
+        info = NULL;
+        pdev_type = "device";
+    }
 
     spin_lock(&pcidevs_lock);
-    pseg = alloc_pseg(0);
+    pseg = alloc_pseg(seg);
     if ( !pseg )
         goto out;
     pdev = alloc_pdev(pseg, bus, devfn);
@@ -251,7 +273,7 @@ int pci_add_device(u8 bus, u8 devfn, const struct pci_dev_info *info)
         pdev->info = *info;
     else if ( !pdev->vf_rlen[0] )
     {
-        unsigned int pos = pci_find_ext_capability(0, bus, devfn,
+        unsigned int pos = pci_find_ext_capability(seg, bus, devfn,
                                                    PCI_EXT_CAP_ID_SRIOV);
         u16 ctrl = pci_conf_read16(bus, slot, func, pos + PCI_SRIOV_CTRL);
 
@@ -271,9 +293,10 @@ int pci_add_device(u8 bus, u8 devfn, const struct pci_dev_info *info)
                 if ( (bar & PCI_BASE_ADDRESS_SPACE) ==
                      PCI_BASE_ADDRESS_SPACE_IO )
                 {
-                    printk(XENLOG_WARNING "SR-IOV device %02x:%02x.%x with vf"
+                    printk(XENLOG_WARNING
-                                          " BAR%u in IO space\n",
+                           "SR-IOV device %04x:%02x:%02x.%u with vf BAR%u"
-                           bus, slot, func, i);
+                           " in IO space\n",
+                           seg, bus, slot, func, i);
                     continue;
                 }
                 pci_conf_write32(bus, slot, func, idx, ~0);
@@ -282,9 +305,10 @@ int pci_add_device(u8 bus, u8 devfn, const struct pci_dev_info *info)
                 {
                     if ( i >= PCI_SRIOV_NUM_BARS )
                     {
-                        printk(XENLOG_WARNING "SR-IOV device %02x:%02x.%x with"
+                        printk(XENLOG_WARNING
-                                              " 64-bit vf BAR in last slot\n",
+                               "SR-IOV device %04x:%02x:%02x.%u with 64-bit"
-                               bus, slot, func);
+                               " vf BAR in last slot\n",
+                               seg, bus, slot, func);
                         break;
                     }
                     hi = pci_conf_read32(bus, slot, func, idx + 4);
@@ -309,9 +333,10 @@ int pci_add_device(u8 bus, u8 devfn, const struct pci_dev_info *info)
             }
         }
         else
-            printk(XENLOG_WARNING "SR-IOV device %02x:%02x.%x has its virtual"
+            printk(XENLOG_WARNING
-                                  " functions already enabled (%04x)\n",
+                   "SR-IOV device %04x:%02x:%02x.%u has its virtual"
-                   bus, slot, func, ctrl);
+                   " functions already enabled (%04x)\n",
+                   seg, bus, slot, func, ctrl);
     }
 
     ret = 0;
@@ -331,14 +356,14 @@ int pci_add_device(u8 bus, u8 devfn, const struct pci_dev_info *info)
 
 out:
     spin_unlock(&pcidevs_lock);
-    printk(XENLOG_DEBUG "PCI add %s %02x:%02x.%x\n", pdev_type,
+    printk(XENLOG_DEBUG "PCI add %s %04x:%02x:%02x.%u\n", pdev_type,
-           bus, slot, func);
+           seg, bus, slot, func);
     return ret;
 }
 
-int pci_remove_device(u8 bus, u8 devfn)
+int pci_remove_device(u16 seg, u8 bus, u8 devfn)
 {
-    struct pci_seg *pseg = get_pseg(0);
+    struct pci_seg *pseg = get_pseg(seg);
     struct pci_dev *pdev;
     int ret = -ENODEV;
 
@@ -354,8 +379,8 @@ int pci_remove_device(u8 bus, u8 devfn)
                 list_del(&pdev->domain_list);
             pci_cleanup_msi(pdev);
             free_pdev(pdev);
-            printk(XENLOG_DEBUG "PCI remove device %02x:%02x.%x\n", bus,
+            printk(XENLOG_DEBUG "PCI remove device %04x:%02x:%02x.%u\n",
-                   PCI_SLOT(devfn), PCI_FUNC(devfn));
+                   seg, bus, PCI_SLOT(devfn), PCI_FUNC(devfn));
             break;
         }
 
@@ -413,7 +438,7 @@ void pci_release_devices(struct domain *d)
 
     spin_lock(&pcidevs_lock);
     pci_clean_dpci_irqs(d);
-    while ( (pdev = pci_get_pdev_by_domain(d, -1, -1)) )
+    while ( (pdev = pci_get_pdev_by_domain(d, -1, -1, -1)) )
     {
         pci_cleanup_msi(pdev);
         bus = pdev->bus; devfn = pdev->devfn;

xen/drivers/passthrough/vtd/iommu.c

@@ -280,7 +280,7 @@ static u64 addr_to_dma_page_maddr(struct domain *domain, u64 addr, int alloc)
          * just get any passthrough device in the domainr - assume user
          * assigns only devices from same node to a given guest.
          */
-        pdev = pci_get_pdev_by_domain(domain, -1, -1);
+        pdev = pci_get_pdev_by_domain(domain, -1, -1, -1);
         drhd = acpi_find_matched_drhd_unit(pdev);
         if ( !alloc || ((hd->pgd_maddr = alloc_pgtable_maddr(drhd, 1)) == 0) )
             goto out;
@@ -297,7 +297,7 @@ static u64 addr_to_dma_page_maddr(struct domain *domain, u64 addr, int alloc)
             if ( !alloc )
                 break;
 
-            pdev = pci_get_pdev_by_domain(domain, -1, -1);
+            pdev = pci_get_pdev_by_domain(domain, -1, -1, -1);
             drhd = acpi_find_matched_drhd_unit(pdev);
             maddr = alloc_pgtable_maddr(drhd, 1);
             if ( !maddr )
@@ -1273,7 +1273,7 @@ int domain_context_mapping_one(
 
         /* First try to get domain ownership from device structure.  If that's
          * not available, try to read it from the context itself. */
-        pdev = pci_get_pdev(bus, devfn);
+        pdev = pci_get_pdev(0, bus, devfn);
         if ( pdev )
         {
             if ( pdev->domain != domain )
@@ -1396,7 +1396,7 @@ static int domain_context_mapping(struct domain *domain, u8 bus, u8 devfn)
     int ret = 0;
     u32 type;
     u8 secbus;
-    struct pci_dev *pdev = pci_get_pdev(bus, devfn);
+    struct pci_dev *pdev = pci_get_pdev(0, bus, devfn);
 
     drhd = acpi_find_matched_drhd_unit(pdev);
     if ( !drhd )
@@ -1521,7 +1521,7 @@ static int domain_context_unmap(struct domain *domain, u8 bus, u8 devfn)
     int ret = 0;
     u32 type;
     u8 tmp_bus, tmp_devfn, secbus;
-    struct pci_dev *pdev = pci_get_pdev(bus, devfn);
+    struct pci_dev *pdev = pci_get_pdev(0, bus, devfn);
     int found = 0;
 
     BUG_ON(!pdev);
@@ -1632,7 +1632,7 @@ static int reassign_device_ownership(
     int ret;
 
     ASSERT(spin_is_locked(&pcidevs_lock));
-    pdev = pci_get_pdev_by_domain(source, bus, devfn);
+    pdev = pci_get_pdev_by_domain(source, 0, bus, devfn);
 
     if (!pdev)
         return -ENODEV;
@@ -1941,7 +1941,7 @@ static void __init setup_dom0_devices(struct domain *d)
     {
         for ( devfn = 0; devfn < 256; devfn++ )
         {
-            pdev = pci_get_pdev(bus, devfn);
+            pdev = pci_get_pdev(0, bus, devfn);
             if ( !pdev )
                 continue;
 
@@ -2175,7 +2175,7 @@ int device_assigned(u8 bus, u8 devfn)
     struct pci_dev *pdev;
 
     spin_lock(&pcidevs_lock);
-    pdev = pci_get_pdev_by_domain(dom0, bus, devfn);
+    pdev = pci_get_pdev_by_domain(dom0, 0, bus, devfn);
     if (!pdev)
     {
         spin_unlock(&pcidevs_lock);
@@ -2197,7 +2197,7 @@ static int intel_iommu_assign_device(struct domain *d, u8 bus, u8 devfn)
         return -ENODEV;
 
     ASSERT(spin_is_locked(&pcidevs_lock));
-    pdev = pci_get_pdev(bus, devfn);
+    pdev = pci_get_pdev(0, bus, devfn);
     if (!pdev)
         return -ENODEV;
 

xen/drivers/passthrough/vtd/quirks.c

@@ -286,7 +286,7 @@ static void map_me_phantom_function(struct domain *domain, u32 dev, int map)
     struct pci_dev *pdev;
 
     /* find ME VT-d engine base on a real ME device */
-    pdev = pci_get_pdev(0, PCI_DEVFN(dev, 0));
+    pdev = pci_get_pdev(0, 0, PCI_DEVFN(dev, 0));
     drhd = acpi_find_matched_drhd_unit(pdev);
 
     /* map or unmap ME phantom function */

xen/drivers/passthrough/vtd/x86/ats.c

@@ -37,6 +37,7 @@ static LIST_HEAD(ats_dev_drhd_units);
 
 struct pci_ats_dev {
     struct list_head list;
+    u16 seg;
     u8 bus;
     u8 devfn;
     u16 ats_queue_depth;    /* ATS device invalidation queue depth */
@@ -91,7 +92,7 @@ int ats_device(int seg, int bus, int devfn)
     if ( !ats_enabled || !iommu_qinval )
         return 0;
 
-    pdev = pci_get_pdev(bus, devfn);
+    pdev = pci_get_pdev(seg, bus, devfn);
     if ( !pdev )
         return 0;
 
@@ -130,8 +131,9 @@ int enable_ats_device(int seg, int bus, int devfn)
     BUG_ON(!pos);
 
     if ( iommu_verbose )
-        dprintk(XENLOG_INFO VTDPREFIX, "%x:%x.%x: ATS capability found\n",
+        dprintk(XENLOG_INFO VTDPREFIX,
-                bus, PCI_SLOT(devfn), PCI_FUNC(devfn));
+                "%04x:%02x:%02x.%u: ATS capability found\n",
+                seg, bus, PCI_SLOT(devfn), PCI_FUNC(devfn));
 
     /* BUGBUG: add back seg when multi-seg platform support is enabled */
     value = pci_conf_read16(bus, PCI_SLOT(devfn),
@@ -140,7 +142,7 @@ int enable_ats_device(int seg, int bus, int devfn)
     {
         list_for_each_entry ( pdev, &ats_devices, list )
         {
-            if ( pdev->bus == bus && pdev->devfn == devfn )
+            if ( pdev->seg == seg && pdev->bus == bus && pdev->devfn == devfn )
             {
                 pos = 0;
                 break;
@@ -161,6 +163,7 @@ int enable_ats_device(int seg, int bus, int devfn)
 
     if ( pos )
     {
+        pdev->seg = seg;
         pdev->bus = bus;
         pdev->devfn = devfn;
         value = pci_conf_read16(bus, PCI_SLOT(devfn),
@@ -170,8 +173,10 @@ int enable_ats_device(int seg, int bus, int devfn)
     }
 
     if ( iommu_verbose )
-        dprintk(XENLOG_INFO VTDPREFIX, "%x:%x.%x: ATS %s enabled\n",
+        dprintk(XENLOG_INFO VTDPREFIX,
-                bus, PCI_SLOT(devfn), PCI_FUNC(devfn), pos ? "is" : "was");
+                "%04x:%02x:%02x.%u: ATS %s enabled\n",
+                seg, bus, PCI_SLOT(devfn), PCI_FUNC(devfn),
+                pos ? "is" : "was");
 
     return pos;
 }
@@ -194,7 +199,7 @@ void disable_ats_device(int seg, int bus, int devfn)
 
     list_for_each_entry ( pdev, &ats_devices, list )
     {
-        if ( pdev->bus == bus && pdev->devfn == devfn )
+        if ( pdev->seg == seg && pdev->bus == bus && pdev->devfn == devfn )
         {
             list_del(&pdev->list);
             xfree(pdev);
@@ -203,8 +208,9 @@ void disable_ats_device(int seg, int bus, int devfn)
     }
 
     if ( iommu_verbose )
-        dprintk(XENLOG_INFO VTDPREFIX, "%x:%x.%x: ATS is disabled\n",
+        dprintk(XENLOG_INFO VTDPREFIX,
-                bus, PCI_SLOT(devfn), PCI_FUNC(devfn));
+                "%04x:%02x:%02x.%u: ATS is disabled\n",
+                seg, bus, PCI_SLOT(devfn), PCI_FUNC(devfn));
 }
 
 

xen/include/asm-x86/msi.h

@@ -59,8 +59,9 @@
 #endif
 
 struct msi_info {
-    int bus;
+    u16 seg;
-    int devfn;
+    u8 bus;
+    u8 devfn;
     int irq;
     int entry_nr;
     uint64_t table_base;

xen/include/public/physdev.h

@@ -142,6 +142,7 @@ 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 MAP_PIRQ_TYPE_MSI_SEG           0x3
 
 #define PHYSDEVOP_map_pirq               13
 struct physdev_map_pirq {
@@ -152,7 +153,7 @@ struct physdev_map_pirq {
     int index;
     /* IN or OUT */
     int pirq;
-    /* IN */
+    /* IN - high 16 bits hold segment for MAP_PIRQ_TYPE_MSI_SEG */
     int bus;
     /* IN */
     int devfn;
@@ -268,6 +269,41 @@ struct physdev_pci_mmcfg_reserved {
 typedef struct physdev_pci_mmcfg_reserved physdev_pci_mmcfg_reserved_t;
 DEFINE_XEN_GUEST_HANDLE(physdev_pci_mmcfg_reserved_t);
 
+#define XEN_PCI_DEV_EXTFN              0x1
+#define XEN_PCI_DEV_VIRTFN             0x2
+#define XEN_PCI_DEV_PXM                0x4
+
+#define PHYSDEVOP_pci_device_add        25
+struct physdev_pci_device_add {
+    /* IN */
+    uint16_t seg;
+    uint8_t bus;
+    uint8_t devfn;
+    uint32_t flags;
+    struct {
+        uint8_t bus;
+        uint8_t devfn;
+    } physfn;
+#if defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
+    uint32_t optarr[];
+#elif defined(__GNUC__)
+    uint32_t optarr[0];
+#endif
+};
+typedef struct physdev_pci_device_add physdev_pci_device_add_t;
+DEFINE_XEN_GUEST_HANDLE(physdev_pci_device_add_t);
+
+#define PHYSDEVOP_pci_device_remove     26
+#define PHYSDEVOP_restore_msi_ext       27
+struct physdev_pci_device {
+    /* IN */
+    uint16_t seg;
+    uint8_t bus;
+    uint8_t devfn;
+};
+typedef struct physdev_pci_device physdev_pci_device_t;
+DEFINE_XEN_GUEST_HANDLE(physdev_pci_device_t);
+
 /*
  * Notify that some PIRQ-bound event channels have been unmasked.
  * ** This command is obsolete since interface version 0x00030202 and is **

xen/include/xen/pci.h

@@ -56,6 +56,7 @@ struct pci_dev {
     spinlock_t msix_table_lock;
 
     struct domain *domain;
+    const u16 seg;
     const u8 bus;
     const u8 devfn;
     struct pci_dev_info info;
@@ -90,10 +91,11 @@ struct pci_dev *pci_lock_domain_pdev(struct domain *d, int bus, int devfn);
 
 void pci_release_devices(struct domain *d);
 int pci_add_segment(u16 seg);
-int pci_add_device(u8 bus, u8 devfn, const struct pci_dev_info *);
+int pci_add_device(u16 seg, u8 bus, u8 devfn, const struct pci_dev_info *);
-int pci_remove_device(u8 bus, u8 devfn);
+int pci_remove_device(u16 seg, u8 bus, u8 devfn);
-struct pci_dev *pci_get_pdev(int bus, int devfn);
+struct pci_dev *pci_get_pdev(int seg, int bus, int devfn);
-struct pci_dev *pci_get_pdev_by_domain(struct domain *d, int bus, int devfn);
+struct pci_dev *pci_get_pdev_by_domain(
+    struct domain *, int seg, int bus, int devfn);
 
 void disconnect_pci_devices(void);
 

xen/include/xlat.lst

@@ -65,6 +65,8 @@
 ?	physdev_irq_status_query	physdev.h
 ?	physdev_manage_pci		physdev.h
 ?	physdev_manage_pci_ext		physdev.h
+?	physdev_pci_device		physdev.h
+?	physdev_pci_device_add		physdev.h
 ?	physdev_pci_mmcfg_reserved	physdev.h
 ?	physdev_unmap_pirq		physdev.h
 ?	physdev_restore_msi		physdev.h