xen/pci: support multi-segment systems

Now that the hypercall interface changes are in -unstable, make the
kernel side code not ignore the segment (aka domain) number anymore
(which results in pretty odd behavior on such systems). Rather, if
only the old interfaces are available, don't call them for devices on
non-zero segments at all.

Signed-off-by: Jan Beulich <jbeulich@suse.com>
[v1: Edited git description]
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>

3 files changed, 136 insertions, 14 deletions

diff --git a/arch/x86/pci/xen.c b/arch/x86/pci/xen.c
index f567965c062..265fa8814cc 100644
--- a/arch/x86/pci/xen.c
+++ b/arch/x86/pci/xen.c
@@ -185,6 +185,8 @@ static void xen_teardown_msi_irq(unsigned int irq)
 }
 
 #ifdef CONFIG_XEN_DOM0
+static bool __read_mostly pci_seg_supported = true;
+
 static int xen_initdom_setup_msi_irqs(struct pci_dev *dev, int nvec, int type)
 {
 	int ret = 0;
@@ -202,10 +204,11 @@ static int xen_initdom_setup_msi_irqs(struct pci_dev *dev, int nvec, int type)
 
 		memset(&map_irq, 0, sizeof(map_irq));
 		map_irq.domid = domid;
-		map_irq.type = MAP_PIRQ_TYPE_MSI;
+		map_irq.type = MAP_PIRQ_TYPE_MSI_SEG;
 		map_irq.index = -1;
 		map_irq.pirq = -1;
-		map_irq.bus = dev->bus->number;
+		map_irq.bus = dev->bus->number |
+			      (pci_domain_nr(dev->bus) << 16);
 		map_irq.devfn = dev->devfn;
 
 		if (type == PCI_CAP_ID_MSIX) {
@@ -222,7 +225,20 @@ static int xen_initdom_setup_msi_irqs(struct pci_dev *dev, int nvec, int type)
 			map_irq.entry_nr = msidesc->msi_attrib.entry_nr;
 		}
 
-		ret = HYPERVISOR_physdev_op(PHYSDEVOP_map_pirq, &map_irq);
+		ret = -EINVAL;
+		if (pci_seg_supported)
+			ret = HYPERVISOR_physdev_op(PHYSDEVOP_map_pirq,
+						    &map_irq);
+		if (ret == -EINVAL && !pci_domain_nr(dev->bus)) {
+			map_irq.type = MAP_PIRQ_TYPE_MSI;
+			map_irq.index = -1;
+			map_irq.pirq = -1;
+			map_irq.bus = dev->bus->number;
+			ret = HYPERVISOR_physdev_op(PHYSDEVOP_map_pirq,
+						    &map_irq);
+			if (ret != -EINVAL)
+				pci_seg_supported = false;
+		}
 		if (ret) {
 			dev_warn(&dev->dev, "xen map irq failed %d for %d domain\n",
 				 ret, domid);
diff --git a/drivers/xen/pci.c b/drivers/xen/pci.c
index 02c402b1ed8..66057075d6e 100644
--- a/drivers/xen/pci.c
+++ b/drivers/xen/pci.c
@@ -18,6 +18,7 @@
  */
 
 #include <linux/pci.h>
+#include <linux/acpi.h>
 #include <xen/xen.h>
 #include <xen/interface/physdev.h>
 #include <xen/interface/xen.h>
@@ -26,26 +27,85 @@
 #include <asm/xen/hypercall.h>
 #include "../pci/pci.h"
 
+static bool __read_mostly pci_seg_supported = true;
+
 static int xen_add_device(struct device *dev)
 {
 	int r;
 	struct pci_dev *pci_dev = to_pci_dev(dev);
+#ifdef CONFIG_PCI_IOV
+	struct pci_dev *physfn = pci_dev->physfn;
+#endif
+
+	if (pci_seg_supported) {
+		struct physdev_pci_device_add add = {
+			.seg = pci_domain_nr(pci_dev->bus),
+			.bus = pci_dev->bus->number,
+			.devfn = pci_dev->devfn
+		};
+#ifdef CONFIG_ACPI
+		acpi_handle handle;
+#endif
 
 #ifdef CONFIG_PCI_IOV
-	if (pci_dev->is_virtfn) {
+		if (pci_dev->is_virtfn) {
+			add.flags = XEN_PCI_DEV_VIRTFN;
+			add.physfn.bus = physfn->bus->number;
+			add.physfn.devfn = physfn->devfn;
+		} else
+#endif
+		if (pci_ari_enabled(pci_dev->bus) && PCI_SLOT(pci_dev->devfn))
+			add.flags = XEN_PCI_DEV_EXTFN;
+
+#ifdef CONFIG_ACPI
+		handle = DEVICE_ACPI_HANDLE(&pci_dev->dev);
+		if (!handle)
+			handle = DEVICE_ACPI_HANDLE(pci_dev->bus->bridge);
+#ifdef CONFIG_PCI_IOV
+		if (!handle && pci_dev->is_virtfn)
+			handle = DEVICE_ACPI_HANDLE(physfn->bus->bridge);
+#endif
+		if (handle) {
+			acpi_status status;
+
+			do {
+				unsigned long long pxm;
+
+				status = acpi_evaluate_integer(handle, "_PXM",
+							       NULL, &pxm);
+				if (ACPI_SUCCESS(status)) {
+					add.optarr[0] = pxm;
+					add.flags |= XEN_PCI_DEV_PXM;
+					break;
+				}
+				status = acpi_get_parent(handle, &handle);
+			} while (ACPI_SUCCESS(status));
+		}
+#endif /* CONFIG_ACPI */
+
+		r = HYPERVISOR_physdev_op(PHYSDEVOP_pci_device_add, &add);
+		if (r != -ENOSYS)
+			return r;
+		pci_seg_supported = false;
+	}
+
+	if (pci_domain_nr(pci_dev->bus))
+		r = -ENOSYS;
+#ifdef CONFIG_PCI_IOV
+	else if (pci_dev->is_virtfn) {
 		struct physdev_manage_pci_ext manage_pci_ext = {
 			.bus		= pci_dev->bus->number,
 			.devfn		= pci_dev->devfn,
 			.is_virtfn 	= 1,
-			.physfn.bus	= pci_dev->physfn->bus->number,
-			.physfn.devfn	= pci_dev->physfn->devfn,
+			.physfn.bus	= physfn->bus->number,
+			.physfn.devfn	= physfn->devfn,
 		};
 
 		r = HYPERVISOR_physdev_op(PHYSDEVOP_manage_pci_add_ext,
 			&manage_pci_ext);
-	} else
+	}
 #endif
-	if (pci_ari_enabled(pci_dev->bus) && PCI_SLOT(pci_dev->devfn)) {
+	else if (pci_ari_enabled(pci_dev->bus) && PCI_SLOT(pci_dev->devfn)) {
 		struct physdev_manage_pci_ext manage_pci_ext = {
 			.bus		= pci_dev->bus->number,
 			.devfn		= pci_dev->devfn,
@@ -71,13 +131,27 @@ static int xen_remove_device(struct device *dev)
 {
 	int r;
 	struct pci_dev *pci_dev = to_pci_dev(dev);
-	struct physdev_manage_pci manage_pci;
 
-	manage_pci.bus = pci_dev->bus->number;
-	manage_pci.devfn = pci_dev->devfn;
+	if (pci_seg_supported) {
+		struct physdev_pci_device device = {
+			.seg = pci_domain_nr(pci_dev->bus),
+			.bus = pci_dev->bus->number,
+			.devfn = pci_dev->devfn
+		};
 
-	r = HYPERVISOR_physdev_op(PHYSDEVOP_manage_pci_remove,
-		&manage_pci);
+		r = HYPERVISOR_physdev_op(PHYSDEVOP_pci_device_remove,
+					  &device);
+	} else if (pci_domain_nr(pci_dev->bus))
+		r = -ENOSYS;
+	else {
+		struct physdev_manage_pci manage_pci = {
+			.bus = pci_dev->bus->number,
+			.devfn = pci_dev->devfn
+		};
+
+		r = HYPERVISOR_physdev_op(PHYSDEVOP_manage_pci_remove,
+					  &manage_pci);
+	}
 
 	return r;
 }
diff --git a/include/xen/interface/physdev.h b/include/xen/interface/physdev.h
index 534cac89a77..c1080d9c705 100644
--- a/include/xen/interface/physdev.h
+++ b/include/xen/interface/physdev.h
@@ -109,6 +109,7 @@ struct physdev_irq {
 #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 {
@@ -119,7 +120,7 @@ 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;
@@ -198,6 +199,37 @@ struct physdev_get_free_pirq {
     uint32_t pirq;
 };
 
+#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
+};
+
+#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;
+};
+
 /*
  * Notify that some PIRQ-bound event channels have been unmasked.
  * ** This command is obsolete since interface version 0x00030202 and is **