libxl: Add interface for querying hypervisor about PCI topology

.. and use this new interface to display it along with CPU topology
and NUMA information when 'xl info -n' command is issued

The output will look like
...
cpu_topology           :
cpu:    core    socket     node
  0:       0        0        0
...
device topology        :
device           node
0000:00:00.0      0
0000:00:01.0      0
...

Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>

tools/libxc/include/xenctrl.h

@@ -1249,6 +1249,7 @@ typedef xen_sysctl_physinfo_t xc_physinfo_t;
 typedef xen_sysctl_cputopo_t xc_cputopo_t;
 typedef xen_sysctl_numainfo_t xc_numainfo_t;
 typedef xen_sysctl_meminfo_t xc_meminfo_t;
+typedef xen_sysctl_pcitopoinfo_t xc_pcitopoinfo_t;
 
 typedef uint32_t xc_cpu_to_node_t;
 typedef uint32_t xc_cpu_to_socket_t;
@@ -1262,6 +1263,8 @@ int xc_cputopoinfo(xc_interface *xch, unsigned *max_cpus,
                    xc_cputopo_t *cputopo);
 int xc_numainfo(xc_interface *xch, unsigned *max_nodes,
                 xc_meminfo_t *meminfo, uint32_t *distance);
+int xc_pcitopoinfo(xc_interface *xch, unsigned num_devs,
+                   physdev_pci_device_t *devs, uint32_t *nodes);
 
 int xc_sched_id(xc_interface *xch,
                 int *sched_id);

tools/libxc/xc_misc.c

@@ -264,6 +264,45 @@ out:
     return ret;
 }
 
+int xc_pcitopoinfo(xc_interface *xch, unsigned num_devs,
+                   physdev_pci_device_t *devs,
+                   uint32_t *nodes)
+{
+    int ret = 0;
+    unsigned processed = 0;
+    DECLARE_SYSCTL;
+    DECLARE_HYPERCALL_BOUNCE(devs, num_devs * sizeof(*devs),
+                             XC_HYPERCALL_BUFFER_BOUNCE_IN);
+    DECLARE_HYPERCALL_BOUNCE(nodes, num_devs* sizeof(*nodes),
+                             XC_HYPERCALL_BUFFER_BOUNCE_BOTH);
+
+    if ( (ret = xc_hypercall_bounce_pre(xch, devs)) )
+        goto out;
+    if ( (ret = xc_hypercall_bounce_pre(xch, nodes)) )
+        goto out;
+
+    sysctl.cmd = XEN_SYSCTL_pcitopoinfo;
+
+    while ( processed < num_devs )
+    {
+        sysctl.u.pcitopoinfo.num_devs = num_devs - processed;
+        set_xen_guest_handle_offset(sysctl.u.pcitopoinfo.devs, devs,
+                                    processed);
+        set_xen_guest_handle_offset(sysctl.u.pcitopoinfo.nodes, nodes,
+                                    processed);
+
+        if ( (ret = do_sysctl(xch, &sysctl)) != 0 )
+                break;
+
+        processed += sysctl.u.pcitopoinfo.num_devs;
+    }
+
+ out:
+    xc_hypercall_bounce_post(xch, devs);
+    xc_hypercall_bounce_post(xch, nodes);
+
+    return ret;
+}
 
 int xc_sched_id(xc_interface *xch,
                 int *sched_id)

tools/libxl/libxl.c

@@ -5139,6 +5139,51 @@ libxl_cputopology *libxl_get_cpu_topology(libxl_ctx *ctx, int *nb_cpu_out)
     return ret;
 }
 
+libxl_pcitopology *libxl_get_pci_topology(libxl_ctx *ctx, int *num_devs)
+{
+    GC_INIT(ctx);
+    physdev_pci_device_t *devs;
+    uint32_t *nodes;
+    libxl_pcitopology *ret = NULL;
+    int i, rc;
+
+    *num_devs = libxl__pci_numdevs(gc);
+    if (*num_devs < 0) {
+        LOG(ERROR, "Unable to determine number of PCI devices, rc %d",
+            *num_devs);
+        goto out;
+    }
+
+    devs = libxl__zalloc(gc, sizeof(*devs) * *num_devs);
+    nodes = libxl__zalloc(gc, sizeof(*nodes) * *num_devs);
+
+    rc = libxl__pci_topology_init(gc, devs, *num_devs);
+    if (rc) {
+        LOG(ERROR, "Cannot initialize PCI hypercall structure, rc %d", rc);
+        goto out;
+    }
+
+    if (xc_pcitopoinfo(ctx->xch, *num_devs, devs, nodes) != 0) {
+        LOGE(ERROR, "PCI topology info hypercall failed");
+        goto out;
+    }
+
+    ret = libxl__zalloc(NOGC, sizeof(libxl_pcitopology) * *num_devs);
+
+    for (i = 0; i < *num_devs; i++) {
+        ret[i].seg = devs[i].seg;
+        ret[i].bus = devs[i].bus;
+        ret[i].devfn = devs[i].devfn;
+        ret[i].node = ((nodes[i] == XEN_INVALID_NODE_ID) ||
+                       (nodes[i] == XEN_INVALID_DEV)) ?
+            LIBXL_PCITOPOLOGY_INVALID_ENTRY : nodes[i];
+    }
+
+ out:
+    GC_FREE;
+    return ret;
+}
+
 libxl_numainfo *libxl_get_numainfo(libxl_ctx *ctx, int *nr)
 {
     GC_INIT(ctx);

tools/libxl/libxl.h

@@ -760,6 +760,14 @@ void libxl_mac_copy(libxl_ctx *ctx, libxl_mac *dst, libxl_mac *src);
 #define LIBXL_HAVE_PSR_MBM 1
 #endif
 
+/*
+ * LIBXL_HAVE_PCITOPOLOGY
+ *
+ * If this is defined, then interface to query hypervisor about PCI device
+ * topology is available.
+ */
+#define LIBXL_HAVE_PCITOPOLOGY 1
+
 typedef char **libxl_string_list;
 void libxl_string_list_dispose(libxl_string_list *sl);
 int libxl_string_list_length(const libxl_string_list *sl);
@@ -1152,6 +1160,10 @@ void libxl_vminfo_list_free(libxl_vminfo *list, int nb_vm);
 libxl_cputopology *libxl_get_cpu_topology(libxl_ctx *ctx, int *nb_cpu_out);
 void libxl_cputopology_list_free(libxl_cputopology *, int nb_cpu);
 
+#define LIBXL_PCITOPOLOGY_INVALID_ENTRY (~(uint32_t)0)
+libxl_pcitopology *libxl_get_pci_topology(libxl_ctx *ctx, int *num_devs);
+void libxl_pcitopology_list_free(libxl_pcitopology *, int num_devs);
+
 #define LIBXL_NUMAINFO_INVALID_ENTRY (~(uint32_t)0)
 libxl_numainfo *libxl_get_numainfo(libxl_ctx *ctx, int *nr);
 void libxl_numainfo_list_free(libxl_numainfo *, int nr);

tools/libxl/libxl_freebsd.c

@@ -131,3 +131,15 @@ libxl_device_model_version libxl__default_device_model(libxl__gc *gc)
 {
     return LIBXL_DEVICE_MODEL_VERSION_QEMU_XEN;
 }
+
+int libxl__pci_numdevs(libxl__gc *gc)
+{
+    return ERROR_NI;
+}
+
+int libxl__pci_topology_init(libxl__gc *gc,
+                             physdev_pci_device_t *devs,
+                             int num_devs)
+{
+    return ERROR_NI;
+}

tools/libxl/libxl_internal.h

@@ -1196,6 +1196,11 @@ _hidden int libxl__try_phy_backend(mode_t st_mode);
 
 _hidden char *libxl__devid_to_localdev(libxl__gc *gc, int devid);
 
+_hidden int libxl__pci_numdevs(libxl__gc *gc);
+_hidden int libxl__pci_topology_init(libxl__gc *gc,
+                                     physdev_pci_device_t *devs,
+                                     int num_devs);
+
 /* from libxl_pci */
 
 _hidden int libxl__device_pci_add(libxl__gc *gc, uint32_t domid, libxl_device_pci *pcidev, int starting);

tools/libxl/libxl_linux.c

@@ -279,3 +279,73 @@ libxl_device_model_version libxl__default_device_model(libxl__gc *gc)
 {
     return LIBXL_DEVICE_MODEL_VERSION_QEMU_XEN;
 }
+
+int libxl__pci_numdevs(libxl__gc *gc)
+{
+    DIR *dir;
+    struct dirent *entry;
+    int num_devs = 0;
+
+    dir = opendir("/sys/bus/pci/devices");
+    if (!dir) {
+        LOGE(ERROR, "Cannot open /sys/bus/pci/devices");
+        return ERROR_FAIL;
+    }
+
+    while ((entry = readdir(dir))) {
+        if (entry->d_name[0] == '.')
+            continue;
+        num_devs++;
+    }
+    closedir(dir);
+
+    return num_devs;
+}
+
+int libxl__pci_topology_init(libxl__gc *gc,
+                             physdev_pci_device_t *devs,
+                             int num_devs)
+{
+
+    DIR *dir;
+    struct dirent *entry;
+    int i, err = 0;
+
+    dir = opendir("/sys/bus/pci/devices");
+    if (!dir) {
+        LOGE(ERROR, "Cannot open /sys/bus/pci/devices");
+        return ERROR_FAIL;
+    }
+
+    i = 0;
+    while ((entry = readdir(dir))) {
+        unsigned int dom, bus, dev, func;
+
+        if (entry->d_name[0] == '.')
+            continue;
+
+        if (i == num_devs) {
+            LOG(ERROR, "Too many devices");
+            err = ERROR_FAIL;
+            errno = -ENOSPC;
+            goto out;
+        }
+
+        if (sscanf(entry->d_name, "%x:%x:%x.%d", &dom, &bus, &dev, &func) < 4) {
+            LOGE(ERROR, "Error processing /sys/bus/pci/devices");
+            err = ERROR_FAIL;
+            goto out;
+        }
+
+        devs[i].seg = dom;
+        devs[i].bus = bus;
+        devs[i].devfn = ((dev & 0x1f) << 3) | (func & 7);
+
+        i++;
+    }
+
+ out:
+    closedir(dir);
+
+    return err;
+}

tools/libxl/libxl_netbsd.c

@@ -95,3 +95,15 @@ libxl_device_model_version libxl__default_device_model(libxl__gc *gc)
 {
     return LIBXL_DEVICE_MODEL_VERSION_QEMU_XEN_TRADITIONAL;
 }
+
+int libxl__pci_numdevs(libxl__gc *gc)
+{
+    return ERROR_NI;
+}
+
+int libxl__pci_topology_init(libxl__gc *gc,
+                             physdev_pci_device_t *devs,
+                             int num_devs)
+{
+    return ERROR_NI;
+}

tools/libxl/libxl_types.idl

@@ -657,6 +657,13 @@ libxl_cputopology = Struct("cputopology", [
     ("node", uint32),
     ], dir=DIR_OUT)
 
+libxl_pcitopology = Struct("pcitopology", [
+    ("seg", uint16),
+    ("bus", uint8),
+    ("devfn", uint8),
+    ("node", uint32),
+    ], dir=DIR_OUT)
+
 libxl_sched_credit_params = Struct("sched_credit_params", [
     ("tslice_ms", integer),
     ("ratelimit_us", integer),

tools/libxl/libxl_utils.c

@@ -952,6 +952,14 @@ void libxl_cputopology_list_free(libxl_cputopology *list, int nr)
     free(list);
 }
 
+void libxl_pcitopology_list_free(libxl_pcitopology *list, int nr)
+{
+    int i;
+    for (i = 0; i < nr; i++)
+        libxl_pcitopology_dispose(&list[i]);
+    free(list);
+}
+
 void libxl_numainfo_list_free(libxl_numainfo *list, int nr)
 {
     int i;

tools/libxl/xl_cmdimpl.c

@@ -5420,12 +5420,15 @@ static void output_numainfo(void)
 
 static void output_topologyinfo(void)
 {
-    libxl_cputopology *info;
+    libxl_cputopology *cpuinfo;
     int i, nr;
+    libxl_pcitopology *pciinfo;
+    int valid_devs;
 
-    info = libxl_get_cpu_topology(ctx, &nr);
-    if (info == NULL) {
-        fprintf(stderr, "libxl_get_topologyinfo failed.\n");
+
+    cpuinfo = libxl_get_cpu_topology(ctx, &nr);
+    if (cpuinfo == NULL) {
+        fprintf(stderr, "libxl_get_cpu_topology failed.\n");
         return;
     }
 
@@ -5433,12 +5436,35 @@ static void output_topologyinfo(void)
     printf("cpu:    core    socket     node\n");
 
     for (i = 0; i < nr; i++) {
-        if (info[i].core != LIBXL_CPUTOPOLOGY_INVALID_ENTRY)
+        if (cpuinfo[i].core != LIBXL_CPUTOPOLOGY_INVALID_ENTRY)
             printf("%3d:    %4d     %4d     %4d\n", i,
-                   info[i].core, info[i].socket, info[i].node);
+                   cpuinfo[i].core, cpuinfo[i].socket, cpuinfo[i].node);
     }
 
-    libxl_cputopology_list_free(info, nr);
+    libxl_cputopology_list_free(cpuinfo, nr);
+
+    pciinfo = libxl_get_pci_topology(ctx, &nr);
+    if (pciinfo == NULL) {
+        fprintf(stderr, "libxl_get_pci_topology failed.\n");
+        return;
+    }
+
+    printf("device topology        :\n");
+    printf("device           node\n");
+    for (i = 0; i < nr; i++) {
+        if (pciinfo[i].node != LIBXL_PCITOPOLOGY_INVALID_ENTRY) {
+            printf("%04x:%02x:%02x.%01x      %d\n", pciinfo[i].seg,
+                   pciinfo[i].bus,
+                   ((pciinfo[i].devfn >> 3) & 0x1f), (pciinfo[i].devfn & 7),
+                   pciinfo[i].node);
+            valid_devs++;
+        }
+    }
+
+    if (valid_devs == 0)
+        printf("No device topology data available\n");
+
+    libxl_pcitopology_list_free(pciinfo, nr);
 
     return;
 }