qemu: Spawn qemu under mount namespace

Prime time. When it comes to spawning qemu process and
relabelling all the devices it's going to touch, there's inherent
race with other applications in the system (e.g. udev). Instead
of trying convincing udev to not touch libvirt managed devices,
we can create a separate mount namespace for the qemu, and mount
our own /dev there. Of course this puts more work onto us as we
have to maintain /dev files on each domain start and device
hot(un-)plug. On the other hand, this enhances security also.

From technical POV, on domain startup process the parent
(libvirtd) creates:

  /var/lib/libvirt/qemu/$domain.dev
  /var/lib/libvirt/qemu/$domain.devpts

The child (which is going to be qemu eventually) calls unshare()
to create new mount namespace. From now on anything that child
does is invisible to the parent. Child then mounts tmpfs on
$domain.dev (so that it still sees original /dev from the host)
and creates some devices (as explained in one of the previous
patches). The devices have to be created exactly as they are in
the host (including perms, seclabels, ACLs, ...). After that it
moves $domain.dev mount to /dev.

What's the $domain.devpts mount there for then you ask? QEMU can
create PTYs for some chardevs. And historically we exposed the
host ends in our domain XML allowing users to connect to them.
Therefore we must preserve devpts mount to be shared with the
host's one.

To make this patch as small as possible, creating of devices
configured for domain in question is implemented in next patches.

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>

diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c
index d3e505b17639f1125e68b3d1c632bffb9a6cc073..3735c35f39dcb35c4049851397e1bda6e4bbc84e 100644 (file)
--- a/src/qemu/qemu_domain.c
+++ b/src/qemu/qemu_domain.c
@@ -55,6 +55,9 @@
 
 #include <sys/time.h>
 #include <fcntl.h>
+#if defined(HAVE_SYS_MOUNT_H)
+# include <sys/mount.h>
+#endif
 
 #include <libxml/xpathInternals.h>
 
@@ -86,6 +89,20 @@ VIR_ENUM_IMPL(qemuDomainAsyncJob, QEMU_ASYNC_JOB_LAST,
               "start",
 );
 
+VIR_ENUM_IMPL(qemuDomainNamespace, QEMU_DOMAIN_NS_LAST,
+              "mount",
+);
+
+
+static struct {
+    const char *path;
+    const char *suffix;
+} devPreserveMounts[] = {
+    {"/dev/pts", "devpts"},
+    {"/dev/shm", "devshm"},
+    {"/dev/mqueue", "mqueue"},
+};
+
 
 struct _qemuDomainLogContext {
     int refs;
@@ -146,6 +163,70 @@ qemuDomainAsyncJobPhaseFromString(qemuDomainAsyncJob job,
 }
 
 
+bool
+qemuDomainNamespaceEnabled(virDomainObjPtr vm,
+                           qemuDomainNamespace ns)
+{
+    qemuDomainObjPrivatePtr priv = vm->privateData;
+
+    return priv->namespaces &&
+        virBitmapIsBitSet(priv->namespaces, ns);
+}
+
+
+static int
+qemuDomainEnableNamespace(virDomainObjPtr vm,
+                          qemuDomainNamespace ns)
+{
+    qemuDomainObjPrivatePtr priv = vm->privateData;
+
+    if (!priv->namespaces &&
+        !(priv->namespaces = virBitmapNew(QEMU_DOMAIN_NS_LAST)))
+        return -1;
+
+    if (virBitmapSetBit(priv->namespaces, ns) < 0) {
+        virReportError(VIR_ERR_INTERNAL_ERROR,
+                       _("Unable to enable namespace: %s"),
+                       qemuDomainNamespaceTypeToString(ns));
+        return -1;
+    }
+
+    return 0;
+}
+
+
+static int
+qemuDomainGetPreservedMounts(virQEMUDriverPtr driver,
+                             virDomainObjPtr vm,
+                             char ***devMountsPath,
+                             size_t *ndevMountsPath)
+{
+    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
+    char **paths;
+    size_t i;
+
+    if (VIR_ALLOC_N(paths, ARRAY_CARDINALITY(devPreserveMounts)) < 0)
+        goto error;
+
+    for (i = 0; i < ARRAY_CARDINALITY(devPreserveMounts); i++) {
+        if (virAsprintf(&paths[i], "%s/%s.%s",
+                        cfg->stateDir, vm->def->name,
+                        devPreserveMounts[i].suffix) < 0)
+            goto error;
+    }
+
+    *devMountsPath = paths;
+    *ndevMountsPath = ARRAY_CARDINALITY(devPreserveMounts);
+    virObjectUnref(cfg);
+    return 0;
+
+ error:
+    virStringListFreeCount(paths, ARRAY_CARDINALITY(devPreserveMounts));
+    virObjectUnref(cfg);
+    return -1;
+}
+
+
 void qemuDomainEventQueue(virQEMUDriverPtr driver,
                           virObjectEventPtr event)
 {
@@ -1541,6 +1622,8 @@ qemuDomainObjPrivateFree(void *data)
 
     virObjectUnref(priv->qemuCaps);
 
+    virBitmapFree(priv->namespaces);
+
     virCgroupFree(&priv->cgroup);
     virDomainPCIAddressSetFree(priv->pciaddrs);
     virDomainUSBAddressSetFree(priv->usbaddrs);
@@ -1627,6 +1710,17 @@ qemuDomainObjPrivateXMLFormat(virBufferPtr buf,
                           virDomainChrTypeToString(priv->monConfig->type));
     }
 
+    if (priv->namespaces) {
+        ssize_t ns = -1;
+
+        virBufferAddLit(buf, "<namespaces>\n");
+        virBufferAdjustIndent(buf, 2);
+        while ((ns = virBitmapNextSetBit(priv->namespaces, ns)) >= 0)
+            virBufferAsprintf(buf, "<%s/>\n", qemuDomainNamespaceTypeToString(ns));
+        virBufferAdjustIndent(buf, -2);
+        virBufferAddLit(buf, "</namespaces>\n");
+    }
+
     qemuDomainObjPrivateXMLFormatVcpus(buf, vm->def);
 
     if (priv->qemuCaps) {
@@ -1771,6 +1865,7 @@ qemuDomainObjPrivateXMLParse(xmlXPathContextPtr ctxt,
     int n;
     size_t i;
     xmlNodePtr *nodes = NULL;
+    xmlNodePtr node = NULL;
     virQEMUCapsPtr qemuCaps = NULL;
     virCapsPtr caps = NULL;
 
@@ -1809,6 +1904,30 @@ qemuDomainObjPrivateXMLParse(xmlXPathContextPtr ctxt,
         goto error;
     }
 
+    if ((node = virXPathNode("./namespaces", ctxt))) {
+        xmlNodePtr next;
+
+        for (next = node->children; next; next = next->next) {
+            int ns = qemuDomainNamespaceTypeFromString((const char *) next->name);
+
+            if (ns < 0) {
+                virReportError(VIR_ERR_INTERNAL_ERROR,
+                               _("malformed namespace name: %s"),
+                               next->name);
+                goto error;
+            }
+
+            if (qemuDomainEnableNamespace(vm, ns) < 0)
+                goto error;
+        }
+    }
+
+    if (priv->namespaces &&
+        virBitmapIsAllClear(priv->namespaces)) {
+        virBitmapFree(priv->namespaces);
+        priv->namespaces = NULL;
+    }
+
     if ((n = virXPathNodeSet("./vcpus/vcpu", ctxt, &nodes)) < 0)
         goto error;
 
@@ -1959,10 +2078,12 @@ qemuDomainObjPrivateXMLParse(xmlXPathContextPtr ctxt,
     return 0;
 
  error:
-    virDomainChrSourceDefFree(priv->monConfig);
-    priv->monConfig = NULL;
     VIR_FREE(nodes);
     VIR_FREE(tmp);
+    virBitmapFree(priv->namespaces);
+    priv->namespaces = NULL;
+    virDomainChrSourceDefFree(priv->monConfig);
+    priv->monConfig = NULL;
     virStringListFree(priv->qemuDevices);
     priv->qemuDevices = NULL;
     virObjectUnref(qemuCaps);
@@ -6653,3 +6774,319 @@ qemuDomainSupportsVideoVga(virDomainVideoDefPtr video,
 
     return true;
 }
+
+
+static int
+qemuDomainCreateDevice(const char *device,
+                       const char *path,
+                       bool allow_noent)
+{
+    char *devicePath = NULL;
+    struct stat sb;
+    int ret = -1;
+
+    if (!STRPREFIX(device, "/dev")) {
+        virReportError(VIR_ERR_INTERNAL_ERROR,
+                       _("invalid device: %s"),
+                       device);
+        goto cleanup;
+    }
+
+    if (virAsprintf(&devicePath, "%s/%s",
+                    path, device + 4) < 0)
+        goto cleanup;
+
+    if (stat(device, &sb) < 0) {
+        if (errno == ENOENT && allow_noent) {
+            /* Ignore non-existent device. */
+            ret = 0;
+            goto cleanup;
+        }
+
+        virReportSystemError(errno, _("Unable to stat %s"), device);
+        goto cleanup;
+    }
+
+    if (virFileMakeParentPath(devicePath) < 0) {
+        virReportSystemError(errno,
+                             _("Unable to create %s"),
+                             devicePath);
+        goto cleanup;
+    }
+
+    if (mknod(devicePath, sb.st_mode, sb.st_rdev) < 0) {
+        virReportSystemError(errno,
+                             _("Failed to make device %s"),
+                             devicePath);
+        goto cleanup;
+    }
+
+    if (chown(devicePath, sb.st_uid, sb.st_gid) < 0) {
+        virReportSystemError(errno,
+                             _("Failed to chown device %s"),
+                             devicePath);
+        goto cleanup;
+    }
+
+    if (virFileCopyACLs(device, devicePath) < 0 &&
+        errno != ENOTSUP) {
+        virReportSystemError(errno,
+                             _("Failed to copy ACLs on device %s"),
+                             devicePath);
+        goto cleanup;
+    }
+
+    ret = 0;
+ cleanup:
+    VIR_FREE(devicePath);
+    return ret;
+}
+
+
+
+static int
+qemuDomainPopulateDevices(virQEMUDriverPtr driver,
+                          virDomainObjPtr vm ATTRIBUTE_UNUSED,
+                          const char *path)
+{
+    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
+    const char *const *devices = (const char *const *) cfg->cgroupDeviceACL;
+    size_t i;
+    int ret = -1;
+
+    if (!devices)
+        devices = defaultDeviceACL;
+
+    for (i = 0; devices[i]; i++) {
+        if (qemuDomainCreateDevice(devices[i], path, true) < 0)
+            goto cleanup;
+    }
+
+    ret = 0;
+ cleanup:
+    virObjectUnref(cfg);
+    return ret;
+}
+
+
+static int
+qemuDomainSetupDev(virQEMUDriverPtr driver,
+                   virDomainObjPtr vm,
+                   const char *path)
+{
+    char *mount_options = NULL;
+    char *opts = NULL;
+    int ret = -1;
+
+    VIR_DEBUG("Setting up /dev/ for domain %s", vm->def->name);
+
+    mount_options = virSecurityManagerGetMountOptions(driver->securityManager,
+                                                      vm->def);
+
+    if (!mount_options &&
+        VIR_STRDUP(mount_options, "") < 0)
+        goto cleanup;
+
+    /*
+     * tmpfs is limited to 64kb, since we only have device nodes in there
+     * and don't want to DOS the entire OS RAM usage
+     */
+    if (virAsprintf(&opts,
+                    "mode=755,size=65536%s", mount_options) < 0)
+        goto cleanup;
+
+    if (virFileSetupDev(path, opts) < 0)
+        goto cleanup;
+
+    if (qemuDomainPopulateDevices(driver, vm, path) < 0)
+        goto cleanup;
+
+    ret = 0;
+ cleanup:
+    VIR_FREE(opts);
+    VIR_FREE(mount_options);
+    return ret;
+}
+
+
+int
+qemuDomainBuildNamespace(virQEMUDriverPtr driver,
+                         virDomainObjPtr vm)
+{
+    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
+    const unsigned long mount_flags = MS_MOVE;
+    char *devPath = NULL;
+    char **devMountsPath = NULL;
+    size_t ndevMountsPath = 0, i;
+    int ret = -1;
+
+    if (!qemuDomainNamespaceEnabled(vm, QEMU_DOMAIN_NS_MOUNT)) {
+        ret = 0;
+        goto cleanup;
+    }
+
+    if (virAsprintf(&devPath, "%s/%s.dev",
+                    cfg->stateDir, vm->def->name) < 0)
+        goto cleanup;
+
+    if (qemuDomainGetPreservedMounts(driver, vm,
+                                     &devMountsPath, &ndevMountsPath) < 0)
+        goto cleanup;
+
+    if (qemuDomainSetupDev(driver, vm, devPath) < 0)
+        goto cleanup;
+
+    /* Save some mount points because we want to share them with the host */
+    for (i = 0; i < ndevMountsPath; i++) {
+        if (mount(devPreserveMounts[i].path, devMountsPath[i],
+                  NULL, mount_flags, NULL) < 0) {
+            virReportSystemError(errno,
+                                 _("Unable to move %s mount"),
+                                 devPreserveMounts[i].path);
+            goto cleanup;
+        }
+    }
+
+    if (mount(devPath, "/dev", NULL, mount_flags, NULL) < 0) {
+        virReportSystemError(errno,
+                             _("Failed to mount %s on /dev"),
+                             devPath);
+        goto cleanup;
+    }
+
+    for (i = 0; i < ndevMountsPath; i++) {
+        if (virFileMakePath(devPreserveMounts[i].path) < 0) {
+            virReportSystemError(errno, _("Cannot create %s"),
+                                 devPreserveMounts[i].path);
+            goto cleanup;
+        }
+
+        if (mount(devMountsPath[i], devPreserveMounts[i].path,
+                  NULL, mount_flags, NULL) < 0) {
+            virReportSystemError(errno,
+                                 _("Failed to mount %s on %s"),
+                                 devMountsPath[i],
+                                 devPreserveMounts[i].path);
+            goto cleanup;
+        }
+    }
+
+    ret = 0;
+ cleanup:
+    virObjectUnref(cfg);
+    VIR_FREE(devPath);
+    virStringListFreeCount(devMountsPath, ndevMountsPath);
+    return ret;
+}
+
+
+#if defined(__linux__)
+int
+qemuDomainCreateNamespace(virQEMUDriverPtr driver,
+                          virDomainObjPtr vm)
+{
+    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
+    int ret = -1;
+    char *devPath = NULL;
+    char **devMountsPath = NULL;
+    size_t ndevMountsPath = 0, i;
+
+    if (!virQEMUDriverIsPrivileged(driver)) {
+        ret = 0;
+        goto cleanup;
+    }
+
+    if (virAsprintf(&devPath, "%s/%s.dev",
+                    cfg->stateDir, vm->def->name) < 0)
+        goto cleanup;
+
+    if (qemuDomainGetPreservedMounts(driver, vm,
+                                     &devMountsPath, &ndevMountsPath) < 0)
+        goto cleanup;
+
+    if (virFileMakePath(devPath) < 0) {
+        virReportSystemError(errno,
+                             _("Failed to create %s"),
+                             devPath);
+        goto cleanup;
+    }
+
+    for (i = 0; i < ndevMountsPath; i++) {
+        if (virFileMakePath(devMountsPath[i]) < 0) {
+            virReportSystemError(errno,
+                                 _("Failed to create %s"),
+                                 devMountsPath[i]);
+            goto cleanup;
+        }
+    }
+
+    /* Enabling of the mount namespace goes here. */
+
+    ret = 0;
+ cleanup:
+    if (ret < 0) {
+        if (devPath)
+            unlink(devPath);
+        for (i = 0; i < ndevMountsPath; i++)
+            unlink(devMountsPath[i]);
+    }
+    virStringListFreeCount(devMountsPath, ndevMountsPath);
+    VIR_FREE(devPath);
+    virObjectUnref(cfg);
+    return ret;
+}
+
+#else /* !defined(__linux__) */
+
+int
+qemuDomainCreateNamespace(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
+                          virDomainObjPtr vm ATTRIBUTE_UNUSED)
+{
+    /* Namespaces are Linux specific. On other platforms just
+     * carry on with the old behaviour. */
+    return 0;
+}
+#endif /* !defined(__linux__) */
+
+
+void
+qemuDomainDeleteNamespace(virQEMUDriverPtr driver,
+                          virDomainObjPtr vm)
+{
+    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
+    char *devPath = NULL;
+    char **devMountsPath = NULL;
+    size_t ndevMountsPath = 0, i;
+
+
+    if (!qemuDomainNamespaceEnabled(vm, QEMU_DOMAIN_NS_MOUNT))
+        return;
+
+    if (virAsprintf(&devPath, "%s/%s.dev",
+                    cfg->stateDir, vm->def->name) < 0)
+        goto cleanup;
+
+    if (qemuDomainGetPreservedMounts(driver, vm,
+                                     &devMountsPath, &ndevMountsPath) < 0)
+        goto cleanup;
+
+    if (rmdir(devPath) < 0) {
+        virReportSystemError(errno,
+                             _("Unable to remove %s"),
+                             devPath);
+        /* Bet effort. Fall through. */
+    }
+
+    for (i = 0; i < ndevMountsPath; i++) {
+        if (rmdir(devMountsPath[i]) < 0) {
+            virReportSystemError(errno,
+                                 _("Unable to remove %s"),
+                                 devMountsPath[i]);
+            /* Bet effort. Fall through. */
+        }
+    }
+ cleanup:
+    virObjectUnref(cfg);
+    virStringListFreeCount(devMountsPath, ndevMountsPath);
+    VIR_FREE(devPath);
+}

diff --git a/src/qemu/qemu_domain.h b/src/qemu/qemu_domain.h
index 7650ff392d0449dbec2400cec4267d1f9c979a2c..9dad5bc7cce9b1f2f7d93c8d95da2361ce9ec9b3 100644 (file)
--- a/src/qemu/qemu_domain.h
+++ b/src/qemu/qemu_domain.h
@@ -165,11 +165,23 @@ struct _qemuDomainUnpluggingDevice {
     qemuDomainUnpluggingDeviceStatus status;
 };
 
+
+typedef enum {
+    QEMU_DOMAIN_NS_MOUNT = 0,
+    QEMU_DOMAIN_NS_LAST
+} qemuDomainNamespace;
+VIR_ENUM_DECL(qemuDomainNamespace)
+
+bool qemuDomainNamespaceEnabled(virDomainObjPtr vm,
+                                qemuDomainNamespace ns);
+
 typedef struct _qemuDomainObjPrivate qemuDomainObjPrivate;
 typedef qemuDomainObjPrivate *qemuDomainObjPrivatePtr;
 struct _qemuDomainObjPrivate {
     struct qemuDomainJobObj job;
 
+    virBitmapPtr namespaces;
+
     qemuMonitorPtr mon;
     virDomainChrSourceDefPtr monConfig;
     bool monJSON;
@@ -785,4 +797,12 @@ int qemuDomainCheckMonitor(virQEMUDriverPtr driver,
 bool qemuDomainSupportsVideoVga(virDomainVideoDefPtr video,
                                 virQEMUCapsPtr qemuCaps);
 
+int qemuDomainBuildNamespace(virQEMUDriverPtr driver,
+                             virDomainObjPtr vm);
+
+int qemuDomainCreateNamespace(virQEMUDriverPtr driver,
+                              virDomainObjPtr vm);
+
+void qemuDomainDeleteNamespace(virQEMUDriverPtr driver,
+                               virDomainObjPtr vm);
 #endif /* __QEMU_DOMAIN_H__ */

diff --git a/src/qemu/qemu_process.c b/src/qemu/qemu_process.c
index 3871034818b119b340db22ccd7eb2a3eff7c1ffa..08859b8880fb24c9e785b82a9c46d1b5ef05c45b 100644 (file)
--- a/src/qemu/qemu_process.c
+++ b/src/qemu/qemu_process.c
@@ -2673,6 +2673,12 @@ static int qemuProcessHook(void *data)
     if (virSecurityManagerClearSocketLabel(h->driver->securityManager, h->vm->def) < 0)
         goto cleanup;
 
+    if (virProcessSetupPrivateMountNS() < 0)
+        goto cleanup;
+
+    if (qemuDomainBuildNamespace(h->driver, h->vm) < 0)
+        goto cleanup;
+
     if (virDomainNumatuneGetMode(h->vm->def->numa, -1, &mode) == 0) {
         if (mode == VIR_DOMAIN_NUMATUNE_MEM_STRICT &&
             h->cfg->cgroupControllers & (1 << VIR_CGROUP_CONTROLLER_CPUSET) &&
@@ -5498,6 +5504,11 @@ qemuProcessLaunch(virConnectPtr conn,
 
     qemuDomainLogContextMarkPosition(logCtxt);
 
+    VIR_DEBUG("Building mount namespace");
+
+    if (qemuDomainCreateNamespace(driver, vm) < 0)
+        goto cleanup;
+
     VIR_DEBUG("Clear emulator capabilities: %d",
               cfg->clearEmulatorCapabilities);
     if (cfg->clearEmulatorCapabilities)
@@ -6269,6 +6280,8 @@ void qemuProcessStop(virQEMUDriverPtr driver,
         }
     }
 
+    qemuDomainDeleteNamespace(driver, vm);
+
     vm->taint = 0;
     vm->pid = -1;
     virDomainObjSetState(vm, VIR_DOMAIN_SHUTOFF, reason);