vhost: Distinguish errors in vhost_backend_init()
authorKevin Wolf <kwolf@redhat.com>
Wed, 9 Jun 2021 15:46:53 +0000 (08:46 -0700)
committerKevin Wolf <kwolf@redhat.com>
Wed, 30 Jun 2021 11:16:03 +0000 (13:16 +0200)
Instead of just returning 0/-1 and letting the caller make up a
meaningless error message, add an Error parameter to allow reporting the
real error and switch to 0/-errno so that different kind of errors can
be distinguished in the caller.

Specifically, in vhost-user, EPROTO is used for all errors that relate
to the connection itself, whereas other error codes are used for errors
relating to the content of the connection. This will allow us later to
automatically reconnect when the connection goes away, without ending up
in an endless loop if it's a permanent error in the configuration.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210609154658.350308-3-kwolf@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
hw/virtio/vhost-backend.c
hw/virtio/vhost-user.c
hw/virtio/vhost-vdpa.c
hw/virtio/vhost.c
include/hw/virtio/vhost-backend.h

index 31b33bde37b276cb8993039d7354463689ea0c88..f4f71cf58ab64d8a2379a431640b09f32be8c8af 100644 (file)
@@ -30,7 +30,7 @@ static int vhost_kernel_call(struct vhost_dev *dev, unsigned long int request,
     return ioctl(fd, request, arg);
 }
 
-static int vhost_kernel_init(struct vhost_dev *dev, void *opaque)
+static int vhost_kernel_init(struct vhost_dev *dev, void *opaque, Error **errp)
 {
     assert(dev->vhost_ops->backend_type == VHOST_BACKEND_TYPE_KERNEL);
 
index ee57abe04526f6c55d983cb0254c95d4262977b3..024cb201bb7c8fca1b5302734339bafae0df7ca3 100644 (file)
@@ -1856,7 +1856,8 @@ static int vhost_user_postcopy_notifier(NotifierWithReturn *notifier,
     return 0;
 }
 
-static int vhost_user_backend_init(struct vhost_dev *dev, void *opaque)
+static int vhost_user_backend_init(struct vhost_dev *dev, void *opaque,
+                                   Error **errp)
 {
     uint64_t features, protocol_features, ram_slots;
     struct vhost_user *u;
@@ -1871,7 +1872,7 @@ static int vhost_user_backend_init(struct vhost_dev *dev, void *opaque)
 
     err = vhost_user_get_features(dev, &features);
     if (err < 0) {
-        return err;
+        return -EPROTO;
     }
 
     if (virtio_has_feature(features, VHOST_USER_F_PROTOCOL_FEATURES)) {
@@ -1880,7 +1881,7 @@ static int vhost_user_backend_init(struct vhost_dev *dev, void *opaque)
         err = vhost_user_get_u64(dev, VHOST_USER_GET_PROTOCOL_FEATURES,
                                  &protocol_features);
         if (err < 0) {
-            return err;
+            return -EPROTO;
         }
 
         dev->protocol_features =
@@ -1891,14 +1892,14 @@ static int vhost_user_backend_init(struct vhost_dev *dev, void *opaque)
             dev->protocol_features &= ~(1ULL << VHOST_USER_PROTOCOL_F_CONFIG);
         } else if (!(protocol_features &
                     (1ULL << VHOST_USER_PROTOCOL_F_CONFIG))) {
-            error_report("Device expects VHOST_USER_PROTOCOL_F_CONFIG "
-                    "but backend does not support it.");
-            return -1;
+            error_setg(errp, "Device expects VHOST_USER_PROTOCOL_F_CONFIG "
+                       "but backend does not support it.");
+            return -EINVAL;
         }
 
         err = vhost_user_set_protocol_features(dev, dev->protocol_features);
         if (err < 0) {
-            return err;
+            return -EPROTO;
         }
 
         /* query the max queues we support if backend supports Multiple Queue */
@@ -1906,12 +1907,12 @@ static int vhost_user_backend_init(struct vhost_dev *dev, void *opaque)
             err = vhost_user_get_u64(dev, VHOST_USER_GET_QUEUE_NUM,
                                      &dev->max_queues);
             if (err < 0) {
-                return err;
+                return -EPROTO;
             }
         }
         if (dev->num_queues && dev->max_queues < dev->num_queues) {
-            error_report("The maximum number of queues supported by the "
-                         "backend is %" PRIu64, dev->max_queues);
+            error_setg(errp, "The maximum number of queues supported by the "
+                       "backend is %" PRIu64, dev->max_queues);
             return -EINVAL;
         }
 
@@ -1920,9 +1921,9 @@ static int vhost_user_backend_init(struct vhost_dev *dev, void *opaque)
                     VHOST_USER_PROTOCOL_F_SLAVE_REQ) &&
                  virtio_has_feature(dev->protocol_features,
                     VHOST_USER_PROTOCOL_F_REPLY_ACK))) {
-            error_report("IOMMU support requires reply-ack and "
-                         "slave-req protocol features.");
-            return -1;
+            error_setg(errp, "IOMMU support requires reply-ack and "
+                       "slave-req protocol features.");
+            return -EINVAL;
         }
 
         /* get max memory regions if backend supports configurable RAM slots */
@@ -1932,15 +1933,15 @@ static int vhost_user_backend_init(struct vhost_dev *dev, void *opaque)
         } else {
             err = vhost_user_get_max_memslots(dev, &ram_slots);
             if (err < 0) {
-                return err;
+                return -EPROTO;
             }
 
             if (ram_slots < u->user->memory_slots) {
-                error_report("The backend specified a max ram slots limit "
-                             "of %" PRIu64", when the prior validated limit was %d. "
-                             "This limit should never decrease.", ram_slots,
-                             u->user->memory_slots);
-                return -1;
+                error_setg(errp, "The backend specified a max ram slots limit "
+                           "of %" PRIu64", when the prior validated limit was "
+                           "%d. This limit should never decrease.", ram_slots,
+                           u->user->memory_slots);
+                return -EINVAL;
             }
 
             u->user->memory_slots = MIN(ram_slots, VHOST_USER_MAX_RAM_SLOTS);
@@ -1958,7 +1959,7 @@ static int vhost_user_backend_init(struct vhost_dev *dev, void *opaque)
     if (dev->vq_index == 0) {
         err = vhost_setup_slave_channel(dev);
         if (err < 0) {
-            return err;
+            return -EPROTO;
         }
     }
 
index 61ba3133317934a263f3734f373293777e232a2c..80827ee0407b0aecf43f1269e7f228378f3008a8 100644 (file)
@@ -265,7 +265,7 @@ static void vhost_vdpa_add_status(struct vhost_dev *dev, uint8_t status)
     vhost_vdpa_call(dev, VHOST_VDPA_SET_STATUS, &s);
 }
 
-static int vhost_vdpa_init(struct vhost_dev *dev, void *opaque)
+static int vhost_vdpa_init(struct vhost_dev *dev, void *opaque, Error **errp)
 {
     struct vhost_vdpa *v;
     assert(dev->vhost_ops->backend_type == VHOST_BACKEND_TYPE_VDPA);
index 991c67ddcd19275a45fc90f758e5807a4046c518..fd131357067abcdad6703c480d7fbf0057492d53 100644 (file)
@@ -1289,9 +1289,9 @@ int vhost_dev_init(struct vhost_dev *hdev, void *opaque,
                    VhostBackendType backend_type, uint32_t busyloop_timeout,
                    Error **errp)
 {
+    ERRP_GUARD();
     uint64_t features;
     int i, r, n_initialized_vqs = 0;
-    Error *local_err = NULL;
 
     hdev->vdev = NULL;
     hdev->migration_blocker = NULL;
@@ -1299,9 +1299,11 @@ int vhost_dev_init(struct vhost_dev *hdev, void *opaque,
     r = vhost_set_backend_type(hdev, backend_type);
     assert(r >= 0);
 
-    r = hdev->vhost_ops->vhost_backend_init(hdev, opaque);
+    r = hdev->vhost_ops->vhost_backend_init(hdev, opaque, errp);
     if (r < 0) {
-        error_setg(errp, "vhost_backend_init failed");
+        if (!*errp) {
+            error_setg_errno(errp, -r, "vhost_backend_init failed");
+        }
         goto fail;
     }
 
@@ -1369,9 +1371,8 @@ int vhost_dev_init(struct vhost_dev *hdev, void *opaque,
     }
 
     if (hdev->migration_blocker != NULL) {
-        r = migrate_add_blocker(hdev->migration_blocker, &local_err);
-        if (local_err) {
-            error_propagate(errp, local_err);
+        r = migrate_add_blocker(hdev->migration_blocker, errp);
+        if (*errp) {
             error_free(hdev->migration_blocker);
             goto fail_busyloop;
         }
index 8a6f8e2a7a30c591c2233f6b4003c129804f0fa2..728ebb0ed9d9132dda9c9e542256c30ff3db773a 100644 (file)
@@ -37,7 +37,8 @@ struct vhost_scsi_target;
 struct vhost_iotlb_msg;
 struct vhost_virtqueue;
 
-typedef int (*vhost_backend_init)(struct vhost_dev *dev, void *opaque);
+typedef int (*vhost_backend_init)(struct vhost_dev *dev, void *opaque,
+                                  Error **errp);
 typedef int (*vhost_backend_cleanup)(struct vhost_dev *dev);
 typedef int (*vhost_backend_memslots_limit)(struct vhost_dev *dev);