]> git.baikalelectronics.ru Git - kernel.git/commitdiff
rbd: fix error handling around rbd_init_disk()
authorIlya Dryomov <idryomov@gmail.com>
Thu, 13 Apr 2017 10:17:38 +0000 (12:17 +0200)
committerIlya Dryomov <idryomov@gmail.com>
Thu, 4 May 2017 07:19:23 +0000 (09:19 +0200)
add_disk() takes an extra reference on disk->queue, which is put in
put_disk() -> disk_release().  Avoiding blk_cleanup_queue() (which also
puts the queue) until add_disk() sets GENHD_FL_UP works for the queue
itself, but leaks various queue internals.  Conditioning tag_set freeing
on GENHD_FL_UP is wrong too: all error paths after rbd_init_disk() leak
the tag_set.

Move the final "announce" steps out of rbd_dev_device_setup() so that
it can be unwound like any other function.  Leave "announce" steps to
do_rbd_add/remove().

Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Jason Dillaman <dillaman@redhat.com>
drivers/block/rbd.c

index b299ed0315f84432f16ddd95863f9021ac080150..50395af7a9a670759b31bc5ef7a879eaba60494a 100644 (file)
@@ -4114,19 +4114,10 @@ static int rbd_queue_rq(struct blk_mq_hw_ctx *hctx,
 
 static void rbd_free_disk(struct rbd_device *rbd_dev)
 {
-       struct gendisk *disk = rbd_dev->disk;
-
-       if (!disk)
-               return;
-
+       blk_cleanup_queue(rbd_dev->disk->queue);
+       blk_mq_free_tag_set(&rbd_dev->tag_set);
+       put_disk(rbd_dev->disk);
        rbd_dev->disk = NULL;
-       if (disk->flags & GENHD_FL_UP) {
-               del_gendisk(disk);
-               if (disk->queue)
-                       blk_cleanup_queue(disk->queue);
-               blk_mq_free_tag_set(&rbd_dev->tag_set);
-       }
-       put_disk(disk);
 }
 
 static int rbd_obj_read_sync(struct rbd_device *rbd_dev,
@@ -4385,8 +4376,12 @@ static int rbd_init_disk(struct rbd_device *rbd_dev)
        if (!ceph_test_opt(rbd_dev->rbd_client->client, NOCRC))
                q->backing_dev_info->capabilities |= BDI_CAP_STABLE_WRITES;
 
+       /*
+        * disk_release() expects a queue ref from add_disk() and will
+        * put it.  Hold an extra ref until add_disk() is called.
+        */
+       WARN_ON(!blk_get_queue(q));
        disk->queue = q;
-
        q->queuedata = rbd_dev;
 
        rbd_dev->disk = disk;
@@ -5875,6 +5870,15 @@ out_err:
        return ret;
 }
 
+static void rbd_dev_device_release(struct rbd_device *rbd_dev)
+{
+       clear_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags);
+       rbd_dev_mapping_clear(rbd_dev);
+       rbd_free_disk(rbd_dev);
+       if (!single_major)
+               unregister_blkdev(rbd_dev->major, rbd_dev->name);
+}
+
 /*
  * rbd_dev->header_rwsem must be locked for write and will be unlocked
  * upon return.
@@ -5910,26 +5914,13 @@ static int rbd_dev_device_setup(struct rbd_device *rbd_dev)
        set_capacity(rbd_dev->disk, rbd_dev->mapping.size / SECTOR_SIZE);
        set_disk_ro(rbd_dev->disk, rbd_dev->mapping.read_only);
 
-       dev_set_name(&rbd_dev->dev, "%d", rbd_dev->dev_id);
-       ret = device_add(&rbd_dev->dev);
+       ret = dev_set_name(&rbd_dev->dev, "%d", rbd_dev->dev_id);
        if (ret)
                goto err_out_mapping;
 
-       /* Everything's ready.  Announce the disk to the world. */
-
        set_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags);
        up_write(&rbd_dev->header_rwsem);
-
-       spin_lock(&rbd_dev_list_lock);
-       list_add_tail(&rbd_dev->node, &rbd_dev_list);
-       spin_unlock(&rbd_dev_list_lock);
-
-       add_disk(rbd_dev->disk);
-       pr_info("%s: capacity %llu features 0x%llx\n", rbd_dev->disk->disk_name,
-               (unsigned long long)get_capacity(rbd_dev->disk) << SECTOR_SHIFT,
-               rbd_dev->header.features);
-
-       return ret;
+       return 0;
 
 err_out_mapping:
        rbd_dev_mapping_clear(rbd_dev);
@@ -6131,11 +6122,30 @@ static ssize_t do_rbd_add(struct bus_type *bus,
        if (rc)
                goto err_out_image_probe;
 
+       /* Everything's ready.  Announce the disk to the world. */
+
+       rc = device_add(&rbd_dev->dev);
+       if (rc)
+               goto err_out_device_setup;
+
+       add_disk(rbd_dev->disk);
+       /* see rbd_init_disk() */
+       blk_put_queue(rbd_dev->disk->queue);
+
+       spin_lock(&rbd_dev_list_lock);
+       list_add_tail(&rbd_dev->node, &rbd_dev_list);
+       spin_unlock(&rbd_dev_list_lock);
+
+       pr_info("%s: capacity %llu features 0x%llx\n", rbd_dev->disk->disk_name,
+               (unsigned long long)get_capacity(rbd_dev->disk) << SECTOR_SHIFT,
+               rbd_dev->header.features);
        rc = count;
 out:
        module_put(THIS_MODULE);
        return rc;
 
+err_out_device_setup:
+       rbd_dev_device_release(rbd_dev);
 err_out_image_probe:
        rbd_dev_image_release(rbd_dev);
 err_out_rbd_dev:
@@ -6165,21 +6175,6 @@ static ssize_t rbd_add_single_major(struct bus_type *bus,
        return do_rbd_add(bus, buf, count);
 }
 
-static void rbd_dev_device_release(struct rbd_device *rbd_dev)
-{
-       rbd_free_disk(rbd_dev);
-
-       spin_lock(&rbd_dev_list_lock);
-       list_del_init(&rbd_dev->node);
-       spin_unlock(&rbd_dev_list_lock);
-
-       clear_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags);
-       device_del(&rbd_dev->dev);
-       rbd_dev_mapping_clear(rbd_dev);
-       if (!single_major)
-               unregister_blkdev(rbd_dev->major, rbd_dev->name);
-}
-
 static void rbd_dev_remove_parent(struct rbd_device *rbd_dev)
 {
        while (rbd_dev->parent) {
@@ -6266,6 +6261,12 @@ static ssize_t do_rbd_remove(struct bus_type *bus,
                blk_set_queue_dying(rbd_dev->disk->queue);
        }
 
+       del_gendisk(rbd_dev->disk);
+       spin_lock(&rbd_dev_list_lock);
+       list_del_init(&rbd_dev->node);
+       spin_unlock(&rbd_dev_list_lock);
+       device_del(&rbd_dev->dev);
+
        down_write(&rbd_dev->lock_rwsem);
        if (__rbd_is_lock_owner(rbd_dev))
                rbd_unlock(rbd_dev);