Btrfs: shared seed device
authorYan Zheng <zheng.yan@oracle.com>
Fri, 12 Dec 2008 15:03:26 +0000 (10:03 -0500)
committerChris Mason <chris.mason@oracle.com>
Fri, 12 Dec 2008 15:03:26 +0000 (10:03 -0500)
This patch makes seed device possible to be shared by
multiple mounted file systems. The sharing is achieved
by cloning seed device's btrfs_fs_devices structure.
Thanks you,

Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
fs/btrfs/disk-io.c
fs/btrfs/extent-tree.c
fs/btrfs/super.c
fs/btrfs/volumes.c
fs/btrfs/volumes.h

index 6f042de1ac43e3c7399057d37ecc131ce74364f5..541a8279ac71ab973458fa2e138ab3096245e70e 100644 (file)
@@ -1711,7 +1711,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
        }
 
        mutex_lock(&fs_info->chunk_mutex);
-       ret = btrfs_read_sys_array(tree_root, btrfs_super_bytenr(disk_super));
+       ret = btrfs_read_sys_array(tree_root);
        mutex_unlock(&fs_info->chunk_mutex);
        if (ret) {
                printk("btrfs: failed to read the system array on %s\n",
index 1cc89246ee2f4ae5a9fabb5b473eebc9cdb85cf5..171057a32679ffa4ba33179ee33f1a6e3ef74486 100644 (file)
@@ -218,7 +218,7 @@ static int cache_block_group(struct btrfs_root *root,
        struct btrfs_key key;
        struct extent_buffer *leaf;
        int slot;
-       u64 last = block_group->key.objectid;
+       u64 last;
 
        if (!block_group)
                return 0;
@@ -239,7 +239,8 @@ static int cache_block_group(struct btrfs_root *root,
         * skip the locking here
         */
        path->skip_locking = 1;
-       key.objectid = max_t(u64, last, BTRFS_SUPER_INFO_OFFSET);
+       last = max_t(u64, block_group->key.objectid, BTRFS_SUPER_INFO_OFFSET);
+       key.objectid = last;
        key.offset = 0;
        btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
        ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
@@ -5335,8 +5336,20 @@ static int noinline relocate_one_extent(struct btrfs_root *extent_root,
                        prev_block = block_start;
                }
 
-               if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID &&
-                   pass >= 2) {
+               btrfs_record_root_in_trans(found_root);
+               if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
+                       /*
+                        * try to update data extent references while
+                        * keeping metadata shared between snapshots.
+                        */
+                       if (pass == 1) {
+                               ret = relocate_one_path(trans, found_root,
+                                               path, &first_key, ref_path,
+                                               group, reloc_inode);
+                               if (ret < 0)
+                                       goto out;
+                               continue;
+                       }
                        /*
                         * use fallback method to process the remaining
                         * references.
@@ -5359,23 +5372,9 @@ static int noinline relocate_one_extent(struct btrfs_root *extent_root,
                                                path, extent_key,
                                                &first_key, ref_path,
                                                new_extents, nr_extents);
-                       if (ret < 0)
-                               goto out;
-                       continue;
-               }
-
-               btrfs_record_root_in_trans(found_root);
-               if (ref_path->owner_objectid < BTRFS_FIRST_FREE_OBJECTID) {
+               } else {
                        ret = relocate_tree_block(trans, found_root, path,
                                                  &first_key, ref_path);
-               } else {
-                       /*
-                        * try to update data extent references while
-                        * keeping metadata shared between snapshots.
-                        */
-                       ret = relocate_one_path(trans, found_root, path,
-                                               &first_key, ref_path,
-                                               group, reloc_inode);
                }
                if (ret < 0)
                        goto out;
index 09908f25fca9c759407377f51206a5cbe15b6a78..84c3b66564d0f16a34712fc4c0803771904ac362 100644 (file)
@@ -58,14 +58,15 @@ static struct super_operations btrfs_super_ops;
 static void btrfs_put_super (struct super_block * sb)
 {
        struct btrfs_root *root = btrfs_sb(sb);
-       struct btrfs_fs_info *fs = root->fs_info;
        int ret;
 
        ret = close_ctree(root);
        if (ret) {
                printk("close ctree returns %d\n", ret);
        }
-       btrfs_sysfs_del_super(fs);
+#if 0
+       btrfs_sysfs_del_super(root->fs_info);
+#endif
        sb->s_fs_info = NULL;
 }
 
@@ -349,11 +350,12 @@ static int btrfs_fill_super(struct super_block * sb,
                err = -ENOMEM;
                goto fail_close;
        }
-
+#if 0
        /* this does the super kobj at the same time */
        err = btrfs_sysfs_add_super(tree_root->fs_info);
        if (err)
                goto fail_close;
+#endif
 
        sb->s_root = root_dentry;
 
index 4d210a731d40ed7850c3e4816f9248148e344433..6672adcec9f827f22fdec1419b1d48f11a044c8e 100644 (file)
@@ -47,7 +47,6 @@ static int init_first_rw_device(struct btrfs_trans_handle *trans,
                                struct btrfs_device *device);
 static int btrfs_relocate_sys_chunks(struct btrfs_root *root);
 
-
 #define map_lookup_size(n) (sizeof(struct map_lookup) + \
                            (sizeof(struct btrfs_bio_stripe) * (n)))
 
@@ -74,34 +73,29 @@ static void unlock_chunks(struct btrfs_root *root)
        mutex_unlock(&root->fs_info->chunk_mutex);
 }
 
+static void free_fs_devices(struct btrfs_fs_devices *fs_devices)
+{
+       struct btrfs_device *device;
+       WARN_ON(fs_devices->opened);
+       while (!list_empty(&fs_devices->devices)) {
+               device = list_entry(fs_devices->devices.next,
+                                   struct btrfs_device, dev_list);
+               list_del(&device->dev_list);
+               kfree(device->name);
+               kfree(device);
+       }
+       kfree(fs_devices);
+}
+
 int btrfs_cleanup_fs_uuids(void)
 {
        struct btrfs_fs_devices *fs_devices;
-       struct btrfs_device *dev;
 
        while (!list_empty(&fs_uuids)) {
                fs_devices = list_entry(fs_uuids.next,
                                        struct btrfs_fs_devices, list);
                list_del(&fs_devices->list);
-               while(!list_empty(&fs_devices->devices)) {
-                       dev = list_entry(fs_devices->devices.next,
-                                        struct btrfs_device, dev_list);
-                       if (dev->bdev) {
-                               close_bdev_exclusive(dev->bdev, dev->mode);
-                               fs_devices->open_devices--;
-                       }
-                       fs_devices->num_devices--;
-                       if (dev->writeable)
-                               fs_devices->rw_devices--;
-                       list_del(&dev->dev_list);
-                       list_del(&dev->dev_alloc_list);
-                       kfree(dev->name);
-                       kfree(dev);
-               }
-               WARN_ON(fs_devices->num_devices);
-               WARN_ON(fs_devices->open_devices);
-               WARN_ON(fs_devices->rw_devices);
-               kfree(fs_devices);
+               free_fs_devices(fs_devices);
        }
        return 0;
 }
@@ -304,12 +298,55 @@ static noinline int device_list_add(const char *path,
        return 0;
 }
 
+static struct btrfs_fs_devices *clone_fs_devices(struct btrfs_fs_devices *orig)
+{
+       struct btrfs_fs_devices *fs_devices;
+       struct btrfs_device *device;
+       struct btrfs_device *orig_dev;
+
+       fs_devices = kzalloc(sizeof(*fs_devices), GFP_NOFS);
+       if (!fs_devices)
+               return ERR_PTR(-ENOMEM);
+
+       INIT_LIST_HEAD(&fs_devices->devices);
+       INIT_LIST_HEAD(&fs_devices->alloc_list);
+       INIT_LIST_HEAD(&fs_devices->list);
+       fs_devices->latest_devid = orig->latest_devid;
+       fs_devices->latest_trans = orig->latest_trans;
+       memcpy(fs_devices->fsid, orig->fsid, sizeof(fs_devices->fsid));
+
+       list_for_each_entry(orig_dev, &orig->devices, dev_list) {
+               device = kzalloc(sizeof(*device), GFP_NOFS);
+               if (!device)
+                       goto error;
+
+               device->name = kstrdup(orig_dev->name, GFP_NOFS);
+               if (!device->name)
+                       goto error;
+
+               device->devid = orig_dev->devid;
+               device->work.func = pending_bios_fn;
+               memcpy(device->uuid, orig_dev->uuid, sizeof(device->uuid));
+               device->barriers = 1;
+               spin_lock_init(&device->io_lock);
+               INIT_LIST_HEAD(&device->dev_list);
+               INIT_LIST_HEAD(&device->dev_alloc_list);
+
+               list_add(&device->dev_list, &fs_devices->devices);
+               device->fs_devices = fs_devices;
+               fs_devices->num_devices++;
+       }
+       return fs_devices;
+error:
+       free_fs_devices(fs_devices);
+       return ERR_PTR(-ENOMEM);
+}
+
 int btrfs_close_extra_devices(struct btrfs_fs_devices *fs_devices)
 {
        struct list_head *tmp;
        struct list_head *cur;
        struct btrfs_device *device;
-       int seed_devices = 0;
 
        mutex_lock(&uuid_mutex);
 again:
@@ -328,17 +365,14 @@ again:
                        device->writeable = 0;
                        fs_devices->rw_devices--;
                }
-               if (!seed_devices) {
-                       list_del_init(&device->dev_list);
-                       fs_devices->num_devices--;
-                       kfree(device->name);
-                       kfree(device);
-               }
+               list_del_init(&device->dev_list);
+               fs_devices->num_devices--;
+               kfree(device->name);
+               kfree(device);
        }
 
        if (fs_devices->seed) {
                fs_devices = fs_devices->seed;
-               seed_devices = 1;
                goto again;
        }
 
@@ -348,10 +382,9 @@ again:
 
 static int __btrfs_close_devices(struct btrfs_fs_devices *fs_devices)
 {
-       struct btrfs_fs_devices *seed_devices;
        struct list_head *cur;
        struct btrfs_device *device;
-again:
+
        if (--fs_devices->opened > 0)
                return 0;
 
@@ -370,31 +403,38 @@ again:
                device->writeable = 0;
                device->in_fs_metadata = 0;
        }
+       WARN_ON(fs_devices->open_devices);
+       WARN_ON(fs_devices->rw_devices);
        fs_devices->opened = 0;
        fs_devices->seeding = 0;
-       fs_devices->sprouted = 0;
 
-       seed_devices = fs_devices->seed;
-       fs_devices->seed = NULL;
-       if (seed_devices) {
-               fs_devices = seed_devices;
-               goto again;
-       }
        return 0;
 }
 
 int btrfs_close_devices(struct btrfs_fs_devices *fs_devices)
 {
+       struct btrfs_fs_devices *seed_devices = NULL;
        int ret;
 
        mutex_lock(&uuid_mutex);
        ret = __btrfs_close_devices(fs_devices);
+       if (!fs_devices->opened) {
+               seed_devices = fs_devices->seed;
+               fs_devices->seed = NULL;
+       }
        mutex_unlock(&uuid_mutex);
+
+       while (seed_devices) {
+               fs_devices = seed_devices;
+               seed_devices = fs_devices->seed;
+               __btrfs_close_devices(fs_devices);
+               free_fs_devices(fs_devices);
+       }
        return ret;
 }
 
-int __btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
-                        fmode_t flags, void *holder)
+static int __btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
+                               fmode_t flags, void *holder)
 {
        struct block_device *bdev;
        struct list_head *head = &fs_devices->devices;
@@ -490,12 +530,8 @@ int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
 
        mutex_lock(&uuid_mutex);
        if (fs_devices->opened) {
-               if (fs_devices->sprouted) {
-                       ret = -EBUSY;
-               } else {
-                       fs_devices->opened++;
-                       ret = 0;
-               }
+               fs_devices->opened++;
+               ret = 0;
        } else {
                ret = __btrfs_open_devices(fs_devices, flags, holder);
        }
@@ -1043,12 +1079,8 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
                goto error_brelse;
 
        device->in_fs_metadata = 0;
-       if (device->fs_devices == root->fs_info->fs_devices) {
-               list_del_init(&device->dev_list);
-               root->fs_info->fs_devices->num_devices--;
-               if (device->bdev)
-                       device->fs_devices->open_devices--;
-       }
+       list_del_init(&device->dev_list);
+       device->fs_devices->num_devices--;
 
        next_device = list_entry(root->fs_info->fs_devices->devices.next,
                                 struct btrfs_device, dev_list);
@@ -1057,34 +1089,27 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
        if (device->bdev == root->fs_info->fs_devices->latest_bdev)
                root->fs_info->fs_devices->latest_bdev = next_device->bdev;
 
+       if (device->bdev) {
+               close_bdev_exclusive(device->bdev, device->mode);
+               device->bdev = NULL;
+               device->fs_devices->open_devices--;
+       }
+
        num_devices = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
        btrfs_set_super_num_devices(&root->fs_info->super_copy, num_devices);
 
-       if (device->fs_devices != root->fs_info->fs_devices) {
-               BUG_ON(device->writeable);
-               brelse(bh);
-               if (bdev)
-                       close_bdev_exclusive(bdev, FMODE_READ);
-
-               if (device->bdev) {
-                       close_bdev_exclusive(device->bdev, device->mode);
-                       device->bdev = NULL;
-                       device->fs_devices->open_devices--;
-               }
-               if (device->fs_devices->open_devices == 0) {
-                       struct btrfs_fs_devices *fs_devices;
-                       fs_devices = root->fs_info->fs_devices;
-                       while (fs_devices) {
-                               if (fs_devices->seed == device->fs_devices)
-                                       break;
-                               fs_devices = fs_devices->seed;
-                       }
-                       fs_devices->seed = device->fs_devices->seed;
-                       device->fs_devices->seed = NULL;
-                       __btrfs_close_devices(device->fs_devices);
+       if (device->fs_devices->open_devices == 0) {
+               struct btrfs_fs_devices *fs_devices;
+               fs_devices = root->fs_info->fs_devices;
+               while (fs_devices) {
+                       if (fs_devices->seed == device->fs_devices)
+                               break;
+                       fs_devices = fs_devices->seed;
                }
-               ret = 0;
-               goto out;
+               fs_devices->seed = device->fs_devices->seed;
+               device->fs_devices->seed = NULL;
+               __btrfs_close_devices(device->fs_devices);
+               free_fs_devices(device->fs_devices);
        }
 
        /*
@@ -1099,20 +1124,10 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
                set_buffer_dirty(bh);
                sync_dirty_buffer(bh);
        }
-       brelse(bh);
 
-       if (device->bdev) {
-               /* one close for the device struct or super_block */
-               close_bdev_exclusive(device->bdev, device->mode);
-       }
-       if (bdev) {
-               /* one close for us */
-               close_bdev_exclusive(bdev, FMODE_READ);
-       }
        kfree(device->name);
        kfree(device);
        ret = 0;
-       goto out;
 
 error_brelse:
        brelse(bh);
@@ -1133,34 +1148,41 @@ static int btrfs_prepare_sprout(struct btrfs_trans_handle *trans,
 {
        struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
        struct btrfs_fs_devices *old_devices;
+       struct btrfs_fs_devices *seed_devices;
        struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
        struct btrfs_device *device;
        u64 super_flags;
 
        BUG_ON(!mutex_is_locked(&uuid_mutex));
-       if (!fs_devices->seeding || fs_devices->opened != 1)
+       if (!fs_devices->seeding)
                return -EINVAL;
 
-       old_devices = kzalloc(sizeof(*fs_devices), GFP_NOFS);
-       if (!old_devices)
+       seed_devices = kzalloc(sizeof(*fs_devices), GFP_NOFS);
+       if (!seed_devices)
                return -ENOMEM;
 
-       memcpy(old_devices, fs_devices, sizeof(*old_devices));
-       old_devices->opened = 1;
-       old_devices->sprouted = 1;
-       INIT_LIST_HEAD(&old_devices->devices);
-       INIT_LIST_HEAD(&old_devices->alloc_list);
-       list_splice_init(&fs_devices->devices, &old_devices->devices);
-       list_splice_init(&fs_devices->alloc_list, &old_devices->alloc_list);
-       list_for_each_entry(device, &old_devices->devices, dev_list) {
-               device->fs_devices = old_devices;
+       old_devices = clone_fs_devices(fs_devices);
+       if (IS_ERR(old_devices)) {
+               kfree(seed_devices);
+               return PTR_ERR(old_devices);
        }
+
        list_add(&old_devices->list, &fs_uuids);
 
+       memcpy(seed_devices, fs_devices, sizeof(*seed_devices));
+       seed_devices->opened = 1;
+       INIT_LIST_HEAD(&seed_devices->devices);
+       INIT_LIST_HEAD(&seed_devices->alloc_list);
+       list_splice_init(&fs_devices->devices, &seed_devices->devices);
+       list_splice_init(&fs_devices->alloc_list, &seed_devices->alloc_list);
+       list_for_each_entry(device, &seed_devices->devices, dev_list) {
+               device->fs_devices = seed_devices;
+       }
+
        fs_devices->seeding = 0;
        fs_devices->num_devices = 0;
        fs_devices->open_devices = 0;
-       fs_devices->seed = old_devices;
+       fs_devices->seed = seed_devices;
 
        generate_random_uuid(fs_devices->fsid);
        memcpy(root->fs_info->fsid, fs_devices->fsid, BTRFS_FSID_SIZE);
@@ -2642,7 +2664,6 @@ int btrfs_unplug_page(struct btrfs_mapping_tree *map_tree,
                                 NULL, 0, page);
 }
 
-
 static void end_bio_multi_stripe(struct bio *bio, int err)
 {
        struct btrfs_multi_bio *multi = bio->bi_private;
@@ -2840,6 +2861,7 @@ static struct btrfs_device *add_missing_dev(struct btrfs_root *root,
        device->dev_root = root->fs_info->dev_root;
        device->devid = devid;
        device->work.func = pending_bios_fn;
+       device->fs_devices = fs_devices;
        fs_devices->num_devices++;
        spin_lock_init(&device->io_lock);
        INIT_LIST_HEAD(&device->dev_alloc_list);
@@ -2980,8 +3002,10 @@ static int open_seed_devices(struct btrfs_root *root, u8 *fsid)
                ret = -ENOENT;
                goto out;
        }
-       if (fs_devices->opened) {
-               ret = -EBUSY;
+
+       fs_devices = clone_fs_devices(fs_devices);
+       if (IS_ERR(fs_devices)) {
+               ret = PTR_ERR(fs_devices);
                goto out;
        }
 
@@ -2992,13 +3016,13 @@ static int open_seed_devices(struct btrfs_root *root, u8 *fsid)
 
        if (!fs_devices->seeding) {
                __btrfs_close_devices(fs_devices);
+               free_fs_devices(fs_devices);
                ret = -EINVAL;
                goto out;
        }
 
        fs_devices->seed = root->fs_info->fs_devices->seed;
        root->fs_info->fs_devices->seed = fs_devices;
-       fs_devices->sprouted = 1;
 out:
        mutex_unlock(&uuid_mutex);
        return ret;
@@ -3011,7 +3035,6 @@ static int read_one_dev(struct btrfs_root *root,
        struct btrfs_device *device;
        u64 devid;
        int ret;
-       int seed_devices = 0;
        u8 fs_uuid[BTRFS_UUID_SIZE];
        u8 dev_uuid[BTRFS_UUID_SIZE];
 
@@ -3025,14 +3048,13 @@ static int read_one_dev(struct btrfs_root *root,
 
        if (memcmp(fs_uuid, root->fs_info->fsid, BTRFS_UUID_SIZE)) {
                ret = open_seed_devices(root, fs_uuid);
-               if (ret)
+               if (ret && !btrfs_test_opt(root, DEGRADED))
                        return ret;
-               seed_devices = 1;
        }
 
        device = btrfs_find_device(root, devid, dev_uuid, fs_uuid);
        if (!device || !device->bdev) {
-               if (!btrfs_test_opt(root, DEGRADED) || seed_devices)
+               if (!btrfs_test_opt(root, DEGRADED))
                        return -EIO;
 
                if (!device) {
@@ -3074,7 +3096,7 @@ int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf)
        return read_one_dev(root, buf, dev_item);
 }
 
-int btrfs_read_sys_array(struct btrfs_root *root, u64 sb_bytenr)
+int btrfs_read_sys_array(struct btrfs_root *root)
 {
        struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
        struct extent_buffer *sb;
@@ -3089,7 +3111,7 @@ int btrfs_read_sys_array(struct btrfs_root *root, u64 sb_bytenr)
        u32 cur;
        struct btrfs_key key;
 
-       sb = btrfs_find_create_tree_block(root, sb_bytenr,
+       sb = btrfs_find_create_tree_block(root, BTRFS_SUPER_INFO_OFFSET,
                                          BTRFS_SUPER_INFO_SIZE);
        if (!sb)
                return -ENOMEM;
index bdebe83c31951622ab43c4daadcbbe89a0a92529..86c44e9ae1101e2332dadad41b6def15932ae96c 100644 (file)
@@ -93,7 +93,6 @@ struct btrfs_fs_devices {
 
        struct btrfs_fs_devices *seed;
        int seeding;
-       int sprouted;
 
        int opened;
 };
@@ -127,7 +126,7 @@ int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw,
 int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree,
                     u64 chunk_start, u64 physical, u64 devid,
                     u64 **logical, int *naddrs, int *stripe_len);
-int btrfs_read_sys_array(struct btrfs_root *root, u64 sb_bytenr);
+int btrfs_read_sys_array(struct btrfs_root *root);
 int btrfs_read_chunk_tree(struct btrfs_root *root);
 int btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
                      struct btrfs_root *extent_root, u64 type);