Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / fs / btrfs / root-tree.c
diff --git a/kernel/fs/btrfs/root-tree.c b/kernel/fs/btrfs/root-tree.c
new file mode 100644 (file)
index 0000000..360a728
--- /dev/null
@@ -0,0 +1,497 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/err.h>
+#include <linux/uuid.h>
+#include "ctree.h"
+#include "transaction.h"
+#include "disk-io.h"
+#include "print-tree.h"
+
+/*
+ * Read a root item from the tree. In case we detect a root item smaller then
+ * sizeof(root_item), we know it's an old version of the root structure and
+ * initialize all new fields to zero. The same happens if we detect mismatching
+ * generation numbers as then we know the root was once mounted with an older
+ * kernel that was not aware of the root item structure change.
+ */
+static void btrfs_read_root_item(struct extent_buffer *eb, int slot,
+                               struct btrfs_root_item *item)
+{
+       uuid_le uuid;
+       int len;
+       int need_reset = 0;
+
+       len = btrfs_item_size_nr(eb, slot);
+       read_extent_buffer(eb, item, btrfs_item_ptr_offset(eb, slot),
+                       min_t(int, len, (int)sizeof(*item)));
+       if (len < sizeof(*item))
+               need_reset = 1;
+       if (!need_reset && btrfs_root_generation(item)
+               != btrfs_root_generation_v2(item)) {
+               if (btrfs_root_generation_v2(item) != 0) {
+                       printk(KERN_WARNING "BTRFS: mismatching "
+                                       "generation and generation_v2 "
+                                       "found in root item. This root "
+                                       "was probably mounted with an "
+                                       "older kernel. Resetting all "
+                                       "new fields.\n");
+               }
+               need_reset = 1;
+       }
+       if (need_reset) {
+               memset(&item->generation_v2, 0,
+                       sizeof(*item) - offsetof(struct btrfs_root_item,
+                                       generation_v2));
+
+               uuid_le_gen(&uuid);
+               memcpy(item->uuid, uuid.b, BTRFS_UUID_SIZE);
+       }
+}
+
+/*
+ * btrfs_find_root - lookup the root by the key.
+ * root: the root of the root tree
+ * search_key: the key to search
+ * path: the path we search
+ * root_item: the root item of the tree we look for
+ * root_key: the reak key of the tree we look for
+ *
+ * If ->offset of 'seach_key' is -1ULL, it means we are not sure the offset
+ * of the search key, just lookup the root with the highest offset for a
+ * given objectid.
+ *
+ * If we find something return 0, otherwise > 0, < 0 on error.
+ */
+int btrfs_find_root(struct btrfs_root *root, struct btrfs_key *search_key,
+                   struct btrfs_path *path, struct btrfs_root_item *root_item,
+                   struct btrfs_key *root_key)
+{
+       struct btrfs_key found_key;
+       struct extent_buffer *l;
+       int ret;
+       int slot;
+
+       ret = btrfs_search_slot(NULL, root, search_key, path, 0, 0);
+       if (ret < 0)
+               return ret;
+
+       if (search_key->offset != -1ULL) {      /* the search key is exact */
+               if (ret > 0)
+                       goto out;
+       } else {
+               BUG_ON(ret == 0);               /* Logical error */
+               if (path->slots[0] == 0)
+                       goto out;
+               path->slots[0]--;
+               ret = 0;
+       }
+
+       l = path->nodes[0];
+       slot = path->slots[0];
+
+       btrfs_item_key_to_cpu(l, &found_key, slot);
+       if (found_key.objectid != search_key->objectid ||
+           found_key.type != BTRFS_ROOT_ITEM_KEY) {
+               ret = 1;
+               goto out;
+       }
+
+       if (root_item)
+               btrfs_read_root_item(l, slot, root_item);
+       if (root_key)
+               memcpy(root_key, &found_key, sizeof(found_key));
+out:
+       btrfs_release_path(path);
+       return ret;
+}
+
+void btrfs_set_root_node(struct btrfs_root_item *item,
+                        struct extent_buffer *node)
+{
+       btrfs_set_root_bytenr(item, node->start);
+       btrfs_set_root_level(item, btrfs_header_level(node));
+       btrfs_set_root_generation(item, btrfs_header_generation(node));
+}
+
+/*
+ * copy the data in 'item' into the btree
+ */
+int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root
+                     *root, struct btrfs_key *key, struct btrfs_root_item
+                     *item)
+{
+       struct btrfs_path *path;
+       struct extent_buffer *l;
+       int ret;
+       int slot;
+       unsigned long ptr;
+       int old_len;
+
+       path = btrfs_alloc_path();
+       if (!path)
+               return -ENOMEM;
+
+       ret = btrfs_search_slot(trans, root, key, path, 0, 1);
+       if (ret < 0) {
+               btrfs_abort_transaction(trans, root, ret);
+               goto out;
+       }
+
+       if (ret != 0) {
+               btrfs_print_leaf(root, path->nodes[0]);
+               btrfs_crit(root->fs_info, "unable to update root key %llu %u %llu",
+                      key->objectid, key->type, key->offset);
+               BUG_ON(1);
+       }
+
+       l = path->nodes[0];
+       slot = path->slots[0];
+       ptr = btrfs_item_ptr_offset(l, slot);
+       old_len = btrfs_item_size_nr(l, slot);
+
+       /*
+        * If this is the first time we update the root item which originated
+        * from an older kernel, we need to enlarge the item size to make room
+        * for the added fields.
+        */
+       if (old_len < sizeof(*item)) {
+               btrfs_release_path(path);
+               ret = btrfs_search_slot(trans, root, key, path,
+                               -1, 1);
+               if (ret < 0) {
+                       btrfs_abort_transaction(trans, root, ret);
+                       goto out;
+               }
+
+               ret = btrfs_del_item(trans, root, path);
+               if (ret < 0) {
+                       btrfs_abort_transaction(trans, root, ret);
+                       goto out;
+               }
+               btrfs_release_path(path);
+               ret = btrfs_insert_empty_item(trans, root, path,
+                               key, sizeof(*item));
+               if (ret < 0) {
+                       btrfs_abort_transaction(trans, root, ret);
+                       goto out;
+               }
+               l = path->nodes[0];
+               slot = path->slots[0];
+               ptr = btrfs_item_ptr_offset(l, slot);
+       }
+
+       /*
+        * Update generation_v2 so at the next mount we know the new root
+        * fields are valid.
+        */
+       btrfs_set_root_generation_v2(item, btrfs_root_generation(item));
+
+       write_extent_buffer(l, item, ptr, sizeof(*item));
+       btrfs_mark_buffer_dirty(path->nodes[0]);
+out:
+       btrfs_free_path(path);
+       return ret;
+}
+
+int btrfs_insert_root(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+                     struct btrfs_key *key, struct btrfs_root_item *item)
+{
+       /*
+        * Make sure generation v1 and v2 match. See update_root for details.
+        */
+       btrfs_set_root_generation_v2(item, btrfs_root_generation(item));
+       return btrfs_insert_item(trans, root, key, item, sizeof(*item));
+}
+
+int btrfs_find_orphan_roots(struct btrfs_root *tree_root)
+{
+       struct extent_buffer *leaf;
+       struct btrfs_path *path;
+       struct btrfs_key key;
+       struct btrfs_key root_key;
+       struct btrfs_root *root;
+       int err = 0;
+       int ret;
+       bool can_recover = true;
+
+       if (tree_root->fs_info->sb->s_flags & MS_RDONLY)
+               can_recover = false;
+
+       path = btrfs_alloc_path();
+       if (!path)
+               return -ENOMEM;
+
+       key.objectid = BTRFS_ORPHAN_OBJECTID;
+       key.type = BTRFS_ORPHAN_ITEM_KEY;
+       key.offset = 0;
+
+       root_key.type = BTRFS_ROOT_ITEM_KEY;
+       root_key.offset = (u64)-1;
+
+       while (1) {
+               ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0);
+               if (ret < 0) {
+                       err = ret;
+                       break;
+               }
+
+               leaf = path->nodes[0];
+               if (path->slots[0] >= btrfs_header_nritems(leaf)) {
+                       ret = btrfs_next_leaf(tree_root, path);
+                       if (ret < 0)
+                               err = ret;
+                       if (ret != 0)
+                               break;
+                       leaf = path->nodes[0];
+               }
+
+               btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
+               btrfs_release_path(path);
+
+               if (key.objectid != BTRFS_ORPHAN_OBJECTID ||
+                   key.type != BTRFS_ORPHAN_ITEM_KEY)
+                       break;
+
+               root_key.objectid = key.offset;
+               key.offset++;
+
+               root = btrfs_read_fs_root(tree_root, &root_key);
+               err = PTR_ERR_OR_ZERO(root);
+               if (err && err != -ENOENT) {
+                       break;
+               } else if (err == -ENOENT) {
+                       struct btrfs_trans_handle *trans;
+
+                       btrfs_release_path(path);
+
+                       trans = btrfs_join_transaction(tree_root);
+                       if (IS_ERR(trans)) {
+                               err = PTR_ERR(trans);
+                               btrfs_error(tree_root->fs_info, err,
+                                           "Failed to start trans to delete "
+                                           "orphan item");
+                               break;
+                       }
+                       err = btrfs_del_orphan_item(trans, tree_root,
+                                                   root_key.objectid);
+                       btrfs_end_transaction(trans, tree_root);
+                       if (err) {
+                               btrfs_error(tree_root->fs_info, err,
+                                           "Failed to delete root orphan "
+                                           "item");
+                               break;
+                       }
+                       continue;
+               }
+
+               err = btrfs_init_fs_root(root);
+               if (err) {
+                       btrfs_free_fs_root(root);
+                       break;
+               }
+
+               set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
+
+               err = btrfs_insert_fs_root(root->fs_info, root);
+               if (err) {
+                       BUG_ON(err == -EEXIST);
+                       btrfs_free_fs_root(root);
+                       break;
+               }
+
+               if (btrfs_root_refs(&root->root_item) == 0)
+                       btrfs_add_dead_root(root);
+       }
+
+       btrfs_free_path(path);
+       return err;
+}
+
+/* drop the root item for 'key' from 'root' */
+int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+                  struct btrfs_key *key)
+{
+       struct btrfs_path *path;
+       int ret;
+
+       path = btrfs_alloc_path();
+       if (!path)
+               return -ENOMEM;
+       ret = btrfs_search_slot(trans, root, key, path, -1, 1);
+       if (ret < 0)
+               goto out;
+
+       BUG_ON(ret != 0);
+
+       ret = btrfs_del_item(trans, root, path);
+out:
+       btrfs_free_path(path);
+       return ret;
+}
+
+int btrfs_del_root_ref(struct btrfs_trans_handle *trans,
+                      struct btrfs_root *tree_root,
+                      u64 root_id, u64 ref_id, u64 dirid, u64 *sequence,
+                      const char *name, int name_len)
+
+{
+       struct btrfs_path *path;
+       struct btrfs_root_ref *ref;
+       struct extent_buffer *leaf;
+       struct btrfs_key key;
+       unsigned long ptr;
+       int err = 0;
+       int ret;
+
+       path = btrfs_alloc_path();
+       if (!path)
+               return -ENOMEM;
+
+       key.objectid = root_id;
+       key.type = BTRFS_ROOT_BACKREF_KEY;
+       key.offset = ref_id;
+again:
+       ret = btrfs_search_slot(trans, tree_root, &key, path, -1, 1);
+       BUG_ON(ret < 0);
+       if (ret == 0) {
+               leaf = path->nodes[0];
+               ref = btrfs_item_ptr(leaf, path->slots[0],
+                                    struct btrfs_root_ref);
+
+               WARN_ON(btrfs_root_ref_dirid(leaf, ref) != dirid);
+               WARN_ON(btrfs_root_ref_name_len(leaf, ref) != name_len);
+               ptr = (unsigned long)(ref + 1);
+               WARN_ON(memcmp_extent_buffer(leaf, name, ptr, name_len));
+               *sequence = btrfs_root_ref_sequence(leaf, ref);
+
+               ret = btrfs_del_item(trans, tree_root, path);
+               if (ret) {
+                       err = ret;
+                       goto out;
+               }
+       } else
+               err = -ENOENT;
+
+       if (key.type == BTRFS_ROOT_BACKREF_KEY) {
+               btrfs_release_path(path);
+               key.objectid = ref_id;
+               key.type = BTRFS_ROOT_REF_KEY;
+               key.offset = root_id;
+               goto again;
+       }
+
+out:
+       btrfs_free_path(path);
+       return err;
+}
+
+/*
+ * add a btrfs_root_ref item.  type is either BTRFS_ROOT_REF_KEY
+ * or BTRFS_ROOT_BACKREF_KEY.
+ *
+ * The dirid, sequence, name and name_len refer to the directory entry
+ * that is referencing the root.
+ *
+ * For a forward ref, the root_id is the id of the tree referencing
+ * the root and ref_id is the id of the subvol  or snapshot.
+ *
+ * For a back ref the root_id is the id of the subvol or snapshot and
+ * ref_id is the id of the tree referencing it.
+ *
+ * Will return 0, -ENOMEM, or anything from the CoW path
+ */
+int btrfs_add_root_ref(struct btrfs_trans_handle *trans,
+                      struct btrfs_root *tree_root,
+                      u64 root_id, u64 ref_id, u64 dirid, u64 sequence,
+                      const char *name, int name_len)
+{
+       struct btrfs_key key;
+       int ret;
+       struct btrfs_path *path;
+       struct btrfs_root_ref *ref;
+       struct extent_buffer *leaf;
+       unsigned long ptr;
+
+       path = btrfs_alloc_path();
+       if (!path)
+               return -ENOMEM;
+
+       key.objectid = root_id;
+       key.type = BTRFS_ROOT_BACKREF_KEY;
+       key.offset = ref_id;
+again:
+       ret = btrfs_insert_empty_item(trans, tree_root, path, &key,
+                                     sizeof(*ref) + name_len);
+       if (ret) {
+               btrfs_abort_transaction(trans, tree_root, ret);
+               btrfs_free_path(path);
+               return ret;
+       }
+
+       leaf = path->nodes[0];
+       ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_ref);
+       btrfs_set_root_ref_dirid(leaf, ref, dirid);
+       btrfs_set_root_ref_sequence(leaf, ref, sequence);
+       btrfs_set_root_ref_name_len(leaf, ref, name_len);
+       ptr = (unsigned long)(ref + 1);
+       write_extent_buffer(leaf, name, ptr, name_len);
+       btrfs_mark_buffer_dirty(leaf);
+
+       if (key.type == BTRFS_ROOT_BACKREF_KEY) {
+               btrfs_release_path(path);
+               key.objectid = ref_id;
+               key.type = BTRFS_ROOT_REF_KEY;
+               key.offset = root_id;
+               goto again;
+       }
+
+       btrfs_free_path(path);
+       return 0;
+}
+
+/*
+ * Old btrfs forgets to init root_item->flags and root_item->byte_limit
+ * for subvolumes. To work around this problem, we steal a bit from
+ * root_item->inode_item->flags, and use it to indicate if those fields
+ * have been properly initialized.
+ */
+void btrfs_check_and_init_root_item(struct btrfs_root_item *root_item)
+{
+       u64 inode_flags = btrfs_stack_inode_flags(&root_item->inode);
+
+       if (!(inode_flags & BTRFS_INODE_ROOT_ITEM_INIT)) {
+               inode_flags |= BTRFS_INODE_ROOT_ITEM_INIT;
+               btrfs_set_stack_inode_flags(&root_item->inode, inode_flags);
+               btrfs_set_root_flags(root_item, 0);
+               btrfs_set_root_limit(root_item, 0);
+       }
+}
+
+void btrfs_update_root_times(struct btrfs_trans_handle *trans,
+                            struct btrfs_root *root)
+{
+       struct btrfs_root_item *item = &root->root_item;
+       struct timespec ct = CURRENT_TIME;
+
+       spin_lock(&root->root_item_lock);
+       btrfs_set_root_ctransid(item, trans->transid);
+       btrfs_set_stack_timespec_sec(&item->ctime, ct.tv_sec);
+       btrfs_set_stack_timespec_nsec(&item->ctime, ct.tv_nsec);
+       spin_unlock(&root->root_item_lock);
+}