Btrfs: fix hash overflow handling

The handling for directory crc hash overflows was fairly obscure,
split_leaf returns EOVERFLOW when we try to extend the item and that is
supposed to bubble up to userland.  For a while it did so, but along the
way we added better handling of errors and forced the FS readonly if we
hit IO errors during the directory insertion.

Along the way, we started testing only for EEXIST and the EOVERFLOW case
was dropped.  The end result is that we may force the FS readonly if we
catch a directory hash bucket overflow.

This fixes a few problem spots.  First I add tests for EOVERFLOW in the
places where we can safely just return the error up the chain.

btrfs_rename is harder though, because it tries to insert the new
directory item only after it has already unlinked anything the rename
was going to overwrite.  Rather than adding very complex logic, I added
a helper to test for the hash overflow case early while it is still safe
to bail out.

Snapshot and subvolume creation had a similar problem, so they are using
the new helper now too.

Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Reported-by: Pascal Junod <pascal@junod.info>

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 44d9bc87e86318d447001825a8070a7b19058a5d..547b7b05727f917dfc2bad6516f5c5b30ea12c68 100644 (file)
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -3283,6 +3283,8 @@ void btrfs_update_root_times(struct btrfs_trans_handle *trans,
                             struct btrfs_root *root);
 
 /* dir-item.c */
+int btrfs_check_dir_item_collision(struct btrfs_root *root, u64 dir,
+                         const char *name, int name_len);
 int btrfs_insert_dir_item(struct btrfs_trans_handle *trans,
                          struct btrfs_root *root, const char *name,
                          int name_len, struct inode *dir,

diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c
index c1a074d0696ff897258c576c127af7fd8d513846..502c2158167c8fb95578f8166a9802a34e762def 100644 (file)
--- a/fs/btrfs/dir-item.c
+++ b/fs/btrfs/dir-item.c
@@ -213,6 +213,65 @@ struct btrfs_dir_item *btrfs_lookup_dir_item(struct btrfs_trans_handle *trans,
        return btrfs_match_dir_item_name(root, path, name, name_len);
 }
 
+int btrfs_check_dir_item_collision(struct btrfs_root *root, u64 dir,
+                                  const char *name, int name_len)
+{
+       int ret;
+       struct btrfs_key key;
+       struct btrfs_dir_item *di;
+       int data_size;
+       struct extent_buffer *leaf;
+       int slot;
+       struct btrfs_path *path;
+
+
+       path = btrfs_alloc_path();
+       if (!path)
+               return -ENOMEM;
+
+       key.objectid = dir;
+       btrfs_set_key_type(&key, BTRFS_DIR_ITEM_KEY);
+       key.offset = btrfs_name_hash(name, name_len);
+
+       ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+
+       /* return back any errors */
+       if (ret < 0)
+               goto out;
+
+       /* nothing found, we're safe */
+       if (ret > 0) {
+               ret = 0;
+               goto out;
+       }
+
+       /* we found an item, look for our name in the item */
+       di = btrfs_match_dir_item_name(root, path, name, name_len);
+       if (di) {
+               /* our exact name was found */
+               ret = -EEXIST;
+               goto out;
+       }
+
+       /*
+        * see if there is room in the item to insert this
+        * name
+        */
+       data_size = sizeof(*di) + name_len + sizeof(struct btrfs_item);
+       leaf = path->nodes[0];
+       slot = path->slots[0];
+       if (data_size + btrfs_item_size_nr(leaf, slot) +
+           sizeof(struct btrfs_item) > BTRFS_LEAF_DATA_SIZE(root)) {
+               ret = -EOVERFLOW;
+       } else {
+               /* plenty of insertion room */
+               ret = 0;
+       }
+out:
+       btrfs_free_path(path);
+       return ret;
+}
+
 /*
  * lookup a directory item based on index.  'dir' is the objectid
  * we're searching in, and 'mod' tells us if you plan on deleting the

diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 2e6918c85b720a06eb54ad82f4953c653717338b..e95b1f90a1f6676b73514877db83fc72986df66a 100644 (file)
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -4885,7 +4885,7 @@ int btrfs_add_link(struct btrfs_trans_handle *trans,
        ret = btrfs_insert_dir_item(trans, root, name, name_len,
                                    parent_inode, &key,
                                    btrfs_inode_type(inode), index);
-       if (ret == -EEXIST)
+       if (ret == -EEXIST || ret == -EOVERFLOW)
                goto fail_dir_item;
        else if (ret) {
                btrfs_abort_transaction(trans, root, ret);
@@ -7336,6 +7336,28 @@ static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry,
        if (S_ISDIR(old_inode->i_mode) && new_inode &&
            new_inode->i_size > BTRFS_EMPTY_DIR_SIZE)
                return -ENOTEMPTY;
+
+
+       /* check for collisions, even if the  name isn't there */
+       ret = btrfs_check_dir_item_collision(root, new_dir->i_ino,
+                            new_dentry->d_name.name,
+                            new_dentry->d_name.len);
+
+       if (ret) {
+               if (ret == -EEXIST) {
+                       /* we shouldn't get
+                        * eexist without a new_inode */
+                       if (!new_inode) {
+                               WARN_ON(1);
+                               return ret;
+                       }
+               } else {
+                       /* maybe -EOVERFLOW */
+                       return ret;
+               }
+       }
+       ret = 0;
+
        /*
         * we're using rename to replace one file with another.
         * and the replacement file is large.  Start IO on it now so

diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index 657d83ca9dea75312406761403080d975f9692a6..d4608ab72b791682803da6761e52335cb367f31e 100644 (file)
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -710,6 +710,16 @@ static noinline int btrfs_mksubvol(struct path *parent,
        if (error)
                goto out_dput;
 
+       /*
+        * even if this name doesn't exist, we may get hash collisions.
+        * check for them now when we can safely fail
+        */
+       error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
+                                              dir->i_ino, name,
+                                              namelen);
+       if (error)
+               goto out_dput;
+
        down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
 
        if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)

diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index e6509b92433b2ac23b7dbbcbfc8c9b54d6107c0f..87fac9a21ea56578625536ac1229678e854ec5f7 100644 (file)
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -1190,7 +1190,7 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
                                    parent_inode, &key,
                                    BTRFS_FT_DIR, index);
        /* We have check then name at the beginning, so it is impossible. */
-       BUG_ON(ret == -EEXIST);
+       BUG_ON(ret == -EEXIST || ret == -EOVERFLOW);
        if (ret) {
                btrfs_abort_transaction(trans, root, ret);
                goto fail;