ocfs2: should reclaim the inode if '__ocfs2_mknod_locked' returns an error
authoralex chen <alex.chen@huawei.com>
Fri, 6 Nov 2015 02:44:10 +0000 (18:44 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 6 Nov 2015 03:34:48 +0000 (19:34 -0800)
In ocfs2_mknod_locked if '__ocfs2_mknod_locke d' returns an error, we
should reclaim the inode successfully claimed above, otherwise, the
inode never be reused. The case is described below:

ocfs2_mknod
    ocfs2_mknod_locked
        ocfs2_claim_new_inode
                Successfully claim the inode
        __ocfs2_mknod_locked
            ocfs2_journal_access_di
            Failed because of -ENOMEM or other reasons, the inode
                        lockres has not been initialized yet.

    iput(inode)
        ocfs2_evict_inode
            ocfs2_delete_inode
                ocfs2_inode_lock
                    ocfs2_inode_lock_full_nested
                        __ocfs2_cluster_lock
                                Return -EINVAL because of the inode
                                lockres has not been initialized.

                So the following operations are not performed
                ocfs2_wipe_inode
                        ocfs2_remove_inode
                                ocfs2_free_dinode
                                        ocfs2_free_suballoc_bits

Signed-off-by: Alex Chen <alex.chen@huawei.com>
Reviewed-by: Joseph Qi <joseph.qi@huawei.com>
Cc: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/ocfs2/namei.c

index 1206392a071ffd9d1bd21312c8e727be29afabfa..3b48ac25d8a7c450ac23360b5f7d28e9fafcc263 100644 (file)
@@ -655,9 +655,18 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb,
                return status;
        }
 
-       return __ocfs2_mknod_locked(dir, inode, dev, new_fe_bh,
+       status = __ocfs2_mknod_locked(dir, inode, dev, new_fe_bh,
                                    parent_fe_bh, handle, inode_ac,
                                    fe_blkno, suballoc_loc, suballoc_bit);
+       if (status < 0) {
+               u64 bg_blkno = ocfs2_which_suballoc_group(fe_blkno, suballoc_bit);
+               int tmp = ocfs2_free_suballoc_bits(handle, inode_ac->ac_inode,
+                               inode_ac->ac_bh, suballoc_bit, bg_blkno, 1);
+               if (tmp)
+                       mlog_errno(tmp);
+       }
+
+       return status;
 }
 
 static int ocfs2_mkdir(struct inode *dir,