ocfs2: fix error handling in ocfs2_ioctl_move_extents()
authorDan Carpenter <dan.carpenter@oracle.com>
Mon, 29 Apr 2013 22:05:58 +0000 (15:05 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Mon, 29 Apr 2013 22:54:27 +0000 (15:54 -0700)
Smatch complains that if we hit an error (for example if the file is
immutable) then "range" has uninitialized stack data and we copy it to
the user.

I've re-written the error handling to avoid this problem and make it a
little cleaner as well.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Jie Liu <jeff.liu@oracle.com>
Reviewed-by: Jie Liu <jeff.liu@oracle.com>
Cc: Mark Fasheh <mfasheh@suse.com>
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/move_extents.c

index 9f8dcad..8f3d3cb 100644 (file)
@@ -1057,42 +1057,40 @@ int ocfs2_ioctl_move_extents(struct file *filp, void __user *argp)
 
        struct inode *inode = file_inode(filp);
        struct ocfs2_move_extents range;
-       struct ocfs2_move_extents_context *context = NULL;
+       struct ocfs2_move_extents_context *context;
+
+       if (!argp)
+               return -EINVAL;
 
        status = mnt_want_write_file(filp);
        if (status)
                return status;
 
        if ((!S_ISREG(inode->i_mode)) || !(filp->f_mode & FMODE_WRITE))
-               goto out;
+               goto out_drop;
 
        if (inode->i_flags & (S_IMMUTABLE|S_APPEND)) {
                status = -EPERM;
-               goto out;
+               goto out_drop;
        }
 
        context = kzalloc(sizeof(struct ocfs2_move_extents_context), GFP_NOFS);
        if (!context) {
                status = -ENOMEM;
                mlog_errno(status);
-               goto out;
+               goto out_drop;
        }
 
        context->inode = inode;
        context->file = filp;
 
-       if (argp) {
-               if (copy_from_user(&range, argp, sizeof(range))) {
-                       status = -EFAULT;
-                       goto out;
-               }
-       } else {
-               status = -EINVAL;
-               goto out;
+       if (copy_from_user(&range, argp, sizeof(range))) {
+               status = -EFAULT;
+               goto out_free;
        }
 
        if (range.me_start > i_size_read(inode))
-               goto out;
+               goto out_free;
 
        if (range.me_start + range.me_len > i_size_read(inode))
                        range.me_len = i_size_read(inode) - range.me_start;
@@ -1124,25 +1122,24 @@ int ocfs2_ioctl_move_extents(struct file *filp, void __user *argp)
 
                status = ocfs2_validate_and_adjust_move_goal(inode, &range);
                if (status)
-                       goto out;
+                       goto out_copy;
        }
 
        status = ocfs2_move_extents(context);
        if (status)
                mlog_errno(status);
-out:
+out_copy:
        /*
         * movement/defragmentation may end up being partially completed,
         * that's the reason why we need to return userspace the finished
         * length and new_offset even if failure happens somewhere.
         */
-       if (argp) {
-               if (copy_to_user(argp, &range, sizeof(range)))
-                       status = -EFAULT;
-       }
+       if (copy_to_user(argp, &range, sizeof(range)))
+               status = -EFAULT;
 
+out_free:
        kfree(context);
-
+out_drop:
        mnt_drop_write_file(filp);
 
        return status;