On 2012-08-27, at 7:30 AM, raphel johnson wrote:

I was referring EXT4_IOC_RESIZE_FS ioctl implementation in linux
kernel ver 3.3.8.

It is given in the code that online shrinking not supported.

if (n_blocks_count < o_blocks_count) {
/* On-line shrinking not supported */
ext4_warning(sb, "can't shrink FS - resize aborted");
return -EINVAL;
}

I am new to ext4 file system. Could you please tell online shrinking
issues in ext4.

Thanks in adavance.

There are several reasons for this:
- if blocks are in use beyond the new end of the filesystem, potentially
  all inodes in the filesystem need to be scanned to find the owner
- this also causes consistency issues with the page cache, since moving
  in-use disk blocks is tricky, and it isn't always possible to remove
  them from memory
- if in-use inodes are beyond the end of the new filesystem, potentially
  all of the directories in the filesystem need to be scanned to find the
  parent directory
- moving inodes in ext4 will also cause the inode number to change, which
  would break NFS and potentially other programs

In the end, rarely do users need to shrink a filesystem while mounted, so
the effort and complexity of doing this isn't worthwhile.

Cheers, Andreas