Re: [PATCH] ext4: Reserve FEATURE_RO_COMPAT and INO flag

[Theodore Ts'o <tytso@mit.edu>]

On Mon, Apr 29, 2013 at 12:35:18PM -0600, Andreas Dilger wrote:
> 
> One real problem that I see is that EXT4_REPLICA_INO uses the
> last reserved inode.  I recall in the past that Ted wanted to
> make this last reserved inode a hidden directory, so that it
> would be possible to put special inodes in there and access them
> by name.  Not quite as easy as accessing a reserved inode by
> number, but still preferable to not being able to have filesystem
> internal inodes at all in the future.

That wasn't my idea, but it's not a bad one.  I'm not too worried
about running out of file system internal inodes, though.  There are
two main benefits of having special inode numbers.  First, it's a low
numbered inode, so it's easy to recognize.  Second, it's easier to add
a new file system feature to an existing file system.  But either way,
it's not impossible.

First of all, "the first non-reserved inode" is actually a superblock
field, so we could actually change that for newer file systems, and
hence effectively create new reserved inodes.  Secondly, it's simple
enough just to use a superblock field to designate a particular inode
as being special.  This is how we originally dealt with the journal
inode, after all.

Cheers,

						- Ted