[Unionfs] Using unionfs w/o modyfing deployed kernel...

[Deepak Saxena]

Hi,

We are looking at deploying UnionFS to some of our users to solve some
disk space issues and need a way to do so without requiring a core kernel 
image update.

What I'm imagining is simply building and shipping unionfs.ko to implement
our solution (mount tmpfs rw + jffs ro union and then provide a way for
user to cleanup files), however, looking at the UnionFS patches (against 
2.6.22), it looks like it depends on some changes to the VFS layer. It 
looks like mose of these I can ignore and update the UnionFS code to use 
the old APIs but I'd like to run this by the experts first:

do_splice_from() -> vfs_splice_from()

  We don't need splice functionality for our use case so I can just not
  include this change and NULL out the UnionFS pointers.

fsstack_copy_attr_all() interface change

  We can just keep the old interface and pass the NULL 

spin-lock in fsstack_copy_inode_size()

  We're !SMP, so I think we can safely ignore this.

do_path_lookup() LOOKUP_ONE support

  AFAICT from the UnionFS patch, this is not actually getting used though 
  the known issues file points this change out.

lookup_one_len_nd() addition

  From the comments in the known issues file, this seems to only be needed
  for NFS which we are not using, so it may be possible to not include this
  change and just call lookup_one_len() in the UnionFS code.

Thanks!
~Deepak

-- 
Deepak Saxena - Kernel Developer - dsaxena at laptop.org