Hello,

some users have complained about the limit of 32 acl entries on ext2 and
ext3. There is no reason for this limit other than enforcing "reasonably
short" acls. The attached patches (in this order) remove this limit.

  ext23-no-ace-read-limit.diff
  ext23-no-ace-write-limit.diff

The first patch ensures that big acls can properly be read. The second
removes the check when writing. With both patches applied, ext[23]
supports up to 123 acl entries with 1k-blocks, and up to 507 entries
with 4k-blocks.

[XFS inherits a limit of 25 acl entries from Irix. The other file
systems that support acls (jfs, reiserfs) enforce no limit, which means
that up to 8181 acl entries are theoretically possible.]

The write patch unfortunately breaks backward compatibility: Older
kernels cannot read ACLs with more than 32 entries; accessing such files
will result in EIO.

I don't think breaking backward compatibility in this way is acceptable,
so IMHO we should  apply only the "read" patch now, as it ensures a more
or less painless upgrade later. (I will also add the read patch to the
2.4 patch set.)


Regards,
-- 
Andreas Gruenbacher <agruen@suse.de>
SUSE Labs, SUSE LINUX AG