From: Herbert Poetzl (herbert_at_13thfloor.at)
Date: Wed 21 Aug 2002 - 19:29:55 BST
On Wed, Aug 21, 2002 at 07:54:45PM +0200, Seiler Thomas wrote:
> Hello everyone
>
>
> I've just come across some bad news for vserver / ctx...
>
> The EXT2_IMMUTABLE_LINK_FL bit (value 0x00008000) has now another
> meaning:
>
> http://e2fsprogs.bkbits.net:8080/e2fsprogs/search/?expr=EXT2_NOTAIL_FL&
> search=ChangeSet+comments
>
> They use it to indicate that the kernel should not merge 2 file's tails into
> one block to save space. (because this would break LILO and the like...)
> I suppose this behavior is planned for the next stable Kernel releases.
>
> This would mean, that such a new Kernel might take a immutable linkage
> invert flag set by some unify script as a notail flag (and i have no idea if
> this
> has the potential to break anything. Probably not, but hey, you never know.)
>
> As there are plenty of bits left unassigned, it should be no problem to use
> another bit instead (i.e value 0x00010000), and it would be easy to write a
> litte perl script, which updates the attributes on existing filesystems. (If
> no
> one else volunteers, i would write one, but expect poor code quality :-)
>
> It would even be possible to logicaly OR the two bits in the Kernel say in
> the next three relases, so that the change would be smother.
>
> Best would be, if the maintainer of the ext2progs package could define the
> bit in the next relase so that every one sees, that this bit is already
> taken.
> Otherwise, the hole situation could repeat.
maybe the vserver project should contact
Theodore Ts'o, if not done so already, and
*ask* him to reserve one bit for immutable
link flag once and for all ...
> Best Regards,
> Thomas Seiler
best,
Herbert