Re: Advanced parted idea

[Serguei Tzukanov]

> Ah, I understand now.  Unfortunately, it's not quite a perfect
> interface like that... rm(1) doesn't free blocks on the loopback...
>
> So, it's probably useful to provide this, but most users will
> be too scared, IMHO.  Is it possible to get yourself out of
> this mess?  I guess you can always move the files back, delete
> the loopback, and create again...

> What if the loopbacked fs fails to allocate metadata blocks?
Do not permit transformation if there is less than (say) 20% free space 
(or have fs driver that counts maximum metadata for given fileset - 
very hard).
Not a perfect solution, I know. The perfect is to do cp(1).

> It would be nice to hack some way of the fs thinking it's
> full... have a daemon updating the reserved block count (UGLY!)
Daemon (or whatever it is) does not know which touched blocks to free.

>
> Any ideas?
Have fs driver that knows which blocks are used. 
 -or-
Run fsck on loopbacked fs and do read pass over the filesystem 
(incredible slow).


-Serguei.