[parted-devel] Understanding AFFS

[Brian C. Lane]

On Mon, Jan 12, 2015 at 03:40:41PM -0500, Phillip Susi wrote:
> I've been investigating why parted's rescue operation is so slow and
> it seems that it is largely due to AFFS, so I am trying to understand
> it a bit.  I notice that it appears to read the first 16 sectors of
> the disk ( not the partition or potential partition ) looking for some
> sort of "rigid disk" block.  Why is this?  Wouldn't this prevent the
> possibility of having such a filesystem on a gpt partitioned disk, or
> a disk with grub installed?  Or a loop image?  I checked the kernel
> affs code and it has a different way of computing the fs block size (
> which seems to be the reason for reading the rigid disk block ).
> 
> Another reason for the slowness is that affs registers 15 different
> variants and each one must be probed separately resulting in 15 times
> the number of reads.  Is it really necessary to differentiate between
> all of these sub types in the output of parted print?
> 
> Also where can I find tools for creating such a filesystem for testing
> purposes?

I don't know any more than the wikipedia page, I was an Atari guy back
then :) But my guess would be that AFFS was never meant to live on a
partition, since it is based on and backwards-compatible with, the
format they used on floppies. If it is causing that much trouble I
wouldn't be opposed to dropping support for it.

-- 
Brian C. Lane | Anaconda Team | IRC: bcl #anaconda | Port Orchard, WA (PST8PDT)