Re: [RFC] ext2_new_block() behaviour

[Andreas Dilger <adilger@turbolinux.com>]

Al, you write:
> 	Folks, there is a pretty strange detail of the allocation policy -
> if cylinder group has no free blocks past the goal ext2 tries very hard to
> avoid allocation in the beginning of the group. I.e. order looks so:
> 
> 	* goal
> 	* goal .. (goal+63) & ~63
> 	* goal .. end of cylinder group
> 	* cylinder groups past one that contains goal
> 	* cylinder groups before one that contains goal
> 	* beginning of cylinder group..goal-1
> 
> It looks somewhat fishy. What's the reason for such policy?

This predates me by a while, but I suspect that it is done this way on
the assumption it is easier to seek forward on the disk while reading
a file rather than seeking backwards.  Also, since with new inodes the
goal is initially the first block of the group where the inode lives,
the blocks at the start of a group will generally be allocated already,
so it is usually a waste of time checking the start of the group for
free blocks.

Cheers, Andreas
-- 
Andreas Dilger  \ "If a man ate a pound of pasta and a pound of antipasto,
                 \  would they cancel out, leaving him still hungry?"
http://www-mddsp.enel.ucalgary.ca/People/adilger/               -- Dogbert
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/