Re: WEIRD ext2fs dir size BUG?

Anthony DeBoer (
2 Mar 1998 19:22:01 -0000

Alan Cox <> writes:
> > Certainly, there may be cases of spewage causing really huge directories,
> > and if steps have been made to prevent that happening again it would be
> > useful to wind them back to a more reasonable size.
> I have to do this sometimes for inn (granted inn should use squid like
> directory trees..). When the issue was pointed out to Stephen Tweedie he
> sensibly replied that it was beyond the design parameters of ext2fs to
> have a 5Mbyte directory

Maybe then the defrag program (which I don't see in the RH5.0
distribution, so if I'm commenting on an actual program I haven't seen I
apologise) should have some sort of reasonable-size parameter or
constant, such that it will reduce the size of larger directories, but
won't throw away contiguous blocks from smaller directories.
Actually, doing so on the basis of empty space percentage (move valid
entries to the beginning of the directory, then deallocate excess blocks
while excess allocated space is more than X% of total directory size)
might work better.

(My own control/cancel dir is only 1.6MB, and I'm working on an INN patch
to short-circuit cancel processing when we don't have the article being
cancelled. Meanwhile, Diablo's reader mode is going through alpha, and
might ultimately send INN down the path trodden by B and C News. You may
get your sane news directory structures yet.)

Anthony DeBoer <>

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to