Subject: Re: [vserver] Possible Hashify Corruption - Update 2
From: Art -kwaak- van Breemen <ard@telegraafnet.nl>
Date: Sat, 23 Oct 2010 19:25:12 +0200

On Thu, Oct 21, 2010 at 02:31:21PM +0200, Rik Bobbaers wrote:
> if userspace can break your filesystem, there is a bug in your filesystem,
> PERIOD.
Well, in the documented cases, there was no filesystem
corruption, nor were there bugs in the filesystem. Just a wrong
view on what should be there by the application programmers.
So userspace didn't break the filesystem, it wasn't broken, but
userspace let it's data become in an unknown state.
(Something with having to do a fsync if you *really* want to be
sure that your data is on the disc).

> This should never under any circumstance happen, there should be as much
> protection as possible against it.

In this case the author has decided to protect the programmers
against wrong assumptions and made an auto fsync on certain
actions, so programmers have to think less hard about data on the
disk.

> Same goes for a binary that gets you root access when specifying a bad
> parameter, it's a root exploit and should be fixed, it's a bug...

That's correct, but that's not what has been referred to. The
page described the bugs in the configuration file managers of
gnome and kde, and how they get corrupted when not fsynced as it
should have been.

Of course, there were bugs in ext4, but this one was not a bug.

Of course there probably were bugs with ext4 and vserver patches,
but those were probably the alpha pre-release because the ext4
has not been marked stable that long in the time those patches
came.

As herbert says: time to move on to a recent kernel+vserver
patches ;-).

(BTW: root access is nothing, *kernel memory* access, that's the
ultimate goal of pwning other peoples Pc's :-) ).