[Noisebridge-discuss] M.A.I.D. was: Freespace / slackspace file system

Sai Emrys noisebridge at saizai.com
Sat Sep 26 02:17:12 UTC 2009


On Fri, Sep 25, 2009 at 4:26 PM, Andy Isaacson <adi at hexapodia.org> wrote:
> Having the cover filesystem and the hidden filesystem mounted (and
> writing!) simultaneously is going to be challenging.  If ssfs starts
> writing to a formerly-slack block at the same time the ext3 filesystem
> allocates the block and starts writing to it, you're likely to corrupt
> the ext3 datablock.  You want to avoid breaking the host filesystem,
> which is a bit challenging without modifying the host filesystem code.

What if the FSFS daemon has a hook into the host FS, and proxies all
operations to the FSFS, such that it can hard abort any writing to a
just-allocated block?

You're right that the timing there might get dicey, but intuitively to
me (not knowing the low level details) it seems possible.

For SSFS the issue is about the same, but a bit easier - you can just
watch files-at-once and probably have a somewhat less paranoid proxy,
'cause at least write operations will first call a write lock on some
file. I think.

For that matter: what lead time are we talking between something in
the host FS acquiring a write lock on a file, or a direct write on a
block, and actually writing to it? We might be talking race condition
issues even if we do have a allocation-monitoring abort hook.

Hm.

- Sai



More information about the Noisebridge-discuss mailing list