[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: nbd-server
From: |
Wouter Verhelst |
Subject: |
Re: nbd-server |
Date: |
Fri, 28 Sep 2001 12:14:15 +0200 (CEST) |
On Fri, 28 Sep 2001, Roland McGrath wrote:
> I took a quick look at nbd and it's so simple that I just implemented the
> client side for the Hurd. That is, I've added an "nbd" store type to
> libstore.
> So to use it do e.g.:
>
> settrans /dev/nb0 /hurd/storeio -Tnbd hostname:1234
>
> to connect to TCP port 1234 on "hostname". That gives the store a
> blocksize of 1, since the server protocol doesn't impose any.
> You can set the block size of the store in the name, e.g.:
>
> settrans /dev/nb0 /hurd/storeio -Tnbd hostname:1234/1024
>
> does the same as the above, but the resulting device will have a block size
> of 1024 (the default for Linux nbd devices).
>
> As usual, I've compiled this but not tested it at all. I'm sure there are
> some stupid bugs. Hurd folks, please give this a try if you feel like it
> (you can point it at nbd-server on your Linux machine).
Looks great; however, I do have a couple questions:
- Where can I find the code? ;-)
- Pavel Machek, the author of NBD under Linux has put special effort in
making his code deadlock-free in case the block device is used as
swapspace device (a deadlock could arise if nbd-client needs more memory
to handle the reads necessary to handle a page fault...). Did you think
about this situation, or should I consider your port as "not deadlock
safe"?
--
wouter dot verhelst at advalvas dot be
"Human knowledge belongs to the world"
-- From the movie "Antitrust"
rm -rf /bin/laden
- Re: nbd-server, Roland McGrath, 2001/09/28
- Re: nbd-server,
Wouter Verhelst <=