From: Brian J. Watson <Brian.J.Watson@co...> - 2001-09-10 21:58:56
Kitrick Sheets wrote:
> I've made some progress with the UML port but have hit a minor block.
> Anyway, system boots to the point where it attempts to open /dev/console
> from init.
> Unfortunately, the console device is in /dev/nodeXXXXX/console because
> the changes
> to devfs. I'm sure that I'm missing something very obvious but I'm just
> seeing the
> problem. Thought you might be able to shed a bit of light on this for
The problem is we don't yet support mounting devfs on /dev, because the
device naming is broken for apps who expect to find /dev/console and the
like. Our plan is to union the nodeXXXXX directories together, and use
devfsd to provide persistent device naming regardless of the order in
which the nodes are booted.
For example, if /dev/sdb is discovered first on node 2, it should
forevermore be the same device on node 2 (assuming the node or the
device is not decommissioned). Anybody on any node who does anything
with /dev/sdb will be working with that particular disk drive attached
to node 2.
A device such as /dev/console will probably be a bit non-SSI, in the
respect that it'll go to the console on the same node as the process
accessing it, rather than some clusterwide console.
John Byrne can shed more light on our device story, since he's the
developer working on it.
The solution to your problem for now is to mount devfs on some non-/dev
mountpoint. I like /devfs myself. Apps will then find the old-school
mknod-style devices they're looking for through /dev, which is fine for
local device access.
Brian Watson | "The common people of England... so
Linux Kernel Developer | jealous of their liberty, but like the
Open SSI Clustering Project | common people of most other countries
Compaq Computer Corp | never rightly considering wherein it
Los Angeles, CA | consists..."
| -Adam Smith, Wealth of Nations, 1776
Get latest updates about Open Source Projects, Conferences and News.