On Wednesday, 12 November 2003 at 18:32:09 -0500, Thor Lancelot Simon wrote:
> On Thu, Nov 13, 2003 at 09:21:53AM +1030, Greg 'groggy' Lehey wrote:
>> On Wednesday, 12 November 2003 at 22:54:29 +0100, TeCeEm@gmx.de wrote:
>>>> On Wednesday, 12 November 2003 at  5:56:34 +0100, TeCeEm@gmx.de wrote:
>>>>> # ls -l /dev/vinum/
>>>>> total 14
>>>>> crw-------  1 root  wheel  108, 262142 Nov 12 05:30 control
>>>>
>>>> Hmm.  This looks like a victim of the change of major device number.
>>>> Vinum now has major number 162.  Try rm -rf /dev/vinum and see if it
>>>> works like that.
>>>
>>> Ok, I figure these numbers are defined somewhere else than /sys/dev/vinum or
>>> /sys/sbin/vinum and my kernel being out of sync to the userland is an issue
>>> here.
>>
>> Yes, that's the problem.  You'll need to rebuild vinum(8).
>
> Why does the vinum executable have special knowledge of the device
> *number* for Vinum, rather than using filesystem naming?  That seems
> like a severe bug.

That's a matter of opinion.  Others see the creation of device nodes
with scripts like MAKEDEV as a bug.  vinum(8) creates devices
dynamically, so it needs to know the major number.  Yes, you could
create /dev/vinum/control with MAKEDEV, but vinum allows you to create
volumes dynamically, and they need a device number.

If people think that this is too tacky, there's a compromise: have
MAKEDEV create /dev/vinum/control, and have vinum create the volumes
with the same major number.

Thoughts?

Greg
--
Finger grog@lemis.com for PGP public key
See complete headers for address and phone numbers