Sorry for the delay, exams time! :)
On 7/3/06, Bdale Garbee <[EMAIL PROTECTED]> wrote:
Is bug #210377 related? It's been a while, but I think I never felt
that discussion reached a concensus on what makedev should do
differently, if anything, and so I haven't acted on it yet.
I've been investigating a little, and this is the situation, as I see it:
- lilo doesn't use makedev, creates 128 block devices with major 43 in
/dev/nbd%d
- makedev/sid (shell version), creates 8 block devices with major 43
in /dev/nb%d
- makedev/experimental (C version), creates 128 block devices with
major 43 in /dev/nb%d
AND, /usr/share/doc/linux-doc-2.6.15/Documentation/devices.txt.gz says clearly:
43 block Network block devices
0 = /dev/nb0 First network block device
1 = /dev/nb1 Second network block device
So, I think the sane thing would be to settle on the official registry
way, which is using nbX without partitions. People using enbd can
create devices with different names (enb%c%d?), and it could even be
incorporated in makedev.d/enbd from the enbd package; and this should
not make any problem.
Of course, lilo has no excuse to ignore makedev :) The problem may be
is that one's tempted to try MAKEDEV nbd, which doesn't work, and you
may think that MAKEDEV has no nbd support.
It is worth mentioning that the mknod in lilo.postinst spawns from #235805
--
Martín Ferrari
