RE: gvinum startup problems

I just tried booting with kern.geom.debugflags=1 and
geom_vinum_load="YES" in /boot/loader.conf (latest RELENG_6), and the
result was not good:

# gvinum list
g_post_event_x(0xc0513ef0, 0xc2116600, 2, 262144)
3 drives:
D vd1                   State: up       /dev/ad3        A: 0/117800 MB
(0%)
D vd0                   State: up       /dev/ad2        A: 0/117800 MB
(0%)
D vd3                   State: up       /dev/ad1        A: 0/117800 MB
(0%)

1 volume:
V 480GB                 State: down     Plexes:       1 Size:        345
GB

1 plex:
P 480GB.p0            S State: down     Subdisks:     3 Size:        345
GB

4 subdisks:
S 480GB.p0.s3           State: stale    D: vd3          Size:        115
GB
S 480GB.p0.s2           State: up       D: vd2          Size:        115
GB
S 480GB.p0.s1           State: up       D: vd1          Size:        115
GB
S 480GB.p0.s0           State: stale    D: vd0          Size:        115
GB


My previous assertion that it was always /dev/ad3 that failed to be
tasted was probably just a matter of bad memory on my part. As you can
see from the above list, one vinumdrive was never found (/dev/ad0), and
two subdisks were marked as stale. Rebooting the machine without
auto-start of gvinum and then manually doing a "gvinum start" from
single- or multi-user brings the array up properly.

I have put three logs online:

1. Normal boot, plain dmesg (just to make it easier to see what devices
are found during probing, they are drowned in debug messages in the two
logs below)
   http://193.11.63.100/normal_boot.txt

2. kern.geom.debugflags=1 + geom_vinum_load="YES" -> single-user
   http://193.11.63.100/gvinum_auto_start.txt

3. kern.geom.debugflags=1 -> single-user -> "gvinum start"
   http://193.11.63.100/gvinum_manual_start.txt

Warning, the two debug logs are rather long (112kB each).

/Daniel Eriksson