At 6:17 PM -0700 7/7/98, Scott Reynolds wrote:
>On Tue, 7 Jul 1998, Henry B. Hotz wrote:
>
>> The last thing before setting tty flags is the fsck of disks.  If the
>> machine prints fscking sd3a before sd2d then the machine always hangs.  If
>> the machine prints fscking sd2d before sd3a then the machine always goes on
>> to complete a normal boot [...]
>> [...]
>> /dev/sd0a       /               ffs     rw 1 1
>> /dev/sd1b       none            swap    sw 0 0
>> /dev/sd2d       /users          ffs     rw 1 2
>> /dev/sd3a       /mirror         ffs     rw 1 0
>> /dev/sd3b       /usr/vice       ffs     rw 0 0
>
>eeeeeeeeeeeK!

Gee, I wasn't expecting that strong a reaction.  I thought this was just an
obscure harware/driver interaction.

>that last column tells fsck the _order_ to fsck things in.  to tell you
>the truth, i'm not even sure what a `0' in that column does, but i can
>assure you that the root fs should be `1', and all others should have a
>value N > 1 in them.

>From the man page "If the sixth field is not present or zero, a value of
zero is returned and fsck will assume that the filesystem does not need to
be checked."  In other words swap and sd3x are not to be fsck'ed.  Putting
the zeros in there does in fact disable fsck on boot for the sd3 partitions.

While I wrote the line above, I'm pretty sure I got the idea of putting
zero by the swap entry from a file created by some version of the installer
once upon a time.  The man page and I agree with you that the standard rule
is root gets a 1 sequence number and all other standard partitions get a 2.

>you should not modify rc files to sequence the fsck's.

Why not?  (No offense.)  It may be non-optimum use of hardware, but it's
what I need to do to make this machine boot reliably.  I tried putting a 3
in the last column for the sd3 partitions, but that seems to behave the
same as a 2.  (It's still random whether sd2d or sd3a get checked first.)
Short of a fix to the esp driver I think it's the best solution for me
since I do want those sd3x partitions fsck'ed every once in a while.

>now, once you've verified that this works correctly, you may indeed want
>to use send-pr and report the behavior.  i believe you've found a bug of
>the `it must not go loopy when given a broken configuration' sort.

If you say so.  I'm up to 6 or 7 reboots without a problem since I put the
zeros in there.  Combined with the ordering correlation I noticed, I'm
pretty convinced I've diagnosed the problem with my machine.

Perhaps I didn't make it clear that the above is my current, working fstab.
The old, 50%-failing-on-reboot fstab looked like this:

/dev/sd0a       /               ffs     rw 1 1
/dev/sd1b       none            swap    sw 0 0
/dev/sd2d       /users          ffs     rw 1 2
/dev/sd3a       /mirror         ffs     rw 1 2
/dev/sd3b       /usr/vice       ffs     rw 0 2

Signature failed Preliminary Design Review.
Feasibility of a new signature is currently being evaluated.
h.b.hotz@jpl.nasa.gov, or hbhotz@oxy.edu