this morning i have encountered a very annoying situation
with a machine i had upgraded to 4.4-Stable yesterday,
I describe it here in case somebody has seen something similar
or it may be useful to someone.
I was writing to a dos floppy when the machine locked solid.
The last message logged by syslog is
Oct 4 09:15:58 niobe /kernel: fd0c: hard error writing fsbn 19 (ST0
Note that the floppy was mounted so this may be a remnant of the infamous
bug about writing to write protected floppies, which i thought had been
eliminated. But the machine did not panic.
After that the screen remained normal, but keyboard and mouse did not
respond. Machine answered ping but i could not log in through network
to reboot and had to hit the reset button.
Unfortunately reboot went very bad since fsck was not able to preen the
root partition and i was asked to run fsck manually. But superblocks
0 and 32 were corrupted with BAD MAGIC NUMBER. A thing i have never seen
previously. Of course i had no copy of the position of alternate superblocks.
After an hour of reading man pages, i discovered that newfs has an option
-N allowing to fake the creation of a filesystem. I ran
newfs -N /dev/da0s1a
which showed that one copy of the superblock was at 65558.
fsck -b 65558 /dev/da0s1a
worked and offered to repair the superblocks. A trick which is useful
to remember! After that the machine was able to reboot.
Conclusion: i suspect there is a * bug in the filesystem management
at present (since the introduction of dirpref?). Note that
softupdates is not the culprit since root was the only partition
running with softupdates disabled. Incidentally, i had fscked manually
all the other partitions (softupdates enabled) without problem.