Subject: Feb12 wd.c: lost interrupt - ouch!
To: None <email@example.com>
From: Simon J. Gerraty <firstname.lastname@example.org>
Date: 02/22/1994 13:10:56
I'm not writing this from zen (he's a bit dead right now) so pls send
any replies directly :-)
I've spent the last 3 days or so, updating my -current from the Feb12
tar_files. All went very well.
I even had one of the modems going much of the time while
re-building... the only hickup was the machine hanging at one point.
I sure I was not out of swap, and I _think_ I was running a new
Anyway, last night I finished the update by installing the rest of
This morning, I noticed that my mail had all landed in the bit bucket
because I forgot to replace /usr/sbin/sendmail with my IDA version.
The new sendmail vomited on my sendmail.cf, I better point IDA
sendmail at a different sendmail.cf and make sure the one in /etc is
sane for the new sendmail.
I fixed that (re-installed IDA sendmail) and mailed off a message to
this list. Then went to have a look as to whether any of my mail was
still available in /usr/MHSnet/_bad ...
I got an error msg that /usr/MHSnet/_messages was not a directory...
A look at the console window showed:
"wd0: lost interrupt status 58 ??? 0"
I think that's right, except I can't recall the ??? bit.
I noticed that several directories under MHSnet were gone or in deep
shit. So I though a reboot and fsck was in order.
All filesystems were ok except that fsck -p refused to touch /usr
Running fsck manually it said that i-nodes 1558 through 1610 were
partially allocated, so I cleared them. (1558 happend to be /usr/sbin
I had seen folk report about damaged i-nodes and lost filesystems, but
had never suffered them myself. I've been running -current since
November and before that 386bsd since Feb 93, while I had to press the
reset button a few times due to wd.c hanging (I assume), I've never
suffered any serious file system damage until now.
At present I boot off a 120Mb IDE drive and have about 1Gb of SCSI
drives that hold everything else.
I did try rebuilding a kernel with the Nov20 wd.c in it, but it would
not compile. I figured an occasional hang requiring a reset was less
hassle than trashed filesystems.
Has any work been done on wd.c in the last week or so?
I left zen restoring /usr from tape, and I've just ordered a 1Gb
SCSI disk... I'll take the IDE drive out for now so that I can boot
off one of the SCSI disks. Then start building again.
Apart from the wd.c stuff, Feb12-current seems ok!