From: ajayshah@alhena.usc.edu (Ajay Shah) Subject: Bug in mkfs suspected (with proposed workaround) Date: 26 May 1992 00:30:45 GMT
I've worked quite a bit on trying to install Linux yesterday and today
(I had to do it over around a dozen times, that counts for lots of
experience!) and I suspect a bug in mkfs as follows. You are
supposed to say
mkfs -c /dev/xxx number-of-blocks
where number of blocks is obtained by looking at "/bin/fdisk".
I repeatedly (but not always) got IO errors from mkfs. I don't know
exactly what to make of them, but the following is a regularity:
almost always, when I got IO errors from mkfs, they were at the _end_
of it's work. Several times I saw him doing an extra (off by one)
block (e.g. if I asked him to do 49876 blocks I would see him complain
and fail on blocks 49876 _and_ 49877).
(This is not always true -- for a few times I got complaints at the
START of his work).
WORKAROUND: I took to always subtracting 2 from the number of blocks
reported by fdisk, and all was well. I suspect something is amiss
with mkfs. Will this cause any trouble, apart from losing 2k of disk??
This is the version of mkfs in rootimage .95a from tsx-11 taken
yesterday. These errors occurred even with the primary partition,
created using DOS fdisk and formatted using DOS. (This primary
partition was 100 Meg so the first thing we would do is use
/bin/pfdisk to break it into pieces < 63 Meg).
Thanks,
-ans.
--Ajay Shah, (213)749-8133, ajayshah@usc.edu