HDD problems: HELP!
Tom Barnes-Lawrence
the only bloke who replied said "try using later version of RH than 4"
even though I had said that this was part of trying to upgrade the
machine to RH5.2... So can someone on *this* NG actually read this
and give me an answer? Please? If there's a fault I'd like to replace the
bloody thing soonish... Contents of original post continues:
Recently, after helping put together this machine for family's
internet access and stuff, I tried sticking RH4.0 (from old CD) onto an old
400MB SCSI disk on it. Worked. Was bootable, usable, etc. Wasn't too
suprised, as I'd installed Linux on my (ie-other) machine about 2 years ago,
which went reasonably well, and is still same version there (albeit patched
and stuck together with string).
But the disk was really crammed, and RH4 didn't seem to accept new
RPMs and I was having trouble sorting email, so I went on hunt for new
HDD to try install RH5.2 (coverdisk on magazine).
Got 2.1 GB IDE disk from computer fair, told they were system-pulls from
new machines that they didn't get to sell, so they were not-quite new,
not-quite
2nd hand, but I could go back to them if I had problems. This is what I did:
Took out SCSI card and disk (they were old, only used them because
lying around).
Stuck new disk in as primary slave, used autodetect from BIOS (in VX
mboard).
It set it to: CHS 524,128,63, in LBA mode. (checking geometry details
later on disk case and website, says 4192,16,63, but sure enough thats
the same when in LBA)
Started Windoze command prompt. Ran Fdisk, removed FAT16 partition.
Restarted, stuck in RH5.2 CD and selected boot from CDROM in BIOS.
When in Linux instally thing, set up the partitions. Linux Fdisk saw disk OK
and didn't worry about geometry or anything. Made a big partition, and then
a 96MB swap partition at end.
When went to format the partitions for e2fs and swap, checked the "check
for bad blocks during format" bit, just in case. When it was formatting
them,
it mentioned NO PROBLEMS- I checked on the other VTs that the RH install
gives, and each time it seemed quite happy.
It then started installing the packages I'd selected.
Now, in past, installing from this CD drive has caused vague problems
that seemed to be to do with it not reading the CD well. The CD drive
doesn't sound well when it does it, and it gives messages which indicate
problems with the packages themselves.
It did this again this time too, a bit, BUT more worrying were the
messages
it gave at several points about:
ext2_new_block:Free blocks count corrupted for block group 36
This was in one of the other VTs, kernel messages, IIRC.
Same thing about 7 times, 1ce for each package which failed. On the actual
RH install VT, it was saying something about out of disk space- but these
were messages from the RH installer, and it was clearly misinterpretting
stuff, due to the way it managed to carry on installing various other
things.
I figured this was maybe some geometry problem, and tried restarting,
unpartitioning (from RH install), changing geometry from BIOS- to Normal
or Large geometry, and finding that of course fdisk wasnt happy about
this as certain entries in partition table got the cylinders rounded
to1023...
So this ended up with the e2 partition and swap overlapping (as it had
warned me it would) and the install got upset and stopped when it realised
it was trying to install over bits of swap space...
Another time, I tried saying it had 4096 Cylinders in normal rather than
4192
-just because it sounds more sensible- and then changing it to LBA, and
seeing if it liked that. It didn't mind, but it gave same errors as before-
but corrupted for block group 34 this time. I tried taking it through
to the end to see what'd happen. Of course, when it tried to reboot,
it realised the FS had errors, and did fsck. It got *very* upset, and asked
me to run it manually. Yeah, right.
As it never seemed to find bad blocks, I figure that isn't the problem
,but
I have no idea. Help me out, someone? I'm stuck in sodding Outlook Express
STILL and it was easier getting Linux to run on a 2MB 386 (dont ask, took
ages) because it was just using a nice simple 40MB disk without these
horible tricks IDE has to use nowadays.
--
For fsck's sake, HELP!!!
--
Tom Barnes-Lawrence (aka Tomble the Bod)
Kindly remove eggs and spam from email address
if you want to email me.