I have a CTX Cyrix/200; the dedicated linux drive is a 3 Gig Maxtor IDE,
with a BIOS add-on called EZBIOS. I use loadlin because this BIOS has to
load before linux boot -- and I like the config.sys dual boot technique.
I use just two partitions (/ and swap).
RedHat 5.2 and 6.0 both work well with this setup -- except that I remember
always getting (during the rc3.d or rc5.d init sequences): "root is mounted.
WARNING - running fsck on a mounted filesystem can cause severe dammage. -
Check aborted." After this, the remainder of the boot was OK.
After trying Upgrade to 6.2 first, and a complete 6.2 install next, the boot
sequence fails at the fsck with this message: "Root is mounted. (or maybe it
says Root is mounted as /dev/hdb1). CANNOT CONTINUE. Dropping you to a
shell." From the shell, I run fsck and I get result code 8 -- "operational
failure," so I understand why it stops the boot. But I think the fsck
result is bogus.
Additional troubleshooting: Reinstalled RH 6.0, works fine. Installed RPMS
for e2fsprogs from RH 6.2 disk(one or two others required for dependencies,
initscripts was one, for example). Rebooted. Boot failed at exactly the same
place, fsck. Something has changed that is incompatible with my system.
Thanks for your effort.
Why do you thin the result code is bogus?
Have you tried running fsck manually? What messages were displayed?
When I run fsck manually using vers. 1.18 (I believe that's the one in RH 6.2),
I get result code 8 and fsck refuses to attempt fixes.
When I run fsck vers. 1.14 (in RH 6.0), I get code 4 if just asking for codes,
or, I get several >fix (y/n)< dialogs with the following types of errors typical:
duplicate/bad blocks (usually two or three of these);
block bitmap differences (a few);
free blocks count wrong for group #_ (a few).
I am speculating that the error code 8 with vers. 1.18 is bogus since vers. 1.14
can fix these problems.
Something is wrong on your system if the root filesystem is already mounted
read-write during the filesystem check. If older tools don't warn or stop, this
would also be a bug in older rpms.
Current rpms seem to do the right thing, so I'll close this bug.