gabriel rosenkoetter on Fri, 25 Jan 2002 19:06:25 +0100


[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: [PLUG] EXT2-fs error


On Fri, Jan 25, 2002 at 11:52:22AM -0500, Samantha Samuel wrote:
> It was loaded as a module. Kernel was built against new one. Old kernel is
> still there.

And the LKM for the network device doesn't work?

> I have no scsi devices, as for overheating, I do shut the comp down
> everynight, and physically feel for heat, so I doubt that has anything to do
> with it. I didn't take any experimental drivers when configuring the kernel,
> so I suppose the problems has to do with kernel(2.4.17) bugs?

Yeah, like I said, those were just examples of plausible hardware
problems, not problems I necessarily thought you had.

> When I included 'Linux 2.4' in the search, one suggestion which was posted
> was to do a e2fsck.
> When I did this on the mounted partition
> e2fsck -p /dev/hda1 (which is '/') I got threatening messages saying that
> everything may be severely damaged if I went any further. I have /home on
> another partition. I am not sure how to do the repairing, if that is the
> solution.

I haven't used e2fsck, so I couldn't tell you whether it's just
being scary or telling the truth.

> I haven't tried to have another os read the linux partition. I will try that
> later today.

Under NetBSD, you'll want to do:

mount_ext2fs -o noatime,nodev,nocoredump,noexec,nosuid,rdonly <dev> <mnt>

This is a little excessive, but it'll keep NetBSD from having any
chance of hosing the file system.

If it mounts successfully, poke around and see what is and isn't
there. (That user's disappearing home directory, for instance.)

Incidentally, adding the sync flag to /etc/fstab for your disk on
the Linux side would be a good idea till things settle down. (This
will force data to actually go to the disk rather than stay in the
cache; I think the reason you keep losing that user's home directory
may be that it never gets sync'ed.)

> I reinstalled netbsd again yesterday, and again I went with the default
> cylinders numbers given. Upon booting into Linux later, the user that I
> had recreated which was earlier hosed, was hosed again. I am able to
> login, but the user has no home directory. Not sure how to word this, but
> I am guessing that sysinst writes into the linux partition.

The cylinder numbers are important, but you would be seeing worse
problems than this if they were causing this. I have a feeling that
this is a remnant of your ext2fs corruption, but let's make sure...

Is it that the user's directory is gone or that /etc/passwd is
borked? (What's the inode number of /etc/passwd? Of the user's home
directory? What cylinder is it on?)

You ARE installing NetBSD on a separate MBR and fdisk partition,
right?

What does fdisk report under Linux? Under NetBSD? What about
disklabel wd0 under NetBSD?

It's possible that Linux and NetBSD are overlapping blocks/
cylinders...

Does recreating the user's home directory under Linux (and doing a
sync) make NetBSD unbootable?

sysinst only writes into partitions you've told it to and into the
space before the first partition (the MBR record area, usually the
first track of the disk)...

Oh, it'll be answered by the commands I asked you to run above, but
does your Linux partition start at block 0 on the disk or one track
in (maybe block 63?). If the former, you'll need to start over; you
have to leave a track for MBR stuff if you want to multi-boot...

> Yes I can boot both oses, but only because of lilo. The bsd's bootman(?)
> doesn't allow me to get linux up, only netbsd.

I've never heard of a piece of software going by the name "bootman".

> As I was typing this msg, I got this:
> 
> RFail=0 RFErr=0 RErr=0 RAXFR=0 RLame=0 ROpts=0 SSysQ=1 SAns=0 SFwdQ=0
> SDupQ=2 SErr=0 RQ=0 RIQ=0 RFwdQ=0 RDupQ=0 RTCP=0 SFwdR=0 SFail=0 SFErr=0
> SNaAns=0 SNXD=0 RUQ=0 RURQ=0 RUXFR=0 RUUpd=0
> 
> Is this because of crap hw? But the messages I read (after searching for
> RFail=0 RFErr=0 Linux 2.4) seems to suggests that this is a kernel prob.
> And the recommedations are all having to do with debugging the kernel.

That sounds like your kernel and your IDE bus aren't being
particularly agreeable.

I would definitely be backing off to using the 2.2 kernel for
anything except trying to make the 2.4 kernel work, if I were you...

-- 
gabriel rosenkoetter
gr@eclipsed.net

Attachment: pgpz8O88AwXSp.pgp
Description: PGP signature