On Fri, 21 Dec 2001 16:39:33 +0100 Rene Engelhard wrote: <snip>

Use the boot disk to start the rescue system.

Try mount your ext2 partition under /mnt. (search all /dev/hdaZ's to find the partition)

Edit /mnt/etc/fstab to coorect the devices there.

How is it now?

That looks like better advice than mine David, I would certainly try it first. Geoff _________________________________________________________ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com

On Sat, 22 Dec 2001 00:37:12 +0000 David wrote:

Bit stuck with Rene's method. Tried fsck but that just returned to the command line, don't know if it should have any parameters with it. Looked it up in the manual, but it was not on the page they said it was.

I still think that Rene's method offers the better way forward and I am reluctant to encourage you to go further with fsck until you have used it. The reason is that if the problem is simply that your root partition cannot be found, rather than being internally corrupt, then fsck will not help. I read your reply to Rene, and he may have his own better response than this, but, bearing in mind that you are just feeling your way here, and if he does not have a quicker solution, may I suggest that you do the following to help us to help you. (1) Start the rescue system. (2) It appears from your earlier posts that linux is on a second physical (as opposed to logical), drive - ie that it is on hdb. If that is right enter, as root, . That should show the existing partition table on hdb. You obviously have to be careful about syntax here, because fdisk is the formatter, but that command is, so_far as_I_am_aware, totally safe and you can run it on any drive. When you have seen the output on screen you can run the command again as fdisk -l /dev/hdb > part_table. As you know that will write the output to the file part_table which you can then send to us here. (3) Also send us your existing /etc/fstab. If Rene's diagnosis is correct, it should be easy to give you instructions from there on. Geoff _________________________________________________________ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com

Hi Geoff Hope you have had a good Christmas and everyone else on the list Back to my problem I used fdisk and Linux is now on /dev/hdb9. The only editor I know of is Vi, but it is a pig to use, and the file is read only. Could not work out what chmod needs to change its use. I tried to copy it across to a floppy using cp /etc/fstab /dev/fd0. It appears to work, but there is nothing on the floppy. That's the only way to get a copy on to the list at the moment. Is there a simpler editor than Vi? Appreciate if you could help with the above Regards, David On Sat, 22 Dec 2001 11:02:12 +0000, Geoff wrote:

(1) Start the rescue system.

(2) It appears from your earlier posts that linux is on a second physical (as opposed to logical), drive - ie that it is on hdb. If that is right enter, as root, . That should show the existing partition table on hdb. You obviously have to be careful about syntax here, because fdisk is the formatter, but that command is, so_far as_I_am_aware, totally safe and you can run it on any drive. When you have seen the output on screen you can run the command again as fdisk -l /dev/hdb > part_table. As you know that will write the output to the file part_table which you can then send to us here.

(3) Also send us your existing /etc/fstab.

Thanks - much simpler On Thu, 27 Dec 2001 07:47:16 +0000, Terence McCarthy wrote:

On Wednesday 26 December 2001 23:20, David wrote:

...

Is there a simpler editor than Vi? Appreciate if you could help with the above

I like pico (and use it with pilot as a file manager). Try it.

Terence

Regards, David

Thanks - didn't know that. Not sure if that was the case though, since the floppy light came on when I did it. I have now managed to copy it across to a win partition. The problem I am having is that the file is read only. If I make a backup first I get 'Read-only file system'. This seems to apply to the disk rather than the file. ls -l fstab gives -rw-r--r-- 1 root How can I change the system to let me edit the file? I note your point about Pico, however it is a lot simpler and the lines in the file are short. Regards, David On Thu, 27 Dec 2001 09:18:32 +0100, Anders Johansson wrote:

You have to mount the floppy. If you do cp /etc/fstab /dev/fd0 you've probably overwritten the device node fd0 with the textfile fstab. ls -l /dev/fd0 should look like

brw------- 1 andjoh users 2, 0 Sep 24 03:54 /dev/fd0

but in your case I suspect the first character isn't b but -. Do rm /dev/fd0 mknod /dev/fd0 b 2 0

to correct it.

Also, how are you mounting /dev/hdb9? Is it perhaps mounted readonly?

...

Is there a simpler editor than Vi? Appreciate if you could help with the above

pico, as others have suggested, is fairly simple, but beware: it has a tendency to wrap lines and corrupt config files. vi has a steep learning curve, but when you master it you'll have complete control over the files you edit.

On Thu, 27 Dec 2001 13:08:10 +0100, Anders Johansson wrote:

Is this on the win partition you're getting that error? If a partition is mounted readonly you can remount it read/write with

mount -o remount,rw /mountpoint

I am not getting anywhere here :) The above gives me the response mount: can't find /mountpoint in /etc/fstab or /etc/mtab This looks like a circle equation, I cannot edit fstab because it is write protected, and cannot use your command because it is not in fstab. The command prompt is (repair filesystem) # If I use mount /dev/hdb9 I get the same response. I can load fstab but cannot save it. Regards, David

On Thu, 27 Dec 2001 14:33:08 +0100, Anders Johansson wrote:

Well, I meant for "mountpoint" to be replaced by whatever you use in practice. If you're at the "repair filesystem" prompt you should use

mount -o remount,rw /

since / is the mount point of the root file system here.

Sorry, taking things to literally. mount -p remount,rw / That did the trick and I have now edited fstab successfully. Now it is back to the original query that it will not boot, coming up with the same error. FAT bread failed iosfs_read_super: bread failed, dev=03:4b, iso_blknum=16, block=32 Kernel panic: VFS: Unable to mount root fs on 03:4b I would have thought that I would have to alter the MBR to point to where Linux now sits Using the Suse boot disk and booting the installed system from that, it is coming up with so many errors it is probably going to be quicker and safer to re install the system. Regards, David

On Thu, 27 Dec 2001 16:54:36 +0100, Anders Johansson wrote:

I'm a late entry in this thread, but have you edited /etc/lilo.conf? If linux changes location you will have to change the root= line in that file.

Yes I have done that, but I think I have done a professional job in corrupting this system. Just for the history, I tried to add another partition to Linux using Partition Magic. I needed to be in Windows to see what was in that partition before deleting it. At the time it seemed logical to use PM and since Lin is at the end of the disk, did not think it would affect it - older and wiser now :) Tried to retrace my steps but had gone a few too many to remember all. That is why the suggestion is to change the partition reference in fstab. Have used fdisk to find out where they are. I'll give it a few more shots if you think it is worthwhile, but I think that it's looking more and more like a re install. Regards, David

Hi Geoff On Thu, 27 Dec 2001 15:59:42 +0000, Geoff wrote:

I have to be a little careful, because YAST2 obviously has some differences from my YAST.

In my version, and I am sure somewhere in yours, YAST has an option to "Boot Installed System". We know that your installed system is, in fact, on /dev/hdb9. If, therefore, you select "Boot Installed System" and enter <hdb9> (or maybe ) at the prompt, your existing system should be booted and will be usable in the normal way. This means that /etc/fstab will be writeable and you should be able to use kedit or whatever editor you prefer.

Yes, can boot the installed system, have edited both fstab & lilo.conf to point to /dev/hdb9

There is something else that you will (I think) also need to do, and that is to tell lilo about the new location of your root device.

Somewhere in your /etc/lilo.conf is a stanza something like :

image = /boot/vmlinuz root = /dev/hdbX <-- NAME OF OLD PARTITION label = linux

You obviously need to edit that to show hdb9 in place of the old partition.

Done that

When you have edited the file, then as root re-run lilo

That gives me Fatal: open /boot/boot.b: No such file or directory /boot is there, but there are no files listed in it.

That should reinstall lilo and the system ought now to reboot properly next time.

I am sorry I did not go into all of the above in this morning's post. Her Ladyship was anxiously waiting to go to the sales.

Don't worry about that - didn't realise you were married to Royalty - respect. I suppose you get a bit of "how much longer are you going to be on that ******* machine" :)

Lastly ... there is obviously an element of risk in all of the above - once lilo has been re-run there is no easy way back. I think it is your best hope though, short of a reinstall. If you feel that there is any chance of all this having an adverse impact upon elements of your setup that you simply dare not compromise, then my obvious advice is not to do it. You cannot, however, do any harm by doing all that I have suggested except re-intalling lilo, and getting that far will at least tell you if there is a workable root file system on /dev/hdb9.

There is nothing of importance on the machine, it is basically the time to set it all up again. I have used part of the update in Yast to create a backup on another partition, which if I understand it correctly will restore my settings. If you have any further ideas they are welcome, if not I will re install later. Regards, David

On Thu, 27 Dec 2001 18:10:25 +0000, Geoff wrote:

Before we get all apocalyptic :

(1) Go back to fstab please. Just as /dev/hdb9 is now shown mounted on / , what do you see mounted on /boot?

(2) Go back to can you see a smallish partition shown as a linux partition?

The reason for these questions is that /boot is normally a separate small partition. If it is still there then we need to mount it on /boot in the same way that you mounted /dev/hdb9 on /. If all goes well lilo should reinstall without an error.

I think you have hit on the problem here. It looks like I managed to delete that partition. fstab gives me various windows partitions, then at the end /dev/hdb8 Linux swap /dev/hdb9 Linux Just the 2 linux partitions Going to the fstab backup the top 2 lines before the CDRom are /dev/hdb11 / ext2 defaults 1 1 /dev/hdb9 /boot ext2 defaults 1 2 which I assume would make /dev/hdb10 the swap. The question now is whether a boot partition can be created. Space is not critical.

We are long past that ... pained resignation and/or "**I** am setting off for the sales in 5 minutes <pained look>" have been the order of the day for many years now.

Same here, different script :) Regards, David

On Thu, 27 Dec 2001 22:27:44 +0000, Geoff wrote:

I am pretty sure that the partition can be created and that (with a little more effort), a bootable kernel can be put on it. I am very happy to give what help I can, but I have never had to do exactly what is planned. It is very much a question of whether you are interested enough to continue. Yes fine with that, but read on

One good reason for doing so is that you are at least learning valuable recovery techniques on a system about which you do not care to much. I can tell you from hard experience that it is better to do it this way than on a system into which you have invested many hours - but then you knew that ;-) You can also take smug satisfaction from the fact that recovery *is* possible - I doubt that any Windows system could be subjected to this kind of abuse and survive - let alone survive well enough to be used as a tool in its own resurrection (/end of sermon).

In the past when Windows has gone funny you just re install, fine with only one program, not so good with a whole collection. Have to do it anyway every so often, it just plain gets constipated :)

If you want to go ahead, we (not least I), need to take it in stages (though if anyone reading this knows a quick and easy way to do everything please feel free to jump in here).

The first stage is to decide where to make space for the new /boot. It can be a *very* small partition - say 5MB and I think that we could steal it from swap or (perhaps better), put it at an earlier point on the disk. Before we try stealing from swap, can any of your Windows partitions be sacrificed instead? If so I think that we could convert the sacrificial partition to the ext2 needed for /boot simply by The -c switch just checks for bad blocks before creating the file system.

Let me know what you think about that plan. If it works, the next stage will (I think, and unless any reader knows better), be to compile a new kernel and put it that partition. I don't know if that is covered in the manual you have, but if you want to get ahead of me here, in my 7.0 pro manual it is section 4.5 "Installation After Recompiling a Kernel".

If we need to steal some space from an existing partition (eg swap), then the parted program can do it - I am not sure if it supplied as part of the 7.3 distro - the version I am running I d/ld and compiled. Anyway, we can cross that bridge if and when ...

Right. Now knowing where the problem lay I thought I would have a play around. You probably won't approve of this, but I went back to P Magic and created a small partition in front of the swap (7.8mb was the smallest), and at the same time thought that I would create the partition I had intended, both in ext2. So after the win partitions I have new lin \ boot \ swap \ root lin Re booted and had the same error. Booted via the floppy disk and used boot installed system. This time it booted into X and loaded KDE. Played around with fstab changing to the new drive id's. Couldn't get Lilo to work, it could not find /boot/boot.b. Created that as an empty file with Pico. Then it said can't mount /dev/hda. Little more head scratching. Restored the original fstab. Went back to P Magic and split a Win partition to bring the Lin partition id's back to their original value. Rebooted from HD and it recorded errors in fstab and went into file repair mode. Checked fstab and it had added an extra partition which didn't exist, so changed that. Lilo still would not work. Created /boot/boot.b again. Re booted and voila - all the way to KDE. Have re booted again to prove it. Bit unconventional but so far it works, but have a nasty feeling that something is going to turn up and have me crying to you again :) I will now have to let Lin know about the partition I added. That will have to wait 'till tomorrow Thanks for all your help and Anders. Regards, David

On Fri, 28 Dec 2001 11:35:12 +0000, Geoff wrote:

Well, as the pilots say, any landing you can walk away from is a good landing. I still can't quite visualise how you come to have a bootable vmlinuz, let alone boot.b. If it has once booted I doubt that there will be any problems until (maybe - I am guessing), the next amendment to lilo.conf and reinstall, and perhaps not even then.

It had a sort of logic to me in that I was attempting to put things back as they were. However although it runs, boots, etc it is not perfect. Lilo still brings up the same error Fatal: read /dev/hda: No such file or directory and on the console log I have linux kernel: Cannot find map file. I have sorted the new partition out and Linux seems happy with it.

On a different subject, sorry I can't help with the kernel change question you posted .. I run vanilla kernels and have never installed a SuSE update.

Don't worry about that. I have had some help in installing a vanilla kernel but I could not get it to boot, so abandoned that. Thought I would try the Suse kernel instead to if I can get that to work. We'll see :) Regards, David

On Fri, 28 Dec 2001 18:26:47 +0000, Geoff wrote:

The second of these messages "linux kernel: Cannot find map file." is the easier one to fix. The kernel is looking in /boot for the System.map, which is created during compilation. I am not certain sure where is is on your distro -specific setup, but it may be in /usr/src/linux. If you can find it you should copy it to your /boot. This file is not to be confused with the one just called "map" that is created by lilo when is successfully run, and which also lives in /boot.

Don't have system.map on the system.

The first error, "Fatal: read /dev/hda: No such file or directory", is, I think, probably a consequence of your fake boot.b. boot.b is part of lilo and is, as I understand it, the linux loader. To be honest, I am not clear what is doing your loading without it. The usual way to get it back is to reinstall lilo from the rpms. For an experiment I will send you mine by private email, which you can copy into /boot for an experiment. It has a brother, chain.b (used to load windows I think - I may be wrong), which you may as well have too.

Thanks for that, but it doesn't like it. Comes up with error Fatal: First boot sector is version 20.0. Expecting version 21.6 Had a look on Google for "create boot.b" Only one result and that was to re install Lilo. Did that and it re populates /boot including system.map. So that bit is cured but when I run Lilo I get Fatal: open /boot/message: No such file or directory On second thoughts have just created an empty file called message in /boot and lilo now works. Are you sure Linux is not a development of Cluedo :) Regards, David

On Wed, 26 Dec 2001 23:20:38 +0000 David wrote:

Hi Geoff Hope you have had a good Christmas and everyone else on the list

Back to my problem I used fdisk and Linux is now on /dev/hdb9.

OK .. let's try to circumvent sending all of fstab to the list. Somewhere in fstab should be a line something like this /dev/hdbX / ext2 defaults 1 1 The final field may not be identical to mine, - the key field that you are looking for is the second one containing a simple "/". That shows which device is mounted as root. If Rene's analysis was right, it will be the old number of your linux root partition. All you should need to do is to edit that line to show /dev/hdb9 as the device mounted on root. Even the very wonderful vi should allow you to do that. Regards (and Happy New Year), Geoff _________________________________________________________ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com

Thanks Rene for the help. I managed to manoeuvre myself to the root of the main on the ext2 partition, but I could not work out what drive I was in. mnt/dev/hdbX gave permission denied. Managed to load /etc/fstab but again could not see what line to edit. So questions are:- do you change drive by /dev/hdbX or mnt/dev/hdbX, What do you use to see which drive you are on What line in fstab is the line to change. Very basic questions but would appreciate your help Regards, David On Fri, 21 Dec 2001 16:39:33 +0100, Rene Engelhard wrote:

I hat the same thing a year ago as I done something similar as you. he problem arose beacause the ext2 partition was /dev/hdaX. Now you deleted Y paritions before that.

Now the ext2 ist /dev/hdaZ. (any number depending on your parition table). So he could not find his root fs anymoure

...

From there you could do nothing.

Use the boot disk to start the rescue system.

Try mount your ext2 partition under /mnt. (search all /dev/hdaZ's to find the partition)

Edit /mnt/etc/fstab to coorect the devices there.

How is it now?

Rene