[lug] mysterious fstab

[D. Stimits]

"Harris, James" wrote:
> 
> Well, I can't answer for everything, but you are correct in assuming that
> the "LABEL=" causes a substitution with the device name.  This is achieved
> by formatting the filesystem with a label such as "/boot", in this example.
> When you specify the mount using the label instead of the device, it goes
> out and mounts the filesystem it found at boot time with that particular
> label.  This is a super cool way of keeping things more dynamic.  I suppose
> these labels could affect a make bzdisk if it is getting information from
> the fstab and doesn't understand how to replace the labels.
> 
> In case you want to verify what your filesystems are you can always take a
> look at their superblocks to see the label:
> /sbin/dumpe2fs -h /dev/hda1
> 
> You may want to do this on all of them, make a backup of the current fstab
> and replace it with one that uses the device names instead of the labels.
> Then give the make bzdisk another try, it'd only take a minute and might
> resolve the make bzdisk problem.

It was a good try, but a lot of experimentation with different fstab and
other things, and it still fails. I begin to wonder if compiling the
scsi drivers in is sufficient. It acts like there is something else it
needs, although the mkbootdisk works (without adding any ramdisk
modules). Incidentally, if anyone has seen this before and knows of a
cure or clue, here is the repeating message right after lilo:
400
AX:0212
BX:6C00
CX:5001
DX:0000.

D. Stimits, stimits at idcomm.com

> 
> -Jim
> 
> -----Original Message-----
> From: D. Stimits [mailto:stimits at idcomm.com]
> Sent: Saturday, May 26, 2001 23:59
> To: BLUG
> Subject: [lug] mysterious fstab
> 
> After installing RH 7.1 on a machine which uses a separate /boot/ and /,
> I went to make a new 2.4.5 kernel for it, and wanted to test it on a
> boot floppy. The boot floppy fails with repeating AX, BX, CX, DX style
> register dump, infinitely repeating. So one thing I did was consider
> mkbootdisk. I noticed in the man page it says it reads /etc/fstab. But
> when I look at /etc/fstab, I see these lines for / and /boot:
> 
> LABEL=/                 /                       ext2    defaults
> 1 1
> LABEL=/boot             /boot                   ext2    defaults
> 1 2
> 
> How can it get away with no mention of the actual device for "/" and
> "/boot"? It should be /dev/sda2 and sda1, but nowhere does it mention
> it. I can only assume there is some new means of doing a form of
> variable substitution on "LABEL"...otherwise I don't see how the
> original install could succeed. What the heck is going on there? Could
> this have anything to do with a "make bzdisk" creating a failed boot
> floppy? FYI, I did switch the compiler named from gcc to kgcc for the
> compile. Has anyone here successfully compiled a stock kernel with RH
> 7.1?
> 
> On another area of kernel builds, more out of curiosity, I notice that
> the default RH kernel install included in /boot/ not just the vmlinuz
> images, and any initial ram disks, it also includes a vmlinux file
> corresponding to each vmlinuz image. What is the purpose of the vmlinux
> file if vmlinuz is the actual boot kernel? Is this part of initial ram
> disks or support for a separate /boot/ partition?
> 
> D. Stimits, stimits at idcomm.com