Bug 51770

there should be more errors (from the mount command, perhaps) further up the log.

Created attachment 29355 [details]
initrd done with mkinitrd initrdttt.img 2.4.8

I have attached the initrdfile created, it panics just after loading the scsi
module, but there are no traces in the logs (which is normal, because the root
drive isn't mounted yet), anyway here are the messages:
---Afterscsimoduledetectionblablablayadayada
Mounting /proc filesystem
Mounting root filesystem
pivotroot: pivot_root(/sysroot,/sysroot/initrd) failed:2
Freeing unused kernel mem: 232k freed
Kernel panic: No init found. Try passing init option to kernel.

here is the version used of mkinitrd:
Creating root device[lethalwp@little RPMS]$ rpm -qa |grep initrd
mkinitrd-3.1.5-1


I'm also putting as attachment a rpm -qa >rpmlist.txt
I have to admit i haven't used the install of the roswell cds when upgrading
from my 7.1, i did the update all manually using rpm -F (which was way 'longer'
to do than wolverine -> 7.1 :/, a lot of new packages needed...)

Again, for me, it is calling some devfile or so that does not exist?

Kernel used if 2.4.8 homecompiled (using the latest rpm of redhat .config, with
a make oldconfig or so, changing some little stuff)

Created attachment 29356 [details]
my   rpm -qa |sort >rpmlist.txt

Are you sure?  It should have said:

Mounting /proc filesystem
Creating root device
Mounting root filesystem

Yup sorry, copy/paste that went wrong it did say:

---Afterscsimoduledetectionblablablayadayada
Mounting /proc filesystem
Creating root device
Mounting root filesystem
pivotroot: pivot_root(/sysroot,/sysroot/initrd) failed:2
Freeing unused kernel mem: 232k freed
Kernel panic: No init found. Try passing init option to kernel.

does your root filesystem have an ext3 journal file on it?

Nope, it is "good old" ext2
but what i don't understand is the following:
[lethalwp@little lethalwp]$ locate mkrootdev
[lethalwp@little lethalwp]$ 

I have no mkrootdev file (same kind of result with find of course)
How can it make the 'rootdev' if the program it is calling is not present?

no, these are special nash commands.  They are built in to nash.  The initrd is
way too small to have a full runtime environment for a shell and commands.

What device is your root filesystem on?  What bootloader are you using?  Can you
upload your bootloader config file?

I have done a little more digging, but can be wrong: and remarked this:

actually: [root@little kernel]# cat real-root-dev
2066
which is  0x812  (hrm, but root is on  brw-rw----    1 root     disk       8, 
18 Jul 18 17:58 /dev/sdb2
so i was expecting 0x818 or so?)
Anyway, in the linuxrc file i can read:  
mkrootdev /dev/root                                                      
echo 0x0100 > /proc/sys/kernel/real-root-dev   

I presume: 0x0100 is brw-rw----    1 root     disk       1,   0 Jul 18 17:58
/dev/ram0) which is not the real needed root?


And, to end, i've read this little info in the
/usr/src/linux/Documentation/initrd.txt:

"Obsolete root change mechanism
------------------------------

The following mechanism was used before the introduction of pivot_root.
Current kernels still support it, but you should _not_ rely on its
continued availability. ... ... ..."
Which explain the use of echo something>real-root-dev

I'm lost on this, why trying to implement the use of real-root-dev? What is the
need of pivot_root? I thought that once the linuxrc was ended, initrd was
cleaned up,and kernel mounted the root by its own (using rdev or lilo setup)

And actually, it's this last method i am using to get it working

Adding lilo.conf & fstab scripts in a tiffy ;)

Created attachment 29375 [details]
my fstab

Created attachment 29376 [details]
my lilo.conf

18 decimal == 0x12.

you'll notice we're using pivot_root now.  What we're doing is tricking the
kernel into thinking that the real-root-device is the ramdisk (which is really
the initrd).  This means that the kernel won't be doing anything at all when
linuxrc ends except for exec'ing /sbin/init.

you've told me initrd was to little to put a bash in it, but that gave me the
idea of doing it... giving it 8MB of ram, bash, ls, cat, + libs

anyway, i have found the explanation, which is quite ridiculous, and was so
evident :(

after the:
pivotroot: pivot_root(/sysroot,/sysroot/initrd) failed:2
in the bash, i've seen my rootdev was correctly mounted in /sysroot
and then checking /sysroot/initrd: no such directory, so i've made one, and the
system was able to boot.

I really do not understand the need of using that 'tempdir', which, if deleted,
will stop the whole bootsequence?


But, if i may ask one little last question:
kernel said: real-root-dev: 2066  which is in hex=818  and you' telling me to
take those number appart?, and convert 8->hex  and 18->hex?
(A Yes/No question ;) )


Thank you for helping, well-done ;)
Greetz, lethalwp

run "rpm -y mkinitrd" - it required a new RPM of filesystem which made the
directory, it seems you don't have that installed...