kernel build problem

Ok, I have RH 9 on a dual boot box with win2k on the other partition. I wanted to add ntfs read support in the kernel. I configured my kernel, and made ntfs read support built-in (not modular). I then did the following as root:

make dep && make clean && make && make install

when i reboot and try the new kernel, i get something along the lines of

kernel panic, no init found, try passing an argument at command line (init = )

at the beginning of boot. I tried again with:

make dep && make clean && make bzImage
then copied bzImage over to /boot, updated lilo.conf manually, and ran lilo && lilo -q... same result.

Also, is it right that I didn't do a 'make modules && make modules_install' since all I did was add something to the kernel that was built-in?

any help would be much appreciated

It looks like the wrong partition got mounted as the root file system. Which partition do you have your root file system on?

As for not making modules, it depends. Sometimes the versioning information in the kernel changes, and that renders all previously compiled modules unusable for the new kernel. The reason is to ensure that modules that won't work with the kernel cannot even be loaded.

I ran fdisk and printed the partition table, and it is using the right partition... so I don't think it's that. Also, I looked more thoroughly at my /etc/lilo.conf and saw some things I've never seen before, and was wondering if anyone could explain it and tell me if I need to put similar lines in the area for my new kernel build

Here is what my current stable build has

Quote:

image=/boot/vmlinuz-2.4.20-13.9
label=linux
append="hdb=ide-scsi root=LABEL=/"
read-only
initrd=/boot/initrd-2.4.20-13.9.img

what is inird=... mean? and why root=.... in with the append and not by it self?
how do you make/find the right '.img' ?

and now the part that really bugs me... as I'm sure most of you know if I do a 'make install' for a new kernel build, it updates /etc/lilo.conf automatically. So, I have the following entry for the new build, which does not work

Quote:

image=/boot/vmlinuz-2.4.20-13.9custom
label=2.4.20-13.9cust
append="hdb=ide-scsi root=LABEL=/"
read-only
initrd=/boot/initrd-2.4.20-13.9custom.img

any ideas why this doesn't work?
thanks again

An initrd is an initialization ramdisk. The file that you specify is loaded into memory by the boot loader (ie. before the kernel is booted), and the kernel is given the address to it. The file contains an image of a filesystem (usually ext2, but I think it can be anything that has support in your kernel), that the kernel mounts as the root file system. RedHat's initrds usually load some filesystem modules and so before the actual root file system is used, so if you have built support for the filesystem on your root partition into your kernel, you don't need it at all.
I have found that sometimes, new kernels fail to locate the partition labels. Try the device number instead (eg. 0305 for hda5).
The root device can be given either in an append directive or in a root directive; they're the same. root=x is basically an alias for append="root=x", just as read-only is an alias for append="ro". I don't really know why they have all those aliases.
Also, might I suggest using GRUB instead of LILO?

If I build a kernel using

Code:

make dep && make clean && make bzImage

how can I create the right '.img' to go in /etc/lilo.conf for the initrd=... entry?

also, I tried to 'make modules' and got a whole bunch of errors... too many to post here. Are there any general reasons why this would fail? I didn't change any modules

finally, how can I change froim lilo to grub?

You don't need to make an initrd. Just remove that directive from your lilo config. Like I said, RedHat's initrds are only for loading additional modules before the root filesystem is mounted, so if you have all the necessary support built into your core kernel, there's no need for an initrd.
You need to post at least a sample of your errors when making modules.
Check GRUB's texinfo for information on how to use and install it.

here's a snapshot of the output from 'make modules'

Code:

/usr/src/linux-2.4.20-13.9/include/linux/netdevice.h:600: parse error before numeric constant /usr/src/linux-2.4.20-13.9/include/linux/netdevice.h:600: `dev_ioctl_R_ver_str' declared as function returning a function /usr/src/linux-2.4.20-13.9/include/linux/netdevice.h:600: warning: function declaration isn't a prototype /usr/src/linux-2.4.20-13.9/include/linux/netdevice.h: In function `netif_rx_ni':/usr/src/linux-2.4.20-13.9/include/linux/netdevice.h:615: `f0a529b7' undeclared (first use in this function) /usr/src/linux-2.4.20-13.9/include/linux/netdevice.h:615: called object is not a function In file included from /usr/src/linux-2.4.20-13.9/include/net/dst.h:12, from /usr/src/linux-2.4.20-13.9/include/net/sock.h:107, from cipe.h:69, from module.c:15: /usr/src/linux-2.4.20-13.9/include/net/neighbour.h: In function `neigh_confirm':/usr/src/linux-2.4.20-13.9/include/net/neighbour.h:233: `jiffies_R0da02d67' undeclared (first use in this function) /usr/src/linux-2.4.20-13.9/include/net/neighbour.h: In function `neigh_event_send': /usr/src/linux-2.4.20-13.9/include/net/neighbour.h:248: `jiffies_R0da02d67' undeclared (first use in this function) In file included from /usr/src/linux-2.4.20-13.9/include/net/sock.h:107, from cipe.h:69, from module.c:15: /usr/src/linux-2.4.20-13.9/include/net/dst.h: In function `dst_set_expires': /usr/src/linux-2.4.20-13.9/include/net/dst.h:149: `jiffies_R0da02d67' undeclared (first use in this function) In file included from cipe.h:69, from module.c:15: /usr/src/linux-2.4.20-13.9/include/net/sock.h: In function `sk_filter_release': /usr/src/linux-2.4.20-13.9/include/net/sock.h:959: warning: implicit declaration of function `kfree_R037a0cba' /usr/src/linux-2.4.20-13.9/include/net/sock.h: In function `sock_orphan': /usr/src/linux-2.4.20-13.9/include/net/sock.h:1041: `do_softirq_Rf0a529b7' undeclared (first use in this function) /usr/src/linux-2.4.20-13.9/include/net/sock.h: In function `sock_graft': /usr/src/linux-2.4.20-13.9/include/net/sock.h:1050: `do_softirq_Rf0a529b7' undeclared (first use in this function) /usr/src/linux-2.4.20-13.9/include/net/sock.h: In function `sock_recv_timestamp': /usr/src/linux-2.4.20-13.9/include/net/sock.h:1293: warning: implicit declaration of function `put_cmsg_Rf39bf4d9' module.c: In function `cipe_check_kernel': module.c:73: warning: implicit declaration of function `printk_R1b7d4074' /usr/src/linux-2.4.20-13.9/include/linux/skbuff.h: In function `kunmap_skb_frag': /usr/src/linux-2.4.20-13.9/include/linux/skbuff.h:1129: warning: asm operand 1 probably doesn't match constraints make[3]: *** [module.o] Error 1 make[3]: Leaving directory `/usr/src/linux-2.4.20-13.9/drivers/addon/cipe' make[2]: *** [_modsubdir_cipe] Error 2 make[2]: Leaving directory `/usr/src/linux-2.4.20-13.9/drivers/addon' make[1]: *** [_modsubdir_addon] Error 2 make[1]: Leaving directory `/usr/src/linux-2.4.20-13.9/drivers' make: *** [_mod_drivers] Error 2

hope this helps

also, i tried to take out the initrd=... entry, and wrote this entry

Code:

image=/boot/vmlinuz-2.4.20-13.9_51703 label=linuxtest append="hdb=ide-scsi" read-only root=/dev/hda2

and got the 'kernel panic: init not found' error like before. What's wrong with this entry?
I'm going to try putting in the partition number (i guess its 0302 for /deev/hda2?) instead of the label.

Yeah, I recognize those module errors. It was the same when I tried to compile a kernel from RH7.1's kernel source tree. Somehow, RH seems to mess things up with the versioning information every now and then. I dunno if that's something you're just supposed to know how to fix, but I do know that when I downloaded a vanilla kernel from kernel.org, it worked perfectly. You could try that, too.

As for your init problem, could you post your partition table? I'd just like to have a look at it. And yes, 0302 is hda2. To see for yourself, ls -l /dev/hda2, and you'll see something like 3, 2 in the output. The 3 is the major number for the device and the 2 is the minor number. To construct a number to pass to the kernel at boot, write the numbers in hexadecimal, fill it with zeroes to two hexadecimal digits, and concatenate them. That way, you get 0302 for hda2 or 210a for hde10.

Here is my parition table, from fdisk

Code:

Command (m for help): p Disk /dev/hda: 80.0 GB, 80026361856 bytes 255 heads, 63 sectors/track, 9729 cylinders Units = cylinders of 16065 * 512 = 8225280 bytes Device Boot Start End Blocks Id System /dev/hda1 * 1 8667 69617646 7 HPFS/NTFS /dev/hda2 8668 9470 6450097+ 83 Linux /dev/hda3 9471 9535 522112+ 82 Linux swap /dev/hda4 9536 9729 1558305 f Win95 Ext'd (LBA) /dev/hda5 9536 9729 1558273+ 83 Linux

just to be clear, hda2 is mounted on / and hda5 is mounted on /home

Can you try passing "root=0302 init=/bin/sh" to the kernel?