[SOLVED] kernel 2.6.31.5 will not boot

diamondsandrain · 10-28-2009, 05:53 PM

Has anyone had any experience installing and running the latest kernel (2.6.31.5)? I was trying to compile and run it (downloaded from kernel.org) so I could take advantage of the X-Fi support that is not in the 2.6.29 kernel found in slackware 13. I had a nice kernel compiled but it would not boot. Then I tried using the stock settings and that would not boot either. It gives me an error about not being able to mount the root partition and gives me the idea of specifying root= the next time I boot.

The slightly tricky thing is that I have an external eSATA drive that mounts as /dev/sda in some situations and /dev/sdc in others. I have tried several ways to do the boot and nothing seems to work. Is anyone else having any issues with newer kernels on slackware? If I revert to the old kernel everything is fine again, but I don't have X-Fi support.

tuxdev · 10-28-2009, 05:59 PM

You probably have to specify the root when you do mkinitrd, like this:

Code:

mkinitrd -c -k 2.6.31.5-smp -m ext4 -f ext4 -r /dev/sda2

Of course, you'll have to run lilo after this.

If you at least have the right modules in your initrd, you can mount the root manually when it asks you to

Code:

mount -o ro /dev/sda2 /mnt

diamondsandrain · 10-28-2009, 06:16 PM

I just tried that using /dev/sdb1 and /dev/sdc1 (if external is sda it will bump everything down by one). I get a slightly different message that /dev/sdb1 could not be mounted on /mnt. Something like "no such file or directory".

Cheesesteak · 10-29-2009, 12:57 AM

Did you add the large-memory option in /etc/lilo.conf ?

from ftp://ftp.slackware.com/pub/slackwar...ADME_FIRST.TXT

Code:

Due to the size of the 2.6.30.5 kernel and initrd, many BIOSes cannot boot
the kernel/ramdisk combination for i686, and most (if not all) will have
problems with the x86_64 kernel/ramdisk unless this option is added to the
/etc/lilo.conf:

large-memory

This allows LILO to use more than the first 15MB of RAM.  LILO does not use
this option by default since it doesn't work with some older systems (BIOSes
from before around 2001).  LILO tries to detect the amount of space that will
be required to decompress the kernel, but some adjustment to the code is
going to be needed, or perhaps we will have to investigate alternatives for
the bootloader.  In any case, that's something for Slackware 13.1 or 14.0, or
whatever happens down the line.

The option can go into the lilo.conf global section.  It actually works just
fine to make it the first uncommented line of the file.

uppman · 10-29-2009, 03:34 AM

Hi!

Please provide some more info.

e.g.

cat /proc/cmdline

cat /rootdev

cat /rootfs

cat /proc/partitions

lsmod

lspci -v

diamondsandrain · 10-29-2009, 10:14 AM

Large-memory didn't make a difference but I am noticing that it is saying something about not being able to mount root device "821". Is this some kind of code for the partition or is something terribly wrong?

diamondsandrain · 10-29-2009, 11:14 AM

I think I've royally screwed up the root partition somehow. I tried putting it back to the 2.6.29.6 kernel but now every time I boot I get prompted to check the filesystem and it will not fully boot. I try running fsck.ext4 /dev/sdb1 but it comes back clean without even looking like it has done anything.

Also, I am not able to open up config files in vi because it says something about /var/tmp being too full or too busy or something like that. I can't even reboot most of the time because it gives me the error "timeout opening/writing control channel /dev/initctl".

Is it time to reinstall?

This was all in my attempt to provide the information requested.

onebuck · 10-29-2009, 02:21 PM

Hi,

What happens when you use the install cd/dvd to boot your kernel?

diamondsandrain · 10-29-2009, 03:54 PM

I tried booting using huge.s root=/dev/sdb1 ro but it just goes directly to the install environment. Isn't it supposed to boot the real partition with the install kernel?

Seems like it is getting some error from a "boot array".

onebuck · 10-29-2009, 04:32 PM

Hi,

What is the name for the image that you are wanting to boot? You will find it in '/boot' of your boot device;

Code:

sample of '/boot';
# ls /boot
README.initrd                boot_message.txt         diag1.img                   slack.bmp
System.map                   config                   diag2.img                   vmlinuz
System.map-generic-2.6.29.6  config-2.6.32-rc5kms-1a  initrd-generic-2.6.30.5.gz  vmlinuz-2.6.32-rc5kms-1a
System.map-generic-2.6.30.5  config-generic-2.6.29.6  initrd-tree                 vmlinuz-generic-2.6.29.6
System.map-huge-2.6.29.6     config-generic-2.6.30.5  initrd.gz                   vmlinuz-generic-2.6.30.5
boot.0800                    config-huge-2.6.29.6     map                         vmlinuz-huge-2.6.29.6

Notice in the above there are several 'images' that are available for this system.

I could supply the 'vmlinuz-2.6.32-rc5kms-1a' image as the kernel.

diamondsandrain · 10-29-2009, 04:39 PM

The image I was trying to boot was /boot/vmlinuz which is a link to /boot/vmlinuz-huge-2.6.29.6. But when trying from the cd I assume I only need to supply huge.s and it will load that kernel from the install cd.

diamondsandrain · 10-29-2009, 04:45 PM

I have tried reinstalling the kernel package from the disk as well.

diamondsandrain · 10-29-2009, 05:59 PM

Well, I think I figured out why it wouldn't boot right. I tried turning off the external hard drive for simplicity and as a result the root partition was being recognized as /dev/sdb1 but in the /etc/fstab the root partition was listed as /dev/sdc1 so it wasn't finding the partition. I reinstalled before I figured this out. I found out because the reinstall and first boot went fine but once I overwrote the /etc directory with my backups I was back to a non booting system. I wouldn't be surprised if the new kernel wasn't booting because it recognized things in a different order. I mean, the Arch Linux install I have recognizes the external hard drive as sdc but the Arch Linux boot cd recognizes it as sda so that is quite strange.

Still no X-Fi working yet but I am doing another recompile to see if I can figure that out.

tuxdev · 10-29-2009, 06:15 PM

I think you can set up your udev rules so that it always assigns your external hard drive to a specific device name rather than letting it (unpredictably) name your drives depending on whether it's plugged in while you're booting.

onebuck · 10-29-2009, 06:49 PM

Hi,

Quote:

Originally Posted by diamondsandrain

The image I was trying to boot was /boot/vmlinuz which is a link to /boot/vmlinuz-huge-2.6.29.6. But when trying from the cd I assume I only need to supply huge.s and it will load that kernel from the install cd.

NO!

By using 'huge.s' you are telling to use the kernel from the cd/dvd.

Your assumptions are wrong. If you want to use your 'working' installed kernel in '/boot' then you need to designate it.

As for your 'udev' recognition you can setup the rules to reflect your setup. It does sound like a miss configuration for the initial install.