Dear Friends,
On my explerimental partition, just for fun, and to test drivers, &c, I sometimes do things either adventursome or stupid, depending upon one's point of view.
I should like to install the grml 2.6.22 kernel, but, having installed the image, sources, some modules I use, &c, after mkinitrdramfs while the package is built and installed, lilo is correct, upon boot everything fails because, at least according to the running text, it does not find /lib/modules...?
Now, lib modules 2.6.22 is there, but very sparsely populated.
Could someone, forgiving my inexperience and inadequacy, tell me, step by step, how to make this kernel work, as I am either failing to install something necessary, or failing to create the proper initrd image?
I apologise for the inconvenience.
All good wishes,
martin In caecus terrae, luscus rex est.
* martin yazdzik yazdzik@nyct.net [20070716 18:11]:
On my explerimental partition, just for fun, and to test drivers, &c, I sometimes do things either adventursome or stupid, depending upon one's point of view.
I should like to install the grml 2.6.22 kernel, but, having installed the image, sources, some modules I use, &c, after mkinitrdramfs while the package is built and installed, lilo is correct, upon boot everything fails because, at least according to the running text, it does not find /lib/modules...?
Now, lib modules 2.6.22 is there, but very sparsely populated.
Could someone, forgiving my inexperience and inadequacy, tell me, step by step, how to make this kernel work, as I am either failing to install something necessary, or failing to create the proper initrd image?
If you are using the grml-testing repository just run 'aptitude install linux-image-2.6.22-grml', otherwise just download http://dufo.tugraz.at/~prokop/grml-kernel/2.6.22-grml/linux-image-2.6.22-grm... and install it using 'dpkg -i linux-image-2.6.22-grml_grml.01_i386.deb'.
Make sure you have a initramfs for booting your box with the new kernel version, in case of question just execute 'update-initramfs -c -t -k 2.6.22-grml'.
Finally make sure you have an entry for 2.6.22-grml in /etc/lilo.conf when using lilo and don't forget to execute lilo then. If you are using Grub just run 'update-grub'.
That's it and that's the way how I upgrade my systems as well. :)
regards, -mika-
This is progress - thanks - much further along.
now, ato boot-up we get /init... cannot find /dev/root and so on.
or, when running same kernel in debug mode, it stops at registering raid personality something(no way to log this, sorry)
I think this means that the kernel thinks I have hda instead of sda?
Funny enough, this appears, to the ignorant martin, to be another one of those uuid issues, or dmraid issues, which I never seem to have the skill to resolve.
Best, m
On Mon, 2007-07-16 at 18:35 +0200, Michael Prokop wrote:
Make sure you have a initramfs for booting your box with the new kernel version, in case of question just execute 'update-initramfs -c -t -k 2.6.22-grml'.
Finally make sure you have an entry for 2.6.22-grml in /etc/lilo.conf when using lilo and don't forget to execute lilo then. If you are using Grub just run 'update-grub'.
Grml mailing list - Grml@mur.at http://lists.mur.at/mailman/listinfo/grml join #grml on irc.freenode.org grml-devel-blog: http://grml.supersized.org/
Okay - found problem, no solution:
in 2.6.22, libata represents sdX as hdX, but, obviously, I cannot change lilo.conf to root=/dev/hda4, since it gives the error that the device does not exist, which, in fact, it does not under any running kernel.
Any ideas?
Best, m
It is easy to make salami from a dog, but you cannot make a dog out of salami.
Dear Friends,
Progress in understanding the issue but no solution -
after proper install of kernel 2.6.22 and initrd, the kernel cannot find /dev/root. googled that for some reason the new libata calls sdx hdx - okay
I did:
edit lilo to append to the .22 entry root=/dev/hda4
now, the system booted, but since the kernel saw the dev as hda, dma was enabled, and so on, meaning that the hdd was slow, as in langsam or noch langsamer.
cat /proc/scsi/scsi revealed only the dvd device was scsi.
Now, I need to have the kernel direct read my sata drive - as in my mission critical grml, d h
yazdzik@grmllap1 ~ % cat /proc/scsi/scsi Attached devices: Host: scsi0 Channel: 00 Id: 00 Lun: 00 Vendor: ATA Model: HTS726060M9AT00 Rev: MH4O Type: Direct-Access ANSI SCSI revision: 05 Host: scsi1 Channel: 00 Id: 00 Lun: 00 Vendor: _NEC Model: DVD+-RW ND-6500A Rev: 203D Type: CD-ROM ANSI SCSI revision: 05
So, is there a way to force lilo or the new kernel to recognise this and work normally?
Any kernel gurus on the list?
Or am just being plain stupid?
MfG, Martin
Chi troppo vuole nulla stringe
* martin yazdzik yazdzik@nyct.net [20070716 20:15]:
This is progress - thanks - much further along. now, ato boot-up we get /init... cannot find /dev/root and so on.
[...]
* martin yazdzik yazdzik@nyct.net [20070716 21:15]:
Okay - found problem, no solution:
in 2.6.22, libata represents sdX as hdX, but, obviously, I cannot change lilo.conf to root=/dev/hda4, since it gives the error that the device does not exist, which, in fact, it does not under any running kernel.
[...]
* martin yazdzik yazdzik@nyct.net [20070716 22:02]:
Progress in understanding the issue but no solution -
after proper install of kernel 2.6.22 and initrd, the kernel cannot find /dev/root. googled that for some reason the new libata calls sdx hdx - okay
I did:
edit lilo to append to the .22 entry root=/dev/hda4
[...]
Martin, try choosing the UUID approach instead. Boot grml live-CD and find out the UUID of your root-partition. On my system that's what's known as /dev/hda2, so I can identify the UUID via running:
# vol_id --uuid /dev/hda2 56d23a4e-6dc3-4b68-9480-6f6b7f474be0
Now use this UUID for your root-device in /boot/grub/menu.lst or /etc/lilo.conf. For example:
# kopt=root=UUID=56d23a4e-6dc3-4b68-9480-6f6b7f474be0 ro
for menu.lst when using grub, or add 'root=UUID=....' to the append-line of lilo.conf when using lilo.
You should add the UUID also to the /etc/fstab of your root-partition in the following form:
/dev/disk/by-uuid/56d23a4e-6dc3-4b68-9480-6f6b7f474be0 / ext3 noatime,errors=remount-ro 0 1
When using an initramfs generated by initramfs-tools (that's the 'update-initramfs -c -t -k 2.6.22-grml' stuff) your system should boot up now.
Important: Please notice, that you should use initramfs-tools version 0.87b and *not* a more recent one. The versions newer than 0.87b seem to have serious problems regarding booting, that's why I've put 0.87b into the grml-repository too. As long as you don't explicitely run 'apt-get install initramfs-tools/unstable' you won't find an upgraded (broken) version on your system. But please make sure you didn't hit one of the few days where initramfs-tools wasn't pinned via grml's repository yet and you upgraded to a broken package version. In case of question just grab and install http://snapshot.debian.net/archive/2007/04/18/debian/pool/main/i/initramfs-t... and re-execute 'update-initramfs -c -t -k 2.6.22-grml'.
JFTR: I'm talking about http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=431291 and AFAIK the issue isn't fixed yet. :-/ Looks like I've to debug initramfs-tools once more on my own. :-(
regards, -mika-
Thanks, Mika -
you solved the problem as usual -
now, with the addition of ide0=noprobe and ide1=noprobe to lilo, which counters the kernel's preference to load ide first, and the (correct, I hope) UUID -
yazdzik@mygrml1 ~ % cat /proc/scsi/scsi Attached devices: Host: scsi0 Channel: 00 Id: 00 Lun: 00 Vendor: ATA Model: HTS726060M9AT00 Rev: MH4O Type: Direct-Access ANSI SCSI revision: 05 Host: scsi1 Channel: 00 Id: 00 Lun: 00 Vendor: _NEC Model: DVD+-RW ND-6500A Rev: 203D Type: CD-ROM ANSI SCSI revision: 05 yazdzik@mygrml1 ~ % uname -a Linux mygrml1 2.6.22-grml #1 SMP PREEMPT Tue Jul 10 00:35:57 CEST 2007 i686 GNU/Linux yazdzik@mygrml1 ~ % su Password: root@mygrml1 /home/yazdzik # hdparm -T /dev/sda4 /dev/sda4: Timing cached reads: 1916 MB in 2.00 seconds = 958.54 MB/sec root@mygrml1 /home/yazdzik #
Mika, you are a genius, and I can go on trying to experiment with things I should not really touch.
The experimental grml 1.0 is so attractive, however, I am thinking of taking the plunge, now that everything works, and redoing the whole network.
At any rate, you have once again proven that intelligence and kindness to newbies overcomes just about any issue that can arise.
Greetings from New York, and much gratitude.
Your,
Martin
Teilnehmer (3)
-
martin yazdzik -
Michael Prokop -
yazdzik