Can't boot Ubuntu 10.04 after EMC2 install

This is on a fresh, new computer (details below). The install of Ubuntu 10.04 went fine and I did all the current updates. Then I downloaded and ran the linuxcnc-install.sh and it appeared to install OK. When I try to restart, Ubuntu won't come up at all. I get a black background screen with the following alert:

ALERT! /dev/disk/by-uuid/....... does not exist. Dropping to a shell.

At this point it drops into Busybox.

I tried using a LiveCD, but it doesn't boot at all on the new system. The LiveCD appears to be OK and I can use it to boot on other (older) computers. I get a black background screen with the following message:

(initramfs) Unable to find a medium containing a live file system.

At this point it drops into BusyBox.

I've never had problems with installs of EMC2 on older PCs. Is there something that is incompatible with newer processors? Any ideas on how to get the install to work?

Hardware:
Intel Pentium G620 Sandy Bridge 2.6GHz
Gigabyte GA-H61MA-D3V motherboard with AMI EFI BIOS
4Gb RAM
500Gb hard drive

Hi

See this post, someone else had identical problems to you with a Sandy Bridge based MB, Ubuntu 10.04 loaded but the rtai kernel would not
www.linuxcnc.org/index.php/english/compo...=16427&catid=9#16427

Also here, Sandy Bridge does not like RTAI
mail.rtai.org/pipermail/rtai/2011-July/024652.html

And here, Sandy Bridge requires 2.6.37 kernel?
permalink.gmane.org/gmane.linux.real-time.rtai/24904

Your machine is too new really for a LinuxCNC controller, you don't need anything that powerful.

If you want to try it out in the house, would just the sim package do you?
wiki.linuxcnc.org/cgi-bin/wiki.pl?LinuxCNC_Pure_Simulator

regards

Thanks for your prompt reply.

After looking through the links you provided, it appears that there is no simple solution offered.

I'm still puzzled by the cause of the failure. Ubuntu 10.04 works fine on the machine (prior to the install of EMC2). So 10.04 doesn't have the problem with the newer hardware. What is it about EMC2 that causes it to fail? Apparently, EMC2 installation modifies the /boot files. I'm a novice to the inner workings of EMC2 but it appears that it needs RTAI which is not a part of the standard Ubuntu 10.04, and the RTAI isn't supported on the newer architectures. Is this the issue?

I've got an old PC running in my shop now, but the software that generates my g-code needs a lot more horsepower (graphics, RAM, newer software, etc). So I'm forced to have an old PC for a linuxcnc controller and a newer computer for generating the g-code. This is a pain because one needs to network the two computers together, copy files, etc. I was hoping to replace them both with a newer machine so I could get back down to one computer for all tasks.

The other issue is that summer is coming, and the old PC just gets too hot to work on summer days (no AC in the shop). A newer, more energy efficient machine would solve that problem too.

Hi

Yes it certainly is strange, but for a nagging idea in the back of my head that I had seen this error before with Sandy Bridge, I would have launched straight in to advising how to sort out GRUB.

On the face of it, the error is a GRUB error and you would suspect that the LinuxCNC package, in installing a new kernel, had screwed up GRUB.
The error would tend to suggest that the UUID in your grub setup for your HDD is incorrect / does not match the fstab entry or /dev/disk/by-uuid/ itself is missing (not likely).

It might be worth treating it as such and doing a bit more digging.

You should be able to press and hold shift, to get up the grub menu at boot time and then choose the original non-rtai kernel.

From there you can check if the UUID in grub.conf and fstab matches that in /dev/disk/by-uuid,
and indeed if the UUID for both kernel listings in grub.conf is the same.

If these match OK, there must be some deeper incompatibility with the current rtai kernel as the previous links indicated.

See this and contained links for info on repairing grub, again in theory, this should work.
askubuntu.com/questions/37127/grub2-wont...on-separate-harddisk

If you can't access your old kernel, then use the Ubuntu Live CD ( not the LinuxCNC one ) and mount your drive, chroot etc as detailed in the link.

Would be interested to see what you find, if you don't mind doing some more experimentation.
Distance diagnosis is always difficult.

regards

Thanks for your reply and helpful suggestions.

By holding SHIFT I got to the grub menu. There were 2 listings for "Ubuntu, with Linux 2.6.32-122-rtai" (one with recovery mode) and there were 4 listings for "Ubuntu, with Linux 2.6.32-41-generic" (two with recovery mode). There were also two memory test options.

By choosing the "Ubuntu, with Linux 2.6.32-41-generic" I was able to boot into Ubuntu. (Thank you -- I don't have to reinstall everything).

Under /dev/disk/by-uuid/ there were 2 listings. The file /boot/grub/grub.cfg showed the 8 menu entries mentioned above, and all of them pointed to the same uuid (which was one of the two listed under /dev/disk/by-uuid/). The first two menu entries used /boot/initrd.img-2.6.32-122-rtai (which exists in the /boot directory) the next 4 used /boot/initrd.img-2.6.32-41-generic (which also existed under /boot of course).

Manually selecting Ubuntu, with Linux 2.6.32-122-rtai (Recovery Mode) from the grub menu listed progress, but ended up with the same error that was mentioned at the beginning of this thread. Interestingly, the uuid it says doesn't exist does in face exist.

This seems to me that the problem was something with the initrd.img-2.6.32-122-rtai.

Any other ideas?

Hi

This seems to me that the problem was something with the initrd.img-2.6.32-122-rtai.

It may be a case of chasing symptoms which don't directly relate to the root problem, but we can test that quite easily.

The initrd image is certainly generic and won't have been built on a Sandy Bridge chip set machine.

You need to get mkinitramfs from the initramfs-tools package

Having saved your existing copy of the initrd.img
run sudo mkinitramfs -o initrd.img-2.6.32-122-rtai 2.6.32-122-rtai in a terminal and copy the resultant file to /boot

(Because I already have it all installed, I don't know if you will need to install build-essentials and the kernel-headers and maybe some others first, hopefully they will be dependencies and apt will install them for you)

Then try to boot the rtai kernel and see if you fare better.

NB
The magma CVS now has a rtai patch for the 2.6.37 and 2.6.38 kernels.
I will have a look later at whether I can match them to a kernel sub-version that will build correctly (it is often not straight forward).
If so I ***MAY** be able to produce a rtai kernel from a later kernel version which allegedly will run OK on Sandy Bridge.

No promises, but it has to be worth trying.

regards

Hi

If so I ***MAY** be able to produce a rtai kernel from a later kernel version which allegedly will run OK on Sandy Bridge.

Managed to build a 2.6.37.6-RTAI kernel with realtime files and Linuxcnc build to match
The main difficulty was not with the kernel or magma but linuxcnc itself.

Quite a few of the drivers in src/hal/drivers/ use a deprecated function pci_find_device() which is now replaced by pci_get_device() in linux/pci.h
This obviously changed in kernels sometime after 2.6.32

Check out the return from uname -r in the underlying terminal



regards

Hi Bill

How are you getting on?

I am looking for someone with a Sandy Bridge chipset motherboard to test the 2.6.37.6-RTAI kernel I have built.

I now have just about finished a set of installation scripts which will custom build the kernel on your machine from a base Ubuntu 10.04 installation.
They should require relatively little Linux knowledge and none about compiling etc. , just being able to follow a few instructions to launch the scripts.

If that kernel works OK with the Sandy Bridge chipset, it could extend the working life of the 10.04 based LinuxCNC installation for newer machines, until the next
generation distro is released.

regards

My apologies for the delay in responding, I've been out of town visiting my new grand-daughter!

Earlier this week, I re-installed the old PC computers in the shop so that I can continue running. So I haven't pushed any further on trying to figure out the issues with the new Sandy Bridge processor.

Yes, I would be happy to test the 2.6.37.6-RTAI kernel. I appreciate your effort in helping to resolve the issue.

I was a bit intimidated by the though of having to patch the kernel myself. I'm not a computer dummy, but I'm not a Ubuntu/Linux expert by any means. It's been a long time since I've done any C programming (mostly I write in Java these days).

Looking ahead just a bit further -- assuming we get over the current hurdle with the kernel/processor compatibility, will I have issues using a PCI Express parallel port?

Let me know the best way to proceed. This week is pretty free for me, but then I'm out on travel for a week+.

Hi

I just finished testing the scripts on a vanilla Ubuntu 10.04 install, so your timing is good.

Inside the posted zip file are 4 scripts, a kernel configuration file and a README file.
Copy the zip to /usr/src and unzip it with sudo unzip 2.6.37.6-RTAI-build.zip

Then open the README, it gives the sequence to follow.

Basically the first script takes you through getting su root access (this is hidden by default in Ubuntu)

The second updates your system for building the kernel and downloads the sources and then builds the new kernel.

Reboot and run the third which builds the magma realtime system

Reboot and run the fourth which builds LinuxCNC against the kernel and realtime libs

If you hit problems, capture the screen output so I know exactly where you are and get back to me

Very interested to see if this runs OK on your chipset

regards