missing link destination

[coley]

I've been battling on and off trying to get an AMD64 MD going with an i386 hybrid core with no success yet.
I've tried

• adding MD via webadmin, leaving arch type as i386, clicking setup diskless MD, wait for it to finish generating MD, quick reload router. Then tried booting MD. Nothing, MD gets no IP addr.
  Add MD MAC via webadmin, save, reload router, booting MD this time gets ip addr and continues to boot until it loops failing a mount call with error 13. Rebuilding of the image via webadmin doesn't help.
• Delete above MD via webadmin, re-add as above, set its arch type to be amd64, rebuild image and reload router. Less this time, ip addr received by the MD but cannot boot.

After some digging found that the tftboot directory had the initrd image and vmlinuz had broken symlinks. The symbolic link had the right content but the destination file was not present.

Do these files get pulled out of the PlutoMD-archxxx.tar.bz2 files? My amd64 file appears to be corrupt, bunzip decompresses it ok, but tar complains - I can't rember the exact error - " missing '.' " or something to that effect?
Am I digging in the right direction?

thanks
-Coley

[Zaerc]

Where are these symbolic links pointing to?  And wat files are there in the directory where the missing files are supposed to be?

[coley]

The links are laid out as follows
tftpboot/###/initrd.img -> /usr/pluto/diskless/###/boot/initrd.img which itself is a symlink to initrd.img-2.6.22-14-generic in that dir.

and the same for vmlinuz
tftpboot/###/vmlinuz -> /usr/pluto/diskless/###/boot/vmlinuz which itself links to vmlinuz-2.6.22-14-generic.

In both the tftp/###/ and the pluto/diskless/###/ the only files in those dirs are the symlinks.

Also noticed that the MD subdir ### doesn't get created when the MD is added via the webadmin, I have to add the MD MAC address before they get created.

-Coley.

[Zaerc]

That sounds exactly how it should be, apart from the missing actual files of course.  Did you install from the DVD or the CDs?  And what specs does your MD have?

You can also try setting the MD up manually but then as i386, maybe that can give some clues as to what is going wrong here.

[coley]

I originally installed from CD. (machine is dual boot)

MD spec is
AMD M2N-VM HDMI AM2 NF630A MATX VGA+SND+GLN
ATHLON 64 X2 4200+
1GB RAM

-Coley.

[Zaerc]

Ah yes, the CD install is unable to generate the needed tarballs for other architectures.  (So is the DVD install, but this one has the prefabricated files already included.) 

Either manually set your MDs up as i386, or copy the file PlutoMD-amd64.tar.bz2 from the diskless-images/ directory of an LMCE DVD to /usr/pluto/install.  I haven't tried that recently but it used to work fine with the betas.

[coley]

Ok, did that. copied PlutoMD-amd64.tar.bz2 as you said, changed MD arch to amd64 in webadmin, rebuit image and quick reload of router. Now the boot directory has files and looks healthier.
Booted the MD, starts to boot the image provided but fails, gets stuck in a loop:

Code:

Begin: Running /scripts/nfs-premount ...
Done.
Mount call failed: 13
Done.
Begin: Retrying nfs mount ...
<Block of info on IP connection>

until it gives up and drops me out to busybox.
I can ping the moon### from the core but not ssh or telnet to it.

at the very end when the scripts give up I get

Code:

Begin: running /scripts/nfs-bottom
Done.
Done.
Begin: running /scripts/init-bottom

Mount fails trying to mount /root/dev, /sys and /proc as the destinations don't exist.
the last error: target filesystem doesn't have /sbin/init

but /usr/pluto/diskless/###/sbin/init exists on the core?

any other insights?

thx
-Coley.

[Zaerc]

Sounds like you have an Unrecognized NIC as well.  I'm affraid you'll have to set the MD to i386 before trying those instructions though as i386 cores are unable to chroot into amd64 MD diskless directories.

[coley]

Back to this again, was away for a week. Only got a brief chance this morning to investigate. Seems the chipset in use is MCP67 and the module should be forcedeth Will try these later and report back.

Would it be an idea to add known unrecognised NICs to that wiki page? along with what mobo they appear on?

How do NICs get from unrecognised to recognised? is it the kernel version on which LMCE is based that determines this?

thx
-Coley.

[Zaerc]

Back to this again, was away for a week. Only got a brief chance this morning to investigate. Seems the chipset in use is MCP67 and the module should be forcedeth Will try these later and report back.

Would it be an idea to add known unrecognised NICs to that wiki page? along with what mobo they appear on?

How do NICs get from unrecognised to recognised? is it the kernel version on which LMCE is based that determines this?

thx
-Coley.

Better would be to add your motherboard as described in: Adding your own hardware and list the needed module there along with a link to the Unrecognized NIC page.

[coley]

I tried adding forcedeth as a module and rebuilding the image.
It doesn't make any difference. Is there any way to log the console output of an MD booting? I do see a reference that forcedeth is loaded as it scrolls past.

After the MD fails to boot and gives up trying to mount I checked what it was trying to do, It attempts to run the scripts in the /scripts/nfs-premount directory in the image. But when I check the dir it is empty! so there is nothing to run, is this correct?

any more pointers appreciated,
-Coley.

[coley]

I'm still trying!
I've disabled the onboard NIC and put in an n/w card from another PC - still no joy.
MD still fails to get past the "mount call failed" repeating error.

I'm running out of ideas...

-Coley.

[coley]

Ok I think my core/hybrid is sick.
I tried network booting a HP laptop I had previously run as an MD with one of the Betas. It comes up and announces itself to the router, runs some scripts and then reboots. Except it re-announces on reboot! When I check the web admin interface I don't see a new MD.
If I set up MD from web interface and then try boot the Laptop I run into the same mount call failed.

Are there any scripts available that can validate a core install? or it it clean install time for me??

-Coley.

[coley]

moved a stage further - MD now announces itself to the router!
I changed the DHCP settings on the core to give out addresses to non-LMCE components, tried rebooting MD and it used the default image to tftp boot!! Any reason why this would spur the pxe boot into action?

Diskless directories get set up on the core - amd64 recognised and see MD via web admin.

But all the joy was short lived, ran into the "mount call failed 13" looping error.

And yes I have the amd64 tarball from the DVD in my /usr/pluto/install directory as the CD install doesn't carry this.

any pointers??

thx
-Coley.

[MarcoZan]

Hi all

I've a similar issue, but the story behind is different.

I recently moved to LMCE 7.10, and I have a MD that used to pxe boot fine with 7.04.
After moving to 7.10 the initial configuration went ok and MD was still able to pxe boot.

I decided to rebuild MD image after some glitches with video driver, and after this MD won't pxe boot anymore.
I get exactly to the same point as Coley, i.e mount call failed: 13

At this point I don't think that this could be a Unrecognized NIC problem, because if that is the case it should be appearing from the beginning.

Googling around I've found that "13" has to do with permission problems, because maybe client is trying to mount something that has not been exported, but when issuing a showmount -e on core it seems that everything is ok.

Is there anyone that have a better idea?

TIA and regards
Marco