Kernel Panic, round 2

[krys]

okay so I am back on this topic again, here is the recap.
I have a dell pc that uses this NIC : Intel(R) 82562V-2 10/100
I ran this PC as my core for a few weeks, and now I want to use it as an MD, problem is when I try to Pxe boot all i get is:
ipconfig: no devices to configure
/init: .: Can't open /tmp/net-eth0.conf
Kernel panic - not syncing: Attempting to kill init!

Now, I went back and looked at the LMCE install (its still on this pc's hard drive) and it looks like it was using the e1000 module, so I followed the wiki and added this module to the list, recreated the init ramdisk, still nothing. I googled this NIC and found some info saying that there was a new source available and that I should rebuild the module from this source and it would work. So i download the source, build the module, add to list recreated init ramdisk, reboot, SAME ERROR. I tried the e1000e module, SAME ERROR..... I want to run this thing over with my truck at this point.

Does anyone have a clue as to what I need to do to get this thing to boot? Or any suggestions as to what I could do to make sure of what my problem really is? I would think that if this NIC worked for the core that surely it is compatible and I can get it to Pxe boot. There are plenty of threads on the net about people getting this exact NIC to pxe boot (none with LMCE, but a few with other versions of Linux)

Help!
-Krys

[hari]

make sure that the exact module you used on the core ends up being included in the initrd for the MD.

br, Hari

[krys]

I booted LMCE on the unit in question and ran dmesg|grep eth and it said e1000, which is what I tried to use. when this didnt work I downloaded a few new sources and built new modules tried them and had no luck either.... I think I am going to scrap this board and get an MSI. I am done wasting my time just to get the same result!!!!!!!!!!

[krys]

crap, it doesnt look like anyone carries that MSI Live MB anymore, so now I dont know what I want to do, I cant believe that I cant get this damn thing to boot.

[colinjones]

You are making sure you don't get mixed up where you are putting the drivers and executing the commands aren't you?

Because the unit won't boot, typically you have to download/compile the driver to your core first (even modprobe it to make sure it loads and everything).

Then copying it to the /usr/pluto/diskless/# tree into the correct location, and using chroot to execute commands as if they were on the MD. Also, depmod is important before recreating the initramfs.

Finally, be conscious that there are 2 locations that the drivers need to be integrated into an initramfs type image. The default image and the actual boot image (both are detailed in the wiki articles that you are probably following already)

The default image is the initramfs image that is loaded the very first time a new MD boots and the core knows nothing about it. This image/boot is the one that coordinates detecting and setting up a new MD and creating the unique diskless image for it to boot from. On the next boot, it will boot from the actual boot image in this newly created diskless image.

So the question is, is it still on the first boot and failing before it finalises the setup for the second boot. Or is it on the second boot? This will tell you which initramfs to update/focus on. Obviously, you can only update the second boot initramfs image once the first boot is successfully completely because prior to that the diskless image that contains the second initramfs doesn't even exist!

[krys]

Collin,
all of that makes sense and seems to be what I have been doing for most of the day. I still have had no success, I am still on the first boot, I have not made it to the point where it announces itself to the router.

I also tried to created the image manually from the webadmin like the bottom part of the wiki for the r8169 and go that route, but had no luck either

[colinjones]

If you have edited the config files, depmod'd and recreated the initramfs after copying the driver in, and the initramfs in default is definitely updated (check the time/date stamp) then it sounds like the driver isn't working for that NIC.

Perhaps do a modinfo e1000 on the machine when booted as a core and note the details, like:

linuxmce@dcerouter:/lib/modules/2.6.22-14-generic$ modinfo r8168
filename:       /lib/modules/2.6.22-14-generic/kernel/drivers/net/r8168.ko
version:        8.008.00-NAPI
license:        GPL
description:    RealTek RTL-8168 Gigabit Ethernet driver
author:         Realtek and the Linux r8168 crew <netdev@vger.kernel.org>
srcversion:     2E2189C8B2FEE22CF89C4B1
alias:          pci:v000010ECd00008168sv*sd*bc*sc*i*
depends:
vermagic:       2.6.22-14-generic SMP mod_unload 586
parm:           speed:force phy operation. Deprecated by ethtool (. (array of int)
parm:           duplex:force phy operation. Deprecated by ethtool (. (array of int)
parm:           autoneg:force phy operation. Deprecated by ethtool (. (array of int)
parm:           rx_copybreak:Copy breakpoint for copy-only-tiny-frames (int)
parm:           use_dac:Enable PCI DAC. Unsafe on 32 bit PCI slot. (int)
parm:           debug:Debug verbosity level (0=none, ..., 16=all) (int)

[krys]

ok, here is the info you requested

Filename: /lib/modules/2.6.22-14-generic/kernel/drivers/net/e1000/e1000.ko
version: 7.3.20-k2-NAPI
license: GPL
description: Intel (R) PRO/1000 Network Driver
author: Intel Corporation, <linux.nics@intel.com>
srcversion: 6DB39DFE385628913292FF9
alias pci:v00008086d000010C5sv*sd*bc*sc*i*
(lots of alias here, didnt want to waste space unless you need to see them)
depends:
vermagic: 2.6.22-14-generic SMP mod_unload 586
(lots of parms, do you need to see those?)


As you stated lets just focus on one thing at a time, lets just try to get me to the first boot

[colinjones]

Now, on the new core, where you have compiled the new driver - modprobe it into the kernel, then modinfo it as well and compare the versions etc with above... esp the bits I highlighted in bold

[krys]

Now, on the new core, where you have compiled the new driver - modprobe it into the kernel, then modinfo it as well and compare the versions etc with above... esp the bits I highlighted in bold

I am not 100% sure how to do this.

** nevermind I think i did it,
modprobe e1000
modinfo e1000

[krys]

ok its different, but that is because i used the latest source as recommended by a few posts on google (after I tried the e1000 that came with LMCE and had no success)

version: 8.0.9-NAPI
srcversion: 0227095D6783D86D736D413
alias: pco:v00008086d00001085sv*sd*bc*sc*i*



on a side note 8.0.9 was the latest release from dell for this NIC that was supposed to incorporate e1000 and e1000e

[colinjones]

But the principle we are working on, is it not, is that the original driver worked perfectly with this machine when it was running as a core? If so, then get that driver and integrate it instead - then you will know if it is a driver issue or just the process you are using..

[krys]

how do I know for sure that this is the module it is trying to use when it boots as an MD? I know I added it to the list and recreated the initrd, but is there a way I can check to make sure i did that correctly? I will have to try and find that original module somewhere, I think I overwrote it with the new one. will post results

[krys]

ok, I loaded the e1000.ko module that came with LMCE, did the modprobe, modinfo, and it matched up with what was in use when this pc is used as a core. I made sure that e1000 was on the list and used Diskless_BuildDefaultImage.sh to recreate the initrd. Then I rebooted and I get the same Kernel Panic as before.

Am i missing something?

[krys]

holy shit, I loaded the e1000e driver and I just got to the point where it announced to the router....
gonna reload and see if it will build

** Great, now i get past the router announcement but I keep getting "Diskless Setup Failed"
**wait, I am idiot i forgot about step 2 cuz it has taken me so long to get here... brb