kernel panic on 2.6.39-gentoo-r3 [SOLVED]

dvh · here is the new entry for 2.6.39:

just updated from 2.6.36-gentoo-r5. I'm seeing a kernel panic during boot, because the process cannot find my root device. here is the grub menu.lst entry from my working 2.6.36 kernel:

Jaglover · Posted: Sat Aug 06, 2011 2:36 am Post subject:

Having the exact error message would help.

I suspect you used make oldconfig and that borked your .config.
_________________
My Gentoo installation notes.
Please learn how to denote units correctly!

dvh · Posted: Sat Aug 06, 2011 2:39 am Post subject:

i'll be happy to try to get it, but I don't know how to capture the output at that point in the process. how can i capture that?

I do recall this part:

dvh · Posted: Sat Aug 06, 2011 2:40 am Post subject:

indeed. i did use make oldconfig. is that no longer allowed?

Jaglover · Posted: Sat Aug 06, 2011 2:42 am Post subject:

VFS: Cannot open root device "sdb1" or unknown-block(x,x)

Well, that (x,x) thingy would give the clue, (0,0) would mean your HDD controller has no correct driver.
_________________
My Gentoo installation notes.
Please learn how to denote units correctly!

dvh · Posted: Sat Aug 06, 2011 2:43 am Post subject:

it is not 0,0. it is either 2,0 or 0,2. i cannot recall. i will need to reboot again to find out. when I do, is there a way to capture the text?

Jaglover · Posted: Sat Aug 06, 2011 2:44 am Post subject:

dvh · Posted: Sat Aug 06, 2011 2:46 am Post subject:

oh. ok. been doing it that way for years without an issue. I am happy to learn...I know there must be a better way than manually creating another .config. can you help me with a reference to the correct way?

Jaglover · Posted: Sat Aug 06, 2011 2:47 am Post subject:

dvh · Posted: Sat Aug 06, 2011 2:50 am Post subject:

i will investigate. as I said, it worked fine for 2.6.36...so I know the .config was OK then. Seems odd that "make oldconfig' would switch something to module from built in, though.

Jaglover · Posted: Sat Aug 06, 2011 2:57 am Post subject:

Not exactly, make oldcofig cannot cope if something is moved in kernel config tree as it may happen between versions. It won't change built-in to module.
_________________
My Gentoo installation notes.
Please learn how to denote units correctly!

dvh · Posted: Sat Aug 06, 2011 4:28 pm Post subject:

so, following the directions for bringing forward an old .config (from the gentoo link provided above), I used "make menuconfig" to create my new .config. unfortunately, same result. still getting the kernel panic at boot time. here is more info from the console output (retyping, so this may not be perfect).

the process seems to go through the discovery of USB devices, including finding a mouse and keyboard, and finally a "new full speed USB device number 3". then, it says

Hu · Administrator Joined: 06 Mar 2007 Posts: 23082

If the root device cannot be found as requested, the kernel may try other ways to get a root device before it gives up. This is probably why it tried to use a floppy.

Have you tried running the rootdelay value up to some large number, such as a minute? If yes, does the kernel wait that full minute before it panics?

NeddySeagoon · Posted: Sat Aug 06, 2011 6:33 pm Post subject:

dvh,

Aroud 2.6.26 there were some new menu options introduced to the SATA menu of device drivers. If you turned those off, you may well have removed your SATA driver from the kernel.

What does lspci say about your hardware?
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

dvh · Posted: Sat Aug 06, 2011 9:36 pm Post subject:

Hu: I upped the rootdelay to 100, and indeed, the kernel waited for the 100 seconds before panic.

Neddy: 2.6.36-r5 works fine for me, so I don't think that I messed up the SATA drivers in the kernel. also, my kernel is resident on an internal drive along with grub, but my root directory is on an external USB drive. again, this works fine with the earlier kernel, but panics here. seems clear that something is amiss with my kernel config, but I cannot figure out what. here is the lspci output:

dvh · Posted: Sat Aug 06, 2011 11:20 pm Post subject:

just in case this helps, here are some of the boot sequence messages from the working 2.6.36 kernel boot:

NeddySeagoon · Posted: Sun Aug 07, 2011 12:30 am Post subject:

dvh,

I missed your root on USB.

Please put your /usr/src/linix/.config file on a pastebin. wgetpaste is your friend.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

dvh

after looking at the boot messaging closer, it appeared that this might have something to do with IEEE-1394 drivers, for which startup messages are in the same vicinity of the log. so I did a "diff" between the old .config from 2.6.36 and the new config from 2.6.39. of course, i had to fight through a lot of legitimate differences, but I found and interesting difference...IEEE-1394 drivers present [*] in the old .config, but NOT in the new .config. so I used menuconfig and put them [*] into the new .config. and to my surprise...IT NOW BOOTS just fine!

this is stange, I thought. I have two of these external drives, and ONE is connected via firewire. I had forgotten all about this, but after the boot worked, I looked at the hookup and of course, the external drive that I needed is indeed firewire, for which the drivers were missing.

not sure why they were missing, as I updated using the recommended menuconfig method (in addition to the make oldconfig method) and they both failed to bring these options forward. is this normal?

in any case, I am a little embarassed about not knowing my HW setup like I thought I did, but at least I have solved the immediate problem. anyone have a better recommendation for upgrading a kernel to avoid such dropouts in the future?

NeddySeagoon · Posted: Sun Aug 07, 2011 12:24 pm Post subject:

dvh,

If things move in the kernel tree, or new menus are added on the path to the drivers you need, make oldconfig does not always get it right.
The further you jump between kernel versions, the bigger the risk.

After make oldconfig, its always good to check the basics with make menuconfig. I do remember a few horrors myself.
One that stands out in my mind is a kernel with no PCI support. I swear it wasn't me too.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

dvh · Posted: Tue Aug 09, 2011 1:21 am Post subject:

thanks, Neddy. Live and learn, I guess. never been bitten by this one before, but I will certainly be aware of it now...

-dvh

paziu · Tux's lil' helper Joined: 24 Nov 2006 Posts: 78

after copying a good .config to the new tree...
make menuconfig & exit / save - even better if you check and verify ... and see what's (NEW)...

oldconfig might be good if you go from -r2 to -r3, but you made quite a big jump...