New installation with genkernel will not load kernel on boot

DONAHUE · Posted: Fri Sep 21, 2012 8:59 pm Post subject:

http://swift.siphos.be/linux_sea/ch07.html#idp35190256 is a very well written discussion of the process we have been following (In a very well written discussion of linux generally). What we have been doing is leading in the nasty direction of some (intermittent? random?) hardware failure. The logic being that the drivers are correct for the hdd controller bus, correct for the hdd motherboard adapter, correct for the file system, and grub was passing the correct root=partition. (Almost) leaving only hardware. gpt/uefi was not part of the reference discussion. While I can vouch for drivers sd, sg, sr, ahci, ata_pIIx, ext4; I have little experience with the efi/gpt drivers -- EFI_PARTITION, EFI, EFI_STUB, FB_EFI, EFI_VARS. These may easily have quirks that I know nothing about.
Intermittent random is justified by multiple cases of panic block-device(0,0), some panic block-device(0,68), one no panic but an oops; however reconfigures were also happening.
Trying again the no panic but an oops kernel may help support or debunk intermittent random.
I'm off to try a gpt install.

BTW

Jickler · n00b Joined: 09 Sep 2011 Posts: 72

Just got back, I'm compiling the "bad config" now. This would be the kernel where I anticipate not hitting a VFS kernel panic and getting a dmesg log out of. Maybe it'll boot if the file system's repaired, who's knows? (Magic Eight Ball says "Don't count on it")

I'll run e2fsck after I can umount this drive. It would be interesting if the filesystem has broken again..

I made clean, so this'll take a second. Seeing that I have time, I'll start burning sysresccd (I don't have a laptop down here, but I do have an external burner).

-EDIT:
e2fsck appears to be clean after reboot:

Jickler · n00b Joined: 09 Sep 2011 Posts: 72

VFS panic at 0,0

Ugh, I just don't understand. I'm pretty positive this is the exact config I used when I got my dmesg this morning. I can accept that the SATA drivers aren't working well here because, well, that's what this means. Then, on the other hand,THE Linux kernel can handle this hardware. This computer has ran both Ubuntu and Gentoo in the past -- clearly there exists a set of configuration options for at least two versions of the Linux kernel that reliably and consistently boots on this hardware.

I also tried on the sysresccd Boot From First Hard Disk, same result. Maybe I'll poke through the UEFI settings some more; maybe there's some bizarre "feature" or something..

DONAHUE · Posted: Fri Sep 21, 2012 10:28 pm Post subject:

e2fsck /dev/sde4?

the random intermittent nature now heading toward repeatable suggests hardware
suggest boot sysresccd to GUI and run smartmontools against sde drive (or smartctl in cli).
http://wiki.gentoo.org/wiki/Smartmontools

memtest if any other os (except windows

) has been flaky.

different kernel?
_________________
Defund the FCC.

Jickler · n00b Joined: 09 Sep 2011 Posts: 72

I just got it.

Vaguely remembered poking around in GRUB settings before I got my dmesg boot. Decided to do so again before starting hw tests. Noticed GRUB had configured root=/dev/sde4, which doesn't make sense bc the UUID root is in GRUB default. Deleted the UUID line and now I'm sitting at login. Will reset to reproduce, then compile seed config to see if this continues to check out. Must have been a typo or syntax error, will figure out when I'm back on liveCD.

DONAHUE · Posted: Fri Sep 21, 2012 10:47 pm Post subject:

ok. don't lose what you have.
my grub.cfg also defaults to root=/dev/sda2 not to a root=UUID=28c39827-2409-48a3-a17a-81b577db9681
mine may work because I run everything on a single partition

Jickler · n00b Joined: 09 Sep 2011 Posts: 72

Alrighty, I'm off my iphone and on my liveCD

It wasn't a typo in the actual UUID address at least, I have the UUID from blkid and the UUID from grub.cfg lined up and they match perfectly.
This is the actual line in grub.cfg:

DONAHUE · Posted: Fri Sep 21, 2012 11:11 pm Post subject:

4 more crossed fingers here. golden folder on a usb in your pocket?

my grub.cfg also defaults to root=/dev/sda2 not to a root=UUID=28c39827-2409-48a3-a17a-81b577db9681 as we are lead to expect by the grub2 advertisements
mine may work because I run everything on a single partition with only one disk making it hard for the kernel to get confused about which disk is which.
your gnulinux-simple-UUID is different from your --set=root UUID's

Jickler · n00b Joined: 09 Sep 2011 Posts: 72

Hey, sorry for the wait. Had to celebrate a bit

I backed it up on lvm#1, which isn't exactly external media, but then on the other hand if lvm#1 goes down then I wouldn't want to be that desktop computer or the nearest sledgehammer, so it won't really matter anyways. :twisted:

I took the seed config and spent some time adding some modules, checking over things, googling things I want and making sure the required kernel config for them matched up, etc. Don't think I changed anything that was actually compiled into the kernel at all.

Now when the computer boots it shows (none)@unknown_domain instead of root and after you log in everything's a read only filesystem. BUT...this is coming back to me. This is exactly what happened when I installed Gentoo the first time, and I figured it out. It had something to do with the ext4 filesystem breaking if I remember correctly. Last time I booted I e2fsck'd and it was indeed broken again. Upon repair I rebooted but the same thing happened. I just rebooted into a liveCD again and:

DONAHUE · Posted: Sat Sep 22, 2012 12:31 am Post subject:

Jickler · n00b Joined: 09 Sep 2011 Posts: 72

From within the running system?

DONAHUE · Posted: Sat Sep 22, 2012 12:39 am Post subject:

as root
may need

DONAHUE · Posted: Sat Sep 22, 2012 12:48 am Post subject:

Jickler · n00b Joined: 09 Sep 2011 Posts: 72

Okay, that command succeeds at making / RW. Domain name doesn't update, bash prompt reads (none) ~ # and this doesn't change after exiting and logging in again. Testing reboot now.

Reboot logged into unknown domain, RO again. Noticed kernel saw filesystem with errors and forced fsck while booting this time.

DONAHUE · Posted: Sat Sep 22, 2012 12:54 am Post subject:

crossed in the mail

you may be missing boot and sysinit services
_________________
Defund the FCC.

Jickler · n00b Joined: 09 Sep 2011 Posts: 72

After remounting and dhcpcd could do it: http://bpaste.net/show/46830/

Also noticed kernel boot sequence got a little more red at the end than I remember the first few times.

DONAHUE · Posted: Sat Sep 22, 2012 1:10 am Post subject:

Jickler · n00b Joined: 09 Sep 2011 Posts: 72

But they're not working? Haha, everything I read on /etc/issue is about boot logos, which is pretty cool, but not what I need. Mine looks like that: \n. \O...etc. How can we figure out why these services aren't working?

I dunno if this is any help but I have rc logging on, here's rc.log: http://bpaste.net/show/46832/

DONAHUE · Posted: Sat Sep 22, 2012 1:52 am Post subject:

fsck failure is blocking services

Jickler · n00b Joined: 09 Sep 2011 Posts: 72

GMTA, just finished setting up NTP.

Yup, yup! That fixed it

Jickler · n00b Joined: 09 Sep 2011 Posts: 72

Writing from Lynx --
Yup, this system feels good

I don't anticipate anymore super weird problems, everything seems to be running well. I'm sure the problems I run into henceforth will be the everyday noob ones that I can work through with the excellent documentation project. Just wanted to say thanks again, I'm sure this has been expressed many times over on this board but everyone who contributes to forum support is a freaking hero in my book <3 If one day I know enough about Gentoo to provide even a fraction of the excellent support that I've recieved in this thread, on these forums and on irc then doing so will make this hobby of mine sooo worth it

DONAHUE · Posted: Sat Sep 22, 2012 2:37 am Post subject:

buon viaggio

nvidia-drivers will require a minimal xorg.conf

nvidia-xconfig will provide it

nvidia-drivers will need an emerge for each kernel change

copy your .config to the new /usr/src/linux and use make oldconfig to adapt it
_________________
Defund the FCC.