Slackware 12.2 + RAID-1 + LVM + LUKS encrypted root

tmerriam · 12-18-2008, 08:19 PM

Quote:

You might have just left it off in the copy/paste or I may just be barking up the wrong tree, but shouldn't there be a -m before the modules list?

Yeah, that's just a typo in my post.

There is definitely a kernel bug during the install, it occurs at about where fontconfig runs, after the packages are installed. I could only salvage part of it, but it looks redundant, look at the Oops line near the bottom. The earlier ones are lower numbered and occur on alternating cores.

Code:

Pid: 19287, comm: sh Tainted: G      D   (2.6.27.7-smp #2)
EIP: 0060:[<c073c020>] EFLAGS: 00010046 CPU: 0
EIP is at _spin_lock_irqsave+0x10/0x30
EAX: 00000000 EBX: f7e04000 ECX: 00000246 EDX: 00000100
ESI: 00000006 EDI: 00000020 EBP: ecf83ef4 ESP: ecf83ef4
 DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0068
Process sh (pid: 19287, ti=ecf82000 task=f7ea8000 task.ti=ecf82000)
Stack: ecf83f08 c059ff47 c059fee0 f5d6f800 00000020 ecf83f48 c045df4e 00000000 
       f1eefb7c eb6a7000 00000000 00000006 00000000 00000000 b806e000 f5d6f800 
       b7f04590 eb6a701f f783bf80 c045de40 fffffffb ecf83f6c c01b7c2d ecf83f9c 
Call Trace:
 [<c059ff47>] ? ata_scsi_user_scan+0x67/0x130
 [<c059fee0>] ? ata_scsi_user_scan+0x0/0x130
 [<c045df4e>] ? proc_scsi_write+0x10e/0x230
 [<c045de40>] ? proc_scsi_write+0x0/0x230
 [<c01b7c2d>] ? proc_reg_write+0x5d/0x80
 [<c017fa8b>] ? vfs_write+0x9b/0x130
 [<c01b7bd0>] ? proc_reg_write+0x0/0x80
 [<c017ffd2>] ? sys_write+0x42/0x70
 [<c010325a>] ? syscall_call+0x7/0xb
 [<c0730000>] ? quirk_usb_early_handoff+0xb0/0x410
 =======================
Code: 7e 9c ff f0 fe 00 8b 45 04 e8 ad ea 9e ff 5d c3 8d 74 26 00 8d bc 27 00 00 00 00 55 89 e5 e8 f0 7d 9c ff 9c 59 fa ba 00 01 00 00 <f0> 66 0f c1 10 38 f2 74 06 f3 90 8a 10 eb f6 5d 89 c8 c3 8d b6 
EIP: [<c073c020>] _spin_lock_irqsave+0x10/0x30 SS:ESP 0068:ecf83ef4
---[ end trace 4a89a228f35c1749 ]---
BUG: unable to handle kernel NULL pointer dereference at 00000000
IP: [<c073c020>] _spin_lock_irqsave+0x10/0x30
*pde = 00000000 
Oops: 0002 [#16] SMP 
Modules linked in: ssb pcmcia pcmcia_core

Pid: 19309, comm: sh Tainted: G      D   (2.6.27.7-smp #2)
EIP: 0060:[<c073c020>] EFLAGS: 00010046 CPU: 0
EIP is at _spin_lock_irqsave+0x10/0x30
EAX: 00000000 EBX: f7e04000 ECX: 00000246 EDX: 00000100
ESI: 00000007 EDI: 00000020 EBP: eb5f1ef4 ESP: eb5f1ef4
 DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0068
Process sh (pid: 19309, ti=eb5f0000 task=f7ea9810 task.ti=eb5f0000)
Stack: eb5f1f08 c059ff47 c059fee0 f5d6f800 00000020 eb5f1f48 c045df4e 00000000 
       ecfddb7c ecf7a000 00000000 00000007 00000000 00000000 b806e000 f5d6f800 
       b7f04590 ecf7a01f f783bf80 c045de40 fffffffb eb5f1f6c c01b7c2d eb5f1f9c 
Call Trace:
 [<c059ff47>] ? ata_scsi_user_scan+0x67/0x130
 [<c059fee0>] ? ata_scsi_user_scan+0x0/0x130
 [<c045df4e>] ? proc_scsi_write+0x10e/0x230
 [<c045de40>] ? proc_scsi_write+0x0/0x230
 [<c01b7c2d>] ? proc_reg_write+0x5d/0x80
 [<c017fa8b>] ? vfs_write+0x9b/0x130
 [<c01b7bd0>] ? proc_reg_write+0x0/0x80
 [<c017ffd2>] ? sys_write+0x42/0x70
 [<c010325a>] ? syscall_call+0x7/0xb
 [<c0730000>] ? quirk_usb_early_handoff+0xb0/0x410
 =======================
Code: 7e 9c ff f0 fe 00 8b 45 04 e8 ad ea 9e ff 5d c3 8d 74 26 00 8d bc 27 00 00 00 00 55 89 e5 e8 f0 7d 9c ff 9c 59 fa ba 00 01 00 00 <f0> 66 0f c1 10 38 f2 74 06 f3 90 8a 10 eb f6 5d 89 c8 c3 8d b6 
EIP: [<c073c020>] _spin_lock_irqsave+0x10/0x30 SS:ESP 0068:eb5f1ef4
---[ end trace 4a89a228f35c1749 ]---

Sorry if this is starting to get a bit off topic....

mostlyharmless · 12-19-2008, 01:43 PM

Quote:

1 Which means that I most likely will get the very samy result.
2 Therefore I will have to redo lilo.
3 GOTO 1.

Seems to be a "dynamic halt"... (synonym for "infinite loop")

Well, I think I prefer the "Hello world" kind of program. Seriously, if you can't get lilo to cooperate, you might give grub in /extras a whirl. Nice thing about it is that you don't have to keep reinstalling it everytime you fiddle with the kernel and initrd, which I find makes debugging the situation easier. Not that I have anything against LILO, used it for years!

mostlyharmless · 12-19-2008, 03:03 PM

I've been wondering if this problem has anything to do with the /dev/root change that was mentioned in a number of other threads, particularly this one:

http://www.linuxquestions.org/questi...evroot-688189/

gargamel · 12-19-2008, 08:09 PM

I don't see an indication that the /dev/root thing is my way, but I really don't know.
My recent "experiments" didn't solve the problem, however, here are my observations:

(1) New install including formatting / and /boot, following the README_CRYPT.TXT word by
from the point where the LVM setup is complete, including creation of initrd and switching
to generic SMP kernel, liloconfig and lilo. Reboot.
==> In fact I was asked to enter a LUKS passphrase. However, my USB keyboard wouldn't work...

(2) So once again boot with the install DVD, add USB support modules to initrd (uhci-hcd,
ehci-hcd,...), lilo. Reboot.
==> Kernel panic, RAID not working.

However, I noticed something strange. After a complete install like in (1) including
formatting the first run of mkinitrd reports more than 9000 blocks written, while running it
again and again it only reports 6500 blocks. Even when I remove the initrd-tree completely,
mkinitrd would only write about 6500 blocks. But after a full install it writes 9000 plus
blocks, it reports, exactly one time.

Then I downloaded the latest version 1.40 of mkinitrd-command-generator.sh. With some of my
attempts this created just about what I already had, but in my recent attempts it doesn't
"see" the RAID anymore. Weird!

Still stuck...

gargamel

gargamel · 12-19-2008, 09:11 PM

SUCCESS!!!

The new mkinitrd 1.3.3 fixes the problem.

So it was really a software bug, not a PEBKAC.

gargamel

mostlyharmless · 12-21-2008, 08:20 PM

Congratulations and good to know it can work. Maybe I'll get brave enough to try it; though at this point 12.1 has everything I need and 12.2 nothing extra that I want..

gargamel · 12-22-2008, 10:48 AM

Thanks. Yes, 12.1 worked flawlessly here, once I had it up and running thanks to the great support
of the folks here. The most valuable part coming from Alien Bob --- thanks, Eric, once again!
Although he doesn't stop emphasizing that he is not here for holding hands, he and his mates in the
Slackware team (Robby Workman and some other people) are very helpful, just like Pat V. himself, too.
Indeed they didn't ever refuse to hold my hand, and I am very grateful for that!

As to 12.2, there's a lot of updates, I like to have, such as Mozilla Firefox 3.x (note that 2.0.20 is
supposed to be the last 2.0.x version of this browser, the 2.x branch will no longer be maintained).
But there some truth in the saying: "Never change a running system!"

Well, I did, and immediately had a reason to return to LQ.org...

So, thanks guys, for doing it again!

gargamel

gargamel · 12-27-2008, 01:56 PM

I have just written an initial draft for a mini-Howto on this subject. Please review it and help me to improve it!
It's available here:

http://www.iverbi.de/slackware/RAID1...2_2_Howto.html
http://www.iverbi.de/slackware/RAID1...12_2_Howto.pdf

Best regards and thanks for all your feedback in advance,

gargamel

rvo · 12-29-2008, 01:00 PM

Well done gargamel, I'll take a look, this is really useful :-)

gargamel · 12-29-2008, 01:45 PM

Quote:

Originally Posted by rvo

Well done gargamel, I'll take a look, this is really useful :-)

Good luck with this initial draft, and don't forget to
report back any errors or suggestions.

gargamel

tmerriam · 01-15-2009, 09:40 PM

Yeah, that looks really good. I have a few minor suggestions though.

1. From experience, the modules needed for USB keyboards are *NOT* added automatically.

2. This is more of a suggestion, but perhaps you would like to include a note about multiple initrds? This is especially relevant if the user is going to be compiling their own kernels. mkinitrd will overwrite the initrd each time it is run, so there should be a unique one for each kernel entry, ie, the initrd for the default slackware 12.2 kernel should be copied to initrd.gz-2.6.27.7-smp. This can also be set with the -o switch to mkinitrd.

gargamel · 01-16-2009, 04:40 PM

Thanks for your feedback!

1. Ok, I'll correct this. Although it once looked to me as if my USB keyboard was working although initrd did not provide modules to the kernel for it.

2. I understand your point, but the topic is already complex, and adding information on multiple initrd setups would make it even more complex and further reduce the number of potetntial readers...
Also I don't have time to try it out myself, at the moment. That's why I am a bit reluctant, here. What do others think?

gargamel

gargamel · 01-16-2009, 05:23 PM

Just uploaded the second draft with minor corrections, looking forward to your feedback!
BTW, I am a non-native English speaker, so any language corrections or improvements would be most welcome, too!

gargamel

EDIT: For your convenience I add the links, again.

http://www.iverbi.de/slackware/RAID1...2_2_Howto.html
http://www.iverbi.de/slackware/RAID1...12_2_Howto.pdf

tmerriam · 01-23-2009, 08:12 PM

'swap is not part of a RAID array...' Should be 'Swap...' (Capital S)

'At least, make a node into a tissue!' Should be 'At least, make a note on a tissue!'

'This command reads the configuration and do the same things as...' Should be 'This command reads the configuration and does the same things as...'

'But how, you may asked, do I...' Should be 'But how, you may ask, do I...'

Everything else looks great. Nice job.

gargamel · 01-24-2009, 10:46 AM

Added the corrections, thanks!!!

http://www.iverbi.de/slackware/RAID1...2_2_Howto.html
http://www.iverbi.de/slackware/RAID1...12_2_Howto.pdf

gargamel

EDIT:
For a future version I plan to clean up the HTML code, currently generated by export from OOo.
Once the content is stable, that is. PDF output from OOo looks better.