Scheduled Maintenance: We are aware of an issue with Google, AOL, and Yahoo services as email providers which are blocking new registrations. We are trying to fix the issue and we have several internal and external support tickets in process to resolve the issue. Please see: viewtopic.php?t=158230

 

 

 

Reboot fails: new 2.6.8 kernel-headers at fault??

If none of the specific sub-forums seem right for your thread, ask here.
Message
Author
zwerg
Posts: 117
Joined: 2006-02-17 12:16
Location: Bangkok, Thailand

Reboot fails: new 2.6.8 kernel-headers at fault??

#1 Post by zwerg »

Apt-get informed me today that it wanted to give me the latest kernel-headers for my Sarge setup, which uses kernel 2.6.8. The new stuff is "16sarge4", which is fine with me, as not so long ago "16sarge3" was provided. I cleaned up after that (putting /boot/grub/menu.lst back the way it belonged, with "noapic" and "nolapic" in the line "kernel /boot/vmlinuz.....") and things were just fine.

This time, however, the commands ...

shutdown -r now
shutdown -r +0
reboot

...do not work. Though the shutdown proceeds normally, the rebooting simply does not happen. (Command "shutdown -h now" works fine, BTW.)

Surely I'm not alone here.

Any ideas?

TIA

ajdlinux
Posts: 2452
Joined: 2006-04-23 09:37
Location: Port Macquarie, NSW, Australia

#2 Post by ajdlinux »

Why exactly is there an APIC problem?

Also were any other packages upgraded along with the headers?

zwerg
Posts: 117
Joined: 2006-02-17 12:16
Location: Bangkok, Thailand

#3 Post by zwerg »

Well, the APIC problem just may be unrelated and beside the point, but I mentioned it because you never, never know.

If I don't modify that line in /boot/grub/menu/list, the command dmesg | less returns rubbish. The fix is a niftly lillte trick I picked up here after I posted the problem.

And yes, there are some new things for Samba that came along with the new kernel-headers. My guess: irrelevant.

I hope that helps.

ajdlinux
Posts: 2452
Joined: 2006-04-23 09:37
Location: Port Macquarie, NSW, Australia

#4 Post by ajdlinux »

Have you tried without the APIC fixes? noapic could be breaking the reboot. Also are you sure it was only Samba and nothing else?

zwerg
Posts: 117
Joined: 2006-02-17 12:16
Location: Bangkok, Thailand

#5 Post by zwerg »

First, I appreciate your sticking with me on this, ajdlinux.

The answers to your good questions are Yes and Yes.

Following your suspicion about the APIC business, I not only removed the fixes, I also relocated them on the line (inside, instead of outside of, "ro"). No help: the problem is there no matter what happens to "noapic nolapic" in /boot/grub/menu.lst.

When I make a change in the boot menu, I (1) edit the file as root, (2) shut down, (3) boot, (4) start Xwindows as user, (5) log in as root in F1, (6) log out of the GUI, (7) try the command "shutdown -r now" as root in F1. I think that pretty well simulates normal use.

Next I checked /var/cache/apt/archives to see what has been done lately, and it's all kernel headers and samba. (I don't use Samba.)

I'm stumped.

ajdlinux
Posts: 2452
Joined: 2006-04-23 09:37
Location: Port Macquarie, NSW, Australia

#6 Post by ajdlinux »

Interestingly enough, yesterday I found that shutdown -h now isn't working very well for me. The services all stop, then I get a message 'System halted' and it doesn't turn off like it should.

If your kernel headers were upgraded, wouldn't that come along with an upgrade of the actual kernel itself?

zwerg
Posts: 117
Joined: 2006-02-17 12:16
Location: Bangkok, Thailand

#7 Post by zwerg »

Are you using an uninterruptible power supply? (We really, really need those here in Bangkok.) If so, that could be the cause of the failure to shut down. Look at the list of all my posts, and then focus on the one where I whine about how my box stopped shutting down.

I agree with you, it is a kernel upgrade. Not everyone sees it that way, as you will discover if you poke around on this forum. But it is still 2.6.8, not a new kernel, and it is upgraded.

I'd sure like to hear from some folks who (1) run Sarge with the 2.6.8 kernel , and (2) just did the same upgrade I did.

ajdlinux
Posts: 2452
Joined: 2006-04-23 09:37
Location: Port Macquarie, NSW, Australia

#8 Post by ajdlinux »

zwerg wrote:Are you using an uninterruptible power supply? (We really, really need those here in Bangkok.) If so, that could be the cause of the failure to shut down. Look at the list of all my posts, and then focus on the one where I whine about how my box stopped shutting down.
No, I don't have one, Australia's power suppliers are quite reliable.
I agree with you, it is a kernel upgrade. Not everyone sees it that way, as you will discover if you poke around on this forum. But it is still 2.6.8, not a new kernel, and it is upgraded.
So is it only headers or the kernel itself? i.e. is it kernel-headers-2.6.8 only or does it include kernel-image-2.6.8 as well?
I'd sure like to hear from some folks who (1) run Sarge with the 2.6.8 kernel , and (2) just did the same upgrade I did.
I suppose that's not me then! I run Etch with 2.6.15.

zwerg
Posts: 117
Joined: 2006-02-17 12:16
Location: Bangkok, Thailand

#9 Post by zwerg »

Aha! We may be onto something here. I see that kernel-image was in fact updated, along with kernel-headers.

The relevant entries in /var/cache/apt/archives are:

kernel-headers-2.6.8-3_2.6.8-16sarge4_i386.deb
kernel-headers-2.6.8-3_686_2.6.8-16sarge4_i386.deb
kernel-image-2.6.8-3-686_2.6.8-16sarge4_i386.deb

Can't say I understand this duplication of kernel-headers. Maybe you can help me grasp the concept?

A list of installed packages (dpkg -l | less) shows these installed:

kernel-image-2.6.8-16sarge4
kernel-headers-2.6.8-16sarge4

Old kernels that were removed don't show on this list, of course.

The problem with "reboot" did not exist when I was running 2.6.8-16sarge3.

I was under the impression that when you added a kernel, your /boot/grub/menu.lst put the new one at the top of the list and continued to show the older one below it, as an option. I recall seeing that when I went from 2.4.x to 2.6.8. Yet when I recently upgraded, that did not happen. I also thought that a 2.6.8 kernel was a 2.6.8 kernel, but evidently not always! Those misconceptions explain why I assumed that only the headers had been modified this time around. My mistake.

BTW I envy you that reliable power.....I have two UPS units.

User avatar
Absent Minded
Posts: 3464
Joined: 2006-07-09 08:50
Location: Washington State U.S.A.
Been thanked: 3 times

Re:

#10 Post by Absent Minded »

Sounds like you may have the equivalent of ACPI=off when you first installed sarge.
Sarge treated one of my boxes as if it were an older APM system rather than ACPI.
Which I might add can still be used with NOAPIC and NOLAPIC. Since Kernel 2.6 is modular you may still be able to install your acpi support and not have to re-compile your kernel.
When I install on that old box of mine I use

Code: Select all

expert26 acpi=on noapic nolapic
and everything works out fine. It is a VIA chipset btw.

Lavene
Site admin
Site admin
Posts: 4958
Joined: 2006-01-04 04:26
Location: Oslo, Norway

#11 Post by Lavene »

ajdlinux wrote:Interestingly enough, yesterday I found that shutdown -h now isn't working very well for me. The services all stop, then I get a message 'System halted' and it doesn't turn off like it should.
First time I'm reading through this thread. I just want to add that I have the same problem with my 2.6.17 kernels (the precompiled linux-image-2.6.17-1-686 and the stock 2.6.17.4) Every now and then the system don't shut down properly.

I've gone back to an older kernel because of that...

Just wanted to mention it.

Tina :)

Edit: Kernel version typo
Last edited by Lavene on 2006-07-19 15:40, edited 1 time in total.

zwerg
Posts: 117
Joined: 2006-02-17 12:16
Location: Bangkok, Thailand

#12 Post by zwerg »

Yikes, I think I'm in over my head here, but I'll try to carry on. Here is the history of the problem:

Installed Sarge with 2.4.x kernel some months ago. Moved to 2.6.8. Noticed that command dmesg did not work. Added "noapic nolapic" to get it to work again. Removed all kernels from grub except 2.6.8. Through all that, command "reboot" worked fine. I then made the most recent upgrade, detailed above, which appears to have messed up "reboot."

Given the above history, and the things I have tried, I'm not at all clear yet on whether "noapic nolapic" is responsible.

I have a file (edited) of the output of dmesg that shows every reference to APIC. It's available to anyone who wants a copy and thinks that might help to solve my curious little problem. Send me a private message and I'll e-mail you the file. TIA!!

ajdlinux
Posts: 2452
Joined: 2006-04-23 09:37
Location: Port Macquarie, NSW, Australia

#13 Post by ajdlinux »

It's probably ACPI, not APIC, that's causing your problem. (Yes, they are both four-letter acronyms, and they have the same four letters, but they are quite different...)

zwerg
Posts: 117
Joined: 2006-02-17 12:16
Location: Bangkok, Thailand

#14 Post by zwerg »

LOL!!

I recall when I was searching for answers to the problem I had with dmesg -- the response to the command was little more than "APIC error on CPU0: 40(40)" endlessly repeated -- I noticed that posters to this forum were constantly confusing "apic" and "acip." My thought at the time: "How silly. The two are very different!"

Now here I find myself making the same careless error! Well, learning a little humility can be an embarrassing lesson, but a valuable one. Thanks for getting my bowl of alphabet soup organized, adjlinux.

So, if we can get back on track here, what's the fix? How do I get the "reboot" command to work?

Moreover, how does acip pull this stunt, and why would this problem crop up now -- at a time that suggests it may well be linked somehow to the upgrades that were made to the kernel and to Samba?? I have not been messing with apic or acip. At least not AFAIK; if I did make changes in either, it must have been done telepathically, and without my conscious awareness. Poltergeists are not part of my universe, so there has to be a rational explanation for the sudden total failure of "reboot."

Tina suggests the problem could be in the kernel itself. I agree; I still can't see any apic/acip connection. Cripes, that's a dismal prospect.... Anybody have more info for us??

I repeat my request to hear from pure Sarge/2.6.8 users who accepted the same upgrade I did after entering "apt-get update" and "apt-get -u upgrade." Again I say, I can't be alone here.

Thanks to all the folks who have posted here so far....

ajdlinux
Posts: 2452
Joined: 2006-04-23 09:37
Location: Port Macquarie, NSW, Australia

#15 Post by ajdlinux »

BTW it's ACPI, not ACIP, and you did misspell my username :)

What's in the changelog between sarge3 and sarge4? I don't have this package so I can't check, and the PDO changelogs are down :( (The changelogs are kept in /usr/share/doc/kernel-image-2.6.8-3-686/changelog.Debian.gz usually.)

User avatar
Absent Minded
Posts: 3464
Joined: 2006-07-09 08:50
Location: Washington State U.S.A.
Been thanked: 3 times

Kernel Headers

#16 Post by Absent Minded »

he relevant entries in /var/cache/apt/archives are:

kernel-headers-2.6.8-3_2.6.8-16sarge4_i386.deb
kernel-headers-2.6.8-3_686_2.6.8-16sarge4_i386.deb
kernel-image-2.6.8-3-686_2.6.8-16sarge4_i386.deb

Can't say I understand this duplication of kernel-headers. Maybe you can help me grasp the concept?
Your headers arnt duplacated ... the first listed are for NON 686,586 (PIII / PII) Optimized systems.

zwerg
Posts: 117
Joined: 2006-02-17 12:16
Location: Bangkok, Thailand

#17 Post by zwerg »

Just checking, ajdlinux, to make sure you are paying attention.

Well, the changelog is small:

proc-environ-race-1.dpatch, proc-environ-race-2.dpatch

[SECURITY] Fix local root vulnerability by a race in proc. See CVE-2006-362.

--dann frazier <dannf@debian.org> Sat, 15 Jul 2006

ajdlinux
Posts: 2452
Joined: 2006-04-23 09:37
Location: Port Macquarie, NSW, Australia

#18 Post by ajdlinux »

So it was a security update. What modules have you got loaded?

Lavene
Site admin
Site admin
Posts: 4958
Joined: 2006-01-04 04:26
Location: Oslo, Norway

#19 Post by Lavene »

zwerg wrote:Tina suggests the problem could be in the kernel itself. I agree; I still can't see any apic/acip connection.
Well, I was wrong. Because just now my system failed to shut down properly while running my trusted old 2.6.15...

Tina

ajdlinux
Posts: 2452
Joined: 2006-04-23 09:37
Location: Port Macquarie, NSW, Australia

#20 Post by ajdlinux »

I find that every so often my system doesn't shutdown or reboot properly on 2.6.15, especially when KDM fails to shutdown because of some process that hasn't died yet, so I have to use the command line and then it doesn't work...

Here's the list of acpi related modules that I found, all of these are loaded or what?:
/lib/modules/2.6.15-1-k7/kernel/arch/i386/kernel/cpu/cpufreq/acpi-cpufreq.ko
/lib/modules/2.6.15-1-k7/kernel/drivers/acpi
/lib/modules/2.6.15-1-k7/kernel/drivers/acpi/asus_acpi.ko
/lib/modules/2.6.15-1-k7/kernel/drivers/acpi/ac.ko
/lib/modules/2.6.15-1-k7/kernel/drivers/acpi/container.ko
/lib/modules/2.6.15-1-k7/kernel/drivers/acpi/battery.ko
/lib/modules/2.6.15-1-k7/kernel/drivers/acpi/button.ko
/lib/modules/2.6.15-1-k7/kernel/drivers/acpi/hotkey.ko
/lib/modules/2.6.15-1-k7/kernel/drivers/acpi/fan.ko
/lib/modules/2.6.15-1-k7/kernel/drivers/acpi/processor.ko
/lib/modules/2.6.15-1-k7/kernel/drivers/acpi/ibm_acpi.ko
/lib/modules/2.6.15-1-k7/kernel/drivers/acpi/toshiba_acpi.ko
/lib/modules/2.6.15-1-k7/kernel/drivers/acpi/thermal.ko
/lib/modules/2.6.15-1-k7/kernel/drivers/acpi/video.ko
/lib/modules/2.6.15-1-k7/kernel/drivers/pci/hotplug/acpiphp_ibm.ko
/lib/modules/2.6.15-1-k7/kernel/drivers/pci/hotplug/acpiphp.ko

Post Reply