Latest udev causes kernel panics (something about libata)

Painless · 2011-01-19 22:17:31

dedanna1029 wrote:

I am so so glad I paid attention at the bug. Per Tobias Powalowski at the bug in comments:
Have you tried the following:
udevadm info --convert-db
and rebuild your initramfs by running mkinitcpio -p kernel26
(etc)

I tried this, laptop booted okay. But when I tried to rip a CD, it locked up solid I've since reverted to udev 164-3 and am ripping as I type.

dedanna1029 · 2011-01-20 05:13:49

Painless wrote:

dedanna1029 wrote:
I am so so glad I paid attention at the bug. Per Tobias Powalowski at the bug in comments:
Have you tried the following:
udevadm info --convert-db
and rebuild your initramfs by running mkinitcpio -p kernel26
(etc)
I tried this, laptop booted okay. But when I tried to rip a CD, it locked up solid I've since reverted to udev 164-3 and am ripping as I type.

Okay, I'm on a desktop, is all I can think would be the diff there. I've been burning dvds with Nero, the works, with no issues with that. I've rebooted several times both with my external usb hdds hotplugged and not, tried things every which way. You sure you followed those steps to the letter?

dedanna1029 · 2011-01-20 05:14:54

mechoid9 wrote:

@dedana1029: Am I right to assume that one would need to apply olive's patch before following the steps you took in your last post?

No, I just jumped in and did it. As I stated before, I am not using patches from here at the forum, I'm only following the bug and trying what's there that comes directly from the devels. However, who knows, others may need the patch. This thing is fickle. It's different for different people/machines.

Last edited by dedanna1029 (2011-01-20 05:28:49)

dedanna1029 · 2011-01-20 05:17:13

skunktrader wrote:

dedanna1029 wrote:
guidito73 wrote:
Sorry, I forgot to make # mkinitcpio -p kernel26. That did the trick.
I don't know much about these things, so please forgive me if I sound ignorant. Doesn't mkinitcpio run on boot anyway if it doesn't when udev and/or the kernel are installed? I don't understand the need to run this I guess manually.
You need to recreate the initrd image so that the version of udev included in the image is the "old" version instead of the "current" version that is in there now.
dedanna1029 wrote:
The next question I have, is do I need to be su or su - to run it? Thanks.
The presence of the "# " characters prefacing the command implies that the command needs to run as root or sudo

That wasn't the question. I was trying to distinguish if it should be run as just plain su, or as su -. Doesn't matter, I ran it as full root anyway in init 3, when in init 3 one is in the equivalent of su - anyway. Please re-read the question. I do not use sudo for my own reasons.

Last edited by dedanna1029 (2011-01-20 05:21:07)

dedanna1029 · 2011-01-20 05:22:18

For those that this didn't work for, I'm curious on this, as I've been able to do things both ways with different results. Replace Ctl+Alt+F1 with Ctl+Alt+F2, and post results, if any? Thanks.

beerad · 2011-01-20 14:03:40

dedanna1029 wrote:

No, I just jumped in and did it. As I stated before, I am not using patches from here at the forum, I'm only following the bug and trying what's there that comes directly from the devels. However, who knows, others may need the patch. This thing is fickle. It's different for different

What comes directly from the devels is summarized in post #41 of this thread:

rworkman wrote:

Kay Sievers (udev maintainer) says that this is definitely a kernel bug, even though the change to ata_id (in udev) is what triggered it.

No offense to Tobias Powalowski, but how do you figure that the Arch packager is more on top of udev than the udev maintainer? Remember that this has also affected Slackware and others, so this isn't an Arch specific bug..

Also, keep in mind that a kernel panic pretty much equals either a kernel bug or deranged hardware. udev is a userspace application, so it's database should not trigger a kernel panic. If rebuilding the database does fix the issue for you, it's only fixed the symptom and not the underlying kernel bug, so it's really no different than applying the patch.

beerad · 2011-01-20 14:23:46

This patch to libata-scsi.c has been running flawlessly for almost 2 days, so I really think it fixes the problem. It may evolve more, but hopefully something gets submitted officially soon. When it does, maybe someone should open a request to get that patch applied to 2.6.37 and put in testing. Then you guys who are waiting for an Arch kernel can try it. Or maybe somebody has already thought of this and is way ahead of me

Edited to add: It just got posted!: http://permalink.gmane.org/gmane.linux. … evel/16492

Last edited by beerad (2011-01-20 14:58:21)

dedanna1029 · 2011-01-20 15:44:50

beerad wrote:

What comes directly from the devels is summarized in post #41 of this thread:
rworkman wrote:
Kay Sievers (udev maintainer) says that this is definitely a kernel bug, even though the change to ata_id (in udev) is what triggered it.
No offense to Tobias Powalowski, but how do you figure that the Arch packager is more on top of udev than the udev maintainer? Remember that this has also affected Slackware and others, so this isn't an Arch specific bug..
Also, keep in mind that a kernel panic pretty much equals either a kernel bug or deranged hardware. udev is a userspace application, so it's database should not trigger a kernel panic. If rebuilding the database does fix the issue for you, it's only fixed the symptom and not the underlying kernel bug, so it's really no different than applying the patch.

Yes, I know, but I also got thinking on something that someone else said here in this thread:

beerad wrote:

Keep in mind that this is not a udev bug. It's a kernel bug that udev165 tickles.

With that in mind, yes, it probably only fixed the symptom, however, it was an easier "workaround" for me, than the patch. To be honest, not knowing a heck of a lot about these things, I did not know, mechanically, how to apply the patch. Is that same patch also listed at the bug so it can be officially used and fixed? I know there is one there, but don't know if it's the same one or not. Tobias is also a developer, at least that is how he's listed at the bug. The only one who's posted to the bug that I can tell, although, admittedly, I haven't gone through everyone's info at the bug:

Flyspray wrote:

Real Name    Tobias Powalowski
Jabber ID
Global Group    --->Developers
Project Group    None
Tasks opened    134
Assigned To    1922
Comments    2367

And yes, I'm already aware that this is an upstream bug, whether udev directly or the kernel; however, I have seen more local devels send patches and fixes to upstream than I can count. In many cases, they are extremely brilliant. At the end of the day, all I can do in a forum setting is post what works for me, and if it works for someone else, great! If not, I'm sorry.

Last edited by dedanna1029 (2011-01-20 15:51:03)

dedanna1029 · 2011-01-20 15:52:53

beerad wrote:

Edited to add: It just got posted!: http://permalink.gmane.org/gmane.linux. … evel/16492

Yep, I posted this to the bug back on Jan. 12th, when I saw it here.

beerad · 2011-01-20 16:29:47

dedanna1029 wrote:

beerad wrote:
Edited to add: It just got posted!: http://permalink.gmane.org/gmane.linux. … evel/16492
Yep, I posted this to the bug back on Jan. 12th, when I saw it here.

nononono, that should be a direct link to the post made an hour or so ago that contains the patch and related info that is being submitted to the upstream kernel. It was cc:'d to stable, too, so maybe the intent is to backport it to the stable series of kernels as well.

And my apologies if I misunderstood your intent. I thought you were looking for a fix, not a workaround.

dedanna1029 · 2011-01-20 20:27:33

beerad wrote:

nononono, that should be a direct link to the post made an hour or so ago that contains the patch and related info that is being submitted to the upstream kernel. It was cc:'d to stable, too, so maybe the intent is to backport it to the stable series of kernels as well.

Ah, okay, I'm sorry, I misread the link. Blame it on poor eyesight, really.

beerad wrote:

And my apologies if I misunderstood your intent. I thought you were looking for a fix, not a workaround.

I am, but a workaround is good for until there's a fix, yes? If it helps us to get into our machines, it's still progress! I just need instructions on how to apply the patches. The current one at the bug, I'd like to try, even though things are working for now. Thanks.

Last edited by dedanna1029 (2011-01-20 20:43:42)

firewalker · 2011-01-22 07:13:24

Count me in for those strange errors (panic and oops).

dedanna1029 · 2011-01-22 08:09:47

There's a patch that has been submitted to the bug, as of a bit ago:

Jonathan Liu (net147) wrote:

Here is a backport of the patch to Linux 2.6.32.
It should also apply cleanly to Linux 2.6.36.
libata-alignment-2.6.32.patch (1.6 KiB)

https://bugs.archlinux.org/task/22343#comment71155

The problem is, I do not know how to apply a .patch file. Is there anyone here who can give me some pointers? Thanks.

Last edited by dedanna1029 (2011-01-22 08:10:21)

Caspian · 2011-01-22 09:33:38

I'm having the same issues. Every second or third boot I'm "greeted" with kernel panic. Is a permanent solution coming soon?

The_infinite_loop · 2011-01-22 18:09:04

Same problem here.
Symptoms :
- Random kernel panics at boot.
- Kernel panic while mounting CD.
Arch fallback images with no problems.
So the patch posted by Jonathan Liu is a kernel patch. I haven't tried any of the solutions mentioned on the whole thread, I just hope that the patch really solves the problem :\. Otherwise I'll downgrade to udev 164...

DCL196 · 2011-01-23 00:34:45

I also have this problem just as I started using Arch a few weeeks ago but luckily It's tolerable for me, happening only once every 5 or so boots.

kronig · 2011-01-23 05:17:24

I had the same problem when I upgraded from kernel26 2.6.36.2-1 to 2.6.36.3-1. Kernel panics on boot, shouting "ata" related stuff on screen, and also the fallback image panics on boot sporadicaly. Now, downgraded to 2.6.36.2-1 and got my Arch running again with no casualties, even when poking the dvd-player on boot.

Didn't try downgrade udev -- using here the udev 165-1.
Didn't try any patches, in particular Olive's solution (which I couldn't find in the comments, anyway).

My system is an old Dell Inspiron E1405.

x33a · 2011-01-23 06:58:00

Well i finally downgraded to udev 164-3, and everything seems to be working fine after 3 successful boots.

hammer · 2011-01-24 10:58:05

I have read the whole thread and I am kind of happy that I am not the only one with such problems.This weird problems appear and I have to reboot 2 or 3 times everyday.Even FallBack doesn't work better.This is a photo I took from my mobile phone.What makes me concerned is that I dont always get the same output.Problems started since last kernel update(I dont remember if I have donwloaded Udev the same time).What do you recommend me to do?Install Daniel_F's UDev downgrade kernel or both?Here is the image I took--->
http://img228.imageshack.us/img228/8245/0031dd8.jpg

Last edited by hammer (2011-01-24 10:58:28)

phucnguyen · 2011-01-24 13:27:31

I attempted to downgrade udev but got this message

error: failed to prepare transaction (could not satisfy dependencies)
:: upower: requires udev>=165

It turned out there was only 1 version of upower (0.9.8-2). The package was required by gdm and gnome-session.

How do I get around this? Many thanks.

master176281 · 2011-01-24 15:27:28

You can use older packages from http://schlunix.org/archlinux/.

Just download, put in /var/cache/pacman/pkg and install with pacman -U *****.pkg.tar.*z

General information about downgrading: https://wiki.archlinux.org/index.php/Do … nc_Mirrors

Good Luck

darkbeanies · 2011-01-24 15:43:46

How to fix problem:

Download following package:

http://schlunix.org/archlinux/core/os/i … pkg.tar.xz

64 bit people: replace "i686" with "x86_64" without the quotes, ha ha...)

Gnome users-you need to downgrade upower, get it here: http://schlunix.org/archlinux/extra/os/ … pkg.tar.xz

Open terminal and navigate to wherever you downloaded the packages.

If you're downgrading upower , run this command first:

sudo pacman -U upower-0.9.7-2-i686.pkg.tar.xz

Then:

sudo pacman -U udev-164-3-i686.pkg.tar.xz

Then do:

sudo mkinitcpio -p kernel26

This seems to work (5 reboots without error so far...)

Last edited by darkbeanies (2011-01-24 15:48:57)

patrickaupperle · 2011-01-24 23:41:37

dedanna1029 wrote:

I am so so glad I paid attention at the bug. Per Tobias Powalowski at the bug in comments:
Have you tried the following:
udevadm info --convert-db
and rebuild your initramfs by running mkinitcpio -p kernel26
... I took it a step further.
I umounted all of my external usb drives, made sure none of my other hard drives were mounted, and made sure there was no cd or dvd in my dvd drive, then:
Logged out
Hit Ctl+Alt+F1
Logged into root
Then, did:
init 3
udevadm info --convert-db
mkinitcpio -p kernel26
Then, rebooted.
... it works. He is a god. I still have several more reboots to go before I can confirm it's a keeper, but I don't think that will be a problem.

Wow, this worked for me.

I can't believe it did, even downgrading udev didn't. Thank you for sharing this bit of knowledge.

Edit: Just to clarify, I upgraded back to the latest before applying this.
Edit 2: I have only rebooted 3 times since applying this, so I am not 100% sure it worked. I was kernel panicking 2/3 reboots before, so this is awfully coincidental if nothing is fixed.

Last edited by patrickaupperle (2011-01-24 23:45:32)

net147 · 2011-01-25 08:41:38

Patched kernel26 2.6.36.3-2 packages for testing here: http://dev.archlinux.org/~tpowa/kernel26/
Reference: http://mailman.archlinux.org/pipermail/ … 18180.html

firewalker · 2011-01-25 10:36:45

With the pached version everything seems to be just fine for me. 3 days without any problem.

Arch Linux

#151 2011-01-19 22:17:31

Re: Latest udev causes kernel panics (something about libata)

#152 2011-01-20 05:13:49

Re: Latest udev causes kernel panics (something about libata)

#153 2011-01-20 05:14:54

Re: Latest udev causes kernel panics (something about libata)

#154 2011-01-20 05:17:13

Re: Latest udev causes kernel panics (something about libata)

#155 2011-01-20 05:22:18

Re: Latest udev causes kernel panics (something about libata)

#156 2011-01-20 14:03:40

Re: Latest udev causes kernel panics (something about libata)

#157 2011-01-20 14:23:46

Re: Latest udev causes kernel panics (something about libata)

#158 2011-01-20 15:44:50

Re: Latest udev causes kernel panics (something about libata)

#159 2011-01-20 15:52:53

Re: Latest udev causes kernel panics (something about libata)

#160 2011-01-20 16:29:47

Re: Latest udev causes kernel panics (something about libata)

#161 2011-01-20 20:27:33

Re: Latest udev causes kernel panics (something about libata)

#162 2011-01-22 07:13:24

Re: Latest udev causes kernel panics (something about libata)

#163 2011-01-22 08:09:47

Re: Latest udev causes kernel panics (something about libata)

#164 2011-01-22 09:33:38

Re: Latest udev causes kernel panics (something about libata)

#165 2011-01-22 18:09:04

Re: Latest udev causes kernel panics (something about libata)

#166 2011-01-23 00:34:45

Re: Latest udev causes kernel panics (something about libata)

#167 2011-01-23 05:17:24

Re: Latest udev causes kernel panics (something about libata)

#168 2011-01-23 06:58:00

Re: Latest udev causes kernel panics (something about libata)

#169 2011-01-24 10:58:05

Re: Latest udev causes kernel panics (something about libata)

#170 2011-01-24 13:27:31

Re: Latest udev causes kernel panics (something about libata)

#171 2011-01-24 15:27:28

Re: Latest udev causes kernel panics (something about libata)

#172 2011-01-24 15:43:46

Re: Latest udev causes kernel panics (something about libata)

#173 2011-01-24 23:41:37

Re: Latest udev causes kernel panics (something about libata)

#174 2011-01-25 08:41:38

Re: Latest udev causes kernel panics (something about libata)

#175 2011-01-25 10:36:45

Re: Latest udev causes kernel panics (something about libata)

Board footer