Latest udev causes kernel panics (something about libata)

Hide · 2011-01-10 10:32:21

(disregard this)

Last edited by Hide (2011-01-10 10:37:03)

archman-cro · 2011-01-10 10:52:04

Update 2: The system seems to boot fine without having BFS. I use udev 164-3.
4 successful boots so far.

zebulon · 2011-01-10 11:05:46

Daniel_F wrote:

@arctos
arctos wrote:
I've got 2 kernel panics and 1 random crash.
Is it related to the bug we are dealing with in this thead, or you just felt like sharing ??
Because if it is related then your post is missing a lot of information to be helpful or even relevant...
Sorry for being rude, but it would be great if people could think before posting !!

Daniel, I think arctos was answering my question earlier. I recently have had many complete freezes and asked if this could be related to this bug, and if the panic that is described in this post only occurs during boot. I have reverted to udev-164-3, but still experience them, so I wonder if the problem is with kernel 2.6.37 too. The freezes occur at random times, but are more frequent when using mplayer (changing fullscreen) or watching youtube in Firefox.

rabid_works · 2011-01-10 11:31:46

Daniel_F wrote:

BTW: My patched UDEV had about a dozen downloads so far and only one feedback (thanks Deadguy1 !). Come on people: if it works for you (or if it doesn't !) then at least let us know !

Started having freezes at boot time again yesterday (after a 2-3 day hiatus - I reboot once a day on average). Downloaded Daniel_F's patched udev and rebooted 4 times. I had another freeze at boot time on the fourth reboot. I have, once again, reverted to udev-164-3. I didn't run mkinitcpio after the upgrade. I will try the upgrade again with mkinitcpio and report back on results.

EDIT : running mkinitcpio after installing the patched udev seems to have made a difference. I rebooted at least 7 times and didn't see any freezes (I usually experience a freeze every 3 - 4 boots). It's obviously still too early to say for sure but, for the moment, the problem seems to be resolved. I'll re-update should that change. In the meantime, thanks very much for the patched udev.

Last edited by rabid_works (2011-01-10 11:48:50)

Gullible Jones · 2011-01-10 13:42:27

Daniel_F wrote:

@Gullible Jones
Gullible Jones wrote:
...Ubuntu has come out with a kernel and a new version of udev...
Would you mind posting some references to that ?

This kernel update appears to fix some panics. Not sure if any of those are relevant, but the Ubuntu machine I have on hand hasn't crashed (yet!) since the update was applied.

nnoell · 2011-01-10 14:53:37

Hi, I downloaded Daniel_F's patched udev and also ran mkinitcpio after the patch installation yesterday. I rebooted many times and burned some cds withount any problem, it seems everything works fine. Thank you very much for patched udev Daniel_F.

jkralj1 · 2011-01-10 15:31:58

I also ran into the same problem

I solved it in the following manner (I have an older computer from 2005):
in /etc/mkinitcpio.conf
under the HOOKS
instead of "pata" i used "ide"

Attention!
/dev/sda becomes /dev/hda etc...
Update your /etc/fstab if you do not use UUID

Keep at least one backup kernel that works,
because if this does not work you can not boot at all!!!!

Last edited by jkralj1 (2011-01-10 15:58:41)

Gullible Jones · 2011-01-10 16:43:16

That unfortunately won't work for people with SATA drives. Also the IDE drivers are deprecated IIRC.

FWIW, this issue was mentioned on the linux-hotplug mailing list four days ago: http://comments.gmane.org/gmane.linux.h … evel/16438

Alas it appears that there's no official fix yet.

beerad · 2011-01-10 17:43:22

@Gullible, thanks for the link to thread on the linux-hotplug list. It may take a while to get fixed, but the kernel developers are aware of it..

@Zebulon, it sounds to me like your problem is different. This particular panic seems to be due to the PATA driver referencing VM that's not properly set up, and I don't see how that code path would be executed while watching online video. Just a guess on my part, though, I'm certainly not an expert.

matejo · 2011-01-10 17:51:00

Changing pata to ide worked for me - 5 restarts and 5 successful boots instead of kernel panic 6 times in a row.
Even though you can have a SATA harddrive, your cd-rom can be connected by ATA.

Gullible Jones · 2011-01-10 19:23:48

Hey, maybe a stupid idea, but... Is it possible to completely disable udev on Arch, and use a static /dev? That would probably kill most forms of automounting, but for those of us who don't use automounting it might be a viable and hopefully not-too-painful workaround.

Last edited by Gullible Jones (2011-01-10 19:24:06)

Kilzool · 2011-01-10 20:02:26

I did: > sudo pacman -Syu and got the new kernel (2.6.36.3-2) and headers the other day.
I then did > sudo pacman -S udev 165-1 and reinstalled that.

I have not had a kernel panic since. Will be back with more info if I do.

QUESTION: Did we finally resolve this? What was the issue exactly or close to?

Correction: the above comments did not solve the issue...after all. I began having kernel panics upon boot up again.. and like everyone else,
using the "fallback" was fine.

I got tired of this, and waited patiently for someone to release, "in an Arch" way, a fix.
I used one here (on 1-12-2011), posted by Olive.

Thank you Olive.

Last edited by Kilzool (2011-01-12 22:50:48)

archman-cro · 2011-01-10 20:17:09

Gullible Jones wrote:

Hey, maybe a stupid idea, but... Is it possible to completely disable udev on Arch, and use a static /dev? That would probably kill most forms of automounting, but for those of us who don't use automounting it might be a viable and hopefully not-too-painful workaround.

Hmm, I don't get any problems with udev 164-3. Have you tried that one?

Daniel_F · 2011-01-10 20:28:40

Why are you guys still looking at messy solutions such as resorting to deprecated legacy code or disabling udev altogether when you have a much cleaner one in the form of patching the current udev ???? This is the simplest solution an the one that has less, if any, side effects and also will ensure that as soon as a better upstream solution is released you will get it through pacman. Don't trust the binary package ? Then do it yourself, just as beerad did. Arch makes it actually pretty easy with ABS and makepkg.

I may be fairly new here, but I've been using (and abusing) linux daily since 1998, back when it was installed using floppy disks (a lot of them !), so I know my way around it...

And again, if you are using the patched udev, binary or self-made, please come forward and say it so people coming here searching for a solution (this thread has had more than 2000 views in the last couple of days) can be assured it works.

@Gullible Jones: Is Ubuntu runnung udev 165 ??? Because if it's not then I don't see how it could be having this issue (btw: the changelog you posted is from a month ago and has "lucid-proposed; urgency=low" on it's first line)

@Kilzool: Read the thread ! The answer to your question is in post #41

archman-cro · 2011-01-10 20:56:49

Daniel_F wrote:

Why are you guys still looking at messy solutions such as resorting to deprecated legacy code...

Is udev 164-3 messy and a deprecated legacy code?

Daniel_F · 2011-01-10 21:34:46

archman-cro wrote:

Is udev 164-3 messy and a deprecated legacy code?

No, but ide is...

Keeping udev 164 works fine too, but to do that and keep your system up-to-date using pacman you need to add it to IgnorePkg in pacman.conf, which is less than ideal.

Now, you got a solution that consists in running just one command and be done with it while still running the most recent version. Explain to me why that is not good enough and I'll remove it and stop posting.

cgarcia · 2011-01-10 22:16:24

In the last month or so udev has been upgraded 4 times. I wonder if this package stays in the testing repository (no udev is now in testing) a few days before going into "production". I still remember the day bluez was upgraded and segfaulted at boot in all systems (I use a bluetooth mouse). Yes, it was a upstream bug, but shouldn't we leave all core packages in testing a few days?. I mean, the computers using the testing repository may not find all the bugs, but in some cases may help.

I just have downgraded to 164-3 but it's too soon to report success. My machine was upgraded to 165-1 on january 4 and I detected the first kernel panics at boot yesterday (january 9) and they were several. I did not have any problem between january 5 and january 8. My first panic was interestingly after removing/inserting a just blanked DVD-RW disk (I don't burn a disk in months). The rest of DVD work was fine, and only *later* boots have ended in panic, including today's first boot a few minutes ago.

Edit to point: I still have 165-1 in another machine (Eee PC) with not a single crash in the past days.

Last edited by cgarcia (2011-01-10 22:21:25)

Gullible Jones · 2011-01-10 23:07:55

Daniel_F: I think less experienced Linux users (myself included) tend to balk at the idea of patching a compiled base system component, or using an unofficial patched version of one. Also, it's a temporary fix either way, since (as far as I understand it) this is a kernel bug that udev just happens to trigger. Thanks all the same, though.

(I ended up downgrading. FWIW, the latest kernel in [core] doesn't fix the bug.)

Re the Ubuntu thing... Yeah, I was wrong on that, it looks like the resident Ubuntu box panicked for some other reason. My bad.

archman-cro · 2011-01-10 23:22:42

I've booted just fine with 2.6.37 and 165-1 now. I'll be reporting about crashes if they happen.

Daniel_F · 2011-01-11 00:24:08

As far as kernels go, I've looked at all the commits in both 2.6.36.3 and 2.6.37 and none are related to this. So no, they won't fix it.

References:

https://lwn.net/Articles/422147/

and

https://lwn.net/Articles/421637/

Keeping udev at 164 or using a patched 165 are the only solutions for now, and since:

a.) both those kernels and udev 165 were just released

and

b.) this is not yet a widespread problem, because it only affects early-adopters like Arch or Slackware-current users (in fact, just a fraction of those)

an official solution will probably take a while. Unless Tobias (tpowa - Arch's udev mantainer) comes up with an Arch's official patched 165 (165-2 ?)

@cgarcia: AFAIK, an eeePC has no ata hardware. That's probably why it is not experiencing this problem.

As for the [testing] policies, if any dev could chime in and comment on that it would be, at least, instructive !

Last edited by Daniel_F (2011-01-11 00:31:31)

Mektub · 2011-01-11 00:51:19

I have been doing several reboots due to some testing, and I only started having panics while booting after
upgrading the kernel:

[2011-01-11 00:29] upgraded kernel26 (2.6.36.2-1 -> 2.6.36.3-1)
[2011-01-11 00:29] upgraded kernel26-headers (2.6.36.2-1 -> 2.6.36.3-1)
[2011-01-11 00:29] upgraded xorg-xdm (1.1.10-1 -> 1.1.10-2)
[2011-01-11 00:29] upgraded qtfm (4.1-1 -> 4.2-1)

udev had been upgraded on the 2001-01-04, no panics until the kernel upgrade:

[2011-01-04 12:49] upgraded udev (164-3 -> 165-1)

So, now I did install the patched udev from Daniel_F.

[2011-01-11 00:39] Running 'pacman -U http://www.electron2k.com/udev165arch/udev-165-1-i686.pkg.tar.xz'
[2011-01-11 00:40] upgraded udev (165-1 -> 165-1)

Mektub

olive · 2011-01-11 09:32:41

I have experienced the same problem. Kernel crash at boot and also a kernel panic when trying to read an audio CD with mplayer cdda://. I have downgraded udev to 164-3 but I just have the crash at boot. The phenomenon is intermittent and not always reproducible.

UPDATE: I did not rebuild the initrd after downgrading udev. This may explain my latest crash (the initrd contained the bad udev). I will see what happen now

Last edited by olive (2011-01-11 11:31:16)

Perre · 2011-01-11 15:52:02

Daniel_Fs patched udev fixed it for me.

Thanks Daniel!!

eded · 2011-01-11 15:52:04

udev patched from Daniel_F works fine for me...

thanks Daniel

olive · 2011-01-11 16:14:03

@Daniel_F. You provided a link with a patched udev ( http://www.electron2k.com/udev165arch/u … pkg.tar.xz ). Could you also post the file required to regenerate the package (PKGBUILD, original source, patch, etc...). I usually do not install precompiled packages if I cannot have the files to regenerate it (providing the source is also required by the GPL).

Arch Linux

#51 2011-01-10 10:32:21

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

#52 2011-01-10 10:52:04

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

#53 2011-01-10 11:05:46

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

#54 2011-01-10 11:31:46

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

#55 2011-01-10 13:42:27

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

#56 2011-01-10 14:53:37

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

#57 2011-01-10 15:31:58

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

#58 2011-01-10 16:43:16

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

#59 2011-01-10 17:43:22

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

#60 2011-01-10 17:51:00

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

#61 2011-01-10 19:23:48

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

#62 2011-01-10 20:02:26

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

#63 2011-01-10 20:17:09

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

#64 2011-01-10 20:28:40

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

#65 2011-01-10 20:56:49

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

#66 2011-01-10 21:34:46

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

#67 2011-01-10 22:16:24

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

#68 2011-01-10 23:07:55

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

#69 2011-01-10 23:22:42

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

#70 2011-01-11 00:24:08

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

#71 2011-01-11 00:51:19

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

#72 2011-01-11 09:32:41

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

#73 2011-01-11 15:52:02

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

#74 2011-01-11 15:52:04

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

#75 2011-01-11 16:14:03

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

Board footer