keyboard leds flashing == kernel panic

Anything in /var/log/* ?

Have you tried a different kernel and/or the latest distro updates?

1 guessed so ... this is why I also left a bug report on my distro bug 
reporting system, since no application should send the kernel into panic.

Could not find anything significant neither in messages nor in syslog.  
Any clue at where to look exactly?

I am at the latest kernel update of my distro, which is a 2.6.32 that 
should be based on the latest 2.6.32 maintainance kernel.

In any case no other application has never given a hiccup to that 
machine with that kernel. Furthermore, it should not be the kernel, 
since other machines where chromium is fine use just the same. Tried on

DELL laptop E6500 - Intel Core II 2 cores - Intel graphics - Ubuntu 
lucid 64bit  2.6.34-24

Chrome 5.x OK ; Chrome 6.x OK

DELLworkstation - Intel Xeon 4 cores - Nvidia graphics - Ubuntu lucid 
64bit 2.6.34-24 - Nvidia drivers 256.x

Chrome 5.x OK ; Chrome 6.x OK

EEEPC - Atom 270 Hyperthreading - Intel graphics - Ubuntu lucid 32bit 
2.6.34-24

Chrome 5.x OK ; Chrome 6.x OK

Workstation - ASRock chipset - AMD Phenom II 4 cores - Nvidia graphics - 
Ubuntu lucid 64bit 2.6.34-24 - Nvidia drivers 256.x or previous

Chrome 5.x OK; Chrome 6.x LOCKUP; Chrome 7.x LOCKUP; Chromium 7.x LOCKUP

The lockup always occurs when one _exits_ chrome/chromium.

Can there be some instruction that is illegal for the Phenom?

Sergio

A non-root process shouldn't be able to crash the kernel. If it can, it's a kernel bug. The only part of chrome that runs as root is the sandbox - it's unlikely to be involved but you could disable the sandbox (rename /opt/google/chrome/chrome-sandbox temporarily for example) to see if that makes a difference. Illegal instructions should just result in the termination of the process btw.

The other possibility is that the X server (which presumably still runs as root on ubuntu, dunno) is stomping over something causing the kernel to crash.

Ultimately you need to see why the kernel is crashing which means we need to see the console output if it's not being logged. There are ways of getting debug info (serial console for example) but I don't think this is necessarily the right forum for trying to debug that as this doesn't look like chrome bug to me (even if it didn't happen with earlier versions of chrome). Googling about how to resolve kernel panics should yield some clues.

I presume you are using binary only nvidia drivers btw. that tend to make people unhappy as it complicates debugging.

Also experiencing this:

System:
Hardware>
AMD Phenom II 550 BE ('quadded')
Ati HD4200 IGP

OS:
Mepis 8.5 32 bit upgraded + latest fglrx 10.8 (8.762) driver installed today
Ubuntu 10.04 32 bit upgraded + latest fglrx 10.8 (8.762)

Chrome:
6.0.472.55-r58392

Same issue on both systems - closing Chrome results in kernel panic.

Tried:
-Removing all extensions
-Disabling compositing
-removing any 'other' version of Flash installed

Issue still occurs.

This has been happening since around last major Flash security update - upgrading Flash then caused similar issue in Chrome. Resolved by removing Flash package and using Chromes version.

Now system kernel panics anyway...

/var/log/ files reveal no clues...

Not tried:
-using Radeon/Vesa instead - crap.
-disabling Chromes internal Flash (is it even possible)

Will use Firefox for now...

OK - seems it's sandbox causing this issue;

Running command:

google-chrome -no-sandbox

allows closure of chrome without kernel panic!

Interestingly the above command also solved the issue of chrome not functioning correctly in an LTSP environment...pesky sandbox!!

Hmmm ... interesting....I didn't expect that would help ... maybe it's tickling a kernel bug. There is another experimental sandbox you can try ... --enable-seccomp-sandbox .... about:sandbox might tell you if it is running ok IIRC. The other really strange thing is why it would trigger on closing chrome...the sandbox isn't doing anything then AFAIK. Dazed and confused .....

Sergio: does --no-sandbox work for you too?

Hi, I am away at a conference., so I cannot test right now... since on 
the atom box I have with me chrome does not crash the machine. But I 
will test as soon as I am back, on Saturday at worst.  In the meantime I 
am taking notice of the experiments to do (no sandbox and experimental 
sandbox).

Sergio

Sergio

If disabling the sandbox stops the kernel from crashing, then you should file a bug with your distro and figure out why the kernel is crashing. All we can do is mark this bug as "ExternalDependecy".

FYI, using binary blobs taint the kernel. I'd give the open source drivers a try just to see if that helps.

google-chrome --enable-seccomp-sandbox does nothing - chrome doesn't start.

I am aware of kernel tainting, but unfortunately the Xorg Ati driver (Radeon) fails on my system (with the debian Lenny based Mepis install at least - don't think it was any better with Ubuntu 10.04..) not worth the effort+performance loss.

I would guess that this is likely caused by the latest Catalyst (fglrx) driver as recent upgrades have rendered compositing unusable on my system now (X server locks up). 

As to the 'coincidence' that the OP has the same CPU as me (well similar) and suffers the same issue but with Nvidia - who knows?

And why disabling the sandbox resolves this issue, well is beyond my speculation.

Perhaps I will revert to older fglrx driver...perhaps future upgrades to Chrome will resolve this issue...I just hope that disabling the sandbox doesn't leave me vulnerable to nefarious websites!!

I'm adding to my checklist to try with the unaccelerated vesa driver.  As 
mentioned before, I will do my best to be back on  Saturday with my testing results.

With reference to filing a bug against the kernel, there is one already opened 
by myself on launchpad. I do not know if they have propagated it upstream, 
though. My feeling is that since I have so far tested only with legacy drivers 
and being the bug probably hard to reproduce, it might have remained on the 
ubuntu tracking system only.

Sergio

OK - booted Ubuntu 10.04 and ripped out fglrx in favour of xserver-xorg-video-ati and can confirm Chrome doesn't cause kernel panic on closure with or without compositing - which is good news! Am actually quite impressed with the performance of the Xorg Ati driver (Radeon) - even full screen Flash video renders well without tearing! Shame Radeon won't work well on my Debian Lenny based system...

So concluding - it is some component of the proprietary Ati Catalyst (fglrx) driver causing the kernel panics in my case - not sure what Nvidia users options are, as I'm not a gamer - I don't *need* top gpu performance (or closed source code blobs) so for now - all is well again.

OK - just happened again using Ubuntu with the Radeon (xserver-xorg-video-ati) driver - so it isn't just with fglrx/nvidia.

What I noticed just before the kernel panic - which this time blanked my whole screen - i.e. crashed Xorg, was that Gmail wasn't behaving - couldn't enter text into message box - then realised that I couldn't enter text anywhere in Chrome - restarting Gmail tab didn't resolve issue. Closing browser caused crash.

This must be a virtual memory issue - I have noticed that Gmail can chomp through memory, so I am guessing this is likely related.

This is running Chrome as standard - sandbox and all.

I will monitor Gmail mem usage (and other tabs) to see if I can ID this issue.

Hi,

I am back with my tests:

1) Using the noveau drivers and no compositing (rather than proprietary 
nvidia and compositing) still I get the lockup as one exits chromium 
(tested with chromium 7.0.515.0~svn20100904r58587).

Interestingly, with the noveau drivers, just before the lockup X crashes 
(or is killed).

2) disabling the sandbox I get no lockup (with the same chromium and the 
proprietary nvidia drivers).

3) with the alternative sandbox, chromium is unable to produce a graphic 
window and seems to hang. However as soon as I try to interrupt it, I 
get the lockup.

The sole other difference that I have on this machine with regards to 
the others I am using (apart from the AMD cpu) seems to be a more 
complex mix of SATA and IDE drivers (sata harddisk, IDE DVD-RW, IDE Zip 
drive).

I wonder if this may help, but I hope so.

Please provide me hints to add to the kernel bug report to help having 
it reported upstream.
Indeed, since 2.6.32 seems to be quite ubiquitous it would be very 
interesting to find out what specifically happens in the chrome sandbox 
to trigger the issue.

Sergio

The modern alternative to serial console is netconsole(1) btw. ... that might get you some actual kernel messages for the panic and shed some light on the issue.

(1) http://www.kernel.org/doc/Documentation/networking/netconsole.txt

PS. Just in case you haven't done so, you should run memtest86 etc. etc.

I have run memtest86, but definitively not for a sufficiently long time 
(30'). I'll try to leave it on for a few hours one of these nights.

I'll also take a look at netconsole.  Thanks for the advice!

Sergio

Caught screenshot of last kernel panic - see attached or here:
http://picasaweb.google.co.uk/lh/photo/_NxGAXlAXmMpwc8eGz0JZA?feat=directlink

I can what appears to be a double kp and mention of phone device (usb connected N900) would/should this have any effect?

Sticking with Radeon for now and continue to monitor and look into netconsole if applicable.

	20100910_001.jpg 83.3 KB View Download

OK - I have found the cause of the kernel panics - as indicated by above crash log - the kernel panic occurs when closing google-chrome *only while I have my Nokia N900 plugged in via usb*. I have tried this on two different distros (Mepis 8.5 + Ubuntu 10.04) with the same result.

So - kernel bug or not - what is it about closing Google-Chrome (sandbox enabled) that causes kernel panic on my systems? I won't be attempting to file kernel bug without significant and comprehensive info on this situation! 

Interesting...

I am also using the browser with a mobile phone based access, with a 
nokia phone attached to usb.

I am again away from home. As soon as I am back I will check again 
without connecting the mobile phone.

Sergio

Possibly related kernel git update for 2.6.35

http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff_plain;h=7dfde179c38056b91d51e60f3d50902387f27c84

??

Got something interesting!

By connecting to the internet with a Nokia E55 attached via usb, at the 
exit of google chrome *also my laptop* (with Intel CPU, Intel graphics, 
free video drivers) locks up!

So it is definitively something wrong in the kernel/network layer...

Interesting bit is that:

1) Only chrome appears to be able to trigger it (no other application 
has ever done so, and I am a heavy user of mobile internet access).
2) Chrome triggers it on exit, when network access should be over.
3) Chrome triggers it only if some page was navigated before exit.
4) Chrome triggers it only if the sandbox is on (this is the part that 
is weirdest to me)
5) Bug is triggered only if the network is accessed via the mobile phone 
(no issue on wifi or ethernet)

As soon as I can, I gonna look if the issue happens also by connecting 
to the mobile via a bluetooth link.
Unfortunately I do not know how to test other types of ppp connections.

Please, if someone has at hand a mobile with internet access, try to see 
if the issue is systematically reproducible, so that the kernel 
developers can get a proper bug report with a test case.

Just tried using the N900 as 3G modem - works fine. Disconnected from mobile internet (via Ubuntu's nm-applet) then closed Chrome - same kernel oops occurs...

This is interesting progress!

It's a pity we can't see the very first chunk of trouble from the kernel. Nothing in /var/log/messages? Can't scroll back? Maybe bumping up the vga/console/framebuffer/whatever resolution will allow it to capture the first couple of lines? (and of course there is netconsole). The kernel people will probably want to see the initial crash.

The 'command' that invoked the last (tainted) chunk of output was 'netns' which seems to be a kernel thread involved in the network namespace stuff which is interesting given that the sandbox used CLONE_NEWNET for a while - that has arguably been backed out again lately since it created issues with the automounter on ubuntu so perhaps it's not involved. I mention this because it's a pretty obscure feature and it's probably chrome's use of relatively obscure kernel features like this that make this bug seem to only show up with chrome...

There is a similar phonet related kernel crash here btw.:

https://bugzilla.redhat.com/show_bug.cgi?format=multiple&id=595043

Maybe someone can try the kernel patch mentioned there (same as comment #20 here) and see if it helps at all?

I have posted the latest infos as a comment to the ubuntu bug I opened before this thread started. Please, also follow https://bugs.launchpad.net/ubuntu/+source/linux/+bug/616745

OK - got round to capturing the kernel oops using netconsole - relevant beginning part:


[  122.789154] BUG: unable to handle kernel NULL pointer dereference at (null)                         
[  122.789175] IP: [<f80dd18a>] phonet_device_destroy+0x8a/0x160 [phonet]                              
[  122.789197] *pde = 6fa5e067                                                                         
[  122.789209] Oops: 0000 [#1] SMP                                                                     
[  122.789222] last sysfs file: /sys/devices/pci0000:00/0000:00:12.2/usb1/1-6/1-6:1.0/net/usbpn0/statistics/collisions                                                                                                    
[  122.789232] Modules linked in: cdc_phonet cdc_ether phonet cdc_acm usbnet usb_storage binfmt_misc vboxdrv netconsole snd_hda_codec_atihdmi fbcon tileblit font bitblit softcursor vga16fb vgastate configfs snd_hda_codec_realtek snd_hda_intel snd_hda_codec snd_hwdep snd_pcm_oss snd_mixer_oss snd_pcm snd_seq_dummy snd_seq_oss snd_seq_midi snd_rawmidi snd_seq_midi_event snd_seq radeon snd_timer snd_seq_device ppdev ttm drm_kms_helper it87 parport_pc hwmon_vid drm agpgart i2c_algo_bit snd i2c_piix4 soundcore snd_page_alloc shpchp lp parport usbhid r8169 ohci1394 8139too hid 8139cp ieee1394 mii ahci pata_atiixp                                        
[  122.789449]                                                                                               
[  122.789459] Pid: 13, comm: netns Not tainted (2.6.32-24-generic #42-Ubuntu) GA-MA785GMT-UD2H              
[  122.789470] EIP: 0060:[<f80dd18a>] EFLAGS: 00010293 CPU: 1                                                
[  122.789481] EIP is at phonet_device_destroy+0x8a/0x160 [phonet]                                           
[  122.789490] EAX: 00000000 EBX: f318fd68 ECX: eecf9400 EDX: f318fd60                                       
[  122.789499] ESI: 00000000 EDI: eecf9400 EBP: f70b7ebc ESP: f70b7e9c                                       
[  122.789507]  DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068                                                 
[  122.789516] Process netns (pid: 13, ti=f70b6000 task=f70aa640 task.ti=f70b6000)                           
[  122.789523] Stack:

For the complete capture from usb connection on N900 to oops at closure of Google-Chrome - see attached file.

Note - the crash only happens if once phone is attached - connection type is then selected on usb device (phone) itself (PC Suite/USB Mass Storage) - not otherwise.

I have not tried kernel patch - presumably this involves installing latest kernel? Or can patch be applied to current version (2.6.32.x)?

kerneloops.chrome.nokia.txt 10.3 KB View Download

I cannot make the same test, since I do not have the second pc at hand 
to try the netconsole.

However... is the solution as easy as blacklisting the phonet module, 
which is anyway not strictly necessary to go online via the mobile?

Can you please try

1) Attaching the nokia
2) modprobe -r phonet
     this may require removing some other module depending on phonet first
3) going online with kppp (or whatever) using /dev/ttyACM0
4) try chrome
5) see what happens on exit

Sergio

I pinged one of the nokia guys to ask for help and he sent me the patch below. Awesome!!!!!!

Is anyone up for rebuilding their kernel from source? I'm not on ubuntu so I'm not sure how easy it is to roll a third party patch into a deb. Alternately running a later kernel that includes this patch or an equivalent might help. 

The patch below is probably whitespace damaged as I'm trying to paste it from my iPad btw. - I'll attach a pristine copy tomorrow.

From b995ecddc515c5e137a6e53632a9aedbf401553f Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?R=C3=A9mi=20Denis-Courmont?= <remi.denis-courmont@nokia.com>
Date: Tue, 2 Feb 2010 17:04:23 +0200
Subject: [PATCH 1/1] Phonet: disable namespace functionality (until 2.6.33)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Network namespace in the Phonet socket stack causes an OOPS when the
a namespace is destroyed. This occurs as the loopback exit_net handler
is called after the Phonet exit_net handler, and re-enters the Phonet
stack. There is no easy way to fix this in kernel <= 2.6.32. As there
is no use for Phonet namespaces yet, disable them.

(This does *not* disable IP namespaces.)

Signed-off-by: Rémi Denis-Courmont <remi.denis-courmont@nokia.com>
---
 net/phonet/pn_dev.c |   24 ++++++++++++++++--------
 1 files changed, 16 insertions(+), 8 deletions(-)

diff --git a/net/phonet/pn_dev.c b/net/phonet/pn_dev.c
index 1c8643d..e4e0327 100644
--- a/net/phonet/pn_dev.c
+++ b/net/phonet/pn_dev.c
@@ -47,7 +47,7 @@ int phonet_net_id __read_mostly;
 
 struct phonet_device_list *phonet_device_list(struct net *net)
 {
-	struct phonet_net *pnn = net_generic(net, phonet_net_id);
+	struct phonet_net *pnn = net_generic(&init_net, phonet_net_id);
 	return &pnn->pndevs;
 }
 
@@ -261,7 +261,7 @@ static int phonet_device_autoconf(struct net_device *dev)
 
 static void phonet_route_autodel(struct net_device *dev)
 {
-	struct phonet_net *pnn = net_generic(dev_net(dev), phonet_net_id);
+	struct phonet_net *pnn = net_generic(&init_net, phonet_net_id);
 	unsigned i;
 	DECLARE_BITMAP(deleted, 64);
 
@@ -312,7 +312,11 @@ static struct notifier_block phonet_device_notifier = {
 /* Per-namespace Phonet devices handling */
 static int phonet_init_net(struct net *net)
 {
-	struct phonet_net *pnn = kzalloc(sizeof(*pnn), GFP_KERNEL);
+	struct phonet_net *pnn;
+
+	if (!net_eq(net, &init_net))
+		return 0;
+	pnn = kzalloc(sizeof(*pnn), GFP_KERNEL);
 	if (!pnn)
 		return -ENOMEM;
 
@@ -330,10 +334,14 @@ static int phonet_init_net(struct net *net)
 
 static void phonet_exit_net(struct net *net)
 {
-	struct phonet_net *pnn = net_generic(net, phonet_net_id);
+	struct phonet_net *pnn;
 	struct net_device *dev;
 	unsigned i;
 
+	if (!net_eq(net, &init_net))
+		return;
+	pnn = net_generic(net, phonet_net_id);
+
 	rtnl_lock();
 	for_each_netdev(net, dev)
 		phonet_device_destroy(dev);
@@ -379,7 +387,7 @@ void phonet_device_exit(void)
 
 int phonet_route_add(struct net_device *dev, u8 daddr)
 {
-	struct phonet_net *pnn = net_generic(dev_net(dev), phonet_net_id);
+	struct phonet_net *pnn = net_generic(&init_net, phonet_net_id);
 	struct phonet_routes *routes = &pnn->routes;
 	int err = -EEXIST;
 
@@ -396,7 +404,7 @@ int phonet_route_add(struct net_device *dev, u8 daddr)
 
 int phonet_route_del(struct net_device *dev, u8 daddr)
 {
-	struct phonet_net *pnn = net_generic(dev_net(dev), phonet_net_id);
+	struct phonet_net *pnn = net_generic(&init_net, phonet_net_id);
 	struct phonet_routes *routes = &pnn->routes;
 
 	daddr = daddr >> 2;
@@ -416,7 +424,7 @@ int phonet_route_del(struct net_device *dev, u8 daddr)
 
 struct net_device *phonet_route_get(struct net *net, u8 daddr)
 {
-	struct phonet_net *pnn = net_generic(net, phonet_net_id);
+	struct phonet_net *pnn = net_generic(&init_net, phonet_net_id);
 	struct phonet_routes *routes = &pnn->routes;
 	struct net_device *dev;
 
@@ -431,7 +439,7 @@ struct net_device *phonet_route_get(struct net *net, u8 daddr)
 
 struct net_device *phonet_route_output(struct net *net, u8 daddr)
 {
-	struct phonet_net *pnn = net_generic(net, phonet_net_id);
+	struct phonet_net *pnn = net_generic(&init_net, phonet_net_id);
 	struct phonet_routes *routes = &pnn->routes;
 	struct net_device *dev;
 
-- 
1.7.0.4

Yes, the Chrome Linux SUID sandbox use network namespaces. (see src/sandbox/linux/suid/sandbox.c) So the patch in comment 37 may fix your problem.

I think that the best thing to do as soon as you can post the clean 
patch is to cross-post it after the ubuntu bug report. Hopefully, the 
ubuntu kernel developers can pick it up for the next maintainance 
release of the lucid kernel.

Also I hope that the extra-efficient Nokia person can push a full fix 
for the current kernel on the LKML (if it is not already fixed there).

However, I have a feeling that the ubuntu kernel team is rather 
reluctant at applying changes to a mainstream LTS kernel unless they are 
security fixes, even if they fix bugs. I still remember when the Hardy 
2.6.24 kernel was freezing on my hardware while 2.6.25 was not. However, 
since PPAs were introduced in the meantime, I guess there might be a 
possibility to build a PPA kernel with the fix if the patch applies 
cleanly on the ubuntu tree.

Ubuntu also provides a "mainline" 2.6.34.0 for Lucid via the mainline 
PPA. However, since running mainline kernels on ubuntu is a pain to all 
those having NVIDIA or other hardware that goes best with "restricted 
modules" since the latter are unavailable on mainline kernels.

I wonder what are the current uses of phonet.  What would a user with a 
Nokia phone loose by temporarily blacklisting it as a workaround?

Sergio

Pristine patch from Rémi attached.

(no launchpad login ... I might create one later and comment on the other bugs and poke in the kernel git tree a bit)

If you're on a rpm based distribution the steps to rebuild a kernel are pretty much:
1. install kernel src rpm
2. place patch file in SOURCES dir
3. edit spec file to refer to patch
4. rebuild new rpm with rpmbuild and then install it using rpm

The cool part about doing it that way rather than building directly from kernel.org source is that you get a working kernel .config file to start with, you get all the distro specific patches and the result is something your package manager can deal with. I'd imagine there is a similarly easy process for deb files you can follow...

0001-Phonet-disable-namespace-functionality-until-2.6.33.patch 3.7 KB View Download

Might have a go at patching kernel over weekend, but probably will just try blacklist module or just not select connection mode on phone once plugged in, and wait till newer kernel arrives - presuming this issue won't be present in newer kernels?

Great to get a (hopefully) working patch for this so quickly! Thinking about this - I have probably been experiencing this issue for some months - but blaming Flash/Ati for it!! I would imaging there are quite a number of other people experiencing this, but with no clue as to the cause - hopefully now they will find this tread + solution.

Just installed and tested a 2.6.35 kernel (Liquorix: Linux mepis8point5 2.6.35-pae-liquorix-i386 #1 ZEN SMP PREEMPT Thu Sep 16 19:41:11 PDT 2010 i686 GNU/Linux)

Google-Chrome no longer causes kernel oops on closure while Nokia device connected *and activated* via usb cable.

Seems 2.6.35 is the way to go if this issue is causing problems.

Patch still useful for those not ready for a newer kernel.

Rémi says the patch from comment #30 unfortunately doesn't apply on a pristine 2.6.32 btw. :(

He recommends upgrading to 2.6.33 or later.

I have provided a link to the patch on the ubuntu bug traking system a 
few days ago.
I hope they will look at this latest bit of info too.

BTW I think that the patch would be useful mostly for ubuntu Lucid LTS, 
RHEL or CentOS, as other distros/releases are anyway likely to
move to more recent kernels in a short timeframe.

In the ubuntu bug tracking system I have also suggested blacklisting 
phonet as a workaround.

Is there any strong counterindication against the workaround?
Does some application critically depend on phonet?

No action taken yet by the ubuntu kernel team.

What is the meaning of "blacklisting phonet ". Will I still be able to connect to internet via my Nokia N85 mobile device?

Having a file with a .conf extension in

blacklist cdc_phonet
blacklist phonet

in /etc/modprobe.d

will prevent phonet and cdc_phonet from auto-loading when a Nokia phone 
is attached (sorry if this explanation is - as I guess - absolutely 
unnecessary to you, but it is probably good to make things as easy as 
possible for all readers).

As an aside note that when my E55 is connected without this blacklisting 
the phonet module is immediately autoloaded, regardless of whether I 
activate or not on the phone the Nokia PCsuite service.

To go back to your question, blackisting phonet will not prevent getting 
a connection to the internet. I am currently sending this email with the 
E55, after having setup a connection with network-manager.

Apparently, phonet is a network protocol that is still not (much) 
exploited by applications and Nokia phones offer alternatives that are 
currently used. From what I could understand, it was added to the kernel 
very much in provision of future use and as it is needed by linux 
powered nokia tablets.

I hope this clarifies. Maybe the same Nokia person who provided the 
patches can give pointers to places explaining the situation a bit better.

In any case, if the above is correct, we are in a nice position.

Those with less recent kernels such as 2.6.32 (lucid) can simply 
blacklist the module.
Those with more recent kernels such as 2.6.35 can rely on a kernel with 
a fixed phonet already.

I only wonder about 2.6.34, since this is going to be the next ubuntu 
kernel.  Is there anyone who can try?

Seergio

should be easy enough to try 2.6.34 from the kernel-ppa source:

http://kernel.ubuntu.com/~kernel-ppa/mainline/v2.6.34-lucid/

Might give it a go later myself, but if you are going to bother upgrading your kernel, why bother with .34 when .35 (and even .36) are available?

Unfortunately on the nvidia machine trying mainline kernels is a bit of 
a pain since you need to discard all the "restricted" stuff and then set 
up that again. Gonna try as soon as I can be on the machine with intel 
graphics.

The reason to bother with .34 is that the next round of distros (ubuntu 
maverik to start with) will ship .34. So if .34 is broken wrt phonet 
namespaces and can as such lockup when using chrome, this is the right 
moment to say so, so that a patch can be incorporated before release.

Sergio

 Issue 55670  has been merged into this issue.

Thanx sergio

The issue also tracked on launchpad: https://bugs.launchpad.net/ubuntu/+source/chromium-browser/+bug/638024

For clarification r.e. post #39 - Ubuntu 10.10. Maverick will ship with the 2.6.35* kernel - so the .34 kernel shouldn't be an issue there:

https://wiki.ubuntu.com/MaverickMeerkat/TechnicalOverview#Linux kernel 2.6.35

Not tested the latest Maverick kernel myself though so can't confirm whether this patch is mainline yet there too or not.

Upgraded to Ubuntu 10.10 Maverick, ripped out Google-Chrome in favour of Ubuntu packaged Chromium-Browser - no kernel oops on closure of Chromium while Nokia device connected and activated via usb.

uname -a
Linux Phenom4x 2.6.35-22-generic #33-Ubuntu SMP Sun Sep 19 20:34:50 UTC 2010 i686 GNU/Linux

dpkg -l chromium-browser

ii  chromium-browser              6.0.472.62~r59676-0ubuntu1    Chromium browser


I think we can call this bug closed for 2.6.35 and newer kernels.

Great, it was me to refer to too old stuff then!

Sergio

Repro:


[ 6295.349006] BUG: soft lockup - CPU#1 stuck for 61s! [netns:13]
[ 6295.349006] Modules linked in: ppp_deflate bsd_comp ppp_async crc_ccitt cdc_ether cdc_phonet phonet
 usbnet mii cdc_acm usb_storage binfmt_misc rfcomm ppdev sco bridge stp bnep l2cap vboxdrv deflate zlib_deflate ctr twofish twofish_
common camellia serpent blowfish cast5 des_generic aes_i586 aes_generic xcbc rmd160 sha256_generic sha1_generic crypto_null af_key d
m_crypt joydev snd_hda_codec_idt snd_hda_intel snd_hda_codec snd_hwdep snd_pcm_oss snd_mixer_oss snd_pcm snd_seq_dummy snd_seq_oss s
nd_seq_midi arc4 snd_rawmidi snd_seq_midi_event snd_seq snd_timer snd_seq_device iwl3945 iwlcore snd dell_wmi uvcvideo videodev v4l1
_compat btusb bluetooth sdhci_pci sdhci ricoh_mmc mac80211 led_class dell_laptop dcdbas psmouse serio_raw soundcore snd_page_alloc c
fg80211 uinput lp parport fbcon tileblit font bitblit softcursor vga16fb vgastate usbhid hid i915 drm_kms_helper ohci1394 drm ieee13
94 intel_agp ahci i2c_algo_bit video output tg3 agpgart
[ 6295.349006] 
[ 6295.349006] Pid: 13, comm: netns Not tainted (2.6.32-25-generic #44-Ubuntu) Inspiron 1420          
         
[ 6295.349006] EIP: 0060:[<c012a4f5>] EFLAGS: 00000217 CPU: 1
[ 6295.349006] EIP is at __ticket_spin_lock+0x15/0x20
[ 6295.349006] EAX: f0d6f878 EBX: f0d6f878 ECX: efdc4c00 EDX: 00009400
[ 6295.349006] ESI: 00000000 EDI: efdc4c00 EBP: f70b1e88 ESP: f70b1e88
[ 6295.349006]  DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068
[ 6295.349006] CR0: 8005003b CR2: b361b000 CR3: 0084e000 CR4: 000006d0
[ 6295.349006] DR0: 00000000 DR1: 00000000 DR2: 00000000 DR3: 00000000
[ 6295.349006] DR6: ffff0ff0 DR7: 00000400
[ 6295.349006] Call Trace:
[ 6295.349006]  [<c058d437>] _spin_lock_bh+0x17/0x20
[ 6295.349006]  [<fabd914c>] phonet_device_destroy+0x4c/0x160 [phonet]
[ 6295.349006]  [<c05469a9>] ? addrconf_notify+0x99/0x470
[ 6295.349006]  [<c0512e34>] ? arp_ifdown+0x14/0x20
[ 6295.349006]  [<c056ab76>] ? packet_notifier+0x26/0x1a0
[ 6295.349006]  [<fabd95d9>] phonet_device_notify+0x19/0x40 [phonet]
[ 6295.349006]  [<c058fa03>] notifier_call_chain+0x43/0x60

 Issue 56894  has been merged into this issue.

 Issue 61305  has been merged into this issue.

 Issue 62343  has been merged into this issue.

I've noticed that the  issue 55848  is about the same problem.

Chrome Version       : Any recent 6.x or 7.x, not 5.x or previous
URLs (if applicable) : Any
Other browsers tested:
Add OK or FAIL after other browsers where you have tested this issue:
     Safari 4: Untested
  Firefox 3.x: OK
         IE 7: Untested
         IE 8: Untested


What steps will reproduce the problem?
1.  Start google chrome
2.  Open some web pages (even locally... i.e. file:/ urls)
3.  Exit chrome

What is the expected result?

   Browser opens, navigates, closes

What happens instead?

   On browser exit, the whole machine goes into a hard lockup, with mouse cursor freezing, display freezing, network freezing keyboard leds flashing.


Please provide any additional information below. Attach a screenshot if
possible.

1) Machine is a PC, with AMD Phenom II X4 cpu and nvidia graphics

2) OS is Ubuntu Linux 10.04 64 bit

3) Problem seems to occur only on the machine with the Phenom CPU. On a machine with the same 64 bit distribution, Nvidia graphics I get no lockup.
Also the software selection is almost the same.

4) The machine where chrome gives the lockup had never had a single hiccup with no other application.

5) Problems happens either with google-chrome from google and with chromium (ubuntu ppa build)

Any clue on how to try debugging this?

disabling sandbox works for me. although terminal should me some error will closing chrome. best part is kernel panic dint happen.

The original bug is on the kernel module cdc_phonet, not in chrome/chromium.

This has been solved in recent kernels. If you are on a kernel with the bug (2.6.32), easiest thing to to is to blacklist the cdc_phonet module.

As mentioned this is as easy as having a file with a .conf extension in

blacklist cdc_phonet
blacklist phonet

in /etc/modprobe.d

If this does not help, then you are probably having another bug, that might deserve being reported separately with full details.

This was fixed in Linux 2.6.32.40.

phonet was introduced in Linux 2.6.28 and I don't think any distributions are based on 2.6.28-2.6.31, so this could perhaps be closed.

It should be closed (since it's not a Chrome bug) as long as the workaround from comment 52 can be found by those who still need it. Maybe that hint should be moved up into the bug description?

FYI - Ubuntu 10.04 (still a supported distro, still used in production environments) still uses kernel 2.6.32.

$ uname -a
Linux lg 2.6.32-45-generic #99-Ubuntu SMP Tue Oct 16 16:26:56 UTC 2012 x86_64 GNU/Linux

It's 2013, Ubuntu 10.04 is no longer supported on the desktop.