Halt during boot after update to 3.4.3-1.fc17

surfsafari · #1 22nd June 2012, 07:38 PM

I just updated my kernel to vmlinuz-3.4.3-1.fc17.x86_64, and now my machine halts during the boot process
after starting the "NFS File Locking Service". I can still boot the machine with vmlinuz-3.4.0-1.fc17.x86_64 with no problems. I have no idea what special settings may affect this, but I have changed very little from the default settings on installation. There are two things that I notice that are different when booting under vmlinuz-3.4.3-1.fc17.x86_64: 1) Starting Network Manager Wait Online fails 2) The system tries to start this NFS file locking service. Looking at /var/log/boot.log when booted under vmlinuz-3.4.0-1.fc17.x86_64 shows no mention of starting any NFS file locking service and no problems with Network Manager Wait Online.

Any ideas? A quick look around the internet suggests that no one else is seeing anything quite like this. Perhaps I am searching in the wrong places or perhaps I have some special settings. For instance, I think that I asked for the NFS file locking service to not ever start after I upgraded to vmlinuz-3.4.0-1.fc17.x86_64, because I have no use for NFS.

... again. Any help would be welcome.

JohnSeers · #2 23rd June 2012, 10:01 AM

Hi

I have just updated to kernel-3.4.3-1.fc17(64 bit) and I am having problems too.
First time just hung. (When "Restartting"). Tried rebooting several times - nothing.
Removed kernel and reinstalled. Failed boot message something about fixing a recursive problem.
Tried rebooting again and what looked like an infinite loop of low level trace messages.
I can still reboot from perevious kernel.
Any help very welcome as well. More info available if required.

John

surfsafari · #3 25th June 2012, 05:51 PM

John, I am not having that problem.

I looked at /var/log/messages, and it appears that X is failing to start which is preventing kdm from starting. I forgot to check whether the video driver is loading or not, but I think that there are entries in /var/log/messages if it has any problems. So, ... I still do not know what is happening and I wonder why/whether we are the only ones experiencing problems.

---------- Post added at 11:51 AM ---------- Previous post was at 11:48 AM ----------

This guy is having yet another problem:
http://forums.fedoraforum.org/showthread.php?p=1587268

JohnSeers · #4 25th June 2012, 07:20 PM

Hi

There seem to be some references to failed kernels but none of them quite match my experience. Here is a bug entry that suggests setting "acpi=off" may allow the kernel to boot though it did not work for me.

https://bugzilla.redhat.com/show_bug.cgi?id=834318

I have tried a few things since and seem to get a variety of symptoms. Including a successful boot sometimes! Sometimes the boot process reboots, sometimes freezes on the fedora screen, sometimes earlier. I have had a message saying unable to open a fonts file. Also message "Detected recursive error - reboot required". I have been trying to find out how to capture the problem but it is so variable I am not having much success. Luckily I can always boot with the previous kernel - which also makes me think I do not have a hardware problem.

I have removed "rhgb" and "quiet" when booting. I have looked in some boot logs but they have not been helpful so far.

Does anybody have any suggestions how to capture this problem?

JS

surfsafari · #5 26th June 2012, 04:15 PM

My immediate problem is that my video driver fails to load, which means that X cannot start. I used the akmod version of the video driver to avoid having to worry about updating the video driver manually when installing a new kernel. I do not remember this being a problem in Fedora. Perhaps with the old system a kernel update would automatically trigger an update of the video driver. However, If this website is correct, then I am not fulfilling my duties as an akmod user.

http://fedorasolved.org/Members/zcat/akmods

So ... my problem may be my own fault.

Incidentally, I am getting the same message about being unable to open a fonts file. I am not sure what triggered that, but because I am seeing it I suspect that that is unrelated to your problem. I am not sure exactly where in the boot process you are having problems, but to me, it sounds as if you have a "late boot" problem. This means that the problems are occurring during the execution of the boot scripts. These are contained in /etc/rc.d. They are written in Python. It should be possible to add comments or new commands to the boot scripts. Make sure you save a copy before you mark one up. The problem with this approach is that this may not be the kind of information that the Fedora developers want to get. However, it might give you some clue as to where the boot process is failing. Alternatively, there may be ways to configure the kernel or various services to be more verbose. Again, this would likely be found in /etc.

Just being able to put a box around the time when your problems are occurring could be useful. To be able to say that you never see any problems >before< a certain time or >afterward< could be illuminating. For instance, it could be that some service is having problems and this service is required by several other services which are started in parallel. As a result, the problem that you see when booting depends upon which of the dependent services start first. If the dependent services boot in a certain order, then you get a successful boot.

sorry I cannot be more helpful

About a month or two ago, I saw that the market share of desktop linux users was in the low single percentages. It is possible that the increased incidence of problems is due to short staffing at Fedora. It may also be that linux is under pressure in the server market as well.

JohnSeers · #6 27th June 2012, 09:49 AM

Hi

Thanks for the reply.

I installed the debug version of the kernel and that worked fine. (Though a little slow). So that did not really help as nothing to work with!

I think it must be some sort of hardware issue. Perhaps something to do with the graphics card. It is strange it works intermittently and hwy only on that version of the kernel.

Not sure what to try next. Perhaps the next rainy day we have I will have another try. Else wait for the next kernel and hope it works.

JS

JohnSeers · #7 28th June 2012, 10:53 AM

Hi

I have managed to get this kernel to boot successfully. It is something to do with the NVIDIA card and the nouveau driver.

I disabled the nouveau driver by appending "rdblacklist=nouveau modeset.nouveau=0" to the kernel line. It would then boot but with a low resolution screen.

I have tried in the past to install the NVIDIA driver from rpmfusion but I have always ended up with an unuseable system But I tried again.

After reinstalling Fedora 17 from scratch I came across these instructions for installing the NVIDIA drivers from NVIDIA:

http://geek.co.il/wp/2011/11/15/inst...-way#more-1852

I thought it was a bit unlikely but I gave it a go and it has worked for Fedora 17. Just remember to install the development source before running the NVIDIA script.

JS

surfsafari · #8 29th June 2012, 03:54 PM

I use the nVidia binary driver. I'm not sure that I noticed any difference in performance between the nouveau and binary driver unless there graphics were 3D, in which case the difference was >very< noticeable. I do not know why some nVidia cards have problems with the nouveau driver. I suspect that this is a known issue because the nouveau driver worked for me on this machine out of the box with Fedora 15. I also suspect that the problem is that X is having trouble finding a screen resolution/refresh rate that the video card will accept. There should be an error message to that effect, but perhaps it has been eliminated. This was always the problem with getting my nVidia video cards to work back in the days when I was writing xorg.conf by hand.

It is really interesting that the debug kernel eliminated your problems. I'm not sure what to make of that, but it seems consistent with the idea that there is some kind of a timing problem. Anyway, glad to hear that your problem is at least partially solved. It now seems that our problems are likely to be related.

---------- Post added at 09:54 AM ---------- Previous post was at 09:17 AM ----------

It turns out that my problem was indeed my fault. I ran the following commands and I am now running the
3.4.3-1.fc17 kernel.

/usr/bin/yum update kernel-devel
/usr/sbin/akmods --kernel 3.4.3-1.fc17.x86_64
/sbin/reboot

JohnSeers · #9 29th June 2012, 04:50 PM

Hi

I am pleased to hear you have sorted out your problems. If the rpmfusion nvidia drivers would work for me I think I would use akmods.

What exactly is the "nVidia binary driver"? I think I have tried most of the options available but I cannot recall seeing an option called that. Perhaps it is what I need! Or do you just mean not the source option and it is not anything with a particular meaning?

JS

surfsafari · #10 29th June 2012, 06:07 PM

Sorry ... what I wrote was confusing at best. The nvidia driver, the nvidia binary driver and the nVidia proprietary driver are the same thing. It is unfortunate that the rpmfusion packaged versions of the driver have not worked for you. I wish that there were official fedora instructions for installing the nVidia binary driver, but fedora only supports nouveau. In your situation, using akmods might be the best route. You could update kernel-devel and run akmods each time you have a kernel update and very seldom update the nVidia driver by the "official" method. However, for me, I am not sure that using akmods is actually solving any problems for me. I should probably be using kmods.

jwmueller · #11 29th June 2012, 10:46 PM

I ditched my nvidia card for radeon over a year ago so I don't know how well this applies... but I haven't seen it referenced yet in this thread. Have you referenced this: http://forums.fedoraforum.org/showthread.php?t=280750