Kernel panic on startup?
I have a very strange and persistent kernel oops when I try to start FF 39.0, This is for a clean install with a new clean profile.
Please understand this problem has suddenly started to happen after months of working without any problem except for an occasional crash while watching programs from www.bbc.co.uk/iplayer. Following a particularly nasty occurrence I was forced to Ctrl Alt F3 login as root and reboot. (ie my X session was locked out) Now whenever I try to start FF 39.0 I get the kernel oops!
The command line result is below and a snippet from /var/log/messages is attached.
The exceeding odd behaviour is that I can run the old Fedora 17 package managed version 22.0 of FF without any problems. I also have a Google Chrome version which runs perfectly.
I have re-installed Firefox, and the flash plugin (11.2.202.468), and even the kernel!
I do not understand much of the following dumps, what can I do to illuminate the problem?
TIA Charles Bradshaw
~o~
Command line result running 39.0 from the install directory:
[brad@p2800 firefox]$ ./firefox
Message from syslogd@p2800 at Jul 7 16:23:40 ...
kernel:[18148.112012] Process firefox (pid: 4215, ti=ece7a000 task=f192e900 task.ti=ece7a000)
Message from syslogd@p2800 at Jul 7 16:23:40 ...
kernel:[18148.112012] Stack:
Message from syslogd@p2800 at Jul 7 16:23:40 ...
kernel:[18148.112012] Call Trace:
Message from syslogd@p2800 at Jul 7 16:23:40 ...
kernel:[18148.112012] Code: 7d fc 89 ec 5d c3 90 8d 74 26 00 c7 43 34 00 00 00 00 b9 50 87 c0 c0 89 da c7 43 38 00 00 00 00 89 f8 e8 34 e8 ff ff 89 c6 eb 91 <0f> 0b 8d b4 26 00 00 00 00 8d bc 27 00 00 00 00 55 89 e5 57 56
Message from syslogd@p2800 at Jul 7 16:23:40 ...
kernel:[18148.112012] EIP: [<c04c0110>] __audit_syscall_entry+0xf0/0x100 SS:ESP 0068:ece7bfa4
Segmentation fault
~o~
Snippet from /var/log/messages Jul 7 16:23:40 p2800 kernel: [18148.111629] ------------[ cut here ]------------ Jul 7 16:23:40 p2800 kernel: [18148.112012] kernel BUG at kernel/auditsc.c:1772! Jul 7 16:23:40 p2800 kernel: [18148.112012] invalid opcode: 0000 [#7] SMP Jul 7 16:23:40 p2800 kernel: [18148.112012] Modules linked in: fuse lockd sunrpc nf_conntrack_netbios_ns nf_conntrack_broadcast ip6t_REJECT nf_conntrack_ipv4 nf_defrag_ipv4 nf_conntrack_ipv6 nf_defrag_ipv6 xt_state nf_conntrack ip6table_filter ip6_tables arc4 rt2800usb uvcvideo rt2x00usb rt2800lib rt2x00lib snd_hda_codec_idt videobuf2_vmalloc videobuf2_memops mac80211 videobuf2_core videodev snd_hda_intel cfg80211 snd_usb_audio snd_usbmidi_lib snd_hda_codec media snd_rawmidi snd_hwdep snd_seq rfkill serio_raw crc_ccitt iTCO_wdt iTCO_vendor_support lpc_ich snd_seq_device mfd_core mperf snd_pcm i2c_i801 dcdbas snd_page_alloc snd_timer snd soundcore e100 mii microcode binfmt_misc hid_logitech_dj i915 video i2c_algo_bit drm_kms_helper drm i2c_core Jul 7 16:23:40 p2800 kernel: [18148.112012] Pid: 4215, comm: firefox Tainted: G D 3.9.10-100.fc17.i686.PAE #1 Dell Inc. Dell DV051 /0JC474 Jul 7 16:23:40 p2800 kernel: [18148.112012] EIP: 0060:[<c04c0110>] EFLAGS: 00210202 CPU: 0 Jul 7 16:23:40 p2800 kernel: [18148.112012] EIP is at __audit_syscall_entry+0xf0/0x100 Jul 7 16:23:40 p2800 kernel: [18148.112012] EAX: 40000003 EBX: ecda9800 ECX: b535bc51 EDX: 00000021 Jul 7 16:23:40 p2800 kernel: [18148.112012] ESI: 00000001 EDI: f192e900 EBP: ece7bfb0 ESP: ece7bfa4 Jul 7 16:23:40 p2800 kernel: [18148.112012] DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068 Jul 7 16:23:40 p2800 kernel: [18148.112012] CR0: 80050033 CR2: b302b680 CR3: 307eb000 CR4: 000007f0 Jul 7 16:23:40 p2800 kernel: [18148.112012] DR0: 00000000 DR1: 00000000 DR2: 00000000 DR3: 00000000 Jul 7 16:23:40 p2800 kernel: [18148.112012] DR6: ffff0ff0 DR7: 00000400 Jul 7 16:23:40 p2800 kernel: [18148.112012] Process firefox (pid: 4215, ti=ece7a000 task=f192e900 task.ti=ece7a000) Jul 7 16:23:40 p2800 kernel: [18148.112012] Stack: Jul 7 16:23:40 p2800 kernel: [18148.112012] b535bc51 00000041 00000001 ece7a000 c09a82da 00000000 b6e338d4 00000041 Jul 7 16:23:40 p2800 kernel: [18148.112012] 00000001 00000061 00000021 0000007b 0000007b 00000000 00000033 00000021 Jul 7 16:23:40 p2800 kernel: [18148.112012] b76ea424 00000073 00200292 bf8e427c 0000007b 00000000 00000000 Jul 7 16:23:40 p2800 kernel: [18148.112012] Call Trace: Jul 7 16:23:40 p2800 kernel: [18148.112012] [<c09a82da>] sysenter_audit+0x1e/0x25 Jul 7 16:23:40 p2800 kernel: [18148.112012] Code: 7d fc 89 ec 5d c3 90 8d 74 26 00 c7 43 34 00 00 00 00 b9 50 87 c0 c0 89 da c7 43 38 00 00 00 00 89 f8 e8 34 e8 ff ff 89 c6 eb 91 <0f> 0b 8d b4 26 00 00 00 00 8d bc 27 00 00 00 00 55 89 e5 57 56 Jul 7 16:23:40 p2800 kernel: [18148.112012] EIP: [<c04c0110>] __audit_syscall_entry+0xf0/0x100 SS:ESP 0068:ece7bfa4 Jul 7 16:23:40 p2800 kernel: [18148.214174] ---[ end trace 646e311cf2d23b89 ]---
Todas as respostas (12)
I googled this because I am also at odds with the kernal logs:
Modules included are definately those used to stream video. The kernal panic is an error when out of memory or CPU, I am unsure about acronymns and I know that there is a stack trace but its a computer jargin I do not understand.
Firefox has been tainted however so you may want to reinstall Firefox.If it is cpu, I am not sure if this will work for linux:
But there was also a bug I found with gstreamer, what version are you using?
Hi, Thanks for the reply.
But that's not what tainted means in this context. See http://unix.stackexchange.com/questions/118116/linux-what-is-a-tainted-kernel In this case it just means that firefox is what previously caused the crash. If you look the kernel oops is "invalid opcode: 0000".
In fact a clean install of 39.00 has been tried, without result!
A clean install of 32.0.3 does work, but ONLY if the first run auto update is disabled. This is a clear indication that 39.0 is broken in my execution context.
Is it possible to start Firefox at all ? What about as far as the profile manager by using firefox -P from a terminal ? and firefox -v to confirm the version. (You may to specify filepaths as you have multiple installs)
Sorry I have never used Fedora, but I run Mozilla & Canonical Firefox & Iceweasel using Ubuntu Debian or Puppy. I presume Fedora has it's own builds of Firefox.
The Mozilla crash reports are what we really need.
Have you tried installing a Mozilla build of Firefox ? That is what we support here.
John99 said
Is it possible to start Firefox at all ? What about as far as the profile manager by using firefox -P from a terminal ? and firefox -v to confirm the version. (You may to specify filepaths as you have multiple installs)
NO. A clean install of 39.0 produces the crash. I'm following the instructions to the letter. The binary is firefox-39.0.tar.bz2 is installed in a private directory. I cd to it and type ./firefox and the crash occurs immediately applying the -P switch makes no difference. I definately have the necessary libraries because a non updated 32 works.
I will say again I have always run manually installed versions of firefox like this.
Sorry I have never used Fedora, but I run Mozilla & Canonical Firefox & Iceweasel using Ubuntu Debian or Puppy. I presume Fedora has it's own builds of Firefox.
Yes indeed Fedora has its own package managed version. Which starts via a shell script containing the comment "## The script will setup all the environment voodoo needed to make mozilla work." Obviously, I bypass that.
The Mozilla crash reports are what we really need.
Chance would be a fine thing.
Have you tried installing a Mozilla build of Firefox ? That is what we support here.
That is what I'm doing.
Experience says I should probably stop trying to get 39.0 to work under Fedora 17 (ie a long out of support version of Linux). My policy is and has always been "if it ain't broke don't fix it". While 32.0 works with all the add-ons and plugins, I have just disable updates.
On the other hand I had hoped that someone could help me get to the bottom of a problem which will certainly, if not addressed, come back to haunt us.
How does a kernel oops (illegal instruction 0000) happen as the result of an update? More to the point how to find out why??
Alterado por bradcan em
As a rule of thumb distros will update to support certain distro Firefox builds, and will be expected to be able to run similar Mozilla builds. There have been changes to the requirement for compiling Firefox but I don't see recent changes to the requirements for running Firefox.
- Current https://www.mozilla.org/firefox/39.0/system-requirements/
- Much the same even Fx22
- Detailed and higher requirements for compiling (you should not need this) here
By the way, as mentioned in my preveous post did you try
and firefox -v to confirm the version.
Is possible Fedora is in fact trying to run the distro version ?!
Also where you are trying to run the new Firefox instead of
firefox
Try
type firefox
Are both of the above confirming you are trying to run the version of Firefox and the Firefox location you expect ?
I presumed your OS would be a supported one. After all Linux is usually available free of charge and often requires relatively low spec machines to run on. It is probably not good practice to run an unsupported OS. It certainly is risky using out dated insecure unsupported versions of Firefox
- See (Don't use or ) Install an older version of Firefox
- and look at the security fixes you will miss https://www.mozilla.org/security/known-vulnerabilities/firefox/
Time to bite the bullet and upgrade to a supported OS I think.
Sorry, since we are clearly unable to communicate constructively we should close this thread now.
Thanks for your help.
It may not help that Fedora 17 is a old version now as it came out in end of May 2012. Unless the packages are kept up to date the newer Firefox Release may not always function on a Linux distro from 3+ years ago, more so if it is a LTS one.
Alterado por James em
guigs2 said
But there was also a bug I found with gstreamer, what version are you using?
Sorry I focused on the misunderstanding of the meaning of tainted and failed to see your specific question.
Answer: gstreamer 0.10.36-1.fc17 as installed by yum
$ /usr/bin/gstreamer-properties --gst-version GStreamer Core Library version 0.10.36
Alterado por bradcan em
Sorry I was trying to help. My comment was intended to be constructive. bradcan said
Sorry, since we are clearly unable to communicate constructively we should close this thread now. Thanks for your help.
I have seen a specific instance where there was a problem because the distro firefox command takes precedence over the Mozilla firefox executable and results in the Mozilla Firefox failing to open. I could imagine something similar resulting in a kernel panic on attempting to run Firefox. So the following was anticipated to be diagnostic steps: John99 said
<snip> By the way, as mentioned in my preveous post did you try and firefox -v to confirm the version.
Is possible Fedora is in fact trying to run the distro version ?!
Also where you are trying to run the new Firefox instead of firefox Try type firefox
Are both of the above confirming you are trying to run the version of Firefox and the Firefox location you expect ? <snip>
I sometimes run outdated software and OS s myself but I realise there are difficulties and Risks doing so.
John99 said
<snip> By the way, as mentioned in my preveous post did you try and firefox -v to confirm the version.
Is possible Fedora is in fact trying to run the distro version ?!
Also where you are trying to run the new Firefox instead of firefox Try type firefox
Are both of the above confirming you are trying to run the version of Firefox and the Firefox location you expect ? <snip>
John to set the scene for you; I am 69 years young and have been programming since PDP11 RT11 and RSX, Linux since RedHat 7. I have very good reasons to run old versions of Linux. We used to say "for a simple life don't keep updating, just enter commands at the prompt". This is good advice and I'm sorry that I am somewhat impatient with folks that advise me to upgrade to latest versions. It becoms a (not very) merry go round.
1 - I know which version(s) I'm running. 2 - My only reason to upgrade from 22.0 (the package managed version) is because Adblock Plus does not work with that version. 3 - Version 32.0 is running and has been running since 12-10-2014 4 - Now I find that I am seeing an increasing number of crashes while viewing http://www.bbc.co.uk/iplayer programs. I decided, albeit reluctantly, to try the latest Firefox.
As far as I can tell version 39.0 ALWAYS produces the kernel oops, as I originally posted. No command line switch makes any difference. Not even -help. I haven't tried them all, only the obvious ones, -P, -v, -h
I have followed the online instructions as best I can, downloaded the bz2 file several times and checked that it is virgo intacta. Removed old profiles completely, etc.
It is difficult to be certain if library dependency versions are as required!
I have these:
GTK+ 1.2.10-72.fc17 ??
$ gtk-config --version 1.2.10
GLib 1.2.10-36.fc17 ??
$ ldd --version ldd (GNU libc) 2.15 Copyright (C) 2012 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. Written by Roland McGrath and Ulrich Drepper.
Pango 1.30.0-1.fc17 X.Org
$ Xorg -version
X.Org X Server 1.12.4 Release Date: 2012-08-27 X Protocol Version 11, Revision 0 Build Operating System: 2.6.32-358.2.1.el6.x86_64 Current Operating System: Linux p2800 3.9.10-100.fc17.i686.PAE #1 SMP Sun Jul 14 01:34:14 UTC 2013 i686 Kernel command line: BOOT_IMAGE=/vmlinuz-3.9.10-100.fc17.i686.PAE root=UUID=d584a32a-588d-48a2-8f2b-1e44269f2152 ro rd.md=0 rd.lvm=0 rd.dm=0 SYSFONT=latarcyrheb-sun16 KEYTABLE=uk rd.luks=0 LANG=en_GB.UTF-8 rhgb quiet Build Date: 17 April 2013 04:53:44AM Build ID: xorg-x11-server 1.12.4-7.fc17
libstdc++ 4.7.2-2.fc17
Looks like I have some version problems with GTK+ and GLib. Were there changes from 32.0 to 39.0?
If you would confirm one of these is problematic I will have a go at a source install of the latest version(s).
However, please note that GLib version update is likely problematic in it self. (see http://www.tldp.org/HOWTO/html_single/Glibc-Install-HOWTO/) ~J99 I fixed link. Trailing bracket appends http://www.tldp.org/HOWTO/html_single/Glibc-Install-HOWTO )
Thanks
Alterado por John99 em
Not a direct answer but my best suggestion:
- Try installing Mozilla Fx22 (? then Fx37) does that install and run ?
That at least proves or otherwise that Mozilla builds of Firefox install Rather than trying to look for documentation resort to an empirical trial and error method. To me it sounds that rather than it being a crash on startup of a newly installed Fx39 you are getting the kernel panic in the install process.
- Fx39 presumably has created no profile of its own ?
If you would confirm one of these is problematic I will have a go at a source install of the latest version(s).
John99 #answer-752528 said
... There have been changes to the requirement for compiling Firefox but I don't see recent changes to the requirements for running Firefox....
- Current https://www.mozilla.org/firefox/39.0/system-requirements/
- Much the same even Fx22
- Detailed and higher requirements for compiling (you should not need this) here
Mozilla probably are not very thorough with keeping documentation, updated, and I am no expert at finding it but the links I posted suggest the requirements have not changed. However I do not rule out the requirements for compiling were updated, but not the system requirement for running Firefox. I did note the remark for compiling on Ubuntu
Firefox codebase now relies on features of C++11 that are only supported in GCC 4.7 or later. (Firefox 37 requires GCC 4.6 or later).
If you follow the instructions i linked to to meet dependencies to compile Firefox that presumably ensues you will have met requirements to run Firefox. The top of the page I linked to had
- https://developer.mozilla.org/en-US/docs/Mozilla/Developer_guide/Build_Instructions/Linux_Prerequisites#All_Distros_-_One_Line_Bootstrap_Command
But I am not sure that is applicable to a basic Fedora
You would probably be better asking elsewhere about the kernel panic. Then again you may run into problems getting answers because an unsupported OS & tainted kernel could mean trying to find a specialist discussion.
Just to be 100% clear about what tainted means. Please see this document: https://www.kernel.org/doc/Documentation/oops-tracing.txt
It does NOT mean tainted as in the CGI sense (ie data is tainted if came from an untrusted source) and it has nothing whatever to do with broken files.
Snip from the above kernel document: "Some oops reports contain the string 'Tainted: ' after the program counter. This indicates that the kernel has been tainted by some mechanism."
In my case there are 2 consecutive oops reports. Both the result of $ ./firefox -v in the 39.0 install/firefox directory. The first contains the line: "Jul 15 16:40:17 p2800 kernel: [ 1224.491010] Pid: 1368, comm: firefox Not tainted 3.9.10-100.fc17.i686.PAE"
The second and subsequent contain the line: "Jul 15 16:40:34 p2800 kernel: [ 1241.160992] Pid: 1371, comm: firefox Tainted: G D 3.9.10-100.fc17.i686.PAE" in which:
G means all modules loaded have a GPL or compatible license. D means the kernel has died recently. (well obviously)
That's the full extent of my understanding tainted and confusing, but I'm sure kernel developers have good reason.
John99 said Firefox codebase now relies on features of C++11 that are only supported in GCC 4.7 or later. (Firefox 37 requires GCC 4.6 or later).
I regularly run gcc 4.7.2-2, but compiler versions don't matter. The cruncher is the glibc version! As I said before manual installation of non package managed versions of glibc is a very dangerous and time consuming prospect.
Before attempting any such I need to know for certain which version of glibc is a requirement?
Alterado por bradcan em