Fatal trap 12: page fault while in kernel mode

[MotorSport]

I have this error in this version of FreeNAS-8.3.2-ALPHA-r13281-x64.

 

[tikos]

Hi,

I just got a HP N40L Microserver, with the following specs:

CPU: AMD Turion II Neo N40L (1.5 Ghz) K625
Bus speed: 200 MHz, HT Clock: 1600 MHz, Multiplier: 7.5x
Chipset: AMD RS785E/SB820M
RAM: 2GB ECC PC3-10600E Unbuffered DDR3 ECC @ 800MHz
Graphics: Onboard AMD HD4200 VGA: 128MB, 1920x1200 @ 60hz
Storage: 250GB SATA
NIC: Onboard NC107i PCI Express Gigabit 10/100/1000
7 X USB 2.0 Ports
No CD/DVD device

I installed Win8 Pro, and then through VirtualBox 4.2.6, I tried to install FreeNAS 8.3.0.
So, I created a virtual machine (1GB RAM, 8GB Virtual Hard Disk) and I mounted the FreeNAS 8.3.0 64bit ISO image to it, in order to install it.
When I started the virtual machine, I got the following lines:

aesni0: No AESNI support.
acpi0: could not allocate interrupt
ACPI Exception: AE_ALREADY_EXISTS, Unable to install System Control Interrupt handler (20101013/evevent-257)
acpi0: Could not enable ACPI: AE_ALREADY_EXISTS
device_attach: acpi0 attach returned 6

Fatal trap 12: page fault while in kernel mode
cpuid=0; apic id = 00
fault virtual address = 0x88
fault code = supervisor write data, page not present
instruction pointer = 0x20:0xfffffff80892f69
stack pointer = 0x28:0xfffffff8178cbc0
frame pointer = 0x28:0xfffffff8178cc00
code segment = base 0x0, limit 0xfffff, type 0x1b
= DPL 0, press 1, long 1, def32 0, gran 1
processor eflags = interrupt enabled, resume, IOPL = 0
current process = 0 (swapper)
[thread pid 0 tid 100000 ]
Stopped at acpi_install_wakeup_handler+0xd9; movq %r15,0x88(%r12)
db>

Any ideas???

 

[survive]

Hi tikos,

Here's the first thing google returned. looks like it applies in your situation:

http://lists.freebsd.org/pipermail/freebsd-current/2011-March/023710.html

Note, 1GB of memory is way to little for FreeNAS. Best thing you could do is to go out & score some 4 or 8GB DIMMS for the system.

-Will

 

[tikos]

survive
Although I've alreadey found the solution, thanks a lot for your immediate response.

 

[nando1485]

Tikos what was your solution?

 

[William Grzybowski]

Tikos what was your solution?

Add more RAM, use x64

 

[nando1485]

thanks for responding!

 

[sysfu]

I just got this same error on an HP Microserver N36L with 8GB ECC ram running 8.3.1 release x64. The problem is easily reproducible.

I can reliably make the system kernel panic with "Fatal trap 12: page fault while in kernel mode" by manually rm -rf'ing certain directories that were created in my home dir dataset with the PCBSD life-preserver backup utility. My gut is that is has something to do with symlinks.

I'll try destroying the whole dataset next to see if that will kill the data w/out a kernel panic.

Update: destroying the dataset succeeded where rm -rf failed.

 

[sysfu]

OK, this *definitely* has to do with symlinks. I ran into the problem again today, and while in the process of whittling down this problem directory, I discovered that the system would kernel panic when removing the /home/user/bin directory inside the PC-BSD life-preserver backup directory.

Guess what's inside the bin dir? Lots of symlinks. (see listing below)

So there appears to be some bug with the way that FreeNAS/ZFS is handling file delete operations on those symlinks.

If there are any alternative tools for removing files or symbolic links other than the rm command, I'm all ears. It would be a huge hassle to have to back, delete and restore this dataset just to be able to delete a few dozen lousy symlinks.

[seth@nas] ~/life-preserver/sure/bin> ls
./ ebook-viewer@ launchy@ proxytrack@ webhttrack@
../ epub-fix@ libreoffice@ pwcsnap@ weechat-curses@
RetroShare@ evolution@ lobase@ pwcview@ widl@
bitcoin-qt@ evolution-settings@ localc@ rdiffdir@ wine@
calibre@ fetch-ebook-metadata@ lodraw@ regedit@ wine64-nvidiasetup@
calibre-complete@ filezilla@ loffice@ regsvr32@ wineboot@
calibre-customize@ firefox@ lofromtemplate@ remmina@ winebuild@
calibre-debug@ function_grep.pl@ loimpress@ rhythmbox@ winecfg@
calibre-mount-helper@ fzputtygen@ lomath@ rhythmbox-client@ wineconsole@
calibre-parallel@ fzsftp@ loweb@ skype@ winecpp@
calibre-server@ galculator@ lowriter@ soffice@ winedbg@
calibre-smtp@ globaltime@ lrf2lrs@ swine@ winedump@
calibre-uninstall@ gmplayer@ lrfviewer@ swinecli@ winefile@
calibredb@ gnc-fq-check@ lrs2lrf@ swinerun@ wineg++@
chrome@ gnc-fq-dump@ markdown-calibre@ tmux@ winegcc@
clipgrab@ gnc-fq-helper@ midori@ tor@ winemaker@
deluge@ gnc-fq-update@ mplayer@ tor-gencert@ winemine@
deluge-console@ gnome-mplayer@ mplayer.sh@ tor-resolve@ winepath@
deluge-gtk@ gnucash@ msiexec@ torify@ wineserver@
deluge-web@ gnucash-env@ msiexec.sh@ trafshow@ wmc@
deluged@ gnucash-make-guids@ mumble@ tz_convert@ wol@
duplicity@ gnucash-valgrind@ notepad@ uninstall-opera@ wol-bootptab@
dvx* htsserver@ opera@ unopkg@ wol-dhcpdconf@
ebook-convert@ httrack@ orage@ update-gnucash-gconf@ wrc@
ebook-device@ ikea@ pidgin@ vidalia@ xombrero@
ebook-meta@ ikec@ pinta@ web2disk@ xombrero-bin@

- - - Updated - - -

Update: I tried removing some symlinks one by one using the unlink command, and this worked. So I tried to automate the process using the find command and the system kernel panicked again.

Code:

find * -exec unlink {} \;

I tried again using the -ok switch to unlink them one by one. I pretty much pared it down to one problem symlink

lrwxr-xr-x 1 seth users 41 Feb 21 12:00 gnc-fq-check@ -> /usr/pbi/gnucash-amd64/.sbin/gnc-fq-check

I have no idea what is special about this symlink, but this is the one that consistently triggers the fatal trap kernel panic.

This FreeBSD forum thread might be the same issue.

 

[sysfu]

I'm still having this issue after upgrading to FreeNAS-9.1.1-RELEASE-x64. IIRC the 9.1 release also upgrades zpool version to v5000.

Currently the only effective workaround I know of is to backup the files in the dataset, delete the dataset, create a new dataset, then restore the files. A major hassle.

I'd like to help troubleshoot the bug to find a fix but this is my production system so it's tough to do, and I don't want to corrupt any data.

If anyone else using FreeNAS has been suffering from this problem, please post on this thread so we can compare notes.

 

[warri]

IIRC the 9.1 release also upgrades zpool version to v5000.

It includes v5000, but you have to upgrade the pool manually with zpool upgrade. Warning: No going back ;)

 

[cyberjock]

It includes v5000, but you have to upgrade the pool manually with zpool upgrade. Warning: No going back ;)

As discussed in the manual. :)

And I'm sure srf21c read the manual so he wouldn't ask that question ;)

 

[sysfu]

I should have been more clear on the zpool version. I did in fact upgrade all pools to v5000 using the zpool upgrade command. I was having difficulty however verifying the pool versions.

# zfs get version show a value of just '5', so I wasn't sure that this was short for v5000.

 

[warri]

You might want to check this thread: No zpool version after zpool ugrade on Freenas 9.1.1
You get your current zpool version by zpool version, not zfs version. But as discussed in the linked thread, it'll be empty for a pool with v5000.

Sorry for the offtopic, if the problem persists you could always open a bug report and see if the developers have a solution.