Fatal trap 12: page failt while in kernel mode -- error while mounting pool

[gregmans]

Hello,

During the destroy of a dataset i got the following error:

Fatal trap 12: page fault while in kernel mode
Fault virtual address = 0x30
Fault code = supervisor write data, page not present
current process = 0 (zio_free_issue_0)
Stopped at ddt_phys_decref+0x1: subq $0x1,0x30(%rdi)

I now get these kernel panics after every reboot or when i try to import the pool on another system or in single user mode during the mounting of the datasets.

The issue started while trying to delete a Mac OS X Time Machine backup which resided on a dataset. My system froze twice during the 'rm -rf'. As it is headless i did not see the error message. After that i tried a destory dataset and am now getting this error.

I am running FreeNAS 8.3.0 .

Is there a way to import a pool without mounting the datasets? The error always occurs during the mounting of the datasets. I already tried 'zpool import -R -f vol' and 'zpool import -N -f vol' but the '-f' option always gets ignored and thus won't be imported because it was last accessed by another system. I can see with the 'zpool import -R vol' command that my pool is online, but cannot access it.

I really don't know what to do anymore...

 

[cyberjock]

We had someone in IRC make an almost similar mistake. If his example is any sign of how this will turn out, you'll wish you had waited out the rm -rf.

Can you post your server specs and zpool configuration?

 

[gregmans]

FreeBSD crashed during the rm -rf, that is why i thought a 'destory dataset' might help.

Server specs:
AMD Fusion C60 APU dualcore @ 1 GHz
16 GB DDR3 Memory
3 x 3TB WD Red disks

Zpool configuration:
RAIDZ1
zpool v28

That's pretty much all i know by heart.

 

[Dusan]

You may want to try to upgrade to FreeNAS 9.1.1 (or even the 9.2.0 RC). Maybe the newer ZFS code will be able to handle your pool without crashing.

 

[gregmans]

Hello,

I tried FreeNAS 9.2.0 RC2. Unfortunately the newer ZFS code can't handle import my pool either. FreeBSD 9.2 however gives me the following error message "Solaris: WARNING: can't open objset for vol/backup". The rest of the error remains the same.

I did find out that i am able to import my pool using the read-only option (even the corrupt dataset)! So luckily my data isn't lost. Unfortunately i do not have sufficient other storage to be able to copy everything off this pool until i can recreate it.

Do you think there is a way to fix this?

 

[Dusan]

You can try running "zpool import -f -F -n -R /mnt <poolname>". If it says that it is possible to import the pool then run "zpool import -f -F -R /mnt <poolname>" to import it. However, be aware that you may lose some data as this will discard last few transactions.

 

[gregmans]

I have already tried this. FreeBSD always crashes when trying to import my pool except when i use "-o rdonly=on"

 

[Dusan]

Did you really also try the -F option (ZFS recovery)? You only mention -f (force) in your original post.

 

[gregmans]

Yes, i did.

Whenever i import without the "-o rdonly=on", it crashes.
The -F option doesn't change anything.

 

[Dusan]

You may want to try the steps here: http://forums.freenas.org/threads/zfs-has-failed-you.11951/

 

[cyberjock]

I don't think that thread is going to help him. Partly because everything he tried is already on that thread.

He's going to have to pay someone for ZFS recovery most likely. I'd offer my services, but I'm not even sure I can fix his mistake. Not to mention recovery is far from cheap.

 

[gregmans]

I kind of gave up fixing my pool. I have imported my pool in read-only mode and am currently migrating my data to different disks.

But i still wonder if there was any way to fix the old pool, instead of recovering the data elsewhere. And what caused this issue?