Panic after updating to U4

[Jason Kotecki]

This morning I upgraded my FreeNAS from 11.2-U3 to U4. Now it throws a Panic and reboots every 3-4 minutes. I re-installed 11.2-U3 and restored my config, but it still won't stay up. I'm not sure if it's because I also did the zpool upgrade or not. I attached a screenshot showing the panic. A Google search has not turned up an answer for me. I would greatly appreciate any help trying to figure this out.

Thanks,
Jason

 

[anodos]

That's a panic related to the ACL on file. If it's a single file and we can figure out which one it is, then I can probably help you repair it. Are you able to get a system debug? What services are you using? Are you able to send me an old config file?

Are you able to perform a `zpool status`?

 

[Jason Kotecki]

To document the fix for me in case someone else encounters this, here's what I did to resolve it.

I had to stop samba (#service samba_server onestop)
Then I ran the following command to traverse the filesystem, when it panic'd I took a screenshot so I knew what the offending file was.
#find -x `mount -t zfs | awk '{print $3}'` -type d -exec ls -la {} \;

I may have been able to fix the ACL, but I decided to just delete the file instead and then restore it from a backup. After removing the file, the system is back up and stable. Thanks Anodos for helping me to get this resolved.

 

[seanm]

Disconcerting that a messed up ACL on one file should panic the entire OS. If I were you, I'd file a bug. I tried to find the code for the acl_from_aces() function in OpenZFS, but could not, maybe it's part of FreeNAS' codebase?

 

[anodos]

Disconcerting that a messed up ACL on one file should panic the entire OS. If I were you, I'd file a bug. I tried to find the code for the acl_from_aces() function in OpenZFS, but could not, maybe it's part of FreeNAS' codebase?

No. It's part of FreeBSD / ZFS proper. I'm investigating the issue and will file a bug ticket / work on it.

 

[anodos]

It's unclear how the invalid a_type occurred. It is impossible to write an ac_entry_type of 0x4000 in FreeBSD / FreeNAS. The issue occurred subsequent to a server power event, but the dataset was imported into Corral at one point and had the private ZFS property "ACL" set. Corral had a custom ZFS aclmode property 'discard_chmod', which (like many things in corral) was probably not thoroughly tested. I am leaning more toward this being related to the power event (although I don't like that answer). And don't see any reason for alarm at this point.