excessive samba backtraces; overlooked configuration? Known problem?

[Kevin Hunter Kesling]

Hello Forum,

We are getting ready to transition from our previous NAS solution to the iXSystems-acquired FreeNAS box we bought last month. I'm currently in the process of copying data to the new server, but am noting what seems to be an excessive number of samba coredumps. Should I be worried?

For the past couple of weeks, I've received a number of security emails notifying me of smbd core dumps. Everything seemed to be working when I went to check it, so I assumed it was a known problem with a well-known response: ignore and restart the errant process. However, I finally logged into the console today and noted 27,000+ backtraces in /var/log/samba4, over the course of 3 weeks. As I'm currently the only user of the new box, I'm worried that I've overlooked an obvious detail or that we possibly have a hardware error.

My method of copying files/triggering these backtraces is a Bash/rsync script, morally equivalent to:

for PROJECT in ...; do

echo -e "\n-----\nCopying project $PROJECT"
rsync -PHhaAXrv --exclude=.AppleDouble --exclude="*.tmp" ... --delete-excluded --delete-before --delete --inplace "$OLD/$PROJECT" "$NEW/"​

done​

Am I doing something obviously incorrect? Is this a known problem? Have I encountered a bug? We are currently running FreeNAS-9.3-STABLE-201506292130.

Thanks,

Kevin

P.S. Please find attached a representative backtrace.

 

[anodos]

It might be worthwhile to post a bug report on bugs.freenas.org.

It might be that your script is reacting poorly to having zfs aclmode set to "restricted". When aclmode is set in this way it prevents chmod operations. See here - https://www.illumos.org/issues/3254. Your rsync script is probably trying to perform some chmod operations, and samba - being what it is - barfs on itself.

That being said, post a bug report.

 

[Kevin Hunter Kesling]

I'm new enough to FreeNAS that I don't know how to check if the aclmode is set to restricted. Is there a zfs command line option that is not jumping out at me? Something in the WUI? (We're using v9.3.)

In any event, I've submitted a bug report -- thanks for that direction. Let's see how it goes.

 

[Kevin Hunter Kesling]

For those who decide against reading the bug report, it turned the issue was (as expected, frankly!) an issue with what I thought I had to add to the Samba share settings. Namely, Samba provides a number of acl modules for working in different environments, but since NFSv4, ACLs are handled natively. Apparently, the acl modules I used were interacting poorly together with the native environment, so was manifesting as a crash in smbd. So, in short, don't load any of the acl_* modules. Problem avoided.