CIFS Kernel Panic

[maglaubig]

I'm able to replicate this one on 8.2-RELEASE-p9. Whenever I right click and select properties from a Win7 machine on multiple files, or on the actual CIFS share I get a kernel panic and the system crashes and has to be rebooted. Anyone else seen this behavior?

I've taken the steps to run memory testers that didn't reveal any issues and still no smoking gun. I could see this crashing samba first, but a kernel panic?

 

[itiec]

i've got the same sympton. and i have a rsync configuration which seems to be the trigger.
What is weird: no log messages of the restart. Just the timestamp in the dmesg.log or messages log files.

 

[William Grzybowski]

Well, obviously CIFS is not causing kernel panic, it runs in userland mode, but it is doing something that triggers it.

What is the kernel panic message? I would bet this is related to memory and kernel tunning (or the lack of it)

 

[maglaubig]

I thought this kind of operation wasn't in the kernel. I'll have to reproduce tonight and re-post results. I know there is a method to get /var to persist instead of having it go to the USB stick. Can I get an assist on a how to set that up? I do have a volume with UFS (with an existing file system I can't overwrite) as well as an ability to get a ZFS device extent setup.

 

[William Grzybowski]

/var persistent is pretty much useless in kernel panics

 

[maglaubig]

Short of writing down the error, isn't there some way to get a crash dump or something? I've seen messages that I'm out of storage space when attempting to dump this, but I figured that's because I was out of space on the 4GB USB stick I have this running off of.

 

[maglaubig]

Looks like I may have rediscovered a bug thought only to be in NFS here.

I find it highly suspect that I get the same trap and same instruction pointer on two separate systems, and having it matching yet a third from another post. Seems like the only fix is to update to 8.3? Waiting to hear back.

 

[William Grzybowski]

Well, if the issue is the same then yes, the fix is only available for 8.3.

 

[maglaubig]

This appears to have been the same issue. The method I could use to consistently crash the system before has disappeared after upgrading to 8.3.