Re: panic: vput: negative ref cnt

Am 05.04.2009 um 13:15 schrieb Mark Powell:

> On Sun, 5 Apr 2009, Stefan Bethke wrote:
>
> Stefan,
>  Thanks for the input.
>
>> Am 04.04.2009 um 13:55 schrieb Mark Powell:
>>> My script takes a zfs snapshot, mounts the snapshot, backs it up  
>>> using star, then umounts the snapshot. The panic seems to occur on  
>>> the umount.
>>
>> This is a different panic from the one I was getting, but appears  
>> to occur under the same circumstances.
>>
>> I managed to avoid the panic by keeping the snapshot until the next  
>> run of my backup script, and not unmounting anything.
>
>  Yeah, I got around it by avoiding the unmount.
>  My script dates from 7-STABLE days, where a zfs bug caused the  
> automounted snapdir e.g. .zfs/snapshot/star_L0_2009-04-04-16:57, to  
> be missing a '..' entry. That prevented star from detecting that I  
> was performing a true full backup.  Thus my script directly mounted  
> the snapshot to get a proper fs with a '..' entry and unmounted it  
> when finished.
>  Now the included version of zfs in 8-CURRENT doesn't have that bug  
> any longer. So I've reverted to backing up the automounted snapdir.  
> Consequently as I don't perform a mount I don't have to unmount  
> either, effectively working around the bug.
>
>> See <http://lists.freebsd.org/pipermail/freebsd-current/2009-February/003513.html 
>> >
>
> Do you know if a PR has been submitted for this one?

When I lasted searched for it about six weeks ago, I couldn't find  
any.  I wanted to reproduce the panic in VMware first to gather more  
details, but I've been unsuccessful in triggering it there at all.   
Maybe I need to switch my production boxes over to VMware :-)


Stefan

-- 
Stefan Bethke <stb_at_lassitu.de>   Fon +49 151 14070811