zfs panic "evicting znode"

[Charles Sprickman]

Howdy,

I have a box running 8.0-RELEASE that recently started panicing every few 
hours with the following message:

panic: evicting znode 0xa1eafe80

cpuid = 0
Uptime: 30m56s
Physical memory: 2034 MB
Dumping 569 MB (counts down to 458, then the box freezes hard)

The dump doesn't finish, so there's nothing for savecore to grab.

It's a very basic zfs config - two scsi drives in a mirror:

[root at h21 /usr/local/etc/pdns]# zpool status
   pool: zroot
  state: ONLINE
  scrub: scrub completed after 0h5m with 0 errors on Tue Jun 15 21:32:11 
2010
config:

         NAME           STATE     READ WRITE CKSUM
         zroot          ONLINE       0     0     0
           mirror       ONLINE       0     0     0
             gpt/disk0  ONLINE       0     0     0
             gpt/disk1  ONLINE       0     0     0

errors: No known data errors

This is running on an older dual-xeon (1.8GHz/32 bit) supermicro 1U server 
w/2GB of RAM.  The following zfs tunables are set in loader.conf:

[root at h21 /usr/local/etc/pdns]# cat /boot/loader.conf
zfs_load="YES"
vm.kmem_size_max="1000M"
vm.kmem_size="1000M"
vfs.zfs.arc_max="400M"
vfs.root.mountfrom="zfs:zroot"

Google is showing me nothing on this panic except for hits on the source 
code that actuall contains the panic message.

Any hints as to what this means?

Here's zdb output:

[root at h21 /usr/local/adm/bin]# zdb zroot
     version=13
     name='zroot'
     state=0
     txg=267691
     pool_guid=12059945251392529754
     hostid=1898595607
     hostname='h21.biglist.com'
     vdev_tree
         type='root'
         id=0
         guid=12059945251392529754
         children[0]
                 type='mirror'
                 id=0
                 guid=14682767316808875040
                 metaslab_array=23
                 metaslab_shift=30
                 ashift=9
                 asize=142515896320
                 is_log=0
                 children[0]
                         type='disk'
                         id=0
                         guid=11600930948623447097
                         path='/dev/gpt/disk0'
                         whole_disk=0
                 children[1]
                         type='disk'
                         id=1
                         guid=4279842263738814989
                         path='/dev/gpt/disk1'
                         whole_disk=0
Assertion failed: (rwlp->rw_count == 0), file 
/usr/src/cddl/lib/libzpool/../../../cddl/contrib/opensolaris/lib/libzpool/common/kernel.c, 
line 203.
Abort trap: 6 (core dumped)

Note: the coredump on zdb only occurs if the pool's name is specified, 
otherwise the output is identical.  I'm guessing this is just an error in 
zdb since it happens on every 8.0 box I've got.

Any help is appreciated - we're really looking to get zfs in production, 
but this problem is a bit odd, and it's not like we can fsck to fix any 
possible problems with the fs.

Thanks,

Charles