#44 — 'zfs get all' core dumps if /mypool/a is bind mounted

Cannot reproduce.

Can you state specifics?

git log -1
uname -a
gcc -v
fusermount -V

contents of /etc/zfs/zfsrc

zfs-fuse launch command (e.g. startup controlled with --no-kstat-mount -a 0 -e 0 --no-daemon)

HERE IS MY TESTRUN:
-----------------------------------
root@karmic:~# ./devsk.sh
forced clean relaunch of zfs-fuse
======== building pool l named mypool at /tmp/mypool_blk =================================
zpool create -O mountpoint=/mypool mypool /tmp/mypool_blk/za1
  pool: mypool
 state: ONLINE
 scrub: none requested
config:

    NAME STATE READ WRITE CKSUM
    mypool ONLINE 0 0 0
      /tmp/mypool_blk/za1 ONLINE 0 0 0

errors: No known data errors
NAME USED AVAIL REFER MOUNTPOINT
mypool 76.5K 1.95G 18K /mypool
root@karmic:~# zfs create mypool/a
root@karmic:~# zfs mount -a
root@karmic:~# mkdir -p /mnt/temp && mount -o bind /mypool/a /mnt/temp
root@karmic:~# zfs get all | wc -l
85
root@karmic:~# zfs get all mypool/a | wc -l
43
root@karmic:~# zpool status
  pool: mypool
 state: ONLINE
 scrub: none requested
config:

    NAME STATE READ WRITE CKSUM
    mypool ONLINE 0 0 0
      /tmp/mypool_blk/za1 ONLINE 0 0 0

errors: No known data errors
root@karmic:~# mount | tail -3
mypool on /mypool type fuse (rw,allow_other)
mypool/a on /mypool/a type fuse (rw,allow_other)
/mypool/a on /mnt/temp type none (rw,bind)
root@karmic:~#

cannot reproduce with 0.5.0 or 0.6.0, or testing.

I _did_ reproduce it with 0.6.9_beta2:

[...]
root@karmic:~# zfs get all | wc -l
zfs: lib/libavl/avl.c:634: avl_add: Assertion `0' failed.
7
root@karmic:~# zfs get all mypool/a | wc -l
zfs: lib/libavl/avl.c:634: avl_add: Assertion `0' failed.

I can confirm this to be a duplicate of issue #42 and as such has been fixed in the testing branch. Kindly retest with that version

Closing after posting back to the list

though the symptoms cannot be reproduced since cc745d5b84, it is agreed that this one warrants more investigation.

Emmanuel has a fix that addresses a more direct cause for the bug (3973ae03391cc6359d80383143dbe54e23075f6b)

Emmanuel was right: the problem was still occurring (but I apparently tested using debug=0 so I didn't see it because I wasn't crashing the daemon anymore)

His patch is going to be included in testing as of tomorrow - pending his review of my latest dev branch

It is here for the moment
http://zfs-fuse.sehe.nl/[…]/sehe

fixed in 0.7.0 (also in 0.6.9-maint)