Home / server / Questions / 479743
In Process
GorillaPatch
Asked: 2013-02-18 11:28:41 +0800 CST

Netatalk 3.0.1 disconnects with Bad file descriptor on FreeBSD

  • 772

I am having trouble with my netatalk server 3.0.1 on FreeBSD-9.1 RELEASE. Netatalk was installed from the ports collection and is sharing some directories over the network.

Update also netatalk version 3.0.2 has the same problems on my machine

afpd.conf

[Global]
; Global server settings

[Homes]
basedir regex = /home
home name = $f - $u 's Home

[testvolume]
path = /var/tmp

[Storage]
path = /tank

[Time Machine Backup]
path=/tank/TM_backup
vol size limit = 2000000
time machine = yes

/tank is in my case the root of a ZFS Pool with the different filesystems underneath it. I connected a mac running Mac OS X 10.7.5 to it and it works perfectly as a Time Machine device. However if I want to use the storage share to create a new aperture library on it Aperture just stalls and cannot create a new library on the server. Looking at the server log it outputs the following:

server log

Feb 17 20:09:09 glacier afpd[3501]: afp_disconnect: primary reconnect failed
Feb 17 20:09:09 glacier afpd[3210]: Terminated disconnected child[3229], client rebooted.
Feb 17 20:09:09 glacier afpd[3505]: AFP3.3 Login by towe
Feb 17 20:09:09 glacier kernel: pid 3501 (afpd), uid 501: exited on signal 6
Feb 17 20:09:09 glacier afpd[3505]: afp_disconnect: trying primary reconnect
Feb 17 20:09:09 glacier afpd[3210]: Reconnect: transfering session to child[3501]
Feb 17 20:09:09 glacier afpd[3210]: read: Bad file descriptor
Feb 17 20:09:09 glacier afpd[3210]: Reconnect: error sending DSI id to child[3501]
Feb 17 20:09:14 glacier afpd[3505]: afp_disconnect: primary reconnect failed
Feb 17 20:09:14 glacier afpd[3210]: Terminated disconnected child[3229], client rebooted.
Feb 17 20:09:14 glacier kernel: pid 3505 (afpd), uid 501: exited on signal 6
Feb 17 20:09:14 glacier afpd[3506]: AFP3.3 Login by towe
Feb 17 20:09:14 glacier afpd[3506]: afp_disconnect: trying primary reconnect
Feb 17 20:09:14 glacier afpd[3210]: Reconnect: transfering session to child[3505]
Feb 17 20:09:14 glacier afpd[3210]: read: Bad file descriptor
Feb 17 20:09:14 glacier afpd[3210]: Reconnect: error sending DSI id to child[3505]

So the server is stuck in a 5 second repeating reconnection loop which always fails.

This is the output of the zfs list command:

NAME                                    USED  AVAIL  REFER  MOUNTPOINT
tank                                    725G  4.62T   682K  /tank
tank/Downloads                         1.61G  4.62T  1.61G  /tank/Downloads
tank/TM_backup                          495G  4.62T   495G  /tank/TM_backup
tank/movies                             169G  4.62T   169G  /tank/movies
tank/music                             26.4G  4.62T  26.4G  /tank/music
tank/pictures                          24.3G  4.62T  24.3G  /tank/pictures

My questions

  1. How can I fix this reconnection loop?
  2. How can I prevent a mounted AFP share from being unmounted after some time of inactivity?

UPDATE What I now found out is that the problem arises only if I write to a RAIDZ filesystem. A plain ZFS volume works.

freebsd

1 Answers

  1. I'm not sure if you're experiencing the same issue, but I had a very similar problem in the recent past (FreeBSD 8 + ZFS + Netatalk). Clients were disconnecting and timing out when reading files on the volume.

    For me the fix was recompiling the netatalk port and disabling the SENDFILE option in the netatalk port. If you have the SENDFILE option enabled, might be worth a shot.

    • 0