Problems with nullfs on 9.0-RC1

[Moritz Wilhelmy]

Good evening,

I have been experiencing problems with accessing a nullfs which has been
mounted inside a nullfs. This is meant to be a very elaborate way to
exchange files between my jails :-)

I have following entries in /etc/fstab on the host system, which runs
FreeBSD 9.0-RC1 from a few days ago:

/usr/jails/foo/share		/shares/foo		nullfs	ro	0	0
/shares 			/usr/jails/foo/shares	nullfs	ro	0	0
/usr/jails/bar/share		/shares/bar		nullfs	ro	0	0
/shares				/usr/jails/bar/shares	nullfs	ro	0	0
etc.

If I create a file in the bar jail in /share, let's call it zot, it is
available on the host system in /shares/bar, but not inside
/usr/jails/foo/shares/bar or /usr/jails/bar/shares/bar.

My syslog is full of messages like these:

Nov  7 16:44:15 machine kernel: null_bypass: vop_getattr
Nov  7 16:44:15 machine kernel: null_bypass: vop_readdir
Nov  7 16:44:15 machine kernel: null_bypass: vop_readdir
Nov  7 16:44:15 machine kernel: null_bypass: vop_close
Nov  7 16:44:15 machine kernel: null_bypass: vop_access
Nov  7 16:44:15 machine kernel: null_bypass: vop_open
Nov  7 16:44:15 machine kernel: null_bypass: vop_getattr
Nov  7 16:44:15 machine kernel: null_bypass: vop_read
Nov  7 16:44:15 machine kernel: null_bypass: vop_close
Nov  7 16:44:15 machine kernel: null_bypass: vop_access
Nov  7 16:44:15 machine kernel: null_bypass: vop_open
Nov  7 16:44:15 machine kernel: null_bypass: vop_getattr
Nov  7 16:44:15 machine kernel: null_bypass: vop_read
Nov  7 16:44:15 machine kernel: null_bypass: vop_close
Nov  7 16:44:15 machine kernel: null_bypass: vop_getattr
Nov  7 16:44:15 machine kernel: null_bypass: vop_access
Nov  7 16:44:15 machine kernel: null_bypass: vop_open
Nov  7 16:44:15 machine kernel: null_bypass: vop_getattr
Nov  7 16:44:15 machine kernel: null_bypass: vop_readdir
Nov  7 16:44:15 machine kernel: null_bypass: vop_readdir
Nov  7 16:44:15 machine kernel: null_bypass: vop_close
Nov  7 16:44:15 machine kernel: null_bypass: vop_getattr
Nov  7 16:44:15 machine last message repeated 2 times
Nov  7 16:44:15 machine kernel: null_bypass: vop_access

I guess I'm either doing something very stupid here, or this is a bug?
The manpage doesn't mention anything about nullfs still being unstable
(I read it used to be, with a huge bugs section in the manpage,
which read like this: "SLIPPERY WHEN WET, BEWARE OF DOG").
Has anyone run into similar problems?


Best regards,

Moritz