Re: Regression with ZFS on recent current

From: Yuri Pankov <yuri.pankov_at_gmail.com>
Date: Thu, 18 Oct 2012 08:38:07 +0400
On Thu, 18 Oct 2012 00:45:27 +0300, Andriy Gapon wrote:
> on 18/10/2012 00:05 Andriy Gapon said the following:
>> on 17/10/2012 20:24 Guido Falsi said the following:
>>> Hi!
>>>
>>> While testing portshaker I found a regression in latest current.
>>>
>>> portshaker depends on ZFS(when enabling ZFS support) remounting
>>> filesystems on the fly when changing the mountpoint property. This is
>>> working as expected on 9.x while on 10-current is not working. Problem
>>> has appeared sometime before October 3rd.
>>
>> Can you clarify what you mean but this date?
>> That the problem didn't exist in head until recently?  Can you name a date when
>> the problem didn't exists in head then?
>
> Oh, hmm, not sure if it is a bug that we imported from upstream or a porting
> mistake.  Before r238391 do_prefix variable in zfs_prop_set (in
> cddl/contrib/opensolaris/lib/libzfs/common/libzfs_dataset.c) was always initialized.
> Now it is either explicitly set to B_FALSE or left alone uninitialized.
>
> http://svnweb.freebsd.org/base/head/cddl/contrib/opensolaris/lib/libzfs/common/libzfs_dataset.c?r1=237119&r2=238391&pathrev=238391

Not sure if it's helpful, but I can't reproduce the issue with pretty 
recent illumos-gate (at r13853):

sirius:root:~# zfs create -o mountpoint=/test rpool/test
sirius:root:~# touch /test/baz
sirius:root:~# zfs set mountpoint=/foo rpool/test
sirius:root:~# ls -ld /foo/
drwxr-xr-x   2 root     root           3 Oct 18 08:36 /foo/
sirius:root:~# ls -l /foo/
total 1
-rw-r--r--   1 root     root           0 Oct 18 08:36 baz
sirius:root:~# zfs list rpool/test
NAME         USED  AVAIL  REFER  MOUNTPOINT
rpool/test    31K   737G    31K  /foo

>>> Here is some sample output from a9.x machine and a 10-current one:
>>>
>>> root_at_micro:~ [0]# uname -a
>>> FreeBSD micro.madpilot.net 9.1-PRERELEASE FreeBSD 9.1-PRERELEASE #0
>>> r239180: Sat Aug 11 00:14:47 CEST 2012
>>> root_at_micro.madpilot.net:/usr/obj/usr/src/sys/MICRO  amd64
>>> root_at_micro:~ [0]# zfs create tank/test
>>> root_at_micro:~ [0]# touch /test/baz
>>> root_at_micro:~ [0]# zfs set mountpoint=/foo tank/test
>>> root_at_micro:~ [0]# ls -ld /foo
>>> drwxr-xr-x  2 root  wheel  3 Oct 17 19:18 /foo
>>> root_at_micro:~ [0]# ls -l /foo/
>>> total 1
>>> -rw-r--r--  1 root  wheel  0 Oct 17 19:18 baz
>>> root_at_micro:~ [0]# zfs list tank/test
>>> NAME        USED  AVAIL  REFER  MOUNTPOINT
>>> tank/test   144K   826G   144K  /foo
>>>
>>>
>>> root_at_marvin:~ [0]# uname -a
>>> FreeBSD marvin.madpilot.net 10.0-CURRENT FreeBSD 10.0-CURRENT #7
>>> r241638: Wed Oct 17 14:33:18 CEST 2012
>>> root_at_marvin.madpilot.net:/usr/obj/usr/src/sys/MARVIN  amd64
>>> root_at_marvin:~ [0]# zfs create tank/test
>>> root_at_marvin:~ [0]# touch /test/baz
>>> root_at_marvin:~ [0]# zfs set mountpoint=/foo tank/test
>>> root_at_marvin:~ [0]# ls -ld /foo
>>> ls: /foo: No such file or directory
>>> root_at_marvin:~ [1]# ls -l /foo/
>>> ls: /foo/: No such file or directory
>>> root_at_marvin:~ [1]# ls -ld /test
>>> drwxr-xr-x  2 root  wheel  3 Oct 17 19:19 /test
>>> root_at_marvin:~ [0]# zfs list tank/test
>>> NAME        USED  AVAIL  REFER  MOUNTPOINT
>>> tank/test    31K   240G    31K  /foo
>>>
>>>
>>> Is this known? is the new behavior expected?
>>>
>>> Am I doing something stupid?
>>>
>>> Thanks in advance
>>>
>>
>>
>
>
Received on Thu Oct 18 2012 - 02:38:11 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:40:31 UTC