panic: __lockmgr_args: unknown lockmgr request 0x0

[Garrett Cooper]

On Thu, Jul 24, 2008 at 8:17 AM, Attilio Rao <attilio at freebsd.org> wrote:
> 2008/7/23 Rink Springer <rink at freebsd.org>:
>> On Sun, Jul 20, 2008 at 06:26:04PM +0200, Attilio Rao wrote:
>>> 2008/7/20, Attilio Rao <attilio at freebsd.org>:
>>> > 2008/7/20, Lothar Braun <lothar at lobraun.de>:
>>> >
>>> > > Hi Attilio,
>>> >  >
>>> >  >
>>> >  >
>>> >  > > can you please try this on the top of -CURRENT:
>>> >  > > http://www.freebsd.org/~attilio/xfs2.diff
>>> >  > >
>>> >  >
>>> >  >  Thank you for the patch. The panic and the dead lock disappeard, but there
>>> >  > is a new problem insteed. The commands
>>> >  >
>>> >  >  mkfs.xfs /dev/ad8s4
>>> >  >  mount -t xfs /dev/ad8s4 /home
>>> >  >  mkdir /home/lothar
>>> >  >  chown lothar:lothar /home/lothar
>>> >
>>> >
>>> > For what I remind, it is likely XFS is still not ready for writing.
>>> >  This means you should only use it in read-only.
>>>
>>> Speaking of which, I think we should mark it again like a read-only fs
>>> until writing is not 100% ready.
>>
>> NTFS suffers from the same issue; it 'kind of' supports writes. The
>> result is that it supports writes in so limited circumstances that the
>> write support is mostly useless (and it even tends to lead to panics...)
>>
>> I think a better solution is to mount such filesystems r/o by default,
>> and only mount them r/w if explicitely asked to do so, for example by '-o
>> rw' - it would make things a lot clearer for our users when trying to
>> use filesystems, and brave souls are always welcome to force r/w that
>> way.
>>
>> What do you think?
>
> As long as you state that the write support is almost useless, I think
> the better thing is that we should simply drop the write support for
> the moment (and leaving the implementation there, of course, so that
> interested hackers can keep solidifying the support).
>
> Thanks,
> Attilio

Knobs per src.conf for fs experimental functionality?
-Garrett