sysinstall creates corrupt filesystems after repartitioning

Eric Anderson anderson at freebsd.org
Fri Mar 2 15:59:32 UTC 2007


On 03/02/07 09:37, Steven Hartland wrote:
> Jeremy Chadwick wrote:
>> On Fri, Mar 02, 2007 at 08:11:52AM -0600, Eric Anderson wrote:
>>> Mounting an NFS share on top of a skimmed down /usr is very common,
>>> and very desirable.  You may mount /usr from a small read-only
>>> partition (vnode file, etc) and then mount a different partition or
>>> NFS over it if you detect the one you want.
>>>
>>> I think this comes down to: if it hurts, stop doing it.  :)
>>>
>>> Maybe sysinstall should warn you that you are double mounting, but I
>>> don't want it to stop letting me do it.
>> Are we absolutely sure overlaying NFS + local UFS filesystems like
>> this is the cause of the filesystem corruption?
>>
>> If Eric's doing it and it's working fine, I'm left wondering if
>> there's maybe sysinstall isn't handling something right.
> 
> I've rerun the test just to confirm but there are definitely
> two seperate issues here:
> 1. The ufs created by sysinstall after a repartition is corrupt.
> This is totally unrelated to the overlay of /usr as both /usr
> and /data ( which didnt previously exist ) where corrupted.
> 
> 2. Once the blank /usr was mounted over the working nfs /usr
> apps under /usr couldnt be run e.g. vim gave me no such file..
> After unmounting the ufs /usr using "umount -f /dev/da0s1f",
> without -f it gave a error due to use even know nothing was
> in use on it, the functionaility returned. Now this could
> be related to the corruption of the underlying ufs partition.
> If this is the case then solving #1 will also fix #2


So try the same test, with *only* the data partition, without messing 
with the /usr stuff..

Eric



More information about the freebsd-stable mailing list