The new logic also appears to be somewhat broken, not sure if this is covered in that burt --

Given (pre-6.5):

/vol/vol0/foo -ro=192.168.0.0/16,rw=192.168.1.1:10.1.1.1

The export would work as expected (192.168.1.1 and 10.1.1.1 have write access, 192.168.30.30 has ro access). However, under 6.5 this results in only having 10.1.1.1 having write access. Presumably this is due to na_exports(5):

11) -ro= trumps -rw=

..which is perfectly reasonable, IFF a specific host was referenced in both cases. In the event that a host is referred to indirectly (ie. via network or netgroup), the specific reference should always win. Hence, example from man page would be extended (perhaps further still to illustrate priority of netgroup vs. network):

12) -ro=DIRECT trumps -rw=INDIRECT

13) -rw=DIRECT trumps -ro=INDIRECT

14) -ro=DIRECT trumps -ro=DIRECT

On Sun, 22 Feb 2004, Haynes, Tom wrote:

This can be tracked as burt 115086.

I'm just about to make the public report available.

The fix will come out in 6.5.1.

...

Hello,

If you plan to upgrade to 6.5 make sure you check your /etc/exports file after rebooting with 6.5 for the first time. I just upgraded an 840 and an 880 and on both machines the exports file has been rewritten giving everyone read-only access.

Running 6.4.2: /vol/vol10 -access=donald.zdv.uni-mainz.de:unixadm.zdv.uni-mainz.de:dschinny.zdv.uni-mainz.de,root=donald.zdv.uni-mainz.de:unixadm.zdv.uni-mainz.de /vol/vol11 -access=donald.zdv.uni-mainz.de:unixadm.zdv.uni-mainz.de:dschinny.zdv.uni-mainz.de,root=donald.zdv.uni-mainz.de:unixadm.zdv.uni-mainz.de ...

Running 6.5P5: /vol/vol10 -ro,rw=donald.zdv.uni-mainz.de:unixadm.zdv.uni-mainz.de:dschinny.zdv.uni-mainz.de,root=donald.zdv.uni-mainz.de:unixadm.zdv.uni-mainz.de /vol/vol11 -ro,rw=donald.zdv.uni-mainz.de:unixadm.zdv.uni-mainz.de:dschinny.zdv.uni-mainz.de,root=donald.zdv.uni-mainz.de:unixadm.zdv.uni-mainz.de ...

Carsten

-- Tom Haynes, ex-cfb thomas@netapp.com

On Sun, 22 Feb 2004, Haynes, Tom wrote:

Yes, this is fixed in burt 113819, which can be gotten as patch 6.5P5.

Fantastic. Already mangled exports files for now, so I'll probably just leave as-is until 6.5.1.

Has anyone played with the new 'actual=' exports option? I think there's been requests for this since multiple volumes were introduces (5.x?) on this list. Don't know if its intentional, but mounts below the root don't appear to work properly. ie, given the exports:

/schoolbus -actual=/vol/bus/yellow,ro /vol/bus/yellow -ro

mount filer:/vol/bus/yellow/short /mnt

..will but you where you would expect to be. but

mount filer:/schoolbus/short /mnt

...won't get you onto the shortbus, instead /mnt will actually be /vol/bus/yellow.

Either way, great feature!

Are you using 6.5 or previous for -actual?

They are busted in 6.4 branches.

6.5 -- first noticed it when digging around for info on export problems.

You can use exportfs -s to test if you are getting the right path or not.

This does show the proper path. Tried only exporting the line w/ 'actual' and it would appear that any attempt to mount a directory below the one exported (ie '/schoolbus/short/') still ends up getting truncated to the 'actual=' option. Ignoring the shortbus, the actual case is trying to shorten mount paths for a jumpstart. ie, instead of mounting:

/vol/install/Solaris/Solaris_8.hwrel-0202

just exporting:

/Solaris -actual=/vol/install/Solaris

and then mounting filer:/Solaris/Solaris_8.hwrel-0202. As-is, what ends up mounted is /vol/install/Solaris, rather than /vol/install/Solaris/Solaris_8.hwrel-0202

Try reversing the order and see which mount succeeds.

In either case, the mount suceeds, just not in the way you'd expect.

...

On Sun, 22 Feb 2004, Haynes, Tom wrote:

...

Yes, this is fixed in burt 113819, which can be gotten as patch 6.5P5.

Fantastic. Already mangled exports files for now, so I'll probably just leave as-is until 6.5.1.

Has anyone played with the new 'actual=' exports option? I think there's been requests for this since multiple volumes were introduces (5.x?) on this list. Don't know if its intentional, but mounts below the root don't appear to work properly. ie, given the exports:

/schoolbus -actual=/vol/bus/yellow,ro /vol/bus/yellow -ro

mount filer:/vol/bus/yellow/short /mnt

..will but you where you would expect to be. but

mount filer:/schoolbus/short /mnt

...won't get you onto the shortbus, instead /mnt will actually be /vol/bus/yellow.

Either way, great feature!

-- Tom Haynes, ex-cfb thomas@netapp.com

Haynes, Tom wrote:

...

This can be tracked as burt 115086.

I'm just about to make the public report available.

The fix will come out in 6.5.1.

Hi

Do you now when 6.5.1 will be available ?

Thanks

With all the normal disclaimers in place, i.e., I'm an individual contributor and I have as much control over this as I do the weather, my understanding is it is due to ship in late April.

Is '-ro,rw ...' still compatible with netgroups? Can you use netgroups for 'root=...' in 6.5? Did they add a feature to the read the machine/shelf/disk temperature(s) in 6.5? Too many questions and right now no spare machine to test it ;-)

No idea about the machine parts, but on the exports:

1) -access is gone. It is the union of the rw and ro hosts

2) All of ro, rw, and root support netgroups and DNS subdomains.

3) ro now supports a list of hosts

I have no idea of the availability of the 6.5 simulator, but if it is available, grab it and play with it for testing exports.

[Charset ISO-8859-15 unsupported, skipping...]