On Fri, Apr 20, 2012 at 11:55:53AM -0700, Jeremy Allison wrote:
> On Fri, Apr 20, 2012 at 12:04:02PM +0200, Ondrej Valousek wrote:
> > Hi List,
> >
> > I know this has been asked many times, but I will ask one more time:
> > I would like to export NFSv4 mounted filesystem with Samba and would
> > like to be able to use windows "security tab" in explorer to modify
> > these ACLs.
> > Do I have any chance with Samba on RHEL-5/6?
>
> Does Linux have local API's to manipulate mounted NFSv4 ACLs ?
>
> Samba has VFS modules to cope with mapping Windows ACLs onto
> NFSv4 ACLs but currently they only work on AIX, Linux with IBM's
> gpfs and Solaris right now.
>
> If you can point me a docs showing the library APIs for Linux
> to allow programatic manipulation of NFSv4 ACLs then I can
> make it work on Linux too.
It's a bit idiosyncratic, but should be easy enough to handle if you
want, I think: on the linux client, on an nfs4 filesystem, the
"system.nfs4_acl" xattr contains the xdr-encoded ACL as it came in on
the wire (network-byte-order and all). See the definition of fattr4_acl
in rfc 3530.
--b.
On Fri, Apr 20, 2012 at 05:26:27PM -0400, J. Bruce Fields wrote:
> On Fri, Apr 20, 2012 at 11:55:53AM -0700, Jeremy Allison wrote:
> > On Fri, Apr 20, 2012 at 12:04:02PM +0200, Ondrej Valousek wrote:
> > > Hi List,
> > >
> > > I know this has been asked many times, but I will ask one more time:
> > > I would like to export NFSv4 mounted filesystem with Samba and would
> > > like to be able to use windows "security tab" in explorer to modify
> > > these ACLs.
> > > Do I have any chance with Samba on RHEL-5/6?
> >
> > Does Linux have local API's to manipulate mounted NFSv4 ACLs ?
> >
> > Samba has VFS modules to cope with mapping Windows ACLs onto
> > NFSv4 ACLs but currently they only work on AIX, Linux with IBM's
> > gpfs and Solaris right now.
> >
> > If you can point me a docs showing the library APIs for Linux
> > to allow programatic manipulation of NFSv4 ACLs then I can
> > make it work on Linux too.
>
> It's a bit idiosyncratic, but should be easy enough to handle if you
> want, I think: on the linux client, on an nfs4 filesystem, the
> "system.nfs4_acl" xattr contains the xdr-encoded ACL as it came in on
> the wire (network-byte-order and all). See the definition of fattr4_acl
> in rfc 3530.
I'm sorry, but that *really* *really* sucks. There needs to
be an API on the Linux side that implements both RichACLs
and NFSv4 ACLs (as they're the same thing essentially).
Both GPFS and ZFS have defined API's to get and set
ACLs. Just sticking them on a "system.nfs4_acl" xattr
is a cop-out of monstrous proportions (and the kind of
crap that we have to pull :-).
Jeremy.
On Fri, 2012-04-20 at 18:08 -0400, J. Bruce Fields wrote:
> On Fri, Apr 20, 2012 at 02:59:45PM -0700, Jeremy Allison wrote:
> > On Fri, Apr 20, 2012 at 05:26:27PM -0400, J. Bruce Fields wrote:
> > > On Fri, Apr 20, 2012 at 11:55:53AM -0700, Jeremy Allison wrote:
> > > > On Fri, Apr 20, 2012 at 12:04:02PM +0200, Ondrej Valousek wrote:
> > > > > Hi List,
> > > > >
> > > > > I know this has been asked many times, but I will ask one more time:
> > > > > I would like to export NFSv4 mounted filesystem with Samba and would
> > > > > like to be able to use windows "security tab" in explorer to modify
> > > > > these ACLs.
> > > > > Do I have any chance with Samba on RHEL-5/6?
> > > >
> > > > Does Linux have local API's to manipulate mounted NFSv4 ACLs ?
> > > >
> > > > Samba has VFS modules to cope with mapping Windows ACLs onto
> > > > NFSv4 ACLs but currently they only work on AIX, Linux with IBM's
> > > > gpfs and Solaris right now.
> > > >
> > > > If you can point me a docs showing the library APIs for Linux
> > > > to allow programatic manipulation of NFSv4 ACLs then I can
> > > > make it work on Linux too.
> > >
> > > It's a bit idiosyncratic, but should be easy enough to handle if you
> > > want, I think: on the linux client, on an nfs4 filesystem, the
> > > "system.nfs4_acl" xattr contains the xdr-encoded ACL as it came in on
> > > the wire (network-byte-order and all). See the definition of fattr4_acl
> > > in rfc 3530.
> >
> > I'm sorry, but that *really* *really* sucks. There needs to
> > be an API on the Linux side that implements both RichACLs
> > and NFSv4 ACLs (as they're the same thing essentially).
>
> Yes, that's the eventual plan.
>
> The richacl patches still need a little more work, and alas there's
> noone working on them currently as far as I know.
>
> So if you wanted something that would work now, you could use
> system.nfs4_acl now and later replace that by something that tried
> richacls and then fell back.
>
> The xdr isn't hard to parse or create--the main problem would be
> figuring out what to do with the names....
Samba does handle id mapping in general and name mapping shouldn't be
too hard to add, SIDs would be easier for us though :-)
> (By the way, is reexporting nfs with samba really going to work well
> anyway?)
Not really, it is generally unsupported, except for read only shares. It
will work in most of the simple cases for writing .. until it breaks :)
Simo.
--
Simo Sorce
Samba Team GPL Compliance Officer <[email protected]>
Principal Software Engineer at Red Hat, Inc. <[email protected]>
On 21/04/12 03:34, simo wrote:
> On Fri, 2012-04-20 at 18:08 -0400, J. Bruce Fields wrote:
>> On Fri, Apr 20, 2012 at 02:59:45PM -0700, Jeremy Allison wrote:
>>> On Fri, Apr 20, 2012 at 05:26:27PM -0400, J. Bruce Fields wrote:
>>>> On Fri, Apr 20, 2012 at 11:55:53AM -0700, Jeremy Allison wrote:
>>>>> On Fri, Apr 20, 2012 at 12:04:02PM +0200, Ondrej Valousek wrote:
>>>>>> Hi List,
>>>>>>
>
>> (By the way, is reexporting nfs with samba really going to work well
>> anyway?)
>
> Not really, it is generally unsupported, except for read only shares. It
> will work in most of the simple cases for writing .. until it breaks :)
>
Hi
That's worrying.
We been rw'ing nfs/samba for years. Maybe we're the only ones. nfs3 for
our Linux clients and samba for the rest. Same shared data. Only Posix
acl's though. Maybe that's the simple bit?
Cheers,
Steve
On 22/04/12 01:57, Andrew Bartlett wrote:
> On Sat, 2012-04-21 at 13:34 +0200, steve wrote:
>> On 21/04/12 03:34, simo wrote:
>>> On Fri, 2012-04-20 at 18:08 -0400, J. Bruce Fields wrote:
>>>> On Fri, Apr 20, 2012 at 02:59:45PM -0700, Jeremy Allison wrote:
>>>>> On Fri, Apr 20, 2012 at 05:26:27PM -0400, J. Bruce Fields wrote:
>>>>>> On Fri, Apr 20, 2012 at 11:55:53AM -0700, Jeremy Allison wrote:
>>>>>>> On Fri, Apr 20, 2012 at 12:04:02PM +0200, Ondrej Valousek wrote:
>>>>>>>> Hi List,
>>>>>>>>
>>
>>>
>>>> (By the way, is reexporting nfs with samba really going to work well
>>>> anyway?)
>>>
>>> Not really, it is generally unsupported, except for read only shares. It
>>> will work in most of the simple cases for writing .. until it breaks :)
>>>
>>
>> Hi
>> That's worrying.
>>
>> We been rw'ing nfs/samba for years. Maybe we're the only ones. nfs3 for
>> our Linux clients and samba for the rest. Same shared data. Only Posix
>> acl's though. Maybe that's the simple bit?
>> Cheers,
>> Steve
>
> Are Samba and NFSd running on the same server, exporting the same data?
> That is what is expected to work. See 'kernel oplocks' (the default of
> which will change for 4.0) to allow NFS and Samba to break each other's
> oplocks.
>
s3 and nfs work fine (maybe others are doing it too?)
We are testing Samba4 and nfs now and yes same data, same box. As yet,
we've not been able to break it no matter what we throw at it.
When will the oplocks default change? Will it affect us?
Cheers,
Steve
On Sat, 2012-04-21 at 13:34 +0200, steve wrote:
> On 21/04/12 03:34, simo wrote:
> > On Fri, 2012-04-20 at 18:08 -0400, J. Bruce Fields wrote:
> >> On Fri, Apr 20, 2012 at 02:59:45PM -0700, Jeremy Allison wrote:
> >>> On Fri, Apr 20, 2012 at 05:26:27PM -0400, J. Bruce Fields wrote:
> >>>> On Fri, Apr 20, 2012 at 11:55:53AM -0700, Jeremy Allison wrote:
> >>>>> On Fri, Apr 20, 2012 at 12:04:02PM +0200, Ondrej Valousek wrote:
> >>>>>> Hi List,
> >>>>>>
>
> >
> >> (By the way, is reexporting nfs with samba really going to work well
> >> anyway?)
> >
> > Not really, it is generally unsupported, except for read only shares. It
> > will work in most of the simple cases for writing .. until it breaks :)
> >
>
> Hi
> That's worrying.
>
> We been rw'ing nfs/samba for years. Maybe we're the only ones. nfs3 for
> our Linux clients and samba for the rest. Same shared data. Only Posix
> acl's though. Maybe that's the simple bit?
> Cheers,
> Steve
Are Samba and NFSd running on the same server, exporting the same data?
That is what is expected to work. See 'kernel oplocks' (the default of
which will change for 4.0) to allow NFS and Samba to break each other's
oplocks.
What Simo cautions about is where an NFS server (say a NAS server)
exports to some direct NFS clients, and then a 'proxy' Samba server is
an NFS client and a Samba server for Windows clients. This 'proxy'
setup is considered dangerous.
Andrew Bartlett
--
Andrew Bartlett http://samba.org/~abartlet/
Authentication Developer, Samba Team http://samba.org
On Fri, Apr 20, 2012 at 02:59:45PM -0700, Jeremy Allison wrote:
> On Fri, Apr 20, 2012 at 05:26:27PM -0400, J. Bruce Fields wrote:
> > On Fri, Apr 20, 2012 at 11:55:53AM -0700, Jeremy Allison wrote:
> > > On Fri, Apr 20, 2012 at 12:04:02PM +0200, Ondrej Valousek wrote:
> > > > Hi List,
> > > >
> > > > I know this has been asked many times, but I will ask one more time:
> > > > I would like to export NFSv4 mounted filesystem with Samba and would
> > > > like to be able to use windows "security tab" in explorer to modify
> > > > these ACLs.
> > > > Do I have any chance with Samba on RHEL-5/6?
> > >
> > > Does Linux have local API's to manipulate mounted NFSv4 ACLs ?
> > >
> > > Samba has VFS modules to cope with mapping Windows ACLs onto
> > > NFSv4 ACLs but currently they only work on AIX, Linux with IBM's
> > > gpfs and Solaris right now.
> > >
> > > If you can point me a docs showing the library APIs for Linux
> > > to allow programatic manipulation of NFSv4 ACLs then I can
> > > make it work on Linux too.
> >
> > It's a bit idiosyncratic, but should be easy enough to handle if you
> > want, I think: on the linux client, on an nfs4 filesystem, the
> > "system.nfs4_acl" xattr contains the xdr-encoded ACL as it came in on
> > the wire (network-byte-order and all). See the definition of fattr4_acl
> > in rfc 3530.
>
> I'm sorry, but that *really* *really* sucks. There needs to
> be an API on the Linux side that implements both RichACLs
> and NFSv4 ACLs (as they're the same thing essentially).
Yes, that's the eventual plan.
The richacl patches still need a little more work, and alas there's
noone working on them currently as far as I know.
So if you wanted something that would work now, you could use
system.nfs4_acl now and later replace that by something that tried
richacls and then fell back.
The xdr isn't hard to parse or create--the main problem would be
figuring out what to do with the names....
(By the way, is reexporting nfs with samba really going to work well
anyway?)
--b.