On 02/23/2017 09:11 AM, Kees Bakker wrote:
On 23-02-17 13:51, Brendan Kearney wrote:
On 02/23/2017 07:32 AM, Kees Bakker wrote:
On 22-02-17 17:33, Brendan Kearney wrote:
On 02/22/2017 10:26 AM, Kees Bakker wrote:
On 22-02-17 14:05, Brendan Kearney wrote:
On 02/22/2017 05:23 AM, Kees Bakker wrote:
On 21-02-17 19:49, Brendan Kearney wrote:
On 02/21/2017 10:57 AM, Kees Bakker wrote:
Hey,

Maybe one of the NFS users on this list could give me a hint what
could be wrong. I'm not sure if it has any relation with FreeIPA/Kerberos.

I've set up an NFS server and I can mount the NFS directory on my client. So, 
I'm
guessing that setting up Kerberos principal was done correctly.

However, only root can actually access the mounted contents. Any other user
only sees question marks as shown below.

The mount command is simple.
$ sudo mount -v -t nfs srv1.example.com:/home /nfshome
mount.nfs: timeout set for Tue Feb 21 16:36:39 2017
mount.nfs: trying text-based options 
'vers=4,addr=172.16.16.45,clientaddr=172.16.16.30'

On the server side /etc/exports looks like this.
/home        *(rw,sync,sec=krb5i,no_subtree_check)

$ sudo mount |grep nfs
srv1.example.com:/home on /nfshome type nfs4 
(rw,relatime,vers=4.0,rsize=1048576,wsize=1048576,namlen=255,hard,proto=tcp,port=0,timeo=600,retrans=2,sec=krb5i,clientaddr=172.16.16.30,local_lock=none,addr=172.16.16.45)

$ sudo ls -ld /nfshome
drwxr-xr-x 1 root root 72 feb 21 04:22 /nfshome
$ sudo ls -l /nfshome
total 0
drwxr-xr-x 1 keesb  keesb  116 jan 27 12:56 keesb

$ ls -l /nfshome
ls: cannot access '/nfshome': Permission denied
$ ls -l / | grep nfshome
ls: cannot access '/nfshome': Permission denied
d?????????   ? ?    ?       ?            ? nfshome

sec=krb* means that the user accessing the mount has to authenticate with a 
kerberos ticket, and has to be the user or in the group granted access to the 
share.  from the looks of things, the user did not authenticate, and that is 
why the permissions are question marks.  check the kerberos tickets that the 
user has (klist output).  Otherwise, the ownership might be user and group that 
the client machine does not recognize (think posix user/group that is not in 
sync between the NFS server and the client)
Thanks for the reply.

In this case the user _is_ authenticated.
keesb@client1:~$ klist
Ticket cache: KEYRING:persistent:60001:60001
Default principal: ke...@example.com

Valid starting     Expires            Service principal
22-02-17 09:20:30  23-02-17 09:20:25  krbtgt/example....@example.com
no, the user has a TGT.  a nfs/host.domain.tld@REALM ticket is needed to 
authenticate.
(( I'm trying to catch up on the acronyms. TGT. Reading wikipedia now. ))

What other grants could be needed? HBAC Rules?

Do I need an nfs principal for the client? (I didn't think so, but many HOWTO's 
say so [2]. Anyway, it
doesn't help to get access for the user.)
there are principals to create and keytabs to be updated on hte NFS sever, if 
not done already.
I did create a principal for the NFS server (using ipa service-add) and
add to the keytab on the NFS server (using ipa-getkeytab) ...
root@srv1# klist -ke
Keytab name: FILE:/etc/krb5.keytab
KVNO Principal
---- --------------------------------------------------------------------------
      1 host/srv1.example....@example.com (aes256-cts-hmac-sha1-96)
      1 host/srv1.example....@example.com (aes128-cts-hmac-sha1-96)
      1 nfs/srv1.example....@example.com (aes256-cts-hmac-sha1-96)
      1 nfs/srv1.example....@example.com (aes128-cts-hmac-sha1-96)

Is this what you mean?
yes, if that is done, the server side components should be done for kerberos.  
have you set things up in /etc/idmapd.conf so your domain, REALM, etc are setup?
I don't think that a change of idmapd.conf (on the NFS server) is needed 
because all host
names are FQDN and everything is in one and the same REALM.
NFS needs to know how to map a user object to an ID and groups. identities 
established by kerberos do not directly translate to users.  usually some sort 
of directory services are leveraged in order to accomplish this, though PAM and 
things like that can be used to.  by setting things in idmapd.conf, you are 
telling NFS who to translate kerberos identities into usernames, so ownership 
and permissions can be sync'd.
Both the NFS server and the client are configured as FreeIPA client.
On the server the users are known (through PAM, SSSD). Only user
"ubuntu" is a local (/etc/passwd) user. All other users are defined on
the IPA server.

root@srv1:~# ls -l /home
total 0
drwxr-xr-x 1 keesb  keesb  116 jan 27 12:56 keesb
drwxr-xr-x 1 ubuntu ubuntu 142 aug 17  2016 ubuntu
root@srv1:~# ls -ln /home
total 0
drwxr-xr-x 1 60001 60001 116 jan 27 12:56 keesb
drwxr-xr-x 1  1000  1000 142 aug 17  2016 ubuntu

On the client, same story

root@client1:~# ls -l /nfshome
total 0
drwxr-xr-x 1 keesb  keesb  116 jan 27 12:56 keesb
drwxr-xr-x 1 ubuntu  ubuntu  142 aug 17  2016 ubuntu
root@client1:~# ls -ln /nfshome
total 0
drwxr-xr-x 1 60001 60001 116 jan 27 12:56 keesb
drwxr-xr-x 1  1000  1000 142 aug 17  2016 ubuntu



     then the user should be able to pull the ticket for auth.
Sorry to ask, but how do I do that? On the client, I suppose, and by the user ??

keesb@client1$ kinit nfs/srv1.example....@example.com
Password for nfs/srv1.example....@example.com:

But I don't have a password for that. Hmm.
there is no need to init on the client side, as long as the TGT is obtained.  
you should never need to init the nfs/blah.. on the client side.
OK
So, it seems to me that all the basics are setup correctly. The mount succeeds. 
The user
has a TGT and still the (non-root) user cannot even stat the mount point, nor 
the directory
entry itself.

What puzzles me is that root can see everything, also without a TGT.
the mount will succeed, but the user does not have access because NFS does not 
know who the user it.  it has an unassociated ID established by kerberos, but 
not a user.
Let's step back a little.

On the NFS client, the user does an ls of / (that's where the mountpoint is). 
At the point
where ls reads the entry (through stat) for "nfshome" it gets (probably) EPERM.
And hence ls prints all the question marks. That leads me to believe that only 
the kernel
(on the NFS client) is involved at this point. I could be wrong but I don't 
think there is
any kerberos TGT involved.

Furthermore, I'm guessing that the host principle which I got after 
ipa-client-install is
good enough. (This [1] wiki suggests that I need to do a ipa-getkeytab for it, 
which I
did not do.)
# klist -ke
Keytab name: FILE:/etc/krb5.keytab
KVNO Principal
---- --------------------------------------------------------------------------
       1 host/client1.example....@example.com (aes256-cts-hmac-sha1-96)
       1 host/client1.example....@example.com (aes128-cts-hmac-sha1-96)

[1] http://wiki.linux-nfs.org/wiki/index.php/NFS_and_FreeIPA
[2] https://www.lisenet.com/2016/kerberised-nfs-server-on-rhel-7/
http://www.itp.uzh.ch/~dpotter/howto/kerberos

added the thread back to the mailing list for the benefit of any who search on 
the subject.


run journalctl -fl on both the NFS server and the NFS client while trying to mount the share. one of them should tell you something.

the mount operation will succeed, because any ACL set in the exports file is being met. root likley has access because the uid/gid is common and local on both devices and therefore no mapping is needed. your virtual user does not exist in either local user store, and therefore needs to be mapped. what does /etc/nsswitch.conf say for passwd, shadow and group?

--
Manage your subscription for the Freeipa-users mailing list:
https://www.redhat.com/mailman/listinfo/freeipa-users
Go to http://freeipa.org for more info on the project

Reply via email to