Re: Kerberised NFSv4 - everyone gets mapped to nobody on file access

Reply: Andreas Kempe : "Re: Kerberised NFSv4 - everyone gets mapped to nobody on file access"
In reply to: Andreas Kempe : "Re: Kerberised NFSv4 - everyone gets mapped to nobody on file access"
Go to: [ bottom of page ] [ top of archives ] [ this month ]
From: Rick Macklem <rick.macklem_at_gmail.com>
Date: Sat, 30 Mar 2024 23:01:50 UTC
On Sat, Mar 30, 2024 at 2:45 PM Andreas Kempe <kempe@lysator.liu.se> wrote:
>
> On Thu, Mar 28, 2024 at 06:13:00AM -0700, Rick Macklem wrote:
> > On Thu, Mar 28, 2024 at 6:09 AM Rick Macklem <rick.macklem@gmail.com> wrote:
> > >
> > > On Thu, Mar 28, 2024 at 3:46 AM Andreas Kempe <kempe@lysator.liu.se> wrote:
> > > >
> > > > On Wed, Mar 27, 2024 at 03:20:03PM -0700, Rick Macklem wrote:
> > > > > On Wed, Mar 27, 2024 at 10:17 AM Andreas Kempe <kempe@lysator.liu.se> wrote:
> > > > > >
> > > > > > On Tue, Mar 26, 2024 at 05:54:38PM -0700, Rick Macklem wrote:
> > > > > > > On Tue, Mar 26, 2024 at 5:33 PM Rick Macklem <rick.macklem@gmail.com> wrote:
> > > > > > > >
> > > > > > > > Take a look at a packet capture in wireshark.
> > > > > > > > Check that the @domain part of Owner and Owner_group attributes are
> > > > > > > > the same and it is not a string of digits.
> > > > > > > Oh, and just fyi, you can use tcpdump to capture the packets, something like:
> > > > > > > # tcpdump -s 0 -w out.pcap host <nfs-server>
> > > > > > > and then you can look at out.pcap whereever it is convenient to
> > > > > > > install wireshark.
> > > > > > > (I run it on this windows laptop.)
> > > > > > > Don't bother to try and look at NFS with tcpdump. It doesn't know how
> > > > > > > to decode it.
> > > > > > >
> > > > > > > > If the domain is not the same, you can use the -domain command line option
> > > > > > > > on nfsuserd to set it.
> > > > > > > > (Since this "domain" is underdefined, I'd suggest only ascii characters and
> > > > > > > > all alphabetics in lower case.)
> > > > > > > > If the client sends a string of digits, check to make sure the sysctl
> > > > > > > > vfs.nfs.enable_uidtostring is set to 0.
> > > > > > > >
> > > > > >
> > > > > > I'm using lysator.liu.se as the domain on both client and server. It
> > > > > > seems to work since listing files give correct owners.
> > > > > >
> > > > > > I have dumped the traffic from mounting and creating a file named
> > > > > > test file that shows up as owned by nobody. I get the following call
> > > > > > made
> > > > > >
> > > > > >         NFS     438     V4 Call (Reply In 131) Open OPEN DH: 0x30a4c0aa/testfil
> > > > > >
> > > > > > In the OPEN (18) opcode, owner is set to
> > > > > >
> > > > > >                 0000   af 16 00 00 93 fc 00 00 07 76 0d 00
> > > > > >
> > > > > > while the server sets owner to ex. kempe@lysator.liu.se as expected
> > > > > > when directory listings are made.
> > > > > Doesn't make sense. What does wireshake show you for the Owner
> > > > > attribute in the setable attributes of the Open arguments. It should flag
> > > > > it as non-UTF8.
> > > > >
> > > >
> > > > I'm afraid I don't really understand how to check this. Wireshark
> > > > secifies "owner: <DATA>" if that says anything.
> > > >
> > > > > If you email me the pcap.out as an attachment, I'll look at it in wireshark.
> > > > > The out.pcap should include both the Open that creates a file and an
> > > > > "ls -l <file>", so there is a Getattr for the file as well.
> > > > >
> > > >
> > > > I'll send you a capture off-list. Thank you for helping!
> > > I looked at the capture. The server is definitely replying
> > > "nobody@lysator.liu.se"
> > > for both owner and owner_group for the file. You can see it in the
> > > reply to Open, Lookup and
> > > the Getattr (you need to go down past where it lists the attributes to
> > > see what their
> > > values are). It does know kempe@lysator.liu.se, since that is reported
> > > for owner for the directory.
> > Just to clarify, when you look at the replies you will first see a list of
> > attribute names (those just indicate which bits were set for the attributes),
> > then after that there is a list of attributes and their values.
> > (At least that is what I see when I use wireshark. I don't think I toggled
> > any setting to get that, but??)
> >
> > rick
> >
> > >
> > > I have no idea why it would do that, but it's a Linux server so???
> > >
>
> Did you have a look at the owner field in the open reqest that created
> the file? To me, it looks very strange. Could it be that the client
> isn't sending a correct owner in the creation request, causing the
> server to map it to nobody?
The only setable attribute specified by the Open request is "mode".
That just means that the server is expected to create the file with
an ownership of the Kerberos principal used in the RPC credentials.

Now, if you are doing the RPC as root, that will result in nobody (or
a failure to create the file, depending upon directory permissions).
There is no way to know what user principal is represented by the
RPCSEC_GSS credentials, since they are a shorthand for the
Kerberos credentials presented in a Null RPC that happens
when there is no credential.

When creating a file, the user creating the file will need to have
a valid TGT in the client's credential cache.

And the user principal name (without @REALM) must be a name
in the passwd database.

rick
>
> > > rick
> > > >
> > > > > rick
> > > > > ps: If that is what is in the Owner field, all I can suggest is that was what
> > > > >       a getpwnam() returned on the client. Possibly some weirdness with LDAP.
> > > > >       (I never use LDAP. Only a local /etc/passwd.)
> > > > >
> > > > > >
> > > > > > vfs.nfs.enable_uidtostring is 0 on the client machine and I am not
> > > > > > quite able to make sense of what the 12 bytes in the owner field are
> > > > > > supposed to be. They are not the ASCII representation and nither my
> > > > > > user's GID and UID that are both 0x7b02.
> > > > > >
> > > > > > // Andreas Kempe
> > > > >
> > > >
> >
>