NFS mount for TQM823L

NFS mount for TQM823L

[RSR - Piero Dominioni]

Hello to everybody.
We have got a STK8xxL400 with TQM823L minimodule on it,
running Linux-2.4.4-2001-07-23 (Denx Software Engineering)
from initial ramdisk.
We are experimenting with NFS: we succesfully mount a remote
directory (from host, the PC LinuxBox) under the target (STK8xxL400)
filesystem.
We give, on the target, the command

# mount -t nfs 126.126.126.106:/home /tmp

(being 126.126.126.106 our host's IP address, /home the
directory exported for other machines on the subnet and
/tmp the target mount point for remote directory) and we get
the following messages:

RPC: sendmsg returned error 101
portmap: RPC call returned 101
RPC: sendmsg returned error 101
portmap: RPC call returned 101
lockd_up: makesock failed, error =-101
RPC: sendmsg returned error 101
portmap: RPC call returned 101

but if we give (from target's command line )

# cat /proc/mounts

we get

/dev/root        /          ext2   rw   0   0
proc               /proc   proc  rw   0   0
126.126.126.106:/home       /tmp     nfs rw, v2, rsize=8192, wsize=8192,
hard, udp, lock, addr=126.126.126.106  0  0

and we correctly see the remote filesystem as (if it would be) local.
[Note that we previously modified the initial ramdisk, adding
"exports", "hosts.allow" and "hosts.deny" files under /etc,
modifying "hosts" file adding IP addresses of our machines
on the subnet (host, target and loopback at least) and finally
modifying the "host.conf" file, adding the line "multi on".
I give these details to best explain what we have done anymore!]


On the other hand, when we give the command (from host's
command line):

# mount -t nfs 126.126.126.107:/  /mydir

we get this error message:

Mount: RPC: Unable to receive; errno = Connection refused.

Our target's kernel's configuration is:
    <*> NFS file system support
    [*]        Provide NFSv3 client support
    [*]        Root file system on NFS
    <*> NFS Server support
    [*]        Provide NFSv3 server support
So we wouldn't need rpc.nfsd and rpc.mountd daemons
running on target...or would we?
Maybe is portmap daemon missing on target?
Or something else is still missing in our configurations?
If anybody could give hints about NFS we would appreciate
very much.
Thank you.

Best regards.

Piero Dominioni
R.S.R. srl


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

RE: NFS mount for TQM823L

[Jean-Denis Boyer]

The portmapper (portmap) and the lock daemon (rpc.lockd) are not running on
your target.
Either you start them, or just add option "-o nolock" when calling mount.
  # mount -o nolock -t nfs 126.126.126.106:/home /tmp


--------------------------------------------
 Jean-Denis Boyer, B.Eng., Technical Leader
 Mediatrix Telecom Inc.
 4229 Garlock Street
 Sherbrooke (Québec)
 J1L 2C8  CANADA
 (819)829-8749 x241
--------------------------------------------

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

RE: NFS mount for TQM823L

[Jean-Denis Boyer]

> but your hint is correct only in one direction:

Sorry, I stopped reading further... ;-)

To be able to export and mount you target's /, enabling the kernel option is
needed but not enough.
You need the portmapper (portmap), NFS daemon (rpc.nfsd), NFS mount daemon
(rpc.mountd).
I'm not sure, but you may also need the remote quota server (rpc.rquotad).
Of course, you should have a valid /etc/exports file.

All this comes from the standard NFS control script.
Refers to the nfs-utils package (http://sourceforge.net/projects/nfs/)
in the script file 'etc/nodist/nfs-server'.
You can even use it directly to start the NFS server.

Regards,
--------------------------------------------
 Jean-Denis Boyer, B.Eng., Technical Leader
 Mediatrix Telecom Inc.
 4229 Garlock Street
 Sherbrooke (Québec)
 J1L 2C8  CANADA
 (819)829-8749 x241
--------------------------------------------


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

RE: NFS mount for TQM823L

[Jean-Denis Boyer]

> I can see that portmap is suspended:
Do you mean that it is sleeping instead of suspended?
In the STAT column of ps, a 'S' means 'sleeping'.
Look at the 'ps' man pages.

> I guess it is correctly waiting for
> RPC calls and for other daemons (such as mountd and nfsd) to register.
I think you are right.

> But if I try to run mountd or nfsd (# ./mountd,  for example)
> I get the same
> errors as portmap wouldn't be running at all.
Be sure that the file '/etc/services' in the target includes the following
entries:
  sunrpc   111/tcp   portmapper   # RPC 4.0 portmapper TCP
  sunrpc   111/udp   portmapper   # RPC 4.0 portmapper UDP

Also, be sure you have the file '/etc/rpc' present in the target.

> Can you explain me how?
> Should I cross-compile it for my PowerPC platform, load it on
> the target and then run it?
This is a script, so it does not need to be compiled.
The only thing you need is the interpreter (first line in the script).
That script invokes the Bourne Shell (usually accomplished by bash).
The script should be invoked at the command line, giving it a parameters.
Valid parameters are: start, stop, restart, reload and status.

Regards,
--------------------------------------------
 Jean-Denis Boyer, B.Eng., Technical Leader
 Mediatrix Telecom Inc.
 4229 Garlock Street
 Sherbrooke (Québec)
 J1L 2C8  CANADA
 (819)829-8749 x241
--------------------------------------------

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

RE: NFS mount for TQM823L

[Jean-Denis Boyer]

> While if we remote-execute it with
> # rpcinfo -p 126.126.126.107  (our target's IP)
> from the host's command line prompt we correctly obtain

Is your loopback interface properly configured with IP address 127.0.0.1?
Try:
  ipconfig lo 127.0.0.1
And then:
  rpcinfo -p

Regards,
--------------------------------------------
 Jean-Denis Boyer, B.Eng., Technical Leader
 Mediatrix Telecom Inc.
 4229 Garlock Street
 Sherbrooke (Québec)
 J1L 2C8  CANADA
 (819)829-8749 x241
--------------------------------------------

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

RE: NFS mount for TQM823L

[Jean-Denis Boyer]

> Try:
>   ipconfig lo 127.0.0.1

Well, of course, I meant 'ifconfig', not 'ipconfig'.
I can't remember where this mistake comes from... ;-)

--------------------------------------------
 Jean-Denis Boyer, B.Eng., Technical Leader
 Mediatrix Telecom Inc.
 4229 Garlock Street
 Sherbrooke (Québec)
 J1L 2C8  CANADA
 (819)829-8749 x241
--------------------------------------------

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

RE: NFS mount for TQM823L

[Jean-Denis Boyer]

> I send you in attach the 'rpc' and 'services' files I loaded
> on the target (I hand-wrote them first, unsuccessfully:
> I copied then these ones from the host).

This is a good approach. The files seem to be correct.
(BTW, I think the rpc file comes with the glibc distribution).


> (Note that I type the 'mountd' command, but it doesn't
> appear in the ps -aux output list!)

I guess it died, even after successfully registered with the portmapper.
Here are some ideas I have to verify everything.

1) Does it produce a core file?

2) You may start mountd with the following options:
     mountd --debug all --foreground
   This will start mountd in foregroud, turning on debugging information.
   I hope it will help...

3) You may also start the syslog daemon, to be sure you don't miss important
information.

4) Also, is the file /etc/exports present?

5) I see that the nfs startup script performs a 'exportfs -r' before
starting the daemons.
   Could it help?


> While if I type the same command for the host the
> mountd uses a different port number...

This is correct, since the port number is assigned randomly, unless you use
the command line option '-p' to force mountd to bind to a specific port. The
rpc port number should be identical, and correspond to what is in /etc/rpc.


> while, exactly meantime, the following message appears
> on the target's console:
>
> svc: unknown program 100005 (me 100021)

The mountd process (rpc port 100005) is not in the process table, because it
unexpectedly died, even if the registration still appears in the portmap...
as if you kill mountd with a SIGKILL (9) :-(

Good luck!
--------------------------------------------
 Jean-Denis Boyer, B.Eng., Technical Leader
 Mediatrix Telecom Inc.
 4229 Garlock Street
 Sherbrooke (Québec)
 J1L 2C8  CANADA
 (819)829-8749 x241
--------------------------------------------

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

RE: NFS mount for TQM823L

[Jean-Denis Boyer]

> Without the syslogd daemon I couldn't have been able to
> discover that the 'etab' file was missing in /var/lib/nfs target's
> directory.
>
> Who does create this file, on the hosts for example?
>  Which command or daemon?

I think that calling
  /usr/sbin/exportfs -r
before starting every daemons related to NFS is the solution.
Well, that is what the NFS control script does on startup...

> > 1) Does it produce a core file?
> What do you mean with 'core file'?

When an application crashes, by receiving either an ABRT signal or a SEGV
signal (segmentation fault), it produces a core dump file. The file name is
'core', and is created in the directory that was current when the
application was started. I don't know all the details, but this core file
saves the state of the application, at the time it crashed. Along with the
executable file, you can load this core file into gdb, and look at the
problem.

> >    mountd --debug all --foreground
> >   This will start mountd in foregroud, turning on debugging
> information.
> >    I hope it will help...
> How should I have debugged then? (I couldn't do it)
> I have seen nothing: should any message have appeared?

You're right. It doesn't look to make any difference. I even had to look
into the source code to know what option to pass to --debug. This was not
described in the man pages. And the --foregroud doesn't print message on
stdout, either. I was hoping that it may help. 'mountd' seems to ignore
these debugging parameters.

Regards,
--------------------------------------------
 Jean-Denis Boyer, B.Eng., Technical Leader
 Mediatrix Telecom Inc.
 4229 Garlock Street
 Sherbrooke (Québec)
 J1L 2C8  CANADA
 (819)829-8749 x241
--------------------------------------------

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/