NFS problems, UMON, missing directories, wrong permissions

NFS problems, UMON, missing directories, wrong permissions

[Cajus Pollmeier]

=2D----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

[Please cc me, I'm currently not subscribed]

Hi!

Sorry to bother you with this kind of stuff, but I currently have no idea w=
hat's going on here.

The facts:

* Fileserver
  Debian Woody,  Kernel 2.4.19 SMP using knfs and Debian nfs-utils,
  export is on partitions with xfs/ext3

  /etc/exports contains something like this:
  /export/home            10.1.0.0/255.255.0.0(rw,no_root_squash)

* Terminal Server(s)
  Debian Woody,  Kernel 2.4.19 SMP using knfs and Debian nfs-utils=20
  (seem to be 1.0)

  fstab tells to mount with these options:
  fileserver:/export/home /home nfs exec,nodev,nosuid,timeo=3D10,rw,hard,re=
trans=3D20,rsize=3D8192,wsize=3D8192 1 1

One terminal server keeps about 15 users, all accessing there home via nfs,=
 shared files via group
permissions inclusive.


The problem(s) (they are massive at all :-/)

* Randomly missing directories / shredded permissions
  Either users don't see shared directories contents or have no permissions=
 to access them. In
  fact they have, "id" shows up with the correct group membership and "ls -=
la" shows the directory
  group writable. After logging out and in again, everything's fine.

* Periodical error messages in system log
  Client:
  May  6 07:12:59 terminalserver kernel: lockd: nlm_lookup_host(0a010002, p=
=3D17, v=3D4)
  May  6 07:12:59 terminalserver kernel: lockd: host garbage collection
  May  6 07:12:59 terminalserver kernel: lockd: nlmsvc_mark_resources
  May  6 07:12:59 terminalserver kernel: lockd: delete host 10.1.0.2
  May  6 07:12:59 terminalserver kernel: lockd: nsm_unmonitor(10.1.0.2)
  May  6 07:12:59 terminalserver kernel: nsm: xdr_encode_mon(0a010002, -124=
9509120, 67108864, 268435456)
  May  6 07:12:59 terminalserver rpc.statd[1932]: Received erroneous SM_UNM=
ON request from terminalserver for 10.1.0.2
  May  6 07:12:59 terminalserver kernel: lockd: creating host entry
  May  6 07:12:59 terminalserver kernel: lockd: nlm_bind_host(0a010002)
  May  6 07:12:59 terminalserver kernel: lockd: nsm_monitor(10.1.0.2)
  May  6 07:12:59 terminalserver kernel: nsm: xdr_encode_mon(0a010002, -124=
9509120, 67108864, 268435456)
  May  6 07:12:59 terminalserver kernel: nsm: xdr_decode_stat_res status 0 =
state 79
  May  6 07:12:59 terminalserver kernel: lockd: nlm_bind_host(0a010002)
  May  6 07:12:59 terminalserver kernel: lockd: release host 10.1.0.2
  May  6 07:12:59 terminalserver kernel: lockd: get host 10.1.0.2
  May  6 07:12:59 terminalserver kernel: lockd: nlm_lookup_host(0a010002, p=
=3D17, v=3D4)
  May  6 07:12:59 terminalserver kernel: lockd: get host 10.1.0.2
  May  6 07:12:59 terminalserver kernel: lockd: nlm_bind_host(0a010002)
  May  6 07:12:59 terminalserver kernel: lockd: release host 10.1.0.2
  May  6 07:12:59 terminalserver kernel: lockd: release host 10.1.0.2
  May  6 07:13:00 terminalserver kernel: lockd: nlm_lookup_host(0a010002, p=
=3D17, v=3D4)
  May  6 07:13:00 terminalserver kernel: lockd: get host 10.1.0.2
  May  6 07:13:00 terminalserver kernel: lockd: nlm_bind_host(0a010002)
  ....

  Server:
  May  6 07:12:59 fileserver kernel: lockd: request from 0a010005
  May  6 07:12:59 fileserver kernel: lockd: nlm_lookup_host(0a010005, p=3D1=
7, v=3D4)
  May  6 07:12:59 fileserver kernel: lockd: creating host entry
  May  6 07:12:59 fileserver kernel: lockd: nsm_monitor(10.1.0.5)
  May  6 07:12:59 fileserver kernel: nsm: xdr_encode_mon(0a010005, -1249509=
120, 67108864, 268435456)
  May  6 07:12:59 fileserver kernel: nsm: xdr_decode_stat_res status 0 stat=
e 91121
  May  6 07:12:59 fileserver kernel: lockd: nlm_file_lookup(02000001 110008=
00 00020001 00324063 53324ecb 00324060)
  May  6 07:12:59 fileserver kernel: lockd: creating file for (02000001 110=
00800 00020001 00324063 53324ecb 00324060)
  May  6 07:12:59 fileserver kernel: lockd: found file e6c1e280 (count 0)
  May  6 07:12:59 fileserver kernel: lockd: nlmsvc_lock(0811/3293283, ty=3D=
0, pi=3D18885, 0-9223372036854775807, bl=3D1)
  May  6 07:12:59 fileserver kernel: lockd: nlmsvc_lookup_block f=3De6c1e28=
0 pd=3D18885 0-9223372036854775807 ty=3D0
  May  6 07:12:59 fileserver kernel: lockd: posix_lock_file returned 0
  May  6 07:12:59 fileserver kernel: lockd: release host 10.1.0.5
  May  6 07:12:59 fileserver kernel: lockd: nlm_release_file(e6c1e280, ct =
=3D 1)
  May  6 07:12:59 fileserver kernel: nlmsvc_retry_blocked(00000000, when=3D=
0)
  May  6 07:12:59 fileserver kernel: nlmsvc_retry_blocked(00000000, when=3D=
0)
  May  6 07:12:59 fileserver kernel: lockd: request from 0a010005
  May  6 07:12:59 fileserver kernel: lockd: nlm_lookup_host(0a010005, p=3D1=
7, v=3D4)
  May  6 07:12:59 fileserver kernel: lockd: get host 10.1.0.5
  May  6 07:12:59 fileserver kernel: lockd: nlm_file_lookup(02000001 110008=
00 00020001 00324063 53324ecb 00324060)
  May  6 07:12:59 fileserver kernel: lockd: found file e6c1e280 (count 0)
  May  6 07:12:59 fileserver kernel: lockd: nlmsvc_unlock(0811/3293283, pi=
=3D18885, 0-9223372036854775807)
  May  6 07:12:59 fileserver kernel: lockd: nlmsvc_cancel(0811/3293283, pi=
=3D18885, 0-9223372036854775807)
  May  6 07:12:59 fileserver kernel: lockd: nlmsvc_lookup_block f=3De6c1e28=
0 pd=3D18885 0-9223372036854775807 ty=3D2
  May  6 07:12:59 fileserver kernel: lockd: release host 10.1.0.5
  May  6 07:12:59 fileserver kernel: lockd: nlm_release_file(e6c1e280, ct =
=3D 1)
  May  6 07:12:59 fileserver kernel: lockd: closing file 08:11/3293283
  May  6 07:12:59 fileserver kernel: nlmsvc_retry_blocked(00000000, when=3D=
0)
  May  6 07:12:59 fileserver kernel: nlmsvc_retry_blocked(00000000, when=3D=
0)
  May  6 07:13:00 fileserver kernel: lockd: request from 0a010005
  May  6 07:13:00 fileserver kernel: lockd: nlm_lookup_host(0a010005, p=3D1=
7, v=3D4)
  May  6 07:13:00 fileserver kernel: lockd: get host 10.1.0.5

* state in /var/lib/nfs/sm
Server:
drwxr-xr-x    2 root     root         4096 May  6 07:30 .
drwxr-xr-x    4 root     root         4096 May  5 11:36 ..
=2D -rw-------    1 root     root            0 Apr 16 07:13 10.1.0.5   (whi=
ch is the terminalserver)

Client:
drwxr-xr-x    2 root     root         4096 16. Apr 07:13 .
drwxr-xr-x    4 root     root         4096 15. Apr 18:38 ..
=2D -rw-------    1 root     root            0 16. Apr 07:13 10.1.0.2  (whi=
ch is the fileserver)

This is no permission problem, since rpc.statd is running as root and there=
fore is able to
write here.


The solution:

Is missing. I'm willing to debug this even deeper, but my knowledge of nfs =
is limited. Are there any obvious
parameters I can tune? I've read many messages about failing UMON requests,=
 but there never was a solution.

Any help is greatly appreciated,
=2D -Cajus Pollmeier

=2D----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)

iD8DBQE+t3VUtyibJ/7Y+CYRAvGNAKDRQF92MX47J98bjM2CT+KXm1HS9ACg1HJl
PdnHq2/pXlELNwEnk/0T3r4=3D
=3DnXpH
=2D----END PGP SIGNATURE-----



-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: NFS problems, UMON, missing directories, wrong permissions

[Trond Myklebust]

>>>>> " " == Cajus Pollmeier <c.pollmeier@gmx.net> writes:

     > * Randomly missing directories / shredded permissions
     >   Either users don't see shared directories contents or have no
     >   permissions to access them. In fact they have, "id" shows up
     >   with the correct group membership and "ls -la" shows the
     >   directory group writable. After logging out and in again,
     >   everything's fine.

Please note that NFS doesn't support more than 16 groups. If the
permissions depend on the 17th group as listed by 'id', then it
probably won't work.

However that may not be the cause of your problem:

     >   May 6 07:12:59 terminalserver kernel: nsm:
     >   xdr_encode_mon(0a010002, -1249509120, 67108864, 268435456)

That's interesting. Looks like a pretty nasty stack corruption to
me. Are you running an Nvidia-tainted kernel by any chance?

Cheers,
  Trond


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: NFS problems, UMON, missing directories, wrong permissions

[Trond Myklebust]

     >>   May 6 07:12:59 terminalserver kernel: nsm:
     >>   xdr_encode_mon(0a010002, -1249509120, 67108864, 268435456)

     > That's interesting. Looks like a pretty nasty stack corruption
     > to me. Are you running an Nvidia-tainted kernel by any chance?

Duh. Forget that... For some reason or another xdr_encode_mon() has
been set up to print its arguments using network-order (and in
decimal). That's almost worth a bugfix in itself...

OK. The bug appears to be in userland then, since the UNMON request is
correct, and the /var/lib/nfs/sm does contain the 10.0.0.2 address.

Cheers,
  Trond


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: NFS problems, UMON, missing directories, wrong permissions

[Cajus Pollmeier]

=2D----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Am Dienstag, 6. Mai 2003 13:48 schrieb Trond Myklebust:
> >>>>> " " =3D=3D Cajus Pollmeier <c.pollmeier@gmx.net> writes:
>      > * Randomly missing directories / shredded permissions
>      >   Either users don't see shared directories contents or have no
>      >   permissions to access them. In fact they have, "id" shows up
>      >   with the correct group membership and "ls -la" shows the
>      >   directory group writable. After logging out and in again,
>      >   everything's fine.
>
> Please note that NFS doesn't support more than 16 groups. If the
> permissions depend on the 17th group as listed by 'id', then it
> probably won't work.

Hi Trond,

thanks for the quick response...
I didn't know that group membership is limited to 16 when using
nfs. Just checked one user affected by this, he's in 18 groups...

> However that may not be the cause of your problem:
>      >   May 6 07:12:59 terminalserver kernel: nsm:
>      >   xdr_encode_mon(0a010002, -1249509120, 67108864, 268435456)
>
> That's interesting. Looks like a pretty nasty stack corruption to
> me. Are you running an Nvidia-tainted kernel by any chance?

Beware. Its a server :-)

=2D --<some merging>--

> Duh. Forget that... For some reason or another xdr_encode_mon() has
> been set up to print its arguments using network-order (and in
> decimal). That's almost worth a bugfix in itself...
>=20
> OK. The bug appears to be in userland then, since the UNMON request is
> correct, and the /var/lib/nfs/sm does contain the 10.0.0.2 address.

So if the above is not the only problem, I can compile a newer set of=20
nfs-utils to check if I run in problems with them.

Again - thanks,
Cajus
=2D----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)

iD8DBQE+t7pqtyibJ/7Y+CYRAk9dAJsH4lm5I2/GEpITvVgS+Uao1Uw7/gCgkj/S
x/9g/Rhlyogh3NA+p7EN7b0=3D
=3DLcxb
=2D----END PGP SIGNATURE-----



-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs