Re: "RPC: program not registered" with new NFS server?

On Sun, May 1, 2011 at 8:50 PM, Garrett Cooper <yanegomi_at_gmail.com> wrote:
> On Sun, May 1, 2011 at 2:37 PM, Rick Macklem <rmacklem_at_uoguelph.ca> wrote:
>>> Hi Rick, et all,
>>> I upgraded to a later kernel on two of my machines and am running
>>> into issues starting up the nfs kernel. Every time I try mounting like
>>> so:
>>>
>>> # mount -t nfs localhost:/scratch/ /mnt/
>>>
>>> or like so:
>>>
>>> # mount -t oldnfs localhost:/scratch/ /mnt/
>>>
>>> I run into this error:
>>>
>>> [tcp] localhost:/scratch: RPCPROG_NFS: RPC: Program not registered
>>> [tcp6] localhost:/scratch: RPCPROG_NFS: RPC: Program not registered
>>>
>>> I kldloaded nfsd, and then could start the mountd and nfsd
>>> services (this made some of my problems go away, in particular
>>> showmount -e looks sane), but things aren't sane. I know that nfs
>>> client capability works because I can mount remote NFS shares via amd
>>> and raw nfs mounts with another machine that I haven't upgraded and
>>> things are fine, but the server appears to be completely broken on my
>>> machines.
>>> Here is the configuration:
>>>
>>> $ grep NFS /root/FALLOUT
>>> #options NFSCL
>>> #options NFSCLIENT # Network Filesystem Client
>>> #options NFSSERVER # Network Filesystem Server
>>> #options NFSLOCKD # Network Lock Manager
>>> #options NFS_ROOT # NFS usable as /, requires NFSCLIENT
>>> $ grep nfs /etc/src.conf
>>> MODULES_OVERRIDE+= krpc nfscommon nfscl nfsclient nfsd nfslockd
>>> nfsserver
>>>
>>> # rc.conf snippet...
>>>
>>> rpcbind_enable="YES"
>>> mountd_enable="YES"
>>> rpc_lockd_enable="YES"
>>> rpc_statd_enable="YES"
>>> nfsd_enable="YES"
>>>
>> Well, I think the rc variable has always been:
>> nfs_server_enable="YES"
>>
>> which still works for the new one as well as old one.
>>
>> The only other thing I can think of is if you are
>> pre-r220510 you need to create an empty stablerestart
>> file before the new nfs server will start.
>>
>> Take a
>> look to see if the nfsd is running via "ps axHl"
>> or similar and check /var/log/messages for nfsd/mountd
>> related errors.
>>
>> If you need to create the stable restart file, just do
>> the following on the server:
>> # install -o root -g wheel -m 600 /dev/null /var/db/nfs-stablerestart
>> (This shouldn't be necessary for post r220510 systems and pre-r220510
>>  systems shouldn't try and run the new server by default, so this
>>  shouldn't be your problem. Just "ls -l /var/db" to see if the file
>>  is there.)
>
> Ok, I finally go things going. Here's what I did:
>
> 1. Added the following modules to MODULES_OVERRIDE:
>
> # grep nfs /etc/src.conf
> MODULES_OVERRIDE+=      krpc nfscl nfscommon nfs_common nfsclient nfsd
> nfslock nfslockd nfsserver nfssvc
>
> Note that there are a lot more nfs services listed here.
>
> 2. Commented out everything sans nfs_server_enable and
> nfs_client_enable -- so now everything NFS is run from the kernel
> instead of userland minus, nfsd.
>
> Now the NFS server and client are working fine via V3 and V4. It's
> important to note this because the module list is less in the old NFS
> server/client setup to get things fully functional. Example:
>
> Old:
>
> $ kldstat -v | egrep 'nfs|krpc'
>                162 nfssvc
>                164 krpc
>                161 nfsserver
>                158 nfs_common
>                163 nfslockd
>                160 nfs
>                159 nfslock
>
> New:
>
> # kldstat -v | egrep 'nfs|krpc'
> 11    1 0xffffffff80e19000 1aab0    nfsclient.ko (/boot/kernel/nfsclient.ko)
>                232 oldnfs
> 12    7 0xffffffff80e34000 13c88    krpc.ko (/boot/kernel/krpc.ko)
>                229 krpc
> 13    2 0xffffffff80e48000 d7e      nfs_common.ko (/boot/kernel/nfs_common.ko)
>                230 nfs_common
> 14    5 0xffffffff80e49000 f52      nfslock.ko (/boot/kernel/nfslock.ko)
>                231 nfslock
> 15    1 0xffffffff80e4a000 1b8f4    nfsserver.ko (/boot/kernel/nfsserver.ko)
>                234 nfsserver
> 16    5 0xffffffff80e66000 422      nfssvc.ko (/boot/kernel/nfssvc.ko)
>                233 nfssvc
> 17    1 0xffffffff80e67000 30887    nfsd.ko (/boot/kernel/nfsd.ko)
>                237 nfsd
> 18    3 0xffffffff80e98000 10a03    nfscommon.ko (/boot/kernel/nfscommon.ko)
>                235 nfscommon
> 19    1 0xffffffff80ea9000 d94a     nfslockd.ko (/boot/kernel/nfslockd.ko)
>                236 nfslockd
> 20    1 0xffffffff80eb7000 35500    nfscl.ko (/boot/kernel/nfscl.ko)
>                239 nfs
>                238 nfscl
>
>    Now the question I have is -- why are nfs_common and nfscommon not
> the same module (I know the sources are different, and the
> functionality is different, so this is more a rhetorical question than
> anything else)?

    And one last parting note for now: I had to whack all of my saved
kernels because it was trying to load mismatched modules, so something
was wonky there with kldload(2), etc. I have a trivial installkernel
script that tacks on the svn tag to $INSTKERNNAME and I swap the
/boot/kernel symlink pointer to test things out from time to time with
different kernels. Not so much nowadays, but back a year ago I did
this frequently because of stability issues.
Thanks!
-Garrett