Re: Question on portmapper source

[Ben Greear <greearb@candelatech.com>]

On 06/07/2011 12:52 AM, Max Matveev wrote:
> On Mon, 06 Jun 2011 20:52:39 -0700, Ben Greear wrote:
>
>   greearb>  On 06/06/2011 06:11 PM, Chuck Lever wrote:
>   >>
>   >>  On Jun 6, 2011, at 8:16 PM, Ben Greear wrote:
>   >>
>   >>>  What is the official upstream project that contains the portmapper
>   >>>  source?
>   >>
>   >>  Do you want pormapper or rpcbind?  The newer rpcbind actually allows RPC services to register at specific addresses, but the legacy portmapper does not.
>
>   greearb>  I don't know.  I was testing my nfs binding patches (which
>   greearb>  fixes up some of the kernel processes, at least), but when
>   greearb>  trying to connect through a router, we noticed that this
>   greearb>  traffic was using the wrong source IP (ie, it wasn't bound):
>
>   greearb>     0.000000 192.168.100.117 ->  172.16.0.100 TCP 33282>  111 [SYN] Seq=0
>   greearb>  Win=14600 Len=0 MSS=1460 SACK_PERM=1 TSV=6981376 TSER=0 WS=10
>   greearb>     3.007677 192.168.100.117 ->  172.16.0.100 TCP 33282>  111 [SYN] Seq=0
>   greearb>  Win=14600 Len=0 MSS=1460 SACK_PERM=1 TSV=6984384 TSER=0 WS=10
>   greearb>     9.023670 192.168.100.117 ->  172.16.0.100 TCP 33282>  111 [SYN] Seq=0
>   greearb>  Win=14600 Len=0 MSS=1460 SACK_PERM=1 TSV=6990400 TSER=0 WS=10
>   greearb>    10.010239 192.168.100.117 ->  172.16.0.100 Portmap V2 GETPORT Call NFS(100003)
>   greearb>  V:3 UDP
>   greearb>    11.011269 192.168.100.117 ->  172.16.0.100 Portmap [RPC retransmission of #4]V2
>   greearb>  GETPORT Call NFS(100003) V:3 UDP
>   greearb>    12.012298 192.168.100.117 ->  172.16.0.100 Portmap [RPC retransmission of #4]V2
>   greearb>  GETPORT Call NFS(100003) V:3 UDP
>   greearb>    20.422093 192.168.100.117 ->  172.16.0.100 TCP 44732>  111 [SYN] Seq=0
>   greearb>  Win=14600 Len=0 MSS=1460 SACK_PERM=1 TSV=7001798 TSER=0 WS=10
>   greearb>    23.423683 192.168.100.117 ->  172.16.0.100 TCP 44732>  111 [SYN] Seq=0
>   greearb>  Win=14600 Len=0 MSS=1460 SACK_PERM=1 TSV=7004800 TSER=0 WS=10
>   greearb>    29.439676 192.168.100.117 ->  172.16.0.100 TCP 44732>  111 [SYN] Seq=0
>   greearb>  Win=14600 Len=0 MSS=1460 SACK_PERM=1 TSV=7010816 TSER=0 WS=10
>   greearb>    30.432383 192.168.100.117 ->  172.16.0.100 Portmap V2 GETPORT Call NFS(100003)
>   greearb>  V:3 UDP
>   greearb>    31.433420 192.168.100.117 ->  172.16.0.100 Portmap [RPC retransmission of
>   greearb>  #10]V2 GETPORT Call NFS(100003) V:3 UDP
>   greearb>    32.434458 192.168.100.117 ->  172.16.0.100 Portmap [RPC retransmission of
>   greearb>  #10]V2 GETPORT Call NFS(100003) V:3 UDP
>
>   greearb>  I need to make whatever is generating that traffic bind to a source IP.
>
> It's hard to say what's "rigth" and "wrong" from just this tcpdump:
> there is only one IP address.
>
> Recently there was a change to libtirpc to fix IP address selection
> for UDP datagrams:
>
> http://git.infradead.org/users/steved/libtirpc.git/commit/74ef3df0236c55185225c62fba34953f2582da72
>
> This is on the server side - it looks to me you're trying to deal with
> client side so it may be irrelevant.

It's not really wrong for most users..just for me because I need it (mount.nfs, evidently)
to use a specific source-IP so that my ip routing rules choose the correct route.

>
>   >>>  Also, the mount.nfs package, as I'll probably need to modify that
>   >>>  as well...
>   >>
>   >>  mount.nfs is in nfs-utils.
>
>   greearb>  What is the git repo for it, though?
>
> http://git.infradead.org/users/steved/nfs-utils.git
>
>   >>>  I am going to try to allow portmapper to bind to a particular
>   >>>  source IP when making outbound requests....
>   >>
>   >>  I don't understand.  portmapper, as an RPC service, receives
>   >>  incoming requests.  But we don't use the rpcbind CALLIT procedure
>   >>  on Linux, thus it doesn't make outbound requests.
>
>   greearb>  I could easily be confused about where the problem lies.
>   greearb>  Any idea what is generating those requests on the nfs client
>   greearb>  above?
>
> mount.nfs does.

Thanks, I'll dig into the mount.nfs code.

Thanks,
Ben

>
> max
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html


-- 
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc  http://www.candelatech.com