[ENBD] Problem with enbd 2.2.27

[Peter T. Breuer]

"A month of sundays ago Richard Sharpe wrote:"
> I am having problems with enbd 2.2.27.

Yes, but what exactly?

> I am running the client and server on RH 6.2 with the RH 2.2.14 kernel that
> RH distributes ...
> I have run up the server on one machine, serving a 1GB file.
> On the client, I have run up nbd-client and am tracing it with Ethereal.
> 
> What I observe is the following:
> 
>    Client           Server
> 
>             ---> SYN from port 1037 to 1077
>             <--- SYN, ACK
>             ---> ACK
>             .... some time passes
>             <--- SYN, ACK
>             ---> ACK
>             .... some time passes
>             <--- SYN, ACK
>             ---> ACK
> 
> and this continues. It is as if the server is not seeing the ACK that comes

There is no need to guess about what the enbd code is doing from a
network trace (interesting as it is!).  Compile with "make DEBUG=1" (or
2) and you will get very complete debugging traces in the syslogs of the
two machines.

> back from the client, and keeps trying to bring up the link. However I can

And does the link come up?

> get telnet back and forth between the two machines ... 
> 
> On the server, I see a connection stuck in SYN_RECV state when I use
> 'netstat -an'

I would guess that it is waiting for a connection. Why are there TWO
different ports involved (1037 and 1077)? One of them must be the
designated  NBD port for a connection, and the other must be the
hidden port supplied by the system during the connection? And why
is there ONLY one nbd port? There would be more if the connection used
multiple channels, as it ought to do for safety.

> Has anyone seen this? Can anyone suggest a fix.

Not without a clue. How about the command lines? 

> One system is a dual-celeron on BP6 mobo with 256MB and the other is an AMD
> K6-2/400 with 128MB. Both have 10/100 PCI NICs along with 10Mb/s ISA NICs.
> I am trying to get the enbd stuff to work over the 100Mb/s links.

Does it work over the 10BT link? Does the "make test" work would be my
first question. If the answer is affirmative, then change one of SERVER
or CLIENT in the makefile to the name of a remote host and tell if
_that_ works. And compile with DEBUG set so you can see what happens.

Send me the debug traces privately, as they can be large. But I suspect
you have a wrong command line and that is all.


Peter