Julian Elischer <[EMAIL PROTECTED]> wrote:
>John Smith wrote:
>>
>> Hello,
>>
>> Recently while I was playing with netgraph code
>> I noted some interesting comments about 'tracing'
>> netgraph packets and the ability to connect nodes
>> located on different machines. I would like to ask
>> if somebody is working on such a code and if it
>> would be worth writing it. Any additional comments
>> and ideas are also welcome.
>
>If you are playing with it, you may want to let me know your thoughts..
>I know many people are using it for this-and-that but we get
>very little feedback.
Well, I'm not a network administrator, so I haven't
used netgraph for real services. In general, I like
the idea, because I haven't seen yet a 'network
problem' which I can't solve using netgraph. Well,
many times my solution requires some code to be
written, but... this is why I use FreeBSD - I can
always do what I want to. :) In this case, netgraph
helps a lot, because it already has some nodes and
because it has a mechanism for easily implementing new
nodes. One has enough power to solve his particular needs
and there are few posibilities for 'design limitations'.
>
>I'm presently rewriting a large part of netgraph to make is suitable for
>running under SMP without the BGL.. (e.g fine grain locking)
>so I'm interested in wha people think in general, and
>specifically what the liked and didn't like.
>Lastly, if you have used netraph, make sure that you have a look at the
>man page for the -current version to see what ha s been changing.
>
I'm currently running -current... so this is the only code/doc I'm
looking at.
>looking at the extra argumants on the prototypes
>in /sys/netgraph/netgraph.h will show you where to look
>for changes inthe source too.
>
>(in particular tha ability to route messages for flow control has
>been a large change).
>
>
>As to your specific questions, these were vague ideas of things that we
>though were proctical but didn't have time to do.
>
>nodes on differnt machines could be connected by using a udp ksocket
>node as a tunnel.
Well, may be I didn't said exactly what I wanted to.
If we use say, ksocket nodes as a tunnel, we will
transfer the data - ok, but what about metadata?
May be I should say 'to connect two netgraphs'?
May be this is a lost cause, but that's why I'm asking.
>Tracing is simply the ability to add a flag to
>a metadata object, and adding code to the data delivery function
>(ng_send_data()) to somehow emit trace information whnever the metadata
>it is moving has that flag set.
>
>The latter would be very easy to do. (but what logging mechanism
>would you use, and how would you set the bits?)
Well, probably there are several ways to do this -
not quite sure.
At this point I'm just asking for
some feedback :)
____________________________________________________________________
Get free email and a permanent address at http://www.netaddress.com/?N=1
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-hackers" in the body of the message
