Am 13.02.2003 23:23:15, schrieb "Dave Hooper" <[EMAIL PROTECTED]>:

>Someone was talking about this earlier, I wasn't involved in that discussion
>because I felt I had nothing to add but a netstat performed earlier showed
>this (edited for conciseness):

This time I used TCPview, it shows the process which created the connection or
listener:
http://www.sysinternals.com/ntw2k/source/tcpview.shtml

[snip: your netstat output]

>  TCP    b-a-d-sector:4111      b-a-d-sector:8481      TIME_WAIT
>  TCP    b-a-d-sector:4116      b-a-d-sector:8481      ESTABLISHED
>  TCP    b-a-d-sector:8481      b-a-d-sector:0         LISTENING
>  TCP    b-a-d-sector:8481      b-a-d-sector:4116      ESTABLISHED

You asked about this connections: Were you using a Freenet client program
over FCP? If yes then everything is ok, 8481 is one endpoint of the TCP-
connection and 4116 is the other end. Both are on localhost, so this
connection is listed twice.

[snip: your node's connection via FNP]
>
>After finishing retrieving, some time (thirty minutes?) later ...:
>[why did it take so long for connection to some of the node refs to be torn
>down?]

There's a timeout on FNP connections. (10 min? look in the configfile)
To disable hundreds of TCP establishing and closing events the nodes
hold a connection for some time.

>  Proto  Local Address          Foreign Address        State
>  TCP    b-a-d-sector:8888      b-a-d-sector:0         LISTENING
>  TCP    b-a-d-sector:8891      b-a-d-sector:0         LISTENING
>  TCP    b-a-d-sector:*****     b-a-d-sector:0         LISTENING
>  TCP    b-a-d-sector:8481      b-a-d-sector:0         LISTENING
>
>
>Is this expected?  Could anyone quickly run me through the phases involved
>here and the reasoning for, e.g. connections from 127.0.0.1:4118 to
>127.0.0.1:8481?  Also if there's any way to control the allocation range of

Freenet uses only the FNP-port to talk to the internet. But look at bottom,
I discovered some strange things.

>ephemeral ports used during data transfer (assuming that's what they are -
>for some reason I was expecting all traffic to be on my 'advertised' port of
>*****, although I don't really know why).

Connections to 8481 (FCP) should only occour over LAN or localhost.
(Except you run a really public node with public FCP)

>
>
>d
>


I found some strange things on Win2k with TCPview:

[snip: Microsoft-stuff]
[snip: FCP-connections to Fred as seen by Fred's process]

[1244 is processnumber of Fred; example of strange entry about FNP:]
java.exe:1244   TCP     acer:4935       acer:0  LISTENING       
java.exe:1244   TCP     acer:4935       216.87.111.151:17034    ESTABLISHED     

[1740 is Frost; example of strange entry about FCP:]
java.exe:1740   TCP     acer:2857       acer:0  LISTENING       
java.exe:1740   TCP     acer:2857       localhost:8481  ESTABLISHED     

[snip: a lot of connections to other nodes]

[Fred is listening and hold on FNP-port at same time (it's OK, like a server) :]
java.exe:1244   TCP     acer:16634      acer:0  LISTENING       
java.exe:1244   TCP     acer:16634      dyn-bway-130-206.dyn.columbia.edu:4216  
ESTABLISHED     

[snip: other inbound contacts]
[snip: other connections]

There are ports listed twice!?!? It seems that every connection established
from a Java process has a listener on the same port. The more FCP/FNP connections,
the more listening ports...
But I could never connect to a port with
Netcat, it's everytime refused. Is the monitoring program too slow
to show the real state of a socket? Is that a problem with Sun's
Java implementation?
Does that happen on Linux too?


Greetings,
Stef





_______________________________________________
devl mailing list
[EMAIL PROTECTED]
http://hawk.freenetproject.org:8080/cgi-bin/mailman/listinfo/devl

Reply via email to