[En-Nut-Discussion] changes in tcpip stack from 3.9.5 to 3.9.6

[William Baker]

I hope you will forgive the bad etiquite of repost, but perhaps the 
subject could have been clearer.

I have an ethernut 2.1b board, and under 3.9.6 the network is unreliable/
unstable.  I have down downgraded to 3.9.5, and the board is not showing
any signs of network problems found in 3.9.6.

There do appear to be some problems with the 3.9.6 tcpip stack.  Whether
these problems are in any way responsible for the unstable behavior, I
can't say.  Instability in 3.9.6 could have been caused by other factors.
However, on my network the ARP behavior is definately much cleaner and 
well behaved in 3.9.5.

As I gain experience with NutOS and this board, I may attempt to diagnose
the problem further.  I suspect that by then there will be a 3.9.6a.  I'm
willing to experament with anyone's suggestions.

It would also be nice to hear if anyone else has had these problems, or
even if the device has been put in an active desktop environment without
problems.

bbaker

----------------------------------------------
previous post below
----------------------------------------------

I installed and ran the httpserv-er from 3.9.6.  When connected to my 
network, communication is very intermittant.  When connected on an 
isolated subnet, communication is clean and reliable.  Here is a simple 
ping log while the device is connected to my primary network.  Notice 
the dropped packets and 3+ sec responses.

64 bytes from 192.168.4.160: icmp_seq=25 ttl=64 time=0.757 ms
64 bytes from 192.168.4.160: icmp_seq=26 ttl=64 time=0.760 ms
64 bytes from 192.168.4.160: icmp_seq=27 ttl=64 time=0.766 ms
64 bytes from 192.168.4.160: icmp_seq=28 ttl=64 time=0.758 ms
64 bytes from 192.168.4.160: icmp_seq=29 ttl=64 time=0.764 ms
64 bytes from 192.168.4.160: icmp_seq=30 ttl=64 time=0.758 ms
64 bytes from 192.168.4.160: icmp_seq=35 ttl=64 time=3698 ms
64 bytes from 192.168.4.160: icmp_seq=43 ttl=64 time=3438 ms
64 bytes from 192.168.4.160: icmp_seq=47 ttl=64 time=0.763 ms
64 bytes from 192.168.4.160: icmp_seq=48 ttl=64 time=0.784 ms
64 bytes from 192.168.4.160: icmp_seq=49 ttl=64 time=0.763 ms

Capturing packets over the network shows that the device is issueing too 
many ARP requests, which appear to be associated with NetBIOS browser 
announcements, ie, the nut/os sends ARP requests each time it sees the 
BROWSER packet.  Here is a sample network capture.  The Ethernut 2.1 
device is IP 192.168.4.160.

     41 14.847438   192.168.4.53          192.168.4.255         BROWSER  
Host Announcement POWDERCOAT-XP, Workstation, Server, NT Workstation, 
Potential Browser
     42 14.848502   192.168.4.160         Broadcast             ARP      
Who has 192.168.4.53?  Tell 192.168.4.160
     43 15.335182   192.168.4.160         Broadcast             ARP      
Who has 192.168.4.53?  Tell 192.168.4.160
     44 16.460034   192.168.4.160         Broadcast             ARP      
Who has 192.168.4.53?  Tell 192.168.4.160
     45 16.846605   192.168.4.1           192.168.4.255         NBNS     
Registration NB PRIEFERT<1d>
     46 18.585031   192.168.4.160         Broadcast             ARP      
Who has 192.168.4.53?  Tell 192.168.4.160
     47 19.022508   192.168.4.102         Broadcast             ARP      
Who has 192.168.4.14?  Tell 192.168.4.102
     48 19.470393   Hewlett-_b3:22:80     09:00:09:00:00:67     HPEXT    
U, func=UI; DSAP HP Extended LLC Individual, SSAP HP Extended LLC 
Command; HPEXT; DXSAP 0623, SXSAP 0623
     49 19.594466   192.168.4.58          Broadcast             ARP      
Who has 192.168.4.14?  Tell 192.168.4.58
     50 20.337508   192.168.4.50          192.168.4.255         NETLOGON 
SAM LOGON request from client
     51 20.675168   192.168.4.151         Broadcast             ARP      
Who has 192.168.4.1?  Tell 192.168.4.151
     52 22.531522   192.168.4.79          Broadcast             ARP      
Who has 192.168.4.14?  Tell 192.168.4.79
     53 24.662008   192.168.4.146         Broadcast             ARP      
Who has 192.168.4.14?  Tell 192.168.4.146
     54 28.022267   0.0.0.0               255.255.255.255       NTP      NTP
     55 28.074988   192.168.4.45          192.168.4.255         BROWSER  
Host Announcement CEASLEY_ZEBRA, Workstation, Server, Domain Member 
Server, Print Queue Server
     56 28.076031   192.168.4.160         Broadcast             ARP      
Who has 192.168.4.45?  Tell 192.168.4.160
     57 28.522581   192.168.4.160         Broadcast             ARP      
Who has 192.168.4.45?  Tell 192.168.4.160
     58 29.647425   192.168.4.160         Broadcast             ARP      
Who has 192.168.4.45?  Tell 192.168.4.160

It's also strange to see the ARP request sent multiple times.  Hence, it 
appears that both ARP and NetBIOS services exhibit some breakage.

I don't think you can disable ARP and still have a functional server 
TCP/IP device, can you?
Can the NetBIOS Wins responder be disabled?  It does not appear to be 
possible with the configurator.  Can I edit UserConf.mk to add 
parameters to disable this feature?  Where do I find the variables that 
are understood in UserConf.mk?

I regret to say I have no experience with 3.9.5 and cannot currently 
verify its behavior.

Other than that, Nut/OS is really impressive, and I look forward to 
seeing it evolve over the years.

bbaker