Re: [ES40-developers] FW: ES40 and TCPIP PING/SIZE=1785 failing (worksup to 1784)

[Hittner, D. T (IS) <dav...@ng...>]

Based on Volker's ethereal observations, I would guess that:

1) The NIC ring buffer processing is probably working OK, since it's
working up 1784, and had to have processed at least two ring buffers
correctly to get to that point.

2) It's likely, based on my SIMH VAX experience, that TCPIP is
performing a split buffer calculation on the ES40 that is returning
incorrect results, or isn't setting overflow/underflow flags correctly,
so it's branching incorrectly and thus setting the ring buffers wrong. I
couldn't get SIMH VAX network processing to work properly until a few
VAX instructions had been corrected as part of another patch; I didn't
have enough networking sources to debug it effectively.

Fortunately, Volker has found a reproducible breaking point!

A way to debug this might be to:

A) Set a breakpoint in the NIC transmission code when executing the PING
/PACKET_SIZE=1783 code, and see what is in the ring buffers.
B) Repeat with /PACKET_SIZE=1784, logging the ES40 instructions executed
during the (correct) transmission of the 1784 packet.
C) Repeat with /PACKET_SIZE=1785, logging the ES40 instructions executed
during the (incorrect) transmission of the 1785 packet.
D) Compare the two instruction streams and see where they deviate.

If they do, hand walk the operations just before and after the deviation
and see if you can spot the errata. If they don't differ, walk all of
the instructions :-( and see if they are setting values and
overflow/underflow processing correctly.

If you have the VMS sources for TCP/IP and/or the NIC drivers, you could
reduce the search by finding the packet splitting code for the ring
buffers and see what instructions need to be checked.

Alternatively, either the ES40 IDB or your local IDE might support
putting a watchpoint (when a data value is modified) on the ring buffer
data, so you can walk the instruction stream during the packet split and
fill.

Dave

> -----Original Message-----
> From: Camiel Vanderhoeven [mailto:ca...@ca...] 
> Sent: Monday, April 13, 2009 12:40 PM
> To: 'ES40 Developer Discussions'
> Subject: [ES40-developers] FW: ES40 and TCPIP PING/SIZE=1785 
> failing (worksup to 1784)
> 
> Hello developers,
> 
> Would any of you care to dive into this? Solving this problem 
> might make the ES40's networking functional at last.
> 
> Kind regards,
> 
> Camiel Vanderhoeven
> 
> -----Original Message-----
> From: Volker Halle [mailto:Vol...@in...]
> To: Camiel Vanderhoeven
> Subject: ES40 and TCPIP PING/SIZE=1785 failing (works up to 1784)
> 
> Camiel,
> 
> I wondered, why FTPing binary files did not work and tried a 
> couple of PING tests with the ES40 emulator. I seem to have 
> stumbled over something:
> 
> If you use $ TCPIP PING/SIZE=n host to ping from the ES40 
> emulator to some other node in the LAN by varying the ping 
> packet size, everything works fine, until you exceed n=1784. 
> Then you can't get any successful ping through !
> 
> If I look at an Ethereal packet trace for n=1784, I see one 
> packet with 1514 bytes followed by another packet of 346 
> bytes on the wire.
> 
> For n=1785, I see the first packet with only 866 bytes, 
> followed by a packet of size 347, although the fragment 
> offset for the 2nd packet correctly says 1480.
> 
> Trying to debug this should be much easier than trying to 
> debug a hanging FTP session...
> 
> Regards,
> 
> Volker Halle
> 
>  
> 
> __________ Information from ESET NOD32 Antivirus, version of 
> virus signature database 4004 (20090413) __________
> 
> The message was checked by ESET NOD32 Antivirus.
> 
> http://www.eset.com
>  
> 
> 
> --------------------------------------------------------------
> ----------------
> This SF.net email is sponsored by:
> High Quality Requirements in a Collaborative Environment.
> Download a free trial of Rational Requirements Composer Now!
> http://p.sf.net/sfu/www-ibm-com
> _______________________________________________
> Es40-developers mailing list
> Es4...@li...
> https://lists.sourceforge.net/lists/listinfo/es40-developers
>