#131 Tests ikev2four(v,4) fail due to check sum on big-endian

v4.1
closed-fixed
nobody
None
5
2013-11-06
2011-06-13
Skalnik
No

The ikev2four(v,4) tests fail because of reverse order of bytes in check sum on big endian machines when double eventually triple -v option is used. This fails on System/390 and ppc64. Ppc didn't fail . This issue was described on tcpdump-workers mailing list here

http://www.mail-archive.com/tcpdump-workers@lists.tcpdump.org/msg05309.html

These tests fail for 4.1.1 version as well as on for latest git sources.

Discussion

  • Guy Harris

    Guy Harris - 2011-06-13

    OK, so if by "System/390" you mean "32-bit System/3x0", then that's a 32-bit big-endian architecture on which it *does* fail. Has it been tried on z/Architecture ("64-bit System/3x0", if you will)?

    Or by "System/390" did you mean "z/Architecture", in which case what the failing architectures have in common is that they're big-endian and 64-bit (and from IBM, but I digress... :-)).

     
  • Guy Harris

    Guy Harris - 2011-06-13

    I've checked into the trunk a change to use Wireshark's BSD kernel-derived Internet checksum routine, and get rid of some "by hand" checksumming being done. Does that fix the problem?

     
  • Skalnik

    Skalnik - 2011-06-14

    Thanks for fast reaction.
    I've tested the latest git sources with the checksum of Wireshark but the bytes are still in reverse order.

    The s390 arch fails while s390x works fine. Maybe a bit strange but what about alignment of structure elements?

     
  • Guy Harris

    Guy Harris - 2011-06-17

    The UDP printer was printing a value that was in network byte order, so the value was actually correct on the big-endian machines. Unfortunately, the output files were generated on little-endian machines....

    Like the TCP printer, the UDP printer now prints the actual checksum field, converted to host byte order, as well as the value it was supposed to have, also in host byte order. I've updated the output files to match the new output format. Those changes are in the trunk, and the 4.2 branch; I also propagated the checksum routine changes to the 4.2 branch.

     
  • Skalnik

    Skalnik - 2011-06-22

    Works fine. Thanks for your fast help.

     
  • Guy Harris

    Guy Harris - 2011-10-14
    • status: open --> closed-fixed
     
  • Denis Ovsienko

    Denis Ovsienko - 2013-11-06

    Administrators of the "tcpdump" SourceForge project have superseded this tracker item (formerly artifact 3315714, now bug 131) with issue 135 of the "tcpdump" GitHub project.

     

Get latest updates about Open Source Projects, Conferences and News.

Sign up for the SourceForge newsletter:





No, thanks