kern/52608: Unexpected ATF test failure and fixes of IPsec on sparc/GuruPlug

[ozaki-r%netbsd.org@localhost]

>Number:         52608
>Category:       kern
>Synopsis:       Unexpected ATF test failure and fixes of IPsec on sparc/GuruPlug
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Tue Oct 10 02:00:00 +0000 2017
>Originator:     Ryota Ozaki
>Release:        -current
>Organization:
IIJ
>Environment:
rump kernels on sparc and GuruPlug
>Description:
Some ATF tests related to IPsec on sparc/GunuPlug had been failing (kernel panic)
for long time:
  http://releng.netbsd.org/b5reports/sparc/build/2017.09.26.04.34.59/test.html#failed-tcs-summary
  http://www.netbsd.org/~martin/evbarm-atf/Y2017_09/177_atf.html#failed-tcs-summary
  
However the tests began to be success unexpectedly:
  http://releng.netbsd.org/b5reports/sparc/build/2017.09.26.07.42.06/test.html#failed-tcs-summary
  http://www.netbsd.org/~martin/evbarm-atf/Y2017_09/178_atf.html#failed-tcs-summary

The shift happened since the following commit:
  http://cvsweb.netbsd.org/bsdweb.cgi/src/sys/sys/mbuf.h?only_with_tag=MAIN#rev1.171

(There are other suspect commits but it seems that only the above commit likely affects
the issue.)

However the change is just to enlarge the size of struct pkthdr (or change the memory
layout of it slightly). I guess that the change actually doesn't fix an issue and just hides
the issue luckily (i.e., the tests had been failing unluckily).

The issue may be rump kernel specific.

>How-To-Repeat:
See the above URLs.
(I have not run the ATF tests on the machines in question by myself. )
>Fix:
n/a