Re: PR/55164 CVS commit: src/sys/netinet6

[Robert Elz <kre%munnari.OZ.AU@localhost>]

The following reply was made to PR kern/55164; it has been noted by GNATS.

From: Robert Elz <kre%munnari.OZ.AU@localhost>
To: gnats-bugs%netbsd.org@localhost
Cc: 
Subject: Re: PR/55164 CVS commit: src/sys/netinet6
Date: Mon, 13 Apr 2020 03:36:30 +0700

     Date:        Sun, 12 Apr 2020 19:30:02 +0000 (UTC)
     From:        Andreas Gustafsson <gson%gson.org@localhost>
     Message-ID:  <20200412193002.D64E21A9219%mollari.NetBSD.org@localhost>
 
   |  Since this is just one test run, there is a possiblity that it was
   |  some kind of random failure,
 
 It looks very peciuliar if that was a rump panic.
 
 As I suspected, the "rn_init 2" panic comes from rn_init() in net/radix.c
 which is the same thing that does the
 
 	rn_init: radix functions require max_keylen be set
 
 printf (log).     In fact. in rump I would expect that printf to occur
 every time (ever time when rn_init() is called, as max_keylen looks like
 it can never be anything but 0 in a rump kernel (unless there's something
 I'm getting incorrect).   If that happens, then rn_init() simply returns,
 and never gets further in where the "rn_init 2" panic appears...
 
         if (rn_inithead((void *)&mask_rnhead, 0) == 0)
                 panic("rn_init 2");
 
 Whatever is happening here, it doesn't look to be related to routing in
 any specific way, and certainly not to Roy's fix for this PR.
 
 It looks as if the only way that rn_inithead() can return 0 is if malloc()
 fails.   The malloc() call (R_Malloc() macro) there uses M_NOWAIT so
 failure really out to be expected, sometimes.
 
 kre