Re: port-arm/50764 - kernel tries to load already active modules

[Martin Husemann <martin%duskware.de@localhost>]

The following reply was made to PR kern/50764; it has been noted by GNATS.

From: Martin Husemann <martin%duskware.de@localhost>
To: Paul Goyette <paul%whooppee.com@localhost>
Cc: gnats-bugs%NetBSD.org@localhost
Subject: Re: port-arm/50764 - kernel tries to load already active modules
Date: Fri, 5 Feb 2016 11:08:47 +0100

 On Fri, Feb 05, 2016 at 05:37:10PM +0800, Paul Goyette wrote:
 > The real question that should be answered is whether or not the 
 > executable being loaded really needs an interpreter other than those 
 > that are already available.  /usr/tests/lib/libc/gen/posix_spawn/t_spawn 
 > is cited as an example of a trigger for this behavior - what does 
 > file(1) say about t_spawn?
 
 It is not t_spawn itself (that runs fine), but an explicit attempt to
 run a bogus binary:
 
 	posix_spawn an invalid binary
 
 hexdump -C h_zero 
 00000000  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
 *
 00000800
 
 So this part is no suprise.
 
 > Of course, the crash you noted in 50760 should not happen, whether or 
 
 Right, it is due to unaligned relocations in debug info. We need to fix
 that.
 
 > not the module is already loaded.  (You might enable sysctl variable 
 > kern.module.verbose to get some additional hints/clues.)
 
 It did not print anything.
 
 I can easily test, just point me at some code where to add printfs and
 I'll check (but I'm not in a hurry - this is a great test platform right
 now to fix the arm unaligned relocs).
 
 Martin