Re: misc/38425 (savecore confused if getbootfile(3) fails)

[Simon Burge <simonb%NetBSD.org@localhost>]

The following reply was made to PR misc/38425; it has been noted by GNATS.

From: Simon Burge <simonb%NetBSD.org@localhost>
To: yamt%mwd.biglobe.ne.jp@localhost (YAMAMOTO Takashi)
Cc: tls%rek.tjls.com@localhost, dogcow%babymeat.com@localhost, 
gnats-bugs%NetBSD.org@localhost,
    misc-bug-people%netbsd.org@localhost, gnats-admin%netbsd.org@localhost,
    netbsd-bugs%netbsd.org@localhost
Subject: Re: misc/38425 (savecore confused if getbootfile(3) fails) 
Date: Mon, 15 Jun 2009 09:00:41 +1000

 YAMAMOTO Takashi wrote:
 
 > > On Sun, Jun 14, 2009 at 06:50:17AM -0700, Tom Spindler wrote:
 > >> I don't know about any of you, but savecore completely fails to work
 > >> for me after 1.78 savecore.c; in particular, failing to set the
 > >> booted kernel causes check_space to die a painful death:
 > > 
 > > I'll poke at it some more.  This works in our local tree -- maybe I
 > > missed a change when integrating it back into NetBSD.
 > > 
 > > It is absolutely, positively *not* the right solution to cause any
 > > of these utilities to try to know the kernel name.  Any operation
 > > they have any business doing can be done via /dev/ksyms -- the
 > > question is where I missed one or more references to the kernel name
 > > that were exposed beyond libkvm, which correctly defaults to use
 > > /dev/ksyms if some other kernel name is not rammed down its throat.
 > 
 > talking about the right solution, savecore should not use /dev/ksyms
 > or info from the live kernel at all.  we should store necessary info
 > in the dump itself.  (as freebsd does?)
 
 I had thought that /dev/ksyms seemed right, but that obviously doesn't
 help if the kernel that left the core wasn't the currently running
 kernel (as Yamaoto aludes to).
 
 Cheers,
 Simon.