From owner-freebsd-current Fri Dec 6 00:26:47 1996 Return-Path: Received: (from root@localhost) by freefall.freebsd.org (8.8.4/8.8.4) id AAA06689 for current-outgoing; Fri, 6 Dec 1996 00:26:47 -0800 (PST) Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19]) by freefall.freebsd.org (8.8.4/8.8.4) with ESMTP id AAA06649 for ; Fri, 6 Dec 1996 00:26:36 -0800 (PST) Received: (from bde@localhost) by godzilla.zeta.org.au (8.8.3/8.6.9) id TAA17211; Fri, 6 Dec 1996 19:23:44 +1100 Date: Fri, 6 Dec 1996 19:23:44 +1100 From: Bruce Evans Message-Id: <199612060823.TAA17211@godzilla.zeta.org.au> To: current@freebsd.org, markm@iafrica.com Subject: Re: 3.0-current Kernel panicking on bootup Sender: owner-current@freebsd.org X-Loop: FreeBSD.org Precedence: bulk >I have a 386sx/40 which is consistently panicking on bootup. This is >happening in the igmp_init() routine, near as I can determine, and >the error message at panic time is "priveliged instruction fault >while in kernel mode". The instruction is probably `bswap'. `bswap' is used in the kernel unless I386_CPU is defined when is included, but I386_CPU is almost never defined then because does not included "opt_cpu.h" where I386_CPU is defined if it is defined at all. This bug was introduced on Nov 29. It can't be fixed easily by including "opt_cpu.h" when KERNEL is defined, since "opt_cpu.h" doesn't exist for LKMs or for applications that #define KERNEL before including some system headers. Such applications may be broken already. It's too easy to #include nested in some other header and get the bswap version even when KERNEL is #undef'ed before including or explicitly. Bruce