From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1946232AbXBIJPG (ORCPT ); Fri, 9 Feb 2007 04:15:06 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1946239AbXBIJPG (ORCPT ); Fri, 9 Feb 2007 04:15:06 -0500 Received: from ozlabs.org ([203.10.76.45]:38688 "EHLO ozlabs.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1946232AbXBIJPE (ORCPT ); Fri, 9 Feb 2007 04:15:04 -0500 Subject: [PATCH 1/10] lguest: Don't rely on last-linked fallthru when no paravirt handler From: Rusty Russell To: lkml - Kernel Mailing List Cc: Andrew Morton , Andi Kleen , virtualization In-Reply-To: <1171012296.2718.26.camel@localhost.localdomain> References: <1171012296.2718.26.camel@localhost.localdomain> Content-Type: text/plain Date: Fri, 09 Feb 2007 20:14:18 +1100 Message-Id: <1171012458.2718.30.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Evolution 2.8.1 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org The current code simply calls "start_kernel" directly if we're under a hypervisor and no paravirt_ops backend wants us, because paravirt.c registers that as a backend and it's linked last. This was always a vain hope; start_kernel won't get far without setup. It's also impossible for paravirt_ops backends which don't sit in the arch/i386/kernel directory: they can't link before paravirt.o anyway. This implements a real fallthrough if we pass all the registered paravirt probes. Signed-off-by: Rusty Russell =================================================================== --- a/arch/i386/kernel/Makefile +++ b/arch/i386/kernel/Makefile @@ -39,8 +39,6 @@ obj-$(CONFIG_EARLY_PRINTK) += early_prin obj-$(CONFIG_EARLY_PRINTK) += early_printk.o obj-$(CONFIG_HPET_TIMER) += hpet.o obj-$(CONFIG_K8_NB) += k8.o - -# Make sure this is linked after any other paravirt_ops structs: see head.S obj-$(CONFIG_PARAVIRT) += paravirt.o EXTRA_AFLAGS := -traditional =================================================================== --- a/arch/i386/kernel/head.S +++ b/arch/i386/kernel/head.S @@ -502,10 +502,11 @@ startup_paravirt: pushl %ecx pushl %eax - /* paravirt.o is last in link, and that probe fn never returns */ pushl $__start_paravirtprobe 1: movl 0(%esp), %eax + cmpl $__stop_paravirtprobe, %eax + je unhandled_paravirt pushl (%eax) movl 8(%esp), %eax call *(%esp) @@ -517,6 +518,12 @@ 1: addl $4, (%esp) jmp 1b + +unhandled_paravirt: + /* Nothing wanted us: try to die with dignity (impossible trap). */ + movl $0x1F, %edx + pushl $0 + jmp early_fault #endif /* =================================================================== --- a/arch/i386/kernel/paravirt.c +++ b/arch/i386/kernel/paravirt.c @@ -481,9 +481,6 @@ static int __init print_banner(void) return 0; } core_initcall(print_banner); - -/* We simply declare start_kernel to be the paravirt probe of last resort. */ -paravirt_probe(start_kernel); struct paravirt_ops paravirt_ops = { .name = "bare hardware",