Re: Today's -current no longer boots on my amd64 machine

[Jean-Yves Migeon <jeanyves.migeon%free.fr@localhost>]

On Mon, 19 Apr 2010 17:37:00 +0800, Dennis Ferguson
<dennis.c.ferguson%gmail.com@localhost> wrote:
> On 19 Apr 2010, at 14:54 , Jean-Yves Migeon wrote:
> 
>> This is probably related to my recent change. Could you try with the
>> following patch?
> 
> Thanks for that, but no luck.  nm on identcpu.o before and after shows
> the call to the function was in fact removed from the new kernel, but
> it still reboots not many milliseconds after printing the cpu0 at
mainbus0
> line.

So right after it jumps inside cpu_hatch() when starting secondary CPUs I
guess.

Try reverting this diff:

Index: src/sys/arch/amd64/amd64/mptramp.S
diff -u src/sys/arch/amd64/amd64/mptramp.S:1.10
src/sys/arch/amd64/amd64/mptramp.S:1.11
--- src/sys/arch/amd64/amd64/mptramp.S:1.10     Fri Nov 27 03:23:04 2009
+++ src/sys/arch/amd64/amd64/mptramp.S  Sun Apr 18 23:47:50 2010
@@ -1,4 +1,4 @@
-/*     $NetBSD: mptramp.S,v 1.10 2009/11/27 03:23:04 rmind Exp $       */
+/*     $NetBSD: mptramp.S,v 1.11 2010/04/18 23:47:50 jym Exp $ */

 /*-
  * Copyright (c) 2000 The NetBSD Foundation, Inc.
@@ -223,15 +223,7 @@
        testq   %rdi, %rdi
        jz      1b

-       movl    _C_LABEL(cpu_feature),%eax
-       andl    $CPUID_NOX,%eax
-       jz      1f
-       movl    $MSR_EFER,%ecx
-       rdmsr
-       orl     $EFER_NXE,%eax
-       wrmsr
 1:
-
        movq    CPU_INFO_IDLELWP(%rdi),%rsi
        movq    L_PCB(%rsi),%rsi

or if you like it in a more violent way, the whole NX fix:
http://mail-index.netbsd.org/source-changes/2010/04/18/msg008972.html for
now.

I'll have a look at it later today.

-- 
Jean-Yves Migeon
jeanyves.migeon%free.fr@localhost