[BACK]Return to locore.S CVS log [TXT][DIR] Up to [cvs.NetBSD.org] / src / sys / arch / amd64 / amd64

File: [cvs.NetBSD.org] / src / sys / arch / amd64 / amd64 / locore.S (download)

Revision 1.22, Thu May 17 14:51:14 2007 UTC (16 years, 11 months ago) by yamt
Branch: MAIN
Changes since 1.21: +109 -425 lines

merge yamt-idlelwp branch.  asked by core@.  some ports still needs work.

from doc/BRANCHES:

	idle lwp, and some changes depending on it.

	1. separate context switching and thread scheduling.
	   (cf. gmcgarry_ctxsw)
	2. implement idle lwp.
	3. clean up related MD/MI interfaces.
	4. make scheduler(s) modular.

/*	$NetBSD: locore.S,v 1.22 2007/05/17 14:51:14 yamt Exp $	*/

/*
 * Copyright-o-rama!
 */

/*
 * Copyright (c) 2001 Wasabi Systems, Inc.
 * All rights reserved.
 *
 * Written by Frank van der Linden for Wasabi Systems, Inc.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *      This product includes software developed for the NetBSD Project by
 *      Wasabi Systems, Inc.
 * 4. The name of Wasabi Systems, Inc. may not be used to endorse
 *    or promote products derived from this software without specific prior
 *    written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
 * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
 * POSSIBILITY OF SUCH DAMAGE.
 */


/*-
 * Copyright (c) 1998, 2000, 2007 The NetBSD Foundation, Inc.
 * All rights reserved.
 *
 * This code is derived from software contributed to The NetBSD Foundation
 * by Charles M. Hannum.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *        This product includes software developed by the NetBSD
 *        Foundation, Inc. and its contributors.
 * 4. Neither the name of The NetBSD Foundation nor the names of its
 *    contributors may be used to endorse or promote products derived
 *    from this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
 * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
 * POSSIBILITY OF SUCH DAMAGE.
 */

/*-
 * Copyright (c) 1990 The Regents of the University of California.
 * All rights reserved.
 *
 * This code is derived from software contributed to Berkeley by
 * William Jolitz.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 *	@(#)locore.s	7.3 (Berkeley) 5/13/91
 */

#include "opt_ddb.h"
#include "opt_ddbparam.h"
#include "opt_realmem.h"

#include "opt_compat_netbsd.h"
#include "opt_compat_netbsd32.h"
#include "opt_compat_ibcs2.h"

#include "assym.h"
#include "lapic.h"
#include "ioapic.h"
#include "ksyms.h"

#include <sys/errno.h>
#include <sys/syscall.h>

#include <machine/param.h>
#include <machine/pte.h>
#include <machine/pmap.h>
#include <machine/segments.h>
#include <machine/specialreg.h>
#include <machine/trap.h>
#include <machine/bootinfo.h>
#include <machine/frameasm.h>

#if NLAPIC > 0
#include <machine/i82489reg.h>
#endif

/*
 * override user-land alignment before including asm.h
 */
#define	ALIGN_DATA	.align	8
#define ALIGN_TEXT	.align 16,0x90
#define _ALIGN_TEXT	ALIGN_TEXT

#include <machine/asm.h>

/* XXX temporary kluge; these should not be here */
/* Get definitions for IOM_BEGIN, IOM_END, and IOM_SIZE */
#include <dev/isa/isareg.h>


/*
 * Initialization
 */
	.data

#if NLAPIC > 0 
	.align  NBPG
	.globl _C_LABEL(local_apic), _C_LABEL(lapic_id), _C_LABEL(lapic_tpr)
_C_LABEL(local_apic):
	.space  LAPIC_ID
_C_LABEL(lapic_id):
	.long   0x00000000
	.space  LAPIC_TPRI-(LAPIC_ID+4)
_C_LABEL(lapic_tpr):
	.space  LAPIC_PPRI-LAPIC_TPRI
_C_LABEL(lapic_ppr):
	.space  LAPIC_ISR-LAPIC_PPRI 
_C_LABEL(lapic_isr):
	.space  NBPG-LAPIC_ISR
#endif

	.globl	_C_LABEL(cpu_id),_C_LABEL(cpu_vendor), _C_LABEL(cpu_brand_id)
	.globl	_C_LABEL(cpuid_level),_C_LABEL(cpu_feature)
	.globl	_C_LABEL(esym),_C_LABEL(boothowto)
	.globl	_C_LABEL(bootinfo),_C_LABEL(atdevbase)
	.globl	_C_LABEL(proc0paddr),_C_LABEL(PTDpaddr)
	.globl	_C_LABEL(biosbasemem),_C_LABEL(biosextmem)
	.globl	_C_LABEL(gdtstore)
_C_LABEL(cpu):		.long	0	# are we 386, 386sx, or 486,
					#   or Pentium, or..
_C_LABEL(cpu_id):	.long	0	# saved from `cpuid' instruction
_C_LABEL(cpu_feature):	.long	0	# feature flags from 'cpuid'
					#   instruction
_C_LABEL(cpuid_level):	.long	-1	# max. level accepted by 'cpuid'
					#   instruction
_C_LABEL(cpu_vendor):	.space	16	# vendor string returned by `cpuid'
					#   instruction
_C_LABEL(cpu_brand_id):	.long	0	# brand ID from 'cpuid' instruction
_C_LABEL(esym):		.quad	0	# ptr to end of syms
_C_LABEL(atdevbase):	.quad	0	# location of start of iomem in virtual
_C_LABEL(proc0paddr):	.quad	0
_C_LABEL(PTDpaddr):	.quad	0	# paddr of PTD, for libkvm
#ifndef REALBASEMEM
_C_LABEL(biosbasemem):	.long	0	# base memory reported by BIOS
#else
_C_LABEL(biosbasemem):	.long	REALBASEMEM
#endif
#ifndef REALEXTMEM
_C_LABEL(biosextmem):	.long	0	# extended memory reported by BIOS
#else
_C_LABEL(biosextmem):	.long	REALEXTMEM
#endif

#define	_RELOC(x)	((x) - KERNBASE)
#define	RELOC(x)	_RELOC(_C_LABEL(x))

	.globl	gdt64

gdt64:
	.word	gdt64_end-gdt64_start
	.quad	_RELOC(gdt64_start)
.align 64

gdt64_start:
	.quad 0x0000000000000000	/* always empty */
	.quad 0x00af9a000000ffff	/* kernel CS */
	.quad 0x00cf92000000ffff	/* kernel DS */
gdt64_end:

farjmp64:
	.long	longmode-KERNBASE
	.word	GSEL(GCODE_SEL, SEL_KPL)
	
	.space 512
tmpstk:

	.globl _C_LABEL(cpu_private)
	.comm _C_LABEL(cpu_private),NBPG,NBPG

/*
 * Some hackage to deal with 64bit symbols in 32 bit mode.
 * This may not be needed it things are cleaned up a little.
 */


	.text
	.globl	_C_LABEL(kernel_text)
	.set	_C_LABEL(kernel_text),KERNTEXTOFF

.code32

	.globl	start
start:	movw	$0x1234,0x472			# warm boot

	/*
	 * Load parameters from stack
	 * (howto, [bootdev], bootinfo, esym, basemem, extmem).
	 */
	movl	4(%esp),%eax
	movl	%eax,RELOC(boothowto)
	movl	12(%esp),%eax
	testl	%eax, %eax
	jz	1f
	movl	(%eax), %ebx		/* number of entries */
	movl	$RELOC(bootinfo),%ebp
	movl	%ebp, %edx
	addl	$BOOTINFO_MAXSIZE,%ebp
	movl	%ebx, (%edx)
	addl	$4, %edx
2:
	testl	%ebx, %ebx
	jz	1f
	addl	$4, %eax
	movl	(%eax), %ecx		/* address of entry */
	pushl	%edi
	pushl	%esi
	pushl	%eax

	movl	(%ecx),%eax	/* len */
	movl	%edx,%edi
	addl	(%ecx), %edx		/* update dest pointer */
	cmpl	%ebp, %edx
	jg	2f
	movl	%ecx,%esi
	movl	%eax,%ecx
	rep
	movsb
	popl	%eax
	popl	%esi
	popl	%edi
	subl	$1, %ebx
	jmp	2b
2:	/* cleanup for overflow case */
	popl	%eax
	popl	%esi
	popl	%edi
	movl	$RELOC(bootinfo),%ebp
	movl	%ebp, %edx
	subl	%ebx, (%edx)		/* correct number of entries */
1:

 	movl	16(%esp),%eax
	testl	%eax,%eax
	jz	1f
	addl	$KERNBASE_LO,%eax
1: 	movl	$RELOC(esym),%ebp
	movl	%eax,(%ebp)
	movl	$KERNBASE_HI,4(%ebp)

	movl	$RELOC(biosextmem),%ebp
	movl	(%ebp),%eax
	testl	%eax,%eax
	jnz	1f
	movl	20(%esp),%eax
	movl	%eax,(%ebp)
1:
	movl	$RELOC(biosbasemem),%ebp
	movl	(%ebp),%eax
	testl	%eax,%eax
	jnz	1f
	movl	24(%esp),%eax
	movl	%eax,(%ebp)
1:

	/* First, reset the PSL. */
	pushl	$PSL_MBO
	popfl

	xorl	%eax,%eax
	cpuid
	movl	%eax,RELOC(cpuid_level)
	movl	$RELOC(cpu_vendor),%ebp
	movl	%ebx,(%ebp)
	movl	%edx,4(%ebp)
	movl	%ecx,8(%ebp)
	movl	$0,  12(%ebp)

	movl	$1,%eax
	cpuid
	movl	%eax,RELOC(cpu_id)
	movl	%edx,RELOC(cpu_feature)

	/* Brand ID is bits 0-7 of %ebx */
	andl	$255,%ebx
	movl	%ebx,RELOC(cpu_brand_id)

	/* add AMD specific feature flags */
	movl	$0x80000001,%eax
	cpuid
	orl	%edx,RELOC(cpu_feature)

	/*
	 * Finished with old stack; load new %esp now instead of later so we
	 * can trace this code without having to worry about the trace trap
	 * clobbering the memory test or the zeroing of the bss+bootstrap page
	 * tables.
	 *
	 * The boot program should check:
	 *	text+data <= &stack_variable - more_space_for_stack
	 *	text+data+bss+pad+space_for_page_tables <= end_of_memory
	 * Oops, the gdt is in the carcass of the boot program so clearing
	 * the rest of memory is still not possible.
	 */
	movl	$RELOC(tmpstk),%esp

/*
 * Virtual address space of kernel:
 *
 * text | data | bss | [syms] | page dir | proc0 kstack | L1 ptp | L2 ptp | L3 
 *			      0          1       2      3
 */

#if L2_SLOT_KERNBASE > 0
#define TABLE_L2_ENTRIES (2 * (NKL2_KIMG_ENTRIES + 1))
#else
#define TABLE_L2_ENTRIES (NKL2_KIMG_ENTRIES + 1)
#endif

#if L3_SLOT_KERNBASE > 0
#define TABLE_L3_ENTRIES (2 * NKL3_KIMG_ENTRIES)
#else
#define TABLE_L3_ENTRIES NKL3_KIMG_ENTRIES
#endif


#define PROC0_PML4_OFF	0
#define PROC0_STK_OFF	(PROC0_PML4_OFF + NBPG)
#define PROC0_PTP3_OFF	(PROC0_STK_OFF + UPAGES * NBPG)
#define PROC0_PTP2_OFF	(PROC0_PTP3_OFF + NKL4_KIMG_ENTRIES * NBPG)
#define PROC0_PTP1_OFF	(PROC0_PTP2_OFF + TABLE_L3_ENTRIES * NBPG)
#define TABLESIZE \
  ((NKL4_KIMG_ENTRIES + TABLE_L3_ENTRIES + TABLE_L2_ENTRIES + 1 + UPAGES) \
    * NBPG)

#define fillkpt	\
1:	movl	%eax,(%ebx)	; 	/* store phys addr */ \
	movl	$0,4(%ebx)	; 	/* upper 32 bits 0 */ \
	addl	$8,%ebx		; 	/* next pte/pde */ \
	addl	$NBPG,%eax	; 	/* next phys page */ \
	loop	1b		;  \


	/* Find end of kernel image. */
	movl	$RELOC(end),%edi
#if (NKSYMS || defined(DDB) || defined(LKM)) && !defined(SYMTAB_SPACE)
	/* Save the symbols (if loaded). */
	movl	RELOC(esym),%eax
	testl	%eax,%eax
	jz	1f
	subl	$KERNBASE_LO,%eax	/* XXX */
	movl	%eax,%edi
1:
#endif
	/* Clear tables */
	movl	%edi,%esi
	addl	$PGOFSET,%esi
	andl	$~PGOFSET,%esi

	movl	%esi,%edi
	xorl	%eax,%eax
	cld
	movl	$TABLESIZE,%ecx
	shrl	$2,%ecx
	rep
	stosl

	leal	(PROC0_PTP1_OFF)(%esi), %ebx

	/*
	 * Compute etext - KERNBASE. This can't be > 4G, or we can't deal
	 * with it anyway, since we can't load it in 32 bit mode. So use
	 * the bottom 32 bits.
	 */
	movl	$RELOC(etext),%edx
	addl	$PGOFSET,%edx
	andl	$~PGOFSET,%edx

	/*
	 * Skip the first MB.
	 */
	movl	$(KERNTEXTOFF_LO - KERNBASE_LO),%eax
	movl	%eax,%ecx
	shrl	$(PGSHIFT-3),%ecx	/* ((n >> PGSHIFT) << 3) for # pdes */
	addl	%ecx,%ebx

	/* Map kernel text read-only */
	movl	%edx,%ecx
	subl	%eax,%ecx
	shrl	$PGSHIFT,%ecx
	orl     $(PG_V|PG_KR),%eax
	fillkpt

	/* Map the data, BSS, and bootstrap tables read-write. */
	leal	(PG_V|PG_KW)(%edx),%eax
	movl	$TABLESIZE,%ecx
	addl	%esi,%ecx		/* %ecx = &end[TABLESIZE] */
	subl	%edx,%ecx		/* %ecx = %ecx - etext */
	shrl	$PGSHIFT,%ecx
	fillkpt

	/* Map ISA I/O mem (later atdevbase) */
	movl	$(IOM_BEGIN|PG_V|PG_KW/*|PG_N*/),%eax
	movl	$(IOM_SIZE>>PGSHIFT),%ecx
	fillkpt

	/* Set up level 2 pages */
	leal    (PROC0_PTP2_OFF)(%esi),%ebx
	leal	(PROC0_PTP1_OFF)(%esi),%eax
	orl	$(PG_V|PG_KW), %eax
	movl	$(NKL2_KIMG_ENTRIES+1),%ecx
	fillkpt

#if L2_SLOT_KERNBASE > 0
	/* If needed, set up level 2 entries for actual kernel mapping */
	leal	(PROC0_PTP2_OFF+ L2_SLOT_KERNBASE*8)(%esi),%ebx
	leal    (PROC0_PTP1_OFF)(%esi),%eax
	orl     $(PG_V|PG_KW), %eax
	movl    $(NKL2_KIMG_ENTRIES+1),%ecx
	fillkpt
#endif

	/* Set up level 3 pages */
	leal    (PROC0_PTP3_OFF)(%esi),%ebx
	leal	(PROC0_PTP2_OFF)(%esi),%eax
	orl	$(PG_V|PG_KW), %eax
	movl	$NKL3_KIMG_ENTRIES,%ecx
	fillkpt

#if L3_SLOT_KERNBASE > 0
	/* If needed, set up level 3 entries for actual kernel mapping */
	leal	(PROC0_PTP3_OFF+ L3_SLOT_KERNBASE*8)(%esi),%ebx
	leal    (PROC0_PTP2_OFF)(%esi),%eax
	orl     $(PG_V|PG_KW), %eax
	movl    $NKL3_KIMG_ENTRIES,%ecx
	fillkpt
#endif

	/* Set up top level entries for identity mapping */
	leal    (PROC0_PML4_OFF)(%esi),%ebx
	leal	(PROC0_PTP3_OFF)(%esi),%eax
	orl	$(PG_V|PG_KW), %eax
	movl	$NKL4_KIMG_ENTRIES,%ecx
	fillkpt

	/* Set up top level entries for actual kernel mapping */
	leal    (PROC0_PML4_OFF + L4_SLOT_KERNBASE*8)(%esi),%ebx
	leal	(PROC0_PTP3_OFF)(%esi),%eax
	orl	$(PG_V|PG_KW), %eax
	movl	$NKL4_KIMG_ENTRIES,%ecx
	fillkpt

	/* Install recursive top level PDE */
	leal    (PROC0_PML4_OFF + PDIR_SLOT_PTE*8)(%esi),%ebx
	leal    (PROC0_PML4_OFF)(%esi),%eax
	orl	$(PG_V|PG_KW),%eax
	movl	%eax,(%ebx)
	movl	$0, 4(%ebx)


	/* Save phys. addr of PTD, for libkvm. */
	movl	$RELOC(PTDpaddr),%ebp
	movl	%esi,(%ebp)
	movl	$0,4(%ebp)

	/*
	 * Startup checklist:
	 * 1. Enable PAE (and SSE while here).
	 */
	movl	%cr4,%eax
	orl	$(CR4_PAE|CR4_OSFXSR|CR4_OSXMMEXCPT),%eax
	movl	%eax,%cr4

	/*
	 * 2. Set Long Mode Enable in EFER. Also enable the
	 *    syscall extensions.
	 */
	movl    $MSR_EFER,%ecx
	rdmsr
	xorl	%eax,%eax	/* XXX */
	orl	$(EFER_LME|EFER_SCE),%eax
	wrmsr

	/*
	 * 3. Load %cr3 with pointer to PML4.
	 */
	movl	%esi,%eax
	movl	%eax,%cr3

	/*
	 * 4. Enable paging and the rest of it.
	 */
	movl	%cr0,%eax
	orl	$(CR0_PE|CR0_PG|CR0_NE|CR0_TS|CR0_MP|CR0_WP),%eax
	movl	%eax,%cr0
	jmp	compat
compat:

	/*
	 * 5.
	 * Not quite done yet, we're now in a compatibility segment,
	 * in legacy mode. We must jump to a long mode segment.
	 * Need to set up a temporary GDT with a long mode segment
	 * in it to do that.
	 */

	movl	$RELOC(gdt64),%eax
	lgdt	(%eax)
	movl	$RELOC(farjmp64),%eax
	ljmp	*(%eax)

.code64
longmode:
	/*
	 * 6.
	 * Finally, we're in long mode. However, we're still
	 * in the identity mapped area (could not jump out
	 * of that earlier because it would have been a > 32bit
	 * jump). We can do that now, so here we go.
	 */
	movabsq	$longmode_hi,%rax
	jmp	*%rax
longmode_hi:
	/*
	 * We have arrived.
	 * There's no need anymore for the identity mapping in low
	 * memory, remove it.
	 */
	movq	$KERNBASE,%r8

#if L2_SLOT_KERNBASE > 0
	movq	$(NKL2_KIMG_ENTRIES+1),%rcx
	leaq	(PROC0_PTP2_OFF)(%rsi),%rbx
	addq	%r8, %rbx
1:	movq	$0,(%rbx)
	addq	$8,%rbx
	loop	1b
#endif

#if L3_SLOT_KERNBASE > 0
	movq	$NKL3_KIMG_ENTRIES,%rcx
	leaq	(PROC0_PTP3_OFF)(%rsi),%rbx
	addq	%r8, %rbx
1:	movq	$0,(%rbx)
	addq	$8,%rbx
	loop	1b
#endif

	movq	$NKL4_KIMG_ENTRIES,%rcx
	leaq	(PROC0_PML4_OFF)(%rsi),%rbx	# old, phys  address of PML4
	addq	%r8, %rbx			# new, virtual address of PML4
1:	movq	$0,(%rbx)
	addq	$8,%rbx
	loop	1b

	/* Relocate atdevbase. */
	movq	$(TABLESIZE+KERNBASE),%rdx
	addq	%rsi,%rdx
	movq	%rdx,_C_LABEL(atdevbase)(%rip)

	/* Set up bootstrap stack. */
	leaq	(PROC0_STK_OFF)(%rsi),%rax
	addq	%r8,%rax
	movq	%rax,_C_LABEL(proc0paddr)(%rip)
	leaq	(USPACE-FRAMESIZE)(%rax),%rsp
	movq	%rsi,PCB_CR3(%rax)	# pcb->pcb_cr3
	xorq	%rbp,%rbp               # mark end of frames

	xorw	%ax,%ax
	movw	%ax,%gs
	movw	%ax,%fs

	/* XXX merge these */
	leaq	TABLESIZE(%rsi),%rdi
	call	_C_LABEL(init_x86_64)

	call 	_C_LABEL(main)

/*****************************************************************************/

/*
 * void lgdt(struct region_descriptor *rdp);
 *
 * Change the global descriptor table.
 */
NENTRY(lgdt)
	/* Reload the descriptor table. */
	movq	%rdi,%rax
	lgdt	(%rax)
	/* Flush the prefetch q. */
	jmp	1f
	nop
1:	/* Reload "stale" selectors. */
	movl	$GSEL(GDATA_SEL, SEL_KPL),%eax
	movl	%eax,%ds
	movl	%eax,%es
	movl	%eax,%ss
	/* Reload code selector by doing intersegment return. */
	popq	%rax
	pushq	$GSEL(GCODE_SEL, SEL_KPL)
	pushq	%rax
	lretq

/*
 * void x86_flush()
 *
 * Flush instruction pipelines by doing an intersegment (far) return.
 */
NENTRY(x86_flush)
	popq	%rax
	pushq	$GSEL(GCODE_SEL, SEL_KPL)
	pushq	%rax
	lretq

/*
 * int setjmp(label_t *)
 *
 * Used primarily by DDB.
 */
ENTRY(setjmp)
	/*
	 * Only save registers that must be preserved across function
	 * calls according to the ABI (%rbx, %rsp, %rbp, %r12-%r15)
	 * and %rip.
	 */
	movq	%rdi,%rax
	movq	%rbx,(%rax)
	movq	%rsp,8(%rax)
	movq	%rbp,16(%rax)
	movq	%r12,24(%rax)
	movq	%r13,32(%rax)
	movq	%r14,40(%rax)
	movq	%r15,48(%rax)
	movq	(%rsp),%rdx
	movq	%rdx,56(%rax)
	xorl	%eax,%eax
	ret

/*
 * int longjmp(label_t *)
 *
 * Used primarily by DDB.
 */
ENTRY(longjmp)
	movq	%rdi,%rax
	movq	(%rax),%rbx
	movq	8(%rax),%rsp
	movq	16(%rax),%rbp
	movq	24(%rax),%r12
	movq	32(%rax),%r13
	movq	40(%rax),%r14
	movq	48(%rax),%r15
	movq	56(%rax),%rdx
	movq	%rdx,(%rsp)
	movl	$1,%eax
	ret

/*
 * struct lwp *cpu_switchto(struct lwp *, struct lwp *)
 *
 *	1. if (oldlwp != NULL), save its context.
 *	2. then, restore context of newlwp.
 *
 * Note that the stack frame layout is known to "struct switchframe" in
 * <machine/frame.h> and to the code in cpu_lwp_fork() which initializes
 * it for a new lwp.
 */
ENTRY(cpu_switchto)
	pushq	%rbx
	pushq	%r12
	pushq	%r13
	pushq	%r14
	pushq	%r15

	movq	%rdi,%r13	# oldlwp
	movq	%rsi,%r12	# newlwp

	testq	%r13,%r13
	jz	1f

	/* Save old context. */
	movq	L_ADDR(%r13),%rax
	movq	%rsp,PCB_RSP(%rax)
	movq	%rbp,PCB_RBP(%rax)

	/* Switch to newlwp's stack. */
1:	movq	L_ADDR(%r12),%r14
	movq	PCB_RSP(%r14),%rsp
	movq	PCB_RBP(%r14),%rbp

	/* Switch TSS. Reset "task busy" flag before loading. */
	movq    CPUVAR(GDT),%rax
	movl	L_MD_TSS_SEL(%r12),%edx
	andl	$~0x0200,4(%rax,%rdx, 1)
	ltr	%dx

	/* Set curlwp. */
	movq	%r12,CPUVAR(CURLWP)

	/* Don't bother with the rest if switching to a system process. */
	testl	$LW_SYSTEM,L_FLAG(%r12)
	jnz	4f

	/* Is this process using RAS (restartable atomic sequences)? */
	movq	L_PROC(%r12),%rdi
	cmpq	$0,P_RASLIST(%rdi)
	jne	5f

	/*
	 * Restore cr0 (including FPU state).  Raise the IPL to IPL_IPI.
	 * FPU IPIs can alter the LWP's saved cr0.  Dropping the priority
	 * is deferred until mi_switch(), when cpu_switchto() returns.
	 */
2:	movl	$IPL_IPI,CPUVAR(ILEVEL)
	movq	PCB_CR0(%r14),%rcx
	movq	%cr0,%rdx

	/*
	 * If our floating point registers are on a different CPU,
	 * set CR0_TS so we'll trap rather than reuse bogus state.
	 */
	movq	PCB_FPCPU(%r14),%rax
	cmpq	CPUVAR(SELF),%rax
	je	3f
	orq	$CR0_TS,%rcx

	/* Reloading CR0 is very expensive - avoid if possible. */
3:	cmpq	%rdx,%rcx
	je	4f
	movq	%rcx,%cr0

	/* Return to the new LWP, returning 'oldlwp' in %rax. */
4:	movq	%r13,%rax
	popq	%r15
	popq	%r14
	popq	%r13
	popq	%r12
	popq	%rbx
	ret

	/* Check for restartable atomic sequences (RAS). */
5:	movq	L_MD_REGS(%r12),%rbx
	movq	TF_RIP(%rbx),%rsi
	call	_C_LABEL(ras_lookup)
	cmpq	$-1,%rax
	je	2b
	movq	%rax,TF_RIP(%rbx)
	jmp	2b

/*
 * void savectx(struct pcb *pcb);
 *
 * Update pcb, saving current processor state.
 */
ENTRY(savectx)
	/* Save stack pointers. */
	movq	%rsp,PCB_RSP(%rdi)
	movq	%rbp,PCB_RBP(%rdi)
	ret

IDTVEC(syscall32)
	sysret		/* go away please */

/*
 * syscall()
 *
 * syscall insn entry. This currently isn't much faster, but
 * it can be made faster in the future.
 */
IDTVEC(syscall)
	swapgs
	movq	%r15,CPUVAR(SCRATCH)
	movq	CPUVAR(CURLWP),%r15
	movq	L_ADDR(%r15),%r15
	movq	PCB_RSP0(%r15),%r15
	xchgq	%r15,%rsp
	sti

	/*
	 * XXX don't need this whole frame, split of the
	 * syscall frame and trapframe is needed.
	 * First, leave some room for the trapno, error,
	 * ss:rsp, etc, so that all GP registers can be
	 * saved. Then, fill in the rest.
	 */
	pushq	$(LSEL(LUDATA_SEL, SEL_UPL))
	pushq	%r15
	subq	$(TF_RSP-TF_TRAPNO),%rsp
	movq	CPUVAR(SCRATCH),%r15
	subq	$32,%rsp
	INTR_SAVE_GPRS
	movw	%es,TF_ES(%rsp)
	movw	$(LSEL(LUDATA_SEL, SEL_UPL)),TF_DS(%rsp)
	movq	%r11, TF_RFLAGS(%rsp)	/* old rflags from syscall insn */
	movq	$(LSEL(LUCODE_SEL, SEL_UPL)), TF_CS(%rsp)
	movq	%rcx,TF_RIP(%rsp)
	movq	$2,TF_ERR(%rsp)
	movq	$T_ASTFLT, TF_TRAPNO(%rsp)

	movq	CPUVAR(CURLWP),%r14
	movq	%rsp,L_MD_REGS(%r14)	# save pointer to frame
	movq	L_PROC(%r14),%r15
	andl	$~MDP_IRET,L_MD_FLAGS(%r14)
	movq	%rsp,%rdi
	call	*P_MD_SYSCALL(%r15)
1:	/* Check for ASTs on exit to user mode. */
	cli
	CHECK_ASTPENDING(%r14)
	je	2f
	/* Always returning to user mode here. */
	CLEAR_ASTPENDING(%r14)
	sti
	/* Pushed T_ASTFLT into tf_trapno on entry. */
	movq	%rsp,%rdi
	call	_C_LABEL(trap)
	jmp	1b
2:
	sti
	testl	$MDP_IRET, L_MD_FLAGS(%r14)
	jne	iret_return;
syscall_return:
#ifdef DIAGNOSTIC
	cmpl	$IPL_NONE,CPUVAR(ILEVEL)
	jne	3f
#endif
	/*
	 * XXX interrupts off longer than they should be here.
	 */
	cli
	swapgs
	movw	TF_ES(%rsp),%es
	INTR_RESTORE_GPRS
	movw	$(LSEL(LUDATA_SEL, SEL_UPL)),%r11
	movw	%r11,%ds
	addq	$48,%rsp
	popq	%rcx	/* return rip */
	addq	$8,%rsp
	popq	%r11	/* flags as set by sysret insn */
	movq	%ss:(%rsp),%rsp
	sysretq

#ifdef DIAGNOSTIC
3:	movabsq	$4f, %rdi
	movl	TF_RAX(%rsp),%esi
	movl	TF_RDI(%rsp),%edx
	movl	%ebx,%ecx
	movl	CPUVAR(ILEVEL),%r8d
	xorq	%rax,%rax
	call	_C_LABEL(printf)
#ifdef DDB
	int	$3
#endif /* DDB */
	movl	$IPL_NONE,CPUVAR(ILEVEL)
	jmp	1b
4:	.asciz	"WARNING: SPL NOT LOWERED ON SYSCALL %d %d EXIT %x %x\n"
#endif

/*
 * void lwp_trampoline(void);
 *
 * This is a trampoline function pushed run by newly created LWPs
 * in order to do additional setup in their context.  32-bit
 * binaries begin life here.
 */
NENTRY(lwp_trampoline)
	movq	%rbp,%rsi
	movq	%rax,%rdi
	xorq	%rbp,%rbp
	call	_C_LABEL(lwp_startup)
	movq	%r13,%rdi
	call	*%r12
	INTRFASTEXIT
	/* NOTREACHED */

/*
 * void child_trampoline(void);
 *
 * As per lwp_trampoline(), but 64-bit binaries start here.
 */
NENTRY(child_trampoline)
	movq	%rbp,%rsi
	movq	%rax,%rdi
	xorq	%rbp,%rbp
	call	_C_LABEL(lwp_startup)
	movq	%r13,%rdi
	call	*%r12
	jmp	syscall_return

	.globl  _C_LABEL(osyscall_return)

/*
 * oosyscall()
 *
 * Old call gate entry for syscall. only needed if we're
 * going to support running old NetBSD or ibcs2 binaries, etc,
 * on NetBSD/amd64.
 */
IDTVEC(oosyscall)
	/* Set rflags in trap frame. */
	pushfq
	popq	8(%rsp)
	pushq	$7		# size of instruction for restart
	jmp	osyscall1

/*
 * osyscall()
 *
 * Trap gate entry for int $80 syscall, also used by sigreturn.
 */
IDTVEC(osyscall)
	pushq	$2		# size of instruction for restart
osyscall1:
	pushq	$T_ASTFLT	# trap # for doing ASTs
	INTRENTRY
	sti
	movq	CPUVAR(CURLWP),%r14
	movq	%rsp,L_MD_REGS(%r14)	# save pointer to frame
	movq	L_PROC(%r14),%rdx
	movq	%rsp,%rdi
	call	*P_MD_SYSCALL(%rdx)
_C_LABEL(osyscall_return):
2:	/* Check for ASTs on exit to user mode. */
	cli
	CHECK_ASTPENDING(%r14)
	je	1f
	/* Always returning to user mode here. */
	CLEAR_ASTPENDING(%r14)
	sti
	/* Pushed T_ASTFLT into tf_trapno on entry. */
	movq	%rsp,%rdi
	call	_C_LABEL(trap)
	jmp	2b

iret_return:
#ifndef DIAGNOSTIC
1:	INTRFASTEXIT
#else /* DIAGNOSTIC */
1:	cmpl	$IPL_NONE,CPUVAR(ILEVEL)
	jne	3f
	INTRFASTEXIT
3:	sti
	movabsq	$4f, %rdi
	xorq	%rax,%rax
	call	_C_LABEL(printf)
#ifdef DDB
	int	$3
#endif /* DDB */
	movl	$IPL_NONE,CPUVAR(ILEVEL)
	jmp	2b
4:	.asciz	"WARNING: SPL NOT LOWERED ON SYSCALL EXIT\n"
#endif /* DIAGNOSTIC */