[dragonfly.git] / sys / platform / pc64 / x86_64 / vm_machdep.c

/*-
 * Copyright (c) 1982, 1986 The Regents of the University of California.
 * Copyright (c) 1989, 1990 William Jolitz
 * Copyright (c) 1994 John Dyson
 * Copyright (c) 2008 The DragonFly Project.
 * All rights reserved.
 *
 * This code is derived from software contributed to Berkeley by
 * the Systems Programming Group of the University of Utah Computer
 * Science Department, and 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. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *	This product includes software developed by the University of
 *	California, Berkeley and its contributors.
 * 4. 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.
 *
 *	from: @(#)vm_machdep.c	7.3 (Berkeley) 5/13/91
 *	Utah $Hdr: vm_machdep.c 1.16.1.1 89/06/23$
 * $FreeBSD: src/sys/i386/i386/vm_machdep.c,v 1.132.2.9 2003/01/25 19:02:23 dillon Exp $
 */

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/malloc.h>
#include <sys/proc.h>
#include <sys/buf.h>
#include <sys/interrupt.h>
#include <sys/vnode.h>
#include <sys/vmmeter.h>
#include <sys/kernel.h>
#include <sys/sysctl.h>
#include <sys/unistd.h>

#include <machine/clock.h>
#include <machine/cpu.h>
#include <machine/md_var.h>
#include <machine/smp.h>
#include <machine/pcb.h>
#include <machine/pcb_ext.h>
#include <machine/segments.h>
#include <machine/globaldata.h>	/* npxthread */

#include <vm/vm.h>
#include <vm/vm_param.h>
#include <sys/lock.h>
#include <vm/vm_kern.h>
#include <vm/vm_page.h>
#include <vm/vm_map.h>
#include <vm/vm_extern.h>

#include <sys/thread2.h>
#include <sys/mplock2.h>

#include <bus/isa/isa.h>

static void	cpu_reset_real (void);
/*
 * Finish a fork operation, with lwp lp2 nearly set up.
 * Copy and update the pcb, set up the stack so that the child
 * ready to run and return to user mode.
 */
void
cpu_fork(struct lwp *lp1, struct lwp *lp2, int flags)
{
	struct pcb *pcb2;

	if ((flags & RFPROC) == 0) {
		if ((flags & RFMEM) == 0) {
			/* unshare user LDT */
			struct pcb *pcb1 = lp1->lwp_thread->td_pcb;
			struct pcb_ldt *pcb_ldt = pcb1->pcb_ldt;
			if (pcb_ldt && pcb_ldt->ldt_refcnt > 1) {
				pcb_ldt = user_ldt_alloc(pcb1,pcb_ldt->ldt_len);
				user_ldt_free(pcb1);
				pcb1->pcb_ldt = pcb_ldt;
				set_user_ldt(pcb1);
			}
		}
		return;
	}

	/* Ensure that lp1's pcb is up to date. */
	if (mdcpu->gd_npxthread == lp1->lwp_thread)
		npxsave(lp1->lwp_thread->td_savefpu);
	
	/*
	 * Copy lp1's PCB.  This really only applies to the
	 * debug registers and FP state, but its faster to just copy the
	 * whole thing.  Because we only save the PCB at switchout time,
	 * the register state may not be current.
	 */
	pcb2 = lp2->lwp_thread->td_pcb;
	*pcb2 = *lp1->lwp_thread->td_pcb;

	/*
	 * Create a new fresh stack for the new process.
	 * Copy the trap frame for the return to user mode as if from a
	 * syscall.  This copies the user mode register values.
	 *
	 * pcb_rsp must allocate an additional call-return pointer below
	 * the trap frame which will be restored by cpu_heavy_restore from
	 * PCB_RIP, and the thread's td_sp pointer must allocate an
	 * additonal two quadwords below the pcb_rsp call-return pointer to
	 * hold the LWKT restore function pointer and rflags.
	 *
	 * The LWKT restore function pointer must be set to cpu_heavy_restore,
	 * which is our standard heavy-weight process switch-in function.
	 * YYY eventually we should shortcut fork_return and fork_trampoline
	 * to use the LWKT restore function directly so we can get rid of
	 * all the extra crap we are setting up.
	 */
	lp2->lwp_md.md_regs = (struct trapframe *)pcb2 - 1;
	bcopy(lp1->lwp_md.md_regs, lp2->lwp_md.md_regs, sizeof(*lp2->lwp_md.md_regs));

	/*
	 * Set registers for trampoline to user mode.  Leave space for the
	 * return address on stack.  These are the kernel mode register values.
	 */
	pcb2->pcb_cr3 = vtophys(vmspace_pmap(lp2->lwp_proc->p_vmspace)->pm_pml4);
	pcb2->pcb_cr3 |= PG_RW | PG_U | PG_V;
	pcb2->pcb_rbx = (unsigned long)fork_return;	/* fork_trampoline argument */
	pcb2->pcb_rbp = 0;
	pcb2->pcb_rsp = (unsigned long)lp2->lwp_md.md_regs - sizeof(void *);
	pcb2->pcb_r12 = (unsigned long)lp2;		/* fork_trampoline argument */
	pcb2->pcb_r13 = 0;
	pcb2->pcb_r14 = 0;
	pcb2->pcb_r15 = 0;
	pcb2->pcb_rip = (unsigned long)fork_trampoline;
	lp2->lwp_thread->td_sp = (char *)(pcb2->pcb_rsp - sizeof(void *));
	*(u_int64_t *)lp2->lwp_thread->td_sp = PSL_USER;
	lp2->lwp_thread->td_sp -= sizeof(void *);
	*(void **)lp2->lwp_thread->td_sp = (void *)cpu_heavy_restore;

	/*
	 * pcb2->pcb_ldt:	duplicated below, if necessary.
	 * pcb2->pcb_savefpu:	cloned above.
	 * pcb2->pcb_flags:	cloned above (always 0 here?).
	 * pcb2->pcb_onfault:	cloned above (always NULL here).
	 * pcb2->pcb_onfault_sp:cloned above (dont care)
	 */

	/*
	 * XXX don't copy the i/o pages.  this should probably be fixed.
	 */
	pcb2->pcb_ext = NULL;

        /* Copy the LDT, if necessary. */
        if (pcb2->pcb_ldt != NULL) {
		if (flags & RFMEM) {
			pcb2->pcb_ldt->ldt_refcnt++;
		} else {
			pcb2->pcb_ldt = user_ldt_alloc(pcb2,
				pcb2->pcb_ldt->ldt_len);
		}
        }
	bcopy(&lp1->lwp_thread->td_tls, &lp2->lwp_thread->td_tls,
	      sizeof(lp2->lwp_thread->td_tls));
	/*
	 * Now, cpu_switch() can schedule the new lwp.
	 * pcb_rsp is loaded pointing to the cpu_switch() stack frame
	 * containing the return address when exiting cpu_switch.
	 * This will normally be to fork_trampoline(), which will have
	 * %rbx loaded with the new lwp's pointer.  fork_trampoline()
	 * will set up a stack to call fork_return(lp, frame); to complete
	 * the return to user-mode.
	 */
}

/*
 * Prepare new lwp to return to the address specified in params.
 */
int
cpu_prepare_lwp(struct lwp *lp, struct lwp_params *params)
{
	struct trapframe *regs = lp->lwp_md.md_regs;
	void *bad_return = NULL;
	int error;

	regs->tf_rip = (long)params->func;
	regs->tf_rsp = (long)params->stack;
	/* Set up argument for function call */
	regs->tf_rdi = (long)params->arg;

	/*
	 * Set up fake return address.  As the lwp function may never return,
	 * we simply copy out a NULL pointer and force the lwp to receive
	 * a SIGSEGV if it returns anyways.
	 */
	regs->tf_rsp -= sizeof(void *);
	error = copyout(&bad_return, (void *)regs->tf_rsp, sizeof(bad_return));
	if (error)
		return (error);

	cpu_set_fork_handler(lp,
	    (void (*)(void *, struct trapframe *))generic_lwp_return, lp);
	return (0);
}

/*
 * Intercept the return address from a freshly forked process that has NOT
 * been scheduled yet.
 *
 * This is needed to make kernel threads stay in kernel mode.
 */
void
cpu_set_fork_handler(struct lwp *lp, void (*func)(void *, struct trapframe *),
		     void *arg)
{
	/*
	 * Note that the trap frame follows the args, so the function
	 * is really called like this:  func(arg, frame);
	 */
	lp->lwp_thread->td_pcb->pcb_rbx = (long)func;	/* function */
	lp->lwp_thread->td_pcb->pcb_r12 = (long)arg;	/* first arg */
}

void
cpu_set_thread_handler(thread_t td, void (*rfunc)(void), void *func, void *arg)
{
	td->td_pcb->pcb_rbx = (long)func;
	td->td_pcb->pcb_r12 = (long)arg;
	td->td_switch = cpu_lwkt_switch;
	td->td_sp -= sizeof(void *);
	*(void **)td->td_sp = rfunc;	/* exit function on return */
	td->td_sp -= sizeof(void *);
	*(void **)td->td_sp = cpu_kthread_restore;
}

void
cpu_lwp_exit(void)
{
	struct thread *td = curthread;
	struct pcb *pcb;

	pcb = td->td_pcb;

	/* Some i386 functionality was dropped */
	KKASSERT(pcb->pcb_ext == NULL);

	/*
	 * disable all hardware breakpoints
	 */
        if (pcb->pcb_flags & PCB_DBREGS) {
                reset_dbregs();
                pcb->pcb_flags &= ~PCB_DBREGS;
        }
	td->td_gd->gd_cnt.v_swtch++;

	crit_enter_quick(td);
	if (td->td_flags & TDF_TSLEEPQ)
		tsleep_remove(td);
	lwkt_deschedule_self(td);
	lwkt_remove_tdallq(td);
	cpu_thread_exit();
}

/*
 * Terminate the current thread.  The caller must have already acquired
 * the thread's rwlock and placed it on a reap list or otherwise notified
 * a reaper of its existance.  We set a special assembly switch function which
 * releases td_rwlock after it has cleaned up the MMU state and switched
 * out the stack.
 *
 * Must be caller from a critical section and with the thread descheduled.
 */
void
cpu_thread_exit(void)
{
	npxexit();
	curthread->td_switch = cpu_exit_switch;
	curthread->td_flags |= TDF_EXITING;
	lwkt_switch();
	panic("cpu_thread_exit: lwkt_switch() unexpectedly returned");
}

void
cpu_reset(void)
{
	cpu_reset_real();
}

static void
cpu_reset_real(void)
{
	/*
	 * Attempt to do a CPU reset via the keyboard controller,
	 * do not turn of the GateA20, as any machine that fails
	 * to do the reset here would then end up in no man's land.
	 */

#if !defined(BROKEN_KEYBOARD_RESET)
	outb(IO_KBD + 4, 0xFE);
	DELAY(500000);	/* wait 0.5 sec to see if that did it */
	kprintf("Keyboard reset did not work, attempting CPU shutdown\n");
	DELAY(1000000);	/* wait 1 sec for kprintf to complete */
#endif
#if JG
	/* force a shutdown by unmapping entire address space ! */
	bzero((caddr_t) PTD, PAGE_SIZE);
#endif

	/* "good night, sweet prince .... <THUNK!>" */
	cpu_invltlb();
	/* NOTREACHED */
	while(1);
}

/*
 * Convert kernel VA to physical address
 */
vm_paddr_t
kvtop(void *addr)
{
	vm_paddr_t pa;

	pa = pmap_kextract((vm_offset_t)addr);
	if (pa == 0)
		panic("kvtop: zero page frame");
	return (pa);
}

static void
swi_vm(void *arg, void *frame)
{
	if (busdma_swi_pending != 0)
		busdma_swi();
}

static void
swi_vm_setup(void *arg)
{
	register_swi(SWI_VM, swi_vm, NULL, "swi_vm", NULL, 0);
}

SYSINIT(vm_setup, SI_BOOT2_MACHDEP, SI_ORDER_ANY, swi_vm_setup, NULL);

/*
 * platform-specific vmspace initialization (nothing for x86_64)
 */
void
cpu_vmspace_alloc(struct vmspace *vm __unused)
{
}

void
cpu_vmspace_free(struct vmspace *vm __unused)
{
}

int
kvm_access_check(vm_offset_t saddr, vm_offset_t eaddr, int prot)
{
	vm_offset_t addr;

	if (saddr < KvaStart)
		return EFAULT;
	if (eaddr >= KvaEnd)
		return EFAULT;
	for (addr = saddr; addr < eaddr; addr += PAGE_SIZE)  {
		if (pmap_extract(&kernel_pmap, addr) == 0)
			return EFAULT;
	}
	if (!kernacc((caddr_t)saddr, eaddr - saddr, prot))
		return EFAULT;
	return 0;
}

#if 0

void _test_frame_enter(struct trapframe *frame);
void _test_frame_exit(struct trapframe *frame);

void
_test_frame_enter(struct trapframe *frame)
{
	thread_t td = curthread;

	if (ISPL(frame->tf_cs) == SEL_UPL) {
		KKASSERT(td->td_lwp);
                KASSERT(td->td_lwp->lwp_md.md_regs == frame,
                        ("_test_frame_exit: Frame mismatch %p %p",
			td->td_lwp->lwp_md.md_regs, frame));
	    td->td_lwp->lwp_saveusp = (void *)frame->tf_rsp;
	    td->td_lwp->lwp_saveupc = (void *)frame->tf_rip;
	}
	if ((char *)frame < td->td_kstack ||
	    (char *)frame > td->td_kstack + td->td_kstack_size) {
		panic("_test_frame_exit: frame not on kstack %p kstack=%p",
			frame, td->td_kstack);
	}
}

void
_test_frame_exit(struct trapframe *frame)
{
	thread_t td = curthread;

	if (ISPL(frame->tf_cs) == SEL_UPL) {
		KKASSERT(td->td_lwp);
                KASSERT(td->td_lwp->lwp_md.md_regs == frame,
                        ("_test_frame_exit: Frame mismatch %p %p",
			td->td_lwp->lwp_md.md_regs, frame));
		if (td->td_lwp->lwp_saveusp != (void *)frame->tf_rsp) {
			kprintf("_test_frame_exit: %s:%d usp mismatch %p/%p\n",
				td->td_comm, td->td_proc->p_pid,
				td->td_lwp->lwp_saveusp,
				(void *)frame->tf_rsp);
		}
		if (td->td_lwp->lwp_saveupc != (void *)frame->tf_rip) {
			kprintf("_test_frame_exit: %s:%d upc mismatch %p/%p\n",
				td->td_comm, td->td_proc->p_pid,
				td->td_lwp->lwp_saveupc,
				(void *)frame->tf_rip);
		}

		/*
		 * adulterate the fields to catch entries that
		 * don't run through test_frame_enter
		 */
		td->td_lwp->lwp_saveusp =
			(void *)~(intptr_t)td->td_lwp->lwp_saveusp;
		td->td_lwp->lwp_saveupc =
			(void *)~(intptr_t)td->td_lwp->lwp_saveupc;
	}
	if ((char *)frame < td->td_kstack ||
	    (char *)frame > td->td_kstack + td->td_kstack_size) {
		panic("_test_frame_exit: frame not on kstack %p kstack=%p",
			frame, td->td_kstack);
	}
}

#endif
Commit	Line	Data
d7f50089 YY	1	/*-
	2	* Copyright (c) 1982, 1986 The Regents of the University of California.
	3	* Copyright (c) 1989, 1990 William Jolitz
	4	* Copyright (c) 1994 John Dyson
c8fe38ae	5	* Copyright (c) 2008 The DragonFly Project.
d7f50089 YY	6	* All rights reserved.
	7	*
	8	* This code is derived from software contributed to Berkeley by
	9	* the Systems Programming Group of the University of Utah Computer
	10	* Science Department, and William Jolitz.
	11	*
	12	* Redistribution and use in source and binary forms, with or without
	13	* modification, are permitted provided that the following conditions
	14	* are met:
	15	* 1. Redistributions of source code must retain the above copyright
	16	* notice, this list of conditions and the following disclaimer.
	17	* 2. Redistributions in binary form must reproduce the above copyright
	18	* notice, this list of conditions and the following disclaimer in the
	19	* documentation and/or other materials provided with the distribution.
	20	* 3. All advertising materials mentioning features or use of this software
	21	* must display the following acknowledgement:
	22	* This product includes software developed by the University of
	23	* California, Berkeley and its contributors.
	24	* 4. Neither the name of the University nor the names of its contributors
	25	* may be used to endorse or promote products derived from this software
	26	* without specific prior written permission.
	27	*
	28	* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
	29	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
	30	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
	31	* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
	32	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
	33	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
	34	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
	35	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
	36	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
	37	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
	38	* SUCH DAMAGE.
	39	*
	40	* from: @(#)vm_machdep.c 7.3 (Berkeley) 5/13/91
	41	* Utah $Hdr: vm_machdep.c 1.16.1.1 89/06/23$
	42	* $FreeBSD: src/sys/i386/i386/vm_machdep.c,v 1.132.2.9 2003/01/25 19:02:23 dillon Exp $
d7f50089 YY	43	*/
	44
	45	#include <sys/param.h>
	46	#include <sys/systm.h>
	47	#include <sys/malloc.h>
	48	#include <sys/proc.h>
	49	#include <sys/buf.h>
	50	#include <sys/interrupt.h>
	51	#include <sys/vnode.h>
	52	#include <sys/vmmeter.h>
	53	#include <sys/kernel.h>
	54	#include <sys/sysctl.h>
	55	#include <sys/unistd.h>
	56
	57	#include <machine/clock.h>
	58	#include <machine/cpu.h>
	59	#include <machine/md_var.h>
	60	#include <machine/smp.h>
d7f50089	61	#include <machine/pcb.h>
0855a2af	62	#include <machine/pcb_ext.h>
d7f50089 YY	63	#include <machine/segments.h>
	64	#include <machine/globaldata.h> /* npxthread */
	65
	66	#include <vm/vm.h>
	67	#include <vm/vm_param.h>
	68	#include <sys/lock.h>
	69	#include <vm/vm_kern.h>
	70	#include <vm/vm_page.h>
	71	#include <vm/vm_map.h>
	72	#include <vm/vm_extern.h>
	73
	74	#include <sys/thread2.h>
684a93c4	75	#include <sys/mplock2.h>
d7f50089	76
0855a2af	77	#include <bus/isa/isa.h>
d7f50089	78
c8fe38ae	79	static void cpu_reset_real (void);
d7f50089 YY	80	/*
	81	* Finish a fork operation, with lwp lp2 nearly set up.
	82	* Copy and update the pcb, set up the stack so that the child
	83	* ready to run and return to user mode.
	84	*/
	85	void
	86	cpu_fork(struct lwp lp1, struct lwp lp2, int flags)
	87	{
c8fe38ae MD	88	struct pcb *pcb2;
	89
	90	if ((flags & RFPROC) == 0) {
	91	if ((flags & RFMEM) == 0) {
	92	/* unshare user LDT */
	93	struct pcb *pcb1 = lp1->lwp_thread->td_pcb;
	94	struct pcb_ldt *pcb_ldt = pcb1->pcb_ldt;
	95	if (pcb_ldt && pcb_ldt->ldt_refcnt > 1) {
	96	pcb_ldt = user_ldt_alloc(pcb1,pcb_ldt->ldt_len);
	97	user_ldt_free(pcb1);
	98	pcb1->pcb_ldt = pcb_ldt;
	99	set_user_ldt(pcb1);
	100	}
	101	}
	102	return;
	103	}
	104
c8fe38ae MD	105	/* Ensure that lp1's pcb is up to date. */
	106	if (mdcpu->gd_npxthread == lp1->lwp_thread)
	107	npxsave(lp1->lwp_thread->td_savefpu);
c8fe38ae MD	108
	109	/*
	110	* Copy lp1's PCB. This really only applies to the
	111	* debug registers and FP state, but its faster to just copy the
	112	* whole thing. Because we only save the PCB at switchout time,
	113	* the register state may not be current.
	114	*/
	115	pcb2 = lp2->lwp_thread->td_pcb;
	116	pcb2 = lp1->lwp_thread->td_pcb;
	117
	118	/*
	119	* Create a new fresh stack for the new process.
	120	* Copy the trap frame for the return to user mode as if from a
	121	* syscall. This copies the user mode register values.
	122	*
	123	* pcb_rsp must allocate an additional call-return pointer below
	124	* the trap frame which will be restored by cpu_heavy_restore from
	125	* PCB_RIP, and the thread's td_sp pointer must allocate an
	126	* additonal two quadwords below the pcb_rsp call-return pointer to
	127	* hold the LWKT restore function pointer and rflags.
	128	*
	129	* The LWKT restore function pointer must be set to cpu_heavy_restore,
	130	* which is our standard heavy-weight process switch-in function.
	131	* YYY eventually we should shortcut fork_return and fork_trampoline
	132	* to use the LWKT restore function directly so we can get rid of
	133	* all the extra crap we are setting up.
	134	*/
	135	lp2->lwp_md.md_regs = (struct trapframe *)pcb2 - 1;
	136	bcopy(lp1->lwp_md.md_regs, lp2->lwp_md.md_regs, sizeof(*lp2->lwp_md.md_regs));
	137
	138	/*
	139	* Set registers for trampoline to user mode. Leave space for the
	140	* return address on stack. These are the kernel mode register values.
	141	*/
48ffc236	142	pcb2->pcb_cr3 = vtophys(vmspace_pmap(lp2->lwp_proc->p_vmspace)->pm_pml4);
c8fe38ae MD	143	pcb2->pcb_cr3 \|= PG_RW \| PG_U \| PG_V;
	144	pcb2->pcb_rbx = (unsigned long)fork_return; /* fork_trampoline argument */
	145	pcb2->pcb_rbp = 0;
	146	pcb2->pcb_rsp = (unsigned long)lp2->lwp_md.md_regs - sizeof(void *);
	147	pcb2->pcb_r12 = (unsigned long)lp2; /* fork_trampoline argument */
	148	pcb2->pcb_r13 = 0;
	149	pcb2->pcb_r14 = 0;
	150	pcb2->pcb_r15 = 0;
	151	pcb2->pcb_rip = (unsigned long)fork_trampoline;
	152	lp2->lwp_thread->td_sp = (char )(pcb2->pcb_rsp - sizeof(void ));
	153	(u_int64_t )lp2->lwp_thread->td_sp = PSL_USER;
	154	lp2->lwp_thread->td_sp -= sizeof(void *);
	155	(void )lp2->lwp_thread->td_sp = (void )cpu_heavy_restore;
	156
	157	/*
	158	* pcb2->pcb_ldt: duplicated below, if necessary.
	159	* pcb2->pcb_savefpu: cloned above.
	160	* pcb2->pcb_flags: cloned above (always 0 here?).
93ad6da2 MD	161	* pcb2->pcb_onfault: cloned above (always NULL here).
93ad6da2 MD	162	* pcb2->pcb_onfault_sp:cloned above (dont care)
c8fe38ae MD	163	*/
	164
	165	/*
	166	* XXX don't copy the i/o pages. this should probably be fixed.
	167	*/
d8061892	168	pcb2->pcb_ext = NULL;
c8fe38ae MD	169
c8fe38ae MD	170	/* Copy the LDT, if necessary. */
d8061892	171	if (pcb2->pcb_ldt != NULL) {
c8fe38ae MD	172	if (flags & RFMEM) {
	173	pcb2->pcb_ldt->ldt_refcnt++;
	174	} else {
	175	pcb2->pcb_ldt = user_ldt_alloc(pcb2,
	176	pcb2->pcb_ldt->ldt_len);
	177	}
	178	}
	179	bcopy(&lp1->lwp_thread->td_tls, &lp2->lwp_thread->td_tls,
	180	sizeof(lp2->lwp_thread->td_tls));
	181	/*
	182	* Now, cpu_switch() can schedule the new lwp.
	183	* pcb_rsp is loaded pointing to the cpu_switch() stack frame
	184	* containing the return address when exiting cpu_switch.
	185	* This will normally be to fork_trampoline(), which will have
	186	* %rbx loaded with the new lwp's pointer. fork_trampoline()
	187	* will set up a stack to call fork_return(lp, frame); to complete
	188	* the return to user-mode.
	189	*/
d7f50089 YY	190	}
	191
	192	/*
	193	* Prepare new lwp to return to the address specified in params.
	194	*/
	195	int
	196	cpu_prepare_lwp(struct lwp lp, struct lwp_params params)
	197	{
42d8ab5f JG	198	struct trapframe *regs = lp->lwp_md.md_regs;
	199	void *bad_return = NULL;
	200	int error;
	201
	202	regs->tf_rip = (long)params->func;
	203	regs->tf_rsp = (long)params->stack;
	204	/* Set up argument for function call */
3338cc67 MD	205	regs->tf_rdi = (long)params->arg;
3338cc67 MD	206
42d8ab5f JG	207	/*
	208	* Set up fake return address. As the lwp function may never return,
	209	* we simply copy out a NULL pointer and force the lwp to receive
	210	* a SIGSEGV if it returns anyways.
	211	*/
	212	regs->tf_rsp -= sizeof(void *);
	213	error = copyout(&bad_return, (void *)regs->tf_rsp, sizeof(bad_return));
	214	if (error)
	215	return (error);
	216
	217	cpu_set_fork_handler(lp,
	218	(void ()(void , struct trapframe *))generic_lwp_return, lp);
d7f50089 YY	219	return (0);
	220	}
	221
	222	/*
	223	* Intercept the return address from a freshly forked process that has NOT
	224	* been scheduled yet.
	225	*
	226	* This is needed to make kernel threads stay in kernel mode.
	227	*/
	228	void
	229	cpu_set_fork_handler(struct lwp lp, void (func)(void , struct trapframe ),
	230	void *arg)
	231	{
c8fe38ae MD	232	/*
	233	* Note that the trap frame follows the args, so the function
	234	* is really called like this: func(arg, frame);
	235	*/
	236	lp->lwp_thread->td_pcb->pcb_rbx = (long)func; /* function */
	237	lp->lwp_thread->td_pcb->pcb_r12 = (long)arg; /* first arg */
d7f50089 YY	238	}
	239
	240	void
	241	cpu_set_thread_handler(thread_t td, void (rfunc)(void), void func, void *arg)
	242	{
c8fe38ae MD	243	td->td_pcb->pcb_rbx = (long)func;
	244	td->td_pcb->pcb_r12 = (long)arg;
	245	td->td_switch = cpu_lwkt_switch;
	246	td->td_sp -= sizeof(void *);
	247	(void )td->td_sp = rfunc; / exit function on return */
	248	td->td_sp -= sizeof(void *);
	249	(void *)td->td_sp = cpu_kthread_restore;
d7f50089 YY	250	}
	251
	252	void
	253	cpu_lwp_exit(void)
	254	{
0855a2af	255	struct thread *td = curthread;
48ffc236	256	struct pcb *pcb;
2883d2d8	257
48ffc236	258	pcb = td->td_pcb;
2883d2d8 MD	259
	260	/* Some i386 functionality was dropped */
	261	KKASSERT(pcb->pcb_ext == NULL);
	262
	263	/*
	264	* disable all hardware breakpoints
	265	*/
0855a2af	266	if (pcb->pcb_flags & PCB_DBREGS) {
0855a2af JG	267	reset_dbregs();
	268	pcb->pcb_flags &= ~PCB_DBREGS;
	269	}
	270	td->td_gd->gd_cnt.v_swtch++;
	271
	272	crit_enter_quick(td);
09b62fa4 MD	273	if (td->td_flags & TDF_TSLEEPQ)
09b62fa4 MD	274	tsleep_remove(td);
0855a2af JG	275	lwkt_deschedule_self(td);
	276	lwkt_remove_tdallq(td);
	277	cpu_thread_exit();
d7f50089 YY	278	}
	279
	280	/*
	281	* Terminate the current thread. The caller must have already acquired
	282	* the thread's rwlock and placed it on a reap list or otherwise notified
	283	* a reaper of its existance. We set a special assembly switch function which
	284	* releases td_rwlock after it has cleaned up the MMU state and switched
	285	* out the stack.
	286	*
	287	* Must be caller from a critical section and with the thread descheduled.
	288	*/
	289	void
	290	cpu_thread_exit(void)
	291	{
9e6e869e	292	npxexit();
0855a2af JG	293	curthread->td_switch = cpu_exit_switch;
	294	curthread->td_flags \|= TDF_EXITING;
	295	lwkt_switch();
	296	panic("cpu_thread_exit: lwkt_switch() unexpectedly returned");
d7f50089 YY	297	}
d7f50089 YY	298
c8fe38ae MD	299	void
	300	cpu_reset(void)
	301	{
	302	cpu_reset_real();
	303	}
	304
	305	static void
	306	cpu_reset_real(void)
	307	{
	308	/*
	309	* Attempt to do a CPU reset via the keyboard controller,
	310	* do not turn of the GateA20, as any machine that fails
	311	* to do the reset here would then end up in no man's land.
	312	*/
	313
	314	#if !defined(BROKEN_KEYBOARD_RESET)
	315	outb(IO_KBD + 4, 0xFE);
	316	DELAY(500000); /* wait 0.5 sec to see if that did it */
	317	kprintf("Keyboard reset did not work, attempting CPU shutdown\n");
	318	DELAY(1000000); /* wait 1 sec for kprintf to complete */
	319	#endif
	320	#if JG
	321	/* force a shutdown by unmapping entire address space ! */
	322	bzero((caddr_t) PTD, PAGE_SIZE);
	323	#endif
	324
	325	/* "good night, sweet prince .... <THUNK!>" */
	326	cpu_invltlb();
	327	/* NOTREACHED */
	328	while(1);
	329	}
	330
00039178 YT	331	/*
	332	* Convert kernel VA to physical address
	333	*/
	334	vm_paddr_t
	335	kvtop(void *addr)
	336	{
	337	vm_paddr_t pa;
	338
	339	pa = pmap_kextract((vm_offset_t)addr);
	340	if (pa == 0)
	341	panic("kvtop: zero page frame");
	342	return (pa);
	343	}
	344
0d4f82f8 MD	345	static void
	346	swi_vm(void arg, void frame)
	347	{
	348	if (busdma_swi_pending != 0)
	349	busdma_swi();
	350	}
	351
	352	static void
	353	swi_vm_setup(void *arg)
	354	{
1da8d52f	355	register_swi(SWI_VM, swi_vm, NULL, "swi_vm", NULL, 0);
0d4f82f8 MD	356	}
	357
	358	SYSINIT(vm_setup, SI_BOOT2_MACHDEP, SI_ORDER_ANY, swi_vm_setup, NULL);
	359
c8fe38ae	360	/*
b2b3ffcd	361	* platform-specific vmspace initialization (nothing for x86_64)
c8fe38ae MD	362	*/
	363	void
	364	cpu_vmspace_alloc(struct vmspace *vm __unused)
	365	{
	366	}
	367
	368	void
	369	cpu_vmspace_free(struct vmspace *vm __unused)
	370	{
d7f50089 YY	371	}
	372
	373	int
	374	kvm_access_check(vm_offset_t saddr, vm_offset_t eaddr, int prot)
	375	{
0855a2af JG	376	vm_offset_t addr;
	377
	378	if (saddr < KvaStart)
	379	return EFAULT;
	380	if (eaddr >= KvaEnd)
	381	return EFAULT;
	382	for (addr = saddr; addr < eaddr; addr += PAGE_SIZE) {
	383	if (pmap_extract(&kernel_pmap, addr) == 0)
	384	return EFAULT;
	385	}
	386	if (!kernacc((caddr_t)saddr, eaddr - saddr, prot))
	387	return EFAULT;
d7f50089 YY	388	return 0;
	389	}
	390
3338cc67 MD	391	#if 0
	392
	393	void _test_frame_enter(struct trapframe *frame);
	394	void _test_frame_exit(struct trapframe *frame);
	395
	396	void
	397	_test_frame_enter(struct trapframe *frame)
	398	{
	399	thread_t td = curthread;
	400
	401	if (ISPL(frame->tf_cs) == SEL_UPL) {
	402	KKASSERT(td->td_lwp);
	403	KASSERT(td->td_lwp->lwp_md.md_regs == frame,
	404	("_test_frame_exit: Frame mismatch %p %p",
	405	td->td_lwp->lwp_md.md_regs, frame));
	406	td->td_lwp->lwp_saveusp = (void *)frame->tf_rsp;
	407	td->td_lwp->lwp_saveupc = (void *)frame->tf_rip;
	408	}
	409	if ((char *)frame < td->td_kstack \|\|
	410	(char *)frame > td->td_kstack + td->td_kstack_size) {
ed20d0e3	411	panic("_test_frame_exit: frame not on kstack %p kstack=%p",
3338cc67 MD	412	frame, td->td_kstack);
	413	}
	414	}
	415
	416	void
	417	_test_frame_exit(struct trapframe *frame)
	418	{
	419	thread_t td = curthread;
	420
	421	if (ISPL(frame->tf_cs) == SEL_UPL) {
	422	KKASSERT(td->td_lwp);
	423	KASSERT(td->td_lwp->lwp_md.md_regs == frame,
	424	("_test_frame_exit: Frame mismatch %p %p",
	425	td->td_lwp->lwp_md.md_regs, frame));
	426	if (td->td_lwp->lwp_saveusp != (void *)frame->tf_rsp) {
	427	kprintf("_test_frame_exit: %s:%d usp mismatch %p/%p\n",
	428	td->td_comm, td->td_proc->p_pid,
	429	td->td_lwp->lwp_saveusp,
	430	(void *)frame->tf_rsp);
	431	}
	432	if (td->td_lwp->lwp_saveupc != (void *)frame->tf_rip) {
	433	kprintf("_test_frame_exit: %s:%d upc mismatch %p/%p\n",
	434	td->td_comm, td->td_proc->p_pid,
	435	td->td_lwp->lwp_saveupc,
	436	(void *)frame->tf_rip);
	437	}
	438
	439	/*
	440	* adulterate the fields to catch entries that
	441	* don't run through test_frame_enter
	442	*/
	443	td->td_lwp->lwp_saveusp =
	444	(void *)~(intptr_t)td->td_lwp->lwp_saveusp;
	445	td->td_lwp->lwp_saveupc =
	446	(void *)~(intptr_t)td->td_lwp->lwp_saveupc;
	447	}
	448	if ((char *)frame < td->td_kstack \|\|
	449	(char *)frame > td->td_kstack + td->td_kstack_size) {
ed20d0e3	450	panic("_test_frame_exit: frame not on kstack %p kstack=%p",
3338cc67 MD	451	frame, td->td_kstack);
	452	}
	453	}
	454
	455	#endif