xref: /src/sys/arch/alpha/alpha/machdep.c

/* $NetBSD: machdep.c,v 1.113 1998/03/17 05:00:18 thorpej Exp $ */

/*-
 * Copyright (c) 1998 The NetBSD Foundation, Inc.
 * All rights reserved.
 *
 * This code is derived from software contributed to The NetBSD Foundation
 * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
 * NASA Ames Research Center and by Chris G. Demetriou.
 *
 * 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) 1994, 1995, 1996 Carnegie-Mellon University.
 * All rights reserved.
 *
 * Author: Chris G. Demetriou
 *
 * Permission to use, copy, modify and distribute this software and
 * its documentation is hereby granted, provided that both the copyright
 * notice and this permission notice appear in all copies of the
 * software, derivative works or modified versions, and any portions
 * thereof, and that both notices appear in supporting documentation.
 *
 * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
 * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
 * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
 *
 * Carnegie Mellon requests users of this software to return to
 *
 *  Software Distribution Coordinator  or  Software.Distribution (at) CS.CMU.EDU
 *  School of Computer Science
 *  Carnegie Mellon University
 *  Pittsburgh PA 15213-3890
 *
 * any improvements or extensions that they make and grant Carnegie the
 * rights to redistribute these changes.
 */

#include "opt_uvm.h"
#include "opt_pmap_new.h"

#include <sys/cdefs.h>			/* RCS ID & Copyright macro defns */

__KERNEL_RCSID(0, "$NetBSD: machdep.c,v 1.113 1998/03/17 05:00:18 thorpej Exp $");

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/signalvar.h>
#include <sys/kernel.h>
#include <sys/map.h>
#include <sys/proc.h>
#include <sys/buf.h>
#include <sys/reboot.h>
#include <sys/device.h>
#include <sys/file.h>
#ifdef REAL_CLISTS
#include <sys/clist.h>
#endif
#include <sys/callout.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/mman.h>
#include <sys/msgbuf.h>
#include <sys/ioctl.h>
#include <sys/tty.h>
#include <sys/user.h>
#include <sys/exec.h>
#include <sys/exec_ecoff.h>
#include <vm/vm.h>
#include <sys/sysctl.h>
#include <sys/core.h>
#include <sys/kcore.h>
#include <machine/kcore.h>
#ifdef SYSVMSG
#include <sys/msg.h>
#endif
#ifdef SYSVSEM
#include <sys/sem.h>
#endif
#ifdef SYSVSHM
#include <sys/shm.h>
#endif

#include <sys/mount.h>
#include <sys/syscallargs.h>

#include <vm/vm_kern.h>

#if defined(UVM)
#include <uvm/uvm_extern.h>
#endif

#include <dev/cons.h>

#include <machine/autoconf.h>
#include <machine/cpu.h>
#include <machine/reg.h>
#include <machine/rpb.h>
#include <machine/prom.h>
#include <machine/conf.h>

#include <net/netisr.h>
#include <net/if.h>

#ifdef INET
#include <netinet/in.h>
#include <netinet/ip_var.h>
#include "arp.h"
#if NARP > 0
#include <netinet/if_inarp.h>
#endif
#endif
#ifdef NS
#include <netns/ns_var.h>
#endif
#ifdef ISO
#include <netiso/iso.h>
#include <netiso/clnp.h>
#endif
#ifdef CCITT
#include <netccitt/x25.h>
#include <netccitt/pk.h>
#include <netccitt/pk_extern.h>
#endif
#ifdef NATM
#include <netnatm/natm.h>
#endif
#ifdef NETATALK
#include <netatalk/at_extern.h>
#endif
#include "ppp.h"
#if NPPP > 0
#include <net/ppp_defs.h>
#include <net/if_ppp.h>
#endif

#ifdef DDB
#include <machine/db_machdep.h>
#include <ddb/db_access.h>
#include <ddb/db_sym.h>
#include <ddb/db_extern.h>
#include <ddb/db_interface.h>
#endif

#if defined(UVM)
vm_map_t exec_map = NULL;
vm_map_t mb_map = NULL;
vm_map_t phys_map = NULL;
#else
vm_map_t buffer_map;
#endif

/*
 * Declare these as initialized data so we can patch them.
 */
int	nswbuf = 0;
#ifdef	NBUF
int	nbuf = NBUF;
#else
int	nbuf = 0;
#endif
#ifdef	BUFPAGES
int	bufpages = BUFPAGES;
#else
int	bufpages = 0;
#endif
caddr_t msgbufaddr;

int	maxmem;			/* max memory per process */

int	totalphysmem;		/* total amount of physical memory in system */
int	physmem;		/* physical memory used by NetBSD + some rsvd */
int	resvmem;		/* amount of memory reserved for PROM */
int	unusedmem;		/* amount of memory for OS that we don't use */
int	unknownmem;		/* amount of memory with an unknown use */

int	cputype;		/* system type, from the RPB */

/*
 * XXX We need an address to which we can assign things so that they
 * won't be optimized away because we didn't use the value.
 */
u_int32_t no_optimize;

/* the following is used externally (sysctl_hw) */
char	machine[] = MACHINE;		/* from <machine/param.h> */
char	machine_arch[] = MACHINE_ARCH;	/* from <machine/param.h> */
char	cpu_model[128];

struct	user *proc0paddr;

/* Number of machine cycles per microsecond */
u_int64_t	cycles_per_usec;

/* number of cpus in the box.  really! */
int		ncpus;

struct bootinfo_kernel bootinfo;

struct platform platform;

u_int32_t vm_mbuf_size = _VM_MBUF_SIZE;
u_int32_t vm_kmem_size = _VM_KMEM_SIZE;
u_int32_t vm_phys_size = _VM_PHYS_SIZE;

#ifdef DDB
/* start and end of kernel symbol table */
void	*ksym_start, *ksym_end;
#endif

/* for cpu_sysctl() */
int	alpha_unaligned_print = 1;	/* warn about unaligned accesses */
int	alpha_unaligned_fix = 1;	/* fix up unaligned accesses */
int	alpha_unaligned_sigbus = 0;	/* don't SIGBUS on fixed-up accesses */

/*
 * XXX This should be dynamically sized, but we have the chicken-egg problem!
 * XXX it should also be larger than it is, because not all of the mddt
 * XXX clusters end up being used for VM.
 */
phys_ram_seg_t mem_clusters[VM_PHYSSEG_MAX];	/* low size bits overloaded */
int	mem_cluster_cnt;

caddr_t	allocsys __P((caddr_t));
int	cpu_dump __P((void));
int	cpu_dumpsize __P((void));
u_long	cpu_dump_mempagecnt __P((void));
void	dumpsys __P((void));
void	identifycpu __P((void));
void	netintr __P((void));
void	printregs __P((struct reg *));

void
alpha_init(pfn, ptb, bim, bip, biv)
	u_long pfn;		/* first free PFN number */
	u_long ptb;		/* PFN of current level 1 page table */
	u_long bim;		/* bootinfo magic */
	u_long bip;		/* bootinfo pointer */
	u_long biv;		/* bootinfo version */
{
	extern char kernel_text[], _end[];
	struct mddt *mddtp;
	struct mddt_cluster *memc;
	int i, mddtweird;
	struct vm_physseg *vps;
	vm_offset_t kernstart, kernend;
	vm_offset_t kernstartpfn, kernendpfn, pfn0, pfn1;
	vm_size_t size;
	char *p;
	caddr_t v;
	char *bootinfo_msg;

	/* NO OUTPUT ALLOWED UNTIL FURTHER NOTICE */

	/*
	 * Turn off interrupts (not mchecks) and floating point.
	 * Make sure the instruction and data streams are consistent.
	 */
	(void)alpha_pal_swpipl(ALPHA_PSL_IPL_HIGH);
	alpha_pal_wrfen(0);
	ALPHA_TBIA();
	alpha_pal_imb();

	/*
	 * Get critical system information (if possible, from the
	 * information provided by the boot program).
	 */
	bootinfo_msg = NULL;
	if (bim == BOOTINFO_MAGIC) {
		if (biv == 0) {		/* backward compat */
			biv = *(u_long *)bip;
			bip += 8;
		}
		switch (biv) {
		case 1: {
			struct bootinfo_v1 *v1p = (struct bootinfo_v1 *)bip;

			bootinfo.ssym = v1p->ssym;
			bootinfo.esym = v1p->esym;
			/* hwrpb may not be provided by boot block in v1 */
			if (v1p->hwrpb != NULL) {
				bootinfo.hwrpb_phys =
				    ((struct rpb *)v1p->hwrpb)->rpb_phys;
				bootinfo.hwrpb_size = v1p->hwrpbsize;
			} else {
				bootinfo.hwrpb_phys =
				    ((struct rpb *)HWRPB_ADDR)->rpb_phys;
				bootinfo.hwrpb_size =
				    ((struct rpb *)HWRPB_ADDR)->rpb_size;
			}
			bcopy(v1p->boot_flags, bootinfo.boot_flags,
			    min(sizeof v1p->boot_flags,
			      sizeof bootinfo.boot_flags));
			bcopy(v1p->booted_kernel, bootinfo.booted_kernel,
			    min(sizeof v1p->booted_kernel,
			      sizeof bootinfo.booted_kernel));
			/* booted dev not provided in bootinfo */
			init_prom_interface((struct rpb *)
			    ALPHA_PHYS_TO_K0SEG(bootinfo.hwrpb_phys));
                	prom_getenv(PROM_E_BOOTED_DEV, bootinfo.booted_dev,
			    sizeof bootinfo.booted_dev);
			break;
		}
		default:
			bootinfo_msg = "unknown bootinfo version";
			goto nobootinfo;
		}
	} else {
		bootinfo_msg = "boot program did not pass bootinfo";
nobootinfo:
		bootinfo.ssym = (u_long)_end;
		bootinfo.esym = (u_long)_end;
		bootinfo.hwrpb_phys = ((struct rpb *)HWRPB_ADDR)->rpb_phys;
		bootinfo.hwrpb_size = ((struct rpb *)HWRPB_ADDR)->rpb_size;
		init_prom_interface((struct rpb *)HWRPB_ADDR);
		prom_getenv(PROM_E_BOOTED_OSFLAGS, bootinfo.boot_flags,
		    sizeof bootinfo.boot_flags);
		prom_getenv(PROM_E_BOOTED_FILE, bootinfo.booted_kernel,
		    sizeof bootinfo.booted_kernel);
		prom_getenv(PROM_E_BOOTED_DEV, bootinfo.booted_dev,
		    sizeof bootinfo.booted_dev);
	}

	/*
	 * Initialize the kernel's mapping of the RPB.  It's needed for
	 * lots of things.
	 */
	hwrpb = (struct rpb *)ALPHA_PHYS_TO_K0SEG(bootinfo.hwrpb_phys);

	/*
	 * Remember how many cycles there are per microsecond,
	 * so that we can use delay().  Round up, for safety.
	 */
	cycles_per_usec = (hwrpb->rpb_cc_freq + 999999) / 1000000;

	/*
	 * Initalize the (temporary) bootstrap console interface, so
	 * we can use printf until the VM system starts being setup.
	 * The real console is initialized before then.
	 */
	init_bootstrap_console();

	/* OUTPUT NOW ALLOWED */

	/* delayed from above */
	if (bootinfo_msg)
		printf("WARNING: %s (0x%lx, 0x%lx, 0x%lx)\n",
		    bootinfo_msg, bim, bip, biv);

	/*
	 * Point interrupt/exception vectors to our own.
	 */
	alpha_pal_wrent(XentInt, ALPHA_KENTRY_INT);
	alpha_pal_wrent(XentArith, ALPHA_KENTRY_ARITH);
	alpha_pal_wrent(XentMM, ALPHA_KENTRY_MM);
	alpha_pal_wrent(XentIF, ALPHA_KENTRY_IF);
	alpha_pal_wrent(XentUna, ALPHA_KENTRY_UNA);
	alpha_pal_wrent(XentSys, ALPHA_KENTRY_SYS);

	/*
	 * Clear pending machine checks and error reports, and enable
	 * system- and processor-correctable error reporting.
	 */
	alpha_pal_wrmces(alpha_pal_rdmces() &
	    ~(ALPHA_MCES_DSC|ALPHA_MCES_DPC));

	/*
	 * Find out what hardware we're on, and do basic initialization.
	 */
	cputype = hwrpb->rpb_type;
	if (cputype >= ncpuinit) {
		platform_not_supported();
		/* NOTREACHED */
	}
	(*cpuinit[cputype].init)();
	strcpy(cpu_model, platform.model);

	/*
	 * Initalize the real console, so the the bootstrap console is
	 * no longer necessary.
	 */
#ifdef _PMAP_MAY_USE_PROM_CONSOLE
	if (!pmap_uses_prom_console())
#endif
		(*platform.cons_init)();

#ifdef DIAGNOSTIC
	/* Paranoid sanity checking */

	/* We should always be running on the the primary. */
	assert(hwrpb->rpb_primary_cpu_id == alpha_pal_whami());

	/* On single-CPU systypes, the primary should always be CPU 0. */
	if (cputype != ST_DEC_21000)
		assert(hwrpb->rpb_primary_cpu_id == 0);
#endif

	/* NO MORE FIRMWARE ACCESS ALLOWED */
#ifdef _PMAP_MAY_USE_PROM_CONSOLE
	/*
	 * XXX (unless _PMAP_MAY_USE_PROM_CONSOLE is defined and
	 * XXX pmap_uses_prom_console() evaluates to non-zero.)
	 */
#endif

	/*
	 * find out this system's page size
	 */
	PAGE_SIZE = hwrpb->rpb_page_size;
	if (PAGE_SIZE != 8192)
		panic("page size %d != 8192?!", PAGE_SIZE);

	/*
	 * Initialize PAGE_SIZE-dependent variables.
	 */
#if defined(UVM)
	uvm_setpagesize();
#else
	vm_set_page_size();
#endif

	/*
	 * Find the beginning and end of the kernel (and leave a
	 * bit of space before the beginning for the bootstrap
	 * stack).
	 */
	kernstart = trunc_page(kernel_text) - 2 * PAGE_SIZE;
#ifdef DDB
	ksym_start = (void *)bootinfo.ssym;
	ksym_end   = (void *)bootinfo.esym;
	kernend = (vm_offset_t)round_page(ksym_end);
#else
	kernend = (vm_offset_t)round_page(_end);
#endif

	kernstartpfn = atop(ALPHA_K0SEG_TO_PHYS(kernstart));
	kernendpfn = atop(ALPHA_K0SEG_TO_PHYS(kernend));

	/*
	 * Find out how much memory is available, by looking at
	 * the memory cluster descriptors.  This also tries to do
	 * its best to detect things things that have never been seen
	 * before...
	 */
	mddtp = (struct mddt *)(((caddr_t)hwrpb) + hwrpb->rpb_memdat_off);

	/* MDDT SANITY CHECKING */
	mddtweird = 0;
	if (mddtp->mddt_cluster_cnt < 2) {
		mddtweird = 1;
		printf("WARNING: weird number of mem clusters: %d\n",
		    mddtp->mddt_cluster_cnt);
	}

#if 0
	printf("Memory cluster count: %d\n", mddtp->mddt_cluster_cnt);
#endif

	for (i = 0; i < mddtp->mddt_cluster_cnt; i++) {
		memc = &mddtp->mddt_clusters[i];
#if 0
		printf("MEMC %d: pfn 0x%lx cnt 0x%lx usage 0x%lx\n", i,
		    memc->mddt_pfn, memc->mddt_pg_cnt, memc->mddt_usage);
#endif
		totalphysmem += memc->mddt_pg_cnt;
		if (mem_cluster_cnt < VM_PHYSSEG_MAX) {	/* XXX */
			mem_clusters[mem_cluster_cnt].start =
			    ptoa(memc->mddt_pfn);
			mem_clusters[mem_cluster_cnt].size =
			    ptoa(memc->mddt_pg_cnt);
			if (memc->mddt_usage & MDDT_mbz ||
			    memc->mddt_usage & MDDT_NONVOLATILE || /* XXX */
			    memc->mddt_usage & MDDT_PALCODE)
				mem_clusters[mem_cluster_cnt].size |=
				    PROT_READ;
			else
				mem_clusters[mem_cluster_cnt].size |=
				    PROT_READ | PROT_WRITE | PROT_EXEC;
			mem_cluster_cnt++;
		}

		if (memc->mddt_usage & MDDT_mbz) {
			mddtweird = 1;
			printf("WARNING: mem cluster %d has weird "
			    "usage 0x%lx\n", i, memc->mddt_usage);
			unknownmem += memc->mddt_pg_cnt;
			continue;
		}
		if (memc->mddt_usage & MDDT_NONVOLATILE) {
			/* XXX should handle these... */
			printf("WARNING: skipping non-volatile mem "
			    "cluster %d\n", i);
			unusedmem += memc->mddt_pg_cnt;
			continue;
		}
		if (memc->mddt_usage & MDDT_PALCODE) {
			resvmem += memc->mddt_pg_cnt;
			continue;
		}

		/*
		 * We have a memory cluster available for system
		 * software use.  We must determine if this cluster
		 * holds the kernel.
		 */
#ifdef _PMAP_MAY_USE_PROM_CONSOLE
		/*
		 * XXX If the kernel uses the PROM console, we only use the
		 * XXX memory after the kernel in the first system segment,
		 * XXX to avoid clobbering prom mapping, data, etc.
		 */
	    if (!pmap_uses_prom_console() || physmem == 0) {
#endif /* _PMAP_MAY_USE_PROM_CONSOLE */
		physmem += memc->mddt_pg_cnt;
		pfn0 = memc->mddt_pfn;
		pfn1 = memc->mddt_pfn + memc->mddt_pg_cnt;
		if (pfn0 <= kernstartpfn && kernendpfn <= pfn1) {
			/*
			 * Must compute the location of the kernel
			 * within the segment.
			 */
#if 0
			printf("Cluster %d contains kernel\n", i);
#endif
#ifdef _PMAP_MAY_USE_PROM_CONSOLE
		    if (!pmap_uses_prom_console()) {
#endif /* _PMAP_MAY_USE_PROM_CONSOLE */
			if (pfn0 < kernstartpfn) {
				/*
				 * There is a chunk before the kernel.
				 */
#if 0
				printf("Loading chunk before kernel: "
				    "0x%lx / 0x%lx\n", pfn0, kernstartpfn);
#endif
#if defined(UVM)
				uvm_page_physload(pfn0, kernstartpfn,
				    pfn0, kernstartpfn);
#else
				vm_page_physload(pfn0, kernstartpfn,
				    pfn0, kernstartpfn);
#endif
			}
#ifdef _PMAP_MAY_USE_PROM_CONSOLE
		    }
#endif /* _PMAP_MAY_USE_PROM_CONSOLE */
			if (kernendpfn < pfn1) {
				/*
				 * There is a chunk after the kernel.
				 */
#if 0
				printf("Loading chunk after kernel: "
				    "0x%lx / 0x%lx\n", kernendpfn, pfn1);
#endif
#if defined(UVM)
				uvm_page_physload(kernendpfn, pfn1,
				    kernendpfn, pfn1);
#else
				vm_page_physload(kernendpfn, pfn1,
				    kernendpfn, pfn1);
#endif
			}
		} else {
			/*
			 * Just load this cluster as one chunk.
			 */
#if 0
			printf("Loading cluster %d: 0x%lx / 0x%lx\n", i,
			    pfn0, pfn1);
#endif
#if defined(UVM)
			uvm_page_physload(pfn0, pfn1, pfn0, pfn1);
#else
			vm_page_physload(pfn0, pfn1, pfn0, pfn1);
#endif
		}
#ifdef _PMAP_MAY_USE_PROM_CONSOLE
	    }
#endif /* _PMAP_MAY_USE_PROM_CONSOLE */
	}

	/*
	 * Dump out the MDDT if it looks odd...
	 */
	if (mddtweird) {
		printf("\n");
		printf("complete memory cluster information:\n");
		for (i = 0; i < mddtp->mddt_cluster_cnt; i++) {
			printf("mddt %d:\n", i);
			printf("\tpfn %lx\n",
			    mddtp->mddt_clusters[i].mddt_pfn);
			printf("\tcnt %lx\n",
			    mddtp->mddt_clusters[i].mddt_pg_cnt);
			printf("\ttest %lx\n",
			    mddtp->mddt_clusters[i].mddt_pg_test);
			printf("\tbva %lx\n",
			    mddtp->mddt_clusters[i].mddt_v_bitaddr);
			printf("\tbpa %lx\n",
			    mddtp->mddt_clusters[i].mddt_p_bitaddr);
			printf("\tbcksum %lx\n",
			    mddtp->mddt_clusters[i].mddt_bit_cksum);
			printf("\tusage %lx\n",
			    mddtp->mddt_clusters[i].mddt_usage);
		}
		printf("\n");
	}

	if (totalphysmem == 0)
		panic("can't happen: system seems to have no memory!");

#ifdef LIMITMEM
	/*
	 * XXX Kludge so we can run on machines with memory larger
	 * XXX than 1G until all device drivers are converted to
	 * XXX use bus_dma.  (Relies on the fact that vm_physmem
	 * XXX sorted in order of increasing addresses.)
	 */
	if (vm_physmem[vm_nphysseg - 1].end > atop(LIMITMEM * 1024 * 1024)) {

		printf("******** LIMITING MEMORY TO %dMB **********\n",
		    LIMITMEM);

		do {
			u_long ovf;

			vps = &vm_physmem[vm_nphysseg - 1];

			if (vps->start >= atop(LIMITMEM * 1024 * 1024)) {
				/*
				 * If the start is too high, just drop
				 * the whole segment.
				 *
				 * XXX can start != avail_start in this
				 * XXX case?  wouldn't that mean that
				 * XXX some memory was stolen above the
				 * XXX limit?  What to do?
				 */
				ovf = vps->end - vps->start;
				vm_nphysseg--;
			} else {
				/*
				 * If the start is OK, calculate how much
				 * to drop and drop it.
				 */
				ovf = vps->end - atop(LIMITMEM * 1024 * 1024);
				vps->end -= ovf;
				vps->avail_end -= ovf;
			}
			physmem -= ovf;
			unusedmem += ovf;
		} while (vps->end > atop(LIMITMEM * 1024 * 1024));
	}
#endif /* LIMITMEM */

	maxmem = physmem;

#if 0
	printf("totalphysmem = %d\n", totalphysmem);
	printf("physmem = %d\n", physmem);
	printf("resvmem = %d\n", resvmem);
	printf("unusedmem = %d\n", unusedmem);
	printf("unknownmem = %d\n", unknownmem);
#endif

	/*
	 * Adjust some parameters if the amount of physmem
	 * available would cause us to croak. This is completely
	 * eyeballed and isn't meant to be the final answer.
	 * vm_phys_size is probably the only one to really worry
	 * about.
 	 *
	 * It's for booting a GENERIC kernel on a large memory platform.
	 */
	if (physmem >= atop(128 * 1024 * 1024)) {
		vm_mbuf_size <<= 1;
		vm_kmem_size <<= 3;
		vm_phys_size <<= 2;
	}

	/*
	 * Initialize error message buffer (at end of core).
	 */
	{
		size_t sz = round_page(MSGBUFSIZE);

		vps = &vm_physmem[vm_nphysseg - 1];

		/* shrink so that it'll fit in the last segment */
		if ((vps->avail_end - vps->avail_start) < atop(sz))
			sz = ptoa(vps->avail_end - vps->avail_start);

		vps->end -= atop(sz);
		vps->avail_end -= atop(sz);
		msgbufaddr = (caddr_t) ALPHA_PHYS_TO_K0SEG(ptoa(vps->end));
		initmsgbuf(msgbufaddr, sz);

		/* Remove the last segment if it now has no pages. */
		if (vps->start == vps->end)
			vm_nphysseg--;

		/* warn if the message buffer had to be shrunk */
		if (sz != round_page(MSGBUFSIZE))
			printf("WARNING: %d bytes not available for msgbuf in last cluster (%d used)\n",
			    round_page(MSGBUFSIZE), sz);

	}

	/*
	 * Init mapping for u page(s) for proc 0
	 */
	proc0.p_addr = proc0paddr =
	    (struct user *)pmap_steal_memory(UPAGES * PAGE_SIZE, NULL, NULL);

	/*
	 * Allocate space for system data structures.  These data structures
	 * are allocated here instead of cpu_startup() because physical
	 * memory is directly addressable.  We don't have to map these into
	 * virtual address space.
	 */
	size = (vm_size_t)allocsys(0);
	v = (caddr_t)pmap_steal_memory(size, NULL, NULL);
	if ((allocsys(v) - v) != size)
		panic("alpha_init: table size inconsistency");

	/*
	 * Initialize the virtual memory system, and set the
	 * page table base register in proc 0's PCB.
	 */
#ifndef NEW_PMAP
	pmap_bootstrap(ALPHA_PHYS_TO_K0SEG(ptb << PGSHIFT));
#else
	pmap_bootstrap(ALPHA_PHYS_TO_K0SEG(ptb << PGSHIFT),
	    hwrpb->rpb_max_asn);
#endif

	/*
	 * Initialize the rest of proc 0's PCB, and cache its physical
	 * address.
	 */
	proc0.p_md.md_pcbpaddr =
	    (struct pcb *)ALPHA_K0SEG_TO_PHYS((vm_offset_t)&proc0paddr->u_pcb);

	/*
	 * Set the kernel sp, reserving space for an (empty) trapframe,
	 * and make proc0's trapframe pointer point to it for sanity.
	 */
	proc0paddr->u_pcb.pcb_hw.apcb_ksp =
	    (u_int64_t)proc0paddr + USPACE - sizeof(struct trapframe);
	proc0.p_md.md_tf =
	    (struct trapframe *)proc0paddr->u_pcb.pcb_hw.apcb_ksp;

#ifdef NEW_PMAP
	/*
	 * Set up the kernel address space in proc0's hwpcb.
	 */
	PMAP_ACTIVATE(kernel_pmap, &proc0paddr->u_pcb.pcb_hw, 0);
#endif

	/*
	 * Look at arguments passed to us and compute boothowto.
	 */

	boothowto = RB_SINGLE;
#ifdef KADB
	boothowto |= RB_KDB;
#endif
	for (p = bootinfo.boot_flags; p && *p != '\0'; p++) {
		/*
		 * Note that we'd really like to differentiate case here,
		 * but the Alpha AXP Architecture Reference Manual
		 * says that we shouldn't.
		 */
		switch (*p) {
		case 'a': /* autoboot */
		case 'A':
			boothowto &= ~RB_SINGLE;
			break;

#ifdef DEBUG
		case 'c': /* crash dump immediately after autoconfig */
		case 'C':
			boothowto |= RB_DUMP;
			break;
#endif

#if defined(KGDB) || defined(DDB)
		case 'd': /* break into the kernel debugger ASAP */
		case 'D':
			boothowto |= RB_KDB;
			break;
#endif

		case 'h': /* always halt, never reboot */
		case 'H':
			boothowto |= RB_HALT;
			break;

#if 0
		case 'm': /* mini root present in memory */
		case 'M':
			boothowto |= RB_MINIROOT;
			break;
#endif

		case 'n': /* askname */
		case 'N':
			boothowto |= RB_ASKNAME;
			break;

		case 's': /* single-user (default, supported for sanity) */
		case 'S':
			boothowto |= RB_SINGLE;
			break;

		default:
			printf("Unrecognized boot flag '%c'.\n", *p);
			break;
		}
	}

	/*
	 * Initialize debuggers, and break into them if appropriate.
	 */
#ifdef DDB
	db_machine_init();
	ddb_init(ksym_start, ksym_end);
	if (boothowto & RB_KDB)
		Debugger();
#endif
#ifdef KGDB
	if (boothowto & RB_KDB)
		kgdb_connect(0);
#endif

	/*
	 * Figure out the number of cpus in the box, from RPB fields.
	 * Really.  We mean it.
	 */
	for (i = 0; i < hwrpb->rpb_pcs_cnt; i++) {
		struct pcs *pcsp;

		pcsp = (struct pcs *)((char *)hwrpb + hwrpb->rpb_pcs_off +
		    (i * hwrpb->rpb_pcs_size));
		if ((pcsp->pcs_flags & PCS_PP) != 0)
			ncpus++;
	}

	/*
	 * Figure out our clock frequency, from RPB fields.
	 */
	hz = hwrpb->rpb_intr_freq >> 12;
	if (!(60 <= hz && hz <= 10240)) {
		hz = 1024;
#ifdef DIAGNOSTIC
		printf("WARNING: unbelievable rpb_intr_freq: %ld (%d hz)\n",
			hwrpb->rpb_intr_freq, hz);
#endif
	}

}

/*
 * Allocate space for system data structures.  We are given
 * a starting virtual address and we return a final virtual
 * address; along the way we set each data structure pointer.
 *
 * We call allocsys() with 0 to find out how much space we want,
 * allocate that much and fill it with zeroes, and the call
 * allocsys() again with the correct base virtual address.
 */
caddr_t
allocsys(v)
	caddr_t v;
{

#define valloc(name, type, num) \
	    (name) = (type *)v; v = (caddr_t)ALIGN((name)+(num))
#ifdef REAL_CLISTS
	valloc(cfree, struct cblock, nclist);
#endif
	valloc(callout, struct callout, ncallout);
#ifdef SYSVSHM
	valloc(shmsegs, struct shmid_ds, shminfo.shmmni);
#endif
#ifdef SYSVSEM
	valloc(sema, struct semid_ds, seminfo.semmni);
	valloc(sem, struct sem, seminfo.semmns);
	/* This is pretty disgusting! */
	valloc(semu, int, (seminfo.semmnu * seminfo.semusz) / sizeof(int));
#endif
#ifdef SYSVMSG
	valloc(msgpool, char, msginfo.msgmax);
	valloc(msgmaps, struct msgmap, msginfo.msgseg);
	valloc(msghdrs, struct msg, msginfo.msgtql);
	valloc(msqids, struct msqid_ds, msginfo.msgmni);
#endif

	/*
	 * Determine how many buffers to allocate.
	 * We allocate 10% of memory for buffer space.  Insure a
	 * minimum of 16 buffers.  We allocate 1/2 as many swap buffer
	 * headers as file i/o buffers.
	 */
	if (bufpages == 0)
		bufpages = (physmem * 10) / (CLSIZE * 100);
	if (nbuf == 0) {
		nbuf = bufpages;
		if (nbuf < 16)
			nbuf = 16;
	}
	if (nswbuf == 0) {
		nswbuf = (nbuf / 2) &~ 1;	/* force even */
		if (nswbuf > 256)
			nswbuf = 256;		/* sanity */
	}
#if !defined(UVM)
	valloc(swbuf, struct buf, nswbuf);
#endif
	valloc(buf, struct buf, nbuf);
	return (v);
#undef valloc
}

void
consinit()
{

	/*
	 * Everything related to console initialization is done
	 * in alpha_init().
	 */
#if defined(DIAGNOSTIC) && defined(_PMAP_MAY_USE_PROM_CONSOLE)
	printf("consinit: %susing prom console\n",
	    pmap_uses_prom_console() ? "" : "not ");
#endif
}

void
cpu_startup()
{
	register unsigned i;
	int base, residual;
	vm_offset_t minaddr, maxaddr;
	vm_size_t size;
#if defined(DEBUG)
	extern int pmapdebug;
	int opmapdebug = pmapdebug;

	pmapdebug = 0;
#endif

	/*
	 * Good {morning,afternoon,evening,night}.
	 */
	printf(version);
	identifycpu();
	printf("real mem = %lu (%lu reserved for PROM, %lu used by NetBSD)\n",
	    ptoa(totalphysmem), ptoa(resvmem), ptoa(physmem));
	if (unusedmem)
		printf("WARNING: unused memory = %d bytes\n", ctob(unusedmem));
	if (unknownmem)
		printf("WARNING: %d bytes of memory with unknown purpose\n",
		    ctob(unknownmem));

	/*
	 * Allocate virtual address space for file I/O buffers.
	 * Note they are different than the array of headers, 'buf',
	 * and usually occupy more virtual memory than physical.
	 */
	size = MAXBSIZE * nbuf;
#if defined(UVM)
	if (uvm_map(kernel_map, (vm_offset_t *) &buffers, round_page(size),
		    NULL, UVM_UNKNOWN_OFFSET,
		    UVM_MAPFLAG(UVM_PROT_NONE, UVM_PROT_NONE, UVM_INH_NONE,
				UVM_ADV_NORMAL, 0)) != KERN_SUCCESS)
		panic("startup: cannot allocate VM for buffers");
#else
	buffer_map = kmem_suballoc(kernel_map, (vm_offset_t *)&buffers,
	    &maxaddr, size, TRUE);
	minaddr = (vm_offset_t)buffers;
	if (vm_map_find(buffer_map, vm_object_allocate(size), (vm_offset_t)0,
			&minaddr, size, FALSE) != KERN_SUCCESS)
		panic("startup: cannot allocate buffers");
#endif /* UVM */
	base = bufpages / nbuf;
	residual = bufpages % nbuf;
	for (i = 0; i < nbuf; i++) {
#if defined(UVM)
		vm_size_t curbufsize;
		vm_offset_t curbuf;
		struct vm_page *pg;

		/*
		 * Each buffer has MAXBSIZE bytes of VM space allocated.  Of
		 * that MAXBSIZE space, we allocate and map (base+1) pages
		 * for the first "residual" buffers, and then we allocate
		 * "base" pages for the rest.
		 */
		curbuf = (vm_offset_t) buffers + (i * MAXBSIZE);
		curbufsize = CLBYTES * ((i < residual) ? (base+1) : base);

		while (curbufsize) {
			pg = uvm_pagealloc(NULL, 0, NULL);
			if (pg == NULL)
				panic("cpu_startup: not enough memory for "
				    "buffer cache");
#if defined(PMAP_NEW)
			pmap_kenter_pgs(curbuf, &pg, 1);
#else
			pmap_enter(kernel_map->pmap, curbuf,
				   VM_PAGE_TO_PHYS(pg), VM_PROT_ALL, TRUE);
#endif
			curbuf += PAGE_SIZE;
			curbufsize -= PAGE_SIZE;
		}
#else /* ! UVM */
		vm_size_t curbufsize;
		vm_offset_t curbuf;

		/*
		 * First <residual> buffers get (base+1) physical pages
		 * allocated for them.  The rest get (base) physical pages.
		 *
		 * The rest of each buffer occupies virtual space,
		 * but has no physical memory allocated for it.
		 */
		curbuf = (vm_offset_t)buffers + i * MAXBSIZE;
		curbufsize = CLBYTES * (i < residual ? base+1 : base);
		vm_map_pageable(buffer_map, curbuf, curbuf+curbufsize, FALSE);
		vm_map_simplify(buffer_map, curbuf);
#endif /* UVM */
	}
	/*
	 * Allocate a submap for exec arguments.  This map effectively
	 * limits the number of processes exec'ing at any time.
	 */
#if defined(UVM)
	exec_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
				   16 * NCARGS, TRUE, FALSE, NULL);
#else
	exec_map = kmem_suballoc(kernel_map, &minaddr, &maxaddr,
				 16 * NCARGS, TRUE);
#endif

	/*
	 * Allocate a submap for physio
	 */
#if defined(UVM)
	phys_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
				   VM_PHYS_SIZE, TRUE, FALSE, NULL);
#else
	phys_map = kmem_suballoc(kernel_map, &minaddr, &maxaddr,
				 VM_PHYS_SIZE, TRUE);
#endif

	/*
	 * Finally, allocate mbuf cluster submap.
	 */
#if defined(UVM)
	mb_map = uvm_km_suballoc(kernel_map, (vm_offset_t *)&mbutl, &maxaddr,
				 VM_MBUF_SIZE, FALSE, FALSE, NULL);
#else
	mb_map = kmem_suballoc(kernel_map, (vm_offset_t *)&mbutl, &maxaddr,
			       VM_MBUF_SIZE, FALSE);
#endif
	/*
	 * Initialize callouts
	 */
	callfree = callout;
	for (i = 1; i < ncallout; i++)
		callout[i-1].c_next = &callout[i];
	callout[i-1].c_next = NULL;

#if defined(DEBUG)
	pmapdebug = opmapdebug;
#endif
#if defined(UVM)
	printf("avail mem = %ld\n", (long)ptoa(uvmexp.free));
#else
	printf("avail mem = %ld\n", (long)ptoa(cnt.v_free_count));
#endif
	printf("using %ld buffers containing %ld bytes of memory\n",
		(long)nbuf, (long)(bufpages * CLBYTES));

	/*
	 * Set up buffers, so they can be used to read disk labels.
	 */
	bufinit();

	/*
	 * Configure the system.
	 */
	configure();

	/*
	 * Note that bootstrapping is finished, and set the HWRPB up
	 * to do restarts.
	 */
	hwrpb_restart_setup();
}

/*
 * Retrieve the platform name from the DSR.
 */
const char *
alpha_dsr_sysname()
{
	struct dsrdb *dsr;
	const char *sysname;

	/*
	 * DSR does not exist on early HWRPB versions.
	 */
	if (hwrpb->rpb_version < HWRPB_DSRDB_MINVERS)
		return (NULL);

	dsr = (struct dsrdb *)(((caddr_t)hwrpb) + hwrpb->rpb_dsrdb_off);
	sysname = (const char *)((caddr_t)dsr + (dsr->dsr_sysname_off +
	    sizeof(u_int64_t)));
	return (sysname);
}

/*
 * Lookup the system specified system variation in the provided table,
 * returning the model string on match.
 */
const char *
alpha_variation_name(variation, avtp)
	u_int64_t variation;
	const struct alpha_variation_table *avtp;
{
	int i;

	for (i = 0; avtp[i].avt_model != NULL; i++)
		if (avtp[i].avt_variation == variation)
			return (avtp[i].avt_model);
	return (NULL);
}

/*
 * Generate a default platform name based for unknown system variations.
 */
const char *
alpha_unknown_sysname()
{
	static char s[128];		/* safe size */

	sprintf(s, "%s family, unknown model variation 0x%lx",
	    platform.family, hwrpb->rpb_variation & SV_ST_MASK);
	return ((const char *)s);
}

void
identifycpu()
{

	/*
	 * print out CPU identification information.
	 */
	printf("%s, %ldMHz\n", cpu_model,
	    hwrpb->rpb_cc_freq / 1000000);	/* XXX true for 21164? */
	printf("%ld byte page size, %d processor%s.\n",
	    hwrpb->rpb_page_size, ncpus, ncpus == 1 ? "" : "s");
#if 0
	/* this isn't defined for any systems that we run on? */
	printf("serial number 0x%lx 0x%lx\n",
	    ((long *)hwrpb->rpb_ssn)[0], ((long *)hwrpb->rpb_ssn)[1]);

	/* and these aren't particularly useful! */
	printf("variation: 0x%lx, revision 0x%lx\n",
	    hwrpb->rpb_variation, *(long *)hwrpb->rpb_revision);
#endif
}

int	waittime = -1;
struct pcb dumppcb;

void
cpu_reboot(howto, bootstr)
	int howto;
	char *bootstr;
{
	extern int cold;

	/* If system is cold, just halt. */
	if (cold) {
		howto |= RB_HALT;
		goto haltsys;
	}

	/* If "always halt" was specified as a boot flag, obey. */
	if ((boothowto & RB_HALT) != 0)
		howto |= RB_HALT;

	boothowto = howto;
	if ((howto & RB_NOSYNC) == 0 && waittime < 0) {
		waittime = 0;
		vfs_shutdown();
		/*
		 * If we've been adjusting the clock, the todr
		 * will be out of synch; adjust it now.
		 */
		resettodr();
	}

	/* Disable interrupts. */
	splhigh();

	/* If rebooting and a dump is requested do it. */
#if 0
	if ((howto & (RB_DUMP | RB_HALT)) == RB_DUMP)
#else
	if (howto & RB_DUMP)
#endif
		dumpsys();

haltsys:

	/* run any shutdown hooks */
	doshutdownhooks();

#ifdef BOOTKEY
	printf("hit any key to %s...\n", howto & RB_HALT ? "halt" : "reboot");
	cngetc();
	printf("\n");
#endif

	/* Finally, halt/reboot the system. */
	printf("%s\n\n", howto & RB_HALT ? "halted." : "rebooting...");
	prom_halt(howto & RB_HALT);
	/*NOTREACHED*/
}

/*
 * These variables are needed by /sbin/savecore
 */
u_long	dumpmag = 0x8fca0101;	/* magic number */
int 	dumpsize = 0;		/* pages */
long	dumplo = 0; 		/* blocks */

/*
 * cpu_dumpsize: calculate size of machine-dependent kernel core dump headers.
 */
int
cpu_dumpsize()
{
	int size;

	size = ALIGN(sizeof(kcore_seg_t)) + ALIGN(sizeof(cpu_kcore_hdr_t)) +
	    ALIGN(mem_cluster_cnt * sizeof(phys_ram_seg_t));
	if (roundup(size, dbtob(1)) != dbtob(1))
		return -1;

	return (1);
}

/*
 * cpu_dump_mempagecnt: calculate size of RAM (in pages) to be dumped.
 */
u_long
cpu_dump_mempagecnt()
{
	u_long i, n;

	n = 0;
	for (i = 0; i < mem_cluster_cnt; i++)
		n += atop(mem_clusters[i].size);
	return (n);
}

/*
 * cpu_dump: dump machine-dependent kernel core dump headers.
 */
int
cpu_dump()
{
	int (*dump) __P((dev_t, daddr_t, caddr_t, size_t));
	char buf[dbtob(1)];
	kcore_seg_t *segp;
	cpu_kcore_hdr_t *cpuhdrp;
	phys_ram_seg_t *memsegp;
	int i;

	dump = bdevsw[major(dumpdev)].d_dump;

	bzero(buf, sizeof buf);
	segp = (kcore_seg_t *)buf;
	cpuhdrp = (cpu_kcore_hdr_t *)&buf[ALIGN(sizeof(*segp))];
	memsegp = (phys_ram_seg_t *)&buf[ ALIGN(sizeof(*segp)) +
	    ALIGN(sizeof(*cpuhdrp))];

	/*
	 * Generate a segment header.
	 */
	CORE_SETMAGIC(*segp, KCORE_MAGIC, MID_MACHINE, CORE_CPU);
	segp->c_size = dbtob(1) - ALIGN(sizeof(*segp));

	/*
	 * Add the machine-dependent header info.
	 */
	cpuhdrp->lev1map_pa = ALPHA_K0SEG_TO_PHYS((vm_offset_t)Lev1map);
	cpuhdrp->page_size = PAGE_SIZE;
	cpuhdrp->nmemsegs = mem_cluster_cnt;

	/*
	 * Fill in the memory segment descriptors.
	 */
	for (i = 0; i < mem_cluster_cnt; i++) {
		memsegp[i].start = mem_clusters[i].start;
		memsegp[i].size = mem_clusters[i].size & ~PAGE_MASK;
	}

	return (dump(dumpdev, dumplo, (caddr_t)buf, dbtob(1)));
}

/*
 * This is called by main to set dumplo and dumpsize.
 * Dumps always skip the first CLBYTES of disk space
 * in case there might be a disk label stored there.
 * If there is extra space, put dump at the end to
 * reduce the chance that swapping trashes it.
 */
void
cpu_dumpconf()
{
	int nblks, dumpblks;	/* size of dump area */
	int maj;

	if (dumpdev == NODEV)
		goto bad;
	maj = major(dumpdev);
	if (maj < 0 || maj >= nblkdev)
		panic("dumpconf: bad dumpdev=0x%x", dumpdev);
	if (bdevsw[maj].d_psize == NULL)
		goto bad;
	nblks = (*bdevsw[maj].d_psize)(dumpdev);
	if (nblks <= ctod(1))
		goto bad;

	dumpblks = cpu_dumpsize();
	if (dumpblks < 0)
		goto bad;
	dumpblks += ctod(cpu_dump_mempagecnt());

	/* If dump won't fit (incl. room for possible label), punt. */
	if (dumpblks > (nblks - ctod(1)))
		goto bad;

	/* Put dump at end of partition */
	dumplo = nblks - dumpblks;

	/* dumpsize is in page units, and doesn't include headers. */
	dumpsize = cpu_dump_mempagecnt();
	return;

bad:
	dumpsize = 0;
	return;
}

/*
 * Dump the kernel's image to the swap partition.
 */
#define	BYTES_PER_DUMP	NBPG

void
dumpsys()
{
	u_long totalbytesleft, bytes, i, n, memcl;
	u_long maddr;
	int psize;
	daddr_t blkno;
	int (*dump) __P((dev_t, daddr_t, caddr_t, size_t));
	int error;

	/* Save registers. */
	savectx(&dumppcb);

	msgbufenabled = 0;	/* don't record dump msgs in msgbuf */
	if (dumpdev == NODEV)
		return;

	/*
	 * For dumps during autoconfiguration,
	 * if dump device has already configured...
	 */
	if (dumpsize == 0)
		cpu_dumpconf();
	if (dumplo <= 0) {
		printf("\ndump to dev %u,%u not possible\n", major(dumpdev),
		    minor(dumpdev));
		return;
	}
	printf("\ndumping to dev %u,%u offset %ld\n", major(dumpdev),
	    minor(dumpdev), dumplo);

	psize = (*bdevsw[major(dumpdev)].d_psize)(dumpdev);
	printf("dump ");
	if (psize == -1) {
		printf("area unavailable\n");
		return;
	}

	/* XXX should purge all outstanding keystrokes. */

	if ((error = cpu_dump()) != 0)
		goto err;

	totalbytesleft = ptoa(cpu_dump_mempagecnt());
	blkno = dumplo + cpu_dumpsize();
	dump = bdevsw[major(dumpdev)].d_dump;
	error = 0;

	for (memcl = 0; memcl < mem_cluster_cnt; memcl++) {
		maddr = mem_clusters[memcl].start;
		bytes = mem_clusters[memcl].size & ~PAGE_MASK;

		for (i = 0; i < bytes; i += n, totalbytesleft -= n) {

			/* Print out how many MBs we to go. */
			if ((totalbytesleft % (1024*1024)) == 0)
				printf("%d ", totalbytesleft / (1024 * 1024));

			/* Limit size for next transfer. */
			n = bytes - i;
			if (n > BYTES_PER_DUMP)
				n =  BYTES_PER_DUMP;

			error = (*dump)(dumpdev, blkno,
			    (caddr_t)ALPHA_PHYS_TO_K0SEG(maddr), n);
			if (error)
				goto err;
			maddr += n;
			blkno += btodb(n);			/* XXX? */

			/* XXX should look for keystrokes, to cancel. */
		}
	}

err:
	switch (error) {

	case ENXIO:
		printf("device bad\n");
		break;

	case EFAULT:
		printf("device not ready\n");
		break;

	case EINVAL:
		printf("area improper\n");
		break;

	case EIO:
		printf("i/o error\n");
		break;

	case EINTR:
		printf("aborted from console\n");
		break;

	case 0:
		printf("succeeded\n");
		break;

	default:
		printf("error %d\n", error);
		break;
	}
	printf("\n\n");
	delay(1000);
}

void
frametoreg(framep, regp)
	struct trapframe *framep;
	struct reg *regp;
{

	regp->r_regs[R_V0] = framep->tf_regs[FRAME_V0];
	regp->r_regs[R_T0] = framep->tf_regs[FRAME_T0];
	regp->r_regs[R_T1] = framep->tf_regs[FRAME_T1];
	regp->r_regs[R_T2] = framep->tf_regs[FRAME_T2];
	regp->r_regs[R_T3] = framep->tf_regs[FRAME_T3];
	regp->r_regs[R_T4] = framep->tf_regs[FRAME_T4];
	regp->r_regs[R_T5] = framep->tf_regs[FRAME_T5];
	regp->r_regs[R_T6] = framep->tf_regs[FRAME_T6];
	regp->r_regs[R_T7] = framep->tf_regs[FRAME_T7];
	regp->r_regs[R_S0] = framep->tf_regs[FRAME_S0];
	regp->r_regs[R_S1] = framep->tf_regs[FRAME_S1];
	regp->r_regs[R_S2] = framep->tf_regs[FRAME_S2];
	regp->r_regs[R_S3] = framep->tf_regs[FRAME_S3];
	regp->r_regs[R_S4] = framep->tf_regs[FRAME_S4];
	regp->r_regs[R_S5] = framep->tf_regs[FRAME_S5];
	regp->r_regs[R_S6] = framep->tf_regs[FRAME_S6];
	regp->r_regs[R_A0] = framep->tf_regs[FRAME_A0];
	regp->r_regs[R_A1] = framep->tf_regs[FRAME_A1];
	regp->r_regs[R_A2] = framep->tf_regs[FRAME_A2];
	regp->r_regs[R_A3] = framep->tf_regs[FRAME_A3];
	regp->r_regs[R_A4] = framep->tf_regs[FRAME_A4];
	regp->r_regs[R_A5] = framep->tf_regs[FRAME_A5];
	regp->r_regs[R_T8] = framep->tf_regs[FRAME_T8];
	regp->r_regs[R_T9] = framep->tf_regs[FRAME_T9];
	regp->r_regs[R_T10] = framep->tf_regs[FRAME_T10];
	regp->r_regs[R_T11] = framep->tf_regs[FRAME_T11];
	regp->r_regs[R_RA] = framep->tf_regs[FRAME_RA];
	regp->r_regs[R_T12] = framep->tf_regs[FRAME_T12];
	regp->r_regs[R_AT] = framep->tf_regs[FRAME_AT];
	regp->r_regs[R_GP] = framep->tf_regs[FRAME_GP];
	/* regp->r_regs[R_SP] = framep->tf_regs[FRAME_SP]; XXX */
	regp->r_regs[R_ZERO] = 0;
}

void
regtoframe(regp, framep)
	struct reg *regp;
	struct trapframe *framep;
{

	framep->tf_regs[FRAME_V0] = regp->r_regs[R_V0];
	framep->tf_regs[FRAME_T0] = regp->r_regs[R_T0];
	framep->tf_regs[FRAME_T1] = regp->r_regs[R_T1];
	framep->tf_regs[FRAME_T2] = regp->r_regs[R_T2];
	framep->tf_regs[FRAME_T3] = regp->r_regs[R_T3];
	framep->tf_regs[FRAME_T4] = regp->r_regs[R_T4];
	framep->tf_regs[FRAME_T5] = regp->r_regs[R_T5];
	framep->tf_regs[FRAME_T6] = regp->r_regs[R_T6];
	framep->tf_regs[FRAME_T7] = regp->r_regs[R_T7];
	framep->tf_regs[FRAME_S0] = regp->r_regs[R_S0];
	framep->tf_regs[FRAME_S1] = regp->r_regs[R_S1];
	framep->tf_regs[FRAME_S2] = regp->r_regs[R_S2];
	framep->tf_regs[FRAME_S3] = regp->r_regs[R_S3];
	framep->tf_regs[FRAME_S4] = regp->r_regs[R_S4];
	framep->tf_regs[FRAME_S5] = regp->r_regs[R_S5];
	framep->tf_regs[FRAME_S6] = regp->r_regs[R_S6];
	framep->tf_regs[FRAME_A0] = regp->r_regs[R_A0];
	framep->tf_regs[FRAME_A1] = regp->r_regs[R_A1];
	framep->tf_regs[FRAME_A2] = regp->r_regs[R_A2];
	framep->tf_regs[FRAME_A3] = regp->r_regs[R_A3];
	framep->tf_regs[FRAME_A4] = regp->r_regs[R_A4];
	framep->tf_regs[FRAME_A5] = regp->r_regs[R_A5];
	framep->tf_regs[FRAME_T8] = regp->r_regs[R_T8];
	framep->tf_regs[FRAME_T9] = regp->r_regs[R_T9];
	framep->tf_regs[FRAME_T10] = regp->r_regs[R_T10];
	framep->tf_regs[FRAME_T11] = regp->r_regs[R_T11];
	framep->tf_regs[FRAME_RA] = regp->r_regs[R_RA];
	framep->tf_regs[FRAME_T12] = regp->r_regs[R_T12];
	framep->tf_regs[FRAME_AT] = regp->r_regs[R_AT];
	framep->tf_regs[FRAME_GP] = regp->r_regs[R_GP];
	/* framep->tf_regs[FRAME_SP] = regp->r_regs[R_SP]; XXX */
	/* ??? = regp->r_regs[R_ZERO]; */
}

void
printregs(regp)
	struct reg *regp;
{
	int i;

	for (i = 0; i < 32; i++)
		printf("R%d:\t0x%016lx%s", i, regp->r_regs[i],
		   i & 1 ? "\n" : "\t");
}

void
regdump(framep)
	struct trapframe *framep;
{
	struct reg reg;

	frametoreg(framep, &reg);
	reg.r_regs[R_SP] = alpha_pal_rdusp();

	printf("REGISTERS:\n");
	printregs(&reg);
}

#ifdef DEBUG
int sigdebug = 0;
int sigpid = 0;
#define	SDB_FOLLOW	0x01
#define	SDB_KSTACK	0x02
#endif

/*
 * Send an interrupt to process.
 */
void
sendsig(catcher, sig, mask, code)
	sig_t catcher;
	int sig, mask;
	u_long code;
{
	struct proc *p = curproc;
	struct sigcontext *scp, ksc;
	struct trapframe *frame;
	struct sigacts *psp = p->p_sigacts;
	int oonstack, fsize, rndfsize;
	extern char sigcode[], esigcode[];
	extern struct proc *fpcurproc;

	frame = p->p_md.md_tf;
	oonstack = psp->ps_sigstk.ss_flags & SS_ONSTACK;
	fsize = sizeof ksc;
	rndfsize = ((fsize + 15) / 16) * 16;
	/*
	 * Allocate and validate space for the signal handler
	 * context. Note that if the stack is in P0 space, the
	 * call to grow() is a nop, and the useracc() check
	 * will fail if the process has not already allocated
	 * the space with a `brk'.
	 */
	if ((psp->ps_flags & SAS_ALTSTACK) && !oonstack &&
	    (psp->ps_sigonstack & sigmask(sig))) {
		scp = (struct sigcontext *)(psp->ps_sigstk.ss_sp +
		    psp->ps_sigstk.ss_size - rndfsize);
		psp->ps_sigstk.ss_flags |= SS_ONSTACK;
	} else
		scp = (struct sigcontext *)(alpha_pal_rdusp() - rndfsize);
	if ((u_long)scp <= USRSTACK - ctob(p->p_vmspace->vm_ssize))
#if defined(UVM)
		(void)uvm_grow(p, (u_long)scp);
#else
		(void)grow(p, (u_long)scp);
#endif
#ifdef DEBUG
	if ((sigdebug & SDB_KSTACK) && p->p_pid == sigpid)
		printf("sendsig(%d): sig %d ssp %p usp %p\n", p->p_pid,
		    sig, &oonstack, scp);
#endif
#if defined(UVM)
	if (uvm_useracc((caddr_t)scp, fsize, B_WRITE) == 0) {
#else
	if (useracc((caddr_t)scp, fsize, B_WRITE) == 0) {
#endif
#ifdef DEBUG
		if ((sigdebug & SDB_KSTACK) && p->p_pid == sigpid)
			printf("sendsig(%d): useracc failed on sig %d\n",
			    p->p_pid, sig);
#endif
		/*
		 * Process has trashed its stack; give it an illegal
		 * instruction to halt it in its tracks.
		 */
		SIGACTION(p, SIGILL) = SIG_DFL;
		sig = sigmask(SIGILL);
		p->p_sigignore &= ~sig;
		p->p_sigcatch &= ~sig;
		p->p_sigmask &= ~sig;
		psignal(p, SIGILL);
		return;
	}

	/*
	 * Build the signal context to be used by sigreturn.
	 */
	ksc.sc_onstack = oonstack;
	ksc.sc_mask = mask;
	ksc.sc_pc = frame->tf_regs[FRAME_PC];
	ksc.sc_ps = frame->tf_regs[FRAME_PS];

	/* copy the registers. */
	frametoreg(frame, (struct reg *)ksc.sc_regs);
	ksc.sc_regs[R_ZERO] = 0xACEDBADE;		/* magic number */
	ksc.sc_regs[R_SP] = alpha_pal_rdusp();

	/* save the floating-point state, if necessary, then copy it. */
	if (p == fpcurproc) {
		alpha_pal_wrfen(1);
		savefpstate(&p->p_addr->u_pcb.pcb_fp);
		alpha_pal_wrfen(0);
		fpcurproc = NULL;
	}
	ksc.sc_ownedfp = p->p_md.md_flags & MDP_FPUSED;
	bcopy(&p->p_addr->u_pcb.pcb_fp, (struct fpreg *)ksc.sc_fpregs,
	    sizeof(struct fpreg));
	ksc.sc_fp_control = 0;					/* XXX ? */
	bzero(ksc.sc_reserved, sizeof ksc.sc_reserved);		/* XXX */
	bzero(ksc.sc_xxx, sizeof ksc.sc_xxx);			/* XXX */


#ifdef COMPAT_OSF1
	/*
	 * XXX Create an OSF/1-style sigcontext and associated goo.
	 */
#endif

	/*
	 * copy the frame out to userland.
	 */
	(void) copyout((caddr_t)&ksc, (caddr_t)scp, fsize);
#ifdef DEBUG
	if (sigdebug & SDB_FOLLOW)
		printf("sendsig(%d): sig %d scp %p code %lx\n", p->p_pid, sig,
		    scp, code);
#endif

	/*
	 * Set up the registers to return to sigcode.
	 */
	frame->tf_regs[FRAME_PC] =
	    (u_int64_t)PS_STRINGS - (esigcode - sigcode);
	frame->tf_regs[FRAME_A0] = sig;
	frame->tf_regs[FRAME_A1] = code;
	frame->tf_regs[FRAME_A2] = (u_int64_t)scp;
	frame->tf_regs[FRAME_T12] = (u_int64_t)catcher;		/* t12 is pv */
	alpha_pal_wrusp((unsigned long)scp);

#ifdef DEBUG
	if (sigdebug & SDB_FOLLOW)
		printf("sendsig(%d): pc %lx, catcher %lx\n", p->p_pid,
		    frame->tf_regs[FRAME_PC], frame->tf_regs[FRAME_A3]);
	if ((sigdebug & SDB_KSTACK) && p->p_pid == sigpid)
		printf("sendsig(%d): sig %d returns\n",
		    p->p_pid, sig);
#endif
}

/*
 * System call to cleanup state after a signal
 * has been taken.  Reset signal mask and
 * stack state from context left by sendsig (above).
 * Return to previous pc and psl as specified by
 * context left by sendsig. Check carefully to
 * make sure that the user has not modified the
 * psl to gain improper priviledges or to cause
 * a machine fault.
 */
/* ARGSUSED */
int
sys_sigreturn(p, v, retval)
	struct proc *p;
	void *v;
	register_t *retval;
{
	struct sys_sigreturn_args /* {
		syscallarg(struct sigcontext *) sigcntxp;
	} */ *uap = v;
	struct sigcontext *scp, ksc;
	extern struct proc *fpcurproc;

	scp = SCARG(uap, sigcntxp);
#ifdef DEBUG
	if (sigdebug & SDB_FOLLOW)
	    printf("sigreturn: pid %d, scp %p\n", p->p_pid, scp);
#endif

	if (ALIGN(scp) != (u_int64_t)scp)
		return (EINVAL);

	/*
	 * Test and fetch the context structure.
	 * We grab it all at once for speed.
	 */
#if defined(UVM)
	if (uvm_useracc((caddr_t)scp, sizeof (*scp), B_WRITE) == 0 ||
	    copyin((caddr_t)scp, (caddr_t)&ksc, sizeof ksc))
		return (EINVAL);
#else
	if (useracc((caddr_t)scp, sizeof (*scp), B_WRITE) == 0 ||
	    copyin((caddr_t)scp, (caddr_t)&ksc, sizeof ksc))
		return (EINVAL);
#endif

	if (ksc.sc_regs[R_ZERO] != 0xACEDBADE)		/* magic number */
		return (EINVAL);
	/*
	 * Restore the user-supplied information
	 */
	if (ksc.sc_onstack)
		p->p_sigacts->ps_sigstk.ss_flags |= SS_ONSTACK;
	else
		p->p_sigacts->ps_sigstk.ss_flags &= ~SS_ONSTACK;
	p->p_sigmask = ksc.sc_mask &~ sigcantmask;

	p->p_md.md_tf->tf_regs[FRAME_PC] = ksc.sc_pc;
	p->p_md.md_tf->tf_regs[FRAME_PS] =
	    (ksc.sc_ps | ALPHA_PSL_USERSET) & ~ALPHA_PSL_USERCLR;

	regtoframe((struct reg *)ksc.sc_regs, p->p_md.md_tf);
	alpha_pal_wrusp(ksc.sc_regs[R_SP]);

	/* XXX ksc.sc_ownedfp ? */
	if (p == fpcurproc)
		fpcurproc = NULL;
	bcopy((struct fpreg *)ksc.sc_fpregs, &p->p_addr->u_pcb.pcb_fp,
	    sizeof(struct fpreg));
	/* XXX ksc.sc_fp_control ? */

#ifdef DEBUG
	if (sigdebug & SDB_FOLLOW)
		printf("sigreturn(%d): returns\n", p->p_pid);
#endif
	return (EJUSTRETURN);
}

/*
 * machine dependent system variables.
 */
int
cpu_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p)
	int *name;
	u_int namelen;
	void *oldp;
	size_t *oldlenp;
	void *newp;
	size_t newlen;
	struct proc *p;
{
	dev_t consdev;

	/* all sysctl names at this level are terminal */
	if (namelen != 1)
		return (ENOTDIR);		/* overloaded */

	switch (name[0]) {
	case CPU_CONSDEV:
		if (cn_tab != NULL)
			consdev = cn_tab->cn_dev;
		else
			consdev = NODEV;
		return (sysctl_rdstruct(oldp, oldlenp, newp, &consdev,
			sizeof consdev));

	case CPU_ROOT_DEVICE:
		return (sysctl_rdstring(oldp, oldlenp, newp,
		    root_device->dv_xname));

	case CPU_UNALIGNED_PRINT:
		return (sysctl_int(oldp, oldlenp, newp, newlen,
		    &alpha_unaligned_print));

	case CPU_UNALIGNED_FIX:
		return (sysctl_int(oldp, oldlenp, newp, newlen,
		    &alpha_unaligned_fix));

	case CPU_UNALIGNED_SIGBUS:
		return (sysctl_int(oldp, oldlenp, newp, newlen,
		    &alpha_unaligned_sigbus));

	case CPU_BOOTED_KERNEL:
		return (sysctl_rdstring(oldp, oldlenp, newp,
		    bootinfo.booted_kernel));

	default:
		return (EOPNOTSUPP);
	}
	/* NOTREACHED */
}

/*
 * Set registers on exec.
 */
void
setregs(p, pack, stack)
	register struct proc *p;
	struct exec_package *pack;
	u_long stack;
{
	struct trapframe *tfp = p->p_md.md_tf;
	extern struct proc *fpcurproc;
#ifdef DEBUG
	int i;
#endif

#ifdef DEBUG
	/*
	 * Crash and dump, if the user requested it.
	 */
	if (boothowto & RB_DUMP)
		panic("crash requested by boot flags");
#endif

#ifdef DEBUG
	for (i = 0; i < FRAME_SIZE; i++)
		tfp->tf_regs[i] = 0xbabefacedeadbeef;
#else
	bzero(tfp->tf_regs, FRAME_SIZE * sizeof tfp->tf_regs[0]);
#endif
	bzero(&p->p_addr->u_pcb.pcb_fp, sizeof p->p_addr->u_pcb.pcb_fp);
#define FP_RN 2 /* XXX */
	p->p_addr->u_pcb.pcb_fp.fpr_cr = (long)FP_RN << 58;
	alpha_pal_wrusp(stack);
	tfp->tf_regs[FRAME_PS] = ALPHA_PSL_USERSET;
	tfp->tf_regs[FRAME_PC] = pack->ep_entry & ~3;

	tfp->tf_regs[FRAME_A0] = stack;			/* a0 = sp */
	tfp->tf_regs[FRAME_A1] = 0;			/* a1 = rtld cleanup */
	tfp->tf_regs[FRAME_A2] = 0;			/* a2 = rtld object */
	tfp->tf_regs[FRAME_A3] = (u_int64_t)PS_STRINGS;	/* a3 = ps_strings */
	tfp->tf_regs[FRAME_T12] = tfp->tf_regs[FRAME_PC];	/* a.k.a. PV */

	p->p_md.md_flags &= ~MDP_FPUSED;
	if (fpcurproc == p)
		fpcurproc = NULL;
}

void
netintr()
{
	int n, s;

	s = splhigh();
	n = netisr;
	netisr = 0;
	splx(s);

#define	DONETISR(bit, fn)						\
	do {								\
		if (n & (1 << (bit)))					\
			fn;						\
	} while (0)

#ifdef INET
#if NARP > 0
	DONETISR(NETISR_ARP, arpintr());
#endif
	DONETISR(NETISR_IP, ipintr());
#endif
#ifdef NETATALK
	DONETISR(NETISR_ATALK, atintr());
#endif
#ifdef NS
	DONETISR(NETISR_NS, nsintr());
#endif
#ifdef ISO
	DONETISR(NETISR_ISO, clnlintr());
#endif
#ifdef CCITT
	DONETISR(NETISR_CCITT, ccittintr());
#endif
#ifdef NATM
	DONETISR(NETISR_NATM, natmintr());
#endif
#if NPPP > 1
	DONETISR(NETISR_PPP, pppintr());
#endif

#undef DONETISR
}

void
do_sir()
{
	u_int64_t n;

	do {
		(void)splhigh();
		n = ssir;
		ssir = 0;
		splsoft();		/* don't recurse through spl0() */

#if defined(UVM)
#define	COUNT_SOFT	uvmexp.softs++
#else
#define	COUNT_SOFT	cnt.v_soft++
#endif

#define	DO_SIR(bit, fn)							\
		do {							\
			if (n & (bit)) {				\
				COUNT_SOFT;				\
				fn;					\
			}						\
		} while (0)

		DO_SIR(SIR_NET, netintr());
		DO_SIR(SIR_CLOCK, softclock());

#undef COUNT_SOFT
#undef DO_SIR
	} while (ssir != 0);
}

int
spl0()
{

	if (ssir)
		do_sir();		/* it lowers the IPL itself */

	return (alpha_pal_swpipl(ALPHA_PSL_IPL_0));
}

/*
 * The following primitives manipulate the run queues.  _whichqs tells which
 * of the 32 queues _qs have processes in them.  Setrunqueue puts processes
 * into queues, Remrunqueue removes them from queues.  The running process is
 * on no queue, other processes are on a queue related to p->p_priority,
 * divided by 4 actually to shrink the 0-127 range of priorities into the 32
 * available queues.
 */
/*
 * setrunqueue(p)
 *	proc *p;
 *
 * Call should be made at splclock(), and p->p_stat should be SRUN.
 */

void
setrunqueue(p)
	struct proc *p;
{
	int bit;

	/* firewall: p->p_back must be NULL */
	if (p->p_back != NULL)
		panic("setrunqueue");

	bit = p->p_priority >> 2;
	whichqs |= (1 << bit);
	p->p_forw = (struct proc *)&qs[bit];
	p->p_back = qs[bit].ph_rlink;
	p->p_back->p_forw = p;
	qs[bit].ph_rlink = p;
}

/*
 * remrunqueue(p)
 *
 * Call should be made at splclock().
 */
void
remrunqueue(p)
	struct proc *p;
{
	int bit;

	bit = p->p_priority >> 2;
	if ((whichqs & (1 << bit)) == 0)
		panic("remrunqueue");

	p->p_back->p_forw = p->p_forw;
	p->p_forw->p_back = p->p_back;
	p->p_back = NULL;	/* for firewall checking. */

	if ((struct proc *)&qs[bit] == qs[bit].ph_link)
		whichqs &= ~(1 << bit);
}

/*
 * Return the best possible estimate of the time in the timeval
 * to which tvp points.  Unfortunately, we can't read the hardware registers.
 * We guarantee that the time will be greater than the value obtained by a
 * previous call.
 */
void
microtime(tvp)
	register struct timeval *tvp;
{
	int s = splclock();
	static struct timeval lasttime;

	*tvp = time;
#ifdef notdef
	tvp->tv_usec += clkread();
	while (tvp->tv_usec > 1000000) {
		tvp->tv_sec++;
		tvp->tv_usec -= 1000000;
	}
#endif
	if (tvp->tv_sec == lasttime.tv_sec &&
	    tvp->tv_usec <= lasttime.tv_usec &&
	    (tvp->tv_usec = lasttime.tv_usec + 1) > 1000000) {
		tvp->tv_sec++;
		tvp->tv_usec -= 1000000;
	}
	lasttime = *tvp;
	splx(s);
}

/*
 * Wait "n" microseconds.
 */
void
delay(n)
	unsigned long n;
{
	long N = cycles_per_usec * (n);

	while (N > 0)				/* XXX */
		N -= 3;				/* XXX */
}

#if defined(COMPAT_OSF1) || 1		/* XXX */
void	cpu_exec_ecoff_setregs __P((struct proc *, struct exec_package *,
	    u_long));

void
cpu_exec_ecoff_setregs(p, epp, stack)
	struct proc *p;
	struct exec_package *epp;
	u_long stack;
{
	struct ecoff_exechdr *execp = (struct ecoff_exechdr *)epp->ep_hdr;

	setregs(p, epp, stack);
	p->p_md.md_tf->tf_regs[FRAME_GP] = execp->a.gp_value;
}

/*
 * cpu_exec_ecoff_hook():
 *	cpu-dependent ECOFF format hook for execve().
 *
 * Do any machine-dependent diddling of the exec package when doing ECOFF.
 *
 */
int
cpu_exec_ecoff_hook(p, epp)
	struct proc *p;
	struct exec_package *epp;
{
	struct ecoff_exechdr *execp = (struct ecoff_exechdr *)epp->ep_hdr;
	extern struct emul emul_netbsd;
#ifdef COMPAT_OSF1
	extern struct emul emul_osf1;
#endif

	switch (execp->f.f_magic) {
#ifdef COMPAT_OSF1
	case ECOFF_MAGIC_ALPHA:
		epp->ep_emul = &emul_osf1;
		break;
#endif

	case ECOFF_MAGIC_NETBSD_ALPHA:
		epp->ep_emul = &emul_netbsd;
		break;

	default:
		return ENOEXEC;
	}
	return 0;
}
#endif

int
alpha_pa_access(pa)
	u_long pa;
{
	int i;

	for (i = 0; i < mem_cluster_cnt; i++) {
		if (pa < mem_clusters[i].start)
			continue;
		if ((pa - mem_clusters[i].start) >=
		    (mem_clusters[i].size & ~PAGE_MASK))
			continue;
		return (mem_clusters[i].size & PAGE_MASK);	/* prot */
	}
	return (PROT_NONE);
}

/* XXX XXX BEGIN XXX XXX */
vm_offset_t alpha_XXX_dmamap_or;				/* XXX */
								/* XXX */
vm_offset_t							/* XXX */
alpha_XXX_dmamap(v)						/* XXX */
	vm_offset_t v;						/* XXX */
{								/* XXX */
								/* XXX */
	return (vtophys(v) | alpha_XXX_dmamap_or);		/* XXX */
}								/* XXX */
/* XXX XXX END XXX XXX */