2005-01-07 02:29:27 +00:00
|
|
|
/*-
|
2001-06-10 02:39:37 +00:00
|
|
|
* Copyright (C) 1995, 1996 Wolfgang Solfrank.
|
|
|
|
* Copyright (C) 1995, 1996 TooLs GmbH.
|
|
|
|
* All rights reserved.
|
|
|
|
*
|
|
|
|
* 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 TooLs GmbH.
|
|
|
|
* 4. The name of TooLs GmbH may not be used to endorse or promote products
|
|
|
|
* derived from this software without specific prior written permission.
|
|
|
|
*
|
|
|
|
* THIS SOFTWARE IS PROVIDED BY TOOLS GMBH ``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 TOOLS GMBH 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.
|
|
|
|
*/
|
2005-01-07 02:29:27 +00:00
|
|
|
/*-
|
2001-06-10 02:39:37 +00:00
|
|
|
* Copyright (C) 2001 Benno Rice
|
|
|
|
* All rights reserved.
|
|
|
|
*
|
|
|
|
* 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.
|
|
|
|
*
|
|
|
|
* THIS SOFTWARE IS PROVIDED BY Benno Rice ``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 TOOLS GMBH 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.
|
|
|
|
* $NetBSD: machdep.c,v 1.74.2.1 2000/11/01 16:13:48 tv Exp $
|
|
|
|
*/
|
|
|
|
|
2003-04-03 21:36:33 +00:00
|
|
|
#include <sys/cdefs.h>
|
|
|
|
__FBSDID("$FreeBSD$");
|
2001-06-10 02:39:37 +00:00
|
|
|
|
|
|
|
#include "opt_compat.h"
|
2004-07-27 03:41:34 +00:00
|
|
|
#include "opt_ddb.h"
|
2003-07-31 01:31:32 +00:00
|
|
|
#include "opt_kstack_pages.h"
|
2001-06-10 02:39:37 +00:00
|
|
|
#include "opt_msgbuf.h"
|
|
|
|
|
|
|
|
#include <sys/param.h>
|
2005-02-06 01:55:08 +00:00
|
|
|
#include <sys/proc.h>
|
2001-06-10 02:39:37 +00:00
|
|
|
#include <sys/systm.h>
|
2005-02-04 06:58:09 +00:00
|
|
|
#include <sys/bio.h>
|
|
|
|
#include <sys/buf.h>
|
|
|
|
#include <sys/bus.h>
|
2005-02-06 01:55:08 +00:00
|
|
|
#include <sys/cons.h>
|
2005-02-04 05:32:56 +00:00
|
|
|
#include <sys/cpu.h>
|
2001-06-10 02:39:37 +00:00
|
|
|
#include <sys/eventhandler.h>
|
2005-02-06 01:55:08 +00:00
|
|
|
#include <sys/exec.h>
|
2002-08-29 06:17:48 +00:00
|
|
|
#include <sys/imgact.h>
|
2005-02-06 01:55:08 +00:00
|
|
|
#include <sys/kdb.h>
|
2001-06-10 02:39:37 +00:00
|
|
|
#include <sys/kernel.h>
|
2005-02-06 01:55:08 +00:00
|
|
|
#include <sys/ktr.h>
|
|
|
|
#include <sys/linker.h>
|
|
|
|
#include <sys/lock.h>
|
2001-06-10 02:39:37 +00:00
|
|
|
#include <sys/malloc.h>
|
|
|
|
#include <sys/mbuf.h>
|
|
|
|
#include <sys/msgbuf.h>
|
2005-02-06 01:55:08 +00:00
|
|
|
#include <sys/mutex.h>
|
|
|
|
#include <sys/ptrace.h>
|
|
|
|
#include <sys/reboot.h>
|
|
|
|
#include <sys/signalvar.h>
|
2010-06-30 18:03:42 +00:00
|
|
|
#include <sys/syscallsubr.h>
|
2001-06-10 02:39:37 +00:00
|
|
|
#include <sys/sysctl.h>
|
2002-09-19 04:30:43 +00:00
|
|
|
#include <sys/sysent.h>
|
2005-02-06 01:55:08 +00:00
|
|
|
#include <sys/sysproto.h>
|
|
|
|
#include <sys/ucontext.h>
|
|
|
|
#include <sys/uio.h>
|
|
|
|
#include <sys/vmmeter.h>
|
|
|
|
#include <sys/vnode.h>
|
|
|
|
|
2001-06-10 02:39:37 +00:00
|
|
|
#include <net/netisr.h>
|
2005-02-06 01:55:08 +00:00
|
|
|
|
2001-06-10 02:39:37 +00:00
|
|
|
#include <vm/vm.h>
|
2005-02-06 01:55:08 +00:00
|
|
|
#include <vm/vm_extern.h>
|
2001-06-10 02:39:37 +00:00
|
|
|
#include <vm/vm_kern.h>
|
|
|
|
#include <vm/vm_page.h>
|
|
|
|
#include <vm/vm_map.h>
|
|
|
|
#include <vm/vm_object.h>
|
|
|
|
#include <vm/vm_pager.h>
|
2005-02-06 01:55:08 +00:00
|
|
|
|
2009-02-20 17:48:40 +00:00
|
|
|
#include <machine/altivec.h>
|
2010-07-13 05:32:19 +00:00
|
|
|
#ifndef __powerpc64__
|
2001-06-10 02:39:37 +00:00
|
|
|
#include <machine/bat.h>
|
2010-07-13 05:32:19 +00:00
|
|
|
#endif
|
2004-02-11 07:27:34 +00:00
|
|
|
#include <machine/cpu.h>
|
2005-02-06 01:55:08 +00:00
|
|
|
#include <machine/elf.h>
|
|
|
|
#include <machine/fpu.h>
|
2008-03-07 22:27:06 +00:00
|
|
|
#include <machine/hid.h>
|
2008-03-02 17:05:57 +00:00
|
|
|
#include <machine/kdb.h>
|
2001-06-10 02:39:37 +00:00
|
|
|
#include <machine/md_var.h>
|
2002-07-10 12:16:48 +00:00
|
|
|
#include <machine/metadata.h>
|
2005-11-08 06:48:08 +00:00
|
|
|
#include <machine/mmuvar.h>
|
2004-11-27 06:51:39 +00:00
|
|
|
#include <machine/pcb.h>
|
2005-02-06 01:55:08 +00:00
|
|
|
#include <machine/reg.h>
|
2001-06-10 02:39:37 +00:00
|
|
|
#include <machine/sigframe.h>
|
2009-04-12 03:03:55 +00:00
|
|
|
#include <machine/spr.h>
|
2005-02-06 01:55:08 +00:00
|
|
|
#include <machine/trap.h>
|
|
|
|
#include <machine/vmparam.h>
|
2001-06-10 02:39:37 +00:00
|
|
|
|
2004-07-12 22:25:09 +00:00
|
|
|
#include <ddb/ddb.h>
|
|
|
|
|
2005-02-06 01:55:08 +00:00
|
|
|
#include <dev/ofw/openfirm.h>
|
|
|
|
|
2004-07-27 03:41:34 +00:00
|
|
|
#ifdef DDB
|
|
|
|
extern vm_offset_t ksym_start, ksym_end;
|
|
|
|
#endif
|
|
|
|
|
2001-06-10 02:39:37 +00:00
|
|
|
int cold = 1;
|
2010-07-13 05:32:19 +00:00
|
|
|
#ifdef __powerpc64__
|
|
|
|
int cacheline_size = 128;
|
|
|
|
#else
|
2008-09-24 00:28:46 +00:00
|
|
|
int cacheline_size = 32;
|
2010-07-13 05:32:19 +00:00
|
|
|
#endif
|
2009-04-04 00:22:44 +00:00
|
|
|
int hw_direct_map = 1;
|
2001-06-10 02:39:37 +00:00
|
|
|
|
2008-04-27 22:33:43 +00:00
|
|
|
struct pcpu __pcpu[MAXCPU];
|
|
|
|
|
2008-02-12 20:55:51 +00:00
|
|
|
static struct trapframe frame0;
|
2002-02-14 01:39:11 +00:00
|
|
|
|
2001-06-10 02:39:37 +00:00
|
|
|
char machine[] = "powerpc";
|
|
|
|
SYSCTL_STRING(_hw, HW_MACHINE, machine, CTLFLAG_RD, machine, 0, "");
|
|
|
|
|
|
|
|
static void cpu_startup(void *);
|
2008-03-16 10:58:09 +00:00
|
|
|
SYSINIT(cpu, SI_SUB_CPU, SI_ORDER_FIRST, cpu_startup, NULL);
|
2001-06-10 02:39:37 +00:00
|
|
|
|
2008-09-24 00:28:46 +00:00
|
|
|
SYSCTL_INT(_machdep, CPU_CACHELINE, cacheline_size,
|
|
|
|
CTLFLAG_RD, &cacheline_size, 0, "");
|
|
|
|
|
2010-07-13 05:32:19 +00:00
|
|
|
uintptr_t powerpc_init(vm_offset_t, vm_offset_t, vm_offset_t, void *);
|
2001-06-10 02:39:37 +00:00
|
|
|
|
2002-09-19 04:30:43 +00:00
|
|
|
int setfault(faultbuf); /* defined in locore.S */
|
2001-06-10 02:39:37 +00:00
|
|
|
|
2002-08-30 04:04:37 +00:00
|
|
|
long Maxmem = 0;
|
2005-03-01 07:59:24 +00:00
|
|
|
long realmem = 0;
|
2001-06-10 02:39:37 +00:00
|
|
|
|
|
|
|
struct pmap ofw_pmap;
|
|
|
|
|
2010-07-13 05:32:19 +00:00
|
|
|
#ifndef __powerpc64__
|
2001-06-10 02:39:37 +00:00
|
|
|
struct bat battable[16];
|
2010-07-13 05:32:19 +00:00
|
|
|
#endif
|
2001-06-10 02:39:37 +00:00
|
|
|
|
2001-09-04 08:42:35 +00:00
|
|
|
struct kva_md_info kmi;
|
2001-06-10 02:39:37 +00:00
|
|
|
|
|
|
|
static void
|
|
|
|
cpu_startup(void *dummy)
|
|
|
|
{
|
|
|
|
|
2007-11-13 15:47:55 +00:00
|
|
|
/*
|
|
|
|
* Initialise the decrementer-based clock.
|
|
|
|
*/
|
|
|
|
decr_init();
|
|
|
|
|
2001-06-10 02:39:37 +00:00
|
|
|
/*
|
|
|
|
* Good {morning,afternoon,evening,night}.
|
|
|
|
*/
|
2003-02-05 12:10:46 +00:00
|
|
|
cpu_setup(PCPU_GET(cpuid));
|
2001-06-10 02:39:37 +00:00
|
|
|
|
|
|
|
#ifdef PERFMON
|
|
|
|
perfmon_init();
|
|
|
|
#endif
|
2005-03-07 01:52:24 +00:00
|
|
|
printf("real memory = %ld (%ld MB)\n", ptoa(physmem),
|
|
|
|
ptoa(physmem) / 1048576);
|
|
|
|
realmem = physmem;
|
2001-06-10 02:39:37 +00:00
|
|
|
|
2010-02-20 16:23:29 +00:00
|
|
|
if (bootverbose)
|
|
|
|
printf("available KVA = %zd (%zd MB)\n",
|
|
|
|
virtual_end - virtual_avail,
|
|
|
|
(virtual_end - virtual_avail) / 1048576);
|
|
|
|
|
2001-06-10 02:39:37 +00:00
|
|
|
/*
|
|
|
|
* Display any holes after the first chunk of extended memory.
|
|
|
|
*/
|
|
|
|
if (bootverbose) {
|
|
|
|
int indx;
|
|
|
|
|
|
|
|
printf("Physical memory chunk(s):\n");
|
|
|
|
for (indx = 0; phys_avail[indx + 1] != 0; indx += 2) {
|
2010-07-13 05:32:19 +00:00
|
|
|
vm_offset_t size1 =
|
|
|
|
phys_avail[indx + 1] - phys_avail[indx];
|
|
|
|
|
|
|
|
#ifdef __powerpc64__
|
|
|
|
printf("0x%16lx - 0x%16lx, %ld bytes (%ld pages)\n",
|
|
|
|
#else
|
|
|
|
printf("0x%08x - 0x%08x, %d bytes (%ld pages)\n",
|
|
|
|
#endif
|
2001-06-10 02:39:37 +00:00
|
|
|
phys_avail[indx], phys_avail[indx + 1] - 1, size1,
|
|
|
|
size1 / PAGE_SIZE);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2001-09-04 08:42:35 +00:00
|
|
|
vm_ksubmap_init(&kmi);
|
2001-06-10 02:39:37 +00:00
|
|
|
|
2007-05-31 22:52:15 +00:00
|
|
|
printf("avail memory = %ld (%ld MB)\n", ptoa(cnt.v_free_count),
|
|
|
|
ptoa(cnt.v_free_count) / 1048576);
|
2001-06-10 02:39:37 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Set up buffers, so they can be used to read disk labels.
|
|
|
|
*/
|
|
|
|
bufinit();
|
|
|
|
vm_pager_bufferinit();
|
|
|
|
}
|
|
|
|
|
|
|
|
extern char kernel_text[], _end[];
|
|
|
|
|
2010-07-13 05:32:19 +00:00
|
|
|
#ifndef __powerpc64__
|
|
|
|
/* Bits for running on 64-bit systems in 32-bit mode. */
|
2009-04-04 00:22:44 +00:00
|
|
|
extern void *testppc64, *testppc64size;
|
|
|
|
extern void *restorebridge, *restorebridgesize;
|
|
|
|
extern void *rfid_patch, *rfi_patch1, *rfi_patch2;
|
2010-07-13 05:32:19 +00:00
|
|
|
extern void *trapcode64;
|
|
|
|
#endif
|
|
|
|
|
2008-04-27 22:33:43 +00:00
|
|
|
#ifdef SMP
|
|
|
|
extern void *rstcode, *rstsize;
|
|
|
|
#endif
|
2010-07-13 05:32:19 +00:00
|
|
|
extern void *trapcode, *trapsize;
|
2001-06-10 02:39:37 +00:00
|
|
|
extern void *alitrap, *alisize;
|
|
|
|
extern void *dsitrap, *dsisize;
|
|
|
|
extern void *decrint, *decrsize;
|
2002-07-09 13:40:37 +00:00
|
|
|
extern void *extint, *extsize;
|
2004-07-12 22:25:09 +00:00
|
|
|
extern void *dblow, *dbsize;
|
2001-06-10 02:39:37 +00:00
|
|
|
|
2010-07-13 05:32:19 +00:00
|
|
|
uintptr_t
|
|
|
|
powerpc_init(vm_offset_t startkernel, vm_offset_t endkernel,
|
|
|
|
vm_offset_t basekernel, void *mdp)
|
2002-02-14 01:39:11 +00:00
|
|
|
{
|
|
|
|
struct pcpu *pc;
|
2008-04-16 23:28:12 +00:00
|
|
|
vm_offset_t end;
|
2009-04-04 00:22:44 +00:00
|
|
|
void *generictrap;
|
|
|
|
size_t trap_offset;
|
2002-07-10 12:16:48 +00:00
|
|
|
void *kmdp;
|
2005-02-28 07:14:13 +00:00
|
|
|
char *env;
|
2010-07-13 05:32:19 +00:00
|
|
|
register_t msr, scratch;
|
2009-04-12 03:03:55 +00:00
|
|
|
uint8_t *cache_check;
|
2010-07-13 05:32:19 +00:00
|
|
|
#ifndef __powerpc64__
|
2009-11-28 17:33:19 +00:00
|
|
|
int ppc64;
|
2010-07-13 05:32:19 +00:00
|
|
|
#endif
|
2002-07-10 12:16:48 +00:00
|
|
|
|
|
|
|
end = 0;
|
|
|
|
kmdp = NULL;
|
2009-04-04 00:22:44 +00:00
|
|
|
trap_offset = 0;
|
2002-07-10 12:16:48 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Parse metadata if present and fetch parameters. Must be done
|
|
|
|
* before console is inited so cninit gets the right value of
|
|
|
|
* boothowto.
|
|
|
|
*/
|
|
|
|
if (mdp != NULL) {
|
|
|
|
preload_metadata = mdp;
|
2003-05-01 03:33:28 +00:00
|
|
|
kmdp = preload_search_by_type("elf kernel");
|
2002-07-10 12:16:48 +00:00
|
|
|
if (kmdp != NULL) {
|
|
|
|
boothowto = MD_FETCH(kmdp, MODINFOMD_HOWTO, int);
|
|
|
|
kern_envp = MD_FETCH(kmdp, MODINFOMD_ENVP, char *);
|
|
|
|
end = MD_FETCH(kmdp, MODINFOMD_KERNEND, vm_offset_t);
|
2004-07-27 03:41:34 +00:00
|
|
|
#ifdef DDB
|
|
|
|
ksym_start = MD_FETCH(kmdp, MODINFOMD_SSYM, uintptr_t);
|
|
|
|
ksym_end = MD_FETCH(kmdp, MODINFOMD_ESYM, uintptr_t);
|
|
|
|
#endif
|
2002-07-10 12:16:48 +00:00
|
|
|
}
|
|
|
|
}
|
2002-02-14 01:39:11 +00:00
|
|
|
|
2004-03-02 06:13:09 +00:00
|
|
|
/*
|
|
|
|
* Init params/tunables that can be overridden by the loader
|
|
|
|
*/
|
|
|
|
init_param1();
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Start initializing proc0 and thread0.
|
|
|
|
*/
|
2007-11-05 11:36:16 +00:00
|
|
|
proc_linkup0(&proc0, &thread0);
|
2004-03-02 06:13:09 +00:00
|
|
|
thread0.td_frame = &frame0;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Set up per-cpu data.
|
|
|
|
*/
|
2008-04-27 22:33:43 +00:00
|
|
|
pc = __pcpu;
|
2004-03-02 06:13:09 +00:00
|
|
|
pcpu_init(pc, 0, sizeof(struct pcpu));
|
|
|
|
pc->pc_curthread = &thread0;
|
|
|
|
pc->pc_cpuid = 0;
|
|
|
|
|
|
|
|
__asm __volatile("mtsprg 0, %0" :: "r"(pc));
|
|
|
|
|
2008-12-20 00:33:10 +00:00
|
|
|
/*
|
|
|
|
* Init mutexes, which we use heavily in PMAP
|
|
|
|
*/
|
|
|
|
|
2004-03-02 06:13:09 +00:00
|
|
|
mutex_init();
|
|
|
|
|
2008-12-20 00:33:10 +00:00
|
|
|
/*
|
|
|
|
* Install the OF client interface
|
|
|
|
*/
|
|
|
|
|
|
|
|
OF_bootstrap();
|
|
|
|
|
2002-02-14 01:39:11 +00:00
|
|
|
/*
|
|
|
|
* Initialize the console before printing anything.
|
|
|
|
*/
|
|
|
|
cninit();
|
|
|
|
|
2002-07-10 12:16:48 +00:00
|
|
|
/*
|
|
|
|
* Complain if there is no metadata.
|
|
|
|
*/
|
|
|
|
if (mdp == NULL || kmdp == NULL) {
|
|
|
|
printf("powerpc_init: no loader metadata.\n");
|
|
|
|
}
|
|
|
|
|
2009-04-04 00:22:44 +00:00
|
|
|
/*
|
2009-04-12 03:03:55 +00:00
|
|
|
* Init KDB
|
2009-04-04 00:22:44 +00:00
|
|
|
*/
|
|
|
|
|
2009-04-12 03:03:55 +00:00
|
|
|
kdb_init();
|
|
|
|
|
|
|
|
/*
|
|
|
|
* PowerPC 970 CPUs have a misfeature requested by Apple that makes
|
|
|
|
* them pretend they have a 32-byte cacheline. Turn this off
|
|
|
|
* before we measure the cacheline size.
|
|
|
|
*/
|
|
|
|
|
|
|
|
switch (mfpvr() >> 16) {
|
2009-04-04 00:22:44 +00:00
|
|
|
case IBM970:
|
|
|
|
case IBM970FX:
|
|
|
|
case IBM970MP:
|
|
|
|
case IBM970GX:
|
2010-07-13 05:32:19 +00:00
|
|
|
scratch = mfspr(SPR_HID5);
|
2009-04-12 03:03:55 +00:00
|
|
|
scratch &= ~HID5_970_DCBZ_SIZE_HI;
|
2010-07-13 05:32:19 +00:00
|
|
|
mtspr(SPR_HID5, scratch);
|
2009-04-04 00:22:44 +00:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2009-04-12 03:03:55 +00:00
|
|
|
* Initialize the interrupt tables and figure out our cache line
|
|
|
|
* size and whether or not we need the 64-bit bridge code.
|
2009-04-04 00:22:44 +00:00
|
|
|
*/
|
|
|
|
|
2002-02-14 01:39:11 +00:00
|
|
|
/*
|
2009-04-12 03:03:55 +00:00
|
|
|
* Disable translation in case the vector area hasn't been
|
|
|
|
* mapped (G5).
|
2002-02-14 01:39:11 +00:00
|
|
|
*/
|
2009-04-12 03:03:55 +00:00
|
|
|
|
2009-04-04 00:22:44 +00:00
|
|
|
msr = mfmsr();
|
2009-06-10 12:47:54 +00:00
|
|
|
mtmsr((msr & ~(PSL_IR | PSL_DR)) | PSL_RI);
|
2004-07-20 02:22:36 +00:00
|
|
|
isync();
|
2009-04-04 00:22:44 +00:00
|
|
|
|
2009-04-12 03:03:55 +00:00
|
|
|
/*
|
|
|
|
* Measure the cacheline size using dcbz
|
|
|
|
*
|
|
|
|
* Use EXC_PGM as a playground. We are about to overwrite it
|
|
|
|
* anyway, we know it exists, and we know it is cache-aligned.
|
|
|
|
*/
|
|
|
|
|
|
|
|
cache_check = (void *)EXC_PGM;
|
|
|
|
|
|
|
|
for (cacheline_size = 0; cacheline_size < 0x100; cacheline_size++)
|
|
|
|
cache_check[cacheline_size] = 0xff;
|
|
|
|
|
2009-12-03 20:55:09 +00:00
|
|
|
__asm __volatile("dcbz 0,%0":: "r" (cache_check) : "memory");
|
2009-04-12 03:03:55 +00:00
|
|
|
|
|
|
|
/* Find the first byte dcbz did not zero to get the cache line size */
|
|
|
|
for (cacheline_size = 0; cacheline_size < 0x100 &&
|
|
|
|
cache_check[cacheline_size] == 0; cacheline_size++);
|
|
|
|
|
2009-06-10 12:47:54 +00:00
|
|
|
/* Work around psim bug */
|
|
|
|
if (cacheline_size == 0) {
|
|
|
|
printf("WARNING: cacheline size undetermined, setting to 32\n");
|
|
|
|
cacheline_size = 32;
|
|
|
|
}
|
|
|
|
|
2010-07-13 05:32:19 +00:00
|
|
|
#ifndef __powerpc64__
|
2009-04-04 00:22:44 +00:00
|
|
|
/*
|
|
|
|
* Figure out whether we need to use the 64 bit PMAP. This works by
|
|
|
|
* executing an instruction that is only legal on 64-bit PPC (mtmsrd),
|
|
|
|
* and setting ppc64 = 0 if that causes a trap.
|
|
|
|
*/
|
|
|
|
|
|
|
|
ppc64 = 1;
|
|
|
|
|
|
|
|
bcopy(&testppc64, (void *)EXC_PGM, (size_t)&testppc64size);
|
|
|
|
__syncicache((void *)EXC_PGM, (size_t)&testppc64size);
|
|
|
|
|
|
|
|
__asm __volatile("\
|
|
|
|
mfmsr %0; \
|
|
|
|
mtsprg2 %1; \
|
|
|
|
\
|
|
|
|
mtmsrd %0; \
|
|
|
|
mfsprg2 %1;"
|
|
|
|
: "=r"(scratch), "=r"(ppc64));
|
|
|
|
|
2009-11-28 17:33:19 +00:00
|
|
|
if (ppc64)
|
|
|
|
cpu_features |= PPC_FEATURE_64;
|
|
|
|
|
2009-04-04 00:22:44 +00:00
|
|
|
/*
|
|
|
|
* Now copy restorebridge into all the handlers, if necessary,
|
|
|
|
* and set up the trap tables.
|
|
|
|
*/
|
|
|
|
|
2009-11-28 17:33:19 +00:00
|
|
|
if (cpu_features & PPC_FEATURE_64) {
|
2009-04-04 00:22:44 +00:00
|
|
|
/* Patch the two instances of rfi -> rfid */
|
|
|
|
bcopy(&rfid_patch,&rfi_patch1,4);
|
2009-04-05 21:52:13 +00:00
|
|
|
#ifdef KDB
|
|
|
|
/* rfi_patch2 is at the end of dbleave */
|
2009-04-04 00:22:44 +00:00
|
|
|
bcopy(&rfid_patch,&rfi_patch2,4);
|
2009-04-05 21:52:13 +00:00
|
|
|
#endif
|
2009-04-04 00:22:44 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Copy a code snippet to restore 32-bit bridge mode
|
|
|
|
* to the top of every non-generic trap handler
|
|
|
|
*/
|
|
|
|
|
|
|
|
trap_offset += (size_t)&restorebridgesize;
|
|
|
|
bcopy(&restorebridge, (void *)EXC_RST, trap_offset);
|
|
|
|
bcopy(&restorebridge, (void *)EXC_DSI, trap_offset);
|
|
|
|
bcopy(&restorebridge, (void *)EXC_ALI, trap_offset);
|
|
|
|
bcopy(&restorebridge, (void *)EXC_PGM, trap_offset);
|
|
|
|
bcopy(&restorebridge, (void *)EXC_MCHK, trap_offset);
|
|
|
|
bcopy(&restorebridge, (void *)EXC_TRC, trap_offset);
|
|
|
|
bcopy(&restorebridge, (void *)EXC_BPT, trap_offset);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Set the common trap entry point to the one that
|
|
|
|
* knows to restore 32-bit operation on execution.
|
|
|
|
*/
|
|
|
|
|
|
|
|
generictrap = &trapcode64;
|
|
|
|
} else {
|
|
|
|
generictrap = &trapcode;
|
|
|
|
}
|
|
|
|
|
2010-07-13 05:32:19 +00:00
|
|
|
#else /* powerpc64 */
|
|
|
|
cpu_features |= PPC_FEATURE_64;
|
|
|
|
generictrap = &trapcode;
|
|
|
|
#endif
|
|
|
|
|
2008-04-27 22:33:43 +00:00
|
|
|
#ifdef SMP
|
2009-04-04 00:22:44 +00:00
|
|
|
bcopy(&rstcode, (void *)(EXC_RST + trap_offset), (size_t)&rstsize);
|
2008-04-27 22:33:43 +00:00
|
|
|
#else
|
2009-04-04 00:22:44 +00:00
|
|
|
bcopy(generictrap, (void *)EXC_RST, (size_t)&trapsize);
|
2008-04-27 22:33:43 +00:00
|
|
|
#endif
|
2009-04-04 00:22:44 +00:00
|
|
|
|
2004-07-12 22:25:09 +00:00
|
|
|
#ifdef KDB
|
2009-04-04 00:22:44 +00:00
|
|
|
bcopy(&dblow, (void *)(EXC_MCHK + trap_offset), (size_t)&dbsize);
|
|
|
|
bcopy(&dblow, (void *)(EXC_PGM + trap_offset), (size_t)&dbsize);
|
|
|
|
bcopy(&dblow, (void *)(EXC_TRC + trap_offset), (size_t)&dbsize);
|
|
|
|
bcopy(&dblow, (void *)(EXC_BPT + trap_offset), (size_t)&dbsize);
|
|
|
|
#else
|
|
|
|
bcopy(generictrap, (void *)EXC_MCHK, (size_t)&trapsize);
|
|
|
|
bcopy(generictrap, (void *)EXC_PGM, (size_t)&trapsize);
|
|
|
|
bcopy(generictrap, (void *)EXC_TRC, (size_t)&trapsize);
|
|
|
|
bcopy(generictrap, (void *)EXC_BPT, (size_t)&trapsize);
|
2004-02-04 13:00:56 +00:00
|
|
|
#endif
|
2009-04-04 00:22:44 +00:00
|
|
|
bcopy(&alitrap, (void *)(EXC_ALI + trap_offset), (size_t)&alisize);
|
2010-07-13 05:32:19 +00:00
|
|
|
bcopy(&dsitrap, (void *)(EXC_DSI + trap_offset), (size_t)&dsisize);
|
2009-04-04 00:22:44 +00:00
|
|
|
bcopy(generictrap, (void *)EXC_ISI, (size_t)&trapsize);
|
2010-07-13 05:32:19 +00:00
|
|
|
#ifdef __powerpc64__
|
|
|
|
bcopy(generictrap, (void *)EXC_DSE, (size_t)&trapsize);
|
|
|
|
bcopy(generictrap, (void *)EXC_ISE, (size_t)&trapsize);
|
|
|
|
#endif
|
2009-04-04 00:22:44 +00:00
|
|
|
bcopy(generictrap, (void *)EXC_EXI, (size_t)&trapsize);
|
|
|
|
bcopy(generictrap, (void *)EXC_FPU, (size_t)&trapsize);
|
|
|
|
bcopy(generictrap, (void *)EXC_DECR, (size_t)&trapsize);
|
|
|
|
bcopy(generictrap, (void *)EXC_SC, (size_t)&trapsize);
|
|
|
|
bcopy(generictrap, (void *)EXC_FPA, (size_t)&trapsize);
|
|
|
|
bcopy(generictrap, (void *)EXC_VEC, (size_t)&trapsize);
|
|
|
|
bcopy(generictrap, (void *)EXC_VECAST, (size_t)&trapsize);
|
|
|
|
bcopy(generictrap, (void *)EXC_THRM, (size_t)&trapsize);
|
2003-01-08 12:27:47 +00:00
|
|
|
__syncicache(EXC_RSVD, EXC_LAST - EXC_RSVD);
|
2002-02-14 01:39:11 +00:00
|
|
|
|
2002-09-19 04:30:43 +00:00
|
|
|
/*
|
2009-04-04 00:22:44 +00:00
|
|
|
* Restore MSR
|
2002-09-19 04:30:43 +00:00
|
|
|
*/
|
2009-04-04 00:22:44 +00:00
|
|
|
mtmsr(msr);
|
2004-07-20 02:22:36 +00:00
|
|
|
isync();
|
2009-05-14 00:34:26 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Choose a platform module so we can get the physical memory map.
|
|
|
|
*/
|
|
|
|
|
|
|
|
platform_probe_and_attach();
|
2002-09-19 04:30:43 +00:00
|
|
|
|
2002-02-28 03:15:49 +00:00
|
|
|
/*
|
2009-05-14 00:34:26 +00:00
|
|
|
* Initialise virtual memory. Use BUS_PROBE_GENERIC priority
|
|
|
|
* in case the platform module had a better idea of what we
|
|
|
|
* should do.
|
2002-02-28 03:15:49 +00:00
|
|
|
*/
|
2009-11-28 17:33:19 +00:00
|
|
|
if (cpu_features & PPC_FEATURE_64)
|
2009-05-14 00:34:26 +00:00
|
|
|
pmap_mmu_install(MMU_TYPE_G5, BUS_PROBE_GENERIC);
|
2009-04-04 00:22:44 +00:00
|
|
|
else
|
2009-05-14 00:34:26 +00:00
|
|
|
pmap_mmu_install(MMU_TYPE_OEA, BUS_PROBE_GENERIC);
|
2009-04-04 00:22:44 +00:00
|
|
|
|
2002-02-28 03:15:49 +00:00
|
|
|
pmap_bootstrap(startkernel, endkernel);
|
2009-04-04 00:22:44 +00:00
|
|
|
mtmsr(mfmsr() | PSL_IR|PSL_DR|PSL_ME|PSL_RI);
|
|
|
|
isync();
|
2002-02-28 03:15:49 +00:00
|
|
|
|
2002-03-07 10:15:17 +00:00
|
|
|
/*
|
2004-03-02 06:13:09 +00:00
|
|
|
* Initialize params/tunables that are derived from memsize
|
2002-03-07 10:15:17 +00:00
|
|
|
*/
|
|
|
|
init_param2(physmem);
|
|
|
|
|
2005-02-28 07:14:13 +00:00
|
|
|
/*
|
|
|
|
* Grab booted kernel's name
|
|
|
|
*/
|
|
|
|
env = getenv("kernelname");
|
|
|
|
if (env != NULL) {
|
|
|
|
strlcpy(kernelname, env, sizeof(kernelname));
|
|
|
|
freeenv(env);
|
|
|
|
}
|
|
|
|
|
2002-02-28 03:15:49 +00:00
|
|
|
/*
|
|
|
|
* Finish setting up thread0.
|
|
|
|
*/
|
|
|
|
thread0.td_pcb = (struct pcb *)
|
2008-04-16 23:28:12 +00:00
|
|
|
((thread0.td_kstack + thread0.td_kstack_pages * PAGE_SIZE -
|
2010-07-13 05:32:19 +00:00
|
|
|
sizeof(struct pcb)) & ~15UL);
|
2009-04-24 08:57:54 +00:00
|
|
|
bzero((void *)thread0.td_pcb, sizeof(struct pcb));
|
2008-04-16 23:28:12 +00:00
|
|
|
pc->pc_curpcb = thread0.td_pcb;
|
2002-02-28 03:15:49 +00:00
|
|
|
|
2008-04-16 23:28:12 +00:00
|
|
|
/* Initialise the message buffer. */
|
2002-02-28 03:15:49 +00:00
|
|
|
msgbufinit(msgbufp, MSGBUF_SIZE);
|
2004-07-12 22:25:09 +00:00
|
|
|
|
|
|
|
#ifdef KDB
|
|
|
|
if (boothowto & RB_KDB)
|
2007-12-25 17:52:02 +00:00
|
|
|
kdb_enter(KDB_WHY_BOOTFLAGS,
|
|
|
|
"Boot flags requested debugger");
|
2004-07-12 22:25:09 +00:00
|
|
|
#endif
|
2008-04-16 23:28:12 +00:00
|
|
|
|
2010-07-13 05:32:19 +00:00
|
|
|
return (((uintptr_t)thread0.td_pcb -
|
|
|
|
(sizeof(struct callframe) - 3*sizeof(register_t))) & ~15UL);
|
2002-02-14 01:39:11 +00:00
|
|
|
}
|
|
|
|
|
2001-06-10 02:39:37 +00:00
|
|
|
void
|
|
|
|
bzero(void *buf, size_t len)
|
|
|
|
{
|
|
|
|
caddr_t p;
|
|
|
|
|
|
|
|
p = buf;
|
|
|
|
|
|
|
|
while (((vm_offset_t) p & (sizeof(u_long) - 1)) && len) {
|
|
|
|
*p++ = 0;
|
|
|
|
len--;
|
|
|
|
}
|
|
|
|
|
|
|
|
while (len >= sizeof(u_long) * 8) {
|
|
|
|
*(u_long*) p = 0;
|
|
|
|
*((u_long*) p + 1) = 0;
|
|
|
|
*((u_long*) p + 2) = 0;
|
|
|
|
*((u_long*) p + 3) = 0;
|
|
|
|
len -= sizeof(u_long) * 8;
|
|
|
|
*((u_long*) p + 4) = 0;
|
|
|
|
*((u_long*) p + 5) = 0;
|
|
|
|
*((u_long*) p + 6) = 0;
|
|
|
|
*((u_long*) p + 7) = 0;
|
|
|
|
p += sizeof(u_long) * 8;
|
|
|
|
}
|
|
|
|
|
|
|
|
while (len >= sizeof(u_long)) {
|
|
|
|
*(u_long*) p = 0;
|
|
|
|
len -= sizeof(u_long);
|
|
|
|
p += sizeof(u_long);
|
|
|
|
}
|
|
|
|
|
|
|
|
while (len) {
|
|
|
|
*p++ = 0;
|
|
|
|
len--;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
cpu_boot(int howto)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
|
2009-05-18 18:37:18 +00:00
|
|
|
/*
|
|
|
|
* Flush the D-cache for non-DMA I/O so that the I-cache can
|
|
|
|
* be made coherent later.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
cpu_flush_dcache(void *ptr, size_t len)
|
|
|
|
{
|
|
|
|
/* TBD */
|
|
|
|
}
|
|
|
|
|
2007-05-27 21:05:35 +00:00
|
|
|
void
|
|
|
|
cpu_initclocks(void)
|
|
|
|
{
|
|
|
|
|
2007-11-13 15:47:55 +00:00
|
|
|
decr_tc_init();
|
2010-09-11 04:45:51 +00:00
|
|
|
cpu_initclocks_bsp();
|
2007-05-27 21:05:35 +00:00
|
|
|
}
|
|
|
|
|
2001-06-10 02:39:37 +00:00
|
|
|
/*
|
|
|
|
* Shutdown the CPU as much as possible.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
cpu_halt(void)
|
|
|
|
{
|
|
|
|
|
|
|
|
OF_exit();
|
|
|
|
}
|
|
|
|
|
2003-10-19 02:36:07 +00:00
|
|
|
void
|
2008-04-25 05:18:50 +00:00
|
|
|
cpu_idle(int busy)
|
2003-10-19 02:36:07 +00:00
|
|
|
{
|
2010-07-13 05:32:19 +00:00
|
|
|
register_t msr;
|
2009-10-29 14:22:09 +00:00
|
|
|
uint16_t vers;
|
2008-03-07 22:27:06 +00:00
|
|
|
|
|
|
|
msr = mfmsr();
|
2009-10-29 14:22:09 +00:00
|
|
|
vers = mfpvr() >> 16;
|
2005-02-28 09:49:00 +00:00
|
|
|
|
|
|
|
#ifdef INVARIANTS
|
2008-03-07 22:27:06 +00:00
|
|
|
if ((msr & PSL_EE) != PSL_EE) {
|
2005-02-28 09:49:00 +00:00
|
|
|
struct thread *td = curthread;
|
2010-07-13 05:32:19 +00:00
|
|
|
printf("td msr %#lx\n", (u_long)td->td_md.md_saved_msr);
|
2005-02-28 09:49:00 +00:00
|
|
|
panic("ints disabled in idleproc!");
|
|
|
|
}
|
|
|
|
#endif
|
Refactor timer management code with priority to one-shot operation mode.
The main goal of this is to generate timer interrupts only when there is
some work to do. When CPU is busy interrupts are generating at full rate
of hz + stathz to fullfill scheduler and timekeeping requirements. But
when CPU is idle, only minimum set of interrupts (down to 8 interrupts per
second per CPU now), needed to handle scheduled callouts is executed.
This allows significantly increase idle CPU sleep time, increasing effect
of static power-saving technologies. Also it should reduce host CPU load
on virtualized systems, when guest system is idle.
There is set of tunables, also available as writable sysctls, allowing to
control wanted event timer subsystem behavior:
kern.eventtimer.timer - allows to choose event timer hardware to use.
On x86 there is up to 4 different kinds of timers. Depending on whether
chosen timer is per-CPU, behavior of other options slightly differs.
kern.eventtimer.periodic - allows to choose periodic and one-shot
operation mode. In periodic mode, current timer hardware taken as the only
source of time for time events. This mode is quite alike to previous kernel
behavior. One-shot mode instead uses currently selected time counter
hardware to schedule all needed events one by one and program timer to
generate interrupt exactly in specified time. Default value depends of
chosen timer capabilities, but one-shot mode is preferred, until other is
forced by user or hardware.
kern.eventtimer.singlemul - in periodic mode specifies how much times
higher timer frequency should be, to not strictly alias hardclock() and
statclock() events. Default values are 2 and 4, but could be reduced to 1
if extra interrupts are unwanted.
kern.eventtimer.idletick - makes each CPU to receive every timer interrupt
independently of whether they busy or not. By default this options is
disabled. If chosen timer is per-CPU and runs in periodic mode, this option
has no effect - all interrupts are generating.
As soon as this patch modifies cpu_idle() on some platforms, I have also
refactored one on x86. Now it makes use of MONITOR/MWAIT instrunctions
(if supported) under high sleep/wakeup rate, as fast alternative to other
methods. It allows SMP scheduler to wake up sleeping CPUs much faster
without using IPI, significantly increasing performance on some highly
task-switching loads.
Tested by: many (on i386, amd64, sparc64 and powerc)
H/W donated by: Gheorghe Ardelean
Sponsored by: iXsystems, Inc.
2010-09-13 07:25:35 +00:00
|
|
|
CTR2(KTR_SPARE2, "cpu_idle(%d) at %d",
|
|
|
|
busy, curcpu);
|
2008-03-07 22:27:06 +00:00
|
|
|
if (powerpc_pow_enabled) {
|
Refactor timer management code with priority to one-shot operation mode.
The main goal of this is to generate timer interrupts only when there is
some work to do. When CPU is busy interrupts are generating at full rate
of hz + stathz to fullfill scheduler and timekeeping requirements. But
when CPU is idle, only minimum set of interrupts (down to 8 interrupts per
second per CPU now), needed to handle scheduled callouts is executed.
This allows significantly increase idle CPU sleep time, increasing effect
of static power-saving technologies. Also it should reduce host CPU load
on virtualized systems, when guest system is idle.
There is set of tunables, also available as writable sysctls, allowing to
control wanted event timer subsystem behavior:
kern.eventtimer.timer - allows to choose event timer hardware to use.
On x86 there is up to 4 different kinds of timers. Depending on whether
chosen timer is per-CPU, behavior of other options slightly differs.
kern.eventtimer.periodic - allows to choose periodic and one-shot
operation mode. In periodic mode, current timer hardware taken as the only
source of time for time events. This mode is quite alike to previous kernel
behavior. One-shot mode instead uses currently selected time counter
hardware to schedule all needed events one by one and program timer to
generate interrupt exactly in specified time. Default value depends of
chosen timer capabilities, but one-shot mode is preferred, until other is
forced by user or hardware.
kern.eventtimer.singlemul - in periodic mode specifies how much times
higher timer frequency should be, to not strictly alias hardclock() and
statclock() events. Default values are 2 and 4, but could be reduced to 1
if extra interrupts are unwanted.
kern.eventtimer.idletick - makes each CPU to receive every timer interrupt
independently of whether they busy or not. By default this options is
disabled. If chosen timer is per-CPU and runs in periodic mode, this option
has no effect - all interrupts are generating.
As soon as this patch modifies cpu_idle() on some platforms, I have also
refactored one on x86. Now it makes use of MONITOR/MWAIT instrunctions
(if supported) under high sleep/wakeup rate, as fast alternative to other
methods. It allows SMP scheduler to wake up sleeping CPUs much faster
without using IPI, significantly increasing performance on some highly
task-switching loads.
Tested by: many (on i386, amd64, sparc64 and powerc)
H/W donated by: Gheorghe Ardelean
Sponsored by: iXsystems, Inc.
2010-09-13 07:25:35 +00:00
|
|
|
if (!busy) {
|
|
|
|
critical_enter();
|
|
|
|
cpu_idleclock();
|
|
|
|
}
|
2009-10-29 14:22:09 +00:00
|
|
|
switch (vers) {
|
|
|
|
case IBM970:
|
|
|
|
case IBM970FX:
|
|
|
|
case IBM970MP:
|
|
|
|
case MPC7447A:
|
|
|
|
case MPC7448:
|
|
|
|
case MPC7450:
|
|
|
|
case MPC7455:
|
|
|
|
case MPC7457:
|
|
|
|
__asm __volatile("\
|
|
|
|
dssall; sync; mtmsr %0; isync"
|
|
|
|
:: "r"(msr | PSL_POW));
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
powerpc_sync();
|
|
|
|
mtmsr(msr | PSL_POW);
|
|
|
|
isync();
|
|
|
|
break;
|
|
|
|
}
|
Refactor timer management code with priority to one-shot operation mode.
The main goal of this is to generate timer interrupts only when there is
some work to do. When CPU is busy interrupts are generating at full rate
of hz + stathz to fullfill scheduler and timekeeping requirements. But
when CPU is idle, only minimum set of interrupts (down to 8 interrupts per
second per CPU now), needed to handle scheduled callouts is executed.
This allows significantly increase idle CPU sleep time, increasing effect
of static power-saving technologies. Also it should reduce host CPU load
on virtualized systems, when guest system is idle.
There is set of tunables, also available as writable sysctls, allowing to
control wanted event timer subsystem behavior:
kern.eventtimer.timer - allows to choose event timer hardware to use.
On x86 there is up to 4 different kinds of timers. Depending on whether
chosen timer is per-CPU, behavior of other options slightly differs.
kern.eventtimer.periodic - allows to choose periodic and one-shot
operation mode. In periodic mode, current timer hardware taken as the only
source of time for time events. This mode is quite alike to previous kernel
behavior. One-shot mode instead uses currently selected time counter
hardware to schedule all needed events one by one and program timer to
generate interrupt exactly in specified time. Default value depends of
chosen timer capabilities, but one-shot mode is preferred, until other is
forced by user or hardware.
kern.eventtimer.singlemul - in periodic mode specifies how much times
higher timer frequency should be, to not strictly alias hardclock() and
statclock() events. Default values are 2 and 4, but could be reduced to 1
if extra interrupts are unwanted.
kern.eventtimer.idletick - makes each CPU to receive every timer interrupt
independently of whether they busy or not. By default this options is
disabled. If chosen timer is per-CPU and runs in periodic mode, this option
has no effect - all interrupts are generating.
As soon as this patch modifies cpu_idle() on some platforms, I have also
refactored one on x86. Now it makes use of MONITOR/MWAIT instrunctions
(if supported) under high sleep/wakeup rate, as fast alternative to other
methods. It allows SMP scheduler to wake up sleeping CPUs much faster
without using IPI, significantly increasing performance on some highly
task-switching loads.
Tested by: many (on i386, amd64, sparc64 and powerc)
H/W donated by: Gheorghe Ardelean
Sponsored by: iXsystems, Inc.
2010-09-13 07:25:35 +00:00
|
|
|
if (!busy) {
|
|
|
|
cpu_activeclock();
|
|
|
|
critical_exit();
|
|
|
|
}
|
2008-03-07 22:27:06 +00:00
|
|
|
}
|
Refactor timer management code with priority to one-shot operation mode.
The main goal of this is to generate timer interrupts only when there is
some work to do. When CPU is busy interrupts are generating at full rate
of hz + stathz to fullfill scheduler and timekeeping requirements. But
when CPU is idle, only minimum set of interrupts (down to 8 interrupts per
second per CPU now), needed to handle scheduled callouts is executed.
This allows significantly increase idle CPU sleep time, increasing effect
of static power-saving technologies. Also it should reduce host CPU load
on virtualized systems, when guest system is idle.
There is set of tunables, also available as writable sysctls, allowing to
control wanted event timer subsystem behavior:
kern.eventtimer.timer - allows to choose event timer hardware to use.
On x86 there is up to 4 different kinds of timers. Depending on whether
chosen timer is per-CPU, behavior of other options slightly differs.
kern.eventtimer.periodic - allows to choose periodic and one-shot
operation mode. In periodic mode, current timer hardware taken as the only
source of time for time events. This mode is quite alike to previous kernel
behavior. One-shot mode instead uses currently selected time counter
hardware to schedule all needed events one by one and program timer to
generate interrupt exactly in specified time. Default value depends of
chosen timer capabilities, but one-shot mode is preferred, until other is
forced by user or hardware.
kern.eventtimer.singlemul - in periodic mode specifies how much times
higher timer frequency should be, to not strictly alias hardclock() and
statclock() events. Default values are 2 and 4, but could be reduced to 1
if extra interrupts are unwanted.
kern.eventtimer.idletick - makes each CPU to receive every timer interrupt
independently of whether they busy or not. By default this options is
disabled. If chosen timer is per-CPU and runs in periodic mode, this option
has no effect - all interrupts are generating.
As soon as this patch modifies cpu_idle() on some platforms, I have also
refactored one on x86. Now it makes use of MONITOR/MWAIT instrunctions
(if supported) under high sleep/wakeup rate, as fast alternative to other
methods. It allows SMP scheduler to wake up sleeping CPUs much faster
without using IPI, significantly increasing performance on some highly
task-switching loads.
Tested by: many (on i386, amd64, sparc64 and powerc)
H/W donated by: Gheorghe Ardelean
Sponsored by: iXsystems, Inc.
2010-09-13 07:25:35 +00:00
|
|
|
CTR2(KTR_SPARE2, "cpu_idle(%d) at %d done",
|
|
|
|
busy, curcpu);
|
2003-10-19 02:36:07 +00:00
|
|
|
}
|
|
|
|
|
2008-04-25 05:18:50 +00:00
|
|
|
int
|
|
|
|
cpu_idle_wakeup(int cpu)
|
|
|
|
{
|
|
|
|
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
2001-06-10 02:39:37 +00:00
|
|
|
int
|
2001-09-20 00:47:17 +00:00
|
|
|
ptrace_set_pc(struct thread *td, unsigned long addr)
|
2001-06-10 02:39:37 +00:00
|
|
|
{
|
2004-07-29 13:34:50 +00:00
|
|
|
struct trapframe *tf;
|
2001-06-10 02:39:37 +00:00
|
|
|
|
2004-07-29 13:34:50 +00:00
|
|
|
tf = td->td_frame;
|
|
|
|
tf->srr0 = (register_t)addr;
|
|
|
|
|
|
|
|
return (0);
|
2001-06-10 02:39:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
int
|
2001-09-20 00:47:17 +00:00
|
|
|
ptrace_single_step(struct thread *td)
|
2001-06-10 02:39:37 +00:00
|
|
|
{
|
2004-07-29 13:34:50 +00:00
|
|
|
struct trapframe *tf;
|
|
|
|
|
|
|
|
tf = td->td_frame;
|
|
|
|
tf->srr1 |= PSL_SE;
|
2001-06-10 02:39:37 +00:00
|
|
|
|
2004-07-29 13:34:50 +00:00
|
|
|
return (0);
|
2001-06-10 02:39:37 +00:00
|
|
|
}
|
|
|
|
|
2004-07-13 07:22:56 +00:00
|
|
|
int
|
|
|
|
ptrace_clear_single_step(struct thread *td)
|
|
|
|
{
|
2004-07-29 13:34:50 +00:00
|
|
|
struct trapframe *tf;
|
2004-07-13 07:22:56 +00:00
|
|
|
|
2004-07-29 13:34:50 +00:00
|
|
|
tf = td->td_frame;
|
|
|
|
tf->srr1 &= ~PSL_SE;
|
|
|
|
|
|
|
|
return (0);
|
2004-07-13 07:22:56 +00:00
|
|
|
}
|
|
|
|
|
2008-03-02 17:05:57 +00:00
|
|
|
void
|
|
|
|
kdb_cpu_clear_singlestep(void)
|
|
|
|
{
|
|
|
|
|
|
|
|
kdb_frame->srr1 &= ~PSL_SE;
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
kdb_cpu_set_singlestep(void)
|
|
|
|
{
|
|
|
|
|
|
|
|
kdb_frame->srr1 |= PSL_SE;
|
|
|
|
}
|
|
|
|
|
2001-06-10 02:39:37 +00:00
|
|
|
/*
|
2001-12-11 23:33:44 +00:00
|
|
|
* Initialise a struct pcpu.
|
2001-06-10 02:39:37 +00:00
|
|
|
*/
|
|
|
|
void
|
2001-12-11 23:33:44 +00:00
|
|
|
cpu_pcpu_init(struct pcpu *pcpu, int cpuid, size_t sz)
|
|
|
|
{
|
2010-07-13 05:32:19 +00:00
|
|
|
#ifdef __powerpc64__
|
|
|
|
/* Copy the SLB contents from the current CPU */
|
|
|
|
memcpy(pcpu->pc_slb, PCPU_GET(slb), sizeof(pcpu->pc_slb));
|
|
|
|
#endif
|
2001-06-10 02:39:37 +00:00
|
|
|
}
|
2002-05-28 07:36:36 +00:00
|
|
|
|
Divorce critical sections from spinlocks. Critical sections as denoted by
critical_enter() and critical_exit() are now solely a mechanism for
deferring kernel preemptions. They no longer have any affect on
interrupts. This means that standalone critical sections are now very
cheap as they are simply unlocked integer increments and decrements for the
common case.
Spin mutexes now use a separate KPI implemented in MD code: spinlock_enter()
and spinlock_exit(). This KPI is responsible for providing whatever MD
guarantees are needed to ensure that a thread holding a spin lock won't
be preempted by any other code that will try to lock the same lock. For
now all archs continue to block interrupts in a "spinlock section" as they
did formerly in all critical sections. Note that I've also taken this
opportunity to push a few things into MD code rather than MI. For example,
critical_fork_exit() no longer exists. Instead, MD code ensures that new
threads have the correct state when they are created. Also, we no longer
try to fixup the idlethreads for APs in MI code. Instead, each arch sets
the initial curthread and adjusts the state of the idle thread it borrows
in order to perform the initial context switch.
This change is largely a big NOP, but the cleaner separation it provides
will allow for more efficient alternative locking schemes in other parts
of the kernel (bare critical sections rather than per-CPU spin mutexes
for per-CPU data for example).
Reviewed by: grehan, cognet, arch@, others
Tested on: i386, alpha, sparc64, powerpc, arm, possibly more
2005-04-04 21:53:56 +00:00
|
|
|
void
|
|
|
|
spinlock_enter(void)
|
|
|
|
{
|
|
|
|
struct thread *td;
|
|
|
|
|
|
|
|
td = curthread;
|
|
|
|
if (td->td_md.md_spinlock_count == 0)
|
|
|
|
td->td_md.md_saved_msr = intr_disable();
|
|
|
|
td->td_md.md_spinlock_count++;
|
|
|
|
critical_enter();
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
spinlock_exit(void)
|
|
|
|
{
|
|
|
|
struct thread *td;
|
|
|
|
|
|
|
|
td = curthread;
|
|
|
|
critical_exit();
|
|
|
|
td->td_md.md_spinlock_count--;
|
|
|
|
if (td->td_md.md_spinlock_count == 0)
|
|
|
|
intr_restore(td->td_md.md_saved_msr);
|
|
|
|
}
|
|
|
|
|
2002-05-28 07:36:36 +00:00
|
|
|
/*
|
|
|
|
* kcopy(const void *src, void *dst, size_t len);
|
|
|
|
*
|
|
|
|
* Copy len bytes from src to dst, aborting if we encounter a fatal
|
|
|
|
* page fault.
|
|
|
|
*
|
|
|
|
* kcopy() _must_ save and restore the old fault handler since it is
|
|
|
|
* called by uiomove(), which may be in the path of servicing a non-fatal
|
|
|
|
* page fault.
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
kcopy(const void *src, void *dst, size_t len)
|
|
|
|
{
|
|
|
|
struct thread *td;
|
|
|
|
faultbuf env, *oldfault;
|
|
|
|
int rv;
|
|
|
|
|
|
|
|
td = PCPU_GET(curthread);
|
|
|
|
oldfault = td->td_pcb->pcb_onfault;
|
|
|
|
if ((rv = setfault(env)) != 0) {
|
|
|
|
td->td_pcb->pcb_onfault = oldfault;
|
|
|
|
return rv;
|
|
|
|
}
|
|
|
|
|
|
|
|
memcpy(dst, src, len);
|
|
|
|
|
|
|
|
td->td_pcb->pcb_onfault = oldfault;
|
|
|
|
return (0);
|
|
|
|
}
|
2003-04-01 00:18:55 +00:00
|
|
|
|
2004-07-12 22:25:09 +00:00
|
|
|
int db_trap_glue(struct trapframe *); /* Called from trap_subr.S */
|
|
|
|
|
|
|
|
int
|
|
|
|
db_trap_glue(struct trapframe *frame)
|
|
|
|
{
|
|
|
|
if (!(frame->srr1 & PSL_PR)
|
|
|
|
&& (frame->exc == EXC_TRC || frame->exc == EXC_RUNMODETRC
|
|
|
|
|| (frame->exc == EXC_PGM
|
|
|
|
&& (frame->srr1 & 0x20000))
|
2004-07-23 05:27:17 +00:00
|
|
|
|| frame->exc == EXC_BPT
|
|
|
|
|| frame->exc == EXC_DSI)) {
|
2004-07-12 22:25:09 +00:00
|
|
|
int type = frame->exc;
|
|
|
|
if (type == EXC_PGM && (frame->srr1 & 0x20000)) {
|
|
|
|
type = T_BREAKPOINT;
|
|
|
|
}
|
|
|
|
return (kdb_trap(type, 0, frame));
|
|
|
|
}
|
|
|
|
|
|
|
|
return (0);
|
|
|
|
}
|
2010-07-13 05:32:19 +00:00
|
|
|
|
|
|
|
#ifndef __powerpc64__
|
|
|
|
|
|
|
|
uint64_t
|
|
|
|
va_to_vsid(pmap_t pm, vm_offset_t va)
|
|
|
|
{
|
|
|
|
return ((pm->pm_sr[(uintptr_t)va >> ADDR_SR_SHFT]) & SR_VSID_MASK);
|
|
|
|
}
|
|
|
|
|
|
|
|
#endif
|