freebsd-dev/sys/amd64/include/mptable.h

2656 lines
62 KiB
C
Raw Normal View History

/*
* Copyright (c) 1996, by Steve Passe
* 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. The name of the developer may NOT be used to endorse or promote products
* derived from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR 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 AUTHOR 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.
*
1999-08-28 01:08:13 +00:00
* $FreeBSD$
*/
#include "opt_cpu.h"
#include "opt_kstack_pages.h"
#ifdef SMP
#include <machine/smptests.h>
#else
#error
#endif
1997-08-25 21:28:08 +00:00
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/bus.h>
#include <sys/cons.h> /* cngetc() */
#include <sys/dkstat.h>
#ifdef GPROF
#include <sys/gmon.h>
#endif
#include <sys/kernel.h>
2001-10-11 17:53:43 +00:00
#include <sys/ktr.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/memrange.h>
#include <sys/mutex.h>
#include <sys/pcpu.h>
#include <sys/proc.h>
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
#include <sys/smp.h>
#include <sys/sysctl.h>
#include <sys/user.h>
#include <vm/vm.h>
#include <vm/vm_param.h>
#include <vm/pmap.h>
#include <vm/vm_kern.h>
#include <vm/vm_extern.h>
#include <vm/vm_map.h>
#include <machine/apic.h>
#include <machine/atomic.h>
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
#include <machine/cpu.h>
#include <machine/cpufunc.h>
#include <machine/mpapic.h>
#include <machine/psl.h>
#include <machine/segments.h>
#include <machine/smp.h>
#include <machine/smptests.h> /** TEST_DEFAULT_CONFIG, TEST_TEST1 */
#include <machine/tss.h>
#include <machine/specialreg.h>
#include <machine/privatespace.h>
#if defined(APIC_IO)
#include <machine/md_var.h> /* setidt() */
#include <i386/isa/icu.h> /* IPIs */
#include <i386/isa/intr_machdep.h> /* IPIs */
#endif /* APIC_IO */
#if defined(TEST_DEFAULT_CONFIG)
#define MPFPS_MPFB1 TEST_DEFAULT_CONFIG
#else
#define MPFPS_MPFB1 mpfps->mpfb1
#endif /* TEST_DEFAULT_CONFIG */
#define WARMBOOT_TARGET 0
#define WARMBOOT_OFF (KERNBASE + 0x0467)
#define WARMBOOT_SEG (KERNBASE + 0x0469)
#ifdef PC98
#define BIOS_BASE (0xe8000)
#define BIOS_SIZE (0x18000)
#else
#define BIOS_BASE (0xf0000)
#define BIOS_SIZE (0x10000)
#endif
#define BIOS_COUNT (BIOS_SIZE/4)
#define CMOS_REG (0x70)
#define CMOS_DATA (0x71)
#define BIOS_RESET (0x0f)
#define BIOS_WARM (0x0a)
#define PROCENTRY_FLAG_EN 0x01
#define PROCENTRY_FLAG_BP 0x02
#define IOAPICENTRY_FLAG_EN 0x01
/* MP Floating Pointer Structure */
typedef struct MPFPS {
char signature[4];
void *pap;
u_char length;
u_char spec_rev;
u_char checksum;
u_char mpfb1;
u_char mpfb2;
u_char mpfb3;
u_char mpfb4;
u_char mpfb5;
} *mpfps_t;
/* MP Configuration Table Header */
typedef struct MPCTH {
char signature[4];
u_short base_table_length;
u_char spec_rev;
u_char checksum;
u_char oem_id[8];
u_char product_id[12];
void *oem_table_pointer;
u_short oem_table_size;
u_short entry_count;
void *apic_address;
u_short extended_table_length;
u_char extended_table_checksum;
u_char reserved;
} *mpcth_t;
typedef struct PROCENTRY {
u_char type;
u_char apic_id;
u_char apic_version;
u_char cpu_flags;
u_long cpu_signature;
u_long feature_flags;
u_long reserved1;
u_long reserved2;
} *proc_entry_ptr;
typedef struct BUSENTRY {
u_char type;
u_char bus_id;
char bus_type[6];
} *bus_entry_ptr;
typedef struct IOAPICENTRY {
u_char type;
u_char apic_id;
u_char apic_version;
u_char apic_flags;
void *apic_address;
} *io_apic_entry_ptr;
typedef struct INTENTRY {
u_char type;
u_char int_type;
u_short int_flags;
u_char src_bus_id;
u_char src_bus_irq;
u_char dst_apic_id;
u_char dst_apic_int;
} *int_entry_ptr;
/* descriptions of MP basetable entries */
typedef struct BASETABLE_ENTRY {
u_char type;
u_char length;
char name[16];
} basetable_entry;
/*
* this code MUST be enabled here and in mpboot.s.
* it follows the very early stages of AP boot by placing values in CMOS ram.
* it NORMALLY will never be needed and thus the primitive method for enabling.
*
#define CHECK_POINTS
*/
1998-10-10 09:38:02 +00:00
#if defined(CHECK_POINTS) && !defined(PC98)
#define CHECK_READ(A) (outb(CMOS_REG, (A)), inb(CMOS_DATA))
#define CHECK_WRITE(A,D) (outb(CMOS_REG, (A)), outb(CMOS_DATA, (D)))
#define CHECK_INIT(D); \
CHECK_WRITE(0x34, (D)); \
CHECK_WRITE(0x35, (D)); \
CHECK_WRITE(0x36, (D)); \
CHECK_WRITE(0x37, (D)); \
CHECK_WRITE(0x38, (D)); \
CHECK_WRITE(0x39, (D));
#define CHECK_PRINT(S); \
printf("%s: %d, %d, %d, %d, %d, %d\n", \
(S), \
CHECK_READ(0x34), \
CHECK_READ(0x35), \
CHECK_READ(0x36), \
CHECK_READ(0x37), \
CHECK_READ(0x38), \
CHECK_READ(0x39));
#else /* CHECK_POINTS */
#define CHECK_INIT(D)
#define CHECK_PRINT(S)
#endif /* CHECK_POINTS */
/*
* Values to send to the POST hardware.
*/
#define MP_BOOTADDRESS_POST 0x10
#define MP_PROBE_POST 0x11
#define MPTABLE_PASS1_POST 0x12
#define MP_START_POST 0x13
#define MP_ENABLE_POST 0x14
#define MPTABLE_PASS2_POST 0x15
#define START_ALL_APS_POST 0x16
#define INSTALL_AP_TRAMP_POST 0x17
#define START_AP_POST 0x18
#define MP_ANNOUNCE_POST 0x19
/* used to hold the AP's until we are ready to release them */
2001-04-20 01:09:05 +00:00
static struct mtx ap_boot_mtx;
/** XXX FIXME: where does this really belong, isa.h/isa.c perhaps? */
int current_postcode;
/** XXX FIXME: what system files declare these??? */
extern struct region_descriptor r_gdt, r_idt;
int bsp_apic_ready = 0; /* flags useability of BSP apic */
int mp_naps; /* # of Applications processors */
int mp_nbusses; /* # of busses */
int mp_napics; /* # of IO APICs */
int boot_cpu_id; /* designated BSP */
vm_offset_t cpu_apic_address;
vm_offset_t io_apic_address[NAPICID]; /* NAPICID is more than enough */
extern int nkpt;
u_int32_t cpu_apic_versions[MAXCPU];
u_int32_t *io_apic_versions;
#ifdef APIC_INTR_REORDER
struct {
volatile int *location;
int bit;
} apic_isrbit_location[32];
#endif
struct apic_intmapinfo int_to_apicintpin[APIC_INTMAPSIZE];
/*
* APIC ID logical/physical mapping structures.
* We oversize these to simplify boot-time config.
*/
int cpu_num_to_apic_id[NAPICID];
int io_num_to_apic_id[NAPICID];
int apic_id_to_logical[NAPICID];
/* AP uses this during bootstrap. Do not staticize. */
char *bootSTK;
static int bootAP;
/* Hotwire a 0->4MB V==P mapping */
extern pt_entry_t *KPTphys;
/* SMP page table page */
extern pt_entry_t *SMPpt;
struct pcb stoppcbs[MAXCPU];
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
#ifdef APIC_IO
/* Variables needed for SMP tlb shootdown. */
vm_offset_t smp_tlb_addr1;
vm_offset_t smp_tlb_addr2;
volatile int smp_tlb_wait;
static struct mtx smp_tlb_mtx;
#endif
/*
* Local data and functions.
*/
/* Set to 1 once we're ready to let the APs out of the pen. */
static volatile int aps_ready = 0;
static int mp_capable;
static u_int boot_address;
static u_int base_memory;
static int picmode; /* 0: virtual wire mode, 1: PIC mode */
static mpfps_t mpfps;
static int search_for_sig(u_int32_t target, int count);
static void mp_enable(u_int boot_addr);
static void mptable_pass1(void);
static int mptable_pass2(void);
static void default_mp_table(int type);
static void fix_mp_table(void);
static void setup_apic_irq_mapping(void);
static void init_locks(void);
static int start_all_aps(u_int boot_addr);
static void install_ap_tramp(u_int boot_addr);
static int start_ap(int logicalCpu, u_int boot_addr);
void ap_init(void);
static int apic_int_is_bus_type(int intr, int bus_type);
static void release_aps(void *dummy);
/*
* initialize all the SMP locks
*/
/* lock region used by kernel profiling */
int mcount_lock;
#ifdef USE_COMLOCK
/* locks com (tty) data/hardware accesses: a FASTINTR() */
struct mtx com_mtx;
#endif /* USE_COMLOCK */
static void
init_locks(void)
{
#ifdef USE_COMLOCK
mtx_init(&com_mtx, "com", NULL, MTX_SPIN);
#endif /* USE_COMLOCK */
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
#ifdef APIC_IO
mtx_init(&smp_tlb_mtx, "tlb", NULL, MTX_SPIN);
#endif
}
/*
* Calculate usable address in base memory for AP trampoline code.
*/
u_int
mp_bootaddress(u_int basemem)
{
POSTCODE(MP_BOOTADDRESS_POST);
base_memory = basemem * 1024; /* convert to bytes */
boot_address = base_memory & ~0xfff; /* round down to 4k boundary */
if ((base_memory - boot_address) < bootMP_size)
boot_address -= 4096; /* not enough, lower by 4k */
return boot_address;
}
/*
* Look for an Intel MP spec table (ie, SMP capable hardware).
*/
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
void
i386_mp_probe(void)
{
int x;
u_long segment;
u_int32_t target;
POSTCODE(MP_PROBE_POST);
/* see if EBDA exists */
if ((segment = (u_long) * (u_short *) (KERNBASE + 0x40e)) != 0) {
/* search first 1K of EBDA */
target = (u_int32_t) (segment << 4);
if ((x = search_for_sig(target, 1024 / 4)) >= 0)
goto found;
} else {
/* last 1K of base memory, effective 'top of base' passed in */
target = (u_int32_t) (base_memory - 0x400);
if ((x = search_for_sig(target, 1024 / 4)) >= 0)
goto found;
}
/* search the BIOS */
target = (u_int32_t) BIOS_BASE;
if ((x = search_for_sig(target, BIOS_COUNT)) >= 0)
goto found;
/* nothing found */
mpfps = (mpfps_t)0;
mp_capable = 0;
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
return;
found:
/* calculate needed resources */
mpfps = (mpfps_t)x;
mptable_pass1();
/* flag fact that we are running multiple processors */
mp_capable = 1;
}
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
int
cpu_mp_probe(void)
{
/*
* Record BSP in CPU map
* This is done here so that MBUF init code works correctly.
*/
all_cpus = 1;
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
return (mp_capable);
}
/*
* Initialize the SMP hardware and the APIC and start up the AP's.
*/
void
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
cpu_mp_start(void)
{
POSTCODE(MP_START_POST);
/* look for MP capable motherboard */
if (mp_capable)
mp_enable(boot_address);
else
panic("MP hardware not found!");
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
cpu_setregs();
}
/*
* Print various information about the SMP system hardware and setup.
*/
void
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
cpu_mp_announce(void)
{
int x;
POSTCODE(MP_ANNOUNCE_POST);
1997-07-18 03:58:14 +00:00
printf(" cpu0 (BSP): apic id: %2d", CPU_TO_ID(0));
printf(", version: 0x%08x", cpu_apic_versions[0]);
printf(", at 0x%08x\n", cpu_apic_address);
for (x = 1; x <= mp_naps; ++x) {
printf(" cpu%d (AP): apic id: %2d", x, CPU_TO_ID(x));
printf(", version: 0x%08x", cpu_apic_versions[x]);
printf(", at 0x%08x\n", cpu_apic_address);
}
#if defined(APIC_IO)
for (x = 0; x < mp_napics; ++x) {
1997-07-18 03:58:14 +00:00
printf(" io%d (APIC): apic id: %2d", x, IO_TO_ID(x));
printf(", version: 0x%08x", io_apic_versions[x]);
printf(", at 0x%08x\n", io_apic_address[x]);
}
#else
printf(" Warning: APIC I/O disabled\n");
#endif /* APIC_IO */
}
/*
* AP cpu's call this to sync up protected mode.
*/
void
init_secondary(void)
{
int gsel_tss;
int x, myid = bootAP;
u_int cr0;
gdt_segs[GPRIV_SEL].ssd_base = (int) &SMP_prvspace[myid];
gdt_segs[GPROC0_SEL].ssd_base =
(int) &SMP_prvspace[myid].pcpu.pc_common_tss;
SMP_prvspace[myid].pcpu.pc_prvspace =
&SMP_prvspace[myid].pcpu;
for (x = 0; x < NGDT; x++) {
ssdtosd(&gdt_segs[x], &gdt[myid * NGDT + x].sd);
}
r_gdt.rd_limit = NGDT * sizeof(gdt[0]) - 1;
r_gdt.rd_base = (int) &gdt[myid * NGDT];
lgdt(&r_gdt); /* does magic intra-segment return */
lidt(&r_idt);
lldt(_default_ldt);
PCPU_SET(currentldt, _default_ldt);
gsel_tss = GSEL(GPROC0_SEL, SEL_KPL);
gdt[myid * NGDT + GPROC0_SEL].sd.sd_type = SDT_SYS386TSS;
PCPU_SET(common_tss.tss_esp0, 0); /* not used until after switch */
PCPU_SET(common_tss.tss_ss0, GSEL(GDATA_SEL, SEL_KPL));
PCPU_SET(common_tss.tss_ioopt, (sizeof (struct i386tss)) << 16);
PCPU_SET(tss_gdt, &gdt[myid * NGDT + GPROC0_SEL].sd);
PCPU_SET(common_tssd, *PCPU_GET(tss_gdt));
ltr(gsel_tss);
/*
* Set to a known state:
* Set by mpboot.s: CR0_PG, CR0_PE
* Set by cpu_setregs: CR0_NE, CR0_MP, CR0_TS, CR0_WP, CR0_AM
*/
cr0 = rcr0();
cr0 &= ~(CR0_CD | CR0_NW | CR0_EM);
load_cr0(cr0);
pmap_set_opt();
}
#if defined(APIC_IO)
/*
* Final configuration of the BSP's local APIC:
* - disable 'pic mode'.
* - disable 'virtual wire mode'.
* - enable NMI.
*/
void
bsp_apic_configure(void)
{
u_char byte;
u_int32_t temp;
/* leave 'pic mode' if necessary */
if (picmode) {
outb(0x22, 0x70); /* select IMCR */
byte = inb(0x23); /* current contents */
byte |= 0x01; /* mask external INTR */
outb(0x23, byte); /* disconnect 8259s/NMI */
}
/* mask lint0 (the 8259 'virtual wire' connection) */
temp = lapic.lvt_lint0;
temp |= APIC_LVT_M; /* set the mask */
lapic.lvt_lint0 = temp;
/* setup lint1 to handle NMI */
temp = lapic.lvt_lint1;
temp &= ~APIC_LVT_M; /* clear the mask */
lapic.lvt_lint1 = temp;
if (bootverbose)
apic_dump("bsp_apic_configure()");
}
#endif /* APIC_IO */
/*******************************************************************
* local functions and data
*/
/*
* start the SMP system
*/
static void
mp_enable(u_int boot_addr)
{
int x;
#if defined(APIC_IO)
int apic;
u_int ux;
#endif /* APIC_IO */
POSTCODE(MP_ENABLE_POST);
/* turn on 4MB of V == P addressing so we can get to MP table */
*(int *)PTD = PG_V | PG_RW | ((uintptr_t)(void *)KPTphys & PG_FRAME);
invltlb();
/* examine the MP table for needed info, uses physical addresses */
x = mptable_pass2();
*(int *)PTD = 0;
invltlb();
/* can't process default configs till the CPU APIC is pmapped */
if (x)
default_mp_table(x);
/* post scan cleanup */
fix_mp_table();
setup_apic_irq_mapping();
#if defined(APIC_IO)
/* fill the LOGICAL io_apic_versions table */
for (apic = 0; apic < mp_napics; ++apic) {
ux = io_apic_read(apic, IOAPIC_VER);
io_apic_versions[apic] = ux;
io_apic_set_id(apic, IO_TO_ID(apic));
}
1997-04-28 01:08:47 +00:00
/* program each IO APIC in the system */
for (apic = 0; apic < mp_napics; ++apic)
if (io_apic_setup(apic) < 0)
panic("IO APIC setup failure");
/* install a 'Spurious INTerrupt' vector */
setidt(XSPURIOUSINT_OFFSET, Xspuriousint,
SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL));
/* install an inter-CPU IPI for TLB invalidation */
setidt(XINVLTLB_OFFSET, Xinvltlb,
1997-04-28 01:08:47 +00:00
SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL));
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
setidt(XINVLPG_OFFSET, Xinvlpg,
SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL));
setidt(XINVLRNG_OFFSET, Xinvlrng,
SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL));
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
/* install an inter-CPU IPI for forwarding hardclock() */
setidt(XHARDCLOCK_OFFSET, Xhardclock,
SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL));
/* install an inter-CPU IPI for forwarding statclock() */
setidt(XSTATCLOCK_OFFSET, Xstatclock,
SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL));
/* install an inter-CPU IPI for all-CPU rendezvous */
setidt(XRENDEZVOUS_OFFSET, Xrendezvous,
SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL));
/* install an inter-CPU IPI for forcing an additional software trap */
setidt(XCPUAST_OFFSET, Xcpuast,
SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL));
/* install an inter-CPU IPI for CPU stop/restart */
setidt(XCPUSTOP_OFFSET, Xcpustop,
SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL));
#if defined(TEST_TEST1)
/* install a "fake hardware INTerrupt" vector */
setidt(XTEST1_OFFSET, Xtest1,
SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL));
#endif /** TEST_TEST1 */
#endif /* APIC_IO */
/* initialize all SMP locks */
init_locks();
/* start each Application Processor */
start_all_aps(boot_addr);
}
/*
* look for the MP spec signature
*/
/* string defined by the Intel MP Spec as identifying the MP table */
#define MP_SIG 0x5f504d5f /* _MP_ */
#define NEXT(X) ((X) += 4)
static int
search_for_sig(u_int32_t target, int count)
{
int x;
u_int32_t *addr = (u_int32_t *) (KERNBASE + target);
for (x = 0; x < count; NEXT(x))
if (addr[x] == MP_SIG)
/* make array index a byte index */
return (target + (x * sizeof(u_int32_t)));
return -1;
}
static basetable_entry basetable_entry_types[] =
{
{0, 20, "Processor"},
{1, 8, "Bus"},
{2, 8, "I/O APIC"},
{3, 8, "I/O INT"},
{4, 8, "Local INT"}
};
typedef struct BUSDATA {
u_char bus_id;
enum busTypes bus_type;
} bus_datum;
typedef struct INTDATA {
u_char int_type;
u_short int_flags;
u_char src_bus_id;
u_char src_bus_irq;
u_char dst_apic_id;
u_char dst_apic_int;
u_char int_vector;
} io_int, local_int;
typedef struct BUSTYPENAME {
u_char type;
char name[7];
} bus_type_name;
static bus_type_name bus_type_table[] =
{
{CBUS, "CBUS"},
{CBUSII, "CBUSII"},
{EISA, "EISA"},
{MCA, "MCA"},
{UNKNOWN_BUSTYPE, "---"},
{ISA, "ISA"},
{MCA, "MCA"},
{UNKNOWN_BUSTYPE, "---"},
{UNKNOWN_BUSTYPE, "---"},
{UNKNOWN_BUSTYPE, "---"},
{UNKNOWN_BUSTYPE, "---"},
{UNKNOWN_BUSTYPE, "---"},
{PCI, "PCI"},
{UNKNOWN_BUSTYPE, "---"},
{UNKNOWN_BUSTYPE, "---"},
{UNKNOWN_BUSTYPE, "---"},
{UNKNOWN_BUSTYPE, "---"},
{XPRESS, "XPRESS"},
{UNKNOWN_BUSTYPE, "---"}
};
/* from MP spec v1.4, table 5-1 */
static int default_data[7][5] =
{
/* nbus, id0, type0, id1, type1 */
{1, 0, ISA, 255, 255},
{1, 0, EISA, 255, 255},
{1, 0, EISA, 255, 255},
{1, 0, MCA, 255, 255},
{2, 0, ISA, 1, PCI},
{2, 0, EISA, 1, PCI},
{2, 0, MCA, 1, PCI}
};
/* the bus data */
static bus_datum *bus_data;
/* the IO INT data, one entry per possible APIC INTerrupt */
static io_int *io_apic_ints;
static int nintrs;
static int processor_entry(proc_entry_ptr entry, int cpu);
static int bus_entry(bus_entry_ptr entry, int bus);
static int io_apic_entry(io_apic_entry_ptr entry, int apic);
static int int_entry(int_entry_ptr entry, int intr);
static int lookup_bus_type(char *name);
/*
* 1st pass on motherboard's Intel MP specification table.
*
* initializes:
* mp_ncpus = 1
*
* determines:
* cpu_apic_address (common to all CPUs)
* io_apic_address[N]
* mp_naps
* mp_nbusses
* mp_napics
* nintrs
*/
static void
mptable_pass1(void)
{
int x;
mpcth_t cth;
int totalSize;
void* position;
int count;
int type;
POSTCODE(MPTABLE_PASS1_POST);
/* clear various tables */
for (x = 0; x < NAPICID; ++x) {
io_apic_address[x] = ~0; /* IO APIC address table */
}
/* init everything to empty */
mp_naps = 0;
mp_nbusses = 0;
mp_napics = 0;
nintrs = 0;
/* check for use of 'default' configuration */
if (MPFPS_MPFB1 != 0) {
/* use default addresses */
cpu_apic_address = DEFAULT_APIC_BASE;
io_apic_address[0] = DEFAULT_IO_APIC_BASE;
/* fill in with defaults */
mp_naps = 2; /* includes BSP */
mp_maxid = 1;
mp_nbusses = default_data[MPFPS_MPFB1 - 1][0];
#if defined(APIC_IO)
mp_napics = 1;
nintrs = 16;
#endif /* APIC_IO */
}
else {
if ((cth = mpfps->pap) == 0)
panic("MP Configuration Table Header MISSING!");
cpu_apic_address = (vm_offset_t) cth->apic_address;
/* walk the table, recording info of interest */
totalSize = cth->base_table_length - sizeof(struct MPCTH);
position = (u_char *) cth + sizeof(struct MPCTH);
count = cth->entry_count;
while (count--) {
switch (type = *(u_char *) position) {
case 0: /* processor_entry */
if (((proc_entry_ptr)position)->cpu_flags
& PROCENTRY_FLAG_EN) {
++mp_naps;
mp_maxid++;
}
break;
case 1: /* bus_entry */
++mp_nbusses;
break;
case 2: /* io_apic_entry */
if (((io_apic_entry_ptr)position)->apic_flags
& IOAPICENTRY_FLAG_EN)
io_apic_address[mp_napics++] =
(vm_offset_t)((io_apic_entry_ptr)
position)->apic_address;
break;
case 3: /* int_entry */
++nintrs;
break;
case 4: /* int_entry */
break;
default:
panic("mpfps Base Table HOSED!");
/* NOTREACHED */
}
totalSize -= basetable_entry_types[type].length;
(u_char*)position += basetable_entry_types[type].length;
}
}
/* qualify the numbers */
if (mp_naps > MAXCPU) {
printf("Warning: only using %d of %d available CPUs!\n",
MAXCPU, mp_naps);
mp_naps = MAXCPU;
}
/*
* Count the BSP.
* This is also used as a counter while starting the APs.
*/
mp_ncpus = 1;
--mp_naps; /* subtract the BSP */
}
/*
* 2nd pass on motherboard's Intel MP specification table.
*
* sets:
* boot_cpu_id
* ID_TO_IO(N), phy APIC ID to log CPU/IO table
* CPU_TO_ID(N), logical CPU to APIC ID table
* IO_TO_ID(N), logical IO to APIC ID table
* bus_data[N]
* io_apic_ints[N]
*/
static int
mptable_pass2(void)
{
int x;
mpcth_t cth;
int totalSize;
void* position;
int count;
int type;
int apic, bus, cpu, intr;
int i, j;
int pgeflag;
POSTCODE(MPTABLE_PASS2_POST);
pgeflag = 0; /* XXX - Not used under SMP yet. */
MALLOC(io_apic_versions, u_int32_t *, sizeof(u_int32_t) * mp_napics,
M_DEVBUF, M_WAITOK);
MALLOC(ioapic, volatile ioapic_t **, sizeof(ioapic_t *) * mp_napics,
M_DEVBUF, M_WAITOK);
MALLOC(io_apic_ints, io_int *, sizeof(io_int) * (nintrs + 1),
M_DEVBUF, M_WAITOK);
MALLOC(bus_data, bus_datum *, sizeof(bus_datum) * mp_nbusses,
M_DEVBUF, M_WAITOK);
bzero(ioapic, sizeof(ioapic_t *) * mp_napics);
for (i = 0; i < mp_napics; i++) {
for (j = 0; j < mp_napics; j++) {
/* same page frame as a previous IO apic? */
if (((vm_offset_t)SMPpt[NPTEPG-2-j] & PG_FRAME) ==
(io_apic_address[i] & PG_FRAME)) {
ioapic[i] = (ioapic_t *)((u_int)SMP_prvspace
+ (NPTEPG-2-j) * PAGE_SIZE
+ (io_apic_address[i] & PAGE_MASK));
break;
}
/* use this slot if available */
if (((vm_offset_t)SMPpt[NPTEPG-2-j] & PG_FRAME) == 0) {
SMPpt[NPTEPG-2-j] = (pt_entry_t)(PG_V | PG_RW |
pgeflag | (io_apic_address[i] & PG_FRAME));
ioapic[i] = (ioapic_t *)((u_int)SMP_prvspace
+ (NPTEPG-2-j) * PAGE_SIZE
+ (io_apic_address[i] & PAGE_MASK));
break;
}
}
}
/* clear various tables */
for (x = 0; x < NAPICID; ++x) {
ID_TO_IO(x) = -1; /* phy APIC ID to log CPU/IO table */
CPU_TO_ID(x) = -1; /* logical CPU to APIC ID table */
IO_TO_ID(x) = -1; /* logical IO to APIC ID table */
}
/* clear bus data table */
for (x = 0; x < mp_nbusses; ++x)
bus_data[x].bus_id = 0xff;
/* clear IO APIC INT table */
for (x = 0; x < (nintrs + 1); ++x) {
io_apic_ints[x].int_type = 0xff;
io_apic_ints[x].int_vector = 0xff;
}
/* setup the cpu/apic mapping arrays */
boot_cpu_id = -1;
/* record whether PIC or virtual-wire mode */
picmode = (mpfps->mpfb2 & 0x80) ? 1 : 0;
/* check for use of 'default' configuration */
if (MPFPS_MPFB1 != 0)
return MPFPS_MPFB1; /* return default configuration type */
if ((cth = mpfps->pap) == 0)
panic("MP Configuration Table Header MISSING!");
/* walk the table, recording info of interest */
totalSize = cth->base_table_length - sizeof(struct MPCTH);
position = (u_char *) cth + sizeof(struct MPCTH);
count = cth->entry_count;
apic = bus = intr = 0;
cpu = 1; /* pre-count the BSP */
while (count--) {
switch (type = *(u_char *) position) {
case 0:
if (processor_entry(position, cpu))
++cpu;
break;
case 1:
if (bus_entry(position, bus))
++bus;
break;
case 2:
if (io_apic_entry(position, apic))
++apic;
break;
case 3:
if (int_entry(position, intr))
++intr;
break;
case 4:
/* int_entry(position); */
break;
default:
panic("mpfps Base Table HOSED!");
/* NOTREACHED */
}
totalSize -= basetable_entry_types[type].length;
(u_char *) position += basetable_entry_types[type].length;
}
if (boot_cpu_id == -1)
panic("NO BSP found!");
/* report fact that its NOT a default configuration */
return 0;
}
void
assign_apic_irq(int apic, int intpin, int irq)
{
int x;
if (int_to_apicintpin[irq].ioapic != -1)
panic("assign_apic_irq: inconsistent table");
int_to_apicintpin[irq].ioapic = apic;
int_to_apicintpin[irq].int_pin = intpin;
int_to_apicintpin[irq].apic_address = ioapic[apic];
int_to_apicintpin[irq].redirindex = IOAPIC_REDTBL + 2 * intpin;
for (x = 0; x < nintrs; x++) {
if ((io_apic_ints[x].int_type == 0 ||
io_apic_ints[x].int_type == 3) &&
io_apic_ints[x].int_vector == 0xff &&
io_apic_ints[x].dst_apic_id == IO_TO_ID(apic) &&
io_apic_ints[x].dst_apic_int == intpin)
io_apic_ints[x].int_vector = irq;
}
}
void
revoke_apic_irq(int irq)
{
int x;
int oldapic;
int oldintpin;
if (int_to_apicintpin[irq].ioapic == -1)
panic("revoke_apic_irq: inconsistent table");
oldapic = int_to_apicintpin[irq].ioapic;
oldintpin = int_to_apicintpin[irq].int_pin;
int_to_apicintpin[irq].ioapic = -1;
int_to_apicintpin[irq].int_pin = 0;
int_to_apicintpin[irq].apic_address = NULL;
int_to_apicintpin[irq].redirindex = 0;
for (x = 0; x < nintrs; x++) {
if ((io_apic_ints[x].int_type == 0 ||
io_apic_ints[x].int_type == 3) &&
io_apic_ints[x].int_vector != 0xff &&
io_apic_ints[x].dst_apic_id == IO_TO_ID(oldapic) &&
io_apic_ints[x].dst_apic_int == oldintpin)
io_apic_ints[x].int_vector = 0xff;
}
}
static void
allocate_apic_irq(int intr)
{
int apic;
int intpin;
int irq;
if (io_apic_ints[intr].int_vector != 0xff)
return; /* Interrupt handler already assigned */
if (io_apic_ints[intr].int_type != 0 &&
(io_apic_ints[intr].int_type != 3 ||
(io_apic_ints[intr].dst_apic_id == IO_TO_ID(0) &&
io_apic_ints[intr].dst_apic_int == 0)))
return; /* Not INT or ExtInt on != (0, 0) */
irq = 0;
while (irq < APIC_INTMAPSIZE &&
int_to_apicintpin[irq].ioapic != -1)
irq++;
if (irq >= APIC_INTMAPSIZE)
return; /* No free interrupt handlers */
apic = ID_TO_IO(io_apic_ints[intr].dst_apic_id);
intpin = io_apic_ints[intr].dst_apic_int;
assign_apic_irq(apic, intpin, irq);
io_apic_setup_intpin(apic, intpin);
}
static void
swap_apic_id(int apic, int oldid, int newid)
{
int x;
int oapic;
if (oldid == newid)
return; /* Nothing to do */
printf("Changing APIC ID for IO APIC #%d from %d to %d in MP table\n",
apic, oldid, newid);
/* Swap physical APIC IDs in interrupt entries */
for (x = 0; x < nintrs; x++) {
if (io_apic_ints[x].dst_apic_id == oldid)
io_apic_ints[x].dst_apic_id = newid;
else if (io_apic_ints[x].dst_apic_id == newid)
io_apic_ints[x].dst_apic_id = oldid;
}
/* Swap physical APIC IDs in IO_TO_ID mappings */
for (oapic = 0; oapic < mp_napics; oapic++)
if (IO_TO_ID(oapic) == newid)
break;
if (oapic < mp_napics) {
printf("Changing APIC ID for IO APIC #%d from "
"%d to %d in MP table\n",
oapic, newid, oldid);
IO_TO_ID(oapic) = oldid;
}
IO_TO_ID(apic) = newid;
}
static void
fix_id_to_io_mapping(void)
{
int x;
for (x = 0; x < NAPICID; x++)
ID_TO_IO(x) = -1;
for (x = 0; x <= mp_naps; x++)
if (CPU_TO_ID(x) < NAPICID)
ID_TO_IO(CPU_TO_ID(x)) = x;
for (x = 0; x < mp_napics; x++)
if (IO_TO_ID(x) < NAPICID)
ID_TO_IO(IO_TO_ID(x)) = x;
}
static int
first_free_apic_id(void)
{
int freeid, x;
for (freeid = 0; freeid < NAPICID; freeid++) {
for (x = 0; x <= mp_naps; x++)
if (CPU_TO_ID(x) == freeid)
break;
if (x <= mp_naps)
continue;
for (x = 0; x < mp_napics; x++)
if (IO_TO_ID(x) == freeid)
break;
if (x < mp_napics)
continue;
return freeid;
}
return freeid;
}
static int
io_apic_id_acceptable(int apic, int id)
{
int cpu; /* Logical CPU number */
int oapic; /* Logical IO APIC number for other IO APIC */
if (id >= NAPICID)
return 0; /* Out of range */
for (cpu = 0; cpu <= mp_naps; cpu++)
if (CPU_TO_ID(cpu) == id)
return 0; /* Conflict with CPU */
for (oapic = 0; oapic < mp_napics && oapic < apic; oapic++)
if (IO_TO_ID(oapic) == id)
return 0; /* Conflict with other APIC */
return 1; /* ID is acceptable for IO APIC */
}
/*
* parse an Intel MP specification table
*/
static void
fix_mp_table(void)
{
int x;
int id;
1999-01-12 00:19:33 +00:00
int bus_0 = 0; /* Stop GCC warning */
int bus_pci = 0; /* Stop GCC warning */
int num_pci_bus;
int apic; /* IO APIC unit number */
int freeid; /* Free physical APIC ID */
int physid; /* Current physical IO APIC ID */
/*
* Fix mis-numbering of the PCI bus and its INT entries if the BIOS
* did it wrong. The MP spec says that when more than 1 PCI bus
* exists the BIOS must begin with bus entries for the PCI bus and use
* actual PCI bus numbering. This implies that when only 1 PCI bus
* exists the BIOS can choose to ignore this ordering, and indeed many
* MP motherboards do ignore it. This causes a problem when the PCI
* sub-system makes requests of the MP sub-system based on PCI bus
* numbers. So here we look for the situation and renumber the
* busses and associated INTs in an effort to "make it right".
*/
/* find bus 0, PCI bus, count the number of PCI busses */
for (num_pci_bus = 0, x = 0; x < mp_nbusses; ++x) {
if (bus_data[x].bus_id == 0) {
bus_0 = x;
}
if (bus_data[x].bus_type == PCI) {
++num_pci_bus;
bus_pci = x;
}
}
/*
* bus_0 == slot of bus with ID of 0
* bus_pci == slot of last PCI bus encountered
*/
/* check the 1 PCI bus case for sanity */
/* if it is number 0 all is well */
if (num_pci_bus == 1 &&
bus_data[bus_pci].bus_id != 0) {
/* mis-numbered, swap with whichever bus uses slot 0 */
/* swap the bus entry types */
bus_data[bus_pci].bus_type = bus_data[bus_0].bus_type;
bus_data[bus_0].bus_type = PCI;
/* swap each relavant INTerrupt entry */
id = bus_data[bus_pci].bus_id;
for (x = 0; x < nintrs; ++x) {
if (io_apic_ints[x].src_bus_id == id) {
io_apic_ints[x].src_bus_id = 0;
}
else if (io_apic_ints[x].src_bus_id == 0) {
io_apic_ints[x].src_bus_id = id;
}
}
}
/* Assign IO APIC IDs.
*
* First try the existing ID. If a conflict is detected, try
* the ID in the MP table. If a conflict is still detected, find
* a free id.
*
* We cannot use the ID_TO_IO table before all conflicts has been
* resolved and the table has been corrected.
*/
for (apic = 0; apic < mp_napics; ++apic) { /* For all IO APICs */
/* First try to use the value set by the BIOS */
physid = io_apic_get_id(apic);
if (io_apic_id_acceptable(apic, physid)) {
if (IO_TO_ID(apic) != physid)
swap_apic_id(apic, IO_TO_ID(apic), physid);
continue;
}
/* Then check if the value in the MP table is acceptable */
if (io_apic_id_acceptable(apic, IO_TO_ID(apic)))
continue;
/* Last resort, find a free APIC ID and use it */
freeid = first_free_apic_id();
if (freeid >= NAPICID)
panic("No free physical APIC IDs found");
if (io_apic_id_acceptable(apic, freeid)) {
swap_apic_id(apic, IO_TO_ID(apic), freeid);
continue;
}
panic("Free physical APIC ID not usable");
}
fix_id_to_io_mapping();
/* detect and fix broken Compaq MP table */
if (apic_int_type(0, 0) == -1) {
printf("APIC_IO: MP table broken: 8259->APIC entry missing!\n");
io_apic_ints[nintrs].int_type = 3; /* ExtInt */
io_apic_ints[nintrs].int_vector = 0xff; /* Unassigned */
/* XXX fixme, set src bus id etc, but it doesn't seem to hurt */
io_apic_ints[nintrs].dst_apic_id = IO_TO_ID(0);
io_apic_ints[nintrs].dst_apic_int = 0; /* Pin 0 */
nintrs++;
}
}
/* Assign low level interrupt handlers */
static void
setup_apic_irq_mapping(void)
{
int x;
int int_vector;
/* Clear array */
for (x = 0; x < APIC_INTMAPSIZE; x++) {
int_to_apicintpin[x].ioapic = -1;
int_to_apicintpin[x].int_pin = 0;
int_to_apicintpin[x].apic_address = NULL;
int_to_apicintpin[x].redirindex = 0;
}
/* First assign ISA/EISA interrupts */
for (x = 0; x < nintrs; x++) {
int_vector = io_apic_ints[x].src_bus_irq;
if (int_vector < APIC_INTMAPSIZE &&
io_apic_ints[x].int_vector == 0xff &&
int_to_apicintpin[int_vector].ioapic == -1 &&
(apic_int_is_bus_type(x, ISA) ||
apic_int_is_bus_type(x, EISA)) &&
io_apic_ints[x].int_type == 0) {
assign_apic_irq(ID_TO_IO(io_apic_ints[x].dst_apic_id),
io_apic_ints[x].dst_apic_int,
int_vector);
}
}
/* Assign ExtInt entry if no ISA/EISA interrupt 0 entry */
for (x = 0; x < nintrs; x++) {
if (io_apic_ints[x].dst_apic_int == 0 &&
io_apic_ints[x].dst_apic_id == IO_TO_ID(0) &&
io_apic_ints[x].int_vector == 0xff &&
int_to_apicintpin[0].ioapic == -1 &&
io_apic_ints[x].int_type == 3) {
assign_apic_irq(0, 0, 0);
break;
}
}
/* PCI interrupt assignment is deferred */
}
static int
processor_entry(proc_entry_ptr entry, int cpu)
{
/* check for usability */
if (!(entry->cpu_flags & PROCENTRY_FLAG_EN))
return 0;
if(entry->apic_id >= NAPICID)
panic("CPU APIC ID out of range (0..%d)", NAPICID - 1);
/* check for BSP flag */
if (entry->cpu_flags & PROCENTRY_FLAG_BP) {
boot_cpu_id = entry->apic_id;
CPU_TO_ID(0) = entry->apic_id;
ID_TO_CPU(entry->apic_id) = 0;
return 0; /* its already been counted */
}
/* add another AP to list, if less than max number of CPUs */
else if (cpu < MAXCPU) {
CPU_TO_ID(cpu) = entry->apic_id;
ID_TO_CPU(entry->apic_id) = cpu;
return 1;
}
return 0;
}
static int
bus_entry(bus_entry_ptr entry, int bus)
{
int x;
char c, name[8];
/* encode the name into an index */
for (x = 0; x < 6; ++x) {
if ((c = entry->bus_type[x]) == ' ')
break;
name[x] = c;
}
name[x] = '\0';
if ((x = lookup_bus_type(name)) == UNKNOWN_BUSTYPE)
panic("unknown bus type: '%s'", name);
bus_data[bus].bus_id = entry->bus_id;
bus_data[bus].bus_type = x;
return 1;
}
static int
io_apic_entry(io_apic_entry_ptr entry, int apic)
{
if (!(entry->apic_flags & IOAPICENTRY_FLAG_EN))
return 0;
IO_TO_ID(apic) = entry->apic_id;
if (entry->apic_id < NAPICID)
ID_TO_IO(entry->apic_id) = apic;
return 1;
}
static int
lookup_bus_type(char *name)
{
int x;
for (x = 0; x < MAX_BUSTYPE; ++x)
if (strcmp(bus_type_table[x].name, name) == 0)
return bus_type_table[x].type;
return UNKNOWN_BUSTYPE;
}
static int
int_entry(int_entry_ptr entry, int intr)
{
int apic;
io_apic_ints[intr].int_type = entry->int_type;
io_apic_ints[intr].int_flags = entry->int_flags;
io_apic_ints[intr].src_bus_id = entry->src_bus_id;
io_apic_ints[intr].src_bus_irq = entry->src_bus_irq;
if (entry->dst_apic_id == 255) {
/* This signal goes to all IO APICS. Select an IO APIC
with sufficient number of interrupt pins */
for (apic = 0; apic < mp_napics; apic++)
if (((io_apic_read(apic, IOAPIC_VER) &
IOART_VER_MAXREDIR) >> MAXREDIRSHIFT) >=
entry->dst_apic_int)
break;
if (apic < mp_napics)
io_apic_ints[intr].dst_apic_id = IO_TO_ID(apic);
else
io_apic_ints[intr].dst_apic_id = entry->dst_apic_id;
} else
io_apic_ints[intr].dst_apic_id = entry->dst_apic_id;
io_apic_ints[intr].dst_apic_int = entry->dst_apic_int;
return 1;
}
static int
apic_int_is_bus_type(int intr, int bus_type)
{
int bus;
for (bus = 0; bus < mp_nbusses; ++bus)
if ((bus_data[bus].bus_id == io_apic_ints[intr].src_bus_id)
&& ((int) bus_data[bus].bus_type == bus_type))
return 1;
return 0;
}
/*
* Given a traditional ISA INT mask, return an APIC mask.
*/
u_int
isa_apic_mask(u_int isa_mask)
{
int isa_irq;
int apic_pin;
#if defined(SKIP_IRQ15_REDIRECT)
if (isa_mask == (1 << 15)) {
printf("skipping ISA IRQ15 redirect\n");
return isa_mask;
}
#endif /* SKIP_IRQ15_REDIRECT */
isa_irq = ffs(isa_mask); /* find its bit position */
if (isa_irq == 0) /* doesn't exist */
return 0;
--isa_irq; /* make it zero based */
apic_pin = isa_apic_irq(isa_irq); /* look for APIC connection */
if (apic_pin == -1)
return 0;
return (1 << apic_pin); /* convert pin# to a mask */
}
/*
* Determine which APIC pin an ISA/EISA INT is attached to.
*/
#define INTTYPE(I) (io_apic_ints[(I)].int_type)
#define INTPIN(I) (io_apic_ints[(I)].dst_apic_int)
#define INTIRQ(I) (io_apic_ints[(I)].int_vector)
#define INTAPIC(I) (ID_TO_IO(io_apic_ints[(I)].dst_apic_id))
#define SRCBUSIRQ(I) (io_apic_ints[(I)].src_bus_irq)
int
isa_apic_irq(int isa_irq)
{
int intr;
for (intr = 0; intr < nintrs; ++intr) { /* check each record */
if (INTTYPE(intr) == 0) { /* standard INT */
if (SRCBUSIRQ(intr) == isa_irq) {
if (apic_int_is_bus_type(intr, ISA) ||
apic_int_is_bus_type(intr, EISA)) {
if (INTIRQ(intr) == 0xff)
return -1; /* unassigned */
return INTIRQ(intr); /* found */
}
}
}
}
return -1; /* NOT found */
}
/*
* Determine which APIC pin a PCI INT is attached to.
*/
#define SRCBUSID(I) (io_apic_ints[(I)].src_bus_id)
#define SRCBUSDEVICE(I) ((io_apic_ints[(I)].src_bus_irq >> 2) & 0x1f)
#define SRCBUSLINE(I) (io_apic_ints[(I)].src_bus_irq & 0x03)
int
pci_apic_irq(int pciBus, int pciDevice, int pciInt)
{
int intr;
--pciInt; /* zero based */
for (intr = 0; intr < nintrs; ++intr) /* check each record */
if ((INTTYPE(intr) == 0) /* standard INT */
&& (SRCBUSID(intr) == pciBus)
&& (SRCBUSDEVICE(intr) == pciDevice)
&& (SRCBUSLINE(intr) == pciInt)) /* a candidate IRQ */
if (apic_int_is_bus_type(intr, PCI)) {
if (INTIRQ(intr) == 0xff)
allocate_apic_irq(intr);
if (INTIRQ(intr) == 0xff)
return -1; /* unassigned */
return INTIRQ(intr); /* exact match */
}
return -1; /* NOT found */
}
int
next_apic_irq(int irq)
{
int intr, ointr;
int bus, bustype;
bus = 0;
bustype = 0;
for (intr = 0; intr < nintrs; intr++) {
if (INTIRQ(intr) != irq || INTTYPE(intr) != 0)
continue;
bus = SRCBUSID(intr);
bustype = apic_bus_type(bus);
if (bustype != ISA &&
bustype != EISA &&
bustype != PCI)
continue;
break;
}
if (intr >= nintrs) {
return -1;
}
for (ointr = intr + 1; ointr < nintrs; ointr++) {
if (INTTYPE(ointr) != 0)
continue;
if (bus != SRCBUSID(ointr))
continue;
if (bustype == PCI) {
if (SRCBUSDEVICE(intr) != SRCBUSDEVICE(ointr))
continue;
if (SRCBUSLINE(intr) != SRCBUSLINE(ointr))
continue;
}
if (bustype == ISA || bustype == EISA) {
if (SRCBUSIRQ(intr) != SRCBUSIRQ(ointr))
continue;
}
if (INTPIN(intr) == INTPIN(ointr))
continue;
break;
}
if (ointr >= nintrs) {
return -1;
}
return INTIRQ(ointr);
}
#undef SRCBUSLINE
#undef SRCBUSDEVICE
#undef SRCBUSID
#undef SRCBUSIRQ
#undef INTPIN
#undef INTIRQ
#undef INTAPIC
#undef INTTYPE
/*
* Reprogram the MB chipset to NOT redirect an ISA INTerrupt.
*
* XXX FIXME:
* Exactly what this means is unclear at this point. It is a solution
* for motherboards that redirect the MBIRQ0 pin. Generically a motherboard
* could route any of the ISA INTs to upper (>15) IRQ values. But most would
* NOT be redirected via MBIRQ0, thus "undirect()ing" them would NOT be an
* option.
*/
int
undirect_isa_irq(int rirq)
{
#if defined(READY)
if (bootverbose)
printf("Freeing redirected ISA irq %d.\n", rirq);
/** FIXME: tickle the MB redirector chip */
return -1;
#else
if (bootverbose)
printf("Freeing (NOT implemented) redirected ISA irq %d.\n", rirq);
return 0;
#endif /* READY */
}
/*
* Reprogram the MB chipset to NOT redirect a PCI INTerrupt
*/
int
undirect_pci_irq(int rirq)
{
#if defined(READY)
if (bootverbose)
printf("Freeing redirected PCI irq %d.\n", rirq);
/** FIXME: tickle the MB redirector chip */
return -1;
#else
if (bootverbose)
printf("Freeing (NOT implemented) redirected PCI irq %d.\n",
rirq);
return 0;
#endif /* READY */
}
/*
* given a bus ID, return:
* the bus type if found
* -1 if NOT found
*/
int
apic_bus_type(int id)
{
int x;
for (x = 0; x < mp_nbusses; ++x)
if (bus_data[x].bus_id == id)
return bus_data[x].bus_type;
return -1;
}
/*
* given a LOGICAL APIC# and pin#, return:
* the associated src bus ID if found
* -1 if NOT found
*/
int
apic_src_bus_id(int apic, int pin)
{
int x;
/* search each of the possible INTerrupt sources */
for (x = 0; x < nintrs; ++x)
if ((apic == ID_TO_IO(io_apic_ints[x].dst_apic_id)) &&
(pin == io_apic_ints[x].dst_apic_int))
return (io_apic_ints[x].src_bus_id);
return -1; /* NOT found */
}
/*
* given a LOGICAL APIC# and pin#, return:
* the associated src bus IRQ if found
* -1 if NOT found
*/
int
apic_src_bus_irq(int apic, int pin)
{
int x;
for (x = 0; x < nintrs; x++)
if ((apic == ID_TO_IO(io_apic_ints[x].dst_apic_id)) &&
(pin == io_apic_ints[x].dst_apic_int))
return (io_apic_ints[x].src_bus_irq);
return -1; /* NOT found */
}
/*
* given a LOGICAL APIC# and pin#, return:
* the associated INTerrupt type if found
* -1 if NOT found
*/
int
apic_int_type(int apic, int pin)
{
int x;
/* search each of the possible INTerrupt sources */
for (x = 0; x < nintrs; ++x)
if ((apic == ID_TO_IO(io_apic_ints[x].dst_apic_id)) &&
(pin == io_apic_ints[x].dst_apic_int))
return (io_apic_ints[x].int_type);
return -1; /* NOT found */
}
int
apic_irq(int apic, int pin)
{
int x;
int res;
for (x = 0; x < nintrs; ++x)
if ((apic == ID_TO_IO(io_apic_ints[x].dst_apic_id)) &&
(pin == io_apic_ints[x].dst_apic_int)) {
res = io_apic_ints[x].int_vector;
if (res == 0xff)
return -1;
if (apic != int_to_apicintpin[res].ioapic)
panic("apic_irq: inconsistent table");
if (pin != int_to_apicintpin[res].int_pin)
panic("apic_irq inconsistent table (2)");
return res;
}
return -1;
}
/*
* given a LOGICAL APIC# and pin#, return:
* the associated trigger mode if found
* -1 if NOT found
*/
int
apic_trigger(int apic, int pin)
{
int x;
/* search each of the possible INTerrupt sources */
for (x = 0; x < nintrs; ++x)
if ((apic == ID_TO_IO(io_apic_ints[x].dst_apic_id)) &&
(pin == io_apic_ints[x].dst_apic_int))
return ((io_apic_ints[x].int_flags >> 2) & 0x03);
return -1; /* NOT found */
}
/*
* given a LOGICAL APIC# and pin#, return:
* the associated 'active' level if found
* -1 if NOT found
*/
int
apic_polarity(int apic, int pin)
{
int x;
/* search each of the possible INTerrupt sources */
for (x = 0; x < nintrs; ++x)
if ((apic == ID_TO_IO(io_apic_ints[x].dst_apic_id)) &&
(pin == io_apic_ints[x].dst_apic_int))
return (io_apic_ints[x].int_flags & 0x03);
return -1; /* NOT found */
}
/*
* set data according to MP defaults
* FIXME: probably not complete yet...
*/
static void
default_mp_table(int type)
{
int ap_cpu_id;
#if defined(APIC_IO)
int io_apic_id;
int pin;
#endif /* APIC_IO */
#if 0
printf(" MP default config type: %d\n", type);
switch (type) {
case 1:
printf(" bus: ISA, APIC: 82489DX\n");
break;
case 2:
printf(" bus: EISA, APIC: 82489DX\n");
break;
case 3:
printf(" bus: EISA, APIC: 82489DX\n");
break;
case 4:
printf(" bus: MCA, APIC: 82489DX\n");
break;
case 5:
printf(" bus: ISA+PCI, APIC: Integrated\n");
break;
case 6:
printf(" bus: EISA+PCI, APIC: Integrated\n");
break;
case 7:
printf(" bus: MCA+PCI, APIC: Integrated\n");
break;
default:
printf(" future type\n");
break;
/* NOTREACHED */
}
#endif /* 0 */
boot_cpu_id = (lapic.id & APIC_ID_MASK) >> 24;
ap_cpu_id = (boot_cpu_id == 0) ? 1 : 0;
/* BSP */
CPU_TO_ID(0) = boot_cpu_id;
ID_TO_CPU(boot_cpu_id) = 0;
/* one and only AP */
CPU_TO_ID(1) = ap_cpu_id;
ID_TO_CPU(ap_cpu_id) = 1;
#if defined(APIC_IO)
/* one and only IO APIC */
io_apic_id = (io_apic_read(0, IOAPIC_ID) & APIC_ID_MASK) >> 24;
/*
* sanity check, refer to MP spec section 3.6.6, last paragraph
* necessary as some hardware isn't properly setting up the IO APIC
*/
#if defined(REALLY_ANAL_IOAPICID_VALUE)
if (io_apic_id != 2) {
#else
if ((io_apic_id == 0) || (io_apic_id == 1) || (io_apic_id == 15)) {
#endif /* REALLY_ANAL_IOAPICID_VALUE */
io_apic_set_id(0, 2);
io_apic_id = 2;
}
IO_TO_ID(0) = io_apic_id;
ID_TO_IO(io_apic_id) = 0;
#endif /* APIC_IO */
/* fill out bus entries */
switch (type) {
case 1:
case 2:
case 3:
case 4:
case 5:
case 6:
case 7:
bus_data[0].bus_id = default_data[type - 1][1];
bus_data[0].bus_type = default_data[type - 1][2];
bus_data[1].bus_id = default_data[type - 1][3];
bus_data[1].bus_type = default_data[type - 1][4];
break;
/* case 4: case 7: MCA NOT supported */
default: /* illegal/reserved */
panic("BAD default MP config: %d", type);
/* NOTREACHED */
}
#if defined(APIC_IO)
/* general cases from MP v1.4, table 5-2 */
for (pin = 0; pin < 16; ++pin) {
io_apic_ints[pin].int_type = 0;
io_apic_ints[pin].int_flags = 0x05; /* edge/active-hi */
io_apic_ints[pin].src_bus_id = 0;
io_apic_ints[pin].src_bus_irq = pin; /* IRQ2 caught below */
io_apic_ints[pin].dst_apic_id = io_apic_id;
io_apic_ints[pin].dst_apic_int = pin; /* 1-to-1 */
}
/* special cases from MP v1.4, table 5-2 */
if (type == 2) {
io_apic_ints[2].int_type = 0xff; /* N/C */
io_apic_ints[13].int_type = 0xff; /* N/C */
#if !defined(APIC_MIXED_MODE)
/** FIXME: ??? */
panic("sorry, can't support type 2 default yet");
#endif /* APIC_MIXED_MODE */
}
else
io_apic_ints[2].src_bus_irq = 0; /* ISA IRQ0 is on APIC INT 2 */
if (type == 7)
io_apic_ints[0].int_type = 0xff; /* N/C */
else
io_apic_ints[0].int_type = 3; /* vectored 8259 */
#endif /* APIC_IO */
}
/*
* start each AP in our list
*/
static int
start_all_aps(u_int boot_addr)
{
int x, i, pg;
u_char mpbiosreason;
u_long mpbioswarmvec;
struct pcpu *pc;
char *stack;
uintptr_t kptbase;
POSTCODE(START_ALL_APS_POST);
mtx_init(&ap_boot_mtx, "ap boot", NULL, MTX_SPIN);
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
/* initialize BSP's local APIC */
apic_initialize();
bsp_apic_ready = 1;
/* install the AP 1st level boot code */
install_ap_tramp(boot_addr);
/* save the current value of the warm-start vector */
mpbioswarmvec = *((u_long *) WARMBOOT_OFF);
#ifndef PC98
outb(CMOS_REG, BIOS_RESET);
mpbiosreason = inb(CMOS_DATA);
1998-10-10 09:38:02 +00:00
#endif
/* set up temporary P==V mapping for AP boot */
/* XXX this is a hack, we should boot the AP on its own stack/PTD */
kptbase = (uintptr_t)(void *)KPTphys;
for (x = 0; x < NKPT; x++)
PTD[x] = (pd_entry_t)(PG_V | PG_RW |
((kptbase + x * PAGE_SIZE) & PG_FRAME));
invltlb();
/* start each AP */
for (x = 1; x <= mp_naps; ++x) {
/* This is a bit verbose, it will go away soon. */
/* first page of AP's private space */
pg = x * i386_btop(sizeof(struct privatespace));
/* allocate a new private data page */
pc = (struct pcpu *)kmem_alloc(kernel_map, PAGE_SIZE);
/* wire it into the private page table page */
SMPpt[pg] = (pt_entry_t)(PG_V | PG_RW | vtophys(pc));
/* allocate and set up an idle stack data page */
stack = (char *)kmem_alloc(kernel_map, KSTACK_PAGES * PAGE_SIZE); /* XXXKSE */
for (i = 0; i < KSTACK_PAGES; i++)
SMPpt[pg + 1 + i] = (pt_entry_t)
(PG_V | PG_RW | vtophys(PAGE_SIZE * i + stack));
/* prime data page for it to use */
pcpu_init(pc, x, sizeof(struct pcpu));
/* setup a vector to our boot code */
*((volatile u_short *) WARMBOOT_OFF) = WARMBOOT_TARGET;
*((volatile u_short *) WARMBOOT_SEG) = (boot_addr >> 4);
#ifndef PC98
outb(CMOS_REG, BIOS_RESET);
outb(CMOS_DATA, BIOS_WARM); /* 'warm-start' */
1998-10-10 09:38:02 +00:00
#endif
bootSTK = &SMP_prvspace[x].idlekstack[KSTACK_PAGES * PAGE_SIZE];
bootAP = x;
/* attempt to start the Application Processor */
CHECK_INIT(99); /* setup checkpoints */
if (!start_ap(x, boot_addr)) {
printf("AP #%d (PHY# %d) failed!\n", x, CPU_TO_ID(x));
CHECK_PRINT("trace"); /* show checkpoints */
/* better panic as the AP may be running loose */
printf("panic y/n? [y] ");
if (cngetc() != 'n')
panic("bye-bye");
}
CHECK_PRINT("trace"); /* show checkpoints */
/* record its version info */
cpu_apic_versions[x] = cpu_apic_versions[0];
all_cpus |= (1 << x); /* record AP in CPU map */
}
/* build our map of 'other' CPUs */
PCPU_SET(other_cpus, all_cpus & ~PCPU_GET(cpumask));
/* fill in our (BSP) APIC version */
cpu_apic_versions[0] = lapic.version;
/* restore the warmstart vector */
*(u_long *) WARMBOOT_OFF = mpbioswarmvec;
#ifndef PC98
outb(CMOS_REG, BIOS_RESET);
outb(CMOS_DATA, mpbiosreason);
1998-10-10 09:38:02 +00:00
#endif
/*
* Set up the idle context for the BSP. Similar to above except
* that some was done by locore, some by pmap.c and some is implicit
* because the BSP is cpu#0 and the page is initially zero, and also
* because we can refer to variables by name on the BSP..
*/
/* Allocate and setup BSP idle stack */
stack = (char *)kmem_alloc(kernel_map, KSTACK_PAGES * PAGE_SIZE);
for (i = 0; i < KSTACK_PAGES; i++)
SMPpt[1 + i] = (pt_entry_t)
(PG_V | PG_RW | vtophys(PAGE_SIZE * i + stack));
for (x = 0; x < NKPT; x++)
PTD[x] = 0;
pmap_set_opt();
/* number of APs actually started */
return mp_ncpus - 1;
}
/*
* load the 1st level AP boot code into base memory.
*/
/* targets for relocation */
extern void bigJump(void);
extern void bootCodeSeg(void);
extern void bootDataSeg(void);
extern void MPentry(void);
extern u_int MP_GDT;
extern u_int mp_gdtbase;
static void
install_ap_tramp(u_int boot_addr)
{
int x;
int size = *(int *) ((u_long) & bootMP_size);
u_char *src = (u_char *) ((u_long) bootMP);
u_char *dst = (u_char *) boot_addr + KERNBASE;
u_int boot_base = (u_int) bootMP;
u_int8_t *dst8;
u_int16_t *dst16;
u_int32_t *dst32;
POSTCODE(INSTALL_AP_TRAMP_POST);
for (x = 0; x < size; ++x)
*dst++ = *src++;
/*
* modify addresses in code we just moved to basemem. unfortunately we
* need fairly detailed info about mpboot.s for this to work. changes
* to mpboot.s might require changes here.
*/
/* boot code is located in KERNEL space */
dst = (u_char *) boot_addr + KERNBASE;
/* modify the lgdt arg */
dst32 = (u_int32_t *) (dst + ((u_int) & mp_gdtbase - boot_base));
*dst32 = boot_addr + ((u_int) & MP_GDT - boot_base);
/* modify the ljmp target for MPentry() */
dst32 = (u_int32_t *) (dst + ((u_int) bigJump - boot_base) + 1);
*dst32 = ((u_int) MPentry - KERNBASE);
/* modify the target for boot code segment */
dst16 = (u_int16_t *) (dst + ((u_int) bootCodeSeg - boot_base));
dst8 = (u_int8_t *) (dst16 + 1);
*dst16 = (u_int) boot_addr & 0xffff;
*dst8 = ((u_int) boot_addr >> 16) & 0xff;
/* modify the target for boot data segment */
dst16 = (u_int16_t *) (dst + ((u_int) bootDataSeg - boot_base));
dst8 = (u_int8_t *) (dst16 + 1);
*dst16 = (u_int) boot_addr & 0xffff;
*dst8 = ((u_int) boot_addr >> 16) & 0xff;
}
/*
* this function starts the AP (application processor) identified
* by the APIC ID 'physicalCpu'. It does quite a "song and dance"
* to accomplish this. This is necessary because of the nuances
* of the different hardware we might encounter. It ain't pretty,
* but it seems to work.
*/
static int
start_ap(int logical_cpu, u_int boot_addr)
{
int physical_cpu;
int vector;
int cpus;
u_long icr_lo, icr_hi;
POSTCODE(START_AP_POST);
/* get the PHYSICAL APIC ID# */
physical_cpu = CPU_TO_ID(logical_cpu);
/* calculate the vector */
vector = (boot_addr >> 12) & 0xff;
/* used as a watchpoint to signal AP startup */
cpus = mp_ncpus;
/*
* first we do an INIT/RESET IPI this INIT IPI might be run, reseting
* and running the target CPU. OR this INIT IPI might be latched (P5
* bug), CPU waiting for STARTUP IPI. OR this INIT IPI might be
* ignored.
*/
/* setup the address for the target AP */
icr_hi = lapic.icr_hi & ~APIC_ID_MASK;
icr_hi |= (physical_cpu << 24);
lapic.icr_hi = icr_hi;
/* do an INIT IPI: assert RESET */
icr_lo = lapic.icr_lo & 0xfff00000;
lapic.icr_lo = icr_lo | 0x0000c500;
/* wait for pending status end */
while (lapic.icr_lo & APIC_DELSTAT_MASK)
/* spin */ ;
/* do an INIT IPI: deassert RESET */
lapic.icr_lo = icr_lo | 0x00008500;
/* wait for pending status end */
u_sleep(10000); /* wait ~10mS */
while (lapic.icr_lo & APIC_DELSTAT_MASK)
/* spin */ ;
/*
* next we do a STARTUP IPI: the previous INIT IPI might still be
* latched, (P5 bug) this 1st STARTUP would then terminate
* immediately, and the previously started INIT IPI would continue. OR
* the previous INIT IPI has already run. and this STARTUP IPI will
* run. OR the previous INIT IPI was ignored. and this STARTUP IPI
* will run.
*/
/* do a STARTUP IPI */
lapic.icr_lo = icr_lo | 0x00000600 | vector;
while (lapic.icr_lo & APIC_DELSTAT_MASK)
/* spin */ ;
u_sleep(200); /* wait ~200uS */
/*
* finally we do a 2nd STARTUP IPI: this 2nd STARTUP IPI should run IF
* the previous STARTUP IPI was cancelled by a latched INIT IPI. OR
* this STARTUP IPI will be ignored, as only ONE STARTUP IPI is
* recognized after hardware RESET or INIT IPI.
*/
lapic.icr_lo = icr_lo | 0x00000600 | vector;
while (lapic.icr_lo & APIC_DELSTAT_MASK)
/* spin */ ;
u_sleep(200); /* wait ~200uS */
/* wait for it to start */
set_apic_timer(5000000);/* == 5 seconds */
while (read_apic_timer())
if (mp_ncpus > cpus)
return 1; /* return SUCCESS */
return 0; /* return FAILURE */
}
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
#if defined(APIC_IO)
#ifdef COUNT_XINVLTLB_HITS
u_int xhits_gbl[MAXCPU];
u_int xhits_pg[MAXCPU];
u_int xhits_rng[MAXCPU];
SYSCTL_NODE(_debug, OID_AUTO, xhits, CTLFLAG_RW, 0, "");
SYSCTL_OPAQUE(_debug_xhits, OID_AUTO, global, CTLFLAG_RW, &xhits_gbl,
sizeof(xhits_gbl), "IU", "");
SYSCTL_OPAQUE(_debug_xhits, OID_AUTO, page, CTLFLAG_RW, &xhits_pg,
sizeof(xhits_pg), "IU", "");
SYSCTL_OPAQUE(_debug_xhits, OID_AUTO, range, CTLFLAG_RW, &xhits_rng,
sizeof(xhits_rng), "IU", "");
u_int ipi_global;
u_int ipi_page;
u_int ipi_range;
u_int ipi_range_size;
SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_global, CTLFLAG_RW, &ipi_global, 0, "");
SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_page, CTLFLAG_RW, &ipi_page, 0, "");
SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_range, CTLFLAG_RW, &ipi_range, 0, "");
SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_range_size, CTLFLAG_RW, &ipi_range_size,
0, "");
u_int ipi_masked_global;
u_int ipi_masked_page;
u_int ipi_masked_range;
u_int ipi_masked_range_size;
SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_masked_global, CTLFLAG_RW,
&ipi_masked_global, 0, "");
SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_masked_page, CTLFLAG_RW,
&ipi_masked_page, 0, "");
SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_masked_range, CTLFLAG_RW,
&ipi_masked_range, 0, "");
SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_masked_range_size, CTLFLAG_RW,
&ipi_masked_range_size, 0, "");
#endif
/*
* Flush the TLB on all other CPU's
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
*/
static void
smp_tlb_shootdown(u_int vector, vm_offset_t addr1, vm_offset_t addr2)
{
u_int ncpu;
register_t eflags;
ncpu = mp_ncpus - 1; /* does not shootdown self */
if (ncpu < 1)
return; /* no other cpus */
eflags = read_eflags();
if ((eflags & PSL_I) == 0)
panic("absolutely cannot call smp_ipi_shootdown with interrupts already disabled");
mtx_lock_spin(&smp_tlb_mtx);
smp_tlb_addr1 = addr1;
smp_tlb_addr2 = addr2;
atomic_store_rel_int(&smp_tlb_wait, 0);
ipi_all_but_self(vector);
while (smp_tlb_wait < ncpu)
ia32_pause();
mtx_unlock_spin(&smp_tlb_mtx);
}
/*
* This is about as magic as it gets. fortune(1) has got similar code
* for reversing bits in a word. Who thinks up this stuff??
*
* Yes, it does appear to be consistently faster than:
* while (i = ffs(m)) {
* m >>= i;
* bits++;
* }
* and
* while (lsb = (m & -m)) { // This is magic too
* m &= ~lsb; // or: m ^= lsb
* bits++;
* }
* Both of these latter forms do some very strange things on gcc-3.1 with
* -mcpu=pentiumpro and/or -march=pentiumpro and/or -O or -O2.
* There is probably an SSE or MMX popcnt instruction.
*
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
* I wonder if this should be in libkern?
*
* XXX Stop the presses! Another one:
* static __inline u_int32_t
* popcnt1(u_int32_t v)
* {
* v -= ((v >> 1) & 0x55555555);
* v = (v & 0x33333333) + ((v >> 2) & 0x33333333);
* v = (v + (v >> 4)) & 0x0F0F0F0F;
* return (v * 0x01010101) >> 24;
* }
* The downside is that it has a multiply. With a pentium3 with
* -mcpu=pentiumpro and -march=pentiumpro then gcc-3.1 will use
* an imull, and in that case it is faster. In most other cases
* it appears slightly slower.
*/
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
static __inline u_int32_t
popcnt(u_int32_t m)
{
m = (m & 0x55555555) + ((m & 0xaaaaaaaa) >> 1);
m = (m & 0x33333333) + ((m & 0xcccccccc) >> 2);
m = (m & 0x0f0f0f0f) + ((m & 0xf0f0f0f0) >> 4);
m = (m & 0x00ff00ff) + ((m & 0xff00ff00) >> 8);
m = (m & 0x0000ffff) + ((m & 0xffff0000) >> 16);
return m;
}
static void
smp_targeted_tlb_shootdown(u_int mask, u_int vector, vm_offset_t addr1, vm_offset_t addr2)
{
int ncpu, othercpus;
register_t eflags;
othercpus = mp_ncpus - 1;
if (mask == (u_int)-1) {
ncpu = othercpus;
if (ncpu < 1)
return;
} else {
/* XXX there should be a pcpu self mask */
mask &= ~(1 << PCPU_GET(cpuid));
if (mask == 0)
return;
ncpu = popcnt(mask);
if (ncpu > othercpus) {
/* XXX this should be a panic offence */
printf("SMP: tlb shootdown to %d other cpus (only have %d)\n",
ncpu, othercpus);
ncpu = othercpus;
}
/* XXX should be a panic, implied by mask == 0 above */
if (ncpu < 1)
return;
}
eflags = read_eflags();
if ((eflags & PSL_I) == 0)
panic("absolutely cannot call smp_targeted_ipi_shootdown with interrupts already disabled");
mtx_lock_spin(&smp_tlb_mtx);
smp_tlb_addr1 = addr1;
smp_tlb_addr2 = addr2;
atomic_store_rel_int(&smp_tlb_wait, 0);
if (mask == (u_int)-1)
ipi_all_but_self(vector);
else
ipi_selected(mask, vector);
while (smp_tlb_wait < ncpu)
ia32_pause();
mtx_unlock_spin(&smp_tlb_mtx);
}
#endif
void
smp_invltlb(void)
{
#if defined(APIC_IO)
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
if (smp_started) {
smp_tlb_shootdown(IPI_INVLTLB, 0, 0);
#ifdef COUNT_XINVLTLB_HITS
ipi_global++;
#endif
}
#endif /* APIC_IO */
}
void
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
smp_invlpg(vm_offset_t addr)
{
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
#if defined(APIC_IO)
if (smp_started) {
smp_tlb_shootdown(IPI_INVLPG, addr, 0);
#ifdef COUNT_XINVLTLB_HITS
ipi_page++;
#endif
}
#endif /* APIC_IO */
}
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
void
smp_invlpg_range(vm_offset_t addr1, vm_offset_t addr2)
{
#if defined(APIC_IO)
if (smp_started) {
smp_tlb_shootdown(IPI_INVLRNG, addr1, addr2);
#ifdef COUNT_XINVLTLB_HITS
ipi_range++;
ipi_range_size += (addr2 - addr1) / PAGE_SIZE;
#endif
}
#endif /* APIC_IO */
}
void
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
smp_masked_invltlb(u_int mask)
{
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
#if defined(APIC_IO)
if (smp_started) {
smp_targeted_tlb_shootdown(mask, IPI_INVLTLB, 0, 0);
#ifdef COUNT_XINVLTLB_HITS
ipi_masked_global++;
#endif
}
#endif /* APIC_IO */
}
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
void
smp_masked_invlpg(u_int mask, vm_offset_t addr)
{
#if defined(APIC_IO)
if (smp_started) {
smp_targeted_tlb_shootdown(mask, IPI_INVLPG, addr, 0);
#ifdef COUNT_XINVLTLB_HITS
ipi_masked_page++;
#endif
}
#endif /* APIC_IO */
}
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
void
smp_masked_invlpg_range(u_int mask, vm_offset_t addr1, vm_offset_t addr2)
{
#if defined(APIC_IO)
if (smp_started) {
smp_targeted_tlb_shootdown(mask, IPI_INVLRNG, addr1, addr2);
#ifdef COUNT_XINVLTLB_HITS
ipi_masked_range++;
ipi_masked_range_size += (addr2 - addr1) / PAGE_SIZE;
#endif
}
#endif /* APIC_IO */
}
/*
* This is called once the rest of the system is up and running and we're
* ready to let the AP's out of the pen.
*/
extern void enable_sse(void);
void
ap_init(void)
{
u_int apic_id;
/* spin until all the AP's are ready */
while (!aps_ready)
/* spin */ ;
/* BSP may have changed PTD while we were waiting */
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
invltlb();
#if defined(I586_CPU) && !defined(NO_F00F_HACK)
lidt(&r_idt);
#endif
Clean up some low level bootstrap code: - stop using the evil 'struct trapframe' argument for mi_startup() (formerly main()). There are much better ways of doing it. - do not use prepare_usermode() - setregs() in execve() will do it all for us as long as the p_md.md_regs pointer is set. (which is now done in machdep.c rather than init_main.c. The Alpha port did it this way all along and is much cleaner). - collect all the magic %cr0 etc register settings into one place and have the AP's call that instead of using magic numbers (!!) that keep changing over and over again. - Make it safe to call kthread_create() earlier, including during the device probe sequence. It doesn't need the callback mechanism that NetBSD's version uses. - kthreads created this way are root-less as they exist before the root filesystem is mounted. init(1) is set up so that it aquires the root pointers prior to running. If other kthreads want filesystem acccess we can make this code more generic. - set all threads start times once we have decided what time it is. - init uses a trampoline rather than the evil prepare_usermode() hack. - kern_descrip.c has a couple of tweaks to deal with forking when there is no rootdir or cwd etc. - adjust the early SYSINIT() sequence so that a few prereqisites are in place. eg: make sure the run queue is initialized before doing forks. With this, the USB code can easily create a kthread to do the device tree discovery. (I have tested it, it works nicely). There are still some open issues before this is truely useful. - tsleep() does not like working before the clock is running. It sort-of tries to spin wait, but it can do more useful things now. - stopping a kthread in kld code at unload time is "interesting" but we have a solution for that. The Alpha code needs no changes for this. It already uses pretty much the same strategies, but a little cleaner.
2000-08-11 09:05:12 +00:00
/* set up CPU registers and state */
cpu_setregs();
/* set up FPU state on the AP */
npxinit(__INITIAL_NPXCW__);
/* set up SSE registers */
enable_sse();
/* A quick check from sanity claus */
apic_id = (apic_id_to_logical[(lapic.id & 0x0f000000) >> 24]);
if (PCPU_GET(cpuid) != apic_id) {
printf("SMP: cpuid = %d\n", PCPU_GET(cpuid));
printf("SMP: apic_id = %d\n", apic_id);
printf("PTD[MPPTDI] = %p\n", (void *)PTD[MPPTDI]);
panic("cpuid mismatch! boom!!");
}
/* Init local apic for irq's */
apic_initialize();
/* Set memory range attributes for this CPU to match the BSP */
mem_range_AP_init();
mtx_lock_spin(&ap_boot_mtx);
CTR1(KTR_SMP, "SMP: AP CPU #%d Launched", PCPU_GET(cpuid));
smp_cpus++;
/* Build our map of 'other' CPUs. */
PCPU_SET(other_cpus, all_cpus & ~PCPU_GET(cpumask));
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
if (bootverbose)
apic_dump("ap_init()");
printf("SMP: AP CPU #%d Launched!\n", PCPU_GET(cpuid));
if (smp_cpus == mp_ncpus) {
smp_started = 1; /* enable IPI's, tlb shootdown, freezes etc */
smp_active = 1; /* historic */
}
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&ap_boot_mtx);
/* wait until all the AP's are up */
while (smp_started == 0)
; /* nothing */
binuptime(PCPU_PTR(switchtime));
PCPU_SET(switchticks, ticks);
/* ok, now grab sched_lock and enter the scheduler */
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sched_lock);
cpu_throw(); /* doesn't return */
panic("scheduler returned us to %s", __func__);
}
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
/*
* For statclock, we send an IPI to all CPU's to have them call this
* function.
*
* WARNING! unpend() will call statclock_process() directly and skip this
* routine.
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
*/
void
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
forwarded_statclock(struct trapframe frame)
{
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
mtx_lock_spin(&sched_lock);
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
statclock_process(curthread->td_kse, TRAPF_PC(&frame),
TRAPF_USERMODE(&frame));
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
mtx_unlock_spin(&sched_lock);
}
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
void
forward_statclock(void)
{
int map;
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
CTR0(KTR_SMP, "forward_statclock");
if (!smp_started || cold || panicstr)
return;
map = PCPU_GET(other_cpus) & ~stopped_cpus ;
if (map != 0)
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
ipi_selected(map, IPI_STATCLOCK);
}
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
/*
* For each hardclock(), we send an IPI to all other CPU's to have them
* execute this function. It would be nice to reduce contention on
* sched_lock if we could simply peek at the CPU to determine the user/kernel
* state and call hardclock_process() on the CPU receiving the clock interrupt
* and then just use a simple IPI to handle any ast's if needed.
*
* WARNING! unpend() will call hardclock_process() directly and skip this
* routine.
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
*/
void
forwarded_hardclock(struct trapframe frame)
{
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sched_lock);
hardclock_process(curthread, TRAPF_USERMODE(&frame));
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
mtx_unlock_spin(&sched_lock);
}
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
void
forward_hardclock(void)
{
u_int map;
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
CTR0(KTR_SMP, "forward_hardclock");
if (!smp_started || cold || panicstr)
return;
Overhaul of the SMP code. Several portions of the SMP kernel support have been made machine independent and various other adjustments have been made to support Alpha SMP. - It splits the per-process portions of hardclock() and statclock() off into hardclock_process() and statclock_process() respectively. hardclock() and statclock() call the *_process() functions for the current process so that UP systems will run as before. For SMP systems, it is simply necessary to ensure that all other processors execute the *_process() functions when the main clock functions are triggered on one CPU by an interrupt. For the alpha 4100, clock interrupts are delievered in a staggered broadcast fashion, so we simply call hardclock/statclock on the boot CPU and call the *_process() functions on the secondaries. For x86, we call statclock and hardclock as usual and then call forward_hardclock/statclock in the MD code to send an IPI to cause the AP's to execute forwared_hardclock/statclock which then call the *_process() functions. - forward_signal() and forward_roundrobin() have been reworked to be MI and to involve less hackery. Now the cpu doing the forward sets any flags, etc. and sends a very simple IPI_AST to the other cpu(s). AST IPIs now just basically return so that they can execute ast() and don't bother with setting the astpending or needresched flags themselves. This also removes the loop in forward_signal() as sched_lock closes the race condition that the loop worked around. - need_resched(), resched_wanted() and clear_resched() have been changed to take a process to act on rather than assuming curproc so that they can be used to implement forward_roundrobin() as described above. - Various other SMP variables have been moved to a MI subr_smp.c and a new header sys/smp.h declares MI SMP variables and API's. The IPI API's from machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h. - The globaldata_register() and globaldata_find() functions as well as the SLIST of globaldata structures has become MI and moved into subr_smp.c. Also, the globaldata list is only available if SMP support is compiled in. Reviewed by: jake, peter Looked over by: eivind
2001-04-27 19:28:25 +00:00
map = PCPU_GET(other_cpus) & ~stopped_cpus ;
if (map != 0)
ipi_selected(map, IPI_HARDCLOCK);
}
#ifdef APIC_INTR_REORDER
/*
* Maintain mapping from softintr vector to isr bit in local apic.
*/
void
set_lapic_isrloc(int intr, int vector)
{
if (intr < 0 || intr > 32)
panic("set_apic_isrloc: bad intr argument: %d",intr);
if (vector < ICU_OFFSET || vector > 255)
panic("set_apic_isrloc: bad vector argument: %d",vector);
apic_isrbit_location[intr].location = &lapic.isr0 + ((vector>>5)<<2);
apic_isrbit_location[intr].bit = (1<<(vector & 31));
}
#endif
/*
* send an IPI to a set of cpus.
*/
void
ipi_selected(u_int32_t cpus, u_int ipi)
{
2002-01-18 04:41:23 +00:00
CTR3(KTR_SMP, "%s: cpus: %x ipi: %x", __func__, cpus, ipi);
selected_apic_ipi(cpus, ipi, APIC_DELMODE_FIXED);
}
/*
* send an IPI INTerrupt containing 'vector' to all CPUs, including myself
*/
void
ipi_all(u_int ipi)
{
2002-01-18 04:41:23 +00:00
CTR2(KTR_SMP, "%s: ipi: %x", __func__, ipi);
apic_ipi(APIC_DEST_ALLISELF, ipi, APIC_DELMODE_FIXED);
}
/*
* send an IPI to all CPUs EXCEPT myself
*/
void
ipi_all_but_self(u_int ipi)
{
2002-01-18 04:41:23 +00:00
CTR2(KTR_SMP, "%s: ipi: %x", __func__, ipi);
apic_ipi(APIC_DEST_ALLESELF, ipi, APIC_DELMODE_FIXED);
}
/*
* send an IPI to myself
*/
void
ipi_self(u_int ipi)
{
2002-01-18 04:41:23 +00:00
CTR2(KTR_SMP, "%s: ipi: %x", __func__, ipi);
apic_ipi(APIC_DEST_SELF, ipi, APIC_DELMODE_FIXED);
}
void
release_aps(void *dummy __unused)
{
atomic_store_rel_int(&aps_ready, 1);
}
SYSINIT(start_aps, SI_SUB_SMP, SI_ORDER_FIRST, release_aps, NULL);