linux/arch/x86/kernel/apic/apic.c
<<
>>
Prefs
   1/*
   2 *      Local APIC handling, local APIC timers
   3 *
   4 *      (c) 1999, 2000, 2009 Ingo Molnar <mingo@redhat.com>
   5 *
   6 *      Fixes
   7 *      Maciej W. Rozycki       :       Bits for genuine 82489DX APICs;
   8 *                                      thanks to Eric Gilmore
   9 *                                      and Rolf G. Tews
  10 *                                      for testing these extensively.
  11 *      Maciej W. Rozycki       :       Various updates and fixes.
  12 *      Mikael Pettersson       :       Power Management for UP-APIC.
  13 *      Pavel Machek and
  14 *      Mikael Pettersson       :       PM converted to driver model.
  15 */
  16
  17#include <linux/perf_event.h>
  18#include <linux/kernel_stat.h>
  19#include <linux/mc146818rtc.h>
  20#include <linux/acpi_pmtmr.h>
  21#include <linux/clockchips.h>
  22#include <linux/interrupt.h>
  23#include <linux/bootmem.h>
  24#include <linux/ftrace.h>
  25#include <linux/ioport.h>
  26#include <linux/module.h>
  27#include <linux/syscore_ops.h>
  28#include <linux/delay.h>
  29#include <linux/timex.h>
  30#include <linux/i8253.h>
  31#include <linux/dmar.h>
  32#include <linux/init.h>
  33#include <linux/cpu.h>
  34#include <linux/dmi.h>
  35#include <linux/smp.h>
  36#include <linux/mm.h>
  37
  38#include <asm/perf_event.h>
  39#include <asm/x86_init.h>
  40#include <asm/pgalloc.h>
  41#include <linux/atomic.h>
  42#include <asm/mpspec.h>
  43#include <asm/i8259.h>
  44#include <asm/proto.h>
  45#include <asm/apic.h>
  46#include <asm/io_apic.h>
  47#include <asm/desc.h>
  48#include <asm/hpet.h>
  49#include <asm/idle.h>
  50#include <asm/mtrr.h>
  51#include <asm/time.h>
  52#include <asm/smp.h>
  53#include <asm/mce.h>
  54#include <asm/tsc.h>
  55#include <asm/hypervisor.h>
  56
  57unsigned int num_processors;
  58
  59unsigned disabled_cpus __cpuinitdata;
  60
  61/* Processor that is doing the boot up */
  62unsigned int boot_cpu_physical_apicid = -1U;
  63
  64/*
  65 * The highest APIC ID seen during enumeration.
  66 */
  67unsigned int max_physical_apicid;
  68
  69/*
  70 * Bitmask of physically existing CPUs:
  71 */
  72physid_mask_t phys_cpu_present_map;
  73
  74/*
  75 * Map cpu index to physical APIC ID
  76 */
  77DEFINE_EARLY_PER_CPU(u16, x86_cpu_to_apicid, BAD_APICID);
  78DEFINE_EARLY_PER_CPU(u16, x86_bios_cpu_apicid, BAD_APICID);
  79EXPORT_EARLY_PER_CPU_SYMBOL(x86_cpu_to_apicid);
  80EXPORT_EARLY_PER_CPU_SYMBOL(x86_bios_cpu_apicid);
  81
  82#ifdef CONFIG_X86_32
  83
  84/*
  85 * On x86_32, the mapping between cpu and logical apicid may vary
  86 * depending on apic in use.  The following early percpu variable is
  87 * used for the mapping.  This is where the behaviors of x86_64 and 32
  88 * actually diverge.  Let's keep it ugly for now.
  89 */
  90DEFINE_EARLY_PER_CPU(int, x86_cpu_to_logical_apicid, BAD_APICID);
  91
  92/*
  93 * Knob to control our willingness to enable the local APIC.
  94 *
  95 * +1=force-enable
  96 */
  97static int force_enable_local_apic __initdata;
  98/*
  99 * APIC command line parameters
 100 */
 101static int __init parse_lapic(char *arg)
 102{
 103        force_enable_local_apic = 1;
 104        return 0;
 105}
 106early_param("lapic", parse_lapic);
 107/* Local APIC was disabled by the BIOS and enabled by the kernel */
 108static int enabled_via_apicbase;
 109
 110/*
 111 * Handle interrupt mode configuration register (IMCR).
 112 * This register controls whether the interrupt signals
 113 * that reach the BSP come from the master PIC or from the
 114 * local APIC. Before entering Symmetric I/O Mode, either
 115 * the BIOS or the operating system must switch out of
 116 * PIC Mode by changing the IMCR.
 117 */
 118static inline void imcr_pic_to_apic(void)
 119{
 120        /* select IMCR register */
 121        outb(0x70, 0x22);
 122        /* NMI and 8259 INTR go through APIC */
 123        outb(0x01, 0x23);
 124}
 125
 126static inline void imcr_apic_to_pic(void)
 127{
 128        /* select IMCR register */
 129        outb(0x70, 0x22);
 130        /* NMI and 8259 INTR go directly to BSP */
 131        outb(0x00, 0x23);
 132}
 133#endif
 134
 135#ifdef CONFIG_X86_64
 136static int apic_calibrate_pmtmr __initdata;
 137static __init int setup_apicpmtimer(char *s)
 138{
 139        apic_calibrate_pmtmr = 1;
 140        notsc_setup(NULL);
 141        return 0;
 142}
 143__setup("apicpmtimer", setup_apicpmtimer);
 144#endif
 145
 146int x2apic_mode;
 147#ifdef CONFIG_X86_X2APIC
 148/* x2apic enabled before OS handover */
 149static int x2apic_preenabled;
 150static __init int setup_nox2apic(char *str)
 151{
 152        if (x2apic_enabled()) {
 153                pr_warning("Bios already enabled x2apic, "
 154                           "can't enforce nox2apic");
 155                return 0;
 156        }
 157
 158        setup_clear_cpu_cap(X86_FEATURE_X2APIC);
 159        return 0;
 160}
 161early_param("nox2apic", setup_nox2apic);
 162#endif
 163
 164unsigned long mp_lapic_addr;
 165int disable_apic;
 166/* Disable local APIC timer from the kernel commandline or via dmi quirk */
 167static int disable_apic_timer __initdata;
 168/* Local APIC timer works in C2 */
 169int local_apic_timer_c2_ok;
 170EXPORT_SYMBOL_GPL(local_apic_timer_c2_ok);
 171
 172int first_system_vector = 0xfe;
 173
 174/*
 175 * Debug level, exported for io_apic.c
 176 */
 177unsigned int apic_verbosity;
 178
 179int pic_mode;
 180
 181/* Have we found an MP table */
 182int smp_found_config;
 183
 184static struct resource lapic_resource = {
 185        .name = "Local APIC",
 186        .flags = IORESOURCE_MEM | IORESOURCE_BUSY,
 187};
 188
 189static unsigned int calibration_result;
 190
 191static void apic_pm_activate(void);
 192
 193static unsigned long apic_phys;
 194
 195/*
 196 * Get the LAPIC version
 197 */
 198static inline int lapic_get_version(void)
 199{
 200        return GET_APIC_VERSION(apic_read(APIC_LVR));
 201}
 202
 203/*
 204 * Check, if the APIC is integrated or a separate chip
 205 */
 206static inline int lapic_is_integrated(void)
 207{
 208#ifdef CONFIG_X86_64
 209        return 1;
 210#else
 211        return APIC_INTEGRATED(lapic_get_version());
 212#endif
 213}
 214
 215/*
 216 * Check, whether this is a modern or a first generation APIC
 217 */
 218static int modern_apic(void)
 219{
 220        /* AMD systems use old APIC versions, so check the CPU */
 221        if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD &&
 222            boot_cpu_data.x86 >= 0xf)
 223                return 1;
 224        return lapic_get_version() >= 0x14;
 225}
 226
 227/*
 228 * right after this call apic become NOOP driven
 229 * so apic->write/read doesn't do anything
 230 */
 231static void __init apic_disable(void)
 232{
 233        pr_info("APIC: switched to apic NOOP\n");
 234        apic = &apic_noop;
 235}
 236
 237void native_apic_wait_icr_idle(void)
 238{
 239        while (apic_read(APIC_ICR) & APIC_ICR_BUSY)
 240                cpu_relax();
 241}
 242
 243u32 native_safe_apic_wait_icr_idle(void)
 244{
 245        u32 send_status;
 246        int timeout;
 247
 248        timeout = 0;
 249        do {
 250                send_status = apic_read(APIC_ICR) & APIC_ICR_BUSY;
 251                if (!send_status)
 252                        break;
 253                udelay(100);
 254        } while (timeout++ < 1000);
 255
 256        return send_status;
 257}
 258
 259void native_apic_icr_write(u32 low, u32 id)
 260{
 261        apic_write(APIC_ICR2, SET_APIC_DEST_FIELD(id));
 262        apic_write(APIC_ICR, low);
 263}
 264
 265u64 native_apic_icr_read(void)
 266{
 267        u32 icr1, icr2;
 268
 269        icr2 = apic_read(APIC_ICR2);
 270        icr1 = apic_read(APIC_ICR);
 271
 272        return icr1 | ((u64)icr2 << 32);
 273}
 274
 275#ifdef CONFIG_X86_32
 276/**
 277 * get_physical_broadcast - Get number of physical broadcast IDs
 278 */
 279int get_physical_broadcast(void)
 280{
 281        return modern_apic() ? 0xff : 0xf;
 282}
 283#endif
 284
 285/**
 286 * lapic_get_maxlvt - get the maximum number of local vector table entries
 287 */
 288int lapic_get_maxlvt(void)
 289{
 290        unsigned int v;
 291
 292        v = apic_read(APIC_LVR);
 293        /*
 294         * - we always have APIC integrated on 64bit mode
 295         * - 82489DXs do not report # of LVT entries
 296         */
 297        return APIC_INTEGRATED(GET_APIC_VERSION(v)) ? GET_APIC_MAXLVT(v) : 2;
 298}
 299
 300/*
 301 * Local APIC timer
 302 */
 303
 304/* Clock divisor */
 305#define APIC_DIVISOR 16
 306
 307/*
 308 * This function sets up the local APIC timer, with a timeout of
 309 * 'clocks' APIC bus clock. During calibration we actually call
 310 * this function twice on the boot CPU, once with a bogus timeout
 311 * value, second time for real. The other (noncalibrating) CPUs
 312 * call this function only once, with the real, calibrated value.
 313 *
 314 * We do reads before writes even if unnecessary, to get around the
 315 * P5 APIC double write bug.
 316 */
 317static void __setup_APIC_LVTT(unsigned int clocks, int oneshot, int irqen)
 318{
 319        unsigned int lvtt_value, tmp_value;
 320
 321        lvtt_value = LOCAL_TIMER_VECTOR;
 322        if (!oneshot)
 323                lvtt_value |= APIC_LVT_TIMER_PERIODIC;
 324        if (!lapic_is_integrated())
 325                lvtt_value |= SET_APIC_TIMER_BASE(APIC_TIMER_BASE_DIV);
 326
 327        if (!irqen)
 328                lvtt_value |= APIC_LVT_MASKED;
 329
 330        apic_write(APIC_LVTT, lvtt_value);
 331
 332        /*
 333         * Divide PICLK by 16
 334         */
 335        tmp_value = apic_read(APIC_TDCR);
 336        apic_write(APIC_TDCR,
 337                (tmp_value & ~(APIC_TDR_DIV_1 | APIC_TDR_DIV_TMBASE)) |
 338                APIC_TDR_DIV_16);
 339
 340        if (!oneshot)
 341                apic_write(APIC_TMICT, clocks / APIC_DIVISOR);
 342}
 343
 344/*
 345 * Setup extended LVT, AMD specific
 346 *
 347 * Software should use the LVT offsets the BIOS provides.  The offsets
 348 * are determined by the subsystems using it like those for MCE
 349 * threshold or IBS.  On K8 only offset 0 (APIC500) and MCE interrupts
 350 * are supported. Beginning with family 10h at least 4 offsets are
 351 * available.
 352 *
 353 * Since the offsets must be consistent for all cores, we keep track
 354 * of the LVT offsets in software and reserve the offset for the same
 355 * vector also to be used on other cores. An offset is freed by
 356 * setting the entry to APIC_EILVT_MASKED.
 357 *
 358 * If the BIOS is right, there should be no conflicts. Otherwise a
 359 * "[Firmware Bug]: ..." error message is generated. However, if
 360 * software does not properly determines the offsets, it is not
 361 * necessarily a BIOS bug.
 362 */
 363
 364static atomic_t eilvt_offsets[APIC_EILVT_NR_MAX];
 365
 366static inline int eilvt_entry_is_changeable(unsigned int old, unsigned int new)
 367{
 368        return (old & APIC_EILVT_MASKED)
 369                || (new == APIC_EILVT_MASKED)
 370                || ((new & ~APIC_EILVT_MASKED) == old);
 371}
 372
 373static unsigned int reserve_eilvt_offset(int offset, unsigned int new)
 374{
 375        unsigned int rsvd;                      /* 0: uninitialized */
 376
 377        if (offset >= APIC_EILVT_NR_MAX)
 378                return ~0;
 379
 380        rsvd = atomic_read(&eilvt_offsets[offset]) & ~APIC_EILVT_MASKED;
 381        do {
 382                if (rsvd &&
 383                    !eilvt_entry_is_changeable(rsvd, new))
 384                        /* may not change if vectors are different */
 385                        return rsvd;
 386                rsvd = atomic_cmpxchg(&eilvt_offsets[offset], rsvd, new);
 387        } while (rsvd != new);
 388
 389        return new;
 390}
 391
 392/*
 393 * If mask=1, the LVT entry does not generate interrupts while mask=0
 394 * enables the vector. See also the BKDGs. Must be called with
 395 * preemption disabled.
 396 */
 397
 398int setup_APIC_eilvt(u8 offset, u8 vector, u8 msg_type, u8 mask)
 399{
 400        unsigned long reg = APIC_EILVTn(offset);
 401        unsigned int new, old, reserved;
 402
 403        new = (mask << 16) | (msg_type << 8) | vector;
 404        old = apic_read(reg);
 405        reserved = reserve_eilvt_offset(offset, new);
 406
 407        if (reserved != new) {
 408                pr_err(FW_BUG "cpu %d, try to use APIC%lX (LVT offset %d) for "
 409                       "vector 0x%x, but the register is already in use for "
 410                       "vector 0x%x on another cpu\n",
 411                       smp_processor_id(), reg, offset, new, reserved);
 412                return -EINVAL;
 413        }
 414
 415        if (!eilvt_entry_is_changeable(old, new)) {
 416                pr_err(FW_BUG "cpu %d, try to use APIC%lX (LVT offset %d) for "
 417                       "vector 0x%x, but the register is already in use for "
 418                       "vector 0x%x on this cpu\n",
 419                       smp_processor_id(), reg, offset, new, old);
 420                return -EBUSY;
 421        }
 422
 423        apic_write(reg, new);
 424
 425        return 0;
 426}
 427EXPORT_SYMBOL_GPL(setup_APIC_eilvt);
 428
 429/*
 430 * Program the next event, relative to now
 431 */
 432static int lapic_next_event(unsigned long delta,
 433                            struct clock_event_device *evt)
 434{
 435        apic_write(APIC_TMICT, delta);
 436        return 0;
 437}
 438
 439/*
 440 * Setup the lapic timer in periodic or oneshot mode
 441 */
 442static void lapic_timer_setup(enum clock_event_mode mode,
 443                              struct clock_event_device *evt)
 444{
 445        unsigned long flags;
 446        unsigned int v;
 447
 448        /* Lapic used as dummy for broadcast ? */
 449        if (evt->features & CLOCK_EVT_FEAT_DUMMY)
 450                return;
 451
 452        local_irq_save(flags);
 453
 454        switch (mode) {
 455        case CLOCK_EVT_MODE_PERIODIC:
 456        case CLOCK_EVT_MODE_ONESHOT:
 457                __setup_APIC_LVTT(calibration_result,
 458                                  mode != CLOCK_EVT_MODE_PERIODIC, 1);
 459                break;
 460        case CLOCK_EVT_MODE_UNUSED:
 461        case CLOCK_EVT_MODE_SHUTDOWN:
 462                v = apic_read(APIC_LVTT);
 463                v |= (APIC_LVT_MASKED | LOCAL_TIMER_VECTOR);
 464                apic_write(APIC_LVTT, v);
 465                apic_write(APIC_TMICT, 0);
 466                break;
 467        case CLOCK_EVT_MODE_RESUME:
 468                /* Nothing to do here */
 469                break;
 470        }
 471
 472        local_irq_restore(flags);
 473}
 474
 475/*
 476 * Local APIC timer broadcast function
 477 */
 478static void lapic_timer_broadcast(const struct cpumask *mask)
 479{
 480#ifdef CONFIG_SMP
 481        apic->send_IPI_mask(mask, LOCAL_TIMER_VECTOR);
 482#endif
 483}
 484
 485
 486/*
 487 * The local apic timer can be used for any function which is CPU local.
 488 */
 489static struct clock_event_device lapic_clockevent = {
 490        .name           = "lapic",
 491        .features       = CLOCK_EVT_FEAT_PERIODIC | CLOCK_EVT_FEAT_ONESHOT
 492                        | CLOCK_EVT_FEAT_C3STOP | CLOCK_EVT_FEAT_DUMMY,
 493        .shift          = 32,
 494        .set_mode       = lapic_timer_setup,
 495        .set_next_event = lapic_next_event,
 496        .broadcast      = lapic_timer_broadcast,
 497        .rating         = 100,
 498        .irq            = -1,
 499};
 500static DEFINE_PER_CPU(struct clock_event_device, lapic_events);
 501
 502/*
 503 * Setup the local APIC timer for this CPU. Copy the initialized values
 504 * of the boot CPU and register the clock event in the framework.
 505 */
 506static void __cpuinit setup_APIC_timer(void)
 507{
 508        struct clock_event_device *levt = &__get_cpu_var(lapic_events);
 509
 510        if (this_cpu_has(X86_FEATURE_ARAT)) {
 511                lapic_clockevent.features &= ~CLOCK_EVT_FEAT_C3STOP;
 512                /* Make LAPIC timer preferrable over percpu HPET */
 513                lapic_clockevent.rating = 150;
 514        }
 515
 516        memcpy(levt, &lapic_clockevent, sizeof(*levt));
 517        levt->cpumask = cpumask_of(smp_processor_id());
 518
 519        clockevents_register_device(levt);
 520}
 521
 522/*
 523 * In this functions we calibrate APIC bus clocks to the external timer.
 524 *
 525 * We want to do the calibration only once since we want to have local timer
 526 * irqs syncron. CPUs connected by the same APIC bus have the very same bus
 527 * frequency.
 528 *
 529 * This was previously done by reading the PIT/HPET and waiting for a wrap
 530 * around to find out, that a tick has elapsed. I have a box, where the PIT
 531 * readout is broken, so it never gets out of the wait loop again. This was
 532 * also reported by others.
 533 *
 534 * Monitoring the jiffies value is inaccurate and the clockevents
 535 * infrastructure allows us to do a simple substitution of the interrupt
 536 * handler.
 537 *
 538 * The calibration routine also uses the pm_timer when possible, as the PIT
 539 * happens to run way too slow (factor 2.3 on my VAIO CoreDuo, which goes
 540 * back to normal later in the boot process).
 541 */
 542
 543#define LAPIC_CAL_LOOPS         (HZ/10)
 544
 545static __initdata int lapic_cal_loops = -1;
 546static __initdata long lapic_cal_t1, lapic_cal_t2;
 547static __initdata unsigned long long lapic_cal_tsc1, lapic_cal_tsc2;
 548static __initdata unsigned long lapic_cal_pm1, lapic_cal_pm2;
 549static __initdata unsigned long lapic_cal_j1, lapic_cal_j2;
 550
 551/*
 552 * Temporary interrupt handler.
 553 */
 554static void __init lapic_cal_handler(struct clock_event_device *dev)
 555{
 556        unsigned long long tsc = 0;
 557        long tapic = apic_read(APIC_TMCCT);
 558        unsigned long pm = acpi_pm_read_early();
 559
 560        if (cpu_has_tsc)
 561                rdtscll(tsc);
 562
 563        switch (lapic_cal_loops++) {
 564        case 0:
 565                lapic_cal_t1 = tapic;
 566                lapic_cal_tsc1 = tsc;
 567                lapic_cal_pm1 = pm;
 568                lapic_cal_j1 = jiffies;
 569                break;
 570
 571        case LAPIC_CAL_LOOPS:
 572                lapic_cal_t2 = tapic;
 573                lapic_cal_tsc2 = tsc;
 574                if (pm < lapic_cal_pm1)
 575                        pm += ACPI_PM_OVRRUN;
 576                lapic_cal_pm2 = pm;
 577                lapic_cal_j2 = jiffies;
 578                break;
 579        }
 580}
 581
 582static int __init
 583calibrate_by_pmtimer(long deltapm, long *delta, long *deltatsc)
 584{
 585        const long pm_100ms = PMTMR_TICKS_PER_SEC / 10;
 586        const long pm_thresh = pm_100ms / 100;
 587        unsigned long mult;
 588        u64 res;
 589
 590#ifndef CONFIG_X86_PM_TIMER
 591        return -1;
 592#endif
 593
 594        apic_printk(APIC_VERBOSE, "... PM-Timer delta = %ld\n", deltapm);
 595
 596        /* Check, if the PM timer is available */
 597        if (!deltapm)
 598                return -1;
 599
 600        mult = clocksource_hz2mult(PMTMR_TICKS_PER_SEC, 22);
 601
 602        if (deltapm > (pm_100ms - pm_thresh) &&
 603            deltapm < (pm_100ms + pm_thresh)) {
 604                apic_printk(APIC_VERBOSE, "... PM-Timer result ok\n");
 605                return 0;
 606        }
 607
 608        res = (((u64)deltapm) *  mult) >> 22;
 609        do_div(res, 1000000);
 610        pr_warning("APIC calibration not consistent "
 611                   "with PM-Timer: %ldms instead of 100ms\n",(long)res);
 612
 613        /* Correct the lapic counter value */
 614        res = (((u64)(*delta)) * pm_100ms);
 615        do_div(res, deltapm);
 616        pr_info("APIC delta adjusted to PM-Timer: "
 617                "%lu (%ld)\n", (unsigned long)res, *delta);
 618        *delta = (long)res;
 619
 620        /* Correct the tsc counter value */
 621        if (cpu_has_tsc) {
 622                res = (((u64)(*deltatsc)) * pm_100ms);
 623                do_div(res, deltapm);
 624                apic_printk(APIC_VERBOSE, "TSC delta adjusted to "
 625                                          "PM-Timer: %lu (%ld)\n",
 626                                        (unsigned long)res, *deltatsc);
 627                *deltatsc = (long)res;
 628        }
 629
 630        return 0;
 631}
 632
 633static int __init calibrate_APIC_clock(void)
 634{
 635        struct clock_event_device *levt = &__get_cpu_var(lapic_events);
 636        void (*real_handler)(struct clock_event_device *dev);
 637        unsigned long deltaj;
 638        long delta, deltatsc;
 639        int pm_referenced = 0;
 640
 641        local_irq_disable();
 642
 643        /* Replace the global interrupt handler */
 644        real_handler = global_clock_event->event_handler;
 645        global_clock_event->event_handler = lapic_cal_handler;
 646
 647        /*
 648         * Setup the APIC counter to maximum. There is no way the lapic
 649         * can underflow in the 100ms detection time frame
 650         */
 651        __setup_APIC_LVTT(0xffffffff, 0, 0);
 652
 653        /* Let the interrupts run */
 654        local_irq_enable();
 655
 656        while (lapic_cal_loops <= LAPIC_CAL_LOOPS)
 657                cpu_relax();
 658
 659        local_irq_disable();
 660
 661        /* Restore the real event handler */
 662        global_clock_event->event_handler = real_handler;
 663
 664        /* Build delta t1-t2 as apic timer counts down */
 665        delta = lapic_cal_t1 - lapic_cal_t2;
 666        apic_printk(APIC_VERBOSE, "... lapic delta = %ld\n", delta);
 667
 668        deltatsc = (long)(lapic_cal_tsc2 - lapic_cal_tsc1);
 669
 670        /* we trust the PM based calibration if possible */
 671        pm_referenced = !calibrate_by_pmtimer(lapic_cal_pm2 - lapic_cal_pm1,
 672                                        &delta, &deltatsc);
 673
 674        /* Calculate the scaled math multiplication factor */
 675        lapic_clockevent.mult = div_sc(delta, TICK_NSEC * LAPIC_CAL_LOOPS,
 676                                       lapic_clockevent.shift);
 677        lapic_clockevent.max_delta_ns =
 678                clockevent_delta2ns(0x7FFFFFFF, &lapic_clockevent);
 679        lapic_clockevent.min_delta_ns =
 680                clockevent_delta2ns(0xF, &lapic_clockevent);
 681
 682        calibration_result = (delta * APIC_DIVISOR) / LAPIC_CAL_LOOPS;
 683
 684        apic_printk(APIC_VERBOSE, "..... delta %ld\n", delta);
 685        apic_printk(APIC_VERBOSE, "..... mult: %u\n", lapic_clockevent.mult);
 686        apic_printk(APIC_VERBOSE, "..... calibration result: %u\n",
 687                    calibration_result);
 688
 689        if (cpu_has_tsc) {
 690                apic_printk(APIC_VERBOSE, "..... CPU clock speed is "
 691                            "%ld.%04ld MHz.\n",
 692                            (deltatsc / LAPIC_CAL_LOOPS) / (1000000 / HZ),
 693                            (deltatsc / LAPIC_CAL_LOOPS) % (1000000 / HZ));
 694        }
 695
 696        apic_printk(APIC_VERBOSE, "..... host bus clock speed is "
 697                    "%u.%04u MHz.\n",
 698                    calibration_result / (1000000 / HZ),
 699                    calibration_result % (1000000 / HZ));
 700
 701        /*
 702         * Do a sanity check on the APIC calibration result
 703         */
 704        if (calibration_result < (1000000 / HZ)) {
 705                local_irq_enable();
 706                pr_warning("APIC frequency too slow, disabling apic timer\n");
 707                return -1;
 708        }
 709
 710        levt->features &= ~CLOCK_EVT_FEAT_DUMMY;
 711
 712        /*
 713         * PM timer calibration failed or not turned on
 714         * so lets try APIC timer based calibration
 715         */
 716        if (!pm_referenced) {
 717                apic_printk(APIC_VERBOSE, "... verify APIC timer\n");
 718
 719                /*
 720                 * Setup the apic timer manually
 721                 */
 722                levt->event_handler = lapic_cal_handler;
 723                lapic_timer_setup(CLOCK_EVT_MODE_PERIODIC, levt);
 724                lapic_cal_loops = -1;
 725
 726                /* Let the interrupts run */
 727                local_irq_enable();
 728
 729                while (lapic_cal_loops <= LAPIC_CAL_LOOPS)
 730                        cpu_relax();
 731
 732                /* Stop the lapic timer */
 733                lapic_timer_setup(CLOCK_EVT_MODE_SHUTDOWN, levt);
 734
 735                /* Jiffies delta */
 736                deltaj = lapic_cal_j2 - lapic_cal_j1;
 737                apic_printk(APIC_VERBOSE, "... jiffies delta = %lu\n", deltaj);
 738
 739                /* Check, if the jiffies result is consistent */
 740                if (deltaj >= LAPIC_CAL_LOOPS-2 && deltaj <= LAPIC_CAL_LOOPS+2)
 741                        apic_printk(APIC_VERBOSE, "... jiffies result ok\n");
 742                else
 743                        levt->features |= CLOCK_EVT_FEAT_DUMMY;
 744        } else
 745                local_irq_enable();
 746
 747        if (levt->features & CLOCK_EVT_FEAT_DUMMY) {
 748                pr_warning("APIC timer disabled due to verification failure\n");
 749                        return -1;
 750        }
 751
 752        return 0;
 753}
 754
 755/*
 756 * Setup the boot APIC
 757 *
 758 * Calibrate and verify the result.
 759 */
 760void __init setup_boot_APIC_clock(void)
 761{
 762        /*
 763         * The local apic timer can be disabled via the kernel
 764         * commandline or from the CPU detection code. Register the lapic
 765         * timer as a dummy clock event source on SMP systems, so the
 766         * broadcast mechanism is used. On UP systems simply ignore it.
 767         */
 768        if (disable_apic_timer) {
 769                pr_info("Disabling APIC timer\n");
 770                /* No broadcast on UP ! */
 771                if (num_possible_cpus() > 1) {
 772                        lapic_clockevent.mult = 1;
 773                        setup_APIC_timer();
 774                }
 775                return;
 776        }
 777
 778        apic_printk(APIC_VERBOSE, "Using local APIC timer interrupts.\n"
 779                    "calibrating APIC timer ...\n");
 780
 781        if (calibrate_APIC_clock()) {
 782                /* No broadcast on UP ! */
 783                if (num_possible_cpus() > 1)
 784                        setup_APIC_timer();
 785                return;
 786        }
 787
 788        /*
 789         * If nmi_watchdog is set to IO_APIC, we need the
 790         * PIT/HPET going.  Otherwise register lapic as a dummy
 791         * device.
 792         */
 793        lapic_clockevent.features &= ~CLOCK_EVT_FEAT_DUMMY;
 794
 795        /* Setup the lapic or request the broadcast */
 796        setup_APIC_timer();
 797}
 798
 799void __cpuinit setup_secondary_APIC_clock(void)
 800{
 801        setup_APIC_timer();
 802}
 803
 804/*
 805 * The guts of the apic timer interrupt
 806 */
 807static void local_apic_timer_interrupt(void)
 808{
 809        int cpu = smp_processor_id();
 810        struct clock_event_device *evt = &per_cpu(lapic_events, cpu);
 811
 812        /*
 813         * Normally we should not be here till LAPIC has been initialized but
 814         * in some cases like kdump, its possible that there is a pending LAPIC
 815         * timer interrupt from previous kernel's context and is delivered in
 816         * new kernel the moment interrupts are enabled.
 817         *
 818         * Interrupts are enabled early and LAPIC is setup much later, hence
 819         * its possible that when we get here evt->event_handler is NULL.
 820         * Check for event_handler being NULL and discard the interrupt as
 821         * spurious.
 822         */
 823        if (!evt->event_handler) {
 824                pr_warning("Spurious LAPIC timer interrupt on cpu %d\n", cpu);
 825                /* Switch it off */
 826                lapic_timer_setup(CLOCK_EVT_MODE_SHUTDOWN, evt);
 827                return;
 828        }
 829
 830        /*
 831         * the NMI deadlock-detector uses this.
 832         */
 833        inc_irq_stat(apic_timer_irqs);
 834
 835        evt->event_handler(evt);
 836}
 837
 838/*
 839 * Local APIC timer interrupt. This is the most natural way for doing
 840 * local interrupts, but local timer interrupts can be emulated by
 841 * broadcast interrupts too. [in case the hw doesn't support APIC timers]
 842 *
 843 * [ if a single-CPU system runs an SMP kernel then we call the local
 844 *   interrupt as well. Thus we cannot inline the local irq ... ]
 845 */
 846void __irq_entry smp_apic_timer_interrupt(struct pt_regs *regs)
 847{
 848        struct pt_regs *old_regs = set_irq_regs(regs);
 849
 850        /*
 851         * NOTE! We'd better ACK the irq immediately,
 852         * because timer handling can be slow.
 853         */
 854        ack_APIC_irq();
 855        /*
 856         * update_process_times() expects us to have done irq_enter().
 857         * Besides, if we don't timer interrupts ignore the global
 858         * interrupt lock, which is the WrongThing (tm) to do.
 859         */
 860        exit_idle();
 861        irq_enter();
 862        local_apic_timer_interrupt();
 863        irq_exit();
 864
 865        set_irq_regs(old_regs);
 866}
 867
 868int setup_profiling_timer(unsigned int multiplier)
 869{
 870        return -EINVAL;
 871}
 872
 873/*
 874 * Local APIC start and shutdown
 875 */
 876
 877/**
 878 * clear_local_APIC - shutdown the local APIC
 879 *
 880 * This is called, when a CPU is disabled and before rebooting, so the state of
 881 * the local APIC has no dangling leftovers. Also used to cleanout any BIOS
 882 * leftovers during boot.
 883 */
 884void clear_local_APIC(void)
 885{
 886        int maxlvt;
 887        u32 v;
 888
 889        /* APIC hasn't been mapped yet */
 890        if (!x2apic_mode && !apic_phys)
 891                return;
 892
 893        maxlvt = lapic_get_maxlvt();
 894        /*
 895         * Masking an LVT entry can trigger a local APIC error
 896         * if the vector is zero. Mask LVTERR first to prevent this.
 897         */
 898        if (maxlvt >= 3) {
 899                v = ERROR_APIC_VECTOR; /* any non-zero vector will do */
 900                apic_write(APIC_LVTERR, v | APIC_LVT_MASKED);
 901        }
 902        /*
 903         * Careful: we have to set masks only first to deassert
 904         * any level-triggered sources.
 905         */
 906        v = apic_read(APIC_LVTT);
 907        apic_write(APIC_LVTT, v | APIC_LVT_MASKED);
 908        v = apic_read(APIC_LVT0);
 909        apic_write(APIC_LVT0, v | APIC_LVT_MASKED);
 910        v = apic_read(APIC_LVT1);
 911        apic_write(APIC_LVT1, v | APIC_LVT_MASKED);
 912        if (maxlvt >= 4) {
 913                v = apic_read(APIC_LVTPC);
 914                apic_write(APIC_LVTPC, v | APIC_LVT_MASKED);
 915        }
 916
 917        /* lets not touch this if we didn't frob it */
 918#ifdef CONFIG_X86_THERMAL_VECTOR
 919        if (maxlvt >= 5) {
 920                v = apic_read(APIC_LVTTHMR);
 921                apic_write(APIC_LVTTHMR, v | APIC_LVT_MASKED);
 922        }
 923#endif
 924#ifdef CONFIG_X86_MCE_INTEL
 925        if (maxlvt >= 6) {
 926                v = apic_read(APIC_LVTCMCI);
 927                if (!(v & APIC_LVT_MASKED))
 928                        apic_write(APIC_LVTCMCI, v | APIC_LVT_MASKED);
 929        }
 930#endif
 931
 932        /*
 933         * Clean APIC state for other OSs:
 934         */
 935        apic_write(APIC_LVTT, APIC_LVT_MASKED);
 936        apic_write(APIC_LVT0, APIC_LVT_MASKED);
 937        apic_write(APIC_LVT1, APIC_LVT_MASKED);
 938        if (maxlvt >= 3)
 939                apic_write(APIC_LVTERR, APIC_LVT_MASKED);
 940        if (maxlvt >= 4)
 941                apic_write(APIC_LVTPC, APIC_LVT_MASKED);
 942
 943        /* Integrated APIC (!82489DX) ? */
 944        if (lapic_is_integrated()) {
 945                if (maxlvt > 3)
 946                        /* Clear ESR due to Pentium errata 3AP and 11AP */
 947                        apic_write(APIC_ESR, 0);
 948                apic_read(APIC_ESR);
 949        }
 950}
 951
 952/**
 953 * disable_local_APIC - clear and disable the local APIC
 954 */
 955void disable_local_APIC(void)
 956{
 957        unsigned int value;
 958
 959        /* APIC hasn't been mapped yet */
 960        if (!x2apic_mode && !apic_phys)
 961                return;
 962
 963        clear_local_APIC();
 964
 965        /*
 966         * Disable APIC (implies clearing of registers
 967         * for 82489DX!).
 968         */
 969        value = apic_read(APIC_SPIV);
 970        value &= ~APIC_SPIV_APIC_ENABLED;
 971        apic_write(APIC_SPIV, value);
 972
 973#ifdef CONFIG_X86_32
 974        /*
 975         * When LAPIC was disabled by the BIOS and enabled by the kernel,
 976         * restore the disabled state.
 977         */
 978        if (enabled_via_apicbase) {
 979                unsigned int l, h;
 980
 981                rdmsr(MSR_IA32_APICBASE, l, h);
 982                l &= ~MSR_IA32_APICBASE_ENABLE;
 983                wrmsr(MSR_IA32_APICBASE, l, h);
 984        }
 985#endif
 986}
 987
 988/*
 989 * If Linux enabled the LAPIC against the BIOS default disable it down before
 990 * re-entering the BIOS on shutdown.  Otherwise the BIOS may get confused and
 991 * not power-off.  Additionally clear all LVT entries before disable_local_APIC
 992 * for the case where Linux didn't enable the LAPIC.
 993 */
 994void lapic_shutdown(void)
 995{
 996        unsigned long flags;
 997
 998        if (!cpu_has_apic && !apic_from_smp_config())
 999                return;
1000
1001        local_irq_save(flags);
1002
1003#ifdef CONFIG_X86_32
1004        if (!enabled_via_apicbase)
1005                clear_local_APIC();
1006        else
1007#endif
1008                disable_local_APIC();
1009
1010
1011        local_irq_restore(flags);
1012}
1013
1014/*
1015 * This is to verify that we're looking at a real local APIC.
1016 * Check these against your board if the CPUs aren't getting
1017 * started for no apparent reason.
1018 */
1019int __init verify_local_APIC(void)
1020{
1021        unsigned int reg0, reg1;
1022
1023        /*
1024         * The version register is read-only in a real APIC.
1025         */
1026        reg0 = apic_read(APIC_LVR);
1027        apic_printk(APIC_DEBUG, "Getting VERSION: %x\n", reg0);
1028        apic_write(APIC_LVR, reg0 ^ APIC_LVR_MASK);
1029        reg1 = apic_read(APIC_LVR);
1030        apic_printk(APIC_DEBUG, "Getting VERSION: %x\n", reg1);
1031
1032        /*
1033         * The two version reads above should print the same
1034         * numbers.  If the second one is different, then we
1035         * poke at a non-APIC.
1036         */
1037        if (reg1 != reg0)
1038                return 0;
1039
1040        /*
1041         * Check if the version looks reasonably.
1042         */
1043        reg1 = GET_APIC_VERSION(reg0);
1044        if (reg1 == 0x00 || reg1 == 0xff)
1045                return 0;
1046        reg1 = lapic_get_maxlvt();
1047        if (reg1 < 0x02 || reg1 == 0xff)
1048                return 0;
1049
1050        /*
1051         * The ID register is read/write in a real APIC.
1052         */
1053        reg0 = apic_read(APIC_ID);
1054        apic_printk(APIC_DEBUG, "Getting ID: %x\n", reg0);
1055        apic_write(APIC_ID, reg0 ^ apic->apic_id_mask);
1056        reg1 = apic_read(APIC_ID);
1057        apic_printk(APIC_DEBUG, "Getting ID: %x\n", reg1);
1058        apic_write(APIC_ID, reg0);
1059        if (reg1 != (reg0 ^ apic->apic_id_mask))
1060                return 0;
1061
1062        /*
1063         * The next two are just to see if we have sane values.
1064         * They're only really relevant if we're in Virtual Wire
1065         * compatibility mode, but most boxes are anymore.
1066         */
1067        reg0 = apic_read(APIC_LVT0);
1068        apic_printk(APIC_DEBUG, "Getting LVT0: %x\n", reg0);
1069        reg1 = apic_read(APIC_LVT1);
1070        apic_printk(APIC_DEBUG, "Getting LVT1: %x\n", reg1);
1071
1072        return 1;
1073}
1074
1075/**
1076 * sync_Arb_IDs - synchronize APIC bus arbitration IDs
1077 */
1078void __init sync_Arb_IDs(void)
1079{
1080        /*
1081         * Unsupported on P4 - see Intel Dev. Manual Vol. 3, Ch. 8.6.1 And not
1082         * needed on AMD.
1083         */
1084        if (modern_apic() || boot_cpu_data.x86_vendor == X86_VENDOR_AMD)
1085                return;
1086
1087        /*
1088         * Wait for idle.
1089         */
1090        apic_wait_icr_idle();
1091
1092        apic_printk(APIC_DEBUG, "Synchronizing Arb IDs.\n");
1093        apic_write(APIC_ICR, APIC_DEST_ALLINC |
1094                        APIC_INT_LEVELTRIG | APIC_DM_INIT);
1095}
1096
1097/*
1098 * An initial setup of the virtual wire mode.
1099 */
1100void __init init_bsp_APIC(void)
1101{
1102        unsigned int value;
1103
1104        /*
1105         * Don't do the setup now if we have a SMP BIOS as the
1106         * through-I/O-APIC virtual wire mode might be active.
1107         */
1108        if (smp_found_config || !cpu_has_apic)
1109                return;
1110
1111        /*
1112         * Do not trust the local APIC being empty at bootup.
1113         */
1114        clear_local_APIC();
1115
1116        /*
1117         * Enable APIC.
1118         */
1119        value = apic_read(APIC_SPIV);
1120        value &= ~APIC_VECTOR_MASK;
1121        value |= APIC_SPIV_APIC_ENABLED;
1122
1123#ifdef CONFIG_X86_32
1124        /* This bit is reserved on P4/Xeon and should be cleared */
1125        if ((boot_cpu_data.x86_vendor == X86_VENDOR_INTEL) &&
1126            (boot_cpu_data.x86 == 15))
1127                value &= ~APIC_SPIV_FOCUS_DISABLED;
1128        else
1129#endif
1130                value |= APIC_SPIV_FOCUS_DISABLED;
1131        value |= SPURIOUS_APIC_VECTOR;
1132        apic_write(APIC_SPIV, value);
1133
1134        /*
1135         * Set up the virtual wire mode.
1136         */
1137        apic_write(APIC_LVT0, APIC_DM_EXTINT);
1138        value = APIC_DM_NMI;
1139        if (!lapic_is_integrated())             /* 82489DX */
1140                value |= APIC_LVT_LEVEL_TRIGGER;
1141        apic_write(APIC_LVT1, value);
1142}
1143
1144static void __cpuinit lapic_setup_esr(void)
1145{
1146        unsigned int oldvalue, value, maxlvt;
1147
1148        if (!lapic_is_integrated()) {
1149                pr_info("No ESR for 82489DX.\n");
1150                return;
1151        }
1152
1153        if (apic->disable_esr) {
1154                /*
1155                 * Something untraceable is creating bad interrupts on
1156                 * secondary quads ... for the moment, just leave the
1157                 * ESR disabled - we can't do anything useful with the
1158                 * errors anyway - mbligh
1159                 */
1160                pr_info("Leaving ESR disabled.\n");
1161                return;
1162        }
1163
1164        maxlvt = lapic_get_maxlvt();
1165        if (maxlvt > 3)         /* Due to the Pentium erratum 3AP. */
1166                apic_write(APIC_ESR, 0);
1167        oldvalue = apic_read(APIC_ESR);
1168
1169        /* enables sending errors */
1170        value = ERROR_APIC_VECTOR;
1171        apic_write(APIC_LVTERR, value);
1172
1173        /*
1174         * spec says clear errors after enabling vector.
1175         */
1176        if (maxlvt > 3)
1177                apic_write(APIC_ESR, 0);
1178        value = apic_read(APIC_ESR);
1179        if (value != oldvalue)
1180                apic_printk(APIC_VERBOSE, "ESR value before enabling "
1181                        "vector: 0x%08x  after: 0x%08x\n",
1182                        oldvalue, value);
1183}
1184
1185/**
1186 * setup_local_APIC - setup the local APIC
1187 *
1188 * Used to setup local APIC while initializing BSP or bringin up APs.
1189 * Always called with preemption disabled.
1190 */
1191void __cpuinit setup_local_APIC(void)
1192{
1193        int cpu = smp_processor_id();
1194        unsigned int value, queued;
1195        int i, j, acked = 0;
1196        unsigned long long tsc = 0, ntsc;
1197        long long max_loops = cpu_khz;
1198
1199        if (cpu_has_tsc)
1200                rdtscll(tsc);
1201
1202        if (disable_apic) {
1203                disable_ioapic_support();
1204                return;
1205        }
1206
1207#ifdef CONFIG_X86_32
1208        /* Pound the ESR really hard over the head with a big hammer - mbligh */
1209        if (lapic_is_integrated() && apic->disable_esr) {
1210                apic_write(APIC_ESR, 0);
1211                apic_write(APIC_ESR, 0);
1212                apic_write(APIC_ESR, 0);
1213                apic_write(APIC_ESR, 0);
1214        }
1215#endif
1216        perf_events_lapic_init();
1217
1218        /*
1219         * Double-check whether this APIC is really registered.
1220         * This is meaningless in clustered apic mode, so we skip it.
1221         */
1222        BUG_ON(!apic->apic_id_registered());
1223
1224        /*
1225         * Intel recommends to set DFR, LDR and TPR before enabling
1226         * an APIC.  See e.g. "AP-388 82489DX User's Manual" (Intel
1227         * document number 292116).  So here it goes...
1228         */
1229        apic->init_apic_ldr();
1230
1231#ifdef CONFIG_X86_32
1232        /*
1233         * APIC LDR is initialized.  If logical_apicid mapping was
1234         * initialized during get_smp_config(), make sure it matches the
1235         * actual value.
1236         */
1237        i = early_per_cpu(x86_cpu_to_logical_apicid, cpu);
1238        WARN_ON(i != BAD_APICID && i != logical_smp_processor_id());
1239        /* always use the value from LDR */
1240        early_per_cpu(x86_cpu_to_logical_apicid, cpu) =
1241                logical_smp_processor_id();
1242
1243        /*
1244         * Some NUMA implementations (NUMAQ) don't initialize apicid to
1245         * node mapping during NUMA init.  Now that logical apicid is
1246         * guaranteed to be known, give it another chance.  This is already
1247         * a bit too late - percpu allocation has already happened without
1248         * proper NUMA affinity.
1249         */
1250        if (apic->x86_32_numa_cpu_node)
1251                set_apicid_to_node(early_per_cpu(x86_cpu_to_apicid, cpu),
1252                                   apic->x86_32_numa_cpu_node(cpu));
1253#endif
1254
1255        /*
1256         * Set Task Priority to 'accept all'. We never change this
1257         * later on.
1258         */
1259        value = apic_read(APIC_TASKPRI);
1260        value &= ~APIC_TPRI_MASK;
1261        apic_write(APIC_TASKPRI, value);
1262
1263        /*
1264         * After a crash, we no longer service the interrupts and a pending
1265         * interrupt from previous kernel might still have ISR bit set.
1266         *
1267         * Most probably by now CPU has serviced that pending interrupt and
1268         * it might not have done the ack_APIC_irq() because it thought,
1269         * interrupt came from i8259 as ExtInt. LAPIC did not get EOI so it
1270         * does not clear the ISR bit and cpu thinks it has already serivced
1271         * the interrupt. Hence a vector might get locked. It was noticed
1272         * for timer irq (vector 0x31). Issue an extra EOI to clear ISR.
1273         */
1274        do {
1275                queued = 0;
1276                for (i = APIC_ISR_NR - 1; i >= 0; i--)
1277                        queued |= apic_read(APIC_IRR + i*0x10);
1278
1279                for (i = APIC_ISR_NR - 1; i >= 0; i--) {
1280                        value = apic_read(APIC_ISR + i*0x10);
1281                        for (j = 31; j >= 0; j--) {
1282                                if (value & (1<<j)) {
1283                                        ack_APIC_irq();
1284                                        acked++;
1285                                }
1286                        }
1287                }
1288                if (acked > 256) {
1289                        printk(KERN_ERR "LAPIC pending interrupts after %d EOI\n",
1290                               acked);
1291                        break;
1292                }
1293                if (cpu_has_tsc) {
1294                        rdtscll(ntsc);
1295                        max_loops = (cpu_khz << 10) - (ntsc - tsc);
1296                } else
1297                        max_loops--;
1298        } while (queued && max_loops > 0);
1299        WARN_ON(max_loops <= 0);
1300
1301        /*
1302         * Now that we are all set up, enable the APIC
1303         */
1304        value = apic_read(APIC_SPIV);
1305        value &= ~APIC_VECTOR_MASK;
1306        /*
1307         * Enable APIC
1308         */
1309        value |= APIC_SPIV_APIC_ENABLED;
1310
1311#ifdef CONFIG_X86_32
1312        /*
1313         * Some unknown Intel IO/APIC (or APIC) errata is biting us with
1314         * certain networking cards. If high frequency interrupts are
1315         * happening on a particular IOAPIC pin, plus the IOAPIC routing
1316         * entry is masked/unmasked at a high rate as well then sooner or
1317         * later IOAPIC line gets 'stuck', no more interrupts are received
1318         * from the device. If focus CPU is disabled then the hang goes
1319         * away, oh well :-(
1320         *
1321         * [ This bug can be reproduced easily with a level-triggered
1322         *   PCI Ne2000 networking cards and PII/PIII processors, dual
1323         *   BX chipset. ]
1324         */
1325        /*
1326         * Actually disabling the focus CPU check just makes the hang less
1327         * frequent as it makes the interrupt distributon model be more
1328         * like LRU than MRU (the short-term load is more even across CPUs).
1329         * See also the comment in end_level_ioapic_irq().  --macro
1330         */
1331
1332        /*
1333         * - enable focus processor (bit==0)
1334         * - 64bit mode always use processor focus
1335         *   so no need to set it
1336         */
1337        value &= ~APIC_SPIV_FOCUS_DISABLED;
1338#endif
1339
1340        /*
1341         * Set spurious IRQ vector
1342         */
1343        value |= SPURIOUS_APIC_VECTOR;
1344        apic_write(APIC_SPIV, value);
1345
1346        /*
1347         * Set up LVT0, LVT1:
1348         *
1349         * set up through-local-APIC on the BP's LINT0. This is not
1350         * strictly necessary in pure symmetric-IO mode, but sometimes
1351         * we delegate interrupts to the 8259A.
1352         */
1353        /*
1354         * TODO: set up through-local-APIC from through-I/O-APIC? --macro
1355         */
1356        value = apic_read(APIC_LVT0) & APIC_LVT_MASKED;
1357        if (!cpu && (pic_mode || !value)) {
1358                value = APIC_DM_EXTINT;
1359                apic_printk(APIC_VERBOSE, "enabled ExtINT on CPU#%d\n", cpu);
1360        } else {
1361                value = APIC_DM_EXTINT | APIC_LVT_MASKED;
1362                apic_printk(APIC_VERBOSE, "masked ExtINT on CPU#%d\n", cpu);
1363        }
1364        apic_write(APIC_LVT0, value);
1365
1366        /*
1367         * only the BP should see the LINT1 NMI signal, obviously.
1368         */
1369        if (!cpu)
1370                value = APIC_DM_NMI;
1371        else
1372                value = APIC_DM_NMI | APIC_LVT_MASKED;
1373        if (!lapic_is_integrated())             /* 82489DX */
1374                value |= APIC_LVT_LEVEL_TRIGGER;
1375        apic_write(APIC_LVT1, value);
1376
1377#ifdef CONFIG_X86_MCE_INTEL
1378        /* Recheck CMCI information after local APIC is up on CPU #0 */
1379        if (!cpu)
1380                cmci_recheck();
1381#endif
1382}
1383
1384void __cpuinit end_local_APIC_setup(void)
1385{
1386        lapic_setup_esr();
1387
1388#ifdef CONFIG_X86_32
1389        {
1390                unsigned int value;
1391                /* Disable the local apic timer */
1392                value = apic_read(APIC_LVTT);
1393                value |= (APIC_LVT_MASKED | LOCAL_TIMER_VECTOR);
1394                apic_write(APIC_LVTT, value);
1395        }
1396#endif
1397
1398        apic_pm_activate();
1399}
1400
1401void __init bsp_end_local_APIC_setup(void)
1402{
1403        end_local_APIC_setup();
1404
1405        /*
1406         * Now that local APIC setup is completed for BP, configure the fault
1407         * handling for interrupt remapping.
1408         */
1409        if (intr_remapping_enabled)
1410                enable_drhd_fault_handling();
1411
1412}
1413
1414#ifdef CONFIG_X86_X2APIC
1415void check_x2apic(void)
1416{
1417        if (x2apic_enabled()) {
1418                pr_info("x2apic enabled by BIOS, switching to x2apic ops\n");
1419                x2apic_preenabled = x2apic_mode = 1;
1420        }
1421}
1422
1423void enable_x2apic(void)
1424{
1425        int msr, msr2;
1426
1427        if (!x2apic_mode)
1428                return;
1429
1430        rdmsr(MSR_IA32_APICBASE, msr, msr2);
1431        if (!(msr & X2APIC_ENABLE)) {
1432                printk_once(KERN_INFO "Enabling x2apic\n");
1433                wrmsr(MSR_IA32_APICBASE, msr | X2APIC_ENABLE, msr2);
1434        }
1435}
1436#endif /* CONFIG_X86_X2APIC */
1437
1438int __init enable_IR(void)
1439{
1440#ifdef CONFIG_INTR_REMAP
1441        if (!intr_remapping_supported()) {
1442                pr_debug("intr-remapping not supported\n");
1443                return 0;
1444        }
1445
1446        if (!x2apic_preenabled && skip_ioapic_setup) {
1447                pr_info("Skipped enabling intr-remap because of skipping "
1448                        "io-apic setup\n");
1449                return 0;
1450        }
1451
1452        if (enable_intr_remapping(x2apic_supported()))
1453                return 0;
1454
1455        pr_info("Enabled Interrupt-remapping\n");
1456
1457        return 1;
1458
1459#endif
1460        return 0;
1461}
1462
1463void __init enable_IR_x2apic(void)
1464{
1465        unsigned long flags;
1466        int ret, x2apic_enabled = 0;
1467        int dmar_table_init_ret;
1468
1469        dmar_table_init_ret = dmar_table_init();
1470        if (dmar_table_init_ret && !x2apic_supported())
1471                return;
1472
1473        ret = save_ioapic_entries();
1474        if (ret) {
1475                pr_info("Saving IO-APIC state failed: %d\n", ret);
1476                goto out;
1477        }
1478
1479        local_irq_save(flags);
1480        legacy_pic->mask_all();
1481        mask_ioapic_entries();
1482
1483        if (dmar_table_init_ret)
1484                ret = 0;
1485        else
1486                ret = enable_IR();
1487
1488        if (!ret) {
1489                /* IR is required if there is APIC ID > 255 even when running
1490                 * under KVM
1491                 */
1492                if (max_physical_apicid > 255 ||
1493                    !hypervisor_x2apic_available())
1494                        goto nox2apic;
1495                /*
1496                 * without IR all CPUs can be addressed by IOAPIC/MSI
1497                 * only in physical mode
1498                 */
1499                x2apic_force_phys();
1500        }
1501
1502        x2apic_enabled = 1;
1503
1504        if (x2apic_supported() && !x2apic_mode) {
1505                x2apic_mode = 1;
1506                enable_x2apic();
1507                pr_info("Enabled x2apic\n");
1508        }
1509
1510nox2apic:
1511        if (!ret) /* IR enabling failed */
1512                restore_ioapic_entries();
1513        legacy_pic->restore_mask();
1514        local_irq_restore(flags);
1515
1516out:
1517        if (x2apic_enabled)
1518                return;
1519
1520        if (x2apic_preenabled)
1521                panic("x2apic: enabled by BIOS but kernel init failed.");
1522        else if (cpu_has_x2apic)
1523                pr_info("Not enabling x2apic, Intr-remapping init failed.\n");
1524}
1525
1526#ifdef CONFIG_X86_64
1527/*
1528 * Detect and enable local APICs on non-SMP boards.
1529 * Original code written by Keir Fraser.
1530 * On AMD64 we trust the BIOS - if it says no APIC it is likely
1531 * not correctly set up (usually the APIC timer won't work etc.)
1532 */
1533static int __init detect_init_APIC(void)
1534{
1535        if (!cpu_has_apic) {
1536                pr_info("No local APIC present\n");
1537                return -1;
1538        }
1539
1540        mp_lapic_addr = APIC_DEFAULT_PHYS_BASE;
1541        return 0;
1542}
1543#else
1544
1545static int __init apic_verify(void)
1546{
1547        u32 features, h, l;
1548
1549        /*
1550         * The APIC feature bit should now be enabled
1551         * in `cpuid'
1552         */
1553        features = cpuid_edx(1);
1554        if (!(features & (1 << X86_FEATURE_APIC))) {
1555                pr_warning("Could not enable APIC!\n");
1556                return -1;
1557        }
1558        set_cpu_cap(&boot_cpu_data, X86_FEATURE_APIC);
1559        mp_lapic_addr = APIC_DEFAULT_PHYS_BASE;
1560
1561        /* The BIOS may have set up the APIC at some other address */
1562        rdmsr(MSR_IA32_APICBASE, l, h);
1563        if (l & MSR_IA32_APICBASE_ENABLE)
1564                mp_lapic_addr = l & MSR_IA32_APICBASE_BASE;
1565
1566        pr_info("Found and enabled local APIC!\n");
1567        return 0;
1568}
1569
1570int __init apic_force_enable(unsigned long addr)
1571{
1572        u32 h, l;
1573
1574        if (disable_apic)
1575                return -1;
1576
1577        /*
1578         * Some BIOSes disable the local APIC in the APIC_BASE
1579         * MSR. This can only be done in software for Intel P6 or later
1580         * and AMD K7 (Model > 1) or later.
1581         */
1582        rdmsr(MSR_IA32_APICBASE, l, h);
1583        if (!(l & MSR_IA32_APICBASE_ENABLE)) {
1584                pr_info("Local APIC disabled by BIOS -- reenabling.\n");
1585                l &= ~MSR_IA32_APICBASE_BASE;
1586                l |= MSR_IA32_APICBASE_ENABLE | addr;
1587                wrmsr(MSR_IA32_APICBASE, l, h);
1588                enabled_via_apicbase = 1;
1589        }
1590        return apic_verify();
1591}
1592
1593/*
1594 * Detect and initialize APIC
1595 */
1596static int __init detect_init_APIC(void)
1597{
1598        /* Disabled by kernel option? */
1599        if (disable_apic)
1600                return -1;
1601
1602        switch (boot_cpu_data.x86_vendor) {
1603        case X86_VENDOR_AMD:
1604                if ((boot_cpu_data.x86 == 6 && boot_cpu_data.x86_model > 1) ||
1605                    (boot_cpu_data.x86 >= 15))
1606                        break;
1607                goto no_apic;
1608        case X86_VENDOR_INTEL:
1609                if (boot_cpu_data.x86 == 6 || boot_cpu_data.x86 == 15 ||
1610                    (boot_cpu_data.x86 == 5 && cpu_has_apic))
1611                        break;
1612                goto no_apic;
1613        default:
1614                goto no_apic;
1615        }
1616
1617        if (!cpu_has_apic) {
1618                /*
1619                 * Over-ride BIOS and try to enable the local APIC only if
1620                 * "lapic" specified.
1621                 */
1622                if (!force_enable_local_apic) {
1623                        pr_info("Local APIC disabled by BIOS -- "
1624                                "you can enable it with \"lapic\"\n");
1625                        return -1;
1626                }
1627                if (apic_force_enable(APIC_DEFAULT_PHYS_BASE))
1628                        return -1;
1629        } else {
1630                if (apic_verify())
1631                        return -1;
1632        }
1633
1634        apic_pm_activate();
1635
1636        return 0;
1637
1638no_apic:
1639        pr_info("No local APIC present or hardware disabled\n");
1640        return -1;
1641}
1642#endif
1643
1644/**
1645 * init_apic_mappings - initialize APIC mappings
1646 */
1647void __init init_apic_mappings(void)
1648{
1649        unsigned int new_apicid;
1650
1651        if (x2apic_mode) {
1652                boot_cpu_physical_apicid = read_apic_id();
1653                return;
1654        }
1655
1656        /* If no local APIC can be found return early */
1657        if (!smp_found_config && detect_init_APIC()) {
1658                /* lets NOP'ify apic operations */
1659                pr_info("APIC: disable apic facility\n");
1660                apic_disable();
1661        } else {
1662                apic_phys = mp_lapic_addr;
1663
1664                /*
1665                 * acpi lapic path already maps that address in
1666                 * acpi_register_lapic_address()
1667                 */
1668                if (!acpi_lapic && !smp_found_config)
1669                        register_lapic_address(apic_phys);
1670        }
1671
1672        /*
1673         * Fetch the APIC ID of the BSP in case we have a
1674         * default configuration (or the MP table is broken).
1675         */
1676        new_apicid = read_apic_id();
1677        if (boot_cpu_physical_apicid != new_apicid) {
1678                boot_cpu_physical_apicid = new_apicid;
1679                /*
1680                 * yeah -- we lie about apic_version
1681                 * in case if apic was disabled via boot option
1682                 * but it's not a problem for SMP compiled kernel
1683                 * since smp_sanity_check is prepared for such a case
1684                 * and disable smp mode
1685                 */
1686                apic_version[new_apicid] =
1687                         GET_APIC_VERSION(apic_read(APIC_LVR));
1688        }
1689}
1690
1691void __init register_lapic_address(unsigned long address)
1692{
1693        mp_lapic_addr = address;
1694
1695        if (!x2apic_mode) {
1696                set_fixmap_nocache(FIX_APIC_BASE, address);
1697                apic_printk(APIC_VERBOSE, "mapped APIC to %16lx (%16lx)\n",
1698                            APIC_BASE, mp_lapic_addr);
1699        }
1700        if (boot_cpu_physical_apicid == -1U) {
1701                boot_cpu_physical_apicid  = read_apic_id();
1702                apic_version[boot_cpu_physical_apicid] =
1703                         GET_APIC_VERSION(apic_read(APIC_LVR));
1704        }
1705}
1706
1707/*
1708 * This initializes the IO-APIC and APIC hardware if this is
1709 * a UP kernel.
1710 */
1711int apic_version[MAX_LOCAL_APIC];
1712
1713int __init APIC_init_uniprocessor(void)
1714{
1715        if (disable_apic) {
1716                pr_info("Apic disabled\n");
1717                return -1;
1718        }
1719#ifdef CONFIG_X86_64
1720        if (!cpu_has_apic) {
1721                disable_apic = 1;
1722                pr_info("Apic disabled by BIOS\n");
1723                return -1;
1724        }
1725#else
1726        if (!smp_found_config && !cpu_has_apic)
1727                return -1;
1728
1729        /*
1730         * Complain if the BIOS pretends there is one.
1731         */
1732        if (!cpu_has_apic &&
1733            APIC_INTEGRATED(apic_version[boot_cpu_physical_apicid])) {
1734                pr_err("BIOS bug, local APIC 0x%x not detected!...\n",
1735                        boot_cpu_physical_apicid);
1736                return -1;
1737        }
1738#endif
1739
1740        default_setup_apic_routing();
1741
1742        verify_local_APIC();
1743        connect_bsp_APIC();
1744
1745#ifdef CONFIG_X86_64
1746        apic_write(APIC_ID, SET_APIC_ID(boot_cpu_physical_apicid));
1747#else
1748        /*
1749         * Hack: In case of kdump, after a crash, kernel might be booting
1750         * on a cpu with non-zero lapic id. But boot_cpu_physical_apicid
1751         * might be zero if read from MP tables. Get it from LAPIC.
1752         */
1753# ifdef CONFIG_CRASH_DUMP
1754        boot_cpu_physical_apicid = read_apic_id();
1755# endif
1756#endif
1757        physid_set_mask_of_physid(boot_cpu_physical_apicid, &phys_cpu_present_map);
1758        setup_local_APIC();
1759
1760#ifdef CONFIG_X86_IO_APIC
1761        /*
1762         * Now enable IO-APICs, actually call clear_IO_APIC
1763         * We need clear_IO_APIC before enabling error vector
1764         */
1765        if (!skip_ioapic_setup && nr_ioapics)
1766                enable_IO_APIC();
1767#endif
1768
1769        bsp_end_local_APIC_setup();
1770
1771#ifdef CONFIG_X86_IO_APIC
1772        if (smp_found_config && !skip_ioapic_setup && nr_ioapics)
1773                setup_IO_APIC();
1774        else {
1775                nr_ioapics = 0;
1776        }
1777#endif
1778
1779        x86_init.timers.setup_percpu_clockev();
1780        return 0;
1781}
1782
1783/*
1784 * Local APIC interrupts
1785 */
1786
1787/*
1788 * This interrupt should _never_ happen with our APIC/SMP architecture
1789 */
1790void smp_spurious_interrupt(struct pt_regs *regs)
1791{
1792        u32 v;
1793
1794        exit_idle();
1795        irq_enter();
1796        /*
1797         * Check if this really is a spurious interrupt and ACK it
1798         * if it is a vectored one.  Just in case...
1799         * Spurious interrupts should not be ACKed.
1800         */
1801        v = apic_read(APIC_ISR + ((SPURIOUS_APIC_VECTOR & ~0x1f) >> 1));
1802        if (v & (1 << (SPURIOUS_APIC_VECTOR & 0x1f)))
1803                ack_APIC_irq();
1804
1805        inc_irq_stat(irq_spurious_count);
1806
1807        /* see sw-dev-man vol 3, chapter 7.4.13.5 */
1808        pr_info("spurious APIC interrupt on CPU#%d, "
1809                "should never happen.\n", smp_processor_id());
1810        irq_exit();
1811}
1812
1813/*
1814 * This interrupt should never happen with our APIC/SMP architecture
1815 */
1816void smp_error_interrupt(struct pt_regs *regs)
1817{
1818        u32 v0, v1;
1819        u32 i = 0;
1820        static const char * const error_interrupt_reason[] = {
1821                "Send CS error",                /* APIC Error Bit 0 */
1822                "Receive CS error",             /* APIC Error Bit 1 */
1823                "Send accept error",            /* APIC Error Bit 2 */
1824                "Receive accept error",         /* APIC Error Bit 3 */
1825                "Redirectable IPI",             /* APIC Error Bit 4 */
1826                "Send illegal vector",          /* APIC Error Bit 5 */
1827                "Received illegal vector",      /* APIC Error Bit 6 */
1828                "Illegal register address",     /* APIC Error Bit 7 */
1829        };
1830
1831        exit_idle();
1832        irq_enter();
1833        /* First tickle the hardware, only then report what went on. -- REW */
1834        v0 = apic_read(APIC_ESR);
1835        apic_write(APIC_ESR, 0);
1836        v1 = apic_read(APIC_ESR);
1837        ack_APIC_irq();
1838        atomic_inc(&irq_err_count);
1839
1840        apic_printk(APIC_DEBUG, KERN_DEBUG "APIC error on CPU%d: %02x(%02x)",
1841                    smp_processor_id(), v0 , v1);
1842
1843        v1 = v1 & 0xff;
1844        while (v1) {
1845                if (v1 & 0x1)
1846                        apic_printk(APIC_DEBUG, KERN_CONT " : %s", error_interrupt_reason[i]);
1847                i++;
1848                v1 >>= 1;
1849        };
1850
1851        apic_printk(APIC_DEBUG, KERN_CONT "\n");
1852
1853        irq_exit();
1854}
1855
1856/**
1857 * connect_bsp_APIC - attach the APIC to the interrupt system
1858 */
1859void __init connect_bsp_APIC(void)
1860{
1861#ifdef CONFIG_X86_32
1862        if (pic_mode) {
1863                /*
1864                 * Do not trust the local APIC being empty at bootup.
1865                 */
1866                clear_local_APIC();
1867                /*
1868                 * PIC mode, enable APIC mode in the IMCR, i.e.  connect BSP's
1869                 * local APIC to INT and NMI lines.
1870                 */
1871                apic_printk(APIC_VERBOSE, "leaving PIC mode, "
1872                                "enabling APIC mode.\n");
1873                imcr_pic_to_apic();
1874        }
1875#endif
1876        if (apic->enable_apic_mode)
1877                apic->enable_apic_mode();
1878}
1879
1880/**
1881 * disconnect_bsp_APIC - detach the APIC from the interrupt system
1882 * @virt_wire_setup:    indicates, whether virtual wire mode is selected
1883 *
1884 * Virtual wire mode is necessary to deliver legacy interrupts even when the
1885 * APIC is disabled.
1886 */
1887void disconnect_bsp_APIC(int virt_wire_setup)
1888{
1889        unsigned int value;
1890
1891#ifdef CONFIG_X86_32
1892        if (pic_mode) {
1893                /*
1894                 * Put the board back into PIC mode (has an effect only on
1895                 * certain older boards).  Note that APIC interrupts, including
1896                 * IPIs, won't work beyond this point!  The only exception are
1897                 * INIT IPIs.
1898                 */
1899                apic_printk(APIC_VERBOSE, "disabling APIC mode, "
1900                                "entering PIC mode.\n");
1901                imcr_apic_to_pic();
1902                return;
1903        }
1904#endif
1905
1906        /* Go back to Virtual Wire compatibility mode */
1907
1908        /* For the spurious interrupt use vector F, and enable it */
1909        value = apic_read(APIC_SPIV);
1910        value &= ~APIC_VECTOR_MASK;
1911        value |= APIC_SPIV_APIC_ENABLED;
1912        value |= 0xf;
1913        apic_write(APIC_SPIV, value);
1914
1915        if (!virt_wire_setup) {
1916                /*
1917                 * For LVT0 make it edge triggered, active high,
1918                 * external and enabled
1919                 */
1920                value = apic_read(APIC_LVT0);
1921                value &= ~(APIC_MODE_MASK | APIC_SEND_PENDING |
1922                        APIC_INPUT_POLARITY | APIC_LVT_REMOTE_IRR |
1923                        APIC_LVT_LEVEL_TRIGGER | APIC_LVT_MASKED);
1924                value |= APIC_LVT_REMOTE_IRR | APIC_SEND_PENDING;
1925                value = SET_APIC_DELIVERY_MODE(value, APIC_MODE_EXTINT);
1926                apic_write(APIC_LVT0, value);
1927        } else {
1928                /* Disable LVT0 */
1929                apic_write(APIC_LVT0, APIC_LVT_MASKED);
1930        }
1931
1932        /*
1933         * For LVT1 make it edge triggered, active high,
1934         * nmi and enabled
1935         */
1936        value = apic_read(APIC_LVT1);
1937        value &= ~(APIC_MODE_MASK | APIC_SEND_PENDING |
1938                        APIC_INPUT_POLARITY | APIC_LVT_REMOTE_IRR |
1939                        APIC_LVT_LEVEL_TRIGGER | APIC_LVT_MASKED);
1940        value |= APIC_LVT_REMOTE_IRR | APIC_SEND_PENDING;
1941        value = SET_APIC_DELIVERY_MODE(value, APIC_MODE_NMI);
1942        apic_write(APIC_LVT1, value);
1943}
1944
1945void __cpuinit generic_processor_info(int apicid, int version)
1946{
1947        int cpu, max = nr_cpu_ids;
1948        bool boot_cpu_detected = physid_isset(boot_cpu_physical_apicid,
1949                                phys_cpu_present_map);
1950
1951        /*
1952         * If boot cpu has not been detected yet, then only allow upto
1953         * nr_cpu_ids - 1 processors and keep one slot free for boot cpu
1954         */
1955        if (!boot_cpu_detected && num_processors >= nr_cpu_ids - 1 &&
1956            apicid != boot_cpu_physical_apicid) {
1957                int thiscpu = max + disabled_cpus - 1;
1958
1959                pr_warning(
1960                        "ACPI: NR_CPUS/possible_cpus limit of %i almost"
1961                        " reached. Keeping one slot for boot cpu."
1962                        "  Processor %d/0x%x ignored.\n", max, thiscpu, apicid);
1963
1964                disabled_cpus++;
1965                return;
1966        }
1967
1968        if (num_processors >= nr_cpu_ids) {
1969                int thiscpu = max + disabled_cpus;
1970
1971                pr_warning(
1972                        "ACPI: NR_CPUS/possible_cpus limit of %i reached."
1973                        "  Processor %d/0x%x ignored.\n", max, thiscpu, apicid);
1974
1975                disabled_cpus++;
1976                return;
1977        }
1978
1979        num_processors++;
1980        if (apicid == boot_cpu_physical_apicid) {
1981                /*
1982                 * x86_bios_cpu_apicid is required to have processors listed
1983                 * in same order as logical cpu numbers. Hence the first
1984                 * entry is BSP, and so on.
1985                 * boot_cpu_init() already hold bit 0 in cpu_present_mask
1986                 * for BSP.
1987                 */
1988                cpu = 0;
1989        } else
1990                cpu = cpumask_next_zero(-1, cpu_present_mask);
1991
1992        /*
1993         * Validate version
1994         */
1995        if (version == 0x0) {
1996                pr_warning("BIOS bug: APIC version is 0 for CPU %d/0x%x, fixing up to 0x10\n",
1997                           cpu, apicid);
1998                version = 0x10;
1999        }
2000        apic_version[apicid] = version;
2001
2002        if (version != apic_version[boot_cpu_physical_apicid]) {
2003                pr_warning("BIOS bug: APIC version mismatch, boot CPU: %x, CPU %d: version %x\n",
2004                        apic_version[boot_cpu_physical_apicid], cpu, version);
2005        }
2006
2007        physid_set(apicid, phys_cpu_present_map);
2008        if (apicid > max_physical_apicid)
2009                max_physical_apicid = apicid;
2010
2011#if defined(CONFIG_SMP) || defined(CONFIG_X86_64)
2012        early_per_cpu(x86_cpu_to_apicid, cpu) = apicid;
2013        early_per_cpu(x86_bios_cpu_apicid, cpu) = apicid;
2014#endif
2015#ifdef CONFIG_X86_32
2016        early_per_cpu(x86_cpu_to_logical_apicid, cpu) =
2017                apic->x86_32_early_logical_apicid(cpu);
2018#endif
2019        set_cpu_possible(cpu, true);
2020        set_cpu_present(cpu, true);
2021}
2022
2023int hard_smp_processor_id(void)
2024{
2025        return read_apic_id();
2026}
2027
2028void default_init_apic_ldr(void)
2029{
2030        unsigned long val;
2031
2032        apic_write(APIC_DFR, APIC_DFR_VALUE);
2033        val = apic_read(APIC_LDR) & ~APIC_LDR_MASK;
2034        val |= SET_APIC_LOGICAL_ID(1UL << smp_processor_id());
2035        apic_write(APIC_LDR, val);
2036}
2037
2038/*
2039 * Power management
2040 */
2041#ifdef CONFIG_PM
2042
2043static struct {
2044        /*
2045         * 'active' is true if the local APIC was enabled by us and
2046         * not the BIOS; this signifies that we are also responsible
2047         * for disabling it before entering apm/acpi suspend
2048         */
2049        int active;
2050        /* r/w apic fields */
2051        unsigned int apic_id;
2052        unsigned int apic_taskpri;
2053        unsigned int apic_ldr;
2054        unsigned int apic_dfr;
2055        unsigned int apic_spiv;
2056        unsigned int apic_lvtt;
2057        unsigned int apic_lvtpc;
2058        unsigned int apic_lvt0;
2059        unsigned int apic_lvt1;
2060        unsigned int apic_lvterr;
2061        unsigned int apic_tmict;
2062        unsigned int apic_tdcr;
2063        unsigned int apic_thmr;
2064} apic_pm_state;
2065
2066static int lapic_suspend(void)
2067{
2068        unsigned long flags;
2069        int maxlvt;
2070
2071        if (!apic_pm_state.active)
2072                return 0;
2073
2074        maxlvt = lapic_get_maxlvt();
2075
2076        apic_pm_state.apic_id = apic_read(APIC_ID);
2077        apic_pm_state.apic_taskpri = apic_read(APIC_TASKPRI);
2078        apic_pm_state.apic_ldr = apic_read(APIC_LDR);
2079        apic_pm_state.apic_dfr = apic_read(APIC_DFR);
2080        apic_pm_state.apic_spiv = apic_read(APIC_SPIV);
2081        apic_pm_state.apic_lvtt = apic_read(APIC_LVTT);
2082        if (maxlvt >= 4)
2083                apic_pm_state.apic_lvtpc = apic_read(APIC_LVTPC);
2084        apic_pm_state.apic_lvt0 = apic_read(APIC_LVT0);
2085        apic_pm_state.apic_lvt1 = apic_read(APIC_LVT1);
2086        apic_pm_state.apic_lvterr = apic_read(APIC_LVTERR);
2087        apic_pm_state.apic_tmict = apic_read(APIC_TMICT);
2088        apic_pm_state.apic_tdcr = apic_read(APIC_TDCR);
2089#ifdef CONFIG_X86_THERMAL_VECTOR
2090        if (maxlvt >= 5)
2091                apic_pm_state.apic_thmr = apic_read(APIC_LVTTHMR);
2092#endif
2093
2094        local_irq_save(flags);
2095        disable_local_APIC();
2096
2097        if (intr_remapping_enabled)
2098                disable_intr_remapping();
2099
2100        local_irq_restore(flags);
2101        return 0;
2102}
2103
2104static void lapic_resume(void)
2105{
2106        unsigned int l, h;
2107        unsigned long flags;
2108        int maxlvt;
2109
2110        if (!apic_pm_state.active)
2111                return;
2112
2113        local_irq_save(flags);
2114        if (intr_remapping_enabled) {
2115                /*
2116                 * IO-APIC and PIC have their own resume routines.
2117                 * We just mask them here to make sure the interrupt
2118                 * subsystem is completely quiet while we enable x2apic
2119                 * and interrupt-remapping.
2120                 */
2121                mask_ioapic_entries();
2122                legacy_pic->mask_all();
2123        }
2124
2125        if (x2apic_mode)
2126                enable_x2apic();
2127        else {
2128                /*
2129                 * Make sure the APICBASE points to the right address
2130                 *
2131                 * FIXME! This will be wrong if we ever support suspend on
2132                 * SMP! We'll need to do this as part of the CPU restore!
2133                 */
2134                rdmsr(MSR_IA32_APICBASE, l, h);
2135                l &= ~MSR_IA32_APICBASE_BASE;
2136                l |= MSR_IA32_APICBASE_ENABLE | mp_lapic_addr;
2137                wrmsr(MSR_IA32_APICBASE, l, h);
2138        }
2139
2140        maxlvt = lapic_get_maxlvt();
2141        apic_write(APIC_LVTERR, ERROR_APIC_VECTOR | APIC_LVT_MASKED);
2142        apic_write(APIC_ID, apic_pm_state.apic_id);
2143        apic_write(APIC_DFR, apic_pm_state.apic_dfr);
2144        apic_write(APIC_LDR, apic_pm_state.apic_ldr);
2145        apic_write(APIC_TASKPRI, apic_pm_state.apic_taskpri);
2146        apic_write(APIC_SPIV, apic_pm_state.apic_spiv);
2147        apic_write(APIC_LVT0, apic_pm_state.apic_lvt0);
2148        apic_write(APIC_LVT1, apic_pm_state.apic_lvt1);
2149#if defined(CONFIG_X86_MCE_P4THERMAL) || defined(CONFIG_X86_MCE_INTEL)
2150        if (maxlvt >= 5)
2151                apic_write(APIC_LVTTHMR, apic_pm_state.apic_thmr);
2152#endif
2153        if (maxlvt >= 4)
2154                apic_write(APIC_LVTPC, apic_pm_state.apic_lvtpc);
2155        apic_write(APIC_LVTT, apic_pm_state.apic_lvtt);
2156        apic_write(APIC_TDCR, apic_pm_state.apic_tdcr);
2157        apic_write(APIC_TMICT, apic_pm_state.apic_tmict);
2158        apic_write(APIC_ESR, 0);
2159        apic_read(APIC_ESR);
2160        apic_write(APIC_LVTERR, apic_pm_state.apic_lvterr);
2161        apic_write(APIC_ESR, 0);
2162        apic_read(APIC_ESR);
2163
2164        if (intr_remapping_enabled)
2165                reenable_intr_remapping(x2apic_mode);
2166
2167        local_irq_restore(flags);
2168}
2169
2170/*
2171 * This device has no shutdown method - fully functioning local APICs
2172 * are needed on every CPU up until machine_halt/restart/poweroff.
2173 */
2174
2175static struct syscore_ops lapic_syscore_ops = {
2176        .resume         = lapic_resume,
2177        .suspend        = lapic_suspend,
2178};
2179
2180static void __cpuinit apic_pm_activate(void)
2181{
2182        apic_pm_state.active = 1;
2183}
2184
2185static int __init init_lapic_sysfs(void)
2186{
2187        /* XXX: remove suspend/resume procs if !apic_pm_state.active? */
2188        if (cpu_has_apic)
2189                register_syscore_ops(&lapic_syscore_ops);
2190
2191        return 0;
2192}
2193
2194/* local apic needs to resume before other devices access its registers. */
2195core_initcall(init_lapic_sysfs);
2196
2197#else   /* CONFIG_PM */
2198
2199static void apic_pm_activate(void) { }
2200
2201#endif  /* CONFIG_PM */
2202
2203#ifdef CONFIG_X86_64
2204
2205static int __cpuinit apic_cluster_num(void)
2206{
2207        int i, clusters, zeros;
2208        unsigned id;
2209        u16 *bios_cpu_apicid;
2210        DECLARE_BITMAP(clustermap, NUM_APIC_CLUSTERS);
2211
2212        bios_cpu_apicid = early_per_cpu_ptr(x86_bios_cpu_apicid);
2213        bitmap_zero(clustermap, NUM_APIC_CLUSTERS);
2214
2215        for (i = 0; i < nr_cpu_ids; i++) {
2216                /* are we being called early in kernel startup? */
2217                if (bios_cpu_apicid) {
2218                        id = bios_cpu_apicid[i];
2219                } else if (i < nr_cpu_ids) {
2220                        if (cpu_present(i))
2221                                id = per_cpu(x86_bios_cpu_apicid, i);
2222                        else
2223                                continue;
2224                } else
2225                        break;
2226
2227                if (id != BAD_APICID)
2228                        __set_bit(APIC_CLUSTERID(id), clustermap);
2229        }
2230
2231        /* Problem:  Partially populated chassis may not have CPUs in some of
2232         * the APIC clusters they have been allocated.  Only present CPUs have
2233         * x86_bios_cpu_apicid entries, thus causing zeroes in the bitmap.
2234         * Since clusters are allocated sequentially, count zeros only if
2235         * they are bounded by ones.
2236         */
2237        clusters = 0;
2238        zeros = 0;
2239        for (i = 0; i < NUM_APIC_CLUSTERS; i++) {
2240                if (test_bit(i, clustermap)) {
2241                        clusters += 1 + zeros;
2242                        zeros = 0;
2243                } else
2244                        ++zeros;
2245        }
2246
2247        return clusters;
2248}
2249
2250static int __cpuinitdata multi_checked;
2251static int __cpuinitdata multi;
2252
2253static int __cpuinit set_multi(const struct dmi_system_id *d)
2254{
2255        if (multi)
2256                return 0;
2257        pr_info("APIC: %s detected, Multi Chassis\n", d->ident);
2258        multi = 1;
2259        return 0;
2260}
2261
2262static const __cpuinitconst struct dmi_system_id multi_dmi_table[] = {
2263        {
2264                .callback = set_multi,
2265                .ident = "IBM System Summit2",
2266                .matches = {
2267                        DMI_MATCH(DMI_SYS_VENDOR, "IBM"),
2268                        DMI_MATCH(DMI_PRODUCT_NAME, "Summit2"),
2269                },
2270        },
2271        {}
2272};
2273
2274static void __cpuinit dmi_check_multi(void)
2275{
2276        if (multi_checked)
2277                return;
2278
2279        dmi_check_system(multi_dmi_table);
2280        multi_checked = 1;
2281}
2282
2283/*
2284 * apic_is_clustered_box() -- Check if we can expect good TSC
2285 *
2286 * Thus far, the major user of this is IBM's Summit2 series:
2287 * Clustered boxes may have unsynced TSC problems if they are
2288 * multi-chassis.
2289 * Use DMI to check them
2290 */
2291__cpuinit int apic_is_clustered_box(void)
2292{
2293        dmi_check_multi();
2294        if (multi)
2295                return 1;
2296
2297        if (!is_vsmp_box())
2298                return 0;
2299
2300        /*
2301         * ScaleMP vSMPowered boxes have one cluster per board and TSCs are
2302         * not guaranteed to be synced between boards
2303         */
2304        if (apic_cluster_num() > 1)
2305                return 1;
2306
2307        return 0;
2308}
2309#endif
2310
2311/*
2312 * APIC command line parameters
2313 */
2314static int __init setup_disableapic(char *arg)
2315{
2316        disable_apic = 1;
2317        setup_clear_cpu_cap(X86_FEATURE_APIC);
2318        return 0;
2319}
2320early_param("disableapic", setup_disableapic);
2321
2322/* same as disableapic, for compatibility */
2323static int __init setup_nolapic(char *arg)
2324{
2325        return setup_disableapic(arg);
2326}
2327early_param("nolapic", setup_nolapic);
2328
2329static int __init parse_lapic_timer_c2_ok(char *arg)
2330{
2331        local_apic_timer_c2_ok = 1;
2332        return 0;
2333}
2334early_param("lapic_timer_c2_ok", parse_lapic_timer_c2_ok);
2335
2336static int __init parse_disable_apic_timer(char *arg)
2337{
2338        disable_apic_timer = 1;
2339        return 0;
2340}
2341early_param("noapictimer", parse_disable_apic_timer);
2342
2343static int __init parse_nolapic_timer(char *arg)
2344{
2345        disable_apic_timer = 1;
2346        return 0;
2347}
2348early_param("nolapic_timer", parse_nolapic_timer);
2349
2350static int __init apic_set_verbosity(char *arg)
2351{
2352        if (!arg)  {
2353#ifdef CONFIG_X86_64
2354                skip_ioapic_setup = 0;
2355                return 0;
2356#endif
2357                return -EINVAL;
2358        }
2359
2360        if (strcmp("debug", arg) == 0)
2361                apic_verbosity = APIC_DEBUG;
2362        else if (strcmp("verbose", arg) == 0)
2363                apic_verbosity = APIC_VERBOSE;
2364        else {
2365                pr_warning("APIC Verbosity level %s not recognised"
2366                        " use apic=verbose or apic=debug\n", arg);
2367                return -EINVAL;
2368        }
2369
2370        return 0;
2371}
2372early_param("apic", apic_set_verbosity);
2373
2374static int __init lapic_insert_resource(void)
2375{
2376        if (!apic_phys)
2377                return -1;
2378
2379        /* Put local APIC into the resource map. */
2380        lapic_resource.start = apic_phys;
2381        lapic_resource.end = lapic_resource.start + PAGE_SIZE - 1;
2382        insert_resource(&iomem_resource, &lapic_resource);
2383
2384        return 0;
2385}
2386
2387/*
2388 * need call insert after e820_reserve_resources()
2389 * that is using request_resource
2390 */
2391late_initcall(lapic_insert_resource);
2392