linux/arch/powerpc/kvm/book3s.c
<<
>>
Prefs
   1/*
   2 * Copyright (C) 2009. SUSE Linux Products GmbH. All rights reserved.
   3 *
   4 * Authors:
   5 *    Alexander Graf <agraf@suse.de>
   6 *    Kevin Wolf <mail@kevin-wolf.de>
   7 *
   8 * Description:
   9 * This file is derived from arch/powerpc/kvm/44x.c,
  10 * by Hollis Blanchard <hollisb@us.ibm.com>.
  11 *
  12 * This program is free software; you can redistribute it and/or modify
  13 * it under the terms of the GNU General Public License, version 2, as
  14 * published by the Free Software Foundation.
  15 */
  16
  17#include <linux/kvm_host.h>
  18#include <linux/err.h>
  19#include <linux/export.h>
  20#include <linux/slab.h>
  21#include <linux/module.h>
  22#include <linux/miscdevice.h>
  23#include <linux/gfp.h>
  24#include <linux/sched.h>
  25#include <linux/vmalloc.h>
  26#include <linux/highmem.h>
  27
  28#include <asm/reg.h>
  29#include <asm/cputable.h>
  30#include <asm/cacheflush.h>
  31#include <linux/uaccess.h>
  32#include <asm/io.h>
  33#include <asm/kvm_ppc.h>
  34#include <asm/kvm_book3s.h>
  35#include <asm/mmu_context.h>
  36#include <asm/page.h>
  37#include <asm/xive.h>
  38
  39#include "book3s.h"
  40#include "trace.h"
  41
  42#define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
  43
  44/* #define EXIT_DEBUG */
  45
  46struct kvm_stats_debugfs_item debugfs_entries[] = {
  47        { "exits",       VCPU_STAT(sum_exits) },
  48        { "mmio",        VCPU_STAT(mmio_exits) },
  49        { "sig",         VCPU_STAT(signal_exits) },
  50        { "sysc",        VCPU_STAT(syscall_exits) },
  51        { "inst_emu",    VCPU_STAT(emulated_inst_exits) },
  52        { "dec",         VCPU_STAT(dec_exits) },
  53        { "ext_intr",    VCPU_STAT(ext_intr_exits) },
  54        { "queue_intr",  VCPU_STAT(queue_intr) },
  55        { "halt_poll_success_ns",       VCPU_STAT(halt_poll_success_ns) },
  56        { "halt_poll_fail_ns",          VCPU_STAT(halt_poll_fail_ns) },
  57        { "halt_wait_ns",               VCPU_STAT(halt_wait_ns) },
  58        { "halt_successful_poll", VCPU_STAT(halt_successful_poll), },
  59        { "halt_attempted_poll", VCPU_STAT(halt_attempted_poll), },
  60        { "halt_successful_wait",       VCPU_STAT(halt_successful_wait) },
  61        { "halt_poll_invalid", VCPU_STAT(halt_poll_invalid) },
  62        { "halt_wakeup", VCPU_STAT(halt_wakeup) },
  63        { "pf_storage",  VCPU_STAT(pf_storage) },
  64        { "sp_storage",  VCPU_STAT(sp_storage) },
  65        { "pf_instruc",  VCPU_STAT(pf_instruc) },
  66        { "sp_instruc",  VCPU_STAT(sp_instruc) },
  67        { "ld",          VCPU_STAT(ld) },
  68        { "ld_slow",     VCPU_STAT(ld_slow) },
  69        { "st",          VCPU_STAT(st) },
  70        { "st_slow",     VCPU_STAT(st_slow) },
  71        { "pthru_all",       VCPU_STAT(pthru_all) },
  72        { "pthru_host",      VCPU_STAT(pthru_host) },
  73        { "pthru_bad_aff",   VCPU_STAT(pthru_bad_aff) },
  74        { NULL }
  75};
  76
  77void kvmppc_unfixup_split_real(struct kvm_vcpu *vcpu)
  78{
  79        if (vcpu->arch.hflags & BOOK3S_HFLAG_SPLIT_HACK) {
  80                ulong pc = kvmppc_get_pc(vcpu);
  81                ulong lr = kvmppc_get_lr(vcpu);
  82                if ((pc & SPLIT_HACK_MASK) == SPLIT_HACK_OFFS)
  83                        kvmppc_set_pc(vcpu, pc & ~SPLIT_HACK_MASK);
  84                if ((lr & SPLIT_HACK_MASK) == SPLIT_HACK_OFFS)
  85                        kvmppc_set_lr(vcpu, lr & ~SPLIT_HACK_MASK);
  86                vcpu->arch.hflags &= ~BOOK3S_HFLAG_SPLIT_HACK;
  87        }
  88}
  89EXPORT_SYMBOL_GPL(kvmppc_unfixup_split_real);
  90
  91static inline unsigned long kvmppc_interrupt_offset(struct kvm_vcpu *vcpu)
  92{
  93        if (!is_kvmppc_hv_enabled(vcpu->kvm))
  94                return to_book3s(vcpu)->hior;
  95        return 0;
  96}
  97
  98static inline void kvmppc_update_int_pending(struct kvm_vcpu *vcpu,
  99                        unsigned long pending_now, unsigned long old_pending)
 100{
 101        if (is_kvmppc_hv_enabled(vcpu->kvm))
 102                return;
 103        if (pending_now)
 104                kvmppc_set_int_pending(vcpu, 1);
 105        else if (old_pending)
 106                kvmppc_set_int_pending(vcpu, 0);
 107}
 108
 109static inline bool kvmppc_critical_section(struct kvm_vcpu *vcpu)
 110{
 111        ulong crit_raw;
 112        ulong crit_r1;
 113        bool crit;
 114
 115        if (is_kvmppc_hv_enabled(vcpu->kvm))
 116                return false;
 117
 118        crit_raw = kvmppc_get_critical(vcpu);
 119        crit_r1 = kvmppc_get_gpr(vcpu, 1);
 120
 121        /* Truncate crit indicators in 32 bit mode */
 122        if (!(kvmppc_get_msr(vcpu) & MSR_SF)) {
 123                crit_raw &= 0xffffffff;
 124                crit_r1 &= 0xffffffff;
 125        }
 126
 127        /* Critical section when crit == r1 */
 128        crit = (crit_raw == crit_r1);
 129        /* ... and we're in supervisor mode */
 130        crit = crit && !(kvmppc_get_msr(vcpu) & MSR_PR);
 131
 132        return crit;
 133}
 134
 135void kvmppc_inject_interrupt(struct kvm_vcpu *vcpu, int vec, u64 flags)
 136{
 137        kvmppc_unfixup_split_real(vcpu);
 138        kvmppc_set_srr0(vcpu, kvmppc_get_pc(vcpu));
 139        kvmppc_set_srr1(vcpu, (kvmppc_get_msr(vcpu) & ~0x783f0000ul) | flags);
 140        kvmppc_set_pc(vcpu, kvmppc_interrupt_offset(vcpu) + vec);
 141        vcpu->arch.mmu.reset_msr(vcpu);
 142}
 143
 144static int kvmppc_book3s_vec2irqprio(unsigned int vec)
 145{
 146        unsigned int prio;
 147
 148        switch (vec) {
 149        case 0x100: prio = BOOK3S_IRQPRIO_SYSTEM_RESET;         break;
 150        case 0x200: prio = BOOK3S_IRQPRIO_MACHINE_CHECK;        break;
 151        case 0x300: prio = BOOK3S_IRQPRIO_DATA_STORAGE;         break;
 152        case 0x380: prio = BOOK3S_IRQPRIO_DATA_SEGMENT;         break;
 153        case 0x400: prio = BOOK3S_IRQPRIO_INST_STORAGE;         break;
 154        case 0x480: prio = BOOK3S_IRQPRIO_INST_SEGMENT;         break;
 155        case 0x500: prio = BOOK3S_IRQPRIO_EXTERNAL;             break;
 156        case 0x600: prio = BOOK3S_IRQPRIO_ALIGNMENT;            break;
 157        case 0x700: prio = BOOK3S_IRQPRIO_PROGRAM;              break;
 158        case 0x800: prio = BOOK3S_IRQPRIO_FP_UNAVAIL;           break;
 159        case 0x900: prio = BOOK3S_IRQPRIO_DECREMENTER;          break;
 160        case 0xc00: prio = BOOK3S_IRQPRIO_SYSCALL;              break;
 161        case 0xd00: prio = BOOK3S_IRQPRIO_DEBUG;                break;
 162        case 0xf20: prio = BOOK3S_IRQPRIO_ALTIVEC;              break;
 163        case 0xf40: prio = BOOK3S_IRQPRIO_VSX;                  break;
 164        case 0xf60: prio = BOOK3S_IRQPRIO_FAC_UNAVAIL;          break;
 165        default:    prio = BOOK3S_IRQPRIO_MAX;                  break;
 166        }
 167
 168        return prio;
 169}
 170
 171void kvmppc_book3s_dequeue_irqprio(struct kvm_vcpu *vcpu,
 172                                          unsigned int vec)
 173{
 174        unsigned long old_pending = vcpu->arch.pending_exceptions;
 175
 176        clear_bit(kvmppc_book3s_vec2irqprio(vec),
 177                  &vcpu->arch.pending_exceptions);
 178
 179        kvmppc_update_int_pending(vcpu, vcpu->arch.pending_exceptions,
 180                                  old_pending);
 181}
 182
 183void kvmppc_book3s_queue_irqprio(struct kvm_vcpu *vcpu, unsigned int vec)
 184{
 185        vcpu->stat.queue_intr++;
 186
 187        set_bit(kvmppc_book3s_vec2irqprio(vec),
 188                &vcpu->arch.pending_exceptions);
 189#ifdef EXIT_DEBUG
 190        printk(KERN_INFO "Queueing interrupt %x\n", vec);
 191#endif
 192}
 193EXPORT_SYMBOL_GPL(kvmppc_book3s_queue_irqprio);
 194
 195void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong flags)
 196{
 197        /* might as well deliver this straight away */
 198        kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_PROGRAM, flags);
 199}
 200EXPORT_SYMBOL_GPL(kvmppc_core_queue_program);
 201
 202void kvmppc_core_queue_fpunavail(struct kvm_vcpu *vcpu)
 203{
 204        /* might as well deliver this straight away */
 205        kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, 0);
 206}
 207
 208void kvmppc_core_queue_vec_unavail(struct kvm_vcpu *vcpu)
 209{
 210        /* might as well deliver this straight away */
 211        kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_ALTIVEC, 0);
 212}
 213
 214void kvmppc_core_queue_vsx_unavail(struct kvm_vcpu *vcpu)
 215{
 216        /* might as well deliver this straight away */
 217        kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_VSX, 0);
 218}
 219
 220void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
 221{
 222        kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
 223}
 224EXPORT_SYMBOL_GPL(kvmppc_core_queue_dec);
 225
 226int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
 227{
 228        return test_bit(BOOK3S_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
 229}
 230EXPORT_SYMBOL_GPL(kvmppc_core_pending_dec);
 231
 232void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
 233{
 234        kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
 235}
 236EXPORT_SYMBOL_GPL(kvmppc_core_dequeue_dec);
 237
 238void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
 239                                struct kvm_interrupt *irq)
 240{
 241        /*
 242         * This case (KVM_INTERRUPT_SET) should never actually arise for
 243         * a pseries guest (because pseries guests expect their interrupt
 244         * controllers to continue asserting an external interrupt request
 245         * until it is acknowledged at the interrupt controller), but is
 246         * included to avoid ABI breakage and potentially for other
 247         * sorts of guest.
 248         *
 249         * There is a subtlety here: HV KVM does not test the
 250         * external_oneshot flag in the code that synthesizes
 251         * external interrupts for the guest just before entering
 252         * the guest.  That is OK even if userspace did do a
 253         * KVM_INTERRUPT_SET on a pseries guest vcpu, because the
 254         * caller (kvm_vcpu_ioctl_interrupt) does a kvm_vcpu_kick()
 255         * which ends up doing a smp_send_reschedule(), which will
 256         * pull the guest all the way out to the host, meaning that
 257         * we will call kvmppc_core_prepare_to_enter() before entering
 258         * the guest again, and that will handle the external_oneshot
 259         * flag correctly.
 260         */
 261        if (irq->irq == KVM_INTERRUPT_SET)
 262                vcpu->arch.external_oneshot = 1;
 263
 264        kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL);
 265}
 266
 267void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu)
 268{
 269        kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL);
 270}
 271
 272void kvmppc_core_queue_data_storage(struct kvm_vcpu *vcpu, ulong dar,
 273                                    ulong flags)
 274{
 275        kvmppc_set_dar(vcpu, dar);
 276        kvmppc_set_dsisr(vcpu, flags);
 277        kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_DATA_STORAGE, 0);
 278}
 279EXPORT_SYMBOL_GPL(kvmppc_core_queue_data_storage);
 280
 281void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu, ulong flags)
 282{
 283        kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_INST_STORAGE, flags);
 284}
 285EXPORT_SYMBOL_GPL(kvmppc_core_queue_inst_storage);
 286
 287static int kvmppc_book3s_irqprio_deliver(struct kvm_vcpu *vcpu,
 288                                         unsigned int priority)
 289{
 290        int deliver = 1;
 291        int vec = 0;
 292        bool crit = kvmppc_critical_section(vcpu);
 293
 294        switch (priority) {
 295        case BOOK3S_IRQPRIO_DECREMENTER:
 296                deliver = (kvmppc_get_msr(vcpu) & MSR_EE) && !crit;
 297                vec = BOOK3S_INTERRUPT_DECREMENTER;
 298                break;
 299        case BOOK3S_IRQPRIO_EXTERNAL:
 300                deliver = (kvmppc_get_msr(vcpu) & MSR_EE) && !crit;
 301                vec = BOOK3S_INTERRUPT_EXTERNAL;
 302                break;
 303        case BOOK3S_IRQPRIO_SYSTEM_RESET:
 304                vec = BOOK3S_INTERRUPT_SYSTEM_RESET;
 305                break;
 306        case BOOK3S_IRQPRIO_MACHINE_CHECK:
 307                vec = BOOK3S_INTERRUPT_MACHINE_CHECK;
 308                break;
 309        case BOOK3S_IRQPRIO_DATA_STORAGE:
 310                vec = BOOK3S_INTERRUPT_DATA_STORAGE;
 311                break;
 312        case BOOK3S_IRQPRIO_INST_STORAGE:
 313                vec = BOOK3S_INTERRUPT_INST_STORAGE;
 314                break;
 315        case BOOK3S_IRQPRIO_DATA_SEGMENT:
 316                vec = BOOK3S_INTERRUPT_DATA_SEGMENT;
 317                break;
 318        case BOOK3S_IRQPRIO_INST_SEGMENT:
 319                vec = BOOK3S_INTERRUPT_INST_SEGMENT;
 320                break;
 321        case BOOK3S_IRQPRIO_ALIGNMENT:
 322                vec = BOOK3S_INTERRUPT_ALIGNMENT;
 323                break;
 324        case BOOK3S_IRQPRIO_PROGRAM:
 325                vec = BOOK3S_INTERRUPT_PROGRAM;
 326                break;
 327        case BOOK3S_IRQPRIO_VSX:
 328                vec = BOOK3S_INTERRUPT_VSX;
 329                break;
 330        case BOOK3S_IRQPRIO_ALTIVEC:
 331                vec = BOOK3S_INTERRUPT_ALTIVEC;
 332                break;
 333        case BOOK3S_IRQPRIO_FP_UNAVAIL:
 334                vec = BOOK3S_INTERRUPT_FP_UNAVAIL;
 335                break;
 336        case BOOK3S_IRQPRIO_SYSCALL:
 337                vec = BOOK3S_INTERRUPT_SYSCALL;
 338                break;
 339        case BOOK3S_IRQPRIO_DEBUG:
 340                vec = BOOK3S_INTERRUPT_TRACE;
 341                break;
 342        case BOOK3S_IRQPRIO_PERFORMANCE_MONITOR:
 343                vec = BOOK3S_INTERRUPT_PERFMON;
 344                break;
 345        case BOOK3S_IRQPRIO_FAC_UNAVAIL:
 346                vec = BOOK3S_INTERRUPT_FAC_UNAVAIL;
 347                break;
 348        default:
 349                deliver = 0;
 350                printk(KERN_ERR "KVM: Unknown interrupt: 0x%x\n", priority);
 351                break;
 352        }
 353
 354#if 0
 355        printk(KERN_INFO "Deliver interrupt 0x%x? %x\n", vec, deliver);
 356#endif
 357
 358        if (deliver)
 359                kvmppc_inject_interrupt(vcpu, vec, 0);
 360
 361        return deliver;
 362}
 363
 364/*
 365 * This function determines if an irqprio should be cleared once issued.
 366 */
 367static bool clear_irqprio(struct kvm_vcpu *vcpu, unsigned int priority)
 368{
 369        switch (priority) {
 370                case BOOK3S_IRQPRIO_DECREMENTER:
 371                        /* DEC interrupts get cleared by mtdec */
 372                        return false;
 373                case BOOK3S_IRQPRIO_EXTERNAL:
 374                        /*
 375                         * External interrupts get cleared by userspace
 376                         * except when set by the KVM_INTERRUPT ioctl with
 377                         * KVM_INTERRUPT_SET (not KVM_INTERRUPT_SET_LEVEL).
 378                         */
 379                        if (vcpu->arch.external_oneshot) {
 380                                vcpu->arch.external_oneshot = 0;
 381                                return true;
 382                        }
 383                        return false;
 384        }
 385
 386        return true;
 387}
 388
 389int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
 390{
 391        unsigned long *pending = &vcpu->arch.pending_exceptions;
 392        unsigned long old_pending = vcpu->arch.pending_exceptions;
 393        unsigned int priority;
 394
 395#ifdef EXIT_DEBUG
 396        if (vcpu->arch.pending_exceptions)
 397                printk(KERN_EMERG "KVM: Check pending: %lx\n", vcpu->arch.pending_exceptions);
 398#endif
 399        priority = __ffs(*pending);
 400        while (priority < BOOK3S_IRQPRIO_MAX) {
 401                if (kvmppc_book3s_irqprio_deliver(vcpu, priority) &&
 402                    clear_irqprio(vcpu, priority)) {
 403                        clear_bit(priority, &vcpu->arch.pending_exceptions);
 404                        break;
 405                }
 406
 407                priority = find_next_bit(pending,
 408                                         BITS_PER_BYTE * sizeof(*pending),
 409                                         priority + 1);
 410        }
 411
 412        /* Tell the guest about our interrupt status */
 413        kvmppc_update_int_pending(vcpu, *pending, old_pending);
 414
 415        return 0;
 416}
 417EXPORT_SYMBOL_GPL(kvmppc_core_prepare_to_enter);
 418
 419kvm_pfn_t kvmppc_gpa_to_pfn(struct kvm_vcpu *vcpu, gpa_t gpa, bool writing,
 420                        bool *writable)
 421{
 422        ulong mp_pa = vcpu->arch.magic_page_pa & KVM_PAM;
 423        gfn_t gfn = gpa >> PAGE_SHIFT;
 424
 425        if (!(kvmppc_get_msr(vcpu) & MSR_SF))
 426                mp_pa = (uint32_t)mp_pa;
 427
 428        /* Magic page override */
 429        gpa &= ~0xFFFULL;
 430        if (unlikely(mp_pa) && unlikely((gpa & KVM_PAM) == mp_pa)) {
 431                ulong shared_page = ((ulong)vcpu->arch.shared) & PAGE_MASK;
 432                kvm_pfn_t pfn;
 433
 434                pfn = (kvm_pfn_t)virt_to_phys((void*)shared_page) >> PAGE_SHIFT;
 435                get_page(pfn_to_page(pfn));
 436                if (writable)
 437                        *writable = true;
 438                return pfn;
 439        }
 440
 441        return gfn_to_pfn_prot(vcpu->kvm, gfn, writing, writable);
 442}
 443EXPORT_SYMBOL_GPL(kvmppc_gpa_to_pfn);
 444
 445int kvmppc_xlate(struct kvm_vcpu *vcpu, ulong eaddr, enum xlate_instdata xlid,
 446                 enum xlate_readwrite xlrw, struct kvmppc_pte *pte)
 447{
 448        bool data = (xlid == XLATE_DATA);
 449        bool iswrite = (xlrw == XLATE_WRITE);
 450        int relocated = (kvmppc_get_msr(vcpu) & (data ? MSR_DR : MSR_IR));
 451        int r;
 452
 453        if (relocated) {
 454                r = vcpu->arch.mmu.xlate(vcpu, eaddr, pte, data, iswrite);
 455        } else {
 456                pte->eaddr = eaddr;
 457                pte->raddr = eaddr & KVM_PAM;
 458                pte->vpage = VSID_REAL | eaddr >> 12;
 459                pte->may_read = true;
 460                pte->may_write = true;
 461                pte->may_execute = true;
 462                r = 0;
 463
 464                if ((kvmppc_get_msr(vcpu) & (MSR_IR | MSR_DR)) == MSR_DR &&
 465                    !data) {
 466                        if ((vcpu->arch.hflags & BOOK3S_HFLAG_SPLIT_HACK) &&
 467                            ((eaddr & SPLIT_HACK_MASK) == SPLIT_HACK_OFFS))
 468                        pte->raddr &= ~SPLIT_HACK_MASK;
 469                }
 470        }
 471
 472        return r;
 473}
 474
 475int kvmppc_load_last_inst(struct kvm_vcpu *vcpu,
 476                enum instruction_fetch_type type, u32 *inst)
 477{
 478        ulong pc = kvmppc_get_pc(vcpu);
 479        int r;
 480
 481        if (type == INST_SC)
 482                pc -= 4;
 483
 484        r = kvmppc_ld(vcpu, &pc, sizeof(u32), inst, false);
 485        if (r == EMULATE_DONE)
 486                return r;
 487        else
 488                return EMULATE_AGAIN;
 489}
 490EXPORT_SYMBOL_GPL(kvmppc_load_last_inst);
 491
 492int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
 493{
 494        return 0;
 495}
 496
 497int kvmppc_subarch_vcpu_init(struct kvm_vcpu *vcpu)
 498{
 499        return 0;
 500}
 501
 502void kvmppc_subarch_vcpu_uninit(struct kvm_vcpu *vcpu)
 503{
 504}
 505
 506int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
 507                                  struct kvm_sregs *sregs)
 508{
 509        int ret;
 510
 511        vcpu_load(vcpu);
 512        ret = vcpu->kvm->arch.kvm_ops->get_sregs(vcpu, sregs);
 513        vcpu_put(vcpu);
 514
 515        return ret;
 516}
 517
 518int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
 519                                  struct kvm_sregs *sregs)
 520{
 521        int ret;
 522
 523        vcpu_load(vcpu);
 524        ret = vcpu->kvm->arch.kvm_ops->set_sregs(vcpu, sregs);
 525        vcpu_put(vcpu);
 526
 527        return ret;
 528}
 529
 530int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
 531{
 532        int i;
 533
 534        regs->pc = kvmppc_get_pc(vcpu);
 535        regs->cr = kvmppc_get_cr(vcpu);
 536        regs->ctr = kvmppc_get_ctr(vcpu);
 537        regs->lr = kvmppc_get_lr(vcpu);
 538        regs->xer = kvmppc_get_xer(vcpu);
 539        regs->msr = kvmppc_get_msr(vcpu);
 540        regs->srr0 = kvmppc_get_srr0(vcpu);
 541        regs->srr1 = kvmppc_get_srr1(vcpu);
 542        regs->pid = vcpu->arch.pid;
 543        regs->sprg0 = kvmppc_get_sprg0(vcpu);
 544        regs->sprg1 = kvmppc_get_sprg1(vcpu);
 545        regs->sprg2 = kvmppc_get_sprg2(vcpu);
 546        regs->sprg3 = kvmppc_get_sprg3(vcpu);
 547        regs->sprg4 = kvmppc_get_sprg4(vcpu);
 548        regs->sprg5 = kvmppc_get_sprg5(vcpu);
 549        regs->sprg6 = kvmppc_get_sprg6(vcpu);
 550        regs->sprg7 = kvmppc_get_sprg7(vcpu);
 551
 552        for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
 553                regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
 554
 555        return 0;
 556}
 557
 558int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
 559{
 560        int i;
 561
 562        kvmppc_set_pc(vcpu, regs->pc);
 563        kvmppc_set_cr(vcpu, regs->cr);
 564        kvmppc_set_ctr(vcpu, regs->ctr);
 565        kvmppc_set_lr(vcpu, regs->lr);
 566        kvmppc_set_xer(vcpu, regs->xer);
 567        kvmppc_set_msr(vcpu, regs->msr);
 568        kvmppc_set_srr0(vcpu, regs->srr0);
 569        kvmppc_set_srr1(vcpu, regs->srr1);
 570        kvmppc_set_sprg0(vcpu, regs->sprg0);
 571        kvmppc_set_sprg1(vcpu, regs->sprg1);
 572        kvmppc_set_sprg2(vcpu, regs->sprg2);
 573        kvmppc_set_sprg3(vcpu, regs->sprg3);
 574        kvmppc_set_sprg4(vcpu, regs->sprg4);
 575        kvmppc_set_sprg5(vcpu, regs->sprg5);
 576        kvmppc_set_sprg6(vcpu, regs->sprg6);
 577        kvmppc_set_sprg7(vcpu, regs->sprg7);
 578
 579        for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
 580                kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
 581
 582        return 0;
 583}
 584
 585int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
 586{
 587        return -ENOTSUPP;
 588}
 589
 590int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
 591{
 592        return -ENOTSUPP;
 593}
 594
 595int kvmppc_get_one_reg(struct kvm_vcpu *vcpu, u64 id,
 596                        union kvmppc_one_reg *val)
 597{
 598        int r = 0;
 599        long int i;
 600
 601        r = vcpu->kvm->arch.kvm_ops->get_one_reg(vcpu, id, val);
 602        if (r == -EINVAL) {
 603                r = 0;
 604                switch (id) {
 605                case KVM_REG_PPC_DAR:
 606                        *val = get_reg_val(id, kvmppc_get_dar(vcpu));
 607                        break;
 608                case KVM_REG_PPC_DSISR:
 609                        *val = get_reg_val(id, kvmppc_get_dsisr(vcpu));
 610                        break;
 611                case KVM_REG_PPC_FPR0 ... KVM_REG_PPC_FPR31:
 612                        i = id - KVM_REG_PPC_FPR0;
 613                        *val = get_reg_val(id, VCPU_FPR(vcpu, i));
 614                        break;
 615                case KVM_REG_PPC_FPSCR:
 616                        *val = get_reg_val(id, vcpu->arch.fp.fpscr);
 617                        break;
 618#ifdef CONFIG_VSX
 619                case KVM_REG_PPC_VSR0 ... KVM_REG_PPC_VSR31:
 620                        if (cpu_has_feature(CPU_FTR_VSX)) {
 621                                i = id - KVM_REG_PPC_VSR0;
 622                                val->vsxval[0] = vcpu->arch.fp.fpr[i][0];
 623                                val->vsxval[1] = vcpu->arch.fp.fpr[i][1];
 624                        } else {
 625                                r = -ENXIO;
 626                        }
 627                        break;
 628#endif /* CONFIG_VSX */
 629                case KVM_REG_PPC_DEBUG_INST:
 630                        *val = get_reg_val(id, INS_TW);
 631                        break;
 632#ifdef CONFIG_KVM_XICS
 633                case KVM_REG_PPC_ICP_STATE:
 634                        if (!vcpu->arch.icp && !vcpu->arch.xive_vcpu) {
 635                                r = -ENXIO;
 636                                break;
 637                        }
 638                        if (xive_enabled())
 639                                *val = get_reg_val(id, kvmppc_xive_get_icp(vcpu));
 640                        else
 641                                *val = get_reg_val(id, kvmppc_xics_get_icp(vcpu));
 642                        break;
 643#endif /* CONFIG_KVM_XICS */
 644                case KVM_REG_PPC_FSCR:
 645                        *val = get_reg_val(id, vcpu->arch.fscr);
 646                        break;
 647                case KVM_REG_PPC_TAR:
 648                        *val = get_reg_val(id, vcpu->arch.tar);
 649                        break;
 650                case KVM_REG_PPC_EBBHR:
 651                        *val = get_reg_val(id, vcpu->arch.ebbhr);
 652                        break;
 653                case KVM_REG_PPC_EBBRR:
 654                        *val = get_reg_val(id, vcpu->arch.ebbrr);
 655                        break;
 656                case KVM_REG_PPC_BESCR:
 657                        *val = get_reg_val(id, vcpu->arch.bescr);
 658                        break;
 659                case KVM_REG_PPC_IC:
 660                        *val = get_reg_val(id, vcpu->arch.ic);
 661                        break;
 662                default:
 663                        r = -EINVAL;
 664                        break;
 665                }
 666        }
 667
 668        return r;
 669}
 670
 671int kvmppc_set_one_reg(struct kvm_vcpu *vcpu, u64 id,
 672                        union kvmppc_one_reg *val)
 673{
 674        int r = 0;
 675        long int i;
 676
 677        r = vcpu->kvm->arch.kvm_ops->set_one_reg(vcpu, id, val);
 678        if (r == -EINVAL) {
 679                r = 0;
 680                switch (id) {
 681                case KVM_REG_PPC_DAR:
 682                        kvmppc_set_dar(vcpu, set_reg_val(id, *val));
 683                        break;
 684                case KVM_REG_PPC_DSISR:
 685                        kvmppc_set_dsisr(vcpu, set_reg_val(id, *val));
 686                        break;
 687                case KVM_REG_PPC_FPR0 ... KVM_REG_PPC_FPR31:
 688                        i = id - KVM_REG_PPC_FPR0;
 689                        VCPU_FPR(vcpu, i) = set_reg_val(id, *val);
 690                        break;
 691                case KVM_REG_PPC_FPSCR:
 692                        vcpu->arch.fp.fpscr = set_reg_val(id, *val);
 693                        break;
 694#ifdef CONFIG_VSX
 695                case KVM_REG_PPC_VSR0 ... KVM_REG_PPC_VSR31:
 696                        if (cpu_has_feature(CPU_FTR_VSX)) {
 697                                i = id - KVM_REG_PPC_VSR0;
 698                                vcpu->arch.fp.fpr[i][0] = val->vsxval[0];
 699                                vcpu->arch.fp.fpr[i][1] = val->vsxval[1];
 700                        } else {
 701                                r = -ENXIO;
 702                        }
 703                        break;
 704#endif /* CONFIG_VSX */
 705#ifdef CONFIG_KVM_XICS
 706                case KVM_REG_PPC_ICP_STATE:
 707                        if (!vcpu->arch.icp && !vcpu->arch.xive_vcpu) {
 708                                r = -ENXIO;
 709                                break;
 710                        }
 711                        if (xive_enabled())
 712                                r = kvmppc_xive_set_icp(vcpu, set_reg_val(id, *val));
 713                        else
 714                                r = kvmppc_xics_set_icp(vcpu, set_reg_val(id, *val));
 715                        break;
 716#endif /* CONFIG_KVM_XICS */
 717                case KVM_REG_PPC_FSCR:
 718                        vcpu->arch.fscr = set_reg_val(id, *val);
 719                        break;
 720                case KVM_REG_PPC_TAR:
 721                        vcpu->arch.tar = set_reg_val(id, *val);
 722                        break;
 723                case KVM_REG_PPC_EBBHR:
 724                        vcpu->arch.ebbhr = set_reg_val(id, *val);
 725                        break;
 726                case KVM_REG_PPC_EBBRR:
 727                        vcpu->arch.ebbrr = set_reg_val(id, *val);
 728                        break;
 729                case KVM_REG_PPC_BESCR:
 730                        vcpu->arch.bescr = set_reg_val(id, *val);
 731                        break;
 732                case KVM_REG_PPC_IC:
 733                        vcpu->arch.ic = set_reg_val(id, *val);
 734                        break;
 735                default:
 736                        r = -EINVAL;
 737                        break;
 738                }
 739        }
 740
 741        return r;
 742}
 743
 744void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
 745{
 746        vcpu->kvm->arch.kvm_ops->vcpu_load(vcpu, cpu);
 747}
 748
 749void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
 750{
 751        vcpu->kvm->arch.kvm_ops->vcpu_put(vcpu);
 752}
 753
 754void kvmppc_set_msr(struct kvm_vcpu *vcpu, u64 msr)
 755{
 756        vcpu->kvm->arch.kvm_ops->set_msr(vcpu, msr);
 757}
 758EXPORT_SYMBOL_GPL(kvmppc_set_msr);
 759
 760int kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
 761{
 762        return vcpu->kvm->arch.kvm_ops->vcpu_run(kvm_run, vcpu);
 763}
 764
 765int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
 766                                  struct kvm_translation *tr)
 767{
 768        return 0;
 769}
 770
 771int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
 772                                        struct kvm_guest_debug *dbg)
 773{
 774        vcpu_load(vcpu);
 775        vcpu->guest_debug = dbg->control;
 776        vcpu_put(vcpu);
 777        return 0;
 778}
 779
 780void kvmppc_decrementer_func(struct kvm_vcpu *vcpu)
 781{
 782        kvmppc_core_queue_dec(vcpu);
 783        kvm_vcpu_kick(vcpu);
 784}
 785
 786struct kvm_vcpu *kvmppc_core_vcpu_create(struct kvm *kvm, unsigned int id)
 787{
 788        return kvm->arch.kvm_ops->vcpu_create(kvm, id);
 789}
 790
 791void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
 792{
 793        vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
 794}
 795
 796int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
 797{
 798        return vcpu->kvm->arch.kvm_ops->check_requests(vcpu);
 799}
 800
 801int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
 802{
 803        return kvm->arch.kvm_ops->get_dirty_log(kvm, log);
 804}
 805
 806void kvmppc_core_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
 807                              struct kvm_memory_slot *dont)
 808{
 809        kvm->arch.kvm_ops->free_memslot(free, dont);
 810}
 811
 812int kvmppc_core_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
 813                               unsigned long npages)
 814{
 815        return kvm->arch.kvm_ops->create_memslot(slot, npages);
 816}
 817
 818void kvmppc_core_flush_memslot(struct kvm *kvm, struct kvm_memory_slot *memslot)
 819{
 820        kvm->arch.kvm_ops->flush_memslot(kvm, memslot);
 821}
 822
 823int kvmppc_core_prepare_memory_region(struct kvm *kvm,
 824                                struct kvm_memory_slot *memslot,
 825                                const struct kvm_userspace_memory_region *mem)
 826{
 827        return kvm->arch.kvm_ops->prepare_memory_region(kvm, memslot, mem);
 828}
 829
 830void kvmppc_core_commit_memory_region(struct kvm *kvm,
 831                                const struct kvm_userspace_memory_region *mem,
 832                                const struct kvm_memory_slot *old,
 833                                const struct kvm_memory_slot *new,
 834                                enum kvm_mr_change change)
 835{
 836        kvm->arch.kvm_ops->commit_memory_region(kvm, mem, old, new, change);
 837}
 838
 839int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end)
 840{
 841        return kvm->arch.kvm_ops->unmap_hva_range(kvm, start, end);
 842}
 843
 844int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
 845{
 846        return kvm->arch.kvm_ops->age_hva(kvm, start, end);
 847}
 848
 849int kvm_test_age_hva(struct kvm *kvm, unsigned long hva)
 850{
 851        return kvm->arch.kvm_ops->test_age_hva(kvm, hva);
 852}
 853
 854int kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte)
 855{
 856        kvm->arch.kvm_ops->set_spte_hva(kvm, hva, pte);
 857        return 0;
 858}
 859
 860void kvmppc_mmu_destroy(struct kvm_vcpu *vcpu)
 861{
 862        vcpu->kvm->arch.kvm_ops->mmu_destroy(vcpu);
 863}
 864
 865int kvmppc_core_init_vm(struct kvm *kvm)
 866{
 867
 868#ifdef CONFIG_PPC64
 869        INIT_LIST_HEAD_RCU(&kvm->arch.spapr_tce_tables);
 870        INIT_LIST_HEAD(&kvm->arch.rtas_tokens);
 871#endif
 872
 873        return kvm->arch.kvm_ops->init_vm(kvm);
 874}
 875
 876void kvmppc_core_destroy_vm(struct kvm *kvm)
 877{
 878        kvm->arch.kvm_ops->destroy_vm(kvm);
 879
 880#ifdef CONFIG_PPC64
 881        kvmppc_rtas_tokens_free(kvm);
 882        WARN_ON(!list_empty(&kvm->arch.spapr_tce_tables));
 883#endif
 884}
 885
 886int kvmppc_h_logical_ci_load(struct kvm_vcpu *vcpu)
 887{
 888        unsigned long size = kvmppc_get_gpr(vcpu, 4);
 889        unsigned long addr = kvmppc_get_gpr(vcpu, 5);
 890        u64 buf;
 891        int srcu_idx;
 892        int ret;
 893
 894        if (!is_power_of_2(size) || (size > sizeof(buf)))
 895                return H_TOO_HARD;
 896
 897        srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
 898        ret = kvm_io_bus_read(vcpu, KVM_MMIO_BUS, addr, size, &buf);
 899        srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
 900        if (ret != 0)
 901                return H_TOO_HARD;
 902
 903        switch (size) {
 904        case 1:
 905                kvmppc_set_gpr(vcpu, 4, *(u8 *)&buf);
 906                break;
 907
 908        case 2:
 909                kvmppc_set_gpr(vcpu, 4, be16_to_cpu(*(__be16 *)&buf));
 910                break;
 911
 912        case 4:
 913                kvmppc_set_gpr(vcpu, 4, be32_to_cpu(*(__be32 *)&buf));
 914                break;
 915
 916        case 8:
 917                kvmppc_set_gpr(vcpu, 4, be64_to_cpu(*(__be64 *)&buf));
 918                break;
 919
 920        default:
 921                BUG();
 922        }
 923
 924        return H_SUCCESS;
 925}
 926EXPORT_SYMBOL_GPL(kvmppc_h_logical_ci_load);
 927
 928int kvmppc_h_logical_ci_store(struct kvm_vcpu *vcpu)
 929{
 930        unsigned long size = kvmppc_get_gpr(vcpu, 4);
 931        unsigned long addr = kvmppc_get_gpr(vcpu, 5);
 932        unsigned long val = kvmppc_get_gpr(vcpu, 6);
 933        u64 buf;
 934        int srcu_idx;
 935        int ret;
 936
 937        switch (size) {
 938        case 1:
 939                *(u8 *)&buf = val;
 940                break;
 941
 942        case 2:
 943                *(__be16 *)&buf = cpu_to_be16(val);
 944                break;
 945
 946        case 4:
 947                *(__be32 *)&buf = cpu_to_be32(val);
 948                break;
 949
 950        case 8:
 951                *(__be64 *)&buf = cpu_to_be64(val);
 952                break;
 953
 954        default:
 955                return H_TOO_HARD;
 956        }
 957
 958        srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
 959        ret = kvm_io_bus_write(vcpu, KVM_MMIO_BUS, addr, size, &buf);
 960        srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
 961        if (ret != 0)
 962                return H_TOO_HARD;
 963
 964        return H_SUCCESS;
 965}
 966EXPORT_SYMBOL_GPL(kvmppc_h_logical_ci_store);
 967
 968int kvmppc_core_check_processor_compat(void)
 969{
 970        /*
 971         * We always return 0 for book3s. We check
 972         * for compatibility while loading the HV
 973         * or PR module
 974         */
 975        return 0;
 976}
 977
 978int kvmppc_book3s_hcall_implemented(struct kvm *kvm, unsigned long hcall)
 979{
 980        return kvm->arch.kvm_ops->hcall_implemented(hcall);
 981}
 982
 983#ifdef CONFIG_KVM_XICS
 984int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
 985                bool line_status)
 986{
 987        if (xive_enabled())
 988                return kvmppc_xive_set_irq(kvm, irq_source_id, irq, level,
 989                                           line_status);
 990        else
 991                return kvmppc_xics_set_irq(kvm, irq_source_id, irq, level,
 992                                           line_status);
 993}
 994
 995int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *irq_entry,
 996                              struct kvm *kvm, int irq_source_id,
 997                              int level, bool line_status)
 998{
 999        return kvm_set_irq(kvm, irq_source_id, irq_entry->gsi,
1000                           level, line_status);
1001}
1002static int kvmppc_book3s_set_irq(struct kvm_kernel_irq_routing_entry *e,
1003                                 struct kvm *kvm, int irq_source_id, int level,
1004                                 bool line_status)
1005{
1006        return kvm_set_irq(kvm, irq_source_id, e->gsi, level, line_status);
1007}
1008
1009int kvm_irq_map_gsi(struct kvm *kvm,
1010                    struct kvm_kernel_irq_routing_entry *entries, int gsi)
1011{
1012        entries->gsi = gsi;
1013        entries->type = KVM_IRQ_ROUTING_IRQCHIP;
1014        entries->set = kvmppc_book3s_set_irq;
1015        entries->irqchip.irqchip = 0;
1016        entries->irqchip.pin = gsi;
1017        return 1;
1018}
1019
1020int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin)
1021{
1022        return pin;
1023}
1024
1025#endif /* CONFIG_KVM_XICS */
1026
1027static int kvmppc_book3s_init(void)
1028{
1029        int r;
1030
1031        r = kvm_init(NULL, sizeof(struct kvm_vcpu), 0, THIS_MODULE);
1032        if (r)
1033                return r;
1034#ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1035        r = kvmppc_book3s_init_pr();
1036#endif
1037
1038#ifdef CONFIG_KVM_XICS
1039#ifdef CONFIG_KVM_XIVE
1040        if (xive_enabled()) {
1041                kvmppc_xive_init_module();
1042                kvm_register_device_ops(&kvm_xive_ops, KVM_DEV_TYPE_XICS);
1043        } else
1044#endif
1045                kvm_register_device_ops(&kvm_xics_ops, KVM_DEV_TYPE_XICS);
1046#endif
1047        return r;
1048}
1049
1050static void kvmppc_book3s_exit(void)
1051{
1052#ifdef CONFIG_KVM_XICS
1053        if (xive_enabled())
1054                kvmppc_xive_exit_module();
1055#endif
1056#ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1057        kvmppc_book3s_exit_pr();
1058#endif
1059        kvm_exit();
1060}
1061
1062module_init(kvmppc_book3s_init);
1063module_exit(kvmppc_book3s_exit);
1064
1065/* On 32bit this is our one and only kernel module */
1066#ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1067MODULE_ALIAS_MISCDEV(KVM_MINOR);
1068MODULE_ALIAS("devname:kvm");
1069#endif
1070