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 <asm/tlbflush.h>
  32#include <linux/uaccess.h>
  33#include <asm/io.h>
  34#include <asm/kvm_ppc.h>
  35#include <asm/kvm_book3s.h>
  36#include <asm/mmu_context.h>
  37#include <asm/page.h>
  38#include <asm/xive.h>
  39
  40#include "book3s.h"
  41#include "trace.h"
  42
  43/* #define EXIT_DEBUG */
  44
  45const struct _kvm_stats_desc kvm_vm_stats_desc[] = {
  46        KVM_GENERIC_VM_STATS(),
  47        STATS_DESC_ICOUNTER(VM, num_2M_pages),
  48        STATS_DESC_ICOUNTER(VM, num_1G_pages)
  49};
  50
  51const struct kvm_stats_header kvm_vm_stats_header = {
  52        .name_size = KVM_STATS_NAME_SIZE,
  53        .num_desc = ARRAY_SIZE(kvm_vm_stats_desc),
  54        .id_offset = sizeof(struct kvm_stats_header),
  55        .desc_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE,
  56        .data_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE +
  57                       sizeof(kvm_vm_stats_desc),
  58};
  59
  60const struct _kvm_stats_desc kvm_vcpu_stats_desc[] = {
  61        KVM_GENERIC_VCPU_STATS(),
  62        STATS_DESC_COUNTER(VCPU, sum_exits),
  63        STATS_DESC_COUNTER(VCPU, mmio_exits),
  64        STATS_DESC_COUNTER(VCPU, signal_exits),
  65        STATS_DESC_COUNTER(VCPU, light_exits),
  66        STATS_DESC_COUNTER(VCPU, itlb_real_miss_exits),
  67        STATS_DESC_COUNTER(VCPU, itlb_virt_miss_exits),
  68        STATS_DESC_COUNTER(VCPU, dtlb_real_miss_exits),
  69        STATS_DESC_COUNTER(VCPU, dtlb_virt_miss_exits),
  70        STATS_DESC_COUNTER(VCPU, syscall_exits),
  71        STATS_DESC_COUNTER(VCPU, isi_exits),
  72        STATS_DESC_COUNTER(VCPU, dsi_exits),
  73        STATS_DESC_COUNTER(VCPU, emulated_inst_exits),
  74        STATS_DESC_COUNTER(VCPU, dec_exits),
  75        STATS_DESC_COUNTER(VCPU, ext_intr_exits),
  76        STATS_DESC_COUNTER(VCPU, halt_successful_wait),
  77        STATS_DESC_COUNTER(VCPU, dbell_exits),
  78        STATS_DESC_COUNTER(VCPU, gdbell_exits),
  79        STATS_DESC_COUNTER(VCPU, ld),
  80        STATS_DESC_COUNTER(VCPU, st),
  81        STATS_DESC_COUNTER(VCPU, pf_storage),
  82        STATS_DESC_COUNTER(VCPU, pf_instruc),
  83        STATS_DESC_COUNTER(VCPU, sp_storage),
  84        STATS_DESC_COUNTER(VCPU, sp_instruc),
  85        STATS_DESC_COUNTER(VCPU, queue_intr),
  86        STATS_DESC_COUNTER(VCPU, ld_slow),
  87        STATS_DESC_COUNTER(VCPU, st_slow),
  88        STATS_DESC_COUNTER(VCPU, pthru_all),
  89        STATS_DESC_COUNTER(VCPU, pthru_host),
  90        STATS_DESC_COUNTER(VCPU, pthru_bad_aff)
  91};
  92
  93const struct kvm_stats_header kvm_vcpu_stats_header = {
  94        .name_size = KVM_STATS_NAME_SIZE,
  95        .num_desc = ARRAY_SIZE(kvm_vcpu_stats_desc),
  96        .id_offset = sizeof(struct kvm_stats_header),
  97        .desc_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE,
  98        .data_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE +
  99                       sizeof(kvm_vcpu_stats_desc),
 100};
 101
 102static inline void kvmppc_update_int_pending(struct kvm_vcpu *vcpu,
 103                        unsigned long pending_now, unsigned long old_pending)
 104{
 105        if (is_kvmppc_hv_enabled(vcpu->kvm))
 106                return;
 107        if (pending_now)
 108                kvmppc_set_int_pending(vcpu, 1);
 109        else if (old_pending)
 110                kvmppc_set_int_pending(vcpu, 0);
 111}
 112
 113static inline bool kvmppc_critical_section(struct kvm_vcpu *vcpu)
 114{
 115        ulong crit_raw;
 116        ulong crit_r1;
 117        bool crit;
 118
 119        if (is_kvmppc_hv_enabled(vcpu->kvm))
 120                return false;
 121
 122        crit_raw = kvmppc_get_critical(vcpu);
 123        crit_r1 = kvmppc_get_gpr(vcpu, 1);
 124
 125        /* Truncate crit indicators in 32 bit mode */
 126        if (!(kvmppc_get_msr(vcpu) & MSR_SF)) {
 127                crit_raw &= 0xffffffff;
 128                crit_r1 &= 0xffffffff;
 129        }
 130
 131        /* Critical section when crit == r1 */
 132        crit = (crit_raw == crit_r1);
 133        /* ... and we're in supervisor mode */
 134        crit = crit && !(kvmppc_get_msr(vcpu) & MSR_PR);
 135
 136        return crit;
 137}
 138
 139void kvmppc_inject_interrupt(struct kvm_vcpu *vcpu, int vec, u64 flags)
 140{
 141        vcpu->kvm->arch.kvm_ops->inject_interrupt(vcpu, vec, flags);
 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_machine_check(struct kvm_vcpu *vcpu, ulong flags)
 196{
 197        /* might as well deliver this straight away */
 198        kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_MACHINE_CHECK, flags);
 199}
 200EXPORT_SYMBOL_GPL(kvmppc_core_queue_machine_check);
 201
 202void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong flags)
 203{
 204        /* might as well deliver this straight away */
 205        kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_PROGRAM, flags);
 206}
 207EXPORT_SYMBOL_GPL(kvmppc_core_queue_program);
 208
 209void kvmppc_core_queue_fpunavail(struct kvm_vcpu *vcpu)
 210{
 211        /* might as well deliver this straight away */
 212        kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, 0);
 213}
 214
 215void kvmppc_core_queue_vec_unavail(struct kvm_vcpu *vcpu)
 216{
 217        /* might as well deliver this straight away */
 218        kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_ALTIVEC, 0);
 219}
 220
 221void kvmppc_core_queue_vsx_unavail(struct kvm_vcpu *vcpu)
 222{
 223        /* might as well deliver this straight away */
 224        kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_VSX, 0);
 225}
 226
 227void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
 228{
 229        kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
 230}
 231EXPORT_SYMBOL_GPL(kvmppc_core_queue_dec);
 232
 233int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
 234{
 235        return test_bit(BOOK3S_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
 236}
 237EXPORT_SYMBOL_GPL(kvmppc_core_pending_dec);
 238
 239void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
 240{
 241        kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
 242}
 243EXPORT_SYMBOL_GPL(kvmppc_core_dequeue_dec);
 244
 245void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
 246                                struct kvm_interrupt *irq)
 247{
 248        /*
 249         * This case (KVM_INTERRUPT_SET) should never actually arise for
 250         * a pseries guest (because pseries guests expect their interrupt
 251         * controllers to continue asserting an external interrupt request
 252         * until it is acknowledged at the interrupt controller), but is
 253         * included to avoid ABI breakage and potentially for other
 254         * sorts of guest.
 255         *
 256         * There is a subtlety here: HV KVM does not test the
 257         * external_oneshot flag in the code that synthesizes
 258         * external interrupts for the guest just before entering
 259         * the guest.  That is OK even if userspace did do a
 260         * KVM_INTERRUPT_SET on a pseries guest vcpu, because the
 261         * caller (kvm_vcpu_ioctl_interrupt) does a kvm_vcpu_kick()
 262         * which ends up doing a smp_send_reschedule(), which will
 263         * pull the guest all the way out to the host, meaning that
 264         * we will call kvmppc_core_prepare_to_enter() before entering
 265         * the guest again, and that will handle the external_oneshot
 266         * flag correctly.
 267         */
 268        if (irq->irq == KVM_INTERRUPT_SET)
 269                vcpu->arch.external_oneshot = 1;
 270
 271        kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL);
 272}
 273
 274void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu)
 275{
 276        kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL);
 277}
 278
 279void kvmppc_core_queue_data_storage(struct kvm_vcpu *vcpu, ulong dar,
 280                                    ulong flags)
 281{
 282        kvmppc_set_dar(vcpu, dar);
 283        kvmppc_set_dsisr(vcpu, flags);
 284        kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_DATA_STORAGE, 0);
 285}
 286EXPORT_SYMBOL_GPL(kvmppc_core_queue_data_storage);
 287
 288void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu, ulong flags)
 289{
 290        kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_INST_STORAGE, flags);
 291}
 292EXPORT_SYMBOL_GPL(kvmppc_core_queue_inst_storage);
 293
 294static int kvmppc_book3s_irqprio_deliver(struct kvm_vcpu *vcpu,
 295                                         unsigned int priority)
 296{
 297        int deliver = 1;
 298        int vec = 0;
 299        bool crit = kvmppc_critical_section(vcpu);
 300
 301        switch (priority) {
 302        case BOOK3S_IRQPRIO_DECREMENTER:
 303                deliver = (kvmppc_get_msr(vcpu) & MSR_EE) && !crit;
 304                vec = BOOK3S_INTERRUPT_DECREMENTER;
 305                break;
 306        case BOOK3S_IRQPRIO_EXTERNAL:
 307                deliver = (kvmppc_get_msr(vcpu) & MSR_EE) && !crit;
 308                vec = BOOK3S_INTERRUPT_EXTERNAL;
 309                break;
 310        case BOOK3S_IRQPRIO_SYSTEM_RESET:
 311                vec = BOOK3S_INTERRUPT_SYSTEM_RESET;
 312                break;
 313        case BOOK3S_IRQPRIO_MACHINE_CHECK:
 314                vec = BOOK3S_INTERRUPT_MACHINE_CHECK;
 315                break;
 316        case BOOK3S_IRQPRIO_DATA_STORAGE:
 317                vec = BOOK3S_INTERRUPT_DATA_STORAGE;
 318                break;
 319        case BOOK3S_IRQPRIO_INST_STORAGE:
 320                vec = BOOK3S_INTERRUPT_INST_STORAGE;
 321                break;
 322        case BOOK3S_IRQPRIO_DATA_SEGMENT:
 323                vec = BOOK3S_INTERRUPT_DATA_SEGMENT;
 324                break;
 325        case BOOK3S_IRQPRIO_INST_SEGMENT:
 326                vec = BOOK3S_INTERRUPT_INST_SEGMENT;
 327                break;
 328        case BOOK3S_IRQPRIO_ALIGNMENT:
 329                vec = BOOK3S_INTERRUPT_ALIGNMENT;
 330                break;
 331        case BOOK3S_IRQPRIO_PROGRAM:
 332                vec = BOOK3S_INTERRUPT_PROGRAM;
 333                break;
 334        case BOOK3S_IRQPRIO_VSX:
 335                vec = BOOK3S_INTERRUPT_VSX;
 336                break;
 337        case BOOK3S_IRQPRIO_ALTIVEC:
 338                vec = BOOK3S_INTERRUPT_ALTIVEC;
 339                break;
 340        case BOOK3S_IRQPRIO_FP_UNAVAIL:
 341                vec = BOOK3S_INTERRUPT_FP_UNAVAIL;
 342                break;
 343        case BOOK3S_IRQPRIO_SYSCALL:
 344                vec = BOOK3S_INTERRUPT_SYSCALL;
 345                break;
 346        case BOOK3S_IRQPRIO_DEBUG:
 347                vec = BOOK3S_INTERRUPT_TRACE;
 348                break;
 349        case BOOK3S_IRQPRIO_PERFORMANCE_MONITOR:
 350                vec = BOOK3S_INTERRUPT_PERFMON;
 351                break;
 352        case BOOK3S_IRQPRIO_FAC_UNAVAIL:
 353                vec = BOOK3S_INTERRUPT_FAC_UNAVAIL;
 354                break;
 355        default:
 356                deliver = 0;
 357                printk(KERN_ERR "KVM: Unknown interrupt: 0x%x\n", priority);
 358                break;
 359        }
 360
 361#if 0
 362        printk(KERN_INFO "Deliver interrupt 0x%x? %x\n", vec, deliver);
 363#endif
 364
 365        if (deliver)
 366                kvmppc_inject_interrupt(vcpu, vec, 0);
 367
 368        return deliver;
 369}
 370
 371/*
 372 * This function determines if an irqprio should be cleared once issued.
 373 */
 374static bool clear_irqprio(struct kvm_vcpu *vcpu, unsigned int priority)
 375{
 376        switch (priority) {
 377                case BOOK3S_IRQPRIO_DECREMENTER:
 378                        /* DEC interrupts get cleared by mtdec */
 379                        return false;
 380                case BOOK3S_IRQPRIO_EXTERNAL:
 381                        /*
 382                         * External interrupts get cleared by userspace
 383                         * except when set by the KVM_INTERRUPT ioctl with
 384                         * KVM_INTERRUPT_SET (not KVM_INTERRUPT_SET_LEVEL).
 385                         */
 386                        if (vcpu->arch.external_oneshot) {
 387                                vcpu->arch.external_oneshot = 0;
 388                                return true;
 389                        }
 390                        return false;
 391        }
 392
 393        return true;
 394}
 395
 396int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
 397{
 398        unsigned long *pending = &vcpu->arch.pending_exceptions;
 399        unsigned long old_pending = vcpu->arch.pending_exceptions;
 400        unsigned int priority;
 401
 402#ifdef EXIT_DEBUG
 403        if (vcpu->arch.pending_exceptions)
 404                printk(KERN_EMERG "KVM: Check pending: %lx\n", vcpu->arch.pending_exceptions);
 405#endif
 406        priority = __ffs(*pending);
 407        while (priority < BOOK3S_IRQPRIO_MAX) {
 408                if (kvmppc_book3s_irqprio_deliver(vcpu, priority) &&
 409                    clear_irqprio(vcpu, priority)) {
 410                        clear_bit(priority, &vcpu->arch.pending_exceptions);
 411                        break;
 412                }
 413
 414                priority = find_next_bit(pending,
 415                                         BITS_PER_BYTE * sizeof(*pending),
 416                                         priority + 1);
 417        }
 418
 419        /* Tell the guest about our interrupt status */
 420        kvmppc_update_int_pending(vcpu, *pending, old_pending);
 421
 422        return 0;
 423}
 424EXPORT_SYMBOL_GPL(kvmppc_core_prepare_to_enter);
 425
 426kvm_pfn_t kvmppc_gpa_to_pfn(struct kvm_vcpu *vcpu, gpa_t gpa, bool writing,
 427                        bool *writable)
 428{
 429        ulong mp_pa = vcpu->arch.magic_page_pa & KVM_PAM;
 430        gfn_t gfn = gpa >> PAGE_SHIFT;
 431
 432        if (!(kvmppc_get_msr(vcpu) & MSR_SF))
 433                mp_pa = (uint32_t)mp_pa;
 434
 435        /* Magic page override */
 436        gpa &= ~0xFFFULL;
 437        if (unlikely(mp_pa) && unlikely((gpa & KVM_PAM) == mp_pa)) {
 438                ulong shared_page = ((ulong)vcpu->arch.shared) & PAGE_MASK;
 439                kvm_pfn_t pfn;
 440
 441                pfn = (kvm_pfn_t)virt_to_phys((void*)shared_page) >> PAGE_SHIFT;
 442                get_page(pfn_to_page(pfn));
 443                if (writable)
 444                        *writable = true;
 445                return pfn;
 446        }
 447
 448        return gfn_to_pfn_prot(vcpu->kvm, gfn, writing, writable);
 449}
 450EXPORT_SYMBOL_GPL(kvmppc_gpa_to_pfn);
 451
 452int kvmppc_xlate(struct kvm_vcpu *vcpu, ulong eaddr, enum xlate_instdata xlid,
 453                 enum xlate_readwrite xlrw, struct kvmppc_pte *pte)
 454{
 455        bool data = (xlid == XLATE_DATA);
 456        bool iswrite = (xlrw == XLATE_WRITE);
 457        int relocated = (kvmppc_get_msr(vcpu) & (data ? MSR_DR : MSR_IR));
 458        int r;
 459
 460        if (relocated) {
 461                r = vcpu->arch.mmu.xlate(vcpu, eaddr, pte, data, iswrite);
 462        } else {
 463                pte->eaddr = eaddr;
 464                pte->raddr = eaddr & KVM_PAM;
 465                pte->vpage = VSID_REAL | eaddr >> 12;
 466                pte->may_read = true;
 467                pte->may_write = true;
 468                pte->may_execute = true;
 469                r = 0;
 470
 471                if ((kvmppc_get_msr(vcpu) & (MSR_IR | MSR_DR)) == MSR_DR &&
 472                    !data) {
 473                        if ((vcpu->arch.hflags & BOOK3S_HFLAG_SPLIT_HACK) &&
 474                            ((eaddr & SPLIT_HACK_MASK) == SPLIT_HACK_OFFS))
 475                        pte->raddr &= ~SPLIT_HACK_MASK;
 476                }
 477        }
 478
 479        return r;
 480}
 481
 482int kvmppc_load_last_inst(struct kvm_vcpu *vcpu,
 483                enum instruction_fetch_type type, u32 *inst)
 484{
 485        ulong pc = kvmppc_get_pc(vcpu);
 486        int r;
 487
 488        if (type == INST_SC)
 489                pc -= 4;
 490
 491        r = kvmppc_ld(vcpu, &pc, sizeof(u32), inst, false);
 492        if (r == EMULATE_DONE)
 493                return r;
 494        else
 495                return EMULATE_AGAIN;
 496}
 497EXPORT_SYMBOL_GPL(kvmppc_load_last_inst);
 498
 499int kvmppc_subarch_vcpu_init(struct kvm_vcpu *vcpu)
 500{
 501        return 0;
 502}
 503
 504void kvmppc_subarch_vcpu_uninit(struct kvm_vcpu *vcpu)
 505{
 506}
 507
 508int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
 509                                  struct kvm_sregs *sregs)
 510{
 511        int ret;
 512
 513        vcpu_load(vcpu);
 514        ret = vcpu->kvm->arch.kvm_ops->get_sregs(vcpu, sregs);
 515        vcpu_put(vcpu);
 516
 517        return ret;
 518}
 519
 520int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
 521                                  struct kvm_sregs *sregs)
 522{
 523        int ret;
 524
 525        vcpu_load(vcpu);
 526        ret = vcpu->kvm->arch.kvm_ops->set_sregs(vcpu, sregs);
 527        vcpu_put(vcpu);
 528
 529        return ret;
 530}
 531
 532int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
 533{
 534        int i;
 535
 536        regs->pc = kvmppc_get_pc(vcpu);
 537        regs->cr = kvmppc_get_cr(vcpu);
 538        regs->ctr = kvmppc_get_ctr(vcpu);
 539        regs->lr = kvmppc_get_lr(vcpu);
 540        regs->xer = kvmppc_get_xer(vcpu);
 541        regs->msr = kvmppc_get_msr(vcpu);
 542        regs->srr0 = kvmppc_get_srr0(vcpu);
 543        regs->srr1 = kvmppc_get_srr1(vcpu);
 544        regs->pid = vcpu->arch.pid;
 545        regs->sprg0 = kvmppc_get_sprg0(vcpu);
 546        regs->sprg1 = kvmppc_get_sprg1(vcpu);
 547        regs->sprg2 = kvmppc_get_sprg2(vcpu);
 548        regs->sprg3 = kvmppc_get_sprg3(vcpu);
 549        regs->sprg4 = kvmppc_get_sprg4(vcpu);
 550        regs->sprg5 = kvmppc_get_sprg5(vcpu);
 551        regs->sprg6 = kvmppc_get_sprg6(vcpu);
 552        regs->sprg7 = kvmppc_get_sprg7(vcpu);
 553
 554        for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
 555                regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
 556
 557        return 0;
 558}
 559
 560int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
 561{
 562        int i;
 563
 564        kvmppc_set_pc(vcpu, regs->pc);
 565        kvmppc_set_cr(vcpu, regs->cr);
 566        kvmppc_set_ctr(vcpu, regs->ctr);
 567        kvmppc_set_lr(vcpu, regs->lr);
 568        kvmppc_set_xer(vcpu, regs->xer);
 569        kvmppc_set_msr(vcpu, regs->msr);
 570        kvmppc_set_srr0(vcpu, regs->srr0);
 571        kvmppc_set_srr1(vcpu, regs->srr1);
 572        kvmppc_set_sprg0(vcpu, regs->sprg0);
 573        kvmppc_set_sprg1(vcpu, regs->sprg1);
 574        kvmppc_set_sprg2(vcpu, regs->sprg2);
 575        kvmppc_set_sprg3(vcpu, regs->sprg3);
 576        kvmppc_set_sprg4(vcpu, regs->sprg4);
 577        kvmppc_set_sprg5(vcpu, regs->sprg5);
 578        kvmppc_set_sprg6(vcpu, regs->sprg6);
 579        kvmppc_set_sprg7(vcpu, regs->sprg7);
 580
 581        for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
 582                kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
 583
 584        return 0;
 585}
 586
 587int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
 588{
 589        return -EOPNOTSUPP;
 590}
 591
 592int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
 593{
 594        return -EOPNOTSUPP;
 595}
 596
 597int kvmppc_get_one_reg(struct kvm_vcpu *vcpu, u64 id,
 598                        union kvmppc_one_reg *val)
 599{
 600        int r = 0;
 601        long int i;
 602
 603        r = vcpu->kvm->arch.kvm_ops->get_one_reg(vcpu, id, val);
 604        if (r == -EINVAL) {
 605                r = 0;
 606                switch (id) {
 607                case KVM_REG_PPC_DAR:
 608                        *val = get_reg_val(id, kvmppc_get_dar(vcpu));
 609                        break;
 610                case KVM_REG_PPC_DSISR:
 611                        *val = get_reg_val(id, kvmppc_get_dsisr(vcpu));
 612                        break;
 613                case KVM_REG_PPC_FPR0 ... KVM_REG_PPC_FPR31:
 614                        i = id - KVM_REG_PPC_FPR0;
 615                        *val = get_reg_val(id, VCPU_FPR(vcpu, i));
 616                        break;
 617                case KVM_REG_PPC_FPSCR:
 618                        *val = get_reg_val(id, vcpu->arch.fp.fpscr);
 619                        break;
 620#ifdef CONFIG_VSX
 621                case KVM_REG_PPC_VSR0 ... KVM_REG_PPC_VSR31:
 622                        if (cpu_has_feature(CPU_FTR_VSX)) {
 623                                i = id - KVM_REG_PPC_VSR0;
 624                                val->vsxval[0] = vcpu->arch.fp.fpr[i][0];
 625                                val->vsxval[1] = vcpu->arch.fp.fpr[i][1];
 626                        } else {
 627                                r = -ENXIO;
 628                        }
 629                        break;
 630#endif /* CONFIG_VSX */
 631                case KVM_REG_PPC_DEBUG_INST:
 632                        *val = get_reg_val(id, INS_TW);
 633                        break;
 634#ifdef CONFIG_KVM_XICS
 635                case KVM_REG_PPC_ICP_STATE:
 636                        if (!vcpu->arch.icp && !vcpu->arch.xive_vcpu) {
 637                                r = -ENXIO;
 638                                break;
 639                        }
 640                        if (xics_on_xive())
 641                                *val = get_reg_val(id, kvmppc_xive_get_icp(vcpu));
 642                        else
 643                                *val = get_reg_val(id, kvmppc_xics_get_icp(vcpu));
 644                        break;
 645#endif /* CONFIG_KVM_XICS */
 646#ifdef CONFIG_KVM_XIVE
 647                case KVM_REG_PPC_VP_STATE:
 648                        if (!vcpu->arch.xive_vcpu) {
 649                                r = -ENXIO;
 650                                break;
 651                        }
 652                        if (xive_enabled())
 653                                r = kvmppc_xive_native_get_vp(vcpu, val);
 654                        else
 655                                r = -ENXIO;
 656                        break;
 657#endif /* CONFIG_KVM_XIVE */
 658                case KVM_REG_PPC_FSCR:
 659                        *val = get_reg_val(id, vcpu->arch.fscr);
 660                        break;
 661                case KVM_REG_PPC_TAR:
 662                        *val = get_reg_val(id, vcpu->arch.tar);
 663                        break;
 664                case KVM_REG_PPC_EBBHR:
 665                        *val = get_reg_val(id, vcpu->arch.ebbhr);
 666                        break;
 667                case KVM_REG_PPC_EBBRR:
 668                        *val = get_reg_val(id, vcpu->arch.ebbrr);
 669                        break;
 670                case KVM_REG_PPC_BESCR:
 671                        *val = get_reg_val(id, vcpu->arch.bescr);
 672                        break;
 673                case KVM_REG_PPC_IC:
 674                        *val = get_reg_val(id, vcpu->arch.ic);
 675                        break;
 676                default:
 677                        r = -EINVAL;
 678                        break;
 679                }
 680        }
 681
 682        return r;
 683}
 684
 685int kvmppc_set_one_reg(struct kvm_vcpu *vcpu, u64 id,
 686                        union kvmppc_one_reg *val)
 687{
 688        int r = 0;
 689        long int i;
 690
 691        r = vcpu->kvm->arch.kvm_ops->set_one_reg(vcpu, id, val);
 692        if (r == -EINVAL) {
 693                r = 0;
 694                switch (id) {
 695                case KVM_REG_PPC_DAR:
 696                        kvmppc_set_dar(vcpu, set_reg_val(id, *val));
 697                        break;
 698                case KVM_REG_PPC_DSISR:
 699                        kvmppc_set_dsisr(vcpu, set_reg_val(id, *val));
 700                        break;
 701                case KVM_REG_PPC_FPR0 ... KVM_REG_PPC_FPR31:
 702                        i = id - KVM_REG_PPC_FPR0;
 703                        VCPU_FPR(vcpu, i) = set_reg_val(id, *val);
 704                        break;
 705                case KVM_REG_PPC_FPSCR:
 706                        vcpu->arch.fp.fpscr = set_reg_val(id, *val);
 707                        break;
 708#ifdef CONFIG_VSX
 709                case KVM_REG_PPC_VSR0 ... KVM_REG_PPC_VSR31:
 710                        if (cpu_has_feature(CPU_FTR_VSX)) {
 711                                i = id - KVM_REG_PPC_VSR0;
 712                                vcpu->arch.fp.fpr[i][0] = val->vsxval[0];
 713                                vcpu->arch.fp.fpr[i][1] = val->vsxval[1];
 714                        } else {
 715                                r = -ENXIO;
 716                        }
 717                        break;
 718#endif /* CONFIG_VSX */
 719#ifdef CONFIG_KVM_XICS
 720                case KVM_REG_PPC_ICP_STATE:
 721                        if (!vcpu->arch.icp && !vcpu->arch.xive_vcpu) {
 722                                r = -ENXIO;
 723                                break;
 724                        }
 725                        if (xics_on_xive())
 726                                r = kvmppc_xive_set_icp(vcpu, set_reg_val(id, *val));
 727                        else
 728                                r = kvmppc_xics_set_icp(vcpu, set_reg_val(id, *val));
 729                        break;
 730#endif /* CONFIG_KVM_XICS */
 731#ifdef CONFIG_KVM_XIVE
 732                case KVM_REG_PPC_VP_STATE:
 733                        if (!vcpu->arch.xive_vcpu) {
 734                                r = -ENXIO;
 735                                break;
 736                        }
 737                        if (xive_enabled())
 738                                r = kvmppc_xive_native_set_vp(vcpu, val);
 739                        else
 740                                r = -ENXIO;
 741                        break;
 742#endif /* CONFIG_KVM_XIVE */
 743                case KVM_REG_PPC_FSCR:
 744                        vcpu->arch.fscr = set_reg_val(id, *val);
 745                        break;
 746                case KVM_REG_PPC_TAR:
 747                        vcpu->arch.tar = set_reg_val(id, *val);
 748                        break;
 749                case KVM_REG_PPC_EBBHR:
 750                        vcpu->arch.ebbhr = set_reg_val(id, *val);
 751                        break;
 752                case KVM_REG_PPC_EBBRR:
 753                        vcpu->arch.ebbrr = set_reg_val(id, *val);
 754                        break;
 755                case KVM_REG_PPC_BESCR:
 756                        vcpu->arch.bescr = set_reg_val(id, *val);
 757                        break;
 758                case KVM_REG_PPC_IC:
 759                        vcpu->arch.ic = set_reg_val(id, *val);
 760                        break;
 761                default:
 762                        r = -EINVAL;
 763                        break;
 764                }
 765        }
 766
 767        return r;
 768}
 769
 770void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
 771{
 772        vcpu->kvm->arch.kvm_ops->vcpu_load(vcpu, cpu);
 773}
 774
 775void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
 776{
 777        vcpu->kvm->arch.kvm_ops->vcpu_put(vcpu);
 778}
 779
 780void kvmppc_set_msr(struct kvm_vcpu *vcpu, u64 msr)
 781{
 782        vcpu->kvm->arch.kvm_ops->set_msr(vcpu, msr);
 783}
 784EXPORT_SYMBOL_GPL(kvmppc_set_msr);
 785
 786int kvmppc_vcpu_run(struct kvm_vcpu *vcpu)
 787{
 788        return vcpu->kvm->arch.kvm_ops->vcpu_run(vcpu);
 789}
 790
 791int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
 792                                  struct kvm_translation *tr)
 793{
 794        return 0;
 795}
 796
 797int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
 798                                        struct kvm_guest_debug *dbg)
 799{
 800        vcpu_load(vcpu);
 801        vcpu->guest_debug = dbg->control;
 802        vcpu_put(vcpu);
 803        return 0;
 804}
 805
 806void kvmppc_decrementer_func(struct kvm_vcpu *vcpu)
 807{
 808        kvmppc_core_queue_dec(vcpu);
 809        kvm_vcpu_kick(vcpu);
 810}
 811
 812int kvmppc_core_vcpu_create(struct kvm_vcpu *vcpu)
 813{
 814        return vcpu->kvm->arch.kvm_ops->vcpu_create(vcpu);
 815}
 816
 817void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
 818{
 819        vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
 820}
 821
 822int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
 823{
 824        return vcpu->kvm->arch.kvm_ops->check_requests(vcpu);
 825}
 826
 827void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot)
 828{
 829
 830}
 831
 832int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
 833{
 834        return kvm->arch.kvm_ops->get_dirty_log(kvm, log);
 835}
 836
 837void kvmppc_core_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot)
 838{
 839        kvm->arch.kvm_ops->free_memslot(slot);
 840}
 841
 842void kvmppc_core_flush_memslot(struct kvm *kvm, struct kvm_memory_slot *memslot)
 843{
 844        kvm->arch.kvm_ops->flush_memslot(kvm, memslot);
 845}
 846
 847int kvmppc_core_prepare_memory_region(struct kvm *kvm,
 848                                struct kvm_memory_slot *memslot,
 849                                const struct kvm_userspace_memory_region *mem,
 850                                enum kvm_mr_change change)
 851{
 852        return kvm->arch.kvm_ops->prepare_memory_region(kvm, memslot, mem,
 853                                                        change);
 854}
 855
 856void kvmppc_core_commit_memory_region(struct kvm *kvm,
 857                                const struct kvm_userspace_memory_region *mem,
 858                                const struct kvm_memory_slot *old,
 859                                const struct kvm_memory_slot *new,
 860                                enum kvm_mr_change change)
 861{
 862        kvm->arch.kvm_ops->commit_memory_region(kvm, mem, old, new, change);
 863}
 864
 865int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end,
 866                        unsigned flags)
 867{
 868        return kvm->arch.kvm_ops->unmap_hva_range(kvm, start, end);
 869}
 870
 871int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
 872{
 873        return kvm->arch.kvm_ops->age_hva(kvm, start, end);
 874}
 875
 876int kvm_test_age_hva(struct kvm *kvm, unsigned long hva)
 877{
 878        return kvm->arch.kvm_ops->test_age_hva(kvm, hva);
 879}
 880
 881int kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte)
 882{
 883        kvm->arch.kvm_ops->set_spte_hva(kvm, hva, pte);
 884        return 0;
 885}
 886
 887int kvmppc_core_init_vm(struct kvm *kvm)
 888{
 889
 890#ifdef CONFIG_PPC64
 891        INIT_LIST_HEAD_RCU(&kvm->arch.spapr_tce_tables);
 892        INIT_LIST_HEAD(&kvm->arch.rtas_tokens);
 893        mutex_init(&kvm->arch.rtas_token_lock);
 894#endif
 895
 896        return kvm->arch.kvm_ops->init_vm(kvm);
 897}
 898
 899void kvmppc_core_destroy_vm(struct kvm *kvm)
 900{
 901        kvm->arch.kvm_ops->destroy_vm(kvm);
 902
 903#ifdef CONFIG_PPC64
 904        kvmppc_rtas_tokens_free(kvm);
 905        WARN_ON(!list_empty(&kvm->arch.spapr_tce_tables));
 906#endif
 907
 908#ifdef CONFIG_KVM_XICS
 909        /*
 910         * Free the XIVE and XICS devices which are not directly freed by the
 911         * device 'release' method
 912         */
 913        kfree(kvm->arch.xive_devices.native);
 914        kvm->arch.xive_devices.native = NULL;
 915        kfree(kvm->arch.xive_devices.xics_on_xive);
 916        kvm->arch.xive_devices.xics_on_xive = NULL;
 917        kfree(kvm->arch.xics_device);
 918        kvm->arch.xics_device = NULL;
 919#endif /* CONFIG_KVM_XICS */
 920}
 921
 922int kvmppc_h_logical_ci_load(struct kvm_vcpu *vcpu)
 923{
 924        unsigned long size = kvmppc_get_gpr(vcpu, 4);
 925        unsigned long addr = kvmppc_get_gpr(vcpu, 5);
 926        u64 buf;
 927        int srcu_idx;
 928        int ret;
 929
 930        if (!is_power_of_2(size) || (size > sizeof(buf)))
 931                return H_TOO_HARD;
 932
 933        srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
 934        ret = kvm_io_bus_read(vcpu, KVM_MMIO_BUS, addr, size, &buf);
 935        srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
 936        if (ret != 0)
 937                return H_TOO_HARD;
 938
 939        switch (size) {
 940        case 1:
 941                kvmppc_set_gpr(vcpu, 4, *(u8 *)&buf);
 942                break;
 943
 944        case 2:
 945                kvmppc_set_gpr(vcpu, 4, be16_to_cpu(*(__be16 *)&buf));
 946                break;
 947
 948        case 4:
 949                kvmppc_set_gpr(vcpu, 4, be32_to_cpu(*(__be32 *)&buf));
 950                break;
 951
 952        case 8:
 953                kvmppc_set_gpr(vcpu, 4, be64_to_cpu(*(__be64 *)&buf));
 954                break;
 955
 956        default:
 957                BUG();
 958        }
 959
 960        return H_SUCCESS;
 961}
 962EXPORT_SYMBOL_GPL(kvmppc_h_logical_ci_load);
 963
 964int kvmppc_h_logical_ci_store(struct kvm_vcpu *vcpu)
 965{
 966        unsigned long size = kvmppc_get_gpr(vcpu, 4);
 967        unsigned long addr = kvmppc_get_gpr(vcpu, 5);
 968        unsigned long val = kvmppc_get_gpr(vcpu, 6);
 969        u64 buf;
 970        int srcu_idx;
 971        int ret;
 972
 973        switch (size) {
 974        case 1:
 975                *(u8 *)&buf = val;
 976                break;
 977
 978        case 2:
 979                *(__be16 *)&buf = cpu_to_be16(val);
 980                break;
 981
 982        case 4:
 983                *(__be32 *)&buf = cpu_to_be32(val);
 984                break;
 985
 986        case 8:
 987                *(__be64 *)&buf = cpu_to_be64(val);
 988                break;
 989
 990        default:
 991                return H_TOO_HARD;
 992        }
 993
 994        srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
 995        ret = kvm_io_bus_write(vcpu, KVM_MMIO_BUS, addr, size, &buf);
 996        srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
 997        if (ret != 0)
 998                return H_TOO_HARD;
 999
1000        return H_SUCCESS;
1001}
1002EXPORT_SYMBOL_GPL(kvmppc_h_logical_ci_store);
1003
1004int kvmppc_core_check_processor_compat(void)
1005{
1006        /*
1007         * We always return 0 for book3s. We check
1008         * for compatibility while loading the HV
1009         * or PR module
1010         */
1011        return 0;
1012}
1013
1014int kvmppc_book3s_hcall_implemented(struct kvm *kvm, unsigned long hcall)
1015{
1016        return kvm->arch.kvm_ops->hcall_implemented(hcall);
1017}
1018
1019#ifdef CONFIG_KVM_XICS
1020int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
1021                bool line_status)
1022{
1023        if (xics_on_xive())
1024                return kvmppc_xive_set_irq(kvm, irq_source_id, irq, level,
1025                                           line_status);
1026        else
1027                return kvmppc_xics_set_irq(kvm, irq_source_id, irq, level,
1028                                           line_status);
1029}
1030
1031int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *irq_entry,
1032                              struct kvm *kvm, int irq_source_id,
1033                              int level, bool line_status)
1034{
1035        return kvm_set_irq(kvm, irq_source_id, irq_entry->gsi,
1036                           level, line_status);
1037}
1038static int kvmppc_book3s_set_irq(struct kvm_kernel_irq_routing_entry *e,
1039                                 struct kvm *kvm, int irq_source_id, int level,
1040                                 bool line_status)
1041{
1042        return kvm_set_irq(kvm, irq_source_id, e->gsi, level, line_status);
1043}
1044
1045int kvm_irq_map_gsi(struct kvm *kvm,
1046                    struct kvm_kernel_irq_routing_entry *entries, int gsi)
1047{
1048        entries->gsi = gsi;
1049        entries->type = KVM_IRQ_ROUTING_IRQCHIP;
1050        entries->set = kvmppc_book3s_set_irq;
1051        entries->irqchip.irqchip = 0;
1052        entries->irqchip.pin = gsi;
1053        return 1;
1054}
1055
1056int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin)
1057{
1058        return pin;
1059}
1060
1061#endif /* CONFIG_KVM_XICS */
1062
1063static int kvmppc_book3s_init(void)
1064{
1065        int r;
1066
1067        r = kvm_init(NULL, sizeof(struct kvm_vcpu), 0, THIS_MODULE);
1068        if (r)
1069                return r;
1070#ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1071        r = kvmppc_book3s_init_pr();
1072#endif
1073
1074#ifdef CONFIG_KVM_XICS
1075#ifdef CONFIG_KVM_XIVE
1076        if (xics_on_xive()) {
1077                kvmppc_xive_init_module();
1078                kvm_register_device_ops(&kvm_xive_ops, KVM_DEV_TYPE_XICS);
1079                if (kvmppc_xive_native_supported()) {
1080                        kvmppc_xive_native_init_module();
1081                        kvm_register_device_ops(&kvm_xive_native_ops,
1082                                                KVM_DEV_TYPE_XIVE);
1083                }
1084        } else
1085#endif
1086                kvm_register_device_ops(&kvm_xics_ops, KVM_DEV_TYPE_XICS);
1087#endif
1088        return r;
1089}
1090
1091static void kvmppc_book3s_exit(void)
1092{
1093#ifdef CONFIG_KVM_XICS
1094        if (xics_on_xive()) {
1095                kvmppc_xive_exit_module();
1096                kvmppc_xive_native_exit_module();
1097        }
1098#endif
1099#ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1100        kvmppc_book3s_exit_pr();
1101#endif
1102        kvm_exit();
1103}
1104
1105module_init(kvmppc_book3s_init);
1106module_exit(kvmppc_book3s_exit);
1107
1108/* On 32bit this is our one and only kernel module */
1109#ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1110MODULE_ALIAS_MISCDEV(KVM_MINOR);
1111MODULE_ALIAS("devname:kvm");
1112#endif
1113