linux/arch/powerpc/kvm/powerpc.c
<<
>>
Prefs
   1/*
   2 * This program is free software; you can redistribute it and/or modify
   3 * it under the terms of the GNU General Public License, version 2, as
   4 * published by the Free Software Foundation.
   5 *
   6 * This program is distributed in the hope that it will be useful,
   7 * but WITHOUT ANY WARRANTY; without even the implied warranty of
   8 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
   9 * GNU General Public License for more details.
  10 *
  11 * You should have received a copy of the GNU General Public License
  12 * along with this program; if not, write to the Free Software
  13 * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
  14 *
  15 * Copyright IBM Corp. 2007
  16 *
  17 * Authors: Hollis Blanchard <hollisb@us.ibm.com>
  18 *          Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
  19 */
  20
  21#include <linux/errno.h>
  22#include <linux/err.h>
  23#include <linux/kvm_host.h>
  24#include <linux/vmalloc.h>
  25#include <linux/hrtimer.h>
  26#include <linux/fs.h>
  27#include <linux/slab.h>
  28#include <linux/file.h>
  29#include <linux/module.h>
  30#include <linux/irqbypass.h>
  31#include <linux/kvm_irqfd.h>
  32#include <asm/cputable.h>
  33#include <asm/uaccess.h>
  34#include <asm/kvm_ppc.h>
  35#include <asm/tlbflush.h>
  36#include <asm/cputhreads.h>
  37#include <asm/irqflags.h>
  38#include <asm/iommu.h>
  39#include "timing.h"
  40#include "irq.h"
  41#include "../mm/mmu_decl.h"
  42
  43#define CREATE_TRACE_POINTS
  44#include "trace.h"
  45
  46struct kvmppc_ops *kvmppc_hv_ops;
  47EXPORT_SYMBOL_GPL(kvmppc_hv_ops);
  48struct kvmppc_ops *kvmppc_pr_ops;
  49EXPORT_SYMBOL_GPL(kvmppc_pr_ops);
  50
  51
  52int kvm_arch_vcpu_runnable(struct kvm_vcpu *v)
  53{
  54        return !!(v->arch.pending_exceptions) ||
  55               v->requests;
  56}
  57
  58int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
  59{
  60        return 1;
  61}
  62
  63/*
  64 * Common checks before entering the guest world.  Call with interrupts
  65 * disabled.
  66 *
  67 * returns:
  68 *
  69 * == 1 if we're ready to go into guest state
  70 * <= 0 if we need to go back to the host with return value
  71 */
  72int kvmppc_prepare_to_enter(struct kvm_vcpu *vcpu)
  73{
  74        int r;
  75
  76        WARN_ON(irqs_disabled());
  77        hard_irq_disable();
  78
  79        while (true) {
  80                if (need_resched()) {
  81                        local_irq_enable();
  82                        cond_resched();
  83                        hard_irq_disable();
  84                        continue;
  85                }
  86
  87                if (signal_pending(current)) {
  88                        kvmppc_account_exit(vcpu, SIGNAL_EXITS);
  89                        vcpu->run->exit_reason = KVM_EXIT_INTR;
  90                        r = -EINTR;
  91                        break;
  92                }
  93
  94                vcpu->mode = IN_GUEST_MODE;
  95
  96                /*
  97                 * Reading vcpu->requests must happen after setting vcpu->mode,
  98                 * so we don't miss a request because the requester sees
  99                 * OUTSIDE_GUEST_MODE and assumes we'll be checking requests
 100                 * before next entering the guest (and thus doesn't IPI).
 101                 * This also orders the write to mode from any reads
 102                 * to the page tables done while the VCPU is running.
 103                 * Please see the comment in kvm_flush_remote_tlbs.
 104                 */
 105                smp_mb();
 106
 107                if (vcpu->requests) {
 108                        /* Make sure we process requests preemptable */
 109                        local_irq_enable();
 110                        trace_kvm_check_requests(vcpu);
 111                        r = kvmppc_core_check_requests(vcpu);
 112                        hard_irq_disable();
 113                        if (r > 0)
 114                                continue;
 115                        break;
 116                }
 117
 118                if (kvmppc_core_prepare_to_enter(vcpu)) {
 119                        /* interrupts got enabled in between, so we
 120                           are back at square 1 */
 121                        continue;
 122                }
 123
 124                guest_enter_irqoff();
 125                return 1;
 126        }
 127
 128        /* return to host */
 129        local_irq_enable();
 130        return r;
 131}
 132EXPORT_SYMBOL_GPL(kvmppc_prepare_to_enter);
 133
 134#if defined(CONFIG_PPC_BOOK3S_64) && defined(CONFIG_KVM_BOOK3S_PR_POSSIBLE)
 135static void kvmppc_swab_shared(struct kvm_vcpu *vcpu)
 136{
 137        struct kvm_vcpu_arch_shared *shared = vcpu->arch.shared;
 138        int i;
 139
 140        shared->sprg0 = swab64(shared->sprg0);
 141        shared->sprg1 = swab64(shared->sprg1);
 142        shared->sprg2 = swab64(shared->sprg2);
 143        shared->sprg3 = swab64(shared->sprg3);
 144        shared->srr0 = swab64(shared->srr0);
 145        shared->srr1 = swab64(shared->srr1);
 146        shared->dar = swab64(shared->dar);
 147        shared->msr = swab64(shared->msr);
 148        shared->dsisr = swab32(shared->dsisr);
 149        shared->int_pending = swab32(shared->int_pending);
 150        for (i = 0; i < ARRAY_SIZE(shared->sr); i++)
 151                shared->sr[i] = swab32(shared->sr[i]);
 152}
 153#endif
 154
 155int kvmppc_kvm_pv(struct kvm_vcpu *vcpu)
 156{
 157        int nr = kvmppc_get_gpr(vcpu, 11);
 158        int r;
 159        unsigned long __maybe_unused param1 = kvmppc_get_gpr(vcpu, 3);
 160        unsigned long __maybe_unused param2 = kvmppc_get_gpr(vcpu, 4);
 161        unsigned long __maybe_unused param3 = kvmppc_get_gpr(vcpu, 5);
 162        unsigned long __maybe_unused param4 = kvmppc_get_gpr(vcpu, 6);
 163        unsigned long r2 = 0;
 164
 165        if (!(kvmppc_get_msr(vcpu) & MSR_SF)) {
 166                /* 32 bit mode */
 167                param1 &= 0xffffffff;
 168                param2 &= 0xffffffff;
 169                param3 &= 0xffffffff;
 170                param4 &= 0xffffffff;
 171        }
 172
 173        switch (nr) {
 174        case KVM_HCALL_TOKEN(KVM_HC_PPC_MAP_MAGIC_PAGE):
 175        {
 176#if defined(CONFIG_PPC_BOOK3S_64) && defined(CONFIG_KVM_BOOK3S_PR_POSSIBLE)
 177                /* Book3S can be little endian, find it out here */
 178                int shared_big_endian = true;
 179                if (vcpu->arch.intr_msr & MSR_LE)
 180                        shared_big_endian = false;
 181                if (shared_big_endian != vcpu->arch.shared_big_endian)
 182                        kvmppc_swab_shared(vcpu);
 183                vcpu->arch.shared_big_endian = shared_big_endian;
 184#endif
 185
 186                if (!(param2 & MAGIC_PAGE_FLAG_NOT_MAPPED_NX)) {
 187                        /*
 188                         * Older versions of the Linux magic page code had
 189                         * a bug where they would map their trampoline code
 190                         * NX. If that's the case, remove !PR NX capability.
 191                         */
 192                        vcpu->arch.disable_kernel_nx = true;
 193                        kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
 194                }
 195
 196                vcpu->arch.magic_page_pa = param1 & ~0xfffULL;
 197                vcpu->arch.magic_page_ea = param2 & ~0xfffULL;
 198
 199#ifdef CONFIG_PPC_64K_PAGES
 200                /*
 201                 * Make sure our 4k magic page is in the same window of a 64k
 202                 * page within the guest and within the host's page.
 203                 */
 204                if ((vcpu->arch.magic_page_pa & 0xf000) !=
 205                    ((ulong)vcpu->arch.shared & 0xf000)) {
 206                        void *old_shared = vcpu->arch.shared;
 207                        ulong shared = (ulong)vcpu->arch.shared;
 208                        void *new_shared;
 209
 210                        shared &= PAGE_MASK;
 211                        shared |= vcpu->arch.magic_page_pa & 0xf000;
 212                        new_shared = (void*)shared;
 213                        memcpy(new_shared, old_shared, 0x1000);
 214                        vcpu->arch.shared = new_shared;
 215                }
 216#endif
 217
 218                r2 = KVM_MAGIC_FEAT_SR | KVM_MAGIC_FEAT_MAS0_TO_SPRG7;
 219
 220                r = EV_SUCCESS;
 221                break;
 222        }
 223        case KVM_HCALL_TOKEN(KVM_HC_FEATURES):
 224                r = EV_SUCCESS;
 225#if defined(CONFIG_PPC_BOOK3S) || defined(CONFIG_KVM_E500V2)
 226                r2 |= (1 << KVM_FEATURE_MAGIC_PAGE);
 227#endif
 228
 229                /* Second return value is in r4 */
 230                break;
 231        case EV_HCALL_TOKEN(EV_IDLE):
 232                r = EV_SUCCESS;
 233                kvm_vcpu_block(vcpu);
 234                clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
 235                break;
 236        default:
 237                r = EV_UNIMPLEMENTED;
 238                break;
 239        }
 240
 241        kvmppc_set_gpr(vcpu, 4, r2);
 242
 243        return r;
 244}
 245EXPORT_SYMBOL_GPL(kvmppc_kvm_pv);
 246
 247int kvmppc_sanity_check(struct kvm_vcpu *vcpu)
 248{
 249        int r = false;
 250
 251        /* We have to know what CPU to virtualize */
 252        if (!vcpu->arch.pvr)
 253                goto out;
 254
 255        /* PAPR only works with book3s_64 */
 256        if ((vcpu->arch.cpu_type != KVM_CPU_3S_64) && vcpu->arch.papr_enabled)
 257                goto out;
 258
 259        /* HV KVM can only do PAPR mode for now */
 260        if (!vcpu->arch.papr_enabled && is_kvmppc_hv_enabled(vcpu->kvm))
 261                goto out;
 262
 263#ifdef CONFIG_KVM_BOOKE_HV
 264        if (!cpu_has_feature(CPU_FTR_EMB_HV))
 265                goto out;
 266#endif
 267
 268        r = true;
 269
 270out:
 271        vcpu->arch.sane = r;
 272        return r ? 0 : -EINVAL;
 273}
 274EXPORT_SYMBOL_GPL(kvmppc_sanity_check);
 275
 276int kvmppc_emulate_mmio(struct kvm_run *run, struct kvm_vcpu *vcpu)
 277{
 278        enum emulation_result er;
 279        int r;
 280
 281        er = kvmppc_emulate_loadstore(vcpu);
 282        switch (er) {
 283        case EMULATE_DONE:
 284                /* Future optimization: only reload non-volatiles if they were
 285                 * actually modified. */
 286                r = RESUME_GUEST_NV;
 287                break;
 288        case EMULATE_AGAIN:
 289                r = RESUME_GUEST;
 290                break;
 291        case EMULATE_DO_MMIO:
 292                run->exit_reason = KVM_EXIT_MMIO;
 293                /* We must reload nonvolatiles because "update" load/store
 294                 * instructions modify register state. */
 295                /* Future optimization: only reload non-volatiles if they were
 296                 * actually modified. */
 297                r = RESUME_HOST_NV;
 298                break;
 299        case EMULATE_FAIL:
 300        {
 301                u32 last_inst;
 302
 303                kvmppc_get_last_inst(vcpu, INST_GENERIC, &last_inst);
 304                /* XXX Deliver Program interrupt to guest. */
 305                pr_emerg("%s: emulation failed (%08x)\n", __func__, last_inst);
 306                r = RESUME_HOST;
 307                break;
 308        }
 309        default:
 310                WARN_ON(1);
 311                r = RESUME_GUEST;
 312        }
 313
 314        return r;
 315}
 316EXPORT_SYMBOL_GPL(kvmppc_emulate_mmio);
 317
 318int kvmppc_st(struct kvm_vcpu *vcpu, ulong *eaddr, int size, void *ptr,
 319              bool data)
 320{
 321        ulong mp_pa = vcpu->arch.magic_page_pa & KVM_PAM & PAGE_MASK;
 322        struct kvmppc_pte pte;
 323        int r;
 324
 325        vcpu->stat.st++;
 326
 327        r = kvmppc_xlate(vcpu, *eaddr, data ? XLATE_DATA : XLATE_INST,
 328                         XLATE_WRITE, &pte);
 329        if (r < 0)
 330                return r;
 331
 332        *eaddr = pte.raddr;
 333
 334        if (!pte.may_write)
 335                return -EPERM;
 336
 337        /* Magic page override */
 338        if (kvmppc_supports_magic_page(vcpu) && mp_pa &&
 339            ((pte.raddr & KVM_PAM & PAGE_MASK) == mp_pa) &&
 340            !(kvmppc_get_msr(vcpu) & MSR_PR)) {
 341                void *magic = vcpu->arch.shared;
 342                magic += pte.eaddr & 0xfff;
 343                memcpy(magic, ptr, size);
 344                return EMULATE_DONE;
 345        }
 346
 347        if (kvm_write_guest(vcpu->kvm, pte.raddr, ptr, size))
 348                return EMULATE_DO_MMIO;
 349
 350        return EMULATE_DONE;
 351}
 352EXPORT_SYMBOL_GPL(kvmppc_st);
 353
 354int kvmppc_ld(struct kvm_vcpu *vcpu, ulong *eaddr, int size, void *ptr,
 355                      bool data)
 356{
 357        ulong mp_pa = vcpu->arch.magic_page_pa & KVM_PAM & PAGE_MASK;
 358        struct kvmppc_pte pte;
 359        int rc;
 360
 361        vcpu->stat.ld++;
 362
 363        rc = kvmppc_xlate(vcpu, *eaddr, data ? XLATE_DATA : XLATE_INST,
 364                          XLATE_READ, &pte);
 365        if (rc)
 366                return rc;
 367
 368        *eaddr = pte.raddr;
 369
 370        if (!pte.may_read)
 371                return -EPERM;
 372
 373        if (!data && !pte.may_execute)
 374                return -ENOEXEC;
 375
 376        /* Magic page override */
 377        if (kvmppc_supports_magic_page(vcpu) && mp_pa &&
 378            ((pte.raddr & KVM_PAM & PAGE_MASK) == mp_pa) &&
 379            !(kvmppc_get_msr(vcpu) & MSR_PR)) {
 380                void *magic = vcpu->arch.shared;
 381                magic += pte.eaddr & 0xfff;
 382                memcpy(ptr, magic, size);
 383                return EMULATE_DONE;
 384        }
 385
 386        if (kvm_read_guest(vcpu->kvm, pte.raddr, ptr, size))
 387                return EMULATE_DO_MMIO;
 388
 389        return EMULATE_DONE;
 390}
 391EXPORT_SYMBOL_GPL(kvmppc_ld);
 392
 393int kvm_arch_hardware_enable(void)
 394{
 395        return 0;
 396}
 397
 398int kvm_arch_hardware_setup(void)
 399{
 400        return 0;
 401}
 402
 403void kvm_arch_check_processor_compat(void *rtn)
 404{
 405        *(int *)rtn = kvmppc_core_check_processor_compat();
 406}
 407
 408int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
 409{
 410        struct kvmppc_ops *kvm_ops = NULL;
 411        /*
 412         * if we have both HV and PR enabled, default is HV
 413         */
 414        if (type == 0) {
 415                if (kvmppc_hv_ops)
 416                        kvm_ops = kvmppc_hv_ops;
 417                else
 418                        kvm_ops = kvmppc_pr_ops;
 419                if (!kvm_ops)
 420                        goto err_out;
 421        } else  if (type == KVM_VM_PPC_HV) {
 422                if (!kvmppc_hv_ops)
 423                        goto err_out;
 424                kvm_ops = kvmppc_hv_ops;
 425        } else if (type == KVM_VM_PPC_PR) {
 426                if (!kvmppc_pr_ops)
 427                        goto err_out;
 428                kvm_ops = kvmppc_pr_ops;
 429        } else
 430                goto err_out;
 431
 432        if (kvm_ops->owner && !try_module_get(kvm_ops->owner))
 433                return -ENOENT;
 434
 435        kvm->arch.kvm_ops = kvm_ops;
 436        return kvmppc_core_init_vm(kvm);
 437err_out:
 438        return -EINVAL;
 439}
 440
 441bool kvm_arch_has_vcpu_debugfs(void)
 442{
 443        return false;
 444}
 445
 446int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu)
 447{
 448        return 0;
 449}
 450
 451void kvm_arch_destroy_vm(struct kvm *kvm)
 452{
 453        unsigned int i;
 454        struct kvm_vcpu *vcpu;
 455
 456#ifdef CONFIG_KVM_XICS
 457        /*
 458         * We call kick_all_cpus_sync() to ensure that all
 459         * CPUs have executed any pending IPIs before we
 460         * continue and free VCPUs structures below.
 461         */
 462        if (is_kvmppc_hv_enabled(kvm))
 463                kick_all_cpus_sync();
 464#endif
 465
 466        kvm_for_each_vcpu(i, vcpu, kvm)
 467                kvm_arch_vcpu_free(vcpu);
 468
 469        mutex_lock(&kvm->lock);
 470        for (i = 0; i < atomic_read(&kvm->online_vcpus); i++)
 471                kvm->vcpus[i] = NULL;
 472
 473        atomic_set(&kvm->online_vcpus, 0);
 474
 475        kvmppc_core_destroy_vm(kvm);
 476
 477        mutex_unlock(&kvm->lock);
 478
 479        /* drop the module reference */
 480        module_put(kvm->arch.kvm_ops->owner);
 481}
 482
 483int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
 484{
 485        int r;
 486        /* Assume we're using HV mode when the HV module is loaded */
 487        int hv_enabled = kvmppc_hv_ops ? 1 : 0;
 488
 489        if (kvm) {
 490                /*
 491                 * Hooray - we know which VM type we're running on. Depend on
 492                 * that rather than the guess above.
 493                 */
 494                hv_enabled = is_kvmppc_hv_enabled(kvm);
 495        }
 496
 497        switch (ext) {
 498#ifdef CONFIG_BOOKE
 499        case KVM_CAP_PPC_BOOKE_SREGS:
 500        case KVM_CAP_PPC_BOOKE_WATCHDOG:
 501        case KVM_CAP_PPC_EPR:
 502#else
 503        case KVM_CAP_PPC_SEGSTATE:
 504        case KVM_CAP_PPC_HIOR:
 505        case KVM_CAP_PPC_PAPR:
 506#endif
 507        case KVM_CAP_PPC_UNSET_IRQ:
 508        case KVM_CAP_PPC_IRQ_LEVEL:
 509        case KVM_CAP_ENABLE_CAP:
 510        case KVM_CAP_ENABLE_CAP_VM:
 511        case KVM_CAP_ONE_REG:
 512        case KVM_CAP_IOEVENTFD:
 513        case KVM_CAP_DEVICE_CTRL:
 514                r = 1;
 515                break;
 516        case KVM_CAP_PPC_PAIRED_SINGLES:
 517        case KVM_CAP_PPC_OSI:
 518        case KVM_CAP_PPC_GET_PVINFO:
 519#if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
 520        case KVM_CAP_SW_TLB:
 521#endif
 522                /* We support this only for PR */
 523                r = !hv_enabled;
 524                break;
 525#ifdef CONFIG_KVM_MMIO
 526        case KVM_CAP_COALESCED_MMIO:
 527                r = KVM_COALESCED_MMIO_PAGE_OFFSET;
 528                break;
 529#endif
 530#ifdef CONFIG_KVM_MPIC
 531        case KVM_CAP_IRQ_MPIC:
 532                r = 1;
 533                break;
 534#endif
 535
 536#ifdef CONFIG_PPC_BOOK3S_64
 537        case KVM_CAP_SPAPR_TCE:
 538        case KVM_CAP_SPAPR_TCE_64:
 539        case KVM_CAP_PPC_ALLOC_HTAB:
 540        case KVM_CAP_PPC_RTAS:
 541        case KVM_CAP_PPC_FIXUP_HCALL:
 542        case KVM_CAP_PPC_ENABLE_HCALL:
 543#ifdef CONFIG_KVM_XICS
 544        case KVM_CAP_IRQ_XICS:
 545#endif
 546                r = 1;
 547                break;
 548#endif /* CONFIG_PPC_BOOK3S_64 */
 549#ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
 550        case KVM_CAP_PPC_SMT:
 551                if (hv_enabled)
 552                        r = threads_per_subcore;
 553                else
 554                        r = 0;
 555                break;
 556        case KVM_CAP_PPC_RMA:
 557                r = 0;
 558                break;
 559        case KVM_CAP_PPC_HWRNG:
 560                r = kvmppc_hwrng_present();
 561                break;
 562#endif
 563        case KVM_CAP_SYNC_MMU:
 564#ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
 565                r = hv_enabled;
 566#elif defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 567                r = 1;
 568#else
 569                r = 0;
 570#endif
 571                break;
 572#ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
 573        case KVM_CAP_PPC_HTAB_FD:
 574                r = hv_enabled;
 575                break;
 576#endif
 577        case KVM_CAP_NR_VCPUS:
 578                /*
 579                 * Recommending a number of CPUs is somewhat arbitrary; we
 580                 * return the number of present CPUs for -HV (since a host
 581                 * will have secondary threads "offline"), and for other KVM
 582                 * implementations just count online CPUs.
 583                 */
 584                if (hv_enabled)
 585                        r = num_present_cpus();
 586                else
 587                        r = num_online_cpus();
 588                break;
 589        case KVM_CAP_NR_MEMSLOTS:
 590                r = KVM_USER_MEM_SLOTS;
 591                break;
 592        case KVM_CAP_MAX_VCPUS:
 593                r = KVM_MAX_VCPUS;
 594                break;
 595#ifdef CONFIG_PPC_BOOK3S_64
 596        case KVM_CAP_PPC_GET_SMMU_INFO:
 597                r = 1;
 598                break;
 599        case KVM_CAP_SPAPR_MULTITCE:
 600                r = 1;
 601                break;
 602#endif
 603        case KVM_CAP_PPC_HTM:
 604                r = cpu_has_feature(CPU_FTR_TM_COMP) &&
 605                    is_kvmppc_hv_enabled(kvm);
 606                break;
 607        default:
 608                r = 0;
 609                break;
 610        }
 611        return r;
 612
 613}
 614
 615long kvm_arch_dev_ioctl(struct file *filp,
 616                        unsigned int ioctl, unsigned long arg)
 617{
 618        return -EINVAL;
 619}
 620
 621void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
 622                           struct kvm_memory_slot *dont)
 623{
 624        kvmppc_core_free_memslot(kvm, free, dont);
 625}
 626
 627int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
 628                            unsigned long npages)
 629{
 630        return kvmppc_core_create_memslot(kvm, slot, npages);
 631}
 632
 633int kvm_arch_prepare_memory_region(struct kvm *kvm,
 634                                   struct kvm_memory_slot *memslot,
 635                                   const struct kvm_userspace_memory_region *mem,
 636                                   enum kvm_mr_change change)
 637{
 638        return kvmppc_core_prepare_memory_region(kvm, memslot, mem);
 639}
 640
 641void kvm_arch_commit_memory_region(struct kvm *kvm,
 642                                   const struct kvm_userspace_memory_region *mem,
 643                                   const struct kvm_memory_slot *old,
 644                                   const struct kvm_memory_slot *new,
 645                                   enum kvm_mr_change change)
 646{
 647        kvmppc_core_commit_memory_region(kvm, mem, old, new);
 648}
 649
 650void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
 651                                   struct kvm_memory_slot *slot)
 652{
 653        kvmppc_core_flush_memslot(kvm, slot);
 654}
 655
 656struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id)
 657{
 658        struct kvm_vcpu *vcpu;
 659        vcpu = kvmppc_core_vcpu_create(kvm, id);
 660        if (!IS_ERR(vcpu)) {
 661                vcpu->arch.wqp = &vcpu->wq;
 662                kvmppc_create_vcpu_debugfs(vcpu, id);
 663        }
 664        return vcpu;
 665}
 666
 667void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu)
 668{
 669}
 670
 671void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu)
 672{
 673        /* Make sure we're not using the vcpu anymore */
 674        hrtimer_cancel(&vcpu->arch.dec_timer);
 675
 676        kvmppc_remove_vcpu_debugfs(vcpu);
 677
 678        switch (vcpu->arch.irq_type) {
 679        case KVMPPC_IRQ_MPIC:
 680                kvmppc_mpic_disconnect_vcpu(vcpu->arch.mpic, vcpu);
 681                break;
 682        case KVMPPC_IRQ_XICS:
 683                kvmppc_xics_free_icp(vcpu);
 684                break;
 685        }
 686
 687        kvmppc_core_vcpu_free(vcpu);
 688}
 689
 690void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
 691{
 692        kvm_arch_vcpu_free(vcpu);
 693}
 694
 695int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
 696{
 697        return kvmppc_core_pending_dec(vcpu);
 698}
 699
 700static enum hrtimer_restart kvmppc_decrementer_wakeup(struct hrtimer *timer)
 701{
 702        struct kvm_vcpu *vcpu;
 703
 704        vcpu = container_of(timer, struct kvm_vcpu, arch.dec_timer);
 705        kvmppc_decrementer_func(vcpu);
 706
 707        return HRTIMER_NORESTART;
 708}
 709
 710int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
 711{
 712        int ret;
 713
 714        hrtimer_init(&vcpu->arch.dec_timer, CLOCK_REALTIME, HRTIMER_MODE_ABS);
 715        vcpu->arch.dec_timer.function = kvmppc_decrementer_wakeup;
 716        vcpu->arch.dec_expires = ~(u64)0;
 717
 718#ifdef CONFIG_KVM_EXIT_TIMING
 719        mutex_init(&vcpu->arch.exit_timing_lock);
 720#endif
 721        ret = kvmppc_subarch_vcpu_init(vcpu);
 722        return ret;
 723}
 724
 725void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu)
 726{
 727        kvmppc_mmu_destroy(vcpu);
 728        kvmppc_subarch_vcpu_uninit(vcpu);
 729}
 730
 731void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
 732{
 733#ifdef CONFIG_BOOKE
 734        /*
 735         * vrsave (formerly usprg0) isn't used by Linux, but may
 736         * be used by the guest.
 737         *
 738         * On non-booke this is associated with Altivec and
 739         * is handled by code in book3s.c.
 740         */
 741        mtspr(SPRN_VRSAVE, vcpu->arch.vrsave);
 742#endif
 743        kvmppc_core_vcpu_load(vcpu, cpu);
 744}
 745
 746void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
 747{
 748        kvmppc_core_vcpu_put(vcpu);
 749#ifdef CONFIG_BOOKE
 750        vcpu->arch.vrsave = mfspr(SPRN_VRSAVE);
 751#endif
 752}
 753
 754/*
 755 * irq_bypass_add_producer and irq_bypass_del_producer are only
 756 * useful if the architecture supports PCI passthrough.
 757 * irq_bypass_stop and irq_bypass_start are not needed and so
 758 * kvm_ops are not defined for them.
 759 */
 760bool kvm_arch_has_irq_bypass(void)
 761{
 762        return ((kvmppc_hv_ops && kvmppc_hv_ops->irq_bypass_add_producer) ||
 763                (kvmppc_pr_ops && kvmppc_pr_ops->irq_bypass_add_producer));
 764}
 765
 766int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *cons,
 767                                     struct irq_bypass_producer *prod)
 768{
 769        struct kvm_kernel_irqfd *irqfd =
 770                container_of(cons, struct kvm_kernel_irqfd, consumer);
 771        struct kvm *kvm = irqfd->kvm;
 772
 773        if (kvm->arch.kvm_ops->irq_bypass_add_producer)
 774                return kvm->arch.kvm_ops->irq_bypass_add_producer(cons, prod);
 775
 776        return 0;
 777}
 778
 779void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *cons,
 780                                      struct irq_bypass_producer *prod)
 781{
 782        struct kvm_kernel_irqfd *irqfd =
 783                container_of(cons, struct kvm_kernel_irqfd, consumer);
 784        struct kvm *kvm = irqfd->kvm;
 785
 786        if (kvm->arch.kvm_ops->irq_bypass_del_producer)
 787                kvm->arch.kvm_ops->irq_bypass_del_producer(cons, prod);
 788}
 789
 790static void kvmppc_complete_mmio_load(struct kvm_vcpu *vcpu,
 791                                      struct kvm_run *run)
 792{
 793        u64 uninitialized_var(gpr);
 794
 795        if (run->mmio.len > sizeof(gpr)) {
 796                printk(KERN_ERR "bad MMIO length: %d\n", run->mmio.len);
 797                return;
 798        }
 799
 800        if (!vcpu->arch.mmio_host_swabbed) {
 801                switch (run->mmio.len) {
 802                case 8: gpr = *(u64 *)run->mmio.data; break;
 803                case 4: gpr = *(u32 *)run->mmio.data; break;
 804                case 2: gpr = *(u16 *)run->mmio.data; break;
 805                case 1: gpr = *(u8 *)run->mmio.data; break;
 806                }
 807        } else {
 808                switch (run->mmio.len) {
 809                case 8: gpr = swab64(*(u64 *)run->mmio.data); break;
 810                case 4: gpr = swab32(*(u32 *)run->mmio.data); break;
 811                case 2: gpr = swab16(*(u16 *)run->mmio.data); break;
 812                case 1: gpr = *(u8 *)run->mmio.data; break;
 813                }
 814        }
 815
 816        if (vcpu->arch.mmio_sign_extend) {
 817                switch (run->mmio.len) {
 818#ifdef CONFIG_PPC64
 819                case 4:
 820                        gpr = (s64)(s32)gpr;
 821                        break;
 822#endif
 823                case 2:
 824                        gpr = (s64)(s16)gpr;
 825                        break;
 826                case 1:
 827                        gpr = (s64)(s8)gpr;
 828                        break;
 829                }
 830        }
 831
 832        kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
 833
 834        switch (vcpu->arch.io_gpr & KVM_MMIO_REG_EXT_MASK) {
 835        case KVM_MMIO_REG_GPR:
 836                kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
 837                break;
 838        case KVM_MMIO_REG_FPR:
 839                VCPU_FPR(vcpu, vcpu->arch.io_gpr & KVM_MMIO_REG_MASK) = gpr;
 840                break;
 841#ifdef CONFIG_PPC_BOOK3S
 842        case KVM_MMIO_REG_QPR:
 843                vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
 844                break;
 845        case KVM_MMIO_REG_FQPR:
 846                VCPU_FPR(vcpu, vcpu->arch.io_gpr & KVM_MMIO_REG_MASK) = gpr;
 847                vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
 848                break;
 849#endif
 850        default:
 851                BUG();
 852        }
 853}
 854
 855static int __kvmppc_handle_load(struct kvm_run *run, struct kvm_vcpu *vcpu,
 856                                unsigned int rt, unsigned int bytes,
 857                                int is_default_endian, int sign_extend)
 858{
 859        int idx, ret;
 860        bool host_swabbed;
 861
 862        /* Pity C doesn't have a logical XOR operator */
 863        if (kvmppc_need_byteswap(vcpu)) {
 864                host_swabbed = is_default_endian;
 865        } else {
 866                host_swabbed = !is_default_endian;
 867        }
 868
 869        if (bytes > sizeof(run->mmio.data)) {
 870                printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
 871                       run->mmio.len);
 872        }
 873
 874        run->mmio.phys_addr = vcpu->arch.paddr_accessed;
 875        run->mmio.len = bytes;
 876        run->mmio.is_write = 0;
 877
 878        vcpu->arch.io_gpr = rt;
 879        vcpu->arch.mmio_host_swabbed = host_swabbed;
 880        vcpu->mmio_needed = 1;
 881        vcpu->mmio_is_write = 0;
 882        vcpu->arch.mmio_sign_extend = sign_extend;
 883
 884        idx = srcu_read_lock(&vcpu->kvm->srcu);
 885
 886        ret = kvm_io_bus_read(vcpu, KVM_MMIO_BUS, run->mmio.phys_addr,
 887                              bytes, &run->mmio.data);
 888
 889        srcu_read_unlock(&vcpu->kvm->srcu, idx);
 890
 891        if (!ret) {
 892                kvmppc_complete_mmio_load(vcpu, run);
 893                vcpu->mmio_needed = 0;
 894                return EMULATE_DONE;
 895        }
 896
 897        return EMULATE_DO_MMIO;
 898}
 899
 900int kvmppc_handle_load(struct kvm_run *run, struct kvm_vcpu *vcpu,
 901                       unsigned int rt, unsigned int bytes,
 902                       int is_default_endian)
 903{
 904        return __kvmppc_handle_load(run, vcpu, rt, bytes, is_default_endian, 0);
 905}
 906EXPORT_SYMBOL_GPL(kvmppc_handle_load);
 907
 908/* Same as above, but sign extends */
 909int kvmppc_handle_loads(struct kvm_run *run, struct kvm_vcpu *vcpu,
 910                        unsigned int rt, unsigned int bytes,
 911                        int is_default_endian)
 912{
 913        return __kvmppc_handle_load(run, vcpu, rt, bytes, is_default_endian, 1);
 914}
 915
 916int kvmppc_handle_store(struct kvm_run *run, struct kvm_vcpu *vcpu,
 917                        u64 val, unsigned int bytes, int is_default_endian)
 918{
 919        void *data = run->mmio.data;
 920        int idx, ret;
 921        bool host_swabbed;
 922
 923        /* Pity C doesn't have a logical XOR operator */
 924        if (kvmppc_need_byteswap(vcpu)) {
 925                host_swabbed = is_default_endian;
 926        } else {
 927                host_swabbed = !is_default_endian;
 928        }
 929
 930        if (bytes > sizeof(run->mmio.data)) {
 931                printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
 932                       run->mmio.len);
 933        }
 934
 935        run->mmio.phys_addr = vcpu->arch.paddr_accessed;
 936        run->mmio.len = bytes;
 937        run->mmio.is_write = 1;
 938        vcpu->mmio_needed = 1;
 939        vcpu->mmio_is_write = 1;
 940
 941        /* Store the value at the lowest bytes in 'data'. */
 942        if (!host_swabbed) {
 943                switch (bytes) {
 944                case 8: *(u64 *)data = val; break;
 945                case 4: *(u32 *)data = val; break;
 946                case 2: *(u16 *)data = val; break;
 947                case 1: *(u8  *)data = val; break;
 948                }
 949        } else {
 950                switch (bytes) {
 951                case 8: *(u64 *)data = swab64(val); break;
 952                case 4: *(u32 *)data = swab32(val); break;
 953                case 2: *(u16 *)data = swab16(val); break;
 954                case 1: *(u8  *)data = val; break;
 955                }
 956        }
 957
 958        idx = srcu_read_lock(&vcpu->kvm->srcu);
 959
 960        ret = kvm_io_bus_write(vcpu, KVM_MMIO_BUS, run->mmio.phys_addr,
 961                               bytes, &run->mmio.data);
 962
 963        srcu_read_unlock(&vcpu->kvm->srcu, idx);
 964
 965        if (!ret) {
 966                vcpu->mmio_needed = 0;
 967                return EMULATE_DONE;
 968        }
 969
 970        return EMULATE_DO_MMIO;
 971}
 972EXPORT_SYMBOL_GPL(kvmppc_handle_store);
 973
 974int kvm_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
 975{
 976        int r = 0;
 977        union kvmppc_one_reg val;
 978        int size;
 979
 980        size = one_reg_size(reg->id);
 981        if (size > sizeof(val))
 982                return -EINVAL;
 983
 984        r = kvmppc_get_one_reg(vcpu, reg->id, &val);
 985        if (r == -EINVAL) {
 986                r = 0;
 987                switch (reg->id) {
 988#ifdef CONFIG_ALTIVEC
 989                case KVM_REG_PPC_VR0 ... KVM_REG_PPC_VR31:
 990                        if (!cpu_has_feature(CPU_FTR_ALTIVEC)) {
 991                                r = -ENXIO;
 992                                break;
 993                        }
 994                        val.vval = vcpu->arch.vr.vr[reg->id - KVM_REG_PPC_VR0];
 995                        break;
 996                case KVM_REG_PPC_VSCR:
 997                        if (!cpu_has_feature(CPU_FTR_ALTIVEC)) {
 998                                r = -ENXIO;
 999                                break;
1000                        }
1001                        val = get_reg_val(reg->id, vcpu->arch.vr.vscr.u[3]);
1002                        break;
1003                case KVM_REG_PPC_VRSAVE:
1004                        val = get_reg_val(reg->id, vcpu->arch.vrsave);
1005                        break;
1006#endif /* CONFIG_ALTIVEC */
1007                default:
1008                        r = -EINVAL;
1009                        break;
1010                }
1011        }
1012
1013        if (r)
1014                return r;
1015
1016        if (copy_to_user((char __user *)(unsigned long)reg->addr, &val, size))
1017                r = -EFAULT;
1018
1019        return r;
1020}
1021
1022int kvm_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1023{
1024        int r;
1025        union kvmppc_one_reg val;
1026        int size;
1027
1028        size = one_reg_size(reg->id);
1029        if (size > sizeof(val))
1030                return -EINVAL;
1031
1032        if (copy_from_user(&val, (char __user *)(unsigned long)reg->addr, size))
1033                return -EFAULT;
1034
1035        r = kvmppc_set_one_reg(vcpu, reg->id, &val);
1036        if (r == -EINVAL) {
1037                r = 0;
1038                switch (reg->id) {
1039#ifdef CONFIG_ALTIVEC
1040                case KVM_REG_PPC_VR0 ... KVM_REG_PPC_VR31:
1041                        if (!cpu_has_feature(CPU_FTR_ALTIVEC)) {
1042                                r = -ENXIO;
1043                                break;
1044                        }
1045                        vcpu->arch.vr.vr[reg->id - KVM_REG_PPC_VR0] = val.vval;
1046                        break;
1047                case KVM_REG_PPC_VSCR:
1048                        if (!cpu_has_feature(CPU_FTR_ALTIVEC)) {
1049                                r = -ENXIO;
1050                                break;
1051                        }
1052                        vcpu->arch.vr.vscr.u[3] = set_reg_val(reg->id, val);
1053                        break;
1054                case KVM_REG_PPC_VRSAVE:
1055                        if (!cpu_has_feature(CPU_FTR_ALTIVEC)) {
1056                                r = -ENXIO;
1057                                break;
1058                        }
1059                        vcpu->arch.vrsave = set_reg_val(reg->id, val);
1060                        break;
1061#endif /* CONFIG_ALTIVEC */
1062                default:
1063                        r = -EINVAL;
1064                        break;
1065                }
1066        }
1067
1068        return r;
1069}
1070
1071int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)
1072{
1073        int r;
1074        sigset_t sigsaved;
1075
1076        if (vcpu->sigset_active)
1077                sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved);
1078
1079        if (vcpu->mmio_needed) {
1080                if (!vcpu->mmio_is_write)
1081                        kvmppc_complete_mmio_load(vcpu, run);
1082                vcpu->mmio_needed = 0;
1083        } else if (vcpu->arch.osi_needed) {
1084                u64 *gprs = run->osi.gprs;
1085                int i;
1086
1087                for (i = 0; i < 32; i++)
1088                        kvmppc_set_gpr(vcpu, i, gprs[i]);
1089                vcpu->arch.osi_needed = 0;
1090        } else if (vcpu->arch.hcall_needed) {
1091                int i;
1092
1093                kvmppc_set_gpr(vcpu, 3, run->papr_hcall.ret);
1094                for (i = 0; i < 9; ++i)
1095                        kvmppc_set_gpr(vcpu, 4 + i, run->papr_hcall.args[i]);
1096                vcpu->arch.hcall_needed = 0;
1097#ifdef CONFIG_BOOKE
1098        } else if (vcpu->arch.epr_needed) {
1099                kvmppc_set_epr(vcpu, run->epr.epr);
1100                vcpu->arch.epr_needed = 0;
1101#endif
1102        }
1103
1104        r = kvmppc_vcpu_run(run, vcpu);
1105
1106        if (vcpu->sigset_active)
1107                sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1108
1109        return r;
1110}
1111
1112int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu, struct kvm_interrupt *irq)
1113{
1114        if (irq->irq == KVM_INTERRUPT_UNSET) {
1115                kvmppc_core_dequeue_external(vcpu);
1116                return 0;
1117        }
1118
1119        kvmppc_core_queue_external(vcpu, irq);
1120
1121        kvm_vcpu_kick(vcpu);
1122
1123        return 0;
1124}
1125
1126static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
1127                                     struct kvm_enable_cap *cap)
1128{
1129        int r;
1130
1131        if (cap->flags)
1132                return -EINVAL;
1133
1134        switch (cap->cap) {
1135        case KVM_CAP_PPC_OSI:
1136                r = 0;
1137                vcpu->arch.osi_enabled = true;
1138                break;
1139        case KVM_CAP_PPC_PAPR:
1140                r = 0;
1141                vcpu->arch.papr_enabled = true;
1142                break;
1143        case KVM_CAP_PPC_EPR:
1144                r = 0;
1145                if (cap->args[0])
1146                        vcpu->arch.epr_flags |= KVMPPC_EPR_USER;
1147                else
1148                        vcpu->arch.epr_flags &= ~KVMPPC_EPR_USER;
1149                break;
1150#ifdef CONFIG_BOOKE
1151        case KVM_CAP_PPC_BOOKE_WATCHDOG:
1152                r = 0;
1153                vcpu->arch.watchdog_enabled = true;
1154                break;
1155#endif
1156#if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
1157        case KVM_CAP_SW_TLB: {
1158                struct kvm_config_tlb cfg;
1159                void __user *user_ptr = (void __user *)(uintptr_t)cap->args[0];
1160
1161                r = -EFAULT;
1162                if (copy_from_user(&cfg, user_ptr, sizeof(cfg)))
1163                        break;
1164
1165                r = kvm_vcpu_ioctl_config_tlb(vcpu, &cfg);
1166                break;
1167        }
1168#endif
1169#ifdef CONFIG_KVM_MPIC
1170        case KVM_CAP_IRQ_MPIC: {
1171                struct fd f;
1172                struct kvm_device *dev;
1173
1174                r = -EBADF;
1175                f = fdget(cap->args[0]);
1176                if (!f.file)
1177                        break;
1178
1179                r = -EPERM;
1180                dev = kvm_device_from_filp(f.file);
1181                if (dev)
1182                        r = kvmppc_mpic_connect_vcpu(dev, vcpu, cap->args[1]);
1183
1184                fdput(f);
1185                break;
1186        }
1187#endif
1188#ifdef CONFIG_KVM_XICS
1189        case KVM_CAP_IRQ_XICS: {
1190                struct fd f;
1191                struct kvm_device *dev;
1192
1193                r = -EBADF;
1194                f = fdget(cap->args[0]);
1195                if (!f.file)
1196                        break;
1197
1198                r = -EPERM;
1199                dev = kvm_device_from_filp(f.file);
1200                if (dev)
1201                        r = kvmppc_xics_connect_vcpu(dev, vcpu, cap->args[1]);
1202
1203                fdput(f);
1204                break;
1205        }
1206#endif /* CONFIG_KVM_XICS */
1207        default:
1208                r = -EINVAL;
1209                break;
1210        }
1211
1212        if (!r)
1213                r = kvmppc_sanity_check(vcpu);
1214
1215        return r;
1216}
1217
1218bool kvm_arch_intc_initialized(struct kvm *kvm)
1219{
1220#ifdef CONFIG_KVM_MPIC
1221        if (kvm->arch.mpic)
1222                return true;
1223#endif
1224#ifdef CONFIG_KVM_XICS
1225        if (kvm->arch.xics)
1226                return true;
1227#endif
1228        return false;
1229}
1230
1231int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
1232                                    struct kvm_mp_state *mp_state)
1233{
1234        return -EINVAL;
1235}
1236
1237int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
1238                                    struct kvm_mp_state *mp_state)
1239{
1240        return -EINVAL;
1241}
1242
1243long kvm_arch_vcpu_ioctl(struct file *filp,
1244                         unsigned int ioctl, unsigned long arg)
1245{
1246        struct kvm_vcpu *vcpu = filp->private_data;
1247        void __user *argp = (void __user *)arg;
1248        long r;
1249
1250        switch (ioctl) {
1251        case KVM_INTERRUPT: {
1252                struct kvm_interrupt irq;
1253                r = -EFAULT;
1254                if (copy_from_user(&irq, argp, sizeof(irq)))
1255                        goto out;
1256                r = kvm_vcpu_ioctl_interrupt(vcpu, &irq);
1257                goto out;
1258        }
1259
1260        case KVM_ENABLE_CAP:
1261        {
1262                struct kvm_enable_cap cap;
1263                r = -EFAULT;
1264                if (copy_from_user(&cap, argp, sizeof(cap)))
1265                        goto out;
1266                r = kvm_vcpu_ioctl_enable_cap(vcpu, &cap);
1267                break;
1268        }
1269
1270        case KVM_SET_ONE_REG:
1271        case KVM_GET_ONE_REG:
1272        {
1273                struct kvm_one_reg reg;
1274                r = -EFAULT;
1275                if (copy_from_user(&reg, argp, sizeof(reg)))
1276                        goto out;
1277                if (ioctl == KVM_SET_ONE_REG)
1278                        r = kvm_vcpu_ioctl_set_one_reg(vcpu, &reg);
1279                else
1280                        r = kvm_vcpu_ioctl_get_one_reg(vcpu, &reg);
1281                break;
1282        }
1283
1284#if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
1285        case KVM_DIRTY_TLB: {
1286                struct kvm_dirty_tlb dirty;
1287                r = -EFAULT;
1288                if (copy_from_user(&dirty, argp, sizeof(dirty)))
1289                        goto out;
1290                r = kvm_vcpu_ioctl_dirty_tlb(vcpu, &dirty);
1291                break;
1292        }
1293#endif
1294        default:
1295                r = -EINVAL;
1296        }
1297
1298out:
1299        return r;
1300}
1301
1302int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
1303{
1304        return VM_FAULT_SIGBUS;
1305}
1306
1307static int kvm_vm_ioctl_get_pvinfo(struct kvm_ppc_pvinfo *pvinfo)
1308{
1309        u32 inst_nop = 0x60000000;
1310#ifdef CONFIG_KVM_BOOKE_HV
1311        u32 inst_sc1 = 0x44000022;
1312        pvinfo->hcall[0] = cpu_to_be32(inst_sc1);
1313        pvinfo->hcall[1] = cpu_to_be32(inst_nop);
1314        pvinfo->hcall[2] = cpu_to_be32(inst_nop);
1315        pvinfo->hcall[3] = cpu_to_be32(inst_nop);
1316#else
1317        u32 inst_lis = 0x3c000000;
1318        u32 inst_ori = 0x60000000;
1319        u32 inst_sc = 0x44000002;
1320        u32 inst_imm_mask = 0xffff;
1321
1322        /*
1323         * The hypercall to get into KVM from within guest context is as
1324         * follows:
1325         *
1326         *    lis r0, r0, KVM_SC_MAGIC_R0@h
1327         *    ori r0, KVM_SC_MAGIC_R0@l
1328         *    sc
1329         *    nop
1330         */
1331        pvinfo->hcall[0] = cpu_to_be32(inst_lis | ((KVM_SC_MAGIC_R0 >> 16) & inst_imm_mask));
1332        pvinfo->hcall[1] = cpu_to_be32(inst_ori | (KVM_SC_MAGIC_R0 & inst_imm_mask));
1333        pvinfo->hcall[2] = cpu_to_be32(inst_sc);
1334        pvinfo->hcall[3] = cpu_to_be32(inst_nop);
1335#endif
1336
1337        pvinfo->flags = KVM_PPC_PVINFO_FLAGS_EV_IDLE;
1338
1339        return 0;
1340}
1341
1342int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_event,
1343                          bool line_status)
1344{
1345        if (!irqchip_in_kernel(kvm))
1346                return -ENXIO;
1347
1348        irq_event->status = kvm_set_irq(kvm, KVM_USERSPACE_IRQ_SOURCE_ID,
1349                                        irq_event->irq, irq_event->level,
1350                                        line_status);
1351        return 0;
1352}
1353
1354
1355static int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
1356                                   struct kvm_enable_cap *cap)
1357{
1358        int r;
1359
1360        if (cap->flags)
1361                return -EINVAL;
1362
1363        switch (cap->cap) {
1364#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1365        case KVM_CAP_PPC_ENABLE_HCALL: {
1366                unsigned long hcall = cap->args[0];
1367
1368                r = -EINVAL;
1369                if (hcall > MAX_HCALL_OPCODE || (hcall & 3) ||
1370                    cap->args[1] > 1)
1371                        break;
1372                if (!kvmppc_book3s_hcall_implemented(kvm, hcall))
1373                        break;
1374                if (cap->args[1])
1375                        set_bit(hcall / 4, kvm->arch.enabled_hcalls);
1376                else
1377                        clear_bit(hcall / 4, kvm->arch.enabled_hcalls);
1378                r = 0;
1379                break;
1380        }
1381#endif
1382        default:
1383                r = -EINVAL;
1384                break;
1385        }
1386
1387        return r;
1388}
1389
1390long kvm_arch_vm_ioctl(struct file *filp,
1391                       unsigned int ioctl, unsigned long arg)
1392{
1393        struct kvm *kvm __maybe_unused = filp->private_data;
1394        void __user *argp = (void __user *)arg;
1395        long r;
1396
1397        switch (ioctl) {
1398        case KVM_PPC_GET_PVINFO: {
1399                struct kvm_ppc_pvinfo pvinfo;
1400                memset(&pvinfo, 0, sizeof(pvinfo));
1401                r = kvm_vm_ioctl_get_pvinfo(&pvinfo);
1402                if (copy_to_user(argp, &pvinfo, sizeof(pvinfo))) {
1403                        r = -EFAULT;
1404                        goto out;
1405                }
1406
1407                break;
1408        }
1409        case KVM_ENABLE_CAP:
1410        {
1411                struct kvm_enable_cap cap;
1412                r = -EFAULT;
1413                if (copy_from_user(&cap, argp, sizeof(cap)))
1414                        goto out;
1415                r = kvm_vm_ioctl_enable_cap(kvm, &cap);
1416                break;
1417        }
1418#ifdef CONFIG_PPC_BOOK3S_64
1419        case KVM_CREATE_SPAPR_TCE_64: {
1420                struct kvm_create_spapr_tce_64 create_tce_64;
1421
1422                r = -EFAULT;
1423                if (copy_from_user(&create_tce_64, argp, sizeof(create_tce_64)))
1424                        goto out;
1425                if (create_tce_64.flags) {
1426                        r = -EINVAL;
1427                        goto out;
1428                }
1429                r = kvm_vm_ioctl_create_spapr_tce(kvm, &create_tce_64);
1430                goto out;
1431        }
1432        case KVM_CREATE_SPAPR_TCE: {
1433                struct kvm_create_spapr_tce create_tce;
1434                struct kvm_create_spapr_tce_64 create_tce_64;
1435
1436                r = -EFAULT;
1437                if (copy_from_user(&create_tce, argp, sizeof(create_tce)))
1438                        goto out;
1439
1440                create_tce_64.liobn = create_tce.liobn;
1441                create_tce_64.page_shift = IOMMU_PAGE_SHIFT_4K;
1442                create_tce_64.offset = 0;
1443                create_tce_64.size = create_tce.window_size >>
1444                                IOMMU_PAGE_SHIFT_4K;
1445                create_tce_64.flags = 0;
1446                r = kvm_vm_ioctl_create_spapr_tce(kvm, &create_tce_64);
1447                goto out;
1448        }
1449        case KVM_PPC_GET_SMMU_INFO: {
1450                struct kvm_ppc_smmu_info info;
1451                struct kvm *kvm = filp->private_data;
1452
1453                memset(&info, 0, sizeof(info));
1454                r = kvm->arch.kvm_ops->get_smmu_info(kvm, &info);
1455                if (r >= 0 && copy_to_user(argp, &info, sizeof(info)))
1456                        r = -EFAULT;
1457                break;
1458        }
1459        case KVM_PPC_RTAS_DEFINE_TOKEN: {
1460                struct kvm *kvm = filp->private_data;
1461
1462                r = kvm_vm_ioctl_rtas_define_token(kvm, argp);
1463                break;
1464        }
1465        default: {
1466                struct kvm *kvm = filp->private_data;
1467                r = kvm->arch.kvm_ops->arch_vm_ioctl(filp, ioctl, arg);
1468        }
1469#else /* CONFIG_PPC_BOOK3S_64 */
1470        default:
1471                r = -ENOTTY;
1472#endif
1473        }
1474out:
1475        return r;
1476}
1477
1478static unsigned long lpid_inuse[BITS_TO_LONGS(KVMPPC_NR_LPIDS)];
1479static unsigned long nr_lpids;
1480
1481long kvmppc_alloc_lpid(void)
1482{
1483        long lpid;
1484
1485        do {
1486                lpid = find_first_zero_bit(lpid_inuse, KVMPPC_NR_LPIDS);
1487                if (lpid >= nr_lpids) {
1488                        pr_err("%s: No LPIDs free\n", __func__);
1489                        return -ENOMEM;
1490                }
1491        } while (test_and_set_bit(lpid, lpid_inuse));
1492
1493        return lpid;
1494}
1495EXPORT_SYMBOL_GPL(kvmppc_alloc_lpid);
1496
1497void kvmppc_claim_lpid(long lpid)
1498{
1499        set_bit(lpid, lpid_inuse);
1500}
1501EXPORT_SYMBOL_GPL(kvmppc_claim_lpid);
1502
1503void kvmppc_free_lpid(long lpid)
1504{
1505        clear_bit(lpid, lpid_inuse);
1506}
1507EXPORT_SYMBOL_GPL(kvmppc_free_lpid);
1508
1509void kvmppc_init_lpid(unsigned long nr_lpids_param)
1510{
1511        nr_lpids = min_t(unsigned long, KVMPPC_NR_LPIDS, nr_lpids_param);
1512        memset(lpid_inuse, 0, sizeof(lpid_inuse));
1513}
1514EXPORT_SYMBOL_GPL(kvmppc_init_lpid);
1515
1516int kvm_arch_init(void *opaque)
1517{
1518        return 0;
1519}
1520
1521EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_ppc_instr);
1522