arch/powerpc/kvm/powerpc.c

   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 <asm/cputable.h>
  31 #include <asm/uaccess.h>
  32 #include <asm/kvm_ppc.h>
  33 #include <asm/tlbflush.h>
  34 #include <asm/cputhreads.h>
  35 #include <asm/irqflags.h>
  36 #include "timing.h"
  37 #include "irq.h"
  38 #include "../mm/mmu_decl.h"
  39
  40 #define CREATE_TRACE_POINTS
  41 #include "trace.h"
  42
  43 struct kvmppc_ops *kvmppc_hv_ops;
  44 EXPORT_SYMBOL_GPL(kvmppc_hv_ops);
  45 struct kvmppc_ops *kvmppc_pr_ops;
  46 EXPORT_SYMBOL_GPL(kvmppc_pr_ops);
  47
  48
  49 int kvm_arch_vcpu_runnable(struct kvm_vcpu *v)
  50 {
  51         return !!(v->arch.pending_exceptions) ||
  52                v->requests;
  53 }
  54
  55 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
  56 {
  57         return 1;
  58 }
  59
  60 /*
  61  * Common checks before entering the guest world.  Call with interrupts
  62  * disabled.
  63  *
  64  * returns:
  65  *
  66  * == 1 if we're ready to go into guest state
  67  * <= 0 if we need to go back to the host with return value
  68  */
  69 int kvmppc_prepare_to_enter(struct kvm_vcpu *vcpu)
  70 {
  71         int r;
  72
  73         WARN_ON(irqs_disabled());
  74         hard_irq_disable();
  75
  76         while (true) {
  77                 if (need_resched()) {
  78                         local_irq_enable();
  79                         cond_resched();
  80                         hard_irq_disable();
  81                         continue;
  82                 }
  83
  84                 if (signal_pending(current)) {
  85                         kvmppc_account_exit(vcpu, SIGNAL_EXITS);
  86                         vcpu->run->exit_reason = KVM_EXIT_INTR;
  87                         r = -EINTR;
  88                         break;
  89                 }
  90
  91                 vcpu->mode = IN_GUEST_MODE;
  92
  93                 /*
  94                  * Reading vcpu->requests must happen after setting vcpu->mode,
  95                  * so we don't miss a request because the requester sees
  96                  * OUTSIDE_GUEST_MODE and assumes we'll be checking requests
  97                  * before next entering the guest (and thus doesn't IPI).
  98                  */
  99                 smp_mb();
 100
 101                 if (vcpu->requests) {
 102                         /* Make sure we process requests preemptable */
 103                         local_irq_enable();
 104                         trace_kvm_check_requests(vcpu);
 105                         r = kvmppc_core_check_requests(vcpu);
 106                         hard_irq_disable();
 107                         if (r > 0)
 108                                 continue;
 109                         break;
 110                 }
 111
 112                 if (kvmppc_core_prepare_to_enter(vcpu)) {
 113                         /* interrupts got enabled in between, so we
 114                            are back at square 1 */
 115                         continue;
 116                 }
 117
 118                 kvm_guest_enter();
 119                 return 1;
 120         }
 121
 122         /* return to host */
 123         local_irq_enable();
 124         return r;
 125 }
 126 EXPORT_SYMBOL_GPL(kvmppc_prepare_to_enter);
 127
 128 #if defined(CONFIG_PPC_BOOK3S_64) && defined(CONFIG_KVM_BOOK3S_PR_POSSIBLE)
 129 static void kvmppc_swab_shared(struct kvm_vcpu *vcpu)
 130 {
 131         struct kvm_vcpu_arch_shared *shared = vcpu->arch.shared;
 132         int i;
 133
 134         shared->sprg0 = swab64(shared->sprg0);
 135         shared->sprg1 = swab64(shared->sprg1);
 136         shared->sprg2 = swab64(shared->sprg2);
 137         shared->sprg3 = swab64(shared->sprg3);
 138         shared->srr0 = swab64(shared->srr0);
 139         shared->srr1 = swab64(shared->srr1);
 140         shared->dar = swab64(shared->dar);
 141         shared->msr = swab64(shared->msr);
 142         shared->dsisr = swab32(shared->dsisr);
 143         shared->int_pending = swab32(shared->int_pending);
 144         for (i = 0; i < ARRAY_SIZE(shared->sr); i++)
 145                 shared->sr[i] = swab32(shared->sr[i]);
 146 }
 147 #endif
 148
 149 int kvmppc_kvm_pv(struct kvm_vcpu *vcpu)
 150 {
 151         int nr = kvmppc_get_gpr(vcpu, 11);
 152         int r;
 153         unsigned long __maybe_unused param1 = kvmppc_get_gpr(vcpu, 3);
 154         unsigned long __maybe_unused param2 = kvmppc_get_gpr(vcpu, 4);
 155         unsigned long __maybe_unused param3 = kvmppc_get_gpr(vcpu, 5);
 156         unsigned long __maybe_unused param4 = kvmppc_get_gpr(vcpu, 6);
 157         unsigned long r2 = 0;
 158
 159         if (!(kvmppc_get_msr(vcpu) & MSR_SF)) {
 160                 /* 32 bit mode */
 161                 param1 &= 0xffffffff;
 162                 param2 &= 0xffffffff;
 163                 param3 &= 0xffffffff;
 164                 param4 &= 0xffffffff;
 165         }
 166
 167         switch (nr) {
 168         case KVM_HCALL_TOKEN(KVM_HC_PPC_MAP_MAGIC_PAGE):
 169         {
 170 #if defined(CONFIG_PPC_BOOK3S_64) && defined(CONFIG_KVM_BOOK3S_PR_POSSIBLE)
 171                 /* Book3S can be little endian, find it out here */
 172                 int shared_big_endian = true;
 173                 if (vcpu->arch.intr_msr & MSR_LE)
 174                         shared_big_endian = false;
 175                 if (shared_big_endian != vcpu->arch.shared_big_endian)
 176                         kvmppc_swab_shared(vcpu);
 177                 vcpu->arch.shared_big_endian = shared_big_endian;
 178 #endif
 179
 180                 if (!(param2 & MAGIC_PAGE_FLAG_NOT_MAPPED_NX)) {
 181                         /*
 182                          * Older versions of the Linux magic page code had
 183                          * a bug where they would map their trampoline code
 184                          * NX. If that's the case, remove !PR NX capability.
 185                          */
 186                         vcpu->arch.disable_kernel_nx = true;
 187                         kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
 188                 }
 189
 190                 vcpu->arch.magic_page_pa = param1 & ~0xfffULL;
 191                 vcpu->arch.magic_page_ea = param2 & ~0xfffULL;
 192
 193 #ifdef CONFIG_PPC_64K_PAGES
 194                 /*
 195                  * Make sure our 4k magic page is in the same window of a 64k
 196                  * page within the guest and within the host's page.
 197                  */
 198                 if ((vcpu->arch.magic_page_pa & 0xf000) !=
 199                     ((ulong)vcpu->arch.shared & 0xf000)) {
 200                         void *old_shared = vcpu->arch.shared;
 201                         ulong shared = (ulong)vcpu->arch.shared;
 202                         void *new_shared;
 203
 204                         shared &= PAGE_MASK;
 205                         shared |= vcpu->arch.magic_page_pa & 0xf000;
 206                         new_shared = (void*)shared;
 207                         memcpy(new_shared, old_shared, 0x1000);
 208                         vcpu->arch.shared = new_shared;
 209                 }
 210 #endif
 211
 212                 r2 = KVM_MAGIC_FEAT_SR | KVM_MAGIC_FEAT_MAS0_TO_SPRG7;
 213
 214                 r = EV_SUCCESS;
 215                 break;
 216         }
 217         case KVM_HCALL_TOKEN(KVM_HC_FEATURES):
 218                 r = EV_SUCCESS;
 219 #if defined(CONFIG_PPC_BOOK3S) || defined(CONFIG_KVM_E500V2)
 220                 r2 |= (1 << KVM_FEATURE_MAGIC_PAGE);
 221 #endif
 222
 223                 /* Second return value is in r4 */
 224                 break;
 225         case EV_HCALL_TOKEN(EV_IDLE):
 226                 r = EV_SUCCESS;
 227                 kvm_vcpu_block(vcpu);
 228                 clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
 229                 break;
 230         default:
 231                 r = EV_UNIMPLEMENTED;
 232                 break;
 233         }
 234
 235         kvmppc_set_gpr(vcpu, 4, r2);
 236
 237         return r;
 238 }
 239 EXPORT_SYMBOL_GPL(kvmppc_kvm_pv);
 240
 241 int kvmppc_sanity_check(struct kvm_vcpu *vcpu)
 242 {
 243         int r = false;
 244
 245         /* We have to know what CPU to virtualize */
 246         if (!vcpu->arch.pvr)
 247                 goto out;
 248
 249         /* PAPR only works with book3s_64 */
 250         if ((vcpu->arch.cpu_type != KVM_CPU_3S_64) && vcpu->arch.papr_enabled)
 251                 goto out;
 252
 253         /* HV KVM can only do PAPR mode for now */
 254         if (!vcpu->arch.papr_enabled && is_kvmppc_hv_enabled(vcpu->kvm))
 255                 goto out;
 256
 257 #ifdef CONFIG_KVM_BOOKE_HV
 258         if (!cpu_has_feature(CPU_FTR_EMB_HV))
 259                 goto out;
 260 #endif
 261
 262         r = true;
 263
 264 out:
 265         vcpu->arch.sane = r;
 266         return r ? 0 : -EINVAL;
 267 }
 268 EXPORT_SYMBOL_GPL(kvmppc_sanity_check);
 269
 270 int kvmppc_emulate_mmio(struct kvm_run *run, struct kvm_vcpu *vcpu)
 271 {
 272         enum emulation_result er;
 273         int r;
 274
 275         er = kvmppc_emulate_loadstore(vcpu);
 276         switch (er) {
 277         case EMULATE_DONE:
 278                 /* Future optimization: only reload non-volatiles if they were
 279                  * actually modified. */
 280                 r = RESUME_GUEST_NV;
 281                 break;
 282         case EMULATE_AGAIN:
 283                 r = RESUME_GUEST;
 284                 break;
 285         case EMULATE_DO_MMIO:
 286                 run->exit_reason = KVM_EXIT_MMIO;
 287                 /* We must reload nonvolatiles because "update" load/store
 288                  * instructions modify register state. */
 289                 /* Future optimization: only reload non-volatiles if they were
 290                  * actually modified. */
 291                 r = RESUME_HOST_NV;
 292                 break;
 293         case EMULATE_FAIL:
 294         {
 295                 u32 last_inst;
 296
 297                 kvmppc_get_last_inst(vcpu, INST_GENERIC, &last_inst);
 298                 /* XXX Deliver Program interrupt to guest. */
 299                 pr_emerg("%s: emulation failed (%08x)\n", __func__, last_inst);
 300                 r = RESUME_HOST;
 301                 break;
 302         }
 303         default:
 304                 WARN_ON(1);
 305                 r = RESUME_GUEST;
 306         }
 307
 308         return r;
 309 }
 310 EXPORT_SYMBOL_GPL(kvmppc_emulate_mmio);
 311
 312 int kvmppc_st(struct kvm_vcpu *vcpu, ulong *eaddr, int size, void *ptr,
 313               bool data)
 314 {
 315         ulong mp_pa = vcpu->arch.magic_page_pa & KVM_PAM & PAGE_MASK;
 316         struct kvmppc_pte pte;
 317         int r;
 318
 319         vcpu->stat.st++;
 320
 321         r = kvmppc_xlate(vcpu, *eaddr, data ? XLATE_DATA : XLATE_INST,
 322                          XLATE_WRITE, &pte);
 323         if (r < 0)
 324                 return r;
 325
 326         *eaddr = pte.raddr;
 327
 328         if (!pte.may_write)
 329                 return -EPERM;
 330
 331         /* Magic page override */
 332         if (kvmppc_supports_magic_page(vcpu) && mp_pa &&
 333             ((pte.raddr & KVM_PAM & PAGE_MASK) == mp_pa) &&
 334             !(kvmppc_get_msr(vcpu) & MSR_PR)) {
 335                 void *magic = vcpu->arch.shared;
 336                 magic += pte.eaddr & 0xfff;
 337                 memcpy(magic, ptr, size);
 338                 return EMULATE_DONE;
 339         }
 340
 341         if (kvm_write_guest(vcpu->kvm, pte.raddr, ptr, size))
 342                 return EMULATE_DO_MMIO;
 343
 344         return EMULATE_DONE;
 345 }
 346 EXPORT_SYMBOL_GPL(kvmppc_st);
 347
 348 int kvmppc_ld(struct kvm_vcpu *vcpu, ulong *eaddr, int size, void *ptr,
 349                       bool data)
 350 {
 351         ulong mp_pa = vcpu->arch.magic_page_pa & KVM_PAM & PAGE_MASK;
 352         struct kvmppc_pte pte;
 353         int rc;
 354
 355         vcpu->stat.ld++;
 356
 357         rc = kvmppc_xlate(vcpu, *eaddr, data ? XLATE_DATA : XLATE_INST,
 358                           XLATE_READ, &pte);
 359         if (rc)
 360                 return rc;
 361
 362         *eaddr = pte.raddr;
 363
 364         if (!pte.may_read)
 365                 return -EPERM;
 366
 367         if (!data && !pte.may_execute)
 368                 return -ENOEXEC;
 369
 370         /* Magic page override */
 371         if (kvmppc_supports_magic_page(vcpu) && mp_pa &&
 372             ((pte.raddr & KVM_PAM & PAGE_MASK) == mp_pa) &&
 373             !(kvmppc_get_msr(vcpu) & MSR_PR)) {
 374                 void *magic = vcpu->arch.shared;
 375                 magic += pte.eaddr & 0xfff;
 376                 memcpy(ptr, magic, size);
 377                 return EMULATE_DONE;
 378         }
 379
 380         if (kvm_read_guest(vcpu->kvm, pte.raddr, ptr, size))
 381                 return EMULATE_DO_MMIO;
 382
 383         return EMULATE_DONE;
 384 }
 385 EXPORT_SYMBOL_GPL(kvmppc_ld);
 386
 387 int kvm_arch_hardware_enable(void)
 388 {
 389         return 0;
 390 }
 391
 392 int kvm_arch_hardware_setup(void)
 393 {
 394         return 0;
 395 }
 396
 397 void kvm_arch_check_processor_compat(void *rtn)
 398 {
 399         *(int *)rtn = kvmppc_core_check_processor_compat();
 400 }
 401
 402 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
 403 {
 404         struct kvmppc_ops *kvm_ops = NULL;
 405         /*
 406          * if we have both HV and PR enabled, default is HV
 407          */
 408         if (type == 0) {
 409                 if (kvmppc_hv_ops)
 410                         kvm_ops = kvmppc_hv_ops;
 411                 else
 412                         kvm_ops = kvmppc_pr_ops;
 413                 if (!kvm_ops)
 414                         goto err_out;
 415         } else  if (type == KVM_VM_PPC_HV) {
 416                 if (!kvmppc_hv_ops)
 417                         goto err_out;
 418                 kvm_ops = kvmppc_hv_ops;
 419         } else if (type == KVM_VM_PPC_PR) {
 420                 if (!kvmppc_pr_ops)
 421                         goto err_out;
 422                 kvm_ops = kvmppc_pr_ops;
 423         } else
 424                 goto err_out;
 425
 426         if (kvm_ops->owner && !try_module_get(kvm_ops->owner))
 427                 return -ENOENT;
 428
 429         kvm->arch.kvm_ops = kvm_ops;
 430         return kvmppc_core_init_vm(kvm);
 431 err_out:
 432         return -EINVAL;
 433 }
 434
 435 void kvm_arch_destroy_vm(struct kvm *kvm)
 436 {
 437         unsigned int i;
 438         struct kvm_vcpu *vcpu;
 439
 440         kvm_for_each_vcpu(i, vcpu, kvm)
 441                 kvm_arch_vcpu_free(vcpu);
 442
 443         mutex_lock(&kvm->lock);
 444         for (i = 0; i < atomic_read(&kvm->online_vcpus); i++)
 445                 kvm->vcpus[i] = NULL;
 446
 447         atomic_set(&kvm->online_vcpus, 0);
 448
 449         kvmppc_core_destroy_vm(kvm);
 450
 451         mutex_unlock(&kvm->lock);
 452
 453         /* drop the module reference */
 454         module_put(kvm->arch.kvm_ops->owner);
 455 }
 456
 457 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
 458 {
 459         int r;
 460         /* Assume we're using HV mode when the HV module is loaded */
 461         int hv_enabled = kvmppc_hv_ops ? 1 : 0;
 462
 463         if (kvm) {
 464                 /*
 465                  * Hooray - we know which VM type we're running on. Depend on
 466                  * that rather than the guess above.
 467                  */
 468                 hv_enabled = is_kvmppc_hv_enabled(kvm);
 469         }
 470
 471         switch (ext) {
 472 #ifdef CONFIG_BOOKE
 473         case KVM_CAP_PPC_BOOKE_SREGS:
 474         case KVM_CAP_PPC_BOOKE_WATCHDOG:
 475         case KVM_CAP_PPC_EPR:
 476 #else
 477         case KVM_CAP_PPC_SEGSTATE:
 478         case KVM_CAP_PPC_HIOR:
 479         case KVM_CAP_PPC_PAPR:
 480 #endif
 481         case KVM_CAP_PPC_UNSET_IRQ:
 482         case KVM_CAP_PPC_IRQ_LEVEL:
 483         case KVM_CAP_ENABLE_CAP:
 484         case KVM_CAP_ENABLE_CAP_VM:
 485         case KVM_CAP_ONE_REG:
 486         case KVM_CAP_IOEVENTFD:
 487         case KVM_CAP_DEVICE_CTRL:
 488                 r = 1;
 489                 break;
 490         case KVM_CAP_PPC_PAIRED_SINGLES:
 491         case KVM_CAP_PPC_OSI:
 492         case KVM_CAP_PPC_GET_PVINFO:
 493 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
 494         case KVM_CAP_SW_TLB:
 495 #endif
 496                 /* We support this only for PR */
 497                 r = !hv_enabled;
 498                 break;
 499 #ifdef CONFIG_KVM_MMIO
 500         case KVM_CAP_COALESCED_MMIO:
 501                 r = KVM_COALESCED_MMIO_PAGE_OFFSET;
 502                 break;
 503 #endif
 504 #ifdef CONFIG_KVM_MPIC
 505         case KVM_CAP_IRQ_MPIC:
 506                 r = 1;
 507                 break;
 508 #endif
 509
 510 #ifdef CONFIG_PPC_BOOK3S_64
 511         case KVM_CAP_SPAPR_TCE:
 512         case KVM_CAP_PPC_ALLOC_HTAB:
 513         case KVM_CAP_PPC_RTAS:
 514         case KVM_CAP_PPC_FIXUP_HCALL:
 515         case KVM_CAP_PPC_ENABLE_HCALL:
 516 #ifdef CONFIG_KVM_XICS
 517         case KVM_CAP_IRQ_XICS:
 518 #endif
 519                 r = 1;
 520                 break;
 521 #endif /* CONFIG_PPC_BOOK3S_64 */
 522 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
 523         case KVM_CAP_PPC_SMT:
 524                 if (hv_enabled)
 525                         r = threads_per_subcore;
 526                 else
 527                         r = 0;
 528                 break;
 529         case KVM_CAP_PPC_RMA:
 530                 r = hv_enabled;
 531                 /* PPC970 requires an RMA */
 532                 if (r && cpu_has_feature(CPU_FTR_ARCH_201))
 533                         r = 2;
 534                 break;
 535 #endif
 536         case KVM_CAP_SYNC_MMU:
 537 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
 538                 if (hv_enabled)
 539                         r = cpu_has_feature(CPU_FTR_ARCH_206) ? 1 : 0;
 540                 else
 541                         r = 0;
 542 #elif defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 543                 r = 1;
 544 #else
 545                 r = 0;
 546 #endif
 547                 break;
 548 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
 549         case KVM_CAP_PPC_HTAB_FD:
 550                 r = hv_enabled;
 551                 break;
 552 #endif
 553         case KVM_CAP_NR_VCPUS:
 554                 /*
 555                  * Recommending a number of CPUs is somewhat arbitrary; we
 556                  * return the number of present CPUs for -HV (since a host
 557                  * will have secondary threads "offline"), and for other KVM
 558                  * implementations just count online CPUs.
 559                  */
 560                 if (hv_enabled)
 561                         r = num_present_cpus();
 562                 else
 563                         r = num_online_cpus();
 564                 break;
 565         case KVM_CAP_MAX_VCPUS:
 566                 r = KVM_MAX_VCPUS;
 567                 break;
 568 #ifdef CONFIG_PPC_BOOK3S_64
 569         case KVM_CAP_PPC_GET_SMMU_INFO:
 570                 r = 1;
 571                 break;
 572 #endif
 573         default:
 574                 r = 0;
 575                 break;
 576         }
 577         return r;
 578
 579 }
 580
 581 long kvm_arch_dev_ioctl(struct file *filp,
 582                         unsigned int ioctl, unsigned long arg)
 583 {
 584         return -EINVAL;
 585 }
 586
 587 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
 588                            struct kvm_memory_slot *dont)
 589 {
 590         kvmppc_core_free_memslot(kvm, free, dont);
 591 }
 592
 593 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
 594                             unsigned long npages)
 595 {
 596         return kvmppc_core_create_memslot(kvm, slot, npages);
 597 }
 598
 599 int kvm_arch_prepare_memory_region(struct kvm *kvm,
 600                                    struct kvm_memory_slot *memslot,
 601                                    struct kvm_userspace_memory_region *mem,
 602                                    enum kvm_mr_change change)
 603 {
 604         return kvmppc_core_prepare_memory_region(kvm, memslot, mem);
 605 }
 606
 607 void kvm_arch_commit_memory_region(struct kvm *kvm,
 608                                    struct kvm_userspace_memory_region *mem,
 609                                    const struct kvm_memory_slot *old,
 610                                    enum kvm_mr_change change)
 611 {
 612         kvmppc_core_commit_memory_region(kvm, mem, old);
 613 }
 614
 615 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
 616                                    struct kvm_memory_slot *slot)
 617 {
 618         kvmppc_core_flush_memslot(kvm, slot);
 619 }
 620
 621 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id)
 622 {
 623         struct kvm_vcpu *vcpu;
 624         vcpu = kvmppc_core_vcpu_create(kvm, id);
 625         if (!IS_ERR(vcpu)) {
 626                 vcpu->arch.wqp = &vcpu->wq;
 627                 kvmppc_create_vcpu_debugfs(vcpu, id);
 628         }
 629         return vcpu;
 630 }
 631
 632 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu)
 633 {
 634         return 0;
 635 }
 636
 637 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu)
 638 {
 639         /* Make sure we're not using the vcpu anymore */
 640         hrtimer_cancel(&vcpu->arch.dec_timer);
 641
 642         kvmppc_remove_vcpu_debugfs(vcpu);
 643
 644         switch (vcpu->arch.irq_type) {
 645         case KVMPPC_IRQ_MPIC:
 646                 kvmppc_mpic_disconnect_vcpu(vcpu->arch.mpic, vcpu);
 647                 break;
 648         case KVMPPC_IRQ_XICS:
 649                 kvmppc_xics_free_icp(vcpu);
 650                 break;
 651         }
 652
 653         kvmppc_core_vcpu_free(vcpu);
 654 }
 655
 656 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
 657 {
 658         kvm_arch_vcpu_free(vcpu);
 659 }
 660
 661 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
 662 {
 663         return kvmppc_core_pending_dec(vcpu);
 664 }
 665
 666 enum hrtimer_restart kvmppc_decrementer_wakeup(struct hrtimer *timer)
 667 {
 668         struct kvm_vcpu *vcpu;
 669
 670         vcpu = container_of(timer, struct kvm_vcpu, arch.dec_timer);
 671         kvmppc_decrementer_func(vcpu);
 672
 673         return HRTIMER_NORESTART;
 674 }
 675
 676 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
 677 {
 678         int ret;
 679
 680         hrtimer_init(&vcpu->arch.dec_timer, CLOCK_REALTIME, HRTIMER_MODE_ABS);
 681         vcpu->arch.dec_timer.function = kvmppc_decrementer_wakeup;
 682         vcpu->arch.dec_expires = ~(u64)0;
 683
 684 #ifdef CONFIG_KVM_EXIT_TIMING
 685         mutex_init(&vcpu->arch.exit_timing_lock);
 686 #endif
 687         ret = kvmppc_subarch_vcpu_init(vcpu);
 688         return ret;
 689 }
 690
 691 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu)
 692 {
 693         kvmppc_mmu_destroy(vcpu);
 694         kvmppc_subarch_vcpu_uninit(vcpu);
 695 }
 696
 697 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
 698 {
 699 #ifdef CONFIG_BOOKE
 700         /*
 701          * vrsave (formerly usprg0) isn't used by Linux, but may
 702          * be used by the guest.
 703          *
 704          * On non-booke this is associated with Altivec and
 705          * is handled by code in book3s.c.
 706          */
 707         mtspr(SPRN_VRSAVE, vcpu->arch.vrsave);
 708 #endif
 709         kvmppc_core_vcpu_load(vcpu, cpu);
 710 }
 711
 712 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
 713 {
 714         kvmppc_core_vcpu_put(vcpu);
 715 #ifdef CONFIG_BOOKE
 716         vcpu->arch.vrsave = mfspr(SPRN_VRSAVE);
 717 #endif
 718 }
 719
 720 static void kvmppc_complete_mmio_load(struct kvm_vcpu *vcpu,
 721                                       struct kvm_run *run)
 722 {
 723         u64 uninitialized_var(gpr);
 724
 725         if (run->mmio.len > sizeof(gpr)) {
 726                 printk(KERN_ERR "bad MMIO length: %d\n", run->mmio.len);
 727                 return;
 728         }
 729
 730         if (vcpu->arch.mmio_is_bigendian) {
 731                 switch (run->mmio.len) {
 732                 case 8: gpr = *(u64 *)run->mmio.data; break;
 733                 case 4: gpr = *(u32 *)run->mmio.data; break;
 734                 case 2: gpr = *(u16 *)run->mmio.data; break;
 735                 case 1: gpr = *(u8 *)run->mmio.data; break;
 736                 }
 737         } else {
 738                 /* Convert BE data from userland back to LE. */
 739                 switch (run->mmio.len) {
 740                 case 4: gpr = ld_le32((u32 *)run->mmio.data); break;
 741                 case 2: gpr = ld_le16((u16 *)run->mmio.data); break;
 742                 case 1: gpr = *(u8 *)run->mmio.data; break;
 743                 }
 744         }
 745
 746         if (vcpu->arch.mmio_sign_extend) {
 747                 switch (run->mmio.len) {
 748 #ifdef CONFIG_PPC64
 749                 case 4:
 750                         gpr = (s64)(s32)gpr;
 751                         break;
 752 #endif
 753                 case 2:
 754                         gpr = (s64)(s16)gpr;
 755                         break;
 756                 case 1:
 757                         gpr = (s64)(s8)gpr;
 758                         break;
 759                 }
 760         }
 761
 762         kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
 763
 764         switch (vcpu->arch.io_gpr & KVM_MMIO_REG_EXT_MASK) {
 765         case KVM_MMIO_REG_GPR:
 766                 kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
 767                 break;
 768         case KVM_MMIO_REG_FPR:
 769                 VCPU_FPR(vcpu, vcpu->arch.io_gpr & KVM_MMIO_REG_MASK) = gpr;
 770                 break;
 771 #ifdef CONFIG_PPC_BOOK3S
 772         case KVM_MMIO_REG_QPR:
 773                 vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
 774                 break;
 775         case KVM_MMIO_REG_FQPR:
 776                 VCPU_FPR(vcpu, vcpu->arch.io_gpr & KVM_MMIO_REG_MASK) = gpr;
 777                 vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
 778                 break;
 779 #endif
 780         default:
 781                 BUG();
 782         }
 783 }
 784
 785 int kvmppc_handle_load(struct kvm_run *run, struct kvm_vcpu *vcpu,
 786                        unsigned int rt, unsigned int bytes,
 787                        int is_default_endian)
 788 {
 789         int idx, ret;
 790         int is_bigendian;
 791
 792         if (kvmppc_need_byteswap(vcpu)) {
 793                 /* Default endianness is "little endian". */
 794                 is_bigendian = !is_default_endian;
 795         } else {
 796                 /* Default endianness is "big endian". */
 797                 is_bigendian = is_default_endian;
 798         }
 799
 800         if (bytes > sizeof(run->mmio.data)) {
 801                 printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
 802                        run->mmio.len);
 803         }
 804
 805         run->mmio.phys_addr = vcpu->arch.paddr_accessed;
 806         run->mmio.len = bytes;
 807         run->mmio.is_write = 0;
 808
 809         vcpu->arch.io_gpr = rt;
 810         vcpu->arch.mmio_is_bigendian = is_bigendian;
 811         vcpu->mmio_needed = 1;
 812         vcpu->mmio_is_write = 0;
 813         vcpu->arch.mmio_sign_extend = 0;
 814
 815         idx = srcu_read_lock(&vcpu->kvm->srcu);
 816
 817         ret = kvm_io_bus_read(vcpu->kvm, KVM_MMIO_BUS, run->mmio.phys_addr,
 818                               bytes, &run->mmio.data);
 819
 820         srcu_read_unlock(&vcpu->kvm->srcu, idx);
 821
 822         if (!ret) {
 823                 kvmppc_complete_mmio_load(vcpu, run);
 824                 vcpu->mmio_needed = 0;
 825                 return EMULATE_DONE;
 826         }
 827
 828         return EMULATE_DO_MMIO;
 829 }
 830 EXPORT_SYMBOL_GPL(kvmppc_handle_load);
 831
 832 /* Same as above, but sign extends */
 833 int kvmppc_handle_loads(struct kvm_run *run, struct kvm_vcpu *vcpu,
 834                         unsigned int rt, unsigned int bytes,
 835                         int is_default_endian)
 836 {
 837         int r;
 838
 839         vcpu->arch.mmio_sign_extend = 1;
 840         r = kvmppc_handle_load(run, vcpu, rt, bytes, is_default_endian);
 841
 842         return r;
 843 }
 844
 845 int kvmppc_handle_store(struct kvm_run *run, struct kvm_vcpu *vcpu,
 846                         u64 val, unsigned int bytes, int is_default_endian)
 847 {
 848         void *data = run->mmio.data;
 849         int idx, ret;
 850         int is_bigendian;
 851
 852         if (kvmppc_need_byteswap(vcpu)) {
 853                 /* Default endianness is "little endian". */
 854                 is_bigendian = !is_default_endian;
 855         } else {
 856                 /* Default endianness is "big endian". */
 857                 is_bigendian = is_default_endian;
 858         }
 859
 860         if (bytes > sizeof(run->mmio.data)) {
 861                 printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
 862                        run->mmio.len);
 863         }
 864
 865         run->mmio.phys_addr = vcpu->arch.paddr_accessed;
 866         run->mmio.len = bytes;
 867         run->mmio.is_write = 1;
 868         vcpu->mmio_needed = 1;
 869         vcpu->mmio_is_write = 1;
 870
 871         /* Store the value at the lowest bytes in 'data'. */
 872         if (is_bigendian) {
 873                 switch (bytes) {
 874                 case 8: *(u64 *)data = val; break;
 875                 case 4: *(u32 *)data = val; break;
 876                 case 2: *(u16 *)data = val; break;
 877                 case 1: *(u8  *)data = val; break;
 878                 }
 879         } else {
 880                 /* Store LE value into 'data'. */
 881                 switch (bytes) {
 882                 case 4: st_le32(data, val); break;
 883                 case 2: st_le16(data, val); break;
 884                 case 1: *(u8 *)data = val; break;
 885                 }
 886         }
 887
 888         idx = srcu_read_lock(&vcpu->kvm->srcu);
 889
 890         ret = kvm_io_bus_write(vcpu->kvm, KVM_MMIO_BUS, run->mmio.phys_addr,
 891                                bytes, &run->mmio.data);
 892
 893         srcu_read_unlock(&vcpu->kvm->srcu, idx);
 894
 895         if (!ret) {
 896                 vcpu->mmio_needed = 0;
 897                 return EMULATE_DONE;
 898         }
 899
 900         return EMULATE_DO_MMIO;
 901 }
 902 EXPORT_SYMBOL_GPL(kvmppc_handle_store);
 903
 904 int kvm_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
 905 {
 906         int r = 0;
 907         union kvmppc_one_reg val;
 908         int size;
 909
 910         size = one_reg_size(reg->id);
 911         if (size > sizeof(val))
 912                 return -EINVAL;
 913
 914         r = kvmppc_get_one_reg(vcpu, reg->id, &val);
 915         if (r == -EINVAL) {
 916                 r = 0;
 917                 switch (reg->id) {
 918 #ifdef CONFIG_ALTIVEC
 919                 case KVM_REG_PPC_VR0 ... KVM_REG_PPC_VR31:
 920                         if (!cpu_has_feature(CPU_FTR_ALTIVEC)) {
 921                                 r = -ENXIO;
 922                                 break;
 923                         }
 924                         vcpu->arch.vr.vr[reg->id - KVM_REG_PPC_VR0] = val.vval;
 925                         break;
 926                 case KVM_REG_PPC_VSCR:
 927                         if (!cpu_has_feature(CPU_FTR_ALTIVEC)) {
 928                                 r = -ENXIO;
 929                                 break;
 930                         }
 931                         vcpu->arch.vr.vscr.u[3] = set_reg_val(reg->id, val);
 932                         break;
 933                 case KVM_REG_PPC_VRSAVE:
 934                         if (!cpu_has_feature(CPU_FTR_ALTIVEC)) {
 935                                 r = -ENXIO;
 936                                 break;
 937                         }
 938                         vcpu->arch.vrsave = set_reg_val(reg->id, val);
 939                         break;
 940 #endif /* CONFIG_ALTIVEC */
 941                 default:
 942                         r = -EINVAL;
 943                         break;
 944                 }
 945         }
 946
 947         if (r)
 948                 return r;
 949
 950         if (copy_to_user((char __user *)(unsigned long)reg->addr, &val, size))
 951                 r = -EFAULT;
 952
 953         return r;
 954 }
 955
 956 int kvm_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
 957 {
 958         int r;
 959         union kvmppc_one_reg val;
 960         int size;
 961
 962         size = one_reg_size(reg->id);
 963         if (size > sizeof(val))
 964                 return -EINVAL;
 965
 966         if (copy_from_user(&val, (char __user *)(unsigned long)reg->addr, size))
 967                 return -EFAULT;
 968
 969         r = kvmppc_set_one_reg(vcpu, reg->id, &val);
 970         if (r == -EINVAL) {
 971                 r = 0;
 972                 switch (reg->id) {
 973 #ifdef CONFIG_ALTIVEC
 974                 case KVM_REG_PPC_VR0 ... KVM_REG_PPC_VR31:
 975                         if (!cpu_has_feature(CPU_FTR_ALTIVEC)) {
 976                                 r = -ENXIO;
 977                                 break;
 978                         }
 979                         val.vval = vcpu->arch.vr.vr[reg->id - KVM_REG_PPC_VR0];
 980                         break;
 981                 case KVM_REG_PPC_VSCR:
 982                         if (!cpu_has_feature(CPU_FTR_ALTIVEC)) {
 983                                 r = -ENXIO;
 984                                 break;
 985                         }
 986                         val = get_reg_val(reg->id, vcpu->arch.vr.vscr.u[3]);
 987                         break;
 988                 case KVM_REG_PPC_VRSAVE:
 989                         val = get_reg_val(reg->id, vcpu->arch.vrsave);
 990                         break;
 991 #endif /* CONFIG_ALTIVEC */
 992                 default:
 993                         r = -EINVAL;
 994                         break;
 995                 }
 996         }
 997
 998         return r;
 999 }
1000
1001 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)
1002 {
1003         int r;
1004         sigset_t sigsaved;
1005
1006         if (vcpu->sigset_active)
1007                 sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved);
1008
1009         if (vcpu->mmio_needed) {
1010                 if (!vcpu->mmio_is_write)
1011                         kvmppc_complete_mmio_load(vcpu, run);
1012                 vcpu->mmio_needed = 0;
1013         } else if (vcpu->arch.osi_needed) {
1014                 u64 *gprs = run->osi.gprs;
1015                 int i;
1016
1017                 for (i = 0; i < 32; i++)
1018                         kvmppc_set_gpr(vcpu, i, gprs[i]);
1019                 vcpu->arch.osi_needed = 0;
1020         } else if (vcpu->arch.hcall_needed) {
1021                 int i;
1022
1023                 kvmppc_set_gpr(vcpu, 3, run->papr_hcall.ret);
1024                 for (i = 0; i < 9; ++i)
1025                         kvmppc_set_gpr(vcpu, 4 + i, run->papr_hcall.args[i]);
1026                 vcpu->arch.hcall_needed = 0;
1027 #ifdef CONFIG_BOOKE
1028         } else if (vcpu->arch.epr_needed) {
1029                 kvmppc_set_epr(vcpu, run->epr.epr);
1030                 vcpu->arch.epr_needed = 0;
1031 #endif
1032         }
1033
1034         r = kvmppc_vcpu_run(run, vcpu);
1035
1036         if (vcpu->sigset_active)
1037                 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1038
1039         return r;
1040 }
1041
1042 int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu, struct kvm_interrupt *irq)
1043 {
1044         if (irq->irq == KVM_INTERRUPT_UNSET) {
1045                 kvmppc_core_dequeue_external(vcpu);
1046                 return 0;
1047         }
1048
1049         kvmppc_core_queue_external(vcpu, irq);
1050
1051         kvm_vcpu_kick(vcpu);
1052
1053         return 0;
1054 }
1055
1056 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
1057                                      struct kvm_enable_cap *cap)
1058 {
1059         int r;
1060
1061         if (cap->flags)
1062                 return -EINVAL;
1063
1064         switch (cap->cap) {
1065         case KVM_CAP_PPC_OSI:
1066                 r = 0;
1067                 vcpu->arch.osi_enabled = true;
1068                 break;
1069         case KVM_CAP_PPC_PAPR:
1070                 r = 0;
1071                 vcpu->arch.papr_enabled = true;
1072                 break;
1073         case KVM_CAP_PPC_EPR:
1074                 r = 0;
1075                 if (cap->args[0])
1076                         vcpu->arch.epr_flags |= KVMPPC_EPR_USER;
1077                 else
1078                         vcpu->arch.epr_flags &= ~KVMPPC_EPR_USER;
1079                 break;
1080 #ifdef CONFIG_BOOKE
1081         case KVM_CAP_PPC_BOOKE_WATCHDOG:
1082                 r = 0;
1083                 vcpu->arch.watchdog_enabled = true;
1084                 break;
1085 #endif
1086 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
1087         case KVM_CAP_SW_TLB: {
1088                 struct kvm_config_tlb cfg;
1089                 void __user *user_ptr = (void __user *)(uintptr_t)cap->args[0];
1090
1091                 r = -EFAULT;
1092                 if (copy_from_user(&cfg, user_ptr, sizeof(cfg)))
1093                         break;
1094
1095                 r = kvm_vcpu_ioctl_config_tlb(vcpu, &cfg);
1096                 break;
1097         }
1098 #endif
1099 #ifdef CONFIG_KVM_MPIC
1100         case KVM_CAP_IRQ_MPIC: {
1101                 struct fd f;
1102                 struct kvm_device *dev;
1103
1104                 r = -EBADF;
1105                 f = fdget(cap->args[0]);
1106                 if (!f.file)
1107                         break;
1108
1109                 r = -EPERM;
1110                 dev = kvm_device_from_filp(f.file);
1111                 if (dev)
1112                         r = kvmppc_mpic_connect_vcpu(dev, vcpu, cap->args[1]);
1113
1114                 fdput(f);
1115                 break;
1116         }
1117 #endif
1118 #ifdef CONFIG_KVM_XICS
1119         case KVM_CAP_IRQ_XICS: {
1120                 struct fd f;
1121                 struct kvm_device *dev;
1122
1123                 r = -EBADF;
1124                 f = fdget(cap->args[0]);
1125                 if (!f.file)
1126                         break;
1127
1128                 r = -EPERM;
1129                 dev = kvm_device_from_filp(f.file);
1130                 if (dev)
1131                         r = kvmppc_xics_connect_vcpu(dev, vcpu, cap->args[1]);
1132
1133                 fdput(f);
1134                 break;
1135         }
1136 #endif /* CONFIG_KVM_XICS */
1137         default:
1138                 r = -EINVAL;
1139                 break;
1140         }
1141
1142         if (!r)
1143                 r = kvmppc_sanity_check(vcpu);
1144
1145         return r;
1146 }
1147
1148 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
1149                                     struct kvm_mp_state *mp_state)
1150 {
1151         return -EINVAL;
1152 }
1153
1154 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
1155                                     struct kvm_mp_state *mp_state)
1156 {
1157         return -EINVAL;
1158 }
1159
1160 long kvm_arch_vcpu_ioctl(struct file *filp,
1161                          unsigned int ioctl, unsigned long arg)
1162 {
1163         struct kvm_vcpu *vcpu = filp->private_data;
1164         void __user *argp = (void __user *)arg;
1165         long r;
1166
1167         switch (ioctl) {
1168         case KVM_INTERRUPT: {
1169                 struct kvm_interrupt irq;
1170                 r = -EFAULT;
1171                 if (copy_from_user(&irq, argp, sizeof(irq)))
1172                         goto out;
1173                 r = kvm_vcpu_ioctl_interrupt(vcpu, &irq);
1174                 goto out;
1175         }
1176
1177         case KVM_ENABLE_CAP:
1178         {
1179                 struct kvm_enable_cap cap;
1180                 r = -EFAULT;
1181                 if (copy_from_user(&cap, argp, sizeof(cap)))
1182                         goto out;
1183                 r = kvm_vcpu_ioctl_enable_cap(vcpu, &cap);
1184                 break;
1185         }
1186
1187         case KVM_SET_ONE_REG:
1188         case KVM_GET_ONE_REG:
1189         {
1190                 struct kvm_one_reg reg;
1191                 r = -EFAULT;
1192                 if (copy_from_user(&reg, argp, sizeof(reg)))
1193                         goto out;
1194                 if (ioctl == KVM_SET_ONE_REG)
1195                         r = kvm_vcpu_ioctl_set_one_reg(vcpu, &reg);
1196                 else
1197                         r = kvm_vcpu_ioctl_get_one_reg(vcpu, &reg);
1198                 break;
1199         }
1200
1201 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
1202         case KVM_DIRTY_TLB: {
1203                 struct kvm_dirty_tlb dirty;
1204                 r = -EFAULT;
1205                 if (copy_from_user(&dirty, argp, sizeof(dirty)))
1206                         goto out;
1207                 r = kvm_vcpu_ioctl_dirty_tlb(vcpu, &dirty);
1208                 break;
1209         }
1210 #endif
1211         default:
1212                 r = -EINVAL;
1213         }
1214
1215 out:
1216         return r;
1217 }
1218
1219 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
1220 {
1221         return VM_FAULT_SIGBUS;
1222 }
1223
1224 static int kvm_vm_ioctl_get_pvinfo(struct kvm_ppc_pvinfo *pvinfo)
1225 {
1226         u32 inst_nop = 0x60000000;
1227 #ifdef CONFIG_KVM_BOOKE_HV
1228         u32 inst_sc1 = 0x44000022;
1229         pvinfo->hcall[0] = cpu_to_be32(inst_sc1);
1230         pvinfo->hcall[1] = cpu_to_be32(inst_nop);
1231         pvinfo->hcall[2] = cpu_to_be32(inst_nop);
1232         pvinfo->hcall[3] = cpu_to_be32(inst_nop);
1233 #else
1234         u32 inst_lis = 0x3c000000;
1235         u32 inst_ori = 0x60000000;
1236         u32 inst_sc = 0x44000002;
1237         u32 inst_imm_mask = 0xffff;
1238
1239         /*
1240          * The hypercall to get into KVM from within guest context is as
1241          * follows:
1242          *
1243          *    lis r0, r0, KVM_SC_MAGIC_R0@h
1244          *    ori r0, KVM_SC_MAGIC_R0@l
1245          *    sc
1246          *    nop
1247          */
1248         pvinfo->hcall[0] = cpu_to_be32(inst_lis | ((KVM_SC_MAGIC_R0 >> 16) & inst_imm_mask));
1249         pvinfo->hcall[1] = cpu_to_be32(inst_ori | (KVM_SC_MAGIC_R0 & inst_imm_mask));
1250         pvinfo->hcall[2] = cpu_to_be32(inst_sc);
1251         pvinfo->hcall[3] = cpu_to_be32(inst_nop);
1252 #endif
1253
1254         pvinfo->flags = KVM_PPC_PVINFO_FLAGS_EV_IDLE;
1255
1256         return 0;
1257 }
1258
1259 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_event,
1260                           bool line_status)
1261 {
1262         if (!irqchip_in_kernel(kvm))
1263                 return -ENXIO;
1264
1265         irq_event->status = kvm_set_irq(kvm, KVM_USERSPACE_IRQ_SOURCE_ID,
1266                                         irq_event->irq, irq_event->level,
1267                                         line_status);
1268         return 0;
1269 }
1270
1271
1272 static int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
1273                                    struct kvm_enable_cap *cap)
1274 {
1275         int r;
1276
1277         if (cap->flags)
1278                 return -EINVAL;
1279
1280         switch (cap->cap) {
1281 #ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1282         case KVM_CAP_PPC_ENABLE_HCALL: {
1283                 unsigned long hcall = cap->args[0];
1284
1285                 r = -EINVAL;
1286                 if (hcall > MAX_HCALL_OPCODE || (hcall & 3) ||
1287                     cap->args[1] > 1)
1288                         break;
1289                 if (!kvmppc_book3s_hcall_implemented(kvm, hcall))
1290                         break;
1291                 if (cap->args[1])
1292                         set_bit(hcall / 4, kvm->arch.enabled_hcalls);
1293                 else
1294                         clear_bit(hcall / 4, kvm->arch.enabled_hcalls);
1295                 r = 0;
1296                 break;
1297         }
1298 #endif
1299         default:
1300                 r = -EINVAL;
1301                 break;
1302         }
1303
1304         return r;
1305 }
1306
1307 long kvm_arch_vm_ioctl(struct file *filp,
1308                        unsigned int ioctl, unsigned long arg)
1309 {
1310         struct kvm *kvm __maybe_unused = filp->private_data;
1311         void __user *argp = (void __user *)arg;
1312         long r;
1313
1314         switch (ioctl) {
1315         case KVM_PPC_GET_PVINFO: {
1316                 struct kvm_ppc_pvinfo pvinfo;
1317                 memset(&pvinfo, 0, sizeof(pvinfo));
1318                 r = kvm_vm_ioctl_get_pvinfo(&pvinfo);
1319                 if (copy_to_user(argp, &pvinfo, sizeof(pvinfo))) {
1320                         r = -EFAULT;
1321                         goto out;
1322                 }
1323
1324                 break;
1325         }
1326         case KVM_ENABLE_CAP:
1327         {
1328                 struct kvm_enable_cap cap;
1329                 r = -EFAULT;
1330                 if (copy_from_user(&cap, argp, sizeof(cap)))
1331                         goto out;
1332                 r = kvm_vm_ioctl_enable_cap(kvm, &cap);
1333                 break;
1334         }
1335 #ifdef CONFIG_PPC_BOOK3S_64
1336         case KVM_CREATE_SPAPR_TCE: {
1337                 struct kvm_create_spapr_tce create_tce;
1338
1339                 r = -EFAULT;
1340                 if (copy_from_user(&create_tce, argp, sizeof(create_tce)))
1341                         goto out;
1342                 r = kvm_vm_ioctl_create_spapr_tce(kvm, &create_tce);
1343                 goto out;
1344         }
1345         case KVM_PPC_GET_SMMU_INFO: {
1346                 struct kvm_ppc_smmu_info info;
1347                 struct kvm *kvm = filp->private_data;
1348
1349                 memset(&info, 0, sizeof(info));
1350                 r = kvm->arch.kvm_ops->get_smmu_info(kvm, &info);
1351                 if (r >= 0 && copy_to_user(argp, &info, sizeof(info)))
1352                         r = -EFAULT;
1353                 break;
1354         }
1355         case KVM_PPC_RTAS_DEFINE_TOKEN: {
1356                 struct kvm *kvm = filp->private_data;
1357
1358                 r = kvm_vm_ioctl_rtas_define_token(kvm, argp);
1359                 break;
1360         }
1361         default: {
1362                 struct kvm *kvm = filp->private_data;
1363                 r = kvm->arch.kvm_ops->arch_vm_ioctl(filp, ioctl, arg);
1364         }
1365 #else /* CONFIG_PPC_BOOK3S_64 */
1366         default:
1367                 r = -ENOTTY;
1368 #endif
1369         }
1370 out:
1371         return r;
1372 }
1373
1374 static unsigned long lpid_inuse[BITS_TO_LONGS(KVMPPC_NR_LPIDS)];
1375 static unsigned long nr_lpids;
1376
1377 long kvmppc_alloc_lpid(void)
1378 {
1379         long lpid;
1380
1381         do {
1382                 lpid = find_first_zero_bit(lpid_inuse, KVMPPC_NR_LPIDS);
1383                 if (lpid >= nr_lpids) {
1384                         pr_err("%s: No LPIDs free\n", __func__);
1385                         return -ENOMEM;
1386                 }
1387         } while (test_and_set_bit(lpid, lpid_inuse));
1388
1389         return lpid;
1390 }
1391 EXPORT_SYMBOL_GPL(kvmppc_alloc_lpid);
1392
1393 void kvmppc_claim_lpid(long lpid)
1394 {
1395         set_bit(lpid, lpid_inuse);
1396 }
1397 EXPORT_SYMBOL_GPL(kvmppc_claim_lpid);
1398
1399 void kvmppc_free_lpid(long lpid)
1400 {
1401         clear_bit(lpid, lpid_inuse);
1402 }
1403 EXPORT_SYMBOL_GPL(kvmppc_free_lpid);
1404
1405 void kvmppc_init_lpid(unsigned long nr_lpids_param)
1406 {
1407         nr_lpids = min_t(unsigned long, KVMPPC_NR_LPIDS, nr_lpids_param);
1408         memset(lpid_inuse, 0, sizeof(lpid_inuse));
1409 }
1410 EXPORT_SYMBOL_GPL(kvmppc_init_lpid);
1411
1412 int kvm_arch_init(void *opaque)
1413 {
1414         return 0;
1415 }
1416
1417 EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_ppc_instr);