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 <linux/irqbypass.h>
  31 #include <linux/kvm_irqfd.h>
  32 #include <asm/cputable.h>
  33 #include <linux/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
  46 struct kvmppc_ops *kvmppc_hv_ops;
  47 EXPORT_SYMBOL_GPL(kvmppc_hv_ops);
  48 struct kvmppc_ops *kvmppc_pr_ops;
  49 EXPORT_SYMBOL_GPL(kvmppc_pr_ops);
  50
  51
  52 int kvm_arch_vcpu_runnable(struct kvm_vcpu *v)
  53 {
  54         return !!(v->arch.pending_exceptions) ||
  55                v->requests;
  56 }
  57
  58 int 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  */
  72 int 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 }
 132 EXPORT_SYMBOL_GPL(kvmppc_prepare_to_enter);
 133
 134 #if defined(CONFIG_PPC_BOOK3S_64) && defined(CONFIG_KVM_BOOK3S_PR_POSSIBLE)
 135 static 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
 155 int 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 }
 245 EXPORT_SYMBOL_GPL(kvmppc_kvm_pv);
 246
 247 int 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
 270 out:
 271         vcpu->arch.sane = r;
 272         return r ? 0 : -EINVAL;
 273 }
 274 EXPORT_SYMBOL_GPL(kvmppc_sanity_check);
 275
 276 int 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 }
 316 EXPORT_SYMBOL_GPL(kvmppc_emulate_mmio);
 317
 318 int 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 }
 352 EXPORT_SYMBOL_GPL(kvmppc_st);
 353
 354 int 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 }
 391 EXPORT_SYMBOL_GPL(kvmppc_ld);
 392
 393 int kvm_arch_hardware_enable(void)
 394 {
 395         return 0;
 396 }
 397
 398 int kvm_arch_hardware_setup(void)
 399 {
 400         return 0;
 401 }
 402
 403 void kvm_arch_check_processor_compat(void *rtn)
 404 {
 405         *(int *)rtn = kvmppc_core_check_processor_compat();
 406 }
 407
 408 int 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);
 437 err_out:
 438         return -EINVAL;
 439 }
 440
 441 bool kvm_arch_has_vcpu_debugfs(void)
 442 {
 443         return false;
 444 }
 445
 446 int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu)
 447 {
 448         return 0;
 449 }
 450
 451 void 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
 483 int 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                 /* fallthrough */
 540         case KVM_CAP_SPAPR_TCE_VFIO:
 541         case KVM_CAP_PPC_RTAS:
 542         case KVM_CAP_PPC_FIXUP_HCALL:
 543         case KVM_CAP_PPC_ENABLE_HCALL:
 544 #ifdef CONFIG_KVM_XICS
 545         case KVM_CAP_IRQ_XICS:
 546 #endif
 547                 r = 1;
 548                 break;
 549
 550         case KVM_CAP_PPC_ALLOC_HTAB:
 551                 r = hv_enabled;
 552                 break;
 553 #endif /* CONFIG_PPC_BOOK3S_64 */
 554 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
 555         case KVM_CAP_PPC_SMT:
 556                 r = 0;
 557                 if (hv_enabled) {
 558                         if (cpu_has_feature(CPU_FTR_ARCH_300))
 559                                 r = 1;
 560                         else
 561                                 r = threads_per_subcore;
 562                 }
 563                 break;
 564         case KVM_CAP_PPC_RMA:
 565                 r = 0;
 566                 break;
 567         case KVM_CAP_PPC_HWRNG:
 568                 r = kvmppc_hwrng_present();
 569                 break;
 570         case KVM_CAP_PPC_MMU_RADIX:
 571                 r = !!(hv_enabled && radix_enabled());
 572                 break;
 573         case KVM_CAP_PPC_MMU_HASH_V3:
 574                 r = !!(hv_enabled && !radix_enabled() &&
 575                        cpu_has_feature(CPU_FTR_ARCH_300));
 576                 break;
 577 #endif
 578         case KVM_CAP_SYNC_MMU:
 579 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
 580                 r = hv_enabled;
 581 #elif defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 582                 r = 1;
 583 #else
 584                 r = 0;
 585 #endif
 586                 break;
 587 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
 588         case KVM_CAP_PPC_HTAB_FD:
 589                 r = hv_enabled;
 590                 break;
 591 #endif
 592         case KVM_CAP_NR_VCPUS:
 593                 /*
 594                  * Recommending a number of CPUs is somewhat arbitrary; we
 595                  * return the number of present CPUs for -HV (since a host
 596                  * will have secondary threads "offline"), and for other KVM
 597                  * implementations just count online CPUs.
 598                  */
 599                 if (hv_enabled)
 600                         r = num_present_cpus();
 601                 else
 602                         r = num_online_cpus();
 603                 break;
 604         case KVM_CAP_NR_MEMSLOTS:
 605                 r = KVM_USER_MEM_SLOTS;
 606                 break;
 607         case KVM_CAP_MAX_VCPUS:
 608                 r = KVM_MAX_VCPUS;
 609                 break;
 610 #ifdef CONFIG_PPC_BOOK3S_64
 611         case KVM_CAP_PPC_GET_SMMU_INFO:
 612                 r = 1;
 613                 break;
 614         case KVM_CAP_SPAPR_MULTITCE:
 615                 r = 1;
 616                 break;
 617 #endif
 618         case KVM_CAP_PPC_HTM:
 619                 r = cpu_has_feature(CPU_FTR_TM_COMP) &&
 620                     is_kvmppc_hv_enabled(kvm);
 621                 break;
 622         default:
 623                 r = 0;
 624                 break;
 625         }
 626         return r;
 627
 628 }
 629
 630 long kvm_arch_dev_ioctl(struct file *filp,
 631                         unsigned int ioctl, unsigned long arg)
 632 {
 633         return -EINVAL;
 634 }
 635
 636 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
 637                            struct kvm_memory_slot *dont)
 638 {
 639         kvmppc_core_free_memslot(kvm, free, dont);
 640 }
 641
 642 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
 643                             unsigned long npages)
 644 {
 645         return kvmppc_core_create_memslot(kvm, slot, npages);
 646 }
 647
 648 int kvm_arch_prepare_memory_region(struct kvm *kvm,
 649                                    struct kvm_memory_slot *memslot,
 650                                    const struct kvm_userspace_memory_region *mem,
 651                                    enum kvm_mr_change change)
 652 {
 653         return kvmppc_core_prepare_memory_region(kvm, memslot, mem);
 654 }
 655
 656 void kvm_arch_commit_memory_region(struct kvm *kvm,
 657                                    const struct kvm_userspace_memory_region *mem,
 658                                    const struct kvm_memory_slot *old,
 659                                    const struct kvm_memory_slot *new,
 660                                    enum kvm_mr_change change)
 661 {
 662         kvmppc_core_commit_memory_region(kvm, mem, old, new);
 663 }
 664
 665 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
 666                                    struct kvm_memory_slot *slot)
 667 {
 668         kvmppc_core_flush_memslot(kvm, slot);
 669 }
 670
 671 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id)
 672 {
 673         struct kvm_vcpu *vcpu;
 674         vcpu = kvmppc_core_vcpu_create(kvm, id);
 675         if (!IS_ERR(vcpu)) {
 676                 vcpu->arch.wqp = &vcpu->wq;
 677                 kvmppc_create_vcpu_debugfs(vcpu, id);
 678         }
 679         return vcpu;
 680 }
 681
 682 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu)
 683 {
 684 }
 685
 686 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu)
 687 {
 688         /* Make sure we're not using the vcpu anymore */
 689         hrtimer_cancel(&vcpu->arch.dec_timer);
 690
 691         kvmppc_remove_vcpu_debugfs(vcpu);
 692
 693         switch (vcpu->arch.irq_type) {
 694         case KVMPPC_IRQ_MPIC:
 695                 kvmppc_mpic_disconnect_vcpu(vcpu->arch.mpic, vcpu);
 696                 break;
 697         case KVMPPC_IRQ_XICS:
 698                 kvmppc_xics_free_icp(vcpu);
 699                 break;
 700         }
 701
 702         kvmppc_core_vcpu_free(vcpu);
 703 }
 704
 705 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
 706 {
 707         kvm_arch_vcpu_free(vcpu);
 708 }
 709
 710 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
 711 {
 712         return kvmppc_core_pending_dec(vcpu);
 713 }
 714
 715 static enum hrtimer_restart kvmppc_decrementer_wakeup(struct hrtimer *timer)
 716 {
 717         struct kvm_vcpu *vcpu;
 718
 719         vcpu = container_of(timer, struct kvm_vcpu, arch.dec_timer);
 720         kvmppc_decrementer_func(vcpu);
 721
 722         return HRTIMER_NORESTART;
 723 }
 724
 725 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
 726 {
 727         int ret;
 728
 729         hrtimer_init(&vcpu->arch.dec_timer, CLOCK_REALTIME, HRTIMER_MODE_ABS);
 730         vcpu->arch.dec_timer.function = kvmppc_decrementer_wakeup;
 731         vcpu->arch.dec_expires = ~(u64)0;
 732
 733 #ifdef CONFIG_KVM_EXIT_TIMING
 734         mutex_init(&vcpu->arch.exit_timing_lock);
 735 #endif
 736         ret = kvmppc_subarch_vcpu_init(vcpu);
 737         return ret;
 738 }
 739
 740 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu)
 741 {
 742         kvmppc_mmu_destroy(vcpu);
 743         kvmppc_subarch_vcpu_uninit(vcpu);
 744 }
 745
 746 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
 747 {
 748 #ifdef CONFIG_BOOKE
 749         /*
 750          * vrsave (formerly usprg0) isn't used by Linux, but may
 751          * be used by the guest.
 752          *
 753          * On non-booke this is associated with Altivec and
 754          * is handled by code in book3s.c.
 755          */
 756         mtspr(SPRN_VRSAVE, vcpu->arch.vrsave);
 757 #endif
 758         kvmppc_core_vcpu_load(vcpu, cpu);
 759 }
 760
 761 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
 762 {
 763         kvmppc_core_vcpu_put(vcpu);
 764 #ifdef CONFIG_BOOKE
 765         vcpu->arch.vrsave = mfspr(SPRN_VRSAVE);
 766 #endif
 767 }
 768
 769 /*
 770  * irq_bypass_add_producer and irq_bypass_del_producer are only
 771  * useful if the architecture supports PCI passthrough.
 772  * irq_bypass_stop and irq_bypass_start are not needed and so
 773  * kvm_ops are not defined for them.
 774  */
 775 bool kvm_arch_has_irq_bypass(void)
 776 {
 777         return ((kvmppc_hv_ops && kvmppc_hv_ops->irq_bypass_add_producer) ||
 778                 (kvmppc_pr_ops && kvmppc_pr_ops->irq_bypass_add_producer));
 779 }
 780
 781 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *cons,
 782                                      struct irq_bypass_producer *prod)
 783 {
 784         struct kvm_kernel_irqfd *irqfd =
 785                 container_of(cons, struct kvm_kernel_irqfd, consumer);
 786         struct kvm *kvm = irqfd->kvm;
 787
 788         if (kvm->arch.kvm_ops->irq_bypass_add_producer)
 789                 return kvm->arch.kvm_ops->irq_bypass_add_producer(cons, prod);
 790
 791         return 0;
 792 }
 793
 794 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *cons,
 795                                       struct irq_bypass_producer *prod)
 796 {
 797         struct kvm_kernel_irqfd *irqfd =
 798                 container_of(cons, struct kvm_kernel_irqfd, consumer);
 799         struct kvm *kvm = irqfd->kvm;
 800
 801         if (kvm->arch.kvm_ops->irq_bypass_del_producer)
 802                 kvm->arch.kvm_ops->irq_bypass_del_producer(cons, prod);
 803 }
 804
 805 static void kvmppc_complete_mmio_load(struct kvm_vcpu *vcpu,
 806                                       struct kvm_run *run)
 807 {
 808         u64 uninitialized_var(gpr);
 809
 810         if (run->mmio.len > sizeof(gpr)) {
 811                 printk(KERN_ERR "bad MMIO length: %d\n", run->mmio.len);
 812                 return;
 813         }
 814
 815         if (!vcpu->arch.mmio_host_swabbed) {
 816                 switch (run->mmio.len) {
 817                 case 8: gpr = *(u64 *)run->mmio.data; break;
 818                 case 4: gpr = *(u32 *)run->mmio.data; break;
 819                 case 2: gpr = *(u16 *)run->mmio.data; break;
 820                 case 1: gpr = *(u8 *)run->mmio.data; break;
 821                 }
 822         } else {
 823                 switch (run->mmio.len) {
 824                 case 8: gpr = swab64(*(u64 *)run->mmio.data); break;
 825                 case 4: gpr = swab32(*(u32 *)run->mmio.data); break;
 826                 case 2: gpr = swab16(*(u16 *)run->mmio.data); break;
 827                 case 1: gpr = *(u8 *)run->mmio.data; break;
 828                 }
 829         }
 830
 831         if (vcpu->arch.mmio_sign_extend) {
 832                 switch (run->mmio.len) {
 833 #ifdef CONFIG_PPC64
 834                 case 4:
 835                         gpr = (s64)(s32)gpr;
 836                         break;
 837 #endif
 838                 case 2:
 839                         gpr = (s64)(s16)gpr;
 840                         break;
 841                 case 1:
 842                         gpr = (s64)(s8)gpr;
 843                         break;
 844                 }
 845         }
 846
 847         kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
 848
 849         switch (vcpu->arch.io_gpr & KVM_MMIO_REG_EXT_MASK) {
 850         case KVM_MMIO_REG_GPR:
 851                 kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
 852                 break;
 853         case KVM_MMIO_REG_FPR:
 854                 VCPU_FPR(vcpu, vcpu->arch.io_gpr & KVM_MMIO_REG_MASK) = gpr;
 855                 break;
 856 #ifdef CONFIG_PPC_BOOK3S
 857         case KVM_MMIO_REG_QPR:
 858                 vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
 859                 break;
 860         case KVM_MMIO_REG_FQPR:
 861                 VCPU_FPR(vcpu, vcpu->arch.io_gpr & KVM_MMIO_REG_MASK) = gpr;
 862                 vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
 863                 break;
 864 #endif
 865         default:
 866                 BUG();
 867         }
 868 }
 869
 870 static int __kvmppc_handle_load(struct kvm_run *run, struct kvm_vcpu *vcpu,
 871                                 unsigned int rt, unsigned int bytes,
 872                                 int is_default_endian, int sign_extend)
 873 {
 874         int idx, ret;
 875         bool host_swabbed;
 876
 877         /* Pity C doesn't have a logical XOR operator */
 878         if (kvmppc_need_byteswap(vcpu)) {
 879                 host_swabbed = is_default_endian;
 880         } else {
 881                 host_swabbed = !is_default_endian;
 882         }
 883
 884         if (bytes > sizeof(run->mmio.data)) {
 885                 printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
 886                        run->mmio.len);
 887         }
 888
 889         run->mmio.phys_addr = vcpu->arch.paddr_accessed;
 890         run->mmio.len = bytes;
 891         run->mmio.is_write = 0;
 892
 893         vcpu->arch.io_gpr = rt;
 894         vcpu->arch.mmio_host_swabbed = host_swabbed;
 895         vcpu->mmio_needed = 1;
 896         vcpu->mmio_is_write = 0;
 897         vcpu->arch.mmio_sign_extend = sign_extend;
 898
 899         idx = srcu_read_lock(&vcpu->kvm->srcu);
 900
 901         ret = kvm_io_bus_read(vcpu, KVM_MMIO_BUS, run->mmio.phys_addr,
 902                               bytes, &run->mmio.data);
 903
 904         srcu_read_unlock(&vcpu->kvm->srcu, idx);
 905
 906         if (!ret) {
 907                 kvmppc_complete_mmio_load(vcpu, run);
 908                 vcpu->mmio_needed = 0;
 909                 return EMULATE_DONE;
 910         }
 911
 912         return EMULATE_DO_MMIO;
 913 }
 914
 915 int kvmppc_handle_load(struct kvm_run *run, struct kvm_vcpu *vcpu,
 916                        unsigned int rt, unsigned int bytes,
 917                        int is_default_endian)
 918 {
 919         return __kvmppc_handle_load(run, vcpu, rt, bytes, is_default_endian, 0);
 920 }
 921 EXPORT_SYMBOL_GPL(kvmppc_handle_load);
 922
 923 /* Same as above, but sign extends */
 924 int kvmppc_handle_loads(struct kvm_run *run, struct kvm_vcpu *vcpu,
 925                         unsigned int rt, unsigned int bytes,
 926                         int is_default_endian)
 927 {
 928         return __kvmppc_handle_load(run, vcpu, rt, bytes, is_default_endian, 1);
 929 }
 930
 931 int kvmppc_handle_store(struct kvm_run *run, struct kvm_vcpu *vcpu,
 932                         u64 val, unsigned int bytes, int is_default_endian)
 933 {
 934         void *data = run->mmio.data;
 935         int idx, ret;
 936         bool host_swabbed;
 937
 938         /* Pity C doesn't have a logical XOR operator */
 939         if (kvmppc_need_byteswap(vcpu)) {
 940                 host_swabbed = is_default_endian;
 941         } else {
 942                 host_swabbed = !is_default_endian;
 943         }
 944
 945         if (bytes > sizeof(run->mmio.data)) {
 946                 printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
 947                        run->mmio.len);
 948         }
 949
 950         run->mmio.phys_addr = vcpu->arch.paddr_accessed;
 951         run->mmio.len = bytes;
 952         run->mmio.is_write = 1;
 953         vcpu->mmio_needed = 1;
 954         vcpu->mmio_is_write = 1;
 955
 956         /* Store the value at the lowest bytes in 'data'. */
 957         if (!host_swabbed) {
 958                 switch (bytes) {
 959                 case 8: *(u64 *)data = val; break;
 960                 case 4: *(u32 *)data = val; break;
 961                 case 2: *(u16 *)data = val; break;
 962                 case 1: *(u8  *)data = val; break;
 963                 }
 964         } else {
 965                 switch (bytes) {
 966                 case 8: *(u64 *)data = swab64(val); break;
 967                 case 4: *(u32 *)data = swab32(val); break;
 968                 case 2: *(u16 *)data = swab16(val); break;
 969                 case 1: *(u8  *)data = val; break;
 970                 }
 971         }
 972
 973         idx = srcu_read_lock(&vcpu->kvm->srcu);
 974
 975         ret = kvm_io_bus_write(vcpu, KVM_MMIO_BUS, run->mmio.phys_addr,
 976                                bytes, &run->mmio.data);
 977
 978         srcu_read_unlock(&vcpu->kvm->srcu, idx);
 979
 980         if (!ret) {
 981                 vcpu->mmio_needed = 0;
 982                 return EMULATE_DONE;
 983         }
 984
 985         return EMULATE_DO_MMIO;
 986 }
 987 EXPORT_SYMBOL_GPL(kvmppc_handle_store);
 988
 989 int kvm_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
 990 {
 991         int r = 0;
 992         union kvmppc_one_reg val;
 993         int size;
 994
 995         size = one_reg_size(reg->id);
 996         if (size > sizeof(val))
 997                 return -EINVAL;
 998
 999         r = kvmppc_get_one_reg(vcpu, reg->id, &val);
1000         if (r == -EINVAL) {
1001                 r = 0;
1002                 switch (reg->id) {
1003 #ifdef CONFIG_ALTIVEC
1004                 case KVM_REG_PPC_VR0 ... KVM_REG_PPC_VR31:
1005                         if (!cpu_has_feature(CPU_FTR_ALTIVEC)) {
1006                                 r = -ENXIO;
1007                                 break;
1008                         }
1009                         val.vval = vcpu->arch.vr.vr[reg->id - KVM_REG_PPC_VR0];
1010                         break;
1011                 case KVM_REG_PPC_VSCR:
1012                         if (!cpu_has_feature(CPU_FTR_ALTIVEC)) {
1013                                 r = -ENXIO;
1014                                 break;
1015                         }
1016                         val = get_reg_val(reg->id, vcpu->arch.vr.vscr.u[3]);
1017                         break;
1018                 case KVM_REG_PPC_VRSAVE:
1019                         val = get_reg_val(reg->id, vcpu->arch.vrsave);
1020                         break;
1021 #endif /* CONFIG_ALTIVEC */
1022                 default:
1023                         r = -EINVAL;
1024                         break;
1025                 }
1026         }
1027
1028         if (r)
1029                 return r;
1030
1031         if (copy_to_user((char __user *)(unsigned long)reg->addr, &val, size))
1032                 r = -EFAULT;
1033
1034         return r;
1035 }
1036
1037 int kvm_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1038 {
1039         int r;
1040         union kvmppc_one_reg val;
1041         int size;
1042
1043         size = one_reg_size(reg->id);
1044         if (size > sizeof(val))
1045                 return -EINVAL;
1046
1047         if (copy_from_user(&val, (char __user *)(unsigned long)reg->addr, size))
1048                 return -EFAULT;
1049
1050         r = kvmppc_set_one_reg(vcpu, reg->id, &val);
1051         if (r == -EINVAL) {
1052                 r = 0;
1053                 switch (reg->id) {
1054 #ifdef CONFIG_ALTIVEC
1055                 case KVM_REG_PPC_VR0 ... KVM_REG_PPC_VR31:
1056                         if (!cpu_has_feature(CPU_FTR_ALTIVEC)) {
1057                                 r = -ENXIO;
1058                                 break;
1059                         }
1060                         vcpu->arch.vr.vr[reg->id - KVM_REG_PPC_VR0] = val.vval;
1061                         break;
1062                 case KVM_REG_PPC_VSCR:
1063                         if (!cpu_has_feature(CPU_FTR_ALTIVEC)) {
1064                                 r = -ENXIO;
1065                                 break;
1066                         }
1067                         vcpu->arch.vr.vscr.u[3] = set_reg_val(reg->id, val);
1068                         break;
1069                 case KVM_REG_PPC_VRSAVE:
1070                         if (!cpu_has_feature(CPU_FTR_ALTIVEC)) {
1071                                 r = -ENXIO;
1072                                 break;
1073                         }
1074                         vcpu->arch.vrsave = set_reg_val(reg->id, val);
1075                         break;
1076 #endif /* CONFIG_ALTIVEC */
1077                 default:
1078                         r = -EINVAL;
1079                         break;
1080                 }
1081         }
1082
1083         return r;
1084 }
1085
1086 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)
1087 {
1088         int r;
1089         sigset_t sigsaved;
1090
1091         if (vcpu->sigset_active)
1092                 sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved);
1093
1094         if (vcpu->mmio_needed) {
1095                 if (!vcpu->mmio_is_write)
1096                         kvmppc_complete_mmio_load(vcpu, run);
1097                 vcpu->mmio_needed = 0;
1098         } else if (vcpu->arch.osi_needed) {
1099                 u64 *gprs = run->osi.gprs;
1100                 int i;
1101
1102                 for (i = 0; i < 32; i++)
1103                         kvmppc_set_gpr(vcpu, i, gprs[i]);
1104                 vcpu->arch.osi_needed = 0;
1105         } else if (vcpu->arch.hcall_needed) {
1106                 int i;
1107
1108                 kvmppc_set_gpr(vcpu, 3, run->papr_hcall.ret);
1109                 for (i = 0; i < 9; ++i)
1110                         kvmppc_set_gpr(vcpu, 4 + i, run->papr_hcall.args[i]);
1111                 vcpu->arch.hcall_needed = 0;
1112 #ifdef CONFIG_BOOKE
1113         } else if (vcpu->arch.epr_needed) {
1114                 kvmppc_set_epr(vcpu, run->epr.epr);
1115                 vcpu->arch.epr_needed = 0;
1116 #endif
1117         }
1118
1119         r = kvmppc_vcpu_run(run, vcpu);
1120
1121         if (vcpu->sigset_active)
1122                 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1123
1124         return r;
1125 }
1126
1127 int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu, struct kvm_interrupt *irq)
1128 {
1129         if (irq->irq == KVM_INTERRUPT_UNSET) {
1130                 kvmppc_core_dequeue_external(vcpu);
1131                 return 0;
1132         }
1133
1134         kvmppc_core_queue_external(vcpu, irq);
1135
1136         kvm_vcpu_kick(vcpu);
1137
1138         return 0;
1139 }
1140
1141 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
1142                                      struct kvm_enable_cap *cap)
1143 {
1144         int r;
1145
1146         if (cap->flags)
1147                 return -EINVAL;
1148
1149         switch (cap->cap) {
1150         case KVM_CAP_PPC_OSI:
1151                 r = 0;
1152                 vcpu->arch.osi_enabled = true;
1153                 break;
1154         case KVM_CAP_PPC_PAPR:
1155                 r = 0;
1156                 vcpu->arch.papr_enabled = true;
1157                 break;
1158         case KVM_CAP_PPC_EPR:
1159                 r = 0;
1160                 if (cap->args[0])
1161                         vcpu->arch.epr_flags |= KVMPPC_EPR_USER;
1162                 else
1163                         vcpu->arch.epr_flags &= ~KVMPPC_EPR_USER;
1164                 break;
1165 #ifdef CONFIG_BOOKE
1166         case KVM_CAP_PPC_BOOKE_WATCHDOG:
1167                 r = 0;
1168                 vcpu->arch.watchdog_enabled = true;
1169                 break;
1170 #endif
1171 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
1172         case KVM_CAP_SW_TLB: {
1173                 struct kvm_config_tlb cfg;
1174                 void __user *user_ptr = (void __user *)(uintptr_t)cap->args[0];
1175
1176                 r = -EFAULT;
1177                 if (copy_from_user(&cfg, user_ptr, sizeof(cfg)))
1178                         break;
1179
1180                 r = kvm_vcpu_ioctl_config_tlb(vcpu, &cfg);
1181                 break;
1182         }
1183 #endif
1184 #ifdef CONFIG_KVM_MPIC
1185         case KVM_CAP_IRQ_MPIC: {
1186                 struct fd f;
1187                 struct kvm_device *dev;
1188
1189                 r = -EBADF;
1190                 f = fdget(cap->args[0]);
1191                 if (!f.file)
1192                         break;
1193
1194                 r = -EPERM;
1195                 dev = kvm_device_from_filp(f.file);
1196                 if (dev)
1197                         r = kvmppc_mpic_connect_vcpu(dev, vcpu, cap->args[1]);
1198
1199                 fdput(f);
1200                 break;
1201         }
1202 #endif
1203 #ifdef CONFIG_KVM_XICS
1204         case KVM_CAP_IRQ_XICS: {
1205                 struct fd f;
1206                 struct kvm_device *dev;
1207
1208                 r = -EBADF;
1209                 f = fdget(cap->args[0]);
1210                 if (!f.file)
1211                         break;
1212
1213                 r = -EPERM;
1214                 dev = kvm_device_from_filp(f.file);
1215                 if (dev)
1216                         r = kvmppc_xics_connect_vcpu(dev, vcpu, cap->args[1]);
1217
1218                 fdput(f);
1219                 break;
1220         }
1221 #endif /* CONFIG_KVM_XICS */
1222         default:
1223                 r = -EINVAL;
1224                 break;
1225         }
1226
1227         if (!r)
1228                 r = kvmppc_sanity_check(vcpu);
1229
1230         return r;
1231 }
1232
1233 bool kvm_arch_intc_initialized(struct kvm *kvm)
1234 {
1235 #ifdef CONFIG_KVM_MPIC
1236         if (kvm->arch.mpic)
1237                 return true;
1238 #endif
1239 #ifdef CONFIG_KVM_XICS
1240         if (kvm->arch.xics)
1241                 return true;
1242 #endif
1243         return false;
1244 }
1245
1246 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
1247                                     struct kvm_mp_state *mp_state)
1248 {
1249         return -EINVAL;
1250 }
1251
1252 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
1253                                     struct kvm_mp_state *mp_state)
1254 {
1255         return -EINVAL;
1256 }
1257
1258 long kvm_arch_vcpu_ioctl(struct file *filp,
1259                          unsigned int ioctl, unsigned long arg)
1260 {
1261         struct kvm_vcpu *vcpu = filp->private_data;
1262         void __user *argp = (void __user *)arg;
1263         long r;
1264
1265         switch (ioctl) {
1266         case KVM_INTERRUPT: {
1267                 struct kvm_interrupt irq;
1268                 r = -EFAULT;
1269                 if (copy_from_user(&irq, argp, sizeof(irq)))
1270                         goto out;
1271                 r = kvm_vcpu_ioctl_interrupt(vcpu, &irq);
1272                 goto out;
1273         }
1274
1275         case KVM_ENABLE_CAP:
1276         {
1277                 struct kvm_enable_cap cap;
1278                 r = -EFAULT;
1279                 if (copy_from_user(&cap, argp, sizeof(cap)))
1280                         goto out;
1281                 r = kvm_vcpu_ioctl_enable_cap(vcpu, &cap);
1282                 break;
1283         }
1284
1285         case KVM_SET_ONE_REG:
1286         case KVM_GET_ONE_REG:
1287         {
1288                 struct kvm_one_reg reg;
1289                 r = -EFAULT;
1290                 if (copy_from_user(&reg, argp, sizeof(reg)))
1291                         goto out;
1292                 if (ioctl == KVM_SET_ONE_REG)
1293                         r = kvm_vcpu_ioctl_set_one_reg(vcpu, &reg);
1294                 else
1295                         r = kvm_vcpu_ioctl_get_one_reg(vcpu, &reg);
1296                 break;
1297         }
1298
1299 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
1300         case KVM_DIRTY_TLB: {
1301                 struct kvm_dirty_tlb dirty;
1302                 r = -EFAULT;
1303                 if (copy_from_user(&dirty, argp, sizeof(dirty)))
1304                         goto out;
1305                 r = kvm_vcpu_ioctl_dirty_tlb(vcpu, &dirty);
1306                 break;
1307         }
1308 #endif
1309         default:
1310                 r = -EINVAL;
1311         }
1312
1313 out:
1314         return r;
1315 }
1316
1317 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
1318 {
1319         return VM_FAULT_SIGBUS;
1320 }
1321
1322 static int kvm_vm_ioctl_get_pvinfo(struct kvm_ppc_pvinfo *pvinfo)
1323 {
1324         u32 inst_nop = 0x60000000;
1325 #ifdef CONFIG_KVM_BOOKE_HV
1326         u32 inst_sc1 = 0x44000022;
1327         pvinfo->hcall[0] = cpu_to_be32(inst_sc1);
1328         pvinfo->hcall[1] = cpu_to_be32(inst_nop);
1329         pvinfo->hcall[2] = cpu_to_be32(inst_nop);
1330         pvinfo->hcall[3] = cpu_to_be32(inst_nop);
1331 #else
1332         u32 inst_lis = 0x3c000000;
1333         u32 inst_ori = 0x60000000;
1334         u32 inst_sc = 0x44000002;
1335         u32 inst_imm_mask = 0xffff;
1336
1337         /*
1338          * The hypercall to get into KVM from within guest context is as
1339          * follows:
1340          *
1341          *    lis r0, r0, KVM_SC_MAGIC_R0@h
1342          *    ori r0, KVM_SC_MAGIC_R0@l
1343          *    sc
1344          *    nop
1345          */
1346         pvinfo->hcall[0] = cpu_to_be32(inst_lis | ((KVM_SC_MAGIC_R0 >> 16) & inst_imm_mask));
1347         pvinfo->hcall[1] = cpu_to_be32(inst_ori | (KVM_SC_MAGIC_R0 & inst_imm_mask));
1348         pvinfo->hcall[2] = cpu_to_be32(inst_sc);
1349         pvinfo->hcall[3] = cpu_to_be32(inst_nop);
1350 #endif
1351
1352         pvinfo->flags = KVM_PPC_PVINFO_FLAGS_EV_IDLE;
1353
1354         return 0;
1355 }
1356
1357 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_event,
1358                           bool line_status)
1359 {
1360         if (!irqchip_in_kernel(kvm))
1361                 return -ENXIO;
1362
1363         irq_event->status = kvm_set_irq(kvm, KVM_USERSPACE_IRQ_SOURCE_ID,
1364                                         irq_event->irq, irq_event->level,
1365                                         line_status);
1366         return 0;
1367 }
1368
1369
1370 static int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
1371                                    struct kvm_enable_cap *cap)
1372 {
1373         int r;
1374
1375         if (cap->flags)
1376                 return -EINVAL;
1377
1378         switch (cap->cap) {
1379 #ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1380         case KVM_CAP_PPC_ENABLE_HCALL: {
1381                 unsigned long hcall = cap->args[0];
1382
1383                 r = -EINVAL;
1384                 if (hcall > MAX_HCALL_OPCODE || (hcall & 3) ||
1385                     cap->args[1] > 1)
1386                         break;
1387                 if (!kvmppc_book3s_hcall_implemented(kvm, hcall))
1388                         break;
1389                 if (cap->args[1])
1390                         set_bit(hcall / 4, kvm->arch.enabled_hcalls);
1391                 else
1392                         clear_bit(hcall / 4, kvm->arch.enabled_hcalls);
1393                 r = 0;
1394                 break;
1395         }
1396 #endif
1397         default:
1398                 r = -EINVAL;
1399                 break;
1400         }
1401
1402         return r;
1403 }
1404
1405 long kvm_arch_vm_ioctl(struct file *filp,
1406                        unsigned int ioctl, unsigned long arg)
1407 {
1408         struct kvm *kvm __maybe_unused = filp->private_data;
1409         void __user *argp = (void __user *)arg;
1410         long r;
1411
1412         switch (ioctl) {
1413         case KVM_PPC_GET_PVINFO: {
1414                 struct kvm_ppc_pvinfo pvinfo;
1415                 memset(&pvinfo, 0, sizeof(pvinfo));
1416                 r = kvm_vm_ioctl_get_pvinfo(&pvinfo);
1417                 if (copy_to_user(argp, &pvinfo, sizeof(pvinfo))) {
1418                         r = -EFAULT;
1419                         goto out;
1420                 }
1421
1422                 break;
1423         }
1424         case KVM_ENABLE_CAP:
1425         {
1426                 struct kvm_enable_cap cap;
1427                 r = -EFAULT;
1428                 if (copy_from_user(&cap, argp, sizeof(cap)))
1429                         goto out;
1430                 r = kvm_vm_ioctl_enable_cap(kvm, &cap);
1431                 break;
1432         }
1433 #ifdef CONFIG_PPC_BOOK3S_64
1434         case KVM_CREATE_SPAPR_TCE_64: {
1435                 struct kvm_create_spapr_tce_64 create_tce_64;
1436
1437                 r = -EFAULT;
1438                 if (copy_from_user(&create_tce_64, argp, sizeof(create_tce_64)))
1439                         goto out;
1440                 if (create_tce_64.flags) {
1441                         r = -EINVAL;
1442                         goto out;
1443                 }
1444                 r = kvm_vm_ioctl_create_spapr_tce(kvm, &create_tce_64);
1445                 goto out;
1446         }
1447         case KVM_CREATE_SPAPR_TCE: {
1448                 struct kvm_create_spapr_tce create_tce;
1449                 struct kvm_create_spapr_tce_64 create_tce_64;
1450
1451                 r = -EFAULT;
1452                 if (copy_from_user(&create_tce, argp, sizeof(create_tce)))
1453                         goto out;
1454
1455                 create_tce_64.liobn = create_tce.liobn;
1456                 create_tce_64.page_shift = IOMMU_PAGE_SHIFT_4K;
1457                 create_tce_64.offset = 0;
1458                 create_tce_64.size = create_tce.window_size >>
1459                                 IOMMU_PAGE_SHIFT_4K;
1460                 create_tce_64.flags = 0;
1461                 r = kvm_vm_ioctl_create_spapr_tce(kvm, &create_tce_64);
1462                 goto out;
1463         }
1464         case KVM_PPC_GET_SMMU_INFO: {
1465                 struct kvm_ppc_smmu_info info;
1466                 struct kvm *kvm = filp->private_data;
1467
1468                 memset(&info, 0, sizeof(info));
1469                 r = kvm->arch.kvm_ops->get_smmu_info(kvm, &info);
1470                 if (r >= 0 && copy_to_user(argp, &info, sizeof(info)))
1471                         r = -EFAULT;
1472                 break;
1473         }
1474         case KVM_PPC_RTAS_DEFINE_TOKEN: {
1475                 struct kvm *kvm = filp->private_data;
1476
1477                 r = kvm_vm_ioctl_rtas_define_token(kvm, argp);
1478                 break;
1479         }
1480         case KVM_PPC_CONFIGURE_V3_MMU: {
1481                 struct kvm *kvm = filp->private_data;
1482                 struct kvm_ppc_mmuv3_cfg cfg;
1483
1484                 r = -EINVAL;
1485                 if (!kvm->arch.kvm_ops->configure_mmu)
1486                         goto out;
1487                 r = -EFAULT;
1488                 if (copy_from_user(&cfg, argp, sizeof(cfg)))
1489                         goto out;
1490                 r = kvm->arch.kvm_ops->configure_mmu(kvm, &cfg);
1491                 break;
1492         }
1493         case KVM_PPC_GET_RMMU_INFO: {
1494                 struct kvm *kvm = filp->private_data;
1495                 struct kvm_ppc_rmmu_info info;
1496
1497                 r = -EINVAL;
1498                 if (!kvm->arch.kvm_ops->get_rmmu_info)
1499                         goto out;
1500                 r = kvm->arch.kvm_ops->get_rmmu_info(kvm, &info);
1501                 if (r >= 0 && copy_to_user(argp, &info, sizeof(info)))
1502                         r = -EFAULT;
1503                 break;
1504         }
1505         default: {
1506                 struct kvm *kvm = filp->private_data;
1507                 r = kvm->arch.kvm_ops->arch_vm_ioctl(filp, ioctl, arg);
1508         }
1509 #else /* CONFIG_PPC_BOOK3S_64 */
1510         default:
1511                 r = -ENOTTY;
1512 #endif
1513         }
1514 out:
1515         return r;
1516 }
1517
1518 static unsigned long lpid_inuse[BITS_TO_LONGS(KVMPPC_NR_LPIDS)];
1519 static unsigned long nr_lpids;
1520
1521 long kvmppc_alloc_lpid(void)
1522 {
1523         long lpid;
1524
1525         do {
1526                 lpid = find_first_zero_bit(lpid_inuse, KVMPPC_NR_LPIDS);
1527                 if (lpid >= nr_lpids) {
1528                         pr_err("%s: No LPIDs free\n", __func__);
1529                         return -ENOMEM;
1530                 }
1531         } while (test_and_set_bit(lpid, lpid_inuse));
1532
1533         return lpid;
1534 }
1535 EXPORT_SYMBOL_GPL(kvmppc_alloc_lpid);
1536
1537 void kvmppc_claim_lpid(long lpid)
1538 {
1539         set_bit(lpid, lpid_inuse);
1540 }
1541 EXPORT_SYMBOL_GPL(kvmppc_claim_lpid);
1542
1543 void kvmppc_free_lpid(long lpid)
1544 {
1545         clear_bit(lpid, lpid_inuse);
1546 }
1547 EXPORT_SYMBOL_GPL(kvmppc_free_lpid);
1548
1549 void kvmppc_init_lpid(unsigned long nr_lpids_param)
1550 {
1551         nr_lpids = min_t(unsigned long, KVMPPC_NR_LPIDS, nr_lpids_param);
1552         memset(lpid_inuse, 0, sizeof(lpid_inuse));
1553 }
1554 EXPORT_SYMBOL_GPL(kvmppc_init_lpid);
1555
1556 int kvm_arch_init(void *opaque)
1557 {
1558         return 0;
1559 }
1560
1561 EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_ppc_instr);