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 <asm/cputable.h>
  30 #include <asm/uaccess.h>
  31 #include <asm/kvm_ppc.h>
  32 #include <asm/tlbflush.h>
  33 #include <asm/cputhreads.h>
  34 #include <asm/irqflags.h>
  35 #include "timing.h"
  36 #include "irq.h"
  37 #include "../mm/mmu_decl.h"
  38
  39 #define CREATE_TRACE_POINTS
  40 #include "trace.h"
  41
  42 struct kvmppc_ops *kvmppc_ops;
  43
  44 int kvm_arch_vcpu_runnable(struct kvm_vcpu *v)
  45 {
  46         return !!(v->arch.pending_exceptions) ||
  47                v->requests;
  48 }
  49
  50 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
  51 {
  52         return 1;
  53 }
  54
  55 #ifndef CONFIG_KVM_BOOK3S_64_HV
  56 /*
  57  * Common checks before entering the guest world.  Call with interrupts
  58  * disabled.
  59  *
  60  * returns:
  61  *
  62  * == 1 if we're ready to go into guest state
  63  * <= 0 if we need to go back to the host with return value
  64  */
  65 int kvmppc_prepare_to_enter(struct kvm_vcpu *vcpu)
  66 {
  67         int r = 1;
  68
  69         WARN_ON_ONCE(!irqs_disabled());
  70         while (true) {
  71                 if (need_resched()) {
  72                         local_irq_enable();
  73                         cond_resched();
  74                         local_irq_disable();
  75                         continue;
  76                 }
  77
  78                 if (signal_pending(current)) {
  79                         kvmppc_account_exit(vcpu, SIGNAL_EXITS);
  80                         vcpu->run->exit_reason = KVM_EXIT_INTR;
  81                         r = -EINTR;
  82                         break;
  83                 }
  84
  85                 vcpu->mode = IN_GUEST_MODE;
  86
  87                 /*
  88                  * Reading vcpu->requests must happen after setting vcpu->mode,
  89                  * so we don't miss a request because the requester sees
  90                  * OUTSIDE_GUEST_MODE and assumes we'll be checking requests
  91                  * before next entering the guest (and thus doesn't IPI).
  92                  */
  93                 smp_mb();
  94
  95                 if (vcpu->requests) {
  96                         /* Make sure we process requests preemptable */
  97                         local_irq_enable();
  98                         trace_kvm_check_requests(vcpu);
  99                         r = kvmppc_core_check_requests(vcpu);
 100                         local_irq_disable();
 101                         if (r > 0)
 102                                 continue;
 103                         break;
 104                 }
 105
 106                 if (kvmppc_core_prepare_to_enter(vcpu)) {
 107                         /* interrupts got enabled in between, so we
 108                            are back at square 1 */
 109                         continue;
 110                 }
 111
 112 #ifdef CONFIG_PPC64
 113                 /* lazy EE magic */
 114                 hard_irq_disable();
 115                 if (lazy_irq_pending()) {
 116                         /* Got an interrupt in between, try again */
 117                         local_irq_enable();
 118                         local_irq_disable();
 119                         kvm_guest_exit();
 120                         continue;
 121                 }
 122 #endif
 123
 124                 kvm_guest_enter();
 125                 break;
 126         }
 127
 128         return r;
 129 }
 130 #endif /* CONFIG_KVM_BOOK3S_64_HV */
 131
 132 int kvmppc_kvm_pv(struct kvm_vcpu *vcpu)
 133 {
 134         int nr = kvmppc_get_gpr(vcpu, 11);
 135         int r;
 136         unsigned long __maybe_unused param1 = kvmppc_get_gpr(vcpu, 3);
 137         unsigned long __maybe_unused param2 = kvmppc_get_gpr(vcpu, 4);
 138         unsigned long __maybe_unused param3 = kvmppc_get_gpr(vcpu, 5);
 139         unsigned long __maybe_unused param4 = kvmppc_get_gpr(vcpu, 6);
 140         unsigned long r2 = 0;
 141
 142         if (!(vcpu->arch.shared->msr & MSR_SF)) {
 143                 /* 32 bit mode */
 144                 param1 &= 0xffffffff;
 145                 param2 &= 0xffffffff;
 146                 param3 &= 0xffffffff;
 147                 param4 &= 0xffffffff;
 148         }
 149
 150         switch (nr) {
 151         case KVM_HCALL_TOKEN(KVM_HC_PPC_MAP_MAGIC_PAGE):
 152         {
 153                 vcpu->arch.magic_page_pa = param1;
 154                 vcpu->arch.magic_page_ea = param2;
 155
 156                 r2 = KVM_MAGIC_FEAT_SR | KVM_MAGIC_FEAT_MAS0_TO_SPRG7;
 157
 158                 r = EV_SUCCESS;
 159                 break;
 160         }
 161         case KVM_HCALL_TOKEN(KVM_HC_FEATURES):
 162                 r = EV_SUCCESS;
 163 #if defined(CONFIG_PPC_BOOK3S) || defined(CONFIG_KVM_E500V2)
 164                 /* XXX Missing magic page on 44x */
 165                 r2 |= (1 << KVM_FEATURE_MAGIC_PAGE);
 166 #endif
 167
 168                 /* Second return value is in r4 */
 169                 break;
 170         case EV_HCALL_TOKEN(EV_IDLE):
 171                 r = EV_SUCCESS;
 172                 kvm_vcpu_block(vcpu);
 173                 clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
 174                 break;
 175         default:
 176                 r = EV_UNIMPLEMENTED;
 177                 break;
 178         }
 179
 180         kvmppc_set_gpr(vcpu, 4, r2);
 181
 182         return r;
 183 }
 184
 185 int kvmppc_sanity_check(struct kvm_vcpu *vcpu)
 186 {
 187         int r = false;
 188
 189         /* We have to know what CPU to virtualize */
 190         if (!vcpu->arch.pvr)
 191                 goto out;
 192
 193         /* PAPR only works with book3s_64 */
 194         if ((vcpu->arch.cpu_type != KVM_CPU_3S_64) && vcpu->arch.papr_enabled)
 195                 goto out;
 196
 197 #ifdef CONFIG_KVM_BOOK3S_64_HV
 198         /* HV KVM can only do PAPR mode for now */
 199         if (!vcpu->arch.papr_enabled)
 200                 goto out;
 201 #endif
 202
 203 #ifdef CONFIG_KVM_BOOKE_HV
 204         if (!cpu_has_feature(CPU_FTR_EMB_HV))
 205                 goto out;
 206 #endif
 207
 208         r = true;
 209
 210 out:
 211         vcpu->arch.sane = r;
 212         return r ? 0 : -EINVAL;
 213 }
 214
 215 int kvmppc_emulate_mmio(struct kvm_run *run, struct kvm_vcpu *vcpu)
 216 {
 217         enum emulation_result er;
 218         int r;
 219
 220         er = kvmppc_emulate_instruction(run, vcpu);
 221         switch (er) {
 222         case EMULATE_DONE:
 223                 /* Future optimization: only reload non-volatiles if they were
 224                  * actually modified. */
 225                 r = RESUME_GUEST_NV;
 226                 break;
 227         case EMULATE_DO_MMIO:
 228                 run->exit_reason = KVM_EXIT_MMIO;
 229                 /* We must reload nonvolatiles because "update" load/store
 230                  * instructions modify register state. */
 231                 /* Future optimization: only reload non-volatiles if they were
 232                  * actually modified. */
 233                 r = RESUME_HOST_NV;
 234                 break;
 235         case EMULATE_FAIL:
 236                 /* XXX Deliver Program interrupt to guest. */
 237                 printk(KERN_EMERG "%s: emulation failed (%08x)\n", __func__,
 238                        kvmppc_get_last_inst(vcpu));
 239                 r = RESUME_HOST;
 240                 break;
 241         default:
 242                 WARN_ON(1);
 243                 r = RESUME_GUEST;
 244         }
 245
 246         return r;
 247 }
 248
 249 int kvm_arch_hardware_enable(void *garbage)
 250 {
 251         return 0;
 252 }
 253
 254 void kvm_arch_hardware_disable(void *garbage)
 255 {
 256 }
 257
 258 int kvm_arch_hardware_setup(void)
 259 {
 260         return 0;
 261 }
 262
 263 void kvm_arch_hardware_unsetup(void)
 264 {
 265 }
 266
 267 void kvm_arch_check_processor_compat(void *rtn)
 268 {
 269         *(int *)rtn = kvmppc_core_check_processor_compat();
 270 }
 271
 272 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
 273 {
 274         if (type)
 275                 return -EINVAL;
 276
 277         return kvmppc_core_init_vm(kvm);
 278 }
 279
 280 void kvm_arch_destroy_vm(struct kvm *kvm)
 281 {
 282         unsigned int i;
 283         struct kvm_vcpu *vcpu;
 284
 285         kvm_for_each_vcpu(i, vcpu, kvm)
 286                 kvm_arch_vcpu_free(vcpu);
 287
 288         mutex_lock(&kvm->lock);
 289         for (i = 0; i < atomic_read(&kvm->online_vcpus); i++)
 290                 kvm->vcpus[i] = NULL;
 291
 292         atomic_set(&kvm->online_vcpus, 0);
 293
 294         kvmppc_core_destroy_vm(kvm);
 295
 296         mutex_unlock(&kvm->lock);
 297 }
 298
 299 void kvm_arch_sync_events(struct kvm *kvm)
 300 {
 301 }
 302
 303 int kvm_dev_ioctl_check_extension(long ext)
 304 {
 305         int r;
 306
 307         switch (ext) {
 308 #ifdef CONFIG_BOOKE
 309         case KVM_CAP_PPC_BOOKE_SREGS:
 310         case KVM_CAP_PPC_BOOKE_WATCHDOG:
 311         case KVM_CAP_PPC_EPR:
 312 #else
 313         case KVM_CAP_PPC_SEGSTATE:
 314         case KVM_CAP_PPC_HIOR:
 315         case KVM_CAP_PPC_PAPR:
 316 #endif
 317         case KVM_CAP_PPC_UNSET_IRQ:
 318         case KVM_CAP_PPC_IRQ_LEVEL:
 319         case KVM_CAP_ENABLE_CAP:
 320         case KVM_CAP_ONE_REG:
 321         case KVM_CAP_IOEVENTFD:
 322         case KVM_CAP_DEVICE_CTRL:
 323                 r = 1;
 324                 break;
 325 #ifndef CONFIG_KVM_BOOK3S_64_HV
 326         case KVM_CAP_PPC_PAIRED_SINGLES:
 327         case KVM_CAP_PPC_OSI:
 328         case KVM_CAP_PPC_GET_PVINFO:
 329 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
 330         case KVM_CAP_SW_TLB:
 331 #endif
 332 #ifdef CONFIG_KVM_MPIC
 333         case KVM_CAP_IRQ_MPIC:
 334 #endif
 335                 r = 1;
 336                 break;
 337         case KVM_CAP_COALESCED_MMIO:
 338                 r = KVM_COALESCED_MMIO_PAGE_OFFSET;
 339                 break;
 340 #endif
 341 #ifdef CONFIG_PPC_BOOK3S_64
 342         case KVM_CAP_SPAPR_TCE:
 343         case KVM_CAP_PPC_ALLOC_HTAB:
 344         case KVM_CAP_PPC_RTAS:
 345 #ifdef CONFIG_KVM_XICS
 346         case KVM_CAP_IRQ_XICS:
 347 #endif
 348                 r = 1;
 349                 break;
 350 #endif /* CONFIG_PPC_BOOK3S_64 */
 351 #ifdef CONFIG_KVM_BOOK3S_64_HV
 352         case KVM_CAP_PPC_SMT:
 353                 r = threads_per_core;
 354                 break;
 355         case KVM_CAP_PPC_RMA:
 356                 r = 1;
 357                 /* PPC970 requires an RMA */
 358                 if (cpu_has_feature(CPU_FTR_ARCH_201))
 359                         r = 2;
 360                 break;
 361 #endif
 362         case KVM_CAP_SYNC_MMU:
 363 #ifdef CONFIG_KVM_BOOK3S_64_HV
 364                 r = cpu_has_feature(CPU_FTR_ARCH_206) ? 1 : 0;
 365 #elif defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 366                 r = 1;
 367 #else
 368                 r = 0;
 369                 break;
 370 #endif
 371 #ifdef CONFIG_KVM_BOOK3S_64_HV
 372         case KVM_CAP_PPC_HTAB_FD:
 373                 r = 1;
 374                 break;
 375 #endif
 376                 break;
 377         case KVM_CAP_NR_VCPUS:
 378                 /*
 379                  * Recommending a number of CPUs is somewhat arbitrary; we
 380                  * return the number of present CPUs for -HV (since a host
 381                  * will have secondary threads "offline"), and for other KVM
 382                  * implementations just count online CPUs.
 383                  */
 384 #ifdef CONFIG_KVM_BOOK3S_64_HV
 385                 r = num_present_cpus();
 386 #else
 387                 r = num_online_cpus();
 388 #endif
 389                 break;
 390         case KVM_CAP_MAX_VCPUS:
 391                 r = KVM_MAX_VCPUS;
 392                 break;
 393 #ifdef CONFIG_PPC_BOOK3S_64
 394         case KVM_CAP_PPC_GET_SMMU_INFO:
 395                 r = 1;
 396                 break;
 397 #endif
 398         default:
 399                 r = 0;
 400                 break;
 401         }
 402         return r;
 403
 404 }
 405
 406 long kvm_arch_dev_ioctl(struct file *filp,
 407                         unsigned int ioctl, unsigned long arg)
 408 {
 409         return -EINVAL;
 410 }
 411
 412 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
 413                            struct kvm_memory_slot *dont)
 414 {
 415         kvmppc_core_free_memslot(free, dont);
 416 }
 417
 418 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages)
 419 {
 420         return kvmppc_core_create_memslot(slot, npages);
 421 }
 422
 423 void kvm_arch_memslots_updated(struct kvm *kvm)
 424 {
 425 }
 426
 427 int kvm_arch_prepare_memory_region(struct kvm *kvm,
 428                                    struct kvm_memory_slot *memslot,
 429                                    struct kvm_userspace_memory_region *mem,
 430                                    enum kvm_mr_change change)
 431 {
 432         return kvmppc_core_prepare_memory_region(kvm, memslot, mem);
 433 }
 434
 435 void kvm_arch_commit_memory_region(struct kvm *kvm,
 436                                    struct kvm_userspace_memory_region *mem,
 437                                    const struct kvm_memory_slot *old,
 438                                    enum kvm_mr_change change)
 439 {
 440         kvmppc_core_commit_memory_region(kvm, mem, old);
 441 }
 442
 443 void kvm_arch_flush_shadow_all(struct kvm *kvm)
 444 {
 445 }
 446
 447 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
 448                                    struct kvm_memory_slot *slot)
 449 {
 450         kvmppc_core_flush_memslot(kvm, slot);
 451 }
 452
 453 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id)
 454 {
 455         struct kvm_vcpu *vcpu;
 456         vcpu = kvmppc_core_vcpu_create(kvm, id);
 457         if (!IS_ERR(vcpu)) {
 458                 vcpu->arch.wqp = &vcpu->wq;
 459                 kvmppc_create_vcpu_debugfs(vcpu, id);
 460         }
 461         return vcpu;
 462 }
 463
 464 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu)
 465 {
 466         return 0;
 467 }
 468
 469 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu)
 470 {
 471         /* Make sure we're not using the vcpu anymore */
 472         hrtimer_cancel(&vcpu->arch.dec_timer);
 473         tasklet_kill(&vcpu->arch.tasklet);
 474
 475         kvmppc_remove_vcpu_debugfs(vcpu);
 476
 477         switch (vcpu->arch.irq_type) {
 478         case KVMPPC_IRQ_MPIC:
 479                 kvmppc_mpic_disconnect_vcpu(vcpu->arch.mpic, vcpu);
 480                 break;
 481         case KVMPPC_IRQ_XICS:
 482                 kvmppc_xics_free_icp(vcpu);
 483                 break;
 484         }
 485
 486         kvmppc_core_vcpu_free(vcpu);
 487 }
 488
 489 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
 490 {
 491         kvm_arch_vcpu_free(vcpu);
 492 }
 493
 494 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
 495 {
 496         return kvmppc_core_pending_dec(vcpu);
 497 }
 498
 499 /*
 500  * low level hrtimer wake routine. Because this runs in hardirq context
 501  * we schedule a tasklet to do the real work.
 502  */
 503 enum hrtimer_restart kvmppc_decrementer_wakeup(struct hrtimer *timer)
 504 {
 505         struct kvm_vcpu *vcpu;
 506
 507         vcpu = container_of(timer, struct kvm_vcpu, arch.dec_timer);
 508         tasklet_schedule(&vcpu->arch.tasklet);
 509
 510         return HRTIMER_NORESTART;
 511 }
 512
 513 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
 514 {
 515         int ret;
 516
 517         hrtimer_init(&vcpu->arch.dec_timer, CLOCK_REALTIME, HRTIMER_MODE_ABS);
 518         tasklet_init(&vcpu->arch.tasklet, kvmppc_decrementer_func, (ulong)vcpu);
 519         vcpu->arch.dec_timer.function = kvmppc_decrementer_wakeup;
 520         vcpu->arch.dec_expires = ~(u64)0;
 521
 522 #ifdef CONFIG_KVM_EXIT_TIMING
 523         mutex_init(&vcpu->arch.exit_timing_lock);
 524 #endif
 525         ret = kvmppc_subarch_vcpu_init(vcpu);
 526         return ret;
 527 }
 528
 529 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu)
 530 {
 531         kvmppc_mmu_destroy(vcpu);
 532         kvmppc_subarch_vcpu_uninit(vcpu);
 533 }
 534
 535 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
 536 {
 537 #ifdef CONFIG_BOOKE
 538         /*
 539          * vrsave (formerly usprg0) isn't used by Linux, but may
 540          * be used by the guest.
 541          *
 542          * On non-booke this is associated with Altivec and
 543          * is handled by code in book3s.c.
 544          */
 545         mtspr(SPRN_VRSAVE, vcpu->arch.vrsave);
 546 #endif
 547         kvmppc_core_vcpu_load(vcpu, cpu);
 548 }
 549
 550 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
 551 {
 552         kvmppc_core_vcpu_put(vcpu);
 553 #ifdef CONFIG_BOOKE
 554         vcpu->arch.vrsave = mfspr(SPRN_VRSAVE);
 555 #endif
 556 }
 557
 558 static void kvmppc_complete_dcr_load(struct kvm_vcpu *vcpu,
 559                                      struct kvm_run *run)
 560 {
 561         kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, run->dcr.data);
 562 }
 563
 564 static void kvmppc_complete_mmio_load(struct kvm_vcpu *vcpu,
 565                                       struct kvm_run *run)
 566 {
 567         u64 uninitialized_var(gpr);
 568
 569         if (run->mmio.len > sizeof(gpr)) {
 570                 printk(KERN_ERR "bad MMIO length: %d\n", run->mmio.len);
 571                 return;
 572         }
 573
 574         if (vcpu->arch.mmio_is_bigendian) {
 575                 switch (run->mmio.len) {
 576                 case 8: gpr = *(u64 *)run->mmio.data; break;
 577                 case 4: gpr = *(u32 *)run->mmio.data; break;
 578                 case 2: gpr = *(u16 *)run->mmio.data; break;
 579                 case 1: gpr = *(u8 *)run->mmio.data; break;
 580                 }
 581         } else {
 582                 /* Convert BE data from userland back to LE. */
 583                 switch (run->mmio.len) {
 584                 case 4: gpr = ld_le32((u32 *)run->mmio.data); break;
 585                 case 2: gpr = ld_le16((u16 *)run->mmio.data); break;
 586                 case 1: gpr = *(u8 *)run->mmio.data; break;
 587                 }
 588         }
 589
 590         if (vcpu->arch.mmio_sign_extend) {
 591                 switch (run->mmio.len) {
 592 #ifdef CONFIG_PPC64
 593                 case 4:
 594                         gpr = (s64)(s32)gpr;
 595                         break;
 596 #endif
 597                 case 2:
 598                         gpr = (s64)(s16)gpr;
 599                         break;
 600                 case 1:
 601                         gpr = (s64)(s8)gpr;
 602                         break;
 603                 }
 604         }
 605
 606         kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
 607
 608         switch (vcpu->arch.io_gpr & KVM_MMIO_REG_EXT_MASK) {
 609         case KVM_MMIO_REG_GPR:
 610                 kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
 611                 break;
 612         case KVM_MMIO_REG_FPR:
 613                 vcpu->arch.fpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
 614                 break;
 615 #ifdef CONFIG_PPC_BOOK3S
 616         case KVM_MMIO_REG_QPR:
 617                 vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
 618                 break;
 619         case KVM_MMIO_REG_FQPR:
 620                 vcpu->arch.fpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
 621                 vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
 622                 break;
 623 #endif
 624         default:
 625                 BUG();
 626         }
 627 }
 628
 629 int kvmppc_handle_load(struct kvm_run *run, struct kvm_vcpu *vcpu,
 630                        unsigned int rt, unsigned int bytes, int is_bigendian)
 631 {
 632         int idx, ret;
 633
 634         if (bytes > sizeof(run->mmio.data)) {
 635                 printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
 636                        run->mmio.len);
 637         }
 638
 639         run->mmio.phys_addr = vcpu->arch.paddr_accessed;
 640         run->mmio.len = bytes;
 641         run->mmio.is_write = 0;
 642
 643         vcpu->arch.io_gpr = rt;
 644         vcpu->arch.mmio_is_bigendian = is_bigendian;
 645         vcpu->mmio_needed = 1;
 646         vcpu->mmio_is_write = 0;
 647         vcpu->arch.mmio_sign_extend = 0;
 648
 649         idx = srcu_read_lock(&vcpu->kvm->srcu);
 650
 651         ret = kvm_io_bus_read(vcpu->kvm, KVM_MMIO_BUS, run->mmio.phys_addr,
 652                               bytes, &run->mmio.data);
 653
 654         srcu_read_unlock(&vcpu->kvm->srcu, idx);
 655
 656         if (!ret) {
 657                 kvmppc_complete_mmio_load(vcpu, run);
 658                 vcpu->mmio_needed = 0;
 659                 return EMULATE_DONE;
 660         }
 661
 662         return EMULATE_DO_MMIO;
 663 }
 664
 665 /* Same as above, but sign extends */
 666 int kvmppc_handle_loads(struct kvm_run *run, struct kvm_vcpu *vcpu,
 667                         unsigned int rt, unsigned int bytes, int is_bigendian)
 668 {
 669         int r;
 670
 671         vcpu->arch.mmio_sign_extend = 1;
 672         r = kvmppc_handle_load(run, vcpu, rt, bytes, is_bigendian);
 673
 674         return r;
 675 }
 676
 677 int kvmppc_handle_store(struct kvm_run *run, struct kvm_vcpu *vcpu,
 678                         u64 val, unsigned int bytes, int is_bigendian)
 679 {
 680         void *data = run->mmio.data;
 681         int idx, ret;
 682
 683         if (bytes > sizeof(run->mmio.data)) {
 684                 printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
 685                        run->mmio.len);
 686         }
 687
 688         run->mmio.phys_addr = vcpu->arch.paddr_accessed;
 689         run->mmio.len = bytes;
 690         run->mmio.is_write = 1;
 691         vcpu->mmio_needed = 1;
 692         vcpu->mmio_is_write = 1;
 693
 694         /* Store the value at the lowest bytes in 'data'. */
 695         if (is_bigendian) {
 696                 switch (bytes) {
 697                 case 8: *(u64 *)data = val; break;
 698                 case 4: *(u32 *)data = val; break;
 699                 case 2: *(u16 *)data = val; break;
 700                 case 1: *(u8  *)data = val; break;
 701                 }
 702         } else {
 703                 /* Store LE value into 'data'. */
 704                 switch (bytes) {
 705                 case 4: st_le32(data, val); break;
 706                 case 2: st_le16(data, val); break;
 707                 case 1: *(u8 *)data = val; break;
 708                 }
 709         }
 710
 711         idx = srcu_read_lock(&vcpu->kvm->srcu);
 712
 713         ret = kvm_io_bus_write(vcpu->kvm, KVM_MMIO_BUS, run->mmio.phys_addr,
 714                                bytes, &run->mmio.data);
 715
 716         srcu_read_unlock(&vcpu->kvm->srcu, idx);
 717
 718         if (!ret) {
 719                 vcpu->mmio_needed = 0;
 720                 return EMULATE_DONE;
 721         }
 722
 723         return EMULATE_DO_MMIO;
 724 }
 725
 726 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)
 727 {
 728         int r;
 729         sigset_t sigsaved;
 730
 731         if (vcpu->sigset_active)
 732                 sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved);
 733
 734         if (vcpu->mmio_needed) {
 735                 if (!vcpu->mmio_is_write)
 736                         kvmppc_complete_mmio_load(vcpu, run);
 737                 vcpu->mmio_needed = 0;
 738         } else if (vcpu->arch.dcr_needed) {
 739                 if (!vcpu->arch.dcr_is_write)
 740                         kvmppc_complete_dcr_load(vcpu, run);
 741                 vcpu->arch.dcr_needed = 0;
 742         } else if (vcpu->arch.osi_needed) {
 743                 u64 *gprs = run->osi.gprs;
 744                 int i;
 745
 746                 for (i = 0; i < 32; i++)
 747                         kvmppc_set_gpr(vcpu, i, gprs[i]);
 748                 vcpu->arch.osi_needed = 0;
 749         } else if (vcpu->arch.hcall_needed) {
 750                 int i;
 751
 752                 kvmppc_set_gpr(vcpu, 3, run->papr_hcall.ret);
 753                 for (i = 0; i < 9; ++i)
 754                         kvmppc_set_gpr(vcpu, 4 + i, run->papr_hcall.args[i]);
 755                 vcpu->arch.hcall_needed = 0;
 756 #ifdef CONFIG_BOOKE
 757         } else if (vcpu->arch.epr_needed) {
 758                 kvmppc_set_epr(vcpu, run->epr.epr);
 759                 vcpu->arch.epr_needed = 0;
 760 #endif
 761         }
 762
 763         r = kvmppc_vcpu_run(run, vcpu);
 764
 765         if (vcpu->sigset_active)
 766                 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
 767
 768         return r;
 769 }
 770
 771 int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu, struct kvm_interrupt *irq)
 772 {
 773         if (irq->irq == KVM_INTERRUPT_UNSET) {
 774                 kvmppc_core_dequeue_external(vcpu);
 775                 return 0;
 776         }
 777
 778         kvmppc_core_queue_external(vcpu, irq);
 779
 780         kvm_vcpu_kick(vcpu);
 781
 782         return 0;
 783 }
 784
 785 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
 786                                      struct kvm_enable_cap *cap)
 787 {
 788         int r;
 789
 790         if (cap->flags)
 791                 return -EINVAL;
 792
 793         switch (cap->cap) {
 794         case KVM_CAP_PPC_OSI:
 795                 r = 0;
 796                 vcpu->arch.osi_enabled = true;
 797                 break;
 798         case KVM_CAP_PPC_PAPR:
 799                 r = 0;
 800                 vcpu->arch.papr_enabled = true;
 801                 break;
 802         case KVM_CAP_PPC_EPR:
 803                 r = 0;
 804                 if (cap->args[0])
 805                         vcpu->arch.epr_flags |= KVMPPC_EPR_USER;
 806                 else
 807                         vcpu->arch.epr_flags &= ~KVMPPC_EPR_USER;
 808                 break;
 809 #ifdef CONFIG_BOOKE
 810         case KVM_CAP_PPC_BOOKE_WATCHDOG:
 811                 r = 0;
 812                 vcpu->arch.watchdog_enabled = true;
 813                 break;
 814 #endif
 815 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
 816         case KVM_CAP_SW_TLB: {
 817                 struct kvm_config_tlb cfg;
 818                 void __user *user_ptr = (void __user *)(uintptr_t)cap->args[0];
 819
 820                 r = -EFAULT;
 821                 if (copy_from_user(&cfg, user_ptr, sizeof(cfg)))
 822                         break;
 823
 824                 r = kvm_vcpu_ioctl_config_tlb(vcpu, &cfg);
 825                 break;
 826         }
 827 #endif
 828 #ifdef CONFIG_KVM_MPIC
 829         case KVM_CAP_IRQ_MPIC: {
 830                 struct fd f;
 831                 struct kvm_device *dev;
 832
 833                 r = -EBADF;
 834                 f = fdget(cap->args[0]);
 835                 if (!f.file)
 836                         break;
 837
 838                 r = -EPERM;
 839                 dev = kvm_device_from_filp(f.file);
 840                 if (dev)
 841                         r = kvmppc_mpic_connect_vcpu(dev, vcpu, cap->args[1]);
 842
 843                 fdput(f);
 844                 break;
 845         }
 846 #endif
 847 #ifdef CONFIG_KVM_XICS
 848         case KVM_CAP_IRQ_XICS: {
 849                 struct fd f;
 850                 struct kvm_device *dev;
 851
 852                 r = -EBADF;
 853                 f = fdget(cap->args[0]);
 854                 if (!f.file)
 855                         break;
 856
 857                 r = -EPERM;
 858                 dev = kvm_device_from_filp(f.file);
 859                 if (dev)
 860                         r = kvmppc_xics_connect_vcpu(dev, vcpu, cap->args[1]);
 861
 862                 fdput(f);
 863                 break;
 864         }
 865 #endif /* CONFIG_KVM_XICS */
 866         default:
 867                 r = -EINVAL;
 868                 break;
 869         }
 870
 871         if (!r)
 872                 r = kvmppc_sanity_check(vcpu);
 873
 874         return r;
 875 }
 876
 877 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
 878                                     struct kvm_mp_state *mp_state)
 879 {
 880         return -EINVAL;
 881 }
 882
 883 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
 884                                     struct kvm_mp_state *mp_state)
 885 {
 886         return -EINVAL;
 887 }
 888
 889 long kvm_arch_vcpu_ioctl(struct file *filp,
 890                          unsigned int ioctl, unsigned long arg)
 891 {
 892         struct kvm_vcpu *vcpu = filp->private_data;
 893         void __user *argp = (void __user *)arg;
 894         long r;
 895
 896         switch (ioctl) {
 897         case KVM_INTERRUPT: {
 898                 struct kvm_interrupt irq;
 899                 r = -EFAULT;
 900                 if (copy_from_user(&irq, argp, sizeof(irq)))
 901                         goto out;
 902                 r = kvm_vcpu_ioctl_interrupt(vcpu, &irq);
 903                 goto out;
 904         }
 905
 906         case KVM_ENABLE_CAP:
 907         {
 908                 struct kvm_enable_cap cap;
 909                 r = -EFAULT;
 910                 if (copy_from_user(&cap, argp, sizeof(cap)))
 911                         goto out;
 912                 r = kvm_vcpu_ioctl_enable_cap(vcpu, &cap);
 913                 break;
 914         }
 915
 916         case KVM_SET_ONE_REG:
 917         case KVM_GET_ONE_REG:
 918         {
 919                 struct kvm_one_reg reg;
 920                 r = -EFAULT;
 921                 if (copy_from_user(&reg, argp, sizeof(reg)))
 922                         goto out;
 923                 if (ioctl == KVM_SET_ONE_REG)
 924                         r = kvm_vcpu_ioctl_set_one_reg(vcpu, &reg);
 925                 else
 926                         r = kvm_vcpu_ioctl_get_one_reg(vcpu, &reg);
 927                 break;
 928         }
 929
 930 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
 931         case KVM_DIRTY_TLB: {
 932                 struct kvm_dirty_tlb dirty;
 933                 r = -EFAULT;
 934                 if (copy_from_user(&dirty, argp, sizeof(dirty)))
 935                         goto out;
 936                 r = kvm_vcpu_ioctl_dirty_tlb(vcpu, &dirty);
 937                 break;
 938         }
 939 #endif
 940         default:
 941                 r = -EINVAL;
 942         }
 943
 944 out:
 945         return r;
 946 }
 947
 948 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
 949 {
 950         return VM_FAULT_SIGBUS;
 951 }
 952
 953 static int kvm_vm_ioctl_get_pvinfo(struct kvm_ppc_pvinfo *pvinfo)
 954 {
 955         u32 inst_nop = 0x60000000;
 956 #ifdef CONFIG_KVM_BOOKE_HV
 957         u32 inst_sc1 = 0x44000022;
 958         pvinfo->hcall[0] = inst_sc1;
 959         pvinfo->hcall[1] = inst_nop;
 960         pvinfo->hcall[2] = inst_nop;
 961         pvinfo->hcall[3] = inst_nop;
 962 #else
 963         u32 inst_lis = 0x3c000000;
 964         u32 inst_ori = 0x60000000;
 965         u32 inst_sc = 0x44000002;
 966         u32 inst_imm_mask = 0xffff;
 967
 968         /*
 969          * The hypercall to get into KVM from within guest context is as
 970          * follows:
 971          *
 972          *    lis r0, r0, KVM_SC_MAGIC_R0@h
 973          *    ori r0, KVM_SC_MAGIC_R0@l
 974          *    sc
 975          *    nop
 976          */
 977         pvinfo->hcall[0] = inst_lis | ((KVM_SC_MAGIC_R0 >> 16) & inst_imm_mask);
 978         pvinfo->hcall[1] = inst_ori | (KVM_SC_MAGIC_R0 & inst_imm_mask);
 979         pvinfo->hcall[2] = inst_sc;
 980         pvinfo->hcall[3] = inst_nop;
 981 #endif
 982
 983         pvinfo->flags = KVM_PPC_PVINFO_FLAGS_EV_IDLE;
 984
 985         return 0;
 986 }
 987
 988 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_event,
 989                           bool line_status)
 990 {
 991         if (!irqchip_in_kernel(kvm))
 992                 return -ENXIO;
 993
 994         irq_event->status = kvm_set_irq(kvm, KVM_USERSPACE_IRQ_SOURCE_ID,
 995                                         irq_event->irq, irq_event->level,
 996                                         line_status);
 997         return 0;
 998 }
 999
1000 long kvm_arch_vm_ioctl(struct file *filp,
1001                        unsigned int ioctl, unsigned long arg)
1002 {
1003         struct kvm *kvm __maybe_unused = filp->private_data;
1004         void __user *argp = (void __user *)arg;
1005         long r;
1006
1007         switch (ioctl) {
1008         case KVM_PPC_GET_PVINFO: {
1009                 struct kvm_ppc_pvinfo pvinfo;
1010                 memset(&pvinfo, 0, sizeof(pvinfo));
1011                 r = kvm_vm_ioctl_get_pvinfo(&pvinfo);
1012                 if (copy_to_user(argp, &pvinfo, sizeof(pvinfo))) {
1013                         r = -EFAULT;
1014                         goto out;
1015                 }
1016
1017                 break;
1018         }
1019 #ifdef CONFIG_PPC_BOOK3S_64
1020         case KVM_CREATE_SPAPR_TCE: {
1021                 struct kvm_create_spapr_tce create_tce;
1022
1023                 r = -EFAULT;
1024                 if (copy_from_user(&create_tce, argp, sizeof(create_tce)))
1025                         goto out;
1026                 r = kvm_vm_ioctl_create_spapr_tce(kvm, &create_tce);
1027                 goto out;
1028         }
1029         case KVM_PPC_GET_SMMU_INFO: {
1030                 struct kvm_ppc_smmu_info info;
1031
1032                 memset(&info, 0, sizeof(info));
1033                 r = kvmppc_ops->get_smmu_info(kvm, &info);
1034                 if (r >= 0 && copy_to_user(argp, &info, sizeof(info)))
1035                         r = -EFAULT;
1036                 break;
1037         }
1038         case KVM_PPC_RTAS_DEFINE_TOKEN: {
1039                 struct kvm *kvm = filp->private_data;
1040
1041                 r = kvm_vm_ioctl_rtas_define_token(kvm, argp);
1042                 break;
1043         }
1044         default:
1045                 r = kvmppc_ops->arch_vm_ioctl(filp, ioctl, arg);
1046
1047 #else /* CONFIG_PPC_BOOK3S_64 */
1048         default:
1049                 r = -ENOTTY;
1050 #endif
1051         }
1052 out:
1053         return r;
1054 }
1055
1056 static unsigned long lpid_inuse[BITS_TO_LONGS(KVMPPC_NR_LPIDS)];
1057 static unsigned long nr_lpids;
1058
1059 long kvmppc_alloc_lpid(void)
1060 {
1061         long lpid;
1062
1063         do {
1064                 lpid = find_first_zero_bit(lpid_inuse, KVMPPC_NR_LPIDS);
1065                 if (lpid >= nr_lpids) {
1066                         pr_err("%s: No LPIDs free\n", __func__);
1067                         return -ENOMEM;
1068                 }
1069         } while (test_and_set_bit(lpid, lpid_inuse));
1070
1071         return lpid;
1072 }
1073
1074 void kvmppc_claim_lpid(long lpid)
1075 {
1076         set_bit(lpid, lpid_inuse);
1077 }
1078
1079 void kvmppc_free_lpid(long lpid)
1080 {
1081         clear_bit(lpid, lpid_inuse);
1082 }
1083
1084 void kvmppc_init_lpid(unsigned long nr_lpids_param)
1085 {
1086         nr_lpids = min_t(unsigned long, KVMPPC_NR_LPIDS, nr_lpids_param);
1087         memset(lpid_inuse, 0, sizeof(lpid_inuse));
1088 }
1089
1090 int kvm_arch_init(void *opaque)
1091 {
1092         if (kvmppc_ops) {
1093                 printk(KERN_ERR "kvm: already loaded the other module\n");
1094                 return -EEXIST;
1095         }
1096         kvmppc_ops = (struct kvmppc_ops *)opaque;
1097         return 0;
1098 }
1099
1100 void kvm_arch_exit(void)
1101 {
1102         kvmppc_ops = NULL;
1103 }