SAFE public projects git trees. - safe/jmp/linux-2.6/blob - arch/powerpc/kvm/book3s.c

   1 /*
   2  * Copyright (C) 2009. SUSE Linux Products GmbH. All rights reserved.
   3  *
   4  * Authors:
   5  *    Alexander Graf <agraf@suse.de>
   6  *    Kevin Wolf <mail@kevin-wolf.de>
   7  *
   8  * Description:
   9  * This file is derived from arch/powerpc/kvm/44x.c,
  10  * by Hollis Blanchard <hollisb@us.ibm.com>.
  11  *
  12  * This program is free software; you can redistribute it and/or modify
  13  * it under the terms of the GNU General Public License, version 2, as
  14  * published by the Free Software Foundation.
  15  */
  16
  17 #include <linux/kvm_host.h>
  18 #include <linux/err.h>
  19
  20 #include <asm/reg.h>
  21 #include <asm/cputable.h>
  22 #include <asm/cacheflush.h>
  23 #include <asm/tlbflush.h>
  24 #include <asm/uaccess.h>
  25 #include <asm/io.h>
  26 #include <asm/kvm_ppc.h>
  27 #include <asm/kvm_book3s.h>
  28 #include <asm/mmu_context.h>
  29 #include <linux/gfp.h>
  30 #include <linux/sched.h>
  31 #include <linux/vmalloc.h>
  32 #include <linux/highmem.h>
  33
  34 #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
  35
  36 /* #define EXIT_DEBUG */
  37 /* #define EXIT_DEBUG_SIMPLE */
  38 /* #define DEBUG_EXT */
  39
  40 static int kvmppc_handle_ext(struct kvm_vcpu *vcpu, unsigned int exit_nr,
  41                              ulong msr);
  42
  43 /* Some compatibility defines */
  44 #ifdef CONFIG_PPC_BOOK3S_32
  45 #define MSR_USER32 MSR_USER
  46 #define MSR_USER64 MSR_USER
  47 #define HW_PAGE_SIZE PAGE_SIZE
  48 #endif
  49
  50 struct kvm_stats_debugfs_item debugfs_entries[] = {
  51         { "exits",       VCPU_STAT(sum_exits) },
  52         { "mmio",        VCPU_STAT(mmio_exits) },
  53         { "sig",         VCPU_STAT(signal_exits) },
  54         { "sysc",        VCPU_STAT(syscall_exits) },
  55         { "inst_emu",    VCPU_STAT(emulated_inst_exits) },
  56         { "dec",         VCPU_STAT(dec_exits) },
  57         { "ext_intr",    VCPU_STAT(ext_intr_exits) },
  58         { "queue_intr",  VCPU_STAT(queue_intr) },
  59         { "halt_wakeup", VCPU_STAT(halt_wakeup) },
  60         { "pf_storage",  VCPU_STAT(pf_storage) },
  61         { "sp_storage",  VCPU_STAT(sp_storage) },
  62         { "pf_instruc",  VCPU_STAT(pf_instruc) },
  63         { "sp_instruc",  VCPU_STAT(sp_instruc) },
  64         { "ld",          VCPU_STAT(ld) },
  65         { "ld_slow",     VCPU_STAT(ld_slow) },
  66         { "st",          VCPU_STAT(st) },
  67         { "st_slow",     VCPU_STAT(st_slow) },
  68         { NULL }
  69 };
  70
  71 void kvmppc_core_load_host_debugstate(struct kvm_vcpu *vcpu)
  72 {
  73 }
  74
  75 void kvmppc_core_load_guest_debugstate(struct kvm_vcpu *vcpu)
  76 {
  77 }
  78
  79 void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
  80 {
  81 #ifdef CONFIG_PPC_BOOK3S_64
  82         memcpy(to_svcpu(vcpu)->slb, to_book3s(vcpu)->slb_shadow, sizeof(to_svcpu(vcpu)->slb));
  83         memcpy(&get_paca()->shadow_vcpu, to_book3s(vcpu)->shadow_vcpu,
  84                sizeof(get_paca()->shadow_vcpu));
  85         to_svcpu(vcpu)->slb_max = to_book3s(vcpu)->slb_shadow_max;
  86 #endif
  87
  88 #ifdef CONFIG_PPC_BOOK3S_32
  89         current->thread.kvm_shadow_vcpu = to_book3s(vcpu)->shadow_vcpu;
  90 #endif
  91 }
  92
  93 void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
  94 {
  95 #ifdef CONFIG_PPC_BOOK3S_64
  96         memcpy(to_book3s(vcpu)->slb_shadow, to_svcpu(vcpu)->slb, sizeof(to_svcpu(vcpu)->slb));
  97         memcpy(to_book3s(vcpu)->shadow_vcpu, &get_paca()->shadow_vcpu,
  98                sizeof(get_paca()->shadow_vcpu));
  99         to_book3s(vcpu)->slb_shadow_max = to_svcpu(vcpu)->slb_max;
 100 #endif
 101
 102         kvmppc_giveup_ext(vcpu, MSR_FP);
 103         kvmppc_giveup_ext(vcpu, MSR_VEC);
 104         kvmppc_giveup_ext(vcpu, MSR_VSX);
 105 }
 106
 107 #if defined(EXIT_DEBUG)
 108 static u32 kvmppc_get_dec(struct kvm_vcpu *vcpu)
 109 {
 110         u64 jd = mftb() - vcpu->arch.dec_jiffies;
 111         return vcpu->arch.dec - jd;
 112 }
 113 #endif
 114
 115 static void kvmppc_recalc_shadow_msr(struct kvm_vcpu *vcpu)
 116 {
 117         vcpu->arch.shadow_msr = vcpu->arch.msr;
 118         /* Guest MSR values */
 119         vcpu->arch.shadow_msr &= MSR_FE0 | MSR_FE1 | MSR_SF | MSR_SE |
 120                                  MSR_BE | MSR_DE;
 121         /* Process MSR values */
 122         vcpu->arch.shadow_msr |= MSR_ME | MSR_RI | MSR_IR | MSR_DR | MSR_PR |
 123                                  MSR_EE;
 124         /* External providers the guest reserved */
 125         vcpu->arch.shadow_msr |= (vcpu->arch.msr & vcpu->arch.guest_owned_ext);
 126         /* 64-bit Process MSR values */
 127 #ifdef CONFIG_PPC_BOOK3S_64
 128         vcpu->arch.shadow_msr |= MSR_ISF | MSR_HV;
 129 #endif
 130 }
 131
 132 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u64 msr)
 133 {
 134         ulong old_msr = vcpu->arch.msr;
 135
 136 #ifdef EXIT_DEBUG
 137         printk(KERN_INFO "KVM: Set MSR to 0x%llx\n", msr);
 138 #endif
 139
 140         msr &= to_book3s(vcpu)->msr_mask;
 141         vcpu->arch.msr = msr;
 142         kvmppc_recalc_shadow_msr(vcpu);
 143
 144         if (msr & (MSR_WE|MSR_POW)) {
 145                 if (!vcpu->arch.pending_exceptions) {
 146                         kvm_vcpu_block(vcpu);
 147                         vcpu->stat.halt_wakeup++;
 148                 }
 149         }
 150
 151         if (((vcpu->arch.msr & (MSR_IR|MSR_DR)) != (old_msr & (MSR_IR|MSR_DR))) ||
 152             (vcpu->arch.msr & MSR_PR) != (old_msr & MSR_PR)) {
 153                 bool dr = (vcpu->arch.msr & MSR_DR) ? true : false;
 154                 bool ir = (vcpu->arch.msr & MSR_IR) ? true : false;
 155
 156                 /* Flush split mode PTEs */
 157                 if (dr != ir)
 158                         kvmppc_mmu_pte_vflush(vcpu, VSID_SPLIT_MASK,
 159                                               VSID_SPLIT_MASK);
 160
 161                 kvmppc_mmu_flush_segments(vcpu);
 162                 kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu));
 163         }
 164
 165         /* Preload FPU if it's enabled */
 166         if (vcpu->arch.msr & MSR_FP)
 167                 kvmppc_handle_ext(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, MSR_FP);
 168 }
 169
 170 void kvmppc_inject_interrupt(struct kvm_vcpu *vcpu, int vec, u64 flags)
 171 {
 172         vcpu->arch.srr0 = kvmppc_get_pc(vcpu);
 173         vcpu->arch.srr1 = vcpu->arch.msr | flags;
 174         kvmppc_set_pc(vcpu, to_book3s(vcpu)->hior + vec);
 175         vcpu->arch.mmu.reset_msr(vcpu);
 176 }
 177
 178 static int kvmppc_book3s_vec2irqprio(unsigned int vec)
 179 {
 180         unsigned int prio;
 181
 182         switch (vec) {
 183         case 0x100: prio = BOOK3S_IRQPRIO_SYSTEM_RESET;         break;
 184         case 0x200: prio = BOOK3S_IRQPRIO_MACHINE_CHECK;        break;
 185         case 0x300: prio = BOOK3S_IRQPRIO_DATA_STORAGE;         break;
 186         case 0x380: prio = BOOK3S_IRQPRIO_DATA_SEGMENT;         break;
 187         case 0x400: prio = BOOK3S_IRQPRIO_INST_STORAGE;         break;
 188         case 0x480: prio = BOOK3S_IRQPRIO_INST_SEGMENT;         break;
 189         case 0x500: prio = BOOK3S_IRQPRIO_EXTERNAL;             break;
 190         case 0x600: prio = BOOK3S_IRQPRIO_ALIGNMENT;            break;
 191         case 0x700: prio = BOOK3S_IRQPRIO_PROGRAM;              break;
 192         case 0x800: prio = BOOK3S_IRQPRIO_FP_UNAVAIL;           break;
 193         case 0x900: prio = BOOK3S_IRQPRIO_DECREMENTER;          break;
 194         case 0xc00: prio = BOOK3S_IRQPRIO_SYSCALL;              break;
 195         case 0xd00: prio = BOOK3S_IRQPRIO_DEBUG;                break;
 196         case 0xf20: prio = BOOK3S_IRQPRIO_ALTIVEC;              break;
 197         case 0xf40: prio = BOOK3S_IRQPRIO_VSX;                  break;
 198         default:    prio = BOOK3S_IRQPRIO_MAX;                  break;
 199         }
 200
 201         return prio;
 202 }
 203
 204 static void kvmppc_book3s_dequeue_irqprio(struct kvm_vcpu *vcpu,
 205                                           unsigned int vec)
 206 {
 207         clear_bit(kvmppc_book3s_vec2irqprio(vec),
 208                   &vcpu->arch.pending_exceptions);
 209 }
 210
 211 void kvmppc_book3s_queue_irqprio(struct kvm_vcpu *vcpu, unsigned int vec)
 212 {
 213         vcpu->stat.queue_intr++;
 214
 215         set_bit(kvmppc_book3s_vec2irqprio(vec),
 216                 &vcpu->arch.pending_exceptions);
 217 #ifdef EXIT_DEBUG
 218         printk(KERN_INFO "Queueing interrupt %x\n", vec);
 219 #endif
 220 }
 221
 222
 223 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong flags)
 224 {
 225         to_book3s(vcpu)->prog_flags = flags;
 226         kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_PROGRAM);
 227 }
 228
 229 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
 230 {
 231         kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
 232 }
 233
 234 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
 235 {
 236         return test_bit(BOOK3S_INTERRUPT_DECREMENTER >> 7, &vcpu->arch.pending_exceptions);
 237 }
 238
 239 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
 240 {
 241         kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
 242 }
 243
 244 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
 245                                 struct kvm_interrupt *irq)
 246 {
 247         kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL);
 248 }
 249
 250 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu,
 251                                   struct kvm_interrupt *irq)
 252 {
 253         kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL);
 254 }
 255
 256 int kvmppc_book3s_irqprio_deliver(struct kvm_vcpu *vcpu, unsigned int priority)
 257 {
 258         int deliver = 1;
 259         int vec = 0;
 260         ulong flags = 0ULL;
 261
 262         switch (priority) {
 263         case BOOK3S_IRQPRIO_DECREMENTER:
 264                 deliver = vcpu->arch.msr & MSR_EE;
 265                 vec = BOOK3S_INTERRUPT_DECREMENTER;
 266                 break;
 267         case BOOK3S_IRQPRIO_EXTERNAL:
 268                 deliver = vcpu->arch.msr & MSR_EE;
 269                 vec = BOOK3S_INTERRUPT_EXTERNAL;
 270                 break;
 271         case BOOK3S_IRQPRIO_SYSTEM_RESET:
 272                 vec = BOOK3S_INTERRUPT_SYSTEM_RESET;
 273                 break;
 274         case BOOK3S_IRQPRIO_MACHINE_CHECK:
 275                 vec = BOOK3S_INTERRUPT_MACHINE_CHECK;
 276                 break;
 277         case BOOK3S_IRQPRIO_DATA_STORAGE:
 278                 vec = BOOK3S_INTERRUPT_DATA_STORAGE;
 279                 break;
 280         case BOOK3S_IRQPRIO_INST_STORAGE:
 281                 vec = BOOK3S_INTERRUPT_INST_STORAGE;
 282                 break;
 283         case BOOK3S_IRQPRIO_DATA_SEGMENT:
 284                 vec = BOOK3S_INTERRUPT_DATA_SEGMENT;
 285                 break;
 286         case BOOK3S_IRQPRIO_INST_SEGMENT:
 287                 vec = BOOK3S_INTERRUPT_INST_SEGMENT;
 288                 break;
 289         case BOOK3S_IRQPRIO_ALIGNMENT:
 290                 vec = BOOK3S_INTERRUPT_ALIGNMENT;
 291                 break;
 292         case BOOK3S_IRQPRIO_PROGRAM:
 293                 vec = BOOK3S_INTERRUPT_PROGRAM;
 294                 flags = to_book3s(vcpu)->prog_flags;
 295                 break;
 296         case BOOK3S_IRQPRIO_VSX:
 297                 vec = BOOK3S_INTERRUPT_VSX;
 298                 break;
 299         case BOOK3S_IRQPRIO_ALTIVEC:
 300                 vec = BOOK3S_INTERRUPT_ALTIVEC;
 301                 break;
 302         case BOOK3S_IRQPRIO_FP_UNAVAIL:
 303                 vec = BOOK3S_INTERRUPT_FP_UNAVAIL;
 304                 break;
 305         case BOOK3S_IRQPRIO_SYSCALL:
 306                 vec = BOOK3S_INTERRUPT_SYSCALL;
 307                 break;
 308         case BOOK3S_IRQPRIO_DEBUG:
 309                 vec = BOOK3S_INTERRUPT_TRACE;
 310                 break;
 311         case BOOK3S_IRQPRIO_PERFORMANCE_MONITOR:
 312                 vec = BOOK3S_INTERRUPT_PERFMON;
 313                 break;
 314         default:
 315                 deliver = 0;
 316                 printk(KERN_ERR "KVM: Unknown interrupt: 0x%x\n", priority);
 317                 break;
 318         }
 319
 320 #if 0
 321         printk(KERN_INFO "Deliver interrupt 0x%x? %x\n", vec, deliver);
 322 #endif
 323
 324         if (deliver)
 325                 kvmppc_inject_interrupt(vcpu, vec, flags);
 326
 327         return deliver;
 328 }
 329
 330 void kvmppc_core_deliver_interrupts(struct kvm_vcpu *vcpu)
 331 {
 332         unsigned long *pending = &vcpu->arch.pending_exceptions;
 333         unsigned int priority;
 334
 335 #ifdef EXIT_DEBUG
 336         if (vcpu->arch.pending_exceptions)
 337                 printk(KERN_EMERG "KVM: Check pending: %lx\n", vcpu->arch.pending_exceptions);
 338 #endif
 339         priority = __ffs(*pending);
 340         while (priority < BOOK3S_IRQPRIO_MAX) {
 341                 if (kvmppc_book3s_irqprio_deliver(vcpu, priority) &&
 342                     (priority != BOOK3S_IRQPRIO_DECREMENTER)) {
 343                         /* DEC interrupts get cleared by mtdec */
 344                         clear_bit(priority, &vcpu->arch.pending_exceptions);
 345                         break;
 346                 }
 347
 348                 priority = find_next_bit(pending,
 349                                          BITS_PER_BYTE * sizeof(*pending),
 350                                          priority + 1);
 351         }
 352 }
 353
 354 void kvmppc_set_pvr(struct kvm_vcpu *vcpu, u32 pvr)
 355 {
 356         vcpu->arch.hflags &= ~BOOK3S_HFLAG_SLB;
 357         vcpu->arch.pvr = pvr;
 358 #ifdef CONFIG_PPC_BOOK3S_64
 359         if ((pvr >= 0x330000) && (pvr < 0x70330000)) {
 360                 kvmppc_mmu_book3s_64_init(vcpu);
 361                 to_book3s(vcpu)->hior = 0xfff00000;
 362                 to_book3s(vcpu)->msr_mask = 0xffffffffffffffffULL;
 363         } else
 364 #endif
 365         {
 366                 kvmppc_mmu_book3s_32_init(vcpu);
 367                 to_book3s(vcpu)->hior = 0;
 368                 to_book3s(vcpu)->msr_mask = 0xffffffffULL;
 369         }
 370
 371         /* If we are in hypervisor level on 970, we can tell the CPU to
 372          * treat DCBZ as 32 bytes store */
 373         vcpu->arch.hflags &= ~BOOK3S_HFLAG_DCBZ32;
 374         if (vcpu->arch.mmu.is_dcbz32(vcpu) && (mfmsr() & MSR_HV) &&
 375             !strcmp(cur_cpu_spec->platform, "ppc970"))
 376                 vcpu->arch.hflags |= BOOK3S_HFLAG_DCBZ32;
 377
 378         /* Cell performs badly if MSR_FEx are set. So let's hope nobody
 379            really needs them in a VM on Cell and force disable them. */
 380         if (!strcmp(cur_cpu_spec->platform, "ppc-cell-be"))
 381                 to_book3s(vcpu)->msr_mask &= ~(MSR_FE0 | MSR_FE1);
 382
 383 #ifdef CONFIG_PPC_BOOK3S_32
 384         /* 32 bit Book3S always has 32 byte dcbz */
 385         vcpu->arch.hflags |= BOOK3S_HFLAG_DCBZ32;
 386 #endif
 387 }
 388
 389 /* Book3s_32 CPUs always have 32 bytes cache line size, which Linux assumes. To
 390  * make Book3s_32 Linux work on Book3s_64, we have to make sure we trap dcbz to
 391  * emulate 32 bytes dcbz length.
 392  *
 393  * The Book3s_64 inventors also realized this case and implemented a special bit
 394  * in the HID5 register, which is a hypervisor ressource. Thus we can't use it.
 395  *
 396  * My approach here is to patch the dcbz instruction on executing pages.
 397  */
 398 static void kvmppc_patch_dcbz(struct kvm_vcpu *vcpu, struct kvmppc_pte *pte)
 399 {
 400         struct page *hpage;
 401         u64 hpage_offset;
 402         u32 *page;
 403         int i;
 404
 405         hpage = gfn_to_page(vcpu->kvm, pte->raddr >> PAGE_SHIFT);
 406         if (is_error_page(hpage))
 407                 return;
 408
 409         hpage_offset = pte->raddr & ~PAGE_MASK;
 410         hpage_offset &= ~0xFFFULL;
 411         hpage_offset /= 4;
 412
 413         get_page(hpage);
 414         page = kmap_atomic(hpage, KM_USER0);
 415
 416         /* patch dcbz into reserved instruction, so we trap */
 417         for (i=hpage_offset; i < hpage_offset + (HW_PAGE_SIZE / 4); i++)
 418                 if ((page[i] & 0xff0007ff) == INS_DCBZ)
 419                         page[i] &= 0xfffffff7;
 420
 421         kunmap_atomic(page, KM_USER0);
 422         put_page(hpage);
 423 }
 424
 425 static int kvmppc_xlate(struct kvm_vcpu *vcpu, ulong eaddr, bool data,
 426                          struct kvmppc_pte *pte)
 427 {
 428         int relocated = (vcpu->arch.msr & (data ? MSR_DR : MSR_IR));
 429         int r;
 430
 431         if (relocated) {
 432                 r = vcpu->arch.mmu.xlate(vcpu, eaddr, pte, data);
 433         } else {
 434                 pte->eaddr = eaddr;
 435                 pte->raddr = eaddr & 0xffffffff;
 436                 pte->vpage = VSID_REAL | eaddr >> 12;
 437                 pte->may_read = true;
 438                 pte->may_write = true;
 439                 pte->may_execute = true;
 440                 r = 0;
 441         }
 442
 443         return r;
 444 }
 445
 446 static hva_t kvmppc_bad_hva(void)
 447 {
 448         return PAGE_OFFSET;
 449 }
 450
 451 static hva_t kvmppc_pte_to_hva(struct kvm_vcpu *vcpu, struct kvmppc_pte *pte,
 452                                bool read)
 453 {
 454         hva_t hpage;
 455
 456         if (read && !pte->may_read)
 457                 goto err;
 458
 459         if (!read && !pte->may_write)
 460                 goto err;
 461
 462         hpage = gfn_to_hva(vcpu->kvm, pte->raddr >> PAGE_SHIFT);
 463         if (kvm_is_error_hva(hpage))
 464                 goto err;
 465
 466         return hpage | (pte->raddr & ~PAGE_MASK);
 467 err:
 468         return kvmppc_bad_hva();
 469 }
 470
 471 int kvmppc_st(struct kvm_vcpu *vcpu, ulong *eaddr, int size, void *ptr,
 472               bool data)
 473 {
 474         struct kvmppc_pte pte;
 475
 476         vcpu->stat.st++;
 477
 478         if (kvmppc_xlate(vcpu, *eaddr, data, &pte))
 479                 return -ENOENT;
 480
 481         *eaddr = pte.raddr;
 482
 483         if (!pte.may_write)
 484                 return -EPERM;
 485
 486         if (kvm_write_guest(vcpu->kvm, pte.raddr, ptr, size))
 487                 return EMULATE_DO_MMIO;
 488
 489         return EMULATE_DONE;
 490 }
 491
 492 int kvmppc_ld(struct kvm_vcpu *vcpu, ulong *eaddr, int size, void *ptr,
 493                       bool data)
 494 {
 495         struct kvmppc_pte pte;
 496         hva_t hva = *eaddr;
 497
 498         vcpu->stat.ld++;
 499
 500         if (kvmppc_xlate(vcpu, *eaddr, data, &pte))
 501                 goto nopte;
 502
 503         *eaddr = pte.raddr;
 504
 505         hva = kvmppc_pte_to_hva(vcpu, &pte, true);
 506         if (kvm_is_error_hva(hva))
 507                 goto mmio;
 508
 509         if (copy_from_user(ptr, (void __user *)hva, size)) {
 510                 printk(KERN_INFO "kvmppc_ld at 0x%lx failed\n", hva);
 511                 goto mmio;
 512         }
 513
 514         return EMULATE_DONE;
 515
 516 nopte:
 517         return -ENOENT;
 518 mmio:
 519         return EMULATE_DO_MMIO;
 520 }
 521
 522 static int kvmppc_visible_gfn(struct kvm_vcpu *vcpu, gfn_t gfn)
 523 {
 524         return kvm_is_visible_gfn(vcpu->kvm, gfn);
 525 }
 526
 527 int kvmppc_handle_pagefault(struct kvm_run *run, struct kvm_vcpu *vcpu,
 528                             ulong eaddr, int vec)
 529 {
 530         bool data = (vec == BOOK3S_INTERRUPT_DATA_STORAGE);
 531         int r = RESUME_GUEST;
 532         int relocated;
 533         int page_found = 0;
 534         struct kvmppc_pte pte;
 535         bool is_mmio = false;
 536         bool dr = (vcpu->arch.msr & MSR_DR) ? true : false;
 537         bool ir = (vcpu->arch.msr & MSR_IR) ? true : false;
 538
 539         relocated = data ? dr : ir;
 540
 541         /* Resolve real address if translation turned on */
 542         if (relocated) {
 543                 page_found = vcpu->arch.mmu.xlate(vcpu, eaddr, &pte, data);
 544         } else {
 545                 pte.may_execute = true;
 546                 pte.may_read = true;
 547                 pte.may_write = true;
 548                 pte.raddr = eaddr & 0xffffffff;
 549                 pte.eaddr = eaddr;
 550                 pte.vpage = eaddr >> 12;
 551         }
 552
 553         switch (vcpu->arch.msr & (MSR_DR|MSR_IR)) {
 554         case 0:
 555                 pte.vpage |= VSID_REAL;
 556                 break;
 557         case MSR_DR:
 558                 pte.vpage |= VSID_REAL_DR;
 559                 break;
 560         case MSR_IR:
 561                 pte.vpage |= VSID_REAL_IR;
 562                 break;
 563         }
 564
 565         if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
 566            (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32))) {
 567                 /*
 568                  * If we do the dcbz hack, we have to NX on every execution,
 569                  * so we can patch the executing code. This renders our guest
 570                  * NX-less.
 571                  */
 572                 pte.may_execute = !data;
 573         }
 574
 575         if (page_found == -ENOENT) {
 576                 /* Page not found in guest PTE entries */
 577                 vcpu->arch.dear = kvmppc_get_fault_dar(vcpu);
 578                 to_book3s(vcpu)->dsisr = to_svcpu(vcpu)->fault_dsisr;
 579                 vcpu->arch.msr |= (to_svcpu(vcpu)->shadow_srr1 & 0x00000000f8000000ULL);
 580                 kvmppc_book3s_queue_irqprio(vcpu, vec);
 581         } else if (page_found == -EPERM) {
 582                 /* Storage protection */
 583                 vcpu->arch.dear = kvmppc_get_fault_dar(vcpu);
 584                 to_book3s(vcpu)->dsisr = to_svcpu(vcpu)->fault_dsisr & ~DSISR_NOHPTE;
 585                 to_book3s(vcpu)->dsisr |= DSISR_PROTFAULT;
 586                 vcpu->arch.msr |= (to_svcpu(vcpu)->shadow_srr1 & 0x00000000f8000000ULL);
 587                 kvmppc_book3s_queue_irqprio(vcpu, vec);
 588         } else if (page_found == -EINVAL) {
 589                 /* Page not found in guest SLB */
 590                 vcpu->arch.dear = kvmppc_get_fault_dar(vcpu);
 591                 kvmppc_book3s_queue_irqprio(vcpu, vec + 0x80);
 592         } else if (!is_mmio &&
 593                    kvmppc_visible_gfn(vcpu, pte.raddr >> PAGE_SHIFT)) {
 594                 /* The guest's PTE is not mapped yet. Map on the host */
 595                 kvmppc_mmu_map_page(vcpu, &pte);
 596                 if (data)
 597                         vcpu->stat.sp_storage++;
 598                 else if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
 599                         (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32)))
 600                         kvmppc_patch_dcbz(vcpu, &pte);
 601         } else {
 602                 /* MMIO */
 603                 vcpu->stat.mmio_exits++;
 604                 vcpu->arch.paddr_accessed = pte.raddr;
 605                 r = kvmppc_emulate_mmio(run, vcpu);
 606                 if ( r == RESUME_HOST_NV )
 607                         r = RESUME_HOST;
 608         }
 609
 610         return r;
 611 }
 612
 613 static inline int get_fpr_index(int i)
 614 {
 615 #ifdef CONFIG_VSX
 616         i *= 2;
 617 #endif
 618         return i;
 619 }
 620
 621 /* Give up external provider (FPU, Altivec, VSX) */
 622 void kvmppc_giveup_ext(struct kvm_vcpu *vcpu, ulong msr)
 623 {
 624         struct thread_struct *t = &current->thread;
 625         u64 *vcpu_fpr = vcpu->arch.fpr;
 626 #ifdef CONFIG_VSX
 627         u64 *vcpu_vsx = vcpu->arch.vsr;
 628 #endif
 629         u64 *thread_fpr = (u64*)t->fpr;
 630         int i;
 631
 632         if (!(vcpu->arch.guest_owned_ext & msr))
 633                 return;
 634
 635 #ifdef DEBUG_EXT
 636         printk(KERN_INFO "Giving up ext 0x%lx\n", msr);
 637 #endif
 638
 639         switch (msr) {
 640         case MSR_FP:
 641                 giveup_fpu(current);
 642                 for (i = 0; i < ARRAY_SIZE(vcpu->arch.fpr); i++)
 643                         vcpu_fpr[i] = thread_fpr[get_fpr_index(i)];
 644
 645                 vcpu->arch.fpscr = t->fpscr.val;
 646                 break;
 647         case MSR_VEC:
 648 #ifdef CONFIG_ALTIVEC
 649                 giveup_altivec(current);
 650                 memcpy(vcpu->arch.vr, t->vr, sizeof(vcpu->arch.vr));
 651                 vcpu->arch.vscr = t->vscr;
 652 #endif
 653                 break;
 654         case MSR_VSX:
 655 #ifdef CONFIG_VSX
 656                 __giveup_vsx(current);
 657                 for (i = 0; i < ARRAY_SIZE(vcpu->arch.vsr); i++)
 658                         vcpu_vsx[i] = thread_fpr[get_fpr_index(i) + 1];
 659 #endif
 660                 break;
 661         default:
 662                 BUG();
 663         }
 664
 665         vcpu->arch.guest_owned_ext &= ~msr;
 666         current->thread.regs->msr &= ~msr;
 667         kvmppc_recalc_shadow_msr(vcpu);
 668 }
 669
 670 static int kvmppc_read_inst(struct kvm_vcpu *vcpu)
 671 {
 672         ulong srr0 = kvmppc_get_pc(vcpu);
 673         u32 last_inst = kvmppc_get_last_inst(vcpu);
 674         int ret;
 675
 676         ret = kvmppc_ld(vcpu, &srr0, sizeof(u32), &last_inst, false);
 677         if (ret == -ENOENT) {
 678                 vcpu->arch.msr = kvmppc_set_field(vcpu->arch.msr, 33, 33, 1);
 679                 vcpu->arch.msr = kvmppc_set_field(vcpu->arch.msr, 34, 36, 0);
 680                 vcpu->arch.msr = kvmppc_set_field(vcpu->arch.msr, 42, 47, 0);
 681                 kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_INST_STORAGE);
 682                 return EMULATE_AGAIN;
 683         }
 684
 685         return EMULATE_DONE;
 686 }
 687
 688 static int kvmppc_check_ext(struct kvm_vcpu *vcpu, unsigned int exit_nr)
 689 {
 690
 691         /* Need to do paired single emulation? */
 692         if (!(vcpu->arch.hflags & BOOK3S_HFLAG_PAIRED_SINGLE))
 693                 return EMULATE_DONE;
 694
 695         /* Read out the instruction */
 696         if (kvmppc_read_inst(vcpu) == EMULATE_DONE)
 697                 /* Need to emulate */
 698                 return EMULATE_FAIL;
 699
 700         return EMULATE_AGAIN;
 701 }
 702
 703 /* Handle external providers (FPU, Altivec, VSX) */
 704 static int kvmppc_handle_ext(struct kvm_vcpu *vcpu, unsigned int exit_nr,
 705                              ulong msr)
 706 {
 707         struct thread_struct *t = &current->thread;
 708         u64 *vcpu_fpr = vcpu->arch.fpr;
 709 #ifdef CONFIG_VSX
 710         u64 *vcpu_vsx = vcpu->arch.vsr;
 711 #endif
 712         u64 *thread_fpr = (u64*)t->fpr;
 713         int i;
 714
 715         /* When we have paired singles, we emulate in software */
 716         if (vcpu->arch.hflags & BOOK3S_HFLAG_PAIRED_SINGLE)
 717                 return RESUME_GUEST;
 718
 719         if (!(vcpu->arch.msr & msr)) {
 720                 kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
 721                 return RESUME_GUEST;
 722         }
 723
 724         /* We already own the ext */
 725         if (vcpu->arch.guest_owned_ext & msr) {
 726                 return RESUME_GUEST;
 727         }
 728
 729 #ifdef DEBUG_EXT
 730         printk(KERN_INFO "Loading up ext 0x%lx\n", msr);
 731 #endif
 732
 733         current->thread.regs->msr |= msr;
 734
 735         switch (msr) {
 736         case MSR_FP:
 737                 for (i = 0; i < ARRAY_SIZE(vcpu->arch.fpr); i++)
 738                         thread_fpr[get_fpr_index(i)] = vcpu_fpr[i];
 739
 740                 t->fpscr.val = vcpu->arch.fpscr;
 741                 t->fpexc_mode = 0;
 742                 kvmppc_load_up_fpu();
 743                 break;
 744         case MSR_VEC:
 745 #ifdef CONFIG_ALTIVEC
 746                 memcpy(t->vr, vcpu->arch.vr, sizeof(vcpu->arch.vr));
 747                 t->vscr = vcpu->arch.vscr;
 748                 t->vrsave = -1;
 749                 kvmppc_load_up_altivec();
 750 #endif
 751                 break;
 752         case MSR_VSX:
 753 #ifdef CONFIG_VSX
 754                 for (i = 0; i < ARRAY_SIZE(vcpu->arch.vsr); i++)
 755                         thread_fpr[get_fpr_index(i) + 1] = vcpu_vsx[i];
 756                 kvmppc_load_up_vsx();
 757 #endif
 758                 break;
 759         default:
 760                 BUG();
 761         }
 762
 763         vcpu->arch.guest_owned_ext |= msr;
 764
 765         kvmppc_recalc_shadow_msr(vcpu);
 766
 767         return RESUME_GUEST;
 768 }
 769
 770 int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
 771                        unsigned int exit_nr)
 772 {
 773         int r = RESUME_HOST;
 774
 775         vcpu->stat.sum_exits++;
 776
 777         run->exit_reason = KVM_EXIT_UNKNOWN;
 778         run->ready_for_interrupt_injection = 1;
 779 #ifdef EXIT_DEBUG
 780         printk(KERN_EMERG "exit_nr=0x%x | pc=0x%lx | dar=0x%lx | dec=0x%x | msr=0x%lx\n",
 781                 exit_nr, kvmppc_get_pc(vcpu), kvmppc_get_fault_dar(vcpu),
 782                 kvmppc_get_dec(vcpu), to_svcpu(vcpu)->shadow_srr1);
 783 #elif defined (EXIT_DEBUG_SIMPLE)
 784         if ((exit_nr != 0x900) && (exit_nr != 0x500))
 785                 printk(KERN_EMERG "exit_nr=0x%x | pc=0x%lx | dar=0x%lx | msr=0x%lx\n",
 786                         exit_nr, kvmppc_get_pc(vcpu), kvmppc_get_fault_dar(vcpu),
 787                         vcpu->arch.msr);
 788 #endif
 789         kvm_resched(vcpu);
 790         switch (exit_nr) {
 791         case BOOK3S_INTERRUPT_INST_STORAGE:
 792                 vcpu->stat.pf_instruc++;
 793
 794 #ifdef CONFIG_PPC_BOOK3S_32
 795                 /* We set segments as unused segments when invalidating them. So
 796                  * treat the respective fault as segment fault. */
 797                 if (to_svcpu(vcpu)->sr[kvmppc_get_pc(vcpu) >> SID_SHIFT]
 798                     == SR_INVALID) {
 799                         kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu));
 800                         r = RESUME_GUEST;
 801                         break;
 802                 }
 803 #endif
 804
 805                 /* only care about PTEG not found errors, but leave NX alone */
 806                 if (to_svcpu(vcpu)->shadow_srr1 & 0x40000000) {
 807                         r = kvmppc_handle_pagefault(run, vcpu, kvmppc_get_pc(vcpu), exit_nr);
 808                         vcpu->stat.sp_instruc++;
 809                 } else if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
 810                           (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32))) {
 811                         /*
 812                          * XXX If we do the dcbz hack we use the NX bit to flush&patch the page,
 813                          *     so we can't use the NX bit inside the guest. Let's cross our fingers,
 814                          *     that no guest that needs the dcbz hack does NX.
 815                          */
 816                         kvmppc_mmu_pte_flush(vcpu, kvmppc_get_pc(vcpu), ~0xFFFUL);
 817                         r = RESUME_GUEST;
 818                 } else {
 819                         vcpu->arch.msr |= to_svcpu(vcpu)->shadow_srr1 & 0x58000000;
 820                         kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
 821                         kvmppc_mmu_pte_flush(vcpu, kvmppc_get_pc(vcpu), ~0xFFFUL);
 822                         r = RESUME_GUEST;
 823                 }
 824                 break;
 825         case BOOK3S_INTERRUPT_DATA_STORAGE:
 826         {
 827                 ulong dar = kvmppc_get_fault_dar(vcpu);
 828                 vcpu->stat.pf_storage++;
 829
 830 #ifdef CONFIG_PPC_BOOK3S_32
 831                 /* We set segments as unused segments when invalidating them. So
 832                  * treat the respective fault as segment fault. */
 833                 if ((to_svcpu(vcpu)->sr[dar >> SID_SHIFT]) == SR_INVALID) {
 834                         kvmppc_mmu_map_segment(vcpu, dar);
 835                         r = RESUME_GUEST;
 836                         break;
 837                 }
 838 #endif
 839
 840                 /* The only case we need to handle is missing shadow PTEs */
 841                 if (to_svcpu(vcpu)->fault_dsisr & DSISR_NOHPTE) {
 842                         r = kvmppc_handle_pagefault(run, vcpu, dar, exit_nr);
 843                 } else {
 844                         vcpu->arch.dear = dar;
 845                         to_book3s(vcpu)->dsisr = to_svcpu(vcpu)->fault_dsisr;
 846                         kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
 847                         kvmppc_mmu_pte_flush(vcpu, vcpu->arch.dear, ~0xFFFUL);
 848                         r = RESUME_GUEST;
 849                 }
 850                 break;
 851         }
 852         case BOOK3S_INTERRUPT_DATA_SEGMENT:
 853                 if (kvmppc_mmu_map_segment(vcpu, kvmppc_get_fault_dar(vcpu)) < 0) {
 854                         vcpu->arch.dear = kvmppc_get_fault_dar(vcpu);
 855                         kvmppc_book3s_queue_irqprio(vcpu,
 856                                 BOOK3S_INTERRUPT_DATA_SEGMENT);
 857                 }
 858                 r = RESUME_GUEST;
 859                 break;
 860         case BOOK3S_INTERRUPT_INST_SEGMENT:
 861                 if (kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu)) < 0) {
 862                         kvmppc_book3s_queue_irqprio(vcpu,
 863                                 BOOK3S_INTERRUPT_INST_SEGMENT);
 864                 }
 865                 r = RESUME_GUEST;
 866                 break;
 867         /* We're good on these - the host merely wanted to get our attention */
 868         case BOOK3S_INTERRUPT_DECREMENTER:
 869                 vcpu->stat.dec_exits++;
 870                 r = RESUME_GUEST;
 871                 break;
 872         case BOOK3S_INTERRUPT_EXTERNAL:
 873                 vcpu->stat.ext_intr_exits++;
 874                 r = RESUME_GUEST;
 875                 break;
 876         case BOOK3S_INTERRUPT_PROGRAM:
 877         {
 878                 enum emulation_result er;
 879                 ulong flags;
 880
 881 program_interrupt:
 882                 flags = to_svcpu(vcpu)->shadow_srr1 & 0x1f0000ull;
 883
 884                 if (vcpu->arch.msr & MSR_PR) {
 885 #ifdef EXIT_DEBUG
 886                         printk(KERN_INFO "Userspace triggered 0x700 exception at 0x%lx (0x%x)\n", kvmppc_get_pc(vcpu), kvmppc_get_last_inst(vcpu));
 887 #endif
 888                         if ((kvmppc_get_last_inst(vcpu) & 0xff0007ff) !=
 889                             (INS_DCBZ & 0xfffffff7)) {
 890                                 kvmppc_core_queue_program(vcpu, flags);
 891                                 r = RESUME_GUEST;
 892                                 break;
 893                         }
 894                 }
 895
 896                 vcpu->stat.emulated_inst_exits++;
 897                 er = kvmppc_emulate_instruction(run, vcpu);
 898                 switch (er) {
 899                 case EMULATE_DONE:
 900                         r = RESUME_GUEST_NV;
 901                         break;
 902                 case EMULATE_AGAIN:
 903                         r = RESUME_GUEST;
 904                         break;
 905                 case EMULATE_FAIL:
 906                         printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
 907                                __func__, kvmppc_get_pc(vcpu), kvmppc_get_last_inst(vcpu));
 908                         kvmppc_core_queue_program(vcpu, flags);
 909                         r = RESUME_GUEST;
 910                         break;
 911                 case EMULATE_DO_MMIO:
 912                         run->exit_reason = KVM_EXIT_MMIO;
 913                         r = RESUME_HOST_NV;
 914                         break;
 915                 default:
 916                         BUG();
 917                 }
 918                 break;
 919         }
 920         case BOOK3S_INTERRUPT_SYSCALL:
 921                 // XXX make user settable
 922                 if (vcpu->arch.osi_enabled &&
 923                     (((u32)kvmppc_get_gpr(vcpu, 3)) == OSI_SC_MAGIC_R3) &&
 924                     (((u32)kvmppc_get_gpr(vcpu, 4)) == OSI_SC_MAGIC_R4)) {
 925                         u64 *gprs = run->osi.gprs;
 926                         int i;
 927
 928                         run->exit_reason = KVM_EXIT_OSI;
 929                         for (i = 0; i < 32; i++)
 930                                 gprs[i] = kvmppc_get_gpr(vcpu, i);
 931                         vcpu->arch.osi_needed = 1;
 932                         r = RESUME_HOST_NV;
 933
 934                 } else {
 935                         vcpu->stat.syscall_exits++;
 936                         kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
 937                         r = RESUME_GUEST;
 938                 }
 939                 break;
 940         case BOOK3S_INTERRUPT_FP_UNAVAIL:
 941         case BOOK3S_INTERRUPT_ALTIVEC:
 942         case BOOK3S_INTERRUPT_VSX:
 943         {
 944                 int ext_msr = 0;
 945
 946                 switch (exit_nr) {
 947                 case BOOK3S_INTERRUPT_FP_UNAVAIL: ext_msr = MSR_FP;  break;
 948                 case BOOK3S_INTERRUPT_ALTIVEC:    ext_msr = MSR_VEC; break;
 949                 case BOOK3S_INTERRUPT_VSX:        ext_msr = MSR_VSX; break;
 950                 }
 951
 952                 switch (kvmppc_check_ext(vcpu, exit_nr)) {
 953                 case EMULATE_DONE:
 954                         /* everything ok - let's enable the ext */
 955                         r = kvmppc_handle_ext(vcpu, exit_nr, ext_msr);
 956                         break;
 957                 case EMULATE_FAIL:
 958                         /* we need to emulate this instruction */
 959                         goto program_interrupt;
 960                         break;
 961                 default:
 962                         /* nothing to worry about - go again */
 963                         break;
 964                 }
 965                 break;
 966         }
 967         case BOOK3S_INTERRUPT_ALIGNMENT:
 968                 if (kvmppc_read_inst(vcpu) == EMULATE_DONE) {
 969                         to_book3s(vcpu)->dsisr = kvmppc_alignment_dsisr(vcpu,
 970                                 kvmppc_get_last_inst(vcpu));
 971                         vcpu->arch.dear = kvmppc_alignment_dar(vcpu,
 972                                 kvmppc_get_last_inst(vcpu));
 973                         kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
 974                 }
 975                 r = RESUME_GUEST;
 976                 break;
 977         case BOOK3S_INTERRUPT_MACHINE_CHECK:
 978         case BOOK3S_INTERRUPT_TRACE:
 979                 kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
 980                 r = RESUME_GUEST;
 981                 break;
 982         default:
 983                 /* Ugh - bork here! What did we get? */
 984                 printk(KERN_EMERG "exit_nr=0x%x | pc=0x%lx | msr=0x%lx\n",
 985                         exit_nr, kvmppc_get_pc(vcpu), to_svcpu(vcpu)->shadow_srr1);
 986                 r = RESUME_HOST;
 987                 BUG();
 988                 break;
 989         }
 990
 991
 992         if (!(r & RESUME_HOST)) {
 993                 /* To avoid clobbering exit_reason, only check for signals if
 994                  * we aren't already exiting to userspace for some other
 995                  * reason. */
 996                 if (signal_pending(current)) {
 997 #ifdef EXIT_DEBUG
 998                         printk(KERN_EMERG "KVM: Going back to host\n");
 999 #endif
1000                         vcpu->stat.signal_exits++;
1001                         run->exit_reason = KVM_EXIT_INTR;
1002                         r = -EINTR;
1003                 } else {
1004                         /* In case an interrupt came in that was triggered
1005                          * from userspace (like DEC), we need to check what
1006                          * to inject now! */
1007                         kvmppc_core_deliver_interrupts(vcpu);
1008                 }
1009         }
1010
1011 #ifdef EXIT_DEBUG
1012         printk(KERN_EMERG "KVM exit: vcpu=0x%p pc=0x%lx r=0x%x\n", vcpu, kvmppc_get_pc(vcpu), r);
1013 #endif
1014
1015         return r;
1016 }
1017
1018 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
1019 {
1020         return 0;
1021 }
1022
1023 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1024 {
1025         int i;
1026
1027         vcpu_load(vcpu);
1028
1029         regs->pc = kvmppc_get_pc(vcpu);
1030         regs->cr = kvmppc_get_cr(vcpu);
1031         regs->ctr = kvmppc_get_ctr(vcpu);
1032         regs->lr = kvmppc_get_lr(vcpu);
1033         regs->xer = kvmppc_get_xer(vcpu);
1034         regs->msr = vcpu->arch.msr;
1035         regs->srr0 = vcpu->arch.srr0;
1036         regs->srr1 = vcpu->arch.srr1;
1037         regs->pid = vcpu->arch.pid;
1038         regs->sprg0 = vcpu->arch.sprg0;
1039         regs->sprg1 = vcpu->arch.sprg1;
1040         regs->sprg2 = vcpu->arch.sprg2;
1041         regs->sprg3 = vcpu->arch.sprg3;
1042         regs->sprg5 = vcpu->arch.sprg4;
1043         regs->sprg6 = vcpu->arch.sprg5;
1044         regs->sprg7 = vcpu->arch.sprg6;
1045
1046         for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1047                 regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
1048
1049         vcpu_put(vcpu);
1050
1051         return 0;
1052 }
1053
1054 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1055 {
1056         int i;
1057
1058         vcpu_load(vcpu);
1059
1060         kvmppc_set_pc(vcpu, regs->pc);
1061         kvmppc_set_cr(vcpu, regs->cr);
1062         kvmppc_set_ctr(vcpu, regs->ctr);
1063         kvmppc_set_lr(vcpu, regs->lr);
1064         kvmppc_set_xer(vcpu, regs->xer);
1065         kvmppc_set_msr(vcpu, regs->msr);
1066         vcpu->arch.srr0 = regs->srr0;
1067         vcpu->arch.srr1 = regs->srr1;
1068         vcpu->arch.sprg0 = regs->sprg0;
1069         vcpu->arch.sprg1 = regs->sprg1;
1070         vcpu->arch.sprg2 = regs->sprg2;
1071         vcpu->arch.sprg3 = regs->sprg3;
1072         vcpu->arch.sprg5 = regs->sprg4;
1073         vcpu->arch.sprg6 = regs->sprg5;
1074         vcpu->arch.sprg7 = regs->sprg6;
1075
1076         for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1077                 kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
1078
1079         vcpu_put(vcpu);
1080
1081         return 0;
1082 }
1083
1084 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1085                                   struct kvm_sregs *sregs)
1086 {
1087         struct kvmppc_vcpu_book3s *vcpu3s = to_book3s(vcpu);
1088         int i;
1089
1090         sregs->pvr = vcpu->arch.pvr;
1091
1092         sregs->u.s.sdr1 = to_book3s(vcpu)->sdr1;
1093         if (vcpu->arch.hflags & BOOK3S_HFLAG_SLB) {
1094                 for (i = 0; i < 64; i++) {
1095                         sregs->u.s.ppc64.slb[i].slbe = vcpu3s->slb[i].orige | i;
1096                         sregs->u.s.ppc64.slb[i].slbv = vcpu3s->slb[i].origv;
1097                 }
1098         } else {
1099                 for (i = 0; i < 16; i++) {
1100                         sregs->u.s.ppc32.sr[i] = vcpu3s->sr[i].raw;
1101                         sregs->u.s.ppc32.sr[i] = vcpu3s->sr[i].raw;
1102                 }
1103                 for (i = 0; i < 8; i++) {
1104                         sregs->u.s.ppc32.ibat[i] = vcpu3s->ibat[i].raw;
1105                         sregs->u.s.ppc32.dbat[i] = vcpu3s->dbat[i].raw;
1106                 }
1107         }
1108         return 0;
1109 }
1110
1111 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1112                                   struct kvm_sregs *sregs)
1113 {
1114         struct kvmppc_vcpu_book3s *vcpu3s = to_book3s(vcpu);
1115         int i;
1116
1117         kvmppc_set_pvr(vcpu, sregs->pvr);
1118
1119         vcpu3s->sdr1 = sregs->u.s.sdr1;
1120         if (vcpu->arch.hflags & BOOK3S_HFLAG_SLB) {
1121                 for (i = 0; i < 64; i++) {
1122                         vcpu->arch.mmu.slbmte(vcpu, sregs->u.s.ppc64.slb[i].slbv,
1123                                                     sregs->u.s.ppc64.slb[i].slbe);
1124                 }
1125         } else {
1126                 for (i = 0; i < 16; i++) {
1127                         vcpu->arch.mmu.mtsrin(vcpu, i, sregs->u.s.ppc32.sr[i]);
1128                 }
1129                 for (i = 0; i < 8; i++) {
1130                         kvmppc_set_bat(vcpu, &(vcpu3s->ibat[i]), false,
1131                                        (u32)sregs->u.s.ppc32.ibat[i]);
1132                         kvmppc_set_bat(vcpu, &(vcpu3s->ibat[i]), true,
1133                                        (u32)(sregs->u.s.ppc32.ibat[i] >> 32));
1134                         kvmppc_set_bat(vcpu, &(vcpu3s->dbat[i]), false,
1135                                        (u32)sregs->u.s.ppc32.dbat[i]);
1136                         kvmppc_set_bat(vcpu, &(vcpu3s->dbat[i]), true,
1137                                        (u32)(sregs->u.s.ppc32.dbat[i] >> 32));
1138                 }
1139         }
1140
1141         /* Flush the MMU after messing with the segments */
1142         kvmppc_mmu_pte_flush(vcpu, 0, 0);
1143         return 0;
1144 }
1145
1146 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1147 {
1148         return -ENOTSUPP;
1149 }
1150
1151 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1152 {
1153         return -ENOTSUPP;
1154 }
1155
1156 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1157                                   struct kvm_translation *tr)
1158 {
1159         return 0;
1160 }
1161
1162 /*
1163  * Get (and clear) the dirty memory log for a memory slot.
1164  */
1165 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
1166                                       struct kvm_dirty_log *log)
1167 {
1168         struct kvm_memory_slot *memslot;
1169         struct kvm_vcpu *vcpu;
1170         ulong ga, ga_end;
1171         int is_dirty = 0;
1172         int r;
1173         unsigned long n;
1174
1175         mutex_lock(&kvm->slots_lock);
1176
1177         r = kvm_get_dirty_log(kvm, log, &is_dirty);
1178         if (r)
1179                 goto out;
1180
1181         /* If nothing is dirty, don't bother messing with page tables. */
1182         if (is_dirty) {
1183                 memslot = &kvm->memslots->memslots[log->slot];
1184
1185                 ga = memslot->base_gfn << PAGE_SHIFT;
1186                 ga_end = ga + (memslot->npages << PAGE_SHIFT);
1187
1188                 kvm_for_each_vcpu(n, vcpu, kvm)
1189                         kvmppc_mmu_pte_pflush(vcpu, ga, ga_end);
1190
1191                 n = kvm_dirty_bitmap_bytes(memslot);
1192                 memset(memslot->dirty_bitmap, 0, n);
1193         }
1194
1195         r = 0;
1196 out:
1197         mutex_unlock(&kvm->slots_lock);
1198         return r;
1199 }
1200
1201 int kvmppc_core_check_processor_compat(void)
1202 {
1203         return 0;
1204 }
1205
1206 struct kvm_vcpu *kvmppc_core_vcpu_create(struct kvm *kvm, unsigned int id)
1207 {
1208         struct kvmppc_vcpu_book3s *vcpu_book3s;
1209         struct kvm_vcpu *vcpu;
1210         int err = -ENOMEM;
1211
1212         vcpu_book3s = vmalloc(sizeof(struct kvmppc_vcpu_book3s));
1213         if (!vcpu_book3s)
1214                 goto out;
1215
1216         memset(vcpu_book3s, 0, sizeof(struct kvmppc_vcpu_book3s));
1217
1218         vcpu_book3s->shadow_vcpu = (struct kvmppc_book3s_shadow_vcpu *)
1219                 kzalloc(sizeof(*vcpu_book3s->shadow_vcpu), GFP_KERNEL);
1220         if (!vcpu_book3s->shadow_vcpu)
1221                 goto free_vcpu;
1222
1223         vcpu = &vcpu_book3s->vcpu;
1224         err = kvm_vcpu_init(vcpu, kvm, id);
1225         if (err)
1226                 goto free_shadow_vcpu;
1227
1228         vcpu->arch.host_retip = kvm_return_point;
1229         vcpu->arch.host_msr = mfmsr();
1230 #ifdef CONFIG_PPC_BOOK3S_64
1231         /* default to book3s_64 (970fx) */
1232         vcpu->arch.pvr = 0x3C0301;
1233 #else
1234         /* default to book3s_32 (750) */
1235         vcpu->arch.pvr = 0x84202;
1236 #endif
1237         kvmppc_set_pvr(vcpu, vcpu->arch.pvr);
1238         vcpu_book3s->slb_nr = 64;
1239
1240         /* remember where some real-mode handlers are */
1241         vcpu->arch.trampoline_lowmem = kvmppc_trampoline_lowmem;
1242         vcpu->arch.trampoline_enter = kvmppc_trampoline_enter;
1243         vcpu->arch.highmem_handler = (ulong)kvmppc_handler_highmem;
1244 #ifdef CONFIG_PPC_BOOK3S_64
1245         vcpu->arch.rmcall = *(ulong*)kvmppc_rmcall;
1246 #else
1247         vcpu->arch.rmcall = (ulong)kvmppc_rmcall;
1248 #endif
1249
1250         vcpu->arch.shadow_msr = MSR_USER64;
1251
1252         err = kvmppc_mmu_init(vcpu);
1253         if (err < 0)
1254                 goto free_shadow_vcpu;
1255
1256         return vcpu;
1257
1258 free_shadow_vcpu:
1259         kfree(vcpu_book3s->shadow_vcpu);
1260 free_vcpu:
1261         vfree(vcpu_book3s);
1262 out:
1263         return ERR_PTR(err);
1264 }
1265
1266 void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
1267 {
1268         struct kvmppc_vcpu_book3s *vcpu_book3s = to_book3s(vcpu);
1269
1270         kvm_vcpu_uninit(vcpu);
1271         kfree(vcpu_book3s->shadow_vcpu);
1272         vfree(vcpu_book3s);
1273 }
1274
1275 extern int __kvmppc_vcpu_entry(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu);
1276 int __kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
1277 {
1278         int ret;
1279         struct thread_struct ext_bkp;
1280 #ifdef CONFIG_ALTIVEC
1281         bool save_vec = current->thread.used_vr;
1282 #endif
1283 #ifdef CONFIG_VSX
1284         bool save_vsx = current->thread.used_vsr;
1285 #endif
1286         ulong ext_msr;
1287
1288         /* No need to go into the guest when all we do is going out */
1289         if (signal_pending(current)) {
1290                 kvm_run->exit_reason = KVM_EXIT_INTR;
1291                 return -EINTR;
1292         }
1293
1294         /* Save FPU state in stack */
1295         if (current->thread.regs->msr & MSR_FP)
1296                 giveup_fpu(current);
1297         memcpy(ext_bkp.fpr, current->thread.fpr, sizeof(current->thread.fpr));
1298         ext_bkp.fpscr = current->thread.fpscr;
1299         ext_bkp.fpexc_mode = current->thread.fpexc_mode;
1300
1301 #ifdef CONFIG_ALTIVEC
1302         /* Save Altivec state in stack */
1303         if (save_vec) {
1304                 if (current->thread.regs->msr & MSR_VEC)
1305                         giveup_altivec(current);
1306                 memcpy(ext_bkp.vr, current->thread.vr, sizeof(ext_bkp.vr));
1307                 ext_bkp.vscr = current->thread.vscr;
1308                 ext_bkp.vrsave = current->thread.vrsave;
1309         }
1310         ext_bkp.used_vr = current->thread.used_vr;
1311 #endif
1312
1313 #ifdef CONFIG_VSX
1314         /* Save VSX state in stack */
1315         if (save_vsx && (current->thread.regs->msr & MSR_VSX))
1316                         __giveup_vsx(current);
1317         ext_bkp.used_vsr = current->thread.used_vsr;
1318 #endif
1319
1320         /* Remember the MSR with disabled extensions */
1321         ext_msr = current->thread.regs->msr;
1322
1323         /* XXX we get called with irq disabled - change that! */
1324         local_irq_enable();
1325
1326         /* Preload FPU if it's enabled */
1327         if (vcpu->arch.msr & MSR_FP)
1328                 kvmppc_handle_ext(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, MSR_FP);
1329
1330         ret = __kvmppc_vcpu_entry(kvm_run, vcpu);
1331
1332         local_irq_disable();
1333
1334         current->thread.regs->msr = ext_msr;
1335
1336         /* Make sure we save the guest FPU/Altivec/VSX state */
1337         kvmppc_giveup_ext(vcpu, MSR_FP);
1338         kvmppc_giveup_ext(vcpu, MSR_VEC);
1339         kvmppc_giveup_ext(vcpu, MSR_VSX);
1340
1341         /* Restore FPU state from stack */
1342         memcpy(current->thread.fpr, ext_bkp.fpr, sizeof(ext_bkp.fpr));
1343         current->thread.fpscr = ext_bkp.fpscr;
1344         current->thread.fpexc_mode = ext_bkp.fpexc_mode;
1345
1346 #ifdef CONFIG_ALTIVEC
1347         /* Restore Altivec state from stack */
1348         if (save_vec && current->thread.used_vr) {
1349                 memcpy(current->thread.vr, ext_bkp.vr, sizeof(ext_bkp.vr));
1350                 current->thread.vscr = ext_bkp.vscr;
1351                 current->thread.vrsave= ext_bkp.vrsave;
1352         }
1353         current->thread.used_vr = ext_bkp.used_vr;
1354 #endif
1355
1356 #ifdef CONFIG_VSX
1357         current->thread.used_vsr = ext_bkp.used_vsr;
1358 #endif
1359
1360         return ret;
1361 }
1362
1363 static int kvmppc_book3s_init(void)
1364 {
1365         return kvm_init(NULL, sizeof(struct kvmppc_vcpu_book3s), THIS_MODULE);
1366 }
1367
1368 static void kvmppc_book3s_exit(void)
1369 {
1370         kvm_exit();
1371 }
1372
1373 module_init(kvmppc_book3s_init);
1374 module_exit(kvmppc_book3s_exit);