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