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