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

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