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