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

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