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