Pileus Git - ~andy/linux/blob - arch/powerpc/kvm/booke.c

   1 /*
   2  * This program is free software; you can redistribute it and/or modify
   3  * it under the terms of the GNU General Public License, version 2, as
   4  * published by the Free Software Foundation.
   5  *
   6  * This program is distributed in the hope that it will be useful,
   7  * but WITHOUT ANY WARRANTY; without even the implied warranty of
   8  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
   9  * GNU General Public License for more details.
  10  *
  11  * You should have received a copy of the GNU General Public License
  12  * along with this program; if not, write to the Free Software
  13  * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
  14  *
  15  * Copyright IBM Corp. 2007
  16  * Copyright 2010-2011 Freescale Semiconductor, Inc.
  17  *
  18  * Authors: Hollis Blanchard <hollisb@us.ibm.com>
  19  *          Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
  20  *          Scott Wood <scottwood@freescale.com>
  21  *          Varun Sethi <varun.sethi@freescale.com>
  22  */
  23
  24 #include <linux/errno.h>
  25 #include <linux/err.h>
  26 #include <linux/kvm_host.h>
  27 #include <linux/gfp.h>
  28 #include <linux/module.h>
  29 #include <linux/vmalloc.h>
  30 #include <linux/fs.h>
  31
  32 #include <asm/cputable.h>
  33 #include <asm/uaccess.h>
  34 #include <asm/kvm_ppc.h>
  35 #include <asm/cacheflush.h>
  36 #include <asm/dbell.h>
  37 #include <asm/hw_irq.h>
  38 #include <asm/irq.h>
  39 #include <asm/time.h>
  40
  41 #include "timing.h"
  42 #include "booke.h"
  43 #include "trace.h"
  44
  45 unsigned long kvmppc_booke_handlers;
  46
  47 #define VM_STAT(x) offsetof(struct kvm, stat.x), KVM_STAT_VM
  48 #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
  49
  50 struct kvm_stats_debugfs_item debugfs_entries[] = {
  51         { "mmio",       VCPU_STAT(mmio_exits) },
  52         { "dcr",        VCPU_STAT(dcr_exits) },
  53         { "sig",        VCPU_STAT(signal_exits) },
  54         { "itlb_r",     VCPU_STAT(itlb_real_miss_exits) },
  55         { "itlb_v",     VCPU_STAT(itlb_virt_miss_exits) },
  56         { "dtlb_r",     VCPU_STAT(dtlb_real_miss_exits) },
  57         { "dtlb_v",     VCPU_STAT(dtlb_virt_miss_exits) },
  58         { "sysc",       VCPU_STAT(syscall_exits) },
  59         { "isi",        VCPU_STAT(isi_exits) },
  60         { "dsi",        VCPU_STAT(dsi_exits) },
  61         { "inst_emu",   VCPU_STAT(emulated_inst_exits) },
  62         { "dec",        VCPU_STAT(dec_exits) },
  63         { "ext_intr",   VCPU_STAT(ext_intr_exits) },
  64         { "halt_wakeup", VCPU_STAT(halt_wakeup) },
  65         { "doorbell", VCPU_STAT(dbell_exits) },
  66         { "guest doorbell", VCPU_STAT(gdbell_exits) },
  67         { "remote_tlb_flush", VM_STAT(remote_tlb_flush) },
  68         { NULL }
  69 };
  70
  71 /* TODO: use vcpu_printf() */
  72 void kvmppc_dump_vcpu(struct kvm_vcpu *vcpu)
  73 {
  74         int i;
  75
  76         printk("pc:   %08lx msr:  %08llx\n", vcpu->arch.pc, vcpu->arch.shared->msr);
  77         printk("lr:   %08lx ctr:  %08lx\n", vcpu->arch.lr, vcpu->arch.ctr);
  78         printk("srr0: %08llx srr1: %08llx\n", vcpu->arch.shared->srr0,
  79                                             vcpu->arch.shared->srr1);
  80
  81         printk("exceptions: %08lx\n", vcpu->arch.pending_exceptions);
  82
  83         for (i = 0; i < 32; i += 4) {
  84                 printk("gpr%02d: %08lx %08lx %08lx %08lx\n", i,
  85                        kvmppc_get_gpr(vcpu, i),
  86                        kvmppc_get_gpr(vcpu, i+1),
  87                        kvmppc_get_gpr(vcpu, i+2),
  88                        kvmppc_get_gpr(vcpu, i+3));
  89         }
  90 }
  91
  92 #ifdef CONFIG_SPE
  93 void kvmppc_vcpu_disable_spe(struct kvm_vcpu *vcpu)
  94 {
  95         preempt_disable();
  96         enable_kernel_spe();
  97         kvmppc_save_guest_spe(vcpu);
  98         vcpu->arch.shadow_msr &= ~MSR_SPE;
  99         preempt_enable();
 100 }
 101
 102 static void kvmppc_vcpu_enable_spe(struct kvm_vcpu *vcpu)
 103 {
 104         preempt_disable();
 105         enable_kernel_spe();
 106         kvmppc_load_guest_spe(vcpu);
 107         vcpu->arch.shadow_msr |= MSR_SPE;
 108         preempt_enable();
 109 }
 110
 111 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
 112 {
 113         if (vcpu->arch.shared->msr & MSR_SPE) {
 114                 if (!(vcpu->arch.shadow_msr & MSR_SPE))
 115                         kvmppc_vcpu_enable_spe(vcpu);
 116         } else if (vcpu->arch.shadow_msr & MSR_SPE) {
 117                 kvmppc_vcpu_disable_spe(vcpu);
 118         }
 119 }
 120 #else
 121 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
 122 {
 123 }
 124 #endif
 125
 126 static void kvmppc_vcpu_sync_fpu(struct kvm_vcpu *vcpu)
 127 {
 128 #if defined(CONFIG_PPC_FPU) && !defined(CONFIG_KVM_BOOKE_HV)
 129         /* We always treat the FP bit as enabled from the host
 130            perspective, so only need to adjust the shadow MSR */
 131         vcpu->arch.shadow_msr &= ~MSR_FP;
 132         vcpu->arch.shadow_msr |= vcpu->arch.shared->msr & MSR_FP;
 133 #endif
 134 }
 135
 136 /*
 137  * Helper function for "full" MSR writes.  No need to call this if only
 138  * EE/CE/ME/DE/RI are changing.
 139  */
 140 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr)
 141 {
 142         u32 old_msr = vcpu->arch.shared->msr;
 143
 144 #ifdef CONFIG_KVM_BOOKE_HV
 145         new_msr |= MSR_GS;
 146 #endif
 147
 148         vcpu->arch.shared->msr = new_msr;
 149
 150         kvmppc_mmu_msr_notify(vcpu, old_msr);
 151         kvmppc_vcpu_sync_spe(vcpu);
 152         kvmppc_vcpu_sync_fpu(vcpu);
 153 }
 154
 155 static void kvmppc_booke_queue_irqprio(struct kvm_vcpu *vcpu,
 156                                        unsigned int priority)
 157 {
 158         trace_kvm_booke_queue_irqprio(vcpu, priority);
 159         set_bit(priority, &vcpu->arch.pending_exceptions);
 160 }
 161
 162 static void kvmppc_core_queue_dtlb_miss(struct kvm_vcpu *vcpu,
 163                                         ulong dear_flags, ulong esr_flags)
 164 {
 165         vcpu->arch.queued_dear = dear_flags;
 166         vcpu->arch.queued_esr = esr_flags;
 167         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DTLB_MISS);
 168 }
 169
 170 static void kvmppc_core_queue_data_storage(struct kvm_vcpu *vcpu,
 171                                            ulong dear_flags, ulong esr_flags)
 172 {
 173         vcpu->arch.queued_dear = dear_flags;
 174         vcpu->arch.queued_esr = esr_flags;
 175         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DATA_STORAGE);
 176 }
 177
 178 static void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu,
 179                                            ulong esr_flags)
 180 {
 181         vcpu->arch.queued_esr = esr_flags;
 182         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_INST_STORAGE);
 183 }
 184
 185 static void kvmppc_core_queue_alignment(struct kvm_vcpu *vcpu, ulong dear_flags,
 186                                         ulong esr_flags)
 187 {
 188         vcpu->arch.queued_dear = dear_flags;
 189         vcpu->arch.queued_esr = esr_flags;
 190         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALIGNMENT);
 191 }
 192
 193 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong esr_flags)
 194 {
 195         vcpu->arch.queued_esr = esr_flags;
 196         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_PROGRAM);
 197 }
 198
 199 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
 200 {
 201         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DECREMENTER);
 202 }
 203
 204 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
 205 {
 206         return test_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
 207 }
 208
 209 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
 210 {
 211         clear_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
 212 }
 213
 214 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
 215                                 struct kvm_interrupt *irq)
 216 {
 217         unsigned int prio = BOOKE_IRQPRIO_EXTERNAL;
 218
 219         if (irq->irq == KVM_INTERRUPT_SET_LEVEL)
 220                 prio = BOOKE_IRQPRIO_EXTERNAL_LEVEL;
 221
 222         kvmppc_booke_queue_irqprio(vcpu, prio);
 223 }
 224
 225 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu)
 226 {
 227         clear_bit(BOOKE_IRQPRIO_EXTERNAL, &vcpu->arch.pending_exceptions);
 228         clear_bit(BOOKE_IRQPRIO_EXTERNAL_LEVEL, &vcpu->arch.pending_exceptions);
 229 }
 230
 231 static void kvmppc_core_queue_watchdog(struct kvm_vcpu *vcpu)
 232 {
 233         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_WATCHDOG);
 234 }
 235
 236 static void kvmppc_core_dequeue_watchdog(struct kvm_vcpu *vcpu)
 237 {
 238         clear_bit(BOOKE_IRQPRIO_WATCHDOG, &vcpu->arch.pending_exceptions);
 239 }
 240
 241 static void set_guest_srr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
 242 {
 243 #ifdef CONFIG_KVM_BOOKE_HV
 244         mtspr(SPRN_GSRR0, srr0);
 245         mtspr(SPRN_GSRR1, srr1);
 246 #else
 247         vcpu->arch.shared->srr0 = srr0;
 248         vcpu->arch.shared->srr1 = srr1;
 249 #endif
 250 }
 251
 252 static void set_guest_csrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
 253 {
 254         vcpu->arch.csrr0 = srr0;
 255         vcpu->arch.csrr1 = srr1;
 256 }
 257
 258 static void set_guest_dsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
 259 {
 260         if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC)) {
 261                 vcpu->arch.dsrr0 = srr0;
 262                 vcpu->arch.dsrr1 = srr1;
 263         } else {
 264                 set_guest_csrr(vcpu, srr0, srr1);
 265         }
 266 }
 267
 268 static void set_guest_mcsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
 269 {
 270         vcpu->arch.mcsrr0 = srr0;
 271         vcpu->arch.mcsrr1 = srr1;
 272 }
 273
 274 static unsigned long get_guest_dear(struct kvm_vcpu *vcpu)
 275 {
 276 #ifdef CONFIG_KVM_BOOKE_HV
 277         return mfspr(SPRN_GDEAR);
 278 #else
 279         return vcpu->arch.shared->dar;
 280 #endif
 281 }
 282
 283 static void set_guest_dear(struct kvm_vcpu *vcpu, unsigned long dear)
 284 {
 285 #ifdef CONFIG_KVM_BOOKE_HV
 286         mtspr(SPRN_GDEAR, dear);
 287 #else
 288         vcpu->arch.shared->dar = dear;
 289 #endif
 290 }
 291
 292 static unsigned long get_guest_esr(struct kvm_vcpu *vcpu)
 293 {
 294 #ifdef CONFIG_KVM_BOOKE_HV
 295         return mfspr(SPRN_GESR);
 296 #else
 297         return vcpu->arch.shared->esr;
 298 #endif
 299 }
 300
 301 static void set_guest_esr(struct kvm_vcpu *vcpu, u32 esr)
 302 {
 303 #ifdef CONFIG_KVM_BOOKE_HV
 304         mtspr(SPRN_GESR, esr);
 305 #else
 306         vcpu->arch.shared->esr = esr;
 307 #endif
 308 }
 309
 310 static unsigned long get_guest_epr(struct kvm_vcpu *vcpu)
 311 {
 312 #ifdef CONFIG_KVM_BOOKE_HV
 313         return mfspr(SPRN_GEPR);
 314 #else
 315         return vcpu->arch.epr;
 316 #endif
 317 }
 318
 319 /* Deliver the interrupt of the corresponding priority, if possible. */
 320 static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu *vcpu,
 321                                         unsigned int priority)
 322 {
 323         int allowed = 0;
 324         ulong msr_mask = 0;
 325         bool update_esr = false, update_dear = false, update_epr = false;
 326         ulong crit_raw = vcpu->arch.shared->critical;
 327         ulong crit_r1 = kvmppc_get_gpr(vcpu, 1);
 328         bool crit;
 329         bool keep_irq = false;
 330         enum int_class int_class;
 331         ulong new_msr = vcpu->arch.shared->msr;
 332
 333         /* Truncate crit indicators in 32 bit mode */
 334         if (!(vcpu->arch.shared->msr & MSR_SF)) {
 335                 crit_raw &= 0xffffffff;
 336                 crit_r1 &= 0xffffffff;
 337         }
 338
 339         /* Critical section when crit == r1 */
 340         crit = (crit_raw == crit_r1);
 341         /* ... and we're in supervisor mode */
 342         crit = crit && !(vcpu->arch.shared->msr & MSR_PR);
 343
 344         if (priority == BOOKE_IRQPRIO_EXTERNAL_LEVEL) {
 345                 priority = BOOKE_IRQPRIO_EXTERNAL;
 346                 keep_irq = true;
 347         }
 348
 349         if ((priority == BOOKE_IRQPRIO_EXTERNAL) && vcpu->arch.epr_enabled)
 350                 update_epr = true;
 351
 352         switch (priority) {
 353         case BOOKE_IRQPRIO_DTLB_MISS:
 354         case BOOKE_IRQPRIO_DATA_STORAGE:
 355         case BOOKE_IRQPRIO_ALIGNMENT:
 356                 update_dear = true;
 357                 /* fall through */
 358         case BOOKE_IRQPRIO_INST_STORAGE:
 359         case BOOKE_IRQPRIO_PROGRAM:
 360                 update_esr = true;
 361                 /* fall through */
 362         case BOOKE_IRQPRIO_ITLB_MISS:
 363         case BOOKE_IRQPRIO_SYSCALL:
 364         case BOOKE_IRQPRIO_FP_UNAVAIL:
 365         case BOOKE_IRQPRIO_SPE_UNAVAIL:
 366         case BOOKE_IRQPRIO_SPE_FP_DATA:
 367         case BOOKE_IRQPRIO_SPE_FP_ROUND:
 368         case BOOKE_IRQPRIO_AP_UNAVAIL:
 369                 allowed = 1;
 370                 msr_mask = MSR_CE | MSR_ME | MSR_DE;
 371                 int_class = INT_CLASS_NONCRIT;
 372                 break;
 373         case BOOKE_IRQPRIO_WATCHDOG:
 374         case BOOKE_IRQPRIO_CRITICAL:
 375         case BOOKE_IRQPRIO_DBELL_CRIT:
 376                 allowed = vcpu->arch.shared->msr & MSR_CE;
 377                 allowed = allowed && !crit;
 378                 msr_mask = MSR_ME;
 379                 int_class = INT_CLASS_CRIT;
 380                 break;
 381         case BOOKE_IRQPRIO_MACHINE_CHECK:
 382                 allowed = vcpu->arch.shared->msr & MSR_ME;
 383                 allowed = allowed && !crit;
 384                 int_class = INT_CLASS_MC;
 385                 break;
 386         case BOOKE_IRQPRIO_DECREMENTER:
 387         case BOOKE_IRQPRIO_FIT:
 388                 keep_irq = true;
 389                 /* fall through */
 390         case BOOKE_IRQPRIO_EXTERNAL:
 391         case BOOKE_IRQPRIO_DBELL:
 392                 allowed = vcpu->arch.shared->msr & MSR_EE;
 393                 allowed = allowed && !crit;
 394                 msr_mask = MSR_CE | MSR_ME | MSR_DE;
 395                 int_class = INT_CLASS_NONCRIT;
 396                 break;
 397         case BOOKE_IRQPRIO_DEBUG:
 398                 allowed = vcpu->arch.shared->msr & MSR_DE;
 399                 allowed = allowed && !crit;
 400                 msr_mask = MSR_ME;
 401                 int_class = INT_CLASS_CRIT;
 402                 break;
 403         }
 404
 405         if (allowed) {
 406                 switch (int_class) {
 407                 case INT_CLASS_NONCRIT:
 408                         set_guest_srr(vcpu, vcpu->arch.pc,
 409                                       vcpu->arch.shared->msr);
 410                         break;
 411                 case INT_CLASS_CRIT:
 412                         set_guest_csrr(vcpu, vcpu->arch.pc,
 413                                        vcpu->arch.shared->msr);
 414                         break;
 415                 case INT_CLASS_DBG:
 416                         set_guest_dsrr(vcpu, vcpu->arch.pc,
 417                                        vcpu->arch.shared->msr);
 418                         break;
 419                 case INT_CLASS_MC:
 420                         set_guest_mcsrr(vcpu, vcpu->arch.pc,
 421                                         vcpu->arch.shared->msr);
 422                         break;
 423                 }
 424
 425                 vcpu->arch.pc = vcpu->arch.ivpr | vcpu->arch.ivor[priority];
 426                 if (update_esr == true)
 427                         set_guest_esr(vcpu, vcpu->arch.queued_esr);
 428                 if (update_dear == true)
 429                         set_guest_dear(vcpu, vcpu->arch.queued_dear);
 430                 if (update_epr == true)
 431                         kvm_make_request(KVM_REQ_EPR_EXIT, vcpu);
 432
 433                 new_msr &= msr_mask;
 434 #if defined(CONFIG_64BIT)
 435                 if (vcpu->arch.epcr & SPRN_EPCR_ICM)
 436                         new_msr |= MSR_CM;
 437 #endif
 438                 kvmppc_set_msr(vcpu, new_msr);
 439
 440                 if (!keep_irq)
 441                         clear_bit(priority, &vcpu->arch.pending_exceptions);
 442         }
 443
 444 #ifdef CONFIG_KVM_BOOKE_HV
 445         /*
 446          * If an interrupt is pending but masked, raise a guest doorbell
 447          * so that we are notified when the guest enables the relevant
 448          * MSR bit.
 449          */
 450         if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_EE)
 451                 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_NONCRIT);
 452         if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_CE)
 453                 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_CRIT);
 454         if (vcpu->arch.pending_exceptions & BOOKE_IRQPRIO_MACHINE_CHECK)
 455                 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_MC);
 456 #endif
 457
 458         return allowed;
 459 }
 460
 461 /*
 462  * Return the number of jiffies until the next timeout.  If the timeout is
 463  * longer than the NEXT_TIMER_MAX_DELTA, then return NEXT_TIMER_MAX_DELTA
 464  * because the larger value can break the timer APIs.
 465  */
 466 static unsigned long watchdog_next_timeout(struct kvm_vcpu *vcpu)
 467 {
 468         u64 tb, wdt_tb, wdt_ticks = 0;
 469         u64 nr_jiffies = 0;
 470         u32 period = TCR_GET_WP(vcpu->arch.tcr);
 471
 472         wdt_tb = 1ULL << (63 - period);
 473         tb = get_tb();
 474         /*
 475          * The watchdog timeout will hapeen when TB bit corresponding
 476          * to watchdog will toggle from 0 to 1.
 477          */
 478         if (tb & wdt_tb)
 479                 wdt_ticks = wdt_tb;
 480
 481         wdt_ticks += wdt_tb - (tb & (wdt_tb - 1));
 482
 483         /* Convert timebase ticks to jiffies */
 484         nr_jiffies = wdt_ticks;
 485
 486         if (do_div(nr_jiffies, tb_ticks_per_jiffy))
 487                 nr_jiffies++;
 488
 489         return min_t(unsigned long long, nr_jiffies, NEXT_TIMER_MAX_DELTA);
 490 }
 491
 492 static void arm_next_watchdog(struct kvm_vcpu *vcpu)
 493 {
 494         unsigned long nr_jiffies;
 495         unsigned long flags;
 496
 497         /*
 498          * If TSR_ENW and TSR_WIS are not set then no need to exit to
 499          * userspace, so clear the KVM_REQ_WATCHDOG request.
 500          */
 501         if ((vcpu->arch.tsr & (TSR_ENW | TSR_WIS)) != (TSR_ENW | TSR_WIS))
 502                 clear_bit(KVM_REQ_WATCHDOG, &vcpu->requests);
 503
 504         spin_lock_irqsave(&vcpu->arch.wdt_lock, flags);
 505         nr_jiffies = watchdog_next_timeout(vcpu);
 506         /*
 507          * If the number of jiffies of watchdog timer >= NEXT_TIMER_MAX_DELTA
 508          * then do not run the watchdog timer as this can break timer APIs.
 509          */
 510         if (nr_jiffies < NEXT_TIMER_MAX_DELTA)
 511                 mod_timer(&vcpu->arch.wdt_timer, jiffies + nr_jiffies);
 512         else
 513                 del_timer(&vcpu->arch.wdt_timer);
 514         spin_unlock_irqrestore(&vcpu->arch.wdt_lock, flags);
 515 }
 516
 517 void kvmppc_watchdog_func(unsigned long data)
 518 {
 519         struct kvm_vcpu *vcpu = (struct kvm_vcpu *)data;
 520         u32 tsr, new_tsr;
 521         int final;
 522
 523         do {
 524                 new_tsr = tsr = vcpu->arch.tsr;
 525                 final = 0;
 526
 527                 /* Time out event */
 528                 if (tsr & TSR_ENW) {
 529                         if (tsr & TSR_WIS)
 530                                 final = 1;
 531                         else
 532                                 new_tsr = tsr | TSR_WIS;
 533                 } else {
 534                         new_tsr = tsr | TSR_ENW;
 535                 }
 536         } while (cmpxchg(&vcpu->arch.tsr, tsr, new_tsr) != tsr);
 537
 538         if (new_tsr & TSR_WIS) {
 539                 smp_wmb();
 540                 kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
 541                 kvm_vcpu_kick(vcpu);
 542         }
 543
 544         /*
 545          * If this is final watchdog expiry and some action is required
 546          * then exit to userspace.
 547          */
 548         if (final && (vcpu->arch.tcr & TCR_WRC_MASK) &&
 549             vcpu->arch.watchdog_enabled) {
 550                 smp_wmb();
 551                 kvm_make_request(KVM_REQ_WATCHDOG, vcpu);
 552                 kvm_vcpu_kick(vcpu);
 553         }
 554
 555         /*
 556          * Stop running the watchdog timer after final expiration to
 557          * prevent the host from being flooded with timers if the
 558          * guest sets a short period.
 559          * Timers will resume when TSR/TCR is updated next time.
 560          */
 561         if (!final)
 562                 arm_next_watchdog(vcpu);
 563 }
 564
 565 static void update_timer_ints(struct kvm_vcpu *vcpu)
 566 {
 567         if ((vcpu->arch.tcr & TCR_DIE) && (vcpu->arch.tsr & TSR_DIS))
 568                 kvmppc_core_queue_dec(vcpu);
 569         else
 570                 kvmppc_core_dequeue_dec(vcpu);
 571
 572         if ((vcpu->arch.tcr & TCR_WIE) && (vcpu->arch.tsr & TSR_WIS))
 573                 kvmppc_core_queue_watchdog(vcpu);
 574         else
 575                 kvmppc_core_dequeue_watchdog(vcpu);
 576 }
 577
 578 static void kvmppc_core_check_exceptions(struct kvm_vcpu *vcpu)
 579 {
 580         unsigned long *pending = &vcpu->arch.pending_exceptions;
 581         unsigned int priority;
 582
 583         priority = __ffs(*pending);
 584         while (priority < BOOKE_IRQPRIO_MAX) {
 585                 if (kvmppc_booke_irqprio_deliver(vcpu, priority))
 586                         break;
 587
 588                 priority = find_next_bit(pending,
 589                                          BITS_PER_BYTE * sizeof(*pending),
 590                                          priority + 1);
 591         }
 592
 593         /* Tell the guest about our interrupt status */
 594         vcpu->arch.shared->int_pending = !!*pending;
 595 }
 596
 597 /* Check pending exceptions and deliver one, if possible. */
 598 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
 599 {
 600         int r = 0;
 601         WARN_ON_ONCE(!irqs_disabled());
 602
 603         kvmppc_core_check_exceptions(vcpu);
 604
 605         if (vcpu->requests) {
 606                 /* Exception delivery raised request; start over */
 607                 return 1;
 608         }
 609
 610         if (vcpu->arch.shared->msr & MSR_WE) {
 611                 local_irq_enable();
 612                 kvm_vcpu_block(vcpu);
 613                 clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
 614                 local_irq_disable();
 615
 616                 kvmppc_set_exit_type(vcpu, EMULATED_MTMSRWE_EXITS);
 617                 r = 1;
 618         };
 619
 620         return r;
 621 }
 622
 623 int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
 624 {
 625         int r = 1; /* Indicate we want to get back into the guest */
 626
 627         if (kvm_check_request(KVM_REQ_PENDING_TIMER, vcpu))
 628                 update_timer_ints(vcpu);
 629 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
 630         if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu))
 631                 kvmppc_core_flush_tlb(vcpu);
 632 #endif
 633
 634         if (kvm_check_request(KVM_REQ_WATCHDOG, vcpu)) {
 635                 vcpu->run->exit_reason = KVM_EXIT_WATCHDOG;
 636                 r = 0;
 637         }
 638
 639         if (kvm_check_request(KVM_REQ_EPR_EXIT, vcpu)) {
 640                 vcpu->run->epr.epr = 0;
 641                 vcpu->arch.epr_needed = true;
 642                 vcpu->run->exit_reason = KVM_EXIT_EPR;
 643                 r = 0;
 644         }
 645
 646         return r;
 647 }
 648
 649 int kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
 650 {
 651         int ret, s;
 652 #ifdef CONFIG_PPC_FPU
 653         unsigned int fpscr;
 654         int fpexc_mode;
 655         u64 fpr[32];
 656 #endif
 657
 658         if (!vcpu->arch.sane) {
 659                 kvm_run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
 660                 return -EINVAL;
 661         }
 662
 663         local_irq_disable();
 664         s = kvmppc_prepare_to_enter(vcpu);
 665         if (s <= 0) {
 666                 local_irq_enable();
 667                 ret = s;
 668                 goto out;
 669         }
 670         kvmppc_lazy_ee_enable();
 671
 672         kvm_guest_enter();
 673
 674 #ifdef CONFIG_PPC_FPU
 675         /* Save userspace FPU state in stack */
 676         enable_kernel_fp();
 677         memcpy(fpr, current->thread.fpr, sizeof(current->thread.fpr));
 678         fpscr = current->thread.fpscr.val;
 679         fpexc_mode = current->thread.fpexc_mode;
 680
 681         /* Restore guest FPU state to thread */
 682         memcpy(current->thread.fpr, vcpu->arch.fpr, sizeof(vcpu->arch.fpr));
 683         current->thread.fpscr.val = vcpu->arch.fpscr;
 684
 685         /*
 686          * Since we can't trap on MSR_FP in GS-mode, we consider the guest
 687          * as always using the FPU.  Kernel usage of FP (via
 688          * enable_kernel_fp()) in this thread must not occur while
 689          * vcpu->fpu_active is set.
 690          */
 691         vcpu->fpu_active = 1;
 692
 693         kvmppc_load_guest_fp(vcpu);
 694 #endif
 695
 696         ret = __kvmppc_vcpu_run(kvm_run, vcpu);
 697
 698         /* No need for kvm_guest_exit. It's done in handle_exit.
 699            We also get here with interrupts enabled. */
 700
 701 #ifdef CONFIG_PPC_FPU
 702         kvmppc_save_guest_fp(vcpu);
 703
 704         vcpu->fpu_active = 0;
 705
 706         /* Save guest FPU state from thread */
 707         memcpy(vcpu->arch.fpr, current->thread.fpr, sizeof(vcpu->arch.fpr));
 708         vcpu->arch.fpscr = current->thread.fpscr.val;
 709
 710         /* Restore userspace FPU state from stack */
 711         memcpy(current->thread.fpr, fpr, sizeof(current->thread.fpr));
 712         current->thread.fpscr.val = fpscr;
 713         current->thread.fpexc_mode = fpexc_mode;
 714 #endif
 715
 716 out:
 717         vcpu->mode = OUTSIDE_GUEST_MODE;
 718         return ret;
 719 }
 720
 721 static int emulation_exit(struct kvm_run *run, struct kvm_vcpu *vcpu)
 722 {
 723         enum emulation_result er;
 724
 725         er = kvmppc_emulate_instruction(run, vcpu);
 726         switch (er) {
 727         case EMULATE_DONE:
 728                 /* don't overwrite subtypes, just account kvm_stats */
 729                 kvmppc_account_exit_stat(vcpu, EMULATED_INST_EXITS);
 730                 /* Future optimization: only reload non-volatiles if
 731                  * they were actually modified by emulation. */
 732                 return RESUME_GUEST_NV;
 733
 734         case EMULATE_DO_DCR:
 735                 run->exit_reason = KVM_EXIT_DCR;
 736                 return RESUME_HOST;
 737
 738         case EMULATE_FAIL:
 739                 printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
 740                        __func__, vcpu->arch.pc, vcpu->arch.last_inst);
 741                 /* For debugging, encode the failing instruction and
 742                  * report it to userspace. */
 743                 run->hw.hardware_exit_reason = ~0ULL << 32;
 744                 run->hw.hardware_exit_reason |= vcpu->arch.last_inst;
 745                 kvmppc_core_queue_program(vcpu, ESR_PIL);
 746                 return RESUME_HOST;
 747
 748         case EMULATE_EXIT_USER:
 749                 return RESUME_HOST;
 750
 751         default:
 752                 BUG();
 753         }
 754 }
 755
 756 static void kvmppc_fill_pt_regs(struct pt_regs *regs)
 757 {
 758         ulong r1, ip, msr, lr;
 759
 760         asm("mr %0, 1" : "=r"(r1));
 761         asm("mflr %0" : "=r"(lr));
 762         asm("mfmsr %0" : "=r"(msr));
 763         asm("bl 1f; 1: mflr %0" : "=r"(ip));
 764
 765         memset(regs, 0, sizeof(*regs));
 766         regs->gpr[1] = r1;
 767         regs->nip = ip;
 768         regs->msr = msr;
 769         regs->link = lr;
 770 }
 771
 772 /*
 773  * For interrupts needed to be handled by host interrupt handlers,
 774  * corresponding host handler are called from here in similar way
 775  * (but not exact) as they are called from low level handler
 776  * (such as from arch/powerpc/kernel/head_fsl_booke.S).
 777  */
 778 static void kvmppc_restart_interrupt(struct kvm_vcpu *vcpu,
 779                                      unsigned int exit_nr)
 780 {
 781         struct pt_regs regs;
 782
 783         switch (exit_nr) {
 784         case BOOKE_INTERRUPT_EXTERNAL:
 785                 kvmppc_fill_pt_regs(&regs);
 786                 do_IRQ(&regs);
 787                 break;
 788         case BOOKE_INTERRUPT_DECREMENTER:
 789                 kvmppc_fill_pt_regs(&regs);
 790                 timer_interrupt(&regs);
 791                 break;
 792 #if defined(CONFIG_PPC_FSL_BOOK3E) || defined(CONFIG_PPC_BOOK3E_64)
 793         case BOOKE_INTERRUPT_DOORBELL:
 794                 kvmppc_fill_pt_regs(&regs);
 795                 doorbell_exception(&regs);
 796                 break;
 797 #endif
 798         case BOOKE_INTERRUPT_MACHINE_CHECK:
 799                 /* FIXME */
 800                 break;
 801         case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
 802                 kvmppc_fill_pt_regs(&regs);
 803                 performance_monitor_exception(&regs);
 804                 break;
 805         case BOOKE_INTERRUPT_WATCHDOG:
 806                 kvmppc_fill_pt_regs(&regs);
 807 #ifdef CONFIG_BOOKE_WDT
 808                 WatchdogException(&regs);
 809 #else
 810                 unknown_exception(&regs);
 811 #endif
 812                 break;
 813         case BOOKE_INTERRUPT_CRITICAL:
 814                 unknown_exception(&regs);
 815                 break;
 816         }
 817 }
 818
 819 /**
 820  * kvmppc_handle_exit
 821  *
 822  * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
 823  */
 824 int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
 825                        unsigned int exit_nr)
 826 {
 827         int r = RESUME_HOST;
 828         int s;
 829
 830         /* update before a new last_exit_type is rewritten */
 831         kvmppc_update_timing_stats(vcpu);
 832
 833         /* restart interrupts if they were meant for the host */
 834         kvmppc_restart_interrupt(vcpu, exit_nr);
 835
 836         local_irq_enable();
 837
 838         trace_kvm_exit(exit_nr, vcpu);
 839         kvm_guest_exit();
 840
 841         run->exit_reason = KVM_EXIT_UNKNOWN;
 842         run->ready_for_interrupt_injection = 1;
 843
 844         switch (exit_nr) {
 845         case BOOKE_INTERRUPT_MACHINE_CHECK:
 846                 printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR));
 847                 kvmppc_dump_vcpu(vcpu);
 848                 /* For debugging, send invalid exit reason to user space */
 849                 run->hw.hardware_exit_reason = ~1ULL << 32;
 850                 run->hw.hardware_exit_reason |= mfspr(SPRN_MCSR);
 851                 r = RESUME_HOST;
 852                 break;
 853
 854         case BOOKE_INTERRUPT_EXTERNAL:
 855                 kvmppc_account_exit(vcpu, EXT_INTR_EXITS);
 856                 r = RESUME_GUEST;
 857                 break;
 858
 859         case BOOKE_INTERRUPT_DECREMENTER:
 860                 kvmppc_account_exit(vcpu, DEC_EXITS);
 861                 r = RESUME_GUEST;
 862                 break;
 863
 864         case BOOKE_INTERRUPT_WATCHDOG:
 865                 r = RESUME_GUEST;
 866                 break;
 867
 868         case BOOKE_INTERRUPT_DOORBELL:
 869                 kvmppc_account_exit(vcpu, DBELL_EXITS);
 870                 r = RESUME_GUEST;
 871                 break;
 872
 873         case BOOKE_INTERRUPT_GUEST_DBELL_CRIT:
 874                 kvmppc_account_exit(vcpu, GDBELL_EXITS);
 875
 876                 /*
 877                  * We are here because there is a pending guest interrupt
 878                  * which could not be delivered as MSR_CE or MSR_ME was not
 879                  * set.  Once we break from here we will retry delivery.
 880                  */
 881                 r = RESUME_GUEST;
 882                 break;
 883
 884         case BOOKE_INTERRUPT_GUEST_DBELL:
 885                 kvmppc_account_exit(vcpu, GDBELL_EXITS);
 886
 887                 /*
 888                  * We are here because there is a pending guest interrupt
 889                  * which could not be delivered as MSR_EE was not set.  Once
 890                  * we break from here we will retry delivery.
 891                  */
 892                 r = RESUME_GUEST;
 893                 break;
 894
 895         case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
 896                 r = RESUME_GUEST;
 897                 break;
 898
 899         case BOOKE_INTERRUPT_HV_PRIV:
 900                 r = emulation_exit(run, vcpu);
 901                 break;
 902
 903         case BOOKE_INTERRUPT_PROGRAM:
 904                 if (vcpu->arch.shared->msr & (MSR_PR | MSR_GS)) {
 905                         /*
 906                          * Program traps generated by user-level software must
 907                          * be handled by the guest kernel.
 908                          *
 909                          * In GS mode, hypervisor privileged instructions trap
 910                          * on BOOKE_INTERRUPT_HV_PRIV, not here, so these are
 911                          * actual program interrupts, handled by the guest.
 912                          */
 913                         kvmppc_core_queue_program(vcpu, vcpu->arch.fault_esr);
 914                         r = RESUME_GUEST;
 915                         kvmppc_account_exit(vcpu, USR_PR_INST);
 916                         break;
 917                 }
 918
 919                 r = emulation_exit(run, vcpu);
 920                 break;
 921
 922         case BOOKE_INTERRUPT_FP_UNAVAIL:
 923                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
 924                 kvmppc_account_exit(vcpu, FP_UNAVAIL);
 925                 r = RESUME_GUEST;
 926                 break;
 927
 928 #ifdef CONFIG_SPE
 929         case BOOKE_INTERRUPT_SPE_UNAVAIL: {
 930                 if (vcpu->arch.shared->msr & MSR_SPE)
 931                         kvmppc_vcpu_enable_spe(vcpu);
 932                 else
 933                         kvmppc_booke_queue_irqprio(vcpu,
 934                                                    BOOKE_IRQPRIO_SPE_UNAVAIL);
 935                 r = RESUME_GUEST;
 936                 break;
 937         }
 938
 939         case BOOKE_INTERRUPT_SPE_FP_DATA:
 940                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_DATA);
 941                 r = RESUME_GUEST;
 942                 break;
 943
 944         case BOOKE_INTERRUPT_SPE_FP_ROUND:
 945                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_ROUND);
 946                 r = RESUME_GUEST;
 947                 break;
 948 #else
 949         case BOOKE_INTERRUPT_SPE_UNAVAIL:
 950                 /*
 951                  * Guest wants SPE, but host kernel doesn't support it.  Send
 952                  * an "unimplemented operation" program check to the guest.
 953                  */
 954                 kvmppc_core_queue_program(vcpu, ESR_PUO | ESR_SPV);
 955                 r = RESUME_GUEST;
 956                 break;
 957
 958         /*
 959          * These really should never happen without CONFIG_SPE,
 960          * as we should never enable the real MSR[SPE] in the guest.
 961          */
 962         case BOOKE_INTERRUPT_SPE_FP_DATA:
 963         case BOOKE_INTERRUPT_SPE_FP_ROUND:
 964                 printk(KERN_CRIT "%s: unexpected SPE interrupt %u at %08lx\n",
 965                        __func__, exit_nr, vcpu->arch.pc);
 966                 run->hw.hardware_exit_reason = exit_nr;
 967                 r = RESUME_HOST;
 968                 break;
 969 #endif
 970
 971         case BOOKE_INTERRUPT_DATA_STORAGE:
 972                 kvmppc_core_queue_data_storage(vcpu, vcpu->arch.fault_dear,
 973                                                vcpu->arch.fault_esr);
 974                 kvmppc_account_exit(vcpu, DSI_EXITS);
 975                 r = RESUME_GUEST;
 976                 break;
 977
 978         case BOOKE_INTERRUPT_INST_STORAGE:
 979                 kvmppc_core_queue_inst_storage(vcpu, vcpu->arch.fault_esr);
 980                 kvmppc_account_exit(vcpu, ISI_EXITS);
 981                 r = RESUME_GUEST;
 982                 break;
 983
 984         case BOOKE_INTERRUPT_ALIGNMENT:
 985                 kvmppc_core_queue_alignment(vcpu, vcpu->arch.fault_dear,
 986                                             vcpu->arch.fault_esr);
 987                 r = RESUME_GUEST;
 988                 break;
 989
 990 #ifdef CONFIG_KVM_BOOKE_HV
 991         case BOOKE_INTERRUPT_HV_SYSCALL:
 992                 if (!(vcpu->arch.shared->msr & MSR_PR)) {
 993                         kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
 994                 } else {
 995                         /*
 996                          * hcall from guest userspace -- send privileged
 997                          * instruction program check.
 998                          */
 999                         kvmppc_core_queue_program(vcpu, ESR_PPR);
1000                 }
1001
1002                 r = RESUME_GUEST;
1003                 break;
1004 #else
1005         case BOOKE_INTERRUPT_SYSCALL:
1006                 if (!(vcpu->arch.shared->msr & MSR_PR) &&
1007                     (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
1008                         /* KVM PV hypercalls */
1009                         kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1010                         r = RESUME_GUEST;
1011                 } else {
1012                         /* Guest syscalls */
1013                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SYSCALL);
1014                 }
1015                 kvmppc_account_exit(vcpu, SYSCALL_EXITS);
1016                 r = RESUME_GUEST;
1017                 break;
1018 #endif
1019
1020         case BOOKE_INTERRUPT_DTLB_MISS: {
1021                 unsigned long eaddr = vcpu->arch.fault_dear;
1022                 int gtlb_index;
1023                 gpa_t gpaddr;
1024                 gfn_t gfn;
1025
1026 #ifdef CONFIG_KVM_E500V2
1027                 if (!(vcpu->arch.shared->msr & MSR_PR) &&
1028                     (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
1029                         kvmppc_map_magic(vcpu);
1030                         kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1031                         r = RESUME_GUEST;
1032
1033                         break;
1034                 }
1035 #endif
1036
1037                 /* Check the guest TLB. */
1038                 gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
1039                 if (gtlb_index < 0) {
1040                         /* The guest didn't have a mapping for it. */
1041                         kvmppc_core_queue_dtlb_miss(vcpu,
1042                                                     vcpu->arch.fault_dear,
1043                                                     vcpu->arch.fault_esr);
1044                         kvmppc_mmu_dtlb_miss(vcpu);
1045                         kvmppc_account_exit(vcpu, DTLB_REAL_MISS_EXITS);
1046                         r = RESUME_GUEST;
1047                         break;
1048                 }
1049
1050                 gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1051                 gfn = gpaddr >> PAGE_SHIFT;
1052
1053                 if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1054                         /* The guest TLB had a mapping, but the shadow TLB
1055                          * didn't, and it is RAM. This could be because:
1056                          * a) the entry is mapping the host kernel, or
1057                          * b) the guest used a large mapping which we're faking
1058                          * Either way, we need to satisfy the fault without
1059                          * invoking the guest. */
1060                         kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1061                         kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1062                         r = RESUME_GUEST;
1063                 } else {
1064                         /* Guest has mapped and accessed a page which is not
1065                          * actually RAM. */
1066                         vcpu->arch.paddr_accessed = gpaddr;
1067                         vcpu->arch.vaddr_accessed = eaddr;
1068                         r = kvmppc_emulate_mmio(run, vcpu);
1069                         kvmppc_account_exit(vcpu, MMIO_EXITS);
1070                 }
1071
1072                 break;
1073         }
1074
1075         case BOOKE_INTERRUPT_ITLB_MISS: {
1076                 unsigned long eaddr = vcpu->arch.pc;
1077                 gpa_t gpaddr;
1078                 gfn_t gfn;
1079                 int gtlb_index;
1080
1081                 r = RESUME_GUEST;
1082
1083                 /* Check the guest TLB. */
1084                 gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
1085                 if (gtlb_index < 0) {
1086                         /* The guest didn't have a mapping for it. */
1087                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
1088                         kvmppc_mmu_itlb_miss(vcpu);
1089                         kvmppc_account_exit(vcpu, ITLB_REAL_MISS_EXITS);
1090                         break;
1091                 }
1092
1093                 kvmppc_account_exit(vcpu, ITLB_VIRT_MISS_EXITS);
1094
1095                 gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1096                 gfn = gpaddr >> PAGE_SHIFT;
1097
1098                 if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1099                         /* The guest TLB had a mapping, but the shadow TLB
1100                          * didn't. This could be because:
1101                          * a) the entry is mapping the host kernel, or
1102                          * b) the guest used a large mapping which we're faking
1103                          * Either way, we need to satisfy the fault without
1104                          * invoking the guest. */
1105                         kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1106                 } else {
1107                         /* Guest mapped and leaped at non-RAM! */
1108                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_MACHINE_CHECK);
1109                 }
1110
1111                 break;
1112         }
1113
1114         case BOOKE_INTERRUPT_DEBUG: {
1115                 u32 dbsr;
1116
1117                 vcpu->arch.pc = mfspr(SPRN_CSRR0);
1118
1119                 /* clear IAC events in DBSR register */
1120                 dbsr = mfspr(SPRN_DBSR);
1121                 dbsr &= DBSR_IAC1 | DBSR_IAC2 | DBSR_IAC3 | DBSR_IAC4;
1122                 mtspr(SPRN_DBSR, dbsr);
1123
1124                 run->exit_reason = KVM_EXIT_DEBUG;
1125                 kvmppc_account_exit(vcpu, DEBUG_EXITS);
1126                 r = RESUME_HOST;
1127                 break;
1128         }
1129
1130         default:
1131                 printk(KERN_EMERG "exit_nr %d\n", exit_nr);
1132                 BUG();
1133         }
1134
1135         /*
1136          * To avoid clobbering exit_reason, only check for signals if we
1137          * aren't already exiting to userspace for some other reason.
1138          */
1139         if (!(r & RESUME_HOST)) {
1140                 local_irq_disable();
1141                 s = kvmppc_prepare_to_enter(vcpu);
1142                 if (s <= 0) {
1143                         local_irq_enable();
1144                         r = (s << 2) | RESUME_HOST | (r & RESUME_FLAG_NV);
1145                 } else {
1146                         kvmppc_lazy_ee_enable();
1147                 }
1148         }
1149
1150         return r;
1151 }
1152
1153 static void kvmppc_set_tsr(struct kvm_vcpu *vcpu, u32 new_tsr)
1154 {
1155         u32 old_tsr = vcpu->arch.tsr;
1156
1157         vcpu->arch.tsr = new_tsr;
1158
1159         if ((old_tsr ^ vcpu->arch.tsr) & (TSR_ENW | TSR_WIS))
1160                 arm_next_watchdog(vcpu);
1161
1162         update_timer_ints(vcpu);
1163 }
1164
1165 /* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
1166 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
1167 {
1168         int i;
1169         int r;
1170
1171         vcpu->arch.pc = 0;
1172         vcpu->arch.shared->pir = vcpu->vcpu_id;
1173         kvmppc_set_gpr(vcpu, 1, (16<<20) - 8); /* -8 for the callee-save LR slot */
1174         kvmppc_set_msr(vcpu, 0);
1175
1176 #ifndef CONFIG_KVM_BOOKE_HV
1177         vcpu->arch.shadow_msr = MSR_USER | MSR_DE | MSR_IS | MSR_DS;
1178         vcpu->arch.shadow_pid = 1;
1179         vcpu->arch.shared->msr = 0;
1180 #endif
1181
1182         /* Eye-catching numbers so we know if the guest takes an interrupt
1183          * before it's programmed its own IVPR/IVORs. */
1184         vcpu->arch.ivpr = 0x55550000;
1185         for (i = 0; i < BOOKE_IRQPRIO_MAX; i++)
1186                 vcpu->arch.ivor[i] = 0x7700 | i * 4;
1187
1188         kvmppc_init_timing_stats(vcpu);
1189
1190         r = kvmppc_core_vcpu_setup(vcpu);
1191         kvmppc_sanity_check(vcpu);
1192         return r;
1193 }
1194
1195 int kvmppc_subarch_vcpu_init(struct kvm_vcpu *vcpu)
1196 {
1197         /* setup watchdog timer once */
1198         spin_lock_init(&vcpu->arch.wdt_lock);
1199         setup_timer(&vcpu->arch.wdt_timer, kvmppc_watchdog_func,
1200                     (unsigned long)vcpu);
1201
1202         return 0;
1203 }
1204
1205 void kvmppc_subarch_vcpu_uninit(struct kvm_vcpu *vcpu)
1206 {
1207         del_timer_sync(&vcpu->arch.wdt_timer);
1208 }
1209
1210 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1211 {
1212         int i;
1213
1214         regs->pc = vcpu->arch.pc;
1215         regs->cr = kvmppc_get_cr(vcpu);
1216         regs->ctr = vcpu->arch.ctr;
1217         regs->lr = vcpu->arch.lr;
1218         regs->xer = kvmppc_get_xer(vcpu);
1219         regs->msr = vcpu->arch.shared->msr;
1220         regs->srr0 = vcpu->arch.shared->srr0;
1221         regs->srr1 = vcpu->arch.shared->srr1;
1222         regs->pid = vcpu->arch.pid;
1223         regs->sprg0 = vcpu->arch.shared->sprg0;
1224         regs->sprg1 = vcpu->arch.shared->sprg1;
1225         regs->sprg2 = vcpu->arch.shared->sprg2;
1226         regs->sprg3 = vcpu->arch.shared->sprg3;
1227         regs->sprg4 = vcpu->arch.shared->sprg4;
1228         regs->sprg5 = vcpu->arch.shared->sprg5;
1229         regs->sprg6 = vcpu->arch.shared->sprg6;
1230         regs->sprg7 = vcpu->arch.shared->sprg7;
1231
1232         for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1233                 regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
1234
1235         return 0;
1236 }
1237
1238 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1239 {
1240         int i;
1241
1242         vcpu->arch.pc = regs->pc;
1243         kvmppc_set_cr(vcpu, regs->cr);
1244         vcpu->arch.ctr = regs->ctr;
1245         vcpu->arch.lr = regs->lr;
1246         kvmppc_set_xer(vcpu, regs->xer);
1247         kvmppc_set_msr(vcpu, regs->msr);
1248         vcpu->arch.shared->srr0 = regs->srr0;
1249         vcpu->arch.shared->srr1 = regs->srr1;
1250         kvmppc_set_pid(vcpu, regs->pid);
1251         vcpu->arch.shared->sprg0 = regs->sprg0;
1252         vcpu->arch.shared->sprg1 = regs->sprg1;
1253         vcpu->arch.shared->sprg2 = regs->sprg2;
1254         vcpu->arch.shared->sprg3 = regs->sprg3;
1255         vcpu->arch.shared->sprg4 = regs->sprg4;
1256         vcpu->arch.shared->sprg5 = regs->sprg5;
1257         vcpu->arch.shared->sprg6 = regs->sprg6;
1258         vcpu->arch.shared->sprg7 = regs->sprg7;
1259
1260         for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1261                 kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
1262
1263         return 0;
1264 }
1265
1266 static void get_sregs_base(struct kvm_vcpu *vcpu,
1267                            struct kvm_sregs *sregs)
1268 {
1269         u64 tb = get_tb();
1270
1271         sregs->u.e.features |= KVM_SREGS_E_BASE;
1272
1273         sregs->u.e.csrr0 = vcpu->arch.csrr0;
1274         sregs->u.e.csrr1 = vcpu->arch.csrr1;
1275         sregs->u.e.mcsr = vcpu->arch.mcsr;
1276         sregs->u.e.esr = get_guest_esr(vcpu);
1277         sregs->u.e.dear = get_guest_dear(vcpu);
1278         sregs->u.e.tsr = vcpu->arch.tsr;
1279         sregs->u.e.tcr = vcpu->arch.tcr;
1280         sregs->u.e.dec = kvmppc_get_dec(vcpu, tb);
1281         sregs->u.e.tb = tb;
1282         sregs->u.e.vrsave = vcpu->arch.vrsave;
1283 }
1284
1285 static int set_sregs_base(struct kvm_vcpu *vcpu,
1286                           struct kvm_sregs *sregs)
1287 {
1288         if (!(sregs->u.e.features & KVM_SREGS_E_BASE))
1289                 return 0;
1290
1291         vcpu->arch.csrr0 = sregs->u.e.csrr0;
1292         vcpu->arch.csrr1 = sregs->u.e.csrr1;
1293         vcpu->arch.mcsr = sregs->u.e.mcsr;
1294         set_guest_esr(vcpu, sregs->u.e.esr);
1295         set_guest_dear(vcpu, sregs->u.e.dear);
1296         vcpu->arch.vrsave = sregs->u.e.vrsave;
1297         kvmppc_set_tcr(vcpu, sregs->u.e.tcr);
1298
1299         if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_DEC) {
1300                 vcpu->arch.dec = sregs->u.e.dec;
1301                 kvmppc_emulate_dec(vcpu);
1302         }
1303
1304         if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_TSR)
1305                 kvmppc_set_tsr(vcpu, sregs->u.e.tsr);
1306
1307         return 0;
1308 }
1309
1310 static void get_sregs_arch206(struct kvm_vcpu *vcpu,
1311                               struct kvm_sregs *sregs)
1312 {
1313         sregs->u.e.features |= KVM_SREGS_E_ARCH206;
1314
1315         sregs->u.e.pir = vcpu->vcpu_id;
1316         sregs->u.e.mcsrr0 = vcpu->arch.mcsrr0;
1317         sregs->u.e.mcsrr1 = vcpu->arch.mcsrr1;
1318         sregs->u.e.decar = vcpu->arch.decar;
1319         sregs->u.e.ivpr = vcpu->arch.ivpr;
1320 }
1321
1322 static int set_sregs_arch206(struct kvm_vcpu *vcpu,
1323                              struct kvm_sregs *sregs)
1324 {
1325         if (!(sregs->u.e.features & KVM_SREGS_E_ARCH206))
1326                 return 0;
1327
1328         if (sregs->u.e.pir != vcpu->vcpu_id)
1329                 return -EINVAL;
1330
1331         vcpu->arch.mcsrr0 = sregs->u.e.mcsrr0;
1332         vcpu->arch.mcsrr1 = sregs->u.e.mcsrr1;
1333         vcpu->arch.decar = sregs->u.e.decar;
1334         vcpu->arch.ivpr = sregs->u.e.ivpr;
1335
1336         return 0;
1337 }
1338
1339 void kvmppc_get_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1340 {
1341         sregs->u.e.features |= KVM_SREGS_E_IVOR;
1342
1343         sregs->u.e.ivor_low[0] = vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL];
1344         sregs->u.e.ivor_low[1] = vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK];
1345         sregs->u.e.ivor_low[2] = vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE];
1346         sregs->u.e.ivor_low[3] = vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE];
1347         sregs->u.e.ivor_low[4] = vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL];
1348         sregs->u.e.ivor_low[5] = vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT];
1349         sregs->u.e.ivor_low[6] = vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM];
1350         sregs->u.e.ivor_low[7] = vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL];
1351         sregs->u.e.ivor_low[8] = vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL];
1352         sregs->u.e.ivor_low[9] = vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL];
1353         sregs->u.e.ivor_low[10] = vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER];
1354         sregs->u.e.ivor_low[11] = vcpu->arch.ivor[BOOKE_IRQPRIO_FIT];
1355         sregs->u.e.ivor_low[12] = vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG];
1356         sregs->u.e.ivor_low[13] = vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS];
1357         sregs->u.e.ivor_low[14] = vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS];
1358         sregs->u.e.ivor_low[15] = vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG];
1359 }
1360
1361 int kvmppc_set_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1362 {
1363         if (!(sregs->u.e.features & KVM_SREGS_E_IVOR))
1364                 return 0;
1365
1366         vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL] = sregs->u.e.ivor_low[0];
1367         vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK] = sregs->u.e.ivor_low[1];
1368         vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE] = sregs->u.e.ivor_low[2];
1369         vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE] = sregs->u.e.ivor_low[3];
1370         vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL] = sregs->u.e.ivor_low[4];
1371         vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT] = sregs->u.e.ivor_low[5];
1372         vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM] = sregs->u.e.ivor_low[6];
1373         vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL] = sregs->u.e.ivor_low[7];
1374         vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL] = sregs->u.e.ivor_low[8];
1375         vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL] = sregs->u.e.ivor_low[9];
1376         vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER] = sregs->u.e.ivor_low[10];
1377         vcpu->arch.ivor[BOOKE_IRQPRIO_FIT] = sregs->u.e.ivor_low[11];
1378         vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG] = sregs->u.e.ivor_low[12];
1379         vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS] = sregs->u.e.ivor_low[13];
1380         vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS] = sregs->u.e.ivor_low[14];
1381         vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG] = sregs->u.e.ivor_low[15];
1382
1383         return 0;
1384 }
1385
1386 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1387                                   struct kvm_sregs *sregs)
1388 {
1389         sregs->pvr = vcpu->arch.pvr;
1390
1391         get_sregs_base(vcpu, sregs);
1392         get_sregs_arch206(vcpu, sregs);
1393         kvmppc_core_get_sregs(vcpu, sregs);
1394         return 0;
1395 }
1396
1397 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1398                                   struct kvm_sregs *sregs)
1399 {
1400         int ret;
1401
1402         if (vcpu->arch.pvr != sregs->pvr)
1403                 return -EINVAL;
1404
1405         ret = set_sregs_base(vcpu, sregs);
1406         if (ret < 0)
1407                 return ret;
1408
1409         ret = set_sregs_arch206(vcpu, sregs);
1410         if (ret < 0)
1411                 return ret;
1412
1413         return kvmppc_core_set_sregs(vcpu, sregs);
1414 }
1415
1416 int kvm_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1417 {
1418         int r = -EINVAL;
1419
1420         switch (reg->id) {
1421         case KVM_REG_PPC_IAC1:
1422         case KVM_REG_PPC_IAC2:
1423         case KVM_REG_PPC_IAC3:
1424         case KVM_REG_PPC_IAC4: {
1425                 int iac = reg->id - KVM_REG_PPC_IAC1;
1426                 r = copy_to_user((u64 __user *)(long)reg->addr,
1427                                  &vcpu->arch.dbg_reg.iac[iac], sizeof(u64));
1428                 break;
1429         }
1430         case KVM_REG_PPC_DAC1:
1431         case KVM_REG_PPC_DAC2: {
1432                 int dac = reg->id - KVM_REG_PPC_DAC1;
1433                 r = copy_to_user((u64 __user *)(long)reg->addr,
1434                                  &vcpu->arch.dbg_reg.dac[dac], sizeof(u64));
1435                 break;
1436         }
1437         case KVM_REG_PPC_EPR: {
1438                 u32 epr = get_guest_epr(vcpu);
1439                 r = put_user(epr, (u32 __user *)(long)reg->addr);
1440                 break;
1441         }
1442 #if defined(CONFIG_64BIT)
1443         case KVM_REG_PPC_EPCR:
1444                 r = put_user(vcpu->arch.epcr, (u32 __user *)(long)reg->addr);
1445                 break;
1446 #endif
1447         case KVM_REG_PPC_TCR:
1448                 r = put_user(vcpu->arch.tcr, (u32 __user *)(long)reg->addr);
1449                 break;
1450         case KVM_REG_PPC_TSR:
1451                 r = put_user(vcpu->arch.tsr, (u32 __user *)(long)reg->addr);
1452                 break;
1453         case KVM_REG_PPC_DEBUG_INST: {
1454                 u32 opcode = KVMPPC_INST_EHPRIV;
1455                 r = copy_to_user((u32 __user *)(long)reg->addr,
1456                                  &opcode, sizeof(u32));
1457                 break;
1458         }
1459         default:
1460                 break;
1461         }
1462         return r;
1463 }
1464
1465 int kvm_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1466 {
1467         int r = -EINVAL;
1468
1469         switch (reg->id) {
1470         case KVM_REG_PPC_IAC1:
1471         case KVM_REG_PPC_IAC2:
1472         case KVM_REG_PPC_IAC3:
1473         case KVM_REG_PPC_IAC4: {
1474                 int iac = reg->id - KVM_REG_PPC_IAC1;
1475                 r = copy_from_user(&vcpu->arch.dbg_reg.iac[iac],
1476                              (u64 __user *)(long)reg->addr, sizeof(u64));
1477                 break;
1478         }
1479         case KVM_REG_PPC_DAC1:
1480         case KVM_REG_PPC_DAC2: {
1481                 int dac = reg->id - KVM_REG_PPC_DAC1;
1482                 r = copy_from_user(&vcpu->arch.dbg_reg.dac[dac],
1483                              (u64 __user *)(long)reg->addr, sizeof(u64));
1484                 break;
1485         }
1486         case KVM_REG_PPC_EPR: {
1487                 u32 new_epr;
1488                 r = get_user(new_epr, (u32 __user *)(long)reg->addr);
1489                 if (!r)
1490                         kvmppc_set_epr(vcpu, new_epr);
1491                 break;
1492         }
1493 #if defined(CONFIG_64BIT)
1494         case KVM_REG_PPC_EPCR: {
1495                 u32 new_epcr;
1496                 r = get_user(new_epcr, (u32 __user *)(long)reg->addr);
1497                 if (r == 0)
1498                         kvmppc_set_epcr(vcpu, new_epcr);
1499                 break;
1500         }
1501 #endif
1502         case KVM_REG_PPC_OR_TSR: {
1503                 u32 tsr_bits;
1504                 r = get_user(tsr_bits, (u32 __user *)(long)reg->addr);
1505                 kvmppc_set_tsr_bits(vcpu, tsr_bits);
1506                 break;
1507         }
1508         case KVM_REG_PPC_CLEAR_TSR: {
1509                 u32 tsr_bits;
1510                 r = get_user(tsr_bits, (u32 __user *)(long)reg->addr);
1511                 kvmppc_clr_tsr_bits(vcpu, tsr_bits);
1512                 break;
1513         }
1514         case KVM_REG_PPC_TSR: {
1515                 u32 tsr;
1516                 r = get_user(tsr, (u32 __user *)(long)reg->addr);
1517                 kvmppc_set_tsr(vcpu, tsr);
1518                 break;
1519         }
1520         case KVM_REG_PPC_TCR: {
1521                 u32 tcr;
1522                 r = get_user(tcr, (u32 __user *)(long)reg->addr);
1523                 kvmppc_set_tcr(vcpu, tcr);
1524                 break;
1525         }
1526         default:
1527                 break;
1528         }
1529         return r;
1530 }
1531
1532 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
1533                                          struct kvm_guest_debug *dbg)
1534 {
1535         return -EINVAL;
1536 }
1537
1538 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1539 {
1540         return -ENOTSUPP;
1541 }
1542
1543 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1544 {
1545         return -ENOTSUPP;
1546 }
1547
1548 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1549                                   struct kvm_translation *tr)
1550 {
1551         int r;
1552
1553         r = kvmppc_core_vcpu_translate(vcpu, tr);
1554         return r;
1555 }
1556
1557 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
1558 {
1559         return -ENOTSUPP;
1560 }
1561
1562 void kvmppc_core_free_memslot(struct kvm_memory_slot *free,
1563                               struct kvm_memory_slot *dont)
1564 {
1565 }
1566
1567 int kvmppc_core_create_memslot(struct kvm_memory_slot *slot,
1568                                unsigned long npages)
1569 {
1570         return 0;
1571 }
1572
1573 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
1574                                       struct kvm_memory_slot *memslot,
1575                                       struct kvm_userspace_memory_region *mem)
1576 {
1577         return 0;
1578 }
1579
1580 void kvmppc_core_commit_memory_region(struct kvm *kvm,
1581                                 struct kvm_userspace_memory_region *mem,
1582                                 const struct kvm_memory_slot *old)
1583 {
1584 }
1585
1586 void kvmppc_core_flush_memslot(struct kvm *kvm, struct kvm_memory_slot *memslot)
1587 {
1588 }
1589
1590 void kvmppc_set_epcr(struct kvm_vcpu *vcpu, u32 new_epcr)
1591 {
1592 #if defined(CONFIG_64BIT)
1593         vcpu->arch.epcr = new_epcr;
1594 #ifdef CONFIG_KVM_BOOKE_HV
1595         vcpu->arch.shadow_epcr &= ~SPRN_EPCR_GICM;
1596         if (vcpu->arch.epcr  & SPRN_EPCR_ICM)
1597                 vcpu->arch.shadow_epcr |= SPRN_EPCR_GICM;
1598 #endif
1599 #endif
1600 }
1601
1602 void kvmppc_set_tcr(struct kvm_vcpu *vcpu, u32 new_tcr)
1603 {
1604         vcpu->arch.tcr = new_tcr;
1605         arm_next_watchdog(vcpu);
1606         update_timer_ints(vcpu);
1607 }
1608
1609 void kvmppc_set_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1610 {
1611         set_bits(tsr_bits, &vcpu->arch.tsr);
1612         smp_wmb();
1613         kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
1614         kvm_vcpu_kick(vcpu);
1615 }
1616
1617 void kvmppc_clr_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1618 {
1619         clear_bits(tsr_bits, &vcpu->arch.tsr);
1620
1621         /*
1622          * We may have stopped the watchdog due to
1623          * being stuck on final expiration.
1624          */
1625         if (tsr_bits & (TSR_ENW | TSR_WIS))
1626                 arm_next_watchdog(vcpu);
1627
1628         update_timer_ints(vcpu);
1629 }
1630
1631 void kvmppc_decrementer_func(unsigned long data)
1632 {
1633         struct kvm_vcpu *vcpu = (struct kvm_vcpu *)data;
1634
1635         if (vcpu->arch.tcr & TCR_ARE) {
1636                 vcpu->arch.dec = vcpu->arch.decar;
1637                 kvmppc_emulate_dec(vcpu);
1638         }
1639
1640         kvmppc_set_tsr_bits(vcpu, TSR_DIS);
1641 }
1642
1643 void kvmppc_booke_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
1644 {
1645         vcpu->cpu = smp_processor_id();
1646         current->thread.kvm_vcpu = vcpu;
1647 }
1648
1649 void kvmppc_booke_vcpu_put(struct kvm_vcpu *vcpu)
1650 {
1651         current->thread.kvm_vcpu = NULL;
1652         vcpu->cpu = -1;
1653 }
1654
1655 int __init kvmppc_booke_init(void)
1656 {
1657 #ifndef CONFIG_KVM_BOOKE_HV
1658         unsigned long ivor[16];
1659         unsigned long *handler = kvmppc_booke_handler_addr;
1660         unsigned long max_ivor = 0;
1661         unsigned long handler_len;
1662         int i;
1663
1664         /* We install our own exception handlers by hijacking IVPR. IVPR must
1665          * be 16-bit aligned, so we need a 64KB allocation. */
1666         kvmppc_booke_handlers = __get_free_pages(GFP_KERNEL | __GFP_ZERO,
1667                                                  VCPU_SIZE_ORDER);
1668         if (!kvmppc_booke_handlers)
1669                 return -ENOMEM;
1670
1671         /* XXX make sure our handlers are smaller than Linux's */
1672
1673         /* Copy our interrupt handlers to match host IVORs. That way we don't
1674          * have to swap the IVORs on every guest/host transition. */
1675         ivor[0] = mfspr(SPRN_IVOR0);
1676         ivor[1] = mfspr(SPRN_IVOR1);
1677         ivor[2] = mfspr(SPRN_IVOR2);
1678         ivor[3] = mfspr(SPRN_IVOR3);
1679         ivor[4] = mfspr(SPRN_IVOR4);
1680         ivor[5] = mfspr(SPRN_IVOR5);
1681         ivor[6] = mfspr(SPRN_IVOR6);
1682         ivor[7] = mfspr(SPRN_IVOR7);
1683         ivor[8] = mfspr(SPRN_IVOR8);
1684         ivor[9] = mfspr(SPRN_IVOR9);
1685         ivor[10] = mfspr(SPRN_IVOR10);
1686         ivor[11] = mfspr(SPRN_IVOR11);
1687         ivor[12] = mfspr(SPRN_IVOR12);
1688         ivor[13] = mfspr(SPRN_IVOR13);
1689         ivor[14] = mfspr(SPRN_IVOR14);
1690         ivor[15] = mfspr(SPRN_IVOR15);
1691
1692         for (i = 0; i < 16; i++) {
1693                 if (ivor[i] > max_ivor)
1694                         max_ivor = i;
1695
1696                 handler_len = handler[i + 1] - handler[i];
1697                 memcpy((void *)kvmppc_booke_handlers + ivor[i],
1698                        (void *)handler[i], handler_len);
1699         }
1700
1701         handler_len = handler[max_ivor + 1] - handler[max_ivor];
1702         flush_icache_range(kvmppc_booke_handlers, kvmppc_booke_handlers +
1703                            ivor[max_ivor] + handler_len);
1704 #endif /* !BOOKE_HV */
1705         return 0;
1706 }
1707
1708 void __exit kvmppc_booke_exit(void)
1709 {
1710         free_pages(kvmppc_booke_handlers, VCPU_SIZE_ORDER);
1711         kvm_exit();
1712 }