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
  40 #include "timing.h"
  41 #include "booke.h"
  42 #include "trace.h"
  43
  44 unsigned long kvmppc_booke_handlers;
  45
  46 #define VM_STAT(x) offsetof(struct kvm, stat.x), KVM_STAT_VM
  47 #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
  48
  49 struct kvm_stats_debugfs_item debugfs_entries[] = {
  50         { "mmio",       VCPU_STAT(mmio_exits) },
  51         { "dcr",        VCPU_STAT(dcr_exits) },
  52         { "sig",        VCPU_STAT(signal_exits) },
  53         { "itlb_r",     VCPU_STAT(itlb_real_miss_exits) },
  54         { "itlb_v",     VCPU_STAT(itlb_virt_miss_exits) },
  55         { "dtlb_r",     VCPU_STAT(dtlb_real_miss_exits) },
  56         { "dtlb_v",     VCPU_STAT(dtlb_virt_miss_exits) },
  57         { "sysc",       VCPU_STAT(syscall_exits) },
  58         { "isi",        VCPU_STAT(isi_exits) },
  59         { "dsi",        VCPU_STAT(dsi_exits) },
  60         { "inst_emu",   VCPU_STAT(emulated_inst_exits) },
  61         { "dec",        VCPU_STAT(dec_exits) },
  62         { "ext_intr",   VCPU_STAT(ext_intr_exits) },
  63         { "halt_wakeup", VCPU_STAT(halt_wakeup) },
  64         { "doorbell", VCPU_STAT(dbell_exits) },
  65         { "guest doorbell", VCPU_STAT(gdbell_exits) },
  66         { "remote_tlb_flush", VM_STAT(remote_tlb_flush) },
  67         { NULL }
  68 };
  69
  70 /* TODO: use vcpu_printf() */
  71 void kvmppc_dump_vcpu(struct kvm_vcpu *vcpu)
  72 {
  73         int i;
  74
  75         printk("pc:   %08lx msr:  %08llx\n", vcpu->arch.pc, vcpu->arch.shared->msr);
  76         printk("lr:   %08lx ctr:  %08lx\n", vcpu->arch.lr, vcpu->arch.ctr);
  77         printk("srr0: %08llx srr1: %08llx\n", vcpu->arch.shared->srr0,
  78                                             vcpu->arch.shared->srr1);
  79
  80         printk("exceptions: %08lx\n", vcpu->arch.pending_exceptions);
  81
  82         for (i = 0; i < 32; i += 4) {
  83                 printk("gpr%02d: %08lx %08lx %08lx %08lx\n", i,
  84                        kvmppc_get_gpr(vcpu, i),
  85                        kvmppc_get_gpr(vcpu, i+1),
  86                        kvmppc_get_gpr(vcpu, i+2),
  87                        kvmppc_get_gpr(vcpu, i+3));
  88         }
  89 }
  90
  91 #ifdef CONFIG_SPE
  92 void kvmppc_vcpu_disable_spe(struct kvm_vcpu *vcpu)
  93 {
  94         preempt_disable();
  95         enable_kernel_spe();
  96         kvmppc_save_guest_spe(vcpu);
  97         vcpu->arch.shadow_msr &= ~MSR_SPE;
  98         preempt_enable();
  99 }
 100
 101 static void kvmppc_vcpu_enable_spe(struct kvm_vcpu *vcpu)
 102 {
 103         preempt_disable();
 104         enable_kernel_spe();
 105         kvmppc_load_guest_spe(vcpu);
 106         vcpu->arch.shadow_msr |= MSR_SPE;
 107         preempt_enable();
 108 }
 109
 110 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
 111 {
 112         if (vcpu->arch.shared->msr & MSR_SPE) {
 113                 if (!(vcpu->arch.shadow_msr & MSR_SPE))
 114                         kvmppc_vcpu_enable_spe(vcpu);
 115         } else if (vcpu->arch.shadow_msr & MSR_SPE) {
 116                 kvmppc_vcpu_disable_spe(vcpu);
 117         }
 118 }
 119 #else
 120 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
 121 {
 122 }
 123 #endif
 124
 125 /*
 126  * Helper function for "full" MSR writes.  No need to call this if only
 127  * EE/CE/ME/DE/RI are changing.
 128  */
 129 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr)
 130 {
 131         u32 old_msr = vcpu->arch.shared->msr;
 132
 133 #ifdef CONFIG_KVM_BOOKE_HV
 134         new_msr |= MSR_GS;
 135 #endif
 136
 137         vcpu->arch.shared->msr = new_msr;
 138
 139         kvmppc_mmu_msr_notify(vcpu, old_msr);
 140         kvmppc_vcpu_sync_spe(vcpu);
 141 }
 142
 143 static void kvmppc_booke_queue_irqprio(struct kvm_vcpu *vcpu,
 144                                        unsigned int priority)
 145 {
 146         trace_kvm_booke_queue_irqprio(vcpu, priority);
 147         set_bit(priority, &vcpu->arch.pending_exceptions);
 148 }
 149
 150 static void kvmppc_core_queue_dtlb_miss(struct kvm_vcpu *vcpu,
 151                                         ulong dear_flags, ulong esr_flags)
 152 {
 153         vcpu->arch.queued_dear = dear_flags;
 154         vcpu->arch.queued_esr = esr_flags;
 155         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DTLB_MISS);
 156 }
 157
 158 static void kvmppc_core_queue_data_storage(struct kvm_vcpu *vcpu,
 159                                            ulong dear_flags, ulong esr_flags)
 160 {
 161         vcpu->arch.queued_dear = dear_flags;
 162         vcpu->arch.queued_esr = esr_flags;
 163         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DATA_STORAGE);
 164 }
 165
 166 static void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu,
 167                                            ulong esr_flags)
 168 {
 169         vcpu->arch.queued_esr = esr_flags;
 170         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_INST_STORAGE);
 171 }
 172
 173 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong esr_flags)
 174 {
 175         vcpu->arch.queued_esr = esr_flags;
 176         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_PROGRAM);
 177 }
 178
 179 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
 180 {
 181         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DECREMENTER);
 182 }
 183
 184 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
 185 {
 186         return test_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
 187 }
 188
 189 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
 190 {
 191         clear_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
 192 }
 193
 194 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
 195                                 struct kvm_interrupt *irq)
 196 {
 197         unsigned int prio = BOOKE_IRQPRIO_EXTERNAL;
 198
 199         if (irq->irq == KVM_INTERRUPT_SET_LEVEL)
 200                 prio = BOOKE_IRQPRIO_EXTERNAL_LEVEL;
 201
 202         kvmppc_booke_queue_irqprio(vcpu, prio);
 203 }
 204
 205 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu,
 206                                   struct kvm_interrupt *irq)
 207 {
 208         clear_bit(BOOKE_IRQPRIO_EXTERNAL, &vcpu->arch.pending_exceptions);
 209         clear_bit(BOOKE_IRQPRIO_EXTERNAL_LEVEL, &vcpu->arch.pending_exceptions);
 210 }
 211
 212 static void set_guest_srr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
 213 {
 214 #ifdef CONFIG_KVM_BOOKE_HV
 215         mtspr(SPRN_GSRR0, srr0);
 216         mtspr(SPRN_GSRR1, srr1);
 217 #else
 218         vcpu->arch.shared->srr0 = srr0;
 219         vcpu->arch.shared->srr1 = srr1;
 220 #endif
 221 }
 222
 223 static void set_guest_csrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
 224 {
 225         vcpu->arch.csrr0 = srr0;
 226         vcpu->arch.csrr1 = srr1;
 227 }
 228
 229 static void set_guest_dsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
 230 {
 231         if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC)) {
 232                 vcpu->arch.dsrr0 = srr0;
 233                 vcpu->arch.dsrr1 = srr1;
 234         } else {
 235                 set_guest_csrr(vcpu, srr0, srr1);
 236         }
 237 }
 238
 239 static void set_guest_mcsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
 240 {
 241         vcpu->arch.mcsrr0 = srr0;
 242         vcpu->arch.mcsrr1 = srr1;
 243 }
 244
 245 static unsigned long get_guest_dear(struct kvm_vcpu *vcpu)
 246 {
 247 #ifdef CONFIG_KVM_BOOKE_HV
 248         return mfspr(SPRN_GDEAR);
 249 #else
 250         return vcpu->arch.shared->dar;
 251 #endif
 252 }
 253
 254 static void set_guest_dear(struct kvm_vcpu *vcpu, unsigned long dear)
 255 {
 256 #ifdef CONFIG_KVM_BOOKE_HV
 257         mtspr(SPRN_GDEAR, dear);
 258 #else
 259         vcpu->arch.shared->dar = dear;
 260 #endif
 261 }
 262
 263 static unsigned long get_guest_esr(struct kvm_vcpu *vcpu)
 264 {
 265 #ifdef CONFIG_KVM_BOOKE_HV
 266         return mfspr(SPRN_GESR);
 267 #else
 268         return vcpu->arch.shared->esr;
 269 #endif
 270 }
 271
 272 static void set_guest_esr(struct kvm_vcpu *vcpu, u32 esr)
 273 {
 274 #ifdef CONFIG_KVM_BOOKE_HV
 275         mtspr(SPRN_GESR, esr);
 276 #else
 277         vcpu->arch.shared->esr = esr;
 278 #endif
 279 }
 280
 281 /* Deliver the interrupt of the corresponding priority, if possible. */
 282 static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu *vcpu,
 283                                         unsigned int priority)
 284 {
 285         int allowed = 0;
 286         ulong msr_mask = 0;
 287         bool update_esr = false, update_dear = false;
 288         ulong crit_raw = vcpu->arch.shared->critical;
 289         ulong crit_r1 = kvmppc_get_gpr(vcpu, 1);
 290         bool crit;
 291         bool keep_irq = false;
 292         enum int_class int_class;
 293
 294         /* Truncate crit indicators in 32 bit mode */
 295         if (!(vcpu->arch.shared->msr & MSR_SF)) {
 296                 crit_raw &= 0xffffffff;
 297                 crit_r1 &= 0xffffffff;
 298         }
 299
 300         /* Critical section when crit == r1 */
 301         crit = (crit_raw == crit_r1);
 302         /* ... and we're in supervisor mode */
 303         crit = crit && !(vcpu->arch.shared->msr & MSR_PR);
 304
 305         if (priority == BOOKE_IRQPRIO_EXTERNAL_LEVEL) {
 306                 priority = BOOKE_IRQPRIO_EXTERNAL;
 307                 keep_irq = true;
 308         }
 309
 310         switch (priority) {
 311         case BOOKE_IRQPRIO_DTLB_MISS:
 312         case BOOKE_IRQPRIO_DATA_STORAGE:
 313                 update_dear = true;
 314                 /* fall through */
 315         case BOOKE_IRQPRIO_INST_STORAGE:
 316         case BOOKE_IRQPRIO_PROGRAM:
 317                 update_esr = true;
 318                 /* fall through */
 319         case BOOKE_IRQPRIO_ITLB_MISS:
 320         case BOOKE_IRQPRIO_SYSCALL:
 321         case BOOKE_IRQPRIO_FP_UNAVAIL:
 322         case BOOKE_IRQPRIO_SPE_UNAVAIL:
 323         case BOOKE_IRQPRIO_SPE_FP_DATA:
 324         case BOOKE_IRQPRIO_SPE_FP_ROUND:
 325         case BOOKE_IRQPRIO_AP_UNAVAIL:
 326         case BOOKE_IRQPRIO_ALIGNMENT:
 327                 allowed = 1;
 328                 msr_mask = MSR_CE | MSR_ME | MSR_DE;
 329                 int_class = INT_CLASS_NONCRIT;
 330                 break;
 331         case BOOKE_IRQPRIO_CRITICAL:
 332         case BOOKE_IRQPRIO_DBELL_CRIT:
 333                 allowed = vcpu->arch.shared->msr & MSR_CE;
 334                 allowed = allowed && !crit;
 335                 msr_mask = MSR_ME;
 336                 int_class = INT_CLASS_CRIT;
 337                 break;
 338         case BOOKE_IRQPRIO_MACHINE_CHECK:
 339                 allowed = vcpu->arch.shared->msr & MSR_ME;
 340                 allowed = allowed && !crit;
 341                 int_class = INT_CLASS_MC;
 342                 break;
 343         case BOOKE_IRQPRIO_DECREMENTER:
 344         case BOOKE_IRQPRIO_FIT:
 345                 keep_irq = true;
 346                 /* fall through */
 347         case BOOKE_IRQPRIO_EXTERNAL:
 348         case BOOKE_IRQPRIO_DBELL:
 349                 allowed = vcpu->arch.shared->msr & MSR_EE;
 350                 allowed = allowed && !crit;
 351                 msr_mask = MSR_CE | MSR_ME | MSR_DE;
 352                 int_class = INT_CLASS_NONCRIT;
 353                 break;
 354         case BOOKE_IRQPRIO_DEBUG:
 355                 allowed = vcpu->arch.shared->msr & MSR_DE;
 356                 allowed = allowed && !crit;
 357                 msr_mask = MSR_ME;
 358                 int_class = INT_CLASS_CRIT;
 359                 break;
 360         }
 361
 362         if (allowed) {
 363                 switch (int_class) {
 364                 case INT_CLASS_NONCRIT:
 365                         set_guest_srr(vcpu, vcpu->arch.pc,
 366                                       vcpu->arch.shared->msr);
 367                         break;
 368                 case INT_CLASS_CRIT:
 369                         set_guest_csrr(vcpu, vcpu->arch.pc,
 370                                        vcpu->arch.shared->msr);
 371                         break;
 372                 case INT_CLASS_DBG:
 373                         set_guest_dsrr(vcpu, vcpu->arch.pc,
 374                                        vcpu->arch.shared->msr);
 375                         break;
 376                 case INT_CLASS_MC:
 377                         set_guest_mcsrr(vcpu, vcpu->arch.pc,
 378                                         vcpu->arch.shared->msr);
 379                         break;
 380                 }
 381
 382                 vcpu->arch.pc = vcpu->arch.ivpr | vcpu->arch.ivor[priority];
 383                 if (update_esr == true)
 384                         set_guest_esr(vcpu, vcpu->arch.queued_esr);
 385                 if (update_dear == true)
 386                         set_guest_dear(vcpu, vcpu->arch.queued_dear);
 387                 kvmppc_set_msr(vcpu, vcpu->arch.shared->msr & msr_mask);
 388
 389                 if (!keep_irq)
 390                         clear_bit(priority, &vcpu->arch.pending_exceptions);
 391         }
 392
 393 #ifdef CONFIG_KVM_BOOKE_HV
 394         /*
 395          * If an interrupt is pending but masked, raise a guest doorbell
 396          * so that we are notified when the guest enables the relevant
 397          * MSR bit.
 398          */
 399         if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_EE)
 400                 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_NONCRIT);
 401         if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_CE)
 402                 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_CRIT);
 403         if (vcpu->arch.pending_exceptions & BOOKE_IRQPRIO_MACHINE_CHECK)
 404                 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_MC);
 405 #endif
 406
 407         return allowed;
 408 }
 409
 410 static void update_timer_ints(struct kvm_vcpu *vcpu)
 411 {
 412         if ((vcpu->arch.tcr & TCR_DIE) && (vcpu->arch.tsr & TSR_DIS))
 413                 kvmppc_core_queue_dec(vcpu);
 414         else
 415                 kvmppc_core_dequeue_dec(vcpu);
 416 }
 417
 418 static void kvmppc_core_check_exceptions(struct kvm_vcpu *vcpu)
 419 {
 420         unsigned long *pending = &vcpu->arch.pending_exceptions;
 421         unsigned int priority;
 422
 423         priority = __ffs(*pending);
 424         while (priority < BOOKE_IRQPRIO_MAX) {
 425                 if (kvmppc_booke_irqprio_deliver(vcpu, priority))
 426                         break;
 427
 428                 priority = find_next_bit(pending,
 429                                          BITS_PER_BYTE * sizeof(*pending),
 430                                          priority + 1);
 431         }
 432
 433         /* Tell the guest about our interrupt status */
 434         vcpu->arch.shared->int_pending = !!*pending;
 435 }
 436
 437 /* Check pending exceptions and deliver one, if possible. */
 438 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
 439 {
 440         int r = 0;
 441         WARN_ON_ONCE(!irqs_disabled());
 442
 443         kvmppc_core_check_exceptions(vcpu);
 444
 445         if (vcpu->arch.shared->msr & MSR_WE) {
 446                 local_irq_enable();
 447                 kvm_vcpu_block(vcpu);
 448                 clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
 449                 local_irq_disable();
 450
 451                 kvmppc_set_exit_type(vcpu, EMULATED_MTMSRWE_EXITS);
 452                 r = 1;
 453         };
 454
 455         return r;
 456 }
 457
 458 void kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
 459 {
 460         if (kvm_check_request(KVM_REQ_PENDING_TIMER, vcpu))
 461                 update_timer_ints(vcpu);
 462 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
 463         if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu))
 464                 kvmppc_core_flush_tlb(vcpu);
 465 #endif
 466 }
 467
 468 int kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
 469 {
 470         int ret;
 471 #ifdef CONFIG_PPC_FPU
 472         unsigned int fpscr;
 473         int fpexc_mode;
 474         u64 fpr[32];
 475 #endif
 476
 477         if (!vcpu->arch.sane) {
 478                 kvm_run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
 479                 return -EINVAL;
 480         }
 481
 482         local_irq_disable();
 483         if (kvmppc_prepare_to_enter(vcpu)) {
 484                 local_irq_enable();
 485                 kvm_run->exit_reason = KVM_EXIT_INTR;
 486                 ret = -EINTR;
 487                 goto out;
 488         }
 489         kvmppc_lazy_ee_enable();
 490
 491         kvm_guest_enter();
 492
 493 #ifdef CONFIG_PPC_FPU
 494         /* Save userspace FPU state in stack */
 495         enable_kernel_fp();
 496         memcpy(fpr, current->thread.fpr, sizeof(current->thread.fpr));
 497         fpscr = current->thread.fpscr.val;
 498         fpexc_mode = current->thread.fpexc_mode;
 499
 500         /* Restore guest FPU state to thread */
 501         memcpy(current->thread.fpr, vcpu->arch.fpr, sizeof(vcpu->arch.fpr));
 502         current->thread.fpscr.val = vcpu->arch.fpscr;
 503
 504         /*
 505          * Since we can't trap on MSR_FP in GS-mode, we consider the guest
 506          * as always using the FPU.  Kernel usage of FP (via
 507          * enable_kernel_fp()) in this thread must not occur while
 508          * vcpu->fpu_active is set.
 509          */
 510         vcpu->fpu_active = 1;
 511
 512         kvmppc_load_guest_fp(vcpu);
 513 #endif
 514
 515         ret = __kvmppc_vcpu_run(kvm_run, vcpu);
 516
 517         /* No need for kvm_guest_exit. It's done in handle_exit.
 518            We also get here with interrupts enabled. */
 519
 520 #ifdef CONFIG_PPC_FPU
 521         kvmppc_save_guest_fp(vcpu);
 522
 523         vcpu->fpu_active = 0;
 524
 525         /* Save guest FPU state from thread */
 526         memcpy(vcpu->arch.fpr, current->thread.fpr, sizeof(vcpu->arch.fpr));
 527         vcpu->arch.fpscr = current->thread.fpscr.val;
 528
 529         /* Restore userspace FPU state from stack */
 530         memcpy(current->thread.fpr, fpr, sizeof(current->thread.fpr));
 531         current->thread.fpscr.val = fpscr;
 532         current->thread.fpexc_mode = fpexc_mode;
 533 #endif
 534
 535 out:
 536         vcpu->mode = OUTSIDE_GUEST_MODE;
 537         smp_wmb();
 538         return ret;
 539 }
 540
 541 static int emulation_exit(struct kvm_run *run, struct kvm_vcpu *vcpu)
 542 {
 543         enum emulation_result er;
 544
 545         er = kvmppc_emulate_instruction(run, vcpu);
 546         switch (er) {
 547         case EMULATE_DONE:
 548                 /* don't overwrite subtypes, just account kvm_stats */
 549                 kvmppc_account_exit_stat(vcpu, EMULATED_INST_EXITS);
 550                 /* Future optimization: only reload non-volatiles if
 551                  * they were actually modified by emulation. */
 552                 return RESUME_GUEST_NV;
 553
 554         case EMULATE_DO_DCR:
 555                 run->exit_reason = KVM_EXIT_DCR;
 556                 return RESUME_HOST;
 557
 558         case EMULATE_FAIL:
 559                 printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
 560                        __func__, vcpu->arch.pc, vcpu->arch.last_inst);
 561                 /* For debugging, encode the failing instruction and
 562                  * report it to userspace. */
 563                 run->hw.hardware_exit_reason = ~0ULL << 32;
 564                 run->hw.hardware_exit_reason |= vcpu->arch.last_inst;
 565                 kvmppc_core_queue_program(vcpu, ESR_PIL);
 566                 return RESUME_HOST;
 567
 568         default:
 569                 BUG();
 570         }
 571 }
 572
 573 static void kvmppc_fill_pt_regs(struct pt_regs *regs)
 574 {
 575         ulong r1, ip, msr, lr;
 576
 577         asm("mr %0, 1" : "=r"(r1));
 578         asm("mflr %0" : "=r"(lr));
 579         asm("mfmsr %0" : "=r"(msr));
 580         asm("bl 1f; 1: mflr %0" : "=r"(ip));
 581
 582         memset(regs, 0, sizeof(*regs));
 583         regs->gpr[1] = r1;
 584         regs->nip = ip;
 585         regs->msr = msr;
 586         regs->link = lr;
 587 }
 588
 589 /*
 590  * For interrupts needed to be handled by host interrupt handlers,
 591  * corresponding host handler are called from here in similar way
 592  * (but not exact) as they are called from low level handler
 593  * (such as from arch/powerpc/kernel/head_fsl_booke.S).
 594  */
 595 static void kvmppc_restart_interrupt(struct kvm_vcpu *vcpu,
 596                                      unsigned int exit_nr)
 597 {
 598         struct pt_regs regs;
 599
 600         switch (exit_nr) {
 601         case BOOKE_INTERRUPT_EXTERNAL:
 602                 kvmppc_fill_pt_regs(&regs);
 603                 do_IRQ(&regs);
 604                 break;
 605         case BOOKE_INTERRUPT_DECREMENTER:
 606                 kvmppc_fill_pt_regs(&regs);
 607                 timer_interrupt(&regs);
 608                 break;
 609 #if defined(CONFIG_PPC_FSL_BOOK3E) || defined(CONFIG_PPC_BOOK3E_64)
 610         case BOOKE_INTERRUPT_DOORBELL:
 611                 kvmppc_fill_pt_regs(&regs);
 612                 doorbell_exception(&regs);
 613                 break;
 614 #endif
 615         case BOOKE_INTERRUPT_MACHINE_CHECK:
 616                 /* FIXME */
 617                 break;
 618         case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
 619                 kvmppc_fill_pt_regs(&regs);
 620                 performance_monitor_exception(&regs);
 621                 break;
 622         case BOOKE_INTERRUPT_WATCHDOG:
 623                 kvmppc_fill_pt_regs(&regs);
 624 #ifdef CONFIG_BOOKE_WDT
 625                 WatchdogException(&regs);
 626 #else
 627                 unknown_exception(&regs);
 628 #endif
 629                 break;
 630         case BOOKE_INTERRUPT_CRITICAL:
 631                 unknown_exception(&regs);
 632                 break;
 633         }
 634 }
 635
 636 /**
 637  * kvmppc_handle_exit
 638  *
 639  * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
 640  */
 641 int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
 642                        unsigned int exit_nr)
 643 {
 644         int r = RESUME_HOST;
 645
 646         /* update before a new last_exit_type is rewritten */
 647         kvmppc_update_timing_stats(vcpu);
 648
 649         /* restart interrupts if they were meant for the host */
 650         kvmppc_restart_interrupt(vcpu, exit_nr);
 651
 652         local_irq_enable();
 653
 654         trace_kvm_exit(exit_nr, vcpu);
 655         kvm_guest_exit();
 656
 657         run->exit_reason = KVM_EXIT_UNKNOWN;
 658         run->ready_for_interrupt_injection = 1;
 659
 660         switch (exit_nr) {
 661         case BOOKE_INTERRUPT_MACHINE_CHECK:
 662                 printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR));
 663                 kvmppc_dump_vcpu(vcpu);
 664                 /* For debugging, send invalid exit reason to user space */
 665                 run->hw.hardware_exit_reason = ~1ULL << 32;
 666                 run->hw.hardware_exit_reason |= mfspr(SPRN_MCSR);
 667                 r = RESUME_HOST;
 668                 break;
 669
 670         case BOOKE_INTERRUPT_EXTERNAL:
 671                 kvmppc_account_exit(vcpu, EXT_INTR_EXITS);
 672                 r = RESUME_GUEST;
 673                 break;
 674
 675         case BOOKE_INTERRUPT_DECREMENTER:
 676                 kvmppc_account_exit(vcpu, DEC_EXITS);
 677                 r = RESUME_GUEST;
 678                 break;
 679
 680         case BOOKE_INTERRUPT_WATCHDOG:
 681                 r = RESUME_GUEST;
 682                 break;
 683
 684         case BOOKE_INTERRUPT_DOORBELL:
 685                 kvmppc_account_exit(vcpu, DBELL_EXITS);
 686                 r = RESUME_GUEST;
 687                 break;
 688
 689         case BOOKE_INTERRUPT_GUEST_DBELL_CRIT:
 690                 kvmppc_account_exit(vcpu, GDBELL_EXITS);
 691
 692                 /*
 693                  * We are here because there is a pending guest interrupt
 694                  * which could not be delivered as MSR_CE or MSR_ME was not
 695                  * set.  Once we break from here we will retry delivery.
 696                  */
 697                 r = RESUME_GUEST;
 698                 break;
 699
 700         case BOOKE_INTERRUPT_GUEST_DBELL:
 701                 kvmppc_account_exit(vcpu, GDBELL_EXITS);
 702
 703                 /*
 704                  * We are here because there is a pending guest interrupt
 705                  * which could not be delivered as MSR_EE was not set.  Once
 706                  * we break from here we will retry delivery.
 707                  */
 708                 r = RESUME_GUEST;
 709                 break;
 710
 711         case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
 712                 r = RESUME_GUEST;
 713                 break;
 714
 715         case BOOKE_INTERRUPT_HV_PRIV:
 716                 r = emulation_exit(run, vcpu);
 717                 break;
 718
 719         case BOOKE_INTERRUPT_PROGRAM:
 720                 if (vcpu->arch.shared->msr & (MSR_PR | MSR_GS)) {
 721                         /*
 722                          * Program traps generated by user-level software must
 723                          * be handled by the guest kernel.
 724                          *
 725                          * In GS mode, hypervisor privileged instructions trap
 726                          * on BOOKE_INTERRUPT_HV_PRIV, not here, so these are
 727                          * actual program interrupts, handled by the guest.
 728                          */
 729                         kvmppc_core_queue_program(vcpu, vcpu->arch.fault_esr);
 730                         r = RESUME_GUEST;
 731                         kvmppc_account_exit(vcpu, USR_PR_INST);
 732                         break;
 733                 }
 734
 735                 r = emulation_exit(run, vcpu);
 736                 break;
 737
 738         case BOOKE_INTERRUPT_FP_UNAVAIL:
 739                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
 740                 kvmppc_account_exit(vcpu, FP_UNAVAIL);
 741                 r = RESUME_GUEST;
 742                 break;
 743
 744 #ifdef CONFIG_SPE
 745         case BOOKE_INTERRUPT_SPE_UNAVAIL: {
 746                 if (vcpu->arch.shared->msr & MSR_SPE)
 747                         kvmppc_vcpu_enable_spe(vcpu);
 748                 else
 749                         kvmppc_booke_queue_irqprio(vcpu,
 750                                                    BOOKE_IRQPRIO_SPE_UNAVAIL);
 751                 r = RESUME_GUEST;
 752                 break;
 753         }
 754
 755         case BOOKE_INTERRUPT_SPE_FP_DATA:
 756                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_DATA);
 757                 r = RESUME_GUEST;
 758                 break;
 759
 760         case BOOKE_INTERRUPT_SPE_FP_ROUND:
 761                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_ROUND);
 762                 r = RESUME_GUEST;
 763                 break;
 764 #else
 765         case BOOKE_INTERRUPT_SPE_UNAVAIL:
 766                 /*
 767                  * Guest wants SPE, but host kernel doesn't support it.  Send
 768                  * an "unimplemented operation" program check to the guest.
 769                  */
 770                 kvmppc_core_queue_program(vcpu, ESR_PUO | ESR_SPV);
 771                 r = RESUME_GUEST;
 772                 break;
 773
 774         /*
 775          * These really should never happen without CONFIG_SPE,
 776          * as we should never enable the real MSR[SPE] in the guest.
 777          */
 778         case BOOKE_INTERRUPT_SPE_FP_DATA:
 779         case BOOKE_INTERRUPT_SPE_FP_ROUND:
 780                 printk(KERN_CRIT "%s: unexpected SPE interrupt %u at %08lx\n",
 781                        __func__, exit_nr, vcpu->arch.pc);
 782                 run->hw.hardware_exit_reason = exit_nr;
 783                 r = RESUME_HOST;
 784                 break;
 785 #endif
 786
 787         case BOOKE_INTERRUPT_DATA_STORAGE:
 788                 kvmppc_core_queue_data_storage(vcpu, vcpu->arch.fault_dear,
 789                                                vcpu->arch.fault_esr);
 790                 kvmppc_account_exit(vcpu, DSI_EXITS);
 791                 r = RESUME_GUEST;
 792                 break;
 793
 794         case BOOKE_INTERRUPT_INST_STORAGE:
 795                 kvmppc_core_queue_inst_storage(vcpu, vcpu->arch.fault_esr);
 796                 kvmppc_account_exit(vcpu, ISI_EXITS);
 797                 r = RESUME_GUEST;
 798                 break;
 799
 800 #ifdef CONFIG_KVM_BOOKE_HV
 801         case BOOKE_INTERRUPT_HV_SYSCALL:
 802                 if (!(vcpu->arch.shared->msr & MSR_PR)) {
 803                         kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
 804                 } else {
 805                         /*
 806                          * hcall from guest userspace -- send privileged
 807                          * instruction program check.
 808                          */
 809                         kvmppc_core_queue_program(vcpu, ESR_PPR);
 810                 }
 811
 812                 r = RESUME_GUEST;
 813                 break;
 814 #else
 815         case BOOKE_INTERRUPT_SYSCALL:
 816                 if (!(vcpu->arch.shared->msr & MSR_PR) &&
 817                     (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
 818                         /* KVM PV hypercalls */
 819                         kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
 820                         r = RESUME_GUEST;
 821                 } else {
 822                         /* Guest syscalls */
 823                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SYSCALL);
 824                 }
 825                 kvmppc_account_exit(vcpu, SYSCALL_EXITS);
 826                 r = RESUME_GUEST;
 827                 break;
 828 #endif
 829
 830         case BOOKE_INTERRUPT_DTLB_MISS: {
 831                 unsigned long eaddr = vcpu->arch.fault_dear;
 832                 int gtlb_index;
 833                 gpa_t gpaddr;
 834                 gfn_t gfn;
 835
 836 #ifdef CONFIG_KVM_E500V2
 837                 if (!(vcpu->arch.shared->msr & MSR_PR) &&
 838                     (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
 839                         kvmppc_map_magic(vcpu);
 840                         kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
 841                         r = RESUME_GUEST;
 842
 843                         break;
 844                 }
 845 #endif
 846
 847                 /* Check the guest TLB. */
 848                 gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
 849                 if (gtlb_index < 0) {
 850                         /* The guest didn't have a mapping for it. */
 851                         kvmppc_core_queue_dtlb_miss(vcpu,
 852                                                     vcpu->arch.fault_dear,
 853                                                     vcpu->arch.fault_esr);
 854                         kvmppc_mmu_dtlb_miss(vcpu);
 855                         kvmppc_account_exit(vcpu, DTLB_REAL_MISS_EXITS);
 856                         r = RESUME_GUEST;
 857                         break;
 858                 }
 859
 860                 gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
 861                 gfn = gpaddr >> PAGE_SHIFT;
 862
 863                 if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
 864                         /* The guest TLB had a mapping, but the shadow TLB
 865                          * didn't, and it is RAM. This could be because:
 866                          * a) the entry is mapping the host kernel, or
 867                          * b) the guest used a large mapping which we're faking
 868                          * Either way, we need to satisfy the fault without
 869                          * invoking the guest. */
 870                         kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
 871                         kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
 872                         r = RESUME_GUEST;
 873                 } else {
 874                         /* Guest has mapped and accessed a page which is not
 875                          * actually RAM. */
 876                         vcpu->arch.paddr_accessed = gpaddr;
 877                         vcpu->arch.vaddr_accessed = eaddr;
 878                         r = kvmppc_emulate_mmio(run, vcpu);
 879                         kvmppc_account_exit(vcpu, MMIO_EXITS);
 880                 }
 881
 882                 break;
 883         }
 884
 885         case BOOKE_INTERRUPT_ITLB_MISS: {
 886                 unsigned long eaddr = vcpu->arch.pc;
 887                 gpa_t gpaddr;
 888                 gfn_t gfn;
 889                 int gtlb_index;
 890
 891                 r = RESUME_GUEST;
 892
 893                 /* Check the guest TLB. */
 894                 gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
 895                 if (gtlb_index < 0) {
 896                         /* The guest didn't have a mapping for it. */
 897                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
 898                         kvmppc_mmu_itlb_miss(vcpu);
 899                         kvmppc_account_exit(vcpu, ITLB_REAL_MISS_EXITS);
 900                         break;
 901                 }
 902
 903                 kvmppc_account_exit(vcpu, ITLB_VIRT_MISS_EXITS);
 904
 905                 gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
 906                 gfn = gpaddr >> PAGE_SHIFT;
 907
 908                 if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
 909                         /* The guest TLB had a mapping, but the shadow TLB
 910                          * didn't. This could be because:
 911                          * a) the entry is mapping the host kernel, or
 912                          * b) the guest used a large mapping which we're faking
 913                          * Either way, we need to satisfy the fault without
 914                          * invoking the guest. */
 915                         kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
 916                 } else {
 917                         /* Guest mapped and leaped at non-RAM! */
 918                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_MACHINE_CHECK);
 919                 }
 920
 921                 break;
 922         }
 923
 924         case BOOKE_INTERRUPT_DEBUG: {
 925                 u32 dbsr;
 926
 927                 vcpu->arch.pc = mfspr(SPRN_CSRR0);
 928
 929                 /* clear IAC events in DBSR register */
 930                 dbsr = mfspr(SPRN_DBSR);
 931                 dbsr &= DBSR_IAC1 | DBSR_IAC2 | DBSR_IAC3 | DBSR_IAC4;
 932                 mtspr(SPRN_DBSR, dbsr);
 933
 934                 run->exit_reason = KVM_EXIT_DEBUG;
 935                 kvmppc_account_exit(vcpu, DEBUG_EXITS);
 936                 r = RESUME_HOST;
 937                 break;
 938         }
 939
 940         default:
 941                 printk(KERN_EMERG "exit_nr %d\n", exit_nr);
 942                 BUG();
 943         }
 944
 945         /*
 946          * To avoid clobbering exit_reason, only check for signals if we
 947          * aren't already exiting to userspace for some other reason.
 948          */
 949         if (!(r & RESUME_HOST)) {
 950                 local_irq_disable();
 951                 if (kvmppc_prepare_to_enter(vcpu)) {
 952                         local_irq_enable();
 953                         run->exit_reason = KVM_EXIT_INTR;
 954                         r = (-EINTR << 2) | RESUME_HOST | (r & RESUME_FLAG_NV);
 955                         kvmppc_account_exit(vcpu, SIGNAL_EXITS);
 956                 } else {
 957                         kvmppc_lazy_ee_enable();
 958                 }
 959         }
 960
 961         return r;
 962 }
 963
 964 /* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
 965 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
 966 {
 967         int i;
 968         int r;
 969
 970         vcpu->arch.pc = 0;
 971         vcpu->arch.shared->pir = vcpu->vcpu_id;
 972         kvmppc_set_gpr(vcpu, 1, (16<<20) - 8); /* -8 for the callee-save LR slot */
 973         kvmppc_set_msr(vcpu, 0);
 974
 975 #ifndef CONFIG_KVM_BOOKE_HV
 976         vcpu->arch.shadow_msr = MSR_USER | MSR_DE | MSR_IS | MSR_DS;
 977         vcpu->arch.shadow_pid = 1;
 978         vcpu->arch.shared->msr = 0;
 979 #endif
 980
 981         /* Eye-catching numbers so we know if the guest takes an interrupt
 982          * before it's programmed its own IVPR/IVORs. */
 983         vcpu->arch.ivpr = 0x55550000;
 984         for (i = 0; i < BOOKE_IRQPRIO_MAX; i++)
 985                 vcpu->arch.ivor[i] = 0x7700 | i * 4;
 986
 987         kvmppc_init_timing_stats(vcpu);
 988
 989         r = kvmppc_core_vcpu_setup(vcpu);
 990         kvmppc_sanity_check(vcpu);
 991         return r;
 992 }
 993
 994 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
 995 {
 996         int i;
 997
 998         regs->pc = vcpu->arch.pc;
 999         regs->cr = kvmppc_get_cr(vcpu);
1000         regs->ctr = vcpu->arch.ctr;
1001         regs->lr = vcpu->arch.lr;
1002         regs->xer = kvmppc_get_xer(vcpu);
1003         regs->msr = vcpu->arch.shared->msr;
1004         regs->srr0 = vcpu->arch.shared->srr0;
1005         regs->srr1 = vcpu->arch.shared->srr1;
1006         regs->pid = vcpu->arch.pid;
1007         regs->sprg0 = vcpu->arch.shared->sprg0;
1008         regs->sprg1 = vcpu->arch.shared->sprg1;
1009         regs->sprg2 = vcpu->arch.shared->sprg2;
1010         regs->sprg3 = vcpu->arch.shared->sprg3;
1011         regs->sprg4 = vcpu->arch.shared->sprg4;
1012         regs->sprg5 = vcpu->arch.shared->sprg5;
1013         regs->sprg6 = vcpu->arch.shared->sprg6;
1014         regs->sprg7 = vcpu->arch.shared->sprg7;
1015
1016         for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1017                 regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
1018
1019         return 0;
1020 }
1021
1022 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1023 {
1024         int i;
1025
1026         vcpu->arch.pc = regs->pc;
1027         kvmppc_set_cr(vcpu, regs->cr);
1028         vcpu->arch.ctr = regs->ctr;
1029         vcpu->arch.lr = regs->lr;
1030         kvmppc_set_xer(vcpu, regs->xer);
1031         kvmppc_set_msr(vcpu, regs->msr);
1032         vcpu->arch.shared->srr0 = regs->srr0;
1033         vcpu->arch.shared->srr1 = regs->srr1;
1034         kvmppc_set_pid(vcpu, regs->pid);
1035         vcpu->arch.shared->sprg0 = regs->sprg0;
1036         vcpu->arch.shared->sprg1 = regs->sprg1;
1037         vcpu->arch.shared->sprg2 = regs->sprg2;
1038         vcpu->arch.shared->sprg3 = regs->sprg3;
1039         vcpu->arch.shared->sprg4 = regs->sprg4;
1040         vcpu->arch.shared->sprg5 = regs->sprg5;
1041         vcpu->arch.shared->sprg6 = regs->sprg6;
1042         vcpu->arch.shared->sprg7 = regs->sprg7;
1043
1044         for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1045                 kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
1046
1047         return 0;
1048 }
1049
1050 static void get_sregs_base(struct kvm_vcpu *vcpu,
1051                            struct kvm_sregs *sregs)
1052 {
1053         u64 tb = get_tb();
1054
1055         sregs->u.e.features |= KVM_SREGS_E_BASE;
1056
1057         sregs->u.e.csrr0 = vcpu->arch.csrr0;
1058         sregs->u.e.csrr1 = vcpu->arch.csrr1;
1059         sregs->u.e.mcsr = vcpu->arch.mcsr;
1060         sregs->u.e.esr = get_guest_esr(vcpu);
1061         sregs->u.e.dear = get_guest_dear(vcpu);
1062         sregs->u.e.tsr = vcpu->arch.tsr;
1063         sregs->u.e.tcr = vcpu->arch.tcr;
1064         sregs->u.e.dec = kvmppc_get_dec(vcpu, tb);
1065         sregs->u.e.tb = tb;
1066         sregs->u.e.vrsave = vcpu->arch.vrsave;
1067 }
1068
1069 static int set_sregs_base(struct kvm_vcpu *vcpu,
1070                           struct kvm_sregs *sregs)
1071 {
1072         if (!(sregs->u.e.features & KVM_SREGS_E_BASE))
1073                 return 0;
1074
1075         vcpu->arch.csrr0 = sregs->u.e.csrr0;
1076         vcpu->arch.csrr1 = sregs->u.e.csrr1;
1077         vcpu->arch.mcsr = sregs->u.e.mcsr;
1078         set_guest_esr(vcpu, sregs->u.e.esr);
1079         set_guest_dear(vcpu, sregs->u.e.dear);
1080         vcpu->arch.vrsave = sregs->u.e.vrsave;
1081         kvmppc_set_tcr(vcpu, sregs->u.e.tcr);
1082
1083         if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_DEC) {
1084                 vcpu->arch.dec = sregs->u.e.dec;
1085                 kvmppc_emulate_dec(vcpu);
1086         }
1087
1088         if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_TSR) {
1089                 vcpu->arch.tsr = sregs->u.e.tsr;
1090                 update_timer_ints(vcpu);
1091         }
1092
1093         return 0;
1094 }
1095
1096 static void get_sregs_arch206(struct kvm_vcpu *vcpu,
1097                               struct kvm_sregs *sregs)
1098 {
1099         sregs->u.e.features |= KVM_SREGS_E_ARCH206;
1100
1101         sregs->u.e.pir = vcpu->vcpu_id;
1102         sregs->u.e.mcsrr0 = vcpu->arch.mcsrr0;
1103         sregs->u.e.mcsrr1 = vcpu->arch.mcsrr1;
1104         sregs->u.e.decar = vcpu->arch.decar;
1105         sregs->u.e.ivpr = vcpu->arch.ivpr;
1106 }
1107
1108 static int set_sregs_arch206(struct kvm_vcpu *vcpu,
1109                              struct kvm_sregs *sregs)
1110 {
1111         if (!(sregs->u.e.features & KVM_SREGS_E_ARCH206))
1112                 return 0;
1113
1114         if (sregs->u.e.pir != vcpu->vcpu_id)
1115                 return -EINVAL;
1116
1117         vcpu->arch.mcsrr0 = sregs->u.e.mcsrr0;
1118         vcpu->arch.mcsrr1 = sregs->u.e.mcsrr1;
1119         vcpu->arch.decar = sregs->u.e.decar;
1120         vcpu->arch.ivpr = sregs->u.e.ivpr;
1121
1122         return 0;
1123 }
1124
1125 void kvmppc_get_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1126 {
1127         sregs->u.e.features |= KVM_SREGS_E_IVOR;
1128
1129         sregs->u.e.ivor_low[0] = vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL];
1130         sregs->u.e.ivor_low[1] = vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK];
1131         sregs->u.e.ivor_low[2] = vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE];
1132         sregs->u.e.ivor_low[3] = vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE];
1133         sregs->u.e.ivor_low[4] = vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL];
1134         sregs->u.e.ivor_low[5] = vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT];
1135         sregs->u.e.ivor_low[6] = vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM];
1136         sregs->u.e.ivor_low[7] = vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL];
1137         sregs->u.e.ivor_low[8] = vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL];
1138         sregs->u.e.ivor_low[9] = vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL];
1139         sregs->u.e.ivor_low[10] = vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER];
1140         sregs->u.e.ivor_low[11] = vcpu->arch.ivor[BOOKE_IRQPRIO_FIT];
1141         sregs->u.e.ivor_low[12] = vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG];
1142         sregs->u.e.ivor_low[13] = vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS];
1143         sregs->u.e.ivor_low[14] = vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS];
1144         sregs->u.e.ivor_low[15] = vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG];
1145 }
1146
1147 int kvmppc_set_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1148 {
1149         if (!(sregs->u.e.features & KVM_SREGS_E_IVOR))
1150                 return 0;
1151
1152         vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL] = sregs->u.e.ivor_low[0];
1153         vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK] = sregs->u.e.ivor_low[1];
1154         vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE] = sregs->u.e.ivor_low[2];
1155         vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE] = sregs->u.e.ivor_low[3];
1156         vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL] = sregs->u.e.ivor_low[4];
1157         vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT] = sregs->u.e.ivor_low[5];
1158         vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM] = sregs->u.e.ivor_low[6];
1159         vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL] = sregs->u.e.ivor_low[7];
1160         vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL] = sregs->u.e.ivor_low[8];
1161         vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL] = sregs->u.e.ivor_low[9];
1162         vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER] = sregs->u.e.ivor_low[10];
1163         vcpu->arch.ivor[BOOKE_IRQPRIO_FIT] = sregs->u.e.ivor_low[11];
1164         vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG] = sregs->u.e.ivor_low[12];
1165         vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS] = sregs->u.e.ivor_low[13];
1166         vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS] = sregs->u.e.ivor_low[14];
1167         vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG] = sregs->u.e.ivor_low[15];
1168
1169         return 0;
1170 }
1171
1172 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1173                                   struct kvm_sregs *sregs)
1174 {
1175         sregs->pvr = vcpu->arch.pvr;
1176
1177         get_sregs_base(vcpu, sregs);
1178         get_sregs_arch206(vcpu, sregs);
1179         kvmppc_core_get_sregs(vcpu, sregs);
1180         return 0;
1181 }
1182
1183 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1184                                   struct kvm_sregs *sregs)
1185 {
1186         int ret;
1187
1188         if (vcpu->arch.pvr != sregs->pvr)
1189                 return -EINVAL;
1190
1191         ret = set_sregs_base(vcpu, sregs);
1192         if (ret < 0)
1193                 return ret;
1194
1195         ret = set_sregs_arch206(vcpu, sregs);
1196         if (ret < 0)
1197                 return ret;
1198
1199         return kvmppc_core_set_sregs(vcpu, sregs);
1200 }
1201
1202 int kvm_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1203 {
1204         return -EINVAL;
1205 }
1206
1207 int kvm_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1208 {
1209         return -EINVAL;
1210 }
1211
1212 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1213 {
1214         return -ENOTSUPP;
1215 }
1216
1217 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1218 {
1219         return -ENOTSUPP;
1220 }
1221
1222 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1223                                   struct kvm_translation *tr)
1224 {
1225         int r;
1226
1227         r = kvmppc_core_vcpu_translate(vcpu, tr);
1228         return r;
1229 }
1230
1231 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
1232 {
1233         return -ENOTSUPP;
1234 }
1235
1236 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
1237                                       struct kvm_userspace_memory_region *mem)
1238 {
1239         return 0;
1240 }
1241
1242 void kvmppc_core_commit_memory_region(struct kvm *kvm,
1243                                 struct kvm_userspace_memory_region *mem)
1244 {
1245 }
1246
1247 void kvmppc_set_tcr(struct kvm_vcpu *vcpu, u32 new_tcr)
1248 {
1249         vcpu->arch.tcr = new_tcr;
1250         update_timer_ints(vcpu);
1251 }
1252
1253 void kvmppc_set_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1254 {
1255         set_bits(tsr_bits, &vcpu->arch.tsr);
1256         smp_wmb();
1257         kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
1258         kvm_vcpu_kick(vcpu);
1259 }
1260
1261 void kvmppc_clr_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1262 {
1263         clear_bits(tsr_bits, &vcpu->arch.tsr);
1264         update_timer_ints(vcpu);
1265 }
1266
1267 void kvmppc_decrementer_func(unsigned long data)
1268 {
1269         struct kvm_vcpu *vcpu = (struct kvm_vcpu *)data;
1270
1271         if (vcpu->arch.tcr & TCR_ARE) {
1272                 vcpu->arch.dec = vcpu->arch.decar;
1273                 kvmppc_emulate_dec(vcpu);
1274         }
1275
1276         kvmppc_set_tsr_bits(vcpu, TSR_DIS);
1277 }
1278
1279 void kvmppc_booke_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
1280 {
1281         current->thread.kvm_vcpu = vcpu;
1282 }
1283
1284 void kvmppc_booke_vcpu_put(struct kvm_vcpu *vcpu)
1285 {
1286         current->thread.kvm_vcpu = NULL;
1287 }
1288
1289 int __init kvmppc_booke_init(void)
1290 {
1291 #ifndef CONFIG_KVM_BOOKE_HV
1292         unsigned long ivor[16];
1293         unsigned long max_ivor = 0;
1294         int i;
1295
1296         /* We install our own exception handlers by hijacking IVPR. IVPR must
1297          * be 16-bit aligned, so we need a 64KB allocation. */
1298         kvmppc_booke_handlers = __get_free_pages(GFP_KERNEL | __GFP_ZERO,
1299                                                  VCPU_SIZE_ORDER);
1300         if (!kvmppc_booke_handlers)
1301                 return -ENOMEM;
1302
1303         /* XXX make sure our handlers are smaller than Linux's */
1304
1305         /* Copy our interrupt handlers to match host IVORs. That way we don't
1306          * have to swap the IVORs on every guest/host transition. */
1307         ivor[0] = mfspr(SPRN_IVOR0);
1308         ivor[1] = mfspr(SPRN_IVOR1);
1309         ivor[2] = mfspr(SPRN_IVOR2);
1310         ivor[3] = mfspr(SPRN_IVOR3);
1311         ivor[4] = mfspr(SPRN_IVOR4);
1312         ivor[5] = mfspr(SPRN_IVOR5);
1313         ivor[6] = mfspr(SPRN_IVOR6);
1314         ivor[7] = mfspr(SPRN_IVOR7);
1315         ivor[8] = mfspr(SPRN_IVOR8);
1316         ivor[9] = mfspr(SPRN_IVOR9);
1317         ivor[10] = mfspr(SPRN_IVOR10);
1318         ivor[11] = mfspr(SPRN_IVOR11);
1319         ivor[12] = mfspr(SPRN_IVOR12);
1320         ivor[13] = mfspr(SPRN_IVOR13);
1321         ivor[14] = mfspr(SPRN_IVOR14);
1322         ivor[15] = mfspr(SPRN_IVOR15);
1323
1324         for (i = 0; i < 16; i++) {
1325                 if (ivor[i] > max_ivor)
1326                         max_ivor = ivor[i];
1327
1328                 memcpy((void *)kvmppc_booke_handlers + ivor[i],
1329                        kvmppc_handlers_start + i * kvmppc_handler_len,
1330                        kvmppc_handler_len);
1331         }
1332         flush_icache_range(kvmppc_booke_handlers,
1333                            kvmppc_booke_handlers + max_ivor + kvmppc_handler_len);
1334 #endif /* !BOOKE_HV */
1335         return 0;
1336 }
1337
1338 void __exit kvmppc_booke_exit(void)
1339 {
1340         free_pages(kvmppc_booke_handlers, VCPU_SIZE_ORDER);
1341         kvm_exit();
1342 }