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