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

   1 /*
   2  * This program is free software; you can redistribute it and/or modify
   3  * it under the terms of the GNU General Public License, version 2, as
   4  * published by the Free Software Foundation.
   5  *
   6  * This program is distributed in the hope that it will be useful,
   7  * but WITHOUT ANY WARRANTY; without even the implied warranty of
   8  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
   9  * GNU General Public License for more details.
  10  *
  11  * You should have received a copy of the GNU General Public License
  12  * along with this program; if not, write to the Free Software
  13  * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
  14  *
  15  * Copyright IBM Corp. 2007
  16  * Copyright 2010-2011 Freescale Semiconductor, Inc.
  17  *
  18  * Authors: Hollis Blanchard <hollisb@us.ibm.com>
  19  *          Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
  20  *          Scott Wood <scottwood@freescale.com>
  21  *          Varun Sethi <varun.sethi@freescale.com>
  22  */
  23
  24 #include <linux/errno.h>
  25 #include <linux/err.h>
  26 #include <linux/kvm_host.h>
  27 #include <linux/gfp.h>
  28 #include <linux/module.h>
  29 #include <linux/vmalloc.h>
  30 #include <linux/fs.h>
  31
  32 #include <asm/cputable.h>
  33 #include <asm/uaccess.h>
  34 #include <asm/kvm_ppc.h>
  35 #include <asm/cacheflush.h>
  36 #include <asm/dbell.h>
  37 #include <asm/hw_irq.h>
  38 #include <asm/irq.h>
  39
  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         { NULL }
  67 };
  68
  69 /* TODO: use vcpu_printf() */
  70 void kvmppc_dump_vcpu(struct kvm_vcpu *vcpu)
  71 {
  72         int i;
  73
  74         printk("pc:   %08lx msr:  %08llx\n", vcpu->arch.pc, vcpu->arch.shared->msr);
  75         printk("lr:   %08lx ctr:  %08lx\n", vcpu->arch.lr, vcpu->arch.ctr);
  76         printk("srr0: %08llx srr1: %08llx\n", vcpu->arch.shared->srr0,
  77                                             vcpu->arch.shared->srr1);
  78
  79         printk("exceptions: %08lx\n", vcpu->arch.pending_exceptions);
  80
  81         for (i = 0; i < 32; i += 4) {
  82                 printk("gpr%02d: %08lx %08lx %08lx %08lx\n", i,
  83                        kvmppc_get_gpr(vcpu, i),
  84                        kvmppc_get_gpr(vcpu, i+1),
  85                        kvmppc_get_gpr(vcpu, i+2),
  86                        kvmppc_get_gpr(vcpu, i+3));
  87         }
  88 }
  89
  90 #ifdef CONFIG_SPE
  91 void kvmppc_vcpu_disable_spe(struct kvm_vcpu *vcpu)
  92 {
  93         preempt_disable();
  94         enable_kernel_spe();
  95         kvmppc_save_guest_spe(vcpu);
  96         vcpu->arch.shadow_msr &= ~MSR_SPE;
  97         preempt_enable();
  98 }
  99
 100 static void kvmppc_vcpu_enable_spe(struct kvm_vcpu *vcpu)
 101 {
 102         preempt_disable();
 103         enable_kernel_spe();
 104         kvmppc_load_guest_spe(vcpu);
 105         vcpu->arch.shadow_msr |= MSR_SPE;
 106         preempt_enable();
 107 }
 108
 109 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
 110 {
 111         if (vcpu->arch.shared->msr & MSR_SPE) {
 112                 if (!(vcpu->arch.shadow_msr & MSR_SPE))
 113                         kvmppc_vcpu_enable_spe(vcpu);
 114         } else if (vcpu->arch.shadow_msr & MSR_SPE) {
 115                 kvmppc_vcpu_disable_spe(vcpu);
 116         }
 117 }
 118 #else
 119 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
 120 {
 121 }
 122 #endif
 123
 124 /*
 125  * Helper function for "full" MSR writes.  No need to call this if only
 126  * EE/CE/ME/DE/RI are changing.
 127  */
 128 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr)
 129 {
 130         u32 old_msr = vcpu->arch.shared->msr;
 131
 132 #ifdef CONFIG_KVM_BOOKE_HV
 133         new_msr |= MSR_GS;
 134 #endif
 135
 136         vcpu->arch.shared->msr = new_msr;
 137
 138         kvmppc_mmu_msr_notify(vcpu, old_msr);
 139         kvmppc_vcpu_sync_spe(vcpu);
 140 }
 141
 142 static void kvmppc_booke_queue_irqprio(struct kvm_vcpu *vcpu,
 143                                        unsigned int priority)
 144 {
 145         set_bit(priority, &vcpu->arch.pending_exceptions);
 146 }
 147
 148 static void kvmppc_core_queue_dtlb_miss(struct kvm_vcpu *vcpu,
 149                                         ulong dear_flags, ulong esr_flags)
 150 {
 151         vcpu->arch.queued_dear = dear_flags;
 152         vcpu->arch.queued_esr = esr_flags;
 153         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DTLB_MISS);
 154 }
 155
 156 static void kvmppc_core_queue_data_storage(struct kvm_vcpu *vcpu,
 157                                            ulong dear_flags, ulong esr_flags)
 158 {
 159         vcpu->arch.queued_dear = dear_flags;
 160         vcpu->arch.queued_esr = esr_flags;
 161         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DATA_STORAGE);
 162 }
 163
 164 static void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu,
 165                                            ulong esr_flags)
 166 {
 167         vcpu->arch.queued_esr = esr_flags;
 168         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_INST_STORAGE);
 169 }
 170
 171 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong esr_flags)
 172 {
 173         vcpu->arch.queued_esr = esr_flags;
 174         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_PROGRAM);
 175 }
 176
 177 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
 178 {
 179         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DECREMENTER);
 180 }
 181
 182 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
 183 {
 184         return test_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
 185 }
 186
 187 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
 188 {
 189         clear_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
 190 }
 191
 192 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
 193                                 struct kvm_interrupt *irq)
 194 {
 195         unsigned int prio = BOOKE_IRQPRIO_EXTERNAL;
 196
 197         if (irq->irq == KVM_INTERRUPT_SET_LEVEL)
 198                 prio = BOOKE_IRQPRIO_EXTERNAL_LEVEL;
 199
 200         kvmppc_booke_queue_irqprio(vcpu, prio);
 201 }
 202
 203 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu,
 204                                   struct kvm_interrupt *irq)
 205 {
 206         clear_bit(BOOKE_IRQPRIO_EXTERNAL, &vcpu->arch.pending_exceptions);
 207         clear_bit(BOOKE_IRQPRIO_EXTERNAL_LEVEL, &vcpu->arch.pending_exceptions);
 208 }
 209
 210 static void set_guest_srr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
 211 {
 212 #ifdef CONFIG_KVM_BOOKE_HV
 213         mtspr(SPRN_GSRR0, srr0);
 214         mtspr(SPRN_GSRR1, srr1);
 215 #else
 216         vcpu->arch.shared->srr0 = srr0;
 217         vcpu->arch.shared->srr1 = srr1;
 218 #endif
 219 }
 220
 221 static void set_guest_csrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
 222 {
 223         vcpu->arch.csrr0 = srr0;
 224         vcpu->arch.csrr1 = srr1;
 225 }
 226
 227 static void set_guest_dsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
 228 {
 229         if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC)) {
 230                 vcpu->arch.dsrr0 = srr0;
 231                 vcpu->arch.dsrr1 = srr1;
 232         } else {
 233                 set_guest_csrr(vcpu, srr0, srr1);
 234         }
 235 }
 236
 237 static void set_guest_mcsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
 238 {
 239         vcpu->arch.mcsrr0 = srr0;
 240         vcpu->arch.mcsrr1 = srr1;
 241 }
 242
 243 static unsigned long get_guest_dear(struct kvm_vcpu *vcpu)
 244 {
 245 #ifdef CONFIG_KVM_BOOKE_HV
 246         return mfspr(SPRN_GDEAR);
 247 #else
 248         return vcpu->arch.shared->dar;
 249 #endif
 250 }
 251
 252 static void set_guest_dear(struct kvm_vcpu *vcpu, unsigned long dear)
 253 {
 254 #ifdef CONFIG_KVM_BOOKE_HV
 255         mtspr(SPRN_GDEAR, dear);
 256 #else
 257         vcpu->arch.shared->dar = dear;
 258 #endif
 259 }
 260
 261 static unsigned long get_guest_esr(struct kvm_vcpu *vcpu)
 262 {
 263 #ifdef CONFIG_KVM_BOOKE_HV
 264         return mfspr(SPRN_GESR);
 265 #else
 266         return vcpu->arch.shared->esr;
 267 #endif
 268 }
 269
 270 static void set_guest_esr(struct kvm_vcpu *vcpu, u32 esr)
 271 {
 272 #ifdef CONFIG_KVM_BOOKE_HV
 273         mtspr(SPRN_GESR, esr);
 274 #else
 275         vcpu->arch.shared->esr = esr;
 276 #endif
 277 }
 278
 279 /* Deliver the interrupt of the corresponding priority, if possible. */
 280 static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu *vcpu,
 281                                         unsigned int priority)
 282 {
 283         int allowed = 0;
 284         ulong msr_mask = 0;
 285         bool update_esr = false, update_dear = false;
 286         ulong crit_raw = vcpu->arch.shared->critical;
 287         ulong crit_r1 = kvmppc_get_gpr(vcpu, 1);
 288         bool crit;
 289         bool keep_irq = false;
 290         enum int_class int_class;
 291
 292         /* Truncate crit indicators in 32 bit mode */
 293         if (!(vcpu->arch.shared->msr & MSR_SF)) {
 294                 crit_raw &= 0xffffffff;
 295                 crit_r1 &= 0xffffffff;
 296         }
 297
 298         /* Critical section when crit == r1 */
 299         crit = (crit_raw == crit_r1);
 300         /* ... and we're in supervisor mode */
 301         crit = crit && !(vcpu->arch.shared->msr & MSR_PR);
 302
 303         if (priority == BOOKE_IRQPRIO_EXTERNAL_LEVEL) {
 304                 priority = BOOKE_IRQPRIO_EXTERNAL;
 305                 keep_irq = true;
 306         }
 307
 308         switch (priority) {
 309         case BOOKE_IRQPRIO_DTLB_MISS:
 310         case BOOKE_IRQPRIO_DATA_STORAGE:
 311                 update_dear = true;
 312                 /* fall through */
 313         case BOOKE_IRQPRIO_INST_STORAGE:
 314         case BOOKE_IRQPRIO_PROGRAM:
 315                 update_esr = true;
 316                 /* fall through */
 317         case BOOKE_IRQPRIO_ITLB_MISS:
 318         case BOOKE_IRQPRIO_SYSCALL:
 319         case BOOKE_IRQPRIO_FP_UNAVAIL:
 320         case BOOKE_IRQPRIO_SPE_UNAVAIL:
 321         case BOOKE_IRQPRIO_SPE_FP_DATA:
 322         case BOOKE_IRQPRIO_SPE_FP_ROUND:
 323         case BOOKE_IRQPRIO_AP_UNAVAIL:
 324         case BOOKE_IRQPRIO_ALIGNMENT:
 325                 allowed = 1;
 326                 msr_mask = MSR_CE | MSR_ME | MSR_DE;
 327                 int_class = INT_CLASS_NONCRIT;
 328                 break;
 329         case BOOKE_IRQPRIO_CRITICAL:
 330         case BOOKE_IRQPRIO_DBELL_CRIT:
 331                 allowed = vcpu->arch.shared->msr & MSR_CE;
 332                 allowed = allowed && !crit;
 333                 msr_mask = MSR_ME;
 334                 int_class = INT_CLASS_CRIT;
 335                 break;
 336         case BOOKE_IRQPRIO_MACHINE_CHECK:
 337                 allowed = vcpu->arch.shared->msr & MSR_ME;
 338                 allowed = allowed && !crit;
 339                 int_class = INT_CLASS_MC;
 340                 break;
 341         case BOOKE_IRQPRIO_DECREMENTER:
 342         case BOOKE_IRQPRIO_FIT:
 343                 keep_irq = true;
 344                 /* fall through */
 345         case BOOKE_IRQPRIO_EXTERNAL:
 346         case BOOKE_IRQPRIO_DBELL:
 347                 allowed = vcpu->arch.shared->msr & MSR_EE;
 348                 allowed = allowed && !crit;
 349                 msr_mask = MSR_CE | MSR_ME | MSR_DE;
 350                 int_class = INT_CLASS_NONCRIT;
 351                 break;
 352         case BOOKE_IRQPRIO_DEBUG:
 353                 allowed = vcpu->arch.shared->msr & MSR_DE;
 354                 allowed = allowed && !crit;
 355                 msr_mask = MSR_ME;
 356                 int_class = INT_CLASS_CRIT;
 357                 break;
 358         }
 359
 360         if (allowed) {
 361                 switch (int_class) {
 362                 case INT_CLASS_NONCRIT:
 363                         set_guest_srr(vcpu, vcpu->arch.pc,
 364                                       vcpu->arch.shared->msr);
 365                         break;
 366                 case INT_CLASS_CRIT:
 367                         set_guest_csrr(vcpu, vcpu->arch.pc,
 368                                        vcpu->arch.shared->msr);
 369                         break;
 370                 case INT_CLASS_DBG:
 371                         set_guest_dsrr(vcpu, vcpu->arch.pc,
 372                                        vcpu->arch.shared->msr);
 373                         break;
 374                 case INT_CLASS_MC:
 375                         set_guest_mcsrr(vcpu, vcpu->arch.pc,
 376                                         vcpu->arch.shared->msr);
 377                         break;
 378                 }
 379
 380                 vcpu->arch.pc = vcpu->arch.ivpr | vcpu->arch.ivor[priority];
 381                 if (update_esr == true)
 382                         set_guest_esr(vcpu, vcpu->arch.queued_esr);
 383                 if (update_dear == true)
 384                         set_guest_dear(vcpu, vcpu->arch.queued_dear);
 385                 kvmppc_set_msr(vcpu, vcpu->arch.shared->msr & msr_mask);
 386
 387                 if (!keep_irq)
 388                         clear_bit(priority, &vcpu->arch.pending_exceptions);
 389         }
 390
 391 #ifdef CONFIG_KVM_BOOKE_HV
 392         /*
 393          * If an interrupt is pending but masked, raise a guest doorbell
 394          * so that we are notified when the guest enables the relevant
 395          * MSR bit.
 396          */
 397         if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_EE)
 398                 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_NONCRIT);
 399         if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_CE)
 400                 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_CRIT);
 401         if (vcpu->arch.pending_exceptions & BOOKE_IRQPRIO_MACHINE_CHECK)
 402                 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_MC);
 403 #endif
 404
 405         return allowed;
 406 }
 407
 408 static void update_timer_ints(struct kvm_vcpu *vcpu)
 409 {
 410         if ((vcpu->arch.tcr & TCR_DIE) && (vcpu->arch.tsr & TSR_DIS))
 411                 kvmppc_core_queue_dec(vcpu);
 412         else
 413                 kvmppc_core_dequeue_dec(vcpu);
 414 }
 415
 416 static void kvmppc_core_check_exceptions(struct kvm_vcpu *vcpu)
 417 {
 418         unsigned long *pending = &vcpu->arch.pending_exceptions;
 419         unsigned int priority;
 420
 421         if (vcpu->requests) {
 422                 if (kvm_check_request(KVM_REQ_PENDING_TIMER, vcpu)) {
 423                         smp_mb();
 424                         update_timer_ints(vcpu);
 425                 }
 426         }
 427
 428         priority = __ffs(*pending);
 429         while (priority < BOOKE_IRQPRIO_MAX) {
 430                 if (kvmppc_booke_irqprio_deliver(vcpu, priority))
 431                         break;
 432
 433                 priority = find_next_bit(pending,
 434                                          BITS_PER_BYTE * sizeof(*pending),
 435                                          priority + 1);
 436         }
 437
 438         /* Tell the guest about our interrupt status */
 439         vcpu->arch.shared->int_pending = !!*pending;
 440 }
 441
 442 /* Check pending exceptions and deliver one, if possible. */
 443 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
 444 {
 445         int r = 0;
 446         WARN_ON_ONCE(!irqs_disabled());
 447
 448         kvmppc_core_check_exceptions(vcpu);
 449
 450         if (vcpu->arch.shared->msr & MSR_WE) {
 451                 local_irq_enable();
 452                 kvm_vcpu_block(vcpu);
 453                 clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
 454                 local_irq_disable();
 455
 456                 kvmppc_set_exit_type(vcpu, EMULATED_MTMSRWE_EXITS);
 457                 r = 1;
 458         };
 459
 460         return r;
 461 }
 462
 463 /*
 464  * Common checks before entering the guest world.  Call with interrupts
 465  * disabled.
 466  *
 467  * returns !0 if a signal is pending and check_signal is true
 468  */
 469 static int kvmppc_prepare_to_enter(struct kvm_vcpu *vcpu)
 470 {
 471         int r = 0;
 472
 473         WARN_ON_ONCE(!irqs_disabled());
 474         while (true) {
 475                 if (need_resched()) {
 476                         local_irq_enable();
 477                         cond_resched();
 478                         local_irq_disable();
 479                         continue;
 480                 }
 481
 482                 if (signal_pending(current)) {
 483                         r = 1;
 484                         break;
 485                 }
 486
 487                 if (kvmppc_core_prepare_to_enter(vcpu)) {
 488                         /* interrupts got enabled in between, so we
 489                            are back at square 1 */
 490                         continue;
 491                 }
 492
 493                 break;
 494         }
 495
 496         return r;
 497 }
 498
 499 int kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
 500 {
 501         int ret;
 502 #ifdef CONFIG_PPC_FPU
 503         unsigned int fpscr;
 504         int fpexc_mode;
 505         u64 fpr[32];
 506 #endif
 507
 508         if (!vcpu->arch.sane) {
 509                 kvm_run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
 510                 return -EINVAL;
 511         }
 512
 513         local_irq_disable();
 514         if (kvmppc_prepare_to_enter(vcpu)) {
 515                 kvm_run->exit_reason = KVM_EXIT_INTR;
 516                 ret = -EINTR;
 517                 goto out;
 518         }
 519
 520         kvm_guest_enter();
 521
 522 #ifdef CONFIG_PPC_FPU
 523         /* Save userspace FPU state in stack */
 524         enable_kernel_fp();
 525         memcpy(fpr, current->thread.fpr, sizeof(current->thread.fpr));
 526         fpscr = current->thread.fpscr.val;
 527         fpexc_mode = current->thread.fpexc_mode;
 528
 529         /* Restore guest FPU state to thread */
 530         memcpy(current->thread.fpr, vcpu->arch.fpr, sizeof(vcpu->arch.fpr));
 531         current->thread.fpscr.val = vcpu->arch.fpscr;
 532
 533         /*
 534          * Since we can't trap on MSR_FP in GS-mode, we consider the guest
 535          * as always using the FPU.  Kernel usage of FP (via
 536          * enable_kernel_fp()) in this thread must not occur while
 537          * vcpu->fpu_active is set.
 538          */
 539         vcpu->fpu_active = 1;
 540
 541         kvmppc_load_guest_fp(vcpu);
 542 #endif
 543
 544         ret = __kvmppc_vcpu_run(kvm_run, vcpu);
 545
 546 #ifdef CONFIG_PPC_FPU
 547         kvmppc_save_guest_fp(vcpu);
 548
 549         vcpu->fpu_active = 0;
 550
 551         /* Save guest FPU state from thread */
 552         memcpy(vcpu->arch.fpr, current->thread.fpr, sizeof(vcpu->arch.fpr));
 553         vcpu->arch.fpscr = current->thread.fpscr.val;
 554
 555         /* Restore userspace FPU state from stack */
 556         memcpy(current->thread.fpr, fpr, sizeof(current->thread.fpr));
 557         current->thread.fpscr.val = fpscr;
 558         current->thread.fpexc_mode = fpexc_mode;
 559 #endif
 560
 561         kvm_guest_exit();
 562
 563 out:
 564         local_irq_enable();
 565         return ret;
 566 }
 567
 568 static int emulation_exit(struct kvm_run *run, struct kvm_vcpu *vcpu)
 569 {
 570         enum emulation_result er;
 571
 572         er = kvmppc_emulate_instruction(run, vcpu);
 573         switch (er) {
 574         case EMULATE_DONE:
 575                 /* don't overwrite subtypes, just account kvm_stats */
 576                 kvmppc_account_exit_stat(vcpu, EMULATED_INST_EXITS);
 577                 /* Future optimization: only reload non-volatiles if
 578                  * they were actually modified by emulation. */
 579                 return RESUME_GUEST_NV;
 580
 581         case EMULATE_DO_DCR:
 582                 run->exit_reason = KVM_EXIT_DCR;
 583                 return RESUME_HOST;
 584
 585         case EMULATE_FAIL:
 586                 printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
 587                        __func__, vcpu->arch.pc, vcpu->arch.last_inst);
 588                 /* For debugging, encode the failing instruction and
 589                  * report it to userspace. */
 590                 run->hw.hardware_exit_reason = ~0ULL << 32;
 591                 run->hw.hardware_exit_reason |= vcpu->arch.last_inst;
 592                 kvmppc_core_queue_program(vcpu, ESR_PIL);
 593                 return RESUME_HOST;
 594
 595         default:
 596                 BUG();
 597         }
 598 }
 599
 600 static void kvmppc_fill_pt_regs(struct pt_regs *regs)
 601 {
 602         ulong r1, ip, msr, lr;
 603
 604         asm("mr %0, 1" : "=r"(r1));
 605         asm("mflr %0" : "=r"(lr));
 606         asm("mfmsr %0" : "=r"(msr));
 607         asm("bl 1f; 1: mflr %0" : "=r"(ip));
 608
 609         memset(regs, 0, sizeof(*regs));
 610         regs->gpr[1] = r1;
 611         regs->nip = ip;
 612         regs->msr = msr;
 613         regs->link = lr;
 614 }
 615
 616 /*
 617  * For interrupts needed to be handled by host interrupt handlers,
 618  * corresponding host handler are called from here in similar way
 619  * (but not exact) as they are called from low level handler
 620  * (such as from arch/powerpc/kernel/head_fsl_booke.S).
 621  */
 622 static void kvmppc_restart_interrupt(struct kvm_vcpu *vcpu,
 623                                      unsigned int exit_nr)
 624 {
 625         struct pt_regs regs;
 626
 627         switch (exit_nr) {
 628         case BOOKE_INTERRUPT_EXTERNAL:
 629                 kvmppc_fill_pt_regs(&regs);
 630                 do_IRQ(&regs);
 631                 break;
 632         case BOOKE_INTERRUPT_DECREMENTER:
 633                 kvmppc_fill_pt_regs(&regs);
 634                 timer_interrupt(&regs);
 635                 break;
 636 #if defined(CONFIG_PPC_FSL_BOOK3E) || defined(CONFIG_PPC_BOOK3E_64)
 637         case BOOKE_INTERRUPT_DOORBELL:
 638                 kvmppc_fill_pt_regs(&regs);
 639                 doorbell_exception(&regs);
 640                 break;
 641 #endif
 642         case BOOKE_INTERRUPT_MACHINE_CHECK:
 643                 /* FIXME */
 644                 break;
 645         case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
 646                 kvmppc_fill_pt_regs(&regs);
 647                 performance_monitor_exception(&regs);
 648                 break;
 649         case BOOKE_INTERRUPT_WATCHDOG:
 650                 kvmppc_fill_pt_regs(&regs);
 651 #ifdef CONFIG_BOOKE_WDT
 652                 WatchdogException(&regs);
 653 #else
 654                 unknown_exception(&regs);
 655 #endif
 656                 break;
 657         case BOOKE_INTERRUPT_CRITICAL:
 658                 unknown_exception(&regs);
 659                 break;
 660         }
 661 }
 662
 663 /**
 664  * kvmppc_handle_exit
 665  *
 666  * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
 667  */
 668 int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
 669                        unsigned int exit_nr)
 670 {
 671         int r = RESUME_HOST;
 672
 673         /* update before a new last_exit_type is rewritten */
 674         kvmppc_update_timing_stats(vcpu);
 675
 676         /* restart interrupts if they were meant for the host */
 677         kvmppc_restart_interrupt(vcpu, exit_nr);
 678
 679         local_irq_enable();
 680
 681         trace_kvm_exit(exit_nr, vcpu);
 682
 683         run->exit_reason = KVM_EXIT_UNKNOWN;
 684         run->ready_for_interrupt_injection = 1;
 685
 686         switch (exit_nr) {
 687         case BOOKE_INTERRUPT_MACHINE_CHECK:
 688                 printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR));
 689                 kvmppc_dump_vcpu(vcpu);
 690                 /* For debugging, send invalid exit reason to user space */
 691                 run->hw.hardware_exit_reason = ~1ULL << 32;
 692                 run->hw.hardware_exit_reason |= mfspr(SPRN_MCSR);
 693                 r = RESUME_HOST;
 694                 break;
 695
 696         case BOOKE_INTERRUPT_EXTERNAL:
 697                 kvmppc_account_exit(vcpu, EXT_INTR_EXITS);
 698                 r = RESUME_GUEST;
 699                 break;
 700
 701         case BOOKE_INTERRUPT_DECREMENTER:
 702                 kvmppc_account_exit(vcpu, DEC_EXITS);
 703                 r = RESUME_GUEST;
 704                 break;
 705
 706         case BOOKE_INTERRUPT_WATCHDOG:
 707                 r = RESUME_GUEST;
 708                 break;
 709
 710         case BOOKE_INTERRUPT_DOORBELL:
 711                 kvmppc_account_exit(vcpu, DBELL_EXITS);
 712                 r = RESUME_GUEST;
 713                 break;
 714
 715         case BOOKE_INTERRUPT_GUEST_DBELL_CRIT:
 716                 kvmppc_account_exit(vcpu, GDBELL_EXITS);
 717
 718                 /*
 719                  * We are here because there is a pending guest interrupt
 720                  * which could not be delivered as MSR_CE or MSR_ME was not
 721                  * set.  Once we break from here we will retry delivery.
 722                  */
 723                 r = RESUME_GUEST;
 724                 break;
 725
 726         case BOOKE_INTERRUPT_GUEST_DBELL:
 727                 kvmppc_account_exit(vcpu, GDBELL_EXITS);
 728
 729                 /*
 730                  * We are here because there is a pending guest interrupt
 731                  * which could not be delivered as MSR_EE was not set.  Once
 732                  * we break from here we will retry delivery.
 733                  */
 734                 r = RESUME_GUEST;
 735                 break;
 736
 737         case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
 738                 r = RESUME_GUEST;
 739                 break;
 740
 741         case BOOKE_INTERRUPT_HV_PRIV:
 742                 r = emulation_exit(run, vcpu);
 743                 break;
 744
 745         case BOOKE_INTERRUPT_PROGRAM:
 746                 if (vcpu->arch.shared->msr & (MSR_PR | MSR_GS)) {
 747                         /*
 748                          * Program traps generated by user-level software must
 749                          * be handled by the guest kernel.
 750                          *
 751                          * In GS mode, hypervisor privileged instructions trap
 752                          * on BOOKE_INTERRUPT_HV_PRIV, not here, so these are
 753                          * actual program interrupts, handled by the guest.
 754                          */
 755                         kvmppc_core_queue_program(vcpu, vcpu->arch.fault_esr);
 756                         r = RESUME_GUEST;
 757                         kvmppc_account_exit(vcpu, USR_PR_INST);
 758                         break;
 759                 }
 760
 761                 r = emulation_exit(run, vcpu);
 762                 break;
 763
 764         case BOOKE_INTERRUPT_FP_UNAVAIL:
 765                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
 766                 kvmppc_account_exit(vcpu, FP_UNAVAIL);
 767                 r = RESUME_GUEST;
 768                 break;
 769
 770 #ifdef CONFIG_SPE
 771         case BOOKE_INTERRUPT_SPE_UNAVAIL: {
 772                 if (vcpu->arch.shared->msr & MSR_SPE)
 773                         kvmppc_vcpu_enable_spe(vcpu);
 774                 else
 775                         kvmppc_booke_queue_irqprio(vcpu,
 776                                                    BOOKE_IRQPRIO_SPE_UNAVAIL);
 777                 r = RESUME_GUEST;
 778                 break;
 779         }
 780
 781         case BOOKE_INTERRUPT_SPE_FP_DATA:
 782                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_DATA);
 783                 r = RESUME_GUEST;
 784                 break;
 785
 786         case BOOKE_INTERRUPT_SPE_FP_ROUND:
 787                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_ROUND);
 788                 r = RESUME_GUEST;
 789                 break;
 790 #else
 791         case BOOKE_INTERRUPT_SPE_UNAVAIL:
 792                 /*
 793                  * Guest wants SPE, but host kernel doesn't support it.  Send
 794                  * an "unimplemented operation" program check to the guest.
 795                  */
 796                 kvmppc_core_queue_program(vcpu, ESR_PUO | ESR_SPV);
 797                 r = RESUME_GUEST;
 798                 break;
 799
 800         /*
 801          * These really should never happen without CONFIG_SPE,
 802          * as we should never enable the real MSR[SPE] in the guest.
 803          */
 804         case BOOKE_INTERRUPT_SPE_FP_DATA:
 805         case BOOKE_INTERRUPT_SPE_FP_ROUND:
 806                 printk(KERN_CRIT "%s: unexpected SPE interrupt %u at %08lx\n",
 807                        __func__, exit_nr, vcpu->arch.pc);
 808                 run->hw.hardware_exit_reason = exit_nr;
 809                 r = RESUME_HOST;
 810                 break;
 811 #endif
 812
 813         case BOOKE_INTERRUPT_DATA_STORAGE:
 814                 kvmppc_core_queue_data_storage(vcpu, vcpu->arch.fault_dear,
 815                                                vcpu->arch.fault_esr);
 816                 kvmppc_account_exit(vcpu, DSI_EXITS);
 817                 r = RESUME_GUEST;
 818                 break;
 819
 820         case BOOKE_INTERRUPT_INST_STORAGE:
 821                 kvmppc_core_queue_inst_storage(vcpu, vcpu->arch.fault_esr);
 822                 kvmppc_account_exit(vcpu, ISI_EXITS);
 823                 r = RESUME_GUEST;
 824                 break;
 825
 826 #ifdef CONFIG_KVM_BOOKE_HV
 827         case BOOKE_INTERRUPT_HV_SYSCALL:
 828                 if (!(vcpu->arch.shared->msr & MSR_PR)) {
 829                         kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
 830                 } else {
 831                         /*
 832                          * hcall from guest userspace -- send privileged
 833                          * instruction program check.
 834                          */
 835                         kvmppc_core_queue_program(vcpu, ESR_PPR);
 836                 }
 837
 838                 r = RESUME_GUEST;
 839                 break;
 840 #else
 841         case BOOKE_INTERRUPT_SYSCALL:
 842                 if (!(vcpu->arch.shared->msr & MSR_PR) &&
 843                     (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
 844                         /* KVM PV hypercalls */
 845                         kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
 846                         r = RESUME_GUEST;
 847                 } else {
 848                         /* Guest syscalls */
 849                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SYSCALL);
 850                 }
 851                 kvmppc_account_exit(vcpu, SYSCALL_EXITS);
 852                 r = RESUME_GUEST;
 853                 break;
 854 #endif
 855
 856         case BOOKE_INTERRUPT_DTLB_MISS: {
 857                 unsigned long eaddr = vcpu->arch.fault_dear;
 858                 int gtlb_index;
 859                 gpa_t gpaddr;
 860                 gfn_t gfn;
 861
 862 #ifdef CONFIG_KVM_E500V2
 863                 if (!(vcpu->arch.shared->msr & MSR_PR) &&
 864                     (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
 865                         kvmppc_map_magic(vcpu);
 866                         kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
 867                         r = RESUME_GUEST;
 868
 869                         break;
 870                 }
 871 #endif
 872
 873                 /* Check the guest TLB. */
 874                 gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
 875                 if (gtlb_index < 0) {
 876                         /* The guest didn't have a mapping for it. */
 877                         kvmppc_core_queue_dtlb_miss(vcpu,
 878                                                     vcpu->arch.fault_dear,
 879                                                     vcpu->arch.fault_esr);
 880                         kvmppc_mmu_dtlb_miss(vcpu);
 881                         kvmppc_account_exit(vcpu, DTLB_REAL_MISS_EXITS);
 882                         r = RESUME_GUEST;
 883                         break;
 884                 }
 885
 886                 gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
 887                 gfn = gpaddr >> PAGE_SHIFT;
 888
 889                 if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
 890                         /* The guest TLB had a mapping, but the shadow TLB
 891                          * didn't, and it is RAM. This could be because:
 892                          * a) the entry is mapping the host kernel, or
 893                          * b) the guest used a large mapping which we're faking
 894                          * Either way, we need to satisfy the fault without
 895                          * invoking the guest. */
 896                         kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
 897                         kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
 898                         r = RESUME_GUEST;
 899                 } else {
 900                         /* Guest has mapped and accessed a page which is not
 901                          * actually RAM. */
 902                         vcpu->arch.paddr_accessed = gpaddr;
 903                         vcpu->arch.vaddr_accessed = eaddr;
 904                         r = kvmppc_emulate_mmio(run, vcpu);
 905                         kvmppc_account_exit(vcpu, MMIO_EXITS);
 906                 }
 907
 908                 break;
 909         }
 910
 911         case BOOKE_INTERRUPT_ITLB_MISS: {
 912                 unsigned long eaddr = vcpu->arch.pc;
 913                 gpa_t gpaddr;
 914                 gfn_t gfn;
 915                 int gtlb_index;
 916
 917                 r = RESUME_GUEST;
 918
 919                 /* Check the guest TLB. */
 920                 gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
 921                 if (gtlb_index < 0) {
 922                         /* The guest didn't have a mapping for it. */
 923                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
 924                         kvmppc_mmu_itlb_miss(vcpu);
 925                         kvmppc_account_exit(vcpu, ITLB_REAL_MISS_EXITS);
 926                         break;
 927                 }
 928
 929                 kvmppc_account_exit(vcpu, ITLB_VIRT_MISS_EXITS);
 930
 931                 gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
 932                 gfn = gpaddr >> PAGE_SHIFT;
 933
 934                 if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
 935                         /* The guest TLB had a mapping, but the shadow TLB
 936                          * didn't. This could be because:
 937                          * a) the entry is mapping the host kernel, or
 938                          * b) the guest used a large mapping which we're faking
 939                          * Either way, we need to satisfy the fault without
 940                          * invoking the guest. */
 941                         kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
 942                 } else {
 943                         /* Guest mapped and leaped at non-RAM! */
 944                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_MACHINE_CHECK);
 945                 }
 946
 947                 break;
 948         }
 949
 950         case BOOKE_INTERRUPT_DEBUG: {
 951                 u32 dbsr;
 952
 953                 vcpu->arch.pc = mfspr(SPRN_CSRR0);
 954
 955                 /* clear IAC events in DBSR register */
 956                 dbsr = mfspr(SPRN_DBSR);
 957                 dbsr &= DBSR_IAC1 | DBSR_IAC2 | DBSR_IAC3 | DBSR_IAC4;
 958                 mtspr(SPRN_DBSR, dbsr);
 959
 960                 run->exit_reason = KVM_EXIT_DEBUG;
 961                 kvmppc_account_exit(vcpu, DEBUG_EXITS);
 962                 r = RESUME_HOST;
 963                 break;
 964         }
 965
 966         default:
 967                 printk(KERN_EMERG "exit_nr %d\n", exit_nr);
 968                 BUG();
 969         }
 970
 971         /*
 972          * To avoid clobbering exit_reason, only check for signals if we
 973          * aren't already exiting to userspace for some other reason.
 974          */
 975         if (!(r & RESUME_HOST)) {
 976                 local_irq_disable();
 977                 if (kvmppc_prepare_to_enter(vcpu)) {
 978                         run->exit_reason = KVM_EXIT_INTR;
 979                         r = (-EINTR << 2) | RESUME_HOST | (r & RESUME_FLAG_NV);
 980                         kvmppc_account_exit(vcpu, SIGNAL_EXITS);
 981                 }
 982         }
 983
 984         return r;
 985 }
 986
 987 /* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
 988 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
 989 {
 990         int i;
 991         int r;
 992
 993         vcpu->arch.pc = 0;
 994         vcpu->arch.shared->pir = vcpu->vcpu_id;
 995         kvmppc_set_gpr(vcpu, 1, (16<<20) - 8); /* -8 for the callee-save LR slot */
 996         kvmppc_set_msr(vcpu, 0);
 997
 998 #ifndef CONFIG_KVM_BOOKE_HV
 999         vcpu->arch.shadow_msr = MSR_USER | MSR_DE | MSR_IS | MSR_DS;
1000         vcpu->arch.shadow_pid = 1;
1001         vcpu->arch.shared->msr = 0;
1002 #endif
1003
1004         /* Eye-catching numbers so we know if the guest takes an interrupt
1005          * before it's programmed its own IVPR/IVORs. */
1006         vcpu->arch.ivpr = 0x55550000;
1007         for (i = 0; i < BOOKE_IRQPRIO_MAX; i++)
1008                 vcpu->arch.ivor[i] = 0x7700 | i * 4;
1009
1010         kvmppc_init_timing_stats(vcpu);
1011
1012         r = kvmppc_core_vcpu_setup(vcpu);
1013         kvmppc_sanity_check(vcpu);
1014         return r;
1015 }
1016
1017 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1018 {
1019         int i;
1020
1021         regs->pc = vcpu->arch.pc;
1022         regs->cr = kvmppc_get_cr(vcpu);
1023         regs->ctr = vcpu->arch.ctr;
1024         regs->lr = vcpu->arch.lr;
1025         regs->xer = kvmppc_get_xer(vcpu);
1026         regs->msr = vcpu->arch.shared->msr;
1027         regs->srr0 = vcpu->arch.shared->srr0;
1028         regs->srr1 = vcpu->arch.shared->srr1;
1029         regs->pid = vcpu->arch.pid;
1030         regs->sprg0 = vcpu->arch.shared->sprg0;
1031         regs->sprg1 = vcpu->arch.shared->sprg1;
1032         regs->sprg2 = vcpu->arch.shared->sprg2;
1033         regs->sprg3 = vcpu->arch.shared->sprg3;
1034         regs->sprg4 = vcpu->arch.shared->sprg4;
1035         regs->sprg5 = vcpu->arch.shared->sprg5;
1036         regs->sprg6 = vcpu->arch.shared->sprg6;
1037         regs->sprg7 = vcpu->arch.shared->sprg7;
1038
1039         for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1040                 regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
1041
1042         return 0;
1043 }
1044
1045 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1046 {
1047         int i;
1048
1049         vcpu->arch.pc = regs->pc;
1050         kvmppc_set_cr(vcpu, regs->cr);
1051         vcpu->arch.ctr = regs->ctr;
1052         vcpu->arch.lr = regs->lr;
1053         kvmppc_set_xer(vcpu, regs->xer);
1054         kvmppc_set_msr(vcpu, regs->msr);
1055         vcpu->arch.shared->srr0 = regs->srr0;
1056         vcpu->arch.shared->srr1 = regs->srr1;
1057         kvmppc_set_pid(vcpu, regs->pid);
1058         vcpu->arch.shared->sprg0 = regs->sprg0;
1059         vcpu->arch.shared->sprg1 = regs->sprg1;
1060         vcpu->arch.shared->sprg2 = regs->sprg2;
1061         vcpu->arch.shared->sprg3 = regs->sprg3;
1062         vcpu->arch.shared->sprg4 = regs->sprg4;
1063         vcpu->arch.shared->sprg5 = regs->sprg5;
1064         vcpu->arch.shared->sprg6 = regs->sprg6;
1065         vcpu->arch.shared->sprg7 = regs->sprg7;
1066
1067         for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1068                 kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
1069
1070         return 0;
1071 }
1072
1073 static void get_sregs_base(struct kvm_vcpu *vcpu,
1074                            struct kvm_sregs *sregs)
1075 {
1076         u64 tb = get_tb();
1077
1078         sregs->u.e.features |= KVM_SREGS_E_BASE;
1079
1080         sregs->u.e.csrr0 = vcpu->arch.csrr0;
1081         sregs->u.e.csrr1 = vcpu->arch.csrr1;
1082         sregs->u.e.mcsr = vcpu->arch.mcsr;
1083         sregs->u.e.esr = get_guest_esr(vcpu);
1084         sregs->u.e.dear = get_guest_dear(vcpu);
1085         sregs->u.e.tsr = vcpu->arch.tsr;
1086         sregs->u.e.tcr = vcpu->arch.tcr;
1087         sregs->u.e.dec = kvmppc_get_dec(vcpu, tb);
1088         sregs->u.e.tb = tb;
1089         sregs->u.e.vrsave = vcpu->arch.vrsave;
1090 }
1091
1092 static int set_sregs_base(struct kvm_vcpu *vcpu,
1093                           struct kvm_sregs *sregs)
1094 {
1095         if (!(sregs->u.e.features & KVM_SREGS_E_BASE))
1096                 return 0;
1097
1098         vcpu->arch.csrr0 = sregs->u.e.csrr0;
1099         vcpu->arch.csrr1 = sregs->u.e.csrr1;
1100         vcpu->arch.mcsr = sregs->u.e.mcsr;
1101         set_guest_esr(vcpu, sregs->u.e.esr);
1102         set_guest_dear(vcpu, sregs->u.e.dear);
1103         vcpu->arch.vrsave = sregs->u.e.vrsave;
1104         kvmppc_set_tcr(vcpu, sregs->u.e.tcr);
1105
1106         if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_DEC) {
1107                 vcpu->arch.dec = sregs->u.e.dec;
1108                 kvmppc_emulate_dec(vcpu);
1109         }
1110
1111         if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_TSR) {
1112                 vcpu->arch.tsr = sregs->u.e.tsr;
1113                 update_timer_ints(vcpu);
1114         }
1115
1116         return 0;
1117 }
1118
1119 static void get_sregs_arch206(struct kvm_vcpu *vcpu,
1120                               struct kvm_sregs *sregs)
1121 {
1122         sregs->u.e.features |= KVM_SREGS_E_ARCH206;
1123
1124         sregs->u.e.pir = vcpu->vcpu_id;
1125         sregs->u.e.mcsrr0 = vcpu->arch.mcsrr0;
1126         sregs->u.e.mcsrr1 = vcpu->arch.mcsrr1;
1127         sregs->u.e.decar = vcpu->arch.decar;
1128         sregs->u.e.ivpr = vcpu->arch.ivpr;
1129 }
1130
1131 static int set_sregs_arch206(struct kvm_vcpu *vcpu,
1132                              struct kvm_sregs *sregs)
1133 {
1134         if (!(sregs->u.e.features & KVM_SREGS_E_ARCH206))
1135                 return 0;
1136
1137         if (sregs->u.e.pir != vcpu->vcpu_id)
1138                 return -EINVAL;
1139
1140         vcpu->arch.mcsrr0 = sregs->u.e.mcsrr0;
1141         vcpu->arch.mcsrr1 = sregs->u.e.mcsrr1;
1142         vcpu->arch.decar = sregs->u.e.decar;
1143         vcpu->arch.ivpr = sregs->u.e.ivpr;
1144
1145         return 0;
1146 }
1147
1148 void kvmppc_get_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1149 {
1150         sregs->u.e.features |= KVM_SREGS_E_IVOR;
1151
1152         sregs->u.e.ivor_low[0] = vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL];
1153         sregs->u.e.ivor_low[1] = vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK];
1154         sregs->u.e.ivor_low[2] = vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE];
1155         sregs->u.e.ivor_low[3] = vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE];
1156         sregs->u.e.ivor_low[4] = vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL];
1157         sregs->u.e.ivor_low[5] = vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT];
1158         sregs->u.e.ivor_low[6] = vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM];
1159         sregs->u.e.ivor_low[7] = vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL];
1160         sregs->u.e.ivor_low[8] = vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL];
1161         sregs->u.e.ivor_low[9] = vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL];
1162         sregs->u.e.ivor_low[10] = vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER];
1163         sregs->u.e.ivor_low[11] = vcpu->arch.ivor[BOOKE_IRQPRIO_FIT];
1164         sregs->u.e.ivor_low[12] = vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG];
1165         sregs->u.e.ivor_low[13] = vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS];
1166         sregs->u.e.ivor_low[14] = vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS];
1167         sregs->u.e.ivor_low[15] = vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG];
1168 }
1169
1170 int kvmppc_set_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1171 {
1172         if (!(sregs->u.e.features & KVM_SREGS_E_IVOR))
1173                 return 0;
1174
1175         vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL] = sregs->u.e.ivor_low[0];
1176         vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK] = sregs->u.e.ivor_low[1];
1177         vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE] = sregs->u.e.ivor_low[2];
1178         vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE] = sregs->u.e.ivor_low[3];
1179         vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL] = sregs->u.e.ivor_low[4];
1180         vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT] = sregs->u.e.ivor_low[5];
1181         vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM] = sregs->u.e.ivor_low[6];
1182         vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL] = sregs->u.e.ivor_low[7];
1183         vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL] = sregs->u.e.ivor_low[8];
1184         vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL] = sregs->u.e.ivor_low[9];
1185         vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER] = sregs->u.e.ivor_low[10];
1186         vcpu->arch.ivor[BOOKE_IRQPRIO_FIT] = sregs->u.e.ivor_low[11];
1187         vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG] = sregs->u.e.ivor_low[12];
1188         vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS] = sregs->u.e.ivor_low[13];
1189         vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS] = sregs->u.e.ivor_low[14];
1190         vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG] = sregs->u.e.ivor_low[15];
1191
1192         return 0;
1193 }
1194
1195 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1196                                   struct kvm_sregs *sregs)
1197 {
1198         sregs->pvr = vcpu->arch.pvr;
1199
1200         get_sregs_base(vcpu, sregs);
1201         get_sregs_arch206(vcpu, sregs);
1202         kvmppc_core_get_sregs(vcpu, sregs);
1203         return 0;
1204 }
1205
1206 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1207                                   struct kvm_sregs *sregs)
1208 {
1209         int ret;
1210
1211         if (vcpu->arch.pvr != sregs->pvr)
1212                 return -EINVAL;
1213
1214         ret = set_sregs_base(vcpu, sregs);
1215         if (ret < 0)
1216                 return ret;
1217
1218         ret = set_sregs_arch206(vcpu, sregs);
1219         if (ret < 0)
1220                 return ret;
1221
1222         return kvmppc_core_set_sregs(vcpu, sregs);
1223 }
1224
1225 int kvm_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1226 {
1227         return -EINVAL;
1228 }
1229
1230 int kvm_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1231 {
1232         return -EINVAL;
1233 }
1234
1235 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1236 {
1237         return -ENOTSUPP;
1238 }
1239
1240 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1241 {
1242         return -ENOTSUPP;
1243 }
1244
1245 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1246                                   struct kvm_translation *tr)
1247 {
1248         int r;
1249
1250         r = kvmppc_core_vcpu_translate(vcpu, tr);
1251         return r;
1252 }
1253
1254 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
1255 {
1256         return -ENOTSUPP;
1257 }
1258
1259 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
1260                                       struct kvm_userspace_memory_region *mem)
1261 {
1262         return 0;
1263 }
1264
1265 void kvmppc_core_commit_memory_region(struct kvm *kvm,
1266                                 struct kvm_userspace_memory_region *mem)
1267 {
1268 }
1269
1270 void kvmppc_set_tcr(struct kvm_vcpu *vcpu, u32 new_tcr)
1271 {
1272         vcpu->arch.tcr = new_tcr;
1273         update_timer_ints(vcpu);
1274 }
1275
1276 void kvmppc_set_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1277 {
1278         set_bits(tsr_bits, &vcpu->arch.tsr);
1279         smp_wmb();
1280         kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
1281         kvm_vcpu_kick(vcpu);
1282 }
1283
1284 void kvmppc_clr_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1285 {
1286         clear_bits(tsr_bits, &vcpu->arch.tsr);
1287         update_timer_ints(vcpu);
1288 }
1289
1290 void kvmppc_decrementer_func(unsigned long data)
1291 {
1292         struct kvm_vcpu *vcpu = (struct kvm_vcpu *)data;
1293
1294         if (vcpu->arch.tcr & TCR_ARE) {
1295                 vcpu->arch.dec = vcpu->arch.decar;
1296                 kvmppc_emulate_dec(vcpu);
1297         }
1298
1299         kvmppc_set_tsr_bits(vcpu, TSR_DIS);
1300 }
1301
1302 void kvmppc_booke_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
1303 {
1304         current->thread.kvm_vcpu = vcpu;
1305 }
1306
1307 void kvmppc_booke_vcpu_put(struct kvm_vcpu *vcpu)
1308 {
1309         current->thread.kvm_vcpu = NULL;
1310 }
1311
1312 int __init kvmppc_booke_init(void)
1313 {
1314 #ifndef CONFIG_KVM_BOOKE_HV
1315         unsigned long ivor[16];
1316         unsigned long max_ivor = 0;
1317         int i;
1318
1319         /* We install our own exception handlers by hijacking IVPR. IVPR must
1320          * be 16-bit aligned, so we need a 64KB allocation. */
1321         kvmppc_booke_handlers = __get_free_pages(GFP_KERNEL | __GFP_ZERO,
1322                                                  VCPU_SIZE_ORDER);
1323         if (!kvmppc_booke_handlers)
1324                 return -ENOMEM;
1325
1326         /* XXX make sure our handlers are smaller than Linux's */
1327
1328         /* Copy our interrupt handlers to match host IVORs. That way we don't
1329          * have to swap the IVORs on every guest/host transition. */
1330         ivor[0] = mfspr(SPRN_IVOR0);
1331         ivor[1] = mfspr(SPRN_IVOR1);
1332         ivor[2] = mfspr(SPRN_IVOR2);
1333         ivor[3] = mfspr(SPRN_IVOR3);
1334         ivor[4] = mfspr(SPRN_IVOR4);
1335         ivor[5] = mfspr(SPRN_IVOR5);
1336         ivor[6] = mfspr(SPRN_IVOR6);
1337         ivor[7] = mfspr(SPRN_IVOR7);
1338         ivor[8] = mfspr(SPRN_IVOR8);
1339         ivor[9] = mfspr(SPRN_IVOR9);
1340         ivor[10] = mfspr(SPRN_IVOR10);
1341         ivor[11] = mfspr(SPRN_IVOR11);
1342         ivor[12] = mfspr(SPRN_IVOR12);
1343         ivor[13] = mfspr(SPRN_IVOR13);
1344         ivor[14] = mfspr(SPRN_IVOR14);
1345         ivor[15] = mfspr(SPRN_IVOR15);
1346
1347         for (i = 0; i < 16; i++) {
1348                 if (ivor[i] > max_ivor)
1349                         max_ivor = ivor[i];
1350
1351                 memcpy((void *)kvmppc_booke_handlers + ivor[i],
1352                        kvmppc_handlers_start + i * kvmppc_handler_len,
1353                        kvmppc_handler_len);
1354         }
1355         flush_icache_range(kvmppc_booke_handlers,
1356                            kvmppc_booke_handlers + max_ivor + kvmppc_handler_len);
1357 #endif /* !BOOKE_HV */
1358         return 0;
1359 }
1360
1361 void __exit kvmppc_booke_exit(void)
1362 {
1363         free_pages(kvmppc_booke_handlers, VCPU_SIZE_ORDER);
1364         kvm_exit();
1365 }