arch/powerpc/kvm/booke.c

   1 /*
   2  * This program is free software; you can redistribute it and/or modify
   3  * it under the terms of the GNU General Public License, version 2, as
   4  * published by the Free Software Foundation.
   5  *
   6  * This program is distributed in the hope that it will be useful,
   7  * but WITHOUT ANY WARRANTY; without even the implied warranty of
   8  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
   9  * GNU General Public License for more details.
  10  *
  11  * You should have received a copy of the GNU General Public License
  12  * along with this program; if not, write to the Free Software
  13  * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
  14  *
  15  * Copyright IBM Corp. 2007
  16  * Copyright 2010-2011 Freescale Semiconductor, Inc.
  17  *
  18  * Authors: Hollis Blanchard <hollisb@us.ibm.com>
  19  *          Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
  20  *          Scott Wood <scottwood@freescale.com>
  21  *          Varun Sethi <varun.sethi@freescale.com>
  22  */
  23
  24 #include <linux/errno.h>
  25 #include <linux/err.h>
  26 #include <linux/kvm_host.h>
  27 #include <linux/gfp.h>
  28 #include <linux/module.h>
  29 #include <linux/vmalloc.h>
  30 #include <linux/fs.h>
  31
  32 #include <asm/cputable.h>
  33 #include <asm/uaccess.h>
  34 #include <asm/kvm_ppc.h>
  35 #include <asm/cacheflush.h>
  36 #include <asm/dbell.h>
  37 #include <asm/hw_irq.h>
  38 #include <asm/irq.h>
  39 #include <asm/time.h>
  40
  41 #include "timing.h"
  42 #include "booke.h"
  43
  44 #define CREATE_TRACE_POINTS
  45 #include "trace_booke.h"
  46
  47 unsigned long kvmppc_booke_handlers;
  48
  49 #define VM_STAT(x) offsetof(struct kvm, stat.x), KVM_STAT_VM
  50 #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
  51
  52 struct kvm_stats_debugfs_item debugfs_entries[] = {
  53         { "mmio",       VCPU_STAT(mmio_exits) },
  54         { "dcr",        VCPU_STAT(dcr_exits) },
  55         { "sig",        VCPU_STAT(signal_exits) },
  56         { "itlb_r",     VCPU_STAT(itlb_real_miss_exits) },
  57         { "itlb_v",     VCPU_STAT(itlb_virt_miss_exits) },
  58         { "dtlb_r",     VCPU_STAT(dtlb_real_miss_exits) },
  59         { "dtlb_v",     VCPU_STAT(dtlb_virt_miss_exits) },
  60         { "sysc",       VCPU_STAT(syscall_exits) },
  61         { "isi",        VCPU_STAT(isi_exits) },
  62         { "dsi",        VCPU_STAT(dsi_exits) },
  63         { "inst_emu",   VCPU_STAT(emulated_inst_exits) },
  64         { "dec",        VCPU_STAT(dec_exits) },
  65         { "ext_intr",   VCPU_STAT(ext_intr_exits) },
  66         { "halt_wakeup", VCPU_STAT(halt_wakeup) },
  67         { "doorbell", VCPU_STAT(dbell_exits) },
  68         { "guest doorbell", VCPU_STAT(gdbell_exits) },
  69         { "remote_tlb_flush", VM_STAT(remote_tlb_flush) },
  70         { NULL }
  71 };
  72
  73 /* TODO: use vcpu_printf() */
  74 void kvmppc_dump_vcpu(struct kvm_vcpu *vcpu)
  75 {
  76         int i;
  77
  78         printk("pc:   %08lx msr:  %08llx\n", vcpu->arch.pc, vcpu->arch.shared->msr);
  79         printk("lr:   %08lx ctr:  %08lx\n", vcpu->arch.lr, vcpu->arch.ctr);
  80         printk("srr0: %08llx srr1: %08llx\n", vcpu->arch.shared->srr0,
  81                                             vcpu->arch.shared->srr1);
  82
  83         printk("exceptions: %08lx\n", vcpu->arch.pending_exceptions);
  84
  85         for (i = 0; i < 32; i += 4) {
  86                 printk("gpr%02d: %08lx %08lx %08lx %08lx\n", i,
  87                        kvmppc_get_gpr(vcpu, i),
  88                        kvmppc_get_gpr(vcpu, i+1),
  89                        kvmppc_get_gpr(vcpu, i+2),
  90                        kvmppc_get_gpr(vcpu, i+3));
  91         }
  92 }
  93
  94 #ifdef CONFIG_SPE
  95 void kvmppc_vcpu_disable_spe(struct kvm_vcpu *vcpu)
  96 {
  97         preempt_disable();
  98         enable_kernel_spe();
  99         kvmppc_save_guest_spe(vcpu);
 100         vcpu->arch.shadow_msr &= ~MSR_SPE;
 101         preempt_enable();
 102 }
 103
 104 static void kvmppc_vcpu_enable_spe(struct kvm_vcpu *vcpu)
 105 {
 106         preempt_disable();
 107         enable_kernel_spe();
 108         kvmppc_load_guest_spe(vcpu);
 109         vcpu->arch.shadow_msr |= MSR_SPE;
 110         preempt_enable();
 111 }
 112
 113 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
 114 {
 115         if (vcpu->arch.shared->msr & MSR_SPE) {
 116                 if (!(vcpu->arch.shadow_msr & MSR_SPE))
 117                         kvmppc_vcpu_enable_spe(vcpu);
 118         } else if (vcpu->arch.shadow_msr & MSR_SPE) {
 119                 kvmppc_vcpu_disable_spe(vcpu);
 120         }
 121 }
 122 #else
 123 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
 124 {
 125 }
 126 #endif
 127
 128 static void kvmppc_vcpu_sync_fpu(struct kvm_vcpu *vcpu)
 129 {
 130 #if defined(CONFIG_PPC_FPU) && !defined(CONFIG_KVM_BOOKE_HV)
 131         /* We always treat the FP bit as enabled from the host
 132            perspective, so only need to adjust the shadow MSR */
 133         vcpu->arch.shadow_msr &= ~MSR_FP;
 134         vcpu->arch.shadow_msr |= vcpu->arch.shared->msr & MSR_FP;
 135 #endif
 136 }
 137
 138 static void kvmppc_vcpu_sync_debug(struct kvm_vcpu *vcpu)
 139 {
 140         /* Synchronize guest's desire to get debug interrupts into shadow MSR */
 141 #ifndef CONFIG_KVM_BOOKE_HV
 142         vcpu->arch.shadow_msr &= ~MSR_DE;
 143         vcpu->arch.shadow_msr |= vcpu->arch.shared->msr & MSR_DE;
 144 #endif
 145
 146         /* Force enable debug interrupts when user space wants to debug */
 147         if (vcpu->guest_debug) {
 148 #ifdef CONFIG_KVM_BOOKE_HV
 149                 /*
 150                  * Since there is no shadow MSR, sync MSR_DE into the guest
 151                  * visible MSR.
 152                  */
 153                 vcpu->arch.shared->msr |= MSR_DE;
 154 #else
 155                 vcpu->arch.shadow_msr |= MSR_DE;
 156                 vcpu->arch.shared->msr &= ~MSR_DE;
 157 #endif
 158         }
 159 }
 160
 161 /*
 162  * Helper function for "full" MSR writes.  No need to call this if only
 163  * EE/CE/ME/DE/RI are changing.
 164  */
 165 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr)
 166 {
 167         u32 old_msr = vcpu->arch.shared->msr;
 168
 169 #ifdef CONFIG_KVM_BOOKE_HV
 170         new_msr |= MSR_GS;
 171 #endif
 172
 173         vcpu->arch.shared->msr = new_msr;
 174
 175         kvmppc_mmu_msr_notify(vcpu, old_msr);
 176         kvmppc_vcpu_sync_spe(vcpu);
 177         kvmppc_vcpu_sync_fpu(vcpu);
 178         kvmppc_vcpu_sync_debug(vcpu);
 179 }
 180
 181 static void kvmppc_booke_queue_irqprio(struct kvm_vcpu *vcpu,
 182                                        unsigned int priority)
 183 {
 184         trace_kvm_booke_queue_irqprio(vcpu, priority);
 185         set_bit(priority, &vcpu->arch.pending_exceptions);
 186 }
 187
 188 static void kvmppc_core_queue_dtlb_miss(struct kvm_vcpu *vcpu,
 189                                         ulong dear_flags, ulong esr_flags)
 190 {
 191         vcpu->arch.queued_dear = dear_flags;
 192         vcpu->arch.queued_esr = esr_flags;
 193         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DTLB_MISS);
 194 }
 195
 196 static void kvmppc_core_queue_data_storage(struct kvm_vcpu *vcpu,
 197                                            ulong dear_flags, ulong esr_flags)
 198 {
 199         vcpu->arch.queued_dear = dear_flags;
 200         vcpu->arch.queued_esr = esr_flags;
 201         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DATA_STORAGE);
 202 }
 203
 204 static void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu,
 205                                            ulong esr_flags)
 206 {
 207         vcpu->arch.queued_esr = esr_flags;
 208         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_INST_STORAGE);
 209 }
 210
 211 static void kvmppc_core_queue_alignment(struct kvm_vcpu *vcpu, ulong dear_flags,
 212                                         ulong esr_flags)
 213 {
 214         vcpu->arch.queued_dear = dear_flags;
 215         vcpu->arch.queued_esr = esr_flags;
 216         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALIGNMENT);
 217 }
 218
 219 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong esr_flags)
 220 {
 221         vcpu->arch.queued_esr = esr_flags;
 222         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_PROGRAM);
 223 }
 224
 225 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
 226 {
 227         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DECREMENTER);
 228 }
 229
 230 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
 231 {
 232         return test_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
 233 }
 234
 235 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
 236 {
 237         clear_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
 238 }
 239
 240 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
 241                                 struct kvm_interrupt *irq)
 242 {
 243         unsigned int prio = BOOKE_IRQPRIO_EXTERNAL;
 244
 245         if (irq->irq == KVM_INTERRUPT_SET_LEVEL)
 246                 prio = BOOKE_IRQPRIO_EXTERNAL_LEVEL;
 247
 248         kvmppc_booke_queue_irqprio(vcpu, prio);
 249 }
 250
 251 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu)
 252 {
 253         clear_bit(BOOKE_IRQPRIO_EXTERNAL, &vcpu->arch.pending_exceptions);
 254         clear_bit(BOOKE_IRQPRIO_EXTERNAL_LEVEL, &vcpu->arch.pending_exceptions);
 255 }
 256
 257 static void kvmppc_core_queue_watchdog(struct kvm_vcpu *vcpu)
 258 {
 259         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_WATCHDOG);
 260 }
 261
 262 static void kvmppc_core_dequeue_watchdog(struct kvm_vcpu *vcpu)
 263 {
 264         clear_bit(BOOKE_IRQPRIO_WATCHDOG, &vcpu->arch.pending_exceptions);
 265 }
 266
 267 static void set_guest_srr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
 268 {
 269 #ifdef CONFIG_KVM_BOOKE_HV
 270         mtspr(SPRN_GSRR0, srr0);
 271         mtspr(SPRN_GSRR1, srr1);
 272 #else
 273         vcpu->arch.shared->srr0 = srr0;
 274         vcpu->arch.shared->srr1 = srr1;
 275 #endif
 276 }
 277
 278 static void set_guest_csrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
 279 {
 280         vcpu->arch.csrr0 = srr0;
 281         vcpu->arch.csrr1 = srr1;
 282 }
 283
 284 static void set_guest_dsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
 285 {
 286         if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC)) {
 287                 vcpu->arch.dsrr0 = srr0;
 288                 vcpu->arch.dsrr1 = srr1;
 289         } else {
 290                 set_guest_csrr(vcpu, srr0, srr1);
 291         }
 292 }
 293
 294 static void set_guest_mcsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
 295 {
 296         vcpu->arch.mcsrr0 = srr0;
 297         vcpu->arch.mcsrr1 = srr1;
 298 }
 299
 300 static unsigned long get_guest_dear(struct kvm_vcpu *vcpu)
 301 {
 302 #ifdef CONFIG_KVM_BOOKE_HV
 303         return mfspr(SPRN_GDEAR);
 304 #else
 305         return vcpu->arch.shared->dar;
 306 #endif
 307 }
 308
 309 static void set_guest_dear(struct kvm_vcpu *vcpu, unsigned long dear)
 310 {
 311 #ifdef CONFIG_KVM_BOOKE_HV
 312         mtspr(SPRN_GDEAR, dear);
 313 #else
 314         vcpu->arch.shared->dar = dear;
 315 #endif
 316 }
 317
 318 static unsigned long get_guest_esr(struct kvm_vcpu *vcpu)
 319 {
 320 #ifdef CONFIG_KVM_BOOKE_HV
 321         return mfspr(SPRN_GESR);
 322 #else
 323         return vcpu->arch.shared->esr;
 324 #endif
 325 }
 326
 327 static void set_guest_esr(struct kvm_vcpu *vcpu, u32 esr)
 328 {
 329 #ifdef CONFIG_KVM_BOOKE_HV
 330         mtspr(SPRN_GESR, esr);
 331 #else
 332         vcpu->arch.shared->esr = esr;
 333 #endif
 334 }
 335
 336 static unsigned long get_guest_epr(struct kvm_vcpu *vcpu)
 337 {
 338 #ifdef CONFIG_KVM_BOOKE_HV
 339         return mfspr(SPRN_GEPR);
 340 #else
 341         return vcpu->arch.epr;
 342 #endif
 343 }
 344
 345 /* Deliver the interrupt of the corresponding priority, if possible. */
 346 static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu *vcpu,
 347                                         unsigned int priority)
 348 {
 349         int allowed = 0;
 350         ulong msr_mask = 0;
 351         bool update_esr = false, update_dear = false, update_epr = false;
 352         ulong crit_raw = vcpu->arch.shared->critical;
 353         ulong crit_r1 = kvmppc_get_gpr(vcpu, 1);
 354         bool crit;
 355         bool keep_irq = false;
 356         enum int_class int_class;
 357         ulong new_msr = vcpu->arch.shared->msr;
 358
 359         /* Truncate crit indicators in 32 bit mode */
 360         if (!(vcpu->arch.shared->msr & MSR_SF)) {
 361                 crit_raw &= 0xffffffff;
 362                 crit_r1 &= 0xffffffff;
 363         }
 364
 365         /* Critical section when crit == r1 */
 366         crit = (crit_raw == crit_r1);
 367         /* ... and we're in supervisor mode */
 368         crit = crit && !(vcpu->arch.shared->msr & MSR_PR);
 369
 370         if (priority == BOOKE_IRQPRIO_EXTERNAL_LEVEL) {
 371                 priority = BOOKE_IRQPRIO_EXTERNAL;
 372                 keep_irq = true;
 373         }
 374
 375         if ((priority == BOOKE_IRQPRIO_EXTERNAL) && vcpu->arch.epr_flags)
 376                 update_epr = true;
 377
 378         switch (priority) {
 379         case BOOKE_IRQPRIO_DTLB_MISS:
 380         case BOOKE_IRQPRIO_DATA_STORAGE:
 381         case BOOKE_IRQPRIO_ALIGNMENT:
 382                 update_dear = true;
 383                 /* fall through */
 384         case BOOKE_IRQPRIO_INST_STORAGE:
 385         case BOOKE_IRQPRIO_PROGRAM:
 386                 update_esr = true;
 387                 /* fall through */
 388         case BOOKE_IRQPRIO_ITLB_MISS:
 389         case BOOKE_IRQPRIO_SYSCALL:
 390         case BOOKE_IRQPRIO_FP_UNAVAIL:
 391         case BOOKE_IRQPRIO_SPE_UNAVAIL:
 392         case BOOKE_IRQPRIO_SPE_FP_DATA:
 393         case BOOKE_IRQPRIO_SPE_FP_ROUND:
 394         case BOOKE_IRQPRIO_AP_UNAVAIL:
 395                 allowed = 1;
 396                 msr_mask = MSR_CE | MSR_ME | MSR_DE;
 397                 int_class = INT_CLASS_NONCRIT;
 398                 break;
 399         case BOOKE_IRQPRIO_WATCHDOG:
 400         case BOOKE_IRQPRIO_CRITICAL:
 401         case BOOKE_IRQPRIO_DBELL_CRIT:
 402                 allowed = vcpu->arch.shared->msr & MSR_CE;
 403                 allowed = allowed && !crit;
 404                 msr_mask = MSR_ME;
 405                 int_class = INT_CLASS_CRIT;
 406                 break;
 407         case BOOKE_IRQPRIO_MACHINE_CHECK:
 408                 allowed = vcpu->arch.shared->msr & MSR_ME;
 409                 allowed = allowed && !crit;
 410                 int_class = INT_CLASS_MC;
 411                 break;
 412         case BOOKE_IRQPRIO_DECREMENTER:
 413         case BOOKE_IRQPRIO_FIT:
 414                 keep_irq = true;
 415                 /* fall through */
 416         case BOOKE_IRQPRIO_EXTERNAL:
 417         case BOOKE_IRQPRIO_DBELL:
 418                 allowed = vcpu->arch.shared->msr & MSR_EE;
 419                 allowed = allowed && !crit;
 420                 msr_mask = MSR_CE | MSR_ME | MSR_DE;
 421                 int_class = INT_CLASS_NONCRIT;
 422                 break;
 423         case BOOKE_IRQPRIO_DEBUG:
 424                 allowed = vcpu->arch.shared->msr & MSR_DE;
 425                 allowed = allowed && !crit;
 426                 msr_mask = MSR_ME;
 427                 int_class = INT_CLASS_CRIT;
 428                 break;
 429         }
 430
 431         if (allowed) {
 432                 switch (int_class) {
 433                 case INT_CLASS_NONCRIT:
 434                         set_guest_srr(vcpu, vcpu->arch.pc,
 435                                       vcpu->arch.shared->msr);
 436                         break;
 437                 case INT_CLASS_CRIT:
 438                         set_guest_csrr(vcpu, vcpu->arch.pc,
 439                                        vcpu->arch.shared->msr);
 440                         break;
 441                 case INT_CLASS_DBG:
 442                         set_guest_dsrr(vcpu, vcpu->arch.pc,
 443                                        vcpu->arch.shared->msr);
 444                         break;
 445                 case INT_CLASS_MC:
 446                         set_guest_mcsrr(vcpu, vcpu->arch.pc,
 447                                         vcpu->arch.shared->msr);
 448                         break;
 449                 }
 450
 451                 vcpu->arch.pc = vcpu->arch.ivpr | vcpu->arch.ivor[priority];
 452                 if (update_esr == true)
 453                         set_guest_esr(vcpu, vcpu->arch.queued_esr);
 454                 if (update_dear == true)
 455                         set_guest_dear(vcpu, vcpu->arch.queued_dear);
 456                 if (update_epr == true) {
 457                         if (vcpu->arch.epr_flags & KVMPPC_EPR_USER)
 458                                 kvm_make_request(KVM_REQ_EPR_EXIT, vcpu);
 459                         else if (vcpu->arch.epr_flags & KVMPPC_EPR_KERNEL) {
 460                                 BUG_ON(vcpu->arch.irq_type != KVMPPC_IRQ_MPIC);
 461                                 kvmppc_mpic_set_epr(vcpu);
 462                         }
 463                 }
 464
 465                 new_msr &= msr_mask;
 466 #if defined(CONFIG_64BIT)
 467                 if (vcpu->arch.epcr & SPRN_EPCR_ICM)
 468                         new_msr |= MSR_CM;
 469 #endif
 470                 kvmppc_set_msr(vcpu, new_msr);
 471
 472                 if (!keep_irq)
 473                         clear_bit(priority, &vcpu->arch.pending_exceptions);
 474         }
 475
 476 #ifdef CONFIG_KVM_BOOKE_HV
 477         /*
 478          * If an interrupt is pending but masked, raise a guest doorbell
 479          * so that we are notified when the guest enables the relevant
 480          * MSR bit.
 481          */
 482         if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_EE)
 483                 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_NONCRIT);
 484         if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_CE)
 485                 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_CRIT);
 486         if (vcpu->arch.pending_exceptions & BOOKE_IRQPRIO_MACHINE_CHECK)
 487                 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_MC);
 488 #endif
 489
 490         return allowed;
 491 }
 492
 493 /*
 494  * Return the number of jiffies until the next timeout.  If the timeout is
 495  * longer than the NEXT_TIMER_MAX_DELTA, then return NEXT_TIMER_MAX_DELTA
 496  * because the larger value can break the timer APIs.
 497  */
 498 static unsigned long watchdog_next_timeout(struct kvm_vcpu *vcpu)
 499 {
 500         u64 tb, wdt_tb, wdt_ticks = 0;
 501         u64 nr_jiffies = 0;
 502         u32 period = TCR_GET_WP(vcpu->arch.tcr);
 503
 504         wdt_tb = 1ULL << (63 - period);
 505         tb = get_tb();
 506         /*
 507          * The watchdog timeout will hapeen when TB bit corresponding
 508          * to watchdog will toggle from 0 to 1.
 509          */
 510         if (tb & wdt_tb)
 511                 wdt_ticks = wdt_tb;
 512
 513         wdt_ticks += wdt_tb - (tb & (wdt_tb - 1));
 514
 515         /* Convert timebase ticks to jiffies */
 516         nr_jiffies = wdt_ticks;
 517
 518         if (do_div(nr_jiffies, tb_ticks_per_jiffy))
 519                 nr_jiffies++;
 520
 521         return min_t(unsigned long long, nr_jiffies, NEXT_TIMER_MAX_DELTA);
 522 }
 523
 524 static void arm_next_watchdog(struct kvm_vcpu *vcpu)
 525 {
 526         unsigned long nr_jiffies;
 527         unsigned long flags;
 528
 529         /*
 530          * If TSR_ENW and TSR_WIS are not set then no need to exit to
 531          * userspace, so clear the KVM_REQ_WATCHDOG request.
 532          */
 533         if ((vcpu->arch.tsr & (TSR_ENW | TSR_WIS)) != (TSR_ENW | TSR_WIS))
 534                 clear_bit(KVM_REQ_WATCHDOG, &vcpu->requests);
 535
 536         spin_lock_irqsave(&vcpu->arch.wdt_lock, flags);
 537         nr_jiffies = watchdog_next_timeout(vcpu);
 538         /*
 539          * If the number of jiffies of watchdog timer >= NEXT_TIMER_MAX_DELTA
 540          * then do not run the watchdog timer as this can break timer APIs.
 541          */
 542         if (nr_jiffies < NEXT_TIMER_MAX_DELTA)
 543                 mod_timer(&vcpu->arch.wdt_timer, jiffies + nr_jiffies);
 544         else
 545                 del_timer(&vcpu->arch.wdt_timer);
 546         spin_unlock_irqrestore(&vcpu->arch.wdt_lock, flags);
 547 }
 548
 549 void kvmppc_watchdog_func(unsigned long data)
 550 {
 551         struct kvm_vcpu *vcpu = (struct kvm_vcpu *)data;
 552         u32 tsr, new_tsr;
 553         int final;
 554
 555         do {
 556                 new_tsr = tsr = vcpu->arch.tsr;
 557                 final = 0;
 558
 559                 /* Time out event */
 560                 if (tsr & TSR_ENW) {
 561                         if (tsr & TSR_WIS)
 562                                 final = 1;
 563                         else
 564                                 new_tsr = tsr | TSR_WIS;
 565                 } else {
 566                         new_tsr = tsr | TSR_ENW;
 567                 }
 568         } while (cmpxchg(&vcpu->arch.tsr, tsr, new_tsr) != tsr);
 569
 570         if (new_tsr & TSR_WIS) {
 571                 smp_wmb();
 572                 kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
 573                 kvm_vcpu_kick(vcpu);
 574         }
 575
 576         /*
 577          * If this is final watchdog expiry and some action is required
 578          * then exit to userspace.
 579          */
 580         if (final && (vcpu->arch.tcr & TCR_WRC_MASK) &&
 581             vcpu->arch.watchdog_enabled) {
 582                 smp_wmb();
 583                 kvm_make_request(KVM_REQ_WATCHDOG, vcpu);
 584                 kvm_vcpu_kick(vcpu);
 585         }
 586
 587         /*
 588          * Stop running the watchdog timer after final expiration to
 589          * prevent the host from being flooded with timers if the
 590          * guest sets a short period.
 591          * Timers will resume when TSR/TCR is updated next time.
 592          */
 593         if (!final)
 594                 arm_next_watchdog(vcpu);
 595 }
 596
 597 static void update_timer_ints(struct kvm_vcpu *vcpu)
 598 {
 599         if ((vcpu->arch.tcr & TCR_DIE) && (vcpu->arch.tsr & TSR_DIS))
 600                 kvmppc_core_queue_dec(vcpu);
 601         else
 602                 kvmppc_core_dequeue_dec(vcpu);
 603
 604         if ((vcpu->arch.tcr & TCR_WIE) && (vcpu->arch.tsr & TSR_WIS))
 605                 kvmppc_core_queue_watchdog(vcpu);
 606         else
 607                 kvmppc_core_dequeue_watchdog(vcpu);
 608 }
 609
 610 static void kvmppc_core_check_exceptions(struct kvm_vcpu *vcpu)
 611 {
 612         unsigned long *pending = &vcpu->arch.pending_exceptions;
 613         unsigned int priority;
 614
 615         priority = __ffs(*pending);
 616         while (priority < BOOKE_IRQPRIO_MAX) {
 617                 if (kvmppc_booke_irqprio_deliver(vcpu, priority))
 618                         break;
 619
 620                 priority = find_next_bit(pending,
 621                                          BITS_PER_BYTE * sizeof(*pending),
 622                                          priority + 1);
 623         }
 624
 625         /* Tell the guest about our interrupt status */
 626         vcpu->arch.shared->int_pending = !!*pending;
 627 }
 628
 629 /* Check pending exceptions and deliver one, if possible. */
 630 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
 631 {
 632         int r = 0;
 633         WARN_ON_ONCE(!irqs_disabled());
 634
 635         kvmppc_core_check_exceptions(vcpu);
 636
 637         if (vcpu->requests) {
 638                 /* Exception delivery raised request; start over */
 639                 return 1;
 640         }
 641
 642         if (vcpu->arch.shared->msr & MSR_WE) {
 643                 local_irq_enable();
 644                 kvm_vcpu_block(vcpu);
 645                 clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
 646                 local_irq_disable();
 647
 648                 kvmppc_set_exit_type(vcpu, EMULATED_MTMSRWE_EXITS);
 649                 r = 1;
 650         };
 651
 652         return r;
 653 }
 654
 655 int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
 656 {
 657         int r = 1; /* Indicate we want to get back into the guest */
 658
 659         if (kvm_check_request(KVM_REQ_PENDING_TIMER, vcpu))
 660                 update_timer_ints(vcpu);
 661 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
 662         if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu))
 663                 kvmppc_core_flush_tlb(vcpu);
 664 #endif
 665
 666         if (kvm_check_request(KVM_REQ_WATCHDOG, vcpu)) {
 667                 vcpu->run->exit_reason = KVM_EXIT_WATCHDOG;
 668                 r = 0;
 669         }
 670
 671         if (kvm_check_request(KVM_REQ_EPR_EXIT, vcpu)) {
 672                 vcpu->run->epr.epr = 0;
 673                 vcpu->arch.epr_needed = true;
 674                 vcpu->run->exit_reason = KVM_EXIT_EPR;
 675                 r = 0;
 676         }
 677
 678         return r;
 679 }
 680
 681 int kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
 682 {
 683         int ret, s;
 684         struct debug_reg debug;
 685 #ifdef CONFIG_PPC_FPU
 686         struct thread_fp_state fp;
 687         int fpexc_mode;
 688 #endif
 689
 690         if (!vcpu->arch.sane) {
 691                 kvm_run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
 692                 return -EINVAL;
 693         }
 694
 695         local_irq_disable();
 696         s = kvmppc_prepare_to_enter(vcpu);
 697         if (s <= 0) {
 698                 local_irq_enable();
 699                 ret = s;
 700                 goto out;
 701         }
 702
 703 #ifdef CONFIG_PPC_FPU
 704         /* Save userspace FPU state in stack */
 705         enable_kernel_fp();
 706         fp = current->thread.fp_state;
 707         fpexc_mode = current->thread.fpexc_mode;
 708
 709         /* Restore guest FPU state to thread */
 710         memcpy(current->thread.fp_state.fpr, vcpu->arch.fpr,
 711                sizeof(vcpu->arch.fpr));
 712         current->thread.fp_state.fpscr = vcpu->arch.fpscr;
 713
 714         /*
 715          * Since we can't trap on MSR_FP in GS-mode, we consider the guest
 716          * as always using the FPU.  Kernel usage of FP (via
 717          * enable_kernel_fp()) in this thread must not occur while
 718          * vcpu->fpu_active is set.
 719          */
 720         vcpu->fpu_active = 1;
 721
 722         kvmppc_load_guest_fp(vcpu);
 723 #endif
 724
 725         /* Switch to guest debug context */
 726         debug = vcpu->arch.shadow_dbg_reg;
 727         switch_booke_debug_regs(&debug);
 728         debug = current->thread.debug;
 729         current->thread.debug = vcpu->arch.shadow_dbg_reg;
 730
 731         kvmppc_fix_ee_before_entry();
 732
 733         ret = __kvmppc_vcpu_run(kvm_run, vcpu);
 734
 735         /* No need for kvm_guest_exit. It's done in handle_exit.
 736            We also get here with interrupts enabled. */
 737
 738         /* Switch back to user space debug context */
 739         switch_booke_debug_regs(&debug);
 740         current->thread.debug = debug;
 741
 742 #ifdef CONFIG_PPC_FPU
 743         kvmppc_save_guest_fp(vcpu);
 744
 745         vcpu->fpu_active = 0;
 746
 747         /* Save guest FPU state from thread */
 748         memcpy(vcpu->arch.fpr, current->thread.fp_state.fpr,
 749                sizeof(vcpu->arch.fpr));
 750         vcpu->arch.fpscr = current->thread.fp_state.fpscr;
 751
 752         /* Restore userspace FPU state from stack */
 753         current->thread.fp_state = fp;
 754         current->thread.fpexc_mode = fpexc_mode;
 755 #endif
 756
 757 out:
 758         vcpu->mode = OUTSIDE_GUEST_MODE;
 759         return ret;
 760 }
 761
 762 static int emulation_exit(struct kvm_run *run, struct kvm_vcpu *vcpu)
 763 {
 764         enum emulation_result er;
 765
 766         er = kvmppc_emulate_instruction(run, vcpu);
 767         switch (er) {
 768         case EMULATE_DONE:
 769                 /* don't overwrite subtypes, just account kvm_stats */
 770                 kvmppc_account_exit_stat(vcpu, EMULATED_INST_EXITS);
 771                 /* Future optimization: only reload non-volatiles if
 772                  * they were actually modified by emulation. */
 773                 return RESUME_GUEST_NV;
 774
 775         case EMULATE_DO_DCR:
 776                 run->exit_reason = KVM_EXIT_DCR;
 777                 return RESUME_HOST;
 778
 779         case EMULATE_FAIL:
 780                 printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
 781                        __func__, vcpu->arch.pc, vcpu->arch.last_inst);
 782                 /* For debugging, encode the failing instruction and
 783                  * report it to userspace. */
 784                 run->hw.hardware_exit_reason = ~0ULL << 32;
 785                 run->hw.hardware_exit_reason |= vcpu->arch.last_inst;
 786                 kvmppc_core_queue_program(vcpu, ESR_PIL);
 787                 return RESUME_HOST;
 788
 789         case EMULATE_EXIT_USER:
 790                 return RESUME_HOST;
 791
 792         default:
 793                 BUG();
 794         }
 795 }
 796
 797 static int kvmppc_handle_debug(struct kvm_run *run, struct kvm_vcpu *vcpu)
 798 {
 799         struct debug_reg *dbg_reg = &(vcpu->arch.shadow_dbg_reg);
 800         u32 dbsr = vcpu->arch.dbsr;
 801
 802         run->debug.arch.status = 0;
 803         run->debug.arch.address = vcpu->arch.pc;
 804
 805         if (dbsr & (DBSR_IAC1 | DBSR_IAC2 | DBSR_IAC3 | DBSR_IAC4)) {
 806                 run->debug.arch.status |= KVMPPC_DEBUG_BREAKPOINT;
 807         } else {
 808                 if (dbsr & (DBSR_DAC1W | DBSR_DAC2W))
 809                         run->debug.arch.status |= KVMPPC_DEBUG_WATCH_WRITE;
 810                 else if (dbsr & (DBSR_DAC1R | DBSR_DAC2R))
 811                         run->debug.arch.status |= KVMPPC_DEBUG_WATCH_READ;
 812                 if (dbsr & (DBSR_DAC1R | DBSR_DAC1W))
 813                         run->debug.arch.address = dbg_reg->dac1;
 814                 else if (dbsr & (DBSR_DAC2R | DBSR_DAC2W))
 815                         run->debug.arch.address = dbg_reg->dac2;
 816         }
 817
 818         return RESUME_HOST;
 819 }
 820
 821 static void kvmppc_fill_pt_regs(struct pt_regs *regs)
 822 {
 823         ulong r1, ip, msr, lr;
 824
 825         asm("mr %0, 1" : "=r"(r1));
 826         asm("mflr %0" : "=r"(lr));
 827         asm("mfmsr %0" : "=r"(msr));
 828         asm("bl 1f; 1: mflr %0" : "=r"(ip));
 829
 830         memset(regs, 0, sizeof(*regs));
 831         regs->gpr[1] = r1;
 832         regs->nip = ip;
 833         regs->msr = msr;
 834         regs->link = lr;
 835 }
 836
 837 /*
 838  * For interrupts needed to be handled by host interrupt handlers,
 839  * corresponding host handler are called from here in similar way
 840  * (but not exact) as they are called from low level handler
 841  * (such as from arch/powerpc/kernel/head_fsl_booke.S).
 842  */
 843 static void kvmppc_restart_interrupt(struct kvm_vcpu *vcpu,
 844                                      unsigned int exit_nr)
 845 {
 846         struct pt_regs regs;
 847
 848         switch (exit_nr) {
 849         case BOOKE_INTERRUPT_EXTERNAL:
 850                 kvmppc_fill_pt_regs(&regs);
 851                 do_IRQ(&regs);
 852                 break;
 853         case BOOKE_INTERRUPT_DECREMENTER:
 854                 kvmppc_fill_pt_regs(&regs);
 855                 timer_interrupt(&regs);
 856                 break;
 857 #if defined(CONFIG_PPC_DOORBELL)
 858         case BOOKE_INTERRUPT_DOORBELL:
 859                 kvmppc_fill_pt_regs(&regs);
 860                 doorbell_exception(&regs);
 861                 break;
 862 #endif
 863         case BOOKE_INTERRUPT_MACHINE_CHECK:
 864                 /* FIXME */
 865                 break;
 866         case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
 867                 kvmppc_fill_pt_regs(&regs);
 868                 performance_monitor_exception(&regs);
 869                 break;
 870         case BOOKE_INTERRUPT_WATCHDOG:
 871                 kvmppc_fill_pt_regs(&regs);
 872 #ifdef CONFIG_BOOKE_WDT
 873                 WatchdogException(&regs);
 874 #else
 875                 unknown_exception(&regs);
 876 #endif
 877                 break;
 878         case BOOKE_INTERRUPT_CRITICAL:
 879                 unknown_exception(&regs);
 880                 break;
 881         case BOOKE_INTERRUPT_DEBUG:
 882                 /* Save DBSR before preemption is enabled */
 883                 vcpu->arch.dbsr = mfspr(SPRN_DBSR);
 884                 kvmppc_clear_dbsr();
 885                 break;
 886         }
 887 }
 888
 889 /**
 890  * kvmppc_handle_exit
 891  *
 892  * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
 893  */
 894 int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
 895                        unsigned int exit_nr)
 896 {
 897         int r = RESUME_HOST;
 898         int s;
 899         int idx;
 900
 901 #ifdef CONFIG_PPC64
 902         WARN_ON(local_paca->irq_happened != 0);
 903 #endif
 904
 905         /*
 906          * We enter with interrupts disabled in hardware, but
 907          * we need to call hard_irq_disable anyway to ensure that
 908          * the software state is kept in sync.
 909          */
 910         hard_irq_disable();
 911
 912         /* update before a new last_exit_type is rewritten */
 913         kvmppc_update_timing_stats(vcpu);
 914
 915         /* restart interrupts if they were meant for the host */
 916         kvmppc_restart_interrupt(vcpu, exit_nr);
 917
 918         local_irq_enable();
 919
 920         trace_kvm_exit(exit_nr, vcpu);
 921         kvm_guest_exit();
 922
 923         run->exit_reason = KVM_EXIT_UNKNOWN;
 924         run->ready_for_interrupt_injection = 1;
 925
 926         switch (exit_nr) {
 927         case BOOKE_INTERRUPT_MACHINE_CHECK:
 928                 printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR));
 929                 kvmppc_dump_vcpu(vcpu);
 930                 /* For debugging, send invalid exit reason to user space */
 931                 run->hw.hardware_exit_reason = ~1ULL << 32;
 932                 run->hw.hardware_exit_reason |= mfspr(SPRN_MCSR);
 933                 r = RESUME_HOST;
 934                 break;
 935
 936         case BOOKE_INTERRUPT_EXTERNAL:
 937                 kvmppc_account_exit(vcpu, EXT_INTR_EXITS);
 938                 r = RESUME_GUEST;
 939                 break;
 940
 941         case BOOKE_INTERRUPT_DECREMENTER:
 942                 kvmppc_account_exit(vcpu, DEC_EXITS);
 943                 r = RESUME_GUEST;
 944                 break;
 945
 946         case BOOKE_INTERRUPT_WATCHDOG:
 947                 r = RESUME_GUEST;
 948                 break;
 949
 950         case BOOKE_INTERRUPT_DOORBELL:
 951                 kvmppc_account_exit(vcpu, DBELL_EXITS);
 952                 r = RESUME_GUEST;
 953                 break;
 954
 955         case BOOKE_INTERRUPT_GUEST_DBELL_CRIT:
 956                 kvmppc_account_exit(vcpu, GDBELL_EXITS);
 957
 958                 /*
 959                  * We are here because there is a pending guest interrupt
 960                  * which could not be delivered as MSR_CE or MSR_ME was not
 961                  * set.  Once we break from here we will retry delivery.
 962                  */
 963                 r = RESUME_GUEST;
 964                 break;
 965
 966         case BOOKE_INTERRUPT_GUEST_DBELL:
 967                 kvmppc_account_exit(vcpu, GDBELL_EXITS);
 968
 969                 /*
 970                  * We are here because there is a pending guest interrupt
 971                  * which could not be delivered as MSR_EE was not set.  Once
 972                  * we break from here we will retry delivery.
 973                  */
 974                 r = RESUME_GUEST;
 975                 break;
 976
 977         case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
 978                 r = RESUME_GUEST;
 979                 break;
 980
 981         case BOOKE_INTERRUPT_HV_PRIV:
 982                 r = emulation_exit(run, vcpu);
 983                 break;
 984
 985         case BOOKE_INTERRUPT_PROGRAM:
 986                 if (vcpu->arch.shared->msr & (MSR_PR | MSR_GS)) {
 987                         /*
 988                          * Program traps generated by user-level software must
 989                          * be handled by the guest kernel.
 990                          *
 991                          * In GS mode, hypervisor privileged instructions trap
 992                          * on BOOKE_INTERRUPT_HV_PRIV, not here, so these are
 993                          * actual program interrupts, handled by the guest.
 994                          */
 995                         kvmppc_core_queue_program(vcpu, vcpu->arch.fault_esr);
 996                         r = RESUME_GUEST;
 997                         kvmppc_account_exit(vcpu, USR_PR_INST);
 998                         break;
 999                 }
1000
1001                 r = emulation_exit(run, vcpu);
1002                 break;
1003
1004         case BOOKE_INTERRUPT_FP_UNAVAIL:
1005                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
1006                 kvmppc_account_exit(vcpu, FP_UNAVAIL);
1007                 r = RESUME_GUEST;
1008                 break;
1009
1010 #ifdef CONFIG_SPE
1011         case BOOKE_INTERRUPT_SPE_UNAVAIL: {
1012                 if (vcpu->arch.shared->msr & MSR_SPE)
1013                         kvmppc_vcpu_enable_spe(vcpu);
1014                 else
1015                         kvmppc_booke_queue_irqprio(vcpu,
1016                                                    BOOKE_IRQPRIO_SPE_UNAVAIL);
1017                 r = RESUME_GUEST;
1018                 break;
1019         }
1020
1021         case BOOKE_INTERRUPT_SPE_FP_DATA:
1022                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_DATA);
1023                 r = RESUME_GUEST;
1024                 break;
1025
1026         case BOOKE_INTERRUPT_SPE_FP_ROUND:
1027                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_ROUND);
1028                 r = RESUME_GUEST;
1029                 break;
1030 #else
1031         case BOOKE_INTERRUPT_SPE_UNAVAIL:
1032                 /*
1033                  * Guest wants SPE, but host kernel doesn't support it.  Send
1034                  * an "unimplemented operation" program check to the guest.
1035                  */
1036                 kvmppc_core_queue_program(vcpu, ESR_PUO | ESR_SPV);
1037                 r = RESUME_GUEST;
1038                 break;
1039
1040         /*
1041          * These really should never happen without CONFIG_SPE,
1042          * as we should never enable the real MSR[SPE] in the guest.
1043          */
1044         case BOOKE_INTERRUPT_SPE_FP_DATA:
1045         case BOOKE_INTERRUPT_SPE_FP_ROUND:
1046                 printk(KERN_CRIT "%s: unexpected SPE interrupt %u at %08lx\n",
1047                        __func__, exit_nr, vcpu->arch.pc);
1048                 run->hw.hardware_exit_reason = exit_nr;
1049                 r = RESUME_HOST;
1050                 break;
1051 #endif
1052
1053         case BOOKE_INTERRUPT_DATA_STORAGE:
1054                 kvmppc_core_queue_data_storage(vcpu, vcpu->arch.fault_dear,
1055                                                vcpu->arch.fault_esr);
1056                 kvmppc_account_exit(vcpu, DSI_EXITS);
1057                 r = RESUME_GUEST;
1058                 break;
1059
1060         case BOOKE_INTERRUPT_INST_STORAGE:
1061                 kvmppc_core_queue_inst_storage(vcpu, vcpu->arch.fault_esr);
1062                 kvmppc_account_exit(vcpu, ISI_EXITS);
1063                 r = RESUME_GUEST;
1064                 break;
1065
1066         case BOOKE_INTERRUPT_ALIGNMENT:
1067                 kvmppc_core_queue_alignment(vcpu, vcpu->arch.fault_dear,
1068                                             vcpu->arch.fault_esr);
1069                 r = RESUME_GUEST;
1070                 break;
1071
1072 #ifdef CONFIG_KVM_BOOKE_HV
1073         case BOOKE_INTERRUPT_HV_SYSCALL:
1074                 if (!(vcpu->arch.shared->msr & MSR_PR)) {
1075                         kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1076                 } else {
1077                         /*
1078                          * hcall from guest userspace -- send privileged
1079                          * instruction program check.
1080                          */
1081                         kvmppc_core_queue_program(vcpu, ESR_PPR);
1082                 }
1083
1084                 r = RESUME_GUEST;
1085                 break;
1086 #else
1087         case BOOKE_INTERRUPT_SYSCALL:
1088                 if (!(vcpu->arch.shared->msr & MSR_PR) &&
1089                     (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
1090                         /* KVM PV hypercalls */
1091                         kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1092                         r = RESUME_GUEST;
1093                 } else {
1094                         /* Guest syscalls */
1095                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SYSCALL);
1096                 }
1097                 kvmppc_account_exit(vcpu, SYSCALL_EXITS);
1098                 r = RESUME_GUEST;
1099                 break;
1100 #endif
1101
1102         case BOOKE_INTERRUPT_DTLB_MISS: {
1103                 unsigned long eaddr = vcpu->arch.fault_dear;
1104                 int gtlb_index;
1105                 gpa_t gpaddr;
1106                 gfn_t gfn;
1107
1108 #ifdef CONFIG_KVM_E500V2
1109                 if (!(vcpu->arch.shared->msr & MSR_PR) &&
1110                     (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
1111                         kvmppc_map_magic(vcpu);
1112                         kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1113                         r = RESUME_GUEST;
1114
1115                         break;
1116                 }
1117 #endif
1118
1119                 /* Check the guest TLB. */
1120                 gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
1121                 if (gtlb_index < 0) {
1122                         /* The guest didn't have a mapping for it. */
1123                         kvmppc_core_queue_dtlb_miss(vcpu,
1124                                                     vcpu->arch.fault_dear,
1125                                                     vcpu->arch.fault_esr);
1126                         kvmppc_mmu_dtlb_miss(vcpu);
1127                         kvmppc_account_exit(vcpu, DTLB_REAL_MISS_EXITS);
1128                         r = RESUME_GUEST;
1129                         break;
1130                 }
1131
1132                 idx = srcu_read_lock(&vcpu->kvm->srcu);
1133
1134                 gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1135                 gfn = gpaddr >> PAGE_SHIFT;
1136
1137                 if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1138                         /* The guest TLB had a mapping, but the shadow TLB
1139                          * didn't, and it is RAM. This could be because:
1140                          * a) the entry is mapping the host kernel, or
1141                          * b) the guest used a large mapping which we're faking
1142                          * Either way, we need to satisfy the fault without
1143                          * invoking the guest. */
1144                         kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1145                         kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1146                         r = RESUME_GUEST;
1147                 } else {
1148                         /* Guest has mapped and accessed a page which is not
1149                          * actually RAM. */
1150                         vcpu->arch.paddr_accessed = gpaddr;
1151                         vcpu->arch.vaddr_accessed = eaddr;
1152                         r = kvmppc_emulate_mmio(run, vcpu);
1153                         kvmppc_account_exit(vcpu, MMIO_EXITS);
1154                 }
1155
1156                 srcu_read_unlock(&vcpu->kvm->srcu, idx);
1157                 break;
1158         }
1159
1160         case BOOKE_INTERRUPT_ITLB_MISS: {
1161                 unsigned long eaddr = vcpu->arch.pc;
1162                 gpa_t gpaddr;
1163                 gfn_t gfn;
1164                 int gtlb_index;
1165
1166                 r = RESUME_GUEST;
1167
1168                 /* Check the guest TLB. */
1169                 gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
1170                 if (gtlb_index < 0) {
1171                         /* The guest didn't have a mapping for it. */
1172                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
1173                         kvmppc_mmu_itlb_miss(vcpu);
1174                         kvmppc_account_exit(vcpu, ITLB_REAL_MISS_EXITS);
1175                         break;
1176                 }
1177
1178                 kvmppc_account_exit(vcpu, ITLB_VIRT_MISS_EXITS);
1179
1180                 idx = srcu_read_lock(&vcpu->kvm->srcu);
1181
1182                 gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1183                 gfn = gpaddr >> PAGE_SHIFT;
1184
1185                 if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1186                         /* The guest TLB had a mapping, but the shadow TLB
1187                          * didn't. This could be because:
1188                          * a) the entry is mapping the host kernel, or
1189                          * b) the guest used a large mapping which we're faking
1190                          * Either way, we need to satisfy the fault without
1191                          * invoking the guest. */
1192                         kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1193                 } else {
1194                         /* Guest mapped and leaped at non-RAM! */
1195                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_MACHINE_CHECK);
1196                 }
1197
1198                 srcu_read_unlock(&vcpu->kvm->srcu, idx);
1199                 break;
1200         }
1201
1202         case BOOKE_INTERRUPT_DEBUG: {
1203                 r = kvmppc_handle_debug(run, vcpu);
1204                 if (r == RESUME_HOST)
1205                         run->exit_reason = KVM_EXIT_DEBUG;
1206                 kvmppc_account_exit(vcpu, DEBUG_EXITS);
1207                 break;
1208         }
1209
1210         default:
1211                 printk(KERN_EMERG "exit_nr %d\n", exit_nr);
1212                 BUG();
1213         }
1214
1215         /*
1216          * To avoid clobbering exit_reason, only check for signals if we
1217          * aren't already exiting to userspace for some other reason.
1218          */
1219         if (!(r & RESUME_HOST)) {
1220                 local_irq_disable();
1221                 s = kvmppc_prepare_to_enter(vcpu);
1222                 if (s <= 0) {
1223                         local_irq_enable();
1224                         r = (s << 2) | RESUME_HOST | (r & RESUME_FLAG_NV);
1225                 } else {
1226                         kvmppc_fix_ee_before_entry();
1227                 }
1228         }
1229
1230         return r;
1231 }
1232
1233 static void kvmppc_set_tsr(struct kvm_vcpu *vcpu, u32 new_tsr)
1234 {
1235         u32 old_tsr = vcpu->arch.tsr;
1236
1237         vcpu->arch.tsr = new_tsr;
1238
1239         if ((old_tsr ^ vcpu->arch.tsr) & (TSR_ENW | TSR_WIS))
1240                 arm_next_watchdog(vcpu);
1241
1242         update_timer_ints(vcpu);
1243 }
1244
1245 /* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
1246 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
1247 {
1248         int i;
1249         int r;
1250
1251         vcpu->arch.pc = 0;
1252         vcpu->arch.shared->pir = vcpu->vcpu_id;
1253         kvmppc_set_gpr(vcpu, 1, (16<<20) - 8); /* -8 for the callee-save LR slot */
1254         kvmppc_set_msr(vcpu, 0);
1255
1256 #ifndef CONFIG_KVM_BOOKE_HV
1257         vcpu->arch.shadow_msr = MSR_USER | MSR_IS | MSR_DS;
1258         vcpu->arch.shadow_pid = 1;
1259         vcpu->arch.shared->msr = 0;
1260 #endif
1261
1262         /* Eye-catching numbers so we know if the guest takes an interrupt
1263          * before it's programmed its own IVPR/IVORs. */
1264         vcpu->arch.ivpr = 0x55550000;
1265         for (i = 0; i < BOOKE_IRQPRIO_MAX; i++)
1266                 vcpu->arch.ivor[i] = 0x7700 | i * 4;
1267
1268         kvmppc_init_timing_stats(vcpu);
1269
1270         r = kvmppc_core_vcpu_setup(vcpu);
1271         kvmppc_sanity_check(vcpu);
1272         return r;
1273 }
1274
1275 int kvmppc_subarch_vcpu_init(struct kvm_vcpu *vcpu)
1276 {
1277         /* setup watchdog timer once */
1278         spin_lock_init(&vcpu->arch.wdt_lock);
1279         setup_timer(&vcpu->arch.wdt_timer, kvmppc_watchdog_func,
1280                     (unsigned long)vcpu);
1281
1282         return 0;
1283 }
1284
1285 void kvmppc_subarch_vcpu_uninit(struct kvm_vcpu *vcpu)
1286 {
1287         del_timer_sync(&vcpu->arch.wdt_timer);
1288 }
1289
1290 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1291 {
1292         int i;
1293
1294         regs->pc = vcpu->arch.pc;
1295         regs->cr = kvmppc_get_cr(vcpu);
1296         regs->ctr = vcpu->arch.ctr;
1297         regs->lr = vcpu->arch.lr;
1298         regs->xer = kvmppc_get_xer(vcpu);
1299         regs->msr = vcpu->arch.shared->msr;
1300         regs->srr0 = vcpu->arch.shared->srr0;
1301         regs->srr1 = vcpu->arch.shared->srr1;
1302         regs->pid = vcpu->arch.pid;
1303         regs->sprg0 = vcpu->arch.shared->sprg0;
1304         regs->sprg1 = vcpu->arch.shared->sprg1;
1305         regs->sprg2 = vcpu->arch.shared->sprg2;
1306         regs->sprg3 = vcpu->arch.shared->sprg3;
1307         regs->sprg4 = vcpu->arch.shared->sprg4;
1308         regs->sprg5 = vcpu->arch.shared->sprg5;
1309         regs->sprg6 = vcpu->arch.shared->sprg6;
1310         regs->sprg7 = vcpu->arch.shared->sprg7;
1311
1312         for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1313                 regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
1314
1315         return 0;
1316 }
1317
1318 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1319 {
1320         int i;
1321
1322         vcpu->arch.pc = regs->pc;
1323         kvmppc_set_cr(vcpu, regs->cr);
1324         vcpu->arch.ctr = regs->ctr;
1325         vcpu->arch.lr = regs->lr;
1326         kvmppc_set_xer(vcpu, regs->xer);
1327         kvmppc_set_msr(vcpu, regs->msr);
1328         vcpu->arch.shared->srr0 = regs->srr0;
1329         vcpu->arch.shared->srr1 = regs->srr1;
1330         kvmppc_set_pid(vcpu, regs->pid);
1331         vcpu->arch.shared->sprg0 = regs->sprg0;
1332         vcpu->arch.shared->sprg1 = regs->sprg1;
1333         vcpu->arch.shared->sprg2 = regs->sprg2;
1334         vcpu->arch.shared->sprg3 = regs->sprg3;
1335         vcpu->arch.shared->sprg4 = regs->sprg4;
1336         vcpu->arch.shared->sprg5 = regs->sprg5;
1337         vcpu->arch.shared->sprg6 = regs->sprg6;
1338         vcpu->arch.shared->sprg7 = regs->sprg7;
1339
1340         for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1341                 kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
1342
1343         return 0;
1344 }
1345
1346 static void get_sregs_base(struct kvm_vcpu *vcpu,
1347                            struct kvm_sregs *sregs)
1348 {
1349         u64 tb = get_tb();
1350
1351         sregs->u.e.features |= KVM_SREGS_E_BASE;
1352
1353         sregs->u.e.csrr0 = vcpu->arch.csrr0;
1354         sregs->u.e.csrr1 = vcpu->arch.csrr1;
1355         sregs->u.e.mcsr = vcpu->arch.mcsr;
1356         sregs->u.e.esr = get_guest_esr(vcpu);
1357         sregs->u.e.dear = get_guest_dear(vcpu);
1358         sregs->u.e.tsr = vcpu->arch.tsr;
1359         sregs->u.e.tcr = vcpu->arch.tcr;
1360         sregs->u.e.dec = kvmppc_get_dec(vcpu, tb);
1361         sregs->u.e.tb = tb;
1362         sregs->u.e.vrsave = vcpu->arch.vrsave;
1363 }
1364
1365 static int set_sregs_base(struct kvm_vcpu *vcpu,
1366                           struct kvm_sregs *sregs)
1367 {
1368         if (!(sregs->u.e.features & KVM_SREGS_E_BASE))
1369                 return 0;
1370
1371         vcpu->arch.csrr0 = sregs->u.e.csrr0;
1372         vcpu->arch.csrr1 = sregs->u.e.csrr1;
1373         vcpu->arch.mcsr = sregs->u.e.mcsr;
1374         set_guest_esr(vcpu, sregs->u.e.esr);
1375         set_guest_dear(vcpu, sregs->u.e.dear);
1376         vcpu->arch.vrsave = sregs->u.e.vrsave;
1377         kvmppc_set_tcr(vcpu, sregs->u.e.tcr);
1378
1379         if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_DEC) {
1380                 vcpu->arch.dec = sregs->u.e.dec;
1381                 kvmppc_emulate_dec(vcpu);
1382         }
1383
1384         if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_TSR)
1385                 kvmppc_set_tsr(vcpu, sregs->u.e.tsr);
1386
1387         return 0;
1388 }
1389
1390 static void get_sregs_arch206(struct kvm_vcpu *vcpu,
1391                               struct kvm_sregs *sregs)
1392 {
1393         sregs->u.e.features |= KVM_SREGS_E_ARCH206;
1394
1395         sregs->u.e.pir = vcpu->vcpu_id;
1396         sregs->u.e.mcsrr0 = vcpu->arch.mcsrr0;
1397         sregs->u.e.mcsrr1 = vcpu->arch.mcsrr1;
1398         sregs->u.e.decar = vcpu->arch.decar;
1399         sregs->u.e.ivpr = vcpu->arch.ivpr;
1400 }
1401
1402 static int set_sregs_arch206(struct kvm_vcpu *vcpu,
1403                              struct kvm_sregs *sregs)
1404 {
1405         if (!(sregs->u.e.features & KVM_SREGS_E_ARCH206))
1406                 return 0;
1407
1408         if (sregs->u.e.pir != vcpu->vcpu_id)
1409                 return -EINVAL;
1410
1411         vcpu->arch.mcsrr0 = sregs->u.e.mcsrr0;
1412         vcpu->arch.mcsrr1 = sregs->u.e.mcsrr1;
1413         vcpu->arch.decar = sregs->u.e.decar;
1414         vcpu->arch.ivpr = sregs->u.e.ivpr;
1415
1416         return 0;
1417 }
1418
1419 int kvmppc_get_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1420 {
1421         sregs->u.e.features |= KVM_SREGS_E_IVOR;
1422
1423         sregs->u.e.ivor_low[0] = vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL];
1424         sregs->u.e.ivor_low[1] = vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK];
1425         sregs->u.e.ivor_low[2] = vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE];
1426         sregs->u.e.ivor_low[3] = vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE];
1427         sregs->u.e.ivor_low[4] = vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL];
1428         sregs->u.e.ivor_low[5] = vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT];
1429         sregs->u.e.ivor_low[6] = vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM];
1430         sregs->u.e.ivor_low[7] = vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL];
1431         sregs->u.e.ivor_low[8] = vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL];
1432         sregs->u.e.ivor_low[9] = vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL];
1433         sregs->u.e.ivor_low[10] = vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER];
1434         sregs->u.e.ivor_low[11] = vcpu->arch.ivor[BOOKE_IRQPRIO_FIT];
1435         sregs->u.e.ivor_low[12] = vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG];
1436         sregs->u.e.ivor_low[13] = vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS];
1437         sregs->u.e.ivor_low[14] = vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS];
1438         sregs->u.e.ivor_low[15] = vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG];
1439         return 0;
1440 }
1441
1442 int kvmppc_set_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1443 {
1444         if (!(sregs->u.e.features & KVM_SREGS_E_IVOR))
1445                 return 0;
1446
1447         vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL] = sregs->u.e.ivor_low[0];
1448         vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK] = sregs->u.e.ivor_low[1];
1449         vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE] = sregs->u.e.ivor_low[2];
1450         vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE] = sregs->u.e.ivor_low[3];
1451         vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL] = sregs->u.e.ivor_low[4];
1452         vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT] = sregs->u.e.ivor_low[5];
1453         vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM] = sregs->u.e.ivor_low[6];
1454         vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL] = sregs->u.e.ivor_low[7];
1455         vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL] = sregs->u.e.ivor_low[8];
1456         vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL] = sregs->u.e.ivor_low[9];
1457         vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER] = sregs->u.e.ivor_low[10];
1458         vcpu->arch.ivor[BOOKE_IRQPRIO_FIT] = sregs->u.e.ivor_low[11];
1459         vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG] = sregs->u.e.ivor_low[12];
1460         vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS] = sregs->u.e.ivor_low[13];
1461         vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS] = sregs->u.e.ivor_low[14];
1462         vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG] = sregs->u.e.ivor_low[15];
1463
1464         return 0;
1465 }
1466
1467 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1468                                   struct kvm_sregs *sregs)
1469 {
1470         sregs->pvr = vcpu->arch.pvr;
1471
1472         get_sregs_base(vcpu, sregs);
1473         get_sregs_arch206(vcpu, sregs);
1474         return vcpu->kvm->arch.kvm_ops->get_sregs(vcpu, sregs);
1475 }
1476
1477 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1478                                   struct kvm_sregs *sregs)
1479 {
1480         int ret;
1481
1482         if (vcpu->arch.pvr != sregs->pvr)
1483                 return -EINVAL;
1484
1485         ret = set_sregs_base(vcpu, sregs);
1486         if (ret < 0)
1487                 return ret;
1488
1489         ret = set_sregs_arch206(vcpu, sregs);
1490         if (ret < 0)
1491                 return ret;
1492
1493         return vcpu->kvm->arch.kvm_ops->set_sregs(vcpu, sregs);
1494 }
1495
1496 int kvm_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1497 {
1498         int r = 0;
1499         union kvmppc_one_reg val;
1500         int size;
1501
1502         size = one_reg_size(reg->id);
1503         if (size > sizeof(val))
1504                 return -EINVAL;
1505
1506         switch (reg->id) {
1507         case KVM_REG_PPC_IAC1:
1508                 val = get_reg_val(reg->id, vcpu->arch.dbg_reg.iac1);
1509                 break;
1510         case KVM_REG_PPC_IAC2:
1511                 val = get_reg_val(reg->id, vcpu->arch.dbg_reg.iac2);
1512                 break;
1513 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1514         case KVM_REG_PPC_IAC3:
1515                 val = get_reg_val(reg->id, vcpu->arch.dbg_reg.iac3);
1516                 break;
1517         case KVM_REG_PPC_IAC4:
1518                 val = get_reg_val(reg->id, vcpu->arch.dbg_reg.iac4);
1519                 break;
1520 #endif
1521         case KVM_REG_PPC_DAC1:
1522                 val = get_reg_val(reg->id, vcpu->arch.dbg_reg.dac1);
1523                 break;
1524         case KVM_REG_PPC_DAC2:
1525                 val = get_reg_val(reg->id, vcpu->arch.dbg_reg.dac2);
1526                 break;
1527         case KVM_REG_PPC_EPR: {
1528                 u32 epr = get_guest_epr(vcpu);
1529                 val = get_reg_val(reg->id, epr);
1530                 break;
1531         }
1532 #if defined(CONFIG_64BIT)
1533         case KVM_REG_PPC_EPCR:
1534                 val = get_reg_val(reg->id, vcpu->arch.epcr);
1535                 break;
1536 #endif
1537         case KVM_REG_PPC_TCR:
1538                 val = get_reg_val(reg->id, vcpu->arch.tcr);
1539                 break;
1540         case KVM_REG_PPC_TSR:
1541                 val = get_reg_val(reg->id, vcpu->arch.tsr);
1542                 break;
1543         case KVM_REG_PPC_DEBUG_INST:
1544                 val = get_reg_val(reg->id, KVMPPC_INST_EHPRIV_DEBUG);
1545                 break;
1546         case KVM_REG_PPC_VRSAVE:
1547                 val = get_reg_val(reg->id, vcpu->arch.vrsave);
1548                 break;
1549         default:
1550                 r = vcpu->kvm->arch.kvm_ops->get_one_reg(vcpu, reg->id, &val);
1551                 break;
1552         }
1553
1554         if (r)
1555                 return r;
1556
1557         if (copy_to_user((char __user *)(unsigned long)reg->addr, &val, size))
1558                 r = -EFAULT;
1559
1560         return r;
1561 }
1562
1563 int kvm_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1564 {
1565         int r = 0;
1566         union kvmppc_one_reg val;
1567         int size;
1568
1569         size = one_reg_size(reg->id);
1570         if (size > sizeof(val))
1571                 return -EINVAL;
1572
1573         if (copy_from_user(&val, (char __user *)(unsigned long)reg->addr, size))
1574                 return -EFAULT;
1575
1576         switch (reg->id) {
1577         case KVM_REG_PPC_IAC1:
1578                 vcpu->arch.dbg_reg.iac1 = set_reg_val(reg->id, val);
1579                 break;
1580         case KVM_REG_PPC_IAC2:
1581                 vcpu->arch.dbg_reg.iac2 = set_reg_val(reg->id, val);
1582                 break;
1583 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1584         case KVM_REG_PPC_IAC3:
1585                 vcpu->arch.dbg_reg.iac3 = set_reg_val(reg->id, val);
1586                 break;
1587         case KVM_REG_PPC_IAC4:
1588                 vcpu->arch.dbg_reg.iac4 = set_reg_val(reg->id, val);
1589                 break;
1590 #endif
1591         case KVM_REG_PPC_DAC1:
1592                 vcpu->arch.dbg_reg.dac1 = set_reg_val(reg->id, val);
1593                 break;
1594         case KVM_REG_PPC_DAC2:
1595                 vcpu->arch.dbg_reg.dac2 = set_reg_val(reg->id, val);
1596                 break;
1597         case KVM_REG_PPC_EPR: {
1598                 u32 new_epr = set_reg_val(reg->id, val);
1599                 kvmppc_set_epr(vcpu, new_epr);
1600                 break;
1601         }
1602 #if defined(CONFIG_64BIT)
1603         case KVM_REG_PPC_EPCR: {
1604                 u32 new_epcr = set_reg_val(reg->id, val);
1605                 kvmppc_set_epcr(vcpu, new_epcr);
1606                 break;
1607         }
1608 #endif
1609         case KVM_REG_PPC_OR_TSR: {
1610                 u32 tsr_bits = set_reg_val(reg->id, val);
1611                 kvmppc_set_tsr_bits(vcpu, tsr_bits);
1612                 break;
1613         }
1614         case KVM_REG_PPC_CLEAR_TSR: {
1615                 u32 tsr_bits = set_reg_val(reg->id, val);
1616                 kvmppc_clr_tsr_bits(vcpu, tsr_bits);
1617                 break;
1618         }
1619         case KVM_REG_PPC_TSR: {
1620                 u32 tsr = set_reg_val(reg->id, val);
1621                 kvmppc_set_tsr(vcpu, tsr);
1622                 break;
1623         }
1624         case KVM_REG_PPC_TCR: {
1625                 u32 tcr = set_reg_val(reg->id, val);
1626                 kvmppc_set_tcr(vcpu, tcr);
1627                 break;
1628         }
1629         case KVM_REG_PPC_VRSAVE:
1630                 vcpu->arch.vrsave = set_reg_val(reg->id, val);
1631                 break;
1632         default:
1633                 r = vcpu->kvm->arch.kvm_ops->set_one_reg(vcpu, reg->id, &val);
1634                 break;
1635         }
1636
1637         return r;
1638 }
1639
1640 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1641 {
1642         return -ENOTSUPP;
1643 }
1644
1645 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1646 {
1647         return -ENOTSUPP;
1648 }
1649
1650 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1651                                   struct kvm_translation *tr)
1652 {
1653         int r;
1654
1655         r = kvmppc_core_vcpu_translate(vcpu, tr);
1656         return r;
1657 }
1658
1659 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
1660 {
1661         return -ENOTSUPP;
1662 }
1663
1664 void kvmppc_core_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
1665                               struct kvm_memory_slot *dont)
1666 {
1667 }
1668
1669 int kvmppc_core_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
1670                                unsigned long npages)
1671 {
1672         return 0;
1673 }
1674
1675 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
1676                                       struct kvm_memory_slot *memslot,
1677                                       struct kvm_userspace_memory_region *mem)
1678 {
1679         return 0;
1680 }
1681
1682 void kvmppc_core_commit_memory_region(struct kvm *kvm,
1683                                 struct kvm_userspace_memory_region *mem,
1684                                 const struct kvm_memory_slot *old)
1685 {
1686 }
1687
1688 void kvmppc_core_flush_memslot(struct kvm *kvm, struct kvm_memory_slot *memslot)
1689 {
1690 }
1691
1692 void kvmppc_set_epcr(struct kvm_vcpu *vcpu, u32 new_epcr)
1693 {
1694 #if defined(CONFIG_64BIT)
1695         vcpu->arch.epcr = new_epcr;
1696 #ifdef CONFIG_KVM_BOOKE_HV
1697         vcpu->arch.shadow_epcr &= ~SPRN_EPCR_GICM;
1698         if (vcpu->arch.epcr  & SPRN_EPCR_ICM)
1699                 vcpu->arch.shadow_epcr |= SPRN_EPCR_GICM;
1700 #endif
1701 #endif
1702 }
1703
1704 void kvmppc_set_tcr(struct kvm_vcpu *vcpu, u32 new_tcr)
1705 {
1706         vcpu->arch.tcr = new_tcr;
1707         arm_next_watchdog(vcpu);
1708         update_timer_ints(vcpu);
1709 }
1710
1711 void kvmppc_set_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1712 {
1713         set_bits(tsr_bits, &vcpu->arch.tsr);
1714         smp_wmb();
1715         kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
1716         kvm_vcpu_kick(vcpu);
1717 }
1718
1719 void kvmppc_clr_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1720 {
1721         clear_bits(tsr_bits, &vcpu->arch.tsr);
1722
1723         /*
1724          * We may have stopped the watchdog due to
1725          * being stuck on final expiration.
1726          */
1727         if (tsr_bits & (TSR_ENW | TSR_WIS))
1728                 arm_next_watchdog(vcpu);
1729
1730         update_timer_ints(vcpu);
1731 }
1732
1733 void kvmppc_decrementer_func(unsigned long data)
1734 {
1735         struct kvm_vcpu *vcpu = (struct kvm_vcpu *)data;
1736
1737         if (vcpu->arch.tcr & TCR_ARE) {
1738                 vcpu->arch.dec = vcpu->arch.decar;
1739                 kvmppc_emulate_dec(vcpu);
1740         }
1741
1742         kvmppc_set_tsr_bits(vcpu, TSR_DIS);
1743 }
1744
1745 static int kvmppc_booke_add_breakpoint(struct debug_reg *dbg_reg,
1746                                        uint64_t addr, int index)
1747 {
1748         switch (index) {
1749         case 0:
1750                 dbg_reg->dbcr0 |= DBCR0_IAC1;
1751                 dbg_reg->iac1 = addr;
1752                 break;
1753         case 1:
1754                 dbg_reg->dbcr0 |= DBCR0_IAC2;
1755                 dbg_reg->iac2 = addr;
1756                 break;
1757 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1758         case 2:
1759                 dbg_reg->dbcr0 |= DBCR0_IAC3;
1760                 dbg_reg->iac3 = addr;
1761                 break;
1762         case 3:
1763                 dbg_reg->dbcr0 |= DBCR0_IAC4;
1764                 dbg_reg->iac4 = addr;
1765                 break;
1766 #endif
1767         default:
1768                 return -EINVAL;
1769         }
1770
1771         dbg_reg->dbcr0 |= DBCR0_IDM;
1772         return 0;
1773 }
1774
1775 static int kvmppc_booke_add_watchpoint(struct debug_reg *dbg_reg, uint64_t addr,
1776                                        int type, int index)
1777 {
1778         switch (index) {
1779         case 0:
1780                 if (type & KVMPPC_DEBUG_WATCH_READ)
1781                         dbg_reg->dbcr0 |= DBCR0_DAC1R;
1782                 if (type & KVMPPC_DEBUG_WATCH_WRITE)
1783                         dbg_reg->dbcr0 |= DBCR0_DAC1W;
1784                 dbg_reg->dac1 = addr;
1785                 break;
1786         case 1:
1787                 if (type & KVMPPC_DEBUG_WATCH_READ)
1788                         dbg_reg->dbcr0 |= DBCR0_DAC2R;
1789                 if (type & KVMPPC_DEBUG_WATCH_WRITE)
1790                         dbg_reg->dbcr0 |= DBCR0_DAC2W;
1791                 dbg_reg->dac2 = addr;
1792                 break;
1793         default:
1794                 return -EINVAL;
1795         }
1796
1797         dbg_reg->dbcr0 |= DBCR0_IDM;
1798         return 0;
1799 }
1800 void kvm_guest_protect_msr(struct kvm_vcpu *vcpu, ulong prot_bitmap, bool set)
1801 {
1802         /* XXX: Add similar MSR protection for BookE-PR */
1803 #ifdef CONFIG_KVM_BOOKE_HV
1804         BUG_ON(prot_bitmap & ~(MSRP_UCLEP | MSRP_DEP | MSRP_PMMP));
1805         if (set) {
1806                 if (prot_bitmap & MSR_UCLE)
1807                         vcpu->arch.shadow_msrp |= MSRP_UCLEP;
1808                 if (prot_bitmap & MSR_DE)
1809                         vcpu->arch.shadow_msrp |= MSRP_DEP;
1810                 if (prot_bitmap & MSR_PMM)
1811                         vcpu->arch.shadow_msrp |= MSRP_PMMP;
1812         } else {
1813                 if (prot_bitmap & MSR_UCLE)
1814                         vcpu->arch.shadow_msrp &= ~MSRP_UCLEP;
1815                 if (prot_bitmap & MSR_DE)
1816                         vcpu->arch.shadow_msrp &= ~MSRP_DEP;
1817                 if (prot_bitmap & MSR_PMM)
1818                         vcpu->arch.shadow_msrp &= ~MSRP_PMMP;
1819         }
1820 #endif
1821 }
1822
1823 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
1824                                          struct kvm_guest_debug *dbg)
1825 {
1826         struct debug_reg *dbg_reg;
1827         int n, b = 0, w = 0;
1828
1829         if (!(dbg->control & KVM_GUESTDBG_ENABLE)) {
1830                 vcpu->arch.shadow_dbg_reg.dbcr0 = 0;
1831                 vcpu->guest_debug = 0;
1832                 kvm_guest_protect_msr(vcpu, MSR_DE, false);
1833                 return 0;
1834         }
1835
1836         kvm_guest_protect_msr(vcpu, MSR_DE, true);
1837         vcpu->guest_debug = dbg->control;
1838         vcpu->arch.shadow_dbg_reg.dbcr0 = 0;
1839         /* Set DBCR0_EDM in guest visible DBCR0 register. */
1840         vcpu->arch.dbg_reg.dbcr0 = DBCR0_EDM;
1841
1842         if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP)
1843                 vcpu->arch.shadow_dbg_reg.dbcr0 |= DBCR0_IDM | DBCR0_IC;
1844
1845         /* Code below handles only HW breakpoints */
1846         dbg_reg = &(vcpu->arch.shadow_dbg_reg);
1847
1848 #ifdef CONFIG_KVM_BOOKE_HV
1849         /*
1850          * On BookE-HV (e500mc) the guest is always executed with MSR.GS=1
1851          * DBCR1 and DBCR2 are set to trigger debug events when MSR.PR is 0
1852          */
1853         dbg_reg->dbcr1 = 0;
1854         dbg_reg->dbcr2 = 0;
1855 #else
1856         /*
1857          * On BookE-PR (e500v2) the guest is always executed with MSR.PR=1
1858          * We set DBCR1 and DBCR2 to only trigger debug events when MSR.PR
1859          * is set.
1860          */
1861         dbg_reg->dbcr1 = DBCR1_IAC1US | DBCR1_IAC2US | DBCR1_IAC3US |
1862                           DBCR1_IAC4US;
1863         dbg_reg->dbcr2 = DBCR2_DAC1US | DBCR2_DAC2US;
1864 #endif
1865
1866         if (!(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP))
1867                 return 0;
1868
1869         for (n = 0; n < (KVMPPC_BOOKE_IAC_NUM + KVMPPC_BOOKE_DAC_NUM); n++) {
1870                 uint64_t addr = dbg->arch.bp[n].addr;
1871                 uint32_t type = dbg->arch.bp[n].type;
1872
1873                 if (type == KVMPPC_DEBUG_NONE)
1874                         continue;
1875
1876                 if (type & !(KVMPPC_DEBUG_WATCH_READ |
1877                              KVMPPC_DEBUG_WATCH_WRITE |
1878                              KVMPPC_DEBUG_BREAKPOINT))
1879                         return -EINVAL;
1880
1881                 if (type & KVMPPC_DEBUG_BREAKPOINT) {
1882                         /* Setting H/W breakpoint */
1883                         if (kvmppc_booke_add_breakpoint(dbg_reg, addr, b++))
1884                                 return -EINVAL;
1885                 } else {
1886                         /* Setting H/W watchpoint */
1887                         if (kvmppc_booke_add_watchpoint(dbg_reg, addr,
1888                                                         type, w++))
1889                                 return -EINVAL;
1890                 }
1891         }
1892
1893         return 0;
1894 }
1895
1896 void kvmppc_booke_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
1897 {
1898         vcpu->cpu = smp_processor_id();
1899         current->thread.kvm_vcpu = vcpu;
1900 }
1901
1902 void kvmppc_booke_vcpu_put(struct kvm_vcpu *vcpu)
1903 {
1904         current->thread.kvm_vcpu = NULL;
1905         vcpu->cpu = -1;
1906
1907         /* Clear pending debug event in DBSR */
1908         kvmppc_clear_dbsr();
1909 }
1910
1911 void kvmppc_mmu_destroy(struct kvm_vcpu *vcpu)
1912 {
1913         vcpu->kvm->arch.kvm_ops->mmu_destroy(vcpu);
1914 }
1915
1916 int kvmppc_core_init_vm(struct kvm *kvm)
1917 {
1918         return kvm->arch.kvm_ops->init_vm(kvm);
1919 }
1920
1921 struct kvm_vcpu *kvmppc_core_vcpu_create(struct kvm *kvm, unsigned int id)
1922 {
1923         return kvm->arch.kvm_ops->vcpu_create(kvm, id);
1924 }
1925
1926 void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
1927 {
1928         vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
1929 }
1930
1931 void kvmppc_core_destroy_vm(struct kvm *kvm)
1932 {
1933         kvm->arch.kvm_ops->destroy_vm(kvm);
1934 }
1935
1936 void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
1937 {
1938         vcpu->kvm->arch.kvm_ops->vcpu_load(vcpu, cpu);
1939 }
1940
1941 void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
1942 {
1943         vcpu->kvm->arch.kvm_ops->vcpu_put(vcpu);
1944 }
1945
1946 int __init kvmppc_booke_init(void)
1947 {
1948 #ifndef CONFIG_KVM_BOOKE_HV
1949         unsigned long ivor[16];
1950         unsigned long *handler = kvmppc_booke_handler_addr;
1951         unsigned long max_ivor = 0;
1952         unsigned long handler_len;
1953         int i;
1954
1955         /* We install our own exception handlers by hijacking IVPR. IVPR must
1956          * be 16-bit aligned, so we need a 64KB allocation. */
1957         kvmppc_booke_handlers = __get_free_pages(GFP_KERNEL | __GFP_ZERO,
1958                                                  VCPU_SIZE_ORDER);
1959         if (!kvmppc_booke_handlers)
1960                 return -ENOMEM;
1961
1962         /* XXX make sure our handlers are smaller than Linux's */
1963
1964         /* Copy our interrupt handlers to match host IVORs. That way we don't
1965          * have to swap the IVORs on every guest/host transition. */
1966         ivor[0] = mfspr(SPRN_IVOR0);
1967         ivor[1] = mfspr(SPRN_IVOR1);
1968         ivor[2] = mfspr(SPRN_IVOR2);
1969         ivor[3] = mfspr(SPRN_IVOR3);
1970         ivor[4] = mfspr(SPRN_IVOR4);
1971         ivor[5] = mfspr(SPRN_IVOR5);
1972         ivor[6] = mfspr(SPRN_IVOR6);
1973         ivor[7] = mfspr(SPRN_IVOR7);
1974         ivor[8] = mfspr(SPRN_IVOR8);
1975         ivor[9] = mfspr(SPRN_IVOR9);
1976         ivor[10] = mfspr(SPRN_IVOR10);
1977         ivor[11] = mfspr(SPRN_IVOR11);
1978         ivor[12] = mfspr(SPRN_IVOR12);
1979         ivor[13] = mfspr(SPRN_IVOR13);
1980         ivor[14] = mfspr(SPRN_IVOR14);
1981         ivor[15] = mfspr(SPRN_IVOR15);
1982
1983         for (i = 0; i < 16; i++) {
1984                 if (ivor[i] > max_ivor)
1985                         max_ivor = i;
1986
1987                 handler_len = handler[i + 1] - handler[i];
1988                 memcpy((void *)kvmppc_booke_handlers + ivor[i],
1989                        (void *)handler[i], handler_len);
1990         }
1991
1992         handler_len = handler[max_ivor + 1] - handler[max_ivor];
1993         flush_icache_range(kvmppc_booke_handlers, kvmppc_booke_handlers +
1994                            ivor[max_ivor] + handler_len);
1995 #endif /* !BOOKE_HV */
1996         return 0;
1997 }
1998
1999 void __exit kvmppc_booke_exit(void)
2000 {
2001         free_pages(kvmppc_booke_handlers, VCPU_SIZE_ORDER);
2002         kvm_exit();
2003 }