target-s390x/kvm.c

   1 /*
   2  * QEMU S390x KVM implementation
   3  *
   4  * Copyright (c) 2009 Alexander Graf <agraf@suse.de>
   5  * Copyright IBM Corp. 2012
   6  *
   7  * This library is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU Lesser General Public
   9  * License as published by the Free Software Foundation; either
  10  * version 2 of the License, or (at your option) any later version.
  11  *
  12  * This library is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * Lesser General Public License for more details.
  16  *
  17  * Contributions after 2012-10-29 are licensed under the terms of the
  18  * GNU GPL, version 2 or (at your option) any later version.
  19  *
  20  * You should have received a copy of the GNU (Lesser) General Public
  21  * License along with this library; if not, see <http://www.gnu.org/licenses/>.
  22  */
  23
  24 #include <sys/types.h>
  25 #include <sys/ioctl.h>
  26 #include <sys/mman.h>
  27
  28 #include <linux/kvm.h>
  29 #include <asm/ptrace.h>
  30
  31 #include "qemu-common.h"
  32 #include "qemu/timer.h"
  33 #include "sysemu/sysemu.h"
  34 #include "sysemu/kvm.h"
  35 #include "cpu.h"
  36 #include "sysemu/device_tree.h"
  37 #include "qapi/qmp/qjson.h"
  38 #include "monitor/monitor.h"
  39 #include "trace.h"
  40
  41 /* #define DEBUG_KVM */
  42
  43 #ifdef DEBUG_KVM
  44 #define DPRINTF(fmt, ...) \
  45     do { fprintf(stderr, fmt, ## __VA_ARGS__); } while (0)
  46 #else
  47 #define DPRINTF(fmt, ...) \
  48     do { } while (0)
  49 #endif
  50
  51 #define IPA0_DIAG                       0x8300
  52 #define IPA0_SIGP                       0xae00
  53 #define IPA0_B2                         0xb200
  54 #define IPA0_B9                         0xb900
  55 #define IPA0_EB                         0xeb00
  56
  57 #define PRIV_B2_SCLP_CALL               0x20
  58 #define PRIV_B2_CSCH                    0x30
  59 #define PRIV_B2_HSCH                    0x31
  60 #define PRIV_B2_MSCH                    0x32
  61 #define PRIV_B2_SSCH                    0x33
  62 #define PRIV_B2_STSCH                   0x34
  63 #define PRIV_B2_TSCH                    0x35
  64 #define PRIV_B2_TPI                     0x36
  65 #define PRIV_B2_SAL                     0x37
  66 #define PRIV_B2_RSCH                    0x38
  67 #define PRIV_B2_STCRW                   0x39
  68 #define PRIV_B2_STCPS                   0x3a
  69 #define PRIV_B2_RCHP                    0x3b
  70 #define PRIV_B2_SCHM                    0x3c
  71 #define PRIV_B2_CHSC                    0x5f
  72 #define PRIV_B2_SIGA                    0x74
  73 #define PRIV_B2_XSCH                    0x76
  74
  75 #define PRIV_EB_SQBS                    0x8a
  76
  77 #define PRIV_B9_EQBS                    0x9c
  78
  79 #define DIAG_IPL                        0x308
  80 #define DIAG_KVM_HYPERCALL              0x500
  81 #define DIAG_KVM_BREAKPOINT             0x501
  82
  83 #define ICPT_INSTRUCTION                0x04
  84 #define ICPT_WAITPSW                    0x1c
  85 #define ICPT_SOFT_INTERCEPT             0x24
  86 #define ICPT_CPU_STOP                   0x28
  87 #define ICPT_IO                         0x40
  88
  89 const KVMCapabilityInfo kvm_arch_required_capabilities[] = {
  90     KVM_CAP_LAST_INFO
  91 };
  92
  93 static int cap_sync_regs;
  94 static int cap_async_pf;
  95
  96 static void *legacy_s390_alloc(size_t size);
  97
  98 int kvm_arch_init(KVMState *s)
  99 {
 100     cap_sync_regs = kvm_check_extension(s, KVM_CAP_SYNC_REGS);
 101     cap_async_pf = kvm_check_extension(s, KVM_CAP_ASYNC_PF);
 102     if (!kvm_check_extension(s, KVM_CAP_S390_GMAP)
 103         || !kvm_check_extension(s, KVM_CAP_S390_COW)) {
 104         phys_mem_set_alloc(legacy_s390_alloc);
 105     }
 106     return 0;
 107 }
 108
 109 unsigned long kvm_arch_vcpu_id(CPUState *cpu)
 110 {
 111     return cpu->cpu_index;
 112 }
 113
 114 int kvm_arch_init_vcpu(CPUState *cpu)
 115 {
 116     /* nothing todo yet */
 117     return 0;
 118 }
 119
 120 void kvm_arch_reset_vcpu(CPUState *cpu)
 121 {
 122     /* The initial reset call is needed here to reset in-kernel
 123      * vcpu data that we can't access directly from QEMU
 124      * (i.e. with older kernels which don't support sync_regs/ONE_REG).
 125      * Before this ioctl cpu_synchronize_state() is called in common kvm
 126      * code (kvm-all) */
 127     if (kvm_vcpu_ioctl(cpu, KVM_S390_INITIAL_RESET, NULL)) {
 128         perror("Can't reset vcpu\n");
 129     }
 130 }
 131
 132 static int kvm_set_one_reg(CPUState *cs, uint64_t id, void *source)
 133 {
 134     struct kvm_one_reg reg;
 135     int r;
 136
 137     reg.id = id;
 138     reg.addr = (uint64_t) source;
 139     r = kvm_vcpu_ioctl(cs, KVM_SET_ONE_REG, &reg);
 140     if (r) {
 141         trace_kvm_failed_reg_set(id, strerror(errno));
 142     }
 143     return r;
 144 }
 145
 146 static int kvm_get_one_reg(CPUState *cs, uint64_t id, void *target)
 147 {
 148     struct kvm_one_reg reg;
 149     int r;
 150
 151     reg.id = id;
 152     reg.addr = (uint64_t) target;
 153     r = kvm_vcpu_ioctl(cs, KVM_GET_ONE_REG, &reg);
 154     if (r) {
 155         trace_kvm_failed_reg_get(id, strerror(errno));
 156     }
 157     return r;
 158 }
 159
 160
 161 int kvm_arch_put_registers(CPUState *cs, int level)
 162 {
 163     S390CPU *cpu = S390_CPU(cs);
 164     CPUS390XState *env = &cpu->env;
 165     struct kvm_sregs sregs;
 166     struct kvm_regs regs;
 167     int r;
 168     int i;
 169
 170     /* always save the PSW  and the GPRS*/
 171     cs->kvm_run->psw_addr = env->psw.addr;
 172     cs->kvm_run->psw_mask = env->psw.mask;
 173
 174     if (cap_sync_regs && cs->kvm_run->kvm_valid_regs & KVM_SYNC_GPRS) {
 175         for (i = 0; i < 16; i++) {
 176             cs->kvm_run->s.regs.gprs[i] = env->regs[i];
 177             cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_GPRS;
 178         }
 179     } else {
 180         for (i = 0; i < 16; i++) {
 181             regs.gprs[i] = env->regs[i];
 182         }
 183         r = kvm_vcpu_ioctl(cs, KVM_SET_REGS, &regs);
 184         if (r < 0) {
 185             return r;
 186         }
 187     }
 188
 189     /* Do we need to save more than that? */
 190     if (level == KVM_PUT_RUNTIME_STATE) {
 191         return 0;
 192     }
 193
 194     /*
 195      * These ONE_REGS are not protected by a capability. As they are only
 196      * necessary for migration we just trace a possible error, but don't
 197      * return with an error return code.
 198      */
 199     kvm_set_one_reg(cs, KVM_REG_S390_CPU_TIMER, &env->cputm);
 200     kvm_set_one_reg(cs, KVM_REG_S390_CLOCK_COMP, &env->ckc);
 201     kvm_set_one_reg(cs, KVM_REG_S390_TODPR, &env->todpr);
 202     kvm_set_one_reg(cs, KVM_REG_S390_GBEA, &env->gbea);
 203     kvm_set_one_reg(cs, KVM_REG_S390_PP, &env->pp);
 204
 205     if (cap_async_pf) {
 206         r = kvm_set_one_reg(cs, KVM_REG_S390_PFTOKEN, &env->pfault_token);
 207         if (r < 0) {
 208             return r;
 209         }
 210         r = kvm_set_one_reg(cs, KVM_REG_S390_PFCOMPARE, &env->pfault_compare);
 211         if (r < 0) {
 212             return r;
 213         }
 214         r = kvm_set_one_reg(cs, KVM_REG_S390_PFSELECT, &env->pfault_select);
 215         if (r < 0) {
 216             return r;
 217         }
 218     }
 219
 220     if (cap_sync_regs &&
 221         cs->kvm_run->kvm_valid_regs & KVM_SYNC_ACRS &&
 222         cs->kvm_run->kvm_valid_regs & KVM_SYNC_CRS) {
 223         for (i = 0; i < 16; i++) {
 224             cs->kvm_run->s.regs.acrs[i] = env->aregs[i];
 225             cs->kvm_run->s.regs.crs[i] = env->cregs[i];
 226         }
 227         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_ACRS;
 228         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_CRS;
 229     } else {
 230         for (i = 0; i < 16; i++) {
 231             sregs.acrs[i] = env->aregs[i];
 232             sregs.crs[i] = env->cregs[i];
 233         }
 234         r = kvm_vcpu_ioctl(cs, KVM_SET_SREGS, &sregs);
 235         if (r < 0) {
 236             return r;
 237         }
 238     }
 239
 240     /* Finally the prefix */
 241     if (cap_sync_regs && cs->kvm_run->kvm_valid_regs & KVM_SYNC_PREFIX) {
 242         cs->kvm_run->s.regs.prefix = env->psa;
 243         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_PREFIX;
 244     } else {
 245         /* prefix is only supported via sync regs */
 246     }
 247     return 0;
 248 }
 249
 250 int kvm_arch_get_registers(CPUState *cs)
 251 {
 252     S390CPU *cpu = S390_CPU(cs);
 253     CPUS390XState *env = &cpu->env;
 254     struct kvm_sregs sregs;
 255     struct kvm_regs regs;
 256     int i, r;
 257
 258     /* get the PSW */
 259     env->psw.addr = cs->kvm_run->psw_addr;
 260     env->psw.mask = cs->kvm_run->psw_mask;
 261
 262     /* the GPRS */
 263     if (cap_sync_regs && cs->kvm_run->kvm_valid_regs & KVM_SYNC_GPRS) {
 264         for (i = 0; i < 16; i++) {
 265             env->regs[i] = cs->kvm_run->s.regs.gprs[i];
 266         }
 267     } else {
 268         r = kvm_vcpu_ioctl(cs, KVM_GET_REGS, &regs);
 269         if (r < 0) {
 270             return r;
 271         }
 272          for (i = 0; i < 16; i++) {
 273             env->regs[i] = regs.gprs[i];
 274         }
 275     }
 276
 277     /* The ACRS and CRS */
 278     if (cap_sync_regs &&
 279         cs->kvm_run->kvm_valid_regs & KVM_SYNC_ACRS &&
 280         cs->kvm_run->kvm_valid_regs & KVM_SYNC_CRS) {
 281         for (i = 0; i < 16; i++) {
 282             env->aregs[i] = cs->kvm_run->s.regs.acrs[i];
 283             env->cregs[i] = cs->kvm_run->s.regs.crs[i];
 284         }
 285     } else {
 286         r = kvm_vcpu_ioctl(cs, KVM_GET_SREGS, &sregs);
 287         if (r < 0) {
 288             return r;
 289         }
 290          for (i = 0; i < 16; i++) {
 291             env->aregs[i] = sregs.acrs[i];
 292             env->cregs[i] = sregs.crs[i];
 293         }
 294     }
 295
 296     /* The prefix */
 297     if (cap_sync_regs && cs->kvm_run->kvm_valid_regs & KVM_SYNC_PREFIX) {
 298         env->psa = cs->kvm_run->s.regs.prefix;
 299     }
 300
 301     /*
 302      * These ONE_REGS are not protected by a capability. As they are only
 303      * necessary for migration we just trace a possible error, but don't
 304      * return with an error return code.
 305      */
 306     kvm_get_one_reg(cs, KVM_REG_S390_CPU_TIMER, &env->cputm);
 307     kvm_get_one_reg(cs, KVM_REG_S390_CLOCK_COMP, &env->ckc);
 308     kvm_get_one_reg(cs, KVM_REG_S390_TODPR, &env->todpr);
 309     kvm_get_one_reg(cs, KVM_REG_S390_GBEA, &env->gbea);
 310     kvm_get_one_reg(cs, KVM_REG_S390_PP, &env->pp);
 311
 312     if (cap_async_pf) {
 313         r = kvm_get_one_reg(cs, KVM_REG_S390_PFTOKEN, &env->pfault_token);
 314         if (r < 0) {
 315             return r;
 316         }
 317         r = kvm_get_one_reg(cs, KVM_REG_S390_PFCOMPARE, &env->pfault_compare);
 318         if (r < 0) {
 319             return r;
 320         }
 321         r = kvm_get_one_reg(cs, KVM_REG_S390_PFSELECT, &env->pfault_select);
 322         if (r < 0) {
 323             return r;
 324         }
 325     }
 326
 327     return 0;
 328 }
 329
 330 /*
 331  * Legacy layout for s390:
 332  * Older S390 KVM requires the topmost vma of the RAM to be
 333  * smaller than an system defined value, which is at least 256GB.
 334  * Larger systems have larger values. We put the guest between
 335  * the end of data segment (system break) and this value. We
 336  * use 32GB as a base to have enough room for the system break
 337  * to grow. We also have to use MAP parameters that avoid
 338  * read-only mapping of guest pages.
 339  */
 340 static void *legacy_s390_alloc(size_t size)
 341 {
 342     void *mem;
 343
 344     mem = mmap((void *) 0x800000000ULL, size,
 345                PROT_EXEC|PROT_READ|PROT_WRITE,
 346                MAP_SHARED | MAP_ANONYMOUS | MAP_FIXED, -1, 0);
 347     return mem == MAP_FAILED ? NULL : mem;
 348 }
 349
 350 int kvm_arch_insert_sw_breakpoint(CPUState *cs, struct kvm_sw_breakpoint *bp)
 351 {
 352     static const uint8_t diag_501[] = {0x83, 0x24, 0x05, 0x01};
 353
 354     if (cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)&bp->saved_insn, 4, 0) ||
 355         cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)diag_501, 4, 1)) {
 356         return -EINVAL;
 357     }
 358     return 0;
 359 }
 360
 361 int kvm_arch_remove_sw_breakpoint(CPUState *cs, struct kvm_sw_breakpoint *bp)
 362 {
 363     uint8_t t[4];
 364     static const uint8_t diag_501[] = {0x83, 0x24, 0x05, 0x01};
 365
 366     if (cpu_memory_rw_debug(cs, bp->pc, t, 4, 0)) {
 367         return -EINVAL;
 368     } else if (memcmp(t, diag_501, 4)) {
 369         return -EINVAL;
 370     } else if (cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)&bp->saved_insn, 1, 1)) {
 371         return -EINVAL;
 372     }
 373
 374     return 0;
 375 }
 376
 377 int kvm_arch_insert_hw_breakpoint(target_ulong addr,
 378                                   target_ulong len, int type)
 379 {
 380     return -ENOSYS;
 381 }
 382
 383 int kvm_arch_remove_hw_breakpoint(target_ulong addr,
 384                                   target_ulong len, int type)
 385 {
 386     return -ENOSYS;
 387 }
 388
 389 void kvm_arch_remove_all_hw_breakpoints(void)
 390 {
 391 }
 392
 393 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg)
 394 {
 395 }
 396
 397 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run)
 398 {
 399 }
 400
 401 void kvm_arch_post_run(CPUState *cpu, struct kvm_run *run)
 402 {
 403 }
 404
 405 int kvm_arch_process_async_events(CPUState *cs)
 406 {
 407     return cs->halted;
 408 }
 409
 410 void kvm_s390_interrupt_internal(S390CPU *cpu, int type, uint32_t parm,
 411                                  uint64_t parm64, int vm)
 412 {
 413     CPUState *cs = CPU(cpu);
 414     struct kvm_s390_interrupt kvmint;
 415     int r;
 416
 417     if (!cs->kvm_state) {
 418         return;
 419     }
 420
 421     kvmint.type = type;
 422     kvmint.parm = parm;
 423     kvmint.parm64 = parm64;
 424
 425     if (vm) {
 426         r = kvm_vm_ioctl(cs->kvm_state, KVM_S390_INTERRUPT, &kvmint);
 427     } else {
 428         r = kvm_vcpu_ioctl(cs, KVM_S390_INTERRUPT, &kvmint);
 429     }
 430
 431     if (r < 0) {
 432         fprintf(stderr, "KVM failed to inject interrupt\n");
 433         exit(1);
 434     }
 435 }
 436
 437 void kvm_s390_virtio_irq(S390CPU *cpu, int config_change, uint64_t token)
 438 {
 439     kvm_s390_interrupt_internal(cpu, KVM_S390_INT_VIRTIO, config_change,
 440                                 token, 1);
 441 }
 442
 443 void kvm_s390_interrupt(S390CPU *cpu, int type, uint32_t code)
 444 {
 445     kvm_s390_interrupt_internal(cpu, type, code, 0, 0);
 446 }
 447
 448 static void enter_pgmcheck(S390CPU *cpu, uint16_t code)
 449 {
 450     kvm_s390_interrupt(cpu, KVM_S390_PROGRAM_INT, code);
 451 }
 452
 453 static int kvm_sclp_service_call(S390CPU *cpu, struct kvm_run *run,
 454                                  uint16_t ipbh0)
 455 {
 456     CPUS390XState *env = &cpu->env;
 457     uint64_t sccb;
 458     uint32_t code;
 459     int r = 0;
 460
 461     cpu_synchronize_state(CPU(cpu));
 462     sccb = env->regs[ipbh0 & 0xf];
 463     code = env->regs[(ipbh0 & 0xf0) >> 4];
 464
 465     r = sclp_service_call(env, sccb, code);
 466     if (r < 0) {
 467         enter_pgmcheck(cpu, -r);
 468     } else {
 469         setcc(cpu, r);
 470     }
 471
 472     return 0;
 473 }
 474
 475 static int handle_b2(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
 476 {
 477     CPUS390XState *env = &cpu->env;
 478     int rc = 0;
 479     uint16_t ipbh0 = (run->s390_sieic.ipb & 0xffff0000) >> 16;
 480
 481     cpu_synchronize_state(CPU(cpu));
 482
 483     switch (ipa1) {
 484     case PRIV_B2_XSCH:
 485         ioinst_handle_xsch(cpu, env->regs[1]);
 486         break;
 487     case PRIV_B2_CSCH:
 488         ioinst_handle_csch(cpu, env->regs[1]);
 489         break;
 490     case PRIV_B2_HSCH:
 491         ioinst_handle_hsch(cpu, env->regs[1]);
 492         break;
 493     case PRIV_B2_MSCH:
 494         ioinst_handle_msch(cpu, env->regs[1], run->s390_sieic.ipb);
 495         break;
 496     case PRIV_B2_SSCH:
 497         ioinst_handle_ssch(cpu, env->regs[1], run->s390_sieic.ipb);
 498         break;
 499     case PRIV_B2_STCRW:
 500         ioinst_handle_stcrw(cpu, run->s390_sieic.ipb);
 501         break;
 502     case PRIV_B2_STSCH:
 503         ioinst_handle_stsch(cpu, env->regs[1], run->s390_sieic.ipb);
 504         break;
 505     case PRIV_B2_TSCH:
 506         /* We should only get tsch via KVM_EXIT_S390_TSCH. */
 507         fprintf(stderr, "Spurious tsch intercept\n");
 508         break;
 509     case PRIV_B2_CHSC:
 510         ioinst_handle_chsc(cpu, run->s390_sieic.ipb);
 511         break;
 512     case PRIV_B2_TPI:
 513         /* This should have been handled by kvm already. */
 514         fprintf(stderr, "Spurious tpi intercept\n");
 515         break;
 516     case PRIV_B2_SCHM:
 517         ioinst_handle_schm(cpu, env->regs[1], env->regs[2],
 518                            run->s390_sieic.ipb);
 519         break;
 520     case PRIV_B2_RSCH:
 521         ioinst_handle_rsch(cpu, env->regs[1]);
 522         break;
 523     case PRIV_B2_RCHP:
 524         ioinst_handle_rchp(cpu, env->regs[1]);
 525         break;
 526     case PRIV_B2_STCPS:
 527         /* We do not provide this instruction, it is suppressed. */
 528         break;
 529     case PRIV_B2_SAL:
 530         ioinst_handle_sal(cpu, env->regs[1]);
 531         break;
 532     case PRIV_B2_SIGA:
 533         /* Not provided, set CC = 3 for subchannel not operational */
 534         setcc(cpu, 3);
 535         break;
 536     case PRIV_B2_SCLP_CALL:
 537         rc = kvm_sclp_service_call(cpu, run, ipbh0);
 538         break;
 539     default:
 540         rc = -1;
 541         DPRINTF("KVM: unhandled PRIV: 0xb2%x\n", ipa1);
 542         break;
 543     }
 544
 545     return rc;
 546 }
 547
 548 static int handle_b9(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
 549 {
 550     int r = 0;
 551
 552     switch (ipa1) {
 553     case PRIV_B9_EQBS:
 554         /* just inject exception */
 555         r = -1;
 556         break;
 557     default:
 558         r = -1;
 559         DPRINTF("KVM: unhandled PRIV: 0xb9%x\n", ipa1);
 560         break;
 561     }
 562
 563     return r;
 564 }
 565
 566 static int handle_eb(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
 567 {
 568     int r = 0;
 569
 570     switch (ipa1) {
 571     case PRIV_EB_SQBS:
 572         /* just inject exception */
 573         r = -1;
 574         break;
 575     default:
 576         r = -1;
 577         DPRINTF("KVM: unhandled PRIV: 0xeb%x\n", ipa1);
 578         break;
 579     }
 580
 581     return r;
 582 }
 583
 584 static int handle_hypercall(S390CPU *cpu, struct kvm_run *run)
 585 {
 586     CPUS390XState *env = &cpu->env;
 587     int ret;
 588
 589     cpu_synchronize_state(CPU(cpu));
 590     ret = s390_virtio_hypercall(env);
 591     if (ret == -EINVAL) {
 592         enter_pgmcheck(cpu, PGM_SPECIFICATION);
 593         return 0;
 594     }
 595
 596     return ret;
 597 }
 598
 599 static void kvm_handle_diag_308(S390CPU *cpu, struct kvm_run *run)
 600 {
 601     uint64_t r1, r3;
 602
 603     cpu_synchronize_state(CPU(cpu));
 604     r1 = (run->s390_sieic.ipa & 0x00f0) >> 8;
 605     r3 = run->s390_sieic.ipa & 0x000f;
 606     handle_diag_308(&cpu->env, r1, r3);
 607 }
 608
 609 #define DIAG_KVM_CODE_MASK 0x000000000000ffff
 610
 611 static int handle_diag(S390CPU *cpu, struct kvm_run *run, uint32_t ipb)
 612 {
 613     int r = 0;
 614     uint16_t func_code;
 615
 616     /*
 617      * For any diagnose call we support, bits 48-63 of the resulting
 618      * address specify the function code; the remainder is ignored.
 619      */
 620     func_code = decode_basedisp_rs(&cpu->env, ipb) & DIAG_KVM_CODE_MASK;
 621     switch (func_code) {
 622     case DIAG_IPL:
 623         kvm_handle_diag_308(cpu, run);
 624         break;
 625     case DIAG_KVM_HYPERCALL:
 626         r = handle_hypercall(cpu, run);
 627         break;
 628     case DIAG_KVM_BREAKPOINT:
 629         sleep(10);
 630         break;
 631     default:
 632         DPRINTF("KVM: unknown DIAG: 0x%x\n", func_code);
 633         r = -1;
 634         break;
 635     }
 636
 637     return r;
 638 }
 639
 640 static int kvm_s390_cpu_start(S390CPU *cpu)
 641 {
 642     s390_add_running_cpu(cpu);
 643     qemu_cpu_kick(CPU(cpu));
 644     DPRINTF("DONE: KVM cpu start: %p\n", &cpu->env);
 645     return 0;
 646 }
 647
 648 int kvm_s390_cpu_restart(S390CPU *cpu)
 649 {
 650     kvm_s390_interrupt(cpu, KVM_S390_RESTART, 0);
 651     s390_add_running_cpu(cpu);
 652     qemu_cpu_kick(CPU(cpu));
 653     DPRINTF("DONE: KVM cpu restart: %p\n", &cpu->env);
 654     return 0;
 655 }
 656
 657 static void sigp_initial_cpu_reset(void *arg)
 658 {
 659     CPUState *cpu = arg;
 660     S390CPUClass *scc = S390_CPU_GET_CLASS(cpu);
 661
 662     cpu_synchronize_state(cpu);
 663     scc->initial_cpu_reset(cpu);
 664 }
 665
 666 static void sigp_cpu_reset(void *arg)
 667 {
 668     CPUState *cpu = arg;
 669     S390CPUClass *scc = S390_CPU_GET_CLASS(cpu);
 670
 671     cpu_synchronize_state(cpu);
 672     scc->cpu_reset(cpu);
 673 }
 674
 675 #define SIGP_ORDER_MASK 0x000000ff
 676
 677 static int handle_sigp(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
 678 {
 679     CPUS390XState *env = &cpu->env;
 680     uint8_t order_code;
 681     uint16_t cpu_addr;
 682     S390CPU *target_cpu;
 683     uint64_t *statusreg = &env->regs[ipa1 >> 4];
 684     int cc;
 685
 686     cpu_synchronize_state(CPU(cpu));
 687
 688     /* get order code */
 689     order_code = decode_basedisp_rs(env, run->s390_sieic.ipb) & SIGP_ORDER_MASK;
 690
 691     cpu_addr = env->regs[ipa1 & 0x0f];
 692     target_cpu = s390_cpu_addr2state(cpu_addr);
 693     if (target_cpu == NULL) {
 694         cc = 3;    /* not operational */
 695         goto out;
 696     }
 697
 698     switch (order_code) {
 699     case SIGP_START:
 700         cc = kvm_s390_cpu_start(target_cpu);
 701         break;
 702     case SIGP_RESTART:
 703         cc = kvm_s390_cpu_restart(target_cpu);
 704         break;
 705     case SIGP_SET_ARCH:
 706         *statusreg &= 0xffffffff00000000UL;
 707         *statusreg |= SIGP_STAT_INVALID_PARAMETER;
 708         cc = 1;   /* status stored */
 709         break;
 710     case SIGP_INITIAL_CPU_RESET:
 711         run_on_cpu(CPU(target_cpu), sigp_initial_cpu_reset, CPU(target_cpu));
 712         cc = 0;
 713         break;
 714     case SIGP_CPU_RESET:
 715         run_on_cpu(CPU(target_cpu), sigp_cpu_reset, CPU(target_cpu));
 716         cc = 0;
 717         break;
 718     default:
 719         DPRINTF("KVM: unknown SIGP: 0x%x\n", order_code);
 720         *statusreg &= 0xffffffff00000000UL;
 721         *statusreg |= SIGP_STAT_INVALID_ORDER;
 722         cc = 1;   /* status stored */
 723         break;
 724     }
 725
 726 out:
 727     setcc(cpu, cc);
 728     return 0;
 729 }
 730
 731 static void handle_instruction(S390CPU *cpu, struct kvm_run *run)
 732 {
 733     unsigned int ipa0 = (run->s390_sieic.ipa & 0xff00);
 734     uint8_t ipa1 = run->s390_sieic.ipa & 0x00ff;
 735     int r = -1;
 736
 737     DPRINTF("handle_instruction 0x%x 0x%x\n",
 738             run->s390_sieic.ipa, run->s390_sieic.ipb);
 739     switch (ipa0) {
 740     case IPA0_B2:
 741         r = handle_b2(cpu, run, ipa1);
 742         break;
 743     case IPA0_B9:
 744         r = handle_b9(cpu, run, ipa1);
 745         break;
 746     case IPA0_EB:
 747         r = handle_eb(cpu, run, ipa1);
 748         break;
 749     case IPA0_DIAG:
 750         r = handle_diag(cpu, run, run->s390_sieic.ipb);
 751         break;
 752     case IPA0_SIGP:
 753         r = handle_sigp(cpu, run, ipa1);
 754         break;
 755     }
 756
 757     if (r < 0) {
 758         enter_pgmcheck(cpu, 0x0001);
 759     }
 760 }
 761
 762 static bool is_special_wait_psw(CPUState *cs)
 763 {
 764     /* signal quiesce */
 765     return cs->kvm_run->psw_addr == 0xfffUL;
 766 }
 767
 768 static int handle_intercept(S390CPU *cpu)
 769 {
 770     CPUState *cs = CPU(cpu);
 771     struct kvm_run *run = cs->kvm_run;
 772     int icpt_code = run->s390_sieic.icptcode;
 773     int r = 0;
 774
 775     DPRINTF("intercept: 0x%x (at 0x%lx)\n", icpt_code,
 776             (long)cs->kvm_run->psw_addr);
 777     switch (icpt_code) {
 778         case ICPT_INSTRUCTION:
 779             handle_instruction(cpu, run);
 780             break;
 781         case ICPT_WAITPSW:
 782             /* disabled wait, since enabled wait is handled in kernel */
 783             if (s390_del_running_cpu(cpu) == 0) {
 784                 if (is_special_wait_psw(cs)) {
 785                     qemu_system_shutdown_request();
 786                 } else {
 787                     QObject *data;
 788
 789                     data = qobject_from_jsonf("{ 'action': %s }", "pause");
 790                     monitor_protocol_event(QEVENT_GUEST_PANICKED, data);
 791                     qobject_decref(data);
 792                     vm_stop(RUN_STATE_GUEST_PANICKED);
 793                 }
 794             }
 795             r = EXCP_HALTED;
 796             break;
 797         case ICPT_CPU_STOP:
 798             if (s390_del_running_cpu(cpu) == 0) {
 799                 qemu_system_shutdown_request();
 800             }
 801             r = EXCP_HALTED;
 802             break;
 803         case ICPT_SOFT_INTERCEPT:
 804             fprintf(stderr, "KVM unimplemented icpt SOFT\n");
 805             exit(1);
 806             break;
 807         case ICPT_IO:
 808             fprintf(stderr, "KVM unimplemented icpt IO\n");
 809             exit(1);
 810             break;
 811         default:
 812             fprintf(stderr, "Unknown intercept code: %d\n", icpt_code);
 813             exit(1);
 814             break;
 815     }
 816
 817     return r;
 818 }
 819
 820 static int handle_tsch(S390CPU *cpu)
 821 {
 822     CPUS390XState *env = &cpu->env;
 823     CPUState *cs = CPU(cpu);
 824     struct kvm_run *run = cs->kvm_run;
 825     int ret;
 826
 827     cpu_synchronize_state(cs);
 828
 829     ret = ioinst_handle_tsch(env, env->regs[1], run->s390_tsch.ipb);
 830     if (ret >= 0) {
 831         /* Success; set condition code. */
 832         setcc(cpu, ret);
 833         ret = 0;
 834     } else if (ret < -1) {
 835         /*
 836          * Failure.
 837          * If an I/O interrupt had been dequeued, we have to reinject it.
 838          */
 839         if (run->s390_tsch.dequeued) {
 840             uint16_t subchannel_id = run->s390_tsch.subchannel_id;
 841             uint16_t subchannel_nr = run->s390_tsch.subchannel_nr;
 842             uint32_t io_int_parm = run->s390_tsch.io_int_parm;
 843             uint32_t io_int_word = run->s390_tsch.io_int_word;
 844             uint32_t type = ((subchannel_id & 0xff00) << 24) |
 845                 ((subchannel_id & 0x00060) << 22) | (subchannel_nr << 16);
 846
 847             kvm_s390_interrupt_internal(cpu, type,
 848                                         ((uint32_t)subchannel_id << 16)
 849                                         | subchannel_nr,
 850                                         ((uint64_t)io_int_parm << 32)
 851                                         | io_int_word, 1);
 852         }
 853         ret = 0;
 854     }
 855     return ret;
 856 }
 857
 858 static int kvm_arch_handle_debug_exit(S390CPU *cpu)
 859 {
 860     return -ENOSYS;
 861 }
 862
 863 int kvm_arch_handle_exit(CPUState *cs, struct kvm_run *run)
 864 {
 865     S390CPU *cpu = S390_CPU(cs);
 866     int ret = 0;
 867
 868     switch (run->exit_reason) {
 869         case KVM_EXIT_S390_SIEIC:
 870             ret = handle_intercept(cpu);
 871             break;
 872         case KVM_EXIT_S390_RESET:
 873             qemu_system_reset_request();
 874             break;
 875         case KVM_EXIT_S390_TSCH:
 876             ret = handle_tsch(cpu);
 877             break;
 878         case KVM_EXIT_DEBUG:
 879             ret = kvm_arch_handle_debug_exit(cpu);
 880             break;
 881         default:
 882             fprintf(stderr, "Unknown KVM exit: %d\n", run->exit_reason);
 883             break;
 884     }
 885
 886     if (ret == 0) {
 887         ret = EXCP_INTERRUPT;
 888     }
 889     return ret;
 890 }
 891
 892 bool kvm_arch_stop_on_emulation_error(CPUState *cpu)
 893 {
 894     return true;
 895 }
 896
 897 int kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr)
 898 {
 899     return 1;
 900 }
 901
 902 int kvm_arch_on_sigbus(int code, void *addr)
 903 {
 904     return 1;
 905 }
 906
 907 void kvm_s390_io_interrupt(S390CPU *cpu, uint16_t subchannel_id,
 908                            uint16_t subchannel_nr, uint32_t io_int_parm,
 909                            uint32_t io_int_word)
 910 {
 911     uint32_t type;
 912
 913     if (io_int_word & IO_INT_WORD_AI) {
 914         type = KVM_S390_INT_IO(1, 0, 0, 0);
 915     } else {
 916         type = ((subchannel_id & 0xff00) << 24) |
 917             ((subchannel_id & 0x00060) << 22) | (subchannel_nr << 16);
 918     }
 919     kvm_s390_interrupt_internal(cpu, type,
 920                                 ((uint32_t)subchannel_id << 16) | subchannel_nr,
 921                                 ((uint64_t)io_int_parm << 32) | io_int_word, 1);
 922 }
 923
 924 void kvm_s390_crw_mchk(S390CPU *cpu)
 925 {
 926     kvm_s390_interrupt_internal(cpu, KVM_S390_MCHK, 1 << 28,
 927                                 0x00400f1d40330000, 1);
 928 }
 929
 930 void kvm_s390_enable_css_support(S390CPU *cpu)
 931 {
 932     int r;
 933
 934     /* Activate host kernel channel subsystem support. */
 935     r = kvm_vcpu_enable_cap(CPU(cpu), KVM_CAP_S390_CSS_SUPPORT, 0);
 936     assert(r == 0);
 937 }
 938
 939 void kvm_arch_init_irq_routing(KVMState *s)
 940 {
 941 }
 942
 943 int kvm_s390_assign_subch_ioeventfd(EventNotifier *notifier, uint32_t sch,
 944                                     int vq, bool assign)
 945 {
 946     struct kvm_ioeventfd kick = {
 947         .flags = KVM_IOEVENTFD_FLAG_VIRTIO_CCW_NOTIFY |
 948         KVM_IOEVENTFD_FLAG_DATAMATCH,
 949         .fd = event_notifier_get_fd(notifier),
 950         .datamatch = vq,
 951         .addr = sch,
 952         .len = 8,
 953     };
 954     if (!kvm_check_extension(kvm_state, KVM_CAP_IOEVENTFD)) {
 955         return -ENOSYS;
 956     }
 957     if (!assign) {
 958         kick.flags |= KVM_IOEVENTFD_FLAG_DEASSIGN;
 959     }
 960     return kvm_vm_ioctl(kvm_state, KVM_IOEVENTFD, &kick);
 961 }