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 "hw/hw.h"
  36 #include "cpu.h"
  37 #include "sysemu/device_tree.h"
  38 #include "qapi/qmp/qjson.h"
  39 #include "monitor/monitor.h"
  40 #include "exec/gdbstub.h"
  41 #include "trace.h"
  42 #include "qapi-event.h"
  43
  44 /* #define DEBUG_KVM */
  45
  46 #ifdef DEBUG_KVM
  47 #define DPRINTF(fmt, ...) \
  48     do { fprintf(stderr, fmt, ## __VA_ARGS__); } while (0)
  49 #else
  50 #define DPRINTF(fmt, ...) \
  51     do { } while (0)
  52 #endif
  53
  54 #define IPA0_DIAG                       0x8300
  55 #define IPA0_SIGP                       0xae00
  56 #define IPA0_B2                         0xb200
  57 #define IPA0_B9                         0xb900
  58 #define IPA0_EB                         0xeb00
  59
  60 #define PRIV_B2_SCLP_CALL               0x20
  61 #define PRIV_B2_CSCH                    0x30
  62 #define PRIV_B2_HSCH                    0x31
  63 #define PRIV_B2_MSCH                    0x32
  64 #define PRIV_B2_SSCH                    0x33
  65 #define PRIV_B2_STSCH                   0x34
  66 #define PRIV_B2_TSCH                    0x35
  67 #define PRIV_B2_TPI                     0x36
  68 #define PRIV_B2_SAL                     0x37
  69 #define PRIV_B2_RSCH                    0x38
  70 #define PRIV_B2_STCRW                   0x39
  71 #define PRIV_B2_STCPS                   0x3a
  72 #define PRIV_B2_RCHP                    0x3b
  73 #define PRIV_B2_SCHM                    0x3c
  74 #define PRIV_B2_CHSC                    0x5f
  75 #define PRIV_B2_SIGA                    0x74
  76 #define PRIV_B2_XSCH                    0x76
  77
  78 #define PRIV_EB_SQBS                    0x8a
  79
  80 #define PRIV_B9_EQBS                    0x9c
  81
  82 #define DIAG_IPL                        0x308
  83 #define DIAG_KVM_HYPERCALL              0x500
  84 #define DIAG_KVM_BREAKPOINT             0x501
  85
  86 #define ICPT_INSTRUCTION                0x04
  87 #define ICPT_PROGRAM                    0x08
  88 #define ICPT_EXT_INT                    0x14
  89 #define ICPT_WAITPSW                    0x1c
  90 #define ICPT_SOFT_INTERCEPT             0x24
  91 #define ICPT_CPU_STOP                   0x28
  92 #define ICPT_IO                         0x40
  93
  94 static CPUWatchpoint hw_watchpoint;
  95 /*
  96  * We don't use a list because this structure is also used to transmit the
  97  * hardware breakpoints to the kernel.
  98  */
  99 static struct kvm_hw_breakpoint *hw_breakpoints;
 100 static int nb_hw_breakpoints;
 101
 102 const KVMCapabilityInfo kvm_arch_required_capabilities[] = {
 103     KVM_CAP_LAST_INFO
 104 };
 105
 106 static int cap_sync_regs;
 107 static int cap_async_pf;
 108
 109 static void *legacy_s390_alloc(size_t size);
 110
 111 static int kvm_s390_check_clear_cmma(KVMState *s)
 112 {
 113     struct kvm_device_attr attr = {
 114         .group = KVM_S390_VM_MEM_CTRL,
 115         .attr = KVM_S390_VM_MEM_CLR_CMMA,
 116     };
 117
 118     return kvm_vm_ioctl(s, KVM_HAS_DEVICE_ATTR, &attr);
 119 }
 120
 121 static int kvm_s390_check_enable_cmma(KVMState *s)
 122 {
 123     struct kvm_device_attr attr = {
 124         .group = KVM_S390_VM_MEM_CTRL,
 125         .attr = KVM_S390_VM_MEM_ENABLE_CMMA,
 126     };
 127
 128     return kvm_vm_ioctl(s, KVM_HAS_DEVICE_ATTR, &attr);
 129 }
 130
 131 void kvm_s390_clear_cmma_callback(void *opaque)
 132 {
 133     int rc;
 134     KVMState *s = opaque;
 135     struct kvm_device_attr attr = {
 136         .group = KVM_S390_VM_MEM_CTRL,
 137         .attr = KVM_S390_VM_MEM_CLR_CMMA,
 138     };
 139
 140     rc = kvm_vm_ioctl(s, KVM_SET_DEVICE_ATTR, &attr);
 141     trace_kvm_clear_cmma(rc);
 142 }
 143
 144 static void kvm_s390_enable_cmma(KVMState *s)
 145 {
 146     int rc;
 147     struct kvm_device_attr attr = {
 148         .group = KVM_S390_VM_MEM_CTRL,
 149         .attr = KVM_S390_VM_MEM_ENABLE_CMMA,
 150     };
 151
 152     if (kvm_s390_check_enable_cmma(s) || kvm_s390_check_clear_cmma(s)) {
 153         return;
 154     }
 155
 156     rc = kvm_vm_ioctl(s, KVM_SET_DEVICE_ATTR, &attr);
 157     if (!rc) {
 158         qemu_register_reset(kvm_s390_clear_cmma_callback, s);
 159     }
 160     trace_kvm_enable_cmma(rc);
 161 }
 162
 163 int kvm_arch_init(KVMState *s)
 164 {
 165     cap_sync_regs = kvm_check_extension(s, KVM_CAP_SYNC_REGS);
 166     cap_async_pf = kvm_check_extension(s, KVM_CAP_ASYNC_PF);
 167
 168     if (kvm_check_extension(s, KVM_CAP_VM_ATTRIBUTES)) {
 169         kvm_s390_enable_cmma(s);
 170     }
 171
 172     if (!kvm_check_extension(s, KVM_CAP_S390_GMAP)
 173         || !kvm_check_extension(s, KVM_CAP_S390_COW)) {
 174         phys_mem_set_alloc(legacy_s390_alloc);
 175     }
 176     return 0;
 177 }
 178
 179 unsigned long kvm_arch_vcpu_id(CPUState *cpu)
 180 {
 181     return cpu->cpu_index;
 182 }
 183
 184 int kvm_arch_init_vcpu(CPUState *cs)
 185 {
 186     S390CPU *cpu = S390_CPU(cs);
 187     kvm_s390_set_cpu_state(cpu, cpu->env.cpu_state);
 188     return 0;
 189 }
 190
 191 void kvm_s390_reset_vcpu(S390CPU *cpu)
 192 {
 193     CPUState *cs = CPU(cpu);
 194
 195     /* The initial reset call is needed here to reset in-kernel
 196      * vcpu data that we can't access directly from QEMU
 197      * (i.e. with older kernels which don't support sync_regs/ONE_REG).
 198      * Before this ioctl cpu_synchronize_state() is called in common kvm
 199      * code (kvm-all) */
 200     if (kvm_vcpu_ioctl(cs, KVM_S390_INITIAL_RESET, NULL)) {
 201         error_report("Initial CPU reset failed on CPU %i\n", cs->cpu_index);
 202     }
 203 }
 204
 205 int kvm_arch_put_registers(CPUState *cs, int level)
 206 {
 207     S390CPU *cpu = S390_CPU(cs);
 208     CPUS390XState *env = &cpu->env;
 209     struct kvm_sregs sregs;
 210     struct kvm_regs regs;
 211     struct kvm_fpu fpu;
 212     int r;
 213     int i;
 214
 215     /* always save the PSW  and the GPRS*/
 216     cs->kvm_run->psw_addr = env->psw.addr;
 217     cs->kvm_run->psw_mask = env->psw.mask;
 218
 219     if (cap_sync_regs && cs->kvm_run->kvm_valid_regs & KVM_SYNC_GPRS) {
 220         for (i = 0; i < 16; i++) {
 221             cs->kvm_run->s.regs.gprs[i] = env->regs[i];
 222             cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_GPRS;
 223         }
 224     } else {
 225         for (i = 0; i < 16; i++) {
 226             regs.gprs[i] = env->regs[i];
 227         }
 228         r = kvm_vcpu_ioctl(cs, KVM_SET_REGS, &regs);
 229         if (r < 0) {
 230             return r;
 231         }
 232     }
 233
 234     /* Floating point */
 235     for (i = 0; i < 16; i++) {
 236         fpu.fprs[i] = env->fregs[i].ll;
 237     }
 238     fpu.fpc = env->fpc;
 239
 240     r = kvm_vcpu_ioctl(cs, KVM_SET_FPU, &fpu);
 241     if (r < 0) {
 242         return r;
 243     }
 244
 245     /* Do we need to save more than that? */
 246     if (level == KVM_PUT_RUNTIME_STATE) {
 247         return 0;
 248     }
 249
 250     /*
 251      * These ONE_REGS are not protected by a capability. As they are only
 252      * necessary for migration we just trace a possible error, but don't
 253      * return with an error return code.
 254      */
 255     kvm_set_one_reg(cs, KVM_REG_S390_CPU_TIMER, &env->cputm);
 256     kvm_set_one_reg(cs, KVM_REG_S390_CLOCK_COMP, &env->ckc);
 257     kvm_set_one_reg(cs, KVM_REG_S390_TODPR, &env->todpr);
 258     kvm_set_one_reg(cs, KVM_REG_S390_GBEA, &env->gbea);
 259     kvm_set_one_reg(cs, KVM_REG_S390_PP, &env->pp);
 260
 261     if (cap_async_pf) {
 262         r = kvm_set_one_reg(cs, KVM_REG_S390_PFTOKEN, &env->pfault_token);
 263         if (r < 0) {
 264             return r;
 265         }
 266         r = kvm_set_one_reg(cs, KVM_REG_S390_PFCOMPARE, &env->pfault_compare);
 267         if (r < 0) {
 268             return r;
 269         }
 270         r = kvm_set_one_reg(cs, KVM_REG_S390_PFSELECT, &env->pfault_select);
 271         if (r < 0) {
 272             return r;
 273         }
 274     }
 275
 276     if (cap_sync_regs &&
 277         cs->kvm_run->kvm_valid_regs & KVM_SYNC_ACRS &&
 278         cs->kvm_run->kvm_valid_regs & KVM_SYNC_CRS) {
 279         for (i = 0; i < 16; i++) {
 280             cs->kvm_run->s.regs.acrs[i] = env->aregs[i];
 281             cs->kvm_run->s.regs.crs[i] = env->cregs[i];
 282         }
 283         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_ACRS;
 284         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_CRS;
 285     } else {
 286         for (i = 0; i < 16; i++) {
 287             sregs.acrs[i] = env->aregs[i];
 288             sregs.crs[i] = env->cregs[i];
 289         }
 290         r = kvm_vcpu_ioctl(cs, KVM_SET_SREGS, &sregs);
 291         if (r < 0) {
 292             return r;
 293         }
 294     }
 295
 296     /* Finally the prefix */
 297     if (cap_sync_regs && cs->kvm_run->kvm_valid_regs & KVM_SYNC_PREFIX) {
 298         cs->kvm_run->s.regs.prefix = env->psa;
 299         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_PREFIX;
 300     } else {
 301         /* prefix is only supported via sync regs */
 302     }
 303     return 0;
 304 }
 305
 306 int kvm_arch_get_registers(CPUState *cs)
 307 {
 308     S390CPU *cpu = S390_CPU(cs);
 309     CPUS390XState *env = &cpu->env;
 310     struct kvm_sregs sregs;
 311     struct kvm_regs regs;
 312     struct kvm_fpu fpu;
 313     int i, r;
 314
 315     /* get the PSW */
 316     env->psw.addr = cs->kvm_run->psw_addr;
 317     env->psw.mask = cs->kvm_run->psw_mask;
 318
 319     /* the GPRS */
 320     if (cap_sync_regs && cs->kvm_run->kvm_valid_regs & KVM_SYNC_GPRS) {
 321         for (i = 0; i < 16; i++) {
 322             env->regs[i] = cs->kvm_run->s.regs.gprs[i];
 323         }
 324     } else {
 325         r = kvm_vcpu_ioctl(cs, KVM_GET_REGS, &regs);
 326         if (r < 0) {
 327             return r;
 328         }
 329          for (i = 0; i < 16; i++) {
 330             env->regs[i] = regs.gprs[i];
 331         }
 332     }
 333
 334     /* The ACRS and CRS */
 335     if (cap_sync_regs &&
 336         cs->kvm_run->kvm_valid_regs & KVM_SYNC_ACRS &&
 337         cs->kvm_run->kvm_valid_regs & KVM_SYNC_CRS) {
 338         for (i = 0; i < 16; i++) {
 339             env->aregs[i] = cs->kvm_run->s.regs.acrs[i];
 340             env->cregs[i] = cs->kvm_run->s.regs.crs[i];
 341         }
 342     } else {
 343         r = kvm_vcpu_ioctl(cs, KVM_GET_SREGS, &sregs);
 344         if (r < 0) {
 345             return r;
 346         }
 347          for (i = 0; i < 16; i++) {
 348             env->aregs[i] = sregs.acrs[i];
 349             env->cregs[i] = sregs.crs[i];
 350         }
 351     }
 352
 353     /* Floating point */
 354     r = kvm_vcpu_ioctl(cs, KVM_GET_FPU, &fpu);
 355     if (r < 0) {
 356         return r;
 357     }
 358     for (i = 0; i < 16; i++) {
 359         env->fregs[i].ll = fpu.fprs[i];
 360     }
 361     env->fpc = fpu.fpc;
 362
 363     /* The prefix */
 364     if (cap_sync_regs && cs->kvm_run->kvm_valid_regs & KVM_SYNC_PREFIX) {
 365         env->psa = cs->kvm_run->s.regs.prefix;
 366     }
 367
 368     /*
 369      * These ONE_REGS are not protected by a capability. As they are only
 370      * necessary for migration we just trace a possible error, but don't
 371      * return with an error return code.
 372      */
 373     kvm_get_one_reg(cs, KVM_REG_S390_CPU_TIMER, &env->cputm);
 374     kvm_get_one_reg(cs, KVM_REG_S390_CLOCK_COMP, &env->ckc);
 375     kvm_get_one_reg(cs, KVM_REG_S390_TODPR, &env->todpr);
 376     kvm_get_one_reg(cs, KVM_REG_S390_GBEA, &env->gbea);
 377     kvm_get_one_reg(cs, KVM_REG_S390_PP, &env->pp);
 378
 379     if (cap_async_pf) {
 380         r = kvm_get_one_reg(cs, KVM_REG_S390_PFTOKEN, &env->pfault_token);
 381         if (r < 0) {
 382             return r;
 383         }
 384         r = kvm_get_one_reg(cs, KVM_REG_S390_PFCOMPARE, &env->pfault_compare);
 385         if (r < 0) {
 386             return r;
 387         }
 388         r = kvm_get_one_reg(cs, KVM_REG_S390_PFSELECT, &env->pfault_select);
 389         if (r < 0) {
 390             return r;
 391         }
 392     }
 393
 394     return 0;
 395 }
 396
 397 /*
 398  * Legacy layout for s390:
 399  * Older S390 KVM requires the topmost vma of the RAM to be
 400  * smaller than an system defined value, which is at least 256GB.
 401  * Larger systems have larger values. We put the guest between
 402  * the end of data segment (system break) and this value. We
 403  * use 32GB as a base to have enough room for the system break
 404  * to grow. We also have to use MAP parameters that avoid
 405  * read-only mapping of guest pages.
 406  */
 407 static void *legacy_s390_alloc(size_t size, , uint64_t *align)
 408 {
 409     void *mem;
 410
 411     mem = mmap((void *) 0x800000000ULL, size,
 412                PROT_EXEC|PROT_READ|PROT_WRITE,
 413                MAP_SHARED | MAP_ANONYMOUS | MAP_FIXED, -1, 0);
 414     return mem == MAP_FAILED ? NULL : mem;
 415 }
 416
 417 /* DIAG 501 is used for sw breakpoints */
 418 static const uint8_t diag_501[] = {0x83, 0x24, 0x05, 0x01};
 419
 420 int kvm_arch_insert_sw_breakpoint(CPUState *cs, struct kvm_sw_breakpoint *bp)
 421 {
 422
 423     if (cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)&bp->saved_insn,
 424                             sizeof(diag_501), 0) ||
 425         cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)diag_501,
 426                             sizeof(diag_501), 1)) {
 427         return -EINVAL;
 428     }
 429     return 0;
 430 }
 431
 432 int kvm_arch_remove_sw_breakpoint(CPUState *cs, struct kvm_sw_breakpoint *bp)
 433 {
 434     uint8_t t[sizeof(diag_501)];
 435
 436     if (cpu_memory_rw_debug(cs, bp->pc, t, sizeof(diag_501), 0)) {
 437         return -EINVAL;
 438     } else if (memcmp(t, diag_501, sizeof(diag_501))) {
 439         return -EINVAL;
 440     } else if (cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)&bp->saved_insn,
 441                                    sizeof(diag_501), 1)) {
 442         return -EINVAL;
 443     }
 444
 445     return 0;
 446 }
 447
 448 static struct kvm_hw_breakpoint *find_hw_breakpoint(target_ulong addr,
 449                                                     int len, int type)
 450 {
 451     int n;
 452
 453     for (n = 0; n < nb_hw_breakpoints; n++) {
 454         if (hw_breakpoints[n].addr == addr && hw_breakpoints[n].type == type &&
 455             (hw_breakpoints[n].len == len || len == -1)) {
 456             return &hw_breakpoints[n];
 457         }
 458     }
 459
 460     return NULL;
 461 }
 462
 463 static int insert_hw_breakpoint(target_ulong addr, int len, int type)
 464 {
 465     int size;
 466
 467     if (find_hw_breakpoint(addr, len, type)) {
 468         return -EEXIST;
 469     }
 470
 471     size = (nb_hw_breakpoints + 1) * sizeof(struct kvm_hw_breakpoint);
 472
 473     if (!hw_breakpoints) {
 474         nb_hw_breakpoints = 0;
 475         hw_breakpoints = (struct kvm_hw_breakpoint *)g_try_malloc(size);
 476     } else {
 477         hw_breakpoints =
 478             (struct kvm_hw_breakpoint *)g_try_realloc(hw_breakpoints, size);
 479     }
 480
 481     if (!hw_breakpoints) {
 482         nb_hw_breakpoints = 0;
 483         return -ENOMEM;
 484     }
 485
 486     hw_breakpoints[nb_hw_breakpoints].addr = addr;
 487     hw_breakpoints[nb_hw_breakpoints].len = len;
 488     hw_breakpoints[nb_hw_breakpoints].type = type;
 489
 490     nb_hw_breakpoints++;
 491
 492     return 0;
 493 }
 494
 495 int kvm_arch_insert_hw_breakpoint(target_ulong addr,
 496                                   target_ulong len, int type)
 497 {
 498     switch (type) {
 499     case GDB_BREAKPOINT_HW:
 500         type = KVM_HW_BP;
 501         break;
 502     case GDB_WATCHPOINT_WRITE:
 503         if (len < 1) {
 504             return -EINVAL;
 505         }
 506         type = KVM_HW_WP_WRITE;
 507         break;
 508     default:
 509         return -ENOSYS;
 510     }
 511     return insert_hw_breakpoint(addr, len, type);
 512 }
 513
 514 int kvm_arch_remove_hw_breakpoint(target_ulong addr,
 515                                   target_ulong len, int type)
 516 {
 517     int size;
 518     struct kvm_hw_breakpoint *bp = find_hw_breakpoint(addr, len, type);
 519
 520     if (bp == NULL) {
 521         return -ENOENT;
 522     }
 523
 524     nb_hw_breakpoints--;
 525     if (nb_hw_breakpoints > 0) {
 526         /*
 527          * In order to trim the array, move the last element to the position to
 528          * be removed - if necessary.
 529          */
 530         if (bp != &hw_breakpoints[nb_hw_breakpoints]) {
 531             *bp = hw_breakpoints[nb_hw_breakpoints];
 532         }
 533         size = nb_hw_breakpoints * sizeof(struct kvm_hw_breakpoint);
 534         hw_breakpoints =
 535              (struct kvm_hw_breakpoint *)g_realloc(hw_breakpoints, size);
 536     } else {
 537         g_free(hw_breakpoints);
 538         hw_breakpoints = NULL;
 539     }
 540
 541     return 0;
 542 }
 543
 544 void kvm_arch_remove_all_hw_breakpoints(void)
 545 {
 546     nb_hw_breakpoints = 0;
 547     g_free(hw_breakpoints);
 548     hw_breakpoints = NULL;
 549 }
 550
 551 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg)
 552 {
 553     int i;
 554
 555     if (nb_hw_breakpoints > 0) {
 556         dbg->arch.nr_hw_bp = nb_hw_breakpoints;
 557         dbg->arch.hw_bp = hw_breakpoints;
 558
 559         for (i = 0; i < nb_hw_breakpoints; ++i) {
 560             hw_breakpoints[i].phys_addr = s390_cpu_get_phys_addr_debug(cpu,
 561                                                        hw_breakpoints[i].addr);
 562         }
 563         dbg->control |= KVM_GUESTDBG_ENABLE | KVM_GUESTDBG_USE_HW_BP;
 564     } else {
 565         dbg->arch.nr_hw_bp = 0;
 566         dbg->arch.hw_bp = NULL;
 567     }
 568 }
 569
 570 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run)
 571 {
 572 }
 573
 574 void kvm_arch_post_run(CPUState *cpu, struct kvm_run *run)
 575 {
 576 }
 577
 578 int kvm_arch_process_async_events(CPUState *cs)
 579 {
 580     return cs->halted;
 581 }
 582
 583 static int s390_kvm_irq_to_interrupt(struct kvm_s390_irq *irq,
 584                                      struct kvm_s390_interrupt *interrupt)
 585 {
 586     int r = 0;
 587
 588     interrupt->type = irq->type;
 589     switch (irq->type) {
 590     case KVM_S390_INT_VIRTIO:
 591         interrupt->parm = irq->u.ext.ext_params;
 592         /* fall through */
 593     case KVM_S390_INT_PFAULT_INIT:
 594     case KVM_S390_INT_PFAULT_DONE:
 595         interrupt->parm64 = irq->u.ext.ext_params2;
 596         break;
 597     case KVM_S390_PROGRAM_INT:
 598         interrupt->parm = irq->u.pgm.code;
 599         break;
 600     case KVM_S390_SIGP_SET_PREFIX:
 601         interrupt->parm = irq->u.prefix.address;
 602         break;
 603     case KVM_S390_INT_SERVICE:
 604         interrupt->parm = irq->u.ext.ext_params;
 605         break;
 606     case KVM_S390_MCHK:
 607         interrupt->parm = irq->u.mchk.cr14;
 608         interrupt->parm64 = irq->u.mchk.mcic;
 609         break;
 610     case KVM_S390_INT_EXTERNAL_CALL:
 611         interrupt->parm = irq->u.extcall.code;
 612         break;
 613     case KVM_S390_INT_EMERGENCY:
 614         interrupt->parm = irq->u.emerg.code;
 615         break;
 616     case KVM_S390_SIGP_STOP:
 617     case KVM_S390_RESTART:
 618         break; /* These types have no parameters */
 619     case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
 620         interrupt->parm = irq->u.io.subchannel_id << 16;
 621         interrupt->parm |= irq->u.io.subchannel_nr;
 622         interrupt->parm64 = (uint64_t)irq->u.io.io_int_parm << 32;
 623         interrupt->parm64 |= irq->u.io.io_int_word;
 624         break;
 625     default:
 626         r = -EINVAL;
 627         break;
 628     }
 629     return r;
 630 }
 631
 632 void kvm_s390_vcpu_interrupt(S390CPU *cpu, struct kvm_s390_irq *irq)
 633 {
 634     struct kvm_s390_interrupt kvmint = {};
 635     CPUState *cs = CPU(cpu);
 636     int r;
 637
 638     r = s390_kvm_irq_to_interrupt(irq, &kvmint);
 639     if (r < 0) {
 640         fprintf(stderr, "%s called with bogus interrupt\n", __func__);
 641         exit(1);
 642     }
 643
 644     r = kvm_vcpu_ioctl(cs, KVM_S390_INTERRUPT, &kvmint);
 645     if (r < 0) {
 646         fprintf(stderr, "KVM failed to inject interrupt\n");
 647         exit(1);
 648     }
 649 }
 650
 651 static void __kvm_s390_floating_interrupt(struct kvm_s390_irq *irq)
 652 {
 653     struct kvm_s390_interrupt kvmint = {};
 654     int r;
 655
 656     r = s390_kvm_irq_to_interrupt(irq, &kvmint);
 657     if (r < 0) {
 658         fprintf(stderr, "%s called with bogus interrupt\n", __func__);
 659         exit(1);
 660     }
 661
 662     r = kvm_vm_ioctl(kvm_state, KVM_S390_INTERRUPT, &kvmint);
 663     if (r < 0) {
 664         fprintf(stderr, "KVM failed to inject interrupt\n");
 665         exit(1);
 666     }
 667 }
 668
 669 void kvm_s390_floating_interrupt(struct kvm_s390_irq *irq)
 670 {
 671     static bool use_flic = true;
 672     int r;
 673
 674     if (use_flic) {
 675         r = kvm_s390_inject_flic(irq);
 676         if (r == -ENOSYS) {
 677             use_flic = false;
 678         }
 679         if (!r) {
 680             return;
 681         }
 682     }
 683     __kvm_s390_floating_interrupt(irq);
 684 }
 685
 686 void kvm_s390_virtio_irq(int config_change, uint64_t token)
 687 {
 688     struct kvm_s390_irq irq = {
 689         .type = KVM_S390_INT_VIRTIO,
 690         .u.ext.ext_params = config_change,
 691         .u.ext.ext_params2 = token,
 692     };
 693
 694     kvm_s390_floating_interrupt(&irq);
 695 }
 696
 697 void kvm_s390_service_interrupt(uint32_t parm)
 698 {
 699     struct kvm_s390_irq irq = {
 700         .type = KVM_S390_INT_SERVICE,
 701         .u.ext.ext_params = parm,
 702     };
 703
 704     kvm_s390_floating_interrupt(&irq);
 705 }
 706
 707 static void enter_pgmcheck(S390CPU *cpu, uint16_t code)
 708 {
 709     struct kvm_s390_irq irq = {
 710         .type = KVM_S390_PROGRAM_INT,
 711         .u.pgm.code = code,
 712     };
 713
 714     kvm_s390_vcpu_interrupt(cpu, &irq);
 715 }
 716
 717 static int kvm_sclp_service_call(S390CPU *cpu, struct kvm_run *run,
 718                                  uint16_t ipbh0)
 719 {
 720     CPUS390XState *env = &cpu->env;
 721     uint64_t sccb;
 722     uint32_t code;
 723     int r = 0;
 724
 725     cpu_synchronize_state(CPU(cpu));
 726     sccb = env->regs[ipbh0 & 0xf];
 727     code = env->regs[(ipbh0 & 0xf0) >> 4];
 728
 729     r = sclp_service_call(env, sccb, code);
 730     if (r < 0) {
 731         enter_pgmcheck(cpu, -r);
 732     } else {
 733         setcc(cpu, r);
 734     }
 735
 736     return 0;
 737 }
 738
 739 static int handle_b2(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
 740 {
 741     CPUS390XState *env = &cpu->env;
 742     int rc = 0;
 743     uint16_t ipbh0 = (run->s390_sieic.ipb & 0xffff0000) >> 16;
 744
 745     cpu_synchronize_state(CPU(cpu));
 746
 747     switch (ipa1) {
 748     case PRIV_B2_XSCH:
 749         ioinst_handle_xsch(cpu, env->regs[1]);
 750         break;
 751     case PRIV_B2_CSCH:
 752         ioinst_handle_csch(cpu, env->regs[1]);
 753         break;
 754     case PRIV_B2_HSCH:
 755         ioinst_handle_hsch(cpu, env->regs[1]);
 756         break;
 757     case PRIV_B2_MSCH:
 758         ioinst_handle_msch(cpu, env->regs[1], run->s390_sieic.ipb);
 759         break;
 760     case PRIV_B2_SSCH:
 761         ioinst_handle_ssch(cpu, env->regs[1], run->s390_sieic.ipb);
 762         break;
 763     case PRIV_B2_STCRW:
 764         ioinst_handle_stcrw(cpu, run->s390_sieic.ipb);
 765         break;
 766     case PRIV_B2_STSCH:
 767         ioinst_handle_stsch(cpu, env->regs[1], run->s390_sieic.ipb);
 768         break;
 769     case PRIV_B2_TSCH:
 770         /* We should only get tsch via KVM_EXIT_S390_TSCH. */
 771         fprintf(stderr, "Spurious tsch intercept\n");
 772         break;
 773     case PRIV_B2_CHSC:
 774         ioinst_handle_chsc(cpu, run->s390_sieic.ipb);
 775         break;
 776     case PRIV_B2_TPI:
 777         /* This should have been handled by kvm already. */
 778         fprintf(stderr, "Spurious tpi intercept\n");
 779         break;
 780     case PRIV_B2_SCHM:
 781         ioinst_handle_schm(cpu, env->regs[1], env->regs[2],
 782                            run->s390_sieic.ipb);
 783         break;
 784     case PRIV_B2_RSCH:
 785         ioinst_handle_rsch(cpu, env->regs[1]);
 786         break;
 787     case PRIV_B2_RCHP:
 788         ioinst_handle_rchp(cpu, env->regs[1]);
 789         break;
 790     case PRIV_B2_STCPS:
 791         /* We do not provide this instruction, it is suppressed. */
 792         break;
 793     case PRIV_B2_SAL:
 794         ioinst_handle_sal(cpu, env->regs[1]);
 795         break;
 796     case PRIV_B2_SIGA:
 797         /* Not provided, set CC = 3 for subchannel not operational */
 798         setcc(cpu, 3);
 799         break;
 800     case PRIV_B2_SCLP_CALL:
 801         rc = kvm_sclp_service_call(cpu, run, ipbh0);
 802         break;
 803     default:
 804         rc = -1;
 805         DPRINTF("KVM: unhandled PRIV: 0xb2%x\n", ipa1);
 806         break;
 807     }
 808
 809     return rc;
 810 }
 811
 812 static int handle_b9(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
 813 {
 814     int r = 0;
 815
 816     switch (ipa1) {
 817     case PRIV_B9_EQBS:
 818         /* just inject exception */
 819         r = -1;
 820         break;
 821     default:
 822         r = -1;
 823         DPRINTF("KVM: unhandled PRIV: 0xb9%x\n", ipa1);
 824         break;
 825     }
 826
 827     return r;
 828 }
 829
 830 static int handle_eb(S390CPU *cpu, struct kvm_run *run, uint8_t ipbl)
 831 {
 832     int r = 0;
 833
 834     switch (ipbl) {
 835     case PRIV_EB_SQBS:
 836         /* just inject exception */
 837         r = -1;
 838         break;
 839     default:
 840         r = -1;
 841         DPRINTF("KVM: unhandled PRIV: 0xeb%x\n", ipbl);
 842         break;
 843     }
 844
 845     return r;
 846 }
 847
 848 static int handle_hypercall(S390CPU *cpu, struct kvm_run *run)
 849 {
 850     CPUS390XState *env = &cpu->env;
 851     int ret;
 852
 853     cpu_synchronize_state(CPU(cpu));
 854     ret = s390_virtio_hypercall(env);
 855     if (ret == -EINVAL) {
 856         enter_pgmcheck(cpu, PGM_SPECIFICATION);
 857         return 0;
 858     }
 859
 860     return ret;
 861 }
 862
 863 static void kvm_handle_diag_308(S390CPU *cpu, struct kvm_run *run)
 864 {
 865     uint64_t r1, r3;
 866
 867     cpu_synchronize_state(CPU(cpu));
 868     r1 = (run->s390_sieic.ipa & 0x00f0) >> 8;
 869     r3 = run->s390_sieic.ipa & 0x000f;
 870     handle_diag_308(&cpu->env, r1, r3);
 871 }
 872
 873 static int handle_sw_breakpoint(S390CPU *cpu, struct kvm_run *run)
 874 {
 875     CPUS390XState *env = &cpu->env;
 876     unsigned long pc;
 877
 878     cpu_synchronize_state(CPU(cpu));
 879
 880     pc = env->psw.addr - 4;
 881     if (kvm_find_sw_breakpoint(CPU(cpu), pc)) {
 882         env->psw.addr = pc;
 883         return EXCP_DEBUG;
 884     }
 885
 886     return -ENOENT;
 887 }
 888
 889 #define DIAG_KVM_CODE_MASK 0x000000000000ffff
 890
 891 static int handle_diag(S390CPU *cpu, struct kvm_run *run, uint32_t ipb)
 892 {
 893     int r = 0;
 894     uint16_t func_code;
 895
 896     /*
 897      * For any diagnose call we support, bits 48-63 of the resulting
 898      * address specify the function code; the remainder is ignored.
 899      */
 900     func_code = decode_basedisp_rs(&cpu->env, ipb) & DIAG_KVM_CODE_MASK;
 901     switch (func_code) {
 902     case DIAG_IPL:
 903         kvm_handle_diag_308(cpu, run);
 904         break;
 905     case DIAG_KVM_HYPERCALL:
 906         r = handle_hypercall(cpu, run);
 907         break;
 908     case DIAG_KVM_BREAKPOINT:
 909         r = handle_sw_breakpoint(cpu, run);
 910         break;
 911     default:
 912         DPRINTF("KVM: unknown DIAG: 0x%x\n", func_code);
 913         r = -1;
 914         break;
 915     }
 916
 917     return r;
 918 }
 919
 920 static void sigp_cpu_start(void *arg)
 921 {
 922     CPUState *cs = arg;
 923     S390CPU *cpu = S390_CPU(cs);
 924
 925     s390_cpu_set_state(CPU_STATE_OPERATING, cpu);
 926     DPRINTF("DONE: KVM cpu start: %p\n", &cpu->env);
 927 }
 928
 929 static void sigp_cpu_restart(void *arg)
 930 {
 931     CPUState *cs = arg;
 932     S390CPU *cpu = S390_CPU(cs);
 933     struct kvm_s390_irq irq = {
 934         .type = KVM_S390_RESTART,
 935     };
 936
 937     kvm_s390_vcpu_interrupt(cpu, &irq);
 938     s390_cpu_set_state(CPU_STATE_OPERATING, cpu);
 939 }
 940
 941 int kvm_s390_cpu_restart(S390CPU *cpu)
 942 {
 943     run_on_cpu(CPU(cpu), sigp_cpu_restart, CPU(cpu));
 944     DPRINTF("DONE: KVM cpu restart: %p\n", &cpu->env);
 945     return 0;
 946 }
 947
 948 static void sigp_initial_cpu_reset(void *arg)
 949 {
 950     CPUState *cpu = arg;
 951     S390CPUClass *scc = S390_CPU_GET_CLASS(cpu);
 952
 953     cpu_synchronize_state(cpu);
 954     scc->initial_cpu_reset(cpu);
 955     cpu_synchronize_post_reset(cpu);
 956 }
 957
 958 static void sigp_cpu_reset(void *arg)
 959 {
 960     CPUState *cpu = arg;
 961     S390CPUClass *scc = S390_CPU_GET_CLASS(cpu);
 962
 963     cpu_synchronize_state(cpu);
 964     scc->cpu_reset(cpu);
 965     cpu_synchronize_post_reset(cpu);
 966 }
 967
 968 #define SIGP_ORDER_MASK 0x000000ff
 969
 970 static int handle_sigp(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
 971 {
 972     CPUS390XState *env = &cpu->env;
 973     uint8_t order_code;
 974     uint16_t cpu_addr;
 975     S390CPU *target_cpu;
 976     uint64_t *statusreg = &env->regs[ipa1 >> 4];
 977     int cc;
 978
 979     cpu_synchronize_state(CPU(cpu));
 980
 981     /* get order code */
 982     order_code = decode_basedisp_rs(env, run->s390_sieic.ipb) & SIGP_ORDER_MASK;
 983
 984     cpu_addr = env->regs[ipa1 & 0x0f];
 985     target_cpu = s390_cpu_addr2state(cpu_addr);
 986     if (target_cpu == NULL) {
 987         cc = 3;    /* not operational */
 988         goto out;
 989     }
 990
 991     switch (order_code) {
 992     case SIGP_START:
 993         run_on_cpu(CPU(target_cpu), sigp_cpu_start, CPU(target_cpu));
 994         cc = 0;
 995         break;
 996     case SIGP_RESTART:
 997         run_on_cpu(CPU(target_cpu), sigp_cpu_restart, CPU(target_cpu));
 998         cc = 0;
 999         break;
1000     case SIGP_SET_ARCH:
1001         *statusreg &= 0xffffffff00000000UL;
1002         *statusreg |= SIGP_STAT_INVALID_PARAMETER;
1003         cc = 1;   /* status stored */
1004         break;
1005     case SIGP_INITIAL_CPU_RESET:
1006         run_on_cpu(CPU(target_cpu), sigp_initial_cpu_reset, CPU(target_cpu));
1007         cc = 0;
1008         break;
1009     case SIGP_CPU_RESET:
1010         run_on_cpu(CPU(target_cpu), sigp_cpu_reset, CPU(target_cpu));
1011         cc = 0;
1012         break;
1013     default:
1014         DPRINTF("KVM: unknown SIGP: 0x%x\n", order_code);
1015         *statusreg &= 0xffffffff00000000UL;
1016         *statusreg |= SIGP_STAT_INVALID_ORDER;
1017         cc = 1;   /* status stored */
1018         break;
1019     }
1020
1021 out:
1022     setcc(cpu, cc);
1023     return 0;
1024 }
1025
1026 static int handle_instruction(S390CPU *cpu, struct kvm_run *run)
1027 {
1028     unsigned int ipa0 = (run->s390_sieic.ipa & 0xff00);
1029     uint8_t ipa1 = run->s390_sieic.ipa & 0x00ff;
1030     int r = -1;
1031
1032     DPRINTF("handle_instruction 0x%x 0x%x\n",
1033             run->s390_sieic.ipa, run->s390_sieic.ipb);
1034     switch (ipa0) {
1035     case IPA0_B2:
1036         r = handle_b2(cpu, run, ipa1);
1037         break;
1038     case IPA0_B9:
1039         r = handle_b9(cpu, run, ipa1);
1040         break;
1041     case IPA0_EB:
1042         r = handle_eb(cpu, run, run->s390_sieic.ipb & 0xff);
1043         break;
1044     case IPA0_DIAG:
1045         r = handle_diag(cpu, run, run->s390_sieic.ipb);
1046         break;
1047     case IPA0_SIGP:
1048         r = handle_sigp(cpu, run, ipa1);
1049         break;
1050     }
1051
1052     if (r < 0) {
1053         r = 0;
1054         enter_pgmcheck(cpu, 0x0001);
1055     }
1056
1057     return r;
1058 }
1059
1060 static bool is_special_wait_psw(CPUState *cs)
1061 {
1062     /* signal quiesce */
1063     return cs->kvm_run->psw_addr == 0xfffUL;
1064 }
1065
1066 static void guest_panicked(void)
1067 {
1068     qapi_event_send_guest_panicked(GUEST_PANIC_ACTION_PAUSE,
1069                                    &error_abort);
1070     vm_stop(RUN_STATE_GUEST_PANICKED);
1071 }
1072
1073 static void unmanageable_intercept(S390CPU *cpu, const char *str, int pswoffset)
1074 {
1075     CPUState *cs = CPU(cpu);
1076
1077     error_report("Unmanageable %s! CPU%i new PSW: 0x%016lx:%016lx",
1078                  str, cs->cpu_index, ldq_phys(cs->as, cpu->env.psa + pswoffset),
1079                  ldq_phys(cs->as, cpu->env.psa + pswoffset + 8));
1080     s390_cpu_halt(cpu);
1081     guest_panicked();
1082 }
1083
1084 static int handle_intercept(S390CPU *cpu)
1085 {
1086     CPUState *cs = CPU(cpu);
1087     struct kvm_run *run = cs->kvm_run;
1088     int icpt_code = run->s390_sieic.icptcode;
1089     int r = 0;
1090
1091     DPRINTF("intercept: 0x%x (at 0x%lx)\n", icpt_code,
1092             (long)cs->kvm_run->psw_addr);
1093     switch (icpt_code) {
1094         case ICPT_INSTRUCTION:
1095             r = handle_instruction(cpu, run);
1096             break;
1097         case ICPT_PROGRAM:
1098             unmanageable_intercept(cpu, "program interrupt",
1099                                    offsetof(LowCore, program_new_psw));
1100             r = EXCP_HALTED;
1101             break;
1102         case ICPT_EXT_INT:
1103             unmanageable_intercept(cpu, "external interrupt",
1104                                    offsetof(LowCore, external_new_psw));
1105             r = EXCP_HALTED;
1106             break;
1107         case ICPT_WAITPSW:
1108             /* disabled wait, since enabled wait is handled in kernel */
1109             cpu_synchronize_state(cs);
1110             if (s390_cpu_halt(cpu) == 0) {
1111                 if (is_special_wait_psw(cs)) {
1112                     qemu_system_shutdown_request();
1113                 } else {
1114                     guest_panicked();
1115                 }
1116             }
1117             r = EXCP_HALTED;
1118             break;
1119         case ICPT_CPU_STOP:
1120             if (s390_cpu_set_state(CPU_STATE_STOPPED, cpu) == 0) {
1121                 qemu_system_shutdown_request();
1122             }
1123             r = EXCP_HALTED;
1124             break;
1125         case ICPT_SOFT_INTERCEPT:
1126             fprintf(stderr, "KVM unimplemented icpt SOFT\n");
1127             exit(1);
1128             break;
1129         case ICPT_IO:
1130             fprintf(stderr, "KVM unimplemented icpt IO\n");
1131             exit(1);
1132             break;
1133         default:
1134             fprintf(stderr, "Unknown intercept code: %d\n", icpt_code);
1135             exit(1);
1136             break;
1137     }
1138
1139     return r;
1140 }
1141
1142 static int handle_tsch(S390CPU *cpu)
1143 {
1144     CPUS390XState *env = &cpu->env;
1145     CPUState *cs = CPU(cpu);
1146     struct kvm_run *run = cs->kvm_run;
1147     int ret;
1148
1149     cpu_synchronize_state(cs);
1150
1151     ret = ioinst_handle_tsch(env, env->regs[1], run->s390_tsch.ipb);
1152     if (ret >= 0) {
1153         /* Success; set condition code. */
1154         setcc(cpu, ret);
1155         ret = 0;
1156     } else if (ret < -1) {
1157         /*
1158          * Failure.
1159          * If an I/O interrupt had been dequeued, we have to reinject it.
1160          */
1161         if (run->s390_tsch.dequeued) {
1162             kvm_s390_io_interrupt(run->s390_tsch.subchannel_id,
1163                                   run->s390_tsch.subchannel_nr,
1164                                   run->s390_tsch.io_int_parm,
1165                                   run->s390_tsch.io_int_word);
1166         }
1167         ret = 0;
1168     }
1169     return ret;
1170 }
1171
1172 static int kvm_arch_handle_debug_exit(S390CPU *cpu)
1173 {
1174     CPUState *cs = CPU(cpu);
1175     struct kvm_run *run = cs->kvm_run;
1176
1177     int ret = 0;
1178     struct kvm_debug_exit_arch *arch_info = &run->debug.arch;
1179
1180     switch (arch_info->type) {
1181     case KVM_HW_WP_WRITE:
1182         if (find_hw_breakpoint(arch_info->addr, -1, arch_info->type)) {
1183             cs->watchpoint_hit = &hw_watchpoint;
1184             hw_watchpoint.vaddr = arch_info->addr;
1185             hw_watchpoint.flags = BP_MEM_WRITE;
1186             ret = EXCP_DEBUG;
1187         }
1188         break;
1189     case KVM_HW_BP:
1190         if (find_hw_breakpoint(arch_info->addr, -1, arch_info->type)) {
1191             ret = EXCP_DEBUG;
1192         }
1193         break;
1194     case KVM_SINGLESTEP:
1195         if (cs->singlestep_enabled) {
1196             ret = EXCP_DEBUG;
1197         }
1198         break;
1199     default:
1200         ret = -ENOSYS;
1201     }
1202
1203     return ret;
1204 }
1205
1206 int kvm_arch_handle_exit(CPUState *cs, struct kvm_run *run)
1207 {
1208     S390CPU *cpu = S390_CPU(cs);
1209     int ret = 0;
1210
1211     switch (run->exit_reason) {
1212         case KVM_EXIT_S390_SIEIC:
1213             ret = handle_intercept(cpu);
1214             break;
1215         case KVM_EXIT_S390_RESET:
1216             qemu_system_reset_request();
1217             break;
1218         case KVM_EXIT_S390_TSCH:
1219             ret = handle_tsch(cpu);
1220             break;
1221         case KVM_EXIT_DEBUG:
1222             ret = kvm_arch_handle_debug_exit(cpu);
1223             break;
1224         default:
1225             fprintf(stderr, "Unknown KVM exit: %d\n", run->exit_reason);
1226             break;
1227     }
1228
1229     if (ret == 0) {
1230         ret = EXCP_INTERRUPT;
1231     }
1232     return ret;
1233 }
1234
1235 bool kvm_arch_stop_on_emulation_error(CPUState *cpu)
1236 {
1237     return true;
1238 }
1239
1240 int kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr)
1241 {
1242     return 1;
1243 }
1244
1245 int kvm_arch_on_sigbus(int code, void *addr)
1246 {
1247     return 1;
1248 }
1249
1250 void kvm_s390_io_interrupt(uint16_t subchannel_id,
1251                            uint16_t subchannel_nr, uint32_t io_int_parm,
1252                            uint32_t io_int_word)
1253 {
1254     struct kvm_s390_irq irq = {
1255         .u.io.subchannel_id = subchannel_id,
1256         .u.io.subchannel_nr = subchannel_nr,
1257         .u.io.io_int_parm = io_int_parm,
1258         .u.io.io_int_word = io_int_word,
1259     };
1260
1261     if (io_int_word & IO_INT_WORD_AI) {
1262         irq.type = KVM_S390_INT_IO(1, 0, 0, 0);
1263     } else {
1264         irq.type = ((subchannel_id & 0xff00) << 24) |
1265             ((subchannel_id & 0x00060) << 22) | (subchannel_nr << 16);
1266     }
1267     kvm_s390_floating_interrupt(&irq);
1268 }
1269
1270 void kvm_s390_crw_mchk(void)
1271 {
1272     struct kvm_s390_irq irq = {
1273         .type = KVM_S390_MCHK,
1274         .u.mchk.cr14 = 1 << 28,
1275         .u.mchk.mcic = 0x00400f1d40330000ULL,
1276     };
1277     kvm_s390_floating_interrupt(&irq);
1278 }
1279
1280 void kvm_s390_enable_css_support(S390CPU *cpu)
1281 {
1282     int r;
1283
1284     /* Activate host kernel channel subsystem support. */
1285     r = kvm_vcpu_enable_cap(CPU(cpu), KVM_CAP_S390_CSS_SUPPORT, 0);
1286     assert(r == 0);
1287 }
1288
1289 void kvm_arch_init_irq_routing(KVMState *s)
1290 {
1291     /*
1292      * Note that while irqchip capabilities generally imply that cpustates
1293      * are handled in-kernel, it is not true for s390 (yet); therefore, we
1294      * have to override the common code kvm_halt_in_kernel_allowed setting.
1295      */
1296     if (kvm_check_extension(s, KVM_CAP_IRQ_ROUTING)) {
1297         kvm_irqfds_allowed = true;
1298         kvm_gsi_routing_allowed = true;
1299         kvm_halt_in_kernel_allowed = false;
1300     }
1301 }
1302
1303 int kvm_s390_assign_subch_ioeventfd(EventNotifier *notifier, uint32_t sch,
1304                                     int vq, bool assign)
1305 {
1306     struct kvm_ioeventfd kick = {
1307         .flags = KVM_IOEVENTFD_FLAG_VIRTIO_CCW_NOTIFY |
1308         KVM_IOEVENTFD_FLAG_DATAMATCH,
1309         .fd = event_notifier_get_fd(notifier),
1310         .datamatch = vq,
1311         .addr = sch,
1312         .len = 8,
1313     };
1314     if (!kvm_check_extension(kvm_state, KVM_CAP_IOEVENTFD)) {
1315         return -ENOSYS;
1316     }
1317     if (!assign) {
1318         kick.flags |= KVM_IOEVENTFD_FLAG_DEASSIGN;
1319     }
1320     return kvm_vm_ioctl(kvm_state, KVM_IOEVENTFD, &kick);
1321 }
1322
1323 int kvm_s390_get_memslot_count(KVMState *s)
1324 {
1325     return kvm_check_extension(s, KVM_CAP_NR_MEMSLOTS);
1326 }
1327
1328 int kvm_s390_set_cpu_state(S390CPU *cpu, uint8_t cpu_state)
1329 {
1330     struct kvm_mp_state mp_state = {};
1331     int ret;
1332
1333     /* the kvm part might not have been initialized yet */
1334     if (CPU(cpu)->kvm_state == NULL) {
1335         return 0;
1336     }
1337
1338     switch (cpu_state) {
1339     case CPU_STATE_STOPPED:
1340         mp_state.mp_state = KVM_MP_STATE_STOPPED;
1341         break;
1342     case CPU_STATE_CHECK_STOP:
1343         mp_state.mp_state = KVM_MP_STATE_CHECK_STOP;
1344         break;
1345     case CPU_STATE_OPERATING:
1346         mp_state.mp_state = KVM_MP_STATE_OPERATING;
1347         break;
1348     case CPU_STATE_LOAD:
1349         mp_state.mp_state = KVM_MP_STATE_LOAD;
1350         break;
1351     default:
1352         error_report("Requested CPU state is not a valid S390 CPU state: %u",
1353                      cpu_state);
1354         exit(1);
1355     }
1356
1357     ret = kvm_vcpu_ioctl(CPU(cpu), KVM_SET_MP_STATE, &mp_state);
1358     if (ret) {
1359         trace_kvm_failed_cpu_state_set(CPU(cpu)->cpu_index, cpu_state,
1360                                        strerror(-ret));
1361     }
1362
1363     return ret;
1364 }