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 #include "hw/s390x/s390-pci-inst.h"
  44 #include "hw/s390x/s390-pci-bus.h"
  45 #include "hw/s390x/ipl.h"
  46
  47 /* #define DEBUG_KVM */
  48
  49 #ifdef DEBUG_KVM
  50 #define DPRINTF(fmt, ...) \
  51     do { fprintf(stderr, fmt, ## __VA_ARGS__); } while (0)
  52 #else
  53 #define DPRINTF(fmt, ...) \
  54     do { } while (0)
  55 #endif
  56
  57 #define IPA0_DIAG                       0x8300
  58 #define IPA0_SIGP                       0xae00
  59 #define IPA0_B2                         0xb200
  60 #define IPA0_B9                         0xb900
  61 #define IPA0_EB                         0xeb00
  62 #define IPA0_E3                         0xe300
  63
  64 #define PRIV_B2_SCLP_CALL               0x20
  65 #define PRIV_B2_CSCH                    0x30
  66 #define PRIV_B2_HSCH                    0x31
  67 #define PRIV_B2_MSCH                    0x32
  68 #define PRIV_B2_SSCH                    0x33
  69 #define PRIV_B2_STSCH                   0x34
  70 #define PRIV_B2_TSCH                    0x35
  71 #define PRIV_B2_TPI                     0x36
  72 #define PRIV_B2_SAL                     0x37
  73 #define PRIV_B2_RSCH                    0x38
  74 #define PRIV_B2_STCRW                   0x39
  75 #define PRIV_B2_STCPS                   0x3a
  76 #define PRIV_B2_RCHP                    0x3b
  77 #define PRIV_B2_SCHM                    0x3c
  78 #define PRIV_B2_CHSC                    0x5f
  79 #define PRIV_B2_SIGA                    0x74
  80 #define PRIV_B2_XSCH                    0x76
  81
  82 #define PRIV_EB_SQBS                    0x8a
  83 #define PRIV_EB_PCISTB                  0xd0
  84 #define PRIV_EB_SIC                     0xd1
  85
  86 #define PRIV_B9_EQBS                    0x9c
  87 #define PRIV_B9_CLP                     0xa0
  88 #define PRIV_B9_PCISTG                  0xd0
  89 #define PRIV_B9_PCILG                   0xd2
  90 #define PRIV_B9_RPCIT                   0xd3
  91
  92 #define PRIV_E3_MPCIFC                  0xd0
  93 #define PRIV_E3_STPCIFC                 0xd4
  94
  95 #define DIAG_IPL                        0x308
  96 #define DIAG_KVM_HYPERCALL              0x500
  97 #define DIAG_KVM_BREAKPOINT             0x501
  98
  99 #define ICPT_INSTRUCTION                0x04
 100 #define ICPT_PROGRAM                    0x08
 101 #define ICPT_EXT_INT                    0x14
 102 #define ICPT_WAITPSW                    0x1c
 103 #define ICPT_SOFT_INTERCEPT             0x24
 104 #define ICPT_CPU_STOP                   0x28
 105 #define ICPT_IO                         0x40
 106
 107 static CPUWatchpoint hw_watchpoint;
 108 /*
 109  * We don't use a list because this structure is also used to transmit the
 110  * hardware breakpoints to the kernel.
 111  */
 112 static struct kvm_hw_breakpoint *hw_breakpoints;
 113 static int nb_hw_breakpoints;
 114
 115 const KVMCapabilityInfo kvm_arch_required_capabilities[] = {
 116     KVM_CAP_LAST_INFO
 117 };
 118
 119 static int cap_sync_regs;
 120 static int cap_async_pf;
 121
 122 static void *legacy_s390_alloc(size_t size, uint64_t *align);
 123
 124 static int kvm_s390_check_clear_cmma(KVMState *s)
 125 {
 126     struct kvm_device_attr attr = {
 127         .group = KVM_S390_VM_MEM_CTRL,
 128         .attr = KVM_S390_VM_MEM_CLR_CMMA,
 129     };
 130
 131     return kvm_vm_ioctl(s, KVM_HAS_DEVICE_ATTR, &attr);
 132 }
 133
 134 static int kvm_s390_check_enable_cmma(KVMState *s)
 135 {
 136     struct kvm_device_attr attr = {
 137         .group = KVM_S390_VM_MEM_CTRL,
 138         .attr = KVM_S390_VM_MEM_ENABLE_CMMA,
 139     };
 140
 141     return kvm_vm_ioctl(s, KVM_HAS_DEVICE_ATTR, &attr);
 142 }
 143
 144 void kvm_s390_clear_cmma_callback(void *opaque)
 145 {
 146     int rc;
 147     KVMState *s = opaque;
 148     struct kvm_device_attr attr = {
 149         .group = KVM_S390_VM_MEM_CTRL,
 150         .attr = KVM_S390_VM_MEM_CLR_CMMA,
 151     };
 152
 153     rc = kvm_vm_ioctl(s, KVM_SET_DEVICE_ATTR, &attr);
 154     trace_kvm_clear_cmma(rc);
 155 }
 156
 157 static void kvm_s390_enable_cmma(KVMState *s)
 158 {
 159     int rc;
 160     struct kvm_device_attr attr = {
 161         .group = KVM_S390_VM_MEM_CTRL,
 162         .attr = KVM_S390_VM_MEM_ENABLE_CMMA,
 163     };
 164
 165     if (kvm_s390_check_enable_cmma(s) || kvm_s390_check_clear_cmma(s)) {
 166         return;
 167     }
 168
 169     rc = kvm_vm_ioctl(s, KVM_SET_DEVICE_ATTR, &attr);
 170     if (!rc) {
 171         qemu_register_reset(kvm_s390_clear_cmma_callback, s);
 172     }
 173     trace_kvm_enable_cmma(rc);
 174 }
 175
 176 int kvm_arch_init(KVMState *s)
 177 {
 178     cap_sync_regs = kvm_check_extension(s, KVM_CAP_SYNC_REGS);
 179     cap_async_pf = kvm_check_extension(s, KVM_CAP_ASYNC_PF);
 180
 181     if (kvm_check_extension(s, KVM_CAP_VM_ATTRIBUTES)) {
 182         kvm_s390_enable_cmma(s);
 183     }
 184
 185     if (!kvm_check_extension(s, KVM_CAP_S390_GMAP)
 186         || !kvm_check_extension(s, KVM_CAP_S390_COW)) {
 187         phys_mem_set_alloc(legacy_s390_alloc);
 188     }
 189     return 0;
 190 }
 191
 192 unsigned long kvm_arch_vcpu_id(CPUState *cpu)
 193 {
 194     return cpu->cpu_index;
 195 }
 196
 197 int kvm_arch_init_vcpu(CPUState *cs)
 198 {
 199     S390CPU *cpu = S390_CPU(cs);
 200     kvm_s390_set_cpu_state(cpu, cpu->env.cpu_state);
 201     return 0;
 202 }
 203
 204 void kvm_s390_reset_vcpu(S390CPU *cpu)
 205 {
 206     CPUState *cs = CPU(cpu);
 207
 208     /* The initial reset call is needed here to reset in-kernel
 209      * vcpu data that we can't access directly from QEMU
 210      * (i.e. with older kernels which don't support sync_regs/ONE_REG).
 211      * Before this ioctl cpu_synchronize_state() is called in common kvm
 212      * code (kvm-all) */
 213     if (kvm_vcpu_ioctl(cs, KVM_S390_INITIAL_RESET, NULL)) {
 214         error_report("Initial CPU reset failed on CPU %i\n", cs->cpu_index);
 215     }
 216 }
 217
 218 static int can_sync_regs(CPUState *cs, int regs)
 219 {
 220     return cap_sync_regs && (cs->kvm_run->kvm_valid_regs & regs) == regs;
 221 }
 222
 223 int kvm_arch_put_registers(CPUState *cs, int level)
 224 {
 225     S390CPU *cpu = S390_CPU(cs);
 226     CPUS390XState *env = &cpu->env;
 227     struct kvm_sregs sregs;
 228     struct kvm_regs regs;
 229     struct kvm_fpu fpu = {};
 230     int r;
 231     int i;
 232
 233     /* always save the PSW  and the GPRS*/
 234     cs->kvm_run->psw_addr = env->psw.addr;
 235     cs->kvm_run->psw_mask = env->psw.mask;
 236
 237     if (can_sync_regs(cs, KVM_SYNC_GPRS)) {
 238         for (i = 0; i < 16; i++) {
 239             cs->kvm_run->s.regs.gprs[i] = env->regs[i];
 240             cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_GPRS;
 241         }
 242     } else {
 243         for (i = 0; i < 16; i++) {
 244             regs.gprs[i] = env->regs[i];
 245         }
 246         r = kvm_vcpu_ioctl(cs, KVM_SET_REGS, &regs);
 247         if (r < 0) {
 248             return r;
 249         }
 250     }
 251
 252     /* Floating point */
 253     for (i = 0; i < 16; i++) {
 254         fpu.fprs[i] = env->fregs[i].ll;
 255     }
 256     fpu.fpc = env->fpc;
 257
 258     r = kvm_vcpu_ioctl(cs, KVM_SET_FPU, &fpu);
 259     if (r < 0) {
 260         return r;
 261     }
 262
 263     /* Do we need to save more than that? */
 264     if (level == KVM_PUT_RUNTIME_STATE) {
 265         return 0;
 266     }
 267
 268     if (can_sync_regs(cs, KVM_SYNC_ARCH0)) {
 269         cs->kvm_run->s.regs.cputm = env->cputm;
 270         cs->kvm_run->s.regs.ckc = env->ckc;
 271         cs->kvm_run->s.regs.todpr = env->todpr;
 272         cs->kvm_run->s.regs.gbea = env->gbea;
 273         cs->kvm_run->s.regs.pp = env->pp;
 274         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_ARCH0;
 275     } else {
 276         /*
 277          * These ONE_REGS are not protected by a capability. As they are only
 278          * necessary for migration we just trace a possible error, but don't
 279          * return with an error return code.
 280          */
 281         kvm_set_one_reg(cs, KVM_REG_S390_CPU_TIMER, &env->cputm);
 282         kvm_set_one_reg(cs, KVM_REG_S390_CLOCK_COMP, &env->ckc);
 283         kvm_set_one_reg(cs, KVM_REG_S390_TODPR, &env->todpr);
 284         kvm_set_one_reg(cs, KVM_REG_S390_GBEA, &env->gbea);
 285         kvm_set_one_reg(cs, KVM_REG_S390_PP, &env->pp);
 286     }
 287
 288     /* pfault parameters */
 289     if (can_sync_regs(cs, KVM_SYNC_PFAULT)) {
 290         cs->kvm_run->s.regs.pft = env->pfault_token;
 291         cs->kvm_run->s.regs.pfs = env->pfault_select;
 292         cs->kvm_run->s.regs.pfc = env->pfault_compare;
 293         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_PFAULT;
 294     } else if (cap_async_pf) {
 295         r = kvm_set_one_reg(cs, KVM_REG_S390_PFTOKEN, &env->pfault_token);
 296         if (r < 0) {
 297             return r;
 298         }
 299         r = kvm_set_one_reg(cs, KVM_REG_S390_PFCOMPARE, &env->pfault_compare);
 300         if (r < 0) {
 301             return r;
 302         }
 303         r = kvm_set_one_reg(cs, KVM_REG_S390_PFSELECT, &env->pfault_select);
 304         if (r < 0) {
 305             return r;
 306         }
 307     }
 308
 309     /* access registers and control registers*/
 310     if (can_sync_regs(cs, KVM_SYNC_ACRS | KVM_SYNC_CRS)) {
 311         for (i = 0; i < 16; i++) {
 312             cs->kvm_run->s.regs.acrs[i] = env->aregs[i];
 313             cs->kvm_run->s.regs.crs[i] = env->cregs[i];
 314         }
 315         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_ACRS;
 316         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_CRS;
 317     } else {
 318         for (i = 0; i < 16; i++) {
 319             sregs.acrs[i] = env->aregs[i];
 320             sregs.crs[i] = env->cregs[i];
 321         }
 322         r = kvm_vcpu_ioctl(cs, KVM_SET_SREGS, &sregs);
 323         if (r < 0) {
 324             return r;
 325         }
 326     }
 327
 328     /* Finally the prefix */
 329     if (can_sync_regs(cs, KVM_SYNC_PREFIX)) {
 330         cs->kvm_run->s.regs.prefix = env->psa;
 331         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_PREFIX;
 332     } else {
 333         /* prefix is only supported via sync regs */
 334     }
 335     return 0;
 336 }
 337
 338 int kvm_arch_get_registers(CPUState *cs)
 339 {
 340     S390CPU *cpu = S390_CPU(cs);
 341     CPUS390XState *env = &cpu->env;
 342     struct kvm_sregs sregs;
 343     struct kvm_regs regs;
 344     struct kvm_fpu fpu;
 345     int i, r;
 346
 347     /* get the PSW */
 348     env->psw.addr = cs->kvm_run->psw_addr;
 349     env->psw.mask = cs->kvm_run->psw_mask;
 350
 351     /* the GPRS */
 352     if (can_sync_regs(cs, KVM_SYNC_GPRS)) {
 353         for (i = 0; i < 16; i++) {
 354             env->regs[i] = cs->kvm_run->s.regs.gprs[i];
 355         }
 356     } else {
 357         r = kvm_vcpu_ioctl(cs, KVM_GET_REGS, &regs);
 358         if (r < 0) {
 359             return r;
 360         }
 361          for (i = 0; i < 16; i++) {
 362             env->regs[i] = regs.gprs[i];
 363         }
 364     }
 365
 366     /* The ACRS and CRS */
 367     if (can_sync_regs(cs, KVM_SYNC_ACRS | KVM_SYNC_CRS)) {
 368         for (i = 0; i < 16; i++) {
 369             env->aregs[i] = cs->kvm_run->s.regs.acrs[i];
 370             env->cregs[i] = cs->kvm_run->s.regs.crs[i];
 371         }
 372     } else {
 373         r = kvm_vcpu_ioctl(cs, KVM_GET_SREGS, &sregs);
 374         if (r < 0) {
 375             return r;
 376         }
 377          for (i = 0; i < 16; i++) {
 378             env->aregs[i] = sregs.acrs[i];
 379             env->cregs[i] = sregs.crs[i];
 380         }
 381     }
 382
 383     /* Floating point */
 384     r = kvm_vcpu_ioctl(cs, KVM_GET_FPU, &fpu);
 385     if (r < 0) {
 386         return r;
 387     }
 388     for (i = 0; i < 16; i++) {
 389         env->fregs[i].ll = fpu.fprs[i];
 390     }
 391     env->fpc = fpu.fpc;
 392
 393     /* The prefix */
 394     if (can_sync_regs(cs, KVM_SYNC_PREFIX)) {
 395         env->psa = cs->kvm_run->s.regs.prefix;
 396     }
 397
 398     if (can_sync_regs(cs, KVM_SYNC_ARCH0)) {
 399         env->cputm = cs->kvm_run->s.regs.cputm;
 400         env->ckc = cs->kvm_run->s.regs.ckc;
 401         env->todpr = cs->kvm_run->s.regs.todpr;
 402         env->gbea = cs->kvm_run->s.regs.gbea;
 403         env->pp = cs->kvm_run->s.regs.pp;
 404     } else {
 405         /*
 406          * These ONE_REGS are not protected by a capability. As they are only
 407          * necessary for migration we just trace a possible error, but don't
 408          * return with an error return code.
 409          */
 410         kvm_get_one_reg(cs, KVM_REG_S390_CPU_TIMER, &env->cputm);
 411         kvm_get_one_reg(cs, KVM_REG_S390_CLOCK_COMP, &env->ckc);
 412         kvm_get_one_reg(cs, KVM_REG_S390_TODPR, &env->todpr);
 413         kvm_get_one_reg(cs, KVM_REG_S390_GBEA, &env->gbea);
 414         kvm_get_one_reg(cs, KVM_REG_S390_PP, &env->pp);
 415     }
 416
 417     /* pfault parameters */
 418     if (can_sync_regs(cs, KVM_SYNC_PFAULT)) {
 419         env->pfault_token = cs->kvm_run->s.regs.pft;
 420         env->pfault_select = cs->kvm_run->s.regs.pfs;
 421         env->pfault_compare = cs->kvm_run->s.regs.pfc;
 422     } else if (cap_async_pf) {
 423         r = kvm_get_one_reg(cs, KVM_REG_S390_PFTOKEN, &env->pfault_token);
 424         if (r < 0) {
 425             return r;
 426         }
 427         r = kvm_get_one_reg(cs, KVM_REG_S390_PFCOMPARE, &env->pfault_compare);
 428         if (r < 0) {
 429             return r;
 430         }
 431         r = kvm_get_one_reg(cs, KVM_REG_S390_PFSELECT, &env->pfault_select);
 432         if (r < 0) {
 433             return r;
 434         }
 435     }
 436
 437     return 0;
 438 }
 439
 440 /*
 441  * Legacy layout for s390:
 442  * Older S390 KVM requires the topmost vma of the RAM to be
 443  * smaller than an system defined value, which is at least 256GB.
 444  * Larger systems have larger values. We put the guest between
 445  * the end of data segment (system break) and this value. We
 446  * use 32GB as a base to have enough room for the system break
 447  * to grow. We also have to use MAP parameters that avoid
 448  * read-only mapping of guest pages.
 449  */
 450 static void *legacy_s390_alloc(size_t size, uint64_t *align)
 451 {
 452     void *mem;
 453
 454     mem = mmap((void *) 0x800000000ULL, size,
 455                PROT_EXEC|PROT_READ|PROT_WRITE,
 456                MAP_SHARED | MAP_ANONYMOUS | MAP_FIXED, -1, 0);
 457     return mem == MAP_FAILED ? NULL : mem;
 458 }
 459
 460 /* DIAG 501 is used for sw breakpoints */
 461 static const uint8_t diag_501[] = {0x83, 0x24, 0x05, 0x01};
 462
 463 int kvm_arch_insert_sw_breakpoint(CPUState *cs, struct kvm_sw_breakpoint *bp)
 464 {
 465
 466     if (cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)&bp->saved_insn,
 467                             sizeof(diag_501), 0) ||
 468         cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)diag_501,
 469                             sizeof(diag_501), 1)) {
 470         return -EINVAL;
 471     }
 472     return 0;
 473 }
 474
 475 int kvm_arch_remove_sw_breakpoint(CPUState *cs, struct kvm_sw_breakpoint *bp)
 476 {
 477     uint8_t t[sizeof(diag_501)];
 478
 479     if (cpu_memory_rw_debug(cs, bp->pc, t, sizeof(diag_501), 0)) {
 480         return -EINVAL;
 481     } else if (memcmp(t, diag_501, sizeof(diag_501))) {
 482         return -EINVAL;
 483     } else if (cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)&bp->saved_insn,
 484                                    sizeof(diag_501), 1)) {
 485         return -EINVAL;
 486     }
 487
 488     return 0;
 489 }
 490
 491 static struct kvm_hw_breakpoint *find_hw_breakpoint(target_ulong addr,
 492                                                     int len, int type)
 493 {
 494     int n;
 495
 496     for (n = 0; n < nb_hw_breakpoints; n++) {
 497         if (hw_breakpoints[n].addr == addr && hw_breakpoints[n].type == type &&
 498             (hw_breakpoints[n].len == len || len == -1)) {
 499             return &hw_breakpoints[n];
 500         }
 501     }
 502
 503     return NULL;
 504 }
 505
 506 static int insert_hw_breakpoint(target_ulong addr, int len, int type)
 507 {
 508     int size;
 509
 510     if (find_hw_breakpoint(addr, len, type)) {
 511         return -EEXIST;
 512     }
 513
 514     size = (nb_hw_breakpoints + 1) * sizeof(struct kvm_hw_breakpoint);
 515
 516     if (!hw_breakpoints) {
 517         nb_hw_breakpoints = 0;
 518         hw_breakpoints = (struct kvm_hw_breakpoint *)g_try_malloc(size);
 519     } else {
 520         hw_breakpoints =
 521             (struct kvm_hw_breakpoint *)g_try_realloc(hw_breakpoints, size);
 522     }
 523
 524     if (!hw_breakpoints) {
 525         nb_hw_breakpoints = 0;
 526         return -ENOMEM;
 527     }
 528
 529     hw_breakpoints[nb_hw_breakpoints].addr = addr;
 530     hw_breakpoints[nb_hw_breakpoints].len = len;
 531     hw_breakpoints[nb_hw_breakpoints].type = type;
 532
 533     nb_hw_breakpoints++;
 534
 535     return 0;
 536 }
 537
 538 int kvm_arch_insert_hw_breakpoint(target_ulong addr,
 539                                   target_ulong len, int type)
 540 {
 541     switch (type) {
 542     case GDB_BREAKPOINT_HW:
 543         type = KVM_HW_BP;
 544         break;
 545     case GDB_WATCHPOINT_WRITE:
 546         if (len < 1) {
 547             return -EINVAL;
 548         }
 549         type = KVM_HW_WP_WRITE;
 550         break;
 551     default:
 552         return -ENOSYS;
 553     }
 554     return insert_hw_breakpoint(addr, len, type);
 555 }
 556
 557 int kvm_arch_remove_hw_breakpoint(target_ulong addr,
 558                                   target_ulong len, int type)
 559 {
 560     int size;
 561     struct kvm_hw_breakpoint *bp = find_hw_breakpoint(addr, len, type);
 562
 563     if (bp == NULL) {
 564         return -ENOENT;
 565     }
 566
 567     nb_hw_breakpoints--;
 568     if (nb_hw_breakpoints > 0) {
 569         /*
 570          * In order to trim the array, move the last element to the position to
 571          * be removed - if necessary.
 572          */
 573         if (bp != &hw_breakpoints[nb_hw_breakpoints]) {
 574             *bp = hw_breakpoints[nb_hw_breakpoints];
 575         }
 576         size = nb_hw_breakpoints * sizeof(struct kvm_hw_breakpoint);
 577         hw_breakpoints =
 578              (struct kvm_hw_breakpoint *)g_realloc(hw_breakpoints, size);
 579     } else {
 580         g_free(hw_breakpoints);
 581         hw_breakpoints = NULL;
 582     }
 583
 584     return 0;
 585 }
 586
 587 void kvm_arch_remove_all_hw_breakpoints(void)
 588 {
 589     nb_hw_breakpoints = 0;
 590     g_free(hw_breakpoints);
 591     hw_breakpoints = NULL;
 592 }
 593
 594 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg)
 595 {
 596     int i;
 597
 598     if (nb_hw_breakpoints > 0) {
 599         dbg->arch.nr_hw_bp = nb_hw_breakpoints;
 600         dbg->arch.hw_bp = hw_breakpoints;
 601
 602         for (i = 0; i < nb_hw_breakpoints; ++i) {
 603             hw_breakpoints[i].phys_addr = s390_cpu_get_phys_addr_debug(cpu,
 604                                                        hw_breakpoints[i].addr);
 605         }
 606         dbg->control |= KVM_GUESTDBG_ENABLE | KVM_GUESTDBG_USE_HW_BP;
 607     } else {
 608         dbg->arch.nr_hw_bp = 0;
 609         dbg->arch.hw_bp = NULL;
 610     }
 611 }
 612
 613 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run)
 614 {
 615 }
 616
 617 void kvm_arch_post_run(CPUState *cpu, struct kvm_run *run)
 618 {
 619 }
 620
 621 int kvm_arch_process_async_events(CPUState *cs)
 622 {
 623     return cs->halted;
 624 }
 625
 626 static int s390_kvm_irq_to_interrupt(struct kvm_s390_irq *irq,
 627                                      struct kvm_s390_interrupt *interrupt)
 628 {
 629     int r = 0;
 630
 631     interrupt->type = irq->type;
 632     switch (irq->type) {
 633     case KVM_S390_INT_VIRTIO:
 634         interrupt->parm = irq->u.ext.ext_params;
 635         /* fall through */
 636     case KVM_S390_INT_PFAULT_INIT:
 637     case KVM_S390_INT_PFAULT_DONE:
 638         interrupt->parm64 = irq->u.ext.ext_params2;
 639         break;
 640     case KVM_S390_PROGRAM_INT:
 641         interrupt->parm = irq->u.pgm.code;
 642         break;
 643     case KVM_S390_SIGP_SET_PREFIX:
 644         interrupt->parm = irq->u.prefix.address;
 645         break;
 646     case KVM_S390_INT_SERVICE:
 647         interrupt->parm = irq->u.ext.ext_params;
 648         break;
 649     case KVM_S390_MCHK:
 650         interrupt->parm = irq->u.mchk.cr14;
 651         interrupt->parm64 = irq->u.mchk.mcic;
 652         break;
 653     case KVM_S390_INT_EXTERNAL_CALL:
 654         interrupt->parm = irq->u.extcall.code;
 655         break;
 656     case KVM_S390_INT_EMERGENCY:
 657         interrupt->parm = irq->u.emerg.code;
 658         break;
 659     case KVM_S390_SIGP_STOP:
 660     case KVM_S390_RESTART:
 661         break; /* These types have no parameters */
 662     case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
 663         interrupt->parm = irq->u.io.subchannel_id << 16;
 664         interrupt->parm |= irq->u.io.subchannel_nr;
 665         interrupt->parm64 = (uint64_t)irq->u.io.io_int_parm << 32;
 666         interrupt->parm64 |= irq->u.io.io_int_word;
 667         break;
 668     default:
 669         r = -EINVAL;
 670         break;
 671     }
 672     return r;
 673 }
 674
 675 void kvm_s390_vcpu_interrupt(S390CPU *cpu, struct kvm_s390_irq *irq)
 676 {
 677     struct kvm_s390_interrupt kvmint = {};
 678     CPUState *cs = CPU(cpu);
 679     int r;
 680
 681     r = s390_kvm_irq_to_interrupt(irq, &kvmint);
 682     if (r < 0) {
 683         fprintf(stderr, "%s called with bogus interrupt\n", __func__);
 684         exit(1);
 685     }
 686
 687     r = kvm_vcpu_ioctl(cs, KVM_S390_INTERRUPT, &kvmint);
 688     if (r < 0) {
 689         fprintf(stderr, "KVM failed to inject interrupt\n");
 690         exit(1);
 691     }
 692 }
 693
 694 static void __kvm_s390_floating_interrupt(struct kvm_s390_irq *irq)
 695 {
 696     struct kvm_s390_interrupt kvmint = {};
 697     int r;
 698
 699     r = s390_kvm_irq_to_interrupt(irq, &kvmint);
 700     if (r < 0) {
 701         fprintf(stderr, "%s called with bogus interrupt\n", __func__);
 702         exit(1);
 703     }
 704
 705     r = kvm_vm_ioctl(kvm_state, KVM_S390_INTERRUPT, &kvmint);
 706     if (r < 0) {
 707         fprintf(stderr, "KVM failed to inject interrupt\n");
 708         exit(1);
 709     }
 710 }
 711
 712 void kvm_s390_floating_interrupt(struct kvm_s390_irq *irq)
 713 {
 714     static bool use_flic = true;
 715     int r;
 716
 717     if (use_flic) {
 718         r = kvm_s390_inject_flic(irq);
 719         if (r == -ENOSYS) {
 720             use_flic = false;
 721         }
 722         if (!r) {
 723             return;
 724         }
 725     }
 726     __kvm_s390_floating_interrupt(irq);
 727 }
 728
 729 void kvm_s390_virtio_irq(int config_change, uint64_t token)
 730 {
 731     struct kvm_s390_irq irq = {
 732         .type = KVM_S390_INT_VIRTIO,
 733         .u.ext.ext_params = config_change,
 734         .u.ext.ext_params2 = token,
 735     };
 736
 737     kvm_s390_floating_interrupt(&irq);
 738 }
 739
 740 void kvm_s390_service_interrupt(uint32_t parm)
 741 {
 742     struct kvm_s390_irq irq = {
 743         .type = KVM_S390_INT_SERVICE,
 744         .u.ext.ext_params = parm,
 745     };
 746
 747     kvm_s390_floating_interrupt(&irq);
 748 }
 749
 750 static void enter_pgmcheck(S390CPU *cpu, uint16_t code)
 751 {
 752     struct kvm_s390_irq irq = {
 753         .type = KVM_S390_PROGRAM_INT,
 754         .u.pgm.code = code,
 755     };
 756
 757     kvm_s390_vcpu_interrupt(cpu, &irq);
 758 }
 759
 760 static int kvm_sclp_service_call(S390CPU *cpu, struct kvm_run *run,
 761                                  uint16_t ipbh0)
 762 {
 763     CPUS390XState *env = &cpu->env;
 764     uint64_t sccb;
 765     uint32_t code;
 766     int r = 0;
 767
 768     cpu_synchronize_state(CPU(cpu));
 769     sccb = env->regs[ipbh0 & 0xf];
 770     code = env->regs[(ipbh0 & 0xf0) >> 4];
 771
 772     r = sclp_service_call(env, sccb, code);
 773     if (r < 0) {
 774         enter_pgmcheck(cpu, -r);
 775     } else {
 776         setcc(cpu, r);
 777     }
 778
 779     return 0;
 780 }
 781
 782 static int handle_b2(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
 783 {
 784     CPUS390XState *env = &cpu->env;
 785     int rc = 0;
 786     uint16_t ipbh0 = (run->s390_sieic.ipb & 0xffff0000) >> 16;
 787
 788     cpu_synchronize_state(CPU(cpu));
 789
 790     switch (ipa1) {
 791     case PRIV_B2_XSCH:
 792         ioinst_handle_xsch(cpu, env->regs[1]);
 793         break;
 794     case PRIV_B2_CSCH:
 795         ioinst_handle_csch(cpu, env->regs[1]);
 796         break;
 797     case PRIV_B2_HSCH:
 798         ioinst_handle_hsch(cpu, env->regs[1]);
 799         break;
 800     case PRIV_B2_MSCH:
 801         ioinst_handle_msch(cpu, env->regs[1], run->s390_sieic.ipb);
 802         break;
 803     case PRIV_B2_SSCH:
 804         ioinst_handle_ssch(cpu, env->regs[1], run->s390_sieic.ipb);
 805         break;
 806     case PRIV_B2_STCRW:
 807         ioinst_handle_stcrw(cpu, run->s390_sieic.ipb);
 808         break;
 809     case PRIV_B2_STSCH:
 810         ioinst_handle_stsch(cpu, env->regs[1], run->s390_sieic.ipb);
 811         break;
 812     case PRIV_B2_TSCH:
 813         /* We should only get tsch via KVM_EXIT_S390_TSCH. */
 814         fprintf(stderr, "Spurious tsch intercept\n");
 815         break;
 816     case PRIV_B2_CHSC:
 817         ioinst_handle_chsc(cpu, run->s390_sieic.ipb);
 818         break;
 819     case PRIV_B2_TPI:
 820         /* This should have been handled by kvm already. */
 821         fprintf(stderr, "Spurious tpi intercept\n");
 822         break;
 823     case PRIV_B2_SCHM:
 824         ioinst_handle_schm(cpu, env->regs[1], env->regs[2],
 825                            run->s390_sieic.ipb);
 826         break;
 827     case PRIV_B2_RSCH:
 828         ioinst_handle_rsch(cpu, env->regs[1]);
 829         break;
 830     case PRIV_B2_RCHP:
 831         ioinst_handle_rchp(cpu, env->regs[1]);
 832         break;
 833     case PRIV_B2_STCPS:
 834         /* We do not provide this instruction, it is suppressed. */
 835         break;
 836     case PRIV_B2_SAL:
 837         ioinst_handle_sal(cpu, env->regs[1]);
 838         break;
 839     case PRIV_B2_SIGA:
 840         /* Not provided, set CC = 3 for subchannel not operational */
 841         setcc(cpu, 3);
 842         break;
 843     case PRIV_B2_SCLP_CALL:
 844         rc = kvm_sclp_service_call(cpu, run, ipbh0);
 845         break;
 846     default:
 847         rc = -1;
 848         DPRINTF("KVM: unhandled PRIV: 0xb2%x\n", ipa1);
 849         break;
 850     }
 851
 852     return rc;
 853 }
 854
 855 static uint64_t get_base_disp_rxy(S390CPU *cpu, struct kvm_run *run)
 856 {
 857     CPUS390XState *env = &cpu->env;
 858     uint32_t x2 = (run->s390_sieic.ipa & 0x000f);
 859     uint32_t base2 = run->s390_sieic.ipb >> 28;
 860     uint32_t disp2 = ((run->s390_sieic.ipb & 0x0fff0000) >> 16) +
 861                      ((run->s390_sieic.ipb & 0xff00) << 4);
 862
 863     if (disp2 & 0x80000) {
 864         disp2 += 0xfff00000;
 865     }
 866
 867     return (base2 ? env->regs[base2] : 0) +
 868            (x2 ? env->regs[x2] : 0) + (long)(int)disp2;
 869 }
 870
 871 static uint64_t get_base_disp_rsy(S390CPU *cpu, struct kvm_run *run)
 872 {
 873     CPUS390XState *env = &cpu->env;
 874     uint32_t base2 = run->s390_sieic.ipb >> 28;
 875     uint32_t disp2 = ((run->s390_sieic.ipb & 0x0fff0000) >> 16) +
 876                      ((run->s390_sieic.ipb & 0xff00) << 4);
 877
 878     if (disp2 & 0x80000) {
 879         disp2 += 0xfff00000;
 880     }
 881
 882     return (base2 ? env->regs[base2] : 0) + (long)(int)disp2;
 883 }
 884
 885 static int kvm_clp_service_call(S390CPU *cpu, struct kvm_run *run)
 886 {
 887     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
 888
 889     return clp_service_call(cpu, r2);
 890 }
 891
 892 static int kvm_pcilg_service_call(S390CPU *cpu, struct kvm_run *run)
 893 {
 894     uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
 895     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
 896
 897     return pcilg_service_call(cpu, r1, r2);
 898 }
 899
 900 static int kvm_pcistg_service_call(S390CPU *cpu, struct kvm_run *run)
 901 {
 902     uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
 903     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
 904
 905     return pcistg_service_call(cpu, r1, r2);
 906 }
 907
 908 static int kvm_stpcifc_service_call(S390CPU *cpu, struct kvm_run *run)
 909 {
 910     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
 911     uint64_t fiba;
 912
 913     cpu_synchronize_state(CPU(cpu));
 914     fiba = get_base_disp_rxy(cpu, run);
 915
 916     return stpcifc_service_call(cpu, r1, fiba);
 917 }
 918
 919 static int kvm_sic_service_call(S390CPU *cpu, struct kvm_run *run)
 920 {
 921     /* NOOP */
 922     return 0;
 923 }
 924
 925 static int kvm_rpcit_service_call(S390CPU *cpu, struct kvm_run *run)
 926 {
 927     uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
 928     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
 929
 930     return rpcit_service_call(cpu, r1, r2);
 931 }
 932
 933 static int kvm_pcistb_service_call(S390CPU *cpu, struct kvm_run *run)
 934 {
 935     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
 936     uint8_t r3 = run->s390_sieic.ipa & 0x000f;
 937     uint64_t gaddr;
 938
 939     cpu_synchronize_state(CPU(cpu));
 940     gaddr = get_base_disp_rsy(cpu, run);
 941
 942     return pcistb_service_call(cpu, r1, r3, gaddr);
 943 }
 944
 945 static int kvm_mpcifc_service_call(S390CPU *cpu, struct kvm_run *run)
 946 {
 947     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
 948     uint64_t fiba;
 949
 950     cpu_synchronize_state(CPU(cpu));
 951     fiba = get_base_disp_rxy(cpu, run);
 952
 953     return mpcifc_service_call(cpu, r1, fiba);
 954 }
 955
 956 static int handle_b9(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
 957 {
 958     int r = 0;
 959
 960     switch (ipa1) {
 961     case PRIV_B9_CLP:
 962         r = kvm_clp_service_call(cpu, run);
 963         break;
 964     case PRIV_B9_PCISTG:
 965         r = kvm_pcistg_service_call(cpu, run);
 966         break;
 967     case PRIV_B9_PCILG:
 968         r = kvm_pcilg_service_call(cpu, run);
 969         break;
 970     case PRIV_B9_RPCIT:
 971         r = kvm_rpcit_service_call(cpu, run);
 972         break;
 973     case PRIV_B9_EQBS:
 974         /* just inject exception */
 975         r = -1;
 976         break;
 977     default:
 978         r = -1;
 979         DPRINTF("KVM: unhandled PRIV: 0xb9%x\n", ipa1);
 980         break;
 981     }
 982
 983     return r;
 984 }
 985
 986 static int handle_eb(S390CPU *cpu, struct kvm_run *run, uint8_t ipbl)
 987 {
 988     int r = 0;
 989
 990     switch (ipbl) {
 991     case PRIV_EB_PCISTB:
 992         r = kvm_pcistb_service_call(cpu, run);
 993         break;
 994     case PRIV_EB_SIC:
 995         r = kvm_sic_service_call(cpu, run);
 996         break;
 997     case PRIV_EB_SQBS:
 998         /* just inject exception */
 999         r = -1;
1000         break;
1001     default:
1002         r = -1;
1003         DPRINTF("KVM: unhandled PRIV: 0xeb%x\n", ipbl);
1004         break;
1005     }
1006
1007     return r;
1008 }
1009
1010 static int handle_e3(S390CPU *cpu, struct kvm_run *run, uint8_t ipbl)
1011 {
1012     int r = 0;
1013
1014     switch (ipbl) {
1015     case PRIV_E3_MPCIFC:
1016         r = kvm_mpcifc_service_call(cpu, run);
1017         break;
1018     case PRIV_E3_STPCIFC:
1019         r = kvm_stpcifc_service_call(cpu, run);
1020         break;
1021     default:
1022         r = -1;
1023         DPRINTF("KVM: unhandled PRIV: 0xe3%x\n", ipbl);
1024         break;
1025     }
1026
1027     return r;
1028 }
1029
1030 static int handle_hypercall(S390CPU *cpu, struct kvm_run *run)
1031 {
1032     CPUS390XState *env = &cpu->env;
1033     int ret;
1034
1035     cpu_synchronize_state(CPU(cpu));
1036     ret = s390_virtio_hypercall(env);
1037     if (ret == -EINVAL) {
1038         enter_pgmcheck(cpu, PGM_SPECIFICATION);
1039         return 0;
1040     }
1041
1042     return ret;
1043 }
1044
1045 static void kvm_handle_diag_308(S390CPU *cpu, struct kvm_run *run)
1046 {
1047     uint64_t r1, r3;
1048
1049     cpu_synchronize_state(CPU(cpu));
1050     r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1051     r3 = run->s390_sieic.ipa & 0x000f;
1052     handle_diag_308(&cpu->env, r1, r3);
1053 }
1054
1055 static int handle_sw_breakpoint(S390CPU *cpu, struct kvm_run *run)
1056 {
1057     CPUS390XState *env = &cpu->env;
1058     unsigned long pc;
1059
1060     cpu_synchronize_state(CPU(cpu));
1061
1062     pc = env->psw.addr - 4;
1063     if (kvm_find_sw_breakpoint(CPU(cpu), pc)) {
1064         env->psw.addr = pc;
1065         return EXCP_DEBUG;
1066     }
1067
1068     return -ENOENT;
1069 }
1070
1071 #define DIAG_KVM_CODE_MASK 0x000000000000ffff
1072
1073 static int handle_diag(S390CPU *cpu, struct kvm_run *run, uint32_t ipb)
1074 {
1075     int r = 0;
1076     uint16_t func_code;
1077
1078     /*
1079      * For any diagnose call we support, bits 48-63 of the resulting
1080      * address specify the function code; the remainder is ignored.
1081      */
1082     func_code = decode_basedisp_rs(&cpu->env, ipb) & DIAG_KVM_CODE_MASK;
1083     switch (func_code) {
1084     case DIAG_IPL:
1085         kvm_handle_diag_308(cpu, run);
1086         break;
1087     case DIAG_KVM_HYPERCALL:
1088         r = handle_hypercall(cpu, run);
1089         break;
1090     case DIAG_KVM_BREAKPOINT:
1091         r = handle_sw_breakpoint(cpu, run);
1092         break;
1093     default:
1094         DPRINTF("KVM: unknown DIAG: 0x%x\n", func_code);
1095         enter_pgmcheck(cpu, PGM_SPECIFICATION);
1096         break;
1097     }
1098
1099     return r;
1100 }
1101
1102 static void sigp_cpu_start(void *arg)
1103 {
1104     CPUState *cs = arg;
1105     S390CPU *cpu = S390_CPU(cs);
1106
1107     s390_cpu_set_state(CPU_STATE_OPERATING, cpu);
1108     DPRINTF("DONE: KVM cpu start: %p\n", &cpu->env);
1109 }
1110
1111 static void sigp_cpu_restart(void *arg)
1112 {
1113     CPUState *cs = arg;
1114     S390CPU *cpu = S390_CPU(cs);
1115     struct kvm_s390_irq irq = {
1116         .type = KVM_S390_RESTART,
1117     };
1118
1119     kvm_s390_vcpu_interrupt(cpu, &irq);
1120     s390_cpu_set_state(CPU_STATE_OPERATING, cpu);
1121 }
1122
1123 int kvm_s390_cpu_restart(S390CPU *cpu)
1124 {
1125     run_on_cpu(CPU(cpu), sigp_cpu_restart, CPU(cpu));
1126     DPRINTF("DONE: KVM cpu restart: %p\n", &cpu->env);
1127     return 0;
1128 }
1129
1130 static void sigp_initial_cpu_reset(void *arg)
1131 {
1132     CPUState *cpu = arg;
1133     S390CPUClass *scc = S390_CPU_GET_CLASS(cpu);
1134
1135     cpu_synchronize_state(cpu);
1136     scc->initial_cpu_reset(cpu);
1137     cpu_synchronize_post_reset(cpu);
1138 }
1139
1140 static void sigp_cpu_reset(void *arg)
1141 {
1142     CPUState *cpu = arg;
1143     S390CPUClass *scc = S390_CPU_GET_CLASS(cpu);
1144
1145     cpu_synchronize_state(cpu);
1146     scc->cpu_reset(cpu);
1147     cpu_synchronize_post_reset(cpu);
1148 }
1149
1150 #define SIGP_ORDER_MASK 0x000000ff
1151
1152 static int handle_sigp(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
1153 {
1154     CPUS390XState *env = &cpu->env;
1155     uint8_t order_code;
1156     uint16_t cpu_addr;
1157     S390CPU *target_cpu;
1158     uint64_t *statusreg = &env->regs[ipa1 >> 4];
1159     int cc;
1160
1161     cpu_synchronize_state(CPU(cpu));
1162
1163     /* get order code */
1164     order_code = decode_basedisp_rs(env, run->s390_sieic.ipb) & SIGP_ORDER_MASK;
1165
1166     cpu_addr = env->regs[ipa1 & 0x0f];
1167     target_cpu = s390_cpu_addr2state(cpu_addr);
1168     if (target_cpu == NULL) {
1169         cc = 3;    /* not operational */
1170         goto out;
1171     }
1172
1173     switch (order_code) {
1174     case SIGP_START:
1175         run_on_cpu(CPU(target_cpu), sigp_cpu_start, CPU(target_cpu));
1176         cc = 0;
1177         break;
1178     case SIGP_RESTART:
1179         run_on_cpu(CPU(target_cpu), sigp_cpu_restart, CPU(target_cpu));
1180         cc = 0;
1181         break;
1182     case SIGP_SET_ARCH:
1183         *statusreg &= 0xffffffff00000000UL;
1184         *statusreg |= SIGP_STAT_INVALID_PARAMETER;
1185         cc = 1;   /* status stored */
1186         break;
1187     case SIGP_INITIAL_CPU_RESET:
1188         run_on_cpu(CPU(target_cpu), sigp_initial_cpu_reset, CPU(target_cpu));
1189         cc = 0;
1190         break;
1191     case SIGP_CPU_RESET:
1192         run_on_cpu(CPU(target_cpu), sigp_cpu_reset, CPU(target_cpu));
1193         cc = 0;
1194         break;
1195     default:
1196         DPRINTF("KVM: unknown SIGP: 0x%x\n", order_code);
1197         *statusreg &= 0xffffffff00000000UL;
1198         *statusreg |= SIGP_STAT_INVALID_ORDER;
1199         cc = 1;   /* status stored */
1200         break;
1201     }
1202
1203 out:
1204     setcc(cpu, cc);
1205     return 0;
1206 }
1207
1208 static int handle_instruction(S390CPU *cpu, struct kvm_run *run)
1209 {
1210     unsigned int ipa0 = (run->s390_sieic.ipa & 0xff00);
1211     uint8_t ipa1 = run->s390_sieic.ipa & 0x00ff;
1212     int r = -1;
1213
1214     DPRINTF("handle_instruction 0x%x 0x%x\n",
1215             run->s390_sieic.ipa, run->s390_sieic.ipb);
1216     switch (ipa0) {
1217     case IPA0_B2:
1218         r = handle_b2(cpu, run, ipa1);
1219         break;
1220     case IPA0_B9:
1221         r = handle_b9(cpu, run, ipa1);
1222         break;
1223     case IPA0_EB:
1224         r = handle_eb(cpu, run, run->s390_sieic.ipb & 0xff);
1225         break;
1226     case IPA0_E3:
1227         r = handle_e3(cpu, run, run->s390_sieic.ipb & 0xff);
1228         break;
1229     case IPA0_DIAG:
1230         r = handle_diag(cpu, run, run->s390_sieic.ipb);
1231         break;
1232     case IPA0_SIGP:
1233         r = handle_sigp(cpu, run, ipa1);
1234         break;
1235     }
1236
1237     if (r < 0) {
1238         r = 0;
1239         enter_pgmcheck(cpu, 0x0001);
1240     }
1241
1242     return r;
1243 }
1244
1245 static bool is_special_wait_psw(CPUState *cs)
1246 {
1247     /* signal quiesce */
1248     return cs->kvm_run->psw_addr == 0xfffUL;
1249 }
1250
1251 static void guest_panicked(void)
1252 {
1253     qapi_event_send_guest_panicked(GUEST_PANIC_ACTION_PAUSE,
1254                                    &error_abort);
1255     vm_stop(RUN_STATE_GUEST_PANICKED);
1256 }
1257
1258 static void unmanageable_intercept(S390CPU *cpu, const char *str, int pswoffset)
1259 {
1260     CPUState *cs = CPU(cpu);
1261
1262     error_report("Unmanageable %s! CPU%i new PSW: 0x%016lx:%016lx",
1263                  str, cs->cpu_index, ldq_phys(cs->as, cpu->env.psa + pswoffset),
1264                  ldq_phys(cs->as, cpu->env.psa + pswoffset + 8));
1265     s390_cpu_halt(cpu);
1266     guest_panicked();
1267 }
1268
1269 static int handle_intercept(S390CPU *cpu)
1270 {
1271     CPUState *cs = CPU(cpu);
1272     struct kvm_run *run = cs->kvm_run;
1273     int icpt_code = run->s390_sieic.icptcode;
1274     int r = 0;
1275
1276     DPRINTF("intercept: 0x%x (at 0x%lx)\n", icpt_code,
1277             (long)cs->kvm_run->psw_addr);
1278     switch (icpt_code) {
1279         case ICPT_INSTRUCTION:
1280             r = handle_instruction(cpu, run);
1281             break;
1282         case ICPT_PROGRAM:
1283             unmanageable_intercept(cpu, "program interrupt",
1284                                    offsetof(LowCore, program_new_psw));
1285             r = EXCP_HALTED;
1286             break;
1287         case ICPT_EXT_INT:
1288             unmanageable_intercept(cpu, "external interrupt",
1289                                    offsetof(LowCore, external_new_psw));
1290             r = EXCP_HALTED;
1291             break;
1292         case ICPT_WAITPSW:
1293             /* disabled wait, since enabled wait is handled in kernel */
1294             cpu_synchronize_state(cs);
1295             if (s390_cpu_halt(cpu) == 0) {
1296                 if (is_special_wait_psw(cs)) {
1297                     qemu_system_shutdown_request();
1298                 } else {
1299                     guest_panicked();
1300                 }
1301             }
1302             r = EXCP_HALTED;
1303             break;
1304         case ICPT_CPU_STOP:
1305             if (s390_cpu_set_state(CPU_STATE_STOPPED, cpu) == 0) {
1306                 qemu_system_shutdown_request();
1307             }
1308             r = EXCP_HALTED;
1309             break;
1310         case ICPT_SOFT_INTERCEPT:
1311             fprintf(stderr, "KVM unimplemented icpt SOFT\n");
1312             exit(1);
1313             break;
1314         case ICPT_IO:
1315             fprintf(stderr, "KVM unimplemented icpt IO\n");
1316             exit(1);
1317             break;
1318         default:
1319             fprintf(stderr, "Unknown intercept code: %d\n", icpt_code);
1320             exit(1);
1321             break;
1322     }
1323
1324     return r;
1325 }
1326
1327 static int handle_tsch(S390CPU *cpu)
1328 {
1329     CPUS390XState *env = &cpu->env;
1330     CPUState *cs = CPU(cpu);
1331     struct kvm_run *run = cs->kvm_run;
1332     int ret;
1333
1334     cpu_synchronize_state(cs);
1335
1336     ret = ioinst_handle_tsch(env, env->regs[1], run->s390_tsch.ipb);
1337     if (ret >= 0) {
1338         /* Success; set condition code. */
1339         setcc(cpu, ret);
1340         ret = 0;
1341     } else if (ret < -1) {
1342         /*
1343          * Failure.
1344          * If an I/O interrupt had been dequeued, we have to reinject it.
1345          */
1346         if (run->s390_tsch.dequeued) {
1347             kvm_s390_io_interrupt(run->s390_tsch.subchannel_id,
1348                                   run->s390_tsch.subchannel_nr,
1349                                   run->s390_tsch.io_int_parm,
1350                                   run->s390_tsch.io_int_word);
1351         }
1352         ret = 0;
1353     }
1354     return ret;
1355 }
1356
1357 static int kvm_arch_handle_debug_exit(S390CPU *cpu)
1358 {
1359     CPUState *cs = CPU(cpu);
1360     struct kvm_run *run = cs->kvm_run;
1361
1362     int ret = 0;
1363     struct kvm_debug_exit_arch *arch_info = &run->debug.arch;
1364
1365     switch (arch_info->type) {
1366     case KVM_HW_WP_WRITE:
1367         if (find_hw_breakpoint(arch_info->addr, -1, arch_info->type)) {
1368             cs->watchpoint_hit = &hw_watchpoint;
1369             hw_watchpoint.vaddr = arch_info->addr;
1370             hw_watchpoint.flags = BP_MEM_WRITE;
1371             ret = EXCP_DEBUG;
1372         }
1373         break;
1374     case KVM_HW_BP:
1375         if (find_hw_breakpoint(arch_info->addr, -1, arch_info->type)) {
1376             ret = EXCP_DEBUG;
1377         }
1378         break;
1379     case KVM_SINGLESTEP:
1380         if (cs->singlestep_enabled) {
1381             ret = EXCP_DEBUG;
1382         }
1383         break;
1384     default:
1385         ret = -ENOSYS;
1386     }
1387
1388     return ret;
1389 }
1390
1391 int kvm_arch_handle_exit(CPUState *cs, struct kvm_run *run)
1392 {
1393     S390CPU *cpu = S390_CPU(cs);
1394     int ret = 0;
1395
1396     switch (run->exit_reason) {
1397         case KVM_EXIT_S390_SIEIC:
1398             ret = handle_intercept(cpu);
1399             break;
1400         case KVM_EXIT_S390_RESET:
1401             s390_reipl_request();
1402             break;
1403         case KVM_EXIT_S390_TSCH:
1404             ret = handle_tsch(cpu);
1405             break;
1406         case KVM_EXIT_DEBUG:
1407             ret = kvm_arch_handle_debug_exit(cpu);
1408             break;
1409         default:
1410             fprintf(stderr, "Unknown KVM exit: %d\n", run->exit_reason);
1411             break;
1412     }
1413
1414     if (ret == 0) {
1415         ret = EXCP_INTERRUPT;
1416     }
1417     return ret;
1418 }
1419
1420 bool kvm_arch_stop_on_emulation_error(CPUState *cpu)
1421 {
1422     return true;
1423 }
1424
1425 int kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr)
1426 {
1427     return 1;
1428 }
1429
1430 int kvm_arch_on_sigbus(int code, void *addr)
1431 {
1432     return 1;
1433 }
1434
1435 void kvm_s390_io_interrupt(uint16_t subchannel_id,
1436                            uint16_t subchannel_nr, uint32_t io_int_parm,
1437                            uint32_t io_int_word)
1438 {
1439     struct kvm_s390_irq irq = {
1440         .u.io.subchannel_id = subchannel_id,
1441         .u.io.subchannel_nr = subchannel_nr,
1442         .u.io.io_int_parm = io_int_parm,
1443         .u.io.io_int_word = io_int_word,
1444     };
1445
1446     if (io_int_word & IO_INT_WORD_AI) {
1447         irq.type = KVM_S390_INT_IO(1, 0, 0, 0);
1448     } else {
1449         irq.type = ((subchannel_id & 0xff00) << 24) |
1450             ((subchannel_id & 0x00060) << 22) | (subchannel_nr << 16);
1451     }
1452     kvm_s390_floating_interrupt(&irq);
1453 }
1454
1455 void kvm_s390_crw_mchk(void)
1456 {
1457     struct kvm_s390_irq irq = {
1458         .type = KVM_S390_MCHK,
1459         .u.mchk.cr14 = 1 << 28,
1460         .u.mchk.mcic = 0x00400f1d40330000ULL,
1461     };
1462     kvm_s390_floating_interrupt(&irq);
1463 }
1464
1465 void kvm_s390_enable_css_support(S390CPU *cpu)
1466 {
1467     int r;
1468
1469     /* Activate host kernel channel subsystem support. */
1470     r = kvm_vcpu_enable_cap(CPU(cpu), KVM_CAP_S390_CSS_SUPPORT, 0);
1471     assert(r == 0);
1472 }
1473
1474 void kvm_arch_init_irq_routing(KVMState *s)
1475 {
1476     /*
1477      * Note that while irqchip capabilities generally imply that cpustates
1478      * are handled in-kernel, it is not true for s390 (yet); therefore, we
1479      * have to override the common code kvm_halt_in_kernel_allowed setting.
1480      */
1481     if (kvm_check_extension(s, KVM_CAP_IRQ_ROUTING)) {
1482         kvm_gsi_routing_allowed = true;
1483         kvm_halt_in_kernel_allowed = false;
1484     }
1485 }
1486
1487 int kvm_s390_assign_subch_ioeventfd(EventNotifier *notifier, uint32_t sch,
1488                                     int vq, bool assign)
1489 {
1490     struct kvm_ioeventfd kick = {
1491         .flags = KVM_IOEVENTFD_FLAG_VIRTIO_CCW_NOTIFY |
1492         KVM_IOEVENTFD_FLAG_DATAMATCH,
1493         .fd = event_notifier_get_fd(notifier),
1494         .datamatch = vq,
1495         .addr = sch,
1496         .len = 8,
1497     };
1498     if (!kvm_check_extension(kvm_state, KVM_CAP_IOEVENTFD)) {
1499         return -ENOSYS;
1500     }
1501     if (!assign) {
1502         kick.flags |= KVM_IOEVENTFD_FLAG_DEASSIGN;
1503     }
1504     return kvm_vm_ioctl(kvm_state, KVM_IOEVENTFD, &kick);
1505 }
1506
1507 int kvm_s390_get_memslot_count(KVMState *s)
1508 {
1509     return kvm_check_extension(s, KVM_CAP_NR_MEMSLOTS);
1510 }
1511
1512 int kvm_s390_set_cpu_state(S390CPU *cpu, uint8_t cpu_state)
1513 {
1514     struct kvm_mp_state mp_state = {};
1515     int ret;
1516
1517     /* the kvm part might not have been initialized yet */
1518     if (CPU(cpu)->kvm_state == NULL) {
1519         return 0;
1520     }
1521
1522     switch (cpu_state) {
1523     case CPU_STATE_STOPPED:
1524         mp_state.mp_state = KVM_MP_STATE_STOPPED;
1525         break;
1526     case CPU_STATE_CHECK_STOP:
1527         mp_state.mp_state = KVM_MP_STATE_CHECK_STOP;
1528         break;
1529     case CPU_STATE_OPERATING:
1530         mp_state.mp_state = KVM_MP_STATE_OPERATING;
1531         break;
1532     case CPU_STATE_LOAD:
1533         mp_state.mp_state = KVM_MP_STATE_LOAD;
1534         break;
1535     default:
1536         error_report("Requested CPU state is not a valid S390 CPU state: %u",
1537                      cpu_state);
1538         exit(1);
1539     }
1540
1541     ret = kvm_vcpu_ioctl(CPU(cpu), KVM_SET_MP_STATE, &mp_state);
1542     if (ret) {
1543         trace_kvm_failed_cpu_state_set(CPU(cpu)->cpu_index, cpu_state,
1544                                        strerror(-ret));
1545     }
1546
1547     return ret;
1548 }
1549
1550 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry *route,
1551                               uint64_t address, uint32_t data)
1552 {
1553     S390PCIBusDevice *pbdev;
1554     uint32_t fid = data >> ZPCI_MSI_VEC_BITS;
1555     uint32_t vec = data & ZPCI_MSI_VEC_MASK;
1556
1557     pbdev = s390_pci_find_dev_by_fid(fid);
1558     if (!pbdev) {
1559         DPRINTF("add_msi_route no dev\n");
1560         return -ENODEV;
1561     }
1562
1563     pbdev->routes.adapter.ind_offset = vec;
1564
1565     route->type = KVM_IRQ_ROUTING_S390_ADAPTER;
1566     route->flags = 0;
1567     route->u.adapter.summary_addr = pbdev->routes.adapter.summary_addr;
1568     route->u.adapter.ind_addr = pbdev->routes.adapter.ind_addr;
1569     route->u.adapter.summary_offset = pbdev->routes.adapter.summary_offset;
1570     route->u.adapter.ind_offset = pbdev->routes.adapter.ind_offset;
1571     route->u.adapter.adapter_id = pbdev->routes.adapter.adapter_id;
1572     return 0;
1573 }