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