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/error-report.h"
  33 #include "qemu/timer.h"
  34 #include "sysemu/sysemu.h"
  35 #include "sysemu/kvm.h"
  36 #include "hw/hw.h"
  37 #include "cpu.h"
  38 #include "sysemu/device_tree.h"
  39 #include "qapi/qmp/qjson.h"
  40 #include "exec/gdbstub.h"
  41 #include "exec/address-spaces.h"
  42 #include "trace.h"
  43 #include "qapi-event.h"
  44 #include "hw/s390x/s390-pci-inst.h"
  45 #include "hw/s390x/s390-pci-bus.h"
  46 #include "hw/s390x/ipl.h"
  47 #include "hw/s390x/ebcdic.h"
  48 #include "exec/memattrs.h"
  49
  50 /* #define DEBUG_KVM */
  51
  52 #ifdef DEBUG_KVM
  53 #define DPRINTF(fmt, ...) \
  54     do { fprintf(stderr, fmt, ## __VA_ARGS__); } while (0)
  55 #else
  56 #define DPRINTF(fmt, ...) \
  57     do { } while (0)
  58 #endif
  59
  60 #define kvm_vm_check_mem_attr(s, attr) \
  61     kvm_vm_check_attr(s, KVM_S390_VM_MEM_CTRL, attr)
  62
  63 #define IPA0_DIAG                       0x8300
  64 #define IPA0_SIGP                       0xae00
  65 #define IPA0_B2                         0xb200
  66 #define IPA0_B9                         0xb900
  67 #define IPA0_EB                         0xeb00
  68 #define IPA0_E3                         0xe300
  69
  70 #define PRIV_B2_SCLP_CALL               0x20
  71 #define PRIV_B2_CSCH                    0x30
  72 #define PRIV_B2_HSCH                    0x31
  73 #define PRIV_B2_MSCH                    0x32
  74 #define PRIV_B2_SSCH                    0x33
  75 #define PRIV_B2_STSCH                   0x34
  76 #define PRIV_B2_TSCH                    0x35
  77 #define PRIV_B2_TPI                     0x36
  78 #define PRIV_B2_SAL                     0x37
  79 #define PRIV_B2_RSCH                    0x38
  80 #define PRIV_B2_STCRW                   0x39
  81 #define PRIV_B2_STCPS                   0x3a
  82 #define PRIV_B2_RCHP                    0x3b
  83 #define PRIV_B2_SCHM                    0x3c
  84 #define PRIV_B2_CHSC                    0x5f
  85 #define PRIV_B2_SIGA                    0x74
  86 #define PRIV_B2_XSCH                    0x76
  87
  88 #define PRIV_EB_SQBS                    0x8a
  89 #define PRIV_EB_PCISTB                  0xd0
  90 #define PRIV_EB_SIC                     0xd1
  91
  92 #define PRIV_B9_EQBS                    0x9c
  93 #define PRIV_B9_CLP                     0xa0
  94 #define PRIV_B9_PCISTG                  0xd0
  95 #define PRIV_B9_PCILG                   0xd2
  96 #define PRIV_B9_RPCIT                   0xd3
  97
  98 #define PRIV_E3_MPCIFC                  0xd0
  99 #define PRIV_E3_STPCIFC                 0xd4
 100
 101 #define DIAG_TIMEREVENT                 0x288
 102 #define DIAG_IPL                        0x308
 103 #define DIAG_KVM_HYPERCALL              0x500
 104 #define DIAG_KVM_BREAKPOINT             0x501
 105
 106 #define ICPT_INSTRUCTION                0x04
 107 #define ICPT_PROGRAM                    0x08
 108 #define ICPT_EXT_INT                    0x14
 109 #define ICPT_WAITPSW                    0x1c
 110 #define ICPT_SOFT_INTERCEPT             0x24
 111 #define ICPT_CPU_STOP                   0x28
 112 #define ICPT_IO                         0x40
 113
 114 #define NR_LOCAL_IRQS 32
 115 /*
 116  * Needs to be big enough to contain max_cpus emergency signals
 117  * and in addition NR_LOCAL_IRQS interrupts
 118  */
 119 #define VCPU_IRQ_BUF_SIZE (sizeof(struct kvm_s390_irq) * \
 120                            (max_cpus + NR_LOCAL_IRQS))
 121
 122 static CPUWatchpoint hw_watchpoint;
 123 /*
 124  * We don't use a list because this structure is also used to transmit the
 125  * hardware breakpoints to the kernel.
 126  */
 127 static struct kvm_hw_breakpoint *hw_breakpoints;
 128 static int nb_hw_breakpoints;
 129
 130 const KVMCapabilityInfo kvm_arch_required_capabilities[] = {
 131     KVM_CAP_LAST_INFO
 132 };
 133
 134 static int cap_sync_regs;
 135 static int cap_async_pf;
 136 static int cap_mem_op;
 137 static int cap_s390_irq;
 138
 139 static void *legacy_s390_alloc(size_t size, uint64_t *align);
 140
 141 static int kvm_s390_query_mem_limit(KVMState *s, uint64_t *memory_limit)
 142 {
 143     struct kvm_device_attr attr = {
 144         .group = KVM_S390_VM_MEM_CTRL,
 145         .attr = KVM_S390_VM_MEM_LIMIT_SIZE,
 146         .addr = (uint64_t) memory_limit,
 147     };
 148
 149     return kvm_vm_ioctl(s, KVM_GET_DEVICE_ATTR, &attr);
 150 }
 151
 152 int kvm_s390_set_mem_limit(KVMState *s, uint64_t new_limit, uint64_t *hw_limit)
 153 {
 154     int rc;
 155
 156     struct kvm_device_attr attr = {
 157         .group = KVM_S390_VM_MEM_CTRL,
 158         .attr = KVM_S390_VM_MEM_LIMIT_SIZE,
 159         .addr = (uint64_t) &new_limit,
 160     };
 161
 162     if (!kvm_vm_check_mem_attr(s, KVM_S390_VM_MEM_LIMIT_SIZE)) {
 163         return 0;
 164     }
 165
 166     rc = kvm_s390_query_mem_limit(s, hw_limit);
 167     if (rc) {
 168         return rc;
 169     } else if (*hw_limit < new_limit) {
 170         return -E2BIG;
 171     }
 172
 173     return kvm_vm_ioctl(s, KVM_SET_DEVICE_ATTR, &attr);
 174 }
 175
 176 void kvm_s390_cmma_reset(void)
 177 {
 178     int rc;
 179     struct kvm_device_attr attr = {
 180         .group = KVM_S390_VM_MEM_CTRL,
 181         .attr = KVM_S390_VM_MEM_CLR_CMMA,
 182     };
 183
 184     rc = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
 185     trace_kvm_clear_cmma(rc);
 186 }
 187
 188 static void kvm_s390_enable_cmma(KVMState *s)
 189 {
 190     int rc;
 191     struct kvm_device_attr attr = {
 192         .group = KVM_S390_VM_MEM_CTRL,
 193         .attr = KVM_S390_VM_MEM_ENABLE_CMMA,
 194     };
 195
 196     if (!kvm_vm_check_mem_attr(s, KVM_S390_VM_MEM_ENABLE_CMMA) ||
 197         !kvm_vm_check_mem_attr(s, KVM_S390_VM_MEM_CLR_CMMA)) {
 198         return;
 199     }
 200
 201     rc = kvm_vm_ioctl(s, KVM_SET_DEVICE_ATTR, &attr);
 202     trace_kvm_enable_cmma(rc);
 203 }
 204
 205 static void kvm_s390_set_attr(uint64_t attr)
 206 {
 207     struct kvm_device_attr attribute = {
 208         .group = KVM_S390_VM_CRYPTO,
 209         .attr  = attr,
 210     };
 211
 212     int ret = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attribute);
 213
 214     if (ret) {
 215         error_report("Failed to set crypto device attribute %lu: %s",
 216                      attr, strerror(-ret));
 217     }
 218 }
 219
 220 static void kvm_s390_init_aes_kw(void)
 221 {
 222     uint64_t attr = KVM_S390_VM_CRYPTO_DISABLE_AES_KW;
 223
 224     if (object_property_get_bool(OBJECT(qdev_get_machine()), "aes-key-wrap",
 225                                  NULL)) {
 226             attr = KVM_S390_VM_CRYPTO_ENABLE_AES_KW;
 227     }
 228
 229     if (kvm_vm_check_attr(kvm_state, KVM_S390_VM_CRYPTO, attr)) {
 230             kvm_s390_set_attr(attr);
 231     }
 232 }
 233
 234 static void kvm_s390_init_dea_kw(void)
 235 {
 236     uint64_t attr = KVM_S390_VM_CRYPTO_DISABLE_DEA_KW;
 237
 238     if (object_property_get_bool(OBJECT(qdev_get_machine()), "dea-key-wrap",
 239                                  NULL)) {
 240             attr = KVM_S390_VM_CRYPTO_ENABLE_DEA_KW;
 241     }
 242
 243     if (kvm_vm_check_attr(kvm_state, KVM_S390_VM_CRYPTO, attr)) {
 244             kvm_s390_set_attr(attr);
 245     }
 246 }
 247
 248 void kvm_s390_crypto_reset(void)
 249 {
 250     kvm_s390_init_aes_kw();
 251     kvm_s390_init_dea_kw();
 252 }
 253
 254 int kvm_arch_init(MachineState *ms, KVMState *s)
 255 {
 256     cap_sync_regs = kvm_check_extension(s, KVM_CAP_SYNC_REGS);
 257     cap_async_pf = kvm_check_extension(s, KVM_CAP_ASYNC_PF);
 258     cap_mem_op = kvm_check_extension(s, KVM_CAP_S390_MEM_OP);
 259     cap_s390_irq = kvm_check_extension(s, KVM_CAP_S390_INJECT_IRQ);
 260
 261     kvm_s390_enable_cmma(s);
 262
 263     if (!kvm_check_extension(s, KVM_CAP_S390_GMAP)
 264         || !kvm_check_extension(s, KVM_CAP_S390_COW)) {
 265         phys_mem_set_alloc(legacy_s390_alloc);
 266     }
 267
 268     kvm_vm_enable_cap(s, KVM_CAP_S390_USER_SIGP, 0);
 269     kvm_vm_enable_cap(s, KVM_CAP_S390_VECTOR_REGISTERS, 0);
 270     kvm_vm_enable_cap(s, KVM_CAP_S390_USER_STSI, 0);
 271
 272     return 0;
 273 }
 274
 275 unsigned long kvm_arch_vcpu_id(CPUState *cpu)
 276 {
 277     return cpu->cpu_index;
 278 }
 279
 280 int kvm_arch_init_vcpu(CPUState *cs)
 281 {
 282     S390CPU *cpu = S390_CPU(cs);
 283     kvm_s390_set_cpu_state(cpu, cpu->env.cpu_state);
 284     cpu->irqstate = g_malloc0(VCPU_IRQ_BUF_SIZE);
 285     return 0;
 286 }
 287
 288 void kvm_s390_reset_vcpu(S390CPU *cpu)
 289 {
 290     CPUState *cs = CPU(cpu);
 291
 292     /* The initial reset call is needed here to reset in-kernel
 293      * vcpu data that we can't access directly from QEMU
 294      * (i.e. with older kernels which don't support sync_regs/ONE_REG).
 295      * Before this ioctl cpu_synchronize_state() is called in common kvm
 296      * code (kvm-all) */
 297     if (kvm_vcpu_ioctl(cs, KVM_S390_INITIAL_RESET, NULL)) {
 298         error_report("Initial CPU reset failed on CPU %i", cs->cpu_index);
 299     }
 300 }
 301
 302 static int can_sync_regs(CPUState *cs, int regs)
 303 {
 304     return cap_sync_regs && (cs->kvm_run->kvm_valid_regs & regs) == regs;
 305 }
 306
 307 int kvm_arch_put_registers(CPUState *cs, int level)
 308 {
 309     S390CPU *cpu = S390_CPU(cs);
 310     CPUS390XState *env = &cpu->env;
 311     struct kvm_sregs sregs;
 312     struct kvm_regs regs;
 313     struct kvm_fpu fpu = {};
 314     int r;
 315     int i;
 316
 317     /* always save the PSW  and the GPRS*/
 318     cs->kvm_run->psw_addr = env->psw.addr;
 319     cs->kvm_run->psw_mask = env->psw.mask;
 320
 321     if (can_sync_regs(cs, KVM_SYNC_GPRS)) {
 322         for (i = 0; i < 16; i++) {
 323             cs->kvm_run->s.regs.gprs[i] = env->regs[i];
 324             cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_GPRS;
 325         }
 326     } else {
 327         for (i = 0; i < 16; i++) {
 328             regs.gprs[i] = env->regs[i];
 329         }
 330         r = kvm_vcpu_ioctl(cs, KVM_SET_REGS, &regs);
 331         if (r < 0) {
 332             return r;
 333         }
 334     }
 335
 336     if (can_sync_regs(cs, KVM_SYNC_VRS)) {
 337         for (i = 0; i < 32; i++) {
 338             cs->kvm_run->s.regs.vrs[i][0] = env->vregs[i][0].ll;
 339             cs->kvm_run->s.regs.vrs[i][1] = env->vregs[i][1].ll;
 340         }
 341         cs->kvm_run->s.regs.fpc = env->fpc;
 342         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_VRS;
 343     } else {
 344         /* Floating point */
 345         for (i = 0; i < 16; i++) {
 346             fpu.fprs[i] = get_freg(env, i)->ll;
 347         }
 348         fpu.fpc = env->fpc;
 349
 350         r = kvm_vcpu_ioctl(cs, KVM_SET_FPU, &fpu);
 351         if (r < 0) {
 352             return r;
 353         }
 354     }
 355
 356     /* Do we need to save more than that? */
 357     if (level == KVM_PUT_RUNTIME_STATE) {
 358         return 0;
 359     }
 360
 361     if (can_sync_regs(cs, KVM_SYNC_ARCH0)) {
 362         cs->kvm_run->s.regs.cputm = env->cputm;
 363         cs->kvm_run->s.regs.ckc = env->ckc;
 364         cs->kvm_run->s.regs.todpr = env->todpr;
 365         cs->kvm_run->s.regs.gbea = env->gbea;
 366         cs->kvm_run->s.regs.pp = env->pp;
 367         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_ARCH0;
 368     } else {
 369         /*
 370          * These ONE_REGS are not protected by a capability. As they are only
 371          * necessary for migration we just trace a possible error, but don't
 372          * return with an error return code.
 373          */
 374         kvm_set_one_reg(cs, KVM_REG_S390_CPU_TIMER, &env->cputm);
 375         kvm_set_one_reg(cs, KVM_REG_S390_CLOCK_COMP, &env->ckc);
 376         kvm_set_one_reg(cs, KVM_REG_S390_TODPR, &env->todpr);
 377         kvm_set_one_reg(cs, KVM_REG_S390_GBEA, &env->gbea);
 378         kvm_set_one_reg(cs, KVM_REG_S390_PP, &env->pp);
 379     }
 380
 381     /* pfault parameters */
 382     if (can_sync_regs(cs, KVM_SYNC_PFAULT)) {
 383         cs->kvm_run->s.regs.pft = env->pfault_token;
 384         cs->kvm_run->s.regs.pfs = env->pfault_select;
 385         cs->kvm_run->s.regs.pfc = env->pfault_compare;
 386         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_PFAULT;
 387     } else if (cap_async_pf) {
 388         r = kvm_set_one_reg(cs, KVM_REG_S390_PFTOKEN, &env->pfault_token);
 389         if (r < 0) {
 390             return r;
 391         }
 392         r = kvm_set_one_reg(cs, KVM_REG_S390_PFCOMPARE, &env->pfault_compare);
 393         if (r < 0) {
 394             return r;
 395         }
 396         r = kvm_set_one_reg(cs, KVM_REG_S390_PFSELECT, &env->pfault_select);
 397         if (r < 0) {
 398             return r;
 399         }
 400     }
 401
 402     /* access registers and control registers*/
 403     if (can_sync_regs(cs, KVM_SYNC_ACRS | KVM_SYNC_CRS)) {
 404         for (i = 0; i < 16; i++) {
 405             cs->kvm_run->s.regs.acrs[i] = env->aregs[i];
 406             cs->kvm_run->s.regs.crs[i] = env->cregs[i];
 407         }
 408         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_ACRS;
 409         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_CRS;
 410     } else {
 411         for (i = 0; i < 16; i++) {
 412             sregs.acrs[i] = env->aregs[i];
 413             sregs.crs[i] = env->cregs[i];
 414         }
 415         r = kvm_vcpu_ioctl(cs, KVM_SET_SREGS, &sregs);
 416         if (r < 0) {
 417             return r;
 418         }
 419     }
 420
 421     /* Finally the prefix */
 422     if (can_sync_regs(cs, KVM_SYNC_PREFIX)) {
 423         cs->kvm_run->s.regs.prefix = env->psa;
 424         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_PREFIX;
 425     } else {
 426         /* prefix is only supported via sync regs */
 427     }
 428     return 0;
 429 }
 430
 431 int kvm_arch_get_registers(CPUState *cs)
 432 {
 433     S390CPU *cpu = S390_CPU(cs);
 434     CPUS390XState *env = &cpu->env;
 435     struct kvm_sregs sregs;
 436     struct kvm_regs regs;
 437     struct kvm_fpu fpu;
 438     int i, r;
 439
 440     /* get the PSW */
 441     env->psw.addr = cs->kvm_run->psw_addr;
 442     env->psw.mask = cs->kvm_run->psw_mask;
 443
 444     /* the GPRS */
 445     if (can_sync_regs(cs, KVM_SYNC_GPRS)) {
 446         for (i = 0; i < 16; i++) {
 447             env->regs[i] = cs->kvm_run->s.regs.gprs[i];
 448         }
 449     } else {
 450         r = kvm_vcpu_ioctl(cs, KVM_GET_REGS, &regs);
 451         if (r < 0) {
 452             return r;
 453         }
 454          for (i = 0; i < 16; i++) {
 455             env->regs[i] = regs.gprs[i];
 456         }
 457     }
 458
 459     /* The ACRS and CRS */
 460     if (can_sync_regs(cs, KVM_SYNC_ACRS | KVM_SYNC_CRS)) {
 461         for (i = 0; i < 16; i++) {
 462             env->aregs[i] = cs->kvm_run->s.regs.acrs[i];
 463             env->cregs[i] = cs->kvm_run->s.regs.crs[i];
 464         }
 465     } else {
 466         r = kvm_vcpu_ioctl(cs, KVM_GET_SREGS, &sregs);
 467         if (r < 0) {
 468             return r;
 469         }
 470          for (i = 0; i < 16; i++) {
 471             env->aregs[i] = sregs.acrs[i];
 472             env->cregs[i] = sregs.crs[i];
 473         }
 474     }
 475
 476     /* Floating point and vector registers */
 477     if (can_sync_regs(cs, KVM_SYNC_VRS)) {
 478         for (i = 0; i < 32; i++) {
 479             env->vregs[i][0].ll = cs->kvm_run->s.regs.vrs[i][0];
 480             env->vregs[i][1].ll = cs->kvm_run->s.regs.vrs[i][1];
 481         }
 482         env->fpc = cs->kvm_run->s.regs.fpc;
 483     } else {
 484         r = kvm_vcpu_ioctl(cs, KVM_GET_FPU, &fpu);
 485         if (r < 0) {
 486             return r;
 487         }
 488         for (i = 0; i < 16; i++) {
 489             get_freg(env, i)->ll = fpu.fprs[i];
 490         }
 491         env->fpc = fpu.fpc;
 492     }
 493
 494     /* The prefix */
 495     if (can_sync_regs(cs, KVM_SYNC_PREFIX)) {
 496         env->psa = cs->kvm_run->s.regs.prefix;
 497     }
 498
 499     if (can_sync_regs(cs, KVM_SYNC_ARCH0)) {
 500         env->cputm = cs->kvm_run->s.regs.cputm;
 501         env->ckc = cs->kvm_run->s.regs.ckc;
 502         env->todpr = cs->kvm_run->s.regs.todpr;
 503         env->gbea = cs->kvm_run->s.regs.gbea;
 504         env->pp = cs->kvm_run->s.regs.pp;
 505     } else {
 506         /*
 507          * These ONE_REGS are not protected by a capability. As they are only
 508          * necessary for migration we just trace a possible error, but don't
 509          * return with an error return code.
 510          */
 511         kvm_get_one_reg(cs, KVM_REG_S390_CPU_TIMER, &env->cputm);
 512         kvm_get_one_reg(cs, KVM_REG_S390_CLOCK_COMP, &env->ckc);
 513         kvm_get_one_reg(cs, KVM_REG_S390_TODPR, &env->todpr);
 514         kvm_get_one_reg(cs, KVM_REG_S390_GBEA, &env->gbea);
 515         kvm_get_one_reg(cs, KVM_REG_S390_PP, &env->pp);
 516     }
 517
 518     /* pfault parameters */
 519     if (can_sync_regs(cs, KVM_SYNC_PFAULT)) {
 520         env->pfault_token = cs->kvm_run->s.regs.pft;
 521         env->pfault_select = cs->kvm_run->s.regs.pfs;
 522         env->pfault_compare = cs->kvm_run->s.regs.pfc;
 523     } else if (cap_async_pf) {
 524         r = kvm_get_one_reg(cs, KVM_REG_S390_PFTOKEN, &env->pfault_token);
 525         if (r < 0) {
 526             return r;
 527         }
 528         r = kvm_get_one_reg(cs, KVM_REG_S390_PFCOMPARE, &env->pfault_compare);
 529         if (r < 0) {
 530             return r;
 531         }
 532         r = kvm_get_one_reg(cs, KVM_REG_S390_PFSELECT, &env->pfault_select);
 533         if (r < 0) {
 534             return r;
 535         }
 536     }
 537
 538     return 0;
 539 }
 540
 541 int kvm_s390_get_clock(uint8_t *tod_high, uint64_t *tod_low)
 542 {
 543     int r;
 544     struct kvm_device_attr attr = {
 545         .group = KVM_S390_VM_TOD,
 546         .attr = KVM_S390_VM_TOD_LOW,
 547         .addr = (uint64_t)tod_low,
 548     };
 549
 550     r = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
 551     if (r) {
 552         return r;
 553     }
 554
 555     attr.attr = KVM_S390_VM_TOD_HIGH;
 556     attr.addr = (uint64_t)tod_high;
 557     return kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
 558 }
 559
 560 int kvm_s390_set_clock(uint8_t *tod_high, uint64_t *tod_low)
 561 {
 562     int r;
 563
 564     struct kvm_device_attr attr = {
 565         .group = KVM_S390_VM_TOD,
 566         .attr = KVM_S390_VM_TOD_LOW,
 567         .addr = (uint64_t)tod_low,
 568     };
 569
 570     r = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
 571     if (r) {
 572         return r;
 573     }
 574
 575     attr.attr = KVM_S390_VM_TOD_HIGH;
 576     attr.addr = (uint64_t)tod_high;
 577     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
 578 }
 579
 580 /**
 581  * kvm_s390_mem_op:
 582  * @addr:      the logical start address in guest memory
 583  * @ar:        the access register number
 584  * @hostbuf:   buffer in host memory. NULL = do only checks w/o copying
 585  * @len:       length that should be transferred
 586  * @is_write:  true = write, false = read
 587  * Returns:    0 on success, non-zero if an exception or error occurred
 588  *
 589  * Use KVM ioctl to read/write from/to guest memory. An access exception
 590  * is injected into the vCPU in case of translation errors.
 591  */
 592 int kvm_s390_mem_op(S390CPU *cpu, vaddr addr, uint8_t ar, void *hostbuf,
 593                     int len, bool is_write)
 594 {
 595     struct kvm_s390_mem_op mem_op = {
 596         .gaddr = addr,
 597         .flags = KVM_S390_MEMOP_F_INJECT_EXCEPTION,
 598         .size = len,
 599         .op = is_write ? KVM_S390_MEMOP_LOGICAL_WRITE
 600                        : KVM_S390_MEMOP_LOGICAL_READ,
 601         .buf = (uint64_t)hostbuf,
 602         .ar = ar,
 603     };
 604     int ret;
 605
 606     if (!cap_mem_op) {
 607         return -ENOSYS;
 608     }
 609     if (!hostbuf) {
 610         mem_op.flags |= KVM_S390_MEMOP_F_CHECK_ONLY;
 611     }
 612
 613     ret = kvm_vcpu_ioctl(CPU(cpu), KVM_S390_MEM_OP, &mem_op);
 614     if (ret < 0) {
 615         error_printf("KVM_S390_MEM_OP failed: %s\n", strerror(-ret));
 616     }
 617     return ret;
 618 }
 619
 620 /*
 621  * Legacy layout for s390:
 622  * Older S390 KVM requires the topmost vma of the RAM to be
 623  * smaller than an system defined value, which is at least 256GB.
 624  * Larger systems have larger values. We put the guest between
 625  * the end of data segment (system break) and this value. We
 626  * use 32GB as a base to have enough room for the system break
 627  * to grow. We also have to use MAP parameters that avoid
 628  * read-only mapping of guest pages.
 629  */
 630 static void *legacy_s390_alloc(size_t size, uint64_t *align)
 631 {
 632     void *mem;
 633
 634     mem = mmap((void *) 0x800000000ULL, size,
 635                PROT_EXEC|PROT_READ|PROT_WRITE,
 636                MAP_SHARED | MAP_ANONYMOUS | MAP_FIXED, -1, 0);
 637     return mem == MAP_FAILED ? NULL : mem;
 638 }
 639
 640 /* DIAG 501 is used for sw breakpoints */
 641 static const uint8_t diag_501[] = {0x83, 0x24, 0x05, 0x01};
 642
 643 int kvm_arch_insert_sw_breakpoint(CPUState *cs, struct kvm_sw_breakpoint *bp)
 644 {
 645
 646     if (cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)&bp->saved_insn,
 647                             sizeof(diag_501), 0) ||
 648         cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)diag_501,
 649                             sizeof(diag_501), 1)) {
 650         return -EINVAL;
 651     }
 652     return 0;
 653 }
 654
 655 int kvm_arch_remove_sw_breakpoint(CPUState *cs, struct kvm_sw_breakpoint *bp)
 656 {
 657     uint8_t t[sizeof(diag_501)];
 658
 659     if (cpu_memory_rw_debug(cs, bp->pc, t, sizeof(diag_501), 0)) {
 660         return -EINVAL;
 661     } else if (memcmp(t, diag_501, sizeof(diag_501))) {
 662         return -EINVAL;
 663     } else if (cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)&bp->saved_insn,
 664                                    sizeof(diag_501), 1)) {
 665         return -EINVAL;
 666     }
 667
 668     return 0;
 669 }
 670
 671 static struct kvm_hw_breakpoint *find_hw_breakpoint(target_ulong addr,
 672                                                     int len, int type)
 673 {
 674     int n;
 675
 676     for (n = 0; n < nb_hw_breakpoints; n++) {
 677         if (hw_breakpoints[n].addr == addr && hw_breakpoints[n].type == type &&
 678             (hw_breakpoints[n].len == len || len == -1)) {
 679             return &hw_breakpoints[n];
 680         }
 681     }
 682
 683     return NULL;
 684 }
 685
 686 static int insert_hw_breakpoint(target_ulong addr, int len, int type)
 687 {
 688     int size;
 689
 690     if (find_hw_breakpoint(addr, len, type)) {
 691         return -EEXIST;
 692     }
 693
 694     size = (nb_hw_breakpoints + 1) * sizeof(struct kvm_hw_breakpoint);
 695
 696     if (!hw_breakpoints) {
 697         nb_hw_breakpoints = 0;
 698         hw_breakpoints = (struct kvm_hw_breakpoint *)g_try_malloc(size);
 699     } else {
 700         hw_breakpoints =
 701             (struct kvm_hw_breakpoint *)g_try_realloc(hw_breakpoints, size);
 702     }
 703
 704     if (!hw_breakpoints) {
 705         nb_hw_breakpoints = 0;
 706         return -ENOMEM;
 707     }
 708
 709     hw_breakpoints[nb_hw_breakpoints].addr = addr;
 710     hw_breakpoints[nb_hw_breakpoints].len = len;
 711     hw_breakpoints[nb_hw_breakpoints].type = type;
 712
 713     nb_hw_breakpoints++;
 714
 715     return 0;
 716 }
 717
 718 int kvm_arch_insert_hw_breakpoint(target_ulong addr,
 719                                   target_ulong len, int type)
 720 {
 721     switch (type) {
 722     case GDB_BREAKPOINT_HW:
 723         type = KVM_HW_BP;
 724         break;
 725     case GDB_WATCHPOINT_WRITE:
 726         if (len < 1) {
 727             return -EINVAL;
 728         }
 729         type = KVM_HW_WP_WRITE;
 730         break;
 731     default:
 732         return -ENOSYS;
 733     }
 734     return insert_hw_breakpoint(addr, len, type);
 735 }
 736
 737 int kvm_arch_remove_hw_breakpoint(target_ulong addr,
 738                                   target_ulong len, int type)
 739 {
 740     int size;
 741     struct kvm_hw_breakpoint *bp = find_hw_breakpoint(addr, len, type);
 742
 743     if (bp == NULL) {
 744         return -ENOENT;
 745     }
 746
 747     nb_hw_breakpoints--;
 748     if (nb_hw_breakpoints > 0) {
 749         /*
 750          * In order to trim the array, move the last element to the position to
 751          * be removed - if necessary.
 752          */
 753         if (bp != &hw_breakpoints[nb_hw_breakpoints]) {
 754             *bp = hw_breakpoints[nb_hw_breakpoints];
 755         }
 756         size = nb_hw_breakpoints * sizeof(struct kvm_hw_breakpoint);
 757         hw_breakpoints =
 758              (struct kvm_hw_breakpoint *)g_realloc(hw_breakpoints, size);
 759     } else {
 760         g_free(hw_breakpoints);
 761         hw_breakpoints = NULL;
 762     }
 763
 764     return 0;
 765 }
 766
 767 void kvm_arch_remove_all_hw_breakpoints(void)
 768 {
 769     nb_hw_breakpoints = 0;
 770     g_free(hw_breakpoints);
 771     hw_breakpoints = NULL;
 772 }
 773
 774 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg)
 775 {
 776     int i;
 777
 778     if (nb_hw_breakpoints > 0) {
 779         dbg->arch.nr_hw_bp = nb_hw_breakpoints;
 780         dbg->arch.hw_bp = hw_breakpoints;
 781
 782         for (i = 0; i < nb_hw_breakpoints; ++i) {
 783             hw_breakpoints[i].phys_addr = s390_cpu_get_phys_addr_debug(cpu,
 784                                                        hw_breakpoints[i].addr);
 785         }
 786         dbg->control |= KVM_GUESTDBG_ENABLE | KVM_GUESTDBG_USE_HW_BP;
 787     } else {
 788         dbg->arch.nr_hw_bp = 0;
 789         dbg->arch.hw_bp = NULL;
 790     }
 791 }
 792
 793 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run)
 794 {
 795 }
 796
 797 MemTxAttrs kvm_arch_post_run(CPUState *cs, struct kvm_run *run)
 798 {
 799     return MEMTXATTRS_UNSPECIFIED;
 800 }
 801
 802 int kvm_arch_process_async_events(CPUState *cs)
 803 {
 804     return cs->halted;
 805 }
 806
 807 static int s390_kvm_irq_to_interrupt(struct kvm_s390_irq *irq,
 808                                      struct kvm_s390_interrupt *interrupt)
 809 {
 810     int r = 0;
 811
 812     interrupt->type = irq->type;
 813     switch (irq->type) {
 814     case KVM_S390_INT_VIRTIO:
 815         interrupt->parm = irq->u.ext.ext_params;
 816         /* fall through */
 817     case KVM_S390_INT_PFAULT_INIT:
 818     case KVM_S390_INT_PFAULT_DONE:
 819         interrupt->parm64 = irq->u.ext.ext_params2;
 820         break;
 821     case KVM_S390_PROGRAM_INT:
 822         interrupt->parm = irq->u.pgm.code;
 823         break;
 824     case KVM_S390_SIGP_SET_PREFIX:
 825         interrupt->parm = irq->u.prefix.address;
 826         break;
 827     case KVM_S390_INT_SERVICE:
 828         interrupt->parm = irq->u.ext.ext_params;
 829         break;
 830     case KVM_S390_MCHK:
 831         interrupt->parm = irq->u.mchk.cr14;
 832         interrupt->parm64 = irq->u.mchk.mcic;
 833         break;
 834     case KVM_S390_INT_EXTERNAL_CALL:
 835         interrupt->parm = irq->u.extcall.code;
 836         break;
 837     case KVM_S390_INT_EMERGENCY:
 838         interrupt->parm = irq->u.emerg.code;
 839         break;
 840     case KVM_S390_SIGP_STOP:
 841     case KVM_S390_RESTART:
 842         break; /* These types have no parameters */
 843     case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
 844         interrupt->parm = irq->u.io.subchannel_id << 16;
 845         interrupt->parm |= irq->u.io.subchannel_nr;
 846         interrupt->parm64 = (uint64_t)irq->u.io.io_int_parm << 32;
 847         interrupt->parm64 |= irq->u.io.io_int_word;
 848         break;
 849     default:
 850         r = -EINVAL;
 851         break;
 852     }
 853     return r;
 854 }
 855
 856 static void inject_vcpu_irq_legacy(CPUState *cs, struct kvm_s390_irq *irq)
 857 {
 858     struct kvm_s390_interrupt kvmint = {};
 859     int r;
 860
 861     r = s390_kvm_irq_to_interrupt(irq, &kvmint);
 862     if (r < 0) {
 863         fprintf(stderr, "%s called with bogus interrupt\n", __func__);
 864         exit(1);
 865     }
 866
 867     r = kvm_vcpu_ioctl(cs, KVM_S390_INTERRUPT, &kvmint);
 868     if (r < 0) {
 869         fprintf(stderr, "KVM failed to inject interrupt\n");
 870         exit(1);
 871     }
 872 }
 873
 874 void kvm_s390_vcpu_interrupt(S390CPU *cpu, struct kvm_s390_irq *irq)
 875 {
 876     CPUState *cs = CPU(cpu);
 877     int r;
 878
 879     if (cap_s390_irq) {
 880         r = kvm_vcpu_ioctl(cs, KVM_S390_IRQ, irq);
 881         if (!r) {
 882             return;
 883         }
 884         error_report("KVM failed to inject interrupt %llx", irq->type);
 885         exit(1);
 886     }
 887
 888     inject_vcpu_irq_legacy(cs, irq);
 889 }
 890
 891 static void __kvm_s390_floating_interrupt(struct kvm_s390_irq *irq)
 892 {
 893     struct kvm_s390_interrupt kvmint = {};
 894     int r;
 895
 896     r = s390_kvm_irq_to_interrupt(irq, &kvmint);
 897     if (r < 0) {
 898         fprintf(stderr, "%s called with bogus interrupt\n", __func__);
 899         exit(1);
 900     }
 901
 902     r = kvm_vm_ioctl(kvm_state, KVM_S390_INTERRUPT, &kvmint);
 903     if (r < 0) {
 904         fprintf(stderr, "KVM failed to inject interrupt\n");
 905         exit(1);
 906     }
 907 }
 908
 909 void kvm_s390_floating_interrupt(struct kvm_s390_irq *irq)
 910 {
 911     static bool use_flic = true;
 912     int r;
 913
 914     if (use_flic) {
 915         r = kvm_s390_inject_flic(irq);
 916         if (r == -ENOSYS) {
 917             use_flic = false;
 918         }
 919         if (!r) {
 920             return;
 921         }
 922     }
 923     __kvm_s390_floating_interrupt(irq);
 924 }
 925
 926 void kvm_s390_virtio_irq(int config_change, uint64_t token)
 927 {
 928     struct kvm_s390_irq irq = {
 929         .type = KVM_S390_INT_VIRTIO,
 930         .u.ext.ext_params = config_change,
 931         .u.ext.ext_params2 = token,
 932     };
 933
 934     kvm_s390_floating_interrupt(&irq);
 935 }
 936
 937 void kvm_s390_service_interrupt(uint32_t parm)
 938 {
 939     struct kvm_s390_irq irq = {
 940         .type = KVM_S390_INT_SERVICE,
 941         .u.ext.ext_params = parm,
 942     };
 943
 944     kvm_s390_floating_interrupt(&irq);
 945 }
 946
 947 static void enter_pgmcheck(S390CPU *cpu, uint16_t code)
 948 {
 949     struct kvm_s390_irq irq = {
 950         .type = KVM_S390_PROGRAM_INT,
 951         .u.pgm.code = code,
 952     };
 953
 954     kvm_s390_vcpu_interrupt(cpu, &irq);
 955 }
 956
 957 void kvm_s390_access_exception(S390CPU *cpu, uint16_t code, uint64_t te_code)
 958 {
 959     struct kvm_s390_irq irq = {
 960         .type = KVM_S390_PROGRAM_INT,
 961         .u.pgm.code = code,
 962         .u.pgm.trans_exc_code = te_code,
 963         .u.pgm.exc_access_id = te_code & 3,
 964     };
 965
 966     kvm_s390_vcpu_interrupt(cpu, &irq);
 967 }
 968
 969 static int kvm_sclp_service_call(S390CPU *cpu, struct kvm_run *run,
 970                                  uint16_t ipbh0)
 971 {
 972     CPUS390XState *env = &cpu->env;
 973     uint64_t sccb;
 974     uint32_t code;
 975     int r = 0;
 976
 977     cpu_synchronize_state(CPU(cpu));
 978     sccb = env->regs[ipbh0 & 0xf];
 979     code = env->regs[(ipbh0 & 0xf0) >> 4];
 980
 981     r = sclp_service_call(env, sccb, code);
 982     if (r < 0) {
 983         enter_pgmcheck(cpu, -r);
 984     } else {
 985         setcc(cpu, r);
 986     }
 987
 988     return 0;
 989 }
 990
 991 static int handle_b2(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
 992 {
 993     CPUS390XState *env = &cpu->env;
 994     int rc = 0;
 995     uint16_t ipbh0 = (run->s390_sieic.ipb & 0xffff0000) >> 16;
 996
 997     cpu_synchronize_state(CPU(cpu));
 998
 999     switch (ipa1) {
1000     case PRIV_B2_XSCH:
1001         ioinst_handle_xsch(cpu, env->regs[1]);
1002         break;
1003     case PRIV_B2_CSCH:
1004         ioinst_handle_csch(cpu, env->regs[1]);
1005         break;
1006     case PRIV_B2_HSCH:
1007         ioinst_handle_hsch(cpu, env->regs[1]);
1008         break;
1009     case PRIV_B2_MSCH:
1010         ioinst_handle_msch(cpu, env->regs[1], run->s390_sieic.ipb);
1011         break;
1012     case PRIV_B2_SSCH:
1013         ioinst_handle_ssch(cpu, env->regs[1], run->s390_sieic.ipb);
1014         break;
1015     case PRIV_B2_STCRW:
1016         ioinst_handle_stcrw(cpu, run->s390_sieic.ipb);
1017         break;
1018     case PRIV_B2_STSCH:
1019         ioinst_handle_stsch(cpu, env->regs[1], run->s390_sieic.ipb);
1020         break;
1021     case PRIV_B2_TSCH:
1022         /* We should only get tsch via KVM_EXIT_S390_TSCH. */
1023         fprintf(stderr, "Spurious tsch intercept\n");
1024         break;
1025     case PRIV_B2_CHSC:
1026         ioinst_handle_chsc(cpu, run->s390_sieic.ipb);
1027         break;
1028     case PRIV_B2_TPI:
1029         /* This should have been handled by kvm already. */
1030         fprintf(stderr, "Spurious tpi intercept\n");
1031         break;
1032     case PRIV_B2_SCHM:
1033         ioinst_handle_schm(cpu, env->regs[1], env->regs[2],
1034                            run->s390_sieic.ipb);
1035         break;
1036     case PRIV_B2_RSCH:
1037         ioinst_handle_rsch(cpu, env->regs[1]);
1038         break;
1039     case PRIV_B2_RCHP:
1040         ioinst_handle_rchp(cpu, env->regs[1]);
1041         break;
1042     case PRIV_B2_STCPS:
1043         /* We do not provide this instruction, it is suppressed. */
1044         break;
1045     case PRIV_B2_SAL:
1046         ioinst_handle_sal(cpu, env->regs[1]);
1047         break;
1048     case PRIV_B2_SIGA:
1049         /* Not provided, set CC = 3 for subchannel not operational */
1050         setcc(cpu, 3);
1051         break;
1052     case PRIV_B2_SCLP_CALL:
1053         rc = kvm_sclp_service_call(cpu, run, ipbh0);
1054         break;
1055     default:
1056         rc = -1;
1057         DPRINTF("KVM: unhandled PRIV: 0xb2%x\n", ipa1);
1058         break;
1059     }
1060
1061     return rc;
1062 }
1063
1064 static uint64_t get_base_disp_rxy(S390CPU *cpu, struct kvm_run *run,
1065                                   uint8_t *ar)
1066 {
1067     CPUS390XState *env = &cpu->env;
1068     uint32_t x2 = (run->s390_sieic.ipa & 0x000f);
1069     uint32_t base2 = run->s390_sieic.ipb >> 28;
1070     uint32_t disp2 = ((run->s390_sieic.ipb & 0x0fff0000) >> 16) +
1071                      ((run->s390_sieic.ipb & 0xff00) << 4);
1072
1073     if (disp2 & 0x80000) {
1074         disp2 += 0xfff00000;
1075     }
1076     if (ar) {
1077         *ar = base2;
1078     }
1079
1080     return (base2 ? env->regs[base2] : 0) +
1081            (x2 ? env->regs[x2] : 0) + (long)(int)disp2;
1082 }
1083
1084 static uint64_t get_base_disp_rsy(S390CPU *cpu, struct kvm_run *run,
1085                                   uint8_t *ar)
1086 {
1087     CPUS390XState *env = &cpu->env;
1088     uint32_t base2 = run->s390_sieic.ipb >> 28;
1089     uint32_t disp2 = ((run->s390_sieic.ipb & 0x0fff0000) >> 16) +
1090                      ((run->s390_sieic.ipb & 0xff00) << 4);
1091
1092     if (disp2 & 0x80000) {
1093         disp2 += 0xfff00000;
1094     }
1095     if (ar) {
1096         *ar = base2;
1097     }
1098
1099     return (base2 ? env->regs[base2] : 0) + (long)(int)disp2;
1100 }
1101
1102 static int kvm_clp_service_call(S390CPU *cpu, struct kvm_run *run)
1103 {
1104     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1105
1106     return clp_service_call(cpu, r2);
1107 }
1108
1109 static int kvm_pcilg_service_call(S390CPU *cpu, struct kvm_run *run)
1110 {
1111     uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
1112     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1113
1114     return pcilg_service_call(cpu, r1, r2);
1115 }
1116
1117 static int kvm_pcistg_service_call(S390CPU *cpu, struct kvm_run *run)
1118 {
1119     uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
1120     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1121
1122     return pcistg_service_call(cpu, r1, r2);
1123 }
1124
1125 static int kvm_stpcifc_service_call(S390CPU *cpu, struct kvm_run *run)
1126 {
1127     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1128     uint64_t fiba;
1129     uint8_t ar;
1130
1131     cpu_synchronize_state(CPU(cpu));
1132     fiba = get_base_disp_rxy(cpu, run, &ar);
1133
1134     return stpcifc_service_call(cpu, r1, fiba, ar);
1135 }
1136
1137 static int kvm_sic_service_call(S390CPU *cpu, struct kvm_run *run)
1138 {
1139     /* NOOP */
1140     return 0;
1141 }
1142
1143 static int kvm_rpcit_service_call(S390CPU *cpu, struct kvm_run *run)
1144 {
1145     uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
1146     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1147
1148     return rpcit_service_call(cpu, r1, r2);
1149 }
1150
1151 static int kvm_pcistb_service_call(S390CPU *cpu, struct kvm_run *run)
1152 {
1153     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1154     uint8_t r3 = run->s390_sieic.ipa & 0x000f;
1155     uint64_t gaddr;
1156     uint8_t ar;
1157
1158     cpu_synchronize_state(CPU(cpu));
1159     gaddr = get_base_disp_rsy(cpu, run, &ar);
1160
1161     return pcistb_service_call(cpu, r1, r3, gaddr, ar);
1162 }
1163
1164 static int kvm_mpcifc_service_call(S390CPU *cpu, struct kvm_run *run)
1165 {
1166     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1167     uint64_t fiba;
1168     uint8_t ar;
1169
1170     cpu_synchronize_state(CPU(cpu));
1171     fiba = get_base_disp_rxy(cpu, run, &ar);
1172
1173     return mpcifc_service_call(cpu, r1, fiba, ar);
1174 }
1175
1176 static int handle_b9(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
1177 {
1178     int r = 0;
1179
1180     switch (ipa1) {
1181     case PRIV_B9_CLP:
1182         r = kvm_clp_service_call(cpu, run);
1183         break;
1184     case PRIV_B9_PCISTG:
1185         r = kvm_pcistg_service_call(cpu, run);
1186         break;
1187     case PRIV_B9_PCILG:
1188         r = kvm_pcilg_service_call(cpu, run);
1189         break;
1190     case PRIV_B9_RPCIT:
1191         r = kvm_rpcit_service_call(cpu, run);
1192         break;
1193     case PRIV_B9_EQBS:
1194         /* just inject exception */
1195         r = -1;
1196         break;
1197     default:
1198         r = -1;
1199         DPRINTF("KVM: unhandled PRIV: 0xb9%x\n", ipa1);
1200         break;
1201     }
1202
1203     return r;
1204 }
1205
1206 static int handle_eb(S390CPU *cpu, struct kvm_run *run, uint8_t ipbl)
1207 {
1208     int r = 0;
1209
1210     switch (ipbl) {
1211     case PRIV_EB_PCISTB:
1212         r = kvm_pcistb_service_call(cpu, run);
1213         break;
1214     case PRIV_EB_SIC:
1215         r = kvm_sic_service_call(cpu, run);
1216         break;
1217     case PRIV_EB_SQBS:
1218         /* just inject exception */
1219         r = -1;
1220         break;
1221     default:
1222         r = -1;
1223         DPRINTF("KVM: unhandled PRIV: 0xeb%x\n", ipbl);
1224         break;
1225     }
1226
1227     return r;
1228 }
1229
1230 static int handle_e3(S390CPU *cpu, struct kvm_run *run, uint8_t ipbl)
1231 {
1232     int r = 0;
1233
1234     switch (ipbl) {
1235     case PRIV_E3_MPCIFC:
1236         r = kvm_mpcifc_service_call(cpu, run);
1237         break;
1238     case PRIV_E3_STPCIFC:
1239         r = kvm_stpcifc_service_call(cpu, run);
1240         break;
1241     default:
1242         r = -1;
1243         DPRINTF("KVM: unhandled PRIV: 0xe3%x\n", ipbl);
1244         break;
1245     }
1246
1247     return r;
1248 }
1249
1250 static int handle_hypercall(S390CPU *cpu, struct kvm_run *run)
1251 {
1252     CPUS390XState *env = &cpu->env;
1253     int ret;
1254
1255     cpu_synchronize_state(CPU(cpu));
1256     ret = s390_virtio_hypercall(env);
1257     if (ret == -EINVAL) {
1258         enter_pgmcheck(cpu, PGM_SPECIFICATION);
1259         return 0;
1260     }
1261
1262     return ret;
1263 }
1264
1265 static void kvm_handle_diag_288(S390CPU *cpu, struct kvm_run *run)
1266 {
1267     uint64_t r1, r3;
1268     int rc;
1269
1270     cpu_synchronize_state(CPU(cpu));
1271     r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1272     r3 = run->s390_sieic.ipa & 0x000f;
1273     rc = handle_diag_288(&cpu->env, r1, r3);
1274     if (rc) {
1275         enter_pgmcheck(cpu, PGM_SPECIFICATION);
1276     }
1277 }
1278
1279 static void kvm_handle_diag_308(S390CPU *cpu, struct kvm_run *run)
1280 {
1281     uint64_t r1, r3;
1282
1283     cpu_synchronize_state(CPU(cpu));
1284     r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1285     r3 = run->s390_sieic.ipa & 0x000f;
1286     handle_diag_308(&cpu->env, r1, r3);
1287 }
1288
1289 static int handle_sw_breakpoint(S390CPU *cpu, struct kvm_run *run)
1290 {
1291     CPUS390XState *env = &cpu->env;
1292     unsigned long pc;
1293
1294     cpu_synchronize_state(CPU(cpu));
1295
1296     pc = env->psw.addr - 4;
1297     if (kvm_find_sw_breakpoint(CPU(cpu), pc)) {
1298         env->psw.addr = pc;
1299         return EXCP_DEBUG;
1300     }
1301
1302     return -ENOENT;
1303 }
1304
1305 #define DIAG_KVM_CODE_MASK 0x000000000000ffff
1306
1307 static int handle_diag(S390CPU *cpu, struct kvm_run *run, uint32_t ipb)
1308 {
1309     int r = 0;
1310     uint16_t func_code;
1311
1312     /*
1313      * For any diagnose call we support, bits 48-63 of the resulting
1314      * address specify the function code; the remainder is ignored.
1315      */
1316     func_code = decode_basedisp_rs(&cpu->env, ipb, NULL) & DIAG_KVM_CODE_MASK;
1317     switch (func_code) {
1318     case DIAG_TIMEREVENT:
1319         kvm_handle_diag_288(cpu, run);
1320         break;
1321     case DIAG_IPL:
1322         kvm_handle_diag_308(cpu, run);
1323         break;
1324     case DIAG_KVM_HYPERCALL:
1325         r = handle_hypercall(cpu, run);
1326         break;
1327     case DIAG_KVM_BREAKPOINT:
1328         r = handle_sw_breakpoint(cpu, run);
1329         break;
1330     default:
1331         DPRINTF("KVM: unknown DIAG: 0x%x\n", func_code);
1332         enter_pgmcheck(cpu, PGM_SPECIFICATION);
1333         break;
1334     }
1335
1336     return r;
1337 }
1338
1339 typedef struct SigpInfo {
1340     S390CPU *cpu;
1341     uint64_t param;
1342     int cc;
1343     uint64_t *status_reg;
1344 } SigpInfo;
1345
1346 static void set_sigp_status(SigpInfo *si, uint64_t status)
1347 {
1348     *si->status_reg &= 0xffffffff00000000ULL;
1349     *si->status_reg |= status;
1350     si->cc = SIGP_CC_STATUS_STORED;
1351 }
1352
1353 static void sigp_start(void *arg)
1354 {
1355     SigpInfo *si = arg;
1356
1357     if (s390_cpu_get_state(si->cpu) != CPU_STATE_STOPPED) {
1358         si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1359         return;
1360     }
1361
1362     s390_cpu_set_state(CPU_STATE_OPERATING, si->cpu);
1363     si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1364 }
1365
1366 static void sigp_stop(void *arg)
1367 {
1368     SigpInfo *si = arg;
1369     struct kvm_s390_irq irq = {
1370         .type = KVM_S390_SIGP_STOP,
1371     };
1372
1373     if (s390_cpu_get_state(si->cpu) != CPU_STATE_OPERATING) {
1374         si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1375         return;
1376     }
1377
1378     /* disabled wait - sleeping in user space */
1379     if (CPU(si->cpu)->halted) {
1380         s390_cpu_set_state(CPU_STATE_STOPPED, si->cpu);
1381     } else {
1382         /* execute the stop function */
1383         si->cpu->env.sigp_order = SIGP_STOP;
1384         kvm_s390_vcpu_interrupt(si->cpu, &irq);
1385     }
1386     si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1387 }
1388
1389 #define ADTL_SAVE_AREA_SIZE 1024
1390 static int kvm_s390_store_adtl_status(S390CPU *cpu, hwaddr addr)
1391 {
1392     void *mem;
1393     hwaddr len = ADTL_SAVE_AREA_SIZE;
1394
1395     mem = cpu_physical_memory_map(addr, &len, 1);
1396     if (!mem) {
1397         return -EFAULT;
1398     }
1399     if (len != ADTL_SAVE_AREA_SIZE) {
1400         cpu_physical_memory_unmap(mem, len, 1, 0);
1401         return -EFAULT;
1402     }
1403
1404     memcpy(mem, &cpu->env.vregs, 512);
1405
1406     cpu_physical_memory_unmap(mem, len, 1, len);
1407
1408     return 0;
1409 }
1410
1411 #define KVM_S390_STORE_STATUS_DEF_ADDR offsetof(LowCore, floating_pt_save_area)
1412 #define SAVE_AREA_SIZE 512
1413 static int kvm_s390_store_status(S390CPU *cpu, hwaddr addr, bool store_arch)
1414 {
1415     static const uint8_t ar_id = 1;
1416     uint64_t ckc = cpu->env.ckc >> 8;
1417     void *mem;
1418     int i;
1419     hwaddr len = SAVE_AREA_SIZE;
1420
1421     mem = cpu_physical_memory_map(addr, &len, 1);
1422     if (!mem) {
1423         return -EFAULT;
1424     }
1425     if (len != SAVE_AREA_SIZE) {
1426         cpu_physical_memory_unmap(mem, len, 1, 0);
1427         return -EFAULT;
1428     }
1429
1430     if (store_arch) {
1431         cpu_physical_memory_write(offsetof(LowCore, ar_access_id), &ar_id, 1);
1432     }
1433     for (i = 0; i < 16; ++i) {
1434         *((uint64 *)mem + i) = get_freg(&cpu->env, i)->ll;
1435     }
1436     memcpy(mem + 128, &cpu->env.regs, 128);
1437     memcpy(mem + 256, &cpu->env.psw, 16);
1438     memcpy(mem + 280, &cpu->env.psa, 4);
1439     memcpy(mem + 284, &cpu->env.fpc, 4);
1440     memcpy(mem + 292, &cpu->env.todpr, 4);
1441     memcpy(mem + 296, &cpu->env.cputm, 8);
1442     memcpy(mem + 304, &ckc, 8);
1443     memcpy(mem + 320, &cpu->env.aregs, 64);
1444     memcpy(mem + 384, &cpu->env.cregs, 128);
1445
1446     cpu_physical_memory_unmap(mem, len, 1, len);
1447
1448     return 0;
1449 }
1450
1451 static void sigp_stop_and_store_status(void *arg)
1452 {
1453     SigpInfo *si = arg;
1454     struct kvm_s390_irq irq = {
1455         .type = KVM_S390_SIGP_STOP,
1456     };
1457
1458     /* disabled wait - sleeping in user space */
1459     if (s390_cpu_get_state(si->cpu) == CPU_STATE_OPERATING &&
1460         CPU(si->cpu)->halted) {
1461         s390_cpu_set_state(CPU_STATE_STOPPED, si->cpu);
1462     }
1463
1464     switch (s390_cpu_get_state(si->cpu)) {
1465     case CPU_STATE_OPERATING:
1466         si->cpu->env.sigp_order = SIGP_STOP_STORE_STATUS;
1467         kvm_s390_vcpu_interrupt(si->cpu, &irq);
1468         /* store will be performed when handling the stop intercept */
1469         break;
1470     case CPU_STATE_STOPPED:
1471         /* already stopped, just store the status */
1472         cpu_synchronize_state(CPU(si->cpu));
1473         kvm_s390_store_status(si->cpu, KVM_S390_STORE_STATUS_DEF_ADDR, true);
1474         break;
1475     }
1476     si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1477 }
1478
1479 static void sigp_store_status_at_address(void *arg)
1480 {
1481     SigpInfo *si = arg;
1482     uint32_t address = si->param & 0x7ffffe00u;
1483
1484     /* cpu has to be stopped */
1485     if (s390_cpu_get_state(si->cpu) != CPU_STATE_STOPPED) {
1486         set_sigp_status(si, SIGP_STAT_INCORRECT_STATE);
1487         return;
1488     }
1489
1490     cpu_synchronize_state(CPU(si->cpu));
1491
1492     if (kvm_s390_store_status(si->cpu, address, false)) {
1493         set_sigp_status(si, SIGP_STAT_INVALID_PARAMETER);
1494         return;
1495     }
1496     si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1497 }
1498
1499 static void sigp_store_adtl_status(void *arg)
1500 {
1501     SigpInfo *si = arg;
1502
1503     if (!kvm_check_extension(kvm_state, KVM_CAP_S390_VECTOR_REGISTERS)) {
1504         set_sigp_status(si, SIGP_STAT_INVALID_ORDER);
1505         return;
1506     }
1507
1508     /* cpu has to be stopped */
1509     if (s390_cpu_get_state(si->cpu) != CPU_STATE_STOPPED) {
1510         set_sigp_status(si, SIGP_STAT_INCORRECT_STATE);
1511         return;
1512     }
1513
1514     /* parameter must be aligned to 1024-byte boundary */
1515     if (si->param & 0x3ff) {
1516         set_sigp_status(si, SIGP_STAT_INVALID_PARAMETER);
1517         return;
1518     }
1519
1520     cpu_synchronize_state(CPU(si->cpu));
1521
1522     if (kvm_s390_store_adtl_status(si->cpu, si->param)) {
1523         set_sigp_status(si, SIGP_STAT_INVALID_PARAMETER);
1524         return;
1525     }
1526     si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1527 }
1528
1529 static void sigp_restart(void *arg)
1530 {
1531     SigpInfo *si = arg;
1532     struct kvm_s390_irq irq = {
1533         .type = KVM_S390_RESTART,
1534     };
1535
1536     switch (s390_cpu_get_state(si->cpu)) {
1537     case CPU_STATE_STOPPED:
1538         /* the restart irq has to be delivered prior to any other pending irq */
1539         cpu_synchronize_state(CPU(si->cpu));
1540         do_restart_interrupt(&si->cpu->env);
1541         s390_cpu_set_state(CPU_STATE_OPERATING, si->cpu);
1542         break;
1543     case CPU_STATE_OPERATING:
1544         kvm_s390_vcpu_interrupt(si->cpu, &irq);
1545         break;
1546     }
1547     si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1548 }
1549
1550 int kvm_s390_cpu_restart(S390CPU *cpu)
1551 {
1552     SigpInfo si = {
1553         .cpu = cpu,
1554     };
1555
1556     run_on_cpu(CPU(cpu), sigp_restart, &si);
1557     DPRINTF("DONE: KVM cpu restart: %p\n", &cpu->env);
1558     return 0;
1559 }
1560
1561 static void sigp_initial_cpu_reset(void *arg)
1562 {
1563     SigpInfo *si = arg;
1564     CPUState *cs = CPU(si->cpu);
1565     S390CPUClass *scc = S390_CPU_GET_CLASS(si->cpu);
1566
1567     cpu_synchronize_state(cs);
1568     scc->initial_cpu_reset(cs);
1569     cpu_synchronize_post_reset(cs);
1570     si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1571 }
1572
1573 static void sigp_cpu_reset(void *arg)
1574 {
1575     SigpInfo *si = arg;
1576     CPUState *cs = CPU(si->cpu);
1577     S390CPUClass *scc = S390_CPU_GET_CLASS(si->cpu);
1578
1579     cpu_synchronize_state(cs);
1580     scc->cpu_reset(cs);
1581     cpu_synchronize_post_reset(cs);
1582     si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1583 }
1584
1585 static void sigp_set_prefix(void *arg)
1586 {
1587     SigpInfo *si = arg;
1588     uint32_t addr = si->param & 0x7fffe000u;
1589
1590     cpu_synchronize_state(CPU(si->cpu));
1591
1592     if (!address_space_access_valid(&address_space_memory, addr,
1593                                     sizeof(struct LowCore), false)) {
1594         set_sigp_status(si, SIGP_STAT_INVALID_PARAMETER);
1595         return;
1596     }
1597
1598     /* cpu has to be stopped */
1599     if (s390_cpu_get_state(si->cpu) != CPU_STATE_STOPPED) {
1600         set_sigp_status(si, SIGP_STAT_INCORRECT_STATE);
1601         return;
1602     }
1603
1604     si->cpu->env.psa = addr;
1605     cpu_synchronize_post_init(CPU(si->cpu));
1606     si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1607 }
1608
1609 static int handle_sigp_single_dst(S390CPU *dst_cpu, uint8_t order,
1610                                   uint64_t param, uint64_t *status_reg)
1611 {
1612     SigpInfo si = {
1613         .cpu = dst_cpu,
1614         .param = param,
1615         .status_reg = status_reg,
1616     };
1617
1618     /* cpu available? */
1619     if (dst_cpu == NULL) {
1620         return SIGP_CC_NOT_OPERATIONAL;
1621     }
1622
1623     /* only resets can break pending orders */
1624     if (dst_cpu->env.sigp_order != 0 &&
1625         order != SIGP_CPU_RESET &&
1626         order != SIGP_INITIAL_CPU_RESET) {
1627         return SIGP_CC_BUSY;
1628     }
1629
1630     switch (order) {
1631     case SIGP_START:
1632         run_on_cpu(CPU(dst_cpu), sigp_start, &si);
1633         break;
1634     case SIGP_STOP:
1635         run_on_cpu(CPU(dst_cpu), sigp_stop, &si);
1636         break;
1637     case SIGP_RESTART:
1638         run_on_cpu(CPU(dst_cpu), sigp_restart, &si);
1639         break;
1640     case SIGP_STOP_STORE_STATUS:
1641         run_on_cpu(CPU(dst_cpu), sigp_stop_and_store_status, &si);
1642         break;
1643     case SIGP_STORE_STATUS_ADDR:
1644         run_on_cpu(CPU(dst_cpu), sigp_store_status_at_address, &si);
1645         break;
1646     case SIGP_STORE_ADTL_STATUS:
1647         run_on_cpu(CPU(dst_cpu), sigp_store_adtl_status, &si);
1648         break;
1649     case SIGP_SET_PREFIX:
1650         run_on_cpu(CPU(dst_cpu), sigp_set_prefix, &si);
1651         break;
1652     case SIGP_INITIAL_CPU_RESET:
1653         run_on_cpu(CPU(dst_cpu), sigp_initial_cpu_reset, &si);
1654         break;
1655     case SIGP_CPU_RESET:
1656         run_on_cpu(CPU(dst_cpu), sigp_cpu_reset, &si);
1657         break;
1658     default:
1659         DPRINTF("KVM: unknown SIGP: 0x%x\n", order);
1660         set_sigp_status(&si, SIGP_STAT_INVALID_ORDER);
1661     }
1662
1663     return si.cc;
1664 }
1665
1666 static int sigp_set_architecture(S390CPU *cpu, uint32_t param,
1667                                  uint64_t *status_reg)
1668 {
1669     CPUState *cur_cs;
1670     S390CPU *cur_cpu;
1671
1672     /* due to the BQL, we are the only active cpu */
1673     CPU_FOREACH(cur_cs) {
1674         cur_cpu = S390_CPU(cur_cs);
1675         if (cur_cpu->env.sigp_order != 0) {
1676             return SIGP_CC_BUSY;
1677         }
1678         cpu_synchronize_state(cur_cs);
1679         /* all but the current one have to be stopped */
1680         if (cur_cpu != cpu &&
1681             s390_cpu_get_state(cur_cpu) != CPU_STATE_STOPPED) {
1682             *status_reg &= 0xffffffff00000000ULL;
1683             *status_reg |= SIGP_STAT_INCORRECT_STATE;
1684             return SIGP_CC_STATUS_STORED;
1685         }
1686     }
1687
1688     switch (param & 0xff) {
1689     case SIGP_MODE_ESA_S390:
1690         /* not supported */
1691         return SIGP_CC_NOT_OPERATIONAL;
1692     case SIGP_MODE_Z_ARCH_TRANS_ALL_PSW:
1693     case SIGP_MODE_Z_ARCH_TRANS_CUR_PSW:
1694         CPU_FOREACH(cur_cs) {
1695             cur_cpu = S390_CPU(cur_cs);
1696             cur_cpu->env.pfault_token = -1UL;
1697         }
1698         break;
1699     default:
1700         *status_reg &= 0xffffffff00000000ULL;
1701         *status_reg |= SIGP_STAT_INVALID_PARAMETER;
1702         return SIGP_CC_STATUS_STORED;
1703     }
1704
1705     return SIGP_CC_ORDER_CODE_ACCEPTED;
1706 }
1707
1708 #define SIGP_ORDER_MASK 0x000000ff
1709
1710 static int handle_sigp(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
1711 {
1712     CPUS390XState *env = &cpu->env;
1713     const uint8_t r1 = ipa1 >> 4;
1714     const uint8_t r3 = ipa1 & 0x0f;
1715     int ret;
1716     uint8_t order;
1717     uint64_t *status_reg;
1718     uint64_t param;
1719     S390CPU *dst_cpu = NULL;
1720
1721     cpu_synchronize_state(CPU(cpu));
1722
1723     /* get order code */
1724     order = decode_basedisp_rs(env, run->s390_sieic.ipb, NULL)
1725         & SIGP_ORDER_MASK;
1726     status_reg = &env->regs[r1];
1727     param = (r1 % 2) ? env->regs[r1] : env->regs[r1 + 1];
1728
1729     switch (order) {
1730     case SIGP_SET_ARCH:
1731         ret = sigp_set_architecture(cpu, param, status_reg);
1732         break;
1733     default:
1734         /* all other sigp orders target a single vcpu */
1735         dst_cpu = s390_cpu_addr2state(env->regs[r3]);
1736         ret = handle_sigp_single_dst(dst_cpu, order, param, status_reg);
1737     }
1738
1739     trace_kvm_sigp_finished(order, CPU(cpu)->cpu_index,
1740                             dst_cpu ? CPU(dst_cpu)->cpu_index : -1, ret);
1741
1742     if (ret >= 0) {
1743         setcc(cpu, ret);
1744         return 0;
1745     }
1746
1747     return ret;
1748 }
1749
1750 static int handle_instruction(S390CPU *cpu, struct kvm_run *run)
1751 {
1752     unsigned int ipa0 = (run->s390_sieic.ipa & 0xff00);
1753     uint8_t ipa1 = run->s390_sieic.ipa & 0x00ff;
1754     int r = -1;
1755
1756     DPRINTF("handle_instruction 0x%x 0x%x\n",
1757             run->s390_sieic.ipa, run->s390_sieic.ipb);
1758     switch (ipa0) {
1759     case IPA0_B2:
1760         r = handle_b2(cpu, run, ipa1);
1761         break;
1762     case IPA0_B9:
1763         r = handle_b9(cpu, run, ipa1);
1764         break;
1765     case IPA0_EB:
1766         r = handle_eb(cpu, run, run->s390_sieic.ipb & 0xff);
1767         break;
1768     case IPA0_E3:
1769         r = handle_e3(cpu, run, run->s390_sieic.ipb & 0xff);
1770         break;
1771     case IPA0_DIAG:
1772         r = handle_diag(cpu, run, run->s390_sieic.ipb);
1773         break;
1774     case IPA0_SIGP:
1775         r = handle_sigp(cpu, run, ipa1);
1776         break;
1777     }
1778
1779     if (r < 0) {
1780         r = 0;
1781         enter_pgmcheck(cpu, 0x0001);
1782     }
1783
1784     return r;
1785 }
1786
1787 static bool is_special_wait_psw(CPUState *cs)
1788 {
1789     /* signal quiesce */
1790     return cs->kvm_run->psw_addr == 0xfffUL;
1791 }
1792
1793 static void unmanageable_intercept(S390CPU *cpu, const char *str, int pswoffset)
1794 {
1795     CPUState *cs = CPU(cpu);
1796
1797     error_report("Unmanageable %s! CPU%i new PSW: 0x%016lx:%016lx",
1798                  str, cs->cpu_index, ldq_phys(cs->as, cpu->env.psa + pswoffset),
1799                  ldq_phys(cs->as, cpu->env.psa + pswoffset + 8));
1800     s390_cpu_halt(cpu);
1801     qemu_system_guest_panicked();
1802 }
1803
1804 static int handle_intercept(S390CPU *cpu)
1805 {
1806     CPUState *cs = CPU(cpu);
1807     struct kvm_run *run = cs->kvm_run;
1808     int icpt_code = run->s390_sieic.icptcode;
1809     int r = 0;
1810
1811     DPRINTF("intercept: 0x%x (at 0x%lx)\n", icpt_code,
1812             (long)cs->kvm_run->psw_addr);
1813     switch (icpt_code) {
1814         case ICPT_INSTRUCTION:
1815             r = handle_instruction(cpu, run);
1816             break;
1817         case ICPT_PROGRAM:
1818             unmanageable_intercept(cpu, "program interrupt",
1819                                    offsetof(LowCore, program_new_psw));
1820             r = EXCP_HALTED;
1821             break;
1822         case ICPT_EXT_INT:
1823             unmanageable_intercept(cpu, "external interrupt",
1824                                    offsetof(LowCore, external_new_psw));
1825             r = EXCP_HALTED;
1826             break;
1827         case ICPT_WAITPSW:
1828             /* disabled wait, since enabled wait is handled in kernel */
1829             cpu_synchronize_state(cs);
1830             if (s390_cpu_halt(cpu) == 0) {
1831                 if (is_special_wait_psw(cs)) {
1832                     qemu_system_shutdown_request();
1833                 } else {
1834                     qemu_system_guest_panicked();
1835                 }
1836             }
1837             r = EXCP_HALTED;
1838             break;
1839         case ICPT_CPU_STOP:
1840             if (s390_cpu_set_state(CPU_STATE_STOPPED, cpu) == 0) {
1841                 qemu_system_shutdown_request();
1842             }
1843             if (cpu->env.sigp_order == SIGP_STOP_STORE_STATUS) {
1844                 kvm_s390_store_status(cpu, KVM_S390_STORE_STATUS_DEF_ADDR,
1845                                       true);
1846             }
1847             cpu->env.sigp_order = 0;
1848             r = EXCP_HALTED;
1849             break;
1850         case ICPT_SOFT_INTERCEPT:
1851             fprintf(stderr, "KVM unimplemented icpt SOFT\n");
1852             exit(1);
1853             break;
1854         case ICPT_IO:
1855             fprintf(stderr, "KVM unimplemented icpt IO\n");
1856             exit(1);
1857             break;
1858         default:
1859             fprintf(stderr, "Unknown intercept code: %d\n", icpt_code);
1860             exit(1);
1861             break;
1862     }
1863
1864     return r;
1865 }
1866
1867 static int handle_tsch(S390CPU *cpu)
1868 {
1869     CPUState *cs = CPU(cpu);
1870     struct kvm_run *run = cs->kvm_run;
1871     int ret;
1872
1873     cpu_synchronize_state(cs);
1874
1875     ret = ioinst_handle_tsch(cpu, cpu->env.regs[1], run->s390_tsch.ipb);
1876     if (ret < 0) {
1877         /*
1878          * Failure.
1879          * If an I/O interrupt had been dequeued, we have to reinject it.
1880          */
1881         if (run->s390_tsch.dequeued) {
1882             kvm_s390_io_interrupt(run->s390_tsch.subchannel_id,
1883                                   run->s390_tsch.subchannel_nr,
1884                                   run->s390_tsch.io_int_parm,
1885                                   run->s390_tsch.io_int_word);
1886         }
1887         ret = 0;
1888     }
1889     return ret;
1890 }
1891
1892 static void insert_stsi_3_2_2(S390CPU *cpu, __u64 addr, uint8_t ar)
1893 {
1894     struct sysib_322 sysib;
1895     int del;
1896
1897     if (s390_cpu_virt_mem_read(cpu, addr, ar, &sysib, sizeof(sysib))) {
1898         return;
1899     }
1900     /* Shift the stack of Extended Names to prepare for our own data */
1901     memmove(&sysib.ext_names[1], &sysib.ext_names[0],
1902             sizeof(sysib.ext_names[0]) * (sysib.count - 1));
1903     /* First virt level, that doesn't provide Ext Names delimits stack. It is
1904      * assumed it's not capable of managing Extended Names for lower levels.
1905      */
1906     for (del = 1; del < sysib.count; del++) {
1907         if (!sysib.vm[del].ext_name_encoding || !sysib.ext_names[del][0]) {
1908             break;
1909         }
1910     }
1911     if (del < sysib.count) {
1912         memset(sysib.ext_names[del], 0,
1913                sizeof(sysib.ext_names[0]) * (sysib.count - del));
1914     }
1915     /* Insert short machine name in EBCDIC, padded with blanks */
1916     if (qemu_name) {
1917         memset(sysib.vm[0].name, 0x40, sizeof(sysib.vm[0].name));
1918         ebcdic_put(sysib.vm[0].name, qemu_name, MIN(sizeof(sysib.vm[0].name),
1919                                                     strlen(qemu_name)));
1920     }
1921     sysib.vm[0].ext_name_encoding = 2; /* 2 = UTF-8 */
1922     memset(sysib.ext_names[0], 0, sizeof(sysib.ext_names[0]));
1923     /* If hypervisor specifies zero Extended Name in STSI322 SYSIB, it's
1924      * considered by s390 as not capable of providing any Extended Name.
1925      * Therefore if no name was specified on qemu invocation, we go with the
1926      * same "KVMguest" default, which KVM has filled into short name field.
1927      */
1928     if (qemu_name) {
1929         strncpy((char *)sysib.ext_names[0], qemu_name,
1930                 sizeof(sysib.ext_names[0]));
1931     } else {
1932         strcpy((char *)sysib.ext_names[0], "KVMguest");
1933     }
1934     /* Insert UUID */
1935     memcpy(sysib.vm[0].uuid, qemu_uuid, sizeof(sysib.vm[0].uuid));
1936
1937     s390_cpu_virt_mem_write(cpu, addr, ar, &sysib, sizeof(sysib));
1938 }
1939
1940 static int handle_stsi(S390CPU *cpu)
1941 {
1942     CPUState *cs = CPU(cpu);
1943     struct kvm_run *run = cs->kvm_run;
1944
1945     switch (run->s390_stsi.fc) {
1946     case 3:
1947         if (run->s390_stsi.sel1 != 2 || run->s390_stsi.sel2 != 2) {
1948             return 0;
1949         }
1950         /* Only sysib 3.2.2 needs post-handling for now. */
1951         insert_stsi_3_2_2(cpu, run->s390_stsi.addr, run->s390_stsi.ar);
1952         return 0;
1953     default:
1954         return 0;
1955     }
1956 }
1957
1958 static int kvm_arch_handle_debug_exit(S390CPU *cpu)
1959 {
1960     CPUState *cs = CPU(cpu);
1961     struct kvm_run *run = cs->kvm_run;
1962
1963     int ret = 0;
1964     struct kvm_debug_exit_arch *arch_info = &run->debug.arch;
1965
1966     switch (arch_info->type) {
1967     case KVM_HW_WP_WRITE:
1968         if (find_hw_breakpoint(arch_info->addr, -1, arch_info->type)) {
1969             cs->watchpoint_hit = &hw_watchpoint;
1970             hw_watchpoint.vaddr = arch_info->addr;
1971             hw_watchpoint.flags = BP_MEM_WRITE;
1972             ret = EXCP_DEBUG;
1973         }
1974         break;
1975     case KVM_HW_BP:
1976         if (find_hw_breakpoint(arch_info->addr, -1, arch_info->type)) {
1977             ret = EXCP_DEBUG;
1978         }
1979         break;
1980     case KVM_SINGLESTEP:
1981         if (cs->singlestep_enabled) {
1982             ret = EXCP_DEBUG;
1983         }
1984         break;
1985     default:
1986         ret = -ENOSYS;
1987     }
1988
1989     return ret;
1990 }
1991
1992 int kvm_arch_handle_exit(CPUState *cs, struct kvm_run *run)
1993 {
1994     S390CPU *cpu = S390_CPU(cs);
1995     int ret = 0;
1996
1997     qemu_mutex_lock_iothread();
1998
1999     switch (run->exit_reason) {
2000         case KVM_EXIT_S390_SIEIC:
2001             ret = handle_intercept(cpu);
2002             break;
2003         case KVM_EXIT_S390_RESET:
2004             s390_reipl_request();
2005             break;
2006         case KVM_EXIT_S390_TSCH:
2007             ret = handle_tsch(cpu);
2008             break;
2009         case KVM_EXIT_S390_STSI:
2010             ret = handle_stsi(cpu);
2011             break;
2012         case KVM_EXIT_DEBUG:
2013             ret = kvm_arch_handle_debug_exit(cpu);
2014             break;
2015         default:
2016             fprintf(stderr, "Unknown KVM exit: %d\n", run->exit_reason);
2017             break;
2018     }
2019     qemu_mutex_unlock_iothread();
2020
2021     if (ret == 0) {
2022         ret = EXCP_INTERRUPT;
2023     }
2024     return ret;
2025 }
2026
2027 bool kvm_arch_stop_on_emulation_error(CPUState *cpu)
2028 {
2029     return true;
2030 }
2031
2032 int kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr)
2033 {
2034     return 1;
2035 }
2036
2037 int kvm_arch_on_sigbus(int code, void *addr)
2038 {
2039     return 1;
2040 }
2041
2042 void kvm_s390_io_interrupt(uint16_t subchannel_id,
2043                            uint16_t subchannel_nr, uint32_t io_int_parm,
2044                            uint32_t io_int_word)
2045 {
2046     struct kvm_s390_irq irq = {
2047         .u.io.subchannel_id = subchannel_id,
2048         .u.io.subchannel_nr = subchannel_nr,
2049         .u.io.io_int_parm = io_int_parm,
2050         .u.io.io_int_word = io_int_word,
2051     };
2052
2053     if (io_int_word & IO_INT_WORD_AI) {
2054         irq.type = KVM_S390_INT_IO(1, 0, 0, 0);
2055     } else {
2056         irq.type = ((subchannel_id & 0xff00) << 24) |
2057             ((subchannel_id & 0x00060) << 22) | (subchannel_nr << 16);
2058     }
2059     kvm_s390_floating_interrupt(&irq);
2060 }
2061
2062 static uint64_t build_channel_report_mcic(void)
2063 {
2064     uint64_t mcic;
2065
2066     /* subclass: indicate channel report pending */
2067     mcic = MCIC_SC_CP |
2068     /* subclass modifiers: none */
2069     /* storage errors: none */
2070     /* validity bits: no damage */
2071         MCIC_VB_WP | MCIC_VB_MS | MCIC_VB_PM | MCIC_VB_IA | MCIC_VB_FP |
2072         MCIC_VB_GR | MCIC_VB_CR | MCIC_VB_ST | MCIC_VB_AR | MCIC_VB_PR |
2073         MCIC_VB_FC | MCIC_VB_CT | MCIC_VB_CC;
2074     if (kvm_check_extension(kvm_state, KVM_CAP_S390_VECTOR_REGISTERS)) {
2075         mcic |= MCIC_VB_VR;
2076     }
2077     return mcic;
2078 }
2079
2080 void kvm_s390_crw_mchk(void)
2081 {
2082     struct kvm_s390_irq irq = {
2083         .type = KVM_S390_MCHK,
2084         .u.mchk.cr14 = 1 << 28,
2085         .u.mchk.mcic = build_channel_report_mcic(),
2086     };
2087     kvm_s390_floating_interrupt(&irq);
2088 }
2089
2090 void kvm_s390_enable_css_support(S390CPU *cpu)
2091 {
2092     int r;
2093
2094     /* Activate host kernel channel subsystem support. */
2095     r = kvm_vcpu_enable_cap(CPU(cpu), KVM_CAP_S390_CSS_SUPPORT, 0);
2096     assert(r == 0);
2097 }
2098
2099 void kvm_arch_init_irq_routing(KVMState *s)
2100 {
2101     /*
2102      * Note that while irqchip capabilities generally imply that cpustates
2103      * are handled in-kernel, it is not true for s390 (yet); therefore, we
2104      * have to override the common code kvm_halt_in_kernel_allowed setting.
2105      */
2106     if (kvm_check_extension(s, KVM_CAP_IRQ_ROUTING)) {
2107         kvm_gsi_routing_allowed = true;
2108         kvm_halt_in_kernel_allowed = false;
2109     }
2110 }
2111
2112 int kvm_s390_assign_subch_ioeventfd(EventNotifier *notifier, uint32_t sch,
2113                                     int vq, bool assign)
2114 {
2115     struct kvm_ioeventfd kick = {
2116         .flags = KVM_IOEVENTFD_FLAG_VIRTIO_CCW_NOTIFY |
2117         KVM_IOEVENTFD_FLAG_DATAMATCH,
2118         .fd = event_notifier_get_fd(notifier),
2119         .datamatch = vq,
2120         .addr = sch,
2121         .len = 8,
2122     };
2123     if (!kvm_check_extension(kvm_state, KVM_CAP_IOEVENTFD)) {
2124         return -ENOSYS;
2125     }
2126     if (!assign) {
2127         kick.flags |= KVM_IOEVENTFD_FLAG_DEASSIGN;
2128     }
2129     return kvm_vm_ioctl(kvm_state, KVM_IOEVENTFD, &kick);
2130 }
2131
2132 int kvm_s390_get_memslot_count(KVMState *s)
2133 {
2134     return kvm_check_extension(s, KVM_CAP_NR_MEMSLOTS);
2135 }
2136
2137 int kvm_s390_set_cpu_state(S390CPU *cpu, uint8_t cpu_state)
2138 {
2139     struct kvm_mp_state mp_state = {};
2140     int ret;
2141
2142     /* the kvm part might not have been initialized yet */
2143     if (CPU(cpu)->kvm_state == NULL) {
2144         return 0;
2145     }
2146
2147     switch (cpu_state) {
2148     case CPU_STATE_STOPPED:
2149         mp_state.mp_state = KVM_MP_STATE_STOPPED;
2150         break;
2151     case CPU_STATE_CHECK_STOP:
2152         mp_state.mp_state = KVM_MP_STATE_CHECK_STOP;
2153         break;
2154     case CPU_STATE_OPERATING:
2155         mp_state.mp_state = KVM_MP_STATE_OPERATING;
2156         break;
2157     case CPU_STATE_LOAD:
2158         mp_state.mp_state = KVM_MP_STATE_LOAD;
2159         break;
2160     default:
2161         error_report("Requested CPU state is not a valid S390 CPU state: %u",
2162                      cpu_state);
2163         exit(1);
2164     }
2165
2166     ret = kvm_vcpu_ioctl(CPU(cpu), KVM_SET_MP_STATE, &mp_state);
2167     if (ret) {
2168         trace_kvm_failed_cpu_state_set(CPU(cpu)->cpu_index, cpu_state,
2169                                        strerror(-ret));
2170     }
2171
2172     return ret;
2173 }
2174
2175 void kvm_s390_vcpu_interrupt_pre_save(S390CPU *cpu)
2176 {
2177     struct kvm_s390_irq_state irq_state;
2178     CPUState *cs = CPU(cpu);
2179     int32_t bytes;
2180
2181     if (!kvm_check_extension(kvm_state, KVM_CAP_S390_IRQ_STATE)) {
2182         return;
2183     }
2184
2185     irq_state.buf = (uint64_t) cpu->irqstate;
2186     irq_state.len = VCPU_IRQ_BUF_SIZE;
2187
2188     bytes = kvm_vcpu_ioctl(cs, KVM_S390_GET_IRQ_STATE, &irq_state);
2189     if (bytes < 0) {
2190         cpu->irqstate_saved_size = 0;
2191         error_report("Migration of interrupt state failed");
2192         return;
2193     }
2194
2195     cpu->irqstate_saved_size = bytes;
2196 }
2197
2198 int kvm_s390_vcpu_interrupt_post_load(S390CPU *cpu)
2199 {
2200     CPUState *cs = CPU(cpu);
2201     struct kvm_s390_irq_state irq_state;
2202     int r;
2203
2204     if (cpu->irqstate_saved_size == 0) {
2205         return 0;
2206     }
2207
2208     if (!kvm_check_extension(kvm_state, KVM_CAP_S390_IRQ_STATE)) {
2209         return -ENOSYS;
2210     }
2211
2212     irq_state.buf = (uint64_t) cpu->irqstate;
2213     irq_state.len = cpu->irqstate_saved_size;
2214
2215     r = kvm_vcpu_ioctl(cs, KVM_S390_SET_IRQ_STATE, &irq_state);
2216     if (r) {
2217         error_report("Setting interrupt state failed %d", r);
2218     }
2219     return r;
2220 }
2221
2222 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry *route,
2223                              uint64_t address, uint32_t data, PCIDevice *dev)
2224 {
2225     S390PCIBusDevice *pbdev;
2226     uint32_t fid = data >> ZPCI_MSI_VEC_BITS;
2227     uint32_t vec = data & ZPCI_MSI_VEC_MASK;
2228
2229     pbdev = s390_pci_find_dev_by_fid(fid);
2230     if (!pbdev) {
2231         DPRINTF("add_msi_route no dev\n");
2232         return -ENODEV;
2233     }
2234
2235     pbdev->routes.adapter.ind_offset = vec;
2236
2237     route->type = KVM_IRQ_ROUTING_S390_ADAPTER;
2238     route->flags = 0;
2239     route->u.adapter.summary_addr = pbdev->routes.adapter.summary_addr;
2240     route->u.adapter.ind_addr = pbdev->routes.adapter.ind_addr;
2241     route->u.adapter.summary_offset = pbdev->routes.adapter.summary_offset;
2242     route->u.adapter.ind_offset = pbdev->routes.adapter.ind_offset;
2243     route->u.adapter.adapter_id = pbdev->routes.adapter.adapter_id;
2244     return 0;
2245 }
2246
2247 int kvm_arch_msi_data_to_gsi(uint32_t data)
2248 {
2249     abort();
2250 }