target-s390x/kvm.c

   1 /*
   2  * QEMU S390x KVM implementation
   3  *
   4  * Copyright (c) 2009 Alexander Graf <agraf@suse.de>
   5  *
   6  * This library is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2 of the License, or (at your option) any later version.
  10  *
  11  * This library is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public
  17  * License along with this library; if not, see <http://www.gnu.org/licenses/>.
  18  */
  19
  20 #include <sys/types.h>
  21 #include <sys/ioctl.h>
  22 #include <sys/mman.h>
  23
  24 #include <linux/kvm.h>
  25 #include <asm/ptrace.h>
  26
  27 #include "qemu-common.h"
  28 #include "qemu-timer.h"
  29 #include "sysemu.h"
  30 #include "kvm.h"
  31 #include "cpu.h"
  32 #include "device_tree.h"
  33
  34 /* #define DEBUG_KVM */
  35
  36 #ifdef DEBUG_KVM
  37 #define dprintf(fmt, ...) \
  38     do { fprintf(stderr, fmt, ## __VA_ARGS__); } while (0)
  39 #else
  40 #define dprintf(fmt, ...) \
  41     do { } while (0)
  42 #endif
  43
  44 #define IPA0_DIAG                       0x8300
  45 #define IPA0_SIGP                       0xae00
  46 #define IPA0_PRIV                       0xb200
  47
  48 #define PRIV_SCLP_CALL                  0x20
  49 #define DIAG_KVM_HYPERCALL              0x500
  50 #define DIAG_KVM_BREAKPOINT             0x501
  51
  52 #define ICPT_INSTRUCTION                0x04
  53 #define ICPT_WAITPSW                    0x1c
  54 #define ICPT_SOFT_INTERCEPT             0x24
  55 #define ICPT_CPU_STOP                   0x28
  56 #define ICPT_IO                         0x40
  57
  58 #define SIGP_RESTART                    0x06
  59 #define SIGP_INITIAL_CPU_RESET          0x0b
  60 #define SIGP_STORE_STATUS_ADDR          0x0e
  61 #define SIGP_SET_ARCH                   0x12
  62
  63 const KVMCapabilityInfo kvm_arch_required_capabilities[] = {
  64     KVM_CAP_LAST_INFO
  65 };
  66
  67 static int cap_sync_regs;
  68
  69 int kvm_arch_init(KVMState *s)
  70 {
  71     cap_sync_regs = kvm_check_extension(s, KVM_CAP_SYNC_REGS);
  72     return 0;
  73 }
  74
  75 int kvm_arch_init_vcpu(CPUS390XState *env)
  76 {
  77     int ret = 0;
  78
  79     if (kvm_vcpu_ioctl(env, KVM_S390_INITIAL_RESET, NULL) < 0) {
  80         perror("cannot init reset vcpu");
  81     }
  82
  83     return ret;
  84 }
  85
  86 void kvm_arch_reset_vcpu(CPUS390XState *env)
  87 {
  88     /* FIXME: add code to reset vcpu. */
  89 }
  90
  91 int kvm_arch_put_registers(CPUS390XState *env, int level)
  92 {
  93     struct kvm_sregs sregs;
  94     struct kvm_regs regs;
  95     int ret;
  96     int i;
  97
  98     /* always save the PSW  and the GPRS*/
  99     env->kvm_run->psw_addr = env->psw.addr;
 100     env->kvm_run->psw_mask = env->psw.mask;
 101
 102     if (cap_sync_regs && env->kvm_run->kvm_valid_regs & KVM_SYNC_GPRS) {
 103         for (i = 0; i < 16; i++) {
 104             env->kvm_run->s.regs.gprs[i] = env->regs[i];
 105             env->kvm_run->kvm_dirty_regs |= KVM_SYNC_GPRS;
 106         }
 107     } else {
 108         for (i = 0; i < 16; i++) {
 109             regs.gprs[i] = env->regs[i];
 110         }
 111         ret = kvm_vcpu_ioctl(env, KVM_SET_REGS, &regs);
 112         if (ret < 0) {
 113             return ret;
 114         }
 115     }
 116
 117     /* Do we need to save more than that? */
 118     if (level == KVM_PUT_RUNTIME_STATE) {
 119         return 0;
 120     }
 121
 122     if (cap_sync_regs &&
 123         env->kvm_run->kvm_valid_regs & KVM_SYNC_ACRS &&
 124         env->kvm_run->kvm_valid_regs & KVM_SYNC_CRS) {
 125         for (i = 0; i < 16; i++) {
 126             env->kvm_run->s.regs.acrs[i] = env->aregs[i];
 127             env->kvm_run->s.regs.crs[i] = env->cregs[i];
 128         }
 129         env->kvm_run->kvm_dirty_regs |= KVM_SYNC_ACRS;
 130         env->kvm_run->kvm_dirty_regs |= KVM_SYNC_CRS;
 131     } else {
 132         for (i = 0; i < 16; i++) {
 133             sregs.acrs[i] = env->aregs[i];
 134             sregs.crs[i] = env->cregs[i];
 135         }
 136         ret = kvm_vcpu_ioctl(env, KVM_SET_SREGS, &sregs);
 137         if (ret < 0) {
 138             return ret;
 139         }
 140     }
 141
 142     /* Finally the prefix */
 143     if (cap_sync_regs && env->kvm_run->kvm_valid_regs & KVM_SYNC_PREFIX) {
 144         env->kvm_run->s.regs.prefix = env->psa;
 145         env->kvm_run->kvm_dirty_regs |= KVM_SYNC_PREFIX;
 146     } else {
 147         /* prefix is only supported via sync regs */
 148     }
 149     return 0;
 150 }
 151
 152 int kvm_arch_get_registers(CPUS390XState *env)
 153 {
 154     struct kvm_sregs sregs;
 155     struct kvm_regs regs;
 156     int ret;
 157     int i;
 158
 159     /* get the PSW */
 160     env->psw.addr = env->kvm_run->psw_addr;
 161     env->psw.mask = env->kvm_run->psw_mask;
 162
 163     /* the GPRS */
 164     if (cap_sync_regs && env->kvm_run->kvm_valid_regs & KVM_SYNC_GPRS) {
 165         for (i = 0; i < 16; i++) {
 166             env->regs[i] = env->kvm_run->s.regs.gprs[i];
 167         }
 168     } else {
 169         ret = kvm_vcpu_ioctl(env, KVM_GET_REGS, &regs);
 170         if (ret < 0) {
 171             return ret;
 172         }
 173          for (i = 0; i < 16; i++) {
 174             env->regs[i] = regs.gprs[i];
 175         }
 176     }
 177
 178     /* The ACRS and CRS */
 179     if (cap_sync_regs &&
 180         env->kvm_run->kvm_valid_regs & KVM_SYNC_ACRS &&
 181         env->kvm_run->kvm_valid_regs & KVM_SYNC_CRS) {
 182         for (i = 0; i < 16; i++) {
 183             env->aregs[i] = env->kvm_run->s.regs.acrs[i];
 184             env->cregs[i] = env->kvm_run->s.regs.crs[i];
 185         }
 186     } else {
 187         ret = kvm_vcpu_ioctl(env, KVM_GET_SREGS, &sregs);
 188         if (ret < 0) {
 189             return ret;
 190         }
 191          for (i = 0; i < 16; i++) {
 192             env->aregs[i] = sregs.acrs[i];
 193             env->cregs[i] = sregs.crs[i];
 194         }
 195     }
 196
 197     /* Finally the prefix */
 198     if (cap_sync_regs && env->kvm_run->kvm_valid_regs & KVM_SYNC_PREFIX) {
 199         env->psa = env->kvm_run->s.regs.prefix;
 200     } else {
 201         /* no prefix without sync regs */
 202     }
 203
 204     return 0;
 205 }
 206
 207 /*
 208  * Legacy layout for s390:
 209  * Older S390 KVM requires the topmost vma of the RAM to be
 210  * smaller than an system defined value, which is at least 256GB.
 211  * Larger systems have larger values. We put the guest between
 212  * the end of data segment (system break) and this value. We
 213  * use 32GB as a base to have enough room for the system break
 214  * to grow. We also have to use MAP parameters that avoid
 215  * read-only mapping of guest pages.
 216  */
 217 static void *legacy_s390_alloc(ram_addr_t size)
 218 {
 219     void *mem;
 220
 221     mem = mmap((void *) 0x800000000ULL, size,
 222                PROT_EXEC|PROT_READ|PROT_WRITE,
 223                MAP_SHARED | MAP_ANONYMOUS | MAP_FIXED, -1, 0);
 224     if (mem == MAP_FAILED) {
 225         fprintf(stderr, "Allocating RAM failed\n");
 226         abort();
 227     }
 228     return mem;
 229 }
 230
 231 void *kvm_arch_vmalloc(ram_addr_t size)
 232 {
 233     /* Can we use the standard allocation ? */
 234     if (kvm_check_extension(kvm_state, KVM_CAP_S390_GMAP) &&
 235         kvm_check_extension(kvm_state, KVM_CAP_S390_COW)) {
 236         return NULL;
 237     } else {
 238         return legacy_s390_alloc(size);
 239     }
 240 }
 241
 242 int kvm_arch_insert_sw_breakpoint(CPUS390XState *env, struct kvm_sw_breakpoint *bp)
 243 {
 244     static const uint8_t diag_501[] = {0x83, 0x24, 0x05, 0x01};
 245
 246     if (cpu_memory_rw_debug(env, bp->pc, (uint8_t *)&bp->saved_insn, 4, 0) ||
 247         cpu_memory_rw_debug(env, bp->pc, (uint8_t *)diag_501, 4, 1)) {
 248         return -EINVAL;
 249     }
 250     return 0;
 251 }
 252
 253 int kvm_arch_remove_sw_breakpoint(CPUS390XState *env, struct kvm_sw_breakpoint *bp)
 254 {
 255     uint8_t t[4];
 256     static const uint8_t diag_501[] = {0x83, 0x24, 0x05, 0x01};
 257
 258     if (cpu_memory_rw_debug(env, bp->pc, t, 4, 0)) {
 259         return -EINVAL;
 260     } else if (memcmp(t, diag_501, 4)) {
 261         return -EINVAL;
 262     } else if (cpu_memory_rw_debug(env, bp->pc, (uint8_t *)&bp->saved_insn, 1, 1)) {
 263         return -EINVAL;
 264     }
 265
 266     return 0;
 267 }
 268
 269 void kvm_arch_pre_run(CPUS390XState *env, struct kvm_run *run)
 270 {
 271 }
 272
 273 void kvm_arch_post_run(CPUS390XState *env, struct kvm_run *run)
 274 {
 275 }
 276
 277 int kvm_arch_process_async_events(CPUS390XState *env)
 278 {
 279     return env->halted;
 280 }
 281
 282 void kvm_s390_interrupt_internal(CPUS390XState *env, int type, uint32_t parm,
 283                                  uint64_t parm64, int vm)
 284 {
 285     struct kvm_s390_interrupt kvmint;
 286     int r;
 287
 288     if (!env->kvm_state) {
 289         return;
 290     }
 291
 292     kvmint.type = type;
 293     kvmint.parm = parm;
 294     kvmint.parm64 = parm64;
 295
 296     if (vm) {
 297         r = kvm_vm_ioctl(env->kvm_state, KVM_S390_INTERRUPT, &kvmint);
 298     } else {
 299         r = kvm_vcpu_ioctl(env, KVM_S390_INTERRUPT, &kvmint);
 300     }
 301
 302     if (r < 0) {
 303         fprintf(stderr, "KVM failed to inject interrupt\n");
 304         exit(1);
 305     }
 306 }
 307
 308 void kvm_s390_virtio_irq(CPUS390XState *env, int config_change, uint64_t token)
 309 {
 310     kvm_s390_interrupt_internal(env, KVM_S390_INT_VIRTIO, config_change,
 311                                 token, 1);
 312 }
 313
 314 void kvm_s390_interrupt(CPUS390XState *env, int type, uint32_t code)
 315 {
 316     kvm_s390_interrupt_internal(env, type, code, 0, 0);
 317 }
 318
 319 static void enter_pgmcheck(CPUS390XState *env, uint16_t code)
 320 {
 321     kvm_s390_interrupt(env, KVM_S390_PROGRAM_INT, code);
 322 }
 323
 324 static inline void setcc(CPUS390XState *env, uint64_t cc)
 325 {
 326     env->kvm_run->psw_mask &= ~(3ull << 44);
 327     env->kvm_run->psw_mask |= (cc & 3) << 44;
 328
 329     env->psw.mask &= ~(3ul << 44);
 330     env->psw.mask |= (cc & 3) << 44;
 331 }
 332
 333 static int kvm_sclp_service_call(CPUS390XState *env, struct kvm_run *run,
 334                                  uint16_t ipbh0)
 335 {
 336     uint32_t sccb;
 337     uint64_t code;
 338     int r = 0;
 339
 340     cpu_synchronize_state(env);
 341     sccb = env->regs[ipbh0 & 0xf];
 342     code = env->regs[(ipbh0 & 0xf0) >> 4];
 343
 344     r = sclp_service_call(sccb, code);
 345     if (r < 0) {
 346         enter_pgmcheck(env, -r);
 347     }
 348     setcc(env, r);
 349
 350     return 0;
 351 }
 352
 353 static int handle_priv(CPUS390XState *env, struct kvm_run *run, uint8_t ipa1)
 354 {
 355     int r = 0;
 356     uint16_t ipbh0 = (run->s390_sieic.ipb & 0xffff0000) >> 16;
 357
 358     dprintf("KVM: PRIV: %d\n", ipa1);
 359     switch (ipa1) {
 360         case PRIV_SCLP_CALL:
 361             r = kvm_sclp_service_call(env, run, ipbh0);
 362             break;
 363         default:
 364             dprintf("KVM: unknown PRIV: 0x%x\n", ipa1);
 365             r = -1;
 366             break;
 367     }
 368
 369     return r;
 370 }
 371
 372 static int handle_hypercall(CPUS390XState *env, struct kvm_run *run)
 373 {
 374     cpu_synchronize_state(env);
 375     env->regs[2] = s390_virtio_hypercall(env, env->regs[2], env->regs[1]);
 376
 377     return 0;
 378 }
 379
 380 static int handle_diag(CPUS390XState *env, struct kvm_run *run, int ipb_code)
 381 {
 382     int r = 0;
 383
 384     switch (ipb_code) {
 385         case DIAG_KVM_HYPERCALL:
 386             r = handle_hypercall(env, run);
 387             break;
 388         case DIAG_KVM_BREAKPOINT:
 389             sleep(10);
 390             break;
 391         default:
 392             dprintf("KVM: unknown DIAG: 0x%x\n", ipb_code);
 393             r = -1;
 394             break;
 395     }
 396
 397     return r;
 398 }
 399
 400 static int s390_cpu_restart(S390CPU *cpu)
 401 {
 402     CPUS390XState *env = &cpu->env;
 403
 404     kvm_s390_interrupt(env, KVM_S390_RESTART, 0);
 405     s390_add_running_cpu(env);
 406     qemu_cpu_kick(CPU(cpu));
 407     dprintf("DONE: SIGP cpu restart: %p\n", env);
 408     return 0;
 409 }
 410
 411 static int s390_store_status(CPUS390XState *env, uint32_t parameter)
 412 {
 413     /* XXX */
 414     fprintf(stderr, "XXX SIGP store status\n");
 415     return -1;
 416 }
 417
 418 static int s390_cpu_initial_reset(CPUS390XState *env)
 419 {
 420     int i;
 421
 422     s390_del_running_cpu(env);
 423     if (kvm_vcpu_ioctl(env, KVM_S390_INITIAL_RESET, NULL) < 0) {
 424         perror("cannot init reset vcpu");
 425     }
 426
 427     /* Manually zero out all registers */
 428     cpu_synchronize_state(env);
 429     for (i = 0; i < 16; i++) {
 430         env->regs[i] = 0;
 431     }
 432
 433     dprintf("DONE: SIGP initial reset: %p\n", env);
 434     return 0;
 435 }
 436
 437 static int handle_sigp(CPUS390XState *env, struct kvm_run *run, uint8_t ipa1)
 438 {
 439     uint8_t order_code;
 440     uint32_t parameter;
 441     uint16_t cpu_addr;
 442     uint8_t t;
 443     int r = -1;
 444     S390CPU *target_cpu;
 445     CPUS390XState *target_env;
 446
 447     cpu_synchronize_state(env);
 448
 449     /* get order code */
 450     order_code = run->s390_sieic.ipb >> 28;
 451     if (order_code > 0) {
 452         order_code = env->regs[order_code];
 453     }
 454     order_code += (run->s390_sieic.ipb & 0x0fff0000) >> 16;
 455
 456     /* get parameters */
 457     t = (ipa1 & 0xf0) >> 4;
 458     if (!(t % 2)) {
 459         t++;
 460     }
 461
 462     parameter = env->regs[t] & 0x7ffffe00;
 463     cpu_addr = env->regs[ipa1 & 0x0f];
 464
 465     target_cpu = s390_cpu_addr2state(cpu_addr);
 466     if (target_cpu == NULL) {
 467         goto out;
 468     }
 469     target_env = &target_cpu->env;
 470
 471     switch (order_code) {
 472         case SIGP_RESTART:
 473             r = s390_cpu_restart(target_cpu);
 474             break;
 475         case SIGP_STORE_STATUS_ADDR:
 476             r = s390_store_status(target_env, parameter);
 477             break;
 478         case SIGP_SET_ARCH:
 479             /* make the caller panic */
 480             return -1;
 481         case SIGP_INITIAL_CPU_RESET:
 482             r = s390_cpu_initial_reset(target_env);
 483             break;
 484         default:
 485             fprintf(stderr, "KVM: unknown SIGP: 0x%x\n", order_code);
 486             break;
 487     }
 488
 489 out:
 490     setcc(env, r ? 3 : 0);
 491     return 0;
 492 }
 493
 494 static int handle_instruction(CPUS390XState *env, struct kvm_run *run)
 495 {
 496     unsigned int ipa0 = (run->s390_sieic.ipa & 0xff00);
 497     uint8_t ipa1 = run->s390_sieic.ipa & 0x00ff;
 498     int ipb_code = (run->s390_sieic.ipb & 0x0fff0000) >> 16;
 499     int r = -1;
 500
 501     dprintf("handle_instruction 0x%x 0x%x\n", run->s390_sieic.ipa, run->s390_sieic.ipb);
 502     switch (ipa0) {
 503         case IPA0_PRIV:
 504             r = handle_priv(env, run, ipa1);
 505             break;
 506         case IPA0_DIAG:
 507             r = handle_diag(env, run, ipb_code);
 508             break;
 509         case IPA0_SIGP:
 510             r = handle_sigp(env, run, ipa1);
 511             break;
 512     }
 513
 514     if (r < 0) {
 515         enter_pgmcheck(env, 0x0001);
 516     }
 517     return 0;
 518 }
 519
 520 static bool is_special_wait_psw(CPUS390XState *env)
 521 {
 522     /* signal quiesce */
 523     return env->kvm_run->psw_addr == 0xfffUL;
 524 }
 525
 526 static int handle_intercept(CPUS390XState *env)
 527 {
 528     struct kvm_run *run = env->kvm_run;
 529     int icpt_code = run->s390_sieic.icptcode;
 530     int r = 0;
 531
 532     dprintf("intercept: 0x%x (at 0x%lx)\n", icpt_code,
 533             (long)env->kvm_run->psw_addr);
 534     switch (icpt_code) {
 535         case ICPT_INSTRUCTION:
 536             r = handle_instruction(env, run);
 537             break;
 538         case ICPT_WAITPSW:
 539             if (s390_del_running_cpu(env) == 0 &&
 540                 is_special_wait_psw(env)) {
 541                 qemu_system_shutdown_request();
 542             }
 543             r = EXCP_HALTED;
 544             break;
 545         case ICPT_CPU_STOP:
 546             if (s390_del_running_cpu(env) == 0) {
 547                 qemu_system_shutdown_request();
 548             }
 549             r = EXCP_HALTED;
 550             break;
 551         case ICPT_SOFT_INTERCEPT:
 552             fprintf(stderr, "KVM unimplemented icpt SOFT\n");
 553             exit(1);
 554             break;
 555         case ICPT_IO:
 556             fprintf(stderr, "KVM unimplemented icpt IO\n");
 557             exit(1);
 558             break;
 559         default:
 560             fprintf(stderr, "Unknown intercept code: %d\n", icpt_code);
 561             exit(1);
 562             break;
 563     }
 564
 565     return r;
 566 }
 567
 568 int kvm_arch_handle_exit(CPUS390XState *env, struct kvm_run *run)
 569 {
 570     int ret = 0;
 571
 572     switch (run->exit_reason) {
 573         case KVM_EXIT_S390_SIEIC:
 574             ret = handle_intercept(env);
 575             break;
 576         case KVM_EXIT_S390_RESET:
 577             qemu_system_reset_request();
 578             break;
 579         default:
 580             fprintf(stderr, "Unknown KVM exit: %d\n", run->exit_reason);
 581             break;
 582     }
 583
 584     if (ret == 0) {
 585         ret = EXCP_INTERRUPT;
 586     }
 587     return ret;
 588 }
 589
 590 bool kvm_arch_stop_on_emulation_error(CPUS390XState *env)
 591 {
 592     return true;
 593 }
 594
 595 int kvm_arch_on_sigbus_vcpu(CPUS390XState *env, int code, void *addr)
 596 {
 597     return 1;
 598 }
 599
 600 int kvm_arch_on_sigbus(int code, void *addr)
 601 {
 602     return 1;
 603 }