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 "cpu.h"
  36 #include "sysemu/device_tree.h"
  37
  38 /* #define DEBUG_KVM */
  39
  40 #ifdef DEBUG_KVM
  41 #define dprintf(fmt, ...) \
  42     do { fprintf(stderr, fmt, ## __VA_ARGS__); } while (0)
  43 #else
  44 #define dprintf(fmt, ...) \
  45     do { } while (0)
  46 #endif
  47
  48 #define IPA0_DIAG                       0x8300
  49 #define IPA0_SIGP                       0xae00
  50 #define IPA0_PRIV                       0xb200
  51
  52 #define PRIV_SCLP_CALL                  0x20
  53 #define DIAG_KVM_HYPERCALL              0x500
  54 #define DIAG_KVM_BREAKPOINT             0x501
  55
  56 #define ICPT_INSTRUCTION                0x04
  57 #define ICPT_WAITPSW                    0x1c
  58 #define ICPT_SOFT_INTERCEPT             0x24
  59 #define ICPT_CPU_STOP                   0x28
  60 #define ICPT_IO                         0x40
  61
  62 #define SIGP_RESTART                    0x06
  63 #define SIGP_INITIAL_CPU_RESET          0x0b
  64 #define SIGP_STORE_STATUS_ADDR          0x0e
  65 #define SIGP_SET_ARCH                   0x12
  66
  67 const KVMCapabilityInfo kvm_arch_required_capabilities[] = {
  68     KVM_CAP_LAST_INFO
  69 };
  70
  71 static int cap_sync_regs;
  72
  73 int kvm_arch_init(KVMState *s)
  74 {
  75     cap_sync_regs = kvm_check_extension(s, KVM_CAP_SYNC_REGS);
  76     return 0;
  77 }
  78
  79 unsigned long kvm_arch_vcpu_id(CPUState *cpu)
  80 {
  81     return cpu->cpu_index;
  82 }
  83
  84 int kvm_arch_init_vcpu(CPUState *cpu)
  85 {
  86     int ret = 0;
  87
  88     if (kvm_vcpu_ioctl(cpu, KVM_S390_INITIAL_RESET, NULL) < 0) {
  89         perror("cannot init reset vcpu");
  90     }
  91
  92     return ret;
  93 }
  94
  95 void kvm_arch_reset_vcpu(CPUState *cpu)
  96 {
  97     /* The initial reset call is needed here to reset in-kernel
  98      * vcpu data that we can't access directly from QEMU
  99      * (i.e. with older kernels which don't support sync_regs/ONE_REG).
 100      * Before this ioctl cpu_synchronize_state() is called in common kvm
 101      * code (kvm-all) */
 102     if (kvm_vcpu_ioctl(cpu, KVM_S390_INITIAL_RESET, NULL)) {
 103         perror("Can't reset vcpu\n");
 104     }
 105 }
 106
 107 int kvm_arch_put_registers(CPUState *cs, int level)
 108 {
 109     S390CPU *cpu = S390_CPU(cs);
 110     CPUS390XState *env = &cpu->env;
 111     struct kvm_sregs sregs;
 112     struct kvm_regs regs;
 113     int ret;
 114     int i;
 115
 116     /* always save the PSW  and the GPRS*/
 117     cs->kvm_run->psw_addr = env->psw.addr;
 118     cs->kvm_run->psw_mask = env->psw.mask;
 119
 120     if (cap_sync_regs && cs->kvm_run->kvm_valid_regs & KVM_SYNC_GPRS) {
 121         for (i = 0; i < 16; i++) {
 122             cs->kvm_run->s.regs.gprs[i] = env->regs[i];
 123             cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_GPRS;
 124         }
 125     } else {
 126         for (i = 0; i < 16; i++) {
 127             regs.gprs[i] = env->regs[i];
 128         }
 129         ret = kvm_vcpu_ioctl(cs, KVM_SET_REGS, &regs);
 130         if (ret < 0) {
 131             return ret;
 132         }
 133     }
 134
 135     /* Do we need to save more than that? */
 136     if (level == KVM_PUT_RUNTIME_STATE) {
 137         return 0;
 138     }
 139
 140     if (cap_sync_regs &&
 141         cs->kvm_run->kvm_valid_regs & KVM_SYNC_ACRS &&
 142         cs->kvm_run->kvm_valid_regs & KVM_SYNC_CRS) {
 143         for (i = 0; i < 16; i++) {
 144             cs->kvm_run->s.regs.acrs[i] = env->aregs[i];
 145             cs->kvm_run->s.regs.crs[i] = env->cregs[i];
 146         }
 147         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_ACRS;
 148         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_CRS;
 149     } else {
 150         for (i = 0; i < 16; i++) {
 151             sregs.acrs[i] = env->aregs[i];
 152             sregs.crs[i] = env->cregs[i];
 153         }
 154         ret = kvm_vcpu_ioctl(cs, KVM_SET_SREGS, &sregs);
 155         if (ret < 0) {
 156             return ret;
 157         }
 158     }
 159
 160     /* Finally the prefix */
 161     if (cap_sync_regs && cs->kvm_run->kvm_valid_regs & KVM_SYNC_PREFIX) {
 162         cs->kvm_run->s.regs.prefix = env->psa;
 163         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_PREFIX;
 164     } else {
 165         /* prefix is only supported via sync regs */
 166     }
 167     return 0;
 168 }
 169
 170 int kvm_arch_get_registers(CPUState *cs)
 171 {
 172     S390CPU *cpu = S390_CPU(cs);
 173     CPUS390XState *env = &cpu->env;
 174     struct kvm_sregs sregs;
 175     struct kvm_regs regs;
 176     int ret;
 177     int i;
 178
 179     /* get the PSW */
 180     env->psw.addr = cs->kvm_run->psw_addr;
 181     env->psw.mask = cs->kvm_run->psw_mask;
 182
 183     /* the GPRS */
 184     if (cap_sync_regs && cs->kvm_run->kvm_valid_regs & KVM_SYNC_GPRS) {
 185         for (i = 0; i < 16; i++) {
 186             env->regs[i] = cs->kvm_run->s.regs.gprs[i];
 187         }
 188     } else {
 189         ret = kvm_vcpu_ioctl(cs, KVM_GET_REGS, &regs);
 190         if (ret < 0) {
 191             return ret;
 192         }
 193          for (i = 0; i < 16; i++) {
 194             env->regs[i] = regs.gprs[i];
 195         }
 196     }
 197
 198     /* The ACRS and CRS */
 199     if (cap_sync_regs &&
 200         cs->kvm_run->kvm_valid_regs & KVM_SYNC_ACRS &&
 201         cs->kvm_run->kvm_valid_regs & KVM_SYNC_CRS) {
 202         for (i = 0; i < 16; i++) {
 203             env->aregs[i] = cs->kvm_run->s.regs.acrs[i];
 204             env->cregs[i] = cs->kvm_run->s.regs.crs[i];
 205         }
 206     } else {
 207         ret = kvm_vcpu_ioctl(cs, KVM_GET_SREGS, &sregs);
 208         if (ret < 0) {
 209             return ret;
 210         }
 211          for (i = 0; i < 16; i++) {
 212             env->aregs[i] = sregs.acrs[i];
 213             env->cregs[i] = sregs.crs[i];
 214         }
 215     }
 216
 217     /* Finally the prefix */
 218     if (cap_sync_regs && cs->kvm_run->kvm_valid_regs & KVM_SYNC_PREFIX) {
 219         env->psa = cs->kvm_run->s.regs.prefix;
 220     } else {
 221         /* no prefix without sync regs */
 222     }
 223
 224     return 0;
 225 }
 226
 227 /*
 228  * Legacy layout for s390:
 229  * Older S390 KVM requires the topmost vma of the RAM to be
 230  * smaller than an system defined value, which is at least 256GB.
 231  * Larger systems have larger values. We put the guest between
 232  * the end of data segment (system break) and this value. We
 233  * use 32GB as a base to have enough room for the system break
 234  * to grow. We also have to use MAP parameters that avoid
 235  * read-only mapping of guest pages.
 236  */
 237 static void *legacy_s390_alloc(ram_addr_t size)
 238 {
 239     void *mem;
 240
 241     mem = mmap((void *) 0x800000000ULL, size,
 242                PROT_EXEC|PROT_READ|PROT_WRITE,
 243                MAP_SHARED | MAP_ANONYMOUS | MAP_FIXED, -1, 0);
 244     if (mem == MAP_FAILED) {
 245         fprintf(stderr, "Allocating RAM failed\n");
 246         abort();
 247     }
 248     return mem;
 249 }
 250
 251 void *kvm_arch_vmalloc(ram_addr_t size)
 252 {
 253     /* Can we use the standard allocation ? */
 254     if (kvm_check_extension(kvm_state, KVM_CAP_S390_GMAP) &&
 255         kvm_check_extension(kvm_state, KVM_CAP_S390_COW)) {
 256         return NULL;
 257     } else {
 258         return legacy_s390_alloc(size);
 259     }
 260 }
 261
 262 int kvm_arch_insert_sw_breakpoint(CPUState *cs, struct kvm_sw_breakpoint *bp)
 263 {
 264     S390CPU *cpu = S390_CPU(cs);
 265     CPUS390XState *env = &cpu->env;
 266     static const uint8_t diag_501[] = {0x83, 0x24, 0x05, 0x01};
 267
 268     if (cpu_memory_rw_debug(env, bp->pc, (uint8_t *)&bp->saved_insn, 4, 0) ||
 269         cpu_memory_rw_debug(env, bp->pc, (uint8_t *)diag_501, 4, 1)) {
 270         return -EINVAL;
 271     }
 272     return 0;
 273 }
 274
 275 int kvm_arch_remove_sw_breakpoint(CPUState *cs, struct kvm_sw_breakpoint *bp)
 276 {
 277     S390CPU *cpu = S390_CPU(cs);
 278     CPUS390XState *env = &cpu->env;
 279     uint8_t t[4];
 280     static const uint8_t diag_501[] = {0x83, 0x24, 0x05, 0x01};
 281
 282     if (cpu_memory_rw_debug(env, bp->pc, t, 4, 0)) {
 283         return -EINVAL;
 284     } else if (memcmp(t, diag_501, 4)) {
 285         return -EINVAL;
 286     } else if (cpu_memory_rw_debug(env, bp->pc, (uint8_t *)&bp->saved_insn, 1, 1)) {
 287         return -EINVAL;
 288     }
 289
 290     return 0;
 291 }
 292
 293 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run)
 294 {
 295 }
 296
 297 void kvm_arch_post_run(CPUState *cpu, struct kvm_run *run)
 298 {
 299 }
 300
 301 int kvm_arch_process_async_events(CPUState *cs)
 302 {
 303     S390CPU *cpu = S390_CPU(cs);
 304     return cpu->env.halted;
 305 }
 306
 307 void kvm_s390_interrupt_internal(S390CPU *cpu, int type, uint32_t parm,
 308                                  uint64_t parm64, int vm)
 309 {
 310     CPUState *cs = CPU(cpu);
 311     struct kvm_s390_interrupt kvmint;
 312     int r;
 313
 314     if (!cs->kvm_state) {
 315         return;
 316     }
 317
 318     kvmint.type = type;
 319     kvmint.parm = parm;
 320     kvmint.parm64 = parm64;
 321
 322     if (vm) {
 323         r = kvm_vm_ioctl(cs->kvm_state, KVM_S390_INTERRUPT, &kvmint);
 324     } else {
 325         r = kvm_vcpu_ioctl(cs, KVM_S390_INTERRUPT, &kvmint);
 326     }
 327
 328     if (r < 0) {
 329         fprintf(stderr, "KVM failed to inject interrupt\n");
 330         exit(1);
 331     }
 332 }
 333
 334 void kvm_s390_virtio_irq(S390CPU *cpu, int config_change, uint64_t token)
 335 {
 336     kvm_s390_interrupt_internal(cpu, KVM_S390_INT_VIRTIO, config_change,
 337                                 token, 1);
 338 }
 339
 340 void kvm_s390_interrupt(S390CPU *cpu, int type, uint32_t code)
 341 {
 342     kvm_s390_interrupt_internal(cpu, type, code, 0, 0);
 343 }
 344
 345 static void enter_pgmcheck(S390CPU *cpu, uint16_t code)
 346 {
 347     kvm_s390_interrupt(cpu, KVM_S390_PROGRAM_INT, code);
 348 }
 349
 350 static inline void setcc(S390CPU *cpu, uint64_t cc)
 351 {
 352     CPUS390XState *env = &cpu->env;
 353     CPUState *cs = CPU(cpu);
 354
 355     cs->kvm_run->psw_mask &= ~(3ull << 44);
 356     cs->kvm_run->psw_mask |= (cc & 3) << 44;
 357
 358     env->psw.mask &= ~(3ul << 44);
 359     env->psw.mask |= (cc & 3) << 44;
 360 }
 361
 362 static int kvm_sclp_service_call(S390CPU *cpu, struct kvm_run *run,
 363                                  uint16_t ipbh0)
 364 {
 365     CPUS390XState *env = &cpu->env;
 366     uint32_t sccb;
 367     uint64_t code;
 368     int r = 0;
 369
 370     cpu_synchronize_state(env);
 371     sccb = env->regs[ipbh0 & 0xf];
 372     code = env->regs[(ipbh0 & 0xf0) >> 4];
 373
 374     r = sclp_service_call(sccb, code);
 375     if (r < 0) {
 376         enter_pgmcheck(cpu, -r);
 377     }
 378     setcc(cpu, r);
 379
 380     return 0;
 381 }
 382
 383 static int handle_priv(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
 384 {
 385     int r = 0;
 386     uint16_t ipbh0 = (run->s390_sieic.ipb & 0xffff0000) >> 16;
 387
 388     dprintf("KVM: PRIV: %d\n", ipa1);
 389     switch (ipa1) {
 390         case PRIV_SCLP_CALL:
 391             r = kvm_sclp_service_call(cpu, run, ipbh0);
 392             break;
 393         default:
 394             dprintf("KVM: unknown PRIV: 0x%x\n", ipa1);
 395             r = -1;
 396             break;
 397     }
 398
 399     return r;
 400 }
 401
 402 static int handle_hypercall(CPUS390XState *env, struct kvm_run *run)
 403 {
 404     cpu_synchronize_state(env);
 405     env->regs[2] = s390_virtio_hypercall(env);
 406
 407     return 0;
 408 }
 409
 410 static int handle_diag(CPUS390XState *env, struct kvm_run *run, int ipb_code)
 411 {
 412     int r = 0;
 413
 414     switch (ipb_code) {
 415         case DIAG_KVM_HYPERCALL:
 416             r = handle_hypercall(env, run);
 417             break;
 418         case DIAG_KVM_BREAKPOINT:
 419             sleep(10);
 420             break;
 421         default:
 422             dprintf("KVM: unknown DIAG: 0x%x\n", ipb_code);
 423             r = -1;
 424             break;
 425     }
 426
 427     return r;
 428 }
 429
 430 static int s390_cpu_restart(S390CPU *cpu)
 431 {
 432     CPUS390XState *env = &cpu->env;
 433
 434     kvm_s390_interrupt(cpu, KVM_S390_RESTART, 0);
 435     s390_add_running_cpu(env);
 436     qemu_cpu_kick(CPU(cpu));
 437     dprintf("DONE: SIGP cpu restart: %p\n", env);
 438     return 0;
 439 }
 440
 441 static int s390_store_status(CPUS390XState *env, uint32_t parameter)
 442 {
 443     /* XXX */
 444     fprintf(stderr, "XXX SIGP store status\n");
 445     return -1;
 446 }
 447
 448 static int s390_cpu_initial_reset(S390CPU *cpu)
 449 {
 450     CPUS390XState *env = &cpu->env;
 451     int i;
 452
 453     s390_del_running_cpu(env);
 454     if (kvm_vcpu_ioctl(CPU(cpu), KVM_S390_INITIAL_RESET, NULL) < 0) {
 455         perror("cannot init reset vcpu");
 456     }
 457
 458     /* Manually zero out all registers */
 459     cpu_synchronize_state(env);
 460     for (i = 0; i < 16; i++) {
 461         env->regs[i] = 0;
 462     }
 463
 464     dprintf("DONE: SIGP initial reset: %p\n", env);
 465     return 0;
 466 }
 467
 468 static int handle_sigp(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
 469 {
 470     CPUS390XState *env = &cpu->env;
 471     uint8_t order_code;
 472     uint32_t parameter;
 473     uint16_t cpu_addr;
 474     uint8_t t;
 475     int r = -1;
 476     S390CPU *target_cpu;
 477     CPUS390XState *target_env;
 478
 479     cpu_synchronize_state(env);
 480
 481     /* get order code */
 482     order_code = run->s390_sieic.ipb >> 28;
 483     if (order_code > 0) {
 484         order_code = env->regs[order_code];
 485     }
 486     order_code += (run->s390_sieic.ipb & 0x0fff0000) >> 16;
 487
 488     /* get parameters */
 489     t = (ipa1 & 0xf0) >> 4;
 490     if (!(t % 2)) {
 491         t++;
 492     }
 493
 494     parameter = env->regs[t] & 0x7ffffe00;
 495     cpu_addr = env->regs[ipa1 & 0x0f];
 496
 497     target_cpu = s390_cpu_addr2state(cpu_addr);
 498     if (target_cpu == NULL) {
 499         goto out;
 500     }
 501     target_env = &target_cpu->env;
 502
 503     switch (order_code) {
 504         case SIGP_RESTART:
 505             r = s390_cpu_restart(target_cpu);
 506             break;
 507         case SIGP_STORE_STATUS_ADDR:
 508             r = s390_store_status(target_env, parameter);
 509             break;
 510         case SIGP_SET_ARCH:
 511             /* make the caller panic */
 512             return -1;
 513         case SIGP_INITIAL_CPU_RESET:
 514             r = s390_cpu_initial_reset(target_cpu);
 515             break;
 516         default:
 517             fprintf(stderr, "KVM: unknown SIGP: 0x%x\n", order_code);
 518             break;
 519     }
 520
 521 out:
 522     setcc(cpu, r ? 3 : 0);
 523     return 0;
 524 }
 525
 526 static int handle_instruction(S390CPU *cpu, struct kvm_run *run)
 527 {
 528     CPUS390XState *env = &cpu->env;
 529     unsigned int ipa0 = (run->s390_sieic.ipa & 0xff00);
 530     uint8_t ipa1 = run->s390_sieic.ipa & 0x00ff;
 531     int ipb_code = (run->s390_sieic.ipb & 0x0fff0000) >> 16;
 532     int r = -1;
 533
 534     dprintf("handle_instruction 0x%x 0x%x\n", run->s390_sieic.ipa, run->s390_sieic.ipb);
 535     switch (ipa0) {
 536         case IPA0_PRIV:
 537             r = handle_priv(cpu, run, ipa1);
 538             break;
 539         case IPA0_DIAG:
 540             r = handle_diag(env, run, ipb_code);
 541             break;
 542         case IPA0_SIGP:
 543             r = handle_sigp(cpu, run, ipa1);
 544             break;
 545     }
 546
 547     if (r < 0) {
 548         enter_pgmcheck(cpu, 0x0001);
 549     }
 550     return 0;
 551 }
 552
 553 static bool is_special_wait_psw(CPUState *cs)
 554 {
 555     /* signal quiesce */
 556     return cs->kvm_run->psw_addr == 0xfffUL;
 557 }
 558
 559 static int handle_intercept(S390CPU *cpu)
 560 {
 561     CPUS390XState *env = &cpu->env;
 562     CPUState *cs = CPU(cpu);
 563     struct kvm_run *run = cs->kvm_run;
 564     int icpt_code = run->s390_sieic.icptcode;
 565     int r = 0;
 566
 567     dprintf("intercept: 0x%x (at 0x%lx)\n", icpt_code,
 568             (long)cs->kvm_run->psw_addr);
 569     switch (icpt_code) {
 570         case ICPT_INSTRUCTION:
 571             r = handle_instruction(cpu, run);
 572             break;
 573         case ICPT_WAITPSW:
 574             if (s390_del_running_cpu(env) == 0 &&
 575                 is_special_wait_psw(cs)) {
 576                 qemu_system_shutdown_request();
 577             }
 578             r = EXCP_HALTED;
 579             break;
 580         case ICPT_CPU_STOP:
 581             if (s390_del_running_cpu(env) == 0) {
 582                 qemu_system_shutdown_request();
 583             }
 584             r = EXCP_HALTED;
 585             break;
 586         case ICPT_SOFT_INTERCEPT:
 587             fprintf(stderr, "KVM unimplemented icpt SOFT\n");
 588             exit(1);
 589             break;
 590         case ICPT_IO:
 591             fprintf(stderr, "KVM unimplemented icpt IO\n");
 592             exit(1);
 593             break;
 594         default:
 595             fprintf(stderr, "Unknown intercept code: %d\n", icpt_code);
 596             exit(1);
 597             break;
 598     }
 599
 600     return r;
 601 }
 602
 603 int kvm_arch_handle_exit(CPUState *cs, struct kvm_run *run)
 604 {
 605     S390CPU *cpu = S390_CPU(cs);
 606     int ret = 0;
 607
 608     switch (run->exit_reason) {
 609         case KVM_EXIT_S390_SIEIC:
 610             ret = handle_intercept(cpu);
 611             break;
 612         case KVM_EXIT_S390_RESET:
 613             qemu_system_reset_request();
 614             break;
 615         default:
 616             fprintf(stderr, "Unknown KVM exit: %d\n", run->exit_reason);
 617             break;
 618     }
 619
 620     if (ret == 0) {
 621         ret = EXCP_INTERRUPT;
 622     }
 623     return ret;
 624 }
 625
 626 bool kvm_arch_stop_on_emulation_error(CPUState *cpu)
 627 {
 628     return true;
 629 }
 630
 631 int kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr)
 632 {
 633     return 1;
 634 }
 635
 636 int kvm_arch_on_sigbus(int code, void *addr)
 637 {
 638     return 1;
 639 }