bsd-user/main.c

   1 /*
   2  *  qemu bsd user main
   3  *
   4  *  Copyright (c) 2003-2008 Fabrice Bellard
   5  *  Copyright (c) 2013-14 Stacey Son
   6  *
   7  *  This program is free software; you can redistribute it and/or modify
   8  *  it under the terms of the GNU General Public License as published by
   9  *  the Free Software Foundation; either version 2 of the License, or
  10  *  (at your option) any later version.
  11  *
  12  *  This program 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
  15  *  GNU General Public License for more details.
  16  *
  17  *  You should have received a copy of the GNU General Public License
  18  *  along with this program; if not, see <http://www.gnu.org/licenses/>.
  19  */
  20
  21 #include "qemu/osdep.h"
  22 #include <sys/resource.h>
  23 #include <sys/sysctl.h>
  24
  25 #include "qemu/help-texts.h"
  26 #include "qemu/units.h"
  27 #include "qemu/accel.h"
  28 #include "qemu-version.h"
  29 #include <machine/trap.h>
  30
  31 #include "qapi/error.h"
  32 #include "qemu.h"
  33 #include "qemu/config-file.h"
  34 #include "qemu/error-report.h"
  35 #include "qemu/path.h"
  36 #include "qemu/help_option.h"
  37 #include "qemu/module.h"
  38 #include "qemu/plugin.h"
  39 #include "user/guest-base.h"
  40 #include "user/page-protection.h"
  41 #include "tcg/startup.h"
  42 #include "qemu/timer.h"
  43 #include "qemu/envlist.h"
  44 #include "qemu/cutils.h"
  45 #include "exec/log.h"
  46 #include "trace/control.h"
  47 #include "crypto/init.h"
  48 #include "qemu/guest-random.h"
  49 #include "gdbstub/user.h"
  50 #include "exec/page-vary.h"
  51
  52 #include "host-os.h"
  53 #include "target_arch_cpu.h"
  54
  55
  56 /*
  57  * TODO: Remove these and rely only on qemu_real_host_page_size().
  58  */
  59 uintptr_t qemu_host_page_size;
  60 intptr_t qemu_host_page_mask;
  61
  62 static bool opt_one_insn_per_tb;
  63 static unsigned long opt_tb_size;
  64 uintptr_t guest_base;
  65 bool have_guest_base;
  66 /*
  67  * When running 32-on-64 we should make sure we can fit all of the possible
  68  * guest address space into a contiguous chunk of virtual host memory.
  69  *
  70  * This way we will never overlap with our own libraries or binaries or stack
  71  * or anything else that QEMU maps.
  72  *
  73  * Many cpus reserve the high bit (or more than one for some 64-bit cpus)
  74  * of the address for the kernel.  Some cpus rely on this and user space
  75  * uses the high bit(s) for pointer tagging and the like.  For them, we
  76  * must preserve the expected address space.
  77  */
  78 #ifndef MAX_RESERVED_VA
  79 # if HOST_LONG_BITS > TARGET_VIRT_ADDR_SPACE_BITS
  80 #  if TARGET_VIRT_ADDR_SPACE_BITS == 32 && \
  81       (TARGET_LONG_BITS == 32 || defined(TARGET_ABI32))
  82 #   define MAX_RESERVED_VA(CPU)  0xfffffffful
  83 #  else
  84 #   define MAX_RESERVED_VA(CPU)  ((1ul << TARGET_VIRT_ADDR_SPACE_BITS) - 1)
  85 #  endif
  86 # else
  87 #  define MAX_RESERVED_VA(CPU)  0
  88 # endif
  89 #endif
  90
  91 unsigned long reserved_va;
  92 unsigned long guest_addr_max;
  93
  94 const char *interp_prefix = CONFIG_QEMU_INTERP_PREFIX;
  95 const char *qemu_uname_release;
  96
  97 unsigned long target_maxtsiz = TARGET_MAXTSIZ;   /* max text size */
  98 unsigned long target_dfldsiz = TARGET_DFLDSIZ;   /* initial data size limit */
  99 unsigned long target_maxdsiz = TARGET_MAXDSIZ;   /* max data size */
 100 unsigned long target_dflssiz = TARGET_DFLSSIZ;   /* initial data size limit */
 101 unsigned long target_maxssiz = TARGET_MAXSSIZ;   /* max stack size */
 102 unsigned long target_sgrowsiz = TARGET_SGROWSIZ; /* amount to grow stack */
 103
 104 /* Helper routines for implementing atomic operations. */
 105
 106 void fork_start(void)
 107 {
 108     start_exclusive();
 109     mmap_fork_start();
 110     cpu_list_lock();
 111     qemu_plugin_user_prefork_lock();
 112     gdbserver_fork_start();
 113 }
 114
 115 void fork_end(pid_t pid)
 116 {
 117     bool child = pid == 0;
 118
 119     qemu_plugin_user_postfork(child);
 120     mmap_fork_end(child);
 121     if (child) {
 122         CPUState *cpu, *next_cpu;
 123         /*
 124          * Child processes created by fork() only have a single thread.
 125          * Discard information about the parent threads.
 126          */
 127         CPU_FOREACH_SAFE(cpu, next_cpu) {
 128             if (cpu != thread_cpu) {
 129                 QTAILQ_REMOVE_RCU(&cpus_queue, cpu, node);
 130             }
 131         }
 132         qemu_init_cpu_list();
 133         get_task_state(thread_cpu)->ts_tid = qemu_get_thread_id();
 134     } else {
 135         cpu_list_unlock();
 136     }
 137     gdbserver_fork_end(thread_cpu, pid);
 138     /*
 139      * qemu_init_cpu_list() reinitialized the child exclusive state, but we
 140      * also need to keep current_cpu consistent, so call end_exclusive() for
 141      * both child and parent.
 142      */
 143     end_exclusive();
 144 }
 145
 146 void cpu_loop(CPUArchState *env)
 147 {
 148     target_cpu_loop(env);
 149 }
 150
 151 static void usage(void)
 152 {
 153     printf("qemu-" TARGET_NAME " version " QEMU_FULL_VERSION
 154            "\n" QEMU_COPYRIGHT "\n"
 155            "usage: qemu-" TARGET_NAME " [options] program [arguments...]\n"
 156            "BSD CPU emulator (compiled for %s emulation)\n"
 157            "\n"
 158            "Standard options:\n"
 159            "-h                print this help\n"
 160            "-g port           wait gdb connection to port\n"
 161            "-L path           set the elf interpreter prefix (default=%s)\n"
 162            "-s size           set the stack size in bytes (default=%ld)\n"
 163            "-cpu model        select CPU (-cpu help for list)\n"
 164            "-drop-ld-preload  drop LD_PRELOAD for target process\n"
 165            "-E var=value      sets/modifies targets environment variable(s)\n"
 166            "-U var            unsets targets environment variable(s)\n"
 167            "-B address        set guest_base address to address\n"
 168            "\n"
 169            "Debug options:\n"
 170            "-d item1[,...]    enable logging of specified items\n"
 171            "                  (use '-d help' for a list of log items)\n"
 172            "-D logfile        write logs to 'logfile' (default stderr)\n"
 173            "-one-insn-per-tb  run with one guest instruction per emulated TB\n"
 174            "-tb-size size     TCG translation block cache size\n"
 175            "-strace           log system calls\n"
 176            "-trace            [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
 177            "                  specify tracing options\n"
 178 #ifdef CONFIG_PLUGIN
 179            "-plugin           [file=]<file>[,<argname>=<argvalue>]\n"
 180 #endif
 181            "\n"
 182            "Environment variables:\n"
 183            "QEMU_STRACE       Print system calls and arguments similar to the\n"
 184            "                  'strace' program.  Enable by setting to any value.\n"
 185            "You can use -E and -U options to set/unset environment variables\n"
 186            "for target process.  It is possible to provide several variables\n"
 187            "by repeating the option.  For example:\n"
 188            "    -E var1=val2 -E var2=val2 -U LD_PRELOAD -U LD_DEBUG\n"
 189            "Note that if you provide several changes to single variable\n"
 190            "last change will stay in effect.\n"
 191            "\n"
 192            QEMU_HELP_BOTTOM "\n"
 193            ,
 194            TARGET_NAME,
 195            interp_prefix,
 196            target_dflssiz);
 197     exit(1);
 198 }
 199
 200 __thread CPUState *thread_cpu;
 201
 202 void stop_all_tasks(void)
 203 {
 204     /*
 205      * We trust when using NPTL (pthreads) start_exclusive() handles thread
 206      * stopping correctly.
 207      */
 208     start_exclusive();
 209 }
 210
 211 bool qemu_cpu_is_self(CPUState *cpu)
 212 {
 213     return thread_cpu == cpu;
 214 }
 215
 216 void qemu_cpu_kick(CPUState *cpu)
 217 {
 218     cpu_exit(cpu);
 219 }
 220
 221 /* Assumes contents are already zeroed.  */
 222 static void init_task_state(TaskState *ts)
 223 {
 224     ts->sigaltstack_used = (struct target_sigaltstack) {
 225         .ss_sp = 0,
 226         .ss_size = 0,
 227         .ss_flags = TARGET_SS_DISABLE,
 228     };
 229 }
 230
 231 static QemuPluginList plugins = QTAILQ_HEAD_INITIALIZER(plugins);
 232
 233 void gemu_log(const char *fmt, ...)
 234 {
 235     va_list ap;
 236
 237     va_start(ap, fmt);
 238     vfprintf(stderr, fmt, ap);
 239     va_end(ap);
 240 }
 241
 242 static void
 243 adjust_ssize(void)
 244 {
 245     struct rlimit rl;
 246
 247     if (getrlimit(RLIMIT_STACK, &rl) != 0) {
 248         return;
 249     }
 250
 251     target_maxssiz = MIN(target_maxssiz, rl.rlim_max);
 252     target_dflssiz = MIN(MAX(target_dflssiz, rl.rlim_cur), target_maxssiz);
 253
 254     rl.rlim_max = target_maxssiz;
 255     rl.rlim_cur = target_dflssiz;
 256     setrlimit(RLIMIT_STACK, &rl);
 257 }
 258
 259 int main(int argc, char **argv)
 260 {
 261     const char *filename;
 262     const char *cpu_model;
 263     const char *cpu_type;
 264     const char *log_file = NULL;
 265     const char *log_mask = NULL;
 266     const char *seed_optarg = NULL;
 267     struct target_pt_regs regs1, *regs = &regs1;
 268     struct image_info info1, *info = &info1;
 269     struct bsd_binprm bprm;
 270     TaskState *ts;
 271     CPUArchState *env;
 272     CPUState *cpu;
 273     int optind, rv;
 274     const char *r;
 275     const char *gdbstub = NULL;
 276     char **target_environ, **wrk;
 277     envlist_t *envlist = NULL;
 278     char *argv0 = NULL;
 279     int host_page_size;
 280     unsigned long max_reserved_va;
 281
 282     adjust_ssize();
 283
 284     if (argc <= 1) {
 285         usage();
 286     }
 287
 288
 289     error_init(argv[0]);
 290     module_call_init(MODULE_INIT_TRACE);
 291     qemu_init_cpu_list();
 292     module_call_init(MODULE_INIT_QOM);
 293
 294     envlist = envlist_create();
 295
 296     /*
 297      * add current environment into the list
 298      * envlist_setenv adds to the front of the list; to preserve environ
 299      * order add from back to front
 300      */
 301     for (wrk = environ; *wrk != NULL; wrk++) {
 302         continue;
 303     }
 304     while (wrk != environ) {
 305         wrk--;
 306         (void) envlist_setenv(envlist, *wrk);
 307     }
 308
 309     qemu_host_page_size = getpagesize();
 310     qemu_host_page_size = MAX(qemu_host_page_size, TARGET_PAGE_SIZE);
 311
 312     cpu_model = NULL;
 313
 314     qemu_add_opts(&qemu_trace_opts);
 315     qemu_plugin_add_opts();
 316
 317     optind = 1;
 318     for (;;) {
 319         if (optind >= argc) {
 320             break;
 321         }
 322         r = argv[optind];
 323         if (r[0] != '-') {
 324             break;
 325         }
 326         optind++;
 327         r++;
 328         if (!strcmp(r, "-")) {
 329             break;
 330         } else if (!strcmp(r, "d")) {
 331             if (optind >= argc) {
 332                 break;
 333             }
 334             log_mask = argv[optind++];
 335         } else if (!strcmp(r, "D")) {
 336             if (optind >= argc) {
 337                 break;
 338             }
 339             log_file = argv[optind++];
 340         } else if (!strcmp(r, "E")) {
 341             r = argv[optind++];
 342             if (envlist_setenv(envlist, r) != 0) {
 343                 usage();
 344             }
 345         } else if (!strcmp(r, "ignore-environment")) {
 346             envlist_free(envlist);
 347             envlist = envlist_create();
 348         } else if (!strcmp(r, "U")) {
 349             r = argv[optind++];
 350             if (envlist_unsetenv(envlist, r) != 0) {
 351                 usage();
 352             }
 353         } else if (!strcmp(r, "s")) {
 354             r = argv[optind++];
 355             rv = qemu_strtoul(r, &r, 0, &target_dflssiz);
 356             if (rv < 0 || target_dflssiz <= 0) {
 357                 usage();
 358             }
 359             if (*r == 'M') {
 360                 target_dflssiz *= 1024 * 1024;
 361             } else if (*r == 'k' || *r == 'K') {
 362                 target_dflssiz *= 1024;
 363             }
 364             if (target_dflssiz > target_maxssiz) {
 365                 usage();
 366             }
 367         } else if (!strcmp(r, "L")) {
 368             interp_prefix = argv[optind++];
 369         } else if (!strcmp(r, "p")) {
 370             unsigned size, want = qemu_real_host_page_size();
 371
 372             r = argv[optind++];
 373             if (qemu_strtoui(r, NULL, 10, &size) || size != want) {
 374                 warn_report("Deprecated page size option cannot "
 375                             "change host page size (%u)", want);
 376             }
 377         } else if (!strcmp(r, "g")) {
 378             gdbstub = g_strdup(argv[optind++]);
 379         } else if (!strcmp(r, "r")) {
 380             qemu_uname_release = argv[optind++];
 381         } else if (!strcmp(r, "cpu")) {
 382             cpu_model = argv[optind++];
 383             if (is_help_option(cpu_model)) {
 384                 list_cpus();
 385                 exit(1);
 386             }
 387         } else if (!strcmp(r, "B")) {
 388             rv = qemu_strtoul(argv[optind++], NULL, 0, &guest_base);
 389             if (rv < 0) {
 390                 usage();
 391             }
 392             have_guest_base = true;
 393         } else if (!strcmp(r, "drop-ld-preload")) {
 394             (void) envlist_unsetenv(envlist, "LD_PRELOAD");
 395         } else if (!strcmp(r, "seed")) {
 396             seed_optarg = optarg;
 397         } else if (!strcmp(r, "one-insn-per-tb")) {
 398             opt_one_insn_per_tb = true;
 399         } else if (!strcmp(r, "tb-size")) {
 400             r = argv[optind++];
 401             if (qemu_strtoul(r, NULL, 0, &opt_tb_size)) {
 402                 usage();
 403             }
 404         } else if (!strcmp(r, "strace")) {
 405             do_strace = 1;
 406         } else if (!strcmp(r, "trace")) {
 407             trace_opt_parse(optarg);
 408 #ifdef CONFIG_PLUGIN
 409         } else if (!strcmp(r, "plugin")) {
 410             r = argv[optind++];
 411             qemu_plugin_opt_parse(r, &plugins);
 412 #endif
 413         } else if (!strcmp(r, "0")) {
 414             argv0 = argv[optind++];
 415         } else {
 416             usage();
 417         }
 418     }
 419
 420     qemu_host_page_mask = -qemu_host_page_size;
 421
 422     /* init debug */
 423     {
 424         int mask = 0;
 425         if (log_mask) {
 426             mask = qemu_str_to_log_mask(log_mask);
 427             if (!mask) {
 428                 qemu_print_log_usage(stdout);
 429                 exit(1);
 430             }
 431         }
 432         qemu_set_log_filename_flags(log_file, mask, &error_fatal);
 433     }
 434
 435     if (optind >= argc) {
 436         usage();
 437     }
 438     filename = argv[optind];
 439     if (argv0) {
 440         argv[optind] = argv0;
 441     }
 442
 443     if (!trace_init_backends()) {
 444         exit(1);
 445     }
 446     trace_init_file();
 447     qemu_plugin_load_list(&plugins, &error_fatal);
 448
 449     /* Zero out regs */
 450     memset(regs, 0, sizeof(struct target_pt_regs));
 451
 452     /* Zero bsd params */
 453     memset(&bprm, 0, sizeof(bprm));
 454
 455     /* Zero out image_info */
 456     memset(info, 0, sizeof(struct image_info));
 457
 458     /* Scan interp_prefix dir for replacement files. */
 459     init_paths(interp_prefix);
 460
 461     if (cpu_model == NULL) {
 462         cpu_model = TARGET_DEFAULT_CPU_MODEL;
 463     }
 464
 465     cpu_type = parse_cpu_option(cpu_model);
 466
 467     /* init tcg before creating CPUs and to get qemu_host_page_size */
 468     {
 469         AccelState *accel = current_accel();
 470         AccelClass *ac = ACCEL_GET_CLASS(accel);
 471
 472         accel_init_interfaces(ac);
 473         object_property_set_bool(OBJECT(accel), "one-insn-per-tb",
 474                                  opt_one_insn_per_tb, &error_abort);
 475         object_property_set_int(OBJECT(accel), "tb-size",
 476                                 opt_tb_size, &error_abort);
 477         ac->init_machine(NULL);
 478     }
 479
 480     /*
 481      * Finalize page size before creating CPUs.
 482      * This will do nothing if !TARGET_PAGE_BITS_VARY.
 483      * The most efficient setting is to match the host.
 484      */
 485     host_page_size = qemu_real_host_page_size();
 486     set_preferred_target_page_bits(ctz32(host_page_size));
 487     finalize_target_page_bits();
 488
 489     cpu = cpu_create(cpu_type);
 490     env = cpu_env(cpu);
 491     cpu_reset(cpu);
 492     thread_cpu = cpu;
 493
 494     /*
 495      * Reserving too much vm space via mmap can run into problems with rlimits,
 496      * oom due to page table creation, etc.  We will still try it, if directed
 497      * by the command-line option, but not by default. Unless we're running a
 498      * target address space of 32 or fewer bits on a host with 64 bits.
 499      */
 500     max_reserved_va = MAX_RESERVED_VA(cpu);
 501     if (reserved_va != 0) {
 502         if ((reserved_va + 1) % host_page_size) {
 503             char *s = size_to_str(host_page_size);
 504             fprintf(stderr, "Reserved virtual address not aligned mod %s\n", s);
 505             g_free(s);
 506             exit(EXIT_FAILURE);
 507         }
 508         if (max_reserved_va && reserved_va > max_reserved_va) {
 509             fprintf(stderr, "Reserved virtual address too big\n");
 510             exit(EXIT_FAILURE);
 511         }
 512     } else if (HOST_LONG_BITS == 64 && TARGET_VIRT_ADDR_SPACE_BITS <= 32) {
 513         /* MAX_RESERVED_VA + 1 is a large power of 2, so is aligned. */
 514         reserved_va = max_reserved_va;
 515     }
 516     if (reserved_va != 0) {
 517         guest_addr_max = reserved_va;
 518     } else if (MIN(TARGET_VIRT_ADDR_SPACE_BITS, TARGET_ABI_BITS) <= 32) {
 519         guest_addr_max = UINT32_MAX;
 520     } else {
 521         guest_addr_max = ~0ul;
 522     }
 523
 524     if (getenv("QEMU_STRACE")) {
 525         do_strace = 1;
 526     }
 527
 528     target_environ = envlist_to_environ(envlist, NULL);
 529     envlist_free(envlist);
 530
 531     {
 532         Error *err = NULL;
 533         if (seed_optarg != NULL) {
 534             qemu_guest_random_seed_main(seed_optarg, &err);
 535         } else {
 536             qcrypto_init(&err);
 537         }
 538         if (err) {
 539             error_reportf_err(err, "cannot initialize crypto: ");
 540             exit(1);
 541         }
 542     }
 543
 544     /*
 545      * Now that page sizes are configured we can do
 546      * proper page alignment for guest_base.
 547      */
 548     if (have_guest_base) {
 549         if (guest_base & ~qemu_host_page_mask) {
 550             error_report("Selected guest base not host page aligned");
 551             exit(1);
 552         }
 553     }
 554
 555     /*
 556      * If reserving host virtual address space, do so now.
 557      * Combined with '-B', ensure that the chosen range is free.
 558      */
 559     if (reserved_va) {
 560         void *p;
 561
 562         if (have_guest_base) {
 563             p = mmap((void *)guest_base, reserved_va + 1, PROT_NONE,
 564                      MAP_ANON | MAP_PRIVATE | MAP_FIXED | MAP_EXCL, -1, 0);
 565         } else {
 566             p = mmap(NULL, reserved_va + 1, PROT_NONE,
 567                      MAP_ANON | MAP_PRIVATE, -1, 0);
 568         }
 569         if (p == MAP_FAILED) {
 570             const char *err = strerror(errno);
 571             char *sz = size_to_str(reserved_va + 1);
 572
 573             if (have_guest_base) {
 574                 error_report("Cannot allocate %s bytes at -B %p for guest "
 575                              "address space: %s", sz, (void *)guest_base, err);
 576             } else {
 577                 error_report("Cannot allocate %s bytes for guest "
 578                              "address space: %s", sz, err);
 579             }
 580             exit(1);
 581         }
 582         guest_base = (uintptr_t)p;
 583         have_guest_base = true;
 584
 585         /* Ensure that mmap_next_start is within range. */
 586         if (reserved_va <= mmap_next_start) {
 587             mmap_next_start = (reserved_va / 4 * 3)
 588                               & TARGET_PAGE_MASK & qemu_host_page_mask;
 589         }
 590     }
 591
 592     if (loader_exec(filename, argv + optind, target_environ, regs, info,
 593                     &bprm) != 0) {
 594         printf("Error loading %s\n", filename);
 595         _exit(1);
 596     }
 597
 598     for (wrk = target_environ; *wrk; wrk++) {
 599         g_free(*wrk);
 600     }
 601
 602     g_free(target_environ);
 603
 604     if (qemu_loglevel_mask(CPU_LOG_PAGE)) {
 605         FILE *f = qemu_log_trylock();
 606         if (f) {
 607             fprintf(f, "guest_base  %p\n", (void *)guest_base);
 608             fprintf(f, "page layout changed following binary load\n");
 609             page_dump(f);
 610
 611             fprintf(f, "end_code    0x" TARGET_ABI_FMT_lx "\n",
 612                     info->end_code);
 613             fprintf(f, "start_code  0x" TARGET_ABI_FMT_lx "\n",
 614                     info->start_code);
 615             fprintf(f, "start_data  0x" TARGET_ABI_FMT_lx "\n",
 616                     info->start_data);
 617             fprintf(f, "end_data    0x" TARGET_ABI_FMT_lx "\n",
 618                     info->end_data);
 619             fprintf(f, "start_stack 0x" TARGET_ABI_FMT_lx "\n",
 620                     info->start_stack);
 621             fprintf(f, "brk         0x" TARGET_ABI_FMT_lx "\n", info->brk);
 622             fprintf(f, "entry       0x" TARGET_ABI_FMT_lx "\n", info->entry);
 623
 624             qemu_log_unlock(f);
 625         }
 626     }
 627
 628     /* build Task State */
 629     ts = g_new0(TaskState, 1);
 630     init_task_state(ts);
 631     ts->info = info;
 632     ts->bprm = &bprm;
 633     ts->ts_tid = qemu_get_thread_id();
 634     cpu->opaque = ts;
 635
 636     target_set_brk(info->brk);
 637     syscall_init();
 638     signal_init();
 639
 640     /*
 641      * Now that we've loaded the binary, GUEST_BASE is fixed.  Delay
 642      * generating the prologue until now so that the prologue can take
 643      * the real value of GUEST_BASE into account.
 644      */
 645     tcg_prologue_init();
 646
 647     target_cpu_init(env, regs);
 648
 649     if (gdbstub) {
 650         gdbserver_start(gdbstub, &error_fatal);
 651     }
 652     cpu_loop(env);
 653     /* never exits */
 654     return 0;
 655 }