xref: /openbmc/qemu/linux-user/main.c (revision 51e47cf8)
1 /*
2  *  qemu user main
3  *
4  *  Copyright (c) 2003-2008 Fabrice Bellard
5  *
6  *  This program is free software; you can redistribute it and/or modify
7  *  it under the terms of the GNU General Public License as published by
8  *  the Free Software Foundation; either version 2 of the License, or
9  *  (at your option) any later version.
10  *
11  *  This program 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
14  *  GNU General Public License for more details.
15  *
16  *  You should have received a copy of the GNU General Public License
17  *  along with this program; if not, see <http://www.gnu.org/licenses/>.
18  */
19 
20 #include "qemu/osdep.h"
21 #include "qemu/help-texts.h"
22 #include "qemu/units.h"
23 #include "qemu/accel.h"
24 #include "qemu-version.h"
25 #include <sys/syscall.h>
26 #include <sys/resource.h>
27 #include <sys/shm.h>
28 #include <linux/binfmts.h>
29 
30 #include "qapi/error.h"
31 #include "qemu.h"
32 #include "user-internals.h"
33 #include "qemu/path.h"
34 #include "qemu/queue.h"
35 #include "qemu/config-file.h"
36 #include "qemu/cutils.h"
37 #include "qemu/error-report.h"
38 #include "qemu/help_option.h"
39 #include "qemu/module.h"
40 #include "qemu/plugin.h"
41 #include "exec/exec-all.h"
42 #include "exec/gdbstub.h"
43 #include "gdbstub/user.h"
44 #include "tcg/tcg.h"
45 #include "qemu/timer.h"
46 #include "qemu/envlist.h"
47 #include "qemu/guest-random.h"
48 #include "elf.h"
49 #include "trace/control.h"
50 #include "target_elf.h"
51 #include "cpu_loop-common.h"
52 #include "crypto/init.h"
53 #include "fd-trans.h"
54 #include "signal-common.h"
55 #include "loader.h"
56 #include "user-mmap.h"
57 #include "accel/tcg/perf.h"
58 
59 #ifdef CONFIG_SEMIHOSTING
60 #include "semihosting/semihost.h"
61 #endif
62 
63 #ifndef AT_FLAGS_PRESERVE_ARGV0
64 #define AT_FLAGS_PRESERVE_ARGV0_BIT 0
65 #define AT_FLAGS_PRESERVE_ARGV0 (1 << AT_FLAGS_PRESERVE_ARGV0_BIT)
66 #endif
67 
68 char *exec_path;
69 char real_exec_path[PATH_MAX];
70 
71 static bool opt_one_insn_per_tb;
72 static const char *argv0;
73 static const char *gdbstub;
74 static envlist_t *envlist;
75 static const char *cpu_model;
76 static const char *cpu_type;
77 static const char *seed_optarg;
78 unsigned long mmap_min_addr;
79 uintptr_t guest_base;
80 bool have_guest_base;
81 
82 /*
83  * Used to implement backwards-compatibility for the `-strace`, and
84  * QEMU_STRACE options. Without this, the QEMU_LOG can be overwritten by
85  * -strace, or vice versa.
86  */
87 static bool enable_strace;
88 
89 /*
90  * The last log mask given by the user in an environment variable or argument.
91  * Used to support command line arguments overriding environment variables.
92  */
93 static int last_log_mask;
94 static const char *last_log_filename;
95 
96 /*
97  * When running 32-on-64 we should make sure we can fit all of the possible
98  * guest address space into a contiguous chunk of virtual host memory.
99  *
100  * This way we will never overlap with our own libraries or binaries or stack
101  * or anything else that QEMU maps.
102  *
103  * Many cpus reserve the high bit (or more than one for some 64-bit cpus)
104  * of the address for the kernel.  Some cpus rely on this and user space
105  * uses the high bit(s) for pointer tagging and the like.  For them, we
106  * must preserve the expected address space.
107  */
108 #ifndef MAX_RESERVED_VA
109 # if HOST_LONG_BITS > TARGET_VIRT_ADDR_SPACE_BITS
110 #  if TARGET_VIRT_ADDR_SPACE_BITS == 32 && \
111       (TARGET_LONG_BITS == 32 || defined(TARGET_ABI32))
112 #   define MAX_RESERVED_VA(CPU)  0xfffffffful
113 #  else
114 #   define MAX_RESERVED_VA(CPU)  ((1ul << TARGET_VIRT_ADDR_SPACE_BITS) - 1)
115 #  endif
116 # else
117 #  define MAX_RESERVED_VA(CPU)  0
118 # endif
119 #endif
120 
121 unsigned long reserved_va;
122 
123 static void usage(int exitcode);
124 
125 static const char *interp_prefix = CONFIG_QEMU_INTERP_PREFIX;
126 const char *qemu_uname_release;
127 
128 #if !defined(TARGET_DEFAULT_STACK_SIZE)
129 /* XXX: on x86 MAP_GROWSDOWN only works if ESP <= address + 32, so
130    we allocate a bigger stack. Need a better solution, for example
131    by remapping the process stack directly at the right place */
132 #define TARGET_DEFAULT_STACK_SIZE	8 * 1024 * 1024UL
133 #endif
134 
135 unsigned long guest_stack_size = TARGET_DEFAULT_STACK_SIZE;
136 
137 /***********************************************************/
138 /* Helper routines for implementing atomic operations.  */
139 
140 /* Make sure everything is in a consistent state for calling fork().  */
141 void fork_start(void)
142 {
143     start_exclusive();
144     mmap_fork_start();
145     cpu_list_lock();
146     qemu_plugin_user_prefork_lock();
147 }
148 
149 void fork_end(int child)
150 {
151     qemu_plugin_user_postfork(child);
152     mmap_fork_end(child);
153     if (child) {
154         CPUState *cpu, *next_cpu;
155         /* Child processes created by fork() only have a single thread.
156            Discard information about the parent threads.  */
157         CPU_FOREACH_SAFE(cpu, next_cpu) {
158             if (cpu != thread_cpu) {
159                 QTAILQ_REMOVE_RCU(&cpus, cpu, node);
160             }
161         }
162         qemu_init_cpu_list();
163         gdbserver_fork(thread_cpu);
164     } else {
165         cpu_list_unlock();
166     }
167     /*
168      * qemu_init_cpu_list() reinitialized the child exclusive state, but we
169      * also need to keep current_cpu consistent, so call end_exclusive() for
170      * both child and parent.
171      */
172     end_exclusive();
173 }
174 
175 __thread CPUState *thread_cpu;
176 
177 bool qemu_cpu_is_self(CPUState *cpu)
178 {
179     return thread_cpu == cpu;
180 }
181 
182 void qemu_cpu_kick(CPUState *cpu)
183 {
184     cpu_exit(cpu);
185 }
186 
187 void task_settid(TaskState *ts)
188 {
189     if (ts->ts_tid == 0) {
190         ts->ts_tid = (pid_t)syscall(SYS_gettid);
191     }
192 }
193 
194 void stop_all_tasks(void)
195 {
196     /*
197      * We trust that when using NPTL, start_exclusive()
198      * handles thread stopping correctly.
199      */
200     start_exclusive();
201 }
202 
203 /* Assumes contents are already zeroed.  */
204 void init_task_state(TaskState *ts)
205 {
206     long ticks_per_sec;
207     struct timespec bt;
208 
209     ts->used = 1;
210     ts->sigaltstack_used = (struct target_sigaltstack) {
211         .ss_sp = 0,
212         .ss_size = 0,
213         .ss_flags = TARGET_SS_DISABLE,
214     };
215 
216     /* Capture task start time relative to system boot */
217 
218     ticks_per_sec = sysconf(_SC_CLK_TCK);
219 
220     if ((ticks_per_sec > 0) && !clock_gettime(CLOCK_BOOTTIME, &bt)) {
221         /* start_boottime is expressed in clock ticks */
222         ts->start_boottime = bt.tv_sec * (uint64_t) ticks_per_sec;
223         ts->start_boottime += bt.tv_nsec * (uint64_t) ticks_per_sec /
224                               NANOSECONDS_PER_SECOND;
225     }
226 }
227 
228 CPUArchState *cpu_copy(CPUArchState *env)
229 {
230     CPUState *cpu = env_cpu(env);
231     CPUState *new_cpu = cpu_create(cpu_type);
232     CPUArchState *new_env = new_cpu->env_ptr;
233     CPUBreakpoint *bp;
234 
235     /* Reset non arch specific state */
236     cpu_reset(new_cpu);
237 
238     new_cpu->tcg_cflags = cpu->tcg_cflags;
239     memcpy(new_env, env, sizeof(CPUArchState));
240 #if defined(TARGET_I386) || defined(TARGET_X86_64)
241     new_env->gdt.base = target_mmap(0, sizeof(uint64_t) * TARGET_GDT_ENTRIES,
242                                     PROT_READ | PROT_WRITE,
243                                     MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
244     memcpy(g2h_untagged(new_env->gdt.base), g2h_untagged(env->gdt.base),
245            sizeof(uint64_t) * TARGET_GDT_ENTRIES);
246     OBJECT(new_cpu)->free = OBJECT(cpu)->free;
247 #endif
248 
249     /* Clone all break/watchpoints.
250        Note: Once we support ptrace with hw-debug register access, make sure
251        BP_CPU break/watchpoints are handled correctly on clone. */
252     QTAILQ_INIT(&new_cpu->breakpoints);
253     QTAILQ_FOREACH(bp, &cpu->breakpoints, entry) {
254         cpu_breakpoint_insert(new_cpu, bp->pc, bp->flags, NULL);
255     }
256 
257     return new_env;
258 }
259 
260 static void handle_arg_help(const char *arg)
261 {
262     usage(EXIT_SUCCESS);
263 }
264 
265 static void handle_arg_log(const char *arg)
266 {
267     last_log_mask = qemu_str_to_log_mask(arg);
268     if (!last_log_mask) {
269         qemu_print_log_usage(stdout);
270         exit(EXIT_FAILURE);
271     }
272 }
273 
274 static void handle_arg_dfilter(const char *arg)
275 {
276     qemu_set_dfilter_ranges(arg, &error_fatal);
277 }
278 
279 static void handle_arg_log_filename(const char *arg)
280 {
281     last_log_filename = arg;
282 }
283 
284 static void handle_arg_set_env(const char *arg)
285 {
286     char *r, *p, *token;
287     r = p = strdup(arg);
288     while ((token = strsep(&p, ",")) != NULL) {
289         if (envlist_setenv(envlist, token) != 0) {
290             usage(EXIT_FAILURE);
291         }
292     }
293     free(r);
294 }
295 
296 static void handle_arg_unset_env(const char *arg)
297 {
298     char *r, *p, *token;
299     r = p = strdup(arg);
300     while ((token = strsep(&p, ",")) != NULL) {
301         if (envlist_unsetenv(envlist, token) != 0) {
302             usage(EXIT_FAILURE);
303         }
304     }
305     free(r);
306 }
307 
308 static void handle_arg_argv0(const char *arg)
309 {
310     argv0 = strdup(arg);
311 }
312 
313 static void handle_arg_stack_size(const char *arg)
314 {
315     char *p;
316     guest_stack_size = strtoul(arg, &p, 0);
317     if (guest_stack_size == 0) {
318         usage(EXIT_FAILURE);
319     }
320 
321     if (*p == 'M') {
322         guest_stack_size *= MiB;
323     } else if (*p == 'k' || *p == 'K') {
324         guest_stack_size *= KiB;
325     }
326 }
327 
328 static void handle_arg_ld_prefix(const char *arg)
329 {
330     interp_prefix = strdup(arg);
331 }
332 
333 static void handle_arg_pagesize(const char *arg)
334 {
335     qemu_host_page_size = atoi(arg);
336     if (qemu_host_page_size == 0 ||
337         (qemu_host_page_size & (qemu_host_page_size - 1)) != 0) {
338         fprintf(stderr, "page size must be a power of two\n");
339         exit(EXIT_FAILURE);
340     }
341 }
342 
343 static void handle_arg_seed(const char *arg)
344 {
345     seed_optarg = arg;
346 }
347 
348 static void handle_arg_gdb(const char *arg)
349 {
350     gdbstub = g_strdup(arg);
351 }
352 
353 static void handle_arg_uname(const char *arg)
354 {
355     qemu_uname_release = strdup(arg);
356 }
357 
358 static void handle_arg_cpu(const char *arg)
359 {
360     cpu_model = strdup(arg);
361     if (cpu_model == NULL || is_help_option(cpu_model)) {
362         /* XXX: implement xxx_cpu_list for targets that still miss it */
363 #if defined(cpu_list)
364         cpu_list();
365 #endif
366         exit(EXIT_FAILURE);
367     }
368 }
369 
370 static void handle_arg_guest_base(const char *arg)
371 {
372     guest_base = strtol(arg, NULL, 0);
373     have_guest_base = true;
374 }
375 
376 static void handle_arg_reserved_va(const char *arg)
377 {
378     char *p;
379     int shift = 0;
380     unsigned long val;
381 
382     val = strtoul(arg, &p, 0);
383     switch (*p) {
384     case 'k':
385     case 'K':
386         shift = 10;
387         break;
388     case 'M':
389         shift = 20;
390         break;
391     case 'G':
392         shift = 30;
393         break;
394     }
395     if (shift) {
396         unsigned long unshifted = val;
397         p++;
398         val <<= shift;
399         if (val >> shift != unshifted) {
400             fprintf(stderr, "Reserved virtual address too big\n");
401             exit(EXIT_FAILURE);
402         }
403     }
404     if (*p) {
405         fprintf(stderr, "Unrecognised -R size suffix '%s'\n", p);
406         exit(EXIT_FAILURE);
407     }
408     /* The representation is size - 1, with 0 remaining "default". */
409     reserved_va = val ? val - 1 : 0;
410 }
411 
412 static void handle_arg_one_insn_per_tb(const char *arg)
413 {
414     opt_one_insn_per_tb = true;
415 }
416 
417 static void handle_arg_strace(const char *arg)
418 {
419     enable_strace = true;
420 }
421 
422 static void handle_arg_version(const char *arg)
423 {
424     printf("qemu-" TARGET_NAME " version " QEMU_FULL_VERSION
425            "\n" QEMU_COPYRIGHT "\n");
426     exit(EXIT_SUCCESS);
427 }
428 
429 static void handle_arg_trace(const char *arg)
430 {
431     trace_opt_parse(arg);
432 }
433 
434 #if defined(TARGET_XTENSA)
435 static void handle_arg_abi_call0(const char *arg)
436 {
437     xtensa_set_abi_call0();
438 }
439 #endif
440 
441 static void handle_arg_perfmap(const char *arg)
442 {
443     perf_enable_perfmap();
444 }
445 
446 static void handle_arg_jitdump(const char *arg)
447 {
448     perf_enable_jitdump();
449 }
450 
451 static QemuPluginList plugins = QTAILQ_HEAD_INITIALIZER(plugins);
452 
453 #ifdef CONFIG_PLUGIN
454 static void handle_arg_plugin(const char *arg)
455 {
456     qemu_plugin_opt_parse(arg, &plugins);
457 }
458 #endif
459 
460 struct qemu_argument {
461     const char *argv;
462     const char *env;
463     bool has_arg;
464     void (*handle_opt)(const char *arg);
465     const char *example;
466     const char *help;
467 };
468 
469 static const struct qemu_argument arg_table[] = {
470     {"h",          "",                 false, handle_arg_help,
471      "",           "print this help"},
472     {"help",       "",                 false, handle_arg_help,
473      "",           ""},
474     {"g",          "QEMU_GDB",         true,  handle_arg_gdb,
475      "port",       "wait gdb connection to 'port'"},
476     {"L",          "QEMU_LD_PREFIX",   true,  handle_arg_ld_prefix,
477      "path",       "set the elf interpreter prefix to 'path'"},
478     {"s",          "QEMU_STACK_SIZE",  true,  handle_arg_stack_size,
479      "size",       "set the stack size to 'size' bytes"},
480     {"cpu",        "QEMU_CPU",         true,  handle_arg_cpu,
481      "model",      "select CPU (-cpu help for list)"},
482     {"E",          "QEMU_SET_ENV",     true,  handle_arg_set_env,
483      "var=value",  "sets targets environment variable (see below)"},
484     {"U",          "QEMU_UNSET_ENV",   true,  handle_arg_unset_env,
485      "var",        "unsets targets environment variable (see below)"},
486     {"0",          "QEMU_ARGV0",       true,  handle_arg_argv0,
487      "argv0",      "forces target process argv[0] to be 'argv0'"},
488     {"r",          "QEMU_UNAME",       true,  handle_arg_uname,
489      "uname",      "set qemu uname release string to 'uname'"},
490     {"B",          "QEMU_GUEST_BASE",  true,  handle_arg_guest_base,
491      "address",    "set guest_base address to 'address'"},
492     {"R",          "QEMU_RESERVED_VA", true,  handle_arg_reserved_va,
493      "size",       "reserve 'size' bytes for guest virtual address space"},
494     {"d",          "QEMU_LOG",         true,  handle_arg_log,
495      "item[,...]", "enable logging of specified items "
496      "(use '-d help' for a list of items)"},
497     {"dfilter",    "QEMU_DFILTER",     true,  handle_arg_dfilter,
498      "range[,...]","filter logging based on address range"},
499     {"D",          "QEMU_LOG_FILENAME", true, handle_arg_log_filename,
500      "logfile",     "write logs to 'logfile' (default stderr)"},
501     {"p",          "QEMU_PAGESIZE",    true,  handle_arg_pagesize,
502      "pagesize",   "set the host page size to 'pagesize'"},
503     {"one-insn-per-tb",
504                    "QEMU_ONE_INSN_PER_TB",  false, handle_arg_one_insn_per_tb,
505      "",           "run with one guest instruction per emulated TB"},
506     {"singlestep", "QEMU_SINGLESTEP",  false, handle_arg_one_insn_per_tb,
507      "",           "deprecated synonym for -one-insn-per-tb"},
508     {"strace",     "QEMU_STRACE",      false, handle_arg_strace,
509      "",           "log system calls"},
510     {"seed",       "QEMU_RAND_SEED",   true,  handle_arg_seed,
511      "",           "Seed for pseudo-random number generator"},
512     {"trace",      "QEMU_TRACE",       true,  handle_arg_trace,
513      "",           "[[enable=]<pattern>][,events=<file>][,file=<file>]"},
514 #ifdef CONFIG_PLUGIN
515     {"plugin",     "QEMU_PLUGIN",      true,  handle_arg_plugin,
516      "",           "[file=]<file>[,<argname>=<argvalue>]"},
517 #endif
518     {"version",    "QEMU_VERSION",     false, handle_arg_version,
519      "",           "display version information and exit"},
520 #if defined(TARGET_XTENSA)
521     {"xtensa-abi-call0", "QEMU_XTENSA_ABI_CALL0", false, handle_arg_abi_call0,
522      "",           "assume CALL0 Xtensa ABI"},
523 #endif
524     {"perfmap",    "QEMU_PERFMAP",     false, handle_arg_perfmap,
525      "",           "Generate a /tmp/perf-${pid}.map file for perf"},
526     {"jitdump",    "QEMU_JITDUMP",     false, handle_arg_jitdump,
527      "",           "Generate a jit-${pid}.dump file for perf"},
528     {NULL, NULL, false, NULL, NULL, NULL}
529 };
530 
531 static void usage(int exitcode)
532 {
533     const struct qemu_argument *arginfo;
534     int maxarglen;
535     int maxenvlen;
536 
537     printf("usage: qemu-" TARGET_NAME " [options] program [arguments...]\n"
538            "Linux CPU emulator (compiled for " TARGET_NAME " emulation)\n"
539            "\n"
540            "Options and associated environment variables:\n"
541            "\n");
542 
543     /* Calculate column widths. We must always have at least enough space
544      * for the column header.
545      */
546     maxarglen = strlen("Argument");
547     maxenvlen = strlen("Env-variable");
548 
549     for (arginfo = arg_table; arginfo->handle_opt != NULL; arginfo++) {
550         int arglen = strlen(arginfo->argv);
551         if (arginfo->has_arg) {
552             arglen += strlen(arginfo->example) + 1;
553         }
554         if (strlen(arginfo->env) > maxenvlen) {
555             maxenvlen = strlen(arginfo->env);
556         }
557         if (arglen > maxarglen) {
558             maxarglen = arglen;
559         }
560     }
561 
562     printf("%-*s %-*s Description\n", maxarglen+1, "Argument",
563             maxenvlen, "Env-variable");
564 
565     for (arginfo = arg_table; arginfo->handle_opt != NULL; arginfo++) {
566         if (arginfo->has_arg) {
567             printf("-%s %-*s %-*s %s\n", arginfo->argv,
568                    (int)(maxarglen - strlen(arginfo->argv) - 1),
569                    arginfo->example, maxenvlen, arginfo->env, arginfo->help);
570         } else {
571             printf("-%-*s %-*s %s\n", maxarglen, arginfo->argv,
572                     maxenvlen, arginfo->env,
573                     arginfo->help);
574         }
575     }
576 
577     printf("\n"
578            "Defaults:\n"
579            "QEMU_LD_PREFIX  = %s\n"
580            "QEMU_STACK_SIZE = %ld byte\n",
581            interp_prefix,
582            guest_stack_size);
583 
584     printf("\n"
585            "You can use -E and -U options or the QEMU_SET_ENV and\n"
586            "QEMU_UNSET_ENV environment variables to set and unset\n"
587            "environment variables for the target process.\n"
588            "It is possible to provide several variables by separating them\n"
589            "by commas in getsubopt(3) style. Additionally it is possible to\n"
590            "provide the -E and -U options multiple times.\n"
591            "The following lines are equivalent:\n"
592            "    -E var1=val2 -E var2=val2 -U LD_PRELOAD -U LD_DEBUG\n"
593            "    -E var1=val2,var2=val2 -U LD_PRELOAD,LD_DEBUG\n"
594            "    QEMU_SET_ENV=var1=val2,var2=val2 QEMU_UNSET_ENV=LD_PRELOAD,LD_DEBUG\n"
595            "Note that if you provide several changes to a single variable\n"
596            "the last change will stay in effect.\n"
597            "\n"
598            QEMU_HELP_BOTTOM "\n");
599 
600     exit(exitcode);
601 }
602 
603 static int parse_args(int argc, char **argv)
604 {
605     const char *r;
606     int optind;
607     const struct qemu_argument *arginfo;
608 
609     for (arginfo = arg_table; arginfo->handle_opt != NULL; arginfo++) {
610         if (arginfo->env == NULL) {
611             continue;
612         }
613 
614         r = getenv(arginfo->env);
615         if (r != NULL) {
616             arginfo->handle_opt(r);
617         }
618     }
619 
620     optind = 1;
621     for (;;) {
622         if (optind >= argc) {
623             break;
624         }
625         r = argv[optind];
626         if (r[0] != '-') {
627             break;
628         }
629         optind++;
630         r++;
631         if (!strcmp(r, "-")) {
632             break;
633         }
634         /* Treat --foo the same as -foo.  */
635         if (r[0] == '-') {
636             r++;
637         }
638 
639         for (arginfo = arg_table; arginfo->handle_opt != NULL; arginfo++) {
640             if (!strcmp(r, arginfo->argv)) {
641                 if (arginfo->has_arg) {
642                     if (optind >= argc) {
643                         (void) fprintf(stderr,
644                             "qemu: missing argument for option '%s'\n", r);
645                         exit(EXIT_FAILURE);
646                     }
647                     arginfo->handle_opt(argv[optind]);
648                     optind++;
649                 } else {
650                     arginfo->handle_opt(NULL);
651                 }
652                 break;
653             }
654         }
655 
656         /* no option matched the current argv */
657         if (arginfo->handle_opt == NULL) {
658             (void) fprintf(stderr, "qemu: unknown option '%s'\n", r);
659             exit(EXIT_FAILURE);
660         }
661     }
662 
663     if (optind >= argc) {
664         (void) fprintf(stderr, "qemu: no user program specified\n");
665         exit(EXIT_FAILURE);
666     }
667 
668     exec_path = argv[optind];
669 
670     return optind;
671 }
672 
673 int main(int argc, char **argv, char **envp)
674 {
675     struct target_pt_regs regs1, *regs = &regs1;
676     struct image_info info1, *info = &info1;
677     struct linux_binprm bprm;
678     TaskState *ts;
679     CPUArchState *env;
680     CPUState *cpu;
681     int optind;
682     char **target_environ, **wrk;
683     char **target_argv;
684     int target_argc;
685     int i;
686     int ret;
687     int execfd;
688     unsigned long max_reserved_va;
689     bool preserve_argv0;
690 
691     error_init(argv[0]);
692     module_call_init(MODULE_INIT_TRACE);
693     qemu_init_cpu_list();
694     module_call_init(MODULE_INIT_QOM);
695 
696     envlist = envlist_create();
697 
698     /* add current environment into the list */
699     for (wrk = environ; *wrk != NULL; wrk++) {
700         (void) envlist_setenv(envlist, *wrk);
701     }
702 
703     /* Read the stack limit from the kernel.  If it's "unlimited",
704        then we can do little else besides use the default.  */
705     {
706         struct rlimit lim;
707         if (getrlimit(RLIMIT_STACK, &lim) == 0
708             && lim.rlim_cur != RLIM_INFINITY
709             && lim.rlim_cur == (target_long)lim.rlim_cur
710             && lim.rlim_cur > guest_stack_size) {
711             guest_stack_size = lim.rlim_cur;
712         }
713     }
714 
715     cpu_model = NULL;
716 
717     qemu_add_opts(&qemu_trace_opts);
718     qemu_plugin_add_opts();
719 
720     optind = parse_args(argc, argv);
721 
722     qemu_set_log_filename_flags(last_log_filename,
723                                 last_log_mask | (enable_strace * LOG_STRACE),
724                                 &error_fatal);
725 
726     if (!trace_init_backends()) {
727         exit(1);
728     }
729     trace_init_file();
730     qemu_plugin_load_list(&plugins, &error_fatal);
731 
732     /* Zero out regs */
733     memset(regs, 0, sizeof(struct target_pt_regs));
734 
735     /* Zero out image_info */
736     memset(info, 0, sizeof(struct image_info));
737 
738     memset(&bprm, 0, sizeof (bprm));
739 
740     /* Scan interp_prefix dir for replacement files. */
741     init_paths(interp_prefix);
742 
743     init_qemu_uname_release();
744 
745     /*
746      * Manage binfmt-misc open-binary flag
747      */
748     execfd = qemu_getauxval(AT_EXECFD);
749     if (execfd == 0) {
750         execfd = open(exec_path, O_RDONLY);
751         if (execfd < 0) {
752             printf("Error while loading %s: %s\n", exec_path, strerror(errno));
753             _exit(EXIT_FAILURE);
754         }
755     }
756 
757     /* Resolve executable file name to full path name */
758     if (realpath(exec_path, real_exec_path)) {
759         exec_path = real_exec_path;
760     }
761 
762     /*
763      * get binfmt_misc flags
764      */
765     preserve_argv0 = !!(qemu_getauxval(AT_FLAGS) & AT_FLAGS_PRESERVE_ARGV0);
766 
767     /*
768      * Manage binfmt-misc preserve-arg[0] flag
769      *    argv[optind]     full path to the binary
770      *    argv[optind + 1] original argv[0]
771      */
772     if (optind + 1 < argc && preserve_argv0) {
773         optind++;
774     }
775 
776     if (cpu_model == NULL) {
777         cpu_model = cpu_get_model(get_elf_eflags(execfd));
778     }
779     cpu_type = parse_cpu_option(cpu_model);
780 
781     /* init tcg before creating CPUs and to get qemu_host_page_size */
782     {
783         AccelState *accel = current_accel();
784         AccelClass *ac = ACCEL_GET_CLASS(accel);
785 
786         accel_init_interfaces(ac);
787         object_property_set_bool(OBJECT(accel), "one-insn-per-tb",
788                                  opt_one_insn_per_tb, &error_abort);
789         ac->init_machine(NULL);
790     }
791     cpu = cpu_create(cpu_type);
792     env = cpu->env_ptr;
793     cpu_reset(cpu);
794     thread_cpu = cpu;
795 
796     /*
797      * Reserving too much vm space via mmap can run into problems
798      * with rlimits, oom due to page table creation, etc.  We will
799      * still try it, if directed by the command-line option, but
800      * not by default.
801      */
802     max_reserved_va = MAX_RESERVED_VA(cpu);
803     if (reserved_va != 0) {
804         if ((reserved_va + 1) % qemu_host_page_size) {
805             char *s = size_to_str(qemu_host_page_size);
806             fprintf(stderr, "Reserved virtual address not aligned mod %s\n", s);
807             g_free(s);
808             exit(EXIT_FAILURE);
809         }
810         if (max_reserved_va && reserved_va > max_reserved_va) {
811             fprintf(stderr, "Reserved virtual address too big\n");
812             exit(EXIT_FAILURE);
813         }
814     } else if (HOST_LONG_BITS == 64 && TARGET_VIRT_ADDR_SPACE_BITS <= 32) {
815         /* MAX_RESERVED_VA + 1 is a large power of 2, so is aligned. */
816         reserved_va = max_reserved_va;
817     }
818 
819     {
820         Error *err = NULL;
821         if (seed_optarg != NULL) {
822             qemu_guest_random_seed_main(seed_optarg, &err);
823         } else {
824             qcrypto_init(&err);
825         }
826         if (err) {
827             error_reportf_err(err, "cannot initialize crypto: ");
828             exit(1);
829         }
830     }
831 
832     target_environ = envlist_to_environ(envlist, NULL);
833     envlist_free(envlist);
834 
835     /*
836      * Read in mmap_min_addr kernel parameter.  This value is used
837      * When loading the ELF image to determine whether guest_base
838      * is needed.  It is also used in mmap_find_vma.
839      */
840     {
841         FILE *fp;
842 
843         if ((fp = fopen("/proc/sys/vm/mmap_min_addr", "r")) != NULL) {
844             unsigned long tmp;
845             if (fscanf(fp, "%lu", &tmp) == 1 && tmp != 0) {
846                 mmap_min_addr = tmp;
847                 qemu_log_mask(CPU_LOG_PAGE, "host mmap_min_addr=0x%lx\n",
848                               mmap_min_addr);
849             }
850             fclose(fp);
851         }
852     }
853 
854     /*
855      * We prefer to not make NULL pointers accessible to QEMU.
856      * If we're in a chroot with no /proc, fall back to 1 page.
857      */
858     if (mmap_min_addr == 0) {
859         mmap_min_addr = qemu_host_page_size;
860         qemu_log_mask(CPU_LOG_PAGE,
861                       "host mmap_min_addr=0x%lx (fallback)\n",
862                       mmap_min_addr);
863     }
864 
865     /*
866      * Prepare copy of argv vector for target.
867      */
868     target_argc = argc - optind;
869     target_argv = calloc(target_argc + 1, sizeof (char *));
870     if (target_argv == NULL) {
871         (void) fprintf(stderr, "Unable to allocate memory for target_argv\n");
872         exit(EXIT_FAILURE);
873     }
874 
875     /*
876      * If argv0 is specified (using '-0' switch) we replace
877      * argv[0] pointer with the given one.
878      */
879     i = 0;
880     if (argv0 != NULL) {
881         target_argv[i++] = strdup(argv0);
882     }
883     for (; i < target_argc; i++) {
884         target_argv[i] = strdup(argv[optind + i]);
885     }
886     target_argv[target_argc] = NULL;
887 
888     ts = g_new0(TaskState, 1);
889     init_task_state(ts);
890     /* build Task State */
891     ts->info = info;
892     ts->bprm = &bprm;
893     cpu->opaque = ts;
894     task_settid(ts);
895 
896     fd_trans_init();
897 
898     ret = loader_exec(execfd, exec_path, target_argv, target_environ, regs,
899         info, &bprm);
900     if (ret != 0) {
901         printf("Error while loading %s: %s\n", exec_path, strerror(-ret));
902         _exit(EXIT_FAILURE);
903     }
904 
905     for (wrk = target_environ; *wrk; wrk++) {
906         g_free(*wrk);
907     }
908 
909     g_free(target_environ);
910 
911     if (qemu_loglevel_mask(CPU_LOG_PAGE)) {
912         FILE *f = qemu_log_trylock();
913         if (f) {
914             fprintf(f, "guest_base  %p\n", (void *)guest_base);
915             fprintf(f, "page layout changed following binary load\n");
916             page_dump(f);
917 
918             fprintf(f, "start_brk   0x" TARGET_ABI_FMT_lx "\n",
919                     info->start_brk);
920             fprintf(f, "end_code    0x" TARGET_ABI_FMT_lx "\n",
921                     info->end_code);
922             fprintf(f, "start_code  0x" TARGET_ABI_FMT_lx "\n",
923                     info->start_code);
924             fprintf(f, "start_data  0x" TARGET_ABI_FMT_lx "\n",
925                     info->start_data);
926             fprintf(f, "end_data    0x" TARGET_ABI_FMT_lx "\n",
927                     info->end_data);
928             fprintf(f, "start_stack 0x" TARGET_ABI_FMT_lx "\n",
929                     info->start_stack);
930             fprintf(f, "brk         0x" TARGET_ABI_FMT_lx "\n",
931                     info->brk);
932             fprintf(f, "entry       0x" TARGET_ABI_FMT_lx "\n",
933                     info->entry);
934             fprintf(f, "argv_start  0x" TARGET_ABI_FMT_lx "\n",
935                     info->argv);
936             fprintf(f, "env_start   0x" TARGET_ABI_FMT_lx "\n",
937                     info->envp);
938             fprintf(f, "auxv_start  0x" TARGET_ABI_FMT_lx "\n",
939                     info->saved_auxv);
940             qemu_log_unlock(f);
941         }
942     }
943 
944     target_set_brk(info->brk);
945     syscall_init();
946     signal_init();
947 
948     /* Now that we've loaded the binary, GUEST_BASE is fixed.  Delay
949        generating the prologue until now so that the prologue can take
950        the real value of GUEST_BASE into account.  */
951     tcg_prologue_init(tcg_ctx);
952 
953     target_cpu_copy_regs(env, regs);
954 
955     if (gdbstub) {
956         if (gdbserver_start(gdbstub) < 0) {
957             fprintf(stderr, "qemu: could not open gdbserver on %s\n",
958                     gdbstub);
959             exit(EXIT_FAILURE);
960         }
961         gdb_handlesig(cpu, 0);
962     }
963 
964 #ifdef CONFIG_SEMIHOSTING
965     qemu_semihosting_guestfd_init();
966 #endif
967 
968     cpu_loop(env);
969     /* never exits */
970     return 0;
971 }
972