xref: /openbmc/qemu/linux-user/main.c (revision 7d87775f)
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 "user/guest-base.h"
42 #include "exec/exec-all.h"
43 #include "exec/gdbstub.h"
44 #include "gdbstub/user.h"
45 #include "tcg/startup.h"
46 #include "qemu/timer.h"
47 #include "qemu/envlist.h"
48 #include "qemu/guest-random.h"
49 #include "elf.h"
50 #include "trace/control.h"
51 #include "target_elf.h"
52 #include "cpu_loop-common.h"
53 #include "crypto/init.h"
54 #include "fd-trans.h"
55 #include "signal-common.h"
56 #include "loader.h"
57 #include "user-mmap.h"
58 #include "tcg/perf.h"
59 #include "exec/page-vary.h"
60 
61 #ifdef CONFIG_SEMIHOSTING
62 #include "semihosting/semihost.h"
63 #endif
64 
65 #ifndef AT_FLAGS_PRESERVE_ARGV0
66 #define AT_FLAGS_PRESERVE_ARGV0_BIT 0
67 #define AT_FLAGS_PRESERVE_ARGV0 (1 << AT_FLAGS_PRESERVE_ARGV0_BIT)
68 #endif
69 
70 char *exec_path;
71 char real_exec_path[PATH_MAX];
72 
73 static bool opt_one_insn_per_tb;
74 static const char *argv0;
75 static const char *gdbstub;
76 static envlist_t *envlist;
77 static const char *cpu_model;
78 static const char *cpu_type;
79 static const char *seed_optarg;
80 unsigned long mmap_min_addr;
81 uintptr_t guest_base;
82 bool have_guest_base;
83 
84 /*
85  * Used to implement backwards-compatibility for the `-strace`, and
86  * QEMU_STRACE options. Without this, the QEMU_LOG can be overwritten by
87  * -strace, or vice versa.
88  */
89 static bool enable_strace;
90 
91 /*
92  * The last log mask given by the user in an environment variable or argument.
93  * Used to support command line arguments overriding environment variables.
94  */
95 static int last_log_mask;
96 static const char *last_log_filename;
97 
98 /*
99  * When running 32-on-64 we should make sure we can fit all of the possible
100  * guest address space into a contiguous chunk of virtual host memory.
101  *
102  * This way we will never overlap with our own libraries or binaries or stack
103  * or anything else that QEMU maps.
104  *
105  * Many cpus reserve the high bit (or more than one for some 64-bit cpus)
106  * of the address for the kernel.  Some cpus rely on this and user space
107  * uses the high bit(s) for pointer tagging and the like.  For them, we
108  * must preserve the expected address space.
109  */
110 #ifndef MAX_RESERVED_VA
111 # if HOST_LONG_BITS > TARGET_VIRT_ADDR_SPACE_BITS
112 #  if TARGET_VIRT_ADDR_SPACE_BITS == 32 && \
113       (TARGET_LONG_BITS == 32 || defined(TARGET_ABI32))
114 #   define MAX_RESERVED_VA(CPU)  0xfffffffful
115 #  else
116 #   define MAX_RESERVED_VA(CPU)  ((1ul << TARGET_VIRT_ADDR_SPACE_BITS) - 1)
117 #  endif
118 # else
119 #  define MAX_RESERVED_VA(CPU)  0
120 # endif
121 #endif
122 
123 unsigned long reserved_va;
124 
125 static void usage(int exitcode);
126 
127 static const char *interp_prefix = CONFIG_QEMU_INTERP_PREFIX;
128 const char *qemu_uname_release;
129 
130 #if !defined(TARGET_DEFAULT_STACK_SIZE)
131 /* XXX: on x86 MAP_GROWSDOWN only works if ESP <= address + 32, so
132    we allocate a bigger stack. Need a better solution, for example
133    by remapping the process stack directly at the right place */
134 #define TARGET_DEFAULT_STACK_SIZE	8 * 1024 * 1024UL
135 #endif
136 
137 unsigned long guest_stack_size = TARGET_DEFAULT_STACK_SIZE;
138 
139 /***********************************************************/
140 /* Helper routines for implementing atomic operations.  */
141 
142 /* Make sure everything is in a consistent state for calling fork().  */
143 void fork_start(void)
144 {
145     start_exclusive();
146     mmap_fork_start();
147     cpu_list_lock();
148     qemu_plugin_user_prefork_lock();
149     gdbserver_fork_start();
150 }
151 
152 void fork_end(pid_t pid)
153 {
154     bool child = pid == 0;
155 
156     qemu_plugin_user_postfork(child);
157     mmap_fork_end(child);
158     if (child) {
159         CPUState *cpu, *next_cpu;
160         /* Child processes created by fork() only have a single thread.
161            Discard information about the parent threads.  */
162         CPU_FOREACH_SAFE(cpu, next_cpu) {
163             if (cpu != thread_cpu) {
164                 QTAILQ_REMOVE_RCU(&cpus_queue, cpu, node);
165             }
166         }
167         qemu_init_cpu_list();
168         get_task_state(thread_cpu)->ts_tid = qemu_get_thread_id();
169     } else {
170         cpu_list_unlock();
171     }
172     gdbserver_fork_end(thread_cpu, pid);
173     /*
174      * qemu_init_cpu_list() reinitialized the child exclusive state, but we
175      * also need to keep current_cpu consistent, so call end_exclusive() for
176      * both child and parent.
177      */
178     end_exclusive();
179 }
180 
181 __thread CPUState *thread_cpu;
182 
183 bool qemu_cpu_is_self(CPUState *cpu)
184 {
185     return thread_cpu == cpu;
186 }
187 
188 void qemu_cpu_kick(CPUState *cpu)
189 {
190     cpu_exit(cpu);
191 }
192 
193 void task_settid(TaskState *ts)
194 {
195     if (ts->ts_tid == 0) {
196         ts->ts_tid = (pid_t)syscall(SYS_gettid);
197     }
198 }
199 
200 void stop_all_tasks(void)
201 {
202     /*
203      * We trust that when using NPTL, start_exclusive()
204      * handles thread stopping correctly.
205      */
206     start_exclusive();
207 }
208 
209 /* Assumes contents are already zeroed.  */
210 void init_task_state(TaskState *ts)
211 {
212     long ticks_per_sec;
213     struct timespec bt;
214 
215     ts->used = 1;
216     ts->sigaltstack_used = (struct target_sigaltstack) {
217         .ss_sp = 0,
218         .ss_size = 0,
219         .ss_flags = TARGET_SS_DISABLE,
220     };
221 
222     /* Capture task start time relative to system boot */
223 
224     ticks_per_sec = sysconf(_SC_CLK_TCK);
225 
226     if ((ticks_per_sec > 0) && !clock_gettime(CLOCK_BOOTTIME, &bt)) {
227         /* start_boottime is expressed in clock ticks */
228         ts->start_boottime = bt.tv_sec * (uint64_t) ticks_per_sec;
229         ts->start_boottime += bt.tv_nsec * (uint64_t) ticks_per_sec /
230                               NANOSECONDS_PER_SECOND;
231     }
232 }
233 
234 CPUArchState *cpu_copy(CPUArchState *env)
235 {
236     CPUState *cpu = env_cpu(env);
237     CPUState *new_cpu = cpu_create(cpu_type);
238     CPUArchState *new_env = cpu_env(new_cpu);
239     CPUBreakpoint *bp;
240 
241     /* Reset non arch specific state */
242     cpu_reset(new_cpu);
243 
244     new_cpu->tcg_cflags = cpu->tcg_cflags;
245     memcpy(new_env, env, sizeof(CPUArchState));
246 #if defined(TARGET_I386) || defined(TARGET_X86_64)
247     new_env->gdt.base = target_mmap(0, sizeof(uint64_t) * TARGET_GDT_ENTRIES,
248                                     PROT_READ | PROT_WRITE,
249                                     MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
250     memcpy(g2h_untagged(new_env->gdt.base), g2h_untagged(env->gdt.base),
251            sizeof(uint64_t) * TARGET_GDT_ENTRIES);
252     OBJECT(new_cpu)->free = OBJECT(cpu)->free;
253 #endif
254 
255     /* Clone all break/watchpoints.
256        Note: Once we support ptrace with hw-debug register access, make sure
257        BP_CPU break/watchpoints are handled correctly on clone. */
258     QTAILQ_INIT(&new_cpu->breakpoints);
259     QTAILQ_FOREACH(bp, &cpu->breakpoints, entry) {
260         cpu_breakpoint_insert(new_cpu, bp->pc, bp->flags, NULL);
261     }
262 
263     return new_env;
264 }
265 
266 static void handle_arg_help(const char *arg)
267 {
268     usage(EXIT_SUCCESS);
269 }
270 
271 static void handle_arg_log(const char *arg)
272 {
273     last_log_mask = qemu_str_to_log_mask(arg);
274     if (!last_log_mask) {
275         qemu_print_log_usage(stdout);
276         exit(EXIT_FAILURE);
277     }
278 }
279 
280 static void handle_arg_dfilter(const char *arg)
281 {
282     qemu_set_dfilter_ranges(arg, &error_fatal);
283 }
284 
285 static void handle_arg_log_filename(const char *arg)
286 {
287     last_log_filename = arg;
288 }
289 
290 static void handle_arg_set_env(const char *arg)
291 {
292     char *r, *p, *token;
293     r = p = strdup(arg);
294     while ((token = strsep(&p, ",")) != NULL) {
295         if (envlist_setenv(envlist, token) != 0) {
296             usage(EXIT_FAILURE);
297         }
298     }
299     free(r);
300 }
301 
302 static void handle_arg_unset_env(const char *arg)
303 {
304     char *r, *p, *token;
305     r = p = strdup(arg);
306     while ((token = strsep(&p, ",")) != NULL) {
307         if (envlist_unsetenv(envlist, token) != 0) {
308             usage(EXIT_FAILURE);
309         }
310     }
311     free(r);
312 }
313 
314 static void handle_arg_argv0(const char *arg)
315 {
316     argv0 = strdup(arg);
317 }
318 
319 static void handle_arg_stack_size(const char *arg)
320 {
321     char *p;
322     guest_stack_size = strtoul(arg, &p, 0);
323     if (guest_stack_size == 0) {
324         usage(EXIT_FAILURE);
325     }
326 
327     if (*p == 'M') {
328         guest_stack_size *= MiB;
329     } else if (*p == 'k' || *p == 'K') {
330         guest_stack_size *= KiB;
331     }
332 }
333 
334 static void handle_arg_ld_prefix(const char *arg)
335 {
336     interp_prefix = strdup(arg);
337 }
338 
339 static void handle_arg_pagesize(const char *arg)
340 {
341     unsigned size, want = qemu_real_host_page_size();
342 
343     if (qemu_strtoui(arg, NULL, 10, &size) || size != want) {
344         warn_report("Deprecated page size option cannot "
345                     "change host page size (%u)", want);
346     }
347 }
348 
349 static void handle_arg_seed(const char *arg)
350 {
351     seed_optarg = arg;
352 }
353 
354 static void handle_arg_gdb(const char *arg)
355 {
356     gdbstub = g_strdup(arg);
357 }
358 
359 static void handle_arg_uname(const char *arg)
360 {
361     qemu_uname_release = strdup(arg);
362 }
363 
364 static void handle_arg_cpu(const char *arg)
365 {
366     cpu_model = strdup(arg);
367     if (cpu_model == NULL || is_help_option(cpu_model)) {
368         list_cpus();
369         exit(EXIT_FAILURE);
370     }
371 }
372 
373 static void handle_arg_guest_base(const char *arg)
374 {
375     guest_base = strtol(arg, NULL, 0);
376     have_guest_base = true;
377 }
378 
379 static void handle_arg_reserved_va(const char *arg)
380 {
381     char *p;
382     int shift = 0;
383     unsigned long val;
384 
385     val = strtoul(arg, &p, 0);
386     switch (*p) {
387     case 'k':
388     case 'K':
389         shift = 10;
390         break;
391     case 'M':
392         shift = 20;
393         break;
394     case 'G':
395         shift = 30;
396         break;
397     }
398     if (shift) {
399         unsigned long unshifted = val;
400         p++;
401         val <<= shift;
402         if (val >> shift != unshifted) {
403             fprintf(stderr, "Reserved virtual address too big\n");
404             exit(EXIT_FAILURE);
405         }
406     }
407     if (*p) {
408         fprintf(stderr, "Unrecognised -R size suffix '%s'\n", p);
409         exit(EXIT_FAILURE);
410     }
411     /* The representation is size - 1, with 0 remaining "default". */
412     reserved_va = val ? val - 1 : 0;
413 }
414 
415 static const char *rtsig_map = CONFIG_QEMU_RTSIG_MAP;
416 
417 static void handle_arg_rtsig_map(const char *arg)
418 {
419     rtsig_map = arg;
420 }
421 
422 static void handle_arg_one_insn_per_tb(const char *arg)
423 {
424     opt_one_insn_per_tb = true;
425 }
426 
427 static void handle_arg_strace(const char *arg)
428 {
429     enable_strace = true;
430 }
431 
432 static void handle_arg_version(const char *arg)
433 {
434     printf("qemu-" TARGET_NAME " version " QEMU_FULL_VERSION
435            "\n" QEMU_COPYRIGHT "\n");
436     exit(EXIT_SUCCESS);
437 }
438 
439 static void handle_arg_trace(const char *arg)
440 {
441     trace_opt_parse(arg);
442 }
443 
444 #if defined(TARGET_XTENSA)
445 static void handle_arg_abi_call0(const char *arg)
446 {
447     xtensa_set_abi_call0();
448 }
449 #endif
450 
451 static void handle_arg_perfmap(const char *arg)
452 {
453     perf_enable_perfmap();
454 }
455 
456 static void handle_arg_jitdump(const char *arg)
457 {
458     perf_enable_jitdump();
459 }
460 
461 static QemuPluginList plugins = QTAILQ_HEAD_INITIALIZER(plugins);
462 
463 #ifdef CONFIG_PLUGIN
464 static void handle_arg_plugin(const char *arg)
465 {
466     qemu_plugin_opt_parse(arg, &plugins);
467 }
468 #endif
469 
470 struct qemu_argument {
471     const char *argv;
472     const char *env;
473     bool has_arg;
474     void (*handle_opt)(const char *arg);
475     const char *example;
476     const char *help;
477 };
478 
479 static const struct qemu_argument arg_table[] = {
480     {"h",          "",                 false, handle_arg_help,
481      "",           "print this help"},
482     {"help",       "",                 false, handle_arg_help,
483      "",           ""},
484     {"g",          "QEMU_GDB",         true,  handle_arg_gdb,
485      "port",       "wait gdb connection to 'port'"},
486     {"L",          "QEMU_LD_PREFIX",   true,  handle_arg_ld_prefix,
487      "path",       "set the elf interpreter prefix to 'path'"},
488     {"s",          "QEMU_STACK_SIZE",  true,  handle_arg_stack_size,
489      "size",       "set the stack size to 'size' bytes"},
490     {"cpu",        "QEMU_CPU",         true,  handle_arg_cpu,
491      "model",      "select CPU (-cpu help for list)"},
492     {"E",          "QEMU_SET_ENV",     true,  handle_arg_set_env,
493      "var=value",  "sets targets environment variable (see below)"},
494     {"U",          "QEMU_UNSET_ENV",   true,  handle_arg_unset_env,
495      "var",        "unsets targets environment variable (see below)"},
496     {"0",          "QEMU_ARGV0",       true,  handle_arg_argv0,
497      "argv0",      "forces target process argv[0] to be 'argv0'"},
498     {"r",          "QEMU_UNAME",       true,  handle_arg_uname,
499      "uname",      "set qemu uname release string to 'uname'"},
500     {"B",          "QEMU_GUEST_BASE",  true,  handle_arg_guest_base,
501      "address",    "set guest_base address to 'address'"},
502     {"R",          "QEMU_RESERVED_VA", true,  handle_arg_reserved_va,
503      "size",       "reserve 'size' bytes for guest virtual address space"},
504     {"t",          "QEMU_RTSIG_MAP",   true,  handle_arg_rtsig_map,
505      "tsig hsig n[,...]",
506                    "map target rt signals [tsig,tsig+n) to [hsig,hsig+n]"},
507     {"d",          "QEMU_LOG",         true,  handle_arg_log,
508      "item[,...]", "enable logging of specified items "
509      "(use '-d help' for a list of items)"},
510     {"dfilter",    "QEMU_DFILTER",     true,  handle_arg_dfilter,
511      "range[,...]","filter logging based on address range"},
512     {"D",          "QEMU_LOG_FILENAME", true, handle_arg_log_filename,
513      "logfile",     "write logs to 'logfile' (default stderr)"},
514     {"p",          "QEMU_PAGESIZE",    true,  handle_arg_pagesize,
515      "pagesize",   "deprecated change to host page size"},
516     {"one-insn-per-tb",
517                    "QEMU_ONE_INSN_PER_TB",  false, handle_arg_one_insn_per_tb,
518      "",           "run with one guest instruction per emulated TB"},
519     {"strace",     "QEMU_STRACE",      false, handle_arg_strace,
520      "",           "log system calls"},
521     {"seed",       "QEMU_RAND_SEED",   true,  handle_arg_seed,
522      "",           "Seed for pseudo-random number generator"},
523     {"trace",      "QEMU_TRACE",       true,  handle_arg_trace,
524      "",           "[[enable=]<pattern>][,events=<file>][,file=<file>]"},
525 #ifdef CONFIG_PLUGIN
526     {"plugin",     "QEMU_PLUGIN",      true,  handle_arg_plugin,
527      "",           "[file=]<file>[,<argname>=<argvalue>]"},
528 #endif
529     {"version",    "QEMU_VERSION",     false, handle_arg_version,
530      "",           "display version information and exit"},
531 #if defined(TARGET_XTENSA)
532     {"xtensa-abi-call0", "QEMU_XTENSA_ABI_CALL0", false, handle_arg_abi_call0,
533      "",           "assume CALL0 Xtensa ABI"},
534 #endif
535     {"perfmap",    "QEMU_PERFMAP",     false, handle_arg_perfmap,
536      "",           "Generate a /tmp/perf-${pid}.map file for perf"},
537     {"jitdump",    "QEMU_JITDUMP",     false, handle_arg_jitdump,
538      "",           "Generate a jit-${pid}.dump file for perf"},
539     {NULL, NULL, false, NULL, NULL, NULL}
540 };
541 
542 static void usage(int exitcode)
543 {
544     const struct qemu_argument *arginfo;
545     int maxarglen;
546     int maxenvlen;
547 
548     printf("usage: qemu-" TARGET_NAME " [options] program [arguments...]\n"
549            "Linux CPU emulator (compiled for " TARGET_NAME " emulation)\n"
550            "\n"
551            "Options and associated environment variables:\n"
552            "\n");
553 
554     /* Calculate column widths. We must always have at least enough space
555      * for the column header.
556      */
557     maxarglen = strlen("Argument");
558     maxenvlen = strlen("Env-variable");
559 
560     for (arginfo = arg_table; arginfo->handle_opt != NULL; arginfo++) {
561         int arglen = strlen(arginfo->argv);
562         if (arginfo->has_arg) {
563             arglen += strlen(arginfo->example) + 1;
564         }
565         if (strlen(arginfo->env) > maxenvlen) {
566             maxenvlen = strlen(arginfo->env);
567         }
568         if (arglen > maxarglen) {
569             maxarglen = arglen;
570         }
571     }
572 
573     printf("%-*s %-*s Description\n", maxarglen+1, "Argument",
574             maxenvlen, "Env-variable");
575 
576     for (arginfo = arg_table; arginfo->handle_opt != NULL; arginfo++) {
577         if (arginfo->has_arg) {
578             printf("-%s %-*s %-*s %s\n", arginfo->argv,
579                    (int)(maxarglen - strlen(arginfo->argv) - 1),
580                    arginfo->example, maxenvlen, arginfo->env, arginfo->help);
581         } else {
582             printf("-%-*s %-*s %s\n", maxarglen, arginfo->argv,
583                     maxenvlen, arginfo->env,
584                     arginfo->help);
585         }
586     }
587 
588     printf("\n"
589            "Defaults:\n"
590            "QEMU_LD_PREFIX  = %s\n"
591            "QEMU_STACK_SIZE = %ld byte\n",
592            interp_prefix,
593            guest_stack_size);
594 
595     printf("\n"
596            "You can use -E and -U options or the QEMU_SET_ENV and\n"
597            "QEMU_UNSET_ENV environment variables to set and unset\n"
598            "environment variables for the target process.\n"
599            "It is possible to provide several variables by separating them\n"
600            "by commas in getsubopt(3) style. Additionally it is possible to\n"
601            "provide the -E and -U options multiple times.\n"
602            "The following lines are equivalent:\n"
603            "    -E var1=val2 -E var2=val2 -U LD_PRELOAD -U LD_DEBUG\n"
604            "    -E var1=val2,var2=val2 -U LD_PRELOAD,LD_DEBUG\n"
605            "    QEMU_SET_ENV=var1=val2,var2=val2 QEMU_UNSET_ENV=LD_PRELOAD,LD_DEBUG\n"
606            "Note that if you provide several changes to a single variable\n"
607            "the last change will stay in effect.\n"
608            "\n"
609            QEMU_HELP_BOTTOM "\n");
610 
611     exit(exitcode);
612 }
613 
614 static int parse_args(int argc, char **argv)
615 {
616     const char *r;
617     int optind;
618     const struct qemu_argument *arginfo;
619 
620     for (arginfo = arg_table; arginfo->handle_opt != NULL; arginfo++) {
621         if (arginfo->env == NULL) {
622             continue;
623         }
624 
625         r = getenv(arginfo->env);
626         if (r != NULL) {
627             arginfo->handle_opt(r);
628         }
629     }
630 
631     optind = 1;
632     for (;;) {
633         if (optind >= argc) {
634             break;
635         }
636         r = argv[optind];
637         if (r[0] != '-') {
638             break;
639         }
640         optind++;
641         r++;
642         if (!strcmp(r, "-")) {
643             break;
644         }
645         /* Treat --foo the same as -foo.  */
646         if (r[0] == '-') {
647             r++;
648         }
649 
650         for (arginfo = arg_table; arginfo->handle_opt != NULL; arginfo++) {
651             if (!strcmp(r, arginfo->argv)) {
652                 if (arginfo->has_arg) {
653                     if (optind >= argc) {
654                         (void) fprintf(stderr,
655                             "qemu: missing argument for option '%s'\n", r);
656                         exit(EXIT_FAILURE);
657                     }
658                     arginfo->handle_opt(argv[optind]);
659                     optind++;
660                 } else {
661                     arginfo->handle_opt(NULL);
662                 }
663                 break;
664             }
665         }
666 
667         /* no option matched the current argv */
668         if (arginfo->handle_opt == NULL) {
669             (void) fprintf(stderr, "qemu: unknown option '%s'\n", r);
670             exit(EXIT_FAILURE);
671         }
672     }
673 
674     if (optind >= argc) {
675         (void) fprintf(stderr, "qemu: no user program specified\n");
676         exit(EXIT_FAILURE);
677     }
678 
679     exec_path = argv[optind];
680 
681     return optind;
682 }
683 
684 int main(int argc, char **argv, char **envp)
685 {
686     struct target_pt_regs regs1, *regs = &regs1;
687     struct image_info info1, *info = &info1;
688     struct linux_binprm bprm;
689     TaskState *ts;
690     CPUArchState *env;
691     CPUState *cpu;
692     int optind;
693     char **target_environ, **wrk;
694     char **target_argv;
695     int target_argc;
696     int i;
697     int ret;
698     int execfd;
699     int host_page_size;
700     unsigned long max_reserved_va;
701     bool preserve_argv0;
702 
703     error_init(argv[0]);
704     module_call_init(MODULE_INIT_TRACE);
705     qemu_init_cpu_list();
706     module_call_init(MODULE_INIT_QOM);
707 
708     envlist = envlist_create();
709 
710     /*
711      * add current environment into the list
712      * envlist_setenv adds to the front of the list; to preserve environ
713      * order add from back to front
714      */
715     for (wrk = environ; *wrk != NULL; wrk++) {
716         continue;
717     }
718     while (wrk != environ) {
719         wrk--;
720         (void) envlist_setenv(envlist, *wrk);
721     }
722 
723     /* Read the stack limit from the kernel.  If it's "unlimited",
724        then we can do little else besides use the default.  */
725     {
726         struct rlimit lim;
727         if (getrlimit(RLIMIT_STACK, &lim) == 0
728             && lim.rlim_cur != RLIM_INFINITY
729             && lim.rlim_cur == (target_long)lim.rlim_cur
730             && lim.rlim_cur > guest_stack_size) {
731             guest_stack_size = lim.rlim_cur;
732         }
733     }
734 
735     cpu_model = NULL;
736 
737     qemu_add_opts(&qemu_trace_opts);
738     qemu_plugin_add_opts();
739 
740     optind = parse_args(argc, argv);
741 
742     qemu_set_log_filename_flags(last_log_filename,
743                                 last_log_mask | (enable_strace * LOG_STRACE),
744                                 &error_fatal);
745 
746     if (!trace_init_backends()) {
747         exit(1);
748     }
749     trace_init_file();
750     qemu_plugin_load_list(&plugins, &error_fatal);
751 
752     /* Zero out regs */
753     memset(regs, 0, sizeof(struct target_pt_regs));
754 
755     /* Zero out image_info */
756     memset(info, 0, sizeof(struct image_info));
757 
758     memset(&bprm, 0, sizeof (bprm));
759 
760     /* Scan interp_prefix dir for replacement files. */
761     init_paths(interp_prefix);
762 
763     init_qemu_uname_release();
764 
765     /*
766      * Manage binfmt-misc open-binary flag
767      */
768     errno = 0;
769     execfd = qemu_getauxval(AT_EXECFD);
770     if (errno != 0) {
771         execfd = open(exec_path, O_RDONLY);
772         if (execfd < 0) {
773             printf("Error while loading %s: %s\n", exec_path, strerror(errno));
774             _exit(EXIT_FAILURE);
775         }
776     }
777 
778     /* Resolve executable file name to full path name */
779     if (realpath(exec_path, real_exec_path)) {
780         exec_path = real_exec_path;
781     }
782 
783     /*
784      * get binfmt_misc flags
785      */
786     preserve_argv0 = !!(qemu_getauxval(AT_FLAGS) & AT_FLAGS_PRESERVE_ARGV0);
787 
788     /*
789      * Manage binfmt-misc preserve-arg[0] flag
790      *    argv[optind]     full path to the binary
791      *    argv[optind + 1] original argv[0]
792      */
793     if (optind + 1 < argc && preserve_argv0) {
794         optind++;
795     }
796 
797     if (cpu_model == NULL) {
798         cpu_model = cpu_get_model(get_elf_eflags(execfd));
799     }
800     cpu_type = parse_cpu_option(cpu_model);
801 
802     /* init tcg before creating CPUs */
803     {
804         AccelState *accel = current_accel();
805         AccelClass *ac = ACCEL_GET_CLASS(accel);
806 
807         accel_init_interfaces(ac);
808         object_property_set_bool(OBJECT(accel), "one-insn-per-tb",
809                                  opt_one_insn_per_tb, &error_abort);
810         ac->init_machine(NULL);
811     }
812 
813     /*
814      * Finalize page size before creating CPUs.
815      * This will do nothing if !TARGET_PAGE_BITS_VARY.
816      * The most efficient setting is to match the host.
817      */
818     host_page_size = qemu_real_host_page_size();
819     set_preferred_target_page_bits(ctz32(host_page_size));
820     finalize_target_page_bits();
821 
822     cpu = cpu_create(cpu_type);
823     env = cpu_env(cpu);
824     cpu_reset(cpu);
825     thread_cpu = cpu;
826 
827     /*
828      * Reserving too much vm space via mmap can run into problems with rlimits,
829      * oom due to page table creation, etc.  We will still try it, if directed
830      * by the command-line option, but not by default. Unless we're running a
831      * target address space of 32 or fewer bits on a host with 64 bits.
832      */
833     max_reserved_va = MAX_RESERVED_VA(cpu);
834     if (reserved_va != 0) {
835         if ((reserved_va + 1) % host_page_size) {
836             char *s = size_to_str(host_page_size);
837             fprintf(stderr, "Reserved virtual address not aligned mod %s\n", s);
838             g_free(s);
839             exit(EXIT_FAILURE);
840         }
841         if (max_reserved_va && reserved_va > max_reserved_va) {
842             fprintf(stderr, "Reserved virtual address too big\n");
843             exit(EXIT_FAILURE);
844         }
845     } else if (HOST_LONG_BITS == 64 && TARGET_VIRT_ADDR_SPACE_BITS <= 32) {
846         /* MAX_RESERVED_VA + 1 is a large power of 2, so is aligned. */
847         reserved_va = max_reserved_va;
848     }
849 
850     /*
851      * Temporarily disable
852      *   "comparison is always false due to limited range of data type"
853      * due to comparison between (possible) uint64_t and uintptr_t.
854      */
855 #pragma GCC diagnostic push
856 #pragma GCC diagnostic ignored "-Wtype-limits"
857 #pragma GCC diagnostic ignored "-Wtautological-compare"
858 
859     /*
860      * Select an initial value for task_unmapped_base that is in range.
861      */
862     if (reserved_va) {
863         if (TASK_UNMAPPED_BASE < reserved_va) {
864             task_unmapped_base = TASK_UNMAPPED_BASE;
865         } else {
866             /* The most common default formula is TASK_SIZE / 3. */
867             task_unmapped_base = TARGET_PAGE_ALIGN(reserved_va / 3);
868         }
869     } else if (TASK_UNMAPPED_BASE < UINTPTR_MAX) {
870         task_unmapped_base = TASK_UNMAPPED_BASE;
871     } else {
872         /* 32-bit host: pick something medium size. */
873         task_unmapped_base = 0x10000000;
874     }
875     mmap_next_start = task_unmapped_base;
876 
877     /* Similarly for elf_et_dyn_base. */
878     if (reserved_va) {
879         if (ELF_ET_DYN_BASE < reserved_va) {
880             elf_et_dyn_base = ELF_ET_DYN_BASE;
881         } else {
882             /* The most common default formula is TASK_SIZE / 3 * 2. */
883             elf_et_dyn_base = TARGET_PAGE_ALIGN(reserved_va / 3) * 2;
884         }
885     } else if (ELF_ET_DYN_BASE < UINTPTR_MAX) {
886         elf_et_dyn_base = ELF_ET_DYN_BASE;
887     } else {
888         /* 32-bit host: pick something medium size. */
889         elf_et_dyn_base = 0x18000000;
890     }
891 
892 #pragma GCC diagnostic pop
893 
894     {
895         Error *err = NULL;
896         if (seed_optarg != NULL) {
897             qemu_guest_random_seed_main(seed_optarg, &err);
898         } else {
899             qcrypto_init(&err);
900         }
901         if (err) {
902             error_reportf_err(err, "cannot initialize crypto: ");
903             exit(1);
904         }
905     }
906 
907     target_environ = envlist_to_environ(envlist, NULL);
908     envlist_free(envlist);
909 
910     /*
911      * Read in mmap_min_addr kernel parameter.  This value is used
912      * When loading the ELF image to determine whether guest_base
913      * is needed.  It is also used in mmap_find_vma.
914      */
915     {
916         FILE *fp;
917 
918         if ((fp = fopen("/proc/sys/vm/mmap_min_addr", "r")) != NULL) {
919             unsigned long tmp;
920             if (fscanf(fp, "%lu", &tmp) == 1 && tmp != 0) {
921                 mmap_min_addr = MAX(tmp, host_page_size);
922                 qemu_log_mask(CPU_LOG_PAGE, "host mmap_min_addr=0x%lx\n",
923                               mmap_min_addr);
924             }
925             fclose(fp);
926         }
927     }
928 
929     /*
930      * We prefer to not make NULL pointers accessible to QEMU.
931      * If we're in a chroot with no /proc, fall back to 1 page.
932      */
933     if (mmap_min_addr == 0) {
934         mmap_min_addr = host_page_size;
935         qemu_log_mask(CPU_LOG_PAGE,
936                       "host mmap_min_addr=0x%lx (fallback)\n",
937                       mmap_min_addr);
938     }
939 
940     /*
941      * Prepare copy of argv vector for target.
942      */
943     target_argc = argc - optind;
944     target_argv = g_new0(char *, target_argc + 1);
945 
946     /*
947      * If argv0 is specified (using '-0' switch) we replace
948      * argv[0] pointer with the given one.
949      */
950     i = 0;
951     if (argv0 != NULL) {
952         target_argv[i++] = strdup(argv0);
953     }
954     for (; i < target_argc; i++) {
955         target_argv[i] = strdup(argv[optind + i]);
956     }
957     target_argv[target_argc] = NULL;
958 
959     ts = g_new0(TaskState, 1);
960     init_task_state(ts);
961     /* build Task State */
962     ts->info = info;
963     ts->bprm = &bprm;
964     cpu->opaque = ts;
965     task_settid(ts);
966 
967     fd_trans_init();
968 
969     ret = loader_exec(execfd, exec_path, target_argv, target_environ, regs,
970         info, &bprm);
971     if (ret != 0) {
972         printf("Error while loading %s: %s\n", exec_path, strerror(-ret));
973         _exit(EXIT_FAILURE);
974     }
975 
976     for (wrk = target_environ; *wrk; wrk++) {
977         g_free(*wrk);
978     }
979 
980     g_free(target_environ);
981 
982     if (qemu_loglevel_mask(CPU_LOG_PAGE)) {
983         FILE *f = qemu_log_trylock();
984         if (f) {
985             fprintf(f, "guest_base  %p\n", (void *)guest_base);
986             fprintf(f, "page layout changed following binary load\n");
987             page_dump(f);
988 
989             fprintf(f, "end_code    0x" TARGET_ABI_FMT_lx "\n",
990                     info->end_code);
991             fprintf(f, "start_code  0x" TARGET_ABI_FMT_lx "\n",
992                     info->start_code);
993             fprintf(f, "start_data  0x" TARGET_ABI_FMT_lx "\n",
994                     info->start_data);
995             fprintf(f, "end_data    0x" TARGET_ABI_FMT_lx "\n",
996                     info->end_data);
997             fprintf(f, "start_stack 0x" TARGET_ABI_FMT_lx "\n",
998                     info->start_stack);
999             fprintf(f, "brk         0x" TARGET_ABI_FMT_lx "\n",
1000                     info->brk);
1001             fprintf(f, "entry       0x" TARGET_ABI_FMT_lx "\n",
1002                     info->entry);
1003             fprintf(f, "argv_start  0x" TARGET_ABI_FMT_lx "\n",
1004                     info->argv);
1005             fprintf(f, "env_start   0x" TARGET_ABI_FMT_lx "\n",
1006                     info->envp);
1007             fprintf(f, "auxv_start  0x" TARGET_ABI_FMT_lx "\n",
1008                     info->saved_auxv);
1009             qemu_log_unlock(f);
1010         }
1011     }
1012 
1013     target_set_brk(info->brk);
1014     syscall_init();
1015     signal_init(rtsig_map);
1016 
1017     /* Now that we've loaded the binary, GUEST_BASE is fixed.  Delay
1018        generating the prologue until now so that the prologue can take
1019        the real value of GUEST_BASE into account.  */
1020     tcg_prologue_init();
1021 
1022     target_cpu_copy_regs(env, regs);
1023 
1024     if (gdbstub) {
1025         if (gdbserver_start(gdbstub) < 0) {
1026             fprintf(stderr, "qemu: could not open gdbserver on %s\n",
1027                     gdbstub);
1028             exit(EXIT_FAILURE);
1029         }
1030         gdb_handlesig(cpu, 0, NULL, NULL, 0);
1031     }
1032 
1033 #ifdef CONFIG_SEMIHOSTING
1034     qemu_semihosting_guestfd_init();
1035 #endif
1036 
1037     cpu_loop(env);
1038     /* never exits */
1039     return 0;
1040 }
1041