xref: /openbmc/qemu/linux-user/main.c (revision 1af52156676065b1fc2d4815bf23b1c4c99938b3)
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 "user/page-protection.h"
43 #include "exec/gdbstub.h"
44 #include "gdbstub/user.h"
45 #include "accel/accel-ops.h"
46 #include "tcg/startup.h"
47 #include "qemu/timer.h"
48 #include "qemu/envlist.h"
49 #include "qemu/guest-random.h"
50 #include "elf.h"
51 #include "trace/control.h"
52 #include "user/cpu_loop.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 unsigned long opt_tb_size;
75 static const char *argv0;
76 static const char *gdbstub;
77 static envlist_t *envlist;
78 static const char *cpu_model;
79 static const char *cpu_type;
80 static const char *seed_optarg;
81 unsigned long mmap_min_addr;
82 uintptr_t guest_base;
83 bool have_guest_base;
84 
85 /*
86  * Used to implement backwards-compatibility for the `-strace`, and
87  * QEMU_STRACE options. Without this, the QEMU_LOG can be overwritten by
88  * -strace, or vice versa.
89  */
90 static bool enable_strace;
91 
92 /*
93  * The last log mask given by the user in an environment variable or argument.
94  * Used to support command line arguments overriding environment variables.
95  */
96 static int last_log_mask;
97 static const char *last_log_filename;
98 
99 /*
100  * When running 32-on-64 we should make sure we can fit all of the possible
101  * guest address space into a contiguous chunk of virtual host memory.
102  *
103  * This way we will never overlap with our own libraries or binaries or stack
104  * or anything else that QEMU maps.
105  *
106  * Many cpus reserve the high bit (or more than one for some 64-bit cpus)
107  * of the address for the kernel.  Some cpus rely on this and user space
108  * uses the high bit(s) for pointer tagging and the like.  For them, we
109  * must preserve the expected address space.
110  */
111 #ifndef MAX_RESERVED_VA
112 # if HOST_LONG_BITS > TARGET_VIRT_ADDR_SPACE_BITS
113 #  if TARGET_VIRT_ADDR_SPACE_BITS == 32 && \
114       (TARGET_LONG_BITS == 32 || defined(TARGET_ABI32))
115 #   define MAX_RESERVED_VA(CPU)  0xfffffffful
116 #  else
117 #   define MAX_RESERVED_VA(CPU)  ((1ul << TARGET_VIRT_ADDR_SPACE_BITS) - 1)
118 #  endif
119 # else
120 #  define MAX_RESERVED_VA(CPU)  0
121 # endif
122 #endif
123 
124 unsigned long reserved_va;
125 unsigned long guest_addr_max;
126 
127 static void usage(int exitcode);
128 
129 static const char *interp_prefix = CONFIG_QEMU_INTERP_PREFIX;
130 const char *qemu_uname_release;
131 
132 #if !defined(TARGET_DEFAULT_STACK_SIZE)
133 /* XXX: on x86 MAP_GROWSDOWN only works if ESP <= address + 32, so
134    we allocate a bigger stack. Need a better solution, for example
135    by remapping the process stack directly at the right place */
136 #define TARGET_DEFAULT_STACK_SIZE	8 * 1024 * 1024UL
137 #endif
138 
139 unsigned long guest_stack_size = TARGET_DEFAULT_STACK_SIZE;
140 
141 /***********************************************************/
142 /* Helper routines for implementing atomic operations.  */
143 
144 /* Make sure everything is in a consistent state for calling fork().  */
fork_start(void)145 void fork_start(void)
146 {
147     start_exclusive();
148     clone_fork_start();
149     mmap_fork_start();
150     cpu_list_lock();
151     qemu_plugin_user_prefork_lock();
152     gdbserver_fork_start();
153     fd_trans_prefork();
154 }
155 
fork_end(pid_t pid)156 void fork_end(pid_t pid)
157 {
158     bool child = pid == 0;
159 
160     fd_trans_postfork();
161     qemu_plugin_user_postfork(child);
162     mmap_fork_end(child);
163     if (child) {
164         CPUState *cpu, *next_cpu;
165         /* Child processes created by fork() only have a single thread.
166            Discard information about the parent threads.  */
167         CPU_FOREACH_SAFE(cpu, next_cpu) {
168             if (cpu != thread_cpu) {
169                 QTAILQ_REMOVE_RCU(&cpus_queue, cpu, node);
170             }
171         }
172         qemu_init_cpu_list();
173         get_task_state(thread_cpu)->ts_tid = qemu_get_thread_id();
174     } else {
175         cpu_list_unlock();
176     }
177     gdbserver_fork_end(thread_cpu, pid);
178     clone_fork_end(child);
179     /*
180      * qemu_init_cpu_list() reinitialized the child exclusive state, but we
181      * also need to keep current_cpu consistent, so call end_exclusive() for
182      * both child and parent.
183      */
184     end_exclusive();
185 }
186 
187 __thread CPUState *thread_cpu;
188 
qemu_cpu_is_self(CPUState * cpu)189 bool qemu_cpu_is_self(CPUState *cpu)
190 {
191     return thread_cpu == cpu;
192 }
193 
task_settid(TaskState * ts)194 void task_settid(TaskState *ts)
195 {
196     if (ts->ts_tid == 0) {
197         ts->ts_tid = (pid_t)syscall(SYS_gettid);
198     }
199 }
200 
stop_all_tasks(void)201 void stop_all_tasks(void)
202 {
203     /*
204      * We trust that when using NPTL, start_exclusive()
205      * handles thread stopping correctly.
206      */
207     start_exclusive();
208 }
209 
210 /* Assumes contents are already zeroed.  */
init_task_state(TaskState * ts)211 void init_task_state(TaskState *ts)
212 {
213     long ticks_per_sec;
214     struct timespec bt;
215 
216     ts->used = 1;
217     ts->sigaltstack_used = (struct target_sigaltstack) {
218         .ss_sp = 0,
219         .ss_size = 0,
220         .ss_flags = TARGET_SS_DISABLE,
221     };
222 
223     /* Capture task start time relative to system boot */
224 
225     ticks_per_sec = sysconf(_SC_CLK_TCK);
226 
227     if ((ticks_per_sec > 0) && !clock_gettime(CLOCK_BOOTTIME, &bt)) {
228         /* start_boottime is expressed in clock ticks */
229         ts->start_boottime = bt.tv_sec * (uint64_t) ticks_per_sec;
230         ts->start_boottime += bt.tv_nsec * (uint64_t) ticks_per_sec /
231                               NANOSECONDS_PER_SECOND;
232     }
233 
234     ts->sys_dispatch_len = -1;
235 }
236 
cpu_copy(CPUArchState * env)237 CPUArchState *cpu_copy(CPUArchState *env)
238 {
239     CPUState *cpu = env_cpu(env);
240     CPUState *new_cpu = cpu_create(cpu_type);
241     CPUArchState *new_env = cpu_env(new_cpu);
242     CPUBreakpoint *bp;
243 
244     /* Reset non arch specific state */
245     cpu_reset(new_cpu);
246 
247     new_cpu->tcg_cflags = cpu->tcg_cflags;
248     memcpy(new_env, env, sizeof(CPUArchState));
249 #if defined(TARGET_I386) || defined(TARGET_X86_64)
250     new_env->gdt.base = target_mmap(0, sizeof(uint64_t) * TARGET_GDT_ENTRIES,
251                                     PROT_READ | PROT_WRITE,
252                                     MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
253     memcpy(g2h_untagged(new_env->gdt.base), g2h_untagged(env->gdt.base),
254            sizeof(uint64_t) * TARGET_GDT_ENTRIES);
255     OBJECT(new_cpu)->free = OBJECT(cpu)->free;
256 #endif
257 
258     /* Clone all break/watchpoints.
259        Note: Once we support ptrace with hw-debug register access, make sure
260        BP_CPU break/watchpoints are handled correctly on clone. */
261     QTAILQ_INIT(&new_cpu->breakpoints);
262     QTAILQ_FOREACH(bp, &cpu->breakpoints, entry) {
263         cpu_breakpoint_insert(new_cpu, bp->pc, bp->flags, NULL);
264     }
265 
266     return new_env;
267 }
268 
handle_arg_help(const char * arg)269 static void handle_arg_help(const char *arg)
270 {
271     usage(EXIT_SUCCESS);
272 }
273 
handle_arg_log(const char * arg)274 static void handle_arg_log(const char *arg)
275 {
276     last_log_mask = qemu_str_to_log_mask(arg);
277     if (!last_log_mask) {
278         qemu_print_log_usage(stdout);
279         exit(EXIT_FAILURE);
280     }
281 }
282 
handle_arg_dfilter(const char * arg)283 static void handle_arg_dfilter(const char *arg)
284 {
285     qemu_set_dfilter_ranges(arg, &error_fatal);
286 }
287 
handle_arg_log_filename(const char * arg)288 static void handle_arg_log_filename(const char *arg)
289 {
290     last_log_filename = arg;
291 }
292 
handle_arg_set_env(const char * arg)293 static void handle_arg_set_env(const char *arg)
294 {
295     char *r, *p, *token;
296     r = p = strdup(arg);
297     while ((token = strsep(&p, ",")) != NULL) {
298         if (envlist_setenv(envlist, token) != 0) {
299             usage(EXIT_FAILURE);
300         }
301     }
302     free(r);
303 }
304 
handle_arg_unset_env(const char * arg)305 static void handle_arg_unset_env(const char *arg)
306 {
307     char *r, *p, *token;
308     r = p = strdup(arg);
309     while ((token = strsep(&p, ",")) != NULL) {
310         if (envlist_unsetenv(envlist, token) != 0) {
311             usage(EXIT_FAILURE);
312         }
313     }
314     free(r);
315 }
316 
handle_arg_argv0(const char * arg)317 static void handle_arg_argv0(const char *arg)
318 {
319     argv0 = strdup(arg);
320 }
321 
handle_arg_stack_size(const char * arg)322 static void handle_arg_stack_size(const char *arg)
323 {
324     char *p;
325     guest_stack_size = strtoul(arg, &p, 0);
326     if (guest_stack_size == 0) {
327         usage(EXIT_FAILURE);
328     }
329 
330     if (*p == 'M') {
331         guest_stack_size *= MiB;
332     } else if (*p == 'k' || *p == 'K') {
333         guest_stack_size *= KiB;
334     }
335 }
336 
handle_arg_ld_prefix(const char * arg)337 static void handle_arg_ld_prefix(const char *arg)
338 {
339     interp_prefix = strdup(arg);
340 }
341 
handle_arg_seed(const char * arg)342 static void handle_arg_seed(const char *arg)
343 {
344     seed_optarg = arg;
345 }
346 
handle_arg_gdb(const char * arg)347 static void handle_arg_gdb(const char *arg)
348 {
349     gdbstub = g_strdup(arg);
350 }
351 
handle_arg_uname(const char * arg)352 static void handle_arg_uname(const char *arg)
353 {
354     qemu_uname_release = strdup(arg);
355 }
356 
handle_arg_cpu(const char * arg)357 static void handle_arg_cpu(const char *arg)
358 {
359     cpu_model = strdup(arg);
360     if (cpu_model == NULL || is_help_option(cpu_model)) {
361         list_cpus();
362         exit(EXIT_FAILURE);
363     }
364 }
365 
handle_arg_guest_base(const char * arg)366 static void handle_arg_guest_base(const char *arg)
367 {
368     guest_base = strtol(arg, NULL, 0);
369     have_guest_base = true;
370 }
371 
handle_arg_reserved_va(const char * arg)372 static void handle_arg_reserved_va(const char *arg)
373 {
374     char *p;
375     int shift = 0;
376     unsigned long val;
377 
378     val = strtoul(arg, &p, 0);
379     switch (*p) {
380     case 'k':
381     case 'K':
382         shift = 10;
383         break;
384     case 'M':
385         shift = 20;
386         break;
387     case 'G':
388         shift = 30;
389         break;
390     }
391     if (shift) {
392         unsigned long unshifted = val;
393         p++;
394         val <<= shift;
395         if (val >> shift != unshifted) {
396             fprintf(stderr, "Reserved virtual address too big\n");
397             exit(EXIT_FAILURE);
398         }
399     }
400     if (*p) {
401         fprintf(stderr, "Unrecognised -R size suffix '%s'\n", p);
402         exit(EXIT_FAILURE);
403     }
404     /* The representation is size - 1, with 0 remaining "default". */
405     reserved_va = val ? val - 1 : 0;
406 }
407 
408 static const char *rtsig_map = CONFIG_QEMU_RTSIG_MAP;
409 
handle_arg_rtsig_map(const char * arg)410 static void handle_arg_rtsig_map(const char *arg)
411 {
412     rtsig_map = arg;
413 }
414 
handle_arg_one_insn_per_tb(const char * arg)415 static void handle_arg_one_insn_per_tb(const char *arg)
416 {
417     opt_one_insn_per_tb = true;
418 }
419 
handle_arg_tb_size(const char * arg)420 static void handle_arg_tb_size(const char *arg)
421 {
422     if (qemu_strtoul(arg, NULL, 0, &opt_tb_size)) {
423         usage(EXIT_FAILURE);
424     }
425 }
426 
handle_arg_strace(const char * arg)427 static void handle_arg_strace(const char *arg)
428 {
429     enable_strace = true;
430 }
431 
handle_arg_version(const char * arg)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 
handle_arg_trace(const char * arg)439 static void handle_arg_trace(const char *arg)
440 {
441     trace_opt_parse(arg);
442 }
443 
444 #if defined(TARGET_XTENSA)
handle_arg_abi_call0(const char * arg)445 static void handle_arg_abi_call0(const char *arg)
446 {
447     xtensa_set_abi_call0();
448 }
449 #endif
450 
handle_arg_perfmap(const char * arg)451 static void handle_arg_perfmap(const char *arg)
452 {
453     perf_enable_perfmap();
454 }
455 
handle_arg_jitdump(const char * arg)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
handle_arg_plugin(const char * arg)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     {"one-insn-per-tb",
515                    "QEMU_ONE_INSN_PER_TB",  false, handle_arg_one_insn_per_tb,
516      "",           "run with one guest instruction per emulated TB"},
517     {"tb-size",    "QEMU_TB_SIZE",     true,  handle_arg_tb_size,
518      "size",       "TCG translation block cache size"},
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 
usage(int exitcode)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 
parse_args(int argc,char ** argv)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 
main(int argc,char ** argv,char ** envp)684 int main(int argc, char **argv, char **envp)
685 {
686     struct image_info info1, *info = &info1;
687     struct linux_binprm bprm;
688     TaskState *ts;
689     CPUArchState *env;
690     CPUState *cpu;
691     int optind;
692     char **target_environ, **wrk;
693     char **target_argv;
694     int target_argc;
695     int i;
696     int ret;
697     int execfd;
698     int host_page_size;
699     unsigned long max_reserved_va;
700     bool preserve_argv0;
701 
702     error_init(argv[0]);
703     module_call_init(MODULE_INIT_TRACE);
704     qemu_init_cpu_list();
705     module_call_init(MODULE_INIT_QOM);
706 
707     envlist = envlist_create();
708 
709     /*
710      * add current environment into the list
711      * envlist_setenv adds to the front of the list; to preserve environ
712      * order add from back to front
713      */
714     for (wrk = environ; *wrk != NULL; wrk++) {
715         continue;
716     }
717     while (wrk != environ) {
718         wrk--;
719         (void) envlist_setenv(envlist, *wrk);
720     }
721 
722     /* Read the stack limit from the kernel.  If it's "unlimited",
723        then we can do little else besides use the default.  */
724     {
725         struct rlimit lim;
726         if (getrlimit(RLIMIT_STACK, &lim) == 0
727             && lim.rlim_cur != RLIM_INFINITY
728             && lim.rlim_cur == (target_long)lim.rlim_cur
729             && lim.rlim_cur > guest_stack_size) {
730             guest_stack_size = lim.rlim_cur;
731         }
732     }
733 
734     cpu_model = NULL;
735 
736     qemu_add_opts(&qemu_trace_opts);
737     qemu_plugin_add_opts();
738 
739     optind = parse_args(argc, argv);
740 
741     qemu_set_log_filename_flags(last_log_filename,
742                                 last_log_mask | (enable_strace * LOG_STRACE),
743                                 &error_fatal);
744 
745     if (!trace_init_backends()) {
746         exit(1);
747     }
748     trace_init_file();
749     qemu_plugin_load_list(&plugins, &error_fatal);
750 
751     /* Zero out image_info */
752     memset(info, 0, sizeof(struct image_info));
753 
754     memset(&bprm, 0, sizeof (bprm));
755 
756     /* Scan interp_prefix dir for replacement files. */
757     init_paths(interp_prefix);
758 
759     init_qemu_uname_release();
760 
761     /*
762      * Manage binfmt-misc open-binary flag
763      */
764     errno = 0;
765     execfd = qemu_getauxval(AT_EXECFD);
766     if (errno != 0) {
767         execfd = open(exec_path, O_RDONLY);
768         if (execfd < 0) {
769             printf("Error while loading %s: %s\n", exec_path, strerror(errno));
770             _exit(EXIT_FAILURE);
771         }
772     }
773 
774     /* Resolve executable file name to full path name */
775     if (realpath(exec_path, real_exec_path)) {
776         exec_path = real_exec_path;
777     }
778 
779     /*
780      * get binfmt_misc flags
781      */
782     preserve_argv0 = !!(qemu_getauxval(AT_FLAGS) & AT_FLAGS_PRESERVE_ARGV0);
783 
784     /*
785      * Manage binfmt-misc preserve-arg[0] flag
786      *    argv[optind]     full path to the binary
787      *    argv[optind + 1] original argv[0]
788      */
789     if (optind + 1 < argc && preserve_argv0) {
790         optind++;
791     }
792 
793     if (cpu_model == NULL) {
794         cpu_model = get_elf_cpu_model(get_elf_eflags(execfd));
795     }
796     cpu_type = parse_cpu_option(cpu_model);
797 
798     /* init tcg before creating CPUs */
799     {
800         AccelState *accel = current_accel();
801         AccelClass *ac = ACCEL_GET_CLASS(accel);
802 
803         accel_init_interfaces(ac);
804         object_property_set_bool(OBJECT(accel), "one-insn-per-tb",
805                                  opt_one_insn_per_tb, &error_abort);
806         object_property_set_int(OBJECT(accel), "tb-size",
807                                 opt_tb_size, &error_abort);
808         ac->init_machine(accel, NULL);
809     }
810 
811     /*
812      * Finalize page size before creating CPUs.
813      * This will do nothing if !TARGET_PAGE_BITS_VARY.
814      * The most efficient setting is to match the host.
815      */
816     host_page_size = qemu_real_host_page_size();
817     set_preferred_target_page_bits(ctz32(host_page_size));
818     finalize_target_page_bits();
819 
820     cpu = cpu_create(cpu_type);
821     env = cpu_env(cpu);
822     cpu_reset(cpu);
823     thread_cpu = cpu;
824 
825     /*
826      * Reserving too much vm space via mmap can run into problems with rlimits,
827      * oom due to page table creation, etc.  We will still try it, if directed
828      * by the command-line option, but not by default. Unless we're running a
829      * target address space of 32 or fewer bits on a host with 64 bits.
830      */
831     max_reserved_va = MAX_RESERVED_VA(cpu);
832     if (reserved_va != 0) {
833         if ((reserved_va + 1) % host_page_size) {
834             char *s = size_to_str(host_page_size);
835             fprintf(stderr, "Reserved virtual address not aligned mod %s\n", s);
836             g_free(s);
837             exit(EXIT_FAILURE);
838         }
839         if (max_reserved_va && reserved_va > max_reserved_va) {
840             fprintf(stderr, "Reserved virtual address too big\n");
841             exit(EXIT_FAILURE);
842         }
843     } else if (HOST_LONG_BITS == 64 && TARGET_VIRT_ADDR_SPACE_BITS <= 32) {
844         /* MAX_RESERVED_VA + 1 is a large power of 2, so is aligned. */
845         reserved_va = max_reserved_va;
846     }
847     if (reserved_va != 0) {
848         guest_addr_max = reserved_va;
849     } else if (MIN(TARGET_VIRT_ADDR_SPACE_BITS, TARGET_ABI_BITS) <= 32) {
850         guest_addr_max = UINT32_MAX;
851     } else {
852         guest_addr_max = ~0ul;
853     }
854 
855     /*
856      * Temporarily disable
857      *   "comparison is always false due to limited range of data type"
858      * due to comparison between (possible) uint64_t and uintptr_t.
859      */
860 #pragma GCC diagnostic push
861 #pragma GCC diagnostic ignored "-Wtype-limits"
862 #pragma GCC diagnostic ignored "-Wtautological-compare"
863 
864     /*
865      * Select an initial value for task_unmapped_base that is in range.
866      */
867     if (reserved_va) {
868         if (TASK_UNMAPPED_BASE < reserved_va) {
869             task_unmapped_base = TASK_UNMAPPED_BASE;
870         } else {
871             /* The most common default formula is TASK_SIZE / 3. */
872             task_unmapped_base = TARGET_PAGE_ALIGN(reserved_va / 3);
873         }
874     } else if (TASK_UNMAPPED_BASE < UINTPTR_MAX) {
875         task_unmapped_base = TASK_UNMAPPED_BASE;
876     } else {
877         /* 32-bit host: pick something medium size. */
878         task_unmapped_base = 0x10000000;
879     }
880     mmap_next_start = task_unmapped_base;
881 
882     /* Similarly for elf_et_dyn_base. */
883     if (reserved_va) {
884         if (ELF_ET_DYN_BASE < reserved_va) {
885             elf_et_dyn_base = ELF_ET_DYN_BASE;
886         } else {
887             /* The most common default formula is TASK_SIZE / 3 * 2. */
888             elf_et_dyn_base = TARGET_PAGE_ALIGN(reserved_va / 3) * 2;
889         }
890     } else if (ELF_ET_DYN_BASE < UINTPTR_MAX) {
891         elf_et_dyn_base = ELF_ET_DYN_BASE;
892     } else {
893         /* 32-bit host: pick something medium size. */
894         elf_et_dyn_base = 0x18000000;
895     }
896 
897 #pragma GCC diagnostic pop
898 
899     {
900         Error *err = NULL;
901         if (seed_optarg != NULL) {
902             qemu_guest_random_seed_main(seed_optarg, &err);
903         } else {
904             qcrypto_init(&err);
905         }
906         if (err) {
907             error_reportf_err(err, "cannot initialize crypto: ");
908             exit(1);
909         }
910     }
911 
912     target_environ = envlist_to_environ(envlist, NULL);
913     envlist_free(envlist);
914 
915     /*
916      * Read in mmap_min_addr kernel parameter.  This value is used
917      * When loading the ELF image to determine whether guest_base
918      * is needed.  It is also used in mmap_find_vma.
919      */
920     {
921         FILE *fp;
922 
923         if ((fp = fopen("/proc/sys/vm/mmap_min_addr", "r")) != NULL) {
924             unsigned long tmp;
925             if (fscanf(fp, "%lu", &tmp) == 1 && tmp != 0) {
926                 mmap_min_addr = MAX(tmp, host_page_size);
927                 qemu_log_mask(CPU_LOG_PAGE, "host mmap_min_addr=0x%lx\n",
928                               mmap_min_addr);
929             }
930             fclose(fp);
931         }
932     }
933 
934     /*
935      * We prefer to not make NULL pointers accessible to QEMU.
936      * If we're in a chroot with no /proc, fall back to 1 page.
937      */
938     if (mmap_min_addr == 0) {
939         mmap_min_addr = host_page_size;
940         qemu_log_mask(CPU_LOG_PAGE,
941                       "host mmap_min_addr=0x%lx (fallback)\n",
942                       mmap_min_addr);
943     }
944 
945     /*
946      * Prepare copy of argv vector for target.
947      */
948     target_argc = argc - optind;
949     target_argv = g_new0(char *, target_argc + 1);
950 
951     /*
952      * If argv0 is specified (using '-0' switch) we replace
953      * argv[0] pointer with the given one.
954      */
955     i = 0;
956     if (argv0 != NULL) {
957         target_argv[i++] = strdup(argv0);
958     }
959     for (; i < target_argc; i++) {
960         target_argv[i] = strdup(argv[optind + i]);
961     }
962     target_argv[target_argc] = NULL;
963 
964     ts = g_new0(TaskState, 1);
965     init_task_state(ts);
966     /* build Task State */
967     ts->info = info;
968     ts->bprm = &bprm;
969     cpu->opaque = ts;
970     task_settid(ts);
971 
972     fd_trans_init();
973 
974     ret = loader_exec(execfd, exec_path, target_argv, target_environ,
975                       info, &bprm);
976     if (ret != 0) {
977         printf("Error while loading %s: %s\n", exec_path, strerror(-ret));
978         _exit(EXIT_FAILURE);
979     }
980 
981     for (wrk = target_environ; *wrk; wrk++) {
982         g_free(*wrk);
983     }
984 
985     g_free(target_environ);
986 
987     if (qemu_loglevel_mask(CPU_LOG_PAGE)) {
988         FILE *f = qemu_log_trylock();
989         if (f) {
990             fprintf(f, "guest_base  %p\n", (void *)guest_base);
991             fprintf(f, "page layout changed following binary load\n");
992             page_dump(f);
993 
994             fprintf(f, "end_code    0x" TARGET_ABI_FMT_lx "\n",
995                     info->end_code);
996             fprintf(f, "start_code  0x" TARGET_ABI_FMT_lx "\n",
997                     info->start_code);
998             fprintf(f, "start_data  0x" TARGET_ABI_FMT_lx "\n",
999                     info->start_data);
1000             fprintf(f, "end_data    0x" TARGET_ABI_FMT_lx "\n",
1001                     info->end_data);
1002             fprintf(f, "start_stack 0x" TARGET_ABI_FMT_lx "\n",
1003                     info->start_stack);
1004             fprintf(f, "brk         0x" TARGET_ABI_FMT_lx "\n",
1005                     info->brk);
1006             fprintf(f, "entry       0x" TARGET_ABI_FMT_lx "\n",
1007                     info->entry);
1008             fprintf(f, "argv_start  0x" TARGET_ABI_FMT_lx "\n",
1009                     info->argv);
1010             fprintf(f, "env_start   0x" TARGET_ABI_FMT_lx "\n",
1011                     info->envp);
1012             fprintf(f, "auxv_start  0x" TARGET_ABI_FMT_lx "\n",
1013                     info->saved_auxv);
1014             qemu_log_unlock(f);
1015         }
1016     }
1017 
1018     target_set_brk(info->brk);
1019     syscall_init();
1020     signal_init(rtsig_map);
1021 
1022     /* Now that we've loaded the binary, GUEST_BASE is fixed.  Delay
1023        generating the prologue until now so that the prologue can take
1024        the real value of GUEST_BASE into account.  */
1025     tcg_prologue_init();
1026 
1027     init_main_thread(cpu, info);
1028 
1029     if (gdbstub) {
1030         gdbserver_start(gdbstub, &error_fatal);
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