xref: /openbmc/qemu/linux-user/main.c (revision 9c4888c9)
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 "tcg/tcg.h"
44 #include "qemu/timer.h"
45 #include "qemu/envlist.h"
46 #include "qemu/guest-random.h"
47 #include "elf.h"
48 #include "trace/control.h"
49 #include "target_elf.h"
50 #include "cpu_loop-common.h"
51 #include "crypto/init.h"
52 #include "fd-trans.h"
53 #include "signal-common.h"
54 #include "loader.h"
55 #include "user-mmap.h"
56 
57 #ifndef AT_FLAGS_PRESERVE_ARGV0
58 #define AT_FLAGS_PRESERVE_ARGV0_BIT 0
59 #define AT_FLAGS_PRESERVE_ARGV0 (1 << AT_FLAGS_PRESERVE_ARGV0_BIT)
60 #endif
61 
62 char *exec_path;
63 
64 int singlestep;
65 static const char *argv0;
66 static const char *gdbstub;
67 static envlist_t *envlist;
68 static const char *cpu_model;
69 static const char *cpu_type;
70 static const char *seed_optarg;
71 unsigned long mmap_min_addr;
72 uintptr_t guest_base;
73 bool have_guest_base;
74 
75 /*
76  * Used to implement backwards-compatibility for the `-strace`, and
77  * QEMU_STRACE options. Without this, the QEMU_LOG can be overwritten by
78  * -strace, or vice versa.
79  */
80 static bool enable_strace;
81 
82 /*
83  * The last log mask given by the user in an environment variable or argument.
84  * Used to support command line arguments overriding environment variables.
85  */
86 static int last_log_mask;
87 static const char *last_log_filename;
88 
89 /*
90  * When running 32-on-64 we should make sure we can fit all of the possible
91  * guest address space into a contiguous chunk of virtual host memory.
92  *
93  * This way we will never overlap with our own libraries or binaries or stack
94  * or anything else that QEMU maps.
95  *
96  * Many cpus reserve the high bit (or more than one for some 64-bit cpus)
97  * of the address for the kernel.  Some cpus rely on this and user space
98  * uses the high bit(s) for pointer tagging and the like.  For them, we
99  * must preserve the expected address space.
100  */
101 #ifndef MAX_RESERVED_VA
102 # if HOST_LONG_BITS > TARGET_VIRT_ADDR_SPACE_BITS
103 #  if TARGET_VIRT_ADDR_SPACE_BITS == 32 && \
104       (TARGET_LONG_BITS == 32 || defined(TARGET_ABI32))
105 /* There are a number of places where we assign reserved_va to a variable
106    of type abi_ulong and expect it to fit.  Avoid the last page.  */
107 #   define MAX_RESERVED_VA(CPU)  (0xfffffffful & TARGET_PAGE_MASK)
108 #  else
109 #   define MAX_RESERVED_VA(CPU)  (1ul << TARGET_VIRT_ADDR_SPACE_BITS)
110 #  endif
111 # else
112 #  define MAX_RESERVED_VA(CPU)  0
113 # endif
114 #endif
115 
116 unsigned long reserved_va;
117 
118 static void usage(int exitcode);
119 
120 static const char *interp_prefix = CONFIG_QEMU_INTERP_PREFIX;
121 const char *qemu_uname_release;
122 
123 /* XXX: on x86 MAP_GROWSDOWN only works if ESP <= address + 32, so
124    we allocate a bigger stack. Need a better solution, for example
125    by remapping the process stack directly at the right place */
126 unsigned long guest_stack_size = 8 * 1024 * 1024UL;
127 
128 /***********************************************************/
129 /* Helper routines for implementing atomic operations.  */
130 
131 /* Make sure everything is in a consistent state for calling fork().  */
132 void fork_start(void)
133 {
134     start_exclusive();
135     mmap_fork_start();
136     cpu_list_lock();
137 }
138 
139 void fork_end(int child)
140 {
141     mmap_fork_end(child);
142     if (child) {
143         CPUState *cpu, *next_cpu;
144         /* Child processes created by fork() only have a single thread.
145            Discard information about the parent threads.  */
146         CPU_FOREACH_SAFE(cpu, next_cpu) {
147             if (cpu != thread_cpu) {
148                 QTAILQ_REMOVE_RCU(&cpus, cpu, node);
149             }
150         }
151         qemu_init_cpu_list();
152         gdbserver_fork(thread_cpu);
153         /* qemu_init_cpu_list() takes care of reinitializing the
154          * exclusive state, so we don't need to end_exclusive() here.
155          */
156     } else {
157         cpu_list_unlock();
158         end_exclusive();
159     }
160 }
161 
162 __thread CPUState *thread_cpu;
163 
164 bool qemu_cpu_is_self(CPUState *cpu)
165 {
166     return thread_cpu == cpu;
167 }
168 
169 void qemu_cpu_kick(CPUState *cpu)
170 {
171     cpu_exit(cpu);
172 }
173 
174 void task_settid(TaskState *ts)
175 {
176     if (ts->ts_tid == 0) {
177         ts->ts_tid = (pid_t)syscall(SYS_gettid);
178     }
179 }
180 
181 void stop_all_tasks(void)
182 {
183     /*
184      * We trust that when using NPTL, start_exclusive()
185      * handles thread stopping correctly.
186      */
187     start_exclusive();
188 }
189 
190 /* Assumes contents are already zeroed.  */
191 void init_task_state(TaskState *ts)
192 {
193     long ticks_per_sec;
194     struct timespec bt;
195 
196     ts->used = 1;
197     ts->sigaltstack_used = (struct target_sigaltstack) {
198         .ss_sp = 0,
199         .ss_size = 0,
200         .ss_flags = TARGET_SS_DISABLE,
201     };
202 
203     /* Capture task start time relative to system boot */
204 
205     ticks_per_sec = sysconf(_SC_CLK_TCK);
206 
207     if ((ticks_per_sec > 0) && !clock_gettime(CLOCK_BOOTTIME, &bt)) {
208         /* start_boottime is expressed in clock ticks */
209         ts->start_boottime = bt.tv_sec * (uint64_t) ticks_per_sec;
210         ts->start_boottime += bt.tv_nsec * (uint64_t) ticks_per_sec /
211                               NANOSECONDS_PER_SECOND;
212     }
213 }
214 
215 CPUArchState *cpu_copy(CPUArchState *env)
216 {
217     CPUState *cpu = env_cpu(env);
218     CPUState *new_cpu = cpu_create(cpu_type);
219     CPUArchState *new_env = new_cpu->env_ptr;
220     CPUBreakpoint *bp;
221 
222     /* Reset non arch specific state */
223     cpu_reset(new_cpu);
224 
225     new_cpu->tcg_cflags = cpu->tcg_cflags;
226     memcpy(new_env, env, sizeof(CPUArchState));
227 
228     /* Clone all break/watchpoints.
229        Note: Once we support ptrace with hw-debug register access, make sure
230        BP_CPU break/watchpoints are handled correctly on clone. */
231     QTAILQ_INIT(&new_cpu->breakpoints);
232     QTAILQ_FOREACH(bp, &cpu->breakpoints, entry) {
233         cpu_breakpoint_insert(new_cpu, bp->pc, bp->flags, NULL);
234     }
235 
236     return new_env;
237 }
238 
239 static void handle_arg_help(const char *arg)
240 {
241     usage(EXIT_SUCCESS);
242 }
243 
244 static void handle_arg_log(const char *arg)
245 {
246     last_log_mask = qemu_str_to_log_mask(arg);
247     if (!last_log_mask) {
248         qemu_print_log_usage(stdout);
249         exit(EXIT_FAILURE);
250     }
251 }
252 
253 static void handle_arg_dfilter(const char *arg)
254 {
255     qemu_set_dfilter_ranges(arg, &error_fatal);
256 }
257 
258 static void handle_arg_log_filename(const char *arg)
259 {
260     last_log_filename = arg;
261 }
262 
263 static void handle_arg_set_env(const char *arg)
264 {
265     char *r, *p, *token;
266     r = p = strdup(arg);
267     while ((token = strsep(&p, ",")) != NULL) {
268         if (envlist_setenv(envlist, token) != 0) {
269             usage(EXIT_FAILURE);
270         }
271     }
272     free(r);
273 }
274 
275 static void handle_arg_unset_env(const char *arg)
276 {
277     char *r, *p, *token;
278     r = p = strdup(arg);
279     while ((token = strsep(&p, ",")) != NULL) {
280         if (envlist_unsetenv(envlist, token) != 0) {
281             usage(EXIT_FAILURE);
282         }
283     }
284     free(r);
285 }
286 
287 static void handle_arg_argv0(const char *arg)
288 {
289     argv0 = strdup(arg);
290 }
291 
292 static void handle_arg_stack_size(const char *arg)
293 {
294     char *p;
295     guest_stack_size = strtoul(arg, &p, 0);
296     if (guest_stack_size == 0) {
297         usage(EXIT_FAILURE);
298     }
299 
300     if (*p == 'M') {
301         guest_stack_size *= MiB;
302     } else if (*p == 'k' || *p == 'K') {
303         guest_stack_size *= KiB;
304     }
305 }
306 
307 static void handle_arg_ld_prefix(const char *arg)
308 {
309     interp_prefix = strdup(arg);
310 }
311 
312 static void handle_arg_pagesize(const char *arg)
313 {
314     qemu_host_page_size = atoi(arg);
315     if (qemu_host_page_size == 0 ||
316         (qemu_host_page_size & (qemu_host_page_size - 1)) != 0) {
317         fprintf(stderr, "page size must be a power of two\n");
318         exit(EXIT_FAILURE);
319     }
320 }
321 
322 static void handle_arg_seed(const char *arg)
323 {
324     seed_optarg = arg;
325 }
326 
327 static void handle_arg_gdb(const char *arg)
328 {
329     gdbstub = g_strdup(arg);
330 }
331 
332 static void handle_arg_uname(const char *arg)
333 {
334     qemu_uname_release = strdup(arg);
335 }
336 
337 static void handle_arg_cpu(const char *arg)
338 {
339     cpu_model = strdup(arg);
340     if (cpu_model == NULL || is_help_option(cpu_model)) {
341         /* XXX: implement xxx_cpu_list for targets that still miss it */
342 #if defined(cpu_list)
343         cpu_list();
344 #endif
345         exit(EXIT_FAILURE);
346     }
347 }
348 
349 static void handle_arg_guest_base(const char *arg)
350 {
351     guest_base = strtol(arg, NULL, 0);
352     have_guest_base = true;
353 }
354 
355 static void handle_arg_reserved_va(const char *arg)
356 {
357     char *p;
358     int shift = 0;
359     reserved_va = strtoul(arg, &p, 0);
360     switch (*p) {
361     case 'k':
362     case 'K':
363         shift = 10;
364         break;
365     case 'M':
366         shift = 20;
367         break;
368     case 'G':
369         shift = 30;
370         break;
371     }
372     if (shift) {
373         unsigned long unshifted = reserved_va;
374         p++;
375         reserved_va <<= shift;
376         if (reserved_va >> shift != unshifted) {
377             fprintf(stderr, "Reserved virtual address too big\n");
378             exit(EXIT_FAILURE);
379         }
380     }
381     if (*p) {
382         fprintf(stderr, "Unrecognised -R size suffix '%s'\n", p);
383         exit(EXIT_FAILURE);
384     }
385 }
386 
387 static void handle_arg_singlestep(const char *arg)
388 {
389     singlestep = 1;
390 }
391 
392 static void handle_arg_strace(const char *arg)
393 {
394     enable_strace = true;
395 }
396 
397 static void handle_arg_version(const char *arg)
398 {
399     printf("qemu-" TARGET_NAME " version " QEMU_FULL_VERSION
400            "\n" QEMU_COPYRIGHT "\n");
401     exit(EXIT_SUCCESS);
402 }
403 
404 static void handle_arg_trace(const char *arg)
405 {
406     trace_opt_parse(arg);
407 }
408 
409 #if defined(TARGET_XTENSA)
410 static void handle_arg_abi_call0(const char *arg)
411 {
412     xtensa_set_abi_call0();
413 }
414 #endif
415 
416 static QemuPluginList plugins = QTAILQ_HEAD_INITIALIZER(plugins);
417 
418 #ifdef CONFIG_PLUGIN
419 static void handle_arg_plugin(const char *arg)
420 {
421     qemu_plugin_opt_parse(arg, &plugins);
422 }
423 #endif
424 
425 struct qemu_argument {
426     const char *argv;
427     const char *env;
428     bool has_arg;
429     void (*handle_opt)(const char *arg);
430     const char *example;
431     const char *help;
432 };
433 
434 static const struct qemu_argument arg_table[] = {
435     {"h",          "",                 false, handle_arg_help,
436      "",           "print this help"},
437     {"help",       "",                 false, handle_arg_help,
438      "",           ""},
439     {"g",          "QEMU_GDB",         true,  handle_arg_gdb,
440      "port",       "wait gdb connection to 'port'"},
441     {"L",          "QEMU_LD_PREFIX",   true,  handle_arg_ld_prefix,
442      "path",       "set the elf interpreter prefix to 'path'"},
443     {"s",          "QEMU_STACK_SIZE",  true,  handle_arg_stack_size,
444      "size",       "set the stack size to 'size' bytes"},
445     {"cpu",        "QEMU_CPU",         true,  handle_arg_cpu,
446      "model",      "select CPU (-cpu help for list)"},
447     {"E",          "QEMU_SET_ENV",     true,  handle_arg_set_env,
448      "var=value",  "sets targets environment variable (see below)"},
449     {"U",          "QEMU_UNSET_ENV",   true,  handle_arg_unset_env,
450      "var",        "unsets targets environment variable (see below)"},
451     {"0",          "QEMU_ARGV0",       true,  handle_arg_argv0,
452      "argv0",      "forces target process argv[0] to be 'argv0'"},
453     {"r",          "QEMU_UNAME",       true,  handle_arg_uname,
454      "uname",      "set qemu uname release string to 'uname'"},
455     {"B",          "QEMU_GUEST_BASE",  true,  handle_arg_guest_base,
456      "address",    "set guest_base address to 'address'"},
457     {"R",          "QEMU_RESERVED_VA", true,  handle_arg_reserved_va,
458      "size",       "reserve 'size' bytes for guest virtual address space"},
459     {"d",          "QEMU_LOG",         true,  handle_arg_log,
460      "item[,...]", "enable logging of specified items "
461      "(use '-d help' for a list of items)"},
462     {"dfilter",    "QEMU_DFILTER",     true,  handle_arg_dfilter,
463      "range[,...]","filter logging based on address range"},
464     {"D",          "QEMU_LOG_FILENAME", true, handle_arg_log_filename,
465      "logfile",     "write logs to 'logfile' (default stderr)"},
466     {"p",          "QEMU_PAGESIZE",    true,  handle_arg_pagesize,
467      "pagesize",   "set the host page size to 'pagesize'"},
468     {"singlestep", "QEMU_SINGLESTEP",  false, handle_arg_singlestep,
469      "",           "run in singlestep mode"},
470     {"strace",     "QEMU_STRACE",      false, handle_arg_strace,
471      "",           "log system calls"},
472     {"seed",       "QEMU_RAND_SEED",   true,  handle_arg_seed,
473      "",           "Seed for pseudo-random number generator"},
474     {"trace",      "QEMU_TRACE",       true,  handle_arg_trace,
475      "",           "[[enable=]<pattern>][,events=<file>][,file=<file>]"},
476 #ifdef CONFIG_PLUGIN
477     {"plugin",     "QEMU_PLUGIN",      true,  handle_arg_plugin,
478      "",           "[file=]<file>[,<argname>=<argvalue>]"},
479 #endif
480     {"version",    "QEMU_VERSION",     false, handle_arg_version,
481      "",           "display version information and exit"},
482 #if defined(TARGET_XTENSA)
483     {"xtensa-abi-call0", "QEMU_XTENSA_ABI_CALL0", false, handle_arg_abi_call0,
484      "",           "assume CALL0 Xtensa ABI"},
485 #endif
486     {NULL, NULL, false, NULL, NULL, NULL}
487 };
488 
489 static void usage(int exitcode)
490 {
491     const struct qemu_argument *arginfo;
492     int maxarglen;
493     int maxenvlen;
494 
495     printf("usage: qemu-" TARGET_NAME " [options] program [arguments...]\n"
496            "Linux CPU emulator (compiled for " TARGET_NAME " emulation)\n"
497            "\n"
498            "Options and associated environment variables:\n"
499            "\n");
500 
501     /* Calculate column widths. We must always have at least enough space
502      * for the column header.
503      */
504     maxarglen = strlen("Argument");
505     maxenvlen = strlen("Env-variable");
506 
507     for (arginfo = arg_table; arginfo->handle_opt != NULL; arginfo++) {
508         int arglen = strlen(arginfo->argv);
509         if (arginfo->has_arg) {
510             arglen += strlen(arginfo->example) + 1;
511         }
512         if (strlen(arginfo->env) > maxenvlen) {
513             maxenvlen = strlen(arginfo->env);
514         }
515         if (arglen > maxarglen) {
516             maxarglen = arglen;
517         }
518     }
519 
520     printf("%-*s %-*s Description\n", maxarglen+1, "Argument",
521             maxenvlen, "Env-variable");
522 
523     for (arginfo = arg_table; arginfo->handle_opt != NULL; arginfo++) {
524         if (arginfo->has_arg) {
525             printf("-%s %-*s %-*s %s\n", arginfo->argv,
526                    (int)(maxarglen - strlen(arginfo->argv) - 1),
527                    arginfo->example, maxenvlen, arginfo->env, arginfo->help);
528         } else {
529             printf("-%-*s %-*s %s\n", maxarglen, arginfo->argv,
530                     maxenvlen, arginfo->env,
531                     arginfo->help);
532         }
533     }
534 
535     printf("\n"
536            "Defaults:\n"
537            "QEMU_LD_PREFIX  = %s\n"
538            "QEMU_STACK_SIZE = %ld byte\n",
539            interp_prefix,
540            guest_stack_size);
541 
542     printf("\n"
543            "You can use -E and -U options or the QEMU_SET_ENV and\n"
544            "QEMU_UNSET_ENV environment variables to set and unset\n"
545            "environment variables for the target process.\n"
546            "It is possible to provide several variables by separating them\n"
547            "by commas in getsubopt(3) style. Additionally it is possible to\n"
548            "provide the -E and -U options multiple times.\n"
549            "The following lines are equivalent:\n"
550            "    -E var1=val2 -E var2=val2 -U LD_PRELOAD -U LD_DEBUG\n"
551            "    -E var1=val2,var2=val2 -U LD_PRELOAD,LD_DEBUG\n"
552            "    QEMU_SET_ENV=var1=val2,var2=val2 QEMU_UNSET_ENV=LD_PRELOAD,LD_DEBUG\n"
553            "Note that if you provide several changes to a single variable\n"
554            "the last change will stay in effect.\n"
555            "\n"
556            QEMU_HELP_BOTTOM "\n");
557 
558     exit(exitcode);
559 }
560 
561 static int parse_args(int argc, char **argv)
562 {
563     const char *r;
564     int optind;
565     const struct qemu_argument *arginfo;
566 
567     for (arginfo = arg_table; arginfo->handle_opt != NULL; arginfo++) {
568         if (arginfo->env == NULL) {
569             continue;
570         }
571 
572         r = getenv(arginfo->env);
573         if (r != NULL) {
574             arginfo->handle_opt(r);
575         }
576     }
577 
578     optind = 1;
579     for (;;) {
580         if (optind >= argc) {
581             break;
582         }
583         r = argv[optind];
584         if (r[0] != '-') {
585             break;
586         }
587         optind++;
588         r++;
589         if (!strcmp(r, "-")) {
590             break;
591         }
592         /* Treat --foo the same as -foo.  */
593         if (r[0] == '-') {
594             r++;
595         }
596 
597         for (arginfo = arg_table; arginfo->handle_opt != NULL; arginfo++) {
598             if (!strcmp(r, arginfo->argv)) {
599                 if (arginfo->has_arg) {
600                     if (optind >= argc) {
601                         (void) fprintf(stderr,
602                             "qemu: missing argument for option '%s'\n", r);
603                         exit(EXIT_FAILURE);
604                     }
605                     arginfo->handle_opt(argv[optind]);
606                     optind++;
607                 } else {
608                     arginfo->handle_opt(NULL);
609                 }
610                 break;
611             }
612         }
613 
614         /* no option matched the current argv */
615         if (arginfo->handle_opt == NULL) {
616             (void) fprintf(stderr, "qemu: unknown option '%s'\n", r);
617             exit(EXIT_FAILURE);
618         }
619     }
620 
621     if (optind >= argc) {
622         (void) fprintf(stderr, "qemu: no user program specified\n");
623         exit(EXIT_FAILURE);
624     }
625 
626     exec_path = argv[optind];
627 
628     return optind;
629 }
630 
631 int main(int argc, char **argv, char **envp)
632 {
633     struct target_pt_regs regs1, *regs = &regs1;
634     struct image_info info1, *info = &info1;
635     struct linux_binprm bprm;
636     TaskState *ts;
637     CPUArchState *env;
638     CPUState *cpu;
639     int optind;
640     char **target_environ, **wrk;
641     char **target_argv;
642     int target_argc;
643     int i;
644     int ret;
645     int execfd;
646     unsigned long max_reserved_va;
647     bool preserve_argv0;
648 
649     error_init(argv[0]);
650     module_call_init(MODULE_INIT_TRACE);
651     qemu_init_cpu_list();
652     module_call_init(MODULE_INIT_QOM);
653 
654     envlist = envlist_create();
655 
656     /* add current environment into the list */
657     for (wrk = environ; *wrk != NULL; wrk++) {
658         (void) envlist_setenv(envlist, *wrk);
659     }
660 
661     /* Read the stack limit from the kernel.  If it's "unlimited",
662        then we can do little else besides use the default.  */
663     {
664         struct rlimit lim;
665         if (getrlimit(RLIMIT_STACK, &lim) == 0
666             && lim.rlim_cur != RLIM_INFINITY
667             && lim.rlim_cur == (target_long)lim.rlim_cur) {
668             guest_stack_size = lim.rlim_cur;
669         }
670     }
671 
672     cpu_model = NULL;
673 
674     qemu_add_opts(&qemu_trace_opts);
675     qemu_plugin_add_opts();
676 
677     optind = parse_args(argc, argv);
678 
679     qemu_set_log_filename_flags(last_log_filename,
680                                 last_log_mask | (enable_strace * LOG_STRACE),
681                                 &error_fatal);
682 
683     if (!trace_init_backends()) {
684         exit(1);
685     }
686     trace_init_file();
687     qemu_plugin_load_list(&plugins, &error_fatal);
688 
689     /* Zero out regs */
690     memset(regs, 0, sizeof(struct target_pt_regs));
691 
692     /* Zero out image_info */
693     memset(info, 0, sizeof(struct image_info));
694 
695     memset(&bprm, 0, sizeof (bprm));
696 
697     /* Scan interp_prefix dir for replacement files. */
698     init_paths(interp_prefix);
699 
700     init_qemu_uname_release();
701 
702     /*
703      * Manage binfmt-misc open-binary flag
704      */
705     execfd = qemu_getauxval(AT_EXECFD);
706     if (execfd == 0) {
707         execfd = open(exec_path, O_RDONLY);
708         if (execfd < 0) {
709             printf("Error while loading %s: %s\n", exec_path, strerror(errno));
710             _exit(EXIT_FAILURE);
711         }
712     }
713 
714     /*
715      * get binfmt_misc flags
716      */
717     preserve_argv0 = !!(qemu_getauxval(AT_FLAGS) & AT_FLAGS_PRESERVE_ARGV0);
718 
719     /*
720      * Manage binfmt-misc preserve-arg[0] flag
721      *    argv[optind]     full path to the binary
722      *    argv[optind + 1] original argv[0]
723      */
724     if (optind + 1 < argc && preserve_argv0) {
725         optind++;
726     }
727 
728     if (cpu_model == NULL) {
729         cpu_model = cpu_get_model(get_elf_eflags(execfd));
730     }
731     cpu_type = parse_cpu_option(cpu_model);
732 
733     /* init tcg before creating CPUs and to get qemu_host_page_size */
734     {
735         AccelClass *ac = ACCEL_GET_CLASS(current_accel());
736 
737         accel_init_interfaces(ac);
738         ac->init_machine(NULL);
739     }
740     cpu = cpu_create(cpu_type);
741     env = cpu->env_ptr;
742     cpu_reset(cpu);
743     thread_cpu = cpu;
744 
745     /*
746      * Reserving too much vm space via mmap can run into problems
747      * with rlimits, oom due to page table creation, etc.  We will
748      * still try it, if directed by the command-line option, but
749      * not by default.
750      */
751     max_reserved_va = MAX_RESERVED_VA(cpu);
752     if (reserved_va != 0) {
753         if (max_reserved_va && reserved_va > max_reserved_va) {
754             fprintf(stderr, "Reserved virtual address too big\n");
755             exit(EXIT_FAILURE);
756         }
757     } else if (HOST_LONG_BITS == 64 && TARGET_VIRT_ADDR_SPACE_BITS <= 32) {
758         /*
759          * reserved_va must be aligned with the host page size
760          * as it is used with mmap()
761          */
762         reserved_va = max_reserved_va & qemu_host_page_mask;
763     }
764 
765     {
766         Error *err = NULL;
767         if (seed_optarg != NULL) {
768             qemu_guest_random_seed_main(seed_optarg, &err);
769         } else {
770             qcrypto_init(&err);
771         }
772         if (err) {
773             error_reportf_err(err, "cannot initialize crypto: ");
774             exit(1);
775         }
776     }
777 
778     target_environ = envlist_to_environ(envlist, NULL);
779     envlist_free(envlist);
780 
781     /*
782      * Read in mmap_min_addr kernel parameter.  This value is used
783      * When loading the ELF image to determine whether guest_base
784      * is needed.  It is also used in mmap_find_vma.
785      */
786     {
787         FILE *fp;
788 
789         if ((fp = fopen("/proc/sys/vm/mmap_min_addr", "r")) != NULL) {
790             unsigned long tmp;
791             if (fscanf(fp, "%lu", &tmp) == 1 && tmp != 0) {
792                 mmap_min_addr = tmp;
793                 qemu_log_mask(CPU_LOG_PAGE, "host mmap_min_addr=0x%lx\n",
794                               mmap_min_addr);
795             }
796             fclose(fp);
797         }
798     }
799 
800     /*
801      * We prefer to not make NULL pointers accessible to QEMU.
802      * If we're in a chroot with no /proc, fall back to 1 page.
803      */
804     if (mmap_min_addr == 0) {
805         mmap_min_addr = qemu_host_page_size;
806         qemu_log_mask(CPU_LOG_PAGE,
807                       "host mmap_min_addr=0x%lx (fallback)\n",
808                       mmap_min_addr);
809     }
810 
811     /*
812      * Prepare copy of argv vector for target.
813      */
814     target_argc = argc - optind;
815     target_argv = calloc(target_argc + 1, sizeof (char *));
816     if (target_argv == NULL) {
817         (void) fprintf(stderr, "Unable to allocate memory for target_argv\n");
818         exit(EXIT_FAILURE);
819     }
820 
821     /*
822      * If argv0 is specified (using '-0' switch) we replace
823      * argv[0] pointer with the given one.
824      */
825     i = 0;
826     if (argv0 != NULL) {
827         target_argv[i++] = strdup(argv0);
828     }
829     for (; i < target_argc; i++) {
830         target_argv[i] = strdup(argv[optind + i]);
831     }
832     target_argv[target_argc] = NULL;
833 
834     ts = g_new0(TaskState, 1);
835     init_task_state(ts);
836     /* build Task State */
837     ts->info = info;
838     ts->bprm = &bprm;
839     cpu->opaque = ts;
840     task_settid(ts);
841 
842     fd_trans_init();
843 
844     ret = loader_exec(execfd, exec_path, target_argv, target_environ, regs,
845         info, &bprm);
846     if (ret != 0) {
847         printf("Error while loading %s: %s\n", exec_path, strerror(-ret));
848         _exit(EXIT_FAILURE);
849     }
850 
851     for (wrk = target_environ; *wrk; wrk++) {
852         g_free(*wrk);
853     }
854 
855     g_free(target_environ);
856 
857     if (qemu_loglevel_mask(CPU_LOG_PAGE)) {
858         FILE *f = qemu_log_trylock();
859         if (f) {
860             fprintf(f, "guest_base  %p\n", (void *)guest_base);
861             fprintf(f, "page layout changed following binary load\n");
862             page_dump(f);
863 
864             fprintf(f, "start_brk   0x" TARGET_ABI_FMT_lx "\n",
865                     info->start_brk);
866             fprintf(f, "end_code    0x" TARGET_ABI_FMT_lx "\n",
867                     info->end_code);
868             fprintf(f, "start_code  0x" TARGET_ABI_FMT_lx "\n",
869                     info->start_code);
870             fprintf(f, "start_data  0x" TARGET_ABI_FMT_lx "\n",
871                     info->start_data);
872             fprintf(f, "end_data    0x" TARGET_ABI_FMT_lx "\n",
873                     info->end_data);
874             fprintf(f, "start_stack 0x" TARGET_ABI_FMT_lx "\n",
875                     info->start_stack);
876             fprintf(f, "brk         0x" TARGET_ABI_FMT_lx "\n",
877                     info->brk);
878             fprintf(f, "entry       0x" TARGET_ABI_FMT_lx "\n",
879                     info->entry);
880             fprintf(f, "argv_start  0x" TARGET_ABI_FMT_lx "\n",
881                     info->arg_start);
882             fprintf(f, "env_start   0x" TARGET_ABI_FMT_lx "\n",
883                     info->arg_end + (abi_ulong)sizeof(abi_ulong));
884             fprintf(f, "auxv_start  0x" TARGET_ABI_FMT_lx "\n",
885                     info->saved_auxv);
886             qemu_log_unlock(f);
887         }
888     }
889 
890     target_set_brk(info->brk);
891     syscall_init();
892     signal_init();
893 
894     /* Now that we've loaded the binary, GUEST_BASE is fixed.  Delay
895        generating the prologue until now so that the prologue can take
896        the real value of GUEST_BASE into account.  */
897     tcg_prologue_init(tcg_ctx);
898 
899     target_cpu_copy_regs(env, regs);
900 
901     if (gdbstub) {
902         if (gdbserver_start(gdbstub) < 0) {
903             fprintf(stderr, "qemu: could not open gdbserver on %s\n",
904                     gdbstub);
905             exit(EXIT_FAILURE);
906         }
907         gdb_handlesig(cpu, 0);
908     }
909     cpu_loop(env);
910     /* never exits */
911     return 0;
912 }
913