1 /*
2 * qemu bsd user main
3 *
4 * Copyright (c) 2003-2008 Fabrice Bellard
5 * Copyright (c) 2013-14 Stacey Son
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, see <http://www.gnu.org/licenses/>.
19 */
20
21 #include "qemu/osdep.h"
22 #include <sys/resource.h>
23 #include <sys/sysctl.h>
24
25 #include "qemu/help-texts.h"
26 #include "qemu/units.h"
27 #include "qemu/accel.h"
28 #include "qemu-version.h"
29 #include <machine/trap.h>
30
31 #include "qapi/error.h"
32 #include "qemu.h"
33 #include "qemu/config-file.h"
34 #include "qemu/error-report.h"
35 #include "qemu/path.h"
36 #include "qemu/help_option.h"
37 #include "qemu/module.h"
38 #include "qemu/plugin.h"
39 #include "exec/exec-all.h"
40 #include "user/guest-base.h"
41 #include "user/page-protection.h"
42 #include "tcg/startup.h"
43 #include "qemu/timer.h"
44 #include "qemu/envlist.h"
45 #include "qemu/cutils.h"
46 #include "exec/log.h"
47 #include "trace/control.h"
48 #include "crypto/init.h"
49 #include "qemu/guest-random.h"
50 #include "gdbstub/user.h"
51 #include "exec/page-vary.h"
52
53 #include "host-os.h"
54 #include "target_arch_cpu.h"
55
56
57 /*
58 * TODO: Remove these and rely only on qemu_real_host_page_size().
59 */
60 uintptr_t qemu_host_page_size;
61 intptr_t qemu_host_page_mask;
62
63 static bool opt_one_insn_per_tb;
64 static unsigned long opt_tb_size;
65 uintptr_t guest_base;
66 bool have_guest_base;
67 /*
68 * When running 32-on-64 we should make sure we can fit all of the possible
69 * guest address space into a contiguous chunk of virtual host memory.
70 *
71 * This way we will never overlap with our own libraries or binaries or stack
72 * or anything else that QEMU maps.
73 *
74 * Many cpus reserve the high bit (or more than one for some 64-bit cpus)
75 * of the address for the kernel. Some cpus rely on this and user space
76 * uses the high bit(s) for pointer tagging and the like. For them, we
77 * must preserve the expected address space.
78 */
79 #ifndef MAX_RESERVED_VA
80 # if HOST_LONG_BITS > TARGET_VIRT_ADDR_SPACE_BITS
81 # if TARGET_VIRT_ADDR_SPACE_BITS == 32 && \
82 (TARGET_LONG_BITS == 32 || defined(TARGET_ABI32))
83 # define MAX_RESERVED_VA(CPU) 0xfffffffful
84 # else
85 # define MAX_RESERVED_VA(CPU) ((1ul << TARGET_VIRT_ADDR_SPACE_BITS) - 1)
86 # endif
87 # else
88 # define MAX_RESERVED_VA(CPU) 0
89 # endif
90 #endif
91
92 unsigned long reserved_va;
93
94 const char *interp_prefix = CONFIG_QEMU_INTERP_PREFIX;
95 const char *qemu_uname_release;
96
97 unsigned long target_maxtsiz = TARGET_MAXTSIZ; /* max text size */
98 unsigned long target_dfldsiz = TARGET_DFLDSIZ; /* initial data size limit */
99 unsigned long target_maxdsiz = TARGET_MAXDSIZ; /* max data size */
100 unsigned long target_dflssiz = TARGET_DFLSSIZ; /* initial data size limit */
101 unsigned long target_maxssiz = TARGET_MAXSSIZ; /* max stack size */
102 unsigned long target_sgrowsiz = TARGET_SGROWSIZ; /* amount to grow stack */
103
104 /* Helper routines for implementing atomic operations. */
105
fork_start(void)106 void fork_start(void)
107 {
108 start_exclusive();
109 mmap_fork_start();
110 cpu_list_lock();
111 qemu_plugin_user_prefork_lock();
112 gdbserver_fork_start();
113 }
114
fork_end(pid_t pid)115 void fork_end(pid_t pid)
116 {
117 bool child = pid == 0;
118
119 qemu_plugin_user_postfork(child);
120 mmap_fork_end(child);
121 if (child) {
122 CPUState *cpu, *next_cpu;
123 /*
124 * Child processes created by fork() only have a single thread.
125 * Discard information about the parent threads.
126 */
127 CPU_FOREACH_SAFE(cpu, next_cpu) {
128 if (cpu != thread_cpu) {
129 QTAILQ_REMOVE_RCU(&cpus_queue, cpu, node);
130 }
131 }
132 qemu_init_cpu_list();
133 get_task_state(thread_cpu)->ts_tid = qemu_get_thread_id();
134 } else {
135 cpu_list_unlock();
136 }
137 gdbserver_fork_end(thread_cpu, pid);
138 /*
139 * qemu_init_cpu_list() reinitialized the child exclusive state, but we
140 * also need to keep current_cpu consistent, so call end_exclusive() for
141 * both child and parent.
142 */
143 end_exclusive();
144 }
145
cpu_loop(CPUArchState * env)146 void cpu_loop(CPUArchState *env)
147 {
148 target_cpu_loop(env);
149 }
150
usage(void)151 static void usage(void)
152 {
153 printf("qemu-" TARGET_NAME " version " QEMU_FULL_VERSION
154 "\n" QEMU_COPYRIGHT "\n"
155 "usage: qemu-" TARGET_NAME " [options] program [arguments...]\n"
156 "BSD CPU emulator (compiled for %s emulation)\n"
157 "\n"
158 "Standard options:\n"
159 "-h print this help\n"
160 "-g port wait gdb connection to port\n"
161 "-L path set the elf interpreter prefix (default=%s)\n"
162 "-s size set the stack size in bytes (default=%ld)\n"
163 "-cpu model select CPU (-cpu help for list)\n"
164 "-drop-ld-preload drop LD_PRELOAD for target process\n"
165 "-E var=value sets/modifies targets environment variable(s)\n"
166 "-U var unsets targets environment variable(s)\n"
167 "-B address set guest_base address to address\n"
168 "\n"
169 "Debug options:\n"
170 "-d item1[,...] enable logging of specified items\n"
171 " (use '-d help' for a list of log items)\n"
172 "-D logfile write logs to 'logfile' (default stderr)\n"
173 "-one-insn-per-tb run with one guest instruction per emulated TB\n"
174 "-tb-size size TCG translation block cache size\n"
175 "-strace log system calls\n"
176 "-trace [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
177 " specify tracing options\n"
178 "\n"
179 "Environment variables:\n"
180 "QEMU_STRACE Print system calls and arguments similar to the\n"
181 " 'strace' program. Enable by setting to any value.\n"
182 "You can use -E and -U options to set/unset environment variables\n"
183 "for target process. It is possible to provide several variables\n"
184 "by repeating the option. For example:\n"
185 " -E var1=val2 -E var2=val2 -U LD_PRELOAD -U LD_DEBUG\n"
186 "Note that if you provide several changes to single variable\n"
187 "last change will stay in effect.\n"
188 "\n"
189 QEMU_HELP_BOTTOM "\n"
190 ,
191 TARGET_NAME,
192 interp_prefix,
193 target_dflssiz);
194 exit(1);
195 }
196
197 __thread CPUState *thread_cpu;
198
stop_all_tasks(void)199 void stop_all_tasks(void)
200 {
201 /*
202 * We trust when using NPTL (pthreads) start_exclusive() handles thread
203 * stopping correctly.
204 */
205 start_exclusive();
206 }
207
qemu_cpu_is_self(CPUState * cpu)208 bool qemu_cpu_is_self(CPUState *cpu)
209 {
210 return thread_cpu == cpu;
211 }
212
qemu_cpu_kick(CPUState * cpu)213 void qemu_cpu_kick(CPUState *cpu)
214 {
215 cpu_exit(cpu);
216 }
217
218 /* Assumes contents are already zeroed. */
init_task_state(TaskState * ts)219 static void init_task_state(TaskState *ts)
220 {
221 ts->sigaltstack_used = (struct target_sigaltstack) {
222 .ss_sp = 0,
223 .ss_size = 0,
224 .ss_flags = TARGET_SS_DISABLE,
225 };
226 }
227
gemu_log(const char * fmt,...)228 void gemu_log(const char *fmt, ...)
229 {
230 va_list ap;
231
232 va_start(ap, fmt);
233 vfprintf(stderr, fmt, ap);
234 va_end(ap);
235 }
236
237 static void
adjust_ssize(void)238 adjust_ssize(void)
239 {
240 struct rlimit rl;
241
242 if (getrlimit(RLIMIT_STACK, &rl) != 0) {
243 return;
244 }
245
246 target_maxssiz = MIN(target_maxssiz, rl.rlim_max);
247 target_dflssiz = MIN(MAX(target_dflssiz, rl.rlim_cur), target_maxssiz);
248
249 rl.rlim_max = target_maxssiz;
250 rl.rlim_cur = target_dflssiz;
251 setrlimit(RLIMIT_STACK, &rl);
252 }
253
main(int argc,char ** argv)254 int main(int argc, char **argv)
255 {
256 const char *filename;
257 const char *cpu_model;
258 const char *cpu_type;
259 const char *log_file = NULL;
260 const char *log_mask = NULL;
261 const char *seed_optarg = NULL;
262 struct target_pt_regs regs1, *regs = ®s1;
263 struct image_info info1, *info = &info1;
264 struct bsd_binprm bprm;
265 TaskState *ts;
266 CPUArchState *env;
267 CPUState *cpu;
268 int optind, rv;
269 const char *r;
270 const char *gdbstub = NULL;
271 char **target_environ, **wrk;
272 envlist_t *envlist = NULL;
273 char *argv0 = NULL;
274 int host_page_size;
275 unsigned long max_reserved_va;
276
277 adjust_ssize();
278
279 if (argc <= 1) {
280 usage();
281 }
282
283
284 error_init(argv[0]);
285 module_call_init(MODULE_INIT_TRACE);
286 qemu_init_cpu_list();
287 module_call_init(MODULE_INIT_QOM);
288
289 envlist = envlist_create();
290
291 /*
292 * add current environment into the list
293 * envlist_setenv adds to the front of the list; to preserve environ
294 * order add from back to front
295 */
296 for (wrk = environ; *wrk != NULL; wrk++) {
297 continue;
298 }
299 while (wrk != environ) {
300 wrk--;
301 (void) envlist_setenv(envlist, *wrk);
302 }
303
304 qemu_host_page_size = getpagesize();
305 qemu_host_page_size = MAX(qemu_host_page_size, TARGET_PAGE_SIZE);
306
307 cpu_model = NULL;
308
309 qemu_add_opts(&qemu_trace_opts);
310
311 optind = 1;
312 for (;;) {
313 if (optind >= argc) {
314 break;
315 }
316 r = argv[optind];
317 if (r[0] != '-') {
318 break;
319 }
320 optind++;
321 r++;
322 if (!strcmp(r, "-")) {
323 break;
324 } else if (!strcmp(r, "d")) {
325 if (optind >= argc) {
326 break;
327 }
328 log_mask = argv[optind++];
329 } else if (!strcmp(r, "D")) {
330 if (optind >= argc) {
331 break;
332 }
333 log_file = argv[optind++];
334 } else if (!strcmp(r, "E")) {
335 r = argv[optind++];
336 if (envlist_setenv(envlist, r) != 0) {
337 usage();
338 }
339 } else if (!strcmp(r, "ignore-environment")) {
340 envlist_free(envlist);
341 envlist = envlist_create();
342 } else if (!strcmp(r, "U")) {
343 r = argv[optind++];
344 if (envlist_unsetenv(envlist, r) != 0) {
345 usage();
346 }
347 } else if (!strcmp(r, "s")) {
348 r = argv[optind++];
349 rv = qemu_strtoul(r, &r, 0, &target_dflssiz);
350 if (rv < 0 || target_dflssiz <= 0) {
351 usage();
352 }
353 if (*r == 'M') {
354 target_dflssiz *= 1024 * 1024;
355 } else if (*r == 'k' || *r == 'K') {
356 target_dflssiz *= 1024;
357 }
358 if (target_dflssiz > target_maxssiz) {
359 usage();
360 }
361 } else if (!strcmp(r, "L")) {
362 interp_prefix = argv[optind++];
363 } else if (!strcmp(r, "p")) {
364 unsigned size, want = qemu_real_host_page_size();
365
366 r = argv[optind++];
367 if (qemu_strtoui(r, NULL, 10, &size) || size != want) {
368 warn_report("Deprecated page size option cannot "
369 "change host page size (%u)", want);
370 }
371 } else if (!strcmp(r, "g")) {
372 gdbstub = g_strdup(argv[optind++]);
373 } else if (!strcmp(r, "r")) {
374 qemu_uname_release = argv[optind++];
375 } else if (!strcmp(r, "cpu")) {
376 cpu_model = argv[optind++];
377 if (is_help_option(cpu_model)) {
378 list_cpus();
379 exit(1);
380 }
381 } else if (!strcmp(r, "B")) {
382 rv = qemu_strtoul(argv[optind++], NULL, 0, &guest_base);
383 if (rv < 0) {
384 usage();
385 }
386 have_guest_base = true;
387 } else if (!strcmp(r, "drop-ld-preload")) {
388 (void) envlist_unsetenv(envlist, "LD_PRELOAD");
389 } else if (!strcmp(r, "seed")) {
390 seed_optarg = optarg;
391 } else if (!strcmp(r, "one-insn-per-tb")) {
392 opt_one_insn_per_tb = true;
393 } else if (!strcmp(r, "tb-size")) {
394 r = argv[optind++];
395 if (qemu_strtoul(r, NULL, 0, &opt_tb_size)) {
396 usage();
397 }
398 } else if (!strcmp(r, "strace")) {
399 do_strace = 1;
400 } else if (!strcmp(r, "trace")) {
401 trace_opt_parse(optarg);
402 } else if (!strcmp(r, "0")) {
403 argv0 = argv[optind++];
404 } else {
405 usage();
406 }
407 }
408
409 qemu_host_page_mask = -qemu_host_page_size;
410
411 /* init debug */
412 {
413 int mask = 0;
414 if (log_mask) {
415 mask = qemu_str_to_log_mask(log_mask);
416 if (!mask) {
417 qemu_print_log_usage(stdout);
418 exit(1);
419 }
420 }
421 qemu_set_log_filename_flags(log_file, mask, &error_fatal);
422 }
423
424 if (optind >= argc) {
425 usage();
426 }
427 filename = argv[optind];
428 if (argv0) {
429 argv[optind] = argv0;
430 }
431
432 if (!trace_init_backends()) {
433 exit(1);
434 }
435 trace_init_file();
436
437 /* Zero out regs */
438 memset(regs, 0, sizeof(struct target_pt_regs));
439
440 /* Zero bsd params */
441 memset(&bprm, 0, sizeof(bprm));
442
443 /* Zero out image_info */
444 memset(info, 0, sizeof(struct image_info));
445
446 /* Scan interp_prefix dir for replacement files. */
447 init_paths(interp_prefix);
448
449 if (cpu_model == NULL) {
450 cpu_model = TARGET_DEFAULT_CPU_MODEL;
451 }
452
453 cpu_type = parse_cpu_option(cpu_model);
454
455 /* init tcg before creating CPUs and to get qemu_host_page_size */
456 {
457 AccelState *accel = current_accel();
458 AccelClass *ac = ACCEL_GET_CLASS(accel);
459
460 accel_init_interfaces(ac);
461 object_property_set_bool(OBJECT(accel), "one-insn-per-tb",
462 opt_one_insn_per_tb, &error_abort);
463 object_property_set_int(OBJECT(accel), "tb-size",
464 opt_tb_size, &error_abort);
465 ac->init_machine(NULL);
466 }
467
468 /*
469 * Finalize page size before creating CPUs.
470 * This will do nothing if !TARGET_PAGE_BITS_VARY.
471 * The most efficient setting is to match the host.
472 */
473 host_page_size = qemu_real_host_page_size();
474 set_preferred_target_page_bits(ctz32(host_page_size));
475 finalize_target_page_bits();
476
477 cpu = cpu_create(cpu_type);
478 env = cpu_env(cpu);
479 cpu_reset(cpu);
480 thread_cpu = cpu;
481
482 /*
483 * Reserving too much vm space via mmap can run into problems with rlimits,
484 * oom due to page table creation, etc. We will still try it, if directed
485 * by the command-line option, but not by default. Unless we're running a
486 * target address space of 32 or fewer bits on a host with 64 bits.
487 */
488 max_reserved_va = MAX_RESERVED_VA(cpu);
489 if (reserved_va != 0) {
490 if ((reserved_va + 1) % host_page_size) {
491 char *s = size_to_str(host_page_size);
492 fprintf(stderr, "Reserved virtual address not aligned mod %s\n", s);
493 g_free(s);
494 exit(EXIT_FAILURE);
495 }
496 if (max_reserved_va && reserved_va > max_reserved_va) {
497 fprintf(stderr, "Reserved virtual address too big\n");
498 exit(EXIT_FAILURE);
499 }
500 } else if (HOST_LONG_BITS == 64 && TARGET_VIRT_ADDR_SPACE_BITS <= 32) {
501 /* MAX_RESERVED_VA + 1 is a large power of 2, so is aligned. */
502 reserved_va = max_reserved_va;
503 }
504
505 if (getenv("QEMU_STRACE")) {
506 do_strace = 1;
507 }
508
509 target_environ = envlist_to_environ(envlist, NULL);
510 envlist_free(envlist);
511
512 {
513 Error *err = NULL;
514 if (seed_optarg != NULL) {
515 qemu_guest_random_seed_main(seed_optarg, &err);
516 } else {
517 qcrypto_init(&err);
518 }
519 if (err) {
520 error_reportf_err(err, "cannot initialize crypto: ");
521 exit(1);
522 }
523 }
524
525 /*
526 * Now that page sizes are configured we can do
527 * proper page alignment for guest_base.
528 */
529 if (have_guest_base) {
530 if (guest_base & ~qemu_host_page_mask) {
531 error_report("Selected guest base not host page aligned");
532 exit(1);
533 }
534 }
535
536 /*
537 * If reserving host virtual address space, do so now.
538 * Combined with '-B', ensure that the chosen range is free.
539 */
540 if (reserved_va) {
541 void *p;
542
543 if (have_guest_base) {
544 p = mmap((void *)guest_base, reserved_va + 1, PROT_NONE,
545 MAP_ANON | MAP_PRIVATE | MAP_FIXED | MAP_EXCL, -1, 0);
546 } else {
547 p = mmap(NULL, reserved_va + 1, PROT_NONE,
548 MAP_ANON | MAP_PRIVATE, -1, 0);
549 }
550 if (p == MAP_FAILED) {
551 const char *err = strerror(errno);
552 char *sz = size_to_str(reserved_va + 1);
553
554 if (have_guest_base) {
555 error_report("Cannot allocate %s bytes at -B %p for guest "
556 "address space: %s", sz, (void *)guest_base, err);
557 } else {
558 error_report("Cannot allocate %s bytes for guest "
559 "address space: %s", sz, err);
560 }
561 exit(1);
562 }
563 guest_base = (uintptr_t)p;
564 have_guest_base = true;
565
566 /* Ensure that mmap_next_start is within range. */
567 if (reserved_va <= mmap_next_start) {
568 mmap_next_start = (reserved_va / 4 * 3)
569 & TARGET_PAGE_MASK & qemu_host_page_mask;
570 }
571 }
572
573 if (loader_exec(filename, argv + optind, target_environ, regs, info,
574 &bprm) != 0) {
575 printf("Error loading %s\n", filename);
576 _exit(1);
577 }
578
579 for (wrk = target_environ; *wrk; wrk++) {
580 g_free(*wrk);
581 }
582
583 g_free(target_environ);
584
585 if (qemu_loglevel_mask(CPU_LOG_PAGE)) {
586 FILE *f = qemu_log_trylock();
587 if (f) {
588 fprintf(f, "guest_base %p\n", (void *)guest_base);
589 fprintf(f, "page layout changed following binary load\n");
590 page_dump(f);
591
592 fprintf(f, "end_code 0x" TARGET_ABI_FMT_lx "\n",
593 info->end_code);
594 fprintf(f, "start_code 0x" TARGET_ABI_FMT_lx "\n",
595 info->start_code);
596 fprintf(f, "start_data 0x" TARGET_ABI_FMT_lx "\n",
597 info->start_data);
598 fprintf(f, "end_data 0x" TARGET_ABI_FMT_lx "\n",
599 info->end_data);
600 fprintf(f, "start_stack 0x" TARGET_ABI_FMT_lx "\n",
601 info->start_stack);
602 fprintf(f, "brk 0x" TARGET_ABI_FMT_lx "\n", info->brk);
603 fprintf(f, "entry 0x" TARGET_ABI_FMT_lx "\n", info->entry);
604
605 qemu_log_unlock(f);
606 }
607 }
608
609 /* build Task State */
610 ts = g_new0(TaskState, 1);
611 init_task_state(ts);
612 ts->info = info;
613 ts->bprm = &bprm;
614 ts->ts_tid = qemu_get_thread_id();
615 cpu->opaque = ts;
616
617 target_set_brk(info->brk);
618 syscall_init();
619 signal_init();
620
621 /*
622 * Now that we've loaded the binary, GUEST_BASE is fixed. Delay
623 * generating the prologue until now so that the prologue can take
624 * the real value of GUEST_BASE into account.
625 */
626 tcg_prologue_init();
627
628 target_cpu_init(env, regs);
629
630 if (gdbstub) {
631 gdbserver_start(gdbstub, &error_fatal);
632 }
633 cpu_loop(env);
634 /* never exits */
635 return 0;
636 }
637