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