xref: /openbmc/qemu/bsd-user/main.c (revision 322b038c)
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 "exec/exec-all.h"
39 #include "tcg/startup.h"
40 #include "qemu/timer.h"
41 #include "qemu/envlist.h"
42 #include "qemu/cutils.h"
43 #include "exec/log.h"
44 #include "trace/control.h"
45 #include "crypto/init.h"
46 #include "qemu/guest-random.h"
47 #include "gdbstub/user.h"
48 
49 #include "host-os.h"
50 #include "target_arch_cpu.h"
51 
52 static bool opt_one_insn_per_tb;
53 uintptr_t guest_base;
54 bool have_guest_base;
55 /*
56  * When running 32-on-64 we should make sure we can fit all of the possible
57  * guest address space into a contiguous chunk of virtual host memory.
58  *
59  * This way we will never overlap with our own libraries or binaries or stack
60  * or anything else that QEMU maps.
61  *
62  * Many cpus reserve the high bit (or more than one for some 64-bit cpus)
63  * of the address for the kernel.  Some cpus rely on this and user space
64  * uses the high bit(s) for pointer tagging and the like.  For them, we
65  * must preserve the expected address space.
66  */
67 #ifndef MAX_RESERVED_VA
68 # if HOST_LONG_BITS > TARGET_VIRT_ADDR_SPACE_BITS
69 #  if TARGET_VIRT_ADDR_SPACE_BITS == 32 && \
70       (TARGET_LONG_BITS == 32 || defined(TARGET_ABI32))
71 #   define MAX_RESERVED_VA  0xfffffffful
72 #  else
73 #   define MAX_RESERVED_VA  ((1ul << TARGET_VIRT_ADDR_SPACE_BITS) - 1)
74 #  endif
75 # else
76 #  define MAX_RESERVED_VA  0
77 # endif
78 #endif
79 
80 /*
81  * That said, reserving *too* much vm space via mmap can run into problems
82  * with rlimits, oom due to page table creation, etc.  We will still try it,
83  * if directed by the command-line option, but not by default.
84  */
85 #if HOST_LONG_BITS == 64 && TARGET_VIRT_ADDR_SPACE_BITS <= 32
86 unsigned long reserved_va = MAX_RESERVED_VA;
87 #else
88 unsigned long reserved_va;
89 #endif
90 
91 const char *interp_prefix = CONFIG_QEMU_INTERP_PREFIX;
92 const char *qemu_uname_release;
93 char qemu_proc_pathname[PATH_MAX];  /* full path to exeutable */
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 
104 void fork_start(void)
105 {
106     start_exclusive();
107     cpu_list_lock();
108     mmap_fork_start();
109 }
110 
111 void fork_end(int child)
112 {
113     if (child) {
114         CPUState *cpu, *next_cpu;
115         /*
116          * Child processes created by fork() only have a single thread.  Discard
117          * information about the parent threads.
118          */
119         CPU_FOREACH_SAFE(cpu, next_cpu) {
120             if (cpu != thread_cpu) {
121                 QTAILQ_REMOVE_RCU(&cpus_queue, cpu, node);
122             }
123         }
124         mmap_fork_end(child);
125         /*
126          * qemu_init_cpu_list() takes care of reinitializing the exclusive
127          * state, so we don't need to end_exclusive() here.
128          */
129         qemu_init_cpu_list();
130         gdbserver_fork(thread_cpu);
131     } else {
132         mmap_fork_end(child);
133         cpu_list_unlock();
134         end_exclusive();
135     }
136 }
137 
138 void cpu_loop(CPUArchState *env)
139 {
140     target_cpu_loop(env);
141 }
142 
143 static void usage(void)
144 {
145     printf("qemu-" TARGET_NAME " version " QEMU_FULL_VERSION
146            "\n" QEMU_COPYRIGHT "\n"
147            "usage: qemu-" TARGET_NAME " [options] program [arguments...]\n"
148            "BSD CPU emulator (compiled for %s emulation)\n"
149            "\n"
150            "Standard options:\n"
151            "-h                print this help\n"
152            "-g port           wait gdb connection to port\n"
153            "-L path           set the elf interpreter prefix (default=%s)\n"
154            "-s size           set the stack size in bytes (default=%ld)\n"
155            "-cpu model        select CPU (-cpu help for list)\n"
156            "-drop-ld-preload  drop LD_PRELOAD for target process\n"
157            "-E var=value      sets/modifies targets environment variable(s)\n"
158            "-U var            unsets targets environment variable(s)\n"
159            "-B address        set guest_base address to address\n"
160            "\n"
161            "Debug options:\n"
162            "-d item1[,...]    enable logging of specified items\n"
163            "                  (use '-d help' for a list of log items)\n"
164            "-D logfile        write logs to 'logfile' (default stderr)\n"
165            "-one-insn-per-tb  run with one guest instruction per emulated TB\n"
166            "-singlestep       deprecated synonym for -one-insn-per-tb\n"
167            "-strace           log system calls\n"
168            "-trace            [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
169            "                  specify tracing options\n"
170            "\n"
171            "Environment variables:\n"
172            "QEMU_STRACE       Print system calls and arguments similar to the\n"
173            "                  'strace' program.  Enable by setting to any value.\n"
174            "You can use -E and -U options to set/unset environment variables\n"
175            "for target process.  It is possible to provide several variables\n"
176            "by repeating the option.  For example:\n"
177            "    -E var1=val2 -E var2=val2 -U LD_PRELOAD -U LD_DEBUG\n"
178            "Note that if you provide several changes to single variable\n"
179            "last change will stay in effect.\n"
180            "\n"
181            QEMU_HELP_BOTTOM "\n"
182            ,
183            TARGET_NAME,
184            interp_prefix,
185            target_dflssiz);
186     exit(1);
187 }
188 
189 __thread CPUState *thread_cpu;
190 
191 void stop_all_tasks(void)
192 {
193     /*
194      * We trust when using NPTL (pthreads) start_exclusive() handles thread
195      * stopping correctly.
196      */
197     start_exclusive();
198 }
199 
200 bool qemu_cpu_is_self(CPUState *cpu)
201 {
202     return thread_cpu == cpu;
203 }
204 
205 void qemu_cpu_kick(CPUState *cpu)
206 {
207     cpu_exit(cpu);
208 }
209 
210 /* Assumes contents are already zeroed.  */
211 static void init_task_state(TaskState *ts)
212 {
213     ts->sigaltstack_used = (struct target_sigaltstack) {
214         .ss_sp = 0,
215         .ss_size = 0,
216         .ss_flags = TARGET_SS_DISABLE,
217     };
218 }
219 
220 void gemu_log(const char *fmt, ...)
221 {
222     va_list ap;
223 
224     va_start(ap, fmt);
225     vfprintf(stderr, fmt, ap);
226     va_end(ap);
227 }
228 
229 static void
230 adjust_ssize(void)
231 {
232     struct rlimit rl;
233 
234     if (getrlimit(RLIMIT_STACK, &rl) != 0) {
235         return;
236     }
237 
238     target_maxssiz = MIN(target_maxssiz, rl.rlim_max);
239     target_dflssiz = MIN(MAX(target_dflssiz, rl.rlim_cur), target_maxssiz);
240 
241     rl.rlim_max = target_maxssiz;
242     rl.rlim_cur = target_dflssiz;
243     setrlimit(RLIMIT_STACK, &rl);
244 }
245 
246 static void save_proc_pathname(char *argv0)
247 {
248     int mib[4];
249     size_t len;
250 
251     mib[0] = CTL_KERN;
252     mib[1] = KERN_PROC;
253     mib[2] = KERN_PROC_PATHNAME;
254     mib[3] = -1;
255 
256     len = sizeof(qemu_proc_pathname);
257     if (sysctl(mib, 4, qemu_proc_pathname, &len, NULL, 0)) {
258         perror("sysctl");
259     }
260 }
261 
262 int main(int argc, char **argv)
263 {
264     const char *filename;
265     const char *cpu_model;
266     const char *cpu_type;
267     const char *log_file = NULL;
268     const char *log_mask = NULL;
269     const char *seed_optarg = NULL;
270     struct target_pt_regs regs1, *regs = &regs1;
271     struct image_info info1, *info = &info1;
272     struct bsd_binprm bprm;
273     TaskState *ts;
274     CPUArchState *env;
275     CPUState *cpu;
276     int optind, rv;
277     const char *r;
278     const char *gdbstub = NULL;
279     char **target_environ, **wrk;
280     envlist_t *envlist = NULL;
281     char *argv0 = NULL;
282 
283     adjust_ssize();
284 
285     if (argc <= 1) {
286         usage();
287     }
288 
289     save_proc_pathname(argv[0]);
290 
291     error_init(argv[0]);
292     module_call_init(MODULE_INIT_TRACE);
293     qemu_init_cpu_list();
294     module_call_init(MODULE_INIT_QOM);
295 
296     envlist = envlist_create();
297 
298     /*
299      * add current environment into the list
300      * envlist_setenv adds to the front of the list; to preserve environ
301      * order add from back to front
302      */
303     for (wrk = environ; *wrk != NULL; wrk++) {
304         continue;
305     }
306     while (wrk != environ) {
307         wrk--;
308         (void) envlist_setenv(envlist, *wrk);
309     }
310 
311     cpu_model = NULL;
312 
313     qemu_add_opts(&qemu_trace_opts);
314 
315     optind = 1;
316     for (;;) {
317         if (optind >= argc) {
318             break;
319         }
320         r = argv[optind];
321         if (r[0] != '-') {
322             break;
323         }
324         optind++;
325         r++;
326         if (!strcmp(r, "-")) {
327             break;
328         } else if (!strcmp(r, "d")) {
329             if (optind >= argc) {
330                 break;
331             }
332             log_mask = argv[optind++];
333         } else if (!strcmp(r, "D")) {
334             if (optind >= argc) {
335                 break;
336             }
337             log_file = argv[optind++];
338         } else if (!strcmp(r, "E")) {
339             r = argv[optind++];
340             if (envlist_setenv(envlist, r) != 0) {
341                 usage();
342             }
343         } else if (!strcmp(r, "ignore-environment")) {
344             envlist_free(envlist);
345             envlist = envlist_create();
346         } else if (!strcmp(r, "U")) {
347             r = argv[optind++];
348             if (envlist_unsetenv(envlist, r) != 0) {
349                 usage();
350             }
351         } else if (!strcmp(r, "s")) {
352             r = argv[optind++];
353             rv = qemu_strtoul(r, &r, 0, &target_dflssiz);
354             if (rv < 0 || target_dflssiz <= 0) {
355                 usage();
356             }
357             if (*r == 'M') {
358                 target_dflssiz *= 1024 * 1024;
359             } else if (*r == 'k' || *r == 'K') {
360                 target_dflssiz *= 1024;
361             }
362             if (target_dflssiz > target_maxssiz) {
363                 usage();
364             }
365         } else if (!strcmp(r, "L")) {
366             interp_prefix = argv[optind++];
367         } else if (!strcmp(r, "p")) {
368             qemu_host_page_size = atoi(argv[optind++]);
369             if (qemu_host_page_size == 0 ||
370                 (qemu_host_page_size & (qemu_host_page_size - 1)) != 0) {
371                 fprintf(stderr, "page size must be a power of two\n");
372                 exit(1);
373             }
374         } else if (!strcmp(r, "g")) {
375             gdbstub = g_strdup(argv[optind++]);
376         } else if (!strcmp(r, "r")) {
377             qemu_uname_release = argv[optind++];
378         } else if (!strcmp(r, "cpu")) {
379             cpu_model = argv[optind++];
380             if (is_help_option(cpu_model)) {
381                 list_cpus();
382                 exit(1);
383             }
384         } else if (!strcmp(r, "B")) {
385             rv = qemu_strtoul(argv[optind++], NULL, 0, &guest_base);
386             if (rv < 0) {
387                 usage();
388             }
389             have_guest_base = true;
390         } else if (!strcmp(r, "drop-ld-preload")) {
391             (void) envlist_unsetenv(envlist, "LD_PRELOAD");
392         } else if (!strcmp(r, "seed")) {
393             seed_optarg = optarg;
394         } else if (!strcmp(r, "singlestep") || !strcmp(r, "one-insn-per-tb")) {
395             opt_one_insn_per_tb = true;
396         } else if (!strcmp(r, "strace")) {
397             do_strace = 1;
398         } else if (!strcmp(r, "trace")) {
399             trace_opt_parse(optarg);
400         } else if (!strcmp(r, "0")) {
401             argv0 = argv[optind++];
402         } else {
403             usage();
404         }
405     }
406 
407     /* init debug */
408     {
409         int mask = 0;
410         if (log_mask) {
411             mask = qemu_str_to_log_mask(log_mask);
412             if (!mask) {
413                 qemu_print_log_usage(stdout);
414                 exit(1);
415             }
416         }
417         qemu_set_log_filename_flags(log_file, mask, &error_fatal);
418     }
419 
420     if (optind >= argc) {
421         usage();
422     }
423     filename = argv[optind];
424     if (argv0) {
425         argv[optind] = argv0;
426     }
427 
428     if (!trace_init_backends()) {
429         exit(1);
430     }
431     trace_init_file();
432 
433     /* Zero out regs */
434     memset(regs, 0, sizeof(struct target_pt_regs));
435 
436     /* Zero bsd params */
437     memset(&bprm, 0, sizeof(bprm));
438 
439     /* Zero out image_info */
440     memset(info, 0, sizeof(struct image_info));
441 
442     /* Scan interp_prefix dir for replacement files. */
443     init_paths(interp_prefix);
444 
445     if (cpu_model == NULL) {
446         cpu_model = TARGET_DEFAULT_CPU_MODEL;
447     }
448 
449     cpu_type = parse_cpu_option(cpu_model);
450 
451     /* init tcg before creating CPUs and to get qemu_host_page_size */
452     {
453         AccelState *accel = current_accel();
454         AccelClass *ac = ACCEL_GET_CLASS(accel);
455 
456         accel_init_interfaces(ac);
457         object_property_set_bool(OBJECT(accel), "one-insn-per-tb",
458                                  opt_one_insn_per_tb, &error_abort);
459         ac->init_machine(NULL);
460     }
461     cpu = cpu_create(cpu_type);
462     env = cpu_env(cpu);
463     cpu_reset(cpu);
464     thread_cpu = cpu;
465 
466     if (getenv("QEMU_STRACE")) {
467         do_strace = 1;
468     }
469 
470     target_environ = envlist_to_environ(envlist, NULL);
471     envlist_free(envlist);
472 
473     {
474         Error *err = NULL;
475         if (seed_optarg != NULL) {
476             qemu_guest_random_seed_main(seed_optarg, &err);
477         } else {
478             qcrypto_init(&err);
479         }
480         if (err) {
481             error_reportf_err(err, "cannot initialize crypto: ");
482             exit(1);
483         }
484     }
485 
486     /*
487      * Now that page sizes are configured we can do
488      * proper page alignment for guest_base.
489      */
490     if (have_guest_base) {
491         if (guest_base & ~qemu_host_page_mask) {
492             error_report("Selected guest base not host page aligned");
493             exit(1);
494         }
495     }
496 
497     /*
498      * If reserving host virtual address space, do so now.
499      * Combined with '-B', ensure that the chosen range is free.
500      */
501     if (reserved_va) {
502         void *p;
503 
504         if (have_guest_base) {
505             p = mmap((void *)guest_base, reserved_va + 1, PROT_NONE,
506                      MAP_ANON | MAP_PRIVATE | MAP_FIXED | MAP_EXCL, -1, 0);
507         } else {
508             p = mmap(NULL, reserved_va + 1, PROT_NONE,
509                      MAP_ANON | MAP_PRIVATE, -1, 0);
510         }
511         if (p == MAP_FAILED) {
512             const char *err = strerror(errno);
513             char *sz = size_to_str(reserved_va + 1);
514 
515             if (have_guest_base) {
516                 error_report("Cannot allocate %s bytes at -B %p for guest "
517                              "address space: %s", sz, (void *)guest_base, err);
518             } else {
519                 error_report("Cannot allocate %s bytes for guest "
520                              "address space: %s", sz, err);
521             }
522             exit(1);
523         }
524         guest_base = (uintptr_t)p;
525         have_guest_base = true;
526 
527         /* Ensure that mmap_next_start is within range. */
528         if (reserved_va <= mmap_next_start) {
529             mmap_next_start = (reserved_va / 4 * 3)
530                               & TARGET_PAGE_MASK & qemu_host_page_mask;
531         }
532     }
533 
534     if (loader_exec(filename, argv + optind, target_environ, regs, info,
535                     &bprm) != 0) {
536         printf("Error loading %s\n", filename);
537         _exit(1);
538     }
539 
540     for (wrk = target_environ; *wrk; wrk++) {
541         g_free(*wrk);
542     }
543 
544     g_free(target_environ);
545 
546     if (qemu_loglevel_mask(CPU_LOG_PAGE)) {
547         FILE *f = qemu_log_trylock();
548         if (f) {
549             fprintf(f, "guest_base  %p\n", (void *)guest_base);
550             fprintf(f, "page layout changed following binary load\n");
551             page_dump(f);
552 
553             fprintf(f, "end_code    0x" TARGET_ABI_FMT_lx "\n",
554                     info->end_code);
555             fprintf(f, "start_code  0x" TARGET_ABI_FMT_lx "\n",
556                     info->start_code);
557             fprintf(f, "start_data  0x" TARGET_ABI_FMT_lx "\n",
558                     info->start_data);
559             fprintf(f, "end_data    0x" TARGET_ABI_FMT_lx "\n",
560                     info->end_data);
561             fprintf(f, "start_stack 0x" TARGET_ABI_FMT_lx "\n",
562                     info->start_stack);
563             fprintf(f, "brk         0x" TARGET_ABI_FMT_lx "\n", info->brk);
564             fprintf(f, "entry       0x" TARGET_ABI_FMT_lx "\n", info->entry);
565 
566             qemu_log_unlock(f);
567         }
568     }
569 
570     /* build Task State */
571     ts = g_new0(TaskState, 1);
572     init_task_state(ts);
573     ts->info = info;
574     ts->bprm = &bprm;
575     cpu->opaque = ts;
576 
577     target_set_brk(info->brk);
578     syscall_init();
579     signal_init();
580 
581     /*
582      * Now that we've loaded the binary, GUEST_BASE is fixed.  Delay
583      * generating the prologue until now so that the prologue can take
584      * the real value of GUEST_BASE into account.
585      */
586     tcg_prologue_init();
587 
588     target_cpu_init(env, regs);
589 
590     if (gdbstub) {
591         gdbserver_start(gdbstub);
592         gdb_handlesig(cpu, 0);
593     }
594     cpu_loop(env);
595     /* never exits */
596     return 0;
597 }
598