xref: /openbmc/qemu/bsd-user/main.c (revision f96b157ebb93f94cd56ebbc99bc20982b8fd86ef)
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 "user/guest-base.h"
40 #include "user/page-protection.h"
41 #include "accel/accel-ops.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 unsigned long guest_addr_max;
94 
95 const char *interp_prefix = CONFIG_QEMU_INTERP_PREFIX;
96 const char *qemu_uname_release;
97 
98 unsigned long target_maxtsiz = TARGET_MAXTSIZ;   /* max text size */
99 unsigned long target_dfldsiz = TARGET_DFLDSIZ;   /* initial data size limit */
100 unsigned long target_maxdsiz = TARGET_MAXDSIZ;   /* max data size */
101 unsigned long target_dflssiz = TARGET_DFLSSIZ;   /* initial data size limit */
102 unsigned long target_maxssiz = TARGET_MAXSSIZ;   /* max stack size */
103 unsigned long target_sgrowsiz = TARGET_SGROWSIZ; /* amount to grow stack */
104 
105 /* Helper routines for implementing atomic operations. */
106 
fork_start(void)107 void fork_start(void)
108 {
109     start_exclusive();
110     mmap_fork_start();
111     cpu_list_lock();
112     qemu_plugin_user_prefork_lock();
113     gdbserver_fork_start();
114 }
115 
fork_end(pid_t pid)116 void fork_end(pid_t pid)
117 {
118     bool child = pid == 0;
119 
120     qemu_plugin_user_postfork(child);
121     mmap_fork_end(child);
122     if (child) {
123         CPUState *cpu, *next_cpu;
124         /*
125          * Child processes created by fork() only have a single thread.
126          * Discard information about the parent threads.
127          */
128         CPU_FOREACH_SAFE(cpu, next_cpu) {
129             if (cpu != thread_cpu) {
130                 QTAILQ_REMOVE_RCU(&cpus_queue, cpu, node);
131             }
132         }
133         qemu_init_cpu_list();
134         get_task_state(thread_cpu)->ts_tid = qemu_get_thread_id();
135     } else {
136         cpu_list_unlock();
137     }
138     gdbserver_fork_end(thread_cpu, pid);
139     /*
140      * qemu_init_cpu_list() reinitialized the child exclusive state, but we
141      * also need to keep current_cpu consistent, so call end_exclusive() for
142      * both child and parent.
143      */
144     end_exclusive();
145 }
146 
cpu_loop(CPUArchState * env)147 void cpu_loop(CPUArchState *env)
148 {
149     target_cpu_loop(env);
150 }
151 
usage(void)152 static void usage(void)
153 {
154     printf("qemu-" TARGET_NAME " version " QEMU_FULL_VERSION
155            "\n" QEMU_COPYRIGHT "\n"
156            "usage: qemu-" TARGET_NAME " [options] program [arguments...]\n"
157            "BSD CPU emulator (compiled for %s emulation)\n"
158            "\n"
159            "Standard options:\n"
160            "-h                print this help\n"
161            "-g port           wait gdb connection to port\n"
162            "-L path           set the elf interpreter prefix (default=%s)\n"
163            "-s size           set the stack size in bytes (default=%ld)\n"
164            "-cpu model        select CPU (-cpu help for list)\n"
165            "-drop-ld-preload  drop LD_PRELOAD for target process\n"
166            "-E var=value      sets/modifies targets environment variable(s)\n"
167            "-U var            unsets targets environment variable(s)\n"
168            "-B address        set guest_base address to address\n"
169            "\n"
170            "Debug options:\n"
171            "-d item1[,...]    enable logging of specified items\n"
172            "                  (use '-d help' for a list of log items)\n"
173            "-D logfile        write logs to 'logfile' (default stderr)\n"
174            "-one-insn-per-tb  run with one guest instruction per emulated TB\n"
175            "-tb-size size     TCG translation block cache size\n"
176            "-strace           log system calls\n"
177            "-trace            [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
178            "                  specify tracing options\n"
179 #ifdef CONFIG_PLUGIN
180            "-plugin           [file=]<file>[,<argname>=<argvalue>]\n"
181 #endif
182            "\n"
183            "Environment variables:\n"
184            "QEMU_STRACE       Print system calls and arguments similar to the\n"
185            "                  'strace' program.  Enable by setting to any value.\n"
186            "You can use -E and -U options to set/unset environment variables\n"
187            "for target process.  It is possible to provide several variables\n"
188            "by repeating the option.  For example:\n"
189            "    -E var1=val2 -E var2=val2 -U LD_PRELOAD -U LD_DEBUG\n"
190            "Note that if you provide several changes to single variable\n"
191            "last change will stay in effect.\n"
192            "\n"
193            QEMU_HELP_BOTTOM "\n"
194            ,
195            TARGET_NAME,
196            interp_prefix,
197            target_dflssiz);
198     exit(1);
199 }
200 
201 __thread CPUState *thread_cpu;
202 
stop_all_tasks(void)203 void stop_all_tasks(void)
204 {
205     /*
206      * We trust when using NPTL (pthreads) start_exclusive() handles thread
207      * stopping correctly.
208      */
209     start_exclusive();
210 }
211 
qemu_cpu_is_self(CPUState * cpu)212 bool qemu_cpu_is_self(CPUState *cpu)
213 {
214     return thread_cpu == cpu;
215 }
216 
qemu_cpu_kick(CPUState * cpu)217 void qemu_cpu_kick(CPUState *cpu)
218 {
219     cpu_exit(cpu);
220 }
221 
222 /* Assumes contents are already zeroed.  */
init_task_state(TaskState * ts)223 static void init_task_state(TaskState *ts)
224 {
225     ts->sigaltstack_used = (struct target_sigaltstack) {
226         .ss_sp = 0,
227         .ss_size = 0,
228         .ss_flags = TARGET_SS_DISABLE,
229     };
230 }
231 
232 static QemuPluginList plugins = QTAILQ_HEAD_INITIALIZER(plugins);
233 
gemu_log(const char * fmt,...)234 void gemu_log(const char *fmt, ...)
235 {
236     va_list ap;
237 
238     va_start(ap, fmt);
239     vfprintf(stderr, fmt, ap);
240     va_end(ap);
241 }
242 
243 static void
adjust_ssize(void)244 adjust_ssize(void)
245 {
246     struct rlimit rl;
247 
248     if (getrlimit(RLIMIT_STACK, &rl) != 0) {
249         return;
250     }
251 
252     target_maxssiz = MIN(target_maxssiz, rl.rlim_max);
253     target_dflssiz = MIN(MAX(target_dflssiz, rl.rlim_cur), target_maxssiz);
254 
255     rl.rlim_max = target_maxssiz;
256     rl.rlim_cur = target_dflssiz;
257     setrlimit(RLIMIT_STACK, &rl);
258 }
259 
main(int argc,char ** argv)260 int main(int argc, char **argv)
261 {
262     const char *filename;
263     const char *cpu_model;
264     const char *cpu_type;
265     const char *log_file = NULL;
266     const char *log_mask = NULL;
267     const char *seed_optarg = NULL;
268     struct target_pt_regs regs1, *regs = &regs1;
269     struct image_info info1, *info = &info1;
270     struct bsd_binprm bprm;
271     TaskState *ts;
272     CPUArchState *env;
273     CPUState *cpu;
274     int optind, rv;
275     const char *r;
276     const char *gdbstub = NULL;
277     char **target_environ, **wrk;
278     envlist_t *envlist = NULL;
279     char *argv0 = NULL;
280     int host_page_size;
281     unsigned long max_reserved_va;
282 
283     adjust_ssize();
284 
285     if (argc <= 1) {
286         usage();
287     }
288 
289 
290     error_init(argv[0]);
291     module_call_init(MODULE_INIT_TRACE);
292     qemu_init_cpu_list();
293     module_call_init(MODULE_INIT_QOM);
294 
295     envlist = envlist_create();
296 
297     /*
298      * add current environment into the list
299      * envlist_setenv adds to the front of the list; to preserve environ
300      * order add from back to front
301      */
302     for (wrk = environ; *wrk != NULL; wrk++) {
303         continue;
304     }
305     while (wrk != environ) {
306         wrk--;
307         (void) envlist_setenv(envlist, *wrk);
308     }
309 
310     qemu_host_page_size = getpagesize();
311     qemu_host_page_size = MAX(qemu_host_page_size, TARGET_PAGE_SIZE);
312 
313     cpu_model = NULL;
314 
315     qemu_add_opts(&qemu_trace_opts);
316     qemu_plugin_add_opts();
317 
318     optind = 1;
319     for (;;) {
320         if (optind >= argc) {
321             break;
322         }
323         r = argv[optind];
324         if (r[0] != '-') {
325             break;
326         }
327         optind++;
328         r++;
329         if (!strcmp(r, "-")) {
330             break;
331         } else if (!strcmp(r, "d")) {
332             if (optind >= argc) {
333                 break;
334             }
335             log_mask = argv[optind++];
336         } else if (!strcmp(r, "D")) {
337             if (optind >= argc) {
338                 break;
339             }
340             log_file = argv[optind++];
341         } else if (!strcmp(r, "E")) {
342             r = argv[optind++];
343             if (envlist_setenv(envlist, r) != 0) {
344                 usage();
345             }
346         } else if (!strcmp(r, "ignore-environment")) {
347             envlist_free(envlist);
348             envlist = envlist_create();
349         } else if (!strcmp(r, "U")) {
350             r = argv[optind++];
351             if (envlist_unsetenv(envlist, r) != 0) {
352                 usage();
353             }
354         } else if (!strcmp(r, "s")) {
355             r = argv[optind++];
356             rv = qemu_strtoul(r, &r, 0, &target_dflssiz);
357             if (rv < 0 || target_dflssiz <= 0) {
358                 usage();
359             }
360             if (*r == 'M') {
361                 target_dflssiz *= 1024 * 1024;
362             } else if (*r == 'k' || *r == 'K') {
363                 target_dflssiz *= 1024;
364             }
365             if (target_dflssiz > target_maxssiz) {
366                 usage();
367             }
368         } else if (!strcmp(r, "L")) {
369             interp_prefix = argv[optind++];
370         } else if (!strcmp(r, "p")) {
371             unsigned size, want = qemu_real_host_page_size();
372 
373             r = argv[optind++];
374             if (qemu_strtoui(r, NULL, 10, &size) || size != want) {
375                 warn_report("Deprecated page size option cannot "
376                             "change host page size (%u)", want);
377             }
378         } else if (!strcmp(r, "g")) {
379             gdbstub = g_strdup(argv[optind++]);
380         } else if (!strcmp(r, "r")) {
381             qemu_uname_release = argv[optind++];
382         } else if (!strcmp(r, "cpu")) {
383             cpu_model = argv[optind++];
384             if (is_help_option(cpu_model)) {
385                 list_cpus();
386                 exit(1);
387             }
388         } else if (!strcmp(r, "B")) {
389             rv = qemu_strtoul(argv[optind++], NULL, 0, &guest_base);
390             if (rv < 0) {
391                 usage();
392             }
393             have_guest_base = true;
394         } else if (!strcmp(r, "drop-ld-preload")) {
395             (void) envlist_unsetenv(envlist, "LD_PRELOAD");
396         } else if (!strcmp(r, "seed")) {
397             seed_optarg = optarg;
398         } else if (!strcmp(r, "one-insn-per-tb")) {
399             opt_one_insn_per_tb = true;
400         } else if (!strcmp(r, "tb-size")) {
401             r = argv[optind++];
402             if (qemu_strtoul(r, NULL, 0, &opt_tb_size)) {
403                 usage();
404             }
405         } else if (!strcmp(r, "strace")) {
406             do_strace = 1;
407         } else if (!strcmp(r, "trace")) {
408             trace_opt_parse(optarg);
409 #ifdef CONFIG_PLUGIN
410         } else if (!strcmp(r, "plugin")) {
411             r = argv[optind++];
412             qemu_plugin_opt_parse(r, &plugins);
413 #endif
414         } else if (!strcmp(r, "0")) {
415             argv0 = argv[optind++];
416         } else {
417             usage();
418         }
419     }
420 
421     qemu_host_page_mask = -qemu_host_page_size;
422 
423     /* init debug */
424     {
425         int mask = 0;
426         if (log_mask) {
427             mask = qemu_str_to_log_mask(log_mask);
428             if (!mask) {
429                 qemu_print_log_usage(stdout);
430                 exit(1);
431             }
432         }
433         qemu_set_log_filename_flags(log_file, mask, &error_fatal);
434     }
435 
436     if (optind >= argc) {
437         usage();
438     }
439     filename = argv[optind];
440     if (argv0) {
441         argv[optind] = argv0;
442     }
443 
444     if (!trace_init_backends()) {
445         exit(1);
446     }
447     trace_init_file();
448     qemu_plugin_load_list(&plugins, &error_fatal);
449 
450     /* Zero out regs */
451     memset(regs, 0, sizeof(struct target_pt_regs));
452 
453     /* Zero bsd params */
454     memset(&bprm, 0, sizeof(bprm));
455 
456     /* Zero out image_info */
457     memset(info, 0, sizeof(struct image_info));
458 
459     /* Scan interp_prefix dir for replacement files. */
460     init_paths(interp_prefix);
461 
462     if (cpu_model == NULL) {
463         cpu_model = TARGET_DEFAULT_CPU_MODEL;
464     }
465 
466     cpu_type = parse_cpu_option(cpu_model);
467 
468     /* init tcg before creating CPUs and to get qemu_host_page_size */
469     {
470         AccelState *accel = current_accel();
471         AccelClass *ac = ACCEL_GET_CLASS(accel);
472 
473         accel_init_interfaces(ac);
474         object_property_set_bool(OBJECT(accel), "one-insn-per-tb",
475                                  opt_one_insn_per_tb, &error_abort);
476         object_property_set_int(OBJECT(accel), "tb-size",
477                                 opt_tb_size, &error_abort);
478         ac->init_machine(accel, NULL);
479     }
480 
481     /*
482      * Finalize page size before creating CPUs.
483      * This will do nothing if !TARGET_PAGE_BITS_VARY.
484      * The most efficient setting is to match the host.
485      */
486     host_page_size = qemu_real_host_page_size();
487     set_preferred_target_page_bits(ctz32(host_page_size));
488     finalize_target_page_bits();
489 
490     cpu = cpu_create(cpu_type);
491     env = cpu_env(cpu);
492     cpu_reset(cpu);
493     thread_cpu = cpu;
494 
495     /*
496      * Reserving too much vm space via mmap can run into problems with rlimits,
497      * oom due to page table creation, etc.  We will still try it, if directed
498      * by the command-line option, but not by default. Unless we're running a
499      * target address space of 32 or fewer bits on a host with 64 bits.
500      */
501     max_reserved_va = MAX_RESERVED_VA(cpu);
502     if (reserved_va != 0) {
503         if ((reserved_va + 1) % host_page_size) {
504             char *s = size_to_str(host_page_size);
505             fprintf(stderr, "Reserved virtual address not aligned mod %s\n", s);
506             g_free(s);
507             exit(EXIT_FAILURE);
508         }
509         if (max_reserved_va && reserved_va > max_reserved_va) {
510             fprintf(stderr, "Reserved virtual address too big\n");
511             exit(EXIT_FAILURE);
512         }
513     } else if (HOST_LONG_BITS == 64 && TARGET_VIRT_ADDR_SPACE_BITS <= 32) {
514         /* MAX_RESERVED_VA + 1 is a large power of 2, so is aligned. */
515         reserved_va = max_reserved_va;
516     }
517     if (reserved_va != 0) {
518         guest_addr_max = reserved_va;
519     } else if (MIN(TARGET_VIRT_ADDR_SPACE_BITS, TARGET_ABI_BITS) <= 32) {
520         guest_addr_max = UINT32_MAX;
521     } else {
522         guest_addr_max = ~0ul;
523     }
524 
525     if (getenv("QEMU_STRACE")) {
526         do_strace = 1;
527     }
528 
529     target_environ = envlist_to_environ(envlist, NULL);
530     envlist_free(envlist);
531 
532     {
533         Error *err = NULL;
534         if (seed_optarg != NULL) {
535             qemu_guest_random_seed_main(seed_optarg, &err);
536         } else {
537             qcrypto_init(&err);
538         }
539         if (err) {
540             error_reportf_err(err, "cannot initialize crypto: ");
541             exit(1);
542         }
543     }
544 
545     /*
546      * Now that page sizes are configured we can do
547      * proper page alignment for guest_base.
548      */
549     if (have_guest_base) {
550         if (guest_base & ~qemu_host_page_mask) {
551             error_report("Selected guest base not host page aligned");
552             exit(1);
553         }
554     }
555 
556     /*
557      * If reserving host virtual address space, do so now.
558      * Combined with '-B', ensure that the chosen range is free.
559      */
560     if (reserved_va) {
561         void *p;
562 
563         if (have_guest_base) {
564             p = mmap((void *)guest_base, reserved_va + 1, PROT_NONE,
565                      MAP_ANON | MAP_PRIVATE | MAP_FIXED | MAP_EXCL, -1, 0);
566         } else {
567             p = mmap(NULL, reserved_va + 1, PROT_NONE,
568                      MAP_ANON | MAP_PRIVATE, -1, 0);
569         }
570         if (p == MAP_FAILED) {
571             const char *err = strerror(errno);
572             char *sz = size_to_str(reserved_va + 1);
573 
574             if (have_guest_base) {
575                 error_report("Cannot allocate %s bytes at -B %p for guest "
576                              "address space: %s", sz, (void *)guest_base, err);
577             } else {
578                 error_report("Cannot allocate %s bytes for guest "
579                              "address space: %s", sz, err);
580             }
581             exit(1);
582         }
583         guest_base = (uintptr_t)p;
584         have_guest_base = true;
585 
586         /* Ensure that mmap_next_start is within range. */
587         if (reserved_va <= mmap_next_start) {
588             mmap_next_start = (reserved_va / 4 * 3)
589                               & TARGET_PAGE_MASK & qemu_host_page_mask;
590         }
591     }
592 
593     if (loader_exec(filename, argv + optind, target_environ, regs, info,
594                     &bprm) != 0) {
595         printf("Error loading %s\n", filename);
596         _exit(1);
597     }
598 
599     for (wrk = target_environ; *wrk; wrk++) {
600         g_free(*wrk);
601     }
602 
603     g_free(target_environ);
604 
605     if (qemu_loglevel_mask(CPU_LOG_PAGE)) {
606         FILE *f = qemu_log_trylock();
607         if (f) {
608             fprintf(f, "guest_base  %p\n", (void *)guest_base);
609             fprintf(f, "page layout changed following binary load\n");
610             page_dump(f);
611 
612             fprintf(f, "end_code    0x" TARGET_ABI_FMT_lx "\n",
613                     info->end_code);
614             fprintf(f, "start_code  0x" TARGET_ABI_FMT_lx "\n",
615                     info->start_code);
616             fprintf(f, "start_data  0x" TARGET_ABI_FMT_lx "\n",
617                     info->start_data);
618             fprintf(f, "end_data    0x" TARGET_ABI_FMT_lx "\n",
619                     info->end_data);
620             fprintf(f, "start_stack 0x" TARGET_ABI_FMT_lx "\n",
621                     info->start_stack);
622             fprintf(f, "brk         0x" TARGET_ABI_FMT_lx "\n", info->brk);
623             fprintf(f, "entry       0x" TARGET_ABI_FMT_lx "\n", info->entry);
624 
625             qemu_log_unlock(f);
626         }
627     }
628 
629     /* build Task State */
630     ts = g_new0(TaskState, 1);
631     init_task_state(ts);
632     ts->info = info;
633     ts->bprm = &bprm;
634     ts->ts_tid = qemu_get_thread_id();
635     cpu->opaque = ts;
636 
637     target_set_brk(info->brk);
638     syscall_init();
639     signal_init();
640 
641     /*
642      * Now that we've loaded the binary, GUEST_BASE is fixed.  Delay
643      * generating the prologue until now so that the prologue can take
644      * the real value of GUEST_BASE into account.
645      */
646     tcg_prologue_init();
647 
648     target_cpu_init(env, regs);
649 
650     if (gdbstub) {
651         gdbserver_start(gdbstub, &error_fatal);
652     }
653     cpu_loop(env);
654     /* never exits */
655     return 0;
656 }
657