xref: /openbmc/qemu/bsd-user/main.c (revision 8eb806a7)
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 <sys/types.h>
22 #include <sys/time.h>
23 #include <sys/resource.h>
24 #include <sys/sysctl.h>
25 
26 #include "qemu/osdep.h"
27 #include "qemu-common.h"
28 #include "qemu/units.h"
29 #include "qemu/accel.h"
30 #include "sysemu/tcg.h"
31 #include "qemu-version.h"
32 #include <machine/trap.h>
33 
34 #include "qapi/error.h"
35 #include "qemu.h"
36 #include "qemu/config-file.h"
37 #include "qemu/error-report.h"
38 #include "qemu/path.h"
39 #include "qemu/help_option.h"
40 #include "qemu/module.h"
41 #include "exec/exec-all.h"
42 #include "tcg/tcg.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 
51 #include "host-os.h"
52 #include "target_arch_cpu.h"
53 
54 int singlestep;
55 uintptr_t guest_base;
56 bool have_guest_base;
57 /*
58  * When running 32-on-64 we should make sure we can fit all of the possible
59  * guest address space into a contiguous chunk of virtual host memory.
60  *
61  * This way we will never overlap with our own libraries or binaries or stack
62  * or anything else that QEMU maps.
63  *
64  * Many cpus reserve the high bit (or more than one for some 64-bit cpus)
65  * of the address for the kernel.  Some cpus rely on this and user space
66  * uses the high bit(s) for pointer tagging and the like.  For them, we
67  * must preserve the expected address space.
68  */
69 #ifndef MAX_RESERVED_VA
70 # if HOST_LONG_BITS > TARGET_VIRT_ADDR_SPACE_BITS
71 #  if TARGET_VIRT_ADDR_SPACE_BITS == 32 && \
72       (TARGET_LONG_BITS == 32 || defined(TARGET_ABI32))
73 /*
74  * There are a number of places where we assign reserved_va to a variable
75  * of type abi_ulong and expect it to fit.  Avoid the last page.
76  */
77 #   define MAX_RESERVED_VA  (0xfffffffful & TARGET_PAGE_MASK)
78 #  else
79 #   define MAX_RESERVED_VA  (1ul << TARGET_VIRT_ADDR_SPACE_BITS)
80 #  endif
81 # else
82 #  define MAX_RESERVED_VA  0
83 # endif
84 #endif
85 
86 /*
87  * That said, reserving *too* much vm space via mmap can run into problems
88  * with rlimits, oom due to page table creation, etc.  We will still try it,
89  * if directed by the command-line option, but not by default.
90  */
91 #if HOST_LONG_BITS == 64 && TARGET_VIRT_ADDR_SPACE_BITS <= 32
92 unsigned long reserved_va = MAX_RESERVED_VA;
93 #else
94 unsigned long reserved_va;
95 #endif
96 
97 static const char *interp_prefix = CONFIG_QEMU_INTERP_PREFIX;
98 const char *qemu_uname_release;
99 char qemu_proc_pathname[PATH_MAX];  /* full path to exeutable */
100 
101 unsigned long target_maxtsiz = TARGET_MAXTSIZ;   /* max text size */
102 unsigned long target_dfldsiz = TARGET_DFLDSIZ;   /* initial data size limit */
103 unsigned long target_maxdsiz = TARGET_MAXDSIZ;   /* max data size */
104 unsigned long target_dflssiz = TARGET_DFLSSIZ;   /* initial data size limit */
105 unsigned long target_maxssiz = TARGET_MAXSSIZ;   /* max stack size */
106 unsigned long target_sgrowsiz = TARGET_SGROWSIZ; /* amount to grow stack */
107 
108 /* Helper routines for implementing atomic operations. */
109 
110 void fork_start(void)
111 {
112     start_exclusive();
113     cpu_list_lock();
114     mmap_fork_start();
115 }
116 
117 void fork_end(int child)
118 {
119     if (child) {
120         CPUState *cpu, *next_cpu;
121         /*
122          * Child processes created by fork() only have a single thread.  Discard
123          * information about the parent threads.
124          */
125         CPU_FOREACH_SAFE(cpu, next_cpu) {
126             if (cpu != thread_cpu) {
127                 QTAILQ_REMOVE_RCU(&cpus, cpu, node);
128             }
129         }
130         mmap_fork_end(child);
131         /*
132          * qemu_init_cpu_list() takes care of reinitializing the exclusive
133          * state, so we don't need to end_exclusive() here.
134          */
135         qemu_init_cpu_list();
136         gdbserver_fork(thread_cpu);
137     } else {
138         mmap_fork_end(child);
139         cpu_list_unlock();
140         end_exclusive();
141     }
142 }
143 
144 void cpu_loop(CPUArchState *env)
145 {
146     target_cpu_loop(env);
147 }
148 
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            "-singlestep       always run in singlestep mode\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 
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 
205 bool qemu_cpu_is_self(CPUState *cpu)
206 {
207     return thread_cpu == cpu;
208 }
209 
210 void qemu_cpu_kick(CPUState *cpu)
211 {
212     cpu_exit(cpu);
213 }
214 
215 /* Assumes contents are already zeroed.  */
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 
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
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 
251 static void save_proc_pathname(char *argv0)
252 {
253     int mib[4];
254     size_t len;
255 
256     mib[0] = CTL_KERN;
257     mib[1] = KERN_PROC;
258     mib[2] = KERN_PROC_PATHNAME;
259     mib[3] = -1;
260 
261     len = sizeof(qemu_proc_pathname);
262     if (sysctl(mib, 4, qemu_proc_pathname, &len, NULL, 0)) {
263         perror("sysctl");
264     }
265 }
266 
267 int main(int argc, char **argv)
268 {
269     const char *filename;
270     const char *cpu_model;
271     const char *cpu_type;
272     const char *log_file = NULL;
273     const char *log_mask = NULL;
274     const char *seed_optarg = NULL;
275     struct target_pt_regs regs1, *regs = &regs1;
276     struct image_info info1, *info = &info1;
277     struct bsd_binprm bprm;
278     TaskState *ts;
279     CPUArchState *env;
280     CPUState *cpu;
281     int optind, rv;
282     const char *r;
283     const char *gdbstub = NULL;
284     char **target_environ, **wrk;
285     envlist_t *envlist = NULL;
286     char *argv0 = NULL;
287 
288     adjust_ssize();
289 
290     if (argc <= 1) {
291         usage();
292     }
293 
294     save_proc_pathname(argv[0]);
295 
296     error_init(argv[0]);
297     module_call_init(MODULE_INIT_TRACE);
298     qemu_init_cpu_list();
299     module_call_init(MODULE_INIT_QOM);
300 
301     envlist = envlist_create();
302 
303     /* add current environment into the list */
304     for (wrk = environ; *wrk != NULL; wrk++) {
305         (void) envlist_setenv(envlist, *wrk);
306     }
307 
308     cpu_model = NULL;
309 
310     qemu_add_opts(&qemu_trace_opts);
311 
312     optind = 1;
313     for (;;) {
314         if (optind >= argc) {
315             break;
316         }
317         r = argv[optind];
318         if (r[0] != '-') {
319             break;
320         }
321         optind++;
322         r++;
323         if (!strcmp(r, "-")) {
324             break;
325         } else if (!strcmp(r, "d")) {
326             if (optind >= argc) {
327                 break;
328             }
329             log_mask = argv[optind++];
330         } else if (!strcmp(r, "D")) {
331             if (optind >= argc) {
332                 break;
333             }
334             log_file = argv[optind++];
335         } else if (!strcmp(r, "E")) {
336             r = argv[optind++];
337             if (envlist_setenv(envlist, r) != 0) {
338                 usage();
339             }
340         } else if (!strcmp(r, "ignore-environment")) {
341             envlist_free(envlist);
342             envlist = envlist_create();
343         } else if (!strcmp(r, "U")) {
344             r = argv[optind++];
345             if (envlist_unsetenv(envlist, r) != 0) {
346                 usage();
347             }
348         } else if (!strcmp(r, "s")) {
349             r = argv[optind++];
350             rv = qemu_strtoul(r, &r, 0, &target_dflssiz);
351             if (rv < 0 || target_dflssiz <= 0) {
352                 usage();
353             }
354             if (*r == 'M') {
355                 target_dflssiz *= 1024 * 1024;
356             } else if (*r == 'k' || *r == 'K') {
357                 target_dflssiz *= 1024;
358             }
359             if (target_dflssiz > target_maxssiz) {
360                 usage();
361             }
362         } else if (!strcmp(r, "L")) {
363             interp_prefix = argv[optind++];
364         } else if (!strcmp(r, "p")) {
365             qemu_host_page_size = atoi(argv[optind++]);
366             if (qemu_host_page_size == 0 ||
367                 (qemu_host_page_size & (qemu_host_page_size - 1)) != 0) {
368                 fprintf(stderr, "page size must be a power of two\n");
369                 exit(1);
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                 /* XXX: implement xxx_cpu_list for targets that still miss it */
379 #if defined(cpu_list)
380                 cpu_list();
381 #endif
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")) {
395             singlestep = 1;
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     qemu_set_log_filename(log_file, &error_fatal);
409     if (log_mask) {
410         int mask;
411 
412         mask = qemu_str_to_log_mask(log_mask);
413         if (!mask) {
414             qemu_print_log_usage(stdout);
415             exit(1);
416         }
417         qemu_set_log(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         AccelClass *ac = ACCEL_GET_CLASS(current_accel());
454 
455         accel_init_interfaces(ac);
456         ac->init_machine(NULL);
457     }
458     cpu = cpu_create(cpu_type);
459     env = cpu->env_ptr;
460     cpu_reset(cpu);
461     thread_cpu = cpu;
462 
463     if (getenv("QEMU_STRACE")) {
464         do_strace = 1;
465     }
466 
467     target_environ = envlist_to_environ(envlist, NULL);
468     envlist_free(envlist);
469 
470     if (reserved_va) {
471             mmap_next_start = reserved_va;
472     }
473 
474     {
475         Error *err = NULL;
476         if (seed_optarg != NULL) {
477             qemu_guest_random_seed_main(seed_optarg, &err);
478         } else {
479             qcrypto_init(&err);
480         }
481         if (err) {
482             error_reportf_err(err, "cannot initialize crypto: ");
483             exit(1);
484         }
485     }
486 
487     /*
488      * Now that page sizes are configured we can do
489      * proper page alignment for guest_base.
490      */
491     guest_base = HOST_PAGE_ALIGN(guest_base);
492 
493     if (loader_exec(filename, argv + optind, target_environ, regs, info,
494                     &bprm) != 0) {
495         printf("Error loading %s\n", filename);
496         _exit(1);
497     }
498 
499     for (wrk = target_environ; *wrk; wrk++) {
500         g_free(*wrk);
501     }
502 
503     g_free(target_environ);
504 
505     if (qemu_loglevel_mask(CPU_LOG_PAGE)) {
506         qemu_log("guest_base  %p\n", (void *)guest_base);
507         log_page_dump("binary load");
508 
509         qemu_log("start_brk   0x" TARGET_ABI_FMT_lx "\n", info->start_brk);
510         qemu_log("end_code    0x" TARGET_ABI_FMT_lx "\n", info->end_code);
511         qemu_log("start_code  0x" TARGET_ABI_FMT_lx "\n",
512                  info->start_code);
513         qemu_log("start_data  0x" TARGET_ABI_FMT_lx "\n",
514                  info->start_data);
515         qemu_log("end_data    0x" TARGET_ABI_FMT_lx "\n", info->end_data);
516         qemu_log("start_stack 0x" TARGET_ABI_FMT_lx "\n",
517                  info->start_stack);
518         qemu_log("brk         0x" TARGET_ABI_FMT_lx "\n", info->brk);
519         qemu_log("entry       0x" TARGET_ABI_FMT_lx "\n", info->entry);
520     }
521 
522     /* build Task State */
523     ts = g_new0(TaskState, 1);
524     init_task_state(ts);
525     ts->info = info;
526     ts->bprm = &bprm;
527     cpu->opaque = ts;
528 
529     target_set_brk(info->brk);
530     syscall_init();
531     signal_init();
532 
533     /*
534      * Now that we've loaded the binary, GUEST_BASE is fixed.  Delay
535      * generating the prologue until now so that the prologue can take
536      * the real value of GUEST_BASE into account.
537      */
538     tcg_prologue_init(tcg_ctx);
539 
540     target_cpu_init(env, regs);
541 
542     if (gdbstub) {
543         gdbserver_start(gdbstub);
544         gdb_handlesig(cpu, 0);
545     }
546     cpu_loop(env);
547     /* never exits */
548     return 0;
549 }
550