]>
Commit | Line | Data |
---|---|---|
1 | // SPDX-License-Identifier: GPL-2.0-only | |
2 | /* | |
3 | * linux/init/main.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | * | |
7 | * GK 2/5/95 - Changed to support mounting root fs via NFS | |
8 | * Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96 | |
9 | * Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96 | |
10 | * Simplified starting of init: Michael A. Griffith <grif@acm.org> | |
11 | */ | |
12 | ||
13 | #define DEBUG /* Enable initcall_debug */ | |
14 | ||
15 | #include <linux/types.h> | |
16 | #include <linux/extable.h> | |
17 | #include <linux/module.h> | |
18 | #include <linux/proc_fs.h> | |
19 | #include <linux/binfmts.h> | |
20 | #include <linux/kernel.h> | |
21 | #include <linux/syscalls.h> | |
22 | #include <linux/stackprotector.h> | |
23 | #include <linux/string.h> | |
24 | #include <linux/ctype.h> | |
25 | #include <linux/delay.h> | |
26 | #include <linux/ioport.h> | |
27 | #include <linux/init.h> | |
28 | #include <linux/initrd.h> | |
29 | #include <linux/memblock.h> | |
30 | #include <linux/acpi.h> | |
31 | #include <linux/bootconfig.h> | |
32 | #include <linux/console.h> | |
33 | #include <linux/nmi.h> | |
34 | #include <linux/percpu.h> | |
35 | #include <linux/kmod.h> | |
36 | #include <linux/vmalloc.h> | |
37 | #include <linux/kernel_stat.h> | |
38 | #include <linux/start_kernel.h> | |
39 | #include <linux/security.h> | |
40 | #include <linux/smp.h> | |
41 | #include <linux/profile.h> | |
42 | #include <linux/rcupdate.h> | |
43 | #include <linux/moduleparam.h> | |
44 | #include <linux/kallsyms.h> | |
45 | #include <linux/writeback.h> | |
46 | #include <linux/cpu.h> | |
47 | #include <linux/cpuset.h> | |
48 | #include <linux/cgroup.h> | |
49 | #include <linux/efi.h> | |
50 | #include <linux/tick.h> | |
51 | #include <linux/sched/isolation.h> | |
52 | #include <linux/interrupt.h> | |
53 | #include <linux/taskstats_kern.h> | |
54 | #include <linux/delayacct.h> | |
55 | #include <linux/unistd.h> | |
56 | #include <linux/utsname.h> | |
57 | #include <linux/rmap.h> | |
58 | #include <linux/mempolicy.h> | |
59 | #include <linux/key.h> | |
60 | #include <linux/buffer_head.h> | |
61 | #include <linux/page_ext.h> | |
62 | #include <linux/debug_locks.h> | |
63 | #include <linux/debugobjects.h> | |
64 | #include <linux/lockdep.h> | |
65 | #include <linux/kmemleak.h> | |
66 | #include <linux/padata.h> | |
67 | #include <linux/pid_namespace.h> | |
68 | #include <linux/device/driver.h> | |
69 | #include <linux/kthread.h> | |
70 | #include <linux/sched.h> | |
71 | #include <linux/sched/init.h> | |
72 | #include <linux/signal.h> | |
73 | #include <linux/idr.h> | |
74 | #include <linux/kgdb.h> | |
75 | #include <linux/ftrace.h> | |
76 | #include <linux/async.h> | |
77 | #include <linux/sfi.h> | |
78 | #include <linux/shmem_fs.h> | |
79 | #include <linux/slab.h> | |
80 | #include <linux/perf_event.h> | |
81 | #include <linux/ptrace.h> | |
82 | #include <linux/pti.h> | |
83 | #include <linux/blkdev.h> | |
84 | #include <linux/elevator.h> | |
85 | #include <linux/sched/clock.h> | |
86 | #include <linux/sched/task.h> | |
87 | #include <linux/sched/task_stack.h> | |
88 | #include <linux/context_tracking.h> | |
89 | #include <linux/random.h> | |
90 | #include <linux/list.h> | |
91 | #include <linux/integrity.h> | |
92 | #include <linux/proc_ns.h> | |
93 | #include <linux/io.h> | |
94 | #include <linux/cache.h> | |
95 | #include <linux/rodata_test.h> | |
96 | #include <linux/jump_label.h> | |
97 | #include <linux/mem_encrypt.h> | |
98 | #include <linux/kcsan.h> | |
99 | ||
100 | #include <asm/io.h> | |
101 | #include <asm/bugs.h> | |
102 | #include <asm/setup.h> | |
103 | #include <asm/sections.h> | |
104 | #include <asm/cacheflush.h> | |
105 | ||
106 | #define CREATE_TRACE_POINTS | |
107 | #include <trace/events/initcall.h> | |
108 | ||
109 | static int kernel_init(void *); | |
110 | ||
111 | extern void init_IRQ(void); | |
112 | extern void radix_tree_init(void); | |
113 | ||
114 | /* | |
115 | * Debug helper: via this flag we know that we are in 'early bootup code' | |
116 | * where only the boot processor is running with IRQ disabled. This means | |
117 | * two things - IRQ must not be enabled before the flag is cleared and some | |
118 | * operations which are not allowed with IRQ disabled are allowed while the | |
119 | * flag is set. | |
120 | */ | |
121 | bool early_boot_irqs_disabled __read_mostly; | |
122 | ||
123 | enum system_states system_state __read_mostly; | |
124 | EXPORT_SYMBOL(system_state); | |
125 | ||
126 | /* | |
127 | * Boot command-line arguments | |
128 | */ | |
129 | #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT | |
130 | #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT | |
131 | ||
132 | extern void time_init(void); | |
133 | /* Default late time init is NULL. archs can override this later. */ | |
134 | void (*__initdata late_time_init)(void); | |
135 | ||
136 | /* Untouched command line saved by arch-specific code. */ | |
137 | char __initdata boot_command_line[COMMAND_LINE_SIZE]; | |
138 | /* Untouched saved command line (eg. for /proc) */ | |
139 | char *saved_command_line; | |
140 | /* Command line for parameter parsing */ | |
141 | static char *static_command_line; | |
142 | /* Untouched extra command line */ | |
143 | static char *extra_command_line; | |
144 | /* Extra init arguments */ | |
145 | static char *extra_init_args; | |
146 | ||
147 | #ifdef CONFIG_BOOT_CONFIG | |
148 | /* Is bootconfig on command line? */ | |
149 | static bool bootconfig_found; | |
150 | static bool initargs_found; | |
151 | #else | |
152 | # define bootconfig_found false | |
153 | # define initargs_found false | |
154 | #endif | |
155 | ||
156 | static char *execute_command; | |
157 | static char *ramdisk_execute_command; | |
158 | ||
159 | /* | |
160 | * Used to generate warnings if static_key manipulation functions are used | |
161 | * before jump_label_init is called. | |
162 | */ | |
163 | bool static_key_initialized __read_mostly; | |
164 | EXPORT_SYMBOL_GPL(static_key_initialized); | |
165 | ||
166 | /* | |
167 | * If set, this is an indication to the drivers that reset the underlying | |
168 | * device before going ahead with the initialization otherwise driver might | |
169 | * rely on the BIOS and skip the reset operation. | |
170 | * | |
171 | * This is useful if kernel is booting in an unreliable environment. | |
172 | * For ex. kdump situation where previous kernel has crashed, BIOS has been | |
173 | * skipped and devices will be in unknown state. | |
174 | */ | |
175 | unsigned int reset_devices; | |
176 | EXPORT_SYMBOL(reset_devices); | |
177 | ||
178 | static int __init set_reset_devices(char *str) | |
179 | { | |
180 | reset_devices = 1; | |
181 | return 1; | |
182 | } | |
183 | ||
184 | __setup("reset_devices", set_reset_devices); | |
185 | ||
186 | static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; | |
187 | const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; | |
188 | static const char *panic_later, *panic_param; | |
189 | ||
190 | extern const struct obs_kernel_param __setup_start[], __setup_end[]; | |
191 | ||
192 | static bool __init obsolete_checksetup(char *line) | |
193 | { | |
194 | const struct obs_kernel_param *p; | |
195 | bool had_early_param = false; | |
196 | ||
197 | p = __setup_start; | |
198 | do { | |
199 | int n = strlen(p->str); | |
200 | if (parameqn(line, p->str, n)) { | |
201 | if (p->early) { | |
202 | /* Already done in parse_early_param? | |
203 | * (Needs exact match on param part). | |
204 | * Keep iterating, as we can have early | |
205 | * params and __setups of same names 8( */ | |
206 | if (line[n] == '\0' || line[n] == '=') | |
207 | had_early_param = true; | |
208 | } else if (!p->setup_func) { | |
209 | pr_warn("Parameter %s is obsolete, ignored\n", | |
210 | p->str); | |
211 | return true; | |
212 | } else if (p->setup_func(line + n)) | |
213 | return true; | |
214 | } | |
215 | p++; | |
216 | } while (p < __setup_end); | |
217 | ||
218 | return had_early_param; | |
219 | } | |
220 | ||
221 | /* | |
222 | * This should be approx 2 Bo*oMips to start (note initial shift), and will | |
223 | * still work even if initially too large, it will just take slightly longer | |
224 | */ | |
225 | unsigned long loops_per_jiffy = (1<<12); | |
226 | EXPORT_SYMBOL(loops_per_jiffy); | |
227 | ||
228 | static int __init debug_kernel(char *str) | |
229 | { | |
230 | console_loglevel = CONSOLE_LOGLEVEL_DEBUG; | |
231 | return 0; | |
232 | } | |
233 | ||
234 | static int __init quiet_kernel(char *str) | |
235 | { | |
236 | console_loglevel = CONSOLE_LOGLEVEL_QUIET; | |
237 | return 0; | |
238 | } | |
239 | ||
240 | early_param("debug", debug_kernel); | |
241 | early_param("quiet", quiet_kernel); | |
242 | ||
243 | static int __init loglevel(char *str) | |
244 | { | |
245 | int newlevel; | |
246 | ||
247 | /* | |
248 | * Only update loglevel value when a correct setting was passed, | |
249 | * to prevent blind crashes (when loglevel being set to 0) that | |
250 | * are quite hard to debug | |
251 | */ | |
252 | if (get_option(&str, &newlevel)) { | |
253 | console_loglevel = newlevel; | |
254 | return 0; | |
255 | } | |
256 | ||
257 | return -EINVAL; | |
258 | } | |
259 | ||
260 | early_param("loglevel", loglevel); | |
261 | ||
262 | #ifdef CONFIG_BLK_DEV_INITRD | |
263 | static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum) | |
264 | { | |
265 | u32 size, csum; | |
266 | char *data; | |
267 | u32 *hdr; | |
268 | ||
269 | if (!initrd_end) | |
270 | return NULL; | |
271 | ||
272 | data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN; | |
273 | if (memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN)) | |
274 | return NULL; | |
275 | ||
276 | hdr = (u32 *)(data - 8); | |
277 | size = hdr[0]; | |
278 | csum = hdr[1]; | |
279 | ||
280 | data = ((void *)hdr) - size; | |
281 | if ((unsigned long)data < initrd_start) { | |
282 | pr_err("bootconfig size %d is greater than initrd size %ld\n", | |
283 | size, initrd_end - initrd_start); | |
284 | return NULL; | |
285 | } | |
286 | ||
287 | /* Remove bootconfig from initramfs/initrd */ | |
288 | initrd_end = (unsigned long)data; | |
289 | if (_size) | |
290 | *_size = size; | |
291 | if (_csum) | |
292 | *_csum = csum; | |
293 | ||
294 | return data; | |
295 | } | |
296 | #else | |
297 | static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum) | |
298 | { | |
299 | return NULL; | |
300 | } | |
301 | #endif | |
302 | ||
303 | #ifdef CONFIG_BOOT_CONFIG | |
304 | ||
305 | char xbc_namebuf[XBC_KEYLEN_MAX] __initdata; | |
306 | ||
307 | #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0) | |
308 | ||
309 | static int __init xbc_snprint_cmdline(char *buf, size_t size, | |
310 | struct xbc_node *root) | |
311 | { | |
312 | struct xbc_node *knode, *vnode; | |
313 | char *end = buf + size; | |
314 | const char *val; | |
315 | int ret; | |
316 | ||
317 | xbc_node_for_each_key_value(root, knode, val) { | |
318 | ret = xbc_node_compose_key_after(root, knode, | |
319 | xbc_namebuf, XBC_KEYLEN_MAX); | |
320 | if (ret < 0) | |
321 | return ret; | |
322 | ||
323 | vnode = xbc_node_get_child(knode); | |
324 | if (!vnode) { | |
325 | ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf); | |
326 | if (ret < 0) | |
327 | return ret; | |
328 | buf += ret; | |
329 | continue; | |
330 | } | |
331 | xbc_array_for_each_value(vnode, val) { | |
332 | ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ", | |
333 | xbc_namebuf, val); | |
334 | if (ret < 0) | |
335 | return ret; | |
336 | buf += ret; | |
337 | } | |
338 | } | |
339 | ||
340 | return buf - (end - size); | |
341 | } | |
342 | #undef rest | |
343 | ||
344 | /* Make an extra command line under given key word */ | |
345 | static char * __init xbc_make_cmdline(const char *key) | |
346 | { | |
347 | struct xbc_node *root; | |
348 | char *new_cmdline; | |
349 | int ret, len = 0; | |
350 | ||
351 | root = xbc_find_node(key); | |
352 | if (!root) | |
353 | return NULL; | |
354 | ||
355 | /* Count required buffer size */ | |
356 | len = xbc_snprint_cmdline(NULL, 0, root); | |
357 | if (len <= 0) | |
358 | return NULL; | |
359 | ||
360 | new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES); | |
361 | if (!new_cmdline) { | |
362 | pr_err("Failed to allocate memory for extra kernel cmdline.\n"); | |
363 | return NULL; | |
364 | } | |
365 | ||
366 | ret = xbc_snprint_cmdline(new_cmdline, len + 1, root); | |
367 | if (ret < 0 || ret > len) { | |
368 | pr_err("Failed to print extra kernel cmdline.\n"); | |
369 | return NULL; | |
370 | } | |
371 | ||
372 | return new_cmdline; | |
373 | } | |
374 | ||
375 | static u32 boot_config_checksum(unsigned char *p, u32 size) | |
376 | { | |
377 | u32 ret = 0; | |
378 | ||
379 | while (size--) | |
380 | ret += *p++; | |
381 | ||
382 | return ret; | |
383 | } | |
384 | ||
385 | static int __init bootconfig_params(char *param, char *val, | |
386 | const char *unused, void *arg) | |
387 | { | |
388 | if (strcmp(param, "bootconfig") == 0) { | |
389 | bootconfig_found = true; | |
390 | } else if (strcmp(param, "--") == 0) { | |
391 | initargs_found = true; | |
392 | } | |
393 | return 0; | |
394 | } | |
395 | ||
396 | static void __init setup_boot_config(const char *cmdline) | |
397 | { | |
398 | static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; | |
399 | const char *msg; | |
400 | int pos; | |
401 | u32 size, csum; | |
402 | char *data, *copy; | |
403 | int ret; | |
404 | ||
405 | /* Cut out the bootconfig data even if we have no bootconfig option */ | |
406 | data = get_boot_config_from_initrd(&size, &csum); | |
407 | ||
408 | strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); | |
409 | parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL, | |
410 | bootconfig_params); | |
411 | ||
412 | if (!bootconfig_found) | |
413 | return; | |
414 | ||
415 | if (!data) { | |
416 | pr_err("'bootconfig' found on command line, but no bootconfig found\n"); | |
417 | return; | |
418 | } | |
419 | ||
420 | if (size >= XBC_DATA_MAX) { | |
421 | pr_err("bootconfig size %d greater than max size %d\n", | |
422 | size, XBC_DATA_MAX); | |
423 | return; | |
424 | } | |
425 | ||
426 | if (boot_config_checksum((unsigned char *)data, size) != csum) { | |
427 | pr_err("bootconfig checksum failed\n"); | |
428 | return; | |
429 | } | |
430 | ||
431 | copy = memblock_alloc(size + 1, SMP_CACHE_BYTES); | |
432 | if (!copy) { | |
433 | pr_err("Failed to allocate memory for bootconfig\n"); | |
434 | return; | |
435 | } | |
436 | ||
437 | memcpy(copy, data, size); | |
438 | copy[size] = '\0'; | |
439 | ||
440 | ret = xbc_init(copy, &msg, &pos); | |
441 | if (ret < 0) { | |
442 | if (pos < 0) | |
443 | pr_err("Failed to init bootconfig: %s.\n", msg); | |
444 | else | |
445 | pr_err("Failed to parse bootconfig: %s at %d.\n", | |
446 | msg, pos); | |
447 | } else { | |
448 | pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret); | |
449 | /* keys starting with "kernel." are passed via cmdline */ | |
450 | extra_command_line = xbc_make_cmdline("kernel"); | |
451 | /* Also, "init." keys are init arguments */ | |
452 | extra_init_args = xbc_make_cmdline("init"); | |
453 | } | |
454 | return; | |
455 | } | |
456 | ||
457 | #else | |
458 | ||
459 | static void __init setup_boot_config(const char *cmdline) | |
460 | { | |
461 | /* Remove bootconfig data from initrd */ | |
462 | get_boot_config_from_initrd(NULL, NULL); | |
463 | } | |
464 | ||
465 | static int __init warn_bootconfig(char *str) | |
466 | { | |
467 | pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOTCONFIG is not set.\n"); | |
468 | return 0; | |
469 | } | |
470 | early_param("bootconfig", warn_bootconfig); | |
471 | ||
472 | #endif | |
473 | ||
474 | /* Change NUL term back to "=", to make "param" the whole string. */ | |
475 | static void __init repair_env_string(char *param, char *val) | |
476 | { | |
477 | if (val) { | |
478 | /* param=val or param="val"? */ | |
479 | if (val == param+strlen(param)+1) | |
480 | val[-1] = '='; | |
481 | else if (val == param+strlen(param)+2) { | |
482 | val[-2] = '='; | |
483 | memmove(val-1, val, strlen(val)+1); | |
484 | } else | |
485 | BUG(); | |
486 | } | |
487 | } | |
488 | ||
489 | /* Anything after -- gets handed straight to init. */ | |
490 | static int __init set_init_arg(char *param, char *val, | |
491 | const char *unused, void *arg) | |
492 | { | |
493 | unsigned int i; | |
494 | ||
495 | if (panic_later) | |
496 | return 0; | |
497 | ||
498 | repair_env_string(param, val); | |
499 | ||
500 | for (i = 0; argv_init[i]; i++) { | |
501 | if (i == MAX_INIT_ARGS) { | |
502 | panic_later = "init"; | |
503 | panic_param = param; | |
504 | return 0; | |
505 | } | |
506 | } | |
507 | argv_init[i] = param; | |
508 | return 0; | |
509 | } | |
510 | ||
511 | /* | |
512 | * Unknown boot options get handed to init, unless they look like | |
513 | * unused parameters (modprobe will find them in /proc/cmdline). | |
514 | */ | |
515 | static int __init unknown_bootoption(char *param, char *val, | |
516 | const char *unused, void *arg) | |
517 | { | |
518 | size_t len = strlen(param); | |
519 | ||
520 | repair_env_string(param, val); | |
521 | ||
522 | /* Handle obsolete-style parameters */ | |
523 | if (obsolete_checksetup(param)) | |
524 | return 0; | |
525 | ||
526 | /* Unused module parameter. */ | |
527 | if (strnchr(param, len, '.')) | |
528 | return 0; | |
529 | ||
530 | if (panic_later) | |
531 | return 0; | |
532 | ||
533 | if (val) { | |
534 | /* Environment option */ | |
535 | unsigned int i; | |
536 | for (i = 0; envp_init[i]; i++) { | |
537 | if (i == MAX_INIT_ENVS) { | |
538 | panic_later = "env"; | |
539 | panic_param = param; | |
540 | } | |
541 | if (!strncmp(param, envp_init[i], len+1)) | |
542 | break; | |
543 | } | |
544 | envp_init[i] = param; | |
545 | } else { | |
546 | /* Command line option */ | |
547 | unsigned int i; | |
548 | for (i = 0; argv_init[i]; i++) { | |
549 | if (i == MAX_INIT_ARGS) { | |
550 | panic_later = "init"; | |
551 | panic_param = param; | |
552 | } | |
553 | } | |
554 | argv_init[i] = param; | |
555 | } | |
556 | return 0; | |
557 | } | |
558 | ||
559 | static int __init init_setup(char *str) | |
560 | { | |
561 | unsigned int i; | |
562 | ||
563 | execute_command = str; | |
564 | /* | |
565 | * In case LILO is going to boot us with default command line, | |
566 | * it prepends "auto" before the whole cmdline which makes | |
567 | * the shell think it should execute a script with such name. | |
568 | * So we ignore all arguments entered _before_ init=... [MJ] | |
569 | */ | |
570 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
571 | argv_init[i] = NULL; | |
572 | return 1; | |
573 | } | |
574 | __setup("init=", init_setup); | |
575 | ||
576 | static int __init rdinit_setup(char *str) | |
577 | { | |
578 | unsigned int i; | |
579 | ||
580 | ramdisk_execute_command = str; | |
581 | /* See "auto" comment in init_setup */ | |
582 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
583 | argv_init[i] = NULL; | |
584 | return 1; | |
585 | } | |
586 | __setup("rdinit=", rdinit_setup); | |
587 | ||
588 | #ifndef CONFIG_SMP | |
589 | static const unsigned int setup_max_cpus = NR_CPUS; | |
590 | static inline void setup_nr_cpu_ids(void) { } | |
591 | static inline void smp_prepare_cpus(unsigned int maxcpus) { } | |
592 | #endif | |
593 | ||
594 | /* | |
595 | * We need to store the untouched command line for future reference. | |
596 | * We also need to store the touched command line since the parameter | |
597 | * parsing is performed in place, and we should allow a component to | |
598 | * store reference of name/value for future reference. | |
599 | */ | |
600 | static void __init setup_command_line(char *command_line) | |
601 | { | |
602 | size_t len, xlen = 0, ilen = 0; | |
603 | ||
604 | if (extra_command_line) | |
605 | xlen = strlen(extra_command_line); | |
606 | if (extra_init_args) | |
607 | ilen = strlen(extra_init_args) + 4; /* for " -- " */ | |
608 | ||
609 | len = xlen + strlen(boot_command_line) + 1; | |
610 | ||
611 | saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES); | |
612 | if (!saved_command_line) | |
613 | panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen); | |
614 | ||
615 | static_command_line = memblock_alloc(len, SMP_CACHE_BYTES); | |
616 | if (!static_command_line) | |
617 | panic("%s: Failed to allocate %zu bytes\n", __func__, len); | |
618 | ||
619 | if (xlen) { | |
620 | /* | |
621 | * We have to put extra_command_line before boot command | |
622 | * lines because there could be dashes (separator of init | |
623 | * command line) in the command lines. | |
624 | */ | |
625 | strcpy(saved_command_line, extra_command_line); | |
626 | strcpy(static_command_line, extra_command_line); | |
627 | } | |
628 | strcpy(saved_command_line + xlen, boot_command_line); | |
629 | strcpy(static_command_line + xlen, command_line); | |
630 | ||
631 | if (ilen) { | |
632 | /* | |
633 | * Append supplemental init boot args to saved_command_line | |
634 | * so that user can check what command line options passed | |
635 | * to init. | |
636 | */ | |
637 | len = strlen(saved_command_line); | |
638 | if (initargs_found) { | |
639 | saved_command_line[len++] = ' '; | |
640 | } else { | |
641 | strcpy(saved_command_line + len, " -- "); | |
642 | len += 4; | |
643 | } | |
644 | ||
645 | strcpy(saved_command_line + len, extra_init_args); | |
646 | } | |
647 | } | |
648 | ||
649 | /* | |
650 | * We need to finalize in a non-__init function or else race conditions | |
651 | * between the root thread and the init thread may cause start_kernel to | |
652 | * be reaped by free_initmem before the root thread has proceeded to | |
653 | * cpu_idle. | |
654 | * | |
655 | * gcc-3.4 accidentally inlines this function, so use noinline. | |
656 | */ | |
657 | ||
658 | static __initdata DECLARE_COMPLETION(kthreadd_done); | |
659 | ||
660 | noinline void __ref rest_init(void) | |
661 | { | |
662 | struct task_struct *tsk; | |
663 | int pid; | |
664 | ||
665 | rcu_scheduler_starting(); | |
666 | /* | |
667 | * We need to spawn init first so that it obtains pid 1, however | |
668 | * the init task will end up wanting to create kthreads, which, if | |
669 | * we schedule it before we create kthreadd, will OOPS. | |
670 | */ | |
671 | pid = kernel_thread(kernel_init, NULL, CLONE_FS); | |
672 | /* | |
673 | * Pin init on the boot CPU. Task migration is not properly working | |
674 | * until sched_init_smp() has been run. It will set the allowed | |
675 | * CPUs for init to the non isolated CPUs. | |
676 | */ | |
677 | rcu_read_lock(); | |
678 | tsk = find_task_by_pid_ns(pid, &init_pid_ns); | |
679 | set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id())); | |
680 | rcu_read_unlock(); | |
681 | ||
682 | numa_default_policy(); | |
683 | pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); | |
684 | rcu_read_lock(); | |
685 | kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); | |
686 | rcu_read_unlock(); | |
687 | ||
688 | /* | |
689 | * Enable might_sleep() and smp_processor_id() checks. | |
690 | * They cannot be enabled earlier because with CONFIG_PREEMPTION=y | |
691 | * kernel_thread() would trigger might_sleep() splats. With | |
692 | * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled | |
693 | * already, but it's stuck on the kthreadd_done completion. | |
694 | */ | |
695 | system_state = SYSTEM_SCHEDULING; | |
696 | ||
697 | complete(&kthreadd_done); | |
698 | ||
699 | /* | |
700 | * The boot idle thread must execute schedule() | |
701 | * at least once to get things moving: | |
702 | */ | |
703 | schedule_preempt_disabled(); | |
704 | /* Call into cpu_idle with preempt disabled */ | |
705 | cpu_startup_entry(CPUHP_ONLINE); | |
706 | } | |
707 | ||
708 | /* Check for early params. */ | |
709 | static int __init do_early_param(char *param, char *val, | |
710 | const char *unused, void *arg) | |
711 | { | |
712 | const struct obs_kernel_param *p; | |
713 | ||
714 | for (p = __setup_start; p < __setup_end; p++) { | |
715 | if ((p->early && parameq(param, p->str)) || | |
716 | (strcmp(param, "console") == 0 && | |
717 | strcmp(p->str, "earlycon") == 0) | |
718 | ) { | |
719 | if (p->setup_func(val) != 0) | |
720 | pr_warn("Malformed early option '%s'\n", param); | |
721 | } | |
722 | } | |
723 | /* We accept everything at this stage. */ | |
724 | return 0; | |
725 | } | |
726 | ||
727 | void __init parse_early_options(char *cmdline) | |
728 | { | |
729 | parse_args("early options", cmdline, NULL, 0, 0, 0, NULL, | |
730 | do_early_param); | |
731 | } | |
732 | ||
733 | /* Arch code calls this early on, or if not, just before other parsing. */ | |
734 | void __init parse_early_param(void) | |
735 | { | |
736 | static int done __initdata; | |
737 | static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; | |
738 | ||
739 | if (done) | |
740 | return; | |
741 | ||
742 | /* All fall through to do_early_param. */ | |
743 | strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); | |
744 | parse_early_options(tmp_cmdline); | |
745 | done = 1; | |
746 | } | |
747 | ||
748 | void __init __weak arch_post_acpi_subsys_init(void) { } | |
749 | ||
750 | void __init __weak smp_setup_processor_id(void) | |
751 | { | |
752 | } | |
753 | ||
754 | # if THREAD_SIZE >= PAGE_SIZE | |
755 | void __init __weak thread_stack_cache_init(void) | |
756 | { | |
757 | } | |
758 | #endif | |
759 | ||
760 | void __init __weak mem_encrypt_init(void) { } | |
761 | ||
762 | void __init __weak poking_init(void) { } | |
763 | ||
764 | void __init __weak pgtable_cache_init(void) { } | |
765 | ||
766 | bool initcall_debug; | |
767 | core_param(initcall_debug, initcall_debug, bool, 0644); | |
768 | ||
769 | #ifdef TRACEPOINTS_ENABLED | |
770 | static void __init initcall_debug_enable(void); | |
771 | #else | |
772 | static inline void initcall_debug_enable(void) | |
773 | { | |
774 | } | |
775 | #endif | |
776 | ||
777 | /* Report memory auto-initialization states for this boot. */ | |
778 | static void __init report_meminit(void) | |
779 | { | |
780 | const char *stack; | |
781 | ||
782 | if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN)) | |
783 | stack = "all(pattern)"; | |
784 | else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO)) | |
785 | stack = "all(zero)"; | |
786 | else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL)) | |
787 | stack = "byref_all(zero)"; | |
788 | else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF)) | |
789 | stack = "byref(zero)"; | |
790 | else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER)) | |
791 | stack = "__user(zero)"; | |
792 | else | |
793 | stack = "off"; | |
794 | ||
795 | pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n", | |
796 | stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off", | |
797 | want_init_on_free() ? "on" : "off"); | |
798 | if (want_init_on_free()) | |
799 | pr_info("mem auto-init: clearing system memory may take some time...\n"); | |
800 | } | |
801 | ||
802 | /* | |
803 | * Set up kernel memory allocators | |
804 | */ | |
805 | static void __init mm_init(void) | |
806 | { | |
807 | /* | |
808 | * page_ext requires contiguous pages, | |
809 | * bigger than MAX_ORDER unless SPARSEMEM. | |
810 | */ | |
811 | page_ext_init_flatmem(); | |
812 | init_debug_pagealloc(); | |
813 | report_meminit(); | |
814 | mem_init(); | |
815 | kmem_cache_init(); | |
816 | kmemleak_init(); | |
817 | pgtable_init(); | |
818 | debug_objects_mem_init(); | |
819 | vmalloc_init(); | |
820 | ioremap_huge_init(); | |
821 | /* Should be run before the first non-init thread is created */ | |
822 | init_espfix_bsp(); | |
823 | /* Should be run after espfix64 is set up. */ | |
824 | pti_init(); | |
825 | } | |
826 | ||
827 | void __init __weak arch_call_rest_init(void) | |
828 | { | |
829 | rest_init(); | |
830 | } | |
831 | ||
832 | asmlinkage __visible void __init start_kernel(void) | |
833 | { | |
834 | char *command_line; | |
835 | char *after_dashes; | |
836 | ||
837 | set_task_stack_end_magic(&init_task); | |
838 | smp_setup_processor_id(); | |
839 | debug_objects_early_init(); | |
840 | ||
841 | cgroup_init_early(); | |
842 | ||
843 | local_irq_disable(); | |
844 | early_boot_irqs_disabled = true; | |
845 | ||
846 | /* | |
847 | * Interrupts are still disabled. Do necessary setups, then | |
848 | * enable them. | |
849 | */ | |
850 | boot_cpu_init(); | |
851 | page_address_init(); | |
852 | pr_notice("%s", linux_banner); | |
853 | early_security_init(); | |
854 | setup_arch(&command_line); | |
855 | setup_boot_config(command_line); | |
856 | setup_command_line(command_line); | |
857 | setup_nr_cpu_ids(); | |
858 | setup_per_cpu_areas(); | |
859 | smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ | |
860 | boot_cpu_hotplug_init(); | |
861 | ||
862 | build_all_zonelists(NULL); | |
863 | page_alloc_init(); | |
864 | ||
865 | pr_notice("Kernel command line: %s\n", saved_command_line); | |
866 | /* parameters may set static keys */ | |
867 | jump_label_init(); | |
868 | parse_early_param(); | |
869 | after_dashes = parse_args("Booting kernel", | |
870 | static_command_line, __start___param, | |
871 | __stop___param - __start___param, | |
872 | -1, -1, NULL, &unknown_bootoption); | |
873 | if (!IS_ERR_OR_NULL(after_dashes)) | |
874 | parse_args("Setting init args", after_dashes, NULL, 0, -1, -1, | |
875 | NULL, set_init_arg); | |
876 | if (extra_init_args) | |
877 | parse_args("Setting extra init args", extra_init_args, | |
878 | NULL, 0, -1, -1, NULL, set_init_arg); | |
879 | ||
880 | /* | |
881 | * These use large bootmem allocations and must precede | |
882 | * kmem_cache_init() | |
883 | */ | |
884 | setup_log_buf(0); | |
885 | vfs_caches_init_early(); | |
886 | sort_main_extable(); | |
887 | trap_init(); | |
888 | mm_init(); | |
889 | ||
890 | ftrace_init(); | |
891 | ||
892 | /* trace_printk can be enabled here */ | |
893 | early_trace_init(); | |
894 | ||
895 | /* | |
896 | * Set up the scheduler prior starting any interrupts (such as the | |
897 | * timer interrupt). Full topology setup happens at smp_init() | |
898 | * time - but meanwhile we still have a functioning scheduler. | |
899 | */ | |
900 | sched_init(); | |
901 | /* | |
902 | * Disable preemption - early bootup scheduling is extremely | |
903 | * fragile until we cpu_idle() for the first time. | |
904 | */ | |
905 | preempt_disable(); | |
906 | if (WARN(!irqs_disabled(), | |
907 | "Interrupts were enabled *very* early, fixing it\n")) | |
908 | local_irq_disable(); | |
909 | radix_tree_init(); | |
910 | ||
911 | /* | |
912 | * Set up housekeeping before setting up workqueues to allow the unbound | |
913 | * workqueue to take non-housekeeping into account. | |
914 | */ | |
915 | housekeeping_init(); | |
916 | ||
917 | /* | |
918 | * Allow workqueue creation and work item queueing/cancelling | |
919 | * early. Work item execution depends on kthreads and starts after | |
920 | * workqueue_init(). | |
921 | */ | |
922 | workqueue_init_early(); | |
923 | ||
924 | rcu_init(); | |
925 | ||
926 | /* Trace events are available after this */ | |
927 | trace_init(); | |
928 | ||
929 | if (initcall_debug) | |
930 | initcall_debug_enable(); | |
931 | ||
932 | context_tracking_init(); | |
933 | /* init some links before init_ISA_irqs() */ | |
934 | early_irq_init(); | |
935 | init_IRQ(); | |
936 | tick_init(); | |
937 | rcu_init_nohz(); | |
938 | init_timers(); | |
939 | hrtimers_init(); | |
940 | softirq_init(); | |
941 | timekeeping_init(); | |
942 | ||
943 | /* | |
944 | * For best initial stack canary entropy, prepare it after: | |
945 | * - setup_arch() for any UEFI RNG entropy and boot cmdline access | |
946 | * - timekeeping_init() for ktime entropy used in rand_initialize() | |
947 | * - rand_initialize() to get any arch-specific entropy like RDRAND | |
948 | * - add_latent_entropy() to get any latent entropy | |
949 | * - adding command line entropy | |
950 | */ | |
951 | rand_initialize(); | |
952 | add_latent_entropy(); | |
953 | add_device_randomness(command_line, strlen(command_line)); | |
954 | boot_init_stack_canary(); | |
955 | ||
956 | time_init(); | |
957 | perf_event_init(); | |
958 | profile_init(); | |
959 | call_function_init(); | |
960 | WARN(!irqs_disabled(), "Interrupts were enabled early\n"); | |
961 | ||
962 | early_boot_irqs_disabled = false; | |
963 | local_irq_enable(); | |
964 | ||
965 | kmem_cache_init_late(); | |
966 | ||
967 | /* | |
968 | * HACK ALERT! This is early. We're enabling the console before | |
969 | * we've done PCI setups etc, and console_init() must be aware of | |
970 | * this. But we do want output early, in case something goes wrong. | |
971 | */ | |
972 | console_init(); | |
973 | if (panic_later) | |
974 | panic("Too many boot %s vars at `%s'", panic_later, | |
975 | panic_param); | |
976 | ||
977 | lockdep_init(); | |
978 | ||
979 | /* | |
980 | * Need to run this when irqs are enabled, because it wants | |
981 | * to self-test [hard/soft]-irqs on/off lock inversion bugs | |
982 | * too: | |
983 | */ | |
984 | locking_selftest(); | |
985 | ||
986 | /* | |
987 | * This needs to be called before any devices perform DMA | |
988 | * operations that might use the SWIOTLB bounce buffers. It will | |
989 | * mark the bounce buffers as decrypted so that their usage will | |
990 | * not cause "plain-text" data to be decrypted when accessed. | |
991 | */ | |
992 | mem_encrypt_init(); | |
993 | ||
994 | #ifdef CONFIG_BLK_DEV_INITRD | |
995 | if (initrd_start && !initrd_below_start_ok && | |
996 | page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { | |
997 | pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", | |
998 | page_to_pfn(virt_to_page((void *)initrd_start)), | |
999 | min_low_pfn); | |
1000 | initrd_start = 0; | |
1001 | } | |
1002 | #endif | |
1003 | setup_per_cpu_pageset(); | |
1004 | numa_policy_init(); | |
1005 | acpi_early_init(); | |
1006 | if (late_time_init) | |
1007 | late_time_init(); | |
1008 | sched_clock_init(); | |
1009 | calibrate_delay(); | |
1010 | pid_idr_init(); | |
1011 | anon_vma_init(); | |
1012 | #ifdef CONFIG_X86 | |
1013 | if (efi_enabled(EFI_RUNTIME_SERVICES)) | |
1014 | efi_enter_virtual_mode(); | |
1015 | #endif | |
1016 | thread_stack_cache_init(); | |
1017 | cred_init(); | |
1018 | fork_init(); | |
1019 | proc_caches_init(); | |
1020 | uts_ns_init(); | |
1021 | buffer_init(); | |
1022 | key_init(); | |
1023 | security_init(); | |
1024 | dbg_late_init(); | |
1025 | vfs_caches_init(); | |
1026 | pagecache_init(); | |
1027 | signals_init(); | |
1028 | seq_file_init(); | |
1029 | proc_root_init(); | |
1030 | nsfs_init(); | |
1031 | cpuset_init(); | |
1032 | cgroup_init(); | |
1033 | taskstats_init_early(); | |
1034 | delayacct_init(); | |
1035 | ||
1036 | poking_init(); | |
1037 | check_bugs(); | |
1038 | ||
1039 | acpi_subsystem_init(); | |
1040 | arch_post_acpi_subsys_init(); | |
1041 | sfi_init_late(); | |
1042 | kcsan_init(); | |
1043 | ||
1044 | /* Do the rest non-__init'ed, we're now alive */ | |
1045 | arch_call_rest_init(); | |
1046 | ||
1047 | prevent_tail_call_optimization(); | |
1048 | } | |
1049 | ||
1050 | /* Call all constructor functions linked into the kernel. */ | |
1051 | static void __init do_ctors(void) | |
1052 | { | |
1053 | #ifdef CONFIG_CONSTRUCTORS | |
1054 | ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; | |
1055 | ||
1056 | for (; fn < (ctor_fn_t *) __ctors_end; fn++) | |
1057 | (*fn)(); | |
1058 | #endif | |
1059 | } | |
1060 | ||
1061 | #ifdef CONFIG_KALLSYMS | |
1062 | struct blacklist_entry { | |
1063 | struct list_head next; | |
1064 | char *buf; | |
1065 | }; | |
1066 | ||
1067 | static __initdata_or_module LIST_HEAD(blacklisted_initcalls); | |
1068 | ||
1069 | static int __init initcall_blacklist(char *str) | |
1070 | { | |
1071 | char *str_entry; | |
1072 | struct blacklist_entry *entry; | |
1073 | ||
1074 | /* str argument is a comma-separated list of functions */ | |
1075 | do { | |
1076 | str_entry = strsep(&str, ","); | |
1077 | if (str_entry) { | |
1078 | pr_debug("blacklisting initcall %s\n", str_entry); | |
1079 | entry = memblock_alloc(sizeof(*entry), | |
1080 | SMP_CACHE_BYTES); | |
1081 | if (!entry) | |
1082 | panic("%s: Failed to allocate %zu bytes\n", | |
1083 | __func__, sizeof(*entry)); | |
1084 | entry->buf = memblock_alloc(strlen(str_entry) + 1, | |
1085 | SMP_CACHE_BYTES); | |
1086 | if (!entry->buf) | |
1087 | panic("%s: Failed to allocate %zu bytes\n", | |
1088 | __func__, strlen(str_entry) + 1); | |
1089 | strcpy(entry->buf, str_entry); | |
1090 | list_add(&entry->next, &blacklisted_initcalls); | |
1091 | } | |
1092 | } while (str_entry); | |
1093 | ||
1094 | return 0; | |
1095 | } | |
1096 | ||
1097 | static bool __init_or_module initcall_blacklisted(initcall_t fn) | |
1098 | { | |
1099 | struct blacklist_entry *entry; | |
1100 | char fn_name[KSYM_SYMBOL_LEN]; | |
1101 | unsigned long addr; | |
1102 | ||
1103 | if (list_empty(&blacklisted_initcalls)) | |
1104 | return false; | |
1105 | ||
1106 | addr = (unsigned long) dereference_function_descriptor(fn); | |
1107 | sprint_symbol_no_offset(fn_name, addr); | |
1108 | ||
1109 | /* | |
1110 | * fn will be "function_name [module_name]" where [module_name] is not | |
1111 | * displayed for built-in init functions. Strip off the [module_name]. | |
1112 | */ | |
1113 | strreplace(fn_name, ' ', '\0'); | |
1114 | ||
1115 | list_for_each_entry(entry, &blacklisted_initcalls, next) { | |
1116 | if (!strcmp(fn_name, entry->buf)) { | |
1117 | pr_debug("initcall %s blacklisted\n", fn_name); | |
1118 | return true; | |
1119 | } | |
1120 | } | |
1121 | ||
1122 | return false; | |
1123 | } | |
1124 | #else | |
1125 | static int __init initcall_blacklist(char *str) | |
1126 | { | |
1127 | pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); | |
1128 | return 0; | |
1129 | } | |
1130 | ||
1131 | static bool __init_or_module initcall_blacklisted(initcall_t fn) | |
1132 | { | |
1133 | return false; | |
1134 | } | |
1135 | #endif | |
1136 | __setup("initcall_blacklist=", initcall_blacklist); | |
1137 | ||
1138 | static __init_or_module void | |
1139 | trace_initcall_start_cb(void *data, initcall_t fn) | |
1140 | { | |
1141 | ktime_t *calltime = (ktime_t *)data; | |
1142 | ||
1143 | printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current)); | |
1144 | *calltime = ktime_get(); | |
1145 | } | |
1146 | ||
1147 | static __init_or_module void | |
1148 | trace_initcall_finish_cb(void *data, initcall_t fn, int ret) | |
1149 | { | |
1150 | ktime_t *calltime = (ktime_t *)data; | |
1151 | ktime_t delta, rettime; | |
1152 | unsigned long long duration; | |
1153 | ||
1154 | rettime = ktime_get(); | |
1155 | delta = ktime_sub(rettime, *calltime); | |
1156 | duration = (unsigned long long) ktime_to_ns(delta) >> 10; | |
1157 | printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n", | |
1158 | fn, ret, duration); | |
1159 | } | |
1160 | ||
1161 | static ktime_t initcall_calltime; | |
1162 | ||
1163 | #ifdef TRACEPOINTS_ENABLED | |
1164 | static void __init initcall_debug_enable(void) | |
1165 | { | |
1166 | int ret; | |
1167 | ||
1168 | ret = register_trace_initcall_start(trace_initcall_start_cb, | |
1169 | &initcall_calltime); | |
1170 | ret |= register_trace_initcall_finish(trace_initcall_finish_cb, | |
1171 | &initcall_calltime); | |
1172 | WARN(ret, "Failed to register initcall tracepoints\n"); | |
1173 | } | |
1174 | # define do_trace_initcall_start trace_initcall_start | |
1175 | # define do_trace_initcall_finish trace_initcall_finish | |
1176 | #else | |
1177 | static inline void do_trace_initcall_start(initcall_t fn) | |
1178 | { | |
1179 | if (!initcall_debug) | |
1180 | return; | |
1181 | trace_initcall_start_cb(&initcall_calltime, fn); | |
1182 | } | |
1183 | static inline void do_trace_initcall_finish(initcall_t fn, int ret) | |
1184 | { | |
1185 | if (!initcall_debug) | |
1186 | return; | |
1187 | trace_initcall_finish_cb(&initcall_calltime, fn, ret); | |
1188 | } | |
1189 | #endif /* !TRACEPOINTS_ENABLED */ | |
1190 | ||
1191 | int __init_or_module do_one_initcall(initcall_t fn) | |
1192 | { | |
1193 | int count = preempt_count(); | |
1194 | char msgbuf[64]; | |
1195 | int ret; | |
1196 | ||
1197 | if (initcall_blacklisted(fn)) | |
1198 | return -EPERM; | |
1199 | ||
1200 | do_trace_initcall_start(fn); | |
1201 | ret = fn(); | |
1202 | do_trace_initcall_finish(fn, ret); | |
1203 | ||
1204 | msgbuf[0] = 0; | |
1205 | ||
1206 | if (preempt_count() != count) { | |
1207 | sprintf(msgbuf, "preemption imbalance "); | |
1208 | preempt_count_set(count); | |
1209 | } | |
1210 | if (irqs_disabled()) { | |
1211 | strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); | |
1212 | local_irq_enable(); | |
1213 | } | |
1214 | WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf); | |
1215 | ||
1216 | add_latent_entropy(); | |
1217 | return ret; | |
1218 | } | |
1219 | ||
1220 | ||
1221 | extern initcall_entry_t __initcall_start[]; | |
1222 | extern initcall_entry_t __initcall0_start[]; | |
1223 | extern initcall_entry_t __initcall1_start[]; | |
1224 | extern initcall_entry_t __initcall2_start[]; | |
1225 | extern initcall_entry_t __initcall3_start[]; | |
1226 | extern initcall_entry_t __initcall4_start[]; | |
1227 | extern initcall_entry_t __initcall5_start[]; | |
1228 | extern initcall_entry_t __initcall6_start[]; | |
1229 | extern initcall_entry_t __initcall7_start[]; | |
1230 | extern initcall_entry_t __initcall_end[]; | |
1231 | ||
1232 | static initcall_entry_t *initcall_levels[] __initdata = { | |
1233 | __initcall0_start, | |
1234 | __initcall1_start, | |
1235 | __initcall2_start, | |
1236 | __initcall3_start, | |
1237 | __initcall4_start, | |
1238 | __initcall5_start, | |
1239 | __initcall6_start, | |
1240 | __initcall7_start, | |
1241 | __initcall_end, | |
1242 | }; | |
1243 | ||
1244 | /* Keep these in sync with initcalls in include/linux/init.h */ | |
1245 | static const char *initcall_level_names[] __initdata = { | |
1246 | "pure", | |
1247 | "core", | |
1248 | "postcore", | |
1249 | "arch", | |
1250 | "subsys", | |
1251 | "fs", | |
1252 | "device", | |
1253 | "late", | |
1254 | }; | |
1255 | ||
1256 | static int __init ignore_unknown_bootoption(char *param, char *val, | |
1257 | const char *unused, void *arg) | |
1258 | { | |
1259 | return 0; | |
1260 | } | |
1261 | ||
1262 | static void __init do_initcall_level(int level, char *command_line) | |
1263 | { | |
1264 | initcall_entry_t *fn; | |
1265 | ||
1266 | parse_args(initcall_level_names[level], | |
1267 | command_line, __start___param, | |
1268 | __stop___param - __start___param, | |
1269 | level, level, | |
1270 | NULL, ignore_unknown_bootoption); | |
1271 | ||
1272 | trace_initcall_level(initcall_level_names[level]); | |
1273 | for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) | |
1274 | do_one_initcall(initcall_from_entry(fn)); | |
1275 | } | |
1276 | ||
1277 | static void __init do_initcalls(void) | |
1278 | { | |
1279 | int level; | |
1280 | size_t len = strlen(saved_command_line) + 1; | |
1281 | char *command_line; | |
1282 | ||
1283 | command_line = kzalloc(len, GFP_KERNEL); | |
1284 | if (!command_line) | |
1285 | panic("%s: Failed to allocate %zu bytes\n", __func__, len); | |
1286 | ||
1287 | for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) { | |
1288 | /* Parser modifies command_line, restore it each time */ | |
1289 | strcpy(command_line, saved_command_line); | |
1290 | do_initcall_level(level, command_line); | |
1291 | } | |
1292 | ||
1293 | kfree(command_line); | |
1294 | } | |
1295 | ||
1296 | /* | |
1297 | * Ok, the machine is now initialized. None of the devices | |
1298 | * have been touched yet, but the CPU subsystem is up and | |
1299 | * running, and memory and process management works. | |
1300 | * | |
1301 | * Now we can finally start doing some real work.. | |
1302 | */ | |
1303 | static void __init do_basic_setup(void) | |
1304 | { | |
1305 | cpuset_init_smp(); | |
1306 | driver_init(); | |
1307 | init_irq_proc(); | |
1308 | do_ctors(); | |
1309 | usermodehelper_enable(); | |
1310 | do_initcalls(); | |
1311 | } | |
1312 | ||
1313 | static void __init do_pre_smp_initcalls(void) | |
1314 | { | |
1315 | initcall_entry_t *fn; | |
1316 | ||
1317 | trace_initcall_level("early"); | |
1318 | for (fn = __initcall_start; fn < __initcall0_start; fn++) | |
1319 | do_one_initcall(initcall_from_entry(fn)); | |
1320 | } | |
1321 | ||
1322 | static int run_init_process(const char *init_filename) | |
1323 | { | |
1324 | const char *const *p; | |
1325 | ||
1326 | argv_init[0] = init_filename; | |
1327 | pr_info("Run %s as init process\n", init_filename); | |
1328 | pr_debug(" with arguments:\n"); | |
1329 | for (p = argv_init; *p; p++) | |
1330 | pr_debug(" %s\n", *p); | |
1331 | pr_debug(" with environment:\n"); | |
1332 | for (p = envp_init; *p; p++) | |
1333 | pr_debug(" %s\n", *p); | |
1334 | return kernel_execve(init_filename, argv_init, envp_init); | |
1335 | } | |
1336 | ||
1337 | static int try_to_run_init_process(const char *init_filename) | |
1338 | { | |
1339 | int ret; | |
1340 | ||
1341 | ret = run_init_process(init_filename); | |
1342 | ||
1343 | if (ret && ret != -ENOENT) { | |
1344 | pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", | |
1345 | init_filename, ret); | |
1346 | } | |
1347 | ||
1348 | return ret; | |
1349 | } | |
1350 | ||
1351 | static noinline void __init kernel_init_freeable(void); | |
1352 | ||
1353 | #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX) | |
1354 | bool rodata_enabled __ro_after_init = true; | |
1355 | static int __init set_debug_rodata(char *str) | |
1356 | { | |
1357 | return strtobool(str, &rodata_enabled); | |
1358 | } | |
1359 | __setup("rodata=", set_debug_rodata); | |
1360 | #endif | |
1361 | ||
1362 | #ifdef CONFIG_STRICT_KERNEL_RWX | |
1363 | static void mark_readonly(void) | |
1364 | { | |
1365 | if (rodata_enabled) { | |
1366 | /* | |
1367 | * load_module() results in W+X mappings, which are cleaned | |
1368 | * up with call_rcu(). Let's make sure that queued work is | |
1369 | * flushed so that we don't hit false positives looking for | |
1370 | * insecure pages which are W+X. | |
1371 | */ | |
1372 | rcu_barrier(); | |
1373 | mark_rodata_ro(); | |
1374 | rodata_test(); | |
1375 | } else | |
1376 | pr_info("Kernel memory protection disabled.\n"); | |
1377 | } | |
1378 | #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX) | |
1379 | static inline void mark_readonly(void) | |
1380 | { | |
1381 | pr_warn("Kernel memory protection not selected by kernel config.\n"); | |
1382 | } | |
1383 | #else | |
1384 | static inline void mark_readonly(void) | |
1385 | { | |
1386 | pr_warn("This architecture does not have kernel memory protection.\n"); | |
1387 | } | |
1388 | #endif | |
1389 | ||
1390 | void __weak free_initmem(void) | |
1391 | { | |
1392 | free_initmem_default(POISON_FREE_INITMEM); | |
1393 | } | |
1394 | ||
1395 | static int __ref kernel_init(void *unused) | |
1396 | { | |
1397 | int ret; | |
1398 | ||
1399 | kernel_init_freeable(); | |
1400 | /* need to finish all async __init code before freeing the memory */ | |
1401 | async_synchronize_full(); | |
1402 | ftrace_free_init_mem(); | |
1403 | free_initmem(); | |
1404 | mark_readonly(); | |
1405 | ||
1406 | /* | |
1407 | * Kernel mappings are now finalized - update the userspace page-table | |
1408 | * to finalize PTI. | |
1409 | */ | |
1410 | pti_finalize(); | |
1411 | ||
1412 | system_state = SYSTEM_RUNNING; | |
1413 | numa_default_policy(); | |
1414 | ||
1415 | rcu_end_inkernel_boot(); | |
1416 | ||
1417 | do_sysctl_args(); | |
1418 | ||
1419 | if (ramdisk_execute_command) { | |
1420 | ret = run_init_process(ramdisk_execute_command); | |
1421 | if (!ret) | |
1422 | return 0; | |
1423 | pr_err("Failed to execute %s (error %d)\n", | |
1424 | ramdisk_execute_command, ret); | |
1425 | } | |
1426 | ||
1427 | /* | |
1428 | * We try each of these until one succeeds. | |
1429 | * | |
1430 | * The Bourne shell can be used instead of init if we are | |
1431 | * trying to recover a really broken machine. | |
1432 | */ | |
1433 | if (execute_command) { | |
1434 | ret = run_init_process(execute_command); | |
1435 | if (!ret) | |
1436 | return 0; | |
1437 | panic("Requested init %s failed (error %d).", | |
1438 | execute_command, ret); | |
1439 | } | |
1440 | ||
1441 | if (CONFIG_DEFAULT_INIT[0] != '\0') { | |
1442 | ret = run_init_process(CONFIG_DEFAULT_INIT); | |
1443 | if (ret) | |
1444 | pr_err("Default init %s failed (error %d)\n", | |
1445 | CONFIG_DEFAULT_INIT, ret); | |
1446 | else | |
1447 | return 0; | |
1448 | } | |
1449 | ||
1450 | if (!try_to_run_init_process("/sbin/init") || | |
1451 | !try_to_run_init_process("/etc/init") || | |
1452 | !try_to_run_init_process("/bin/init") || | |
1453 | !try_to_run_init_process("/bin/sh")) | |
1454 | return 0; | |
1455 | ||
1456 | panic("No working init found. Try passing init= option to kernel. " | |
1457 | "See Linux Documentation/admin-guide/init.rst for guidance."); | |
1458 | } | |
1459 | ||
1460 | void console_on_rootfs(void) | |
1461 | { | |
1462 | /* Open the /dev/console as stdin, this should never fail */ | |
1463 | if (ksys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) | |
1464 | pr_err("Warning: unable to open an initial console.\n"); | |
1465 | ||
1466 | /* create stdout/stderr */ | |
1467 | (void) ksys_dup(0); | |
1468 | (void) ksys_dup(0); | |
1469 | } | |
1470 | ||
1471 | static noinline void __init kernel_init_freeable(void) | |
1472 | { | |
1473 | /* | |
1474 | * Wait until kthreadd is all set-up. | |
1475 | */ | |
1476 | wait_for_completion(&kthreadd_done); | |
1477 | ||
1478 | /* Now the scheduler is fully set up and can do blocking allocations */ | |
1479 | gfp_allowed_mask = __GFP_BITS_MASK; | |
1480 | ||
1481 | /* | |
1482 | * init can allocate pages on any node | |
1483 | */ | |
1484 | set_mems_allowed(node_states[N_MEMORY]); | |
1485 | ||
1486 | cad_pid = task_pid(current); | |
1487 | ||
1488 | smp_prepare_cpus(setup_max_cpus); | |
1489 | ||
1490 | workqueue_init(); | |
1491 | ||
1492 | init_mm_internals(); | |
1493 | ||
1494 | do_pre_smp_initcalls(); | |
1495 | lockup_detector_init(); | |
1496 | ||
1497 | smp_init(); | |
1498 | sched_init_smp(); | |
1499 | ||
1500 | padata_init(); | |
1501 | page_alloc_init_late(); | |
1502 | /* Initialize page ext after all struct pages are initialized. */ | |
1503 | page_ext_init(); | |
1504 | ||
1505 | do_basic_setup(); | |
1506 | ||
1507 | console_on_rootfs(); | |
1508 | ||
1509 | /* | |
1510 | * check if there is an early userspace init. If yes, let it do all | |
1511 | * the work | |
1512 | */ | |
1513 | ||
1514 | if (!ramdisk_execute_command) | |
1515 | ramdisk_execute_command = "/init"; | |
1516 | ||
1517 | if (ksys_access((const char __user *) | |
1518 | ramdisk_execute_command, 0) != 0) { | |
1519 | ramdisk_execute_command = NULL; | |
1520 | prepare_namespace(); | |
1521 | } | |
1522 | ||
1523 | /* | |
1524 | * Ok, we have completed the initial bootup, and | |
1525 | * we're essentially up and running. Get rid of the | |
1526 | * initmem segments and start the user-mode stuff.. | |
1527 | * | |
1528 | * rootfs is available now, try loading the public keys | |
1529 | * and default modules | |
1530 | */ | |
1531 | ||
1532 | integrity_load_keys(); | |
1533 | } |