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/kprobes.h>
37 #include <linux/vmalloc.h>
38 #include <linux/kernel_stat.h>
39 #include <linux/start_kernel.h>
40 #include <linux/security.h>
41 #include <linux/smp.h>
42 #include <linux/profile.h>
43 #include <linux/kfence.h>
44 #include <linux/rcupdate.h>
45 #include <linux/srcu.h>
46 #include <linux/moduleparam.h>
47 #include <linux/kallsyms.h>
48 #include <linux/buildid.h>
49 #include <linux/writeback.h>
50 #include <linux/cpu.h>
51 #include <linux/cpuset.h>
52 #include <linux/cgroup.h>
53 #include <linux/efi.h>
54 #include <linux/tick.h>
55 #include <linux/sched/isolation.h>
56 #include <linux/interrupt.h>
57 #include <linux/taskstats_kern.h>
58 #include <linux/delayacct.h>
59 #include <linux/unistd.h>
60 #include <linux/utsname.h>
61 #include <linux/rmap.h>
62 #include <linux/mempolicy.h>
63 #include <linux/key.h>
64 #include <linux/page_ext.h>
65 #include <linux/debug_locks.h>
66 #include <linux/debugobjects.h>
67 #include <linux/lockdep.h>
68 #include <linux/kmemleak.h>
69 #include <linux/padata.h>
70 #include <linux/pid_namespace.h>
71 #include <linux/device/driver.h>
72 #include <linux/kthread.h>
73 #include <linux/sched.h>
74 #include <linux/sched/init.h>
75 #include <linux/signal.h>
76 #include <linux/idr.h>
77 #include <linux/kgdb.h>
78 #include <linux/ftrace.h>
79 #include <linux/async.h>
80 #include <linux/shmem_fs.h>
81 #include <linux/slab.h>
82 #include <linux/perf_event.h>
83 #include <linux/ptrace.h>
84 #include <linux/pti.h>
85 #include <linux/blkdev.h>
86 #include <linux/elevator.h>
87 #include <linux/sched/clock.h>
88 #include <linux/sched/task.h>
89 #include <linux/sched/task_stack.h>
90 #include <linux/context_tracking.h>
91 #include <linux/random.h>
92 #include <linux/list.h>
93 #include <linux/integrity.h>
94 #include <linux/proc_ns.h>
95 #include <linux/io.h>
96 #include <linux/cache.h>
97 #include <linux/rodata_test.h>
98 #include <linux/jump_label.h>
99 #include <linux/kcsan.h>
100 #include <linux/init_syscalls.h>
101 #include <linux/stackdepot.h>
102 #include <linux/randomize_kstack.h>
103 #include <net/net_namespace.h>
104
105 #include <asm/io.h>
106 #include <asm/setup.h>
107 #include <asm/sections.h>
108 #include <asm/cacheflush.h>
109
110 #define CREATE_TRACE_POINTS
111 #include <trace/events/initcall.h>
112
113 #include <kunit/test.h>
114
115 static int kernel_init(void *);
116
117 extern void init_IRQ(void);
118 extern void radix_tree_init(void);
119
120 /*
121 * Debug helper: via this flag we know that we are in 'early bootup code'
122 * where only the boot processor is running with IRQ disabled. This means
123 * two things - IRQ must not be enabled before the flag is cleared and some
124 * operations which are not allowed with IRQ disabled are allowed while the
125 * flag is set.
126 */
127 bool early_boot_irqs_disabled __read_mostly;
128
129 enum system_states system_state __read_mostly;
130 EXPORT_SYMBOL(system_state);
131
132 /*
133 * Boot command-line arguments
134 */
135 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
136 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
137
138 extern void time_init(void);
139 /* Default late time init is NULL. archs can override this later. */
140 void (*__initdata late_time_init)(void);
141
142 /* Untouched command line saved by arch-specific code. */
143 char __initdata boot_command_line[COMMAND_LINE_SIZE];
144 /* Untouched saved command line (eg. for /proc) */
145 char *saved_command_line;
146 /* Command line for parameter parsing */
147 static char *static_command_line;
148 /* Untouched extra command line */
149 static char *extra_command_line;
150 /* Extra init arguments */
151 static char *extra_init_args;
152
153 #ifdef CONFIG_BOOT_CONFIG
154 /* Is bootconfig on command line? */
155 static bool bootconfig_found;
156 static size_t initargs_offs;
157 #else
158 # define bootconfig_found false
159 # define initargs_offs 0
160 #endif
161
162 static char *execute_command;
163 static char *ramdisk_execute_command = "/init";
164
165 /*
166 * Used to generate warnings if static_key manipulation functions are used
167 * before jump_label_init is called.
168 */
169 bool static_key_initialized __read_mostly;
170 EXPORT_SYMBOL_GPL(static_key_initialized);
171
172 /*
173 * If set, this is an indication to the drivers that reset the underlying
174 * device before going ahead with the initialization otherwise driver might
175 * rely on the BIOS and skip the reset operation.
176 *
177 * This is useful if kernel is booting in an unreliable environment.
178 * For ex. kdump situation where previous kernel has crashed, BIOS has been
179 * skipped and devices will be in unknown state.
180 */
181 unsigned int reset_devices;
182 EXPORT_SYMBOL(reset_devices);
183
set_reset_devices(char * str)184 static int __init set_reset_devices(char *str)
185 {
186 reset_devices = 1;
187 return 1;
188 }
189
190 __setup("reset_devices", set_reset_devices);
191
192 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
193 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
194 static const char *panic_later, *panic_param;
195
196 extern const struct obs_kernel_param __setup_start[], __setup_end[];
197
obsolete_checksetup(char * line)198 static bool __init obsolete_checksetup(char *line)
199 {
200 const struct obs_kernel_param *p;
201 bool had_early_param = false;
202
203 p = __setup_start;
204 do {
205 int n = strlen(p->str);
206 if (parameqn(line, p->str, n)) {
207 if (p->early) {
208 /* Already done in parse_early_param?
209 * (Needs exact match on param part).
210 * Keep iterating, as we can have early
211 * params and __setups of same names 8( */
212 if (line[n] == '\0' || line[n] == '=')
213 had_early_param = true;
214 } else if (!p->setup_func) {
215 pr_warn("Parameter %s is obsolete, ignored\n",
216 p->str);
217 return true;
218 } else if (p->setup_func(line + n))
219 return true;
220 }
221 p++;
222 } while (p < __setup_end);
223
224 return had_early_param;
225 }
226
227 /*
228 * This should be approx 2 Bo*oMips to start (note initial shift), and will
229 * still work even if initially too large, it will just take slightly longer
230 */
231 unsigned long loops_per_jiffy = (1<<12);
232 EXPORT_SYMBOL(loops_per_jiffy);
233
debug_kernel(char * str)234 static int __init debug_kernel(char *str)
235 {
236 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
237 return 0;
238 }
239
quiet_kernel(char * str)240 static int __init quiet_kernel(char *str)
241 {
242 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
243 return 0;
244 }
245
246 early_param("debug", debug_kernel);
247 early_param("quiet", quiet_kernel);
248
loglevel(char * str)249 static int __init loglevel(char *str)
250 {
251 int newlevel;
252
253 /*
254 * Only update loglevel value when a correct setting was passed,
255 * to prevent blind crashes (when loglevel being set to 0) that
256 * are quite hard to debug
257 */
258 if (get_option(&str, &newlevel)) {
259 console_loglevel = newlevel;
260 return 0;
261 }
262
263 return -EINVAL;
264 }
265
266 early_param("loglevel", loglevel);
267
268 #ifdef CONFIG_BLK_DEV_INITRD
get_boot_config_from_initrd(u32 * _size,u32 * _csum)269 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
270 {
271 u32 size, csum;
272 char *data;
273 u32 *hdr;
274 int i;
275
276 if (!initrd_end)
277 return NULL;
278
279 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
280 /*
281 * Since Grub may align the size of initrd to 4, we must
282 * check the preceding 3 bytes as well.
283 */
284 for (i = 0; i < 4; i++) {
285 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
286 goto found;
287 data--;
288 }
289 return NULL;
290
291 found:
292 hdr = (u32 *)(data - 8);
293 size = le32_to_cpu(hdr[0]);
294 csum = le32_to_cpu(hdr[1]);
295
296 data = ((void *)hdr) - size;
297 if ((unsigned long)data < initrd_start) {
298 pr_err("bootconfig size %d is greater than initrd size %ld\n",
299 size, initrd_end - initrd_start);
300 return NULL;
301 }
302
303 /* Remove bootconfig from initramfs/initrd */
304 initrd_end = (unsigned long)data;
305 if (_size)
306 *_size = size;
307 if (_csum)
308 *_csum = csum;
309
310 return data;
311 }
312 #else
get_boot_config_from_initrd(u32 * _size,u32 * _csum)313 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
314 {
315 return NULL;
316 }
317 #endif
318
319 #ifdef CONFIG_BOOT_CONFIG
320
321 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
322
323 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
324
xbc_snprint_cmdline(char * buf,size_t size,struct xbc_node * root)325 static int __init xbc_snprint_cmdline(char *buf, size_t size,
326 struct xbc_node *root)
327 {
328 struct xbc_node *knode, *vnode;
329 char *end = buf + size;
330 const char *val;
331 int ret;
332
333 xbc_node_for_each_key_value(root, knode, val) {
334 ret = xbc_node_compose_key_after(root, knode,
335 xbc_namebuf, XBC_KEYLEN_MAX);
336 if (ret < 0)
337 return ret;
338
339 vnode = xbc_node_get_child(knode);
340 if (!vnode) {
341 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
342 if (ret < 0)
343 return ret;
344 buf += ret;
345 continue;
346 }
347 xbc_array_for_each_value(vnode, val) {
348 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
349 xbc_namebuf, val);
350 if (ret < 0)
351 return ret;
352 buf += ret;
353 }
354 }
355
356 return buf - (end - size);
357 }
358 #undef rest
359
360 /* Make an extra command line under given key word */
xbc_make_cmdline(const char * key)361 static char * __init xbc_make_cmdline(const char *key)
362 {
363 struct xbc_node *root;
364 char *new_cmdline;
365 int ret, len = 0;
366
367 root = xbc_find_node(key);
368 if (!root)
369 return NULL;
370
371 /* Count required buffer size */
372 len = xbc_snprint_cmdline(NULL, 0, root);
373 if (len <= 0)
374 return NULL;
375
376 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
377 if (!new_cmdline) {
378 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
379 return NULL;
380 }
381
382 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
383 if (ret < 0 || ret > len) {
384 pr_err("Failed to print extra kernel cmdline.\n");
385 memblock_free_ptr(new_cmdline, len + 1);
386 return NULL;
387 }
388
389 return new_cmdline;
390 }
391
bootconfig_params(char * param,char * val,const char * unused,void * arg)392 static int __init bootconfig_params(char *param, char *val,
393 const char *unused, void *arg)
394 {
395 if (strcmp(param, "bootconfig") == 0) {
396 bootconfig_found = true;
397 }
398 return 0;
399 }
400
warn_bootconfig(char * str)401 static int __init warn_bootconfig(char *str)
402 {
403 /* The 'bootconfig' has been handled by bootconfig_params(). */
404 return 0;
405 }
406
setup_boot_config(void)407 static void __init setup_boot_config(void)
408 {
409 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
410 const char *msg;
411 int pos;
412 u32 size, csum;
413 char *data, *copy, *err;
414 int ret;
415
416 /* Cut out the bootconfig data even if we have no bootconfig option */
417 data = get_boot_config_from_initrd(&size, &csum);
418
419 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
420 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
421 bootconfig_params);
422
423 if (IS_ERR(err) || !bootconfig_found)
424 return;
425
426 /* parse_args() stops at the next param of '--' and returns an address */
427 if (err)
428 initargs_offs = err - tmp_cmdline;
429
430 if (!data) {
431 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
432 return;
433 }
434
435 if (size >= XBC_DATA_MAX) {
436 pr_err("bootconfig size %d greater than max size %d\n",
437 size, XBC_DATA_MAX);
438 return;
439 }
440
441 if (xbc_calc_checksum(data, size) != csum) {
442 pr_err("bootconfig checksum failed\n");
443 return;
444 }
445
446 copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
447 if (!copy) {
448 pr_err("Failed to allocate memory for bootconfig\n");
449 return;
450 }
451
452 memcpy(copy, data, size);
453 copy[size] = '\0';
454
455 ret = xbc_init(copy, &msg, &pos);
456 if (ret < 0) {
457 if (pos < 0)
458 pr_err("Failed to init bootconfig: %s.\n", msg);
459 else
460 pr_err("Failed to parse bootconfig: %s at %d.\n",
461 msg, pos);
462 } else {
463 pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
464 /* keys starting with "kernel." are passed via cmdline */
465 extra_command_line = xbc_make_cmdline("kernel");
466 /* Also, "init." keys are init arguments */
467 extra_init_args = xbc_make_cmdline("init");
468 }
469 return;
470 }
471
exit_boot_config(void)472 static void __init exit_boot_config(void)
473 {
474 xbc_destroy_all();
475 }
476
477 #else /* !CONFIG_BOOT_CONFIG */
478
setup_boot_config(void)479 static void __init setup_boot_config(void)
480 {
481 /* Remove bootconfig data from initrd */
482 get_boot_config_from_initrd(NULL, NULL);
483 }
484
warn_bootconfig(char * str)485 static int __init warn_bootconfig(char *str)
486 {
487 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
488 return 0;
489 }
490
491 #define exit_boot_config() do {} while (0)
492
493 #endif /* CONFIG_BOOT_CONFIG */
494
495 early_param("bootconfig", warn_bootconfig);
496
497 /* Change NUL term back to "=", to make "param" the whole string. */
repair_env_string(char * param,char * val)498 static void __init repair_env_string(char *param, char *val)
499 {
500 if (val) {
501 /* param=val or param="val"? */
502 if (val == param+strlen(param)+1)
503 val[-1] = '=';
504 else if (val == param+strlen(param)+2) {
505 val[-2] = '=';
506 memmove(val-1, val, strlen(val)+1);
507 } else
508 BUG();
509 }
510 }
511
512 /* Anything after -- gets handed straight to init. */
set_init_arg(char * param,char * val,const char * unused,void * arg)513 static int __init set_init_arg(char *param, char *val,
514 const char *unused, void *arg)
515 {
516 unsigned int i;
517
518 if (panic_later)
519 return 0;
520
521 repair_env_string(param, val);
522
523 for (i = 0; argv_init[i]; i++) {
524 if (i == MAX_INIT_ARGS) {
525 panic_later = "init";
526 panic_param = param;
527 return 0;
528 }
529 }
530 argv_init[i] = param;
531 return 0;
532 }
533
534 /*
535 * Unknown boot options get handed to init, unless they look like
536 * unused parameters (modprobe will find them in /proc/cmdline).
537 */
unknown_bootoption(char * param,char * val,const char * unused,void * arg)538 static int __init unknown_bootoption(char *param, char *val,
539 const char *unused, void *arg)
540 {
541 size_t len = strlen(param);
542
543 /* Handle params aliased to sysctls */
544 if (sysctl_is_alias(param))
545 return 0;
546
547 repair_env_string(param, val);
548
549 /* Handle obsolete-style parameters */
550 if (obsolete_checksetup(param))
551 return 0;
552
553 /* Unused module parameter. */
554 if (strnchr(param, len, '.'))
555 return 0;
556
557 if (panic_later)
558 return 0;
559
560 if (val) {
561 /* Environment option */
562 unsigned int i;
563 for (i = 0; envp_init[i]; i++) {
564 if (i == MAX_INIT_ENVS) {
565 panic_later = "env";
566 panic_param = param;
567 }
568 if (!strncmp(param, envp_init[i], len+1))
569 break;
570 }
571 envp_init[i] = param;
572 } else {
573 /* Command line option */
574 unsigned int i;
575 for (i = 0; argv_init[i]; i++) {
576 if (i == MAX_INIT_ARGS) {
577 panic_later = "init";
578 panic_param = param;
579 }
580 }
581 argv_init[i] = param;
582 }
583 return 0;
584 }
585
init_setup(char * str)586 static int __init init_setup(char *str)
587 {
588 unsigned int i;
589
590 execute_command = str;
591 /*
592 * In case LILO is going to boot us with default command line,
593 * it prepends "auto" before the whole cmdline which makes
594 * the shell think it should execute a script with such name.
595 * So we ignore all arguments entered _before_ init=... [MJ]
596 */
597 for (i = 1; i < MAX_INIT_ARGS; i++)
598 argv_init[i] = NULL;
599 return 1;
600 }
601 __setup("init=", init_setup);
602
rdinit_setup(char * str)603 static int __init rdinit_setup(char *str)
604 {
605 unsigned int i;
606
607 ramdisk_execute_command = str;
608 /* See "auto" comment in init_setup */
609 for (i = 1; i < MAX_INIT_ARGS; i++)
610 argv_init[i] = NULL;
611 return 1;
612 }
613 __setup("rdinit=", rdinit_setup);
614
615 #ifndef CONFIG_SMP
616 static const unsigned int setup_max_cpus = NR_CPUS;
setup_nr_cpu_ids(void)617 static inline void setup_nr_cpu_ids(void) { }
smp_prepare_cpus(unsigned int maxcpus)618 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
619 #endif
620
621 /*
622 * We need to store the untouched command line for future reference.
623 * We also need to store the touched command line since the parameter
624 * parsing is performed in place, and we should allow a component to
625 * store reference of name/value for future reference.
626 */
setup_command_line(char * command_line)627 static void __init setup_command_line(char *command_line)
628 {
629 size_t len, xlen = 0, ilen = 0;
630
631 if (extra_command_line)
632 xlen = strlen(extra_command_line);
633 if (extra_init_args)
634 ilen = strlen(extra_init_args) + 4; /* for " -- " */
635
636 len = xlen + strlen(boot_command_line) + 1;
637
638 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
639 if (!saved_command_line)
640 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
641
642 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
643 if (!static_command_line)
644 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
645
646 if (xlen) {
647 /*
648 * We have to put extra_command_line before boot command
649 * lines because there could be dashes (separator of init
650 * command line) in the command lines.
651 */
652 strcpy(saved_command_line, extra_command_line);
653 strcpy(static_command_line, extra_command_line);
654 }
655 strcpy(saved_command_line + xlen, boot_command_line);
656 strcpy(static_command_line + xlen, command_line);
657
658 if (ilen) {
659 /*
660 * Append supplemental init boot args to saved_command_line
661 * so that user can check what command line options passed
662 * to init.
663 * The order should always be
664 * " -- "[bootconfig init-param][cmdline init-param]
665 */
666 if (initargs_offs) {
667 len = xlen + initargs_offs;
668 strcpy(saved_command_line + len, extra_init_args);
669 len += ilen - 4; /* strlen(extra_init_args) */
670 strcpy(saved_command_line + len,
671 boot_command_line + initargs_offs - 1);
672 } else {
673 len = strlen(saved_command_line);
674 strcpy(saved_command_line + len, " -- ");
675 len += 4;
676 strcpy(saved_command_line + len, extra_init_args);
677 }
678 }
679 }
680
681 /*
682 * We need to finalize in a non-__init function or else race conditions
683 * between the root thread and the init thread may cause start_kernel to
684 * be reaped by free_initmem before the root thread has proceeded to
685 * cpu_idle.
686 *
687 * gcc-3.4 accidentally inlines this function, so use noinline.
688 */
689
690 static __initdata DECLARE_COMPLETION(kthreadd_done);
691
rest_init(void)692 noinline void __ref rest_init(void)
693 {
694 struct task_struct *tsk;
695 int pid;
696
697 rcu_scheduler_starting();
698 /*
699 * We need to spawn init first so that it obtains pid 1, however
700 * the init task will end up wanting to create kthreads, which, if
701 * we schedule it before we create kthreadd, will OOPS.
702 */
703 pid = kernel_thread(kernel_init, NULL, CLONE_FS);
704 /*
705 * Pin init on the boot CPU. Task migration is not properly working
706 * until sched_init_smp() has been run. It will set the allowed
707 * CPUs for init to the non isolated CPUs.
708 */
709 rcu_read_lock();
710 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
711 tsk->flags |= PF_NO_SETAFFINITY;
712 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
713 rcu_read_unlock();
714
715 numa_default_policy();
716 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
717 rcu_read_lock();
718 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
719 rcu_read_unlock();
720
721 /*
722 * Enable might_sleep() and smp_processor_id() checks.
723 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
724 * kernel_thread() would trigger might_sleep() splats. With
725 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
726 * already, but it's stuck on the kthreadd_done completion.
727 */
728 system_state = SYSTEM_SCHEDULING;
729
730 complete(&kthreadd_done);
731
732 /*
733 * The boot idle thread must execute schedule()
734 * at least once to get things moving:
735 */
736 schedule_preempt_disabled();
737 /* Call into cpu_idle with preempt disabled */
738 cpu_startup_entry(CPUHP_ONLINE);
739 }
740
741 /* Check for early params. */
do_early_param(char * param,char * val,const char * unused,void * arg)742 static int __init do_early_param(char *param, char *val,
743 const char *unused, void *arg)
744 {
745 const struct obs_kernel_param *p;
746
747 for (p = __setup_start; p < __setup_end; p++) {
748 if ((p->early && parameq(param, p->str)) ||
749 (strcmp(param, "console") == 0 &&
750 strcmp(p->str, "earlycon") == 0)
751 ) {
752 if (p->setup_func(val) != 0)
753 pr_warn("Malformed early option '%s'\n", param);
754 }
755 }
756 /* We accept everything at this stage. */
757 return 0;
758 }
759
parse_early_options(char * cmdline)760 void __init parse_early_options(char *cmdline)
761 {
762 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
763 do_early_param);
764 }
765
766 /* Arch code calls this early on, or if not, just before other parsing. */
parse_early_param(void)767 void __init parse_early_param(void)
768 {
769 static int done __initdata;
770 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
771
772 if (done)
773 return;
774
775 /* All fall through to do_early_param. */
776 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
777 parse_early_options(tmp_cmdline);
778 done = 1;
779 }
780
arch_post_acpi_subsys_init(void)781 void __init __weak arch_post_acpi_subsys_init(void) { }
782
smp_setup_processor_id(void)783 void __init __weak smp_setup_processor_id(void)
784 {
785 }
786
787 # if THREAD_SIZE >= PAGE_SIZE
thread_stack_cache_init(void)788 void __init __weak thread_stack_cache_init(void)
789 {
790 }
791 #endif
792
poking_init(void)793 void __init __weak poking_init(void) { }
794
pgtable_cache_init(void)795 void __init __weak pgtable_cache_init(void) { }
796
trap_init(void)797 void __init __weak trap_init(void) { }
798
799 bool initcall_debug;
800 core_param(initcall_debug, initcall_debug, bool, 0644);
801
802 #ifdef TRACEPOINTS_ENABLED
803 static void __init initcall_debug_enable(void);
804 #else
initcall_debug_enable(void)805 static inline void initcall_debug_enable(void)
806 {
807 }
808 #endif
809
810 /* Report memory auto-initialization states for this boot. */
report_meminit(void)811 static void __init report_meminit(void)
812 {
813 const char *stack;
814
815 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
816 stack = "all(pattern)";
817 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
818 stack = "all(zero)";
819 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
820 stack = "byref_all(zero)";
821 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
822 stack = "byref(zero)";
823 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
824 stack = "__user(zero)";
825 else
826 stack = "off";
827
828 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
829 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
830 want_init_on_free() ? "on" : "off");
831 if (want_init_on_free())
832 pr_info("mem auto-init: clearing system memory may take some time...\n");
833 }
834
835 /*
836 * Set up kernel memory allocators
837 */
mm_init(void)838 static void __init mm_init(void)
839 {
840 /*
841 * page_ext requires contiguous pages,
842 * bigger than MAX_ORDER unless SPARSEMEM.
843 */
844 page_ext_init_flatmem();
845 init_mem_debugging_and_hardening();
846 kfence_alloc_pool();
847 report_meminit();
848 stack_depot_init();
849 mem_init();
850 mem_init_print_info();
851 /* page_owner must be initialized after buddy is ready */
852 page_ext_init_flatmem_late();
853 kmem_cache_init();
854 kmemleak_init();
855 pgtable_init();
856 debug_objects_mem_init();
857 vmalloc_init();
858 /* Should be run before the first non-init thread is created */
859 init_espfix_bsp();
860 /* Should be run after espfix64 is set up. */
861 pti_init();
862 mm_cache_init();
863 }
864
865 #ifdef CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
866 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
867 randomize_kstack_offset);
868 DEFINE_PER_CPU(u32, kstack_offset);
869
early_randomize_kstack_offset(char * buf)870 static int __init early_randomize_kstack_offset(char *buf)
871 {
872 int ret;
873 bool bool_result;
874
875 ret = kstrtobool(buf, &bool_result);
876 if (ret)
877 return ret;
878
879 if (bool_result)
880 static_branch_enable(&randomize_kstack_offset);
881 else
882 static_branch_disable(&randomize_kstack_offset);
883 return 0;
884 }
885 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
886 #endif
887
arch_call_rest_init(void)888 void __init __weak arch_call_rest_init(void)
889 {
890 rest_init();
891 }
892
print_unknown_bootoptions(void)893 static void __init print_unknown_bootoptions(void)
894 {
895 char *unknown_options;
896 char *end;
897 const char *const *p;
898 size_t len;
899
900 if (panic_later || (!argv_init[1] && !envp_init[2]))
901 return;
902
903 /*
904 * Determine how many options we have to print out, plus a space
905 * before each
906 */
907 len = 1; /* null terminator */
908 for (p = &argv_init[1]; *p; p++) {
909 len++;
910 len += strlen(*p);
911 }
912 for (p = &envp_init[2]; *p; p++) {
913 len++;
914 len += strlen(*p);
915 }
916
917 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
918 if (!unknown_options) {
919 pr_err("%s: Failed to allocate %zu bytes\n",
920 __func__, len);
921 return;
922 }
923 end = unknown_options;
924
925 for (p = &argv_init[1]; *p; p++)
926 end += sprintf(end, " %s", *p);
927 for (p = &envp_init[2]; *p; p++)
928 end += sprintf(end, " %s", *p);
929
930 /* Start at unknown_options[1] to skip the initial space */
931 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
932 &unknown_options[1]);
933 memblock_free_ptr(unknown_options, len);
934 }
935
start_kernel(void)936 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
937 {
938 char *command_line;
939 char *after_dashes;
940
941 set_task_stack_end_magic(&init_task);
942 smp_setup_processor_id();
943 debug_objects_early_init();
944 init_vmlinux_build_id();
945
946 cgroup_init_early();
947
948 local_irq_disable();
949 early_boot_irqs_disabled = true;
950
951 /*
952 * Interrupts are still disabled. Do necessary setups, then
953 * enable them.
954 */
955 boot_cpu_init();
956 page_address_init();
957 pr_notice("%s", linux_banner);
958 early_security_init();
959 setup_arch(&command_line);
960 setup_boot_config();
961 setup_command_line(command_line);
962 setup_nr_cpu_ids();
963 setup_per_cpu_areas();
964 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
965 boot_cpu_hotplug_init();
966
967 build_all_zonelists(NULL);
968 page_alloc_init();
969
970 pr_notice("Kernel command line: %s\n", saved_command_line);
971 /* parameters may set static keys */
972 jump_label_init();
973 parse_early_param();
974 after_dashes = parse_args("Booting kernel",
975 static_command_line, __start___param,
976 __stop___param - __start___param,
977 -1, -1, NULL, &unknown_bootoption);
978 print_unknown_bootoptions();
979 if (!IS_ERR_OR_NULL(after_dashes))
980 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
981 NULL, set_init_arg);
982 if (extra_init_args)
983 parse_args("Setting extra init args", extra_init_args,
984 NULL, 0, -1, -1, NULL, set_init_arg);
985
986 /*
987 * These use large bootmem allocations and must precede
988 * kmem_cache_init()
989 */
990 setup_log_buf(0);
991 vfs_caches_init_early();
992 sort_main_extable();
993 trap_init();
994 mm_init();
995 poking_init();
996 ftrace_init();
997
998 /* trace_printk can be enabled here */
999 early_trace_init();
1000
1001 /*
1002 * Set up the scheduler prior starting any interrupts (such as the
1003 * timer interrupt). Full topology setup happens at smp_init()
1004 * time - but meanwhile we still have a functioning scheduler.
1005 */
1006 sched_init();
1007
1008 if (WARN(!irqs_disabled(),
1009 "Interrupts were enabled *very* early, fixing it\n"))
1010 local_irq_disable();
1011 radix_tree_init();
1012
1013 /*
1014 * Set up housekeeping before setting up workqueues to allow the unbound
1015 * workqueue to take non-housekeeping into account.
1016 */
1017 housekeeping_init();
1018
1019 /*
1020 * Allow workqueue creation and work item queueing/cancelling
1021 * early. Work item execution depends on kthreads and starts after
1022 * workqueue_init().
1023 */
1024 workqueue_init_early();
1025
1026 rcu_init();
1027
1028 /* Trace events are available after this */
1029 trace_init();
1030
1031 if (initcall_debug)
1032 initcall_debug_enable();
1033
1034 context_tracking_init();
1035 /* init some links before init_ISA_irqs() */
1036 early_irq_init();
1037 init_IRQ();
1038 tick_init();
1039 rcu_init_nohz();
1040 init_timers();
1041 srcu_init();
1042 hrtimers_init();
1043 softirq_init();
1044 timekeeping_init();
1045 kfence_init();
1046 time_init();
1047
1048 /*
1049 * For best initial stack canary entropy, prepare it after:
1050 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
1051 * - timekeeping_init() for ktime entropy used in random_init()
1052 * - time_init() for making random_get_entropy() work on some platforms
1053 * - random_init() to initialize the RNG from from early entropy sources
1054 */
1055 random_init(command_line);
1056 boot_init_stack_canary();
1057
1058 perf_event_init();
1059 profile_init();
1060 call_function_init();
1061 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1062
1063 early_boot_irqs_disabled = false;
1064 local_irq_enable();
1065
1066 kmem_cache_init_late();
1067
1068 /*
1069 * HACK ALERT! This is early. We're enabling the console before
1070 * we've done PCI setups etc, and console_init() must be aware of
1071 * this. But we do want output early, in case something goes wrong.
1072 */
1073 console_init();
1074 if (panic_later)
1075 panic("Too many boot %s vars at `%s'", panic_later,
1076 panic_param);
1077
1078 lockdep_init();
1079
1080 /*
1081 * Need to run this when irqs are enabled, because it wants
1082 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1083 * too:
1084 */
1085 locking_selftest();
1086
1087 #ifdef CONFIG_BLK_DEV_INITRD
1088 if (initrd_start && !initrd_below_start_ok &&
1089 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1090 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1091 page_to_pfn(virt_to_page((void *)initrd_start)),
1092 min_low_pfn);
1093 initrd_start = 0;
1094 }
1095 #endif
1096 setup_per_cpu_pageset();
1097 numa_policy_init();
1098 acpi_early_init();
1099 if (late_time_init)
1100 late_time_init();
1101 sched_clock_init();
1102 calibrate_delay();
1103
1104 arch_cpu_finalize_init();
1105
1106 pid_idr_init();
1107 anon_vma_init();
1108 #ifdef CONFIG_X86
1109 if (efi_enabled(EFI_RUNTIME_SERVICES))
1110 efi_enter_virtual_mode();
1111 #endif
1112 thread_stack_cache_init();
1113 cred_init();
1114 fork_init();
1115 proc_caches_init();
1116 uts_ns_init();
1117 key_init();
1118 security_init();
1119 dbg_late_init();
1120 net_ns_init();
1121 vfs_caches_init();
1122 pagecache_init();
1123 signals_init();
1124 seq_file_init();
1125 proc_root_init();
1126 nsfs_init();
1127 cpuset_init();
1128 cgroup_init();
1129 taskstats_init_early();
1130 delayacct_init();
1131
1132 acpi_subsystem_init();
1133 arch_post_acpi_subsys_init();
1134 kcsan_init();
1135
1136 /* Do the rest non-__init'ed, we're now alive */
1137 arch_call_rest_init();
1138
1139 prevent_tail_call_optimization();
1140 }
1141
1142 /* Call all constructor functions linked into the kernel. */
do_ctors(void)1143 static void __init do_ctors(void)
1144 {
1145 /*
1146 * For UML, the constructors have already been called by the
1147 * normal setup code as it's just a normal ELF binary, so we
1148 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1149 * even on UML for modules.
1150 */
1151 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1152 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1153
1154 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1155 (*fn)();
1156 #endif
1157 }
1158
1159 #ifdef CONFIG_KALLSYMS
1160 struct blacklist_entry {
1161 struct list_head next;
1162 char *buf;
1163 };
1164
1165 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1166
initcall_blacklist(char * str)1167 static int __init initcall_blacklist(char *str)
1168 {
1169 char *str_entry;
1170 struct blacklist_entry *entry;
1171
1172 /* str argument is a comma-separated list of functions */
1173 do {
1174 str_entry = strsep(&str, ",");
1175 if (str_entry) {
1176 pr_debug("blacklisting initcall %s\n", str_entry);
1177 entry = memblock_alloc(sizeof(*entry),
1178 SMP_CACHE_BYTES);
1179 if (!entry)
1180 panic("%s: Failed to allocate %zu bytes\n",
1181 __func__, sizeof(*entry));
1182 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1183 SMP_CACHE_BYTES);
1184 if (!entry->buf)
1185 panic("%s: Failed to allocate %zu bytes\n",
1186 __func__, strlen(str_entry) + 1);
1187 strcpy(entry->buf, str_entry);
1188 list_add(&entry->next, &blacklisted_initcalls);
1189 }
1190 } while (str_entry);
1191
1192 return 1;
1193 }
1194
initcall_blacklisted(initcall_t fn)1195 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1196 {
1197 struct blacklist_entry *entry;
1198 char fn_name[KSYM_SYMBOL_LEN];
1199 unsigned long addr;
1200
1201 if (list_empty(&blacklisted_initcalls))
1202 return false;
1203
1204 addr = (unsigned long) dereference_function_descriptor(fn);
1205 sprint_symbol_no_offset(fn_name, addr);
1206
1207 /*
1208 * fn will be "function_name [module_name]" where [module_name] is not
1209 * displayed for built-in init functions. Strip off the [module_name].
1210 */
1211 strreplace(fn_name, ' ', '\0');
1212
1213 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1214 if (!strcmp(fn_name, entry->buf)) {
1215 pr_debug("initcall %s blacklisted\n", fn_name);
1216 return true;
1217 }
1218 }
1219
1220 return false;
1221 }
1222 #else
initcall_blacklist(char * str)1223 static int __init initcall_blacklist(char *str)
1224 {
1225 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1226 return 0;
1227 }
1228
initcall_blacklisted(initcall_t fn)1229 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1230 {
1231 return false;
1232 }
1233 #endif
1234 __setup("initcall_blacklist=", initcall_blacklist);
1235
1236 static __init_or_module void
trace_initcall_start_cb(void * data,initcall_t fn)1237 trace_initcall_start_cb(void *data, initcall_t fn)
1238 {
1239 ktime_t *calltime = (ktime_t *)data;
1240
1241 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1242 *calltime = ktime_get();
1243 }
1244
1245 static __init_or_module void
trace_initcall_finish_cb(void * data,initcall_t fn,int ret)1246 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1247 {
1248 ktime_t *calltime = (ktime_t *)data;
1249 ktime_t delta, rettime;
1250 unsigned long long duration;
1251
1252 rettime = ktime_get();
1253 delta = ktime_sub(rettime, *calltime);
1254 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1255 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1256 fn, ret, duration);
1257 }
1258
1259 static ktime_t initcall_calltime;
1260
1261 #ifdef TRACEPOINTS_ENABLED
initcall_debug_enable(void)1262 static void __init initcall_debug_enable(void)
1263 {
1264 int ret;
1265
1266 ret = register_trace_initcall_start(trace_initcall_start_cb,
1267 &initcall_calltime);
1268 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1269 &initcall_calltime);
1270 WARN(ret, "Failed to register initcall tracepoints\n");
1271 }
1272 # define do_trace_initcall_start trace_initcall_start
1273 # define do_trace_initcall_finish trace_initcall_finish
1274 #else
do_trace_initcall_start(initcall_t fn)1275 static inline void do_trace_initcall_start(initcall_t fn)
1276 {
1277 if (!initcall_debug)
1278 return;
1279 trace_initcall_start_cb(&initcall_calltime, fn);
1280 }
do_trace_initcall_finish(initcall_t fn,int ret)1281 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1282 {
1283 if (!initcall_debug)
1284 return;
1285 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1286 }
1287 #endif /* !TRACEPOINTS_ENABLED */
1288
do_one_initcall(initcall_t fn)1289 int __init_or_module do_one_initcall(initcall_t fn)
1290 {
1291 int count = preempt_count();
1292 char msgbuf[64];
1293 int ret;
1294
1295 if (initcall_blacklisted(fn))
1296 return -EPERM;
1297
1298 do_trace_initcall_start(fn);
1299 ret = fn();
1300 do_trace_initcall_finish(fn, ret);
1301
1302 msgbuf[0] = 0;
1303
1304 if (preempt_count() != count) {
1305 sprintf(msgbuf, "preemption imbalance ");
1306 preempt_count_set(count);
1307 }
1308 if (irqs_disabled()) {
1309 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1310 local_irq_enable();
1311 }
1312 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1313
1314 add_latent_entropy();
1315 return ret;
1316 }
1317
1318
1319 extern initcall_entry_t __initcall_start[];
1320 extern initcall_entry_t __initcall0_start[];
1321 extern initcall_entry_t __initcall1_start[];
1322 extern initcall_entry_t __initcall2_start[];
1323 extern initcall_entry_t __initcall3_start[];
1324 extern initcall_entry_t __initcall4_start[];
1325 extern initcall_entry_t __initcall5_start[];
1326 extern initcall_entry_t __initcall6_start[];
1327 extern initcall_entry_t __initcall7_start[];
1328 extern initcall_entry_t __initcall_end[];
1329
1330 static initcall_entry_t *initcall_levels[] __initdata = {
1331 __initcall0_start,
1332 __initcall1_start,
1333 __initcall2_start,
1334 __initcall3_start,
1335 __initcall4_start,
1336 __initcall5_start,
1337 __initcall6_start,
1338 __initcall7_start,
1339 __initcall_end,
1340 };
1341
1342 /* Keep these in sync with initcalls in include/linux/init.h */
1343 static const char *initcall_level_names[] __initdata = {
1344 "pure",
1345 "core",
1346 "postcore",
1347 "arch",
1348 "subsys",
1349 "fs",
1350 "device",
1351 "late",
1352 };
1353
ignore_unknown_bootoption(char * param,char * val,const char * unused,void * arg)1354 static int __init ignore_unknown_bootoption(char *param, char *val,
1355 const char *unused, void *arg)
1356 {
1357 return 0;
1358 }
1359
do_initcall_level(int level,char * command_line)1360 static void __init do_initcall_level(int level, char *command_line)
1361 {
1362 initcall_entry_t *fn;
1363
1364 parse_args(initcall_level_names[level],
1365 command_line, __start___param,
1366 __stop___param - __start___param,
1367 level, level,
1368 NULL, ignore_unknown_bootoption);
1369
1370 trace_initcall_level(initcall_level_names[level]);
1371 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1372 do_one_initcall(initcall_from_entry(fn));
1373 }
1374
do_initcalls(void)1375 static void __init do_initcalls(void)
1376 {
1377 int level;
1378 size_t len = strlen(saved_command_line) + 1;
1379 char *command_line;
1380
1381 command_line = kzalloc(len, GFP_KERNEL);
1382 if (!command_line)
1383 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1384
1385 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1386 /* Parser modifies command_line, restore it each time */
1387 strcpy(command_line, saved_command_line);
1388 do_initcall_level(level, command_line);
1389 }
1390
1391 kfree(command_line);
1392 }
1393
1394 /*
1395 * Ok, the machine is now initialized. None of the devices
1396 * have been touched yet, but the CPU subsystem is up and
1397 * running, and memory and process management works.
1398 *
1399 * Now we can finally start doing some real work..
1400 */
do_basic_setup(void)1401 static void __init do_basic_setup(void)
1402 {
1403 cpuset_init_smp();
1404 driver_init();
1405 init_irq_proc();
1406 do_ctors();
1407 do_initcalls();
1408 }
1409
do_pre_smp_initcalls(void)1410 static void __init do_pre_smp_initcalls(void)
1411 {
1412 initcall_entry_t *fn;
1413
1414 trace_initcall_level("early");
1415 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1416 do_one_initcall(initcall_from_entry(fn));
1417 }
1418
run_init_process(const char * init_filename)1419 static int run_init_process(const char *init_filename)
1420 {
1421 const char *const *p;
1422
1423 argv_init[0] = init_filename;
1424 pr_info("Run %s as init process\n", init_filename);
1425 pr_debug(" with arguments:\n");
1426 for (p = argv_init; *p; p++)
1427 pr_debug(" %s\n", *p);
1428 pr_debug(" with environment:\n");
1429 for (p = envp_init; *p; p++)
1430 pr_debug(" %s\n", *p);
1431 return kernel_execve(init_filename, argv_init, envp_init);
1432 }
1433
try_to_run_init_process(const char * init_filename)1434 static int try_to_run_init_process(const char *init_filename)
1435 {
1436 int ret;
1437
1438 ret = run_init_process(init_filename);
1439
1440 if (ret && ret != -ENOENT) {
1441 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1442 init_filename, ret);
1443 }
1444
1445 return ret;
1446 }
1447
1448 static noinline void __init kernel_init_freeable(void);
1449
1450 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1451 bool rodata_enabled __ro_after_init = true;
set_debug_rodata(char * str)1452 static int __init set_debug_rodata(char *str)
1453 {
1454 if (strtobool(str, &rodata_enabled))
1455 pr_warn("Invalid option string for rodata: '%s'\n", str);
1456 return 1;
1457 }
1458 __setup("rodata=", set_debug_rodata);
1459 #endif
1460
1461 #ifdef CONFIG_STRICT_KERNEL_RWX
mark_readonly(void)1462 static void mark_readonly(void)
1463 {
1464 if (rodata_enabled) {
1465 /*
1466 * load_module() results in W+X mappings, which are cleaned
1467 * up with call_rcu(). Let's make sure that queued work is
1468 * flushed so that we don't hit false positives looking for
1469 * insecure pages which are W+X.
1470 */
1471 rcu_barrier();
1472 mark_rodata_ro();
1473 rodata_test();
1474 } else
1475 pr_info("Kernel memory protection disabled.\n");
1476 }
1477 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
mark_readonly(void)1478 static inline void mark_readonly(void)
1479 {
1480 pr_warn("Kernel memory protection not selected by kernel config.\n");
1481 }
1482 #else
mark_readonly(void)1483 static inline void mark_readonly(void)
1484 {
1485 pr_warn("This architecture does not have kernel memory protection.\n");
1486 }
1487 #endif
1488
free_initmem(void)1489 void __weak free_initmem(void)
1490 {
1491 free_initmem_default(POISON_FREE_INITMEM);
1492 }
1493
kernel_init(void * unused)1494 static int __ref kernel_init(void *unused)
1495 {
1496 int ret;
1497
1498 /*
1499 * Wait until kthreadd is all set-up.
1500 */
1501 wait_for_completion(&kthreadd_done);
1502
1503 kernel_init_freeable();
1504 /* need to finish all async __init code before freeing the memory */
1505 async_synchronize_full();
1506 kprobe_free_init_mem();
1507 ftrace_free_init_mem();
1508 kgdb_free_init_mem();
1509 exit_boot_config();
1510 free_initmem();
1511 mark_readonly();
1512
1513 /*
1514 * Kernel mappings are now finalized - update the userspace page-table
1515 * to finalize PTI.
1516 */
1517 pti_finalize();
1518
1519 system_state = SYSTEM_RUNNING;
1520 numa_default_policy();
1521
1522 rcu_end_inkernel_boot();
1523
1524 do_sysctl_args();
1525
1526 if (ramdisk_execute_command) {
1527 ret = run_init_process(ramdisk_execute_command);
1528 if (!ret)
1529 return 0;
1530 pr_err("Failed to execute %s (error %d)\n",
1531 ramdisk_execute_command, ret);
1532 }
1533
1534 /*
1535 * We try each of these until one succeeds.
1536 *
1537 * The Bourne shell can be used instead of init if we are
1538 * trying to recover a really broken machine.
1539 */
1540 if (execute_command) {
1541 ret = run_init_process(execute_command);
1542 if (!ret)
1543 return 0;
1544 panic("Requested init %s failed (error %d).",
1545 execute_command, ret);
1546 }
1547
1548 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1549 ret = run_init_process(CONFIG_DEFAULT_INIT);
1550 if (ret)
1551 pr_err("Default init %s failed (error %d)\n",
1552 CONFIG_DEFAULT_INIT, ret);
1553 else
1554 return 0;
1555 }
1556
1557 if (!try_to_run_init_process("/sbin/init") ||
1558 !try_to_run_init_process("/etc/init") ||
1559 !try_to_run_init_process("/bin/init") ||
1560 !try_to_run_init_process("/bin/sh"))
1561 return 0;
1562
1563 panic("No working init found. Try passing init= option to kernel. "
1564 "See Linux Documentation/admin-guide/init.rst for guidance.");
1565 }
1566
1567 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
console_on_rootfs(void)1568 void __init console_on_rootfs(void)
1569 {
1570 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1571
1572 if (IS_ERR(file)) {
1573 pr_err("Warning: unable to open an initial console.\n");
1574 return;
1575 }
1576 init_dup(file);
1577 init_dup(file);
1578 init_dup(file);
1579 fput(file);
1580 }
1581
kernel_init_freeable(void)1582 static noinline void __init kernel_init_freeable(void)
1583 {
1584 /* Now the scheduler is fully set up and can do blocking allocations */
1585 gfp_allowed_mask = __GFP_BITS_MASK;
1586
1587 /*
1588 * init can allocate pages on any node
1589 */
1590 set_mems_allowed(node_states[N_MEMORY]);
1591
1592 cad_pid = get_pid(task_pid(current));
1593
1594 smp_prepare_cpus(setup_max_cpus);
1595
1596 workqueue_init();
1597
1598 init_mm_internals();
1599
1600 rcu_init_tasks_generic();
1601 do_pre_smp_initcalls();
1602 lockup_detector_init();
1603
1604 smp_init();
1605 sched_init_smp();
1606
1607 padata_init();
1608 page_alloc_init_late();
1609 /* Initialize page ext after all struct pages are initialized. */
1610 page_ext_init();
1611
1612 do_basic_setup();
1613
1614 kunit_run_all_tests();
1615
1616 wait_for_initramfs();
1617 console_on_rootfs();
1618
1619 /*
1620 * check if there is an early userspace init. If yes, let it do all
1621 * the work
1622 */
1623 if (init_eaccess(ramdisk_execute_command) != 0) {
1624 ramdisk_execute_command = NULL;
1625 prepare_namespace();
1626 }
1627
1628 /*
1629 * Ok, we have completed the initial bootup, and
1630 * we're essentially up and running. Get rid of the
1631 * initmem segments and start the user-mode stuff..
1632 *
1633 * rootfs is available now, try loading the public keys
1634 * and default modules
1635 */
1636
1637 integrity_load_keys();
1638 }
1639