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/kmsan.h>
38 #include <linux/vmalloc.h>
39 #include <linux/kernel_stat.h>
40 #include <linux/start_kernel.h>
41 #include <linux/security.h>
42 #include <linux/smp.h>
43 #include <linux/profile.h>
44 #include <linux/kfence.h>
45 #include <linux/rcupdate.h>
46 #include <linux/srcu.h>
47 #include <linux/moduleparam.h>
48 #include <linux/kallsyms.h>
49 #include <linux/buildid.h>
50 #include <linux/writeback.h>
51 #include <linux/cpu.h>
52 #include <linux/cpuset.h>
53 #include <linux/cgroup.h>
54 #include <linux/efi.h>
55 #include <linux/tick.h>
56 #include <linux/sched/isolation.h>
57 #include <linux/interrupt.h>
58 #include <linux/taskstats_kern.h>
59 #include <linux/delayacct.h>
60 #include <linux/unistd.h>
61 #include <linux/utsname.h>
62 #include <linux/rmap.h>
63 #include <linux/mempolicy.h>
64 #include <linux/key.h>
65 #include <linux/page_ext.h>
66 #include <linux/debug_locks.h>
67 #include <linux/debugobjects.h>
68 #include <linux/lockdep.h>
69 #include <linux/kmemleak.h>
70 #include <linux/padata.h>
71 #include <linux/pid_namespace.h>
72 #include <linux/device/driver.h>
73 #include <linux/kthread.h>
74 #include <linux/sched.h>
75 #include <linux/sched/init.h>
76 #include <linux/signal.h>
77 #include <linux/idr.h>
78 #include <linux/kgdb.h>
79 #include <linux/ftrace.h>
80 #include <linux/async.h>
81 #include <linux/shmem_fs.h>
82 #include <linux/slab.h>
83 #include <linux/perf_event.h>
84 #include <linux/ptrace.h>
85 #include <linux/pti.h>
86 #include <linux/blkdev.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 static int kernel_init(void *);
114
115 extern void init_IRQ(void);
116 extern void radix_tree_init(void);
117 extern void maple_tree_init(void);
118
119 /*
120 * Debug helper: via this flag we know that we are in 'early bootup code'
121 * where only the boot processor is running with IRQ disabled. This means
122 * two things - IRQ must not be enabled before the flag is cleared and some
123 * operations which are not allowed with IRQ disabled are allowed while the
124 * flag is set.
125 */
126 bool early_boot_irqs_disabled __read_mostly;
127
128 enum system_states system_state __read_mostly;
129 EXPORT_SYMBOL(system_state);
130
131 /*
132 * Boot command-line arguments
133 */
134 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
135 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
136
137 extern void time_init(void);
138 /* Default late time init is NULL. archs can override this later. */
139 void (*__initdata late_time_init)(void);
140
141 /* Untouched command line saved by arch-specific code. */
142 char __initdata boot_command_line[COMMAND_LINE_SIZE];
143 /* Untouched saved command line (eg. for /proc) */
144 char *saved_command_line;
145 /* Command line for parameter parsing */
146 static char *static_command_line;
147 /* Untouched extra command line */
148 static char *extra_command_line;
149 /* Extra init arguments */
150 static char *extra_init_args;
151
152 #ifdef CONFIG_BOOT_CONFIG
153 /* Is bootconfig on command line? */
154 static bool bootconfig_found;
155 static size_t initargs_offs;
156 #else
157 # define bootconfig_found false
158 # define initargs_offs 0
159 #endif
160
161 static char *execute_command;
162 static char *ramdisk_execute_command = "/init";
163
164 /*
165 * Used to generate warnings if static_key manipulation functions are used
166 * before jump_label_init is called.
167 */
168 bool static_key_initialized __read_mostly;
169 EXPORT_SYMBOL_GPL(static_key_initialized);
170
171 /*
172 * If set, this is an indication to the drivers that reset the underlying
173 * device before going ahead with the initialization otherwise driver might
174 * rely on the BIOS and skip the reset operation.
175 *
176 * This is useful if kernel is booting in an unreliable environment.
177 * For ex. kdump situation where previous kernel has crashed, BIOS has been
178 * skipped and devices will be in unknown state.
179 */
180 unsigned int reset_devices;
181 EXPORT_SYMBOL(reset_devices);
182
set_reset_devices(char * str)183 static int __init set_reset_devices(char *str)
184 {
185 reset_devices = 1;
186 return 1;
187 }
188
189 __setup("reset_devices", set_reset_devices);
190
191 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
192 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
193 static const char *panic_later, *panic_param;
194
195 extern const struct obs_kernel_param __setup_start[], __setup_end[];
196
obsolete_checksetup(char * line)197 static bool __init obsolete_checksetup(char *line)
198 {
199 const struct obs_kernel_param *p;
200 bool had_early_param = false;
201
202 p = __setup_start;
203 do {
204 int n = strlen(p->str);
205 if (parameqn(line, p->str, n)) {
206 if (p->early) {
207 /* Already done in parse_early_param?
208 * (Needs exact match on param part).
209 * Keep iterating, as we can have early
210 * params and __setups of same names 8( */
211 if (line[n] == '\0' || line[n] == '=')
212 had_early_param = true;
213 } else if (!p->setup_func) {
214 pr_warn("Parameter %s is obsolete, ignored\n",
215 p->str);
216 return true;
217 } else if (p->setup_func(line + n))
218 return true;
219 }
220 p++;
221 } while (p < __setup_end);
222
223 return had_early_param;
224 }
225
226 /*
227 * This should be approx 2 Bo*oMips to start (note initial shift), and will
228 * still work even if initially too large, it will just take slightly longer
229 */
230 unsigned long loops_per_jiffy = (1<<12);
231 EXPORT_SYMBOL(loops_per_jiffy);
232
debug_kernel(char * str)233 static int __init debug_kernel(char *str)
234 {
235 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
236 return 0;
237 }
238
quiet_kernel(char * str)239 static int __init quiet_kernel(char *str)
240 {
241 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
242 return 0;
243 }
244
245 early_param("debug", debug_kernel);
246 early_param("quiet", quiet_kernel);
247
loglevel(char * str)248 static int __init loglevel(char *str)
249 {
250 int newlevel;
251
252 /*
253 * Only update loglevel value when a correct setting was passed,
254 * to prevent blind crashes (when loglevel being set to 0) that
255 * are quite hard to debug
256 */
257 if (get_option(&str, &newlevel)) {
258 console_loglevel = newlevel;
259 return 0;
260 }
261
262 return -EINVAL;
263 }
264
265 early_param("loglevel", loglevel);
266
267 #ifdef CONFIG_BLK_DEV_INITRD
get_boot_config_from_initrd(size_t * _size)268 static void * __init get_boot_config_from_initrd(size_t *_size)
269 {
270 u32 size, csum;
271 char *data;
272 u32 *hdr;
273 int i;
274
275 if (!initrd_end)
276 return NULL;
277
278 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
279 /*
280 * Since Grub may align the size of initrd to 4, we must
281 * check the preceding 3 bytes as well.
282 */
283 for (i = 0; i < 4; i++) {
284 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
285 goto found;
286 data--;
287 }
288 return NULL;
289
290 found:
291 hdr = (u32 *)(data - 8);
292 size = le32_to_cpu(hdr[0]);
293 csum = le32_to_cpu(hdr[1]);
294
295 data = ((void *)hdr) - size;
296 if ((unsigned long)data < initrd_start) {
297 pr_err("bootconfig size %d is greater than initrd size %ld\n",
298 size, initrd_end - initrd_start);
299 return NULL;
300 }
301
302 if (xbc_calc_checksum(data, size) != csum) {
303 pr_err("bootconfig checksum failed\n");
304 return NULL;
305 }
306
307 /* Remove bootconfig from initramfs/initrd */
308 initrd_end = (unsigned long)data;
309 if (_size)
310 *_size = size;
311
312 return data;
313 }
314 #else
get_boot_config_from_initrd(size_t * _size)315 static void * __init get_boot_config_from_initrd(size_t *_size)
316 {
317 return NULL;
318 }
319 #endif
320
321 #ifdef CONFIG_BOOT_CONFIG
322
323 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
324
325 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
326
xbc_snprint_cmdline(char * buf,size_t size,struct xbc_node * root)327 static int __init xbc_snprint_cmdline(char *buf, size_t size,
328 struct xbc_node *root)
329 {
330 struct xbc_node *knode, *vnode;
331 char *end = buf + size;
332 const char *val;
333 int ret;
334
335 xbc_node_for_each_key_value(root, knode, val) {
336 ret = xbc_node_compose_key_after(root, knode,
337 xbc_namebuf, XBC_KEYLEN_MAX);
338 if (ret < 0)
339 return ret;
340
341 vnode = xbc_node_get_child(knode);
342 if (!vnode) {
343 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
344 if (ret < 0)
345 return ret;
346 buf += ret;
347 continue;
348 }
349 xbc_array_for_each_value(vnode, val) {
350 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
351 xbc_namebuf, val);
352 if (ret < 0)
353 return ret;
354 buf += ret;
355 }
356 }
357
358 return buf - (end - size);
359 }
360 #undef rest
361
362 /* Make an extra command line under given key word */
xbc_make_cmdline(const char * key)363 static char * __init xbc_make_cmdline(const char *key)
364 {
365 struct xbc_node *root;
366 char *new_cmdline;
367 int ret, len = 0;
368
369 root = xbc_find_node(key);
370 if (!root)
371 return NULL;
372
373 /* Count required buffer size */
374 len = xbc_snprint_cmdline(NULL, 0, root);
375 if (len <= 0)
376 return NULL;
377
378 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
379 if (!new_cmdline) {
380 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
381 return NULL;
382 }
383
384 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
385 if (ret < 0 || ret > len) {
386 pr_err("Failed to print extra kernel cmdline.\n");
387 memblock_free(new_cmdline, len + 1);
388 return NULL;
389 }
390
391 return new_cmdline;
392 }
393
bootconfig_params(char * param,char * val,const char * unused,void * arg)394 static int __init bootconfig_params(char *param, char *val,
395 const char *unused, void *arg)
396 {
397 if (strcmp(param, "bootconfig") == 0) {
398 bootconfig_found = true;
399 }
400 return 0;
401 }
402
warn_bootconfig(char * str)403 static int __init warn_bootconfig(char *str)
404 {
405 /* The 'bootconfig' has been handled by bootconfig_params(). */
406 return 0;
407 }
408
setup_boot_config(void)409 static void __init setup_boot_config(void)
410 {
411 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
412 const char *msg, *data;
413 int pos, ret;
414 size_t size;
415 char *err;
416
417 /* Cut out the bootconfig data even if we have no bootconfig option */
418 data = get_boot_config_from_initrd(&size);
419 /* If there is no bootconfig in initrd, try embedded one. */
420 if (!data)
421 data = xbc_get_embedded_bootconfig(&size);
422
423 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
424 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
425 bootconfig_params);
426
427 if (IS_ERR(err) || !bootconfig_found)
428 return;
429
430 /* parse_args() stops at the next param of '--' and returns an address */
431 if (err)
432 initargs_offs = err - tmp_cmdline;
433
434 if (!data) {
435 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
436 return;
437 }
438
439 if (size >= XBC_DATA_MAX) {
440 pr_err("bootconfig size %ld greater than max size %d\n",
441 (long)size, XBC_DATA_MAX);
442 return;
443 }
444
445 ret = xbc_init(data, size, &msg, &pos);
446 if (ret < 0) {
447 if (pos < 0)
448 pr_err("Failed to init bootconfig: %s.\n", msg);
449 else
450 pr_err("Failed to parse bootconfig: %s at %d.\n",
451 msg, pos);
452 } else {
453 xbc_get_info(&ret, NULL);
454 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
455 /* keys starting with "kernel." are passed via cmdline */
456 extra_command_line = xbc_make_cmdline("kernel");
457 /* Also, "init." keys are init arguments */
458 extra_init_args = xbc_make_cmdline("init");
459 }
460 return;
461 }
462
exit_boot_config(void)463 static void __init exit_boot_config(void)
464 {
465 xbc_exit();
466 }
467
468 #else /* !CONFIG_BOOT_CONFIG */
469
setup_boot_config(void)470 static void __init setup_boot_config(void)
471 {
472 /* Remove bootconfig data from initrd */
473 get_boot_config_from_initrd(NULL);
474 }
475
warn_bootconfig(char * str)476 static int __init warn_bootconfig(char *str)
477 {
478 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
479 return 0;
480 }
481
482 #define exit_boot_config() do {} while (0)
483
484 #endif /* CONFIG_BOOT_CONFIG */
485
486 early_param("bootconfig", warn_bootconfig);
487
488 /* Change NUL term back to "=", to make "param" the whole string. */
repair_env_string(char * param,char * val)489 static void __init repair_env_string(char *param, char *val)
490 {
491 if (val) {
492 /* param=val or param="val"? */
493 if (val == param+strlen(param)+1)
494 val[-1] = '=';
495 else if (val == param+strlen(param)+2) {
496 val[-2] = '=';
497 memmove(val-1, val, strlen(val)+1);
498 } else
499 BUG();
500 }
501 }
502
503 /* Anything after -- gets handed straight to init. */
set_init_arg(char * param,char * val,const char * unused,void * arg)504 static int __init set_init_arg(char *param, char *val,
505 const char *unused, void *arg)
506 {
507 unsigned int i;
508
509 if (panic_later)
510 return 0;
511
512 repair_env_string(param, val);
513
514 for (i = 0; argv_init[i]; i++) {
515 if (i == MAX_INIT_ARGS) {
516 panic_later = "init";
517 panic_param = param;
518 return 0;
519 }
520 }
521 argv_init[i] = param;
522 return 0;
523 }
524
525 /*
526 * Unknown boot options get handed to init, unless they look like
527 * unused parameters (modprobe will find them in /proc/cmdline).
528 */
unknown_bootoption(char * param,char * val,const char * unused,void * arg)529 static int __init unknown_bootoption(char *param, char *val,
530 const char *unused, void *arg)
531 {
532 size_t len = strlen(param);
533
534 /* Handle params aliased to sysctls */
535 if (sysctl_is_alias(param))
536 return 0;
537
538 repair_env_string(param, val);
539
540 /* Handle obsolete-style parameters */
541 if (obsolete_checksetup(param))
542 return 0;
543
544 /* Unused module parameter. */
545 if (strnchr(param, len, '.'))
546 return 0;
547
548 if (panic_later)
549 return 0;
550
551 if (val) {
552 /* Environment option */
553 unsigned int i;
554 for (i = 0; envp_init[i]; i++) {
555 if (i == MAX_INIT_ENVS) {
556 panic_later = "env";
557 panic_param = param;
558 }
559 if (!strncmp(param, envp_init[i], len+1))
560 break;
561 }
562 envp_init[i] = param;
563 } else {
564 /* Command line option */
565 unsigned int i;
566 for (i = 0; argv_init[i]; i++) {
567 if (i == MAX_INIT_ARGS) {
568 panic_later = "init";
569 panic_param = param;
570 }
571 }
572 argv_init[i] = param;
573 }
574 return 0;
575 }
576
init_setup(char * str)577 static int __init init_setup(char *str)
578 {
579 unsigned int i;
580
581 execute_command = str;
582 /*
583 * In case LILO is going to boot us with default command line,
584 * it prepends "auto" before the whole cmdline which makes
585 * the shell think it should execute a script with such name.
586 * So we ignore all arguments entered _before_ init=... [MJ]
587 */
588 for (i = 1; i < MAX_INIT_ARGS; i++)
589 argv_init[i] = NULL;
590 return 1;
591 }
592 __setup("init=", init_setup);
593
rdinit_setup(char * str)594 static int __init rdinit_setup(char *str)
595 {
596 unsigned int i;
597
598 ramdisk_execute_command = str;
599 /* See "auto" comment in init_setup */
600 for (i = 1; i < MAX_INIT_ARGS; i++)
601 argv_init[i] = NULL;
602 return 1;
603 }
604 __setup("rdinit=", rdinit_setup);
605
606 #ifndef CONFIG_SMP
607 static const unsigned int setup_max_cpus = NR_CPUS;
setup_nr_cpu_ids(void)608 static inline void setup_nr_cpu_ids(void) { }
smp_prepare_cpus(unsigned int maxcpus)609 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
610 #endif
611
612 /*
613 * We need to store the untouched command line for future reference.
614 * We also need to store the touched command line since the parameter
615 * parsing is performed in place, and we should allow a component to
616 * store reference of name/value for future reference.
617 */
setup_command_line(char * command_line)618 static void __init setup_command_line(char *command_line)
619 {
620 size_t len, xlen = 0, ilen = 0;
621
622 if (extra_command_line)
623 xlen = strlen(extra_command_line);
624 if (extra_init_args)
625 ilen = strlen(extra_init_args) + 4; /* for " -- " */
626
627 len = xlen + strlen(boot_command_line) + 1;
628
629 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
630 if (!saved_command_line)
631 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
632
633 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
634 if (!static_command_line)
635 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
636
637 if (xlen) {
638 /*
639 * We have to put extra_command_line before boot command
640 * lines because there could be dashes (separator of init
641 * command line) in the command lines.
642 */
643 strcpy(saved_command_line, extra_command_line);
644 strcpy(static_command_line, extra_command_line);
645 }
646 strcpy(saved_command_line + xlen, boot_command_line);
647 strcpy(static_command_line + xlen, command_line);
648
649 if (ilen) {
650 /*
651 * Append supplemental init boot args to saved_command_line
652 * so that user can check what command line options passed
653 * to init.
654 * The order should always be
655 * " -- "[bootconfig init-param][cmdline init-param]
656 */
657 if (initargs_offs) {
658 len = xlen + initargs_offs;
659 strcpy(saved_command_line + len, extra_init_args);
660 len += ilen - 4; /* strlen(extra_init_args) */
661 strcpy(saved_command_line + len,
662 boot_command_line + initargs_offs - 1);
663 } else {
664 len = strlen(saved_command_line);
665 strcpy(saved_command_line + len, " -- ");
666 len += 4;
667 strcpy(saved_command_line + len, extra_init_args);
668 }
669 }
670 }
671
672 /*
673 * We need to finalize in a non-__init function or else race conditions
674 * between the root thread and the init thread may cause start_kernel to
675 * be reaped by free_initmem before the root thread has proceeded to
676 * cpu_idle.
677 *
678 * gcc-3.4 accidentally inlines this function, so use noinline.
679 */
680
681 static __initdata DECLARE_COMPLETION(kthreadd_done);
682
rest_init(void)683 noinline void __ref rest_init(void)
684 {
685 struct task_struct *tsk;
686 int pid;
687
688 rcu_scheduler_starting();
689 /*
690 * We need to spawn init first so that it obtains pid 1, however
691 * the init task will end up wanting to create kthreads, which, if
692 * we schedule it before we create kthreadd, will OOPS.
693 */
694 pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
695 /*
696 * Pin init on the boot CPU. Task migration is not properly working
697 * until sched_init_smp() has been run. It will set the allowed
698 * CPUs for init to the non isolated CPUs.
699 */
700 rcu_read_lock();
701 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
702 tsk->flags |= PF_NO_SETAFFINITY;
703 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
704 rcu_read_unlock();
705
706 numa_default_policy();
707 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
708 rcu_read_lock();
709 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
710 rcu_read_unlock();
711
712 /*
713 * Enable might_sleep() and smp_processor_id() checks.
714 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
715 * kernel_thread() would trigger might_sleep() splats. With
716 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
717 * already, but it's stuck on the kthreadd_done completion.
718 */
719 system_state = SYSTEM_SCHEDULING;
720
721 complete(&kthreadd_done);
722
723 /*
724 * The boot idle thread must execute schedule()
725 * at least once to get things moving:
726 */
727 schedule_preempt_disabled();
728 /* Call into cpu_idle with preempt disabled */
729 cpu_startup_entry(CPUHP_ONLINE);
730 }
731
732 /* Check for early params. */
do_early_param(char * param,char * val,const char * unused,void * arg)733 static int __init do_early_param(char *param, char *val,
734 const char *unused, void *arg)
735 {
736 const struct obs_kernel_param *p;
737
738 for (p = __setup_start; p < __setup_end; p++) {
739 if ((p->early && parameq(param, p->str)) ||
740 (strcmp(param, "console") == 0 &&
741 strcmp(p->str, "earlycon") == 0)
742 ) {
743 if (p->setup_func(val) != 0)
744 pr_warn("Malformed early option '%s'\n", param);
745 }
746 }
747 /* We accept everything at this stage. */
748 return 0;
749 }
750
parse_early_options(char * cmdline)751 void __init parse_early_options(char *cmdline)
752 {
753 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
754 do_early_param);
755 }
756
757 /* Arch code calls this early on, or if not, just before other parsing. */
parse_early_param(void)758 void __init parse_early_param(void)
759 {
760 static int done __initdata;
761 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
762
763 if (done)
764 return;
765
766 /* All fall through to do_early_param. */
767 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
768 parse_early_options(tmp_cmdline);
769 done = 1;
770 }
771
arch_post_acpi_subsys_init(void)772 void __init __weak arch_post_acpi_subsys_init(void) { }
773
smp_setup_processor_id(void)774 void __init __weak smp_setup_processor_id(void)
775 {
776 }
777
778 # if THREAD_SIZE >= PAGE_SIZE
thread_stack_cache_init(void)779 void __init __weak thread_stack_cache_init(void)
780 {
781 }
782 #endif
783
poking_init(void)784 void __init __weak poking_init(void) { }
785
pgtable_cache_init(void)786 void __init __weak pgtable_cache_init(void) { }
787
trap_init(void)788 void __init __weak trap_init(void) { }
789
790 bool initcall_debug;
791 core_param(initcall_debug, initcall_debug, bool, 0644);
792
793 #ifdef TRACEPOINTS_ENABLED
794 static void __init initcall_debug_enable(void);
795 #else
initcall_debug_enable(void)796 static inline void initcall_debug_enable(void)
797 {
798 }
799 #endif
800
801 /* Report memory auto-initialization states for this boot. */
report_meminit(void)802 static void __init report_meminit(void)
803 {
804 const char *stack;
805
806 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
807 stack = "all(pattern)";
808 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
809 stack = "all(zero)";
810 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
811 stack = "byref_all(zero)";
812 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
813 stack = "byref(zero)";
814 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
815 stack = "__user(zero)";
816 else
817 stack = "off";
818
819 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
820 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
821 want_init_on_free() ? "on" : "off");
822 if (want_init_on_free())
823 pr_info("mem auto-init: clearing system memory may take some time...\n");
824 }
825
826 /*
827 * Set up kernel memory allocators
828 */
mm_init(void)829 static void __init mm_init(void)
830 {
831 /*
832 * page_ext requires contiguous pages,
833 * bigger than MAX_ORDER unless SPARSEMEM.
834 */
835 page_ext_init_flatmem();
836 init_mem_debugging_and_hardening();
837 kfence_alloc_pool();
838 report_meminit();
839 kmsan_init_shadow();
840 stack_depot_early_init();
841 mem_init();
842 mem_init_print_info();
843 kmem_cache_init();
844 /*
845 * page_owner must be initialized after buddy is ready, and also after
846 * slab is ready so that stack_depot_init() works properly
847 */
848 page_ext_init_flatmem_late();
849 kmemleak_init();
850 pgtable_init();
851 debug_objects_mem_init();
852 vmalloc_init();
853 /* Should be run after vmap initialization */
854 if (early_page_ext_enabled())
855 page_ext_init();
856 /* Should be run before the first non-init thread is created */
857 init_espfix_bsp();
858 /* Should be run after espfix64 is set up. */
859 pti_init();
860 kmsan_init_runtime();
861 mm_cache_init();
862 }
863
864 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
865 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
866 randomize_kstack_offset);
867 DEFINE_PER_CPU(u32, kstack_offset);
868
early_randomize_kstack_offset(char * buf)869 static int __init early_randomize_kstack_offset(char *buf)
870 {
871 int ret;
872 bool bool_result;
873
874 ret = kstrtobool(buf, &bool_result);
875 if (ret)
876 return ret;
877
878 if (bool_result)
879 static_branch_enable(&randomize_kstack_offset);
880 else
881 static_branch_disable(&randomize_kstack_offset);
882 return 0;
883 }
884 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
885 #endif
886
arch_call_rest_init(void)887 void __init __weak arch_call_rest_init(void)
888 {
889 rest_init();
890 }
891
print_unknown_bootoptions(void)892 static void __init print_unknown_bootoptions(void)
893 {
894 char *unknown_options;
895 char *end;
896 const char *const *p;
897 size_t len;
898
899 if (panic_later || (!argv_init[1] && !envp_init[2]))
900 return;
901
902 /*
903 * Determine how many options we have to print out, plus a space
904 * before each
905 */
906 len = 1; /* null terminator */
907 for (p = &argv_init[1]; *p; p++) {
908 len++;
909 len += strlen(*p);
910 }
911 for (p = &envp_init[2]; *p; p++) {
912 len++;
913 len += strlen(*p);
914 }
915
916 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
917 if (!unknown_options) {
918 pr_err("%s: Failed to allocate %zu bytes\n",
919 __func__, len);
920 return;
921 }
922 end = unknown_options;
923
924 for (p = &argv_init[1]; *p; p++)
925 end += sprintf(end, " %s", *p);
926 for (p = &envp_init[2]; *p; p++)
927 end += sprintf(end, " %s", *p);
928
929 /* Start at unknown_options[1] to skip the initial space */
930 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
931 &unknown_options[1]);
932 memblock_free(unknown_options, len);
933 }
934
start_kernel(void)935 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
936 {
937 char *command_line;
938 char *after_dashes;
939
940 set_task_stack_end_magic(&init_task);
941 smp_setup_processor_id();
942 debug_objects_early_init();
943 init_vmlinux_build_id();
944
945 cgroup_init_early();
946
947 local_irq_disable();
948 early_boot_irqs_disabled = true;
949
950 /*
951 * Interrupts are still disabled. Do necessary setups, then
952 * enable them.
953 */
954 boot_cpu_init();
955 page_address_init();
956 pr_notice("%s", linux_banner);
957 early_security_init();
958 setup_arch(&command_line);
959 setup_boot_config();
960 setup_command_line(command_line);
961 setup_nr_cpu_ids();
962 setup_per_cpu_areas();
963 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
964 boot_cpu_hotplug_init();
965
966 build_all_zonelists(NULL);
967 page_alloc_init();
968
969 pr_notice("Kernel command line: %s\n", saved_command_line);
970 /* parameters may set static keys */
971 jump_label_init();
972 parse_early_param();
973 after_dashes = parse_args("Booting kernel",
974 static_command_line, __start___param,
975 __stop___param - __start___param,
976 -1, -1, NULL, &unknown_bootoption);
977 print_unknown_bootoptions();
978 if (!IS_ERR_OR_NULL(after_dashes))
979 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
980 NULL, set_init_arg);
981 if (extra_init_args)
982 parse_args("Setting extra init args", extra_init_args,
983 NULL, 0, -1, -1, NULL, set_init_arg);
984
985 /* Architectural and non-timekeeping rng init, before allocator init */
986 random_init_early(command_line);
987
988 /*
989 * These use large bootmem allocations and must precede
990 * kmem_cache_init()
991 */
992 setup_log_buf(0);
993 vfs_caches_init_early();
994 sort_main_extable();
995 trap_init();
996 mm_init();
997 poking_init();
998 ftrace_init();
999
1000 /* trace_printk can be enabled here */
1001 early_trace_init();
1002
1003 /*
1004 * Set up the scheduler prior starting any interrupts (such as the
1005 * timer interrupt). Full topology setup happens at smp_init()
1006 * time - but meanwhile we still have a functioning scheduler.
1007 */
1008 sched_init();
1009
1010 if (WARN(!irqs_disabled(),
1011 "Interrupts were enabled *very* early, fixing it\n"))
1012 local_irq_disable();
1013 radix_tree_init();
1014 maple_tree_init();
1015
1016 /*
1017 * Set up housekeeping before setting up workqueues to allow the unbound
1018 * workqueue to take non-housekeeping into account.
1019 */
1020 housekeeping_init();
1021
1022 /*
1023 * Allow workqueue creation and work item queueing/cancelling
1024 * early. Work item execution depends on kthreads and starts after
1025 * workqueue_init().
1026 */
1027 workqueue_init_early();
1028
1029 rcu_init();
1030
1031 /* Trace events are available after this */
1032 trace_init();
1033
1034 if (initcall_debug)
1035 initcall_debug_enable();
1036
1037 context_tracking_init();
1038 /* init some links before init_ISA_irqs() */
1039 early_irq_init();
1040 init_IRQ();
1041 tick_init();
1042 rcu_init_nohz();
1043 init_timers();
1044 srcu_init();
1045 hrtimers_init();
1046 softirq_init();
1047 timekeeping_init();
1048 time_init();
1049
1050 /* This must be after timekeeping is initialized */
1051 random_init();
1052
1053 /* These make use of the fully initialized rng */
1054 kfence_init();
1055 boot_init_stack_canary();
1056
1057 perf_event_init();
1058 profile_init();
1059 call_function_init();
1060 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1061
1062 early_boot_irqs_disabled = false;
1063 local_irq_enable();
1064
1065 kmem_cache_init_late();
1066
1067 /*
1068 * HACK ALERT! This is early. We're enabling the console before
1069 * we've done PCI setups etc, and console_init() must be aware of
1070 * this. But we do want output early, in case something goes wrong.
1071 */
1072 console_init();
1073 if (panic_later)
1074 panic("Too many boot %s vars at `%s'", panic_later,
1075 panic_param);
1076
1077 lockdep_init();
1078
1079 /*
1080 * Need to run this when irqs are enabled, because it wants
1081 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1082 * too:
1083 */
1084 locking_selftest();
1085
1086 #ifdef CONFIG_BLK_DEV_INITRD
1087 if (initrd_start && !initrd_below_start_ok &&
1088 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1089 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1090 page_to_pfn(virt_to_page((void *)initrd_start)),
1091 min_low_pfn);
1092 initrd_start = 0;
1093 }
1094 #endif
1095 setup_per_cpu_pageset();
1096 numa_policy_init();
1097 acpi_early_init();
1098 if (late_time_init)
1099 late_time_init();
1100 sched_clock_init();
1101 calibrate_delay();
1102
1103 arch_cpu_finalize_init();
1104
1105 pid_idr_init();
1106 anon_vma_init();
1107 #ifdef CONFIG_X86
1108 if (efi_enabled(EFI_RUNTIME_SERVICES))
1109 efi_enter_virtual_mode();
1110 #endif
1111 thread_stack_cache_init();
1112 cred_init();
1113 fork_init();
1114 proc_caches_init();
1115 uts_ns_init();
1116 key_init();
1117 security_init();
1118 dbg_late_init();
1119 net_ns_init();
1120 vfs_caches_init();
1121 pagecache_init();
1122 signals_init();
1123 seq_file_init();
1124 proc_root_init();
1125 nsfs_init();
1126 cpuset_init();
1127 cgroup_init();
1128 taskstats_init_early();
1129 delayacct_init();
1130
1131 acpi_subsystem_init();
1132 arch_post_acpi_subsys_init();
1133 kcsan_init();
1134
1135 /* Do the rest non-__init'ed, we're now alive */
1136 arch_call_rest_init();
1137
1138 prevent_tail_call_optimization();
1139 }
1140
1141 /* Call all constructor functions linked into the kernel. */
do_ctors(void)1142 static void __init do_ctors(void)
1143 {
1144 /*
1145 * For UML, the constructors have already been called by the
1146 * normal setup code as it's just a normal ELF binary, so we
1147 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1148 * even on UML for modules.
1149 */
1150 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1151 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1152
1153 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1154 (*fn)();
1155 #endif
1156 }
1157
1158 #ifdef CONFIG_KALLSYMS
1159 struct blacklist_entry {
1160 struct list_head next;
1161 char *buf;
1162 };
1163
1164 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1165
initcall_blacklist(char * str)1166 static int __init initcall_blacklist(char *str)
1167 {
1168 char *str_entry;
1169 struct blacklist_entry *entry;
1170
1171 /* str argument is a comma-separated list of functions */
1172 do {
1173 str_entry = strsep(&str, ",");
1174 if (str_entry) {
1175 pr_debug("blacklisting initcall %s\n", str_entry);
1176 entry = memblock_alloc(sizeof(*entry),
1177 SMP_CACHE_BYTES);
1178 if (!entry)
1179 panic("%s: Failed to allocate %zu bytes\n",
1180 __func__, sizeof(*entry));
1181 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1182 SMP_CACHE_BYTES);
1183 if (!entry->buf)
1184 panic("%s: Failed to allocate %zu bytes\n",
1185 __func__, strlen(str_entry) + 1);
1186 strcpy(entry->buf, str_entry);
1187 list_add(&entry->next, &blacklisted_initcalls);
1188 }
1189 } while (str_entry);
1190
1191 return 1;
1192 }
1193
initcall_blacklisted(initcall_t fn)1194 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1195 {
1196 struct blacklist_entry *entry;
1197 char fn_name[KSYM_SYMBOL_LEN];
1198 unsigned long addr;
1199
1200 if (list_empty(&blacklisted_initcalls))
1201 return false;
1202
1203 addr = (unsigned long) dereference_function_descriptor(fn);
1204 sprint_symbol_no_offset(fn_name, addr);
1205
1206 /*
1207 * fn will be "function_name [module_name]" where [module_name] is not
1208 * displayed for built-in init functions. Strip off the [module_name].
1209 */
1210 strreplace(fn_name, ' ', '\0');
1211
1212 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1213 if (!strcmp(fn_name, entry->buf)) {
1214 pr_debug("initcall %s blacklisted\n", fn_name);
1215 return true;
1216 }
1217 }
1218
1219 return false;
1220 }
1221 #else
initcall_blacklist(char * str)1222 static int __init initcall_blacklist(char *str)
1223 {
1224 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1225 return 0;
1226 }
1227
initcall_blacklisted(initcall_t fn)1228 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1229 {
1230 return false;
1231 }
1232 #endif
1233 __setup("initcall_blacklist=", initcall_blacklist);
1234
1235 static __init_or_module void
trace_initcall_start_cb(void * data,initcall_t fn)1236 trace_initcall_start_cb(void *data, initcall_t fn)
1237 {
1238 ktime_t *calltime = data;
1239
1240 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1241 *calltime = ktime_get();
1242 }
1243
1244 static __init_or_module void
trace_initcall_finish_cb(void * data,initcall_t fn,int ret)1245 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1246 {
1247 ktime_t rettime, *calltime = data;
1248
1249 rettime = ktime_get();
1250 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1251 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1252 }
1253
1254 static ktime_t initcall_calltime;
1255
1256 #ifdef TRACEPOINTS_ENABLED
initcall_debug_enable(void)1257 static void __init initcall_debug_enable(void)
1258 {
1259 int ret;
1260
1261 ret = register_trace_initcall_start(trace_initcall_start_cb,
1262 &initcall_calltime);
1263 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1264 &initcall_calltime);
1265 WARN(ret, "Failed to register initcall tracepoints\n");
1266 }
1267 # define do_trace_initcall_start trace_initcall_start
1268 # define do_trace_initcall_finish trace_initcall_finish
1269 #else
do_trace_initcall_start(initcall_t fn)1270 static inline void do_trace_initcall_start(initcall_t fn)
1271 {
1272 if (!initcall_debug)
1273 return;
1274 trace_initcall_start_cb(&initcall_calltime, fn);
1275 }
do_trace_initcall_finish(initcall_t fn,int ret)1276 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1277 {
1278 if (!initcall_debug)
1279 return;
1280 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1281 }
1282 #endif /* !TRACEPOINTS_ENABLED */
1283
do_one_initcall(initcall_t fn)1284 int __init_or_module do_one_initcall(initcall_t fn)
1285 {
1286 int count = preempt_count();
1287 char msgbuf[64];
1288 int ret;
1289
1290 if (initcall_blacklisted(fn))
1291 return -EPERM;
1292
1293 do_trace_initcall_start(fn);
1294 ret = fn();
1295 do_trace_initcall_finish(fn, ret);
1296
1297 msgbuf[0] = 0;
1298
1299 if (preempt_count() != count) {
1300 sprintf(msgbuf, "preemption imbalance ");
1301 preempt_count_set(count);
1302 }
1303 if (irqs_disabled()) {
1304 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1305 local_irq_enable();
1306 }
1307 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1308
1309 add_latent_entropy();
1310 return ret;
1311 }
1312
1313
1314 extern initcall_entry_t __initcall_start[];
1315 extern initcall_entry_t __initcall0_start[];
1316 extern initcall_entry_t __initcall1_start[];
1317 extern initcall_entry_t __initcall2_start[];
1318 extern initcall_entry_t __initcall3_start[];
1319 extern initcall_entry_t __initcall4_start[];
1320 extern initcall_entry_t __initcall5_start[];
1321 extern initcall_entry_t __initcall6_start[];
1322 extern initcall_entry_t __initcall7_start[];
1323 extern initcall_entry_t __initcall_end[];
1324
1325 static initcall_entry_t *initcall_levels[] __initdata = {
1326 __initcall0_start,
1327 __initcall1_start,
1328 __initcall2_start,
1329 __initcall3_start,
1330 __initcall4_start,
1331 __initcall5_start,
1332 __initcall6_start,
1333 __initcall7_start,
1334 __initcall_end,
1335 };
1336
1337 /* Keep these in sync with initcalls in include/linux/init.h */
1338 static const char *initcall_level_names[] __initdata = {
1339 "pure",
1340 "core",
1341 "postcore",
1342 "arch",
1343 "subsys",
1344 "fs",
1345 "device",
1346 "late",
1347 };
1348
ignore_unknown_bootoption(char * param,char * val,const char * unused,void * arg)1349 static int __init ignore_unknown_bootoption(char *param, char *val,
1350 const char *unused, void *arg)
1351 {
1352 return 0;
1353 }
1354
do_initcall_level(int level,char * command_line)1355 static void __init do_initcall_level(int level, char *command_line)
1356 {
1357 initcall_entry_t *fn;
1358
1359 parse_args(initcall_level_names[level],
1360 command_line, __start___param,
1361 __stop___param - __start___param,
1362 level, level,
1363 NULL, ignore_unknown_bootoption);
1364
1365 trace_initcall_level(initcall_level_names[level]);
1366 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1367 do_one_initcall(initcall_from_entry(fn));
1368 }
1369
do_initcalls(void)1370 static void __init do_initcalls(void)
1371 {
1372 int level;
1373 size_t len = strlen(saved_command_line) + 1;
1374 char *command_line;
1375
1376 command_line = kzalloc(len, GFP_KERNEL);
1377 if (!command_line)
1378 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1379
1380 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1381 /* Parser modifies command_line, restore it each time */
1382 strcpy(command_line, saved_command_line);
1383 do_initcall_level(level, command_line);
1384 }
1385
1386 kfree(command_line);
1387 }
1388
1389 /*
1390 * Ok, the machine is now initialized. None of the devices
1391 * have been touched yet, but the CPU subsystem is up and
1392 * running, and memory and process management works.
1393 *
1394 * Now we can finally start doing some real work..
1395 */
do_basic_setup(void)1396 static void __init do_basic_setup(void)
1397 {
1398 cpuset_init_smp();
1399 driver_init();
1400 init_irq_proc();
1401 do_ctors();
1402 do_initcalls();
1403 }
1404
do_pre_smp_initcalls(void)1405 static void __init do_pre_smp_initcalls(void)
1406 {
1407 initcall_entry_t *fn;
1408
1409 trace_initcall_level("early");
1410 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1411 do_one_initcall(initcall_from_entry(fn));
1412 }
1413
run_init_process(const char * init_filename)1414 static int run_init_process(const char *init_filename)
1415 {
1416 const char *const *p;
1417
1418 argv_init[0] = init_filename;
1419 pr_info("Run %s as init process\n", init_filename);
1420 pr_debug(" with arguments:\n");
1421 for (p = argv_init; *p; p++)
1422 pr_debug(" %s\n", *p);
1423 pr_debug(" with environment:\n");
1424 for (p = envp_init; *p; p++)
1425 pr_debug(" %s\n", *p);
1426 return kernel_execve(init_filename, argv_init, envp_init);
1427 }
1428
try_to_run_init_process(const char * init_filename)1429 static int try_to_run_init_process(const char *init_filename)
1430 {
1431 int ret;
1432
1433 ret = run_init_process(init_filename);
1434
1435 if (ret && ret != -ENOENT) {
1436 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1437 init_filename, ret);
1438 }
1439
1440 return ret;
1441 }
1442
1443 static noinline void __init kernel_init_freeable(void);
1444
1445 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1446 bool rodata_enabled __ro_after_init = true;
1447
1448 #ifndef arch_parse_debug_rodata
arch_parse_debug_rodata(char * str)1449 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1450 #endif
1451
set_debug_rodata(char * str)1452 static int __init set_debug_rodata(char *str)
1453 {
1454 if (arch_parse_debug_rodata(str))
1455 return 0;
1456
1457 if (str && !strcmp(str, "on"))
1458 rodata_enabled = true;
1459 else if (str && !strcmp(str, "off"))
1460 rodata_enabled = false;
1461 else
1462 pr_warn("Invalid option string for rodata: '%s'\n", str);
1463 return 0;
1464 }
1465 early_param("rodata", set_debug_rodata);
1466 #endif
1467
1468 #ifdef CONFIG_STRICT_KERNEL_RWX
mark_readonly(void)1469 static void mark_readonly(void)
1470 {
1471 if (rodata_enabled) {
1472 /*
1473 * load_module() results in W+X mappings, which are cleaned
1474 * up with call_rcu(). Let's make sure that queued work is
1475 * flushed so that we don't hit false positives looking for
1476 * insecure pages which are W+X.
1477 */
1478 rcu_barrier();
1479 mark_rodata_ro();
1480 rodata_test();
1481 } else
1482 pr_info("Kernel memory protection disabled.\n");
1483 }
1484 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
mark_readonly(void)1485 static inline void mark_readonly(void)
1486 {
1487 pr_warn("Kernel memory protection not selected by kernel config.\n");
1488 }
1489 #else
mark_readonly(void)1490 static inline void mark_readonly(void)
1491 {
1492 pr_warn("This architecture does not have kernel memory protection.\n");
1493 }
1494 #endif
1495
free_initmem(void)1496 void __weak free_initmem(void)
1497 {
1498 free_initmem_default(POISON_FREE_INITMEM);
1499 }
1500
kernel_init(void * unused)1501 static int __ref kernel_init(void *unused)
1502 {
1503 int ret;
1504
1505 /*
1506 * Wait until kthreadd is all set-up.
1507 */
1508 wait_for_completion(&kthreadd_done);
1509
1510 kernel_init_freeable();
1511 /* need to finish all async __init code before freeing the memory */
1512 async_synchronize_full();
1513
1514 system_state = SYSTEM_FREEING_INITMEM;
1515 kprobe_free_init_mem();
1516 ftrace_free_init_mem();
1517 kgdb_free_init_mem();
1518 exit_boot_config();
1519 free_initmem();
1520 mark_readonly();
1521
1522 /*
1523 * Kernel mappings are now finalized - update the userspace page-table
1524 * to finalize PTI.
1525 */
1526 pti_finalize();
1527
1528 system_state = SYSTEM_RUNNING;
1529 numa_default_policy();
1530
1531 rcu_end_inkernel_boot();
1532
1533 do_sysctl_args();
1534
1535 if (ramdisk_execute_command) {
1536 ret = run_init_process(ramdisk_execute_command);
1537 if (!ret)
1538 return 0;
1539 pr_err("Failed to execute %s (error %d)\n",
1540 ramdisk_execute_command, ret);
1541 }
1542
1543 /*
1544 * We try each of these until one succeeds.
1545 *
1546 * The Bourne shell can be used instead of init if we are
1547 * trying to recover a really broken machine.
1548 */
1549 if (execute_command) {
1550 ret = run_init_process(execute_command);
1551 if (!ret)
1552 return 0;
1553 panic("Requested init %s failed (error %d).",
1554 execute_command, ret);
1555 }
1556
1557 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1558 ret = run_init_process(CONFIG_DEFAULT_INIT);
1559 if (ret)
1560 pr_err("Default init %s failed (error %d)\n",
1561 CONFIG_DEFAULT_INIT, ret);
1562 else
1563 return 0;
1564 }
1565
1566 if (!try_to_run_init_process("/sbin/init") ||
1567 !try_to_run_init_process("/etc/init") ||
1568 !try_to_run_init_process("/bin/init") ||
1569 !try_to_run_init_process("/bin/sh"))
1570 return 0;
1571
1572 panic("No working init found. Try passing init= option to kernel. "
1573 "See Linux Documentation/admin-guide/init.rst for guidance.");
1574 }
1575
1576 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
console_on_rootfs(void)1577 void __init console_on_rootfs(void)
1578 {
1579 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1580
1581 if (IS_ERR(file)) {
1582 pr_err("Warning: unable to open an initial console.\n");
1583 return;
1584 }
1585 init_dup(file);
1586 init_dup(file);
1587 init_dup(file);
1588 fput(file);
1589 }
1590
kernel_init_freeable(void)1591 static noinline void __init kernel_init_freeable(void)
1592 {
1593 /* Now the scheduler is fully set up and can do blocking allocations */
1594 gfp_allowed_mask = __GFP_BITS_MASK;
1595
1596 /*
1597 * init can allocate pages on any node
1598 */
1599 set_mems_allowed(node_states[N_MEMORY]);
1600
1601 cad_pid = get_pid(task_pid(current));
1602
1603 smp_prepare_cpus(setup_max_cpus);
1604
1605 workqueue_init();
1606
1607 init_mm_internals();
1608
1609 rcu_init_tasks_generic();
1610 do_pre_smp_initcalls();
1611 lockup_detector_init();
1612
1613 smp_init();
1614 sched_init_smp();
1615
1616 padata_init();
1617 page_alloc_init_late();
1618 /* Initialize page ext after all struct pages are initialized. */
1619 if (!early_page_ext_enabled())
1620 page_ext_init();
1621
1622 do_basic_setup();
1623
1624 wait_for_initramfs();
1625 console_on_rootfs();
1626
1627 /*
1628 * check if there is an early userspace init. If yes, let it do all
1629 * the work
1630 */
1631 if (init_eaccess(ramdisk_execute_command) != 0) {
1632 ramdisk_execute_command = NULL;
1633 prepare_namespace();
1634 }
1635
1636 /*
1637 * Ok, we have completed the initial bootup, and
1638 * we're essentially up and running. Get rid of the
1639 * initmem segments and start the user-mode stuff..
1640 *
1641 * rootfs is available now, try loading the public keys
1642 * and default modules
1643 */
1644
1645 integrity_load_keys();
1646 }
1647