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/mem_encrypt.h>
100 #include <linux/kcsan.h>
101 #include <linux/init_syscalls.h>
102 #include <linux/stackdepot.h>
103
104 #include <asm/io.h>
105 #include <asm/bugs.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 repair_env_string(param, val);
544
545 /* Handle obsolete-style parameters */
546 if (obsolete_checksetup(param))
547 return 0;
548
549 /* Unused module parameter. */
550 if (strnchr(param, len, '.'))
551 return 0;
552
553 if (panic_later)
554 return 0;
555
556 if (val) {
557 /* Environment option */
558 unsigned int i;
559 for (i = 0; envp_init[i]; i++) {
560 if (i == MAX_INIT_ENVS) {
561 panic_later = "env";
562 panic_param = param;
563 }
564 if (!strncmp(param, envp_init[i], len+1))
565 break;
566 }
567 envp_init[i] = param;
568 } else {
569 /* Command line option */
570 unsigned int i;
571 for (i = 0; argv_init[i]; i++) {
572 if (i == MAX_INIT_ARGS) {
573 panic_later = "init";
574 panic_param = param;
575 }
576 }
577 argv_init[i] = param;
578 }
579 return 0;
580 }
581
init_setup(char * str)582 static int __init init_setup(char *str)
583 {
584 unsigned int i;
585
586 execute_command = str;
587 /*
588 * In case LILO is going to boot us with default command line,
589 * it prepends "auto" before the whole cmdline which makes
590 * the shell think it should execute a script with such name.
591 * So we ignore all arguments entered _before_ init=... [MJ]
592 */
593 for (i = 1; i < MAX_INIT_ARGS; i++)
594 argv_init[i] = NULL;
595 return 1;
596 }
597 __setup("init=", init_setup);
598
rdinit_setup(char * str)599 static int __init rdinit_setup(char *str)
600 {
601 unsigned int i;
602
603 ramdisk_execute_command = str;
604 /* See "auto" comment in init_setup */
605 for (i = 1; i < MAX_INIT_ARGS; i++)
606 argv_init[i] = NULL;
607 return 1;
608 }
609 __setup("rdinit=", rdinit_setup);
610
611 #ifndef CONFIG_SMP
612 static const unsigned int setup_max_cpus = NR_CPUS;
setup_nr_cpu_ids(void)613 static inline void setup_nr_cpu_ids(void) { }
smp_prepare_cpus(unsigned int maxcpus)614 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
615 #endif
616
617 /*
618 * We need to store the untouched command line for future reference.
619 * We also need to store the touched command line since the parameter
620 * parsing is performed in place, and we should allow a component to
621 * store reference of name/value for future reference.
622 */
setup_command_line(char * command_line)623 static void __init setup_command_line(char *command_line)
624 {
625 size_t len, xlen = 0, ilen = 0;
626
627 if (extra_command_line)
628 xlen = strlen(extra_command_line);
629 if (extra_init_args)
630 ilen = strlen(extra_init_args) + 4; /* for " -- " */
631
632 len = xlen + strlen(boot_command_line) + 1;
633
634 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
635 if (!saved_command_line)
636 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
637
638 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
639 if (!static_command_line)
640 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
641
642 if (xlen) {
643 /*
644 * We have to put extra_command_line before boot command
645 * lines because there could be dashes (separator of init
646 * command line) in the command lines.
647 */
648 strcpy(saved_command_line, extra_command_line);
649 strcpy(static_command_line, extra_command_line);
650 }
651 strcpy(saved_command_line + xlen, boot_command_line);
652 strcpy(static_command_line + xlen, command_line);
653
654 if (ilen) {
655 /*
656 * Append supplemental init boot args to saved_command_line
657 * so that user can check what command line options passed
658 * to init.
659 * The order should always be
660 * " -- "[bootconfig init-param][cmdline init-param]
661 */
662 if (initargs_offs) {
663 len = xlen + initargs_offs;
664 strcpy(saved_command_line + len, extra_init_args);
665 len += ilen - 4; /* strlen(extra_init_args) */
666 strcpy(saved_command_line + len,
667 boot_command_line + initargs_offs - 1);
668 } else {
669 len = strlen(saved_command_line);
670 strcpy(saved_command_line + len, " -- ");
671 len += 4;
672 strcpy(saved_command_line + len, extra_init_args);
673 }
674 }
675 }
676
677 /*
678 * We need to finalize in a non-__init function or else race conditions
679 * between the root thread and the init thread may cause start_kernel to
680 * be reaped by free_initmem before the root thread has proceeded to
681 * cpu_idle.
682 *
683 * gcc-3.4 accidentally inlines this function, so use noinline.
684 */
685
686 static __initdata DECLARE_COMPLETION(kthreadd_done);
687
rest_init(void)688 noinline void __ref rest_init(void)
689 {
690 struct task_struct *tsk;
691 int pid;
692
693 rcu_scheduler_starting();
694 /*
695 * We need to spawn init first so that it obtains pid 1, however
696 * the init task will end up wanting to create kthreads, which, if
697 * we schedule it before we create kthreadd, will OOPS.
698 */
699 pid = kernel_thread(kernel_init, NULL, CLONE_FS);
700 /*
701 * Pin init on the boot CPU. Task migration is not properly working
702 * until sched_init_smp() has been run. It will set the allowed
703 * CPUs for init to the non isolated CPUs.
704 */
705 rcu_read_lock();
706 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
707 tsk->flags |= PF_NO_SETAFFINITY;
708 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
709 rcu_read_unlock();
710
711 numa_default_policy();
712 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
713 rcu_read_lock();
714 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
715 rcu_read_unlock();
716
717 /*
718 * Enable might_sleep() and smp_processor_id() checks.
719 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
720 * kernel_thread() would trigger might_sleep() splats. With
721 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
722 * already, but it's stuck on the kthreadd_done completion.
723 */
724 system_state = SYSTEM_SCHEDULING;
725
726 complete(&kthreadd_done);
727
728 /*
729 * The boot idle thread must execute schedule()
730 * at least once to get things moving:
731 */
732 schedule_preempt_disabled();
733 /* Call into cpu_idle with preempt disabled */
734 cpu_startup_entry(CPUHP_ONLINE);
735 }
736
737 /* Check for early params. */
do_early_param(char * param,char * val,const char * unused,void * arg)738 static int __init do_early_param(char *param, char *val,
739 const char *unused, void *arg)
740 {
741 const struct obs_kernel_param *p;
742
743 for (p = __setup_start; p < __setup_end; p++) {
744 if ((p->early && parameq(param, p->str)) ||
745 (strcmp(param, "console") == 0 &&
746 strcmp(p->str, "earlycon") == 0)
747 ) {
748 if (p->setup_func(val) != 0)
749 pr_warn("Malformed early option '%s'\n", param);
750 }
751 }
752 /* We accept everything at this stage. */
753 return 0;
754 }
755
parse_early_options(char * cmdline)756 void __init parse_early_options(char *cmdline)
757 {
758 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
759 do_early_param);
760 }
761
762 /* Arch code calls this early on, or if not, just before other parsing. */
parse_early_param(void)763 void __init parse_early_param(void)
764 {
765 static int done __initdata;
766 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
767
768 if (done)
769 return;
770
771 /* All fall through to do_early_param. */
772 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
773 parse_early_options(tmp_cmdline);
774 done = 1;
775 }
776
arch_post_acpi_subsys_init(void)777 void __init __weak arch_post_acpi_subsys_init(void) { }
778
smp_setup_processor_id(void)779 void __init __weak smp_setup_processor_id(void)
780 {
781 }
782
783 # if THREAD_SIZE >= PAGE_SIZE
thread_stack_cache_init(void)784 void __init __weak thread_stack_cache_init(void)
785 {
786 }
787 #endif
788
mem_encrypt_init(void)789 void __init __weak mem_encrypt_init(void) { }
790
poking_init(void)791 void __init __weak poking_init(void) { }
792
pgtable_cache_init(void)793 void __init __weak pgtable_cache_init(void) { }
794
trap_init(void)795 void __init __weak trap_init(void) { }
796
797 bool initcall_debug;
798 core_param(initcall_debug, initcall_debug, bool, 0644);
799
800 #ifdef TRACEPOINTS_ENABLED
801 static void __init initcall_debug_enable(void);
802 #else
initcall_debug_enable(void)803 static inline void initcall_debug_enable(void)
804 {
805 }
806 #endif
807
808 /* Report memory auto-initialization states for this boot. */
report_meminit(void)809 static void __init report_meminit(void)
810 {
811 const char *stack;
812
813 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
814 stack = "all(pattern)";
815 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
816 stack = "all(zero)";
817 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
818 stack = "byref_all(zero)";
819 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
820 stack = "byref(zero)";
821 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
822 stack = "__user(zero)";
823 else
824 stack = "off";
825
826 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
827 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
828 want_init_on_free() ? "on" : "off");
829 if (want_init_on_free())
830 pr_info("mem auto-init: clearing system memory may take some time...\n");
831 }
832
833 /*
834 * Set up kernel memory allocators
835 */
mm_init(void)836 static void __init mm_init(void)
837 {
838 /*
839 * page_ext requires contiguous pages,
840 * bigger than MAX_ORDER unless SPARSEMEM.
841 */
842 page_ext_init_flatmem();
843 init_mem_debugging_and_hardening();
844 kfence_alloc_pool();
845 report_meminit();
846 stack_depot_init();
847 mem_init();
848 mem_init_print_info();
849 /* page_owner must be initialized after buddy is ready */
850 page_ext_init_flatmem_late();
851 kmem_cache_init();
852 kmemleak_init();
853 pgtable_init();
854 debug_objects_mem_init();
855 vmalloc_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 }
861
862 #ifdef CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
863 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
864 randomize_kstack_offset);
865 DEFINE_PER_CPU(u32, kstack_offset);
866
early_randomize_kstack_offset(char * buf)867 static int __init early_randomize_kstack_offset(char *buf)
868 {
869 int ret;
870 bool bool_result;
871
872 ret = kstrtobool(buf, &bool_result);
873 if (ret)
874 return ret;
875
876 if (bool_result)
877 static_branch_enable(&randomize_kstack_offset);
878 else
879 static_branch_disable(&randomize_kstack_offset);
880 return 0;
881 }
882 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
883 #endif
884
arch_call_rest_init(void)885 void __init __weak arch_call_rest_init(void)
886 {
887 rest_init();
888 }
889
print_unknown_bootoptions(void)890 static void __init print_unknown_bootoptions(void)
891 {
892 char *unknown_options;
893 char *end;
894 const char *const *p;
895 size_t len;
896
897 if (panic_later || (!argv_init[1] && !envp_init[2]))
898 return;
899
900 /*
901 * Determine how many options we have to print out, plus a space
902 * before each
903 */
904 len = 1; /* null terminator */
905 for (p = &argv_init[1]; *p; p++) {
906 len++;
907 len += strlen(*p);
908 }
909 for (p = &envp_init[2]; *p; p++) {
910 len++;
911 len += strlen(*p);
912 }
913
914 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
915 if (!unknown_options) {
916 pr_err("%s: Failed to allocate %zu bytes\n",
917 __func__, len);
918 return;
919 }
920 end = unknown_options;
921
922 for (p = &argv_init[1]; *p; p++)
923 end += sprintf(end, " %s", *p);
924 for (p = &envp_init[2]; *p; p++)
925 end += sprintf(end, " %s", *p);
926
927 pr_notice("Unknown command line parameters:%s\n", unknown_options);
928 memblock_free_ptr(unknown_options, len);
929 }
930
start_kernel(void)931 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
932 {
933 char *command_line;
934 char *after_dashes;
935
936 set_task_stack_end_magic(&init_task);
937 smp_setup_processor_id();
938 debug_objects_early_init();
939 init_vmlinux_build_id();
940
941 cgroup_init_early();
942
943 local_irq_disable();
944 early_boot_irqs_disabled = true;
945
946 /*
947 * Interrupts are still disabled. Do necessary setups, then
948 * enable them.
949 */
950 boot_cpu_init();
951 page_address_init();
952 pr_notice("%s", linux_banner);
953 early_security_init();
954 setup_arch(&command_line);
955 setup_boot_config();
956 setup_command_line(command_line);
957 setup_nr_cpu_ids();
958 setup_per_cpu_areas();
959 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
960 boot_cpu_hotplug_init();
961
962 build_all_zonelists(NULL);
963 page_alloc_init();
964
965 pr_notice("Kernel command line: %s\n", saved_command_line);
966 /* parameters may set static keys */
967 jump_label_init();
968 parse_early_param();
969 after_dashes = parse_args("Booting kernel",
970 static_command_line, __start___param,
971 __stop___param - __start___param,
972 -1, -1, NULL, &unknown_bootoption);
973 print_unknown_bootoptions();
974 if (!IS_ERR_OR_NULL(after_dashes))
975 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
976 NULL, set_init_arg);
977 if (extra_init_args)
978 parse_args("Setting extra init args", extra_init_args,
979 NULL, 0, -1, -1, NULL, set_init_arg);
980
981 /*
982 * These use large bootmem allocations and must precede
983 * kmem_cache_init()
984 */
985 setup_log_buf(0);
986 vfs_caches_init_early();
987 sort_main_extable();
988 trap_init();
989 mm_init();
990
991 ftrace_init();
992
993 /* trace_printk can be enabled here */
994 early_trace_init();
995
996 /*
997 * Set up the scheduler prior starting any interrupts (such as the
998 * timer interrupt). Full topology setup happens at smp_init()
999 * time - but meanwhile we still have a functioning scheduler.
1000 */
1001 sched_init();
1002
1003 if (WARN(!irqs_disabled(),
1004 "Interrupts were enabled *very* early, fixing it\n"))
1005 local_irq_disable();
1006 radix_tree_init();
1007
1008 /*
1009 * Set up housekeeping before setting up workqueues to allow the unbound
1010 * workqueue to take non-housekeeping into account.
1011 */
1012 housekeeping_init();
1013
1014 /*
1015 * Allow workqueue creation and work item queueing/cancelling
1016 * early. Work item execution depends on kthreads and starts after
1017 * workqueue_init().
1018 */
1019 workqueue_init_early();
1020
1021 rcu_init();
1022
1023 /* Trace events are available after this */
1024 trace_init();
1025
1026 if (initcall_debug)
1027 initcall_debug_enable();
1028
1029 context_tracking_init();
1030 /* init some links before init_ISA_irqs() */
1031 early_irq_init();
1032 init_IRQ();
1033 tick_init();
1034 rcu_init_nohz();
1035 init_timers();
1036 srcu_init();
1037 hrtimers_init();
1038 softirq_init();
1039 timekeeping_init();
1040 kfence_init();
1041
1042 /*
1043 * For best initial stack canary entropy, prepare it after:
1044 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
1045 * - timekeeping_init() for ktime entropy used in rand_initialize()
1046 * - rand_initialize() to get any arch-specific entropy like RDRAND
1047 * - add_latent_entropy() to get any latent entropy
1048 * - adding command line entropy
1049 */
1050 rand_initialize();
1051 add_latent_entropy();
1052 add_device_randomness(command_line, strlen(command_line));
1053 boot_init_stack_canary();
1054
1055 time_init();
1056 perf_event_init();
1057 profile_init();
1058 call_function_init();
1059 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1060
1061 early_boot_irqs_disabled = false;
1062 local_irq_enable();
1063
1064 kmem_cache_init_late();
1065
1066 /*
1067 * HACK ALERT! This is early. We're enabling the console before
1068 * we've done PCI setups etc, and console_init() must be aware of
1069 * this. But we do want output early, in case something goes wrong.
1070 */
1071 console_init();
1072 if (panic_later)
1073 panic("Too many boot %s vars at `%s'", panic_later,
1074 panic_param);
1075
1076 lockdep_init();
1077
1078 /*
1079 * Need to run this when irqs are enabled, because it wants
1080 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1081 * too:
1082 */
1083 locking_selftest();
1084
1085 /*
1086 * This needs to be called before any devices perform DMA
1087 * operations that might use the SWIOTLB bounce buffers. It will
1088 * mark the bounce buffers as decrypted so that their usage will
1089 * not cause "plain-text" data to be decrypted when accessed.
1090 */
1091 mem_encrypt_init();
1092
1093 #ifdef CONFIG_BLK_DEV_INITRD
1094 if (initrd_start && !initrd_below_start_ok &&
1095 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1096 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1097 page_to_pfn(virt_to_page((void *)initrd_start)),
1098 min_low_pfn);
1099 initrd_start = 0;
1100 }
1101 #endif
1102 setup_per_cpu_pageset();
1103 numa_policy_init();
1104 acpi_early_init();
1105 if (late_time_init)
1106 late_time_init();
1107 sched_clock_init();
1108 calibrate_delay();
1109 pid_idr_init();
1110 anon_vma_init();
1111 #ifdef CONFIG_X86
1112 if (efi_enabled(EFI_RUNTIME_SERVICES))
1113 efi_enter_virtual_mode();
1114 #endif
1115 thread_stack_cache_init();
1116 cred_init();
1117 fork_init();
1118 proc_caches_init();
1119 uts_ns_init();
1120 key_init();
1121 security_init();
1122 dbg_late_init();
1123 vfs_caches_init();
1124 pagecache_init();
1125 signals_init();
1126 seq_file_init();
1127 proc_root_init();
1128 nsfs_init();
1129 cpuset_init();
1130 cgroup_init();
1131 taskstats_init_early();
1132 delayacct_init();
1133
1134 poking_init();
1135 check_bugs();
1136
1137 acpi_subsystem_init();
1138 arch_post_acpi_subsys_init();
1139 kcsan_init();
1140
1141 /* Do the rest non-__init'ed, we're now alive */
1142 arch_call_rest_init();
1143
1144 prevent_tail_call_optimization();
1145 }
1146
1147 /* Call all constructor functions linked into the kernel. */
do_ctors(void)1148 static void __init do_ctors(void)
1149 {
1150 /*
1151 * For UML, the constructors have already been called by the
1152 * normal setup code as it's just a normal ELF binary, so we
1153 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1154 * even on UML for modules.
1155 */
1156 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1157 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1158
1159 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1160 (*fn)();
1161 #endif
1162 }
1163
1164 #ifdef CONFIG_KALLSYMS
1165 struct blacklist_entry {
1166 struct list_head next;
1167 char *buf;
1168 };
1169
1170 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1171
initcall_blacklist(char * str)1172 static int __init initcall_blacklist(char *str)
1173 {
1174 char *str_entry;
1175 struct blacklist_entry *entry;
1176
1177 /* str argument is a comma-separated list of functions */
1178 do {
1179 str_entry = strsep(&str, ",");
1180 if (str_entry) {
1181 pr_debug("blacklisting initcall %s\n", str_entry);
1182 entry = memblock_alloc(sizeof(*entry),
1183 SMP_CACHE_BYTES);
1184 if (!entry)
1185 panic("%s: Failed to allocate %zu bytes\n",
1186 __func__, sizeof(*entry));
1187 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1188 SMP_CACHE_BYTES);
1189 if (!entry->buf)
1190 panic("%s: Failed to allocate %zu bytes\n",
1191 __func__, strlen(str_entry) + 1);
1192 strcpy(entry->buf, str_entry);
1193 list_add(&entry->next, &blacklisted_initcalls);
1194 }
1195 } while (str_entry);
1196
1197 return 0;
1198 }
1199
initcall_blacklisted(initcall_t fn)1200 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1201 {
1202 struct blacklist_entry *entry;
1203 char fn_name[KSYM_SYMBOL_LEN];
1204 unsigned long addr;
1205
1206 if (list_empty(&blacklisted_initcalls))
1207 return false;
1208
1209 addr = (unsigned long) dereference_function_descriptor(fn);
1210 sprint_symbol_no_offset(fn_name, addr);
1211
1212 /*
1213 * fn will be "function_name [module_name]" where [module_name] is not
1214 * displayed for built-in init functions. Strip off the [module_name].
1215 */
1216 strreplace(fn_name, ' ', '\0');
1217
1218 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1219 if (!strcmp(fn_name, entry->buf)) {
1220 pr_debug("initcall %s blacklisted\n", fn_name);
1221 return true;
1222 }
1223 }
1224
1225 return false;
1226 }
1227 #else
initcall_blacklist(char * str)1228 static int __init initcall_blacklist(char *str)
1229 {
1230 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1231 return 0;
1232 }
1233
initcall_blacklisted(initcall_t fn)1234 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1235 {
1236 return false;
1237 }
1238 #endif
1239 __setup("initcall_blacklist=", initcall_blacklist);
1240
1241 static __init_or_module void
trace_initcall_start_cb(void * data,initcall_t fn)1242 trace_initcall_start_cb(void *data, initcall_t fn)
1243 {
1244 ktime_t *calltime = (ktime_t *)data;
1245
1246 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1247 *calltime = ktime_get();
1248 }
1249
1250 static __init_or_module void
trace_initcall_finish_cb(void * data,initcall_t fn,int ret)1251 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1252 {
1253 ktime_t *calltime = (ktime_t *)data;
1254 ktime_t delta, rettime;
1255 unsigned long long duration;
1256
1257 rettime = ktime_get();
1258 delta = ktime_sub(rettime, *calltime);
1259 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1260 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1261 fn, ret, duration);
1262 }
1263
1264 static ktime_t initcall_calltime;
1265
1266 #ifdef TRACEPOINTS_ENABLED
initcall_debug_enable(void)1267 static void __init initcall_debug_enable(void)
1268 {
1269 int ret;
1270
1271 ret = register_trace_initcall_start(trace_initcall_start_cb,
1272 &initcall_calltime);
1273 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1274 &initcall_calltime);
1275 WARN(ret, "Failed to register initcall tracepoints\n");
1276 }
1277 # define do_trace_initcall_start trace_initcall_start
1278 # define do_trace_initcall_finish trace_initcall_finish
1279 #else
do_trace_initcall_start(initcall_t fn)1280 static inline void do_trace_initcall_start(initcall_t fn)
1281 {
1282 if (!initcall_debug)
1283 return;
1284 trace_initcall_start_cb(&initcall_calltime, fn);
1285 }
do_trace_initcall_finish(initcall_t fn,int ret)1286 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1287 {
1288 if (!initcall_debug)
1289 return;
1290 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1291 }
1292 #endif /* !TRACEPOINTS_ENABLED */
1293
do_one_initcall(initcall_t fn)1294 int __init_or_module do_one_initcall(initcall_t fn)
1295 {
1296 int count = preempt_count();
1297 char msgbuf[64];
1298 int ret;
1299
1300 if (initcall_blacklisted(fn))
1301 return -EPERM;
1302
1303 do_trace_initcall_start(fn);
1304 ret = fn();
1305 do_trace_initcall_finish(fn, ret);
1306
1307 msgbuf[0] = 0;
1308
1309 if (preempt_count() != count) {
1310 sprintf(msgbuf, "preemption imbalance ");
1311 preempt_count_set(count);
1312 }
1313 if (irqs_disabled()) {
1314 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1315 local_irq_enable();
1316 }
1317 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1318
1319 add_latent_entropy();
1320 return ret;
1321 }
1322
1323
1324 extern initcall_entry_t __initcall_start[];
1325 extern initcall_entry_t __initcall0_start[];
1326 extern initcall_entry_t __initcall1_start[];
1327 extern initcall_entry_t __initcall2_start[];
1328 extern initcall_entry_t __initcall3_start[];
1329 extern initcall_entry_t __initcall4_start[];
1330 extern initcall_entry_t __initcall5_start[];
1331 extern initcall_entry_t __initcall6_start[];
1332 extern initcall_entry_t __initcall7_start[];
1333 extern initcall_entry_t __initcall_end[];
1334
1335 static initcall_entry_t *initcall_levels[] __initdata = {
1336 __initcall0_start,
1337 __initcall1_start,
1338 __initcall2_start,
1339 __initcall3_start,
1340 __initcall4_start,
1341 __initcall5_start,
1342 __initcall6_start,
1343 __initcall7_start,
1344 __initcall_end,
1345 };
1346
1347 /* Keep these in sync with initcalls in include/linux/init.h */
1348 static const char *initcall_level_names[] __initdata = {
1349 "pure",
1350 "core",
1351 "postcore",
1352 "arch",
1353 "subsys",
1354 "fs",
1355 "device",
1356 "late",
1357 };
1358
ignore_unknown_bootoption(char * param,char * val,const char * unused,void * arg)1359 static int __init ignore_unknown_bootoption(char *param, char *val,
1360 const char *unused, void *arg)
1361 {
1362 return 0;
1363 }
1364
do_initcall_level(int level,char * command_line)1365 static void __init do_initcall_level(int level, char *command_line)
1366 {
1367 initcall_entry_t *fn;
1368
1369 parse_args(initcall_level_names[level],
1370 command_line, __start___param,
1371 __stop___param - __start___param,
1372 level, level,
1373 NULL, ignore_unknown_bootoption);
1374
1375 trace_initcall_level(initcall_level_names[level]);
1376 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1377 do_one_initcall(initcall_from_entry(fn));
1378 }
1379
do_initcalls(void)1380 static void __init do_initcalls(void)
1381 {
1382 int level;
1383 size_t len = strlen(saved_command_line) + 1;
1384 char *command_line;
1385
1386 command_line = kzalloc(len, GFP_KERNEL);
1387 if (!command_line)
1388 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1389
1390 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1391 /* Parser modifies command_line, restore it each time */
1392 strcpy(command_line, saved_command_line);
1393 do_initcall_level(level, command_line);
1394 }
1395
1396 kfree(command_line);
1397 }
1398
1399 /*
1400 * Ok, the machine is now initialized. None of the devices
1401 * have been touched yet, but the CPU subsystem is up and
1402 * running, and memory and process management works.
1403 *
1404 * Now we can finally start doing some real work..
1405 */
do_basic_setup(void)1406 static void __init do_basic_setup(void)
1407 {
1408 cpuset_init_smp();
1409 driver_init();
1410 init_irq_proc();
1411 do_ctors();
1412 do_initcalls();
1413 }
1414
do_pre_smp_initcalls(void)1415 static void __init do_pre_smp_initcalls(void)
1416 {
1417 initcall_entry_t *fn;
1418
1419 trace_initcall_level("early");
1420 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1421 do_one_initcall(initcall_from_entry(fn));
1422 }
1423
run_init_process(const char * init_filename)1424 static int run_init_process(const char *init_filename)
1425 {
1426 const char *const *p;
1427
1428 argv_init[0] = init_filename;
1429 pr_info("Run %s as init process\n", init_filename);
1430 pr_debug(" with arguments:\n");
1431 for (p = argv_init; *p; p++)
1432 pr_debug(" %s\n", *p);
1433 pr_debug(" with environment:\n");
1434 for (p = envp_init; *p; p++)
1435 pr_debug(" %s\n", *p);
1436 return kernel_execve(init_filename, argv_init, envp_init);
1437 }
1438
try_to_run_init_process(const char * init_filename)1439 static int try_to_run_init_process(const char *init_filename)
1440 {
1441 int ret;
1442
1443 ret = run_init_process(init_filename);
1444
1445 if (ret && ret != -ENOENT) {
1446 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1447 init_filename, ret);
1448 }
1449
1450 return ret;
1451 }
1452
1453 static noinline void __init kernel_init_freeable(void);
1454
1455 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1456 bool rodata_enabled __ro_after_init = true;
set_debug_rodata(char * str)1457 static int __init set_debug_rodata(char *str)
1458 {
1459 return strtobool(str, &rodata_enabled);
1460 }
1461 __setup("rodata=", set_debug_rodata);
1462 #endif
1463
1464 #ifdef CONFIG_STRICT_KERNEL_RWX
mark_readonly(void)1465 static void mark_readonly(void)
1466 {
1467 if (rodata_enabled) {
1468 /*
1469 * load_module() results in W+X mappings, which are cleaned
1470 * up with call_rcu(). Let's make sure that queued work is
1471 * flushed so that we don't hit false positives looking for
1472 * insecure pages which are W+X.
1473 */
1474 rcu_barrier();
1475 mark_rodata_ro();
1476 rodata_test();
1477 } else
1478 pr_info("Kernel memory protection disabled.\n");
1479 }
1480 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
mark_readonly(void)1481 static inline void mark_readonly(void)
1482 {
1483 pr_warn("Kernel memory protection not selected by kernel config.\n");
1484 }
1485 #else
mark_readonly(void)1486 static inline void mark_readonly(void)
1487 {
1488 pr_warn("This architecture does not have kernel memory protection.\n");
1489 }
1490 #endif
1491
free_initmem(void)1492 void __weak free_initmem(void)
1493 {
1494 free_initmem_default(POISON_FREE_INITMEM);
1495 }
1496
kernel_init(void * unused)1497 static int __ref kernel_init(void *unused)
1498 {
1499 int ret;
1500
1501 /*
1502 * Wait until kthreadd is all set-up.
1503 */
1504 wait_for_completion(&kthreadd_done);
1505
1506 kernel_init_freeable();
1507 /* need to finish all async __init code before freeing the memory */
1508 async_synchronize_full();
1509 kprobe_free_init_mem();
1510 ftrace_free_init_mem();
1511 kgdb_free_init_mem();
1512 exit_boot_config();
1513 free_initmem();
1514 mark_readonly();
1515
1516 /*
1517 * Kernel mappings are now finalized - update the userspace page-table
1518 * to finalize PTI.
1519 */
1520 pti_finalize();
1521
1522 system_state = SYSTEM_RUNNING;
1523 numa_default_policy();
1524
1525 rcu_end_inkernel_boot();
1526
1527 do_sysctl_args();
1528
1529 if (ramdisk_execute_command) {
1530 ret = run_init_process(ramdisk_execute_command);
1531 if (!ret)
1532 return 0;
1533 pr_err("Failed to execute %s (error %d)\n",
1534 ramdisk_execute_command, ret);
1535 }
1536
1537 /*
1538 * We try each of these until one succeeds.
1539 *
1540 * The Bourne shell can be used instead of init if we are
1541 * trying to recover a really broken machine.
1542 */
1543 if (execute_command) {
1544 ret = run_init_process(execute_command);
1545 if (!ret)
1546 return 0;
1547 panic("Requested init %s failed (error %d).",
1548 execute_command, ret);
1549 }
1550
1551 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1552 ret = run_init_process(CONFIG_DEFAULT_INIT);
1553 if (ret)
1554 pr_err("Default init %s failed (error %d)\n",
1555 CONFIG_DEFAULT_INIT, ret);
1556 else
1557 return 0;
1558 }
1559
1560 if (!try_to_run_init_process("/sbin/init") ||
1561 !try_to_run_init_process("/etc/init") ||
1562 !try_to_run_init_process("/bin/init") ||
1563 !try_to_run_init_process("/bin/sh"))
1564 return 0;
1565
1566 panic("No working init found. Try passing init= option to kernel. "
1567 "See Linux Documentation/admin-guide/init.rst for guidance.");
1568 }
1569
1570 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
console_on_rootfs(void)1571 void __init console_on_rootfs(void)
1572 {
1573 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1574
1575 if (IS_ERR(file)) {
1576 pr_err("Warning: unable to open an initial console.\n");
1577 return;
1578 }
1579 init_dup(file);
1580 init_dup(file);
1581 init_dup(file);
1582 fput(file);
1583 }
1584
kernel_init_freeable(void)1585 static noinline void __init kernel_init_freeable(void)
1586 {
1587 /* Now the scheduler is fully set up and can do blocking allocations */
1588 gfp_allowed_mask = __GFP_BITS_MASK;
1589
1590 /*
1591 * init can allocate pages on any node
1592 */
1593 set_mems_allowed(node_states[N_MEMORY]);
1594
1595 cad_pid = get_pid(task_pid(current));
1596
1597 smp_prepare_cpus(setup_max_cpus);
1598
1599 workqueue_init();
1600
1601 init_mm_internals();
1602
1603 rcu_init_tasks_generic();
1604 do_pre_smp_initcalls();
1605 lockup_detector_init();
1606
1607 smp_init();
1608 sched_init_smp();
1609
1610 padata_init();
1611 page_alloc_init_late();
1612 /* Initialize page ext after all struct pages are initialized. */
1613 page_ext_init();
1614
1615 do_basic_setup();
1616
1617 kunit_run_all_tests();
1618
1619 wait_for_initramfs();
1620 console_on_rootfs();
1621
1622 /*
1623 * check if there is an early userspace init. If yes, let it do all
1624 * the work
1625 */
1626 if (init_eaccess(ramdisk_execute_command) != 0) {
1627 ramdisk_execute_command = NULL;
1628 prepare_namespace();
1629 }
1630
1631 /*
1632 * Ok, we have completed the initial bootup, and
1633 * we're essentially up and running. Get rid of the
1634 * initmem segments and start the user-mode stuff..
1635 *
1636 * rootfs is available now, try loading the public keys
1637 * and default modules
1638 */
1639
1640 integrity_load_keys();
1641 }
1642
