Loading...
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
115static int kernel_init(void *);
116
117extern void init_IRQ(void);
118extern 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 */
127bool early_boot_irqs_disabled __read_mostly;
128
129enum system_states system_state __read_mostly;
130EXPORT_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
138extern void time_init(void);
139/* Default late time init is NULL. archs can override this later. */
140void (*__initdata late_time_init)(void);
141
142/* Untouched command line saved by arch-specific code. */
143char __initdata boot_command_line[COMMAND_LINE_SIZE];
144/* Untouched saved command line (eg. for /proc) */
145char *saved_command_line;
146/* Command line for parameter parsing */
147static char *static_command_line;
148/* Untouched extra command line */
149static char *extra_command_line;
150/* Extra init arguments */
151static char *extra_init_args;
152
153#ifdef CONFIG_BOOT_CONFIG
154/* Is bootconfig on command line? */
155static bool bootconfig_found;
156static bool initargs_found;
157#else
158# define bootconfig_found false
159# define initargs_found false
160#endif
161
162static char *execute_command;
163static 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 */
169bool static_key_initialized __read_mostly;
170EXPORT_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 */
181unsigned int reset_devices;
182EXPORT_SYMBOL(reset_devices);
183
184static 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
192static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
193const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
194static const char *panic_later, *panic_param;
195
196extern const struct obs_kernel_param __setup_start[], __setup_end[];
197
198static 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 */
231unsigned long loops_per_jiffy = (1<<12);
232EXPORT_SYMBOL(loops_per_jiffy);
233
234static int __init debug_kernel(char *str)
235{
236 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
237 return 0;
238}
239
240static int __init quiet_kernel(char *str)
241{
242 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
243 return 0;
244}
245
246early_param("debug", debug_kernel);
247early_param("quiet", quiet_kernel);
248
249static 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
266early_param("loglevel", loglevel);
267
268#ifdef CONFIG_BLK_DEV_INITRD
269static 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
291found:
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
313static 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
321static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
322
323#define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
324
325static 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 */
361static 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 return NULL;
386 }
387
388 return new_cmdline;
389}
390
391static int __init bootconfig_params(char *param, char *val,
392 const char *unused, void *arg)
393{
394 if (strcmp(param, "bootconfig") == 0) {
395 bootconfig_found = true;
396 }
397 return 0;
398}
399
400static int __init warn_bootconfig(char *str)
401{
402 /* The 'bootconfig' has been handled by bootconfig_params(). */
403 return 0;
404}
405
406static void __init setup_boot_config(void)
407{
408 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
409 const char *msg;
410 int pos;
411 u32 size, csum;
412 char *data, *copy, *err;
413 int ret;
414
415 /* Cut out the bootconfig data even if we have no bootconfig option */
416 data = get_boot_config_from_initrd(&size, &csum);
417
418 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
419 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
420 bootconfig_params);
421
422 if (IS_ERR(err) || !bootconfig_found)
423 return;
424
425 /* parse_args() stops at '--' and returns an address */
426 if (err)
427 initargs_found = true;
428
429 if (!data) {
430 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
431 return;
432 }
433
434 if (size >= XBC_DATA_MAX) {
435 pr_err("bootconfig size %d greater than max size %d\n",
436 size, XBC_DATA_MAX);
437 return;
438 }
439
440 if (xbc_calc_checksum(data, size) != csum) {
441 pr_err("bootconfig checksum failed\n");
442 return;
443 }
444
445 copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
446 if (!copy) {
447 pr_err("Failed to allocate memory for bootconfig\n");
448 return;
449 }
450
451 memcpy(copy, data, size);
452 copy[size] = '\0';
453
454 ret = xbc_init(copy, &msg, &pos);
455 if (ret < 0) {
456 if (pos < 0)
457 pr_err("Failed to init bootconfig: %s.\n", msg);
458 else
459 pr_err("Failed to parse bootconfig: %s at %d.\n",
460 msg, pos);
461 } else {
462 pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
463 /* keys starting with "kernel." are passed via cmdline */
464 extra_command_line = xbc_make_cmdline("kernel");
465 /* Also, "init." keys are init arguments */
466 extra_init_args = xbc_make_cmdline("init");
467 }
468 return;
469}
470
471#else
472
473static void __init setup_boot_config(void)
474{
475 /* Remove bootconfig data from initrd */
476 get_boot_config_from_initrd(NULL, NULL);
477}
478
479static int __init warn_bootconfig(char *str)
480{
481 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
482 return 0;
483}
484#endif
485early_param("bootconfig", warn_bootconfig);
486
487/* Change NUL term back to "=", to make "param" the whole string. */
488static void __init repair_env_string(char *param, char *val)
489{
490 if (val) {
491 /* param=val or param="val"? */
492 if (val == param+strlen(param)+1)
493 val[-1] = '=';
494 else if (val == param+strlen(param)+2) {
495 val[-2] = '=';
496 memmove(val-1, val, strlen(val)+1);
497 } else
498 BUG();
499 }
500}
501
502/* Anything after -- gets handed straight to init. */
503static int __init set_init_arg(char *param, char *val,
504 const char *unused, void *arg)
505{
506 unsigned int i;
507
508 if (panic_later)
509 return 0;
510
511 repair_env_string(param, val);
512
513 for (i = 0; argv_init[i]; i++) {
514 if (i == MAX_INIT_ARGS) {
515 panic_later = "init";
516 panic_param = param;
517 return 0;
518 }
519 }
520 argv_init[i] = param;
521 return 0;
522}
523
524/*
525 * Unknown boot options get handed to init, unless they look like
526 * unused parameters (modprobe will find them in /proc/cmdline).
527 */
528static int __init unknown_bootoption(char *param, char *val,
529 const char *unused, void *arg)
530{
531 size_t len = strlen(param);
532
533 repair_env_string(param, val);
534
535 /* Handle obsolete-style parameters */
536 if (obsolete_checksetup(param))
537 return 0;
538
539 /* Unused module parameter. */
540 if (strnchr(param, len, '.'))
541 return 0;
542
543 if (panic_later)
544 return 0;
545
546 if (val) {
547 /* Environment option */
548 unsigned int i;
549 for (i = 0; envp_init[i]; i++) {
550 if (i == MAX_INIT_ENVS) {
551 panic_later = "env";
552 panic_param = param;
553 }
554 if (!strncmp(param, envp_init[i], len+1))
555 break;
556 }
557 envp_init[i] = param;
558 } else {
559 /* Command line option */
560 unsigned int i;
561 for (i = 0; argv_init[i]; i++) {
562 if (i == MAX_INIT_ARGS) {
563 panic_later = "init";
564 panic_param = param;
565 }
566 }
567 argv_init[i] = param;
568 }
569 return 0;
570}
571
572static int __init init_setup(char *str)
573{
574 unsigned int i;
575
576 execute_command = str;
577 /*
578 * In case LILO is going to boot us with default command line,
579 * it prepends "auto" before the whole cmdline which makes
580 * the shell think it should execute a script with such name.
581 * So we ignore all arguments entered _before_ init=... [MJ]
582 */
583 for (i = 1; i < MAX_INIT_ARGS; i++)
584 argv_init[i] = NULL;
585 return 1;
586}
587__setup("init=", init_setup);
588
589static int __init rdinit_setup(char *str)
590{
591 unsigned int i;
592
593 ramdisk_execute_command = str;
594 /* See "auto" comment in init_setup */
595 for (i = 1; i < MAX_INIT_ARGS; i++)
596 argv_init[i] = NULL;
597 return 1;
598}
599__setup("rdinit=", rdinit_setup);
600
601#ifndef CONFIG_SMP
602static const unsigned int setup_max_cpus = NR_CPUS;
603static inline void setup_nr_cpu_ids(void) { }
604static inline void smp_prepare_cpus(unsigned int maxcpus) { }
605#endif
606
607/*
608 * We need to store the untouched command line for future reference.
609 * We also need to store the touched command line since the parameter
610 * parsing is performed in place, and we should allow a component to
611 * store reference of name/value for future reference.
612 */
613static void __init setup_command_line(char *command_line)
614{
615 size_t len, xlen = 0, ilen = 0;
616
617 if (extra_command_line)
618 xlen = strlen(extra_command_line);
619 if (extra_init_args)
620 ilen = strlen(extra_init_args) + 4; /* for " -- " */
621
622 len = xlen + strlen(boot_command_line) + 1;
623
624 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
625 if (!saved_command_line)
626 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
627
628 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
629 if (!static_command_line)
630 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
631
632 if (xlen) {
633 /*
634 * We have to put extra_command_line before boot command
635 * lines because there could be dashes (separator of init
636 * command line) in the command lines.
637 */
638 strcpy(saved_command_line, extra_command_line);
639 strcpy(static_command_line, extra_command_line);
640 }
641 strcpy(saved_command_line + xlen, boot_command_line);
642 strcpy(static_command_line + xlen, command_line);
643
644 if (ilen) {
645 /*
646 * Append supplemental init boot args to saved_command_line
647 * so that user can check what command line options passed
648 * to init.
649 */
650 len = strlen(saved_command_line);
651 if (initargs_found) {
652 saved_command_line[len++] = ' ';
653 } else {
654 strcpy(saved_command_line + len, " -- ");
655 len += 4;
656 }
657
658 strcpy(saved_command_line + len, extra_init_args);
659 }
660}
661
662/*
663 * We need to finalize in a non-__init function or else race conditions
664 * between the root thread and the init thread may cause start_kernel to
665 * be reaped by free_initmem before the root thread has proceeded to
666 * cpu_idle.
667 *
668 * gcc-3.4 accidentally inlines this function, so use noinline.
669 */
670
671static __initdata DECLARE_COMPLETION(kthreadd_done);
672
673noinline void __ref rest_init(void)
674{
675 struct task_struct *tsk;
676 int pid;
677
678 rcu_scheduler_starting();
679 /*
680 * We need to spawn init first so that it obtains pid 1, however
681 * the init task will end up wanting to create kthreads, which, if
682 * we schedule it before we create kthreadd, will OOPS.
683 */
684 pid = kernel_thread(kernel_init, NULL, CLONE_FS);
685 /*
686 * Pin init on the boot CPU. Task migration is not properly working
687 * until sched_init_smp() has been run. It will set the allowed
688 * CPUs for init to the non isolated CPUs.
689 */
690 rcu_read_lock();
691 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
692 tsk->flags |= PF_NO_SETAFFINITY;
693 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
694 rcu_read_unlock();
695
696 numa_default_policy();
697 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
698 rcu_read_lock();
699 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
700 rcu_read_unlock();
701
702 /*
703 * Enable might_sleep() and smp_processor_id() checks.
704 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
705 * kernel_thread() would trigger might_sleep() splats. With
706 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
707 * already, but it's stuck on the kthreadd_done completion.
708 */
709 system_state = SYSTEM_SCHEDULING;
710
711 complete(&kthreadd_done);
712
713 /*
714 * The boot idle thread must execute schedule()
715 * at least once to get things moving:
716 */
717 schedule_preempt_disabled();
718 /* Call into cpu_idle with preempt disabled */
719 cpu_startup_entry(CPUHP_ONLINE);
720}
721
722/* Check for early params. */
723static int __init do_early_param(char *param, char *val,
724 const char *unused, void *arg)
725{
726 const struct obs_kernel_param *p;
727
728 for (p = __setup_start; p < __setup_end; p++) {
729 if ((p->early && parameq(param, p->str)) ||
730 (strcmp(param, "console") == 0 &&
731 strcmp(p->str, "earlycon") == 0)
732 ) {
733 if (p->setup_func(val) != 0)
734 pr_warn("Malformed early option '%s'\n", param);
735 }
736 }
737 /* We accept everything at this stage. */
738 return 0;
739}
740
741void __init parse_early_options(char *cmdline)
742{
743 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
744 do_early_param);
745}
746
747/* Arch code calls this early on, or if not, just before other parsing. */
748void __init parse_early_param(void)
749{
750 static int done __initdata;
751 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
752
753 if (done)
754 return;
755
756 /* All fall through to do_early_param. */
757 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
758 parse_early_options(tmp_cmdline);
759 done = 1;
760}
761
762void __init __weak arch_post_acpi_subsys_init(void) { }
763
764void __init __weak smp_setup_processor_id(void)
765{
766}
767
768# if THREAD_SIZE >= PAGE_SIZE
769void __init __weak thread_stack_cache_init(void)
770{
771}
772#endif
773
774void __init __weak mem_encrypt_init(void) { }
775
776void __init __weak poking_init(void) { }
777
778void __init __weak pgtable_cache_init(void) { }
779
780bool initcall_debug;
781core_param(initcall_debug, initcall_debug, bool, 0644);
782
783#ifdef TRACEPOINTS_ENABLED
784static void __init initcall_debug_enable(void);
785#else
786static inline void initcall_debug_enable(void)
787{
788}
789#endif
790
791/* Report memory auto-initialization states for this boot. */
792static void __init report_meminit(void)
793{
794 const char *stack;
795
796 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
797 stack = "all(pattern)";
798 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
799 stack = "all(zero)";
800 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
801 stack = "byref_all(zero)";
802 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
803 stack = "byref(zero)";
804 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
805 stack = "__user(zero)";
806 else
807 stack = "off";
808
809 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
810 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
811 want_init_on_free() ? "on" : "off");
812 if (want_init_on_free())
813 pr_info("mem auto-init: clearing system memory may take some time...\n");
814}
815
816/*
817 * Set up kernel memory allocators
818 */
819static void __init mm_init(void)
820{
821 /*
822 * page_ext requires contiguous pages,
823 * bigger than MAX_ORDER unless SPARSEMEM.
824 */
825 page_ext_init_flatmem();
826 init_mem_debugging_and_hardening();
827 kfence_alloc_pool();
828 report_meminit();
829 stack_depot_init();
830 mem_init();
831 mem_init_print_info();
832 /* page_owner must be initialized after buddy is ready */
833 page_ext_init_flatmem_late();
834 kmem_cache_init();
835 kmemleak_init();
836 pgtable_init();
837 debug_objects_mem_init();
838 vmalloc_init();
839 /* Should be run before the first non-init thread is created */
840 init_espfix_bsp();
841 /* Should be run after espfix64 is set up. */
842 pti_init();
843}
844
845#ifdef CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
846DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
847 randomize_kstack_offset);
848DEFINE_PER_CPU(u32, kstack_offset);
849
850static int __init early_randomize_kstack_offset(char *buf)
851{
852 int ret;
853 bool bool_result;
854
855 ret = kstrtobool(buf, &bool_result);
856 if (ret)
857 return ret;
858
859 if (bool_result)
860 static_branch_enable(&randomize_kstack_offset);
861 else
862 static_branch_disable(&randomize_kstack_offset);
863 return 0;
864}
865early_param("randomize_kstack_offset", early_randomize_kstack_offset);
866#endif
867
868void __init __weak arch_call_rest_init(void)
869{
870 rest_init();
871}
872
873static void __init print_unknown_bootoptions(void)
874{
875 char *unknown_options;
876 char *end;
877 const char *const *p;
878 size_t len;
879
880 if (panic_later || (!argv_init[1] && !envp_init[2]))
881 return;
882
883 /*
884 * Determine how many options we have to print out, plus a space
885 * before each
886 */
887 len = 1; /* null terminator */
888 for (p = &argv_init[1]; *p; p++) {
889 len++;
890 len += strlen(*p);
891 }
892 for (p = &envp_init[2]; *p; p++) {
893 len++;
894 len += strlen(*p);
895 }
896
897 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
898 if (!unknown_options) {
899 pr_err("%s: Failed to allocate %zu bytes\n",
900 __func__, len);
901 return;
902 }
903 end = unknown_options;
904
905 for (p = &argv_init[1]; *p; p++)
906 end += sprintf(end, " %s", *p);
907 for (p = &envp_init[2]; *p; p++)
908 end += sprintf(end, " %s", *p);
909
910 pr_notice("Unknown command line parameters:%s\n", unknown_options);
911 memblock_free(__pa(unknown_options), len);
912}
913
914asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
915{
916 char *command_line;
917 char *after_dashes;
918
919 set_task_stack_end_magic(&init_task);
920 smp_setup_processor_id();
921 debug_objects_early_init();
922 init_vmlinux_build_id();
923
924 cgroup_init_early();
925
926 local_irq_disable();
927 early_boot_irqs_disabled = true;
928
929 /*
930 * Interrupts are still disabled. Do necessary setups, then
931 * enable them.
932 */
933 boot_cpu_init();
934 page_address_init();
935 pr_notice("%s", linux_banner);
936 early_security_init();
937 setup_arch(&command_line);
938 setup_boot_config();
939 setup_command_line(command_line);
940 setup_nr_cpu_ids();
941 setup_per_cpu_areas();
942 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
943 boot_cpu_hotplug_init();
944
945 build_all_zonelists(NULL);
946 page_alloc_init();
947
948 pr_notice("Kernel command line: %s\n", saved_command_line);
949 /* parameters may set static keys */
950 jump_label_init();
951 parse_early_param();
952 after_dashes = parse_args("Booting kernel",
953 static_command_line, __start___param,
954 __stop___param - __start___param,
955 -1, -1, NULL, &unknown_bootoption);
956 print_unknown_bootoptions();
957 if (!IS_ERR_OR_NULL(after_dashes))
958 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
959 NULL, set_init_arg);
960 if (extra_init_args)
961 parse_args("Setting extra init args", extra_init_args,
962 NULL, 0, -1, -1, NULL, set_init_arg);
963
964 /*
965 * These use large bootmem allocations and must precede
966 * kmem_cache_init()
967 */
968 setup_log_buf(0);
969 vfs_caches_init_early();
970 sort_main_extable();
971 trap_init();
972 mm_init();
973
974 ftrace_init();
975
976 /* trace_printk can be enabled here */
977 early_trace_init();
978
979 /*
980 * Set up the scheduler prior starting any interrupts (such as the
981 * timer interrupt). Full topology setup happens at smp_init()
982 * time - but meanwhile we still have a functioning scheduler.
983 */
984 sched_init();
985
986 if (WARN(!irqs_disabled(),
987 "Interrupts were enabled *very* early, fixing it\n"))
988 local_irq_disable();
989 radix_tree_init();
990
991 /*
992 * Set up housekeeping before setting up workqueues to allow the unbound
993 * workqueue to take non-housekeeping into account.
994 */
995 housekeeping_init();
996
997 /*
998 * Allow workqueue creation and work item queueing/cancelling
999 * early. Work item execution depends on kthreads and starts after
1000 * workqueue_init().
1001 */
1002 workqueue_init_early();
1003
1004 rcu_init();
1005
1006 /* Trace events are available after this */
1007 trace_init();
1008
1009 if (initcall_debug)
1010 initcall_debug_enable();
1011
1012 context_tracking_init();
1013 /* init some links before init_ISA_irqs() */
1014 early_irq_init();
1015 init_IRQ();
1016 tick_init();
1017 rcu_init_nohz();
1018 init_timers();
1019 srcu_init();
1020 hrtimers_init();
1021 softirq_init();
1022 timekeeping_init();
1023 kfence_init();
1024
1025 /*
1026 * For best initial stack canary entropy, prepare it after:
1027 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
1028 * - timekeeping_init() for ktime entropy used in rand_initialize()
1029 * - rand_initialize() to get any arch-specific entropy like RDRAND
1030 * - add_latent_entropy() to get any latent entropy
1031 * - adding command line entropy
1032 */
1033 rand_initialize();
1034 add_latent_entropy();
1035 add_device_randomness(command_line, strlen(command_line));
1036 boot_init_stack_canary();
1037
1038 time_init();
1039 perf_event_init();
1040 profile_init();
1041 call_function_init();
1042 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1043
1044 early_boot_irqs_disabled = false;
1045 local_irq_enable();
1046
1047 kmem_cache_init_late();
1048
1049 /*
1050 * HACK ALERT! This is early. We're enabling the console before
1051 * we've done PCI setups etc, and console_init() must be aware of
1052 * this. But we do want output early, in case something goes wrong.
1053 */
1054 console_init();
1055 if (panic_later)
1056 panic("Too many boot %s vars at `%s'", panic_later,
1057 panic_param);
1058
1059 lockdep_init();
1060
1061 /*
1062 * Need to run this when irqs are enabled, because it wants
1063 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1064 * too:
1065 */
1066 locking_selftest();
1067
1068 /*
1069 * This needs to be called before any devices perform DMA
1070 * operations that might use the SWIOTLB bounce buffers. It will
1071 * mark the bounce buffers as decrypted so that their usage will
1072 * not cause "plain-text" data to be decrypted when accessed.
1073 */
1074 mem_encrypt_init();
1075
1076#ifdef CONFIG_BLK_DEV_INITRD
1077 if (initrd_start && !initrd_below_start_ok &&
1078 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1079 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1080 page_to_pfn(virt_to_page((void *)initrd_start)),
1081 min_low_pfn);
1082 initrd_start = 0;
1083 }
1084#endif
1085 setup_per_cpu_pageset();
1086 numa_policy_init();
1087 acpi_early_init();
1088 if (late_time_init)
1089 late_time_init();
1090 sched_clock_init();
1091 calibrate_delay();
1092 pid_idr_init();
1093 anon_vma_init();
1094#ifdef CONFIG_X86
1095 if (efi_enabled(EFI_RUNTIME_SERVICES))
1096 efi_enter_virtual_mode();
1097#endif
1098 thread_stack_cache_init();
1099 cred_init();
1100 fork_init();
1101 proc_caches_init();
1102 uts_ns_init();
1103 key_init();
1104 security_init();
1105 dbg_late_init();
1106 vfs_caches_init();
1107 pagecache_init();
1108 signals_init();
1109 seq_file_init();
1110 proc_root_init();
1111 nsfs_init();
1112 cpuset_init();
1113 cgroup_init();
1114 taskstats_init_early();
1115 delayacct_init();
1116
1117 poking_init();
1118 check_bugs();
1119
1120 acpi_subsystem_init();
1121 arch_post_acpi_subsys_init();
1122 kcsan_init();
1123
1124 /* Do the rest non-__init'ed, we're now alive */
1125 arch_call_rest_init();
1126
1127 prevent_tail_call_optimization();
1128}
1129
1130/* Call all constructor functions linked into the kernel. */
1131static void __init do_ctors(void)
1132{
1133/*
1134 * For UML, the constructors have already been called by the
1135 * normal setup code as it's just a normal ELF binary, so we
1136 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1137 * even on UML for modules.
1138 */
1139#if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1140 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1141
1142 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1143 (*fn)();
1144#endif
1145}
1146
1147#ifdef CONFIG_KALLSYMS
1148struct blacklist_entry {
1149 struct list_head next;
1150 char *buf;
1151};
1152
1153static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1154
1155static int __init initcall_blacklist(char *str)
1156{
1157 char *str_entry;
1158 struct blacklist_entry *entry;
1159
1160 /* str argument is a comma-separated list of functions */
1161 do {
1162 str_entry = strsep(&str, ",");
1163 if (str_entry) {
1164 pr_debug("blacklisting initcall %s\n", str_entry);
1165 entry = memblock_alloc(sizeof(*entry),
1166 SMP_CACHE_BYTES);
1167 if (!entry)
1168 panic("%s: Failed to allocate %zu bytes\n",
1169 __func__, sizeof(*entry));
1170 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1171 SMP_CACHE_BYTES);
1172 if (!entry->buf)
1173 panic("%s: Failed to allocate %zu bytes\n",
1174 __func__, strlen(str_entry) + 1);
1175 strcpy(entry->buf, str_entry);
1176 list_add(&entry->next, &blacklisted_initcalls);
1177 }
1178 } while (str_entry);
1179
1180 return 0;
1181}
1182
1183static bool __init_or_module initcall_blacklisted(initcall_t fn)
1184{
1185 struct blacklist_entry *entry;
1186 char fn_name[KSYM_SYMBOL_LEN];
1187 unsigned long addr;
1188
1189 if (list_empty(&blacklisted_initcalls))
1190 return false;
1191
1192 addr = (unsigned long) dereference_function_descriptor(fn);
1193 sprint_symbol_no_offset(fn_name, addr);
1194
1195 /*
1196 * fn will be "function_name [module_name]" where [module_name] is not
1197 * displayed for built-in init functions. Strip off the [module_name].
1198 */
1199 strreplace(fn_name, ' ', '\0');
1200
1201 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1202 if (!strcmp(fn_name, entry->buf)) {
1203 pr_debug("initcall %s blacklisted\n", fn_name);
1204 return true;
1205 }
1206 }
1207
1208 return false;
1209}
1210#else
1211static int __init initcall_blacklist(char *str)
1212{
1213 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1214 return 0;
1215}
1216
1217static bool __init_or_module initcall_blacklisted(initcall_t fn)
1218{
1219 return false;
1220}
1221#endif
1222__setup("initcall_blacklist=", initcall_blacklist);
1223
1224static __init_or_module void
1225trace_initcall_start_cb(void *data, initcall_t fn)
1226{
1227 ktime_t *calltime = (ktime_t *)data;
1228
1229 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1230 *calltime = ktime_get();
1231}
1232
1233static __init_or_module void
1234trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1235{
1236 ktime_t *calltime = (ktime_t *)data;
1237 ktime_t delta, rettime;
1238 unsigned long long duration;
1239
1240 rettime = ktime_get();
1241 delta = ktime_sub(rettime, *calltime);
1242 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1243 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1244 fn, ret, duration);
1245}
1246
1247static ktime_t initcall_calltime;
1248
1249#ifdef TRACEPOINTS_ENABLED
1250static void __init initcall_debug_enable(void)
1251{
1252 int ret;
1253
1254 ret = register_trace_initcall_start(trace_initcall_start_cb,
1255 &initcall_calltime);
1256 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1257 &initcall_calltime);
1258 WARN(ret, "Failed to register initcall tracepoints\n");
1259}
1260# define do_trace_initcall_start trace_initcall_start
1261# define do_trace_initcall_finish trace_initcall_finish
1262#else
1263static inline void do_trace_initcall_start(initcall_t fn)
1264{
1265 if (!initcall_debug)
1266 return;
1267 trace_initcall_start_cb(&initcall_calltime, fn);
1268}
1269static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1270{
1271 if (!initcall_debug)
1272 return;
1273 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1274}
1275#endif /* !TRACEPOINTS_ENABLED */
1276
1277int __init_or_module do_one_initcall(initcall_t fn)
1278{
1279 int count = preempt_count();
1280 char msgbuf[64];
1281 int ret;
1282
1283 if (initcall_blacklisted(fn))
1284 return -EPERM;
1285
1286 do_trace_initcall_start(fn);
1287 ret = fn();
1288 do_trace_initcall_finish(fn, ret);
1289
1290 msgbuf[0] = 0;
1291
1292 if (preempt_count() != count) {
1293 sprintf(msgbuf, "preemption imbalance ");
1294 preempt_count_set(count);
1295 }
1296 if (irqs_disabled()) {
1297 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1298 local_irq_enable();
1299 }
1300 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1301
1302 add_latent_entropy();
1303 return ret;
1304}
1305
1306
1307extern initcall_entry_t __initcall_start[];
1308extern initcall_entry_t __initcall0_start[];
1309extern initcall_entry_t __initcall1_start[];
1310extern initcall_entry_t __initcall2_start[];
1311extern initcall_entry_t __initcall3_start[];
1312extern initcall_entry_t __initcall4_start[];
1313extern initcall_entry_t __initcall5_start[];
1314extern initcall_entry_t __initcall6_start[];
1315extern initcall_entry_t __initcall7_start[];
1316extern initcall_entry_t __initcall_end[];
1317
1318static initcall_entry_t *initcall_levels[] __initdata = {
1319 __initcall0_start,
1320 __initcall1_start,
1321 __initcall2_start,
1322 __initcall3_start,
1323 __initcall4_start,
1324 __initcall5_start,
1325 __initcall6_start,
1326 __initcall7_start,
1327 __initcall_end,
1328};
1329
1330/* Keep these in sync with initcalls in include/linux/init.h */
1331static const char *initcall_level_names[] __initdata = {
1332 "pure",
1333 "core",
1334 "postcore",
1335 "arch",
1336 "subsys",
1337 "fs",
1338 "device",
1339 "late",
1340};
1341
1342static int __init ignore_unknown_bootoption(char *param, char *val,
1343 const char *unused, void *arg)
1344{
1345 return 0;
1346}
1347
1348static void __init do_initcall_level(int level, char *command_line)
1349{
1350 initcall_entry_t *fn;
1351
1352 parse_args(initcall_level_names[level],
1353 command_line, __start___param,
1354 __stop___param - __start___param,
1355 level, level,
1356 NULL, ignore_unknown_bootoption);
1357
1358 trace_initcall_level(initcall_level_names[level]);
1359 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1360 do_one_initcall(initcall_from_entry(fn));
1361}
1362
1363static void __init do_initcalls(void)
1364{
1365 int level;
1366 size_t len = strlen(saved_command_line) + 1;
1367 char *command_line;
1368
1369 command_line = kzalloc(len, GFP_KERNEL);
1370 if (!command_line)
1371 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1372
1373 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1374 /* Parser modifies command_line, restore it each time */
1375 strcpy(command_line, saved_command_line);
1376 do_initcall_level(level, command_line);
1377 }
1378
1379 kfree(command_line);
1380}
1381
1382/*
1383 * Ok, the machine is now initialized. None of the devices
1384 * have been touched yet, but the CPU subsystem is up and
1385 * running, and memory and process management works.
1386 *
1387 * Now we can finally start doing some real work..
1388 */
1389static void __init do_basic_setup(void)
1390{
1391 cpuset_init_smp();
1392 driver_init();
1393 init_irq_proc();
1394 do_ctors();
1395 do_initcalls();
1396}
1397
1398static void __init do_pre_smp_initcalls(void)
1399{
1400 initcall_entry_t *fn;
1401
1402 trace_initcall_level("early");
1403 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1404 do_one_initcall(initcall_from_entry(fn));
1405}
1406
1407static int run_init_process(const char *init_filename)
1408{
1409 const char *const *p;
1410
1411 argv_init[0] = init_filename;
1412 pr_info("Run %s as init process\n", init_filename);
1413 pr_debug(" with arguments:\n");
1414 for (p = argv_init; *p; p++)
1415 pr_debug(" %s\n", *p);
1416 pr_debug(" with environment:\n");
1417 for (p = envp_init; *p; p++)
1418 pr_debug(" %s\n", *p);
1419 return kernel_execve(init_filename, argv_init, envp_init);
1420}
1421
1422static int try_to_run_init_process(const char *init_filename)
1423{
1424 int ret;
1425
1426 ret = run_init_process(init_filename);
1427
1428 if (ret && ret != -ENOENT) {
1429 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1430 init_filename, ret);
1431 }
1432
1433 return ret;
1434}
1435
1436static noinline void __init kernel_init_freeable(void);
1437
1438#if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1439bool rodata_enabled __ro_after_init = true;
1440static int __init set_debug_rodata(char *str)
1441{
1442 return strtobool(str, &rodata_enabled);
1443}
1444__setup("rodata=", set_debug_rodata);
1445#endif
1446
1447#ifdef CONFIG_STRICT_KERNEL_RWX
1448static void mark_readonly(void)
1449{
1450 if (rodata_enabled) {
1451 /*
1452 * load_module() results in W+X mappings, which are cleaned
1453 * up with call_rcu(). Let's make sure that queued work is
1454 * flushed so that we don't hit false positives looking for
1455 * insecure pages which are W+X.
1456 */
1457 rcu_barrier();
1458 mark_rodata_ro();
1459 rodata_test();
1460 } else
1461 pr_info("Kernel memory protection disabled.\n");
1462}
1463#elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1464static inline void mark_readonly(void)
1465{
1466 pr_warn("Kernel memory protection not selected by kernel config.\n");
1467}
1468#else
1469static inline void mark_readonly(void)
1470{
1471 pr_warn("This architecture does not have kernel memory protection.\n");
1472}
1473#endif
1474
1475void __weak free_initmem(void)
1476{
1477 free_initmem_default(POISON_FREE_INITMEM);
1478}
1479
1480static int __ref kernel_init(void *unused)
1481{
1482 int ret;
1483
1484 /*
1485 * Wait until kthreadd is all set-up.
1486 */
1487 wait_for_completion(&kthreadd_done);
1488
1489 kernel_init_freeable();
1490 /* need to finish all async __init code before freeing the memory */
1491 async_synchronize_full();
1492 kprobe_free_init_mem();
1493 ftrace_free_init_mem();
1494 kgdb_free_init_mem();
1495 free_initmem();
1496 mark_readonly();
1497
1498 /*
1499 * Kernel mappings are now finalized - update the userspace page-table
1500 * to finalize PTI.
1501 */
1502 pti_finalize();
1503
1504 system_state = SYSTEM_RUNNING;
1505 numa_default_policy();
1506
1507 rcu_end_inkernel_boot();
1508
1509 do_sysctl_args();
1510
1511 if (ramdisk_execute_command) {
1512 ret = run_init_process(ramdisk_execute_command);
1513 if (!ret)
1514 return 0;
1515 pr_err("Failed to execute %s (error %d)\n",
1516 ramdisk_execute_command, ret);
1517 }
1518
1519 /*
1520 * We try each of these until one succeeds.
1521 *
1522 * The Bourne shell can be used instead of init if we are
1523 * trying to recover a really broken machine.
1524 */
1525 if (execute_command) {
1526 ret = run_init_process(execute_command);
1527 if (!ret)
1528 return 0;
1529 panic("Requested init %s failed (error %d).",
1530 execute_command, ret);
1531 }
1532
1533 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1534 ret = run_init_process(CONFIG_DEFAULT_INIT);
1535 if (ret)
1536 pr_err("Default init %s failed (error %d)\n",
1537 CONFIG_DEFAULT_INIT, ret);
1538 else
1539 return 0;
1540 }
1541
1542 if (!try_to_run_init_process("/sbin/init") ||
1543 !try_to_run_init_process("/etc/init") ||
1544 !try_to_run_init_process("/bin/init") ||
1545 !try_to_run_init_process("/bin/sh"))
1546 return 0;
1547
1548 panic("No working init found. Try passing init= option to kernel. "
1549 "See Linux Documentation/admin-guide/init.rst for guidance.");
1550}
1551
1552/* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1553void __init console_on_rootfs(void)
1554{
1555 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1556
1557 if (IS_ERR(file)) {
1558 pr_err("Warning: unable to open an initial console.\n");
1559 return;
1560 }
1561 init_dup(file);
1562 init_dup(file);
1563 init_dup(file);
1564 fput(file);
1565}
1566
1567static noinline void __init kernel_init_freeable(void)
1568{
1569 /* Now the scheduler is fully set up and can do blocking allocations */
1570 gfp_allowed_mask = __GFP_BITS_MASK;
1571
1572 /*
1573 * init can allocate pages on any node
1574 */
1575 set_mems_allowed(node_states[N_MEMORY]);
1576
1577 cad_pid = get_pid(task_pid(current));
1578
1579 smp_prepare_cpus(setup_max_cpus);
1580
1581 workqueue_init();
1582
1583 init_mm_internals();
1584
1585 rcu_init_tasks_generic();
1586 do_pre_smp_initcalls();
1587 lockup_detector_init();
1588
1589 smp_init();
1590 sched_init_smp();
1591
1592 padata_init();
1593 page_alloc_init_late();
1594 /* Initialize page ext after all struct pages are initialized. */
1595 page_ext_init();
1596
1597 do_basic_setup();
1598
1599 kunit_run_all_tests();
1600
1601 wait_for_initramfs();
1602 console_on_rootfs();
1603
1604 /*
1605 * check if there is an early userspace init. If yes, let it do all
1606 * the work
1607 */
1608 if (init_eaccess(ramdisk_execute_command) != 0) {
1609 ramdisk_execute_command = NULL;
1610 prepare_namespace();
1611 }
1612
1613 /*
1614 * Ok, we have completed the initial bootup, and
1615 * we're essentially up and running. Get rid of the
1616 * initmem segments and start the user-mode stuff..
1617 *
1618 * rootfs is available now, try loading the public keys
1619 * and default modules
1620 */
1621
1622 integrity_load_keys();
1623}
1/*
2 * linux/init/main.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 *
6 * GK 2/5/95 - Changed to support mounting root fs via NFS
7 * Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
8 * Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
9 * Simplified starting of init: Michael A. Griffith <grif@acm.org>
10 */
11
12#define DEBUG /* Enable initcall_debug */
13
14#include <linux/types.h>
15#include <linux/module.h>
16#include <linux/proc_fs.h>
17#include <linux/kernel.h>
18#include <linux/syscalls.h>
19#include <linux/stackprotector.h>
20#include <linux/string.h>
21#include <linux/ctype.h>
22#include <linux/delay.h>
23#include <linux/ioport.h>
24#include <linux/init.h>
25#include <linux/initrd.h>
26#include <linux/bootmem.h>
27#include <linux/acpi.h>
28#include <linux/tty.h>
29#include <linux/percpu.h>
30#include <linux/kmod.h>
31#include <linux/vmalloc.h>
32#include <linux/kernel_stat.h>
33#include <linux/start_kernel.h>
34#include <linux/security.h>
35#include <linux/smp.h>
36#include <linux/profile.h>
37#include <linux/rcupdate.h>
38#include <linux/moduleparam.h>
39#include <linux/kallsyms.h>
40#include <linux/writeback.h>
41#include <linux/cpu.h>
42#include <linux/cpuset.h>
43#include <linux/cgroup.h>
44#include <linux/efi.h>
45#include <linux/tick.h>
46#include <linux/interrupt.h>
47#include <linux/taskstats_kern.h>
48#include <linux/delayacct.h>
49#include <linux/unistd.h>
50#include <linux/rmap.h>
51#include <linux/mempolicy.h>
52#include <linux/key.h>
53#include <linux/buffer_head.h>
54#include <linux/page_ext.h>
55#include <linux/debug_locks.h>
56#include <linux/debugobjects.h>
57#include <linux/lockdep.h>
58#include <linux/kmemleak.h>
59#include <linux/pid_namespace.h>
60#include <linux/device.h>
61#include <linux/kthread.h>
62#include <linux/sched.h>
63#include <linux/signal.h>
64#include <linux/idr.h>
65#include <linux/kgdb.h>
66#include <linux/ftrace.h>
67#include <linux/async.h>
68#include <linux/kmemcheck.h>
69#include <linux/sfi.h>
70#include <linux/shmem_fs.h>
71#include <linux/slab.h>
72#include <linux/perf_event.h>
73#include <linux/file.h>
74#include <linux/ptrace.h>
75#include <linux/blkdev.h>
76#include <linux/elevator.h>
77#include <linux/sched_clock.h>
78#include <linux/context_tracking.h>
79#include <linux/random.h>
80#include <linux/list.h>
81#include <linux/integrity.h>
82#include <linux/proc_ns.h>
83#include <linux/io.h>
84
85#include <asm/io.h>
86#include <asm/bugs.h>
87#include <asm/setup.h>
88#include <asm/sections.h>
89#include <asm/cacheflush.h>
90
91static int kernel_init(void *);
92
93extern void init_IRQ(void);
94extern void fork_init(void);
95extern void radix_tree_init(void);
96
97/*
98 * Debug helper: via this flag we know that we are in 'early bootup code'
99 * where only the boot processor is running with IRQ disabled. This means
100 * two things - IRQ must not be enabled before the flag is cleared and some
101 * operations which are not allowed with IRQ disabled are allowed while the
102 * flag is set.
103 */
104bool early_boot_irqs_disabled __read_mostly;
105
106enum system_states system_state __read_mostly;
107EXPORT_SYMBOL(system_state);
108
109/*
110 * Boot command-line arguments
111 */
112#define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
113#define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
114
115extern void time_init(void);
116/* Default late time init is NULL. archs can override this later. */
117void (*__initdata late_time_init)(void);
118
119/* Untouched command line saved by arch-specific code. */
120char __initdata boot_command_line[COMMAND_LINE_SIZE];
121/* Untouched saved command line (eg. for /proc) */
122char *saved_command_line;
123/* Command line for parameter parsing */
124static char *static_command_line;
125/* Command line for per-initcall parameter parsing */
126static char *initcall_command_line;
127
128static char *execute_command;
129static char *ramdisk_execute_command;
130
131/*
132 * Used to generate warnings if static_key manipulation functions are used
133 * before jump_label_init is called.
134 */
135bool static_key_initialized __read_mostly;
136EXPORT_SYMBOL_GPL(static_key_initialized);
137
138/*
139 * If set, this is an indication to the drivers that reset the underlying
140 * device before going ahead with the initialization otherwise driver might
141 * rely on the BIOS and skip the reset operation.
142 *
143 * This is useful if kernel is booting in an unreliable environment.
144 * For ex. kdump situation where previous kernel has crashed, BIOS has been
145 * skipped and devices will be in unknown state.
146 */
147unsigned int reset_devices;
148EXPORT_SYMBOL(reset_devices);
149
150static int __init set_reset_devices(char *str)
151{
152 reset_devices = 1;
153 return 1;
154}
155
156__setup("reset_devices", set_reset_devices);
157
158static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
159const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
160static const char *panic_later, *panic_param;
161
162extern const struct obs_kernel_param __setup_start[], __setup_end[];
163
164static bool __init obsolete_checksetup(char *line)
165{
166 const struct obs_kernel_param *p;
167 bool had_early_param = false;
168
169 p = __setup_start;
170 do {
171 int n = strlen(p->str);
172 if (parameqn(line, p->str, n)) {
173 if (p->early) {
174 /* Already done in parse_early_param?
175 * (Needs exact match on param part).
176 * Keep iterating, as we can have early
177 * params and __setups of same names 8( */
178 if (line[n] == '\0' || line[n] == '=')
179 had_early_param = true;
180 } else if (!p->setup_func) {
181 pr_warn("Parameter %s is obsolete, ignored\n",
182 p->str);
183 return true;
184 } else if (p->setup_func(line + n))
185 return true;
186 }
187 p++;
188 } while (p < __setup_end);
189
190 return had_early_param;
191}
192
193/*
194 * This should be approx 2 Bo*oMips to start (note initial shift), and will
195 * still work even if initially too large, it will just take slightly longer
196 */
197unsigned long loops_per_jiffy = (1<<12);
198EXPORT_SYMBOL(loops_per_jiffy);
199
200static int __init debug_kernel(char *str)
201{
202 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
203 return 0;
204}
205
206static int __init quiet_kernel(char *str)
207{
208 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
209 return 0;
210}
211
212early_param("debug", debug_kernel);
213early_param("quiet", quiet_kernel);
214
215static int __init loglevel(char *str)
216{
217 int newlevel;
218
219 /*
220 * Only update loglevel value when a correct setting was passed,
221 * to prevent blind crashes (when loglevel being set to 0) that
222 * are quite hard to debug
223 */
224 if (get_option(&str, &newlevel)) {
225 console_loglevel = newlevel;
226 return 0;
227 }
228
229 return -EINVAL;
230}
231
232early_param("loglevel", loglevel);
233
234/* Change NUL term back to "=", to make "param" the whole string. */
235static int __init repair_env_string(char *param, char *val,
236 const char *unused, void *arg)
237{
238 if (val) {
239 /* param=val or param="val"? */
240 if (val == param+strlen(param)+1)
241 val[-1] = '=';
242 else if (val == param+strlen(param)+2) {
243 val[-2] = '=';
244 memmove(val-1, val, strlen(val)+1);
245 val--;
246 } else
247 BUG();
248 }
249 return 0;
250}
251
252/* Anything after -- gets handed straight to init. */
253static int __init set_init_arg(char *param, char *val,
254 const char *unused, void *arg)
255{
256 unsigned int i;
257
258 if (panic_later)
259 return 0;
260
261 repair_env_string(param, val, unused, NULL);
262
263 for (i = 0; argv_init[i]; i++) {
264 if (i == MAX_INIT_ARGS) {
265 panic_later = "init";
266 panic_param = param;
267 return 0;
268 }
269 }
270 argv_init[i] = param;
271 return 0;
272}
273
274/*
275 * Unknown boot options get handed to init, unless they look like
276 * unused parameters (modprobe will find them in /proc/cmdline).
277 */
278static int __init unknown_bootoption(char *param, char *val,
279 const char *unused, void *arg)
280{
281 repair_env_string(param, val, unused, NULL);
282
283 /* Handle obsolete-style parameters */
284 if (obsolete_checksetup(param))
285 return 0;
286
287 /* Unused module parameter. */
288 if (strchr(param, '.') && (!val || strchr(param, '.') < val))
289 return 0;
290
291 if (panic_later)
292 return 0;
293
294 if (val) {
295 /* Environment option */
296 unsigned int i;
297 for (i = 0; envp_init[i]; i++) {
298 if (i == MAX_INIT_ENVS) {
299 panic_later = "env";
300 panic_param = param;
301 }
302 if (!strncmp(param, envp_init[i], val - param))
303 break;
304 }
305 envp_init[i] = param;
306 } else {
307 /* Command line option */
308 unsigned int i;
309 for (i = 0; argv_init[i]; i++) {
310 if (i == MAX_INIT_ARGS) {
311 panic_later = "init";
312 panic_param = param;
313 }
314 }
315 argv_init[i] = param;
316 }
317 return 0;
318}
319
320static int __init init_setup(char *str)
321{
322 unsigned int i;
323
324 execute_command = str;
325 /*
326 * In case LILO is going to boot us with default command line,
327 * it prepends "auto" before the whole cmdline which makes
328 * the shell think it should execute a script with such name.
329 * So we ignore all arguments entered _before_ init=... [MJ]
330 */
331 for (i = 1; i < MAX_INIT_ARGS; i++)
332 argv_init[i] = NULL;
333 return 1;
334}
335__setup("init=", init_setup);
336
337static int __init rdinit_setup(char *str)
338{
339 unsigned int i;
340
341 ramdisk_execute_command = str;
342 /* See "auto" comment in init_setup */
343 for (i = 1; i < MAX_INIT_ARGS; i++)
344 argv_init[i] = NULL;
345 return 1;
346}
347__setup("rdinit=", rdinit_setup);
348
349#ifndef CONFIG_SMP
350static const unsigned int setup_max_cpus = NR_CPUS;
351static inline void setup_nr_cpu_ids(void) { }
352static inline void smp_prepare_cpus(unsigned int maxcpus) { }
353#endif
354
355/*
356 * We need to store the untouched command line for future reference.
357 * We also need to store the touched command line since the parameter
358 * parsing is performed in place, and we should allow a component to
359 * store reference of name/value for future reference.
360 */
361static void __init setup_command_line(char *command_line)
362{
363 saved_command_line =
364 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
365 initcall_command_line =
366 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
367 static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0);
368 strcpy(saved_command_line, boot_command_line);
369 strcpy(static_command_line, command_line);
370}
371
372/*
373 * We need to finalize in a non-__init function or else race conditions
374 * between the root thread and the init thread may cause start_kernel to
375 * be reaped by free_initmem before the root thread has proceeded to
376 * cpu_idle.
377 *
378 * gcc-3.4 accidentally inlines this function, so use noinline.
379 */
380
381static __initdata DECLARE_COMPLETION(kthreadd_done);
382
383static noinline void __init_refok rest_init(void)
384{
385 int pid;
386
387 rcu_scheduler_starting();
388 /*
389 * We need to spawn init first so that it obtains pid 1, however
390 * the init task will end up wanting to create kthreads, which, if
391 * we schedule it before we create kthreadd, will OOPS.
392 */
393 kernel_thread(kernel_init, NULL, CLONE_FS);
394 numa_default_policy();
395 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
396 rcu_read_lock();
397 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
398 rcu_read_unlock();
399 complete(&kthreadd_done);
400
401 /*
402 * The boot idle thread must execute schedule()
403 * at least once to get things moving:
404 */
405 init_idle_bootup_task(current);
406 schedule_preempt_disabled();
407 /* Call into cpu_idle with preempt disabled */
408 cpu_startup_entry(CPUHP_ONLINE);
409}
410
411/* Check for early params. */
412static int __init do_early_param(char *param, char *val,
413 const char *unused, void *arg)
414{
415 const struct obs_kernel_param *p;
416
417 for (p = __setup_start; p < __setup_end; p++) {
418 if ((p->early && parameq(param, p->str)) ||
419 (strcmp(param, "console") == 0 &&
420 strcmp(p->str, "earlycon") == 0)
421 ) {
422 if (p->setup_func(val) != 0)
423 pr_warn("Malformed early option '%s'\n", param);
424 }
425 }
426 /* We accept everything at this stage. */
427 return 0;
428}
429
430void __init parse_early_options(char *cmdline)
431{
432 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
433 do_early_param);
434}
435
436/* Arch code calls this early on, or if not, just before other parsing. */
437void __init parse_early_param(void)
438{
439 static int done __initdata;
440 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
441
442 if (done)
443 return;
444
445 /* All fall through to do_early_param. */
446 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
447 parse_early_options(tmp_cmdline);
448 done = 1;
449}
450
451void __init __weak smp_setup_processor_id(void)
452{
453}
454
455# if THREAD_SIZE >= PAGE_SIZE
456void __init __weak thread_info_cache_init(void)
457{
458}
459#endif
460
461/*
462 * Set up kernel memory allocators
463 */
464static void __init mm_init(void)
465{
466 /*
467 * page_ext requires contiguous pages,
468 * bigger than MAX_ORDER unless SPARSEMEM.
469 */
470 page_ext_init_flatmem();
471 mem_init();
472 kmem_cache_init();
473 percpu_init_late();
474 pgtable_init();
475 vmalloc_init();
476 ioremap_huge_init();
477}
478
479asmlinkage __visible void __init start_kernel(void)
480{
481 char *command_line;
482 char *after_dashes;
483
484 set_task_stack_end_magic(&init_task);
485 smp_setup_processor_id();
486 debug_objects_early_init();
487
488 /*
489 * Set up the the initial canary ASAP:
490 */
491 boot_init_stack_canary();
492
493 cgroup_init_early();
494
495 local_irq_disable();
496 early_boot_irqs_disabled = true;
497
498/*
499 * Interrupts are still disabled. Do necessary setups, then
500 * enable them
501 */
502 boot_cpu_init();
503 page_address_init();
504 pr_notice("%s", linux_banner);
505 setup_arch(&command_line);
506 mm_init_cpumask(&init_mm);
507 setup_command_line(command_line);
508 setup_nr_cpu_ids();
509 setup_per_cpu_areas();
510 boot_cpu_state_init();
511 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
512
513 build_all_zonelists(NULL, NULL);
514 page_alloc_init();
515
516 pr_notice("Kernel command line: %s\n", boot_command_line);
517 parse_early_param();
518 after_dashes = parse_args("Booting kernel",
519 static_command_line, __start___param,
520 __stop___param - __start___param,
521 -1, -1, NULL, &unknown_bootoption);
522 if (!IS_ERR_OR_NULL(after_dashes))
523 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
524 NULL, set_init_arg);
525
526 jump_label_init();
527
528 /*
529 * These use large bootmem allocations and must precede
530 * kmem_cache_init()
531 */
532 setup_log_buf(0);
533 pidhash_init();
534 vfs_caches_init_early();
535 sort_main_extable();
536 trap_init();
537 mm_init();
538
539 /*
540 * Set up the scheduler prior starting any interrupts (such as the
541 * timer interrupt). Full topology setup happens at smp_init()
542 * time - but meanwhile we still have a functioning scheduler.
543 */
544 sched_init();
545 /*
546 * Disable preemption - early bootup scheduling is extremely
547 * fragile until we cpu_idle() for the first time.
548 */
549 preempt_disable();
550 if (WARN(!irqs_disabled(),
551 "Interrupts were enabled *very* early, fixing it\n"))
552 local_irq_disable();
553 idr_init_cache();
554 rcu_init();
555
556 /* trace_printk() and trace points may be used after this */
557 trace_init();
558
559 context_tracking_init();
560 radix_tree_init();
561 /* init some links before init_ISA_irqs() */
562 early_irq_init();
563 init_IRQ();
564 tick_init();
565 rcu_init_nohz();
566 init_timers();
567 hrtimers_init();
568 softirq_init();
569 timekeeping_init();
570 time_init();
571 sched_clock_postinit();
572 perf_event_init();
573 profile_init();
574 call_function_init();
575 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
576 early_boot_irqs_disabled = false;
577 local_irq_enable();
578
579 kmem_cache_init_late();
580
581 /*
582 * HACK ALERT! This is early. We're enabling the console before
583 * we've done PCI setups etc, and console_init() must be aware of
584 * this. But we do want output early, in case something goes wrong.
585 */
586 console_init();
587 if (panic_later)
588 panic("Too many boot %s vars at `%s'", panic_later,
589 panic_param);
590
591 lockdep_info();
592
593 /*
594 * Need to run this when irqs are enabled, because it wants
595 * to self-test [hard/soft]-irqs on/off lock inversion bugs
596 * too:
597 */
598 locking_selftest();
599
600#ifdef CONFIG_BLK_DEV_INITRD
601 if (initrd_start && !initrd_below_start_ok &&
602 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
603 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
604 page_to_pfn(virt_to_page((void *)initrd_start)),
605 min_low_pfn);
606 initrd_start = 0;
607 }
608#endif
609 page_ext_init();
610 debug_objects_mem_init();
611 kmemleak_init();
612 setup_per_cpu_pageset();
613 numa_policy_init();
614 if (late_time_init)
615 late_time_init();
616 sched_clock_init();
617 calibrate_delay();
618 pidmap_init();
619 anon_vma_init();
620 acpi_early_init();
621#ifdef CONFIG_X86
622 if (efi_enabled(EFI_RUNTIME_SERVICES))
623 efi_enter_virtual_mode();
624#endif
625#ifdef CONFIG_X86_ESPFIX64
626 /* Should be run before the first non-init thread is created */
627 init_espfix_bsp();
628#endif
629 thread_info_cache_init();
630 cred_init();
631 fork_init();
632 proc_caches_init();
633 buffer_init();
634 key_init();
635 security_init();
636 dbg_late_init();
637 vfs_caches_init();
638 signals_init();
639 /* rootfs populating might need page-writeback */
640 page_writeback_init();
641 proc_root_init();
642 nsfs_init();
643 cpuset_init();
644 cgroup_init();
645 taskstats_init_early();
646 delayacct_init();
647
648 check_bugs();
649
650 acpi_subsystem_init();
651 sfi_init_late();
652
653 if (efi_enabled(EFI_RUNTIME_SERVICES)) {
654 efi_late_init();
655 efi_free_boot_services();
656 }
657
658 ftrace_init();
659
660 /* Do the rest non-__init'ed, we're now alive */
661 rest_init();
662}
663
664/* Call all constructor functions linked into the kernel. */
665static void __init do_ctors(void)
666{
667#ifdef CONFIG_CONSTRUCTORS
668 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
669
670 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
671 (*fn)();
672#endif
673}
674
675bool initcall_debug;
676core_param(initcall_debug, initcall_debug, bool, 0644);
677
678#ifdef CONFIG_KALLSYMS
679struct blacklist_entry {
680 struct list_head next;
681 char *buf;
682};
683
684static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
685
686static int __init initcall_blacklist(char *str)
687{
688 char *str_entry;
689 struct blacklist_entry *entry;
690
691 /* str argument is a comma-separated list of functions */
692 do {
693 str_entry = strsep(&str, ",");
694 if (str_entry) {
695 pr_debug("blacklisting initcall %s\n", str_entry);
696 entry = alloc_bootmem(sizeof(*entry));
697 entry->buf = alloc_bootmem(strlen(str_entry) + 1);
698 strcpy(entry->buf, str_entry);
699 list_add(&entry->next, &blacklisted_initcalls);
700 }
701 } while (str_entry);
702
703 return 0;
704}
705
706static bool __init_or_module initcall_blacklisted(initcall_t fn)
707{
708 struct blacklist_entry *entry;
709 char *fn_name;
710
711 fn_name = kasprintf(GFP_KERNEL, "%pf", fn);
712 if (!fn_name)
713 return false;
714
715 list_for_each_entry(entry, &blacklisted_initcalls, next) {
716 if (!strcmp(fn_name, entry->buf)) {
717 pr_debug("initcall %s blacklisted\n", fn_name);
718 kfree(fn_name);
719 return true;
720 }
721 }
722
723 kfree(fn_name);
724 return false;
725}
726#else
727static int __init initcall_blacklist(char *str)
728{
729 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
730 return 0;
731}
732
733static bool __init_or_module initcall_blacklisted(initcall_t fn)
734{
735 return false;
736}
737#endif
738__setup("initcall_blacklist=", initcall_blacklist);
739
740static int __init_or_module do_one_initcall_debug(initcall_t fn)
741{
742 ktime_t calltime, delta, rettime;
743 unsigned long long duration;
744 int ret;
745
746 printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current));
747 calltime = ktime_get();
748 ret = fn();
749 rettime = ktime_get();
750 delta = ktime_sub(rettime, calltime);
751 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
752 printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n",
753 fn, ret, duration);
754
755 return ret;
756}
757
758int __init_or_module do_one_initcall(initcall_t fn)
759{
760 int count = preempt_count();
761 int ret;
762 char msgbuf[64];
763
764 if (initcall_blacklisted(fn))
765 return -EPERM;
766
767 if (initcall_debug)
768 ret = do_one_initcall_debug(fn);
769 else
770 ret = fn();
771
772 msgbuf[0] = 0;
773
774 if (preempt_count() != count) {
775 sprintf(msgbuf, "preemption imbalance ");
776 preempt_count_set(count);
777 }
778 if (irqs_disabled()) {
779 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
780 local_irq_enable();
781 }
782 WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
783
784 return ret;
785}
786
787
788extern initcall_t __initcall_start[];
789extern initcall_t __initcall0_start[];
790extern initcall_t __initcall1_start[];
791extern initcall_t __initcall2_start[];
792extern initcall_t __initcall3_start[];
793extern initcall_t __initcall4_start[];
794extern initcall_t __initcall5_start[];
795extern initcall_t __initcall6_start[];
796extern initcall_t __initcall7_start[];
797extern initcall_t __initcall_end[];
798
799static initcall_t *initcall_levels[] __initdata = {
800 __initcall0_start,
801 __initcall1_start,
802 __initcall2_start,
803 __initcall3_start,
804 __initcall4_start,
805 __initcall5_start,
806 __initcall6_start,
807 __initcall7_start,
808 __initcall_end,
809};
810
811/* Keep these in sync with initcalls in include/linux/init.h */
812static char *initcall_level_names[] __initdata = {
813 "early",
814 "core",
815 "postcore",
816 "arch",
817 "subsys",
818 "fs",
819 "device",
820 "late",
821};
822
823static void __init do_initcall_level(int level)
824{
825 initcall_t *fn;
826
827 strcpy(initcall_command_line, saved_command_line);
828 parse_args(initcall_level_names[level],
829 initcall_command_line, __start___param,
830 __stop___param - __start___param,
831 level, level,
832 NULL, &repair_env_string);
833
834 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
835 do_one_initcall(*fn);
836}
837
838static void __init do_initcalls(void)
839{
840 int level;
841
842 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
843 do_initcall_level(level);
844}
845
846/*
847 * Ok, the machine is now initialized. None of the devices
848 * have been touched yet, but the CPU subsystem is up and
849 * running, and memory and process management works.
850 *
851 * Now we can finally start doing some real work..
852 */
853static void __init do_basic_setup(void)
854{
855 cpuset_init_smp();
856 shmem_init();
857 driver_init();
858 init_irq_proc();
859 do_ctors();
860 usermodehelper_enable();
861 do_initcalls();
862 random_int_secret_init();
863}
864
865static void __init do_pre_smp_initcalls(void)
866{
867 initcall_t *fn;
868
869 for (fn = __initcall_start; fn < __initcall0_start; fn++)
870 do_one_initcall(*fn);
871}
872
873/*
874 * This function requests modules which should be loaded by default and is
875 * called twice right after initrd is mounted and right before init is
876 * exec'd. If such modules are on either initrd or rootfs, they will be
877 * loaded before control is passed to userland.
878 */
879void __init load_default_modules(void)
880{
881 load_default_elevator_module();
882}
883
884static int run_init_process(const char *init_filename)
885{
886 argv_init[0] = init_filename;
887 return do_execve(getname_kernel(init_filename),
888 (const char __user *const __user *)argv_init,
889 (const char __user *const __user *)envp_init);
890}
891
892static int try_to_run_init_process(const char *init_filename)
893{
894 int ret;
895
896 ret = run_init_process(init_filename);
897
898 if (ret && ret != -ENOENT) {
899 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
900 init_filename, ret);
901 }
902
903 return ret;
904}
905
906static noinline void __init kernel_init_freeable(void);
907
908#ifdef CONFIG_DEBUG_RODATA
909static bool rodata_enabled = true;
910static int __init set_debug_rodata(char *str)
911{
912 return strtobool(str, &rodata_enabled);
913}
914__setup("rodata=", set_debug_rodata);
915
916static void mark_readonly(void)
917{
918 if (rodata_enabled)
919 mark_rodata_ro();
920 else
921 pr_info("Kernel memory protection disabled.\n");
922}
923#else
924static inline void mark_readonly(void)
925{
926 pr_warn("This architecture does not have kernel memory protection.\n");
927}
928#endif
929
930static int __ref kernel_init(void *unused)
931{
932 int ret;
933
934 kernel_init_freeable();
935 /* need to finish all async __init code before freeing the memory */
936 async_synchronize_full();
937 free_initmem();
938 mark_readonly();
939 system_state = SYSTEM_RUNNING;
940 numa_default_policy();
941
942 flush_delayed_fput();
943
944 rcu_end_inkernel_boot();
945
946 if (ramdisk_execute_command) {
947 ret = run_init_process(ramdisk_execute_command);
948 if (!ret)
949 return 0;
950 pr_err("Failed to execute %s (error %d)\n",
951 ramdisk_execute_command, ret);
952 }
953
954 /*
955 * We try each of these until one succeeds.
956 *
957 * The Bourne shell can be used instead of init if we are
958 * trying to recover a really broken machine.
959 */
960 if (execute_command) {
961 ret = run_init_process(execute_command);
962 if (!ret)
963 return 0;
964 panic("Requested init %s failed (error %d).",
965 execute_command, ret);
966 }
967 if (!try_to_run_init_process("/sbin/init") ||
968 !try_to_run_init_process("/etc/init") ||
969 !try_to_run_init_process("/bin/init") ||
970 !try_to_run_init_process("/bin/sh"))
971 return 0;
972
973 panic("No working init found. Try passing init= option to kernel. "
974 "See Linux Documentation/init.txt for guidance.");
975}
976
977static noinline void __init kernel_init_freeable(void)
978{
979 /*
980 * Wait until kthreadd is all set-up.
981 */
982 wait_for_completion(&kthreadd_done);
983
984 /* Now the scheduler is fully set up and can do blocking allocations */
985 gfp_allowed_mask = __GFP_BITS_MASK;
986
987 /*
988 * init can allocate pages on any node
989 */
990 set_mems_allowed(node_states[N_MEMORY]);
991 /*
992 * init can run on any cpu.
993 */
994 set_cpus_allowed_ptr(current, cpu_all_mask);
995
996 cad_pid = task_pid(current);
997
998 smp_prepare_cpus(setup_max_cpus);
999
1000 do_pre_smp_initcalls();
1001 lockup_detector_init();
1002
1003 smp_init();
1004 sched_init_smp();
1005
1006 page_alloc_init_late();
1007
1008 do_basic_setup();
1009
1010 /* Open the /dev/console on the rootfs, this should never fail */
1011 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1012 pr_err("Warning: unable to open an initial console.\n");
1013
1014 (void) sys_dup(0);
1015 (void) sys_dup(0);
1016 /*
1017 * check if there is an early userspace init. If yes, let it do all
1018 * the work
1019 */
1020
1021 if (!ramdisk_execute_command)
1022 ramdisk_execute_command = "/init";
1023
1024 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
1025 ramdisk_execute_command = NULL;
1026 prepare_namespace();
1027 }
1028
1029 /*
1030 * Ok, we have completed the initial bootup, and
1031 * we're essentially up and running. Get rid of the
1032 * initmem segments and start the user-mode stuff..
1033 *
1034 * rootfs is available now, try loading the public keys
1035 * and default modules
1036 */
1037
1038 integrity_load_keys();
1039 load_default_modules();
1040}