]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - init/main.c
Merge branch 'acpi-ec'
[mirror_ubuntu-bionic-kernel.git] / init / main.c
CommitLineData
1da177e4
LT
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
ea676e84
AM
12#define DEBUG /* Enable initcall_debug */
13
1da177e4
LT
14#include <linux/types.h>
15#include <linux/module.h>
16#include <linux/proc_fs.h>
1da177e4
LT
17#include <linux/kernel.h>
18#include <linux/syscalls.h>
9b5609fd 19#include <linux/stackprotector.h>
1da177e4
LT
20#include <linux/string.h>
21#include <linux/ctype.h>
22#include <linux/delay.h>
1da177e4
LT
23#include <linux/ioport.h>
24#include <linux/init.h>
1da177e4 25#include <linux/initrd.h>
1da177e4 26#include <linux/bootmem.h>
4a7a16dc 27#include <linux/acpi.h>
1da177e4 28#include <linux/tty.h>
1da177e4
LT
29#include <linux/percpu.h>
30#include <linux/kmod.h>
db64fe02 31#include <linux/vmalloc.h>
1da177e4 32#include <linux/kernel_stat.h>
d7cd5611 33#include <linux/start_kernel.h>
1da177e4 34#include <linux/security.h>
3d442233 35#include <linux/smp.h>
1da177e4
LT
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>
ddbcc7e8 43#include <linux/cgroup.h>
1da177e4 44#include <linux/efi.h>
906568c9 45#include <linux/tick.h>
6168a702 46#include <linux/interrupt.h>
c757249a 47#include <linux/taskstats_kern.h>
ca74e92b 48#include <linux/delayacct.h>
1da177e4
LT
49#include <linux/unistd.h>
50#include <linux/rmap.h>
51#include <linux/mempolicy.h>
52#include <linux/key.h>
b6cd0b77 53#include <linux/buffer_head.h>
94b6da5a 54#include <linux/page_cgroup.h>
9a11b49a 55#include <linux/debug_locks.h>
3ac7fe5a 56#include <linux/debugobjects.h>
fbb9ce95 57#include <linux/lockdep.h>
3c7b4e6b 58#include <linux/kmemleak.h>
84d73786 59#include <linux/pid_namespace.h>
1f21782e 60#include <linux/device.h>
73c27992 61#include <linux/kthread.h>
e6fe6649 62#include <linux/sched.h>
a1c9eea9 63#include <linux/signal.h>
199f0ca5 64#include <linux/idr.h>
0b4b3827 65#include <linux/kgdb.h>
68bf21aa 66#include <linux/ftrace.h>
22a9d645 67#include <linux/async.h>
dfec072e 68#include <linux/kmemcheck.h>
6ae6996a 69#include <linux/sfi.h>
2b2af54a 70#include <linux/shmem_fs.h>
5a0e3ad6 71#include <linux/slab.h>
24a24bb6 72#include <linux/perf_event.h>
4a9d4b02 73#include <linux/file.h>
a74fb73c 74#include <linux/ptrace.h>
bb813f4c
TH
75#include <linux/blkdev.h>
76#include <linux/elevator.h>
38ff87f7 77#include <linux/sched_clock.h>
65f382fd 78#include <linux/context_tracking.h>
47d06e53 79#include <linux/random.h>
1da177e4
LT
80
81#include <asm/io.h>
82#include <asm/bugs.h>
83#include <asm/setup.h>
a940199f 84#include <asm/sections.h>
37b73c82 85#include <asm/cacheflush.h>
1da177e4 86
1da177e4
LT
87#ifdef CONFIG_X86_LOCAL_APIC
88#include <asm/smp.h>
89#endif
90
aae5f662 91static int kernel_init(void *);
1da177e4
LT
92
93extern void init_IRQ(void);
1da177e4
LT
94extern void fork_init(unsigned long);
95extern void mca_init(void);
96extern void sbus_init(void);
1da177e4 97extern void radix_tree_init(void);
37b73c82
AV
98#ifndef CONFIG_DEBUG_RODATA
99static inline void mark_rodata_ro(void) { }
100#endif
1da177e4
LT
101
102#ifdef CONFIG_TC
103extern void tc_init(void);
104#endif
105
2ce802f6
TH
106/*
107 * Debug helper: via this flag we know that we are in 'early bootup code'
108 * where only the boot processor is running with IRQ disabled. This means
109 * two things - IRQ must not be enabled before the flag is cleared and some
110 * operations which are not allowed with IRQ disabled are allowed while the
111 * flag is set.
112 */
113bool early_boot_irqs_disabled __read_mostly;
114
a6826048 115enum system_states system_state __read_mostly;
1da177e4
LT
116EXPORT_SYMBOL(system_state);
117
118/*
119 * Boot command-line arguments
120 */
121#define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
122#define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
123
124extern void time_init(void);
125/* Default late time init is NULL. archs can override this later. */
d2e3192b 126void (*__initdata late_time_init)(void);
1da177e4 127
30d7e0d4
ABL
128/* Untouched command line saved by arch-specific code. */
129char __initdata boot_command_line[COMMAND_LINE_SIZE];
130/* Untouched saved command line (eg. for /proc) */
131char *saved_command_line;
132/* Command line for parameter parsing */
133static char *static_command_line;
1da177e4
LT
134
135static char *execute_command;
ffdfc409 136static char *ramdisk_execute_command;
1da177e4 137
c4b2c0c5
HFS
138/*
139 * Used to generate warnings if static_key manipulation functions are used
140 * before jump_label_init is called.
141 */
142bool static_key_initialized __read_mostly = false;
143EXPORT_SYMBOL_GPL(static_key_initialized);
144
8b3b2955
JB
145/*
146 * If set, this is an indication to the drivers that reset the underlying
147 * device before going ahead with the initialization otherwise driver might
148 * rely on the BIOS and skip the reset operation.
149 *
150 * This is useful if kernel is booting in an unreliable environment.
151 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
152 * skipped and devices will be in unknown state.
153 */
154unsigned int reset_devices;
155EXPORT_SYMBOL(reset_devices);
1da177e4 156
7e96287d
VG
157static int __init set_reset_devices(char *str)
158{
159 reset_devices = 1;
160 return 1;
161}
162
163__setup("reset_devices", set_reset_devices);
164
d7627467
DH
165static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
166const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
1da177e4
LT
167static const char *panic_later, *panic_param;
168
914dcaa8 169extern const struct obs_kernel_param __setup_start[], __setup_end[];
1da177e4
LT
170
171static int __init obsolete_checksetup(char *line)
172{
914dcaa8 173 const struct obs_kernel_param *p;
33df0d19 174 int had_early_param = 0;
1da177e4
LT
175
176 p = __setup_start;
177 do {
178 int n = strlen(p->str);
b1e4d20c 179 if (parameqn(line, p->str, n)) {
1da177e4 180 if (p->early) {
33df0d19
RR
181 /* Already done in parse_early_param?
182 * (Needs exact match on param part).
183 * Keep iterating, as we can have early
184 * params and __setups of same names 8( */
1da177e4 185 if (line[n] == '\0' || line[n] == '=')
33df0d19 186 had_early_param = 1;
1da177e4 187 } else if (!p->setup_func) {
ea676e84
AM
188 pr_warn("Parameter %s is obsolete, ignored\n",
189 p->str);
1da177e4
LT
190 return 1;
191 } else if (p->setup_func(line + n))
192 return 1;
193 }
194 p++;
195 } while (p < __setup_end);
33df0d19
RR
196
197 return had_early_param;
1da177e4
LT
198}
199
200/*
201 * This should be approx 2 Bo*oMips to start (note initial shift), and will
202 * still work even if initially too large, it will just take slightly longer
203 */
204unsigned long loops_per_jiffy = (1<<12);
205
206EXPORT_SYMBOL(loops_per_jiffy);
207
208static int __init debug_kernel(char *str)
209{
1da177e4 210 console_loglevel = 10;
f6f21c81 211 return 0;
1da177e4
LT
212}
213
214static int __init quiet_kernel(char *str)
215{
1da177e4 216 console_loglevel = 4;
f6f21c81 217 return 0;
1da177e4
LT
218}
219
f6f21c81
YL
220early_param("debug", debug_kernel);
221early_param("quiet", quiet_kernel);
1da177e4
LT
222
223static int __init loglevel(char *str)
224{
808bf29b
AS
225 int newlevel;
226
227 /*
228 * Only update loglevel value when a correct setting was passed,
229 * to prevent blind crashes (when loglevel being set to 0) that
230 * are quite hard to debug
231 */
232 if (get_option(&str, &newlevel)) {
233 console_loglevel = newlevel;
234 return 0;
235 }
236
237 return -EINVAL;
1da177e4
LT
238}
239
f6f21c81 240early_param("loglevel", loglevel);
1da177e4 241
a99cd112 242/* Change NUL term back to "=", to make "param" the whole string. */
eb157427 243static int __init repair_env_string(char *param, char *val, const char *unused)
1da177e4 244{
1da177e4
LT
245 if (val) {
246 /* param=val or param="val"? */
247 if (val == param+strlen(param)+1)
248 val[-1] = '=';
249 else if (val == param+strlen(param)+2) {
250 val[-2] = '=';
251 memmove(val-1, val, strlen(val)+1);
252 val--;
253 } else
254 BUG();
255 }
a99cd112
CM
256 return 0;
257}
258
259/*
260 * Unknown boot options get handed to init, unless they look like
261 * unused parameters (modprobe will find them in /proc/cmdline).
262 */
eb157427 263static int __init unknown_bootoption(char *param, char *val, const char *unused)
a99cd112 264{
eb157427 265 repair_env_string(param, val, unused);
1da177e4
LT
266
267 /* Handle obsolete-style parameters */
268 if (obsolete_checksetup(param))
269 return 0;
270
f066a4f6
RR
271 /* Unused module parameter. */
272 if (strchr(param, '.') && (!val || strchr(param, '.') < val))
1da177e4 273 return 0;
1da177e4
LT
274
275 if (panic_later)
276 return 0;
277
278 if (val) {
279 /* Environment option */
280 unsigned int i;
281 for (i = 0; envp_init[i]; i++) {
282 if (i == MAX_INIT_ENVS) {
283 panic_later = "Too many boot env vars at `%s'";
284 panic_param = param;
285 }
286 if (!strncmp(param, envp_init[i], val - param))
287 break;
288 }
289 envp_init[i] = param;
290 } else {
291 /* Command line option */
292 unsigned int i;
293 for (i = 0; argv_init[i]; i++) {
294 if (i == MAX_INIT_ARGS) {
295 panic_later = "Too many boot init vars at `%s'";
296 panic_param = param;
297 }
298 }
299 argv_init[i] = param;
300 }
301 return 0;
302}
303
304static int __init init_setup(char *str)
305{
306 unsigned int i;
307
308 execute_command = str;
309 /*
310 * In case LILO is going to boot us with default command line,
311 * it prepends "auto" before the whole cmdline which makes
312 * the shell think it should execute a script with such name.
313 * So we ignore all arguments entered _before_ init=... [MJ]
314 */
315 for (i = 1; i < MAX_INIT_ARGS; i++)
316 argv_init[i] = NULL;
317 return 1;
318}
319__setup("init=", init_setup);
320
ffdfc409
OJ
321static int __init rdinit_setup(char *str)
322{
323 unsigned int i;
324
325 ramdisk_execute_command = str;
326 /* See "auto" comment in init_setup */
327 for (i = 1; i < MAX_INIT_ARGS; i++)
328 argv_init[i] = NULL;
329 return 1;
330}
331__setup("rdinit=", rdinit_setup);
332
1da177e4 333#ifndef CONFIG_SMP
34db18a0 334static const unsigned int setup_max_cpus = NR_CPUS;
1da177e4
LT
335#ifdef CONFIG_X86_LOCAL_APIC
336static void __init smp_init(void)
337{
338 APIC_init_uniprocessor();
339}
340#else
341#define smp_init() do { } while (0)
342#endif
343
e0982e90 344static inline void setup_nr_cpu_ids(void) { }
1da177e4 345static inline void smp_prepare_cpus(unsigned int maxcpus) { }
1da177e4
LT
346#endif
347
30d7e0d4
ABL
348/*
349 * We need to store the untouched command line for future reference.
350 * We also need to store the touched command line since the parameter
351 * parsing is performed in place, and we should allow a component to
352 * store reference of name/value for future reference.
353 */
354static void __init setup_command_line(char *command_line)
355{
356 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
357 static_command_line = alloc_bootmem(strlen (command_line)+1);
358 strcpy (saved_command_line, boot_command_line);
359 strcpy (static_command_line, command_line);
360}
361
1da177e4
LT
362/*
363 * We need to finalize in a non-__init function or else race conditions
364 * between the root thread and the init thread may cause start_kernel to
365 * be reaped by free_initmem before the root thread has proceeded to
366 * cpu_idle.
367 *
368 * gcc-3.4 accidentally inlines this function, so use noinline.
369 */
370
b433c3d4
PZ
371static __initdata DECLARE_COMPLETION(kthreadd_done);
372
f99ebf0a 373static noinline void __init_refok rest_init(void)
1da177e4 374{
73c27992
EB
375 int pid;
376
7db905e6 377 rcu_scheduler_starting();
b433c3d4 378 /*
97158569 379 * We need to spawn init first so that it obtains pid 1, however
b433c3d4
PZ
380 * the init task will end up wanting to create kthreads, which, if
381 * we schedule it before we create kthreadd, will OOPS.
382 */
aae5f662 383 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
1da177e4 384 numa_default_policy();
73c27992 385 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
d11c563d 386 rcu_read_lock();
5cd20455 387 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
d11c563d 388 rcu_read_unlock();
b433c3d4 389 complete(&kthreadd_done);
f340c0d1
IM
390
391 /*
392 * The boot idle thread must execute schedule()
1df21055 393 * at least once to get things moving:
f340c0d1 394 */
1df21055 395 init_idle_bootup_task(current);
bd2f5536 396 schedule_preempt_disabled();
5bfb5d69 397 /* Call into cpu_idle with preempt disabled */
a1a04ec3 398 cpu_startup_entry(CPUHP_ONLINE);
1df21055 399}
1da177e4
LT
400
401/* Check for early params. */
9fb48c74 402static int __init do_early_param(char *param, char *val, const char *unused)
1da177e4 403{
914dcaa8 404 const struct obs_kernel_param *p;
1da177e4
LT
405
406 for (p = __setup_start; p < __setup_end; p++) {
b1e4d20c 407 if ((p->early && parameq(param, p->str)) ||
18a8bd94
YL
408 (strcmp(param, "console") == 0 &&
409 strcmp(p->str, "earlycon") == 0)
410 ) {
1da177e4 411 if (p->setup_func(val) != 0)
ea676e84 412 pr_warn("Malformed early option '%s'\n", param);
1da177e4
LT
413 }
414 }
415 /* We accept everything at this stage. */
416 return 0;
417}
418
13977091
MD
419void __init parse_early_options(char *cmdline)
420{
026cee00 421 parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
13977091
MD
422}
423
1da177e4
LT
424/* Arch code calls this early on, or if not, just before other parsing. */
425void __init parse_early_param(void)
426{
427 static __initdata int done = 0;
428 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
429
430 if (done)
431 return;
432
433 /* All fall through to do_early_param. */
30d7e0d4 434 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
13977091 435 parse_early_options(tmp_cmdline);
1da177e4
LT
436 done = 1;
437}
438
439/*
440 * Activate the first processor.
441 */
442
44fd2299
SS
443static void __init boot_cpu_init(void)
444{
445 int cpu = smp_processor_id();
446 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
915441b6 447 set_cpu_online(cpu, true);
933b0618 448 set_cpu_active(cpu, true);
915441b6
RR
449 set_cpu_present(cpu, true);
450 set_cpu_possible(cpu, true);
44fd2299
SS
451}
452
839ad62e 453void __init __weak smp_setup_processor_id(void)
033ab7f8
AM
454{
455}
456
eded09cc 457# if THREAD_SIZE >= PAGE_SIZE
8c9843e5
BH
458void __init __weak thread_info_cache_init(void)
459{
460}
eded09cc 461#endif
8c9843e5 462
444f478f
PE
463/*
464 * Set up kernel memory allocators
465 */
466static void __init mm_init(void)
467{
ca371c0d 468 /*
7fa87ce7
JC
469 * page_cgroup requires contiguous pages,
470 * bigger than MAX_ORDER unless SPARSEMEM.
ca371c0d
KH
471 */
472 page_cgroup_init_flatmem();
444f478f
PE
473 mem_init();
474 kmem_cache_init();
099a19d9 475 percpu_init_late();
c868d550 476 pgtable_cache_init();
444f478f
PE
477 vmalloc_init();
478}
479
1da177e4
LT
480asmlinkage void __init start_kernel(void)
481{
482 char * command_line;
914dcaa8 483 extern const struct kernel_param __start___param[], __stop___param[];
033ab7f8 484
fbb9ce95
IM
485 /*
486 * Need to run as early as possible, to initialize the
487 * lockdep hash:
488 */
489 lockdep_init();
73839c5b 490 smp_setup_processor_id();
3ac7fe5a 491 debug_objects_early_init();
42059429
IM
492
493 /*
494 * Set up the the initial canary ASAP:
495 */
496 boot_init_stack_canary();
497
ddbcc7e8 498 cgroup_init_early();
fbb9ce95
IM
499
500 local_irq_disable();
2ce802f6 501 early_boot_irqs_disabled = true;
fbb9ce95 502
1da177e4
LT
503/*
504 * Interrupts are still disabled. Do necessary setups, then
505 * enable them
506 */
44fd2299 507 boot_cpu_init();
1da177e4 508 page_address_init();
ea676e84 509 pr_notice("%s", linux_banner);
1da177e4 510 setup_arch(&command_line);
cf475ad2 511 mm_init_owner(&init_mm, &init_task);
6345d24d 512 mm_init_cpumask(&init_mm);
30d7e0d4 513 setup_command_line(command_line);
e0982e90 514 setup_nr_cpu_ids();
d6647bdf 515 setup_per_cpu_areas();
44fd2299 516 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
1da177e4 517
9adb62a5 518 build_all_zonelists(NULL, NULL);
83b519e8
PE
519 page_alloc_init();
520
ea676e84 521 pr_notice("Kernel command line: %s\n", boot_command_line);
83b519e8
PE
522 parse_early_param();
523 parse_args("Booting kernel", static_command_line, __start___param,
524 __stop___param - __start___param,
ae82fdb1 525 -1, -1, &unknown_bootoption);
97ce2c88
JF
526
527 jump_label_init();
528
83b519e8
PE
529 /*
530 * These use large bootmem allocations and must precede
531 * kmem_cache_init()
532 */
162a7e75 533 setup_log_buf(0);
83b519e8 534 pidhash_init();
83b519e8
PE
535 vfs_caches_init_early();
536 sort_main_extable();
537 trap_init();
444f478f 538 mm_init();
de03c72c 539
1da177e4
LT
540 /*
541 * Set up the scheduler prior starting any interrupts (such as the
542 * timer interrupt). Full topology setup happens at smp_init()
543 * time - but meanwhile we still have a functioning scheduler.
544 */
545 sched_init();
546 /*
547 * Disable preemption - early bootup scheduling is extremely
548 * fragile until we cpu_idle() for the first time.
549 */
550 preempt_disable();
f91eb62f 551 if (WARN(!irqs_disabled(), "Interrupts were enabled *very* early, fixing it\n"))
c4a68306 552 local_irq_disable();
9f58a205 553 idr_init_cache();
1da177e4 554 rcu_init();
d1e43fa5 555 tick_nohz_init();
65f382fd 556 context_tracking_init();
773e3eb7 557 radix_tree_init();
0b8f1efa
YL
558 /* init some links before init_ISA_irqs() */
559 early_irq_init();
1da177e4 560 init_IRQ();
ad2b1353 561 tick_init();
1da177e4 562 init_timers();
c0a31329 563 hrtimers_init();
1da177e4 564 softirq_init();
ad596171 565 timekeeping_init();
88fecaa2 566 time_init();
38ff87f7 567 sched_clock_postinit();
9e630205 568 perf_event_init();
93e02814 569 profile_init();
d8ad7d11 570 call_function_init();
f91eb62f 571 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
2ce802f6 572 early_boot_irqs_disabled = false;
93e02814 573 local_irq_enable();
dcce284a 574
7e85ee0c 575 kmem_cache_init_late();
1da177e4
LT
576
577 /*
578 * HACK ALERT! This is early. We're enabling the console before
579 * we've done PCI setups etc, and console_init() must be aware of
580 * this. But we do want output early, in case something goes wrong.
581 */
582 console_init();
583 if (panic_later)
584 panic(panic_later, panic_param);
fbb9ce95
IM
585
586 lockdep_info();
587
9a11b49a
IM
588 /*
589 * Need to run this when irqs are enabled, because it wants
590 * to self-test [hard/soft]-irqs on/off lock inversion bugs
591 * too:
592 */
593 locking_selftest();
594
1da177e4
LT
595#ifdef CONFIG_BLK_DEV_INITRD
596 if (initrd_start && !initrd_below_start_ok &&
bd673c7c 597 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
ea676e84 598 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
bd673c7c
GU
599 page_to_pfn(virt_to_page((void *)initrd_start)),
600 min_low_pfn);
1da177e4
LT
601 initrd_start = 0;
602 }
603#endif
94b6da5a 604 page_cgroup_init();
3ac7fe5a 605 debug_objects_mem_init();
9b090f2d 606 kmemleak_init();
e7c8d5c9 607 setup_per_cpu_pageset();
1da177e4
LT
608 numa_policy_init();
609 if (late_time_init)
610 late_time_init();
fa84e9ee 611 sched_clock_init();
1da177e4
LT
612 calibrate_delay();
613 pidmap_init();
1da177e4 614 anon_vma_init();
11520e5e 615#ifdef CONFIG_X86
83e68189 616 if (efi_enabled(EFI_RUNTIME_SERVICES))
11520e5e
LT
617 efi_enter_virtual_mode();
618#endif
8c9843e5 619 thread_info_cache_init();
d84f4f99 620 cred_init();
4481374c 621 fork_init(totalram_pages);
1da177e4
LT
622 proc_caches_init();
623 buffer_init();
1da177e4
LT
624 key_init();
625 security_init();
0b4b3827 626 dbg_late_init();
4481374c 627 vfs_caches_init(totalram_pages);
1da177e4
LT
628 signals_init();
629 /* rootfs populating might need page-writeback */
630 page_writeback_init();
631#ifdef CONFIG_PROC_FS
632 proc_root_init();
633#endif
ddbcc7e8 634 cgroup_init();
1da177e4 635 cpuset_init();
c757249a 636 taskstats_init_early();
ca74e92b 637 delayacct_init();
1da177e4
LT
638
639 check_bugs();
640
641 acpi_early_init(); /* before LAPIC and SMP init */
6ae6996a 642 sfi_init_late();
1da177e4 643
83e68189 644 if (efi_enabled(EFI_RUNTIME_SERVICES)) {
2223af38 645 efi_late_init();
78510792 646 efi_free_boot_services();
2223af38 647 }
78510792 648
68bf21aa
SR
649 ftrace_init();
650
1da177e4
LT
651 /* Do the rest non-__init'ed, we're now alive */
652 rest_init();
653}
654
b99b87f7
PO
655/* Call all constructor functions linked into the kernel. */
656static void __init do_ctors(void)
657{
658#ifdef CONFIG_CONSTRUCTORS
196a15b4 659 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
b99b87f7 660
196a15b4
HS
661 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
662 (*fn)();
b99b87f7
PO
663#endif
664}
665
2329abfa 666bool initcall_debug;
d0ea3d7d 667core_param(initcall_debug, initcall_debug, bool, 0644);
1da177e4 668
e4461271 669static int __init_or_module do_one_initcall_debug(initcall_t fn)
1da177e4 670{
74239072 671 ktime_t calltime, delta, rettime;
30dbb20e
AW
672 unsigned long long duration;
673 int ret;
1da177e4 674
ea676e84 675 pr_debug("calling %pF @ %i\n", fn, task_pid_nr(current));
22c5c03b 676 calltime = ktime_get();
30dbb20e 677 ret = fn();
22c5c03b
KW
678 rettime = ktime_get();
679 delta = ktime_sub(rettime, calltime);
680 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
ea676e84
AM
681 pr_debug("initcall %pF returned %d after %lld usecs\n",
682 fn, ret, duration);
1da177e4 683
22c5c03b
KW
684 return ret;
685}
686
e4461271 687int __init_or_module do_one_initcall(initcall_t fn)
22c5c03b
KW
688{
689 int count = preempt_count();
690 int ret;
ff1c8fac 691 char msgbuf[64];
22c5c03b
KW
692
693 if (initcall_debug)
694 ret = do_one_initcall_debug(fn);
695 else
696 ret = fn();
8f0c45cd 697
e0df154f 698 msgbuf[0] = 0;
e662e1cf 699
e0df154f 700 if (preempt_count() != count) {
bf5d770b 701 sprintf(msgbuf, "preemption imbalance ");
4a2b4b22 702 preempt_count_set(count);
1da177e4 703 }
e0df154f 704 if (irqs_disabled()) {
a76bfd0d 705 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
e0df154f
LT
706 local_irq_enable();
707 }
f91eb62f 708 WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
59f9415f 709
30dbb20e 710 return ret;
e0df154f
LT
711}
712
713
026cee00
PM
714extern initcall_t __initcall_start[];
715extern initcall_t __initcall0_start[];
716extern initcall_t __initcall1_start[];
717extern initcall_t __initcall2_start[];
718extern initcall_t __initcall3_start[];
719extern initcall_t __initcall4_start[];
720extern initcall_t __initcall5_start[];
721extern initcall_t __initcall6_start[];
722extern initcall_t __initcall7_start[];
723extern initcall_t __initcall_end[];
724
725static initcall_t *initcall_levels[] __initdata = {
726 __initcall0_start,
727 __initcall1_start,
728 __initcall2_start,
729 __initcall3_start,
730 __initcall4_start,
731 __initcall5_start,
732 __initcall6_start,
733 __initcall7_start,
734 __initcall_end,
735};
736
96263d28 737/* Keep these in sync with initcalls in include/linux/init.h */
026cee00 738static char *initcall_level_names[] __initdata = {
9fb48c74
JC
739 "early",
740 "core",
741 "postcore",
742 "arch",
743 "subsys",
744 "fs",
745 "device",
746 "late",
026cee00
PM
747};
748
026cee00 749static void __init do_initcall_level(int level)
e0df154f 750{
026cee00 751 extern const struct kernel_param __start___param[], __stop___param[];
196a15b4 752 initcall_t *fn;
e0df154f 753
026cee00
PM
754 strcpy(static_command_line, saved_command_line);
755 parse_args(initcall_level_names[level],
756 static_command_line, __start___param,
757 __stop___param - __start___param,
758 level, level,
eb157427 759 &repair_env_string);
026cee00
PM
760
761 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
196a15b4 762 do_one_initcall(*fn);
1da177e4
LT
763}
764
026cee00
PM
765static void __init do_initcalls(void)
766{
767 int level;
768
19efb72f 769 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
026cee00
PM
770 do_initcall_level(level);
771}
772
1da177e4
LT
773/*
774 * Ok, the machine is now initialized. None of the devices
775 * have been touched yet, but the CPU subsystem is up and
776 * running, and memory and process management works.
777 *
778 * Now we can finally start doing some real work..
779 */
780static void __init do_basic_setup(void)
781{
759ee091 782 cpuset_init_smp();
1da177e4 783 usermodehelper_init();
41ffe5d5 784 shmem_init();
1da177e4 785 driver_init();
b04c3afb 786 init_irq_proc();
b99b87f7 787 do_ctors();
d5767c53 788 usermodehelper_enable();
b0f84374 789 do_initcalls();
47d06e53 790 random_int_secret_init();
1da177e4
LT
791}
792
7babe8db 793static void __init do_pre_smp_initcalls(void)
c2147a50 794{
196a15b4 795 initcall_t *fn;
c2147a50 796
026cee00 797 for (fn = __initcall_start; fn < __initcall0_start; fn++)
196a15b4 798 do_one_initcall(*fn);
c2147a50
EGM
799}
800
bb813f4c
TH
801/*
802 * This function requests modules which should be loaded by default and is
803 * called twice right after initrd is mounted and right before init is
804 * exec'd. If such modules are on either initrd or rootfs, they will be
805 * loaded before control is passed to userland.
806 */
807void __init load_default_modules(void)
808{
809 load_default_elevator_module();
810}
811
a74fb73c 812static int run_init_process(const char *init_filename)
1da177e4
LT
813{
814 argv_init[0] = init_filename;
ae903caa
AV
815 return do_execve(init_filename,
816 (const char __user *const __user *)argv_init,
817 (const char __user *const __user *)envp_init);
1da177e4
LT
818}
819
ba24762b
MO
820static int try_to_run_init_process(const char *init_filename)
821{
822 int ret;
823
824 ret = run_init_process(init_filename);
825
826 if (ret && ret != -ENOENT) {
827 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
828 init_filename, ret);
829 }
830
831 return ret;
832}
833
f80b0c90 834static noinline void __init kernel_init_freeable(void);
d6b21238
AV
835
836static int __ref kernel_init(void *unused)
ee5bfa64 837{
ba24762b
MO
838 int ret;
839
d6b21238 840 kernel_init_freeable();
22a9d645
AV
841 /* need to finish all async __init code before freeing the memory */
842 async_synchronize_full();
ee5bfa64 843 free_initmem();
ee5bfa64
VG
844 mark_rodata_ro();
845 system_state = SYSTEM_RUNNING;
846 numa_default_policy();
847
4a9d4b02 848 flush_delayed_fput();
fae5fa44 849
ee5bfa64 850 if (ramdisk_execute_command) {
ba24762b
MO
851 ret = run_init_process(ramdisk_execute_command);
852 if (!ret)
a74fb73c 853 return 0;
ba24762b
MO
854 pr_err("Failed to execute %s (error %d)\n",
855 ramdisk_execute_command, ret);
ee5bfa64
VG
856 }
857
858 /*
859 * We try each of these until one succeeds.
860 *
861 * The Bourne shell can be used instead of init if we are
862 * trying to recover a really broken machine.
863 */
864 if (execute_command) {
ba24762b
MO
865 ret = run_init_process(execute_command);
866 if (!ret)
a74fb73c 867 return 0;
ba24762b
MO
868 pr_err("Failed to execute %s (error %d). Attempting defaults...\n",
869 execute_command, ret);
ee5bfa64 870 }
ba24762b
MO
871 if (!try_to_run_init_process("/sbin/init") ||
872 !try_to_run_init_process("/etc/init") ||
873 !try_to_run_init_process("/bin/init") ||
874 !try_to_run_init_process("/bin/sh"))
a74fb73c 875 return 0;
ee5bfa64 876
ba24762b 877 panic("No working init found. Try passing init= option to kernel. "
9a85b8d6 878 "See Linux Documentation/init.txt for guidance.");
ee5bfa64
VG
879}
880
f80b0c90 881static noinline void __init kernel_init_freeable(void)
1da177e4 882{
b433c3d4
PZ
883 /*
884 * Wait until kthreadd is all set-up.
885 */
886 wait_for_completion(&kthreadd_done);
31a67102
LT
887
888 /* Now the scheduler is fully set up and can do blocking allocations */
889 gfp_allowed_mask = __GFP_BITS_MASK;
890
58568d2a
MX
891 /*
892 * init can allocate pages on any node
893 */
3c466d46 894 set_mems_allowed(node_states[N_MEMORY]);
1da177e4
LT
895 /*
896 * init can run on any cpu.
897 */
1a2142af 898 set_cpus_allowed_ptr(current, cpu_all_mask);
1da177e4 899
9ec52099
CLG
900 cad_pid = task_pid(current);
901
ca74a6f8 902 smp_prepare_cpus(setup_max_cpus);
1da177e4
LT
903
904 do_pre_smp_initcalls();
004417a6 905 lockup_detector_init();
1da177e4 906
1da177e4
LT
907 smp_init();
908 sched_init_smp();
909
1da177e4
LT
910 do_basic_setup();
911
2bd3a997
EB
912 /* Open the /dev/console on the rootfs, this should never fail */
913 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
ea676e84 914 pr_err("Warning: unable to open an initial console.\n");
2bd3a997
EB
915
916 (void) sys_dup(0);
917 (void) sys_dup(0);
1da177e4
LT
918 /*
919 * check if there is an early userspace init. If yes, let it do all
920 * the work
921 */
ffdfc409
OJ
922
923 if (!ramdisk_execute_command)
924 ramdisk_execute_command = "/init";
925
926 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
927 ramdisk_execute_command = NULL;
1da177e4 928 prepare_namespace();
ffdfc409 929 }
1da177e4
LT
930
931 /*
932 * Ok, we have completed the initial bootup, and
933 * we're essentially up and running. Get rid of the
934 * initmem segments and start the user-mode stuff..
935 */
bb813f4c
TH
936
937 /* rootfs is available now, try loading default modules */
938 load_default_modules();
1da177e4 939}