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182 pr_warn("Parameter %s is obsolete, ignored\n",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