Linux 3.10.54
[GitHub/mt8127/android_kernel_alcatel_ttab.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>
b7a52f51 77#include <linux/random.h>
1da177e4
LT
78
79#include <asm/io.h>
80#include <asm/bugs.h>
81#include <asm/setup.h>
a940199f 82#include <asm/sections.h>
37b73c82 83#include <asm/cacheflush.h>
1da177e4 84
1da177e4
LT
85#ifdef CONFIG_X86_LOCAL_APIC
86#include <asm/smp.h>
87#endif
88
aae5f662 89static int kernel_init(void *);
1da177e4
LT
90
91extern void init_IRQ(void);
1da177e4
LT
92extern void fork_init(unsigned long);
93extern void mca_init(void);
94extern void sbus_init(void);
1da177e4 95extern void radix_tree_init(void);
37b73c82
AV
96#ifndef CONFIG_DEBUG_RODATA
97static inline void mark_rodata_ro(void) { }
98#endif
1da177e4
LT
99
100#ifdef CONFIG_TC
101extern void tc_init(void);
102#endif
103
2ce802f6
TH
104/*
105 * Debug helper: via this flag we know that we are in 'early bootup code'
106 * where only the boot processor is running with IRQ disabled. This means
107 * two things - IRQ must not be enabled before the flag is cleared and some
108 * operations which are not allowed with IRQ disabled are allowed while the
109 * flag is set.
110 */
111bool early_boot_irqs_disabled __read_mostly;
112
a6826048 113enum system_states system_state __read_mostly;
1da177e4
LT
114EXPORT_SYMBOL(system_state);
115
116/*
117 * Boot command-line arguments
118 */
119#define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
120#define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
121
122extern void time_init(void);
123/* Default late time init is NULL. archs can override this later. */
d2e3192b 124void (*__initdata late_time_init)(void);
1da177e4
LT
125extern void softirq_init(void);
126
30d7e0d4
ABL
127/* Untouched command line saved by arch-specific code. */
128char __initdata boot_command_line[COMMAND_LINE_SIZE];
129/* Untouched saved command line (eg. for /proc) */
130char *saved_command_line;
131/* Command line for parameter parsing */
132static char *static_command_line;
1da177e4
LT
133
134static char *execute_command;
ffdfc409 135static char *ramdisk_execute_command;
1da177e4 136
8b3b2955
JB
137/*
138 * If set, this is an indication to the drivers that reset the underlying
139 * device before going ahead with the initialization otherwise driver might
140 * rely on the BIOS and skip the reset operation.
141 *
142 * This is useful if kernel is booting in an unreliable environment.
143 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
144 * skipped and devices will be in unknown state.
145 */
146unsigned int reset_devices;
147EXPORT_SYMBOL(reset_devices);
1da177e4 148
7e96287d
VG
149static int __init set_reset_devices(char *str)
150{
151 reset_devices = 1;
152 return 1;
153}
154
155__setup("reset_devices", set_reset_devices);
156
d7627467
DH
157static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
158const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
1da177e4
LT
159static const char *panic_later, *panic_param;
160
914dcaa8 161extern const struct obs_kernel_param __setup_start[], __setup_end[];
1da177e4
LT
162
163static int __init obsolete_checksetup(char *line)
164{
914dcaa8 165 const struct obs_kernel_param *p;
33df0d19 166 int had_early_param = 0;
1da177e4
LT
167
168 p = __setup_start;
169 do {
170 int n = strlen(p->str);
b1e4d20c 171 if (parameqn(line, p->str, n)) {
1da177e4 172 if (p->early) {
33df0d19
RR
173 /* Already done in parse_early_param?
174 * (Needs exact match on param part).
175 * Keep iterating, as we can have early
176 * params and __setups of same names 8( */
1da177e4 177 if (line[n] == '\0' || line[n] == '=')
33df0d19 178 had_early_param = 1;
1da177e4 179 } else if (!p->setup_func) {
ea676e84
AM
180 pr_warn("Parameter %s is obsolete, ignored\n",
181 p->str);
1da177e4
LT
182 return 1;
183 } else if (p->setup_func(line + n))
184 return 1;
185 }
186 p++;
187 } while (p < __setup_end);
33df0d19
RR
188
189 return had_early_param;
1da177e4
LT
190}
191
192/*
193 * This should be approx 2 Bo*oMips to start (note initial shift), and will
194 * still work even if initially too large, it will just take slightly longer
195 */
196unsigned long loops_per_jiffy = (1<<12);
197
198EXPORT_SYMBOL(loops_per_jiffy);
199
200static int __init debug_kernel(char *str)
201{
1da177e4 202 console_loglevel = 10;
f6f21c81 203 return 0;
1da177e4
LT
204}
205
206static int __init quiet_kernel(char *str)
207{
1da177e4 208 console_loglevel = 4;
f6f21c81 209 return 0;
1da177e4
LT
210}
211
f6f21c81
YL
212early_param("debug", debug_kernel);
213early_param("quiet", quiet_kernel);
1da177e4
LT
214
215static int __init loglevel(char *str)
216{
808bf29b
AS
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;
1da177e4
LT
230}
231
f6f21c81 232early_param("loglevel", loglevel);
1da177e4 233
a99cd112 234/* Change NUL term back to "=", to make "param" the whole string. */
eb157427 235static int __init repair_env_string(char *param, char *val, const char *unused)
1da177e4 236{
1da177e4
LT
237 if (val) {
238 /* param=val or param="val"? */
239 if (val == param+strlen(param)+1)
240 val[-1] = '=';
241 else if (val == param+strlen(param)+2) {
242 val[-2] = '=';
243 memmove(val-1, val, strlen(val)+1);
244 val--;
245 } else
246 BUG();
247 }
a99cd112
CM
248 return 0;
249}
250
251/*
252 * Unknown boot options get handed to init, unless they look like
253 * unused parameters (modprobe will find them in /proc/cmdline).
254 */
eb157427 255static int __init unknown_bootoption(char *param, char *val, const char *unused)
a99cd112 256{
eb157427 257 repair_env_string(param, val, unused);
1da177e4
LT
258
259 /* Handle obsolete-style parameters */
260 if (obsolete_checksetup(param))
261 return 0;
262
f066a4f6
RR
263 /* Unused module parameter. */
264 if (strchr(param, '.') && (!val || strchr(param, '.') < val))
1da177e4 265 return 0;
1da177e4
LT
266
267 if (panic_later)
268 return 0;
269
270 if (val) {
271 /* Environment option */
272 unsigned int i;
273 for (i = 0; envp_init[i]; i++) {
274 if (i == MAX_INIT_ENVS) {
275 panic_later = "Too many boot env vars at `%s'";
276 panic_param = param;
277 }
278 if (!strncmp(param, envp_init[i], val - param))
279 break;
280 }
281 envp_init[i] = param;
282 } else {
283 /* Command line option */
284 unsigned int i;
285 for (i = 0; argv_init[i]; i++) {
286 if (i == MAX_INIT_ARGS) {
287 panic_later = "Too many boot init vars at `%s'";
288 panic_param = param;
289 }
290 }
291 argv_init[i] = param;
292 }
293 return 0;
294}
295
296static int __init init_setup(char *str)
297{
298 unsigned int i;
299
300 execute_command = str;
301 /*
302 * In case LILO is going to boot us with default command line,
303 * it prepends "auto" before the whole cmdline which makes
304 * the shell think it should execute a script with such name.
305 * So we ignore all arguments entered _before_ init=... [MJ]
306 */
307 for (i = 1; i < MAX_INIT_ARGS; i++)
308 argv_init[i] = NULL;
309 return 1;
310}
311__setup("init=", init_setup);
312
ffdfc409
OJ
313static int __init rdinit_setup(char *str)
314{
315 unsigned int i;
316
317 ramdisk_execute_command = str;
318 /* See "auto" comment in init_setup */
319 for (i = 1; i < MAX_INIT_ARGS; i++)
320 argv_init[i] = NULL;
321 return 1;
322}
323__setup("rdinit=", rdinit_setup);
324
1da177e4 325#ifndef CONFIG_SMP
34db18a0 326static const unsigned int setup_max_cpus = NR_CPUS;
1da177e4
LT
327#ifdef CONFIG_X86_LOCAL_APIC
328static void __init smp_init(void)
329{
330 APIC_init_uniprocessor();
331}
332#else
333#define smp_init() do { } while (0)
334#endif
335
e0982e90 336static inline void setup_nr_cpu_ids(void) { }
1da177e4 337static inline void smp_prepare_cpus(unsigned int maxcpus) { }
1da177e4
LT
338#endif
339
30d7e0d4
ABL
340/*
341 * We need to store the untouched command line for future reference.
342 * We also need to store the touched command line since the parameter
343 * parsing is performed in place, and we should allow a component to
344 * store reference of name/value for future reference.
345 */
346static void __init setup_command_line(char *command_line)
347{
348 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
349 static_command_line = alloc_bootmem(strlen (command_line)+1);
350 strcpy (saved_command_line, boot_command_line);
351 strcpy (static_command_line, command_line);
352}
353
1da177e4
LT
354/*
355 * We need to finalize in a non-__init function or else race conditions
356 * between the root thread and the init thread may cause start_kernel to
357 * be reaped by free_initmem before the root thread has proceeded to
358 * cpu_idle.
359 *
360 * gcc-3.4 accidentally inlines this function, so use noinline.
361 */
362
b433c3d4
PZ
363static __initdata DECLARE_COMPLETION(kthreadd_done);
364
f99ebf0a 365static noinline void __init_refok rest_init(void)
1da177e4 366{
73c27992
EB
367 int pid;
368
7db905e6 369 rcu_scheduler_starting();
b433c3d4 370 /*
97158569 371 * We need to spawn init first so that it obtains pid 1, however
b433c3d4
PZ
372 * the init task will end up wanting to create kthreads, which, if
373 * we schedule it before we create kthreadd, will OOPS.
374 */
aae5f662 375 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
1da177e4 376 numa_default_policy();
73c27992 377 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
d11c563d 378 rcu_read_lock();
5cd20455 379 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
d11c563d 380 rcu_read_unlock();
b433c3d4 381 complete(&kthreadd_done);
f340c0d1
IM
382
383 /*
384 * The boot idle thread must execute schedule()
1df21055 385 * at least once to get things moving:
f340c0d1 386 */
1df21055 387 init_idle_bootup_task(current);
bd2f5536 388 schedule_preempt_disabled();
5bfb5d69 389 /* Call into cpu_idle with preempt disabled */
a1a04ec3 390 cpu_startup_entry(CPUHP_ONLINE);
1df21055 391}
1da177e4
LT
392
393/* Check for early params. */
9fb48c74 394static int __init do_early_param(char *param, char *val, const char *unused)
1da177e4 395{
914dcaa8 396 const struct obs_kernel_param *p;
1da177e4
LT
397
398 for (p = __setup_start; p < __setup_end; p++) {
b1e4d20c 399 if ((p->early && parameq(param, p->str)) ||
18a8bd94
YL
400 (strcmp(param, "console") == 0 &&
401 strcmp(p->str, "earlycon") == 0)
402 ) {
1da177e4 403 if (p->setup_func(val) != 0)
ea676e84 404 pr_warn("Malformed early option '%s'\n", param);
1da177e4
LT
405 }
406 }
407 /* We accept everything at this stage. */
408 return 0;
409}
410
13977091
MD
411void __init parse_early_options(char *cmdline)
412{
026cee00 413 parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
13977091
MD
414}
415
1da177e4
LT
416/* Arch code calls this early on, or if not, just before other parsing. */
417void __init parse_early_param(void)
418{
419 static __initdata int done = 0;
420 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
421
422 if (done)
423 return;
424
425 /* All fall through to do_early_param. */
30d7e0d4 426 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
13977091 427 parse_early_options(tmp_cmdline);
1da177e4
LT
428 done = 1;
429}
430
431/*
432 * Activate the first processor.
433 */
434
44fd2299
SS
435static void __init boot_cpu_init(void)
436{
437 int cpu = smp_processor_id();
438 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
915441b6 439 set_cpu_online(cpu, true);
933b0618 440 set_cpu_active(cpu, true);
915441b6
RR
441 set_cpu_present(cpu, true);
442 set_cpu_possible(cpu, true);
44fd2299
SS
443}
444
839ad62e 445void __init __weak smp_setup_processor_id(void)
033ab7f8
AM
446{
447}
448
eded09cc 449# if THREAD_SIZE >= PAGE_SIZE
8c9843e5
BH
450void __init __weak thread_info_cache_init(void)
451{
452}
eded09cc 453#endif
8c9843e5 454
444f478f
PE
455/*
456 * Set up kernel memory allocators
457 */
458static void __init mm_init(void)
459{
ca371c0d 460 /*
7fa87ce7
JC
461 * page_cgroup requires contiguous pages,
462 * bigger than MAX_ORDER unless SPARSEMEM.
ca371c0d
KH
463 */
464 page_cgroup_init_flatmem();
444f478f
PE
465 mem_init();
466 kmem_cache_init();
099a19d9 467 percpu_init_late();
c868d550 468 pgtable_cache_init();
444f478f
PE
469 vmalloc_init();
470}
471
1da177e4
LT
472asmlinkage void __init start_kernel(void)
473{
474 char * command_line;
914dcaa8 475 extern const struct kernel_param __start___param[], __stop___param[];
033ab7f8 476
fbb9ce95
IM
477 /*
478 * Need to run as early as possible, to initialize the
479 * lockdep hash:
480 */
481 lockdep_init();
73839c5b 482 smp_setup_processor_id();
3ac7fe5a 483 debug_objects_early_init();
42059429
IM
484
485 /*
486 * Set up the the initial canary ASAP:
487 */
488 boot_init_stack_canary();
489
ddbcc7e8 490 cgroup_init_early();
fbb9ce95
IM
491
492 local_irq_disable();
2ce802f6 493 early_boot_irqs_disabled = true;
fbb9ce95 494
1da177e4
LT
495/*
496 * Interrupts are still disabled. Do necessary setups, then
497 * enable them
498 */
44fd2299 499 boot_cpu_init();
1da177e4 500 page_address_init();
ea676e84 501 pr_notice("%s", linux_banner);
1da177e4 502 setup_arch(&command_line);
cf475ad2 503 mm_init_owner(&init_mm, &init_task);
6345d24d 504 mm_init_cpumask(&init_mm);
30d7e0d4 505 setup_command_line(command_line);
e0982e90 506 setup_nr_cpu_ids();
d6647bdf 507 setup_per_cpu_areas();
44fd2299 508 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
1da177e4 509
9adb62a5 510 build_all_zonelists(NULL, NULL);
83b519e8
PE
511 page_alloc_init();
512
ea676e84 513 pr_notice("Kernel command line: %s\n", boot_command_line);
83b519e8
PE
514 parse_early_param();
515 parse_args("Booting kernel", static_command_line, __start___param,
516 __stop___param - __start___param,
ae82fdb1 517 -1, -1, &unknown_bootoption);
97ce2c88
JF
518
519 jump_label_init();
520
83b519e8
PE
521 /*
522 * These use large bootmem allocations and must precede
523 * kmem_cache_init()
524 */
162a7e75 525 setup_log_buf(0);
83b519e8 526 pidhash_init();
83b519e8
PE
527 vfs_caches_init_early();
528 sort_main_extable();
529 trap_init();
444f478f 530 mm_init();
de03c72c 531
1da177e4
LT
532 /*
533 * Set up the scheduler prior starting any interrupts (such as the
534 * timer interrupt). Full topology setup happens at smp_init()
535 * time - but meanwhile we still have a functioning scheduler.
536 */
537 sched_init();
538 /*
539 * Disable preemption - early bootup scheduling is extremely
540 * fragile until we cpu_idle() for the first time.
541 */
542 preempt_disable();
f91eb62f 543 if (WARN(!irqs_disabled(), "Interrupts were enabled *very* early, fixing it\n"))
c4a68306 544 local_irq_disable();
9f58a205 545 idr_init_cache();
24a24bb6 546 perf_event_init();
1da177e4 547 rcu_init();
d1e43fa5 548 tick_nohz_init();
773e3eb7 549 radix_tree_init();
0b8f1efa
YL
550 /* init some links before init_ISA_irqs() */
551 early_irq_init();
1da177e4 552 init_IRQ();
ad2b1353 553 tick_init();
1da177e4 554 init_timers();
c0a31329 555 hrtimers_init();
1da177e4 556 softirq_init();
ad596171 557 timekeeping_init();
88fecaa2 558 time_init();
93e02814 559 profile_init();
d8ad7d11 560 call_function_init();
f91eb62f 561 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
2ce802f6 562 early_boot_irqs_disabled = false;
93e02814 563 local_irq_enable();
dcce284a 564
7e85ee0c 565 kmem_cache_init_late();
1da177e4
LT
566
567 /*
568 * HACK ALERT! This is early. We're enabling the console before
569 * we've done PCI setups etc, and console_init() must be aware of
570 * this. But we do want output early, in case something goes wrong.
571 */
572 console_init();
573 if (panic_later)
574 panic(panic_later, panic_param);
fbb9ce95
IM
575
576 lockdep_info();
577
9a11b49a
IM
578 /*
579 * Need to run this when irqs are enabled, because it wants
580 * to self-test [hard/soft]-irqs on/off lock inversion bugs
581 * too:
582 */
583 locking_selftest();
584
1da177e4
LT
585#ifdef CONFIG_BLK_DEV_INITRD
586 if (initrd_start && !initrd_below_start_ok &&
bd673c7c 587 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
ea676e84 588 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
bd673c7c
GU
589 page_to_pfn(virt_to_page((void *)initrd_start)),
590 min_low_pfn);
1da177e4
LT
591 initrd_start = 0;
592 }
593#endif
94b6da5a 594 page_cgroup_init();
3ac7fe5a 595 debug_objects_mem_init();
9b090f2d 596 kmemleak_init();
e7c8d5c9 597 setup_per_cpu_pageset();
1da177e4
LT
598 numa_policy_init();
599 if (late_time_init)
600 late_time_init();
fa84e9ee 601 sched_clock_init();
1da177e4
LT
602 calibrate_delay();
603 pidmap_init();
1da177e4 604 anon_vma_init();
11520e5e 605#ifdef CONFIG_X86
83e68189 606 if (efi_enabled(EFI_RUNTIME_SERVICES))
11520e5e 607 efi_enter_virtual_mode();
a7b854c9 608#endif
cd4033c9 609#ifdef CONFIG_X86_ESPFIX64
a7b854c9
PA
610 /* Should be run before the first non-init thread is created */
611 init_espfix_bsp();
11520e5e 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
4a683bf9 663static char msgbuf[64];
4a683bf9 664
e4461271 665static int __init_or_module do_one_initcall_debug(initcall_t fn)
1da177e4 666{
74239072 667 ktime_t calltime, delta, rettime;
30dbb20e
AW
668 unsigned long long duration;
669 int ret;
1da177e4 670
ea676e84 671 pr_debug("calling %pF @ %i\n", fn, task_pid_nr(current));
22c5c03b 672 calltime = ktime_get();
30dbb20e 673 ret = fn();
22c5c03b
KW
674 rettime = ktime_get();
675 delta = ktime_sub(rettime, calltime);
676 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
ea676e84
AM
677 pr_debug("initcall %pF returned %d after %lld usecs\n",
678 fn, ret, duration);
1da177e4 679
22c5c03b
KW
680 return ret;
681}
682
e4461271 683int __init_or_module do_one_initcall(initcall_t fn)
22c5c03b
KW
684{
685 int count = preempt_count();
686 int ret;
687
688 if (initcall_debug)
689 ret = do_one_initcall_debug(fn);
690 else
691 ret = fn();
8f0c45cd 692
e0df154f 693 msgbuf[0] = 0;
e662e1cf 694
e0df154f 695 if (preempt_count() != count) {
bf5d770b 696 sprintf(msgbuf, "preemption imbalance ");
e0df154f 697 preempt_count() = count;
1da177e4 698 }
e0df154f 699 if (irqs_disabled()) {
a76bfd0d 700 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
e0df154f
LT
701 local_irq_enable();
702 }
f91eb62f 703 WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
59f9415f 704
30dbb20e 705 return ret;
e0df154f
LT
706}
707
708
026cee00
PM
709extern initcall_t __initcall_start[];
710extern initcall_t __initcall0_start[];
711extern initcall_t __initcall1_start[];
712extern initcall_t __initcall2_start[];
713extern initcall_t __initcall3_start[];
714extern initcall_t __initcall4_start[];
715extern initcall_t __initcall5_start[];
716extern initcall_t __initcall6_start[];
717extern initcall_t __initcall7_start[];
718extern initcall_t __initcall_end[];
719
720static initcall_t *initcall_levels[] __initdata = {
721 __initcall0_start,
722 __initcall1_start,
723 __initcall2_start,
724 __initcall3_start,
725 __initcall4_start,
726 __initcall5_start,
727 __initcall6_start,
728 __initcall7_start,
729 __initcall_end,
730};
731
96263d28 732/* Keep these in sync with initcalls in include/linux/init.h */
026cee00 733static char *initcall_level_names[] __initdata = {
9fb48c74
JC
734 "early",
735 "core",
736 "postcore",
737 "arch",
738 "subsys",
739 "fs",
740 "device",
741 "late",
026cee00
PM
742};
743
026cee00 744static void __init do_initcall_level(int level)
e0df154f 745{
026cee00 746 extern const struct kernel_param __start___param[], __stop___param[];
196a15b4 747 initcall_t *fn;
e0df154f 748
026cee00
PM
749 strcpy(static_command_line, saved_command_line);
750 parse_args(initcall_level_names[level],
751 static_command_line, __start___param,
752 __stop___param - __start___param,
753 level, level,
eb157427 754 &repair_env_string);
026cee00
PM
755
756 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
196a15b4 757 do_one_initcall(*fn);
1da177e4
LT
758}
759
026cee00
PM
760static void __init do_initcalls(void)
761{
762 int level;
763
19efb72f 764 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
026cee00
PM
765 do_initcall_level(level);
766}
767
1da177e4
LT
768/*
769 * Ok, the machine is now initialized. None of the devices
770 * have been touched yet, but the CPU subsystem is up and
771 * running, and memory and process management works.
772 *
773 * Now we can finally start doing some real work..
774 */
775static void __init do_basic_setup(void)
776{
759ee091 777 cpuset_init_smp();
1da177e4 778 usermodehelper_init();
41ffe5d5 779 shmem_init();
1da177e4 780 driver_init();
b04c3afb 781 init_irq_proc();
b99b87f7 782 do_ctors();
d5767c53 783 usermodehelper_enable();
b0f84374 784 do_initcalls();
b7a52f51 785 random_int_secret_init();
1da177e4
LT
786}
787
7babe8db 788static void __init do_pre_smp_initcalls(void)
c2147a50 789{
196a15b4 790 initcall_t *fn;
c2147a50 791
026cee00 792 for (fn = __initcall_start; fn < __initcall0_start; fn++)
196a15b4 793 do_one_initcall(*fn);
c2147a50
EGM
794}
795
bb813f4c
TH
796/*
797 * This function requests modules which should be loaded by default and is
798 * called twice right after initrd is mounted and right before init is
799 * exec'd. If such modules are on either initrd or rootfs, they will be
800 * loaded before control is passed to userland.
801 */
802void __init load_default_modules(void)
803{
804 load_default_elevator_module();
805}
806
a74fb73c 807static int run_init_process(const char *init_filename)
1da177e4
LT
808{
809 argv_init[0] = init_filename;
ae903caa
AV
810 return do_execve(init_filename,
811 (const char __user *const __user *)argv_init,
812 (const char __user *const __user *)envp_init);
1da177e4
LT
813}
814
f80b0c90 815static noinline void __init kernel_init_freeable(void);
d6b21238
AV
816
817static int __ref kernel_init(void *unused)
ee5bfa64 818{
d6b21238 819 kernel_init_freeable();
22a9d645
AV
820 /* need to finish all async __init code before freeing the memory */
821 async_synchronize_full();
ee5bfa64 822 free_initmem();
ee5bfa64
VG
823 mark_rodata_ro();
824 system_state = SYSTEM_RUNNING;
825 numa_default_policy();
826
4a9d4b02 827 flush_delayed_fput();
fae5fa44 828
ee5bfa64 829 if (ramdisk_execute_command) {
a74fb73c
AV
830 if (!run_init_process(ramdisk_execute_command))
831 return 0;
ea676e84 832 pr_err("Failed to execute %s\n", ramdisk_execute_command);
ee5bfa64
VG
833 }
834
835 /*
836 * We try each of these until one succeeds.
837 *
838 * The Bourne shell can be used instead of init if we are
839 * trying to recover a really broken machine.
840 */
841 if (execute_command) {
a74fb73c
AV
842 if (!run_init_process(execute_command))
843 return 0;
ea676e84
AM
844 pr_err("Failed to execute %s. Attempting defaults...\n",
845 execute_command);
ee5bfa64 846 }
a74fb73c
AV
847 if (!run_init_process("/sbin/init") ||
848 !run_init_process("/etc/init") ||
849 !run_init_process("/bin/init") ||
850 !run_init_process("/bin/sh"))
851 return 0;
ee5bfa64 852
9a85b8d6
AM
853 panic("No init found. Try passing init= option to kernel. "
854 "See Linux Documentation/init.txt for guidance.");
ee5bfa64
VG
855}
856
f80b0c90 857static noinline void __init kernel_init_freeable(void)
1da177e4 858{
b433c3d4
PZ
859 /*
860 * Wait until kthreadd is all set-up.
861 */
862 wait_for_completion(&kthreadd_done);
31a67102
LT
863
864 /* Now the scheduler is fully set up and can do blocking allocations */
865 gfp_allowed_mask = __GFP_BITS_MASK;
866
58568d2a
MX
867 /*
868 * init can allocate pages on any node
869 */
3c466d46 870 set_mems_allowed(node_states[N_MEMORY]);
1da177e4
LT
871 /*
872 * init can run on any cpu.
873 */
1a2142af 874 set_cpus_allowed_ptr(current, cpu_all_mask);
1da177e4 875
9ec52099
CLG
876 cad_pid = task_pid(current);
877
ca74a6f8 878 smp_prepare_cpus(setup_max_cpus);
1da177e4
LT
879
880 do_pre_smp_initcalls();
004417a6 881 lockup_detector_init();
1da177e4 882
1da177e4
LT
883 smp_init();
884 sched_init_smp();
885
1da177e4
LT
886 do_basic_setup();
887
2bd3a997
EB
888 /* Open the /dev/console on the rootfs, this should never fail */
889 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
ea676e84 890 pr_err("Warning: unable to open an initial console.\n");
2bd3a997
EB
891
892 (void) sys_dup(0);
893 (void) sys_dup(0);
1da177e4
LT
894 /*
895 * check if there is an early userspace init. If yes, let it do all
896 * the work
897 */
ffdfc409
OJ
898
899 if (!ramdisk_execute_command)
900 ramdisk_execute_command = "/init";
901
902 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
903 ramdisk_execute_command = NULL;
1da177e4 904 prepare_namespace();
ffdfc409 905 }
1da177e4
LT
906
907 /*
908 * Ok, we have completed the initial bootup, and
909 * we're essentially up and running. Get rid of the
910 * initmem segments and start the user-mode stuff..
911 */
bb813f4c
TH
912
913 /* rootfs is available now, try loading default modules */
914 load_default_modules();
1da177e4 915}