blob: 3820323c4c8467c07831a8844532fb85976e9a9f [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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
Linus Torvalds1da177e2005-04-16 15:20:36 -070012#include <linux/types.h>
13#include <linux/module.h>
14#include <linux/proc_fs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070015#include <linux/kernel.h>
16#include <linux/syscalls.h>
17#include <linux/string.h>
18#include <linux/ctype.h>
19#include <linux/delay.h>
20#include <linux/utsname.h>
21#include <linux/ioport.h>
22#include <linux/init.h>
23#include <linux/smp_lock.h>
24#include <linux/initrd.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/bootmem.h>
26#include <linux/tty.h>
27#include <linux/gfp.h>
28#include <linux/percpu.h>
29#include <linux/kmod.h>
30#include <linux/kernel_stat.h>
Rusty Russelld7cd5612006-12-07 02:14:08 +010031#include <linux/start_kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070032#include <linux/security.h>
Jens Axboe3d442232008-06-26 11:21:34 +020033#include <linux/smp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070034#include <linux/workqueue.h>
35#include <linux/profile.h>
36#include <linux/rcupdate.h>
37#include <linux/moduleparam.h>
38#include <linux/kallsyms.h>
39#include <linux/writeback.h>
40#include <linux/cpu.h>
41#include <linux/cpuset.h>
Paul Menageddbcc7e2007-10-18 23:39:30 -070042#include <linux/cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070043#include <linux/efi.h>
Thomas Gleixner906568c2007-02-16 01:28:01 -080044#include <linux/tick.h>
Andrew Morton6168a702007-02-17 21:22:39 -080045#include <linux/interrupt.h>
Shailabh Nagarc7572492006-07-14 00:24:40 -070046#include <linux/taskstats_kern.h>
Shailabh Nagarca74e922006-07-14 00:24:36 -070047#include <linux/delayacct.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070048#include <linux/unistd.h>
49#include <linux/rmap.h>
50#include <linux/mempolicy.h>
51#include <linux/key.h>
Jan Beulich4552d5d2006-06-26 13:57:28 +020052#include <linux/unwind.h>
Adrian Bunkb6cd0b72006-06-27 02:53:54 -070053#include <linux/buffer_head.h>
Ingo Molnar9a11b49a2006-07-03 00:24:33 -070054#include <linux/debug_locks.h>
Thomas Gleixner3ac7fe52008-04-30 00:55:01 -070055#include <linux/debugobjects.h>
Ingo Molnarfbb9ce92006-07-03 00:24:50 -070056#include <linux/lockdep.h>
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -080057#include <linux/pid_namespace.h>
Adrian Bunk1f217822006-12-19 13:01:28 -080058#include <linux/device.h>
Eric W. Biederman73c27992007-05-09 02:34:32 -070059#include <linux/kthread.h>
Adrian Bunke6fe6642007-11-09 22:39:39 +010060#include <linux/sched.h>
Adrian Bunka1c9eea2008-02-06 01:36:44 -080061#include <linux/signal.h>
Akinobu Mita199f0ca2008-04-29 01:03:13 -070062#include <linux/idr.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070063
64#include <asm/io.h>
65#include <asm/bugs.h>
66#include <asm/setup.h>
Andi Kleena9401992005-07-28 21:15:30 -070067#include <asm/sections.h>
Arjan van de Ven37b73c82006-01-06 00:12:01 -080068#include <asm/cacheflush.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070069
Linus Torvalds1da177e2005-04-16 15:20:36 -070070#ifdef CONFIG_X86_LOCAL_APIC
71#include <asm/smp.h>
72#endif
73
74/*
Coywolf Qi Hunt69c99ac2006-01-14 13:21:11 -080075 * This is one of the first .c files built. Error out early if we have compiler
76 * trouble.
Linus Torvalds1da177e2005-04-16 15:20:36 -070077 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070078
Andrew Mortonf1a60db2006-12-06 20:39:59 -080079#if __GNUC__ == 4 && __GNUC_MINOR__ == 1 && __GNUC_PATCHLEVEL__ == 0
80#warning gcc-4.1.0 is known to miscompile the kernel. A different compiler version is recommended.
81#endif
82
Sam Ravnborgaae5f662007-02-26 16:45:41 +010083static int kernel_init(void *);
Linus Torvalds1da177e2005-04-16 15:20:36 -070084
85extern void init_IRQ(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -070086extern void fork_init(unsigned long);
87extern void mca_init(void);
88extern void sbus_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -070089extern void prio_tree_init(void);
90extern void radix_tree_init(void);
91extern void free_initmem(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -070092#ifdef CONFIG_ACPI
93extern void acpi_early_init(void);
94#else
95static inline void acpi_early_init(void) { }
96#endif
Arjan van de Ven37b73c82006-01-06 00:12:01 -080097#ifndef CONFIG_DEBUG_RODATA
98static inline void mark_rodata_ro(void) { }
99#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100
101#ifdef CONFIG_TC
102extern void tc_init(void);
103#endif
104
105enum system_states system_state;
106EXPORT_SYMBOL(system_state);
107
108/*
109 * Boot command-line arguments
110 */
111#define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
112#define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
113
114extern void time_init(void);
115/* Default late time init is NULL. archs can override this later. */
116void (*late_time_init)(void);
117extern void softirq_init(void);
118
Alon Bar-Lev30d7e0d2007-02-12 00:53:52 -0800119/* 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;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125
126static char *execute_command;
Olof Johanssonffdfc402005-09-06 15:17:19 -0700127static char *ramdisk_execute_command;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128
Jan Beulich8b3b2952007-07-15 23:41:07 -0700129#ifdef CONFIG_SMP
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130/* Setup configured maximum number of CPUs to activate */
Andi Kleenca74a6f2008-01-30 13:33:17 +0100131unsigned int __initdata setup_max_cpus = NR_CPUS;
Vivek Goyal7e962872006-09-27 01:50:44 -0700132
133/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134 * Setup routine for controlling SMP activation
135 *
136 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
137 * activation entirely (the MPS table probe still happens, though).
138 *
139 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
140 * greater than 0, limits the maximum number of CPUs activated in
141 * SMP mode to <NUM>.
142 */
Len Brown61ec7562007-08-16 03:34:22 -0400143#ifndef CONFIG_X86_IO_APIC
144static inline void disable_ioapic_setup(void) {};
145#endif
146
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147static int __init nosmp(char *str)
148{
Andi Kleenca74a6f2008-01-30 13:33:17 +0100149 setup_max_cpus = 0;
Len Brown61ec7562007-08-16 03:34:22 -0400150 disable_ioapic_setup();
Jan Beulich8b3b2952007-07-15 23:41:07 -0700151 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152}
153
Jan Beulich8b3b2952007-07-15 23:41:07 -0700154early_param("nosmp", nosmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155
156static int __init maxcpus(char *str)
157{
Andi Kleenca74a6f2008-01-30 13:33:17 +0100158 get_option(&str, &setup_max_cpus);
159 if (setup_max_cpus == 0)
Len Brown61ec7562007-08-16 03:34:22 -0400160 disable_ioapic_setup();
161
162 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163}
164
Hugh Dickins81340972007-08-27 16:02:12 +0100165early_param("maxcpus", maxcpus);
Jan Beulich8b3b2952007-07-15 23:41:07 -0700166#else
Andi Kleenca74a6f2008-01-30 13:33:17 +0100167#define setup_max_cpus NR_CPUS
Jan Beulich8b3b2952007-07-15 23:41:07 -0700168#endif
169
170/*
171 * If set, this is an indication to the drivers that reset the underlying
172 * device before going ahead with the initialization otherwise driver might
173 * rely on the BIOS and skip the reset operation.
174 *
175 * This is useful if kernel is booting in an unreliable environment.
176 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
177 * skipped and devices will be in unknown state.
178 */
179unsigned int reset_devices;
180EXPORT_SYMBOL(reset_devices);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181
Vivek Goyal7e962872006-09-27 01:50:44 -0700182static int __init set_reset_devices(char *str)
183{
184 reset_devices = 1;
185 return 1;
186}
187
188__setup("reset_devices", set_reset_devices);
189
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
191char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
192static const char *panic_later, *panic_param;
193
194extern struct obs_kernel_param __setup_start[], __setup_end[];
195
196static int __init obsolete_checksetup(char *line)
197{
198 struct obs_kernel_param *p;
Rusty Russell33df0d12006-09-26 10:52:32 +0200199 int had_early_param = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200
201 p = __setup_start;
202 do {
203 int n = strlen(p->str);
204 if (!strncmp(line, p->str, n)) {
205 if (p->early) {
Rusty Russell33df0d12006-09-26 10:52:32 +0200206 /* Already done in parse_early_param?
207 * (Needs exact match on param part).
208 * Keep iterating, as we can have early
209 * params and __setups of same names 8( */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210 if (line[n] == '\0' || line[n] == '=')
Rusty Russell33df0d12006-09-26 10:52:32 +0200211 had_early_param = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212 } else if (!p->setup_func) {
213 printk(KERN_WARNING "Parameter %s is obsolete,"
214 " ignored\n", p->str);
215 return 1;
216 } else if (p->setup_func(line + n))
217 return 1;
218 }
219 p++;
220 } while (p < __setup_end);
Rusty Russell33df0d12006-09-26 10:52:32 +0200221
222 return had_early_param;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223}
224
225/*
226 * This should be approx 2 Bo*oMips to start (note initial shift), and will
227 * still work even if initially too large, it will just take slightly longer
228 */
229unsigned long loops_per_jiffy = (1<<12);
230
231EXPORT_SYMBOL(loops_per_jiffy);
232
233static int __init debug_kernel(char *str)
234{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 console_loglevel = 10;
Yinghai Luf6f21c812008-02-08 04:21:58 -0800236 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237}
238
239static int __init quiet_kernel(char *str)
240{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 console_loglevel = 4;
Yinghai Luf6f21c812008-02-08 04:21:58 -0800242 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243}
244
Yinghai Luf6f21c812008-02-08 04:21:58 -0800245early_param("debug", debug_kernel);
246early_param("quiet", quiet_kernel);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247
248static int __init loglevel(char *str)
249{
250 get_option(&str, &console_loglevel);
Alex Riesend9d4fcf2008-03-04 14:28:31 -0800251 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252}
253
Yinghai Luf6f21c812008-02-08 04:21:58 -0800254early_param("loglevel", loglevel);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255
256/*
257 * Unknown boot options get handed to init, unless they look like
258 * failed parameters
259 */
260static int __init unknown_bootoption(char *param, char *val)
261{
262 /* Change NUL term back to "=", to make "param" the whole string. */
263 if (val) {
264 /* param=val or param="val"? */
265 if (val == param+strlen(param)+1)
266 val[-1] = '=';
267 else if (val == param+strlen(param)+2) {
268 val[-2] = '=';
269 memmove(val-1, val, strlen(val)+1);
270 val--;
271 } else
272 BUG();
273 }
274
275 /* Handle obsolete-style parameters */
276 if (obsolete_checksetup(param))
277 return 0;
278
279 /*
Simon Arlott211fee82007-10-20 01:28:29 +0200280 * Preemptive maintenance for "why didn't my misspelled command
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 * line work?"
282 */
283 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
284 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
285 return 0;
286 }
287
288 if (panic_later)
289 return 0;
290
291 if (val) {
292 /* Environment option */
293 unsigned int i;
294 for (i = 0; envp_init[i]; i++) {
295 if (i == MAX_INIT_ENVS) {
296 panic_later = "Too many boot env vars at `%s'";
297 panic_param = param;
298 }
299 if (!strncmp(param, envp_init[i], val - param))
300 break;
301 }
302 envp_init[i] = param;
303 } else {
304 /* Command line option */
305 unsigned int i;
306 for (i = 0; argv_init[i]; i++) {
307 if (i == MAX_INIT_ARGS) {
308 panic_later = "Too many boot init vars at `%s'";
309 panic_param = param;
310 }
311 }
312 argv_init[i] = param;
313 }
314 return 0;
315}
316
Ingo Molnar12d6f212008-01-30 13:33:58 +0100317#ifdef CONFIG_DEBUG_PAGEALLOC
318int __read_mostly debug_pagealloc_enabled = 0;
319#endif
320
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321static int __init init_setup(char *str)
322{
323 unsigned int i;
324
325 execute_command = str;
326 /*
327 * In case LILO is going to boot us with default command line,
328 * it prepends "auto" before the whole cmdline which makes
329 * the shell think it should execute a script with such name.
330 * So we ignore all arguments entered _before_ init=... [MJ]
331 */
332 for (i = 1; i < MAX_INIT_ARGS; i++)
333 argv_init[i] = NULL;
334 return 1;
335}
336__setup("init=", init_setup);
337
Olof Johanssonffdfc402005-09-06 15:17:19 -0700338static int __init rdinit_setup(char *str)
339{
340 unsigned int i;
341
342 ramdisk_execute_command = str;
343 /* See "auto" comment in init_setup */
344 for (i = 1; i < MAX_INIT_ARGS; i++)
345 argv_init[i] = NULL;
346 return 1;
347}
348__setup("rdinit=", rdinit_setup);
349
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350#ifndef CONFIG_SMP
351
352#ifdef CONFIG_X86_LOCAL_APIC
353static void __init smp_init(void)
354{
355 APIC_init_uniprocessor();
356}
357#else
358#define smp_init() do { } while (0)
359#endif
360
361static inline void setup_per_cpu_areas(void) { }
Mike Travise0982e92008-03-26 14:23:48 -0700362static inline void setup_nr_cpu_ids(void) { }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363static inline void smp_prepare_cpus(unsigned int maxcpus) { }
364
365#else
366
Mike Travis321a8e92008-04-04 18:11:02 -0700367#if NR_CPUS > BITS_PER_LONG
368cpumask_t cpu_mask_all __read_mostly = CPU_MASK_ALL;
369EXPORT_SYMBOL(cpu_mask_all);
370#endif
371
Mike Travise0982e92008-03-26 14:23:48 -0700372/* Setup number of possible processor ids */
373int nr_cpu_ids __read_mostly = NR_CPUS;
374EXPORT_SYMBOL(nr_cpu_ids);
375
376/* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
377static void __init setup_nr_cpu_ids(void)
378{
379 int cpu, highest_cpu = 0;
380
381 for_each_possible_cpu(cpu)
382 highest_cpu = cpu;
383
384 nr_cpu_ids = highest_cpu + 1;
385}
386
Mike Travisdd5af902008-01-30 13:33:32 +0100387#ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA
Eric Dumazetb73b4592006-03-23 03:01:07 -0800388unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389
390EXPORT_SYMBOL(__per_cpu_offset);
391
392static void __init setup_per_cpu_areas(void)
393{
394 unsigned long size, i;
395 char *ptr;
Eric Dumazet63872f82006-03-23 03:01:04 -0800396 unsigned long nr_possible_cpus = num_possible_cpus();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397
398 /* Copy section for each CPU (we discard the original) */
Jeremy Fitzhardingeb6e35902007-05-02 19:27:12 +0200399 size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
400 ptr = alloc_bootmem_pages(size * nr_possible_cpus);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401
KAMEZAWA Hiroyuki0a945022006-03-28 01:56:37 -0800402 for_each_possible_cpu(i) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 __per_cpu_offset[i] = ptr - __per_cpu_start;
404 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
Eric Dumazet63872f82006-03-23 03:01:04 -0800405 ptr += size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406 }
407}
Mike Travisdd5af902008-01-30 13:33:32 +0100408#endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409
410/* Called by boot processor to activate the rest. */
411static void __init smp_init(void)
412{
Christoph Lameter53b8a312007-02-20 13:57:51 -0800413 unsigned int cpu;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414
Max Krasnyanskye761b772008-07-15 04:43:49 -0700415 /*
416 * Set up the current CPU as possible to migrate to.
417 * The other ones will be done by cpu_up/cpu_down()
418 */
419 cpu = smp_processor_id();
420 cpu_set(cpu, cpu_active_map);
421
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 /* FIXME: This should be done in userspace --RR */
Christoph Lameter53b8a312007-02-20 13:57:51 -0800423 for_each_present_cpu(cpu) {
Andi Kleenca74a6f2008-01-30 13:33:17 +0100424 if (num_online_cpus() >= setup_max_cpus)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 break;
Christoph Lameter53b8a312007-02-20 13:57:51 -0800426 if (!cpu_online(cpu))
427 cpu_up(cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428 }
429
430 /* Any cleanup work */
431 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
Andi Kleenca74a6f2008-01-30 13:33:17 +0100432 smp_cpus_done(setup_max_cpus);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433}
434
435#endif
436
437/*
Alon Bar-Lev30d7e0d2007-02-12 00:53:52 -0800438 * We need to store the untouched command line for future reference.
439 * We also need to store the touched command line since the parameter
440 * parsing is performed in place, and we should allow a component to
441 * store reference of name/value for future reference.
442 */
443static void __init setup_command_line(char *command_line)
444{
445 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
446 static_command_line = alloc_bootmem(strlen (command_line)+1);
447 strcpy (saved_command_line, boot_command_line);
448 strcpy (static_command_line, command_line);
449}
450
451/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 * We need to finalize in a non-__init function or else race conditions
453 * between the root thread and the init thread may cause start_kernel to
454 * be reaped by free_initmem before the root thread has proceeded to
455 * cpu_idle.
456 *
457 * gcc-3.4 accidentally inlines this function, so use noinline.
458 */
459
Sam Ravnborg92080302007-05-17 20:43:54 +0200460static void noinline __init_refok rest_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461 __releases(kernel_lock)
462{
Eric W. Biederman73c27992007-05-09 02:34:32 -0700463 int pid;
464
Sam Ravnborgaae5f662007-02-26 16:45:41 +0100465 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466 numa_default_policy();
Eric W. Biederman73c27992007-05-09 02:34:32 -0700467 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
Pavel Emelyanov5cd20452008-04-30 00:54:24 -0700468 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 unlock_kernel();
Ingo Molnarf340c0d2005-06-28 16:40:42 +0200470
471 /*
472 * The boot idle thread must execute schedule()
Ingo Molnar1df21052007-07-09 18:51:58 +0200473 * at least once to get things moving:
Ingo Molnarf340c0d2005-06-28 16:40:42 +0200474 */
Ingo Molnar1df21052007-07-09 18:51:58 +0200475 init_idle_bootup_task(current);
Nick Piggin5bfb5d62005-11-08 21:39:01 -0800476 preempt_enable_no_resched();
Ingo Molnarf340c0d2005-06-28 16:40:42 +0200477 schedule();
Nick Piggin5bfb5d62005-11-08 21:39:01 -0800478 preempt_disable();
Ingo Molnarf340c0d2005-06-28 16:40:42 +0200479
Nick Piggin5bfb5d62005-11-08 21:39:01 -0800480 /* Call into cpu_idle with preempt disabled */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 cpu_idle();
Ingo Molnar1df21052007-07-09 18:51:58 +0200482}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483
484/* Check for early params. */
485static int __init do_early_param(char *param, char *val)
486{
487 struct obs_kernel_param *p;
488
489 for (p = __setup_start; p < __setup_end; p++) {
Yinghai Lu18a8bd92007-07-15 23:37:59 -0700490 if ((p->early && strcmp(param, p->str) == 0) ||
491 (strcmp(param, "console") == 0 &&
492 strcmp(p->str, "earlycon") == 0)
493 ) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 if (p->setup_func(val) != 0)
495 printk(KERN_WARNING
496 "Malformed early option '%s'\n", param);
497 }
498 }
499 /* We accept everything at this stage. */
500 return 0;
501}
502
503/* Arch code calls this early on, or if not, just before other parsing. */
504void __init parse_early_param(void)
505{
506 static __initdata int done = 0;
507 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
508
509 if (done)
510 return;
511
512 /* All fall through to do_early_param. */
Alon Bar-Lev30d7e0d2007-02-12 00:53:52 -0800513 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
515 done = 1;
516}
517
518/*
519 * Activate the first processor.
520 */
521
Stas Sergeev44fd2292006-03-23 02:59:44 -0800522static void __init boot_cpu_init(void)
523{
524 int cpu = smp_processor_id();
525 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
526 cpu_set(cpu, cpu_online_map);
527 cpu_set(cpu, cpu_present_map);
528 cpu_set(cpu, cpu_possible_map);
529}
530
Benjamin Herrenschmidt839ad622008-04-18 16:56:18 +1000531void __init __weak smp_setup_processor_id(void)
Andrew Morton033ab7f2006-06-30 01:55:50 -0700532{
533}
534
Benjamin Herrenschmidt8c9843e2008-04-18 16:56:15 +1000535void __init __weak thread_info_cache_init(void)
536{
537}
538
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539asmlinkage void __init start_kernel(void)
540{
541 char * command_line;
542 extern struct kernel_param __start___param[], __stop___param[];
Andrew Morton033ab7f2006-06-30 01:55:50 -0700543
544 smp_setup_processor_id();
545
Ingo Molnarfbb9ce92006-07-03 00:24:50 -0700546 /*
547 * Need to run as early as possible, to initialize the
548 * lockdep hash:
549 */
Andi Kleenc9538ed2006-09-26 10:52:34 +0200550 unwind_init();
Ingo Molnarfbb9ce92006-07-03 00:24:50 -0700551 lockdep_init();
Thomas Gleixner3ac7fe52008-04-30 00:55:01 -0700552 debug_objects_early_init();
Paul Menageddbcc7e2007-10-18 23:39:30 -0700553 cgroup_init_early();
Ingo Molnarfbb9ce92006-07-03 00:24:50 -0700554
555 local_irq_disable();
556 early_boot_irqs_off();
Ingo Molnar243c7622006-07-03 00:25:06 -0700557 early_init_irq_lock_class();
Ingo Molnarfbb9ce92006-07-03 00:24:50 -0700558
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559/*
560 * Interrupts are still disabled. Do necessary setups, then
561 * enable them
562 */
563 lock_kernel();
Thomas Gleixner906568c2007-02-16 01:28:01 -0800564 tick_init();
Stas Sergeev44fd2292006-03-23 02:59:44 -0800565 boot_cpu_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 page_address_init();
567 printk(KERN_NOTICE);
Linus Torvalds89937802006-12-11 09:28:46 -0800568 printk(linux_banner);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569 setup_arch(&command_line);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700570 mm_init_owner(&init_mm, &init_task);
Alon Bar-Lev30d7e0d2007-02-12 00:53:52 -0800571 setup_command_line(command_line);
Jan Beulich690a9732006-10-21 18:37:01 +0200572 unwind_setup();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 setup_per_cpu_areas();
Mike Travise0982e92008-03-26 14:23:48 -0700574 setup_nr_cpu_ids();
Stas Sergeev44fd2292006-03-23 02:59:44 -0800575 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576
577 /*
578 * Set up the scheduler prior starting any interrupts (such as the
579 * timer interrupt). Full topology setup happens at smp_init()
580 * time - but meanwhile we still have a functioning scheduler.
581 */
582 sched_init();
583 /*
584 * Disable preemption - early bootup scheduling is extremely
585 * fragile until we cpu_idle() for the first time.
586 */
587 preempt_disable();
588 build_all_zonelists();
589 page_alloc_init();
Alon Bar-Lev30d7e0d2007-02-12 00:53:52 -0800590 printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 parse_early_param();
Alon Bar-Lev30d7e0d2007-02-12 00:53:52 -0800592 parse_args("Booting kernel", static_command_line, __start___param,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593 __stop___param - __start___param,
594 &unknown_bootoption);
Ard van Breemenc4a68302007-01-05 16:36:19 -0800595 if (!irqs_disabled()) {
596 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
597 "enabled *very* early, fixing it\n");
598 local_irq_disable();
599 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600 sort_main_extable();
601 trap_init();
602 rcu_init();
603 init_IRQ();
604 pidhash_init();
605 init_timers();
Thomas Gleixnerc0a31322006-01-09 20:52:32 -0800606 hrtimers_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607 softirq_init();
john stultzad596172006-06-26 00:25:06 -0700608 timekeeping_init();
john stultz88fecaa2006-07-03 00:24:04 -0700609 time_init();
Peter Zijlstra3e51f332008-05-03 18:29:28 +0200610 sched_clock_init();
Heiko Carstens93e02812006-07-03 00:24:24 -0700611 profile_init();
612 if (!irqs_disabled())
613 printk("start_kernel(): bug: interrupts were enabled early\n");
Ingo Molnarfbb9ce92006-07-03 00:24:50 -0700614 early_boot_irqs_on();
Heiko Carstens93e02812006-07-03 00:24:24 -0700615 local_irq_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616
617 /*
618 * HACK ALERT! This is early. We're enabling the console before
619 * we've done PCI setups etc, and console_init() must be aware of
620 * this. But we do want output early, in case something goes wrong.
621 */
622 console_init();
623 if (panic_later)
624 panic(panic_later, panic_param);
Ingo Molnarfbb9ce92006-07-03 00:24:50 -0700625
626 lockdep_info();
627
Ingo Molnar9a11b49a2006-07-03 00:24:33 -0700628 /*
629 * Need to run this when irqs are enabled, because it wants
630 * to self-test [hard/soft]-irqs on/off lock inversion bugs
631 * too:
632 */
633 locking_selftest();
634
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635#ifdef CONFIG_BLK_DEV_INITRD
636 if (initrd_start && !initrd_below_start_ok &&
Geert Uytterhoevenbd673c72008-07-29 22:33:36 -0700637 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
Geert Uytterhoevenfb6624e2008-07-17 21:16:36 +0200639 "disabling it.\n",
Geert Uytterhoevenbd673c72008-07-29 22:33:36 -0700640 page_to_pfn(virt_to_page((void *)initrd_start)),
641 min_low_pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642 initrd_start = 0;
643 }
644#endif
645 vfs_caches_init_early();
Paul Jacksonc417f022006-01-08 01:02:01 -0800646 cpuset_init_early();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647 mem_init();
Thomas Gleixnera03c2a42008-02-09 23:24:09 +0100648 enable_debug_pagealloc();
Gautham R Shenoyd2219382008-01-25 21:08:01 +0100649 cpu_hotplug_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650 kmem_cache_init();
Thomas Gleixner3ac7fe52008-04-30 00:55:01 -0700651 debug_objects_mem_init();
Akinobu Mita199f0ca2008-04-29 01:03:13 -0700652 idr_init_cache();
Christoph Lametere7c8d5c2005-06-21 17:14:47 -0700653 setup_per_cpu_pageset();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 numa_policy_init();
655 if (late_time_init)
656 late_time_init();
657 calibrate_delay();
658 pidmap_init();
659 pgtable_cache_init();
660 prio_tree_init();
661 anon_vma_init();
662#ifdef CONFIG_X86
663 if (efi_enabled)
664 efi_enter_virtual_mode();
665#endif
Benjamin Herrenschmidt8c9843e2008-04-18 16:56:15 +1000666 thread_info_cache_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667 fork_init(num_physpages);
668 proc_caches_init();
669 buffer_init();
670 unnamed_dev_init();
671 key_init();
672 security_init();
673 vfs_caches_init(num_physpages);
674 radix_tree_init();
675 signals_init();
676 /* rootfs populating might need page-writeback */
677 page_writeback_init();
678#ifdef CONFIG_PROC_FS
679 proc_root_init();
680#endif
Paul Menageddbcc7e2007-10-18 23:39:30 -0700681 cgroup_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682 cpuset_init();
Shailabh Nagarc7572492006-07-14 00:24:40 -0700683 taskstats_init_early();
Shailabh Nagarca74e922006-07-14 00:24:36 -0700684 delayacct_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685
686 check_bugs();
687
688 acpi_early_init(); /* before LAPIC and SMP init */
689
690 /* Do the rest non-__init'ed, we're now alive */
691 rest_init();
692}
693
Arjan van de Ven59f94152008-07-30 12:49:02 -0700694static int initcall_debug;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695
696static int __init initcall_debug_setup(char *str)
697{
698 initcall_debug = 1;
699 return 1;
700}
701__setup("initcall_debug", initcall_debug_setup);
702
Arjan van de Ven59f94152008-07-30 12:49:02 -0700703int do_one_initcall(initcall_t fn)
Linus Torvaldse0df1542008-05-15 18:14:01 -0700704{
705 int count = preempt_count();
706 ktime_t t0, t1, delta;
Cyrill Gorcunova76bfd02008-05-15 13:52:41 -0700707 char msgbuf[64];
Linus Torvaldse0df1542008-05-15 18:14:01 -0700708 int result;
709
710 if (initcall_debug) {
Linus Torvalds96d746c2008-10-03 13:38:07 -0700711 printk("calling %pF\n", fn);
Linus Torvaldse0df1542008-05-15 18:14:01 -0700712 t0 = ktime_get();
713 }
714
715 result = fn();
716
717 if (initcall_debug) {
718 t1 = ktime_get();
719 delta = ktime_sub(t1, t0);
720
Linus Torvalds96d746c2008-10-03 13:38:07 -0700721 printk("initcall %pF returned %d after %Ld msecs\n",
722 fn, result,
Linus Torvaldse0df1542008-05-15 18:14:01 -0700723 (unsigned long long) delta.tv64 >> 20);
724 }
725
726 msgbuf[0] = 0;
727
728 if (result && result != -ENODEV && initcall_debug)
729 sprintf(msgbuf, "error code %d ", result);
730
731 if (preempt_count() != count) {
Cyrill Gorcunova76bfd02008-05-15 13:52:41 -0700732 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
Linus Torvaldse0df1542008-05-15 18:14:01 -0700733 preempt_count() = count;
734 }
735 if (irqs_disabled()) {
Cyrill Gorcunova76bfd02008-05-15 13:52:41 -0700736 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
Linus Torvaldse0df1542008-05-15 18:14:01 -0700737 local_irq_enable();
738 }
739 if (msgbuf[0]) {
Linus Torvalds96d746c2008-10-03 13:38:07 -0700740 printk("initcall %pF returned with %s\n", fn, msgbuf);
Linus Torvaldse0df1542008-05-15 18:14:01 -0700741 }
Arjan van de Ven59f94152008-07-30 12:49:02 -0700742
743 return result;
Linus Torvaldse0df1542008-05-15 18:14:01 -0700744}
745
746
Eduard - Gabriel Munteanuc2147a52008-07-25 19:45:11 -0700747extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748
749static void __init do_initcalls(void)
750{
751 initcall_t *call;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752
Eduard - Gabriel Munteanuc2147a52008-07-25 19:45:11 -0700753 for (call = __early_initcall_end; call < __initcall_end; call++)
Linus Torvaldse0df1542008-05-15 18:14:01 -0700754 do_one_initcall(*call);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755
756 /* Make sure there is no pending stuff from the initcall sequence */
757 flush_scheduled_work();
758}
759
760/*
761 * Ok, the machine is now initialized. None of the devices
762 * have been touched yet, but the CPU subsystem is up and
763 * running, and memory and process management works.
764 *
765 * Now we can finally start doing some real work..
766 */
767static void __init do_basic_setup(void)
768{
Paul E. McKenney4446a362008-05-12 21:21:05 +0200769 rcu_init_sched(); /* needed by module_init stage. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770 /* drivers will send hotplug events */
771 init_workqueues();
772 usermodehelper_init();
773 driver_init();
Eric W. Biedermanb04c3af2007-02-14 00:33:57 -0800774 init_irq_proc();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 do_initcalls();
776}
777
Eduard - Gabriel Munteanu7babe8d2008-07-25 19:45:11 -0700778static void __init do_pre_smp_initcalls(void)
Eduard - Gabriel Munteanuc2147a52008-07-25 19:45:11 -0700779{
780 initcall_t *call;
781
782 for (call = __initcall_start; call < __early_initcall_end; call++)
783 do_one_initcall(*call);
784}
785
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786static void run_init_process(char *init_filename)
787{
788 argv_init[0] = init_filename;
Arnd Bergmann67608562006-10-02 02:18:26 -0700789 kernel_execve(init_filename, argv_init, envp_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790}
791
Vivek Goyalee5bfa62007-02-13 13:26:22 +0100792/* This is a non __init function. Force it to be noinline otherwise gcc
793 * makes it inline to init() and it becomes part of init.text section
794 */
795static int noinline init_post(void)
796{
797 free_initmem();
798 unlock_kernel();
799 mark_rodata_ro();
800 system_state = SYSTEM_RUNNING;
801 numa_default_policy();
802
803 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
804 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
805
806 (void) sys_dup(0);
807 (void) sys_dup(0);
808
Oleg Nesterovfae5fa42008-04-30 00:53:03 -0700809 current->signal->flags |= SIGNAL_UNKILLABLE;
810
Vivek Goyalee5bfa62007-02-13 13:26:22 +0100811 if (ramdisk_execute_command) {
812 run_init_process(ramdisk_execute_command);
813 printk(KERN_WARNING "Failed to execute %s\n",
814 ramdisk_execute_command);
815 }
816
817 /*
818 * We try each of these until one succeeds.
819 *
820 * The Bourne shell can be used instead of init if we are
821 * trying to recover a really broken machine.
822 */
823 if (execute_command) {
824 run_init_process(execute_command);
825 printk(KERN_WARNING "Failed to execute %s. Attempting "
826 "defaults...\n", execute_command);
827 }
828 run_init_process("/sbin/init");
829 run_init_process("/etc/init");
830 run_init_process("/bin/init");
831 run_init_process("/bin/sh");
832
833 panic("No init found. Try passing init= option to kernel.");
834}
835
Sam Ravnborgaae5f662007-02-26 16:45:41 +0100836static int __init kernel_init(void * unused)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837{
838 lock_kernel();
839 /*
840 * init can run on any cpu.
841 */
Mike Travis321a8e92008-04-04 18:11:02 -0700842 set_cpus_allowed_ptr(current, CPU_MASK_ALL_PTR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843 /*
844 * Tell the world that we're going to be the grim
845 * reaper of innocent orphaned children.
846 *
847 * We don't want people to have to make incorrect
848 * assumptions about where in the task array this
849 * can be found.
850 */
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -0800851 init_pid_ns.child_reaper = current;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852
Cedric Le Goater9ec52092006-10-02 02:19:00 -0700853 cad_pid = task_pid(current);
854
Andi Kleenca74a6f2008-01-30 13:33:17 +0100855 smp_prepare_cpus(setup_max_cpus);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856
857 do_pre_smp_initcalls();
858
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 smp_init();
860 sched_init_smp();
861
862 cpuset_init_smp();
863
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 do_basic_setup();
865
866 /*
867 * check if there is an early userspace init. If yes, let it do all
868 * the work
869 */
Olof Johanssonffdfc402005-09-06 15:17:19 -0700870
871 if (!ramdisk_execute_command)
872 ramdisk_execute_command = "/init";
873
874 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
875 ramdisk_execute_command = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876 prepare_namespace();
Olof Johanssonffdfc402005-09-06 15:17:19 -0700877 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878
879 /*
880 * Ok, we have completed the initial bootup, and
881 * we're essentially up and running. Get rid of the
882 * initmem segments and start the user-mode stuff..
883 */
Vivek Goyalee5bfa62007-02-13 13:26:22 +0100884 init_post();
885 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886}