Commit | Line | Data |
---|---|---|
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 | ||
1da177e4 LT |
12 | #include <linux/types.h> |
13 | #include <linux/module.h> | |
14 | #include <linux/proc_fs.h> | |
1da177e4 LT |
15 | #include <linux/kernel.h> |
16 | #include <linux/syscalls.h> | |
9b5609fd | 17 | #include <linux/stackprotector.h> |
1da177e4 LT |
18 | #include <linux/string.h> |
19 | #include <linux/ctype.h> | |
20 | #include <linux/delay.h> | |
1da177e4 LT |
21 | #include <linux/ioport.h> |
22 | #include <linux/init.h> | |
23 | #include <linux/smp_lock.h> | |
24 | #include <linux/initrd.h> | |
1da177e4 | 25 | #include <linux/bootmem.h> |
4a7a16dc | 26 | #include <linux/acpi.h> |
1da177e4 | 27 | #include <linux/tty.h> |
1da177e4 LT |
28 | #include <linux/percpu.h> |
29 | #include <linux/kmod.h> | |
db64fe02 | 30 | #include <linux/vmalloc.h> |
1da177e4 | 31 | #include <linux/kernel_stat.h> |
d7cd5611 | 32 | #include <linux/start_kernel.h> |
1da177e4 | 33 | #include <linux/security.h> |
3d442233 | 34 | #include <linux/smp.h> |
1da177e4 LT |
35 | #include <linux/workqueue.h> |
36 | #include <linux/profile.h> | |
37 | #include <linux/rcupdate.h> | |
38 | #include <linux/moduleparam.h> | |
39 | #include <linux/kallsyms.h> | |
40 | #include <linux/writeback.h> | |
41 | #include <linux/cpu.h> | |
42 | #include <linux/cpuset.h> | |
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> |
02af61bb | 69 | #include <linux/kmemtrace.h> |
6ae6996a | 70 | #include <linux/sfi.h> |
2b2af54a | 71 | #include <linux/shmem_fs.h> |
5a0e3ad6 | 72 | #include <linux/slab.h> |
3f5ec136 | 73 | #include <trace/boot.h> |
1da177e4 LT |
74 | |
75 | #include <asm/io.h> | |
76 | #include <asm/bugs.h> | |
77 | #include <asm/setup.h> | |
a940199f | 78 | #include <asm/sections.h> |
37b73c82 | 79 | #include <asm/cacheflush.h> |
1da177e4 | 80 | |
1da177e4 LT |
81 | #ifdef CONFIG_X86_LOCAL_APIC |
82 | #include <asm/smp.h> | |
83 | #endif | |
84 | ||
aae5f662 | 85 | static int kernel_init(void *); |
1da177e4 LT |
86 | |
87 | extern void init_IRQ(void); | |
1da177e4 LT |
88 | extern void fork_init(unsigned long); |
89 | extern void mca_init(void); | |
90 | extern void sbus_init(void); | |
1da177e4 LT |
91 | extern void prio_tree_init(void); |
92 | extern void radix_tree_init(void); | |
93 | extern void free_initmem(void); | |
37b73c82 AV |
94 | #ifndef CONFIG_DEBUG_RODATA |
95 | static inline void mark_rodata_ro(void) { } | |
96 | #endif | |
1da177e4 LT |
97 | |
98 | #ifdef CONFIG_TC | |
99 | extern void tc_init(void); | |
100 | #endif | |
101 | ||
a6826048 | 102 | enum system_states system_state __read_mostly; |
1da177e4 LT |
103 | EXPORT_SYMBOL(system_state); |
104 | ||
105 | /* | |
106 | * Boot command-line arguments | |
107 | */ | |
108 | #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT | |
109 | #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT | |
110 | ||
111 | extern void time_init(void); | |
112 | /* Default late time init is NULL. archs can override this later. */ | |
d2e3192b | 113 | void (*__initdata late_time_init)(void); |
1da177e4 LT |
114 | extern void softirq_init(void); |
115 | ||
30d7e0d4 ABL |
116 | /* Untouched command line saved by arch-specific code. */ |
117 | char __initdata boot_command_line[COMMAND_LINE_SIZE]; | |
118 | /* Untouched saved command line (eg. for /proc) */ | |
119 | char *saved_command_line; | |
120 | /* Command line for parameter parsing */ | |
121 | static char *static_command_line; | |
1da177e4 LT |
122 | |
123 | static char *execute_command; | |
ffdfc409 | 124 | static char *ramdisk_execute_command; |
1da177e4 | 125 | |
8b3b2955 | 126 | #ifdef CONFIG_SMP |
1da177e4 | 127 | /* Setup configured maximum number of CPUs to activate */ |
75cbfb97 TR |
128 | unsigned int setup_max_cpus = NR_CPUS; |
129 | EXPORT_SYMBOL(setup_max_cpus); | |
130 | ||
7e96287d | 131 | |
1da177e4 LT |
132 | /* |
133 | * Setup routine for controlling SMP activation | |
134 | * | |
135 | * Command-line option of "nosmp" or "maxcpus=0" will disable SMP | |
136 | * activation entirely (the MPS table probe still happens, though). | |
137 | * | |
138 | * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer | |
139 | * greater than 0, limits the maximum number of CPUs activated in | |
140 | * SMP mode to <NUM>. | |
141 | */ | |
65a4e574 IM |
142 | |
143 | void __weak arch_disable_smp_support(void) { } | |
61ec7567 | 144 | |
1da177e4 LT |
145 | static int __init nosmp(char *str) |
146 | { | |
ca74a6f8 | 147 | setup_max_cpus = 0; |
65a4e574 IM |
148 | arch_disable_smp_support(); |
149 | ||
8b3b2955 | 150 | return 0; |
1da177e4 LT |
151 | } |
152 | ||
8b3b2955 | 153 | early_param("nosmp", nosmp); |
1da177e4 | 154 | |
2b633e3f YL |
155 | /* this is hard limit */ |
156 | static int __init nrcpus(char *str) | |
157 | { | |
158 | int nr_cpus; | |
159 | ||
160 | get_option(&str, &nr_cpus); | |
161 | if (nr_cpus > 0 && nr_cpus < nr_cpu_ids) | |
162 | nr_cpu_ids = nr_cpus; | |
163 | ||
164 | return 0; | |
165 | } | |
166 | ||
167 | early_param("nr_cpus", nrcpus); | |
168 | ||
1da177e4 LT |
169 | static int __init maxcpus(char *str) |
170 | { | |
ca74a6f8 AK |
171 | get_option(&str, &setup_max_cpus); |
172 | if (setup_max_cpus == 0) | |
65a4e574 | 173 | arch_disable_smp_support(); |
61ec7567 LB |
174 | |
175 | return 0; | |
1da177e4 LT |
176 | } |
177 | ||
81340977 | 178 | early_param("maxcpus", maxcpus); |
8b3b2955 | 179 | #else |
7463e633 | 180 | static const unsigned int setup_max_cpus = NR_CPUS; |
8b3b2955 JB |
181 | #endif |
182 | ||
183 | /* | |
184 | * If set, this is an indication to the drivers that reset the underlying | |
185 | * device before going ahead with the initialization otherwise driver might | |
186 | * rely on the BIOS and skip the reset operation. | |
187 | * | |
188 | * This is useful if kernel is booting in an unreliable environment. | |
189 | * For ex. kdump situaiton where previous kernel has crashed, BIOS has been | |
190 | * skipped and devices will be in unknown state. | |
191 | */ | |
192 | unsigned int reset_devices; | |
193 | EXPORT_SYMBOL(reset_devices); | |
1da177e4 | 194 | |
7e96287d VG |
195 | static int __init set_reset_devices(char *str) |
196 | { | |
197 | reset_devices = 1; | |
198 | return 1; | |
199 | } | |
200 | ||
201 | __setup("reset_devices", set_reset_devices); | |
202 | ||
1da177e4 LT |
203 | static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; |
204 | char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; | |
205 | static const char *panic_later, *panic_param; | |
206 | ||
207 | extern struct obs_kernel_param __setup_start[], __setup_end[]; | |
208 | ||
209 | static int __init obsolete_checksetup(char *line) | |
210 | { | |
211 | struct obs_kernel_param *p; | |
33df0d19 | 212 | int had_early_param = 0; |
1da177e4 LT |
213 | |
214 | p = __setup_start; | |
215 | do { | |
216 | int n = strlen(p->str); | |
217 | if (!strncmp(line, p->str, n)) { | |
218 | if (p->early) { | |
33df0d19 RR |
219 | /* Already done in parse_early_param? |
220 | * (Needs exact match on param part). | |
221 | * Keep iterating, as we can have early | |
222 | * params and __setups of same names 8( */ | |
1da177e4 | 223 | if (line[n] == '\0' || line[n] == '=') |
33df0d19 | 224 | had_early_param = 1; |
1da177e4 LT |
225 | } else if (!p->setup_func) { |
226 | printk(KERN_WARNING "Parameter %s is obsolete," | |
227 | " ignored\n", p->str); | |
228 | return 1; | |
229 | } else if (p->setup_func(line + n)) | |
230 | return 1; | |
231 | } | |
232 | p++; | |
233 | } while (p < __setup_end); | |
33df0d19 RR |
234 | |
235 | return had_early_param; | |
1da177e4 LT |
236 | } |
237 | ||
238 | /* | |
239 | * This should be approx 2 Bo*oMips to start (note initial shift), and will | |
240 | * still work even if initially too large, it will just take slightly longer | |
241 | */ | |
242 | unsigned long loops_per_jiffy = (1<<12); | |
243 | ||
244 | EXPORT_SYMBOL(loops_per_jiffy); | |
245 | ||
246 | static int __init debug_kernel(char *str) | |
247 | { | |
1da177e4 | 248 | console_loglevel = 10; |
f6f21c81 | 249 | return 0; |
1da177e4 LT |
250 | } |
251 | ||
252 | static int __init quiet_kernel(char *str) | |
253 | { | |
1da177e4 | 254 | console_loglevel = 4; |
f6f21c81 | 255 | return 0; |
1da177e4 LT |
256 | } |
257 | ||
f6f21c81 YL |
258 | early_param("debug", debug_kernel); |
259 | early_param("quiet", quiet_kernel); | |
1da177e4 LT |
260 | |
261 | static int __init loglevel(char *str) | |
262 | { | |
263 | get_option(&str, &console_loglevel); | |
d9d4fcfe | 264 | return 0; |
1da177e4 LT |
265 | } |
266 | ||
f6f21c81 | 267 | early_param("loglevel", loglevel); |
1da177e4 LT |
268 | |
269 | /* | |
270 | * Unknown boot options get handed to init, unless they look like | |
f066a4f6 | 271 | * unused parameters (modprobe will find them in /proc/cmdline). |
1da177e4 LT |
272 | */ |
273 | static int __init unknown_bootoption(char *param, char *val) | |
274 | { | |
275 | /* Change NUL term back to "=", to make "param" the whole string. */ | |
276 | if (val) { | |
277 | /* param=val or param="val"? */ | |
278 | if (val == param+strlen(param)+1) | |
279 | val[-1] = '='; | |
280 | else if (val == param+strlen(param)+2) { | |
281 | val[-2] = '='; | |
282 | memmove(val-1, val, strlen(val)+1); | |
283 | val--; | |
284 | } else | |
285 | BUG(); | |
286 | } | |
287 | ||
288 | /* Handle obsolete-style parameters */ | |
289 | if (obsolete_checksetup(param)) | |
290 | return 0; | |
291 | ||
f066a4f6 RR |
292 | /* Unused module parameter. */ |
293 | if (strchr(param, '.') && (!val || strchr(param, '.') < val)) | |
1da177e4 | 294 | return 0; |
1da177e4 LT |
295 | |
296 | if (panic_later) | |
297 | return 0; | |
298 | ||
299 | if (val) { | |
300 | /* Environment option */ | |
301 | unsigned int i; | |
302 | for (i = 0; envp_init[i]; i++) { | |
303 | if (i == MAX_INIT_ENVS) { | |
304 | panic_later = "Too many boot env vars at `%s'"; | |
305 | panic_param = param; | |
306 | } | |
307 | if (!strncmp(param, envp_init[i], val - param)) | |
308 | break; | |
309 | } | |
310 | envp_init[i] = param; | |
311 | } else { | |
312 | /* Command line option */ | |
313 | unsigned int i; | |
314 | for (i = 0; argv_init[i]; i++) { | |
315 | if (i == MAX_INIT_ARGS) { | |
316 | panic_later = "Too many boot init vars at `%s'"; | |
317 | panic_param = param; | |
318 | } | |
319 | } | |
320 | argv_init[i] = param; | |
321 | } | |
322 | return 0; | |
323 | } | |
324 | ||
12d6f21e IM |
325 | #ifdef CONFIG_DEBUG_PAGEALLOC |
326 | int __read_mostly debug_pagealloc_enabled = 0; | |
327 | #endif | |
328 | ||
1da177e4 LT |
329 | static int __init init_setup(char *str) |
330 | { | |
331 | unsigned int i; | |
332 | ||
333 | execute_command = str; | |
334 | /* | |
335 | * In case LILO is going to boot us with default command line, | |
336 | * it prepends "auto" before the whole cmdline which makes | |
337 | * the shell think it should execute a script with such name. | |
338 | * So we ignore all arguments entered _before_ init=... [MJ] | |
339 | */ | |
340 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
341 | argv_init[i] = NULL; | |
342 | return 1; | |
343 | } | |
344 | __setup("init=", init_setup); | |
345 | ||
ffdfc409 OJ |
346 | static int __init rdinit_setup(char *str) |
347 | { | |
348 | unsigned int i; | |
349 | ||
350 | ramdisk_execute_command = str; | |
351 | /* See "auto" comment in init_setup */ | |
352 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
353 | argv_init[i] = NULL; | |
354 | return 1; | |
355 | } | |
356 | __setup("rdinit=", rdinit_setup); | |
357 | ||
1da177e4 LT |
358 | #ifndef CONFIG_SMP |
359 | ||
360 | #ifdef CONFIG_X86_LOCAL_APIC | |
361 | static void __init smp_init(void) | |
362 | { | |
363 | APIC_init_uniprocessor(); | |
364 | } | |
365 | #else | |
366 | #define smp_init() do { } while (0) | |
367 | #endif | |
368 | ||
e0982e90 | 369 | static inline void setup_nr_cpu_ids(void) { } |
1da177e4 LT |
370 | static inline void smp_prepare_cpus(unsigned int maxcpus) { } |
371 | ||
372 | #else | |
373 | ||
e0982e90 MT |
374 | /* Setup number of possible processor ids */ |
375 | int nr_cpu_ids __read_mostly = NR_CPUS; | |
376 | EXPORT_SYMBOL(nr_cpu_ids); | |
377 | ||
378 | /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */ | |
379 | static void __init setup_nr_cpu_ids(void) | |
380 | { | |
e0c0ba73 | 381 | nr_cpu_ids = find_last_bit(cpumask_bits(cpu_possible_mask),NR_CPUS) + 1; |
e0982e90 MT |
382 | } |
383 | ||
1da177e4 LT |
384 | /* Called by boot processor to activate the rest. */ |
385 | static void __init smp_init(void) | |
386 | { | |
53b8a315 | 387 | unsigned int cpu; |
1da177e4 LT |
388 | |
389 | /* FIXME: This should be done in userspace --RR */ | |
53b8a315 | 390 | for_each_present_cpu(cpu) { |
ca74a6f8 | 391 | if (num_online_cpus() >= setup_max_cpus) |
1da177e4 | 392 | break; |
53b8a315 CL |
393 | if (!cpu_online(cpu)) |
394 | cpu_up(cpu); | |
1da177e4 LT |
395 | } |
396 | ||
397 | /* Any cleanup work */ | |
398 | printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus()); | |
ca74a6f8 | 399 | smp_cpus_done(setup_max_cpus); |
1da177e4 LT |
400 | } |
401 | ||
402 | #endif | |
403 | ||
30d7e0d4 ABL |
404 | /* |
405 | * We need to store the untouched command line for future reference. | |
406 | * We also need to store the touched command line since the parameter | |
407 | * parsing is performed in place, and we should allow a component to | |
408 | * store reference of name/value for future reference. | |
409 | */ | |
410 | static void __init setup_command_line(char *command_line) | |
411 | { | |
412 | saved_command_line = alloc_bootmem(strlen (boot_command_line)+1); | |
413 | static_command_line = alloc_bootmem(strlen (command_line)+1); | |
414 | strcpy (saved_command_line, boot_command_line); | |
415 | strcpy (static_command_line, command_line); | |
416 | } | |
417 | ||
1da177e4 LT |
418 | /* |
419 | * We need to finalize in a non-__init function or else race conditions | |
420 | * between the root thread and the init thread may cause start_kernel to | |
421 | * be reaped by free_initmem before the root thread has proceeded to | |
422 | * cpu_idle. | |
423 | * | |
424 | * gcc-3.4 accidentally inlines this function, so use noinline. | |
425 | */ | |
426 | ||
b433c3d4 PZ |
427 | static __initdata DECLARE_COMPLETION(kthreadd_done); |
428 | ||
f99ebf0a | 429 | static noinline void __init_refok rest_init(void) |
1da177e4 LT |
430 | __releases(kernel_lock) |
431 | { | |
73c27992 EB |
432 | int pid; |
433 | ||
7db905e6 | 434 | rcu_scheduler_starting(); |
b433c3d4 | 435 | /* |
97158569 | 436 | * We need to spawn init first so that it obtains pid 1, however |
b433c3d4 PZ |
437 | * the init task will end up wanting to create kthreads, which, if |
438 | * we schedule it before we create kthreadd, will OOPS. | |
439 | */ | |
aae5f662 | 440 | kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND); |
1da177e4 | 441 | numa_default_policy(); |
73c27992 | 442 | pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); |
d11c563d | 443 | rcu_read_lock(); |
5cd20455 | 444 | kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); |
d11c563d | 445 | rcu_read_unlock(); |
b433c3d4 | 446 | complete(&kthreadd_done); |
1da177e4 | 447 | unlock_kernel(); |
f340c0d1 IM |
448 | |
449 | /* | |
450 | * The boot idle thread must execute schedule() | |
1df21055 | 451 | * at least once to get things moving: |
f340c0d1 | 452 | */ |
1df21055 | 453 | init_idle_bootup_task(current); |
5bfb5d69 | 454 | preempt_enable_no_resched(); |
f340c0d1 | 455 | schedule(); |
5bfb5d69 | 456 | preempt_disable(); |
f340c0d1 | 457 | |
5bfb5d69 | 458 | /* Call into cpu_idle with preempt disabled */ |
1da177e4 | 459 | cpu_idle(); |
1df21055 | 460 | } |
1da177e4 LT |
461 | |
462 | /* Check for early params. */ | |
463 | static int __init do_early_param(char *param, char *val) | |
464 | { | |
465 | struct obs_kernel_param *p; | |
466 | ||
467 | for (p = __setup_start; p < __setup_end; p++) { | |
18a8bd94 YL |
468 | if ((p->early && strcmp(param, p->str) == 0) || |
469 | (strcmp(param, "console") == 0 && | |
470 | strcmp(p->str, "earlycon") == 0) | |
471 | ) { | |
1da177e4 LT |
472 | if (p->setup_func(val) != 0) |
473 | printk(KERN_WARNING | |
474 | "Malformed early option '%s'\n", param); | |
475 | } | |
476 | } | |
477 | /* We accept everything at this stage. */ | |
478 | return 0; | |
479 | } | |
480 | ||
13977091 MD |
481 | void __init parse_early_options(char *cmdline) |
482 | { | |
483 | parse_args("early options", cmdline, NULL, 0, do_early_param); | |
484 | } | |
485 | ||
1da177e4 LT |
486 | /* Arch code calls this early on, or if not, just before other parsing. */ |
487 | void __init parse_early_param(void) | |
488 | { | |
489 | static __initdata int done = 0; | |
490 | static __initdata char tmp_cmdline[COMMAND_LINE_SIZE]; | |
491 | ||
492 | if (done) | |
493 | return; | |
494 | ||
495 | /* All fall through to do_early_param. */ | |
30d7e0d4 | 496 | strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); |
13977091 | 497 | parse_early_options(tmp_cmdline); |
1da177e4 LT |
498 | done = 1; |
499 | } | |
500 | ||
501 | /* | |
502 | * Activate the first processor. | |
503 | */ | |
504 | ||
44fd2299 SS |
505 | static void __init boot_cpu_init(void) |
506 | { | |
507 | int cpu = smp_processor_id(); | |
508 | /* Mark the boot cpu "present", "online" etc for SMP and UP case */ | |
915441b6 | 509 | set_cpu_online(cpu, true); |
933b0618 | 510 | set_cpu_active(cpu, true); |
915441b6 RR |
511 | set_cpu_present(cpu, true); |
512 | set_cpu_possible(cpu, true); | |
44fd2299 SS |
513 | } |
514 | ||
839ad62e | 515 | void __init __weak smp_setup_processor_id(void) |
033ab7f8 AM |
516 | { |
517 | } | |
518 | ||
8c9843e5 BH |
519 | void __init __weak thread_info_cache_init(void) |
520 | { | |
521 | } | |
522 | ||
444f478f PE |
523 | /* |
524 | * Set up kernel memory allocators | |
525 | */ | |
526 | static void __init mm_init(void) | |
527 | { | |
ca371c0d KH |
528 | /* |
529 | * page_cgroup requires countinous pages as memmap | |
530 | * and it's bigger than MAX_ORDER unless SPARSEMEM. | |
531 | */ | |
532 | page_cgroup_init_flatmem(); | |
444f478f PE |
533 | mem_init(); |
534 | kmem_cache_init(); | |
099a19d9 | 535 | percpu_init_late(); |
c868d550 | 536 | pgtable_cache_init(); |
444f478f PE |
537 | vmalloc_init(); |
538 | } | |
539 | ||
1da177e4 LT |
540 | asmlinkage void __init start_kernel(void) |
541 | { | |
542 | char * command_line; | |
543 | extern struct kernel_param __start___param[], __stop___param[]; | |
033ab7f8 AM |
544 | |
545 | smp_setup_processor_id(); | |
546 | ||
fbb9ce95 IM |
547 | /* |
548 | * Need to run as early as possible, to initialize the | |
549 | * lockdep hash: | |
550 | */ | |
551 | lockdep_init(); | |
3ac7fe5a | 552 | debug_objects_early_init(); |
42059429 IM |
553 | |
554 | /* | |
555 | * Set up the the initial canary ASAP: | |
556 | */ | |
557 | boot_init_stack_canary(); | |
558 | ||
ddbcc7e8 | 559 | cgroup_init_early(); |
fbb9ce95 IM |
560 | |
561 | local_irq_disable(); | |
562 | early_boot_irqs_off(); | |
243c7621 | 563 | early_init_irq_lock_class(); |
fbb9ce95 | 564 | |
1da177e4 LT |
565 | /* |
566 | * Interrupts are still disabled. Do necessary setups, then | |
567 | * enable them | |
568 | */ | |
569 | lock_kernel(); | |
906568c9 | 570 | tick_init(); |
44fd2299 | 571 | boot_cpu_init(); |
1da177e4 | 572 | page_address_init(); |
657cafa6 | 573 | printk(KERN_NOTICE "%s", linux_banner); |
1da177e4 | 574 | setup_arch(&command_line); |
cf475ad2 | 575 | mm_init_owner(&init_mm, &init_task); |
30d7e0d4 | 576 | setup_command_line(command_line); |
e0982e90 | 577 | setup_nr_cpu_ids(); |
d6647bdf | 578 | setup_per_cpu_areas(); |
44fd2299 | 579 | smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ |
1da177e4 | 580 | |
1f522509 | 581 | build_all_zonelists(NULL); |
83b519e8 PE |
582 | page_alloc_init(); |
583 | ||
584 | printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line); | |
585 | parse_early_param(); | |
586 | parse_args("Booting kernel", static_command_line, __start___param, | |
587 | __stop___param - __start___param, | |
588 | &unknown_bootoption); | |
589 | /* | |
590 | * These use large bootmem allocations and must precede | |
591 | * kmem_cache_init() | |
592 | */ | |
593 | pidhash_init(); | |
83b519e8 PE |
594 | vfs_caches_init_early(); |
595 | sort_main_extable(); | |
596 | trap_init(); | |
444f478f | 597 | mm_init(); |
1da177e4 LT |
598 | /* |
599 | * Set up the scheduler prior starting any interrupts (such as the | |
600 | * timer interrupt). Full topology setup happens at smp_init() | |
601 | * time - but meanwhile we still have a functioning scheduler. | |
602 | */ | |
603 | sched_init(); | |
604 | /* | |
605 | * Disable preemption - early bootup scheduling is extremely | |
606 | * fragile until we cpu_idle() for the first time. | |
607 | */ | |
608 | preempt_disable(); | |
c4a68306 AB |
609 | if (!irqs_disabled()) { |
610 | printk(KERN_WARNING "start_kernel(): bug: interrupts were " | |
611 | "enabled *very* early, fixing it\n"); | |
612 | local_irq_disable(); | |
613 | } | |
1da177e4 | 614 | rcu_init(); |
773e3eb7 | 615 | radix_tree_init(); |
0b8f1efa YL |
616 | /* init some links before init_ISA_irqs() */ |
617 | early_irq_init(); | |
1da177e4 | 618 | init_IRQ(); |
3c7b4e6b | 619 | prio_tree_init(); |
1da177e4 | 620 | init_timers(); |
c0a31329 | 621 | hrtimers_init(); |
1da177e4 | 622 | softirq_init(); |
ad596171 | 623 | timekeeping_init(); |
88fecaa2 | 624 | time_init(); |
93e02814 HC |
625 | profile_init(); |
626 | if (!irqs_disabled()) | |
24d431d0 RL |
627 | printk(KERN_CRIT "start_kernel(): bug: interrupts were " |
628 | "enabled early\n"); | |
fbb9ce95 | 629 | early_boot_irqs_on(); |
93e02814 | 630 | local_irq_enable(); |
dcce284a BH |
631 | |
632 | /* Interrupts are enabled now so all GFP allocations are safe. */ | |
452aa699 | 633 | gfp_allowed_mask = __GFP_BITS_MASK; |
dcce284a | 634 | |
7e85ee0c | 635 | kmem_cache_init_late(); |
1da177e4 LT |
636 | |
637 | /* | |
638 | * HACK ALERT! This is early. We're enabling the console before | |
639 | * we've done PCI setups etc, and console_init() must be aware of | |
640 | * this. But we do want output early, in case something goes wrong. | |
641 | */ | |
642 | console_init(); | |
643 | if (panic_later) | |
644 | panic(panic_later, panic_param); | |
fbb9ce95 IM |
645 | |
646 | lockdep_info(); | |
647 | ||
9a11b49a IM |
648 | /* |
649 | * Need to run this when irqs are enabled, because it wants | |
650 | * to self-test [hard/soft]-irqs on/off lock inversion bugs | |
651 | * too: | |
652 | */ | |
653 | locking_selftest(); | |
654 | ||
1da177e4 LT |
655 | #ifdef CONFIG_BLK_DEV_INITRD |
656 | if (initrd_start && !initrd_below_start_ok && | |
bd673c7c | 657 | page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { |
1da177e4 | 658 | printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - " |
fb6624eb | 659 | "disabling it.\n", |
bd673c7c GU |
660 | page_to_pfn(virt_to_page((void *)initrd_start)), |
661 | min_low_pfn); | |
1da177e4 LT |
662 | initrd_start = 0; |
663 | } | |
664 | #endif | |
94b6da5a | 665 | page_cgroup_init(); |
a03c2a48 | 666 | enable_debug_pagealloc(); |
b9ce08c0 | 667 | kmemtrace_init(); |
3c7b4e6b | 668 | kmemleak_init(); |
3ac7fe5a | 669 | debug_objects_mem_init(); |
199f0ca5 | 670 | idr_init_cache(); |
e7c8d5c9 | 671 | setup_per_cpu_pageset(); |
1da177e4 LT |
672 | numa_policy_init(); |
673 | if (late_time_init) | |
674 | late_time_init(); | |
fa84e9ee | 675 | sched_clock_init(); |
1da177e4 LT |
676 | calibrate_delay(); |
677 | pidmap_init(); | |
1da177e4 LT |
678 | anon_vma_init(); |
679 | #ifdef CONFIG_X86 | |
680 | if (efi_enabled) | |
681 | efi_enter_virtual_mode(); | |
682 | #endif | |
8c9843e5 | 683 | thread_info_cache_init(); |
d84f4f99 | 684 | cred_init(); |
4481374c | 685 | fork_init(totalram_pages); |
1da177e4 LT |
686 | proc_caches_init(); |
687 | buffer_init(); | |
1da177e4 LT |
688 | key_init(); |
689 | security_init(); | |
0b4b3827 | 690 | dbg_late_init(); |
4481374c | 691 | vfs_caches_init(totalram_pages); |
1da177e4 LT |
692 | signals_init(); |
693 | /* rootfs populating might need page-writeback */ | |
694 | page_writeback_init(); | |
695 | #ifdef CONFIG_PROC_FS | |
696 | proc_root_init(); | |
697 | #endif | |
ddbcc7e8 | 698 | cgroup_init(); |
1da177e4 | 699 | cpuset_init(); |
c757249a | 700 | taskstats_init_early(); |
ca74e92b | 701 | delayacct_init(); |
1da177e4 LT |
702 | |
703 | check_bugs(); | |
704 | ||
705 | acpi_early_init(); /* before LAPIC and SMP init */ | |
6ae6996a | 706 | sfi_init_late(); |
1da177e4 | 707 | |
68bf21aa SR |
708 | ftrace_init(); |
709 | ||
1da177e4 LT |
710 | /* Do the rest non-__init'ed, we're now alive */ |
711 | rest_init(); | |
712 | } | |
713 | ||
b99b87f7 PO |
714 | /* Call all constructor functions linked into the kernel. */ |
715 | static void __init do_ctors(void) | |
716 | { | |
717 | #ifdef CONFIG_CONSTRUCTORS | |
196a15b4 | 718 | ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; |
b99b87f7 | 719 | |
196a15b4 HS |
720 | for (; fn < (ctor_fn_t *) __ctors_end; fn++) |
721 | (*fn)(); | |
b99b87f7 PO |
722 | #endif |
723 | } | |
724 | ||
22a9d645 | 725 | int initcall_debug; |
d0ea3d7d | 726 | core_param(initcall_debug, initcall_debug, bool, 0644); |
1da177e4 | 727 | |
4a683bf9 IM |
728 | static char msgbuf[64]; |
729 | static struct boot_trace_call call; | |
730 | static struct boot_trace_ret ret; | |
731 | ||
59f9415f | 732 | int do_one_initcall(initcall_t fn) |
1da177e4 | 733 | { |
1da177e4 | 734 | int count = preempt_count(); |
74239072 | 735 | ktime_t calltime, delta, rettime; |
1da177e4 | 736 | |
e0df154f | 737 | if (initcall_debug) { |
74239072 FW |
738 | call.caller = task_pid_nr(current); |
739 | printk("calling %pF @ %i\n", fn, call.caller); | |
740 | calltime = ktime_get(); | |
741 | trace_boot_call(&call, fn); | |
71566a0d | 742 | enable_boot_trace(); |
e0df154f | 743 | } |
1da177e4 | 744 | |
74239072 | 745 | ret.result = fn(); |
1da177e4 | 746 | |
e0df154f | 747 | if (initcall_debug) { |
71566a0d | 748 | disable_boot_trace(); |
74239072 FW |
749 | rettime = ktime_get(); |
750 | delta = ktime_sub(rettime, calltime); | |
1d926f27 | 751 | ret.duration = (unsigned long long) ktime_to_ns(delta) >> 10; |
74239072 | 752 | trace_boot_ret(&ret, fn); |
ca538f6b | 753 | printk("initcall %pF returned %d after %Ld usecs\n", fn, |
74239072 | 754 | ret.result, ret.duration); |
e0df154f | 755 | } |
8f0c45cd | 756 | |
e0df154f | 757 | msgbuf[0] = 0; |
e662e1cf | 758 | |
74239072 FW |
759 | if (ret.result && ret.result != -ENODEV && initcall_debug) |
760 | sprintf(msgbuf, "error code %d ", ret.result); | |
e662e1cf | 761 | |
e0df154f | 762 | if (preempt_count() != count) { |
a76bfd0d | 763 | strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf)); |
e0df154f | 764 | preempt_count() = count; |
1da177e4 | 765 | } |
e0df154f | 766 | if (irqs_disabled()) { |
a76bfd0d | 767 | strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); |
e0df154f LT |
768 | local_irq_enable(); |
769 | } | |
770 | if (msgbuf[0]) { | |
96d746c6 | 771 | printk("initcall %pF returned with %s\n", fn, msgbuf); |
e0df154f | 772 | } |
59f9415f | 773 | |
74239072 | 774 | return ret.result; |
e0df154f LT |
775 | } |
776 | ||
777 | ||
c2147a50 | 778 | extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[]; |
e0df154f LT |
779 | |
780 | static void __init do_initcalls(void) | |
781 | { | |
196a15b4 | 782 | initcall_t *fn; |
e0df154f | 783 | |
196a15b4 HS |
784 | for (fn = __early_initcall_end; fn < __initcall_end; fn++) |
785 | do_one_initcall(*fn); | |
1da177e4 LT |
786 | |
787 | /* Make sure there is no pending stuff from the initcall sequence */ | |
788 | flush_scheduled_work(); | |
789 | } | |
790 | ||
791 | /* | |
792 | * Ok, the machine is now initialized. None of the devices | |
793 | * have been touched yet, but the CPU subsystem is up and | |
794 | * running, and memory and process management works. | |
795 | * | |
796 | * Now we can finally start doing some real work.. | |
797 | */ | |
798 | static void __init do_basic_setup(void) | |
799 | { | |
4403b406 | 800 | init_workqueues(); |
759ee091 | 801 | cpuset_init_smp(); |
1da177e4 | 802 | usermodehelper_init(); |
2b2af54a | 803 | init_tmpfs(); |
1da177e4 | 804 | driver_init(); |
b04c3afb | 805 | init_irq_proc(); |
b99b87f7 | 806 | do_ctors(); |
1da177e4 LT |
807 | do_initcalls(); |
808 | } | |
809 | ||
7babe8db | 810 | static void __init do_pre_smp_initcalls(void) |
c2147a50 | 811 | { |
196a15b4 | 812 | initcall_t *fn; |
c2147a50 | 813 | |
196a15b4 HS |
814 | for (fn = __initcall_start; fn < __early_initcall_end; fn++) |
815 | do_one_initcall(*fn); | |
c2147a50 EGM |
816 | } |
817 | ||
1da177e4 LT |
818 | static void run_init_process(char *init_filename) |
819 | { | |
820 | argv_init[0] = init_filename; | |
67608567 | 821 | kernel_execve(init_filename, argv_init, envp_init); |
1da177e4 LT |
822 | } |
823 | ||
ee5bfa64 VG |
824 | /* This is a non __init function. Force it to be noinline otherwise gcc |
825 | * makes it inline to init() and it becomes part of init.text section | |
826 | */ | |
f99ebf0a | 827 | static noinline int init_post(void) |
acdd052a | 828 | __releases(kernel_lock) |
ee5bfa64 | 829 | { |
22a9d645 AV |
830 | /* need to finish all async __init code before freeing the memory */ |
831 | async_synchronize_full(); | |
ee5bfa64 VG |
832 | free_initmem(); |
833 | unlock_kernel(); | |
834 | mark_rodata_ro(); | |
835 | system_state = SYSTEM_RUNNING; | |
836 | numa_default_policy(); | |
837 | ||
ee5bfa64 | 838 | |
fae5fa44 ON |
839 | current->signal->flags |= SIGNAL_UNKILLABLE; |
840 | ||
ee5bfa64 VG |
841 | if (ramdisk_execute_command) { |
842 | run_init_process(ramdisk_execute_command); | |
843 | printk(KERN_WARNING "Failed to execute %s\n", | |
844 | ramdisk_execute_command); | |
845 | } | |
846 | ||
847 | /* | |
848 | * We try each of these until one succeeds. | |
849 | * | |
850 | * The Bourne shell can be used instead of init if we are | |
851 | * trying to recover a really broken machine. | |
852 | */ | |
853 | if (execute_command) { | |
854 | run_init_process(execute_command); | |
855 | printk(KERN_WARNING "Failed to execute %s. Attempting " | |
856 | "defaults...\n", execute_command); | |
857 | } | |
858 | run_init_process("/sbin/init"); | |
859 | run_init_process("/etc/init"); | |
860 | run_init_process("/bin/init"); | |
861 | run_init_process("/bin/sh"); | |
862 | ||
9a85b8d6 AM |
863 | panic("No init found. Try passing init= option to kernel. " |
864 | "See Linux Documentation/init.txt for guidance."); | |
ee5bfa64 VG |
865 | } |
866 | ||
aae5f662 | 867 | static int __init kernel_init(void * unused) |
1da177e4 | 868 | { |
b433c3d4 PZ |
869 | /* |
870 | * Wait until kthreadd is all set-up. | |
871 | */ | |
872 | wait_for_completion(&kthreadd_done); | |
1da177e4 | 873 | lock_kernel(); |
58568d2a MX |
874 | |
875 | /* | |
876 | * init can allocate pages on any node | |
877 | */ | |
5ab116c9 | 878 | set_mems_allowed(node_states[N_HIGH_MEMORY]); |
1da177e4 LT |
879 | /* |
880 | * init can run on any cpu. | |
881 | */ | |
1a2142af | 882 | set_cpus_allowed_ptr(current, cpu_all_mask); |
1da177e4 LT |
883 | /* |
884 | * Tell the world that we're going to be the grim | |
885 | * reaper of innocent orphaned children. | |
886 | * | |
887 | * We don't want people to have to make incorrect | |
888 | * assumptions about where in the task array this | |
889 | * can be found. | |
890 | */ | |
84d73786 | 891 | init_pid_ns.child_reaper = current; |
1da177e4 | 892 | |
9ec52099 CLG |
893 | cad_pid = task_pid(current); |
894 | ||
ca74a6f8 | 895 | smp_prepare_cpus(setup_max_cpus); |
1da177e4 LT |
896 | |
897 | do_pre_smp_initcalls(); | |
3bf77af6 | 898 | start_boot_trace(); |
1da177e4 | 899 | |
1da177e4 LT |
900 | smp_init(); |
901 | sched_init_smp(); | |
902 | ||
1da177e4 LT |
903 | do_basic_setup(); |
904 | ||
2bd3a997 EB |
905 | /* Open the /dev/console on the rootfs, this should never fail */ |
906 | if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) | |
907 | printk(KERN_WARNING "Warning: unable to open an initial console.\n"); | |
908 | ||
909 | (void) sys_dup(0); | |
910 | (void) sys_dup(0); | |
1da177e4 LT |
911 | /* |
912 | * check if there is an early userspace init. If yes, let it do all | |
913 | * the work | |
914 | */ | |
ffdfc409 OJ |
915 | |
916 | if (!ramdisk_execute_command) | |
917 | ramdisk_execute_command = "/init"; | |
918 | ||
919 | if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) { | |
920 | ramdisk_execute_command = NULL; | |
1da177e4 | 921 | prepare_namespace(); |
ffdfc409 | 922 | } |
1da177e4 LT |
923 | |
924 | /* | |
925 | * Ok, we have completed the initial bootup, and | |
926 | * we're essentially up and running. Get rid of the | |
927 | * initmem segments and start the user-mode stuff.. | |
928 | */ | |
71566a0d | 929 | |
ee5bfa64 VG |
930 | init_post(); |
931 | return 0; | |
1da177e4 | 932 | } |