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