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> | |
1da177e4 | 23 | #include <linux/initrd.h> |
1da177e4 | 24 | #include <linux/bootmem.h> |
4a7a16dc | 25 | #include <linux/acpi.h> |
1da177e4 | 26 | #include <linux/tty.h> |
1da177e4 LT |
27 | #include <linux/percpu.h> |
28 | #include <linux/kmod.h> | |
db64fe02 | 29 | #include <linux/vmalloc.h> |
1da177e4 | 30 | #include <linux/kernel_stat.h> |
d7cd5611 | 31 | #include <linux/start_kernel.h> |
1da177e4 | 32 | #include <linux/security.h> |
3d442233 | 33 | #include <linux/smp.h> |
1da177e4 LT |
34 | #include <linux/profile.h> |
35 | #include <linux/rcupdate.h> | |
36 | #include <linux/moduleparam.h> | |
37 | #include <linux/kallsyms.h> | |
38 | #include <linux/writeback.h> | |
39 | #include <linux/cpu.h> | |
40 | #include <linux/cpuset.h> | |
ddbcc7e8 | 41 | #include <linux/cgroup.h> |
1da177e4 | 42 | #include <linux/efi.h> |
906568c9 | 43 | #include <linux/tick.h> |
6168a702 | 44 | #include <linux/interrupt.h> |
c757249a | 45 | #include <linux/taskstats_kern.h> |
ca74e92b | 46 | #include <linux/delayacct.h> |
1da177e4 LT |
47 | #include <linux/unistd.h> |
48 | #include <linux/rmap.h> | |
49 | #include <linux/mempolicy.h> | |
50 | #include <linux/key.h> | |
b6cd0b77 | 51 | #include <linux/buffer_head.h> |
94b6da5a | 52 | #include <linux/page_cgroup.h> |
9a11b49a | 53 | #include <linux/debug_locks.h> |
3ac7fe5a | 54 | #include <linux/debugobjects.h> |
fbb9ce95 | 55 | #include <linux/lockdep.h> |
3c7b4e6b | 56 | #include <linux/kmemleak.h> |
84d73786 | 57 | #include <linux/pid_namespace.h> |
1f21782e | 58 | #include <linux/device.h> |
73c27992 | 59 | #include <linux/kthread.h> |
e6fe6649 | 60 | #include <linux/sched.h> |
a1c9eea9 | 61 | #include <linux/signal.h> |
199f0ca5 | 62 | #include <linux/idr.h> |
0b4b3827 | 63 | #include <linux/kgdb.h> |
68bf21aa | 64 | #include <linux/ftrace.h> |
22a9d645 | 65 | #include <linux/async.h> |
dfec072e | 66 | #include <linux/kmemcheck.h> |
6ae6996a | 67 | #include <linux/sfi.h> |
2b2af54a | 68 | #include <linux/shmem_fs.h> |
5a0e3ad6 | 69 | #include <linux/slab.h> |
24a24bb6 | 70 | #include <linux/perf_event.h> |
4a9d4b02 | 71 | #include <linux/file.h> |
a74fb73c | 72 | #include <linux/ptrace.h> |
1da177e4 LT |
73 | |
74 | #include <asm/io.h> | |
75 | #include <asm/bugs.h> | |
76 | #include <asm/setup.h> | |
a940199f | 77 | #include <asm/sections.h> |
37b73c82 | 78 | #include <asm/cacheflush.h> |
1da177e4 | 79 | |
1da177e4 LT |
80 | #ifdef CONFIG_X86_LOCAL_APIC |
81 | #include <asm/smp.h> | |
82 | #endif | |
83 | ||
aae5f662 | 84 | static int kernel_init(void *); |
1da177e4 LT |
85 | |
86 | extern void init_IRQ(void); | |
1da177e4 LT |
87 | extern void fork_init(unsigned long); |
88 | extern void mca_init(void); | |
89 | extern void sbus_init(void); | |
1da177e4 LT |
90 | extern void prio_tree_init(void); |
91 | extern void radix_tree_init(void); | |
37b73c82 AV |
92 | #ifndef CONFIG_DEBUG_RODATA |
93 | static inline void mark_rodata_ro(void) { } | |
94 | #endif | |
1da177e4 LT |
95 | |
96 | #ifdef CONFIG_TC | |
97 | extern void tc_init(void); | |
98 | #endif | |
99 | ||
2ce802f6 TH |
100 | /* |
101 | * Debug helper: via this flag we know that we are in 'early bootup code' | |
102 | * where only the boot processor is running with IRQ disabled. This means | |
103 | * two things - IRQ must not be enabled before the flag is cleared and some | |
104 | * operations which are not allowed with IRQ disabled are allowed while the | |
105 | * flag is set. | |
106 | */ | |
107 | bool early_boot_irqs_disabled __read_mostly; | |
108 | ||
a6826048 | 109 | enum system_states system_state __read_mostly; |
1da177e4 LT |
110 | EXPORT_SYMBOL(system_state); |
111 | ||
112 | /* | |
113 | * Boot command-line arguments | |
114 | */ | |
115 | #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT | |
116 | #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT | |
117 | ||
118 | extern void time_init(void); | |
119 | /* Default late time init is NULL. archs can override this later. */ | |
d2e3192b | 120 | void (*__initdata late_time_init)(void); |
1da177e4 LT |
121 | extern void softirq_init(void); |
122 | ||
30d7e0d4 ABL |
123 | /* Untouched command line saved by arch-specific code. */ |
124 | char __initdata boot_command_line[COMMAND_LINE_SIZE]; | |
125 | /* Untouched saved command line (eg. for /proc) */ | |
126 | char *saved_command_line; | |
127 | /* Command line for parameter parsing */ | |
128 | static char *static_command_line; | |
1da177e4 LT |
129 | |
130 | static char *execute_command; | |
ffdfc409 | 131 | static char *ramdisk_execute_command; |
1da177e4 | 132 | |
8b3b2955 JB |
133 | /* |
134 | * If set, this is an indication to the drivers that reset the underlying | |
135 | * device before going ahead with the initialization otherwise driver might | |
136 | * rely on the BIOS and skip the reset operation. | |
137 | * | |
138 | * This is useful if kernel is booting in an unreliable environment. | |
139 | * For ex. kdump situaiton where previous kernel has crashed, BIOS has been | |
140 | * skipped and devices will be in unknown state. | |
141 | */ | |
142 | unsigned int reset_devices; | |
143 | EXPORT_SYMBOL(reset_devices); | |
1da177e4 | 144 | |
7e96287d VG |
145 | static int __init set_reset_devices(char *str) |
146 | { | |
147 | reset_devices = 1; | |
148 | return 1; | |
149 | } | |
150 | ||
151 | __setup("reset_devices", set_reset_devices); | |
152 | ||
d7627467 DH |
153 | static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; |
154 | const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; | |
1da177e4 LT |
155 | static const char *panic_later, *panic_param; |
156 | ||
914dcaa8 | 157 | extern const struct obs_kernel_param __setup_start[], __setup_end[]; |
1da177e4 LT |
158 | |
159 | static int __init obsolete_checksetup(char *line) | |
160 | { | |
914dcaa8 | 161 | const struct obs_kernel_param *p; |
33df0d19 | 162 | int had_early_param = 0; |
1da177e4 LT |
163 | |
164 | p = __setup_start; | |
165 | do { | |
166 | int n = strlen(p->str); | |
b1e4d20c | 167 | if (parameqn(line, p->str, n)) { |
1da177e4 | 168 | if (p->early) { |
33df0d19 RR |
169 | /* Already done in parse_early_param? |
170 | * (Needs exact match on param part). | |
171 | * Keep iterating, as we can have early | |
172 | * params and __setups of same names 8( */ | |
1da177e4 | 173 | if (line[n] == '\0' || line[n] == '=') |
33df0d19 | 174 | had_early_param = 1; |
1da177e4 LT |
175 | } else if (!p->setup_func) { |
176 | printk(KERN_WARNING "Parameter %s is obsolete," | |
177 | " ignored\n", p->str); | |
178 | return 1; | |
179 | } else if (p->setup_func(line + n)) | |
180 | return 1; | |
181 | } | |
182 | p++; | |
183 | } while (p < __setup_end); | |
33df0d19 RR |
184 | |
185 | return had_early_param; | |
1da177e4 LT |
186 | } |
187 | ||
188 | /* | |
189 | * This should be approx 2 Bo*oMips to start (note initial shift), and will | |
190 | * still work even if initially too large, it will just take slightly longer | |
191 | */ | |
192 | unsigned long loops_per_jiffy = (1<<12); | |
193 | ||
194 | EXPORT_SYMBOL(loops_per_jiffy); | |
195 | ||
196 | static int __init debug_kernel(char *str) | |
197 | { | |
1da177e4 | 198 | console_loglevel = 10; |
f6f21c81 | 199 | return 0; |
1da177e4 LT |
200 | } |
201 | ||
202 | static int __init quiet_kernel(char *str) | |
203 | { | |
1da177e4 | 204 | console_loglevel = 4; |
f6f21c81 | 205 | return 0; |
1da177e4 LT |
206 | } |
207 | ||
f6f21c81 YL |
208 | early_param("debug", debug_kernel); |
209 | early_param("quiet", quiet_kernel); | |
1da177e4 LT |
210 | |
211 | static int __init loglevel(char *str) | |
212 | { | |
808bf29b AS |
213 | int newlevel; |
214 | ||
215 | /* | |
216 | * Only update loglevel value when a correct setting was passed, | |
217 | * to prevent blind crashes (when loglevel being set to 0) that | |
218 | * are quite hard to debug | |
219 | */ | |
220 | if (get_option(&str, &newlevel)) { | |
221 | console_loglevel = newlevel; | |
222 | return 0; | |
223 | } | |
224 | ||
225 | return -EINVAL; | |
1da177e4 LT |
226 | } |
227 | ||
f6f21c81 | 228 | early_param("loglevel", loglevel); |
1da177e4 | 229 | |
a99cd112 | 230 | /* Change NUL term back to "=", to make "param" the whole string. */ |
eb157427 | 231 | static int __init repair_env_string(char *param, char *val, const char *unused) |
1da177e4 | 232 | { |
1da177e4 LT |
233 | if (val) { |
234 | /* param=val or param="val"? */ | |
235 | if (val == param+strlen(param)+1) | |
236 | val[-1] = '='; | |
237 | else if (val == param+strlen(param)+2) { | |
238 | val[-2] = '='; | |
239 | memmove(val-1, val, strlen(val)+1); | |
240 | val--; | |
241 | } else | |
242 | BUG(); | |
243 | } | |
a99cd112 CM |
244 | return 0; |
245 | } | |
246 | ||
247 | /* | |
248 | * Unknown boot options get handed to init, unless they look like | |
249 | * unused parameters (modprobe will find them in /proc/cmdline). | |
250 | */ | |
eb157427 | 251 | static int __init unknown_bootoption(char *param, char *val, const char *unused) |
a99cd112 | 252 | { |
eb157427 | 253 | repair_env_string(param, val, unused); |
1da177e4 LT |
254 | |
255 | /* Handle obsolete-style parameters */ | |
256 | if (obsolete_checksetup(param)) | |
257 | return 0; | |
258 | ||
f066a4f6 RR |
259 | /* Unused module parameter. */ |
260 | if (strchr(param, '.') && (!val || strchr(param, '.') < val)) | |
1da177e4 | 261 | return 0; |
1da177e4 LT |
262 | |
263 | if (panic_later) | |
264 | return 0; | |
265 | ||
266 | if (val) { | |
267 | /* Environment option */ | |
268 | unsigned int i; | |
269 | for (i = 0; envp_init[i]; i++) { | |
270 | if (i == MAX_INIT_ENVS) { | |
271 | panic_later = "Too many boot env vars at `%s'"; | |
272 | panic_param = param; | |
273 | } | |
274 | if (!strncmp(param, envp_init[i], val - param)) | |
275 | break; | |
276 | } | |
277 | envp_init[i] = param; | |
278 | } else { | |
279 | /* Command line option */ | |
280 | unsigned int i; | |
281 | for (i = 0; argv_init[i]; i++) { | |
282 | if (i == MAX_INIT_ARGS) { | |
283 | panic_later = "Too many boot init vars at `%s'"; | |
284 | panic_param = param; | |
285 | } | |
286 | } | |
287 | argv_init[i] = param; | |
288 | } | |
289 | return 0; | |
290 | } | |
291 | ||
292 | static int __init init_setup(char *str) | |
293 | { | |
294 | unsigned int i; | |
295 | ||
296 | execute_command = str; | |
297 | /* | |
298 | * In case LILO is going to boot us with default command line, | |
299 | * it prepends "auto" before the whole cmdline which makes | |
300 | * the shell think it should execute a script with such name. | |
301 | * So we ignore all arguments entered _before_ init=... [MJ] | |
302 | */ | |
303 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
304 | argv_init[i] = NULL; | |
305 | return 1; | |
306 | } | |
307 | __setup("init=", init_setup); | |
308 | ||
ffdfc409 OJ |
309 | static int __init rdinit_setup(char *str) |
310 | { | |
311 | unsigned int i; | |
312 | ||
313 | ramdisk_execute_command = str; | |
314 | /* See "auto" comment in init_setup */ | |
315 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
316 | argv_init[i] = NULL; | |
317 | return 1; | |
318 | } | |
319 | __setup("rdinit=", rdinit_setup); | |
320 | ||
1da177e4 | 321 | #ifndef CONFIG_SMP |
34db18a0 | 322 | static const unsigned int setup_max_cpus = NR_CPUS; |
1da177e4 LT |
323 | #ifdef CONFIG_X86_LOCAL_APIC |
324 | static void __init smp_init(void) | |
325 | { | |
326 | APIC_init_uniprocessor(); | |
327 | } | |
328 | #else | |
329 | #define smp_init() do { } while (0) | |
330 | #endif | |
331 | ||
e0982e90 | 332 | static inline void setup_nr_cpu_ids(void) { } |
1da177e4 | 333 | static inline void smp_prepare_cpus(unsigned int maxcpus) { } |
1da177e4 LT |
334 | #endif |
335 | ||
30d7e0d4 ABL |
336 | /* |
337 | * We need to store the untouched command line for future reference. | |
338 | * We also need to store the touched command line since the parameter | |
339 | * parsing is performed in place, and we should allow a component to | |
340 | * store reference of name/value for future reference. | |
341 | */ | |
342 | static void __init setup_command_line(char *command_line) | |
343 | { | |
344 | saved_command_line = alloc_bootmem(strlen (boot_command_line)+1); | |
345 | static_command_line = alloc_bootmem(strlen (command_line)+1); | |
346 | strcpy (saved_command_line, boot_command_line); | |
347 | strcpy (static_command_line, command_line); | |
348 | } | |
349 | ||
1da177e4 LT |
350 | /* |
351 | * We need to finalize in a non-__init function or else race conditions | |
352 | * between the root thread and the init thread may cause start_kernel to | |
353 | * be reaped by free_initmem before the root thread has proceeded to | |
354 | * cpu_idle. | |
355 | * | |
356 | * gcc-3.4 accidentally inlines this function, so use noinline. | |
357 | */ | |
358 | ||
b433c3d4 PZ |
359 | static __initdata DECLARE_COMPLETION(kthreadd_done); |
360 | ||
f99ebf0a | 361 | static noinline void __init_refok rest_init(void) |
1da177e4 | 362 | { |
73c27992 EB |
363 | int pid; |
364 | ||
7db905e6 | 365 | rcu_scheduler_starting(); |
b433c3d4 | 366 | /* |
97158569 | 367 | * We need to spawn init first so that it obtains pid 1, however |
b433c3d4 PZ |
368 | * the init task will end up wanting to create kthreads, which, if |
369 | * we schedule it before we create kthreadd, will OOPS. | |
370 | */ | |
aae5f662 | 371 | kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND); |
1da177e4 | 372 | numa_default_policy(); |
73c27992 | 373 | pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); |
d11c563d | 374 | rcu_read_lock(); |
5cd20455 | 375 | kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); |
d11c563d | 376 | rcu_read_unlock(); |
b433c3d4 | 377 | complete(&kthreadd_done); |
f340c0d1 IM |
378 | |
379 | /* | |
380 | * The boot idle thread must execute schedule() | |
1df21055 | 381 | * at least once to get things moving: |
f340c0d1 | 382 | */ |
1df21055 | 383 | init_idle_bootup_task(current); |
bd2f5536 | 384 | schedule_preempt_disabled(); |
5bfb5d69 | 385 | /* Call into cpu_idle with preempt disabled */ |
1da177e4 | 386 | cpu_idle(); |
1df21055 | 387 | } |
1da177e4 LT |
388 | |
389 | /* Check for early params. */ | |
9fb48c74 | 390 | static int __init do_early_param(char *param, char *val, const char *unused) |
1da177e4 | 391 | { |
914dcaa8 | 392 | const struct obs_kernel_param *p; |
1da177e4 LT |
393 | |
394 | for (p = __setup_start; p < __setup_end; p++) { | |
b1e4d20c | 395 | if ((p->early && parameq(param, p->str)) || |
18a8bd94 YL |
396 | (strcmp(param, "console") == 0 && |
397 | strcmp(p->str, "earlycon") == 0) | |
398 | ) { | |
1da177e4 LT |
399 | if (p->setup_func(val) != 0) |
400 | printk(KERN_WARNING | |
401 | "Malformed early option '%s'\n", param); | |
402 | } | |
403 | } | |
404 | /* We accept everything at this stage. */ | |
405 | return 0; | |
406 | } | |
407 | ||
13977091 MD |
408 | void __init parse_early_options(char *cmdline) |
409 | { | |
026cee00 | 410 | parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param); |
13977091 MD |
411 | } |
412 | ||
1da177e4 LT |
413 | /* Arch code calls this early on, or if not, just before other parsing. */ |
414 | void __init parse_early_param(void) | |
415 | { | |
416 | static __initdata int done = 0; | |
417 | static __initdata char tmp_cmdline[COMMAND_LINE_SIZE]; | |
418 | ||
419 | if (done) | |
420 | return; | |
421 | ||
422 | /* All fall through to do_early_param. */ | |
30d7e0d4 | 423 | strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); |
13977091 | 424 | parse_early_options(tmp_cmdline); |
1da177e4 LT |
425 | done = 1; |
426 | } | |
427 | ||
428 | /* | |
429 | * Activate the first processor. | |
430 | */ | |
431 | ||
44fd2299 SS |
432 | static void __init boot_cpu_init(void) |
433 | { | |
434 | int cpu = smp_processor_id(); | |
435 | /* Mark the boot cpu "present", "online" etc for SMP and UP case */ | |
915441b6 | 436 | set_cpu_online(cpu, true); |
933b0618 | 437 | set_cpu_active(cpu, true); |
915441b6 RR |
438 | set_cpu_present(cpu, true); |
439 | set_cpu_possible(cpu, true); | |
44fd2299 SS |
440 | } |
441 | ||
839ad62e | 442 | void __init __weak smp_setup_processor_id(void) |
033ab7f8 AM |
443 | { |
444 | } | |
445 | ||
8c9843e5 BH |
446 | void __init __weak thread_info_cache_init(void) |
447 | { | |
448 | } | |
449 | ||
444f478f PE |
450 | /* |
451 | * Set up kernel memory allocators | |
452 | */ | |
453 | static void __init mm_init(void) | |
454 | { | |
ca371c0d | 455 | /* |
7fa87ce7 JC |
456 | * page_cgroup requires contiguous pages, |
457 | * bigger than MAX_ORDER unless SPARSEMEM. | |
ca371c0d KH |
458 | */ |
459 | page_cgroup_init_flatmem(); | |
444f478f PE |
460 | mem_init(); |
461 | kmem_cache_init(); | |
099a19d9 | 462 | percpu_init_late(); |
c868d550 | 463 | pgtable_cache_init(); |
444f478f PE |
464 | vmalloc_init(); |
465 | } | |
466 | ||
1da177e4 LT |
467 | asmlinkage void __init start_kernel(void) |
468 | { | |
469 | char * command_line; | |
914dcaa8 | 470 | extern const struct kernel_param __start___param[], __stop___param[]; |
033ab7f8 | 471 | |
fbb9ce95 IM |
472 | /* |
473 | * Need to run as early as possible, to initialize the | |
474 | * lockdep hash: | |
475 | */ | |
476 | lockdep_init(); | |
73839c5b | 477 | smp_setup_processor_id(); |
3ac7fe5a | 478 | debug_objects_early_init(); |
42059429 IM |
479 | |
480 | /* | |
481 | * Set up the the initial canary ASAP: | |
482 | */ | |
483 | boot_init_stack_canary(); | |
484 | ||
ddbcc7e8 | 485 | cgroup_init_early(); |
fbb9ce95 IM |
486 | |
487 | local_irq_disable(); | |
2ce802f6 | 488 | early_boot_irqs_disabled = true; |
fbb9ce95 | 489 | |
1da177e4 LT |
490 | /* |
491 | * Interrupts are still disabled. Do necessary setups, then | |
492 | * enable them | |
493 | */ | |
906568c9 | 494 | tick_init(); |
44fd2299 | 495 | boot_cpu_init(); |
1da177e4 | 496 | page_address_init(); |
657cafa6 | 497 | printk(KERN_NOTICE "%s", linux_banner); |
1da177e4 | 498 | setup_arch(&command_line); |
cf475ad2 | 499 | mm_init_owner(&init_mm, &init_task); |
6345d24d | 500 | mm_init_cpumask(&init_mm); |
30d7e0d4 | 501 | setup_command_line(command_line); |
e0982e90 | 502 | setup_nr_cpu_ids(); |
d6647bdf | 503 | setup_per_cpu_areas(); |
44fd2299 | 504 | smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ |
1da177e4 | 505 | |
9adb62a5 | 506 | build_all_zonelists(NULL, NULL); |
83b519e8 PE |
507 | page_alloc_init(); |
508 | ||
509 | printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line); | |
510 | parse_early_param(); | |
511 | parse_args("Booting kernel", static_command_line, __start___param, | |
512 | __stop___param - __start___param, | |
ae82fdb1 | 513 | -1, -1, &unknown_bootoption); |
97ce2c88 JF |
514 | |
515 | jump_label_init(); | |
516 | ||
83b519e8 PE |
517 | /* |
518 | * These use large bootmem allocations and must precede | |
519 | * kmem_cache_init() | |
520 | */ | |
162a7e75 | 521 | setup_log_buf(0); |
83b519e8 | 522 | pidhash_init(); |
83b519e8 PE |
523 | vfs_caches_init_early(); |
524 | sort_main_extable(); | |
525 | trap_init(); | |
444f478f | 526 | mm_init(); |
de03c72c | 527 | |
1da177e4 LT |
528 | /* |
529 | * Set up the scheduler prior starting any interrupts (such as the | |
530 | * timer interrupt). Full topology setup happens at smp_init() | |
531 | * time - but meanwhile we still have a functioning scheduler. | |
532 | */ | |
533 | sched_init(); | |
534 | /* | |
535 | * Disable preemption - early bootup scheduling is extremely | |
536 | * fragile until we cpu_idle() for the first time. | |
537 | */ | |
538 | preempt_disable(); | |
c4a68306 AB |
539 | if (!irqs_disabled()) { |
540 | printk(KERN_WARNING "start_kernel(): bug: interrupts were " | |
541 | "enabled *very* early, fixing it\n"); | |
542 | local_irq_disable(); | |
543 | } | |
9f58a205 | 544 | idr_init_cache(); |
24a24bb6 | 545 | perf_event_init(); |
1da177e4 | 546 | rcu_init(); |
773e3eb7 | 547 | radix_tree_init(); |
0b8f1efa YL |
548 | /* init some links before init_ISA_irqs() */ |
549 | early_irq_init(); | |
1da177e4 | 550 | init_IRQ(); |
3c7b4e6b | 551 | prio_tree_init(); |
1da177e4 | 552 | init_timers(); |
c0a31329 | 553 | hrtimers_init(); |
1da177e4 | 554 | softirq_init(); |
ad596171 | 555 | timekeeping_init(); |
88fecaa2 | 556 | time_init(); |
93e02814 | 557 | profile_init(); |
d8ad7d11 | 558 | call_function_init(); |
93e02814 | 559 | if (!irqs_disabled()) |
24d431d0 RL |
560 | printk(KERN_CRIT "start_kernel(): bug: interrupts were " |
561 | "enabled early\n"); | |
2ce802f6 | 562 | early_boot_irqs_disabled = false; |
93e02814 | 563 | local_irq_enable(); |
dcce284a | 564 | |
7e85ee0c | 565 | kmem_cache_init_late(); |
1da177e4 LT |
566 | |
567 | /* | |
568 | * HACK ALERT! This is early. We're enabling the console before | |
569 | * we've done PCI setups etc, and console_init() must be aware of | |
570 | * this. But we do want output early, in case something goes wrong. | |
571 | */ | |
572 | console_init(); | |
573 | if (panic_later) | |
574 | panic(panic_later, panic_param); | |
fbb9ce95 IM |
575 | |
576 | lockdep_info(); | |
577 | ||
9a11b49a IM |
578 | /* |
579 | * Need to run this when irqs are enabled, because it wants | |
580 | * to self-test [hard/soft]-irqs on/off lock inversion bugs | |
581 | * too: | |
582 | */ | |
583 | locking_selftest(); | |
584 | ||
1da177e4 LT |
585 | #ifdef CONFIG_BLK_DEV_INITRD |
586 | if (initrd_start && !initrd_below_start_ok && | |
bd673c7c | 587 | page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { |
1da177e4 | 588 | printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - " |
fb6624eb | 589 | "disabling it.\n", |
bd673c7c GU |
590 | page_to_pfn(virt_to_page((void *)initrd_start)), |
591 | min_low_pfn); | |
1da177e4 LT |
592 | initrd_start = 0; |
593 | } | |
594 | #endif | |
94b6da5a | 595 | page_cgroup_init(); |
3ac7fe5a | 596 | debug_objects_mem_init(); |
9b090f2d | 597 | kmemleak_init(); |
e7c8d5c9 | 598 | setup_per_cpu_pageset(); |
1da177e4 LT |
599 | numa_policy_init(); |
600 | if (late_time_init) | |
601 | late_time_init(); | |
fa84e9ee | 602 | sched_clock_init(); |
1da177e4 LT |
603 | calibrate_delay(); |
604 | pidmap_init(); | |
1da177e4 | 605 | anon_vma_init(); |
f026cfa8 PA |
606 | #ifdef CONFIG_X86 |
607 | if (efi_enabled) | |
608 | efi_enter_virtual_mode(); | |
609 | #endif | |
8c9843e5 | 610 | thread_info_cache_init(); |
d84f4f99 | 611 | cred_init(); |
4481374c | 612 | fork_init(totalram_pages); |
1da177e4 LT |
613 | proc_caches_init(); |
614 | buffer_init(); | |
1da177e4 LT |
615 | key_init(); |
616 | security_init(); | |
0b4b3827 | 617 | dbg_late_init(); |
4481374c | 618 | vfs_caches_init(totalram_pages); |
1da177e4 LT |
619 | signals_init(); |
620 | /* rootfs populating might need page-writeback */ | |
621 | page_writeback_init(); | |
622 | #ifdef CONFIG_PROC_FS | |
623 | proc_root_init(); | |
624 | #endif | |
ddbcc7e8 | 625 | cgroup_init(); |
1da177e4 | 626 | cpuset_init(); |
c757249a | 627 | taskstats_init_early(); |
ca74e92b | 628 | delayacct_init(); |
1da177e4 LT |
629 | |
630 | check_bugs(); | |
631 | ||
632 | acpi_early_init(); /* before LAPIC and SMP init */ | |
6ae6996a | 633 | sfi_init_late(); |
1da177e4 | 634 | |
68bf21aa SR |
635 | ftrace_init(); |
636 | ||
1da177e4 LT |
637 | /* Do the rest non-__init'ed, we're now alive */ |
638 | rest_init(); | |
639 | } | |
640 | ||
b99b87f7 PO |
641 | /* Call all constructor functions linked into the kernel. */ |
642 | static void __init do_ctors(void) | |
643 | { | |
644 | #ifdef CONFIG_CONSTRUCTORS | |
196a15b4 | 645 | ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; |
b99b87f7 | 646 | |
196a15b4 HS |
647 | for (; fn < (ctor_fn_t *) __ctors_end; fn++) |
648 | (*fn)(); | |
b99b87f7 PO |
649 | #endif |
650 | } | |
651 | ||
2329abfa | 652 | bool initcall_debug; |
d0ea3d7d | 653 | core_param(initcall_debug, initcall_debug, bool, 0644); |
1da177e4 | 654 | |
4a683bf9 | 655 | static char msgbuf[64]; |
4a683bf9 | 656 | |
e4461271 | 657 | static int __init_or_module do_one_initcall_debug(initcall_t fn) |
1da177e4 | 658 | { |
74239072 | 659 | ktime_t calltime, delta, rettime; |
30dbb20e AW |
660 | unsigned long long duration; |
661 | int ret; | |
1da177e4 | 662 | |
22c5c03b KW |
663 | printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current)); |
664 | calltime = ktime_get(); | |
30dbb20e | 665 | ret = fn(); |
22c5c03b KW |
666 | rettime = ktime_get(); |
667 | delta = ktime_sub(rettime, calltime); | |
668 | duration = (unsigned long long) ktime_to_ns(delta) >> 10; | |
669 | printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", fn, | |
670 | ret, duration); | |
1da177e4 | 671 | |
22c5c03b KW |
672 | return ret; |
673 | } | |
674 | ||
e4461271 | 675 | int __init_or_module do_one_initcall(initcall_t fn) |
22c5c03b KW |
676 | { |
677 | int count = preempt_count(); | |
678 | int ret; | |
679 | ||
680 | if (initcall_debug) | |
681 | ret = do_one_initcall_debug(fn); | |
682 | else | |
683 | ret = fn(); | |
8f0c45cd | 684 | |
e0df154f | 685 | msgbuf[0] = 0; |
e662e1cf | 686 | |
30dbb20e AW |
687 | if (ret && ret != -ENODEV && initcall_debug) |
688 | sprintf(msgbuf, "error code %d ", ret); | |
e662e1cf | 689 | |
e0df154f | 690 | if (preempt_count() != count) { |
a76bfd0d | 691 | strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf)); |
e0df154f | 692 | preempt_count() = count; |
1da177e4 | 693 | } |
e0df154f | 694 | if (irqs_disabled()) { |
a76bfd0d | 695 | strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); |
e0df154f LT |
696 | local_irq_enable(); |
697 | } | |
698 | if (msgbuf[0]) { | |
96d746c6 | 699 | printk("initcall %pF returned with %s\n", fn, msgbuf); |
e0df154f | 700 | } |
59f9415f | 701 | |
30dbb20e | 702 | return ret; |
e0df154f LT |
703 | } |
704 | ||
705 | ||
026cee00 PM |
706 | extern initcall_t __initcall_start[]; |
707 | extern initcall_t __initcall0_start[]; | |
708 | extern initcall_t __initcall1_start[]; | |
709 | extern initcall_t __initcall2_start[]; | |
710 | extern initcall_t __initcall3_start[]; | |
711 | extern initcall_t __initcall4_start[]; | |
712 | extern initcall_t __initcall5_start[]; | |
713 | extern initcall_t __initcall6_start[]; | |
714 | extern initcall_t __initcall7_start[]; | |
715 | extern initcall_t __initcall_end[]; | |
716 | ||
717 | static initcall_t *initcall_levels[] __initdata = { | |
718 | __initcall0_start, | |
719 | __initcall1_start, | |
720 | __initcall2_start, | |
721 | __initcall3_start, | |
722 | __initcall4_start, | |
723 | __initcall5_start, | |
724 | __initcall6_start, | |
725 | __initcall7_start, | |
726 | __initcall_end, | |
727 | }; | |
728 | ||
96263d28 | 729 | /* Keep these in sync with initcalls in include/linux/init.h */ |
026cee00 | 730 | static char *initcall_level_names[] __initdata = { |
9fb48c74 JC |
731 | "early", |
732 | "core", | |
733 | "postcore", | |
734 | "arch", | |
735 | "subsys", | |
736 | "fs", | |
737 | "device", | |
738 | "late", | |
026cee00 PM |
739 | }; |
740 | ||
026cee00 | 741 | static void __init do_initcall_level(int level) |
e0df154f | 742 | { |
026cee00 | 743 | extern const struct kernel_param __start___param[], __stop___param[]; |
196a15b4 | 744 | initcall_t *fn; |
e0df154f | 745 | |
026cee00 PM |
746 | strcpy(static_command_line, saved_command_line); |
747 | parse_args(initcall_level_names[level], | |
748 | static_command_line, __start___param, | |
749 | __stop___param - __start___param, | |
750 | level, level, | |
eb157427 | 751 | &repair_env_string); |
026cee00 PM |
752 | |
753 | for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) | |
196a15b4 | 754 | do_one_initcall(*fn); |
1da177e4 LT |
755 | } |
756 | ||
026cee00 PM |
757 | static void __init do_initcalls(void) |
758 | { | |
759 | int level; | |
760 | ||
19efb72f | 761 | for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) |
026cee00 PM |
762 | do_initcall_level(level); |
763 | } | |
764 | ||
1da177e4 LT |
765 | /* |
766 | * Ok, the machine is now initialized. None of the devices | |
767 | * have been touched yet, but the CPU subsystem is up and | |
768 | * running, and memory and process management works. | |
769 | * | |
770 | * Now we can finally start doing some real work.. | |
771 | */ | |
772 | static void __init do_basic_setup(void) | |
773 | { | |
759ee091 | 774 | cpuset_init_smp(); |
1da177e4 | 775 | usermodehelper_init(); |
41ffe5d5 | 776 | shmem_init(); |
1da177e4 | 777 | driver_init(); |
b04c3afb | 778 | init_irq_proc(); |
b99b87f7 | 779 | do_ctors(); |
d5767c53 | 780 | usermodehelper_enable(); |
b0f84374 | 781 | do_initcalls(); |
1da177e4 LT |
782 | } |
783 | ||
7babe8db | 784 | static void __init do_pre_smp_initcalls(void) |
c2147a50 | 785 | { |
196a15b4 | 786 | initcall_t *fn; |
c2147a50 | 787 | |
026cee00 | 788 | for (fn = __initcall_start; fn < __initcall0_start; fn++) |
196a15b4 | 789 | do_one_initcall(*fn); |
c2147a50 EGM |
790 | } |
791 | ||
a74fb73c | 792 | static int run_init_process(const char *init_filename) |
1da177e4 LT |
793 | { |
794 | argv_init[0] = init_filename; | |
a74fb73c | 795 | return kernel_execve(init_filename, argv_init, envp_init); |
1da177e4 LT |
796 | } |
797 | ||
d6b21238 AV |
798 | static void __init kernel_init_freeable(void); |
799 | ||
800 | static int __ref kernel_init(void *unused) | |
ee5bfa64 | 801 | { |
d6b21238 | 802 | kernel_init_freeable(); |
22a9d645 AV |
803 | /* need to finish all async __init code before freeing the memory */ |
804 | async_synchronize_full(); | |
ee5bfa64 | 805 | free_initmem(); |
ee5bfa64 VG |
806 | mark_rodata_ro(); |
807 | system_state = SYSTEM_RUNNING; | |
808 | numa_default_policy(); | |
809 | ||
fae5fa44 | 810 | current->signal->flags |= SIGNAL_UNKILLABLE; |
4a9d4b02 | 811 | flush_delayed_fput(); |
fae5fa44 | 812 | |
ee5bfa64 | 813 | if (ramdisk_execute_command) { |
a74fb73c AV |
814 | if (!run_init_process(ramdisk_execute_command)) |
815 | return 0; | |
ee5bfa64 VG |
816 | printk(KERN_WARNING "Failed to execute %s\n", |
817 | ramdisk_execute_command); | |
818 | } | |
819 | ||
820 | /* | |
821 | * We try each of these until one succeeds. | |
822 | * | |
823 | * The Bourne shell can be used instead of init if we are | |
824 | * trying to recover a really broken machine. | |
825 | */ | |
826 | if (execute_command) { | |
a74fb73c AV |
827 | if (!run_init_process(execute_command)) |
828 | return 0; | |
ee5bfa64 VG |
829 | printk(KERN_WARNING "Failed to execute %s. Attempting " |
830 | "defaults...\n", execute_command); | |
831 | } | |
a74fb73c AV |
832 | if (!run_init_process("/sbin/init") || |
833 | !run_init_process("/etc/init") || | |
834 | !run_init_process("/bin/init") || | |
835 | !run_init_process("/bin/sh")) | |
836 | return 0; | |
ee5bfa64 | 837 | |
9a85b8d6 AM |
838 | panic("No init found. Try passing init= option to kernel. " |
839 | "See Linux Documentation/init.txt for guidance."); | |
ee5bfa64 VG |
840 | } |
841 | ||
d6b21238 | 842 | static void __init kernel_init_freeable(void) |
1da177e4 | 843 | { |
b433c3d4 PZ |
844 | /* |
845 | * Wait until kthreadd is all set-up. | |
846 | */ | |
847 | wait_for_completion(&kthreadd_done); | |
31a67102 LT |
848 | |
849 | /* Now the scheduler is fully set up and can do blocking allocations */ | |
850 | gfp_allowed_mask = __GFP_BITS_MASK; | |
851 | ||
58568d2a MX |
852 | /* |
853 | * init can allocate pages on any node | |
854 | */ | |
5ab116c9 | 855 | set_mems_allowed(node_states[N_HIGH_MEMORY]); |
1da177e4 LT |
856 | /* |
857 | * init can run on any cpu. | |
858 | */ | |
1a2142af | 859 | set_cpus_allowed_ptr(current, cpu_all_mask); |
1da177e4 | 860 | |
9ec52099 CLG |
861 | cad_pid = task_pid(current); |
862 | ||
ca74a6f8 | 863 | smp_prepare_cpus(setup_max_cpus); |
1da177e4 LT |
864 | |
865 | do_pre_smp_initcalls(); | |
004417a6 | 866 | lockup_detector_init(); |
1da177e4 | 867 | |
1da177e4 LT |
868 | smp_init(); |
869 | sched_init_smp(); | |
870 | ||
1da177e4 LT |
871 | do_basic_setup(); |
872 | ||
2bd3a997 EB |
873 | /* Open the /dev/console on the rootfs, this should never fail */ |
874 | if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) | |
875 | printk(KERN_WARNING "Warning: unable to open an initial console.\n"); | |
876 | ||
877 | (void) sys_dup(0); | |
878 | (void) sys_dup(0); | |
1da177e4 LT |
879 | /* |
880 | * check if there is an early userspace init. If yes, let it do all | |
881 | * the work | |
882 | */ | |
ffdfc409 OJ |
883 | |
884 | if (!ramdisk_execute_command) | |
885 | ramdisk_execute_command = "/init"; | |
886 | ||
887 | if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) { | |
888 | ramdisk_execute_command = NULL; | |
1da177e4 | 889 | prepare_namespace(); |
ffdfc409 | 890 | } |
1da177e4 LT |
891 | |
892 | /* | |
893 | * Ok, we have completed the initial bootup, and | |
894 | * we're essentially up and running. Get rid of the | |
895 | * initmem segments and start the user-mode stuff.. | |
896 | */ | |
1da177e4 | 897 | } |