]>
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 | ||
ea676e84 AM |
12 | #define DEBUG /* Enable initcall_debug */ |
13 | ||
1da177e4 LT |
14 | #include <linux/types.h> |
15 | #include <linux/module.h> | |
16 | #include <linux/proc_fs.h> | |
1da177e4 LT |
17 | #include <linux/kernel.h> |
18 | #include <linux/syscalls.h> | |
9b5609fd | 19 | #include <linux/stackprotector.h> |
1da177e4 LT |
20 | #include <linux/string.h> |
21 | #include <linux/ctype.h> | |
22 | #include <linux/delay.h> | |
1da177e4 LT |
23 | #include <linux/ioport.h> |
24 | #include <linux/init.h> | |
1da177e4 | 25 | #include <linux/initrd.h> |
1da177e4 | 26 | #include <linux/bootmem.h> |
4a7a16dc | 27 | #include <linux/acpi.h> |
1da177e4 | 28 | #include <linux/tty.h> |
1da177e4 LT |
29 | #include <linux/percpu.h> |
30 | #include <linux/kmod.h> | |
db64fe02 | 31 | #include <linux/vmalloc.h> |
1da177e4 | 32 | #include <linux/kernel_stat.h> |
d7cd5611 | 33 | #include <linux/start_kernel.h> |
1da177e4 | 34 | #include <linux/security.h> |
3d442233 | 35 | #include <linux/smp.h> |
1da177e4 LT |
36 | #include <linux/profile.h> |
37 | #include <linux/rcupdate.h> | |
38 | #include <linux/moduleparam.h> | |
39 | #include <linux/kallsyms.h> | |
40 | #include <linux/writeback.h> | |
41 | #include <linux/cpu.h> | |
42 | #include <linux/cpuset.h> | |
ddbcc7e8 | 43 | #include <linux/cgroup.h> |
1da177e4 | 44 | #include <linux/efi.h> |
906568c9 | 45 | #include <linux/tick.h> |
6168a702 | 46 | #include <linux/interrupt.h> |
c757249a | 47 | #include <linux/taskstats_kern.h> |
ca74e92b | 48 | #include <linux/delayacct.h> |
1da177e4 LT |
49 | #include <linux/unistd.h> |
50 | #include <linux/rmap.h> | |
51 | #include <linux/mempolicy.h> | |
52 | #include <linux/key.h> | |
b6cd0b77 | 53 | #include <linux/buffer_head.h> |
94b6da5a | 54 | #include <linux/page_cgroup.h> |
9a11b49a | 55 | #include <linux/debug_locks.h> |
3ac7fe5a | 56 | #include <linux/debugobjects.h> |
fbb9ce95 | 57 | #include <linux/lockdep.h> |
3c7b4e6b | 58 | #include <linux/kmemleak.h> |
84d73786 | 59 | #include <linux/pid_namespace.h> |
1f21782e | 60 | #include <linux/device.h> |
73c27992 | 61 | #include <linux/kthread.h> |
e6fe6649 | 62 | #include <linux/sched.h> |
a1c9eea9 | 63 | #include <linux/signal.h> |
199f0ca5 | 64 | #include <linux/idr.h> |
0b4b3827 | 65 | #include <linux/kgdb.h> |
68bf21aa | 66 | #include <linux/ftrace.h> |
22a9d645 | 67 | #include <linux/async.h> |
dfec072e | 68 | #include <linux/kmemcheck.h> |
6ae6996a | 69 | #include <linux/sfi.h> |
2b2af54a | 70 | #include <linux/shmem_fs.h> |
5a0e3ad6 | 71 | #include <linux/slab.h> |
24a24bb6 | 72 | #include <linux/perf_event.h> |
4a9d4b02 | 73 | #include <linux/file.h> |
a74fb73c | 74 | #include <linux/ptrace.h> |
bb813f4c TH |
75 | #include <linux/blkdev.h> |
76 | #include <linux/elevator.h> | |
38ff87f7 | 77 | #include <linux/sched_clock.h> |
65f382fd | 78 | #include <linux/context_tracking.h> |
47d06e53 | 79 | #include <linux/random.h> |
1da177e4 LT |
80 | |
81 | #include <asm/io.h> | |
82 | #include <asm/bugs.h> | |
83 | #include <asm/setup.h> | |
a940199f | 84 | #include <asm/sections.h> |
37b73c82 | 85 | #include <asm/cacheflush.h> |
1da177e4 | 86 | |
1da177e4 LT |
87 | #ifdef CONFIG_X86_LOCAL_APIC |
88 | #include <asm/smp.h> | |
89 | #endif | |
90 | ||
aae5f662 | 91 | static int kernel_init(void *); |
1da177e4 LT |
92 | |
93 | extern void init_IRQ(void); | |
1da177e4 LT |
94 | extern void fork_init(unsigned long); |
95 | extern void mca_init(void); | |
96 | extern void sbus_init(void); | |
1da177e4 | 97 | extern void radix_tree_init(void); |
37b73c82 AV |
98 | #ifndef CONFIG_DEBUG_RODATA |
99 | static inline void mark_rodata_ro(void) { } | |
100 | #endif | |
1da177e4 LT |
101 | |
102 | #ifdef CONFIG_TC | |
103 | extern void tc_init(void); | |
104 | #endif | |
105 | ||
2ce802f6 TH |
106 | /* |
107 | * Debug helper: via this flag we know that we are in 'early bootup code' | |
108 | * where only the boot processor is running with IRQ disabled. This means | |
109 | * two things - IRQ must not be enabled before the flag is cleared and some | |
110 | * operations which are not allowed with IRQ disabled are allowed while the | |
111 | * flag is set. | |
112 | */ | |
113 | bool early_boot_irqs_disabled __read_mostly; | |
114 | ||
a6826048 | 115 | enum system_states system_state __read_mostly; |
1da177e4 LT |
116 | EXPORT_SYMBOL(system_state); |
117 | ||
118 | /* | |
119 | * Boot command-line arguments | |
120 | */ | |
121 | #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT | |
122 | #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT | |
123 | ||
124 | extern void time_init(void); | |
125 | /* Default late time init is NULL. archs can override this later. */ | |
d2e3192b | 126 | void (*__initdata late_time_init)(void); |
1da177e4 | 127 | |
30d7e0d4 ABL |
128 | /* Untouched command line saved by arch-specific code. */ |
129 | char __initdata boot_command_line[COMMAND_LINE_SIZE]; | |
130 | /* Untouched saved command line (eg. for /proc) */ | |
131 | char *saved_command_line; | |
132 | /* Command line for parameter parsing */ | |
133 | static char *static_command_line; | |
08746a65 KM |
134 | /* Command line for per-initcall parameter parsing */ |
135 | static char *initcall_command_line; | |
1da177e4 LT |
136 | |
137 | static char *execute_command; | |
ffdfc409 | 138 | static char *ramdisk_execute_command; |
1da177e4 | 139 | |
c4b2c0c5 HFS |
140 | /* |
141 | * Used to generate warnings if static_key manipulation functions are used | |
142 | * before jump_label_init is called. | |
143 | */ | |
144 | bool static_key_initialized __read_mostly = false; | |
145 | EXPORT_SYMBOL_GPL(static_key_initialized); | |
146 | ||
8b3b2955 JB |
147 | /* |
148 | * If set, this is an indication to the drivers that reset the underlying | |
149 | * device before going ahead with the initialization otherwise driver might | |
150 | * rely on the BIOS and skip the reset operation. | |
151 | * | |
152 | * This is useful if kernel is booting in an unreliable environment. | |
153 | * For ex. kdump situaiton where previous kernel has crashed, BIOS has been | |
154 | * skipped and devices will be in unknown state. | |
155 | */ | |
156 | unsigned int reset_devices; | |
157 | EXPORT_SYMBOL(reset_devices); | |
1da177e4 | 158 | |
7e96287d VG |
159 | static int __init set_reset_devices(char *str) |
160 | { | |
161 | reset_devices = 1; | |
162 | return 1; | |
163 | } | |
164 | ||
165 | __setup("reset_devices", set_reset_devices); | |
166 | ||
d7627467 DH |
167 | static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; |
168 | const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; | |
1da177e4 LT |
169 | static const char *panic_later, *panic_param; |
170 | ||
914dcaa8 | 171 | extern const struct obs_kernel_param __setup_start[], __setup_end[]; |
1da177e4 LT |
172 | |
173 | static int __init obsolete_checksetup(char *line) | |
174 | { | |
914dcaa8 | 175 | const struct obs_kernel_param *p; |
33df0d19 | 176 | int had_early_param = 0; |
1da177e4 LT |
177 | |
178 | p = __setup_start; | |
179 | do { | |
180 | int n = strlen(p->str); | |
b1e4d20c | 181 | if (parameqn(line, p->str, n)) { |
1da177e4 | 182 | if (p->early) { |
33df0d19 RR |
183 | /* Already done in parse_early_param? |
184 | * (Needs exact match on param part). | |
185 | * Keep iterating, as we can have early | |
186 | * params and __setups of same names 8( */ | |
1da177e4 | 187 | if (line[n] == '\0' || line[n] == '=') |
33df0d19 | 188 | had_early_param = 1; |
1da177e4 | 189 | } else if (!p->setup_func) { |
ea676e84 AM |
190 | pr_warn("Parameter %s is obsolete, ignored\n", |
191 | p->str); | |
1da177e4 LT |
192 | return 1; |
193 | } else if (p->setup_func(line + n)) | |
194 | return 1; | |
195 | } | |
196 | p++; | |
197 | } while (p < __setup_end); | |
33df0d19 RR |
198 | |
199 | return had_early_param; | |
1da177e4 LT |
200 | } |
201 | ||
202 | /* | |
203 | * This should be approx 2 Bo*oMips to start (note initial shift), and will | |
204 | * still work even if initially too large, it will just take slightly longer | |
205 | */ | |
206 | unsigned long loops_per_jiffy = (1<<12); | |
207 | ||
208 | EXPORT_SYMBOL(loops_per_jiffy); | |
209 | ||
210 | static int __init debug_kernel(char *str) | |
211 | { | |
1da177e4 | 212 | console_loglevel = 10; |
f6f21c81 | 213 | return 0; |
1da177e4 LT |
214 | } |
215 | ||
216 | static int __init quiet_kernel(char *str) | |
217 | { | |
1da177e4 | 218 | console_loglevel = 4; |
f6f21c81 | 219 | return 0; |
1da177e4 LT |
220 | } |
221 | ||
f6f21c81 YL |
222 | early_param("debug", debug_kernel); |
223 | early_param("quiet", quiet_kernel); | |
1da177e4 LT |
224 | |
225 | static int __init loglevel(char *str) | |
226 | { | |
808bf29b AS |
227 | int newlevel; |
228 | ||
229 | /* | |
230 | * Only update loglevel value when a correct setting was passed, | |
231 | * to prevent blind crashes (when loglevel being set to 0) that | |
232 | * are quite hard to debug | |
233 | */ | |
234 | if (get_option(&str, &newlevel)) { | |
235 | console_loglevel = newlevel; | |
236 | return 0; | |
237 | } | |
238 | ||
239 | return -EINVAL; | |
1da177e4 LT |
240 | } |
241 | ||
f6f21c81 | 242 | early_param("loglevel", loglevel); |
1da177e4 | 243 | |
a99cd112 | 244 | /* Change NUL term back to "=", to make "param" the whole string. */ |
eb157427 | 245 | static int __init repair_env_string(char *param, char *val, const char *unused) |
1da177e4 | 246 | { |
1da177e4 LT |
247 | if (val) { |
248 | /* param=val or param="val"? */ | |
249 | if (val == param+strlen(param)+1) | |
250 | val[-1] = '='; | |
251 | else if (val == param+strlen(param)+2) { | |
252 | val[-2] = '='; | |
253 | memmove(val-1, val, strlen(val)+1); | |
254 | val--; | |
255 | } else | |
256 | BUG(); | |
257 | } | |
a99cd112 CM |
258 | return 0; |
259 | } | |
260 | ||
261 | /* | |
262 | * Unknown boot options get handed to init, unless they look like | |
263 | * unused parameters (modprobe will find them in /proc/cmdline). | |
264 | */ | |
eb157427 | 265 | static int __init unknown_bootoption(char *param, char *val, const char *unused) |
a99cd112 | 266 | { |
eb157427 | 267 | repair_env_string(param, val, unused); |
1da177e4 LT |
268 | |
269 | /* Handle obsolete-style parameters */ | |
270 | if (obsolete_checksetup(param)) | |
271 | return 0; | |
272 | ||
f066a4f6 RR |
273 | /* Unused module parameter. */ |
274 | if (strchr(param, '.') && (!val || strchr(param, '.') < val)) | |
1da177e4 | 275 | return 0; |
1da177e4 LT |
276 | |
277 | if (panic_later) | |
278 | return 0; | |
279 | ||
280 | if (val) { | |
281 | /* Environment option */ | |
282 | unsigned int i; | |
283 | for (i = 0; envp_init[i]; i++) { | |
284 | if (i == MAX_INIT_ENVS) { | |
285 | panic_later = "Too many boot env vars at `%s'"; | |
286 | panic_param = param; | |
287 | } | |
288 | if (!strncmp(param, envp_init[i], val - param)) | |
289 | break; | |
290 | } | |
291 | envp_init[i] = param; | |
292 | } else { | |
293 | /* Command line option */ | |
294 | unsigned int i; | |
295 | for (i = 0; argv_init[i]; i++) { | |
296 | if (i == MAX_INIT_ARGS) { | |
297 | panic_later = "Too many boot init vars at `%s'"; | |
298 | panic_param = param; | |
299 | } | |
300 | } | |
301 | argv_init[i] = param; | |
302 | } | |
303 | return 0; | |
304 | } | |
305 | ||
306 | static int __init init_setup(char *str) | |
307 | { | |
308 | unsigned int i; | |
309 | ||
310 | execute_command = str; | |
311 | /* | |
312 | * In case LILO is going to boot us with default command line, | |
313 | * it prepends "auto" before the whole cmdline which makes | |
314 | * the shell think it should execute a script with such name. | |
315 | * So we ignore all arguments entered _before_ init=... [MJ] | |
316 | */ | |
317 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
318 | argv_init[i] = NULL; | |
319 | return 1; | |
320 | } | |
321 | __setup("init=", init_setup); | |
322 | ||
ffdfc409 OJ |
323 | static int __init rdinit_setup(char *str) |
324 | { | |
325 | unsigned int i; | |
326 | ||
327 | ramdisk_execute_command = str; | |
328 | /* See "auto" comment in init_setup */ | |
329 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
330 | argv_init[i] = NULL; | |
331 | return 1; | |
332 | } | |
333 | __setup("rdinit=", rdinit_setup); | |
334 | ||
1da177e4 | 335 | #ifndef CONFIG_SMP |
34db18a0 | 336 | static const unsigned int setup_max_cpus = NR_CPUS; |
1da177e4 LT |
337 | #ifdef CONFIG_X86_LOCAL_APIC |
338 | static void __init smp_init(void) | |
339 | { | |
340 | APIC_init_uniprocessor(); | |
341 | } | |
342 | #else | |
343 | #define smp_init() do { } while (0) | |
344 | #endif | |
345 | ||
e0982e90 | 346 | static inline void setup_nr_cpu_ids(void) { } |
1da177e4 | 347 | static inline void smp_prepare_cpus(unsigned int maxcpus) { } |
1da177e4 LT |
348 | #endif |
349 | ||
30d7e0d4 ABL |
350 | /* |
351 | * We need to store the untouched command line for future reference. | |
352 | * We also need to store the touched command line since the parameter | |
353 | * parsing is performed in place, and we should allow a component to | |
354 | * store reference of name/value for future reference. | |
355 | */ | |
356 | static void __init setup_command_line(char *command_line) | |
357 | { | |
098b081b SS |
358 | saved_command_line = |
359 | memblock_virt_alloc(strlen(boot_command_line) + 1, 0); | |
360 | initcall_command_line = | |
361 | memblock_virt_alloc(strlen(boot_command_line) + 1, 0); | |
362 | static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0); | |
30d7e0d4 ABL |
363 | strcpy (saved_command_line, boot_command_line); |
364 | strcpy (static_command_line, command_line); | |
365 | } | |
366 | ||
1da177e4 LT |
367 | /* |
368 | * We need to finalize in a non-__init function or else race conditions | |
369 | * between the root thread and the init thread may cause start_kernel to | |
370 | * be reaped by free_initmem before the root thread has proceeded to | |
371 | * cpu_idle. | |
372 | * | |
373 | * gcc-3.4 accidentally inlines this function, so use noinline. | |
374 | */ | |
375 | ||
b433c3d4 PZ |
376 | static __initdata DECLARE_COMPLETION(kthreadd_done); |
377 | ||
f99ebf0a | 378 | static noinline void __init_refok rest_init(void) |
1da177e4 | 379 | { |
73c27992 EB |
380 | int pid; |
381 | ||
7db905e6 | 382 | rcu_scheduler_starting(); |
b433c3d4 | 383 | /* |
97158569 | 384 | * We need to spawn init first so that it obtains pid 1, however |
b433c3d4 PZ |
385 | * the init task will end up wanting to create kthreads, which, if |
386 | * we schedule it before we create kthreadd, will OOPS. | |
387 | */ | |
aae5f662 | 388 | kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND); |
1da177e4 | 389 | numa_default_policy(); |
73c27992 | 390 | pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); |
d11c563d | 391 | rcu_read_lock(); |
5cd20455 | 392 | kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); |
d11c563d | 393 | rcu_read_unlock(); |
b433c3d4 | 394 | complete(&kthreadd_done); |
f340c0d1 IM |
395 | |
396 | /* | |
397 | * The boot idle thread must execute schedule() | |
1df21055 | 398 | * at least once to get things moving: |
f340c0d1 | 399 | */ |
1df21055 | 400 | init_idle_bootup_task(current); |
bd2f5536 | 401 | schedule_preempt_disabled(); |
5bfb5d69 | 402 | /* Call into cpu_idle with preempt disabled */ |
a1a04ec3 | 403 | cpu_startup_entry(CPUHP_ONLINE); |
1df21055 | 404 | } |
1da177e4 LT |
405 | |
406 | /* Check for early params. */ | |
9fb48c74 | 407 | static int __init do_early_param(char *param, char *val, const char *unused) |
1da177e4 | 408 | { |
914dcaa8 | 409 | const struct obs_kernel_param *p; |
1da177e4 LT |
410 | |
411 | for (p = __setup_start; p < __setup_end; p++) { | |
b1e4d20c | 412 | if ((p->early && parameq(param, p->str)) || |
18a8bd94 YL |
413 | (strcmp(param, "console") == 0 && |
414 | strcmp(p->str, "earlycon") == 0) | |
415 | ) { | |
1da177e4 | 416 | if (p->setup_func(val) != 0) |
ea676e84 | 417 | pr_warn("Malformed early option '%s'\n", param); |
1da177e4 LT |
418 | } |
419 | } | |
420 | /* We accept everything at this stage. */ | |
421 | return 0; | |
422 | } | |
423 | ||
13977091 MD |
424 | void __init parse_early_options(char *cmdline) |
425 | { | |
026cee00 | 426 | parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param); |
13977091 MD |
427 | } |
428 | ||
1da177e4 LT |
429 | /* Arch code calls this early on, or if not, just before other parsing. */ |
430 | void __init parse_early_param(void) | |
431 | { | |
432 | static __initdata int done = 0; | |
433 | static __initdata char tmp_cmdline[COMMAND_LINE_SIZE]; | |
434 | ||
435 | if (done) | |
436 | return; | |
437 | ||
438 | /* All fall through to do_early_param. */ | |
30d7e0d4 | 439 | strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); |
13977091 | 440 | parse_early_options(tmp_cmdline); |
1da177e4 LT |
441 | done = 1; |
442 | } | |
443 | ||
444 | /* | |
445 | * Activate the first processor. | |
446 | */ | |
447 | ||
44fd2299 SS |
448 | static void __init boot_cpu_init(void) |
449 | { | |
450 | int cpu = smp_processor_id(); | |
451 | /* Mark the boot cpu "present", "online" etc for SMP and UP case */ | |
915441b6 | 452 | set_cpu_online(cpu, true); |
933b0618 | 453 | set_cpu_active(cpu, true); |
915441b6 RR |
454 | set_cpu_present(cpu, true); |
455 | set_cpu_possible(cpu, true); | |
44fd2299 SS |
456 | } |
457 | ||
839ad62e | 458 | void __init __weak smp_setup_processor_id(void) |
033ab7f8 AM |
459 | { |
460 | } | |
461 | ||
eded09cc | 462 | # if THREAD_SIZE >= PAGE_SIZE |
8c9843e5 BH |
463 | void __init __weak thread_info_cache_init(void) |
464 | { | |
465 | } | |
eded09cc | 466 | #endif |
8c9843e5 | 467 | |
444f478f PE |
468 | /* |
469 | * Set up kernel memory allocators | |
470 | */ | |
471 | static void __init mm_init(void) | |
472 | { | |
ca371c0d | 473 | /* |
7fa87ce7 JC |
474 | * page_cgroup requires contiguous pages, |
475 | * bigger than MAX_ORDER unless SPARSEMEM. | |
ca371c0d KH |
476 | */ |
477 | page_cgroup_init_flatmem(); | |
444f478f PE |
478 | mem_init(); |
479 | kmem_cache_init(); | |
099a19d9 | 480 | percpu_init_late(); |
b35f1819 | 481 | pgtable_init(); |
444f478f PE |
482 | vmalloc_init(); |
483 | } | |
484 | ||
1da177e4 LT |
485 | asmlinkage void __init start_kernel(void) |
486 | { | |
487 | char * command_line; | |
914dcaa8 | 488 | extern const struct kernel_param __start___param[], __stop___param[]; |
033ab7f8 | 489 | |
fbb9ce95 IM |
490 | /* |
491 | * Need to run as early as possible, to initialize the | |
492 | * lockdep hash: | |
493 | */ | |
494 | lockdep_init(); | |
73839c5b | 495 | smp_setup_processor_id(); |
3ac7fe5a | 496 | debug_objects_early_init(); |
42059429 IM |
497 | |
498 | /* | |
499 | * Set up the the initial canary ASAP: | |
500 | */ | |
501 | boot_init_stack_canary(); | |
502 | ||
ddbcc7e8 | 503 | cgroup_init_early(); |
fbb9ce95 IM |
504 | |
505 | local_irq_disable(); | |
2ce802f6 | 506 | early_boot_irqs_disabled = true; |
fbb9ce95 | 507 | |
1da177e4 LT |
508 | /* |
509 | * Interrupts are still disabled. Do necessary setups, then | |
510 | * enable them | |
511 | */ | |
44fd2299 | 512 | boot_cpu_init(); |
1da177e4 | 513 | page_address_init(); |
ea676e84 | 514 | pr_notice("%s", linux_banner); |
1da177e4 | 515 | setup_arch(&command_line); |
cf475ad2 | 516 | mm_init_owner(&init_mm, &init_task); |
6345d24d | 517 | mm_init_cpumask(&init_mm); |
30d7e0d4 | 518 | setup_command_line(command_line); |
e0982e90 | 519 | setup_nr_cpu_ids(); |
d6647bdf | 520 | setup_per_cpu_areas(); |
44fd2299 | 521 | smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ |
1da177e4 | 522 | |
9adb62a5 | 523 | build_all_zonelists(NULL, NULL); |
83b519e8 PE |
524 | page_alloc_init(); |
525 | ||
ea676e84 | 526 | pr_notice("Kernel command line: %s\n", boot_command_line); |
83b519e8 PE |
527 | parse_early_param(); |
528 | parse_args("Booting kernel", static_command_line, __start___param, | |
529 | __stop___param - __start___param, | |
ae82fdb1 | 530 | -1, -1, &unknown_bootoption); |
97ce2c88 JF |
531 | |
532 | jump_label_init(); | |
533 | ||
83b519e8 PE |
534 | /* |
535 | * These use large bootmem allocations and must precede | |
536 | * kmem_cache_init() | |
537 | */ | |
162a7e75 | 538 | setup_log_buf(0); |
83b519e8 | 539 | pidhash_init(); |
83b519e8 PE |
540 | vfs_caches_init_early(); |
541 | sort_main_extable(); | |
542 | trap_init(); | |
444f478f | 543 | mm_init(); |
de03c72c | 544 | |
1da177e4 LT |
545 | /* |
546 | * Set up the scheduler prior starting any interrupts (such as the | |
547 | * timer interrupt). Full topology setup happens at smp_init() | |
548 | * time - but meanwhile we still have a functioning scheduler. | |
549 | */ | |
550 | sched_init(); | |
551 | /* | |
552 | * Disable preemption - early bootup scheduling is extremely | |
553 | * fragile until we cpu_idle() for the first time. | |
554 | */ | |
555 | preempt_disable(); | |
f91eb62f | 556 | if (WARN(!irqs_disabled(), "Interrupts were enabled *very* early, fixing it\n")) |
c4a68306 | 557 | local_irq_disable(); |
9f58a205 | 558 | idr_init_cache(); |
1da177e4 | 559 | rcu_init(); |
d1e43fa5 | 560 | tick_nohz_init(); |
65f382fd | 561 | context_tracking_init(); |
773e3eb7 | 562 | radix_tree_init(); |
0b8f1efa YL |
563 | /* init some links before init_ISA_irqs() */ |
564 | early_irq_init(); | |
1da177e4 | 565 | init_IRQ(); |
ad2b1353 | 566 | tick_init(); |
1da177e4 | 567 | init_timers(); |
c0a31329 | 568 | hrtimers_init(); |
1da177e4 | 569 | softirq_init(); |
ad596171 | 570 | timekeeping_init(); |
88fecaa2 | 571 | time_init(); |
38ff87f7 | 572 | sched_clock_postinit(); |
9e630205 | 573 | perf_event_init(); |
93e02814 | 574 | profile_init(); |
d8ad7d11 | 575 | call_function_init(); |
f91eb62f | 576 | WARN(!irqs_disabled(), "Interrupts were enabled early\n"); |
2ce802f6 | 577 | early_boot_irqs_disabled = false; |
93e02814 | 578 | local_irq_enable(); |
dcce284a | 579 | |
7e85ee0c | 580 | kmem_cache_init_late(); |
1da177e4 LT |
581 | |
582 | /* | |
583 | * HACK ALERT! This is early. We're enabling the console before | |
584 | * we've done PCI setups etc, and console_init() must be aware of | |
585 | * this. But we do want output early, in case something goes wrong. | |
586 | */ | |
587 | console_init(); | |
588 | if (panic_later) | |
589 | panic(panic_later, panic_param); | |
fbb9ce95 IM |
590 | |
591 | lockdep_info(); | |
592 | ||
9a11b49a IM |
593 | /* |
594 | * Need to run this when irqs are enabled, because it wants | |
595 | * to self-test [hard/soft]-irqs on/off lock inversion bugs | |
596 | * too: | |
597 | */ | |
598 | locking_selftest(); | |
599 | ||
1da177e4 LT |
600 | #ifdef CONFIG_BLK_DEV_INITRD |
601 | if (initrd_start && !initrd_below_start_ok && | |
bd673c7c | 602 | page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { |
ea676e84 | 603 | pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", |
bd673c7c GU |
604 | page_to_pfn(virt_to_page((void *)initrd_start)), |
605 | min_low_pfn); | |
1da177e4 LT |
606 | initrd_start = 0; |
607 | } | |
608 | #endif | |
94b6da5a | 609 | page_cgroup_init(); |
3ac7fe5a | 610 | debug_objects_mem_init(); |
9b090f2d | 611 | kmemleak_init(); |
e7c8d5c9 | 612 | setup_per_cpu_pageset(); |
1da177e4 LT |
613 | numa_policy_init(); |
614 | if (late_time_init) | |
615 | late_time_init(); | |
fa84e9ee | 616 | sched_clock_init(); |
1da177e4 LT |
617 | calibrate_delay(); |
618 | pidmap_init(); | |
1da177e4 | 619 | anon_vma_init(); |
11520e5e | 620 | #ifdef CONFIG_X86 |
83e68189 | 621 | if (efi_enabled(EFI_RUNTIME_SERVICES)) |
11520e5e LT |
622 | efi_enter_virtual_mode(); |
623 | #endif | |
8c9843e5 | 624 | thread_info_cache_init(); |
d84f4f99 | 625 | cred_init(); |
4481374c | 626 | fork_init(totalram_pages); |
1da177e4 LT |
627 | proc_caches_init(); |
628 | buffer_init(); | |
1da177e4 LT |
629 | key_init(); |
630 | security_init(); | |
0b4b3827 | 631 | dbg_late_init(); |
4481374c | 632 | vfs_caches_init(totalram_pages); |
1da177e4 LT |
633 | signals_init(); |
634 | /* rootfs populating might need page-writeback */ | |
635 | page_writeback_init(); | |
636 | #ifdef CONFIG_PROC_FS | |
637 | proc_root_init(); | |
638 | #endif | |
ddbcc7e8 | 639 | cgroup_init(); |
1da177e4 | 640 | cpuset_init(); |
c757249a | 641 | taskstats_init_early(); |
ca74e92b | 642 | delayacct_init(); |
1da177e4 LT |
643 | |
644 | check_bugs(); | |
645 | ||
646 | acpi_early_init(); /* before LAPIC and SMP init */ | |
6ae6996a | 647 | sfi_init_late(); |
1da177e4 | 648 | |
83e68189 | 649 | if (efi_enabled(EFI_RUNTIME_SERVICES)) { |
2223af38 | 650 | efi_late_init(); |
78510792 | 651 | efi_free_boot_services(); |
2223af38 | 652 | } |
78510792 | 653 | |
68bf21aa SR |
654 | ftrace_init(); |
655 | ||
1da177e4 LT |
656 | /* Do the rest non-__init'ed, we're now alive */ |
657 | rest_init(); | |
658 | } | |
659 | ||
b99b87f7 PO |
660 | /* Call all constructor functions linked into the kernel. */ |
661 | static void __init do_ctors(void) | |
662 | { | |
663 | #ifdef CONFIG_CONSTRUCTORS | |
196a15b4 | 664 | ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; |
b99b87f7 | 665 | |
196a15b4 HS |
666 | for (; fn < (ctor_fn_t *) __ctors_end; fn++) |
667 | (*fn)(); | |
b99b87f7 PO |
668 | #endif |
669 | } | |
670 | ||
2329abfa | 671 | bool initcall_debug; |
d0ea3d7d | 672 | core_param(initcall_debug, initcall_debug, bool, 0644); |
1da177e4 | 673 | |
e4461271 | 674 | static int __init_or_module do_one_initcall_debug(initcall_t fn) |
1da177e4 | 675 | { |
74239072 | 676 | ktime_t calltime, delta, rettime; |
30dbb20e AW |
677 | unsigned long long duration; |
678 | int ret; | |
1da177e4 | 679 | |
ea676e84 | 680 | pr_debug("calling %pF @ %i\n", fn, task_pid_nr(current)); |
22c5c03b | 681 | calltime = ktime_get(); |
30dbb20e | 682 | ret = fn(); |
22c5c03b KW |
683 | rettime = ktime_get(); |
684 | delta = ktime_sub(rettime, calltime); | |
685 | duration = (unsigned long long) ktime_to_ns(delta) >> 10; | |
ea676e84 AM |
686 | pr_debug("initcall %pF returned %d after %lld usecs\n", |
687 | fn, ret, duration); | |
1da177e4 | 688 | |
22c5c03b KW |
689 | return ret; |
690 | } | |
691 | ||
e4461271 | 692 | int __init_or_module do_one_initcall(initcall_t fn) |
22c5c03b KW |
693 | { |
694 | int count = preempt_count(); | |
695 | int ret; | |
ff1c8fac | 696 | char msgbuf[64]; |
22c5c03b KW |
697 | |
698 | if (initcall_debug) | |
699 | ret = do_one_initcall_debug(fn); | |
700 | else | |
701 | ret = fn(); | |
8f0c45cd | 702 | |
e0df154f | 703 | msgbuf[0] = 0; |
e662e1cf | 704 | |
e0df154f | 705 | if (preempt_count() != count) { |
bf5d770b | 706 | sprintf(msgbuf, "preemption imbalance "); |
4a2b4b22 | 707 | preempt_count_set(count); |
1da177e4 | 708 | } |
e0df154f | 709 | if (irqs_disabled()) { |
a76bfd0d | 710 | strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); |
e0df154f LT |
711 | local_irq_enable(); |
712 | } | |
f91eb62f | 713 | WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf); |
59f9415f | 714 | |
30dbb20e | 715 | return ret; |
e0df154f LT |
716 | } |
717 | ||
718 | ||
026cee00 PM |
719 | extern initcall_t __initcall_start[]; |
720 | extern initcall_t __initcall0_start[]; | |
721 | extern initcall_t __initcall1_start[]; | |
722 | extern initcall_t __initcall2_start[]; | |
723 | extern initcall_t __initcall3_start[]; | |
724 | extern initcall_t __initcall4_start[]; | |
725 | extern initcall_t __initcall5_start[]; | |
726 | extern initcall_t __initcall6_start[]; | |
727 | extern initcall_t __initcall7_start[]; | |
728 | extern initcall_t __initcall_end[]; | |
729 | ||
730 | static initcall_t *initcall_levels[] __initdata = { | |
731 | __initcall0_start, | |
732 | __initcall1_start, | |
733 | __initcall2_start, | |
734 | __initcall3_start, | |
735 | __initcall4_start, | |
736 | __initcall5_start, | |
737 | __initcall6_start, | |
738 | __initcall7_start, | |
739 | __initcall_end, | |
740 | }; | |
741 | ||
96263d28 | 742 | /* Keep these in sync with initcalls in include/linux/init.h */ |
026cee00 | 743 | static char *initcall_level_names[] __initdata = { |
9fb48c74 JC |
744 | "early", |
745 | "core", | |
746 | "postcore", | |
747 | "arch", | |
748 | "subsys", | |
749 | "fs", | |
750 | "device", | |
751 | "late", | |
026cee00 PM |
752 | }; |
753 | ||
026cee00 | 754 | static void __init do_initcall_level(int level) |
e0df154f | 755 | { |
026cee00 | 756 | extern const struct kernel_param __start___param[], __stop___param[]; |
196a15b4 | 757 | initcall_t *fn; |
e0df154f | 758 | |
08746a65 | 759 | strcpy(initcall_command_line, saved_command_line); |
026cee00 | 760 | parse_args(initcall_level_names[level], |
08746a65 | 761 | initcall_command_line, __start___param, |
026cee00 PM |
762 | __stop___param - __start___param, |
763 | level, level, | |
eb157427 | 764 | &repair_env_string); |
026cee00 PM |
765 | |
766 | for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) | |
196a15b4 | 767 | do_one_initcall(*fn); |
1da177e4 LT |
768 | } |
769 | ||
026cee00 PM |
770 | static void __init do_initcalls(void) |
771 | { | |
772 | int level; | |
773 | ||
19efb72f | 774 | for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) |
026cee00 PM |
775 | do_initcall_level(level); |
776 | } | |
777 | ||
1da177e4 LT |
778 | /* |
779 | * Ok, the machine is now initialized. None of the devices | |
780 | * have been touched yet, but the CPU subsystem is up and | |
781 | * running, and memory and process management works. | |
782 | * | |
783 | * Now we can finally start doing some real work.. | |
784 | */ | |
785 | static void __init do_basic_setup(void) | |
786 | { | |
759ee091 | 787 | cpuset_init_smp(); |
1da177e4 | 788 | usermodehelper_init(); |
41ffe5d5 | 789 | shmem_init(); |
1da177e4 | 790 | driver_init(); |
b04c3afb | 791 | init_irq_proc(); |
b99b87f7 | 792 | do_ctors(); |
d5767c53 | 793 | usermodehelper_enable(); |
b0f84374 | 794 | do_initcalls(); |
47d06e53 | 795 | random_int_secret_init(); |
1da177e4 LT |
796 | } |
797 | ||
7babe8db | 798 | static void __init do_pre_smp_initcalls(void) |
c2147a50 | 799 | { |
196a15b4 | 800 | initcall_t *fn; |
c2147a50 | 801 | |
026cee00 | 802 | for (fn = __initcall_start; fn < __initcall0_start; fn++) |
196a15b4 | 803 | do_one_initcall(*fn); |
c2147a50 EGM |
804 | } |
805 | ||
bb813f4c TH |
806 | /* |
807 | * This function requests modules which should be loaded by default and is | |
808 | * called twice right after initrd is mounted and right before init is | |
809 | * exec'd. If such modules are on either initrd or rootfs, they will be | |
810 | * loaded before control is passed to userland. | |
811 | */ | |
812 | void __init load_default_modules(void) | |
813 | { | |
814 | load_default_elevator_module(); | |
815 | } | |
816 | ||
a74fb73c | 817 | static int run_init_process(const char *init_filename) |
1da177e4 LT |
818 | { |
819 | argv_init[0] = init_filename; | |
ae903caa AV |
820 | return do_execve(init_filename, |
821 | (const char __user *const __user *)argv_init, | |
822 | (const char __user *const __user *)envp_init); | |
1da177e4 LT |
823 | } |
824 | ||
ba24762b MO |
825 | static int try_to_run_init_process(const char *init_filename) |
826 | { | |
827 | int ret; | |
828 | ||
829 | ret = run_init_process(init_filename); | |
830 | ||
831 | if (ret && ret != -ENOENT) { | |
832 | pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", | |
833 | init_filename, ret); | |
834 | } | |
835 | ||
836 | return ret; | |
837 | } | |
838 | ||
f80b0c90 | 839 | static noinline void __init kernel_init_freeable(void); |
d6b21238 AV |
840 | |
841 | static int __ref kernel_init(void *unused) | |
ee5bfa64 | 842 | { |
ba24762b MO |
843 | int ret; |
844 | ||
d6b21238 | 845 | kernel_init_freeable(); |
22a9d645 AV |
846 | /* need to finish all async __init code before freeing the memory */ |
847 | async_synchronize_full(); | |
ee5bfa64 | 848 | free_initmem(); |
ee5bfa64 VG |
849 | mark_rodata_ro(); |
850 | system_state = SYSTEM_RUNNING; | |
851 | numa_default_policy(); | |
852 | ||
4a9d4b02 | 853 | flush_delayed_fput(); |
fae5fa44 | 854 | |
ee5bfa64 | 855 | if (ramdisk_execute_command) { |
ba24762b MO |
856 | ret = run_init_process(ramdisk_execute_command); |
857 | if (!ret) | |
a74fb73c | 858 | return 0; |
ba24762b MO |
859 | pr_err("Failed to execute %s (error %d)\n", |
860 | ramdisk_execute_command, ret); | |
ee5bfa64 VG |
861 | } |
862 | ||
863 | /* | |
864 | * We try each of these until one succeeds. | |
865 | * | |
866 | * The Bourne shell can be used instead of init if we are | |
867 | * trying to recover a really broken machine. | |
868 | */ | |
869 | if (execute_command) { | |
ba24762b MO |
870 | ret = run_init_process(execute_command); |
871 | if (!ret) | |
a74fb73c | 872 | return 0; |
ba24762b MO |
873 | pr_err("Failed to execute %s (error %d). Attempting defaults...\n", |
874 | execute_command, ret); | |
ee5bfa64 | 875 | } |
ba24762b MO |
876 | if (!try_to_run_init_process("/sbin/init") || |
877 | !try_to_run_init_process("/etc/init") || | |
878 | !try_to_run_init_process("/bin/init") || | |
879 | !try_to_run_init_process("/bin/sh")) | |
a74fb73c | 880 | return 0; |
ee5bfa64 | 881 | |
ba24762b | 882 | panic("No working init found. Try passing init= option to kernel. " |
9a85b8d6 | 883 | "See Linux Documentation/init.txt for guidance."); |
ee5bfa64 VG |
884 | } |
885 | ||
f80b0c90 | 886 | static noinline void __init kernel_init_freeable(void) |
1da177e4 | 887 | { |
b433c3d4 PZ |
888 | /* |
889 | * Wait until kthreadd is all set-up. | |
890 | */ | |
891 | wait_for_completion(&kthreadd_done); | |
31a67102 LT |
892 | |
893 | /* Now the scheduler is fully set up and can do blocking allocations */ | |
894 | gfp_allowed_mask = __GFP_BITS_MASK; | |
895 | ||
58568d2a MX |
896 | /* |
897 | * init can allocate pages on any node | |
898 | */ | |
3c466d46 | 899 | set_mems_allowed(node_states[N_MEMORY]); |
1da177e4 LT |
900 | /* |
901 | * init can run on any cpu. | |
902 | */ | |
1a2142af | 903 | set_cpus_allowed_ptr(current, cpu_all_mask); |
1da177e4 | 904 | |
9ec52099 CLG |
905 | cad_pid = task_pid(current); |
906 | ||
ca74a6f8 | 907 | smp_prepare_cpus(setup_max_cpus); |
1da177e4 LT |
908 | |
909 | do_pre_smp_initcalls(); | |
004417a6 | 910 | lockup_detector_init(); |
1da177e4 | 911 | |
1da177e4 LT |
912 | smp_init(); |
913 | sched_init_smp(); | |
914 | ||
1da177e4 LT |
915 | do_basic_setup(); |
916 | ||
2bd3a997 EB |
917 | /* Open the /dev/console on the rootfs, this should never fail */ |
918 | if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) | |
ea676e84 | 919 | pr_err("Warning: unable to open an initial console.\n"); |
2bd3a997 EB |
920 | |
921 | (void) sys_dup(0); | |
922 | (void) sys_dup(0); | |
1da177e4 LT |
923 | /* |
924 | * check if there is an early userspace init. If yes, let it do all | |
925 | * the work | |
926 | */ | |
ffdfc409 OJ |
927 | |
928 | if (!ramdisk_execute_command) | |
929 | ramdisk_execute_command = "/init"; | |
930 | ||
931 | if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) { | |
932 | ramdisk_execute_command = NULL; | |
1da177e4 | 933 | prepare_namespace(); |
ffdfc409 | 934 | } |
1da177e4 LT |
935 | |
936 | /* | |
937 | * Ok, we have completed the initial bootup, and | |
938 | * we're essentially up and running. Get rid of the | |
939 | * initmem segments and start the user-mode stuff.. | |
940 | */ | |
bb813f4c TH |
941 | |
942 | /* rootfs is available now, try loading default modules */ | |
943 | load_default_modules(); | |
1da177e4 | 944 | } |