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> |
7b0b73d7 | 80 | #include <linux/list.h> |
1da177e4 LT |
81 | |
82 | #include <asm/io.h> | |
83 | #include <asm/bugs.h> | |
84 | #include <asm/setup.h> | |
a940199f | 85 | #include <asm/sections.h> |
37b73c82 | 86 | #include <asm/cacheflush.h> |
1da177e4 | 87 | |
1da177e4 LT |
88 | #ifdef CONFIG_X86_LOCAL_APIC |
89 | #include <asm/smp.h> | |
90 | #endif | |
91 | ||
aae5f662 | 92 | static int kernel_init(void *); |
1da177e4 LT |
93 | |
94 | extern void init_IRQ(void); | |
1da177e4 | 95 | extern void fork_init(unsigned long); |
1da177e4 | 96 | extern void radix_tree_init(void); |
37b73c82 AV |
97 | #ifndef CONFIG_DEBUG_RODATA |
98 | static inline void mark_rodata_ro(void) { } | |
99 | #endif | |
1da177e4 | 100 | |
2ce802f6 TH |
101 | /* |
102 | * Debug helper: via this flag we know that we are in 'early bootup code' | |
103 | * where only the boot processor is running with IRQ disabled. This means | |
104 | * two things - IRQ must not be enabled before the flag is cleared and some | |
105 | * operations which are not allowed with IRQ disabled are allowed while the | |
106 | * flag is set. | |
107 | */ | |
108 | bool early_boot_irqs_disabled __read_mostly; | |
109 | ||
a6826048 | 110 | enum system_states system_state __read_mostly; |
1da177e4 LT |
111 | EXPORT_SYMBOL(system_state); |
112 | ||
113 | /* | |
114 | * Boot command-line arguments | |
115 | */ | |
116 | #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT | |
117 | #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT | |
118 | ||
119 | extern void time_init(void); | |
120 | /* Default late time init is NULL. archs can override this later. */ | |
d2e3192b | 121 | void (*__initdata late_time_init)(void); |
1da177e4 | 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; | |
08746a65 KM |
129 | /* Command line for per-initcall parameter parsing */ |
130 | static char *initcall_command_line; | |
1da177e4 LT |
131 | |
132 | static char *execute_command; | |
ffdfc409 | 133 | static char *ramdisk_execute_command; |
1da177e4 | 134 | |
c4b2c0c5 HFS |
135 | /* |
136 | * Used to generate warnings if static_key manipulation functions are used | |
137 | * before jump_label_init is called. | |
138 | */ | |
139 | bool static_key_initialized __read_mostly = false; | |
140 | EXPORT_SYMBOL_GPL(static_key_initialized); | |
141 | ||
8b3b2955 JB |
142 | /* |
143 | * If set, this is an indication to the drivers that reset the underlying | |
144 | * device before going ahead with the initialization otherwise driver might | |
145 | * rely on the BIOS and skip the reset operation. | |
146 | * | |
147 | * This is useful if kernel is booting in an unreliable environment. | |
148 | * For ex. kdump situaiton where previous kernel has crashed, BIOS has been | |
149 | * skipped and devices will be in unknown state. | |
150 | */ | |
151 | unsigned int reset_devices; | |
152 | EXPORT_SYMBOL(reset_devices); | |
1da177e4 | 153 | |
7e96287d VG |
154 | static int __init set_reset_devices(char *str) |
155 | { | |
156 | reset_devices = 1; | |
157 | return 1; | |
158 | } | |
159 | ||
160 | __setup("reset_devices", set_reset_devices); | |
161 | ||
d7627467 DH |
162 | static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; |
163 | const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; | |
1da177e4 LT |
164 | static const char *panic_later, *panic_param; |
165 | ||
914dcaa8 | 166 | extern const struct obs_kernel_param __setup_start[], __setup_end[]; |
1da177e4 LT |
167 | |
168 | static int __init obsolete_checksetup(char *line) | |
169 | { | |
914dcaa8 | 170 | const struct obs_kernel_param *p; |
33df0d19 | 171 | int had_early_param = 0; |
1da177e4 LT |
172 | |
173 | p = __setup_start; | |
174 | do { | |
175 | int n = strlen(p->str); | |
b1e4d20c | 176 | if (parameqn(line, p->str, n)) { |
1da177e4 | 177 | if (p->early) { |
33df0d19 RR |
178 | /* Already done in parse_early_param? |
179 | * (Needs exact match on param part). | |
180 | * Keep iterating, as we can have early | |
181 | * params and __setups of same names 8( */ | |
1da177e4 | 182 | if (line[n] == '\0' || line[n] == '=') |
33df0d19 | 183 | had_early_param = 1; |
1da177e4 | 184 | } else if (!p->setup_func) { |
ea676e84 AM |
185 | pr_warn("Parameter %s is obsolete, ignored\n", |
186 | p->str); | |
1da177e4 LT |
187 | return 1; |
188 | } else if (p->setup_func(line + n)) | |
189 | return 1; | |
190 | } | |
191 | p++; | |
192 | } while (p < __setup_end); | |
33df0d19 RR |
193 | |
194 | return had_early_param; | |
1da177e4 LT |
195 | } |
196 | ||
197 | /* | |
198 | * This should be approx 2 Bo*oMips to start (note initial shift), and will | |
199 | * still work even if initially too large, it will just take slightly longer | |
200 | */ | |
201 | unsigned long loops_per_jiffy = (1<<12); | |
202 | ||
203 | EXPORT_SYMBOL(loops_per_jiffy); | |
204 | ||
205 | static int __init debug_kernel(char *str) | |
206 | { | |
a8fe19eb | 207 | console_loglevel = CONSOLE_LOGLEVEL_DEBUG; |
f6f21c81 | 208 | return 0; |
1da177e4 LT |
209 | } |
210 | ||
211 | static int __init quiet_kernel(char *str) | |
212 | { | |
a8fe19eb | 213 | console_loglevel = CONSOLE_LOGLEVEL_QUIET; |
f6f21c81 | 214 | return 0; |
1da177e4 LT |
215 | } |
216 | ||
f6f21c81 YL |
217 | early_param("debug", debug_kernel); |
218 | early_param("quiet", quiet_kernel); | |
1da177e4 LT |
219 | |
220 | static int __init loglevel(char *str) | |
221 | { | |
808bf29b AS |
222 | int newlevel; |
223 | ||
224 | /* | |
225 | * Only update loglevel value when a correct setting was passed, | |
226 | * to prevent blind crashes (when loglevel being set to 0) that | |
227 | * are quite hard to debug | |
228 | */ | |
229 | if (get_option(&str, &newlevel)) { | |
230 | console_loglevel = newlevel; | |
231 | return 0; | |
232 | } | |
233 | ||
234 | return -EINVAL; | |
1da177e4 LT |
235 | } |
236 | ||
f6f21c81 | 237 | early_param("loglevel", loglevel); |
1da177e4 | 238 | |
a99cd112 | 239 | /* Change NUL term back to "=", to make "param" the whole string. */ |
eb157427 | 240 | static int __init repair_env_string(char *param, char *val, const char *unused) |
1da177e4 | 241 | { |
1da177e4 LT |
242 | if (val) { |
243 | /* param=val or param="val"? */ | |
244 | if (val == param+strlen(param)+1) | |
245 | val[-1] = '='; | |
246 | else if (val == param+strlen(param)+2) { | |
247 | val[-2] = '='; | |
248 | memmove(val-1, val, strlen(val)+1); | |
249 | val--; | |
250 | } else | |
251 | BUG(); | |
252 | } | |
a99cd112 CM |
253 | return 0; |
254 | } | |
255 | ||
256 | /* | |
257 | * Unknown boot options get handed to init, unless they look like | |
258 | * unused parameters (modprobe will find them in /proc/cmdline). | |
259 | */ | |
eb157427 | 260 | static int __init unknown_bootoption(char *param, char *val, const char *unused) |
a99cd112 | 261 | { |
eb157427 | 262 | repair_env_string(param, val, unused); |
1da177e4 LT |
263 | |
264 | /* Handle obsolete-style parameters */ | |
265 | if (obsolete_checksetup(param)) | |
266 | return 0; | |
267 | ||
f066a4f6 RR |
268 | /* Unused module parameter. */ |
269 | if (strchr(param, '.') && (!val || strchr(param, '.') < val)) | |
1da177e4 | 270 | return 0; |
1da177e4 LT |
271 | |
272 | if (panic_later) | |
273 | return 0; | |
274 | ||
275 | if (val) { | |
276 | /* Environment option */ | |
277 | unsigned int i; | |
278 | for (i = 0; envp_init[i]; i++) { | |
279 | if (i == MAX_INIT_ENVS) { | |
499a4584 | 280 | panic_later = "env"; |
1da177e4 LT |
281 | panic_param = param; |
282 | } | |
283 | if (!strncmp(param, envp_init[i], val - param)) | |
284 | break; | |
285 | } | |
286 | envp_init[i] = param; | |
287 | } else { | |
288 | /* Command line option */ | |
289 | unsigned int i; | |
290 | for (i = 0; argv_init[i]; i++) { | |
291 | if (i == MAX_INIT_ARGS) { | |
499a4584 | 292 | panic_later = "init"; |
1da177e4 LT |
293 | panic_param = param; |
294 | } | |
295 | } | |
296 | argv_init[i] = param; | |
297 | } | |
298 | return 0; | |
299 | } | |
300 | ||
301 | static int __init init_setup(char *str) | |
302 | { | |
303 | unsigned int i; | |
304 | ||
305 | execute_command = str; | |
306 | /* | |
307 | * In case LILO is going to boot us with default command line, | |
308 | * it prepends "auto" before the whole cmdline which makes | |
309 | * the shell think it should execute a script with such name. | |
310 | * So we ignore all arguments entered _before_ init=... [MJ] | |
311 | */ | |
312 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
313 | argv_init[i] = NULL; | |
314 | return 1; | |
315 | } | |
316 | __setup("init=", init_setup); | |
317 | ||
ffdfc409 OJ |
318 | static int __init rdinit_setup(char *str) |
319 | { | |
320 | unsigned int i; | |
321 | ||
322 | ramdisk_execute_command = str; | |
323 | /* See "auto" comment in init_setup */ | |
324 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
325 | argv_init[i] = NULL; | |
326 | return 1; | |
327 | } | |
328 | __setup("rdinit=", rdinit_setup); | |
329 | ||
1da177e4 | 330 | #ifndef CONFIG_SMP |
34db18a0 | 331 | static const unsigned int setup_max_cpus = NR_CPUS; |
1da177e4 LT |
332 | #ifdef CONFIG_X86_LOCAL_APIC |
333 | static void __init smp_init(void) | |
334 | { | |
335 | APIC_init_uniprocessor(); | |
336 | } | |
337 | #else | |
338 | #define smp_init() do { } while (0) | |
339 | #endif | |
340 | ||
e0982e90 | 341 | static inline void setup_nr_cpu_ids(void) { } |
1da177e4 | 342 | static inline void smp_prepare_cpus(unsigned int maxcpus) { } |
1da177e4 LT |
343 | #endif |
344 | ||
30d7e0d4 ABL |
345 | /* |
346 | * We need to store the untouched command line for future reference. | |
347 | * We also need to store the touched command line since the parameter | |
348 | * parsing is performed in place, and we should allow a component to | |
349 | * store reference of name/value for future reference. | |
350 | */ | |
351 | static void __init setup_command_line(char *command_line) | |
352 | { | |
098b081b SS |
353 | saved_command_line = |
354 | memblock_virt_alloc(strlen(boot_command_line) + 1, 0); | |
355 | initcall_command_line = | |
356 | memblock_virt_alloc(strlen(boot_command_line) + 1, 0); | |
357 | static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0); | |
30d7e0d4 ABL |
358 | strcpy (saved_command_line, boot_command_line); |
359 | strcpy (static_command_line, command_line); | |
360 | } | |
361 | ||
1da177e4 LT |
362 | /* |
363 | * We need to finalize in a non-__init function or else race conditions | |
364 | * between the root thread and the init thread may cause start_kernel to | |
365 | * be reaped by free_initmem before the root thread has proceeded to | |
366 | * cpu_idle. | |
367 | * | |
368 | * gcc-3.4 accidentally inlines this function, so use noinline. | |
369 | */ | |
370 | ||
b433c3d4 PZ |
371 | static __initdata DECLARE_COMPLETION(kthreadd_done); |
372 | ||
f99ebf0a | 373 | static noinline void __init_refok rest_init(void) |
1da177e4 | 374 | { |
73c27992 EB |
375 | int pid; |
376 | ||
7db905e6 | 377 | rcu_scheduler_starting(); |
b433c3d4 | 378 | /* |
97158569 | 379 | * We need to spawn init first so that it obtains pid 1, however |
b433c3d4 PZ |
380 | * the init task will end up wanting to create kthreads, which, if |
381 | * we schedule it before we create kthreadd, will OOPS. | |
382 | */ | |
aae5f662 | 383 | kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND); |
1da177e4 | 384 | numa_default_policy(); |
73c27992 | 385 | pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); |
d11c563d | 386 | rcu_read_lock(); |
5cd20455 | 387 | kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); |
d11c563d | 388 | rcu_read_unlock(); |
b433c3d4 | 389 | complete(&kthreadd_done); |
f340c0d1 IM |
390 | |
391 | /* | |
392 | * The boot idle thread must execute schedule() | |
1df21055 | 393 | * at least once to get things moving: |
f340c0d1 | 394 | */ |
1df21055 | 395 | init_idle_bootup_task(current); |
bd2f5536 | 396 | schedule_preempt_disabled(); |
5bfb5d69 | 397 | /* Call into cpu_idle with preempt disabled */ |
a1a04ec3 | 398 | cpu_startup_entry(CPUHP_ONLINE); |
1df21055 | 399 | } |
1da177e4 LT |
400 | |
401 | /* Check for early params. */ | |
9fb48c74 | 402 | static int __init do_early_param(char *param, char *val, const char *unused) |
1da177e4 | 403 | { |
914dcaa8 | 404 | const struct obs_kernel_param *p; |
1da177e4 LT |
405 | |
406 | for (p = __setup_start; p < __setup_end; p++) { | |
b1e4d20c | 407 | if ((p->early && parameq(param, p->str)) || |
18a8bd94 YL |
408 | (strcmp(param, "console") == 0 && |
409 | strcmp(p->str, "earlycon") == 0) | |
410 | ) { | |
1da177e4 | 411 | if (p->setup_func(val) != 0) |
ea676e84 | 412 | pr_warn("Malformed early option '%s'\n", param); |
1da177e4 LT |
413 | } |
414 | } | |
415 | /* We accept everything at this stage. */ | |
416 | return 0; | |
417 | } | |
418 | ||
13977091 MD |
419 | void __init parse_early_options(char *cmdline) |
420 | { | |
026cee00 | 421 | parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param); |
13977091 MD |
422 | } |
423 | ||
1da177e4 LT |
424 | /* Arch code calls this early on, or if not, just before other parsing. */ |
425 | void __init parse_early_param(void) | |
426 | { | |
427 | static __initdata int done = 0; | |
428 | static __initdata char tmp_cmdline[COMMAND_LINE_SIZE]; | |
429 | ||
430 | if (done) | |
431 | return; | |
432 | ||
433 | /* All fall through to do_early_param. */ | |
30d7e0d4 | 434 | strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); |
13977091 | 435 | parse_early_options(tmp_cmdline); |
1da177e4 LT |
436 | done = 1; |
437 | } | |
438 | ||
439 | /* | |
440 | * Activate the first processor. | |
441 | */ | |
442 | ||
44fd2299 SS |
443 | static void __init boot_cpu_init(void) |
444 | { | |
445 | int cpu = smp_processor_id(); | |
446 | /* Mark the boot cpu "present", "online" etc for SMP and UP case */ | |
915441b6 | 447 | set_cpu_online(cpu, true); |
933b0618 | 448 | set_cpu_active(cpu, true); |
915441b6 RR |
449 | set_cpu_present(cpu, true); |
450 | set_cpu_possible(cpu, true); | |
44fd2299 SS |
451 | } |
452 | ||
839ad62e | 453 | void __init __weak smp_setup_processor_id(void) |
033ab7f8 AM |
454 | { |
455 | } | |
456 | ||
eded09cc | 457 | # if THREAD_SIZE >= PAGE_SIZE |
8c9843e5 BH |
458 | void __init __weak thread_info_cache_init(void) |
459 | { | |
460 | } | |
eded09cc | 461 | #endif |
8c9843e5 | 462 | |
444f478f PE |
463 | /* |
464 | * Set up kernel memory allocators | |
465 | */ | |
466 | static void __init mm_init(void) | |
467 | { | |
ca371c0d | 468 | /* |
7fa87ce7 JC |
469 | * page_cgroup requires contiguous pages, |
470 | * bigger than MAX_ORDER unless SPARSEMEM. | |
ca371c0d KH |
471 | */ |
472 | page_cgroup_init_flatmem(); | |
444f478f PE |
473 | mem_init(); |
474 | kmem_cache_init(); | |
099a19d9 | 475 | percpu_init_late(); |
b35f1819 | 476 | pgtable_init(); |
444f478f PE |
477 | vmalloc_init(); |
478 | } | |
479 | ||
722a9f92 | 480 | asmlinkage __visible void __init start_kernel(void) |
1da177e4 LT |
481 | { |
482 | char * command_line; | |
914dcaa8 | 483 | extern const struct kernel_param __start___param[], __stop___param[]; |
033ab7f8 | 484 | |
fbb9ce95 IM |
485 | /* |
486 | * Need to run as early as possible, to initialize the | |
487 | * lockdep hash: | |
488 | */ | |
489 | lockdep_init(); | |
73839c5b | 490 | smp_setup_processor_id(); |
3ac7fe5a | 491 | debug_objects_early_init(); |
42059429 IM |
492 | |
493 | /* | |
494 | * Set up the the initial canary ASAP: | |
495 | */ | |
496 | boot_init_stack_canary(); | |
497 | ||
ddbcc7e8 | 498 | cgroup_init_early(); |
fbb9ce95 IM |
499 | |
500 | local_irq_disable(); | |
2ce802f6 | 501 | early_boot_irqs_disabled = true; |
fbb9ce95 | 502 | |
1da177e4 LT |
503 | /* |
504 | * Interrupts are still disabled. Do necessary setups, then | |
505 | * enable them | |
506 | */ | |
44fd2299 | 507 | boot_cpu_init(); |
1da177e4 | 508 | page_address_init(); |
ea676e84 | 509 | pr_notice("%s", linux_banner); |
1da177e4 | 510 | setup_arch(&command_line); |
6345d24d | 511 | mm_init_cpumask(&init_mm); |
30d7e0d4 | 512 | setup_command_line(command_line); |
e0982e90 | 513 | setup_nr_cpu_ids(); |
d6647bdf | 514 | setup_per_cpu_areas(); |
44fd2299 | 515 | smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ |
1da177e4 | 516 | |
9adb62a5 | 517 | build_all_zonelists(NULL, NULL); |
83b519e8 PE |
518 | page_alloc_init(); |
519 | ||
ea676e84 | 520 | pr_notice("Kernel command line: %s\n", boot_command_line); |
83b519e8 PE |
521 | parse_early_param(); |
522 | parse_args("Booting kernel", static_command_line, __start___param, | |
523 | __stop___param - __start___param, | |
ae82fdb1 | 524 | -1, -1, &unknown_bootoption); |
97ce2c88 JF |
525 | |
526 | jump_label_init(); | |
527 | ||
83b519e8 PE |
528 | /* |
529 | * These use large bootmem allocations and must precede | |
530 | * kmem_cache_init() | |
531 | */ | |
162a7e75 | 532 | setup_log_buf(0); |
83b519e8 | 533 | pidhash_init(); |
83b519e8 PE |
534 | vfs_caches_init_early(); |
535 | sort_main_extable(); | |
536 | trap_init(); | |
444f478f | 537 | mm_init(); |
de03c72c | 538 | |
1da177e4 LT |
539 | /* |
540 | * Set up the scheduler prior starting any interrupts (such as the | |
541 | * timer interrupt). Full topology setup happens at smp_init() | |
542 | * time - but meanwhile we still have a functioning scheduler. | |
543 | */ | |
544 | sched_init(); | |
545 | /* | |
546 | * Disable preemption - early bootup scheduling is extremely | |
547 | * fragile until we cpu_idle() for the first time. | |
548 | */ | |
549 | preempt_disable(); | |
f91eb62f | 550 | if (WARN(!irqs_disabled(), "Interrupts were enabled *very* early, fixing it\n")) |
c4a68306 | 551 | local_irq_disable(); |
9f58a205 | 552 | idr_init_cache(); |
1da177e4 | 553 | rcu_init(); |
d1e43fa5 | 554 | tick_nohz_init(); |
65f382fd | 555 | context_tracking_init(); |
773e3eb7 | 556 | radix_tree_init(); |
0b8f1efa YL |
557 | /* init some links before init_ISA_irqs() */ |
558 | early_irq_init(); | |
1da177e4 | 559 | init_IRQ(); |
ad2b1353 | 560 | tick_init(); |
1da177e4 | 561 | init_timers(); |
c0a31329 | 562 | hrtimers_init(); |
1da177e4 | 563 | softirq_init(); |
ad596171 | 564 | timekeeping_init(); |
88fecaa2 | 565 | time_init(); |
38ff87f7 | 566 | sched_clock_postinit(); |
9e630205 | 567 | perf_event_init(); |
93e02814 | 568 | profile_init(); |
d8ad7d11 | 569 | call_function_init(); |
f91eb62f | 570 | WARN(!irqs_disabled(), "Interrupts were enabled early\n"); |
2ce802f6 | 571 | early_boot_irqs_disabled = false; |
93e02814 | 572 | local_irq_enable(); |
dcce284a | 573 | |
7e85ee0c | 574 | kmem_cache_init_late(); |
1da177e4 LT |
575 | |
576 | /* | |
577 | * HACK ALERT! This is early. We're enabling the console before | |
578 | * we've done PCI setups etc, and console_init() must be aware of | |
579 | * this. But we do want output early, in case something goes wrong. | |
580 | */ | |
581 | console_init(); | |
582 | if (panic_later) | |
499a4584 TH |
583 | panic("Too many boot %s vars at `%s'", panic_later, |
584 | panic_param); | |
fbb9ce95 IM |
585 | |
586 | lockdep_info(); | |
587 | ||
9a11b49a IM |
588 | /* |
589 | * Need to run this when irqs are enabled, because it wants | |
590 | * to self-test [hard/soft]-irqs on/off lock inversion bugs | |
591 | * too: | |
592 | */ | |
593 | locking_selftest(); | |
594 | ||
1da177e4 LT |
595 | #ifdef CONFIG_BLK_DEV_INITRD |
596 | if (initrd_start && !initrd_below_start_ok && | |
bd673c7c | 597 | page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { |
ea676e84 | 598 | pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", |
bd673c7c GU |
599 | page_to_pfn(virt_to_page((void *)initrd_start)), |
600 | min_low_pfn); | |
1da177e4 LT |
601 | initrd_start = 0; |
602 | } | |
603 | #endif | |
94b6da5a | 604 | page_cgroup_init(); |
3ac7fe5a | 605 | debug_objects_mem_init(); |
9b090f2d | 606 | kmemleak_init(); |
e7c8d5c9 | 607 | setup_per_cpu_pageset(); |
1da177e4 LT |
608 | numa_policy_init(); |
609 | if (late_time_init) | |
610 | late_time_init(); | |
fa84e9ee | 611 | sched_clock_init(); |
1da177e4 LT |
612 | calibrate_delay(); |
613 | pidmap_init(); | |
1da177e4 | 614 | anon_vma_init(); |
c4e1acbb | 615 | acpi_early_init(); |
11520e5e | 616 | #ifdef CONFIG_X86 |
83e68189 | 617 | if (efi_enabled(EFI_RUNTIME_SERVICES)) |
11520e5e LT |
618 | efi_enter_virtual_mode(); |
619 | #endif | |
8c9843e5 | 620 | thread_info_cache_init(); |
d84f4f99 | 621 | cred_init(); |
4481374c | 622 | fork_init(totalram_pages); |
1da177e4 LT |
623 | proc_caches_init(); |
624 | buffer_init(); | |
1da177e4 LT |
625 | key_init(); |
626 | security_init(); | |
0b4b3827 | 627 | dbg_late_init(); |
4481374c | 628 | vfs_caches_init(totalram_pages); |
1da177e4 LT |
629 | signals_init(); |
630 | /* rootfs populating might need page-writeback */ | |
631 | page_writeback_init(); | |
632 | #ifdef CONFIG_PROC_FS | |
633 | proc_root_init(); | |
634 | #endif | |
ddbcc7e8 | 635 | cgroup_init(); |
1da177e4 | 636 | cpuset_init(); |
c757249a | 637 | taskstats_init_early(); |
ca74e92b | 638 | delayacct_init(); |
1da177e4 LT |
639 | |
640 | check_bugs(); | |
641 | ||
6ae6996a | 642 | sfi_init_late(); |
1da177e4 | 643 | |
83e68189 | 644 | if (efi_enabled(EFI_RUNTIME_SERVICES)) { |
2223af38 | 645 | efi_late_init(); |
78510792 | 646 | efi_free_boot_services(); |
2223af38 | 647 | } |
78510792 | 648 | |
68bf21aa SR |
649 | ftrace_init(); |
650 | ||
1da177e4 LT |
651 | /* Do the rest non-__init'ed, we're now alive */ |
652 | rest_init(); | |
653 | } | |
654 | ||
b99b87f7 PO |
655 | /* Call all constructor functions linked into the kernel. */ |
656 | static void __init do_ctors(void) | |
657 | { | |
658 | #ifdef CONFIG_CONSTRUCTORS | |
196a15b4 | 659 | ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; |
b99b87f7 | 660 | |
196a15b4 HS |
661 | for (; fn < (ctor_fn_t *) __ctors_end; fn++) |
662 | (*fn)(); | |
b99b87f7 PO |
663 | #endif |
664 | } | |
665 | ||
2329abfa | 666 | bool initcall_debug; |
d0ea3d7d | 667 | core_param(initcall_debug, initcall_debug, bool, 0644); |
1da177e4 | 668 | |
7b0b73d7 PB |
669 | #ifdef CONFIG_KALLSYMS |
670 | struct blacklist_entry { | |
671 | struct list_head next; | |
672 | char *buf; | |
673 | }; | |
674 | ||
675 | static __initdata_or_module LIST_HEAD(blacklisted_initcalls); | |
676 | ||
677 | static int __init initcall_blacklist(char *str) | |
678 | { | |
679 | char *str_entry; | |
680 | struct blacklist_entry *entry; | |
681 | ||
682 | /* str argument is a comma-separated list of functions */ | |
683 | do { | |
684 | str_entry = strsep(&str, ","); | |
685 | if (str_entry) { | |
686 | pr_debug("blacklisting initcall %s\n", str_entry); | |
687 | entry = alloc_bootmem(sizeof(*entry)); | |
688 | entry->buf = alloc_bootmem(strlen(str_entry) + 1); | |
689 | strcpy(entry->buf, str_entry); | |
690 | list_add(&entry->next, &blacklisted_initcalls); | |
691 | } | |
692 | } while (str_entry); | |
693 | ||
694 | return 0; | |
695 | } | |
696 | ||
697 | static bool __init_or_module initcall_blacklisted(initcall_t fn) | |
698 | { | |
699 | struct list_head *tmp; | |
700 | struct blacklist_entry *entry; | |
701 | char *fn_name; | |
702 | ||
703 | fn_name = kasprintf(GFP_KERNEL, "%pf", fn); | |
704 | if (!fn_name) | |
705 | return false; | |
706 | ||
707 | list_for_each(tmp, &blacklisted_initcalls) { | |
708 | entry = list_entry(tmp, struct blacklist_entry, next); | |
709 | if (!strcmp(fn_name, entry->buf)) { | |
710 | pr_debug("initcall %s blacklisted\n", fn_name); | |
711 | kfree(fn_name); | |
712 | return true; | |
713 | } | |
714 | } | |
715 | ||
716 | kfree(fn_name); | |
717 | return false; | |
718 | } | |
719 | #else | |
720 | static int __init initcall_blacklist(char *str) | |
721 | { | |
722 | pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); | |
723 | return 0; | |
724 | } | |
725 | ||
726 | static bool __init_or_module initcall_blacklisted(initcall_t fn) | |
727 | { | |
728 | return false; | |
729 | } | |
730 | #endif | |
731 | __setup("initcall_blacklist=", initcall_blacklist); | |
732 | ||
e4461271 | 733 | static int __init_or_module do_one_initcall_debug(initcall_t fn) |
1da177e4 | 734 | { |
74239072 | 735 | ktime_t calltime, delta, rettime; |
30dbb20e AW |
736 | unsigned long long duration; |
737 | int ret; | |
1da177e4 | 738 | |
d62cf815 | 739 | printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current)); |
22c5c03b | 740 | calltime = ktime_get(); |
30dbb20e | 741 | ret = fn(); |
22c5c03b KW |
742 | rettime = ktime_get(); |
743 | delta = ktime_sub(rettime, calltime); | |
744 | duration = (unsigned long long) ktime_to_ns(delta) >> 10; | |
d62cf815 | 745 | printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", |
ea676e84 | 746 | fn, ret, duration); |
1da177e4 | 747 | |
22c5c03b KW |
748 | return ret; |
749 | } | |
750 | ||
e4461271 | 751 | int __init_or_module do_one_initcall(initcall_t fn) |
22c5c03b KW |
752 | { |
753 | int count = preempt_count(); | |
754 | int ret; | |
ff1c8fac | 755 | char msgbuf[64]; |
22c5c03b | 756 | |
7b0b73d7 PB |
757 | if (initcall_blacklisted(fn)) |
758 | return -EPERM; | |
759 | ||
22c5c03b KW |
760 | if (initcall_debug) |
761 | ret = do_one_initcall_debug(fn); | |
762 | else | |
763 | ret = fn(); | |
8f0c45cd | 764 | |
e0df154f | 765 | msgbuf[0] = 0; |
e662e1cf | 766 | |
e0df154f | 767 | if (preempt_count() != count) { |
bf5d770b | 768 | sprintf(msgbuf, "preemption imbalance "); |
4a2b4b22 | 769 | preempt_count_set(count); |
1da177e4 | 770 | } |
e0df154f | 771 | if (irqs_disabled()) { |
a76bfd0d | 772 | strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); |
e0df154f LT |
773 | local_irq_enable(); |
774 | } | |
f91eb62f | 775 | WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf); |
59f9415f | 776 | |
30dbb20e | 777 | return ret; |
e0df154f LT |
778 | } |
779 | ||
780 | ||
026cee00 PM |
781 | extern initcall_t __initcall_start[]; |
782 | extern initcall_t __initcall0_start[]; | |
783 | extern initcall_t __initcall1_start[]; | |
784 | extern initcall_t __initcall2_start[]; | |
785 | extern initcall_t __initcall3_start[]; | |
786 | extern initcall_t __initcall4_start[]; | |
787 | extern initcall_t __initcall5_start[]; | |
788 | extern initcall_t __initcall6_start[]; | |
789 | extern initcall_t __initcall7_start[]; | |
790 | extern initcall_t __initcall_end[]; | |
791 | ||
792 | static initcall_t *initcall_levels[] __initdata = { | |
793 | __initcall0_start, | |
794 | __initcall1_start, | |
795 | __initcall2_start, | |
796 | __initcall3_start, | |
797 | __initcall4_start, | |
798 | __initcall5_start, | |
799 | __initcall6_start, | |
800 | __initcall7_start, | |
801 | __initcall_end, | |
802 | }; | |
803 | ||
96263d28 | 804 | /* Keep these in sync with initcalls in include/linux/init.h */ |
026cee00 | 805 | static char *initcall_level_names[] __initdata = { |
9fb48c74 JC |
806 | "early", |
807 | "core", | |
808 | "postcore", | |
809 | "arch", | |
810 | "subsys", | |
811 | "fs", | |
812 | "device", | |
813 | "late", | |
026cee00 PM |
814 | }; |
815 | ||
026cee00 | 816 | static void __init do_initcall_level(int level) |
e0df154f | 817 | { |
026cee00 | 818 | extern const struct kernel_param __start___param[], __stop___param[]; |
196a15b4 | 819 | initcall_t *fn; |
e0df154f | 820 | |
08746a65 | 821 | strcpy(initcall_command_line, saved_command_line); |
026cee00 | 822 | parse_args(initcall_level_names[level], |
08746a65 | 823 | initcall_command_line, __start___param, |
026cee00 PM |
824 | __stop___param - __start___param, |
825 | level, level, | |
eb157427 | 826 | &repair_env_string); |
026cee00 PM |
827 | |
828 | for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) | |
196a15b4 | 829 | do_one_initcall(*fn); |
1da177e4 LT |
830 | } |
831 | ||
026cee00 PM |
832 | static void __init do_initcalls(void) |
833 | { | |
834 | int level; | |
835 | ||
19efb72f | 836 | for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) |
026cee00 PM |
837 | do_initcall_level(level); |
838 | } | |
839 | ||
1da177e4 LT |
840 | /* |
841 | * Ok, the machine is now initialized. None of the devices | |
842 | * have been touched yet, but the CPU subsystem is up and | |
843 | * running, and memory and process management works. | |
844 | * | |
845 | * Now we can finally start doing some real work.. | |
846 | */ | |
847 | static void __init do_basic_setup(void) | |
848 | { | |
759ee091 | 849 | cpuset_init_smp(); |
1da177e4 | 850 | usermodehelper_init(); |
41ffe5d5 | 851 | shmem_init(); |
1da177e4 | 852 | driver_init(); |
b04c3afb | 853 | init_irq_proc(); |
b99b87f7 | 854 | do_ctors(); |
d5767c53 | 855 | usermodehelper_enable(); |
b0f84374 | 856 | do_initcalls(); |
47d06e53 | 857 | random_int_secret_init(); |
1da177e4 LT |
858 | } |
859 | ||
7babe8db | 860 | static void __init do_pre_smp_initcalls(void) |
c2147a50 | 861 | { |
196a15b4 | 862 | initcall_t *fn; |
c2147a50 | 863 | |
026cee00 | 864 | for (fn = __initcall_start; fn < __initcall0_start; fn++) |
196a15b4 | 865 | do_one_initcall(*fn); |
c2147a50 EGM |
866 | } |
867 | ||
bb813f4c TH |
868 | /* |
869 | * This function requests modules which should be loaded by default and is | |
870 | * called twice right after initrd is mounted and right before init is | |
871 | * exec'd. If such modules are on either initrd or rootfs, they will be | |
872 | * loaded before control is passed to userland. | |
873 | */ | |
874 | void __init load_default_modules(void) | |
875 | { | |
876 | load_default_elevator_module(); | |
877 | } | |
878 | ||
a74fb73c | 879 | static int run_init_process(const char *init_filename) |
1da177e4 LT |
880 | { |
881 | argv_init[0] = init_filename; | |
c4ad8f98 | 882 | return do_execve(getname_kernel(init_filename), |
ae903caa AV |
883 | (const char __user *const __user *)argv_init, |
884 | (const char __user *const __user *)envp_init); | |
1da177e4 LT |
885 | } |
886 | ||
ba24762b MO |
887 | static int try_to_run_init_process(const char *init_filename) |
888 | { | |
889 | int ret; | |
890 | ||
891 | ret = run_init_process(init_filename); | |
892 | ||
893 | if (ret && ret != -ENOENT) { | |
894 | pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", | |
895 | init_filename, ret); | |
896 | } | |
897 | ||
898 | return ret; | |
899 | } | |
900 | ||
f80b0c90 | 901 | static noinline void __init kernel_init_freeable(void); |
d6b21238 AV |
902 | |
903 | static int __ref kernel_init(void *unused) | |
ee5bfa64 | 904 | { |
ba24762b MO |
905 | int ret; |
906 | ||
d6b21238 | 907 | kernel_init_freeable(); |
22a9d645 AV |
908 | /* need to finish all async __init code before freeing the memory */ |
909 | async_synchronize_full(); | |
ee5bfa64 | 910 | free_initmem(); |
ee5bfa64 VG |
911 | mark_rodata_ro(); |
912 | system_state = SYSTEM_RUNNING; | |
913 | numa_default_policy(); | |
914 | ||
4a9d4b02 | 915 | flush_delayed_fput(); |
fae5fa44 | 916 | |
ee5bfa64 | 917 | if (ramdisk_execute_command) { |
ba24762b MO |
918 | ret = run_init_process(ramdisk_execute_command); |
919 | if (!ret) | |
a74fb73c | 920 | return 0; |
ba24762b MO |
921 | pr_err("Failed to execute %s (error %d)\n", |
922 | ramdisk_execute_command, ret); | |
ee5bfa64 VG |
923 | } |
924 | ||
925 | /* | |
926 | * We try each of these until one succeeds. | |
927 | * | |
928 | * The Bourne shell can be used instead of init if we are | |
929 | * trying to recover a really broken machine. | |
930 | */ | |
931 | if (execute_command) { | |
ba24762b MO |
932 | ret = run_init_process(execute_command); |
933 | if (!ret) | |
a74fb73c | 934 | return 0; |
ba24762b MO |
935 | pr_err("Failed to execute %s (error %d). Attempting defaults...\n", |
936 | execute_command, ret); | |
ee5bfa64 | 937 | } |
ba24762b MO |
938 | if (!try_to_run_init_process("/sbin/init") || |
939 | !try_to_run_init_process("/etc/init") || | |
940 | !try_to_run_init_process("/bin/init") || | |
941 | !try_to_run_init_process("/bin/sh")) | |
a74fb73c | 942 | return 0; |
ee5bfa64 | 943 | |
ba24762b | 944 | panic("No working init found. Try passing init= option to kernel. " |
9a85b8d6 | 945 | "See Linux Documentation/init.txt for guidance."); |
ee5bfa64 VG |
946 | } |
947 | ||
f80b0c90 | 948 | static noinline void __init kernel_init_freeable(void) |
1da177e4 | 949 | { |
b433c3d4 PZ |
950 | /* |
951 | * Wait until kthreadd is all set-up. | |
952 | */ | |
953 | wait_for_completion(&kthreadd_done); | |
31a67102 LT |
954 | |
955 | /* Now the scheduler is fully set up and can do blocking allocations */ | |
956 | gfp_allowed_mask = __GFP_BITS_MASK; | |
957 | ||
58568d2a MX |
958 | /* |
959 | * init can allocate pages on any node | |
960 | */ | |
3c466d46 | 961 | set_mems_allowed(node_states[N_MEMORY]); |
1da177e4 LT |
962 | /* |
963 | * init can run on any cpu. | |
964 | */ | |
1a2142af | 965 | set_cpus_allowed_ptr(current, cpu_all_mask); |
1da177e4 | 966 | |
9ec52099 CLG |
967 | cad_pid = task_pid(current); |
968 | ||
ca74a6f8 | 969 | smp_prepare_cpus(setup_max_cpus); |
1da177e4 LT |
970 | |
971 | do_pre_smp_initcalls(); | |
004417a6 | 972 | lockup_detector_init(); |
1da177e4 | 973 | |
1da177e4 LT |
974 | smp_init(); |
975 | sched_init_smp(); | |
976 | ||
1da177e4 LT |
977 | do_basic_setup(); |
978 | ||
2bd3a997 EB |
979 | /* Open the /dev/console on the rootfs, this should never fail */ |
980 | if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) | |
ea676e84 | 981 | pr_err("Warning: unable to open an initial console.\n"); |
2bd3a997 EB |
982 | |
983 | (void) sys_dup(0); | |
984 | (void) sys_dup(0); | |
1da177e4 LT |
985 | /* |
986 | * check if there is an early userspace init. If yes, let it do all | |
987 | * the work | |
988 | */ | |
ffdfc409 OJ |
989 | |
990 | if (!ramdisk_execute_command) | |
991 | ramdisk_execute_command = "/init"; | |
992 | ||
993 | if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) { | |
994 | ramdisk_execute_command = NULL; | |
1da177e4 | 995 | prepare_namespace(); |
ffdfc409 | 996 | } |
1da177e4 LT |
997 | |
998 | /* | |
999 | * Ok, we have completed the initial bootup, and | |
1000 | * we're essentially up and running. Get rid of the | |
1001 | * initmem segments and start the user-mode stuff.. | |
1002 | */ | |
bb813f4c TH |
1003 | |
1004 | /* rootfs is available now, try loading default modules */ | |
1005 | load_default_modules(); | |
1da177e4 | 1006 | } |