]>
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> | |
21 | #include <linux/utsname.h> | |
22 | #include <linux/ioport.h> | |
23 | #include <linux/init.h> | |
24 | #include <linux/smp_lock.h> | |
25 | #include <linux/initrd.h> | |
1da177e4 | 26 | #include <linux/bootmem.h> |
4a7a16dc | 27 | #include <linux/acpi.h> |
1da177e4 LT |
28 | #include <linux/tty.h> |
29 | #include <linux/gfp.h> | |
30 | #include <linux/percpu.h> | |
31 | #include <linux/kmod.h> | |
db64fe02 | 32 | #include <linux/vmalloc.h> |
1da177e4 | 33 | #include <linux/kernel_stat.h> |
d7cd5611 | 34 | #include <linux/start_kernel.h> |
1da177e4 | 35 | #include <linux/security.h> |
3d442233 | 36 | #include <linux/smp.h> |
1da177e4 LT |
37 | #include <linux/workqueue.h> |
38 | #include <linux/profile.h> | |
39 | #include <linux/rcupdate.h> | |
40 | #include <linux/moduleparam.h> | |
41 | #include <linux/kallsyms.h> | |
42 | #include <linux/writeback.h> | |
43 | #include <linux/cpu.h> | |
44 | #include <linux/cpuset.h> | |
ddbcc7e8 | 45 | #include <linux/cgroup.h> |
1da177e4 | 46 | #include <linux/efi.h> |
906568c9 | 47 | #include <linux/tick.h> |
6168a702 | 48 | #include <linux/interrupt.h> |
c757249a | 49 | #include <linux/taskstats_kern.h> |
ca74e92b | 50 | #include <linux/delayacct.h> |
1da177e4 LT |
51 | #include <linux/unistd.h> |
52 | #include <linux/rmap.h> | |
53 | #include <linux/mempolicy.h> | |
54 | #include <linux/key.h> | |
b6cd0b77 | 55 | #include <linux/buffer_head.h> |
94b6da5a | 56 | #include <linux/page_cgroup.h> |
9a11b49a | 57 | #include <linux/debug_locks.h> |
3ac7fe5a | 58 | #include <linux/debugobjects.h> |
fbb9ce95 | 59 | #include <linux/lockdep.h> |
3c7b4e6b | 60 | #include <linux/kmemleak.h> |
84d73786 | 61 | #include <linux/pid_namespace.h> |
1f21782e | 62 | #include <linux/device.h> |
73c27992 | 63 | #include <linux/kthread.h> |
e6fe6649 | 64 | #include <linux/sched.h> |
a1c9eea9 | 65 | #include <linux/signal.h> |
199f0ca5 | 66 | #include <linux/idr.h> |
68bf21aa | 67 | #include <linux/ftrace.h> |
22a9d645 | 68 | #include <linux/async.h> |
dfec072e | 69 | #include <linux/kmemcheck.h> |
02af61bb | 70 | #include <linux/kmemtrace.h> |
2b2af54a | 71 | #include <linux/shmem_fs.h> |
3f5ec136 | 72 | #include <trace/boot.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); | |
92 | extern void free_initmem(void); | |
37b73c82 AV |
93 | #ifndef CONFIG_DEBUG_RODATA |
94 | static inline void mark_rodata_ro(void) { } | |
95 | #endif | |
1da177e4 LT |
96 | |
97 | #ifdef CONFIG_TC | |
98 | extern void tc_init(void); | |
99 | #endif | |
100 | ||
a6826048 | 101 | enum system_states system_state __read_mostly; |
1da177e4 LT |
102 | EXPORT_SYMBOL(system_state); |
103 | ||
104 | /* | |
105 | * Boot command-line arguments | |
106 | */ | |
107 | #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT | |
108 | #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT | |
109 | ||
110 | extern void time_init(void); | |
111 | /* Default late time init is NULL. archs can override this later. */ | |
d2e3192b | 112 | void (*__initdata late_time_init)(void); |
1da177e4 LT |
113 | extern void softirq_init(void); |
114 | ||
30d7e0d4 ABL |
115 | /* Untouched command line saved by arch-specific code. */ |
116 | char __initdata boot_command_line[COMMAND_LINE_SIZE]; | |
117 | /* Untouched saved command line (eg. for /proc) */ | |
118 | char *saved_command_line; | |
119 | /* Command line for parameter parsing */ | |
120 | static char *static_command_line; | |
1da177e4 LT |
121 | |
122 | static char *execute_command; | |
ffdfc409 | 123 | static char *ramdisk_execute_command; |
1da177e4 | 124 | |
8b3b2955 | 125 | #ifdef CONFIG_SMP |
1da177e4 | 126 | /* Setup configured maximum number of CPUs to activate */ |
ca74a6f8 | 127 | unsigned int __initdata setup_max_cpus = NR_CPUS; |
7e96287d | 128 | |
1da177e4 LT |
129 | /* |
130 | * Setup routine for controlling SMP activation | |
131 | * | |
132 | * Command-line option of "nosmp" or "maxcpus=0" will disable SMP | |
133 | * activation entirely (the MPS table probe still happens, though). | |
134 | * | |
135 | * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer | |
136 | * greater than 0, limits the maximum number of CPUs activated in | |
137 | * SMP mode to <NUM>. | |
138 | */ | |
65a4e574 IM |
139 | |
140 | void __weak arch_disable_smp_support(void) { } | |
61ec7567 | 141 | |
1da177e4 LT |
142 | static int __init nosmp(char *str) |
143 | { | |
ca74a6f8 | 144 | setup_max_cpus = 0; |
65a4e574 IM |
145 | arch_disable_smp_support(); |
146 | ||
8b3b2955 | 147 | return 0; |
1da177e4 LT |
148 | } |
149 | ||
8b3b2955 | 150 | early_param("nosmp", nosmp); |
1da177e4 LT |
151 | |
152 | static int __init maxcpus(char *str) | |
153 | { | |
ca74a6f8 AK |
154 | get_option(&str, &setup_max_cpus); |
155 | if (setup_max_cpus == 0) | |
65a4e574 | 156 | arch_disable_smp_support(); |
61ec7567 LB |
157 | |
158 | return 0; | |
1da177e4 LT |
159 | } |
160 | ||
81340977 | 161 | early_param("maxcpus", maxcpus); |
8b3b2955 | 162 | #else |
65a4e574 | 163 | const unsigned int setup_max_cpus = NR_CPUS; |
8b3b2955 JB |
164 | #endif |
165 | ||
166 | /* | |
167 | * If set, this is an indication to the drivers that reset the underlying | |
168 | * device before going ahead with the initialization otherwise driver might | |
169 | * rely on the BIOS and skip the reset operation. | |
170 | * | |
171 | * This is useful if kernel is booting in an unreliable environment. | |
172 | * For ex. kdump situaiton where previous kernel has crashed, BIOS has been | |
173 | * skipped and devices will be in unknown state. | |
174 | */ | |
175 | unsigned int reset_devices; | |
176 | EXPORT_SYMBOL(reset_devices); | |
1da177e4 | 177 | |
7e96287d VG |
178 | static int __init set_reset_devices(char *str) |
179 | { | |
180 | reset_devices = 1; | |
181 | return 1; | |
182 | } | |
183 | ||
184 | __setup("reset_devices", set_reset_devices); | |
185 | ||
1da177e4 LT |
186 | static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; |
187 | char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; | |
188 | static const char *panic_later, *panic_param; | |
189 | ||
190 | extern struct obs_kernel_param __setup_start[], __setup_end[]; | |
191 | ||
192 | static int __init obsolete_checksetup(char *line) | |
193 | { | |
194 | struct obs_kernel_param *p; | |
33df0d19 | 195 | int had_early_param = 0; |
1da177e4 LT |
196 | |
197 | p = __setup_start; | |
198 | do { | |
199 | int n = strlen(p->str); | |
200 | if (!strncmp(line, p->str, n)) { | |
201 | if (p->early) { | |
33df0d19 RR |
202 | /* Already done in parse_early_param? |
203 | * (Needs exact match on param part). | |
204 | * Keep iterating, as we can have early | |
205 | * params and __setups of same names 8( */ | |
1da177e4 | 206 | if (line[n] == '\0' || line[n] == '=') |
33df0d19 | 207 | had_early_param = 1; |
1da177e4 LT |
208 | } else if (!p->setup_func) { |
209 | printk(KERN_WARNING "Parameter %s is obsolete," | |
210 | " ignored\n", p->str); | |
211 | return 1; | |
212 | } else if (p->setup_func(line + n)) | |
213 | return 1; | |
214 | } | |
215 | p++; | |
216 | } while (p < __setup_end); | |
33df0d19 RR |
217 | |
218 | return had_early_param; | |
1da177e4 LT |
219 | } |
220 | ||
221 | /* | |
222 | * This should be approx 2 Bo*oMips to start (note initial shift), and will | |
223 | * still work even if initially too large, it will just take slightly longer | |
224 | */ | |
225 | unsigned long loops_per_jiffy = (1<<12); | |
226 | ||
227 | EXPORT_SYMBOL(loops_per_jiffy); | |
228 | ||
229 | static int __init debug_kernel(char *str) | |
230 | { | |
1da177e4 | 231 | console_loglevel = 10; |
f6f21c81 | 232 | return 0; |
1da177e4 LT |
233 | } |
234 | ||
235 | static int __init quiet_kernel(char *str) | |
236 | { | |
1da177e4 | 237 | console_loglevel = 4; |
f6f21c81 | 238 | return 0; |
1da177e4 LT |
239 | } |
240 | ||
f6f21c81 YL |
241 | early_param("debug", debug_kernel); |
242 | early_param("quiet", quiet_kernel); | |
1da177e4 LT |
243 | |
244 | static int __init loglevel(char *str) | |
245 | { | |
246 | get_option(&str, &console_loglevel); | |
d9d4fcfe | 247 | return 0; |
1da177e4 LT |
248 | } |
249 | ||
f6f21c81 | 250 | early_param("loglevel", loglevel); |
1da177e4 LT |
251 | |
252 | /* | |
253 | * Unknown boot options get handed to init, unless they look like | |
254 | * failed parameters | |
255 | */ | |
256 | static int __init unknown_bootoption(char *param, char *val) | |
257 | { | |
258 | /* Change NUL term back to "=", to make "param" the whole string. */ | |
259 | if (val) { | |
260 | /* param=val or param="val"? */ | |
261 | if (val == param+strlen(param)+1) | |
262 | val[-1] = '='; | |
263 | else if (val == param+strlen(param)+2) { | |
264 | val[-2] = '='; | |
265 | memmove(val-1, val, strlen(val)+1); | |
266 | val--; | |
267 | } else | |
268 | BUG(); | |
269 | } | |
270 | ||
271 | /* Handle obsolete-style parameters */ | |
272 | if (obsolete_checksetup(param)) | |
273 | return 0; | |
274 | ||
275 | /* | |
211fee8a | 276 | * Preemptive maintenance for "why didn't my misspelled command |
1da177e4 LT |
277 | * line work?" |
278 | */ | |
279 | if (strchr(param, '.') && (!val || strchr(param, '.') < val)) { | |
280 | printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param); | |
281 | return 0; | |
282 | } | |
283 | ||
284 | if (panic_later) | |
285 | return 0; | |
286 | ||
287 | if (val) { | |
288 | /* Environment option */ | |
289 | unsigned int i; | |
290 | for (i = 0; envp_init[i]; i++) { | |
291 | if (i == MAX_INIT_ENVS) { | |
292 | panic_later = "Too many boot env vars at `%s'"; | |
293 | panic_param = param; | |
294 | } | |
295 | if (!strncmp(param, envp_init[i], val - param)) | |
296 | break; | |
297 | } | |
298 | envp_init[i] = param; | |
299 | } else { | |
300 | /* Command line option */ | |
301 | unsigned int i; | |
302 | for (i = 0; argv_init[i]; i++) { | |
303 | if (i == MAX_INIT_ARGS) { | |
304 | panic_later = "Too many boot init vars at `%s'"; | |
305 | panic_param = param; | |
306 | } | |
307 | } | |
308 | argv_init[i] = param; | |
309 | } | |
310 | return 0; | |
311 | } | |
312 | ||
12d6f21e IM |
313 | #ifdef CONFIG_DEBUG_PAGEALLOC |
314 | int __read_mostly debug_pagealloc_enabled = 0; | |
315 | #endif | |
316 | ||
1da177e4 LT |
317 | static int __init init_setup(char *str) |
318 | { | |
319 | unsigned int i; | |
320 | ||
321 | execute_command = str; | |
322 | /* | |
323 | * In case LILO is going to boot us with default command line, | |
324 | * it prepends "auto" before the whole cmdline which makes | |
325 | * the shell think it should execute a script with such name. | |
326 | * So we ignore all arguments entered _before_ init=... [MJ] | |
327 | */ | |
328 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
329 | argv_init[i] = NULL; | |
330 | return 1; | |
331 | } | |
332 | __setup("init=", init_setup); | |
333 | ||
ffdfc409 OJ |
334 | static int __init rdinit_setup(char *str) |
335 | { | |
336 | unsigned int i; | |
337 | ||
338 | ramdisk_execute_command = str; | |
339 | /* See "auto" comment in init_setup */ | |
340 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
341 | argv_init[i] = NULL; | |
342 | return 1; | |
343 | } | |
344 | __setup("rdinit=", rdinit_setup); | |
345 | ||
1da177e4 LT |
346 | #ifndef CONFIG_SMP |
347 | ||
348 | #ifdef CONFIG_X86_LOCAL_APIC | |
349 | static void __init smp_init(void) | |
350 | { | |
351 | APIC_init_uniprocessor(); | |
352 | } | |
353 | #else | |
354 | #define smp_init() do { } while (0) | |
355 | #endif | |
356 | ||
e0982e90 | 357 | static inline void setup_nr_cpu_ids(void) { } |
1da177e4 LT |
358 | static inline void smp_prepare_cpus(unsigned int maxcpus) { } |
359 | ||
360 | #else | |
361 | ||
321a8e9d MT |
362 | #if NR_CPUS > BITS_PER_LONG |
363 | cpumask_t cpu_mask_all __read_mostly = CPU_MASK_ALL; | |
364 | EXPORT_SYMBOL(cpu_mask_all); | |
365 | #endif | |
366 | ||
e0982e90 MT |
367 | /* Setup number of possible processor ids */ |
368 | int nr_cpu_ids __read_mostly = NR_CPUS; | |
369 | EXPORT_SYMBOL(nr_cpu_ids); | |
370 | ||
371 | /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */ | |
372 | static void __init setup_nr_cpu_ids(void) | |
373 | { | |
e0c0ba73 | 374 | nr_cpu_ids = find_last_bit(cpumask_bits(cpu_possible_mask),NR_CPUS) + 1; |
e0982e90 MT |
375 | } |
376 | ||
1da177e4 LT |
377 | /* Called by boot processor to activate the rest. */ |
378 | static void __init smp_init(void) | |
379 | { | |
53b8a315 | 380 | unsigned int cpu; |
1da177e4 | 381 | |
e761b772 MK |
382 | /* |
383 | * Set up the current CPU as possible to migrate to. | |
384 | * The other ones will be done by cpu_up/cpu_down() | |
385 | */ | |
2b17fa50 | 386 | set_cpu_active(smp_processor_id(), true); |
e761b772 | 387 | |
1da177e4 | 388 | /* FIXME: This should be done in userspace --RR */ |
53b8a315 | 389 | for_each_present_cpu(cpu) { |
ca74a6f8 | 390 | if (num_online_cpus() >= setup_max_cpus) |
1da177e4 | 391 | break; |
53b8a315 CL |
392 | if (!cpu_online(cpu)) |
393 | cpu_up(cpu); | |
1da177e4 LT |
394 | } |
395 | ||
396 | /* Any cleanup work */ | |
397 | printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus()); | |
ca74a6f8 | 398 | smp_cpus_done(setup_max_cpus); |
1da177e4 LT |
399 | } |
400 | ||
401 | #endif | |
402 | ||
30d7e0d4 ABL |
403 | /* |
404 | * We need to store the untouched command line for future reference. | |
405 | * We also need to store the touched command line since the parameter | |
406 | * parsing is performed in place, and we should allow a component to | |
407 | * store reference of name/value for future reference. | |
408 | */ | |
409 | static void __init setup_command_line(char *command_line) | |
410 | { | |
411 | saved_command_line = alloc_bootmem(strlen (boot_command_line)+1); | |
412 | static_command_line = alloc_bootmem(strlen (command_line)+1); | |
413 | strcpy (saved_command_line, boot_command_line); | |
414 | strcpy (static_command_line, command_line); | |
415 | } | |
416 | ||
1da177e4 LT |
417 | /* |
418 | * We need to finalize in a non-__init function or else race conditions | |
419 | * between the root thread and the init thread may cause start_kernel to | |
420 | * be reaped by free_initmem before the root thread has proceeded to | |
421 | * cpu_idle. | |
422 | * | |
423 | * gcc-3.4 accidentally inlines this function, so use noinline. | |
424 | */ | |
425 | ||
f99ebf0a | 426 | static noinline void __init_refok rest_init(void) |
1da177e4 LT |
427 | __releases(kernel_lock) |
428 | { | |
73c27992 EB |
429 | int pid; |
430 | ||
7db905e6 | 431 | rcu_scheduler_starting(); |
aae5f662 | 432 | kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND); |
1da177e4 | 433 | numa_default_policy(); |
73c27992 | 434 | pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); |
5cd20455 | 435 | kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); |
1da177e4 | 436 | unlock_kernel(); |
f340c0d1 IM |
437 | |
438 | /* | |
439 | * The boot idle thread must execute schedule() | |
1df21055 | 440 | * at least once to get things moving: |
f340c0d1 | 441 | */ |
1df21055 | 442 | init_idle_bootup_task(current); |
5bfb5d69 | 443 | preempt_enable_no_resched(); |
f340c0d1 | 444 | schedule(); |
5bfb5d69 | 445 | preempt_disable(); |
f340c0d1 | 446 | |
5bfb5d69 | 447 | /* Call into cpu_idle with preempt disabled */ |
1da177e4 | 448 | cpu_idle(); |
1df21055 | 449 | } |
1da177e4 LT |
450 | |
451 | /* Check for early params. */ | |
452 | static int __init do_early_param(char *param, char *val) | |
453 | { | |
454 | struct obs_kernel_param *p; | |
455 | ||
456 | for (p = __setup_start; p < __setup_end; p++) { | |
18a8bd94 YL |
457 | if ((p->early && strcmp(param, p->str) == 0) || |
458 | (strcmp(param, "console") == 0 && | |
459 | strcmp(p->str, "earlycon") == 0) | |
460 | ) { | |
1da177e4 LT |
461 | if (p->setup_func(val) != 0) |
462 | printk(KERN_WARNING | |
463 | "Malformed early option '%s'\n", param); | |
464 | } | |
465 | } | |
466 | /* We accept everything at this stage. */ | |
467 | return 0; | |
468 | } | |
469 | ||
13977091 MD |
470 | void __init parse_early_options(char *cmdline) |
471 | { | |
472 | parse_args("early options", cmdline, NULL, 0, do_early_param); | |
473 | } | |
474 | ||
1da177e4 LT |
475 | /* Arch code calls this early on, or if not, just before other parsing. */ |
476 | void __init parse_early_param(void) | |
477 | { | |
478 | static __initdata int done = 0; | |
479 | static __initdata char tmp_cmdline[COMMAND_LINE_SIZE]; | |
480 | ||
481 | if (done) | |
482 | return; | |
483 | ||
484 | /* All fall through to do_early_param. */ | |
30d7e0d4 | 485 | strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); |
13977091 | 486 | parse_early_options(tmp_cmdline); |
1da177e4 LT |
487 | done = 1; |
488 | } | |
489 | ||
490 | /* | |
491 | * Activate the first processor. | |
492 | */ | |
493 | ||
44fd2299 SS |
494 | static void __init boot_cpu_init(void) |
495 | { | |
496 | int cpu = smp_processor_id(); | |
497 | /* Mark the boot cpu "present", "online" etc for SMP and UP case */ | |
915441b6 RR |
498 | set_cpu_online(cpu, true); |
499 | set_cpu_present(cpu, true); | |
500 | set_cpu_possible(cpu, true); | |
44fd2299 SS |
501 | } |
502 | ||
839ad62e | 503 | void __init __weak smp_setup_processor_id(void) |
033ab7f8 AM |
504 | { |
505 | } | |
506 | ||
8c9843e5 BH |
507 | void __init __weak thread_info_cache_init(void) |
508 | { | |
509 | } | |
510 | ||
444f478f PE |
511 | /* |
512 | * Set up kernel memory allocators | |
513 | */ | |
514 | static void __init mm_init(void) | |
515 | { | |
ca371c0d KH |
516 | /* |
517 | * page_cgroup requires countinous pages as memmap | |
518 | * and it's bigger than MAX_ORDER unless SPARSEMEM. | |
519 | */ | |
520 | page_cgroup_init_flatmem(); | |
444f478f PE |
521 | mem_init(); |
522 | kmem_cache_init(); | |
c868d550 | 523 | pgtable_cache_init(); |
444f478f PE |
524 | vmalloc_init(); |
525 | } | |
526 | ||
1da177e4 LT |
527 | asmlinkage void __init start_kernel(void) |
528 | { | |
529 | char * command_line; | |
530 | extern struct kernel_param __start___param[], __stop___param[]; | |
033ab7f8 AM |
531 | |
532 | smp_setup_processor_id(); | |
533 | ||
fbb9ce95 IM |
534 | /* |
535 | * Need to run as early as possible, to initialize the | |
536 | * lockdep hash: | |
537 | */ | |
538 | lockdep_init(); | |
3ac7fe5a | 539 | debug_objects_early_init(); |
42059429 IM |
540 | |
541 | /* | |
542 | * Set up the the initial canary ASAP: | |
543 | */ | |
544 | boot_init_stack_canary(); | |
545 | ||
ddbcc7e8 | 546 | cgroup_init_early(); |
fbb9ce95 IM |
547 | |
548 | local_irq_disable(); | |
549 | early_boot_irqs_off(); | |
243c7621 | 550 | early_init_irq_lock_class(); |
fbb9ce95 | 551 | |
1da177e4 LT |
552 | /* |
553 | * Interrupts are still disabled. Do necessary setups, then | |
554 | * enable them | |
555 | */ | |
556 | lock_kernel(); | |
906568c9 | 557 | tick_init(); |
44fd2299 | 558 | boot_cpu_init(); |
1da177e4 | 559 | page_address_init(); |
657cafa6 | 560 | printk(KERN_NOTICE "%s", linux_banner); |
1da177e4 | 561 | setup_arch(&command_line); |
cf475ad2 | 562 | mm_init_owner(&init_mm, &init_task); |
30d7e0d4 | 563 | setup_command_line(command_line); |
e0982e90 | 564 | setup_nr_cpu_ids(); |
d6647bdf | 565 | setup_per_cpu_areas(); |
44fd2299 | 566 | smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ |
1da177e4 | 567 | |
83b519e8 PE |
568 | build_all_zonelists(); |
569 | page_alloc_init(); | |
570 | ||
571 | printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line); | |
572 | parse_early_param(); | |
573 | parse_args("Booting kernel", static_command_line, __start___param, | |
574 | __stop___param - __start___param, | |
575 | &unknown_bootoption); | |
576 | /* | |
577 | * These use large bootmem allocations and must precede | |
578 | * kmem_cache_init() | |
579 | */ | |
580 | pidhash_init(); | |
83b519e8 PE |
581 | vfs_caches_init_early(); |
582 | sort_main_extable(); | |
583 | trap_init(); | |
444f478f | 584 | mm_init(); |
1da177e4 LT |
585 | /* |
586 | * Set up the scheduler prior starting any interrupts (such as the | |
587 | * timer interrupt). Full topology setup happens at smp_init() | |
588 | * time - but meanwhile we still have a functioning scheduler. | |
589 | */ | |
590 | sched_init(); | |
591 | /* | |
592 | * Disable preemption - early bootup scheduling is extremely | |
593 | * fragile until we cpu_idle() for the first time. | |
594 | */ | |
595 | preempt_disable(); | |
c4a68306 AB |
596 | if (!irqs_disabled()) { |
597 | printk(KERN_WARNING "start_kernel(): bug: interrupts were " | |
598 | "enabled *very* early, fixing it\n"); | |
599 | local_irq_disable(); | |
600 | } | |
1da177e4 | 601 | rcu_init(); |
0b8f1efa YL |
602 | /* init some links before init_ISA_irqs() */ |
603 | early_irq_init(); | |
1da177e4 | 604 | init_IRQ(); |
3c7b4e6b | 605 | prio_tree_init(); |
1da177e4 | 606 | init_timers(); |
c0a31329 | 607 | hrtimers_init(); |
1da177e4 | 608 | softirq_init(); |
ad596171 | 609 | timekeeping_init(); |
88fecaa2 | 610 | time_init(); |
93e02814 HC |
611 | profile_init(); |
612 | if (!irqs_disabled()) | |
24d431d0 RL |
613 | printk(KERN_CRIT "start_kernel(): bug: interrupts were " |
614 | "enabled early\n"); | |
fbb9ce95 | 615 | early_boot_irqs_on(); |
93e02814 | 616 | local_irq_enable(); |
dcce284a BH |
617 | |
618 | /* Interrupts are enabled now so all GFP allocations are safe. */ | |
619 | set_gfp_allowed_mask(__GFP_BITS_MASK); | |
620 | ||
7e85ee0c | 621 | kmem_cache_init_late(); |
1da177e4 LT |
622 | |
623 | /* | |
624 | * HACK ALERT! This is early. We're enabling the console before | |
625 | * we've done PCI setups etc, and console_init() must be aware of | |
626 | * this. But we do want output early, in case something goes wrong. | |
627 | */ | |
628 | console_init(); | |
629 | if (panic_later) | |
630 | panic(panic_later, panic_param); | |
fbb9ce95 IM |
631 | |
632 | lockdep_info(); | |
633 | ||
9a11b49a IM |
634 | /* |
635 | * Need to run this when irqs are enabled, because it wants | |
636 | * to self-test [hard/soft]-irqs on/off lock inversion bugs | |
637 | * too: | |
638 | */ | |
639 | locking_selftest(); | |
640 | ||
1da177e4 LT |
641 | #ifdef CONFIG_BLK_DEV_INITRD |
642 | if (initrd_start && !initrd_below_start_ok && | |
bd673c7c | 643 | page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { |
1da177e4 | 644 | printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - " |
fb6624eb | 645 | "disabling it.\n", |
bd673c7c GU |
646 | page_to_pfn(virt_to_page((void *)initrd_start)), |
647 | min_low_pfn); | |
1da177e4 LT |
648 | initrd_start = 0; |
649 | } | |
650 | #endif | |
94b6da5a | 651 | page_cgroup_init(); |
a03c2a48 | 652 | enable_debug_pagealloc(); |
b9ce08c0 | 653 | kmemtrace_init(); |
3c7b4e6b | 654 | kmemleak_init(); |
3ac7fe5a | 655 | debug_objects_mem_init(); |
199f0ca5 | 656 | idr_init_cache(); |
e7c8d5c9 | 657 | setup_per_cpu_pageset(); |
1da177e4 LT |
658 | numa_policy_init(); |
659 | if (late_time_init) | |
660 | late_time_init(); | |
fa84e9ee | 661 | sched_clock_init(); |
1da177e4 LT |
662 | calibrate_delay(); |
663 | pidmap_init(); | |
1da177e4 LT |
664 | anon_vma_init(); |
665 | #ifdef CONFIG_X86 | |
666 | if (efi_enabled) | |
667 | efi_enter_virtual_mode(); | |
668 | #endif | |
8c9843e5 | 669 | thread_info_cache_init(); |
d84f4f99 | 670 | cred_init(); |
1da177e4 LT |
671 | fork_init(num_physpages); |
672 | proc_caches_init(); | |
673 | buffer_init(); | |
1da177e4 LT |
674 | key_init(); |
675 | security_init(); | |
676 | vfs_caches_init(num_physpages); | |
677 | radix_tree_init(); | |
678 | signals_init(); | |
679 | /* rootfs populating might need page-writeback */ | |
680 | page_writeback_init(); | |
681 | #ifdef CONFIG_PROC_FS | |
682 | proc_root_init(); | |
683 | #endif | |
ddbcc7e8 | 684 | cgroup_init(); |
1da177e4 | 685 | cpuset_init(); |
c757249a | 686 | taskstats_init_early(); |
ca74e92b | 687 | delayacct_init(); |
1da177e4 LT |
688 | |
689 | check_bugs(); | |
690 | ||
691 | acpi_early_init(); /* before LAPIC and SMP init */ | |
692 | ||
68bf21aa SR |
693 | ftrace_init(); |
694 | ||
1da177e4 LT |
695 | /* Do the rest non-__init'ed, we're now alive */ |
696 | rest_init(); | |
697 | } | |
698 | ||
b99b87f7 PO |
699 | /* Call all constructor functions linked into the kernel. */ |
700 | static void __init do_ctors(void) | |
701 | { | |
702 | #ifdef CONFIG_CONSTRUCTORS | |
703 | ctor_fn_t *call = (ctor_fn_t *) __ctors_start; | |
704 | ||
705 | for (; call < (ctor_fn_t *) __ctors_end; call++) | |
706 | (*call)(); | |
707 | #endif | |
708 | } | |
709 | ||
22a9d645 | 710 | int initcall_debug; |
d0ea3d7d | 711 | core_param(initcall_debug, initcall_debug, bool, 0644); |
1da177e4 | 712 | |
4a683bf9 IM |
713 | static char msgbuf[64]; |
714 | static struct boot_trace_call call; | |
715 | static struct boot_trace_ret ret; | |
716 | ||
59f9415f | 717 | int do_one_initcall(initcall_t fn) |
1da177e4 | 718 | { |
1da177e4 | 719 | int count = preempt_count(); |
74239072 | 720 | ktime_t calltime, delta, rettime; |
1da177e4 | 721 | |
e0df154f | 722 | if (initcall_debug) { |
74239072 FW |
723 | call.caller = task_pid_nr(current); |
724 | printk("calling %pF @ %i\n", fn, call.caller); | |
725 | calltime = ktime_get(); | |
726 | trace_boot_call(&call, fn); | |
71566a0d | 727 | enable_boot_trace(); |
e0df154f | 728 | } |
1da177e4 | 729 | |
74239072 | 730 | ret.result = fn(); |
1da177e4 | 731 | |
e0df154f | 732 | if (initcall_debug) { |
71566a0d | 733 | disable_boot_trace(); |
74239072 FW |
734 | rettime = ktime_get(); |
735 | delta = ktime_sub(rettime, calltime); | |
1d926f27 | 736 | ret.duration = (unsigned long long) ktime_to_ns(delta) >> 10; |
74239072 | 737 | trace_boot_ret(&ret, fn); |
ca538f6b | 738 | printk("initcall %pF returned %d after %Ld usecs\n", fn, |
74239072 | 739 | ret.result, ret.duration); |
e0df154f | 740 | } |
8f0c45cd | 741 | |
e0df154f | 742 | msgbuf[0] = 0; |
e662e1cf | 743 | |
74239072 FW |
744 | if (ret.result && ret.result != -ENODEV && initcall_debug) |
745 | sprintf(msgbuf, "error code %d ", ret.result); | |
e662e1cf | 746 | |
e0df154f | 747 | if (preempt_count() != count) { |
a76bfd0d | 748 | strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf)); |
e0df154f | 749 | preempt_count() = count; |
1da177e4 | 750 | } |
e0df154f | 751 | if (irqs_disabled()) { |
a76bfd0d | 752 | strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); |
e0df154f LT |
753 | local_irq_enable(); |
754 | } | |
755 | if (msgbuf[0]) { | |
96d746c6 | 756 | printk("initcall %pF returned with %s\n", fn, msgbuf); |
e0df154f | 757 | } |
59f9415f | 758 | |
74239072 | 759 | return ret.result; |
e0df154f LT |
760 | } |
761 | ||
762 | ||
c2147a50 | 763 | extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[]; |
e0df154f LT |
764 | |
765 | static void __init do_initcalls(void) | |
766 | { | |
767 | initcall_t *call; | |
768 | ||
c2147a50 | 769 | for (call = __early_initcall_end; call < __initcall_end; call++) |
e0df154f | 770 | do_one_initcall(*call); |
1da177e4 LT |
771 | |
772 | /* Make sure there is no pending stuff from the initcall sequence */ | |
773 | flush_scheduled_work(); | |
774 | } | |
775 | ||
776 | /* | |
777 | * Ok, the machine is now initialized. None of the devices | |
778 | * have been touched yet, but the CPU subsystem is up and | |
779 | * running, and memory and process management works. | |
780 | * | |
781 | * Now we can finally start doing some real work.. | |
782 | */ | |
783 | static void __init do_basic_setup(void) | |
784 | { | |
4446a36f | 785 | rcu_init_sched(); /* needed by module_init stage. */ |
4403b406 | 786 | init_workqueues(); |
759ee091 | 787 | cpuset_init_smp(); |
1da177e4 | 788 | usermodehelper_init(); |
2b2af54a | 789 | init_tmpfs(); |
1da177e4 | 790 | driver_init(); |
b04c3afb | 791 | init_irq_proc(); |
b99b87f7 | 792 | do_ctors(); |
1da177e4 LT |
793 | do_initcalls(); |
794 | } | |
795 | ||
7babe8db | 796 | static void __init do_pre_smp_initcalls(void) |
c2147a50 EGM |
797 | { |
798 | initcall_t *call; | |
799 | ||
800 | for (call = __initcall_start; call < __early_initcall_end; call++) | |
801 | do_one_initcall(*call); | |
802 | } | |
803 | ||
1da177e4 LT |
804 | static void run_init_process(char *init_filename) |
805 | { | |
806 | argv_init[0] = init_filename; | |
67608567 | 807 | kernel_execve(init_filename, argv_init, envp_init); |
1da177e4 LT |
808 | } |
809 | ||
ee5bfa64 VG |
810 | /* This is a non __init function. Force it to be noinline otherwise gcc |
811 | * makes it inline to init() and it becomes part of init.text section | |
812 | */ | |
f99ebf0a | 813 | static noinline int init_post(void) |
acdd052a | 814 | __releases(kernel_lock) |
ee5bfa64 | 815 | { |
22a9d645 AV |
816 | /* need to finish all async __init code before freeing the memory */ |
817 | async_synchronize_full(); | |
ee5bfa64 VG |
818 | free_initmem(); |
819 | unlock_kernel(); | |
820 | mark_rodata_ro(); | |
821 | system_state = SYSTEM_RUNNING; | |
822 | numa_default_policy(); | |
823 | ||
824 | if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) | |
825 | printk(KERN_WARNING "Warning: unable to open an initial console.\n"); | |
826 | ||
827 | (void) sys_dup(0); | |
828 | (void) sys_dup(0); | |
829 | ||
fae5fa44 ON |
830 | current->signal->flags |= SIGNAL_UNKILLABLE; |
831 | ||
ee5bfa64 VG |
832 | if (ramdisk_execute_command) { |
833 | run_init_process(ramdisk_execute_command); | |
834 | printk(KERN_WARNING "Failed to execute %s\n", | |
835 | ramdisk_execute_command); | |
836 | } | |
837 | ||
838 | /* | |
839 | * We try each of these until one succeeds. | |
840 | * | |
841 | * The Bourne shell can be used instead of init if we are | |
842 | * trying to recover a really broken machine. | |
843 | */ | |
844 | if (execute_command) { | |
845 | run_init_process(execute_command); | |
846 | printk(KERN_WARNING "Failed to execute %s. Attempting " | |
847 | "defaults...\n", execute_command); | |
848 | } | |
849 | run_init_process("/sbin/init"); | |
850 | run_init_process("/etc/init"); | |
851 | run_init_process("/bin/init"); | |
852 | run_init_process("/bin/sh"); | |
853 | ||
854 | panic("No init found. Try passing init= option to kernel."); | |
855 | } | |
856 | ||
aae5f662 | 857 | static int __init kernel_init(void * unused) |
1da177e4 LT |
858 | { |
859 | lock_kernel(); | |
58568d2a MX |
860 | |
861 | /* | |
862 | * init can allocate pages on any node | |
863 | */ | |
864 | set_mems_allowed(node_possible_map); | |
1da177e4 LT |
865 | /* |
866 | * init can run on any cpu. | |
867 | */ | |
1a2142af | 868 | set_cpus_allowed_ptr(current, cpu_all_mask); |
1da177e4 LT |
869 | /* |
870 | * Tell the world that we're going to be the grim | |
871 | * reaper of innocent orphaned children. | |
872 | * | |
873 | * We don't want people to have to make incorrect | |
874 | * assumptions about where in the task array this | |
875 | * can be found. | |
876 | */ | |
84d73786 | 877 | init_pid_ns.child_reaper = current; |
1da177e4 | 878 | |
9ec52099 CLG |
879 | cad_pid = task_pid(current); |
880 | ||
ca74a6f8 | 881 | smp_prepare_cpus(setup_max_cpus); |
1da177e4 LT |
882 | |
883 | do_pre_smp_initcalls(); | |
3bf77af6 | 884 | start_boot_trace(); |
1da177e4 | 885 | |
1da177e4 LT |
886 | smp_init(); |
887 | sched_init_smp(); | |
888 | ||
1da177e4 LT |
889 | do_basic_setup(); |
890 | ||
891 | /* | |
892 | * check if there is an early userspace init. If yes, let it do all | |
893 | * the work | |
894 | */ | |
ffdfc409 OJ |
895 | |
896 | if (!ramdisk_execute_command) | |
897 | ramdisk_execute_command = "/init"; | |
898 | ||
899 | if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) { | |
900 | ramdisk_execute_command = NULL; | |
1da177e4 | 901 | prepare_namespace(); |
ffdfc409 | 902 | } |
1da177e4 LT |
903 | |
904 | /* | |
905 | * Ok, we have completed the initial bootup, and | |
906 | * we're essentially up and running. Get rid of the | |
907 | * initmem segments and start the user-mode stuff.. | |
908 | */ | |
71566a0d | 909 | |
ee5bfa64 VG |
910 | init_post(); |
911 | return 0; | |
1da177e4 | 912 | } |