]>
Commit | Line | Data |
---|---|---|
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 | ||
12 | #include <linux/types.h> | |
13 | #include <linux/module.h> | |
14 | #include <linux/proc_fs.h> | |
15 | #include <linux/kernel.h> | |
16 | #include <linux/syscalls.h> | |
17 | #include <linux/stackprotector.h> | |
18 | #include <linux/string.h> | |
19 | #include <linux/ctype.h> | |
20 | #include <linux/delay.h> | |
21 | #include <linux/ioport.h> | |
22 | #include <linux/init.h> | |
23 | #include <linux/smp_lock.h> | |
24 | #include <linux/initrd.h> | |
25 | #include <linux/bootmem.h> | |
26 | #include <linux/acpi.h> | |
27 | #include <linux/tty.h> | |
28 | #include <linux/percpu.h> | |
29 | #include <linux/kmod.h> | |
30 | #include <linux/vmalloc.h> | |
31 | #include <linux/kernel_stat.h> | |
32 | #include <linux/start_kernel.h> | |
33 | #include <linux/security.h> | |
34 | #include <linux/smp.h> | |
35 | #include <linux/workqueue.h> | |
36 | #include <linux/profile.h> | |
37 | #include <linux/rcupdate.h> | |
38 | #include <linux/moduleparam.h> | |
39 | #include <linux/kallsyms.h> | |
40 | #include <linux/writeback.h> | |
41 | #include <linux/cpu.h> | |
42 | #include <linux/cpuset.h> | |
43 | #include <linux/cgroup.h> | |
44 | #include <linux/efi.h> | |
45 | #include <linux/tick.h> | |
46 | #include <linux/interrupt.h> | |
47 | #include <linux/taskstats_kern.h> | |
48 | #include <linux/delayacct.h> | |
49 | #include <linux/unistd.h> | |
50 | #include <linux/rmap.h> | |
51 | #include <linux/mempolicy.h> | |
52 | #include <linux/key.h> | |
53 | #include <linux/buffer_head.h> | |
54 | #include <linux/page_cgroup.h> | |
55 | #include <linux/debug_locks.h> | |
56 | #include <linux/debugobjects.h> | |
57 | #include <linux/lockdep.h> | |
58 | #include <linux/kmemleak.h> | |
59 | #include <linux/pid_namespace.h> | |
60 | #include <linux/device.h> | |
61 | #include <linux/kthread.h> | |
62 | #include <linux/sched.h> | |
63 | #include <linux/signal.h> | |
64 | #include <linux/idr.h> | |
65 | #include <linux/kgdb.h> | |
66 | #include <linux/ftrace.h> | |
67 | #include <linux/async.h> | |
68 | #include <linux/kmemcheck.h> | |
69 | #include <linux/kmemtrace.h> | |
70 | #include <linux/sfi.h> | |
71 | #include <linux/shmem_fs.h> | |
72 | #include <linux/slab.h> | |
73 | #include <trace/boot.h> | |
74 | ||
75 | #include <asm/io.h> | |
76 | #include <asm/bugs.h> | |
77 | #include <asm/setup.h> | |
78 | #include <asm/sections.h> | |
79 | #include <asm/cacheflush.h> | |
80 | ||
81 | #ifdef CONFIG_X86_LOCAL_APIC | |
82 | #include <asm/smp.h> | |
83 | #endif | |
84 | ||
85 | static int kernel_init(void *); | |
86 | ||
87 | extern void init_IRQ(void); | |
88 | extern void fork_init(unsigned long); | |
89 | extern void mca_init(void); | |
90 | extern void sbus_init(void); | |
91 | extern void prio_tree_init(void); | |
92 | extern void radix_tree_init(void); | |
93 | extern void free_initmem(void); | |
94 | #ifndef CONFIG_DEBUG_RODATA | |
95 | static inline void mark_rodata_ro(void) { } | |
96 | #endif | |
97 | ||
98 | #ifdef CONFIG_TC | |
99 | extern void tc_init(void); | |
100 | #endif | |
101 | ||
102 | enum system_states system_state __read_mostly; | |
103 | EXPORT_SYMBOL(system_state); | |
104 | ||
105 | /* | |
106 | * Boot command-line arguments | |
107 | */ | |
108 | #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT | |
109 | #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT | |
110 | ||
111 | extern void time_init(void); | |
112 | /* Default late time init is NULL. archs can override this later. */ | |
113 | void (*__initdata late_time_init)(void); | |
114 | extern void softirq_init(void); | |
115 | ||
116 | /* Untouched command line saved by arch-specific code. */ | |
117 | char __initdata boot_command_line[COMMAND_LINE_SIZE]; | |
118 | /* Untouched saved command line (eg. for /proc) */ | |
119 | char *saved_command_line; | |
120 | /* Command line for parameter parsing */ | |
121 | static char *static_command_line; | |
122 | ||
123 | static char *execute_command; | |
124 | static char *ramdisk_execute_command; | |
125 | ||
126 | #ifdef CONFIG_SMP | |
127 | /* Setup configured maximum number of CPUs to activate */ | |
128 | unsigned int __initdata setup_max_cpus = NR_CPUS; | |
129 | ||
130 | /* | |
131 | * Setup routine for controlling SMP activation | |
132 | * | |
133 | * Command-line option of "nosmp" or "maxcpus=0" will disable SMP | |
134 | * activation entirely (the MPS table probe still happens, though). | |
135 | * | |
136 | * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer | |
137 | * greater than 0, limits the maximum number of CPUs activated in | |
138 | * SMP mode to <NUM>. | |
139 | */ | |
140 | ||
141 | void __weak arch_disable_smp_support(void) { } | |
142 | ||
143 | static int __init nosmp(char *str) | |
144 | { | |
145 | setup_max_cpus = 0; | |
146 | arch_disable_smp_support(); | |
147 | ||
148 | return 0; | |
149 | } | |
150 | ||
151 | early_param("nosmp", nosmp); | |
152 | ||
153 | /* this is hard limit */ | |
154 | static int __init nrcpus(char *str) | |
155 | { | |
156 | int nr_cpus; | |
157 | ||
158 | get_option(&str, &nr_cpus); | |
159 | if (nr_cpus > 0 && nr_cpus < nr_cpu_ids) | |
160 | nr_cpu_ids = nr_cpus; | |
161 | ||
162 | return 0; | |
163 | } | |
164 | ||
165 | early_param("nr_cpus", nrcpus); | |
166 | ||
167 | static int __init maxcpus(char *str) | |
168 | { | |
169 | get_option(&str, &setup_max_cpus); | |
170 | if (setup_max_cpus == 0) | |
171 | arch_disable_smp_support(); | |
172 | ||
173 | return 0; | |
174 | } | |
175 | ||
176 | early_param("maxcpus", maxcpus); | |
177 | #else | |
178 | static const unsigned int setup_max_cpus = NR_CPUS; | |
179 | #endif | |
180 | ||
181 | /* | |
182 | * If set, this is an indication to the drivers that reset the underlying | |
183 | * device before going ahead with the initialization otherwise driver might | |
184 | * rely on the BIOS and skip the reset operation. | |
185 | * | |
186 | * This is useful if kernel is booting in an unreliable environment. | |
187 | * For ex. kdump situaiton where previous kernel has crashed, BIOS has been | |
188 | * skipped and devices will be in unknown state. | |
189 | */ | |
190 | unsigned int reset_devices; | |
191 | EXPORT_SYMBOL(reset_devices); | |
192 | ||
193 | static int __init set_reset_devices(char *str) | |
194 | { | |
195 | reset_devices = 1; | |
196 | return 1; | |
197 | } | |
198 | ||
199 | __setup("reset_devices", set_reset_devices); | |
200 | ||
201 | static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; | |
202 | char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; | |
203 | static const char *panic_later, *panic_param; | |
204 | ||
205 | extern struct obs_kernel_param __setup_start[], __setup_end[]; | |
206 | ||
207 | static int __init obsolete_checksetup(char *line) | |
208 | { | |
209 | struct obs_kernel_param *p; | |
210 | int had_early_param = 0; | |
211 | ||
212 | p = __setup_start; | |
213 | do { | |
214 | int n = strlen(p->str); | |
215 | if (!strncmp(line, p->str, n)) { | |
216 | if (p->early) { | |
217 | /* Already done in parse_early_param? | |
218 | * (Needs exact match on param part). | |
219 | * Keep iterating, as we can have early | |
220 | * params and __setups of same names 8( */ | |
221 | if (line[n] == '\0' || line[n] == '=') | |
222 | had_early_param = 1; | |
223 | } else if (!p->setup_func) { | |
224 | printk(KERN_WARNING "Parameter %s is obsolete," | |
225 | " ignored\n", p->str); | |
226 | return 1; | |
227 | } else if (p->setup_func(line + n)) | |
228 | return 1; | |
229 | } | |
230 | p++; | |
231 | } while (p < __setup_end); | |
232 | ||
233 | return had_early_param; | |
234 | } | |
235 | ||
236 | /* | |
237 | * This should be approx 2 Bo*oMips to start (note initial shift), and will | |
238 | * still work even if initially too large, it will just take slightly longer | |
239 | */ | |
240 | unsigned long loops_per_jiffy = (1<<12); | |
241 | ||
242 | EXPORT_SYMBOL(loops_per_jiffy); | |
243 | ||
244 | static int __init debug_kernel(char *str) | |
245 | { | |
246 | console_loglevel = 10; | |
247 | return 0; | |
248 | } | |
249 | ||
250 | static int __init quiet_kernel(char *str) | |
251 | { | |
252 | console_loglevel = 4; | |
253 | return 0; | |
254 | } | |
255 | ||
256 | early_param("debug", debug_kernel); | |
257 | early_param("quiet", quiet_kernel); | |
258 | ||
259 | static int __init loglevel(char *str) | |
260 | { | |
261 | get_option(&str, &console_loglevel); | |
262 | return 0; | |
263 | } | |
264 | ||
265 | early_param("loglevel", loglevel); | |
266 | ||
267 | /* | |
268 | * Unknown boot options get handed to init, unless they look like | |
269 | * unused parameters (modprobe will find them in /proc/cmdline). | |
270 | */ | |
271 | static int __init unknown_bootoption(char *param, char *val) | |
272 | { | |
273 | /* Change NUL term back to "=", to make "param" the whole string. */ | |
274 | if (val) { | |
275 | /* param=val or param="val"? */ | |
276 | if (val == param+strlen(param)+1) | |
277 | val[-1] = '='; | |
278 | else if (val == param+strlen(param)+2) { | |
279 | val[-2] = '='; | |
280 | memmove(val-1, val, strlen(val)+1); | |
281 | val--; | |
282 | } else | |
283 | BUG(); | |
284 | } | |
285 | ||
286 | /* Handle obsolete-style parameters */ | |
287 | if (obsolete_checksetup(param)) | |
288 | return 0; | |
289 | ||
290 | /* Unused module parameter. */ | |
291 | if (strchr(param, '.') && (!val || strchr(param, '.') < val)) | |
292 | return 0; | |
293 | ||
294 | if (panic_later) | |
295 | return 0; | |
296 | ||
297 | if (val) { | |
298 | /* Environment option */ | |
299 | unsigned int i; | |
300 | for (i = 0; envp_init[i]; i++) { | |
301 | if (i == MAX_INIT_ENVS) { | |
302 | panic_later = "Too many boot env vars at `%s'"; | |
303 | panic_param = param; | |
304 | } | |
305 | if (!strncmp(param, envp_init[i], val - param)) | |
306 | break; | |
307 | } | |
308 | envp_init[i] = param; | |
309 | } else { | |
310 | /* Command line option */ | |
311 | unsigned int i; | |
312 | for (i = 0; argv_init[i]; i++) { | |
313 | if (i == MAX_INIT_ARGS) { | |
314 | panic_later = "Too many boot init vars at `%s'"; | |
315 | panic_param = param; | |
316 | } | |
317 | } | |
318 | argv_init[i] = param; | |
319 | } | |
320 | return 0; | |
321 | } | |
322 | ||
323 | #ifdef CONFIG_DEBUG_PAGEALLOC | |
324 | int __read_mostly debug_pagealloc_enabled = 0; | |
325 | #endif | |
326 | ||
327 | static int __init init_setup(char *str) | |
328 | { | |
329 | unsigned int i; | |
330 | ||
331 | execute_command = str; | |
332 | /* | |
333 | * In case LILO is going to boot us with default command line, | |
334 | * it prepends "auto" before the whole cmdline which makes | |
335 | * the shell think it should execute a script with such name. | |
336 | * So we ignore all arguments entered _before_ init=... [MJ] | |
337 | */ | |
338 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
339 | argv_init[i] = NULL; | |
340 | return 1; | |
341 | } | |
342 | __setup("init=", init_setup); | |
343 | ||
344 | static int __init rdinit_setup(char *str) | |
345 | { | |
346 | unsigned int i; | |
347 | ||
348 | ramdisk_execute_command = str; | |
349 | /* See "auto" comment in init_setup */ | |
350 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
351 | argv_init[i] = NULL; | |
352 | return 1; | |
353 | } | |
354 | __setup("rdinit=", rdinit_setup); | |
355 | ||
356 | #ifndef CONFIG_SMP | |
357 | ||
358 | #ifdef CONFIG_X86_LOCAL_APIC | |
359 | static void __init smp_init(void) | |
360 | { | |
361 | APIC_init_uniprocessor(); | |
362 | } | |
363 | #else | |
364 | #define smp_init() do { } while (0) | |
365 | #endif | |
366 | ||
367 | static inline void setup_nr_cpu_ids(void) { } | |
368 | static inline void smp_prepare_cpus(unsigned int maxcpus) { } | |
369 | ||
370 | #else | |
371 | ||
372 | /* Setup number of possible processor ids */ | |
373 | int nr_cpu_ids __read_mostly = NR_CPUS; | |
374 | EXPORT_SYMBOL(nr_cpu_ids); | |
375 | ||
376 | /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */ | |
377 | static void __init setup_nr_cpu_ids(void) | |
378 | { | |
379 | nr_cpu_ids = find_last_bit(cpumask_bits(cpu_possible_mask),NR_CPUS) + 1; | |
380 | } | |
381 | ||
382 | /* Called by boot processor to activate the rest. */ | |
383 | static void __init smp_init(void) | |
384 | { | |
385 | unsigned int cpu; | |
386 | ||
387 | /* FIXME: This should be done in userspace --RR */ | |
388 | for_each_present_cpu(cpu) { | |
389 | if (num_online_cpus() >= setup_max_cpus) | |
390 | break; | |
391 | if (!cpu_online(cpu)) | |
392 | cpu_up(cpu); | |
393 | } | |
394 | ||
395 | /* Any cleanup work */ | |
396 | printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus()); | |
397 | smp_cpus_done(setup_max_cpus); | |
398 | } | |
399 | ||
400 | #endif | |
401 | ||
402 | /* | |
403 | * We need to store the untouched command line for future reference. | |
404 | * We also need to store the touched command line since the parameter | |
405 | * parsing is performed in place, and we should allow a component to | |
406 | * store reference of name/value for future reference. | |
407 | */ | |
408 | static void __init setup_command_line(char *command_line) | |
409 | { | |
410 | saved_command_line = alloc_bootmem(strlen (boot_command_line)+1); | |
411 | static_command_line = alloc_bootmem(strlen (command_line)+1); | |
412 | strcpy (saved_command_line, boot_command_line); | |
413 | strcpy (static_command_line, command_line); | |
414 | } | |
415 | ||
416 | /* | |
417 | * We need to finalize in a non-__init function or else race conditions | |
418 | * between the root thread and the init thread may cause start_kernel to | |
419 | * be reaped by free_initmem before the root thread has proceeded to | |
420 | * cpu_idle. | |
421 | * | |
422 | * gcc-3.4 accidentally inlines this function, so use noinline. | |
423 | */ | |
424 | ||
425 | static noinline void __init_refok rest_init(void) | |
426 | __releases(kernel_lock) | |
427 | { | |
428 | int pid; | |
429 | ||
430 | rcu_scheduler_starting(); | |
431 | kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND); | |
432 | numa_default_policy(); | |
433 | pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); | |
434 | rcu_read_lock(); | |
435 | kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); | |
436 | rcu_read_unlock(); | |
437 | unlock_kernel(); | |
438 | ||
439 | /* | |
440 | * The boot idle thread must execute schedule() | |
441 | * at least once to get things moving: | |
442 | */ | |
443 | init_idle_bootup_task(current); | |
444 | preempt_enable_no_resched(); | |
445 | schedule(); | |
446 | preempt_disable(); | |
447 | ||
448 | /* Call into cpu_idle with preempt disabled */ | |
449 | cpu_idle(); | |
450 | } | |
451 | ||
452 | /* Check for early params. */ | |
453 | static int __init do_early_param(char *param, char *val) | |
454 | { | |
455 | struct obs_kernel_param *p; | |
456 | ||
457 | for (p = __setup_start; p < __setup_end; p++) { | |
458 | if ((p->early && strcmp(param, p->str) == 0) || | |
459 | (strcmp(param, "console") == 0 && | |
460 | strcmp(p->str, "earlycon") == 0) | |
461 | ) { | |
462 | if (p->setup_func(val) != 0) | |
463 | printk(KERN_WARNING | |
464 | "Malformed early option '%s'\n", param); | |
465 | } | |
466 | } | |
467 | /* We accept everything at this stage. */ | |
468 | return 0; | |
469 | } | |
470 | ||
471 | void __init parse_early_options(char *cmdline) | |
472 | { | |
473 | parse_args("early options", cmdline, NULL, 0, do_early_param); | |
474 | } | |
475 | ||
476 | /* Arch code calls this early on, or if not, just before other parsing. */ | |
477 | void __init parse_early_param(void) | |
478 | { | |
479 | static __initdata int done = 0; | |
480 | static __initdata char tmp_cmdline[COMMAND_LINE_SIZE]; | |
481 | ||
482 | if (done) | |
483 | return; | |
484 | ||
485 | /* All fall through to do_early_param. */ | |
486 | strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); | |
487 | parse_early_options(tmp_cmdline); | |
488 | done = 1; | |
489 | } | |
490 | ||
491 | /* | |
492 | * Activate the first processor. | |
493 | */ | |
494 | ||
495 | static void __init boot_cpu_init(void) | |
496 | { | |
497 | int cpu = smp_processor_id(); | |
498 | /* Mark the boot cpu "present", "online" etc for SMP and UP case */ | |
499 | set_cpu_online(cpu, true); | |
500 | set_cpu_active(cpu, true); | |
501 | set_cpu_present(cpu, true); | |
502 | set_cpu_possible(cpu, true); | |
503 | } | |
504 | ||
505 | void __init __weak smp_setup_processor_id(void) | |
506 | { | |
507 | } | |
508 | ||
509 | void __init __weak thread_info_cache_init(void) | |
510 | { | |
511 | } | |
512 | ||
513 | /* | |
514 | * Set up kernel memory allocators | |
515 | */ | |
516 | static void __init mm_init(void) | |
517 | { | |
518 | /* | |
519 | * page_cgroup requires countinous pages as memmap | |
520 | * and it's bigger than MAX_ORDER unless SPARSEMEM. | |
521 | */ | |
522 | page_cgroup_init_flatmem(); | |
523 | mem_init(); | |
524 | kmem_cache_init(); | |
525 | pgtable_cache_init(); | |
526 | vmalloc_init(); | |
527 | } | |
528 | ||
529 | asmlinkage void __init start_kernel(void) | |
530 | { | |
531 | char * command_line; | |
532 | extern struct kernel_param __start___param[], __stop___param[]; | |
533 | ||
534 | smp_setup_processor_id(); | |
535 | ||
536 | /* | |
537 | * Need to run as early as possible, to initialize the | |
538 | * lockdep hash: | |
539 | */ | |
540 | lockdep_init(); | |
541 | debug_objects_early_init(); | |
542 | ||
543 | /* | |
544 | * Set up the the initial canary ASAP: | |
545 | */ | |
546 | boot_init_stack_canary(); | |
547 | ||
548 | cgroup_init_early(); | |
549 | ||
550 | local_irq_disable(); | |
551 | early_boot_irqs_off(); | |
552 | early_init_irq_lock_class(); | |
553 | ||
554 | /* | |
555 | * Interrupts are still disabled. Do necessary setups, then | |
556 | * enable them | |
557 | */ | |
558 | lock_kernel(); | |
559 | tick_init(); | |
560 | boot_cpu_init(); | |
561 | page_address_init(); | |
562 | printk(KERN_NOTICE "%s", linux_banner); | |
563 | setup_arch(&command_line); | |
564 | mm_init_owner(&init_mm, &init_task); | |
565 | setup_command_line(command_line); | |
566 | setup_nr_cpu_ids(); | |
567 | setup_per_cpu_areas(); | |
568 | smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ | |
569 | ||
570 | build_all_zonelists(NULL); | |
571 | page_alloc_init(); | |
572 | ||
573 | printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line); | |
574 | parse_early_param(); | |
575 | parse_args("Booting kernel", static_command_line, __start___param, | |
576 | __stop___param - __start___param, | |
577 | &unknown_bootoption); | |
578 | /* | |
579 | * These use large bootmem allocations and must precede | |
580 | * kmem_cache_init() | |
581 | */ | |
582 | pidhash_init(); | |
583 | vfs_caches_init_early(); | |
584 | sort_main_extable(); | |
585 | trap_init(); | |
586 | mm_init(); | |
587 | /* | |
588 | * Set up the scheduler prior starting any interrupts (such as the | |
589 | * timer interrupt). Full topology setup happens at smp_init() | |
590 | * time - but meanwhile we still have a functioning scheduler. | |
591 | */ | |
592 | sched_init(); | |
593 | /* | |
594 | * Disable preemption - early bootup scheduling is extremely | |
595 | * fragile until we cpu_idle() for the first time. | |
596 | */ | |
597 | preempt_disable(); | |
598 | if (!irqs_disabled()) { | |
599 | printk(KERN_WARNING "start_kernel(): bug: interrupts were " | |
600 | "enabled *very* early, fixing it\n"); | |
601 | local_irq_disable(); | |
602 | } | |
603 | rcu_init(); | |
604 | radix_tree_init(); | |
605 | /* init some links before init_ISA_irqs() */ | |
606 | early_irq_init(); | |
607 | init_IRQ(); | |
608 | prio_tree_init(); | |
609 | init_timers(); | |
610 | hrtimers_init(); | |
611 | softirq_init(); | |
612 | timekeeping_init(); | |
613 | time_init(); | |
614 | profile_init(); | |
615 | if (!irqs_disabled()) | |
616 | printk(KERN_CRIT "start_kernel(): bug: interrupts were " | |
617 | "enabled early\n"); | |
618 | early_boot_irqs_on(); | |
619 | local_irq_enable(); | |
620 | ||
621 | /* Interrupts are enabled now so all GFP allocations are safe. */ | |
622 | gfp_allowed_mask = __GFP_BITS_MASK; | |
623 | ||
624 | kmem_cache_init_late(); | |
625 | ||
626 | /* | |
627 | * HACK ALERT! This is early. We're enabling the console before | |
628 | * we've done PCI setups etc, and console_init() must be aware of | |
629 | * this. But we do want output early, in case something goes wrong. | |
630 | */ | |
631 | console_init(); | |
632 | if (panic_later) | |
633 | panic(panic_later, panic_param); | |
634 | ||
635 | lockdep_info(); | |
636 | ||
637 | /* | |
638 | * Need to run this when irqs are enabled, because it wants | |
639 | * to self-test [hard/soft]-irqs on/off lock inversion bugs | |
640 | * too: | |
641 | */ | |
642 | locking_selftest(); | |
643 | ||
644 | #ifdef CONFIG_BLK_DEV_INITRD | |
645 | if (initrd_start && !initrd_below_start_ok && | |
646 | page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { | |
647 | printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - " | |
648 | "disabling it.\n", | |
649 | page_to_pfn(virt_to_page((void *)initrd_start)), | |
650 | min_low_pfn); | |
651 | initrd_start = 0; | |
652 | } | |
653 | #endif | |
654 | page_cgroup_init(); | |
655 | enable_debug_pagealloc(); | |
656 | kmemtrace_init(); | |
657 | kmemleak_init(); | |
658 | debug_objects_mem_init(); | |
659 | idr_init_cache(); | |
660 | setup_per_cpu_pageset(); | |
661 | numa_policy_init(); | |
662 | if (late_time_init) | |
663 | late_time_init(); | |
664 | sched_clock_init(); | |
665 | calibrate_delay(); | |
666 | pidmap_init(); | |
667 | anon_vma_init(); | |
668 | #ifdef CONFIG_X86 | |
669 | if (efi_enabled) | |
670 | efi_enter_virtual_mode(); | |
671 | #endif | |
672 | thread_info_cache_init(); | |
673 | cred_init(); | |
674 | fork_init(totalram_pages); | |
675 | proc_caches_init(); | |
676 | buffer_init(); | |
677 | key_init(); | |
678 | security_init(); | |
679 | dbg_late_init(); | |
680 | vfs_caches_init(totalram_pages); | |
681 | signals_init(); | |
682 | /* rootfs populating might need page-writeback */ | |
683 | page_writeback_init(); | |
684 | #ifdef CONFIG_PROC_FS | |
685 | proc_root_init(); | |
686 | #endif | |
687 | cgroup_init(); | |
688 | cpuset_init(); | |
689 | taskstats_init_early(); | |
690 | delayacct_init(); | |
691 | ||
692 | check_bugs(); | |
693 | ||
694 | acpi_early_init(); /* before LAPIC and SMP init */ | |
695 | sfi_init_late(); | |
696 | ||
697 | ftrace_init(); | |
698 | ||
699 | /* Do the rest non-__init'ed, we're now alive */ | |
700 | rest_init(); | |
701 | } | |
702 | ||
703 | /* Call all constructor functions linked into the kernel. */ | |
704 | static void __init do_ctors(void) | |
705 | { | |
706 | #ifdef CONFIG_CONSTRUCTORS | |
707 | ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; | |
708 | ||
709 | for (; fn < (ctor_fn_t *) __ctors_end; fn++) | |
710 | (*fn)(); | |
711 | #endif | |
712 | } | |
713 | ||
714 | int initcall_debug; | |
715 | core_param(initcall_debug, initcall_debug, bool, 0644); | |
716 | ||
717 | static char msgbuf[64]; | |
718 | static struct boot_trace_call call; | |
719 | static struct boot_trace_ret ret; | |
720 | ||
721 | int do_one_initcall(initcall_t fn) | |
722 | { | |
723 | int count = preempt_count(); | |
724 | ktime_t calltime, delta, rettime; | |
725 | ||
726 | if (initcall_debug) { | |
727 | call.caller = task_pid_nr(current); | |
728 | printk("calling %pF @ %i\n", fn, call.caller); | |
729 | calltime = ktime_get(); | |
730 | trace_boot_call(&call, fn); | |
731 | enable_boot_trace(); | |
732 | } | |
733 | ||
734 | ret.result = fn(); | |
735 | ||
736 | if (initcall_debug) { | |
737 | disable_boot_trace(); | |
738 | rettime = ktime_get(); | |
739 | delta = ktime_sub(rettime, calltime); | |
740 | ret.duration = (unsigned long long) ktime_to_ns(delta) >> 10; | |
741 | trace_boot_ret(&ret, fn); | |
742 | printk("initcall %pF returned %d after %Ld usecs\n", fn, | |
743 | ret.result, ret.duration); | |
744 | } | |
745 | ||
746 | msgbuf[0] = 0; | |
747 | ||
748 | if (ret.result && ret.result != -ENODEV && initcall_debug) | |
749 | sprintf(msgbuf, "error code %d ", ret.result); | |
750 | ||
751 | if (preempt_count() != count) { | |
752 | strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf)); | |
753 | preempt_count() = count; | |
754 | } | |
755 | if (irqs_disabled()) { | |
756 | strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); | |
757 | local_irq_enable(); | |
758 | } | |
759 | if (msgbuf[0]) { | |
760 | printk("initcall %pF returned with %s\n", fn, msgbuf); | |
761 | } | |
762 | ||
763 | return ret.result; | |
764 | } | |
765 | ||
766 | ||
767 | extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[]; | |
768 | ||
769 | static void __init do_initcalls(void) | |
770 | { | |
771 | initcall_t *fn; | |
772 | ||
773 | for (fn = __early_initcall_end; fn < __initcall_end; fn++) | |
774 | do_one_initcall(*fn); | |
775 | ||
776 | /* Make sure there is no pending stuff from the initcall sequence */ | |
777 | flush_scheduled_work(); | |
778 | } | |
779 | ||
780 | /* | |
781 | * Ok, the machine is now initialized. None of the devices | |
782 | * have been touched yet, but the CPU subsystem is up and | |
783 | * running, and memory and process management works. | |
784 | * | |
785 | * Now we can finally start doing some real work.. | |
786 | */ | |
787 | static void __init do_basic_setup(void) | |
788 | { | |
789 | init_workqueues(); | |
790 | cpuset_init_smp(); | |
791 | usermodehelper_init(); | |
792 | init_tmpfs(); | |
793 | driver_init(); | |
794 | init_irq_proc(); | |
795 | do_ctors(); | |
796 | do_initcalls(); | |
797 | } | |
798 | ||
799 | static void __init do_pre_smp_initcalls(void) | |
800 | { | |
801 | initcall_t *fn; | |
802 | ||
803 | for (fn = __initcall_start; fn < __early_initcall_end; fn++) | |
804 | do_one_initcall(*fn); | |
805 | } | |
806 | ||
807 | static void run_init_process(char *init_filename) | |
808 | { | |
809 | argv_init[0] = init_filename; | |
810 | kernel_execve(init_filename, argv_init, envp_init); | |
811 | } | |
812 | ||
813 | /* This is a non __init function. Force it to be noinline otherwise gcc | |
814 | * makes it inline to init() and it becomes part of init.text section | |
815 | */ | |
816 | static noinline int init_post(void) | |
817 | __releases(kernel_lock) | |
818 | { | |
819 | /* need to finish all async __init code before freeing the memory */ | |
820 | async_synchronize_full(); | |
821 | free_initmem(); | |
822 | unlock_kernel(); | |
823 | mark_rodata_ro(); | |
824 | system_state = SYSTEM_RUNNING; | |
825 | numa_default_policy(); | |
826 | ||
827 | ||
828 | current->signal->flags |= SIGNAL_UNKILLABLE; | |
829 | ||
830 | if (ramdisk_execute_command) { | |
831 | run_init_process(ramdisk_execute_command); | |
832 | printk(KERN_WARNING "Failed to execute %s\n", | |
833 | ramdisk_execute_command); | |
834 | } | |
835 | ||
836 | /* | |
837 | * We try each of these until one succeeds. | |
838 | * | |
839 | * The Bourne shell can be used instead of init if we are | |
840 | * trying to recover a really broken machine. | |
841 | */ | |
842 | if (execute_command) { | |
843 | run_init_process(execute_command); | |
844 | printk(KERN_WARNING "Failed to execute %s. Attempting " | |
845 | "defaults...\n", execute_command); | |
846 | } | |
847 | run_init_process("/sbin/init"); | |
848 | run_init_process("/etc/init"); | |
849 | run_init_process("/bin/init"); | |
850 | run_init_process("/bin/sh"); | |
851 | ||
852 | panic("No init found. Try passing init= option to kernel. " | |
853 | "See Linux Documentation/init.txt for guidance."); | |
854 | } | |
855 | ||
856 | static int __init kernel_init(void * unused) | |
857 | { | |
858 | lock_kernel(); | |
859 | ||
860 | /* | |
861 | * init can allocate pages on any node | |
862 | */ | |
863 | set_mems_allowed(node_states[N_HIGH_MEMORY]); | |
864 | /* | |
865 | * init can run on any cpu. | |
866 | */ | |
867 | set_cpus_allowed_ptr(current, cpu_all_mask); | |
868 | /* | |
869 | * Tell the world that we're going to be the grim | |
870 | * reaper of innocent orphaned children. | |
871 | * | |
872 | * We don't want people to have to make incorrect | |
873 | * assumptions about where in the task array this | |
874 | * can be found. | |
875 | */ | |
876 | init_pid_ns.child_reaper = current; | |
877 | ||
878 | cad_pid = task_pid(current); | |
879 | ||
880 | smp_prepare_cpus(setup_max_cpus); | |
881 | ||
882 | do_pre_smp_initcalls(); | |
883 | start_boot_trace(); | |
884 | ||
885 | smp_init(); | |
886 | sched_init_smp(); | |
887 | ||
888 | do_basic_setup(); | |
889 | ||
890 | /* Open the /dev/console on the rootfs, this should never fail */ | |
891 | if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) | |
892 | printk(KERN_WARNING "Warning: unable to open an initial console.\n"); | |
893 | ||
894 | (void) sys_dup(0); | |
895 | (void) sys_dup(0); | |
896 | /* | |
897 | * check if there is an early userspace init. If yes, let it do all | |
898 | * the work | |
899 | */ | |
900 | ||
901 | if (!ramdisk_execute_command) | |
902 | ramdisk_execute_command = "/init"; | |
903 | ||
904 | if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) { | |
905 | ramdisk_execute_command = NULL; | |
906 | prepare_namespace(); | |
907 | } | |
908 | ||
909 | /* | |
910 | * Ok, we have completed the initial bootup, and | |
911 | * we're essentially up and running. Get rid of the | |
912 | * initmem segments and start the user-mode stuff.. | |
913 | */ | |
914 | ||
915 | init_post(); | |
916 | return 0; | |
917 | } |