]>
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 | #define DEBUG /* Enable initcall_debug */ | |
13 | ||
14 | #include <linux/types.h> | |
15 | #include <linux/module.h> | |
16 | #include <linux/proc_fs.h> | |
17 | #include <linux/kernel.h> | |
18 | #include <linux/syscalls.h> | |
19 | #include <linux/stackprotector.h> | |
20 | #include <linux/string.h> | |
21 | #include <linux/ctype.h> | |
22 | #include <linux/delay.h> | |
23 | #include <linux/ioport.h> | |
24 | #include <linux/init.h> | |
25 | #include <linux/initrd.h> | |
26 | #include <linux/bootmem.h> | |
27 | #include <linux/acpi.h> | |
28 | #include <linux/tty.h> | |
29 | #include <linux/percpu.h> | |
30 | #include <linux/kmod.h> | |
31 | #include <linux/vmalloc.h> | |
32 | #include <linux/kernel_stat.h> | |
33 | #include <linux/start_kernel.h> | |
34 | #include <linux/security.h> | |
35 | #include <linux/smp.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_ext.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/sfi.h> | |
70 | #include <linux/shmem_fs.h> | |
71 | #include <linux/slab.h> | |
72 | #include <linux/perf_event.h> | |
73 | #include <linux/file.h> | |
74 | #include <linux/ptrace.h> | |
75 | #include <linux/blkdev.h> | |
76 | #include <linux/elevator.h> | |
77 | #include <linux/sched_clock.h> | |
78 | #include <linux/context_tracking.h> | |
79 | #include <linux/random.h> | |
80 | #include <linux/list.h> | |
81 | #include <linux/integrity.h> | |
82 | #include <linux/proc_ns.h> | |
83 | #include <linux/io.h> | |
84 | ||
85 | #include <asm/io.h> | |
86 | #include <asm/bugs.h> | |
87 | #include <asm/setup.h> | |
88 | #include <asm/sections.h> | |
89 | #include <asm/cacheflush.h> | |
90 | ||
91 | static int kernel_init(void *); | |
92 | ||
93 | extern void init_IRQ(void); | |
94 | extern void fork_init(void); | |
95 | extern void radix_tree_init(void); | |
96 | #ifndef CONFIG_DEBUG_RODATA | |
97 | static inline void mark_rodata_ro(void) { } | |
98 | #endif | |
99 | ||
100 | /* | |
101 | * Debug helper: via this flag we know that we are in 'early bootup code' | |
102 | * where only the boot processor is running with IRQ disabled. This means | |
103 | * two things - IRQ must not be enabled before the flag is cleared and some | |
104 | * operations which are not allowed with IRQ disabled are allowed while the | |
105 | * flag is set. | |
106 | */ | |
107 | bool early_boot_irqs_disabled __read_mostly; | |
108 | ||
109 | enum system_states system_state __read_mostly; | |
110 | EXPORT_SYMBOL(system_state); | |
111 | ||
112 | /* | |
113 | * Boot command-line arguments | |
114 | */ | |
115 | #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT | |
116 | #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT | |
117 | ||
118 | extern void time_init(void); | |
119 | /* Default late time init is NULL. archs can override this later. */ | |
120 | void (*__initdata late_time_init)(void); | |
121 | ||
122 | /* Untouched command line saved by arch-specific code. */ | |
123 | char __initdata boot_command_line[COMMAND_LINE_SIZE]; | |
124 | /* Untouched saved command line (eg. for /proc) */ | |
125 | char *saved_command_line; | |
126 | /* Command line for parameter parsing */ | |
127 | static char *static_command_line; | |
128 | /* Command line for per-initcall parameter parsing */ | |
129 | static char *initcall_command_line; | |
130 | ||
131 | static char *execute_command; | |
132 | static char *ramdisk_execute_command; | |
133 | ||
134 | /* | |
135 | * Used to generate warnings if static_key manipulation functions are used | |
136 | * before jump_label_init is called. | |
137 | */ | |
138 | bool static_key_initialized __read_mostly; | |
139 | EXPORT_SYMBOL_GPL(static_key_initialized); | |
140 | ||
141 | /* | |
142 | * If set, this is an indication to the drivers that reset the underlying | |
143 | * device before going ahead with the initialization otherwise driver might | |
144 | * rely on the BIOS and skip the reset operation. | |
145 | * | |
146 | * This is useful if kernel is booting in an unreliable environment. | |
147 | * For ex. kdump situation where previous kernel has crashed, BIOS has been | |
148 | * skipped and devices will be in unknown state. | |
149 | */ | |
150 | unsigned int reset_devices; | |
151 | EXPORT_SYMBOL(reset_devices); | |
152 | ||
153 | static int __init set_reset_devices(char *str) | |
154 | { | |
155 | reset_devices = 1; | |
156 | return 1; | |
157 | } | |
158 | ||
159 | __setup("reset_devices", set_reset_devices); | |
160 | ||
161 | static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; | |
162 | const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; | |
163 | static const char *panic_later, *panic_param; | |
164 | ||
165 | extern const struct obs_kernel_param __setup_start[], __setup_end[]; | |
166 | ||
167 | static int __init obsolete_checksetup(char *line) | |
168 | { | |
169 | const struct obs_kernel_param *p; | |
170 | int had_early_param = 0; | |
171 | ||
172 | p = __setup_start; | |
173 | do { | |
174 | int n = strlen(p->str); | |
175 | if (parameqn(line, p->str, n)) { | |
176 | if (p->early) { | |
177 | /* Already done in parse_early_param? | |
178 | * (Needs exact match on param part). | |
179 | * Keep iterating, as we can have early | |
180 | * params and __setups of same names 8( */ | |
181 | if (line[n] == '\0' || line[n] == '=') | |
182 | had_early_param = 1; | |
183 | } else if (!p->setup_func) { | |
184 | pr_warn("Parameter %s is obsolete, ignored\n", | |
185 | p->str); | |
186 | return 1; | |
187 | } else if (p->setup_func(line + n)) | |
188 | return 1; | |
189 | } | |
190 | p++; | |
191 | } while (p < __setup_end); | |
192 | ||
193 | return had_early_param; | |
194 | } | |
195 | ||
196 | /* | |
197 | * This should be approx 2 Bo*oMips to start (note initial shift), and will | |
198 | * still work even if initially too large, it will just take slightly longer | |
199 | */ | |
200 | unsigned long loops_per_jiffy = (1<<12); | |
201 | EXPORT_SYMBOL(loops_per_jiffy); | |
202 | ||
203 | static int __init debug_kernel(char *str) | |
204 | { | |
205 | console_loglevel = CONSOLE_LOGLEVEL_DEBUG; | |
206 | return 0; | |
207 | } | |
208 | ||
209 | static int __init quiet_kernel(char *str) | |
210 | { | |
211 | console_loglevel = CONSOLE_LOGLEVEL_QUIET; | |
212 | return 0; | |
213 | } | |
214 | ||
215 | early_param("debug", debug_kernel); | |
216 | early_param("quiet", quiet_kernel); | |
217 | ||
218 | static int __init loglevel(char *str) | |
219 | { | |
220 | int newlevel; | |
221 | ||
222 | /* | |
223 | * Only update loglevel value when a correct setting was passed, | |
224 | * to prevent blind crashes (when loglevel being set to 0) that | |
225 | * are quite hard to debug | |
226 | */ | |
227 | if (get_option(&str, &newlevel)) { | |
228 | console_loglevel = newlevel; | |
229 | return 0; | |
230 | } | |
231 | ||
232 | return -EINVAL; | |
233 | } | |
234 | ||
235 | early_param("loglevel", loglevel); | |
236 | ||
237 | /* Change NUL term back to "=", to make "param" the whole string. */ | |
238 | static int __init repair_env_string(char *param, char *val, | |
239 | const char *unused, void *arg) | |
240 | { | |
241 | if (val) { | |
242 | /* param=val or param="val"? */ | |
243 | if (val == param+strlen(param)+1) | |
244 | val[-1] = '='; | |
245 | else if (val == param+strlen(param)+2) { | |
246 | val[-2] = '='; | |
247 | memmove(val-1, val, strlen(val)+1); | |
248 | val--; | |
249 | } else | |
250 | BUG(); | |
251 | } | |
252 | return 0; | |
253 | } | |
254 | ||
255 | /* Anything after -- gets handed straight to init. */ | |
256 | static int __init set_init_arg(char *param, char *val, | |
257 | const char *unused, void *arg) | |
258 | { | |
259 | unsigned int i; | |
260 | ||
261 | if (panic_later) | |
262 | return 0; | |
263 | ||
264 | repair_env_string(param, val, unused, NULL); | |
265 | ||
266 | for (i = 0; argv_init[i]; i++) { | |
267 | if (i == MAX_INIT_ARGS) { | |
268 | panic_later = "init"; | |
269 | panic_param = param; | |
270 | return 0; | |
271 | } | |
272 | } | |
273 | argv_init[i] = param; | |
274 | return 0; | |
275 | } | |
276 | ||
277 | /* | |
278 | * Unknown boot options get handed to init, unless they look like | |
279 | * unused parameters (modprobe will find them in /proc/cmdline). | |
280 | */ | |
281 | static int __init unknown_bootoption(char *param, char *val, | |
282 | const char *unused, void *arg) | |
283 | { | |
284 | repair_env_string(param, val, unused, NULL); | |
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 = "env"; | |
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 = "init"; | |
315 | panic_param = param; | |
316 | } | |
317 | } | |
318 | argv_init[i] = param; | |
319 | } | |
320 | return 0; | |
321 | } | |
322 | ||
323 | static int __init init_setup(char *str) | |
324 | { | |
325 | unsigned int i; | |
326 | ||
327 | execute_command = str; | |
328 | /* | |
329 | * In case LILO is going to boot us with default command line, | |
330 | * it prepends "auto" before the whole cmdline which makes | |
331 | * the shell think it should execute a script with such name. | |
332 | * So we ignore all arguments entered _before_ init=... [MJ] | |
333 | */ | |
334 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
335 | argv_init[i] = NULL; | |
336 | return 1; | |
337 | } | |
338 | __setup("init=", init_setup); | |
339 | ||
340 | static int __init rdinit_setup(char *str) | |
341 | { | |
342 | unsigned int i; | |
343 | ||
344 | ramdisk_execute_command = str; | |
345 | /* See "auto" comment in init_setup */ | |
346 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
347 | argv_init[i] = NULL; | |
348 | return 1; | |
349 | } | |
350 | __setup("rdinit=", rdinit_setup); | |
351 | ||
352 | #ifndef CONFIG_SMP | |
353 | static const unsigned int setup_max_cpus = NR_CPUS; | |
354 | static inline void setup_nr_cpu_ids(void) { } | |
355 | static inline void smp_prepare_cpus(unsigned int maxcpus) { } | |
356 | #endif | |
357 | ||
358 | /* | |
359 | * We need to store the untouched command line for future reference. | |
360 | * We also need to store the touched command line since the parameter | |
361 | * parsing is performed in place, and we should allow a component to | |
362 | * store reference of name/value for future reference. | |
363 | */ | |
364 | static void __init setup_command_line(char *command_line) | |
365 | { | |
366 | saved_command_line = | |
367 | memblock_virt_alloc(strlen(boot_command_line) + 1, 0); | |
368 | initcall_command_line = | |
369 | memblock_virt_alloc(strlen(boot_command_line) + 1, 0); | |
370 | static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0); | |
371 | strcpy(saved_command_line, boot_command_line); | |
372 | strcpy(static_command_line, command_line); | |
373 | } | |
374 | ||
375 | /* | |
376 | * We need to finalize in a non-__init function or else race conditions | |
377 | * between the root thread and the init thread may cause start_kernel to | |
378 | * be reaped by free_initmem before the root thread has proceeded to | |
379 | * cpu_idle. | |
380 | * | |
381 | * gcc-3.4 accidentally inlines this function, so use noinline. | |
382 | */ | |
383 | ||
384 | static __initdata DECLARE_COMPLETION(kthreadd_done); | |
385 | ||
386 | static noinline void __init_refok rest_init(void) | |
387 | { | |
388 | int pid; | |
389 | ||
390 | rcu_scheduler_starting(); | |
391 | smpboot_thread_init(); | |
392 | /* | |
393 | * We need to spawn init first so that it obtains pid 1, however | |
394 | * the init task will end up wanting to create kthreads, which, if | |
395 | * we schedule it before we create kthreadd, will OOPS. | |
396 | */ | |
397 | kernel_thread(kernel_init, NULL, CLONE_FS); | |
398 | numa_default_policy(); | |
399 | pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); | |
400 | rcu_read_lock(); | |
401 | kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); | |
402 | rcu_read_unlock(); | |
403 | complete(&kthreadd_done); | |
404 | ||
405 | /* | |
406 | * The boot idle thread must execute schedule() | |
407 | * at least once to get things moving: | |
408 | */ | |
409 | init_idle_bootup_task(current); | |
410 | schedule_preempt_disabled(); | |
411 | /* Call into cpu_idle with preempt disabled */ | |
412 | cpu_startup_entry(CPUHP_ONLINE); | |
413 | } | |
414 | ||
415 | /* Check for early params. */ | |
416 | static int __init do_early_param(char *param, char *val, | |
417 | const char *unused, void *arg) | |
418 | { | |
419 | const struct obs_kernel_param *p; | |
420 | ||
421 | for (p = __setup_start; p < __setup_end; p++) { | |
422 | if ((p->early && parameq(param, p->str)) || | |
423 | (strcmp(param, "console") == 0 && | |
424 | strcmp(p->str, "earlycon") == 0) | |
425 | ) { | |
426 | if (p->setup_func(val) != 0) | |
427 | pr_warn("Malformed early option '%s'\n", param); | |
428 | } | |
429 | } | |
430 | /* We accept everything at this stage. */ | |
431 | return 0; | |
432 | } | |
433 | ||
434 | void __init parse_early_options(char *cmdline) | |
435 | { | |
436 | parse_args("early options", cmdline, NULL, 0, 0, 0, NULL, | |
437 | do_early_param); | |
438 | } | |
439 | ||
440 | /* Arch code calls this early on, or if not, just before other parsing. */ | |
441 | void __init parse_early_param(void) | |
442 | { | |
443 | static int done __initdata; | |
444 | static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; | |
445 | ||
446 | if (done) | |
447 | return; | |
448 | ||
449 | /* All fall through to do_early_param. */ | |
450 | strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); | |
451 | parse_early_options(tmp_cmdline); | |
452 | done = 1; | |
453 | } | |
454 | ||
455 | /* | |
456 | * Activate the first processor. | |
457 | */ | |
458 | ||
459 | static void __init boot_cpu_init(void) | |
460 | { | |
461 | int cpu = smp_processor_id(); | |
462 | /* Mark the boot cpu "present", "online" etc for SMP and UP case */ | |
463 | set_cpu_online(cpu, true); | |
464 | set_cpu_active(cpu, true); | |
465 | set_cpu_present(cpu, true); | |
466 | set_cpu_possible(cpu, true); | |
467 | } | |
468 | ||
469 | void __init __weak smp_setup_processor_id(void) | |
470 | { | |
471 | } | |
472 | ||
473 | # if THREAD_SIZE >= PAGE_SIZE | |
474 | void __init __weak thread_info_cache_init(void) | |
475 | { | |
476 | } | |
477 | #endif | |
478 | ||
479 | /* | |
480 | * Set up kernel memory allocators | |
481 | */ | |
482 | static void __init mm_init(void) | |
483 | { | |
484 | /* | |
485 | * page_ext requires contiguous pages, | |
486 | * bigger than MAX_ORDER unless SPARSEMEM. | |
487 | */ | |
488 | page_ext_init_flatmem(); | |
489 | mem_init(); | |
490 | kmem_cache_init(); | |
491 | percpu_init_late(); | |
492 | pgtable_init(); | |
493 | vmalloc_init(); | |
494 | ioremap_huge_init(); | |
495 | } | |
496 | ||
497 | asmlinkage __visible void __init start_kernel(void) | |
498 | { | |
499 | char *command_line; | |
500 | char *after_dashes; | |
501 | ||
502 | /* | |
503 | * Need to run as early as possible, to initialize the | |
504 | * lockdep hash: | |
505 | */ | |
506 | lockdep_init(); | |
507 | set_task_stack_end_magic(&init_task); | |
508 | smp_setup_processor_id(); | |
509 | debug_objects_early_init(); | |
510 | ||
511 | /* | |
512 | * Set up the the initial canary ASAP: | |
513 | */ | |
514 | boot_init_stack_canary(); | |
515 | ||
516 | cgroup_init_early(); | |
517 | ||
518 | local_irq_disable(); | |
519 | early_boot_irqs_disabled = true; | |
520 | ||
521 | /* | |
522 | * Interrupts are still disabled. Do necessary setups, then | |
523 | * enable them | |
524 | */ | |
525 | boot_cpu_init(); | |
526 | page_address_init(); | |
527 | pr_notice("%s", linux_banner); | |
528 | setup_arch(&command_line); | |
529 | mm_init_cpumask(&init_mm); | |
530 | setup_command_line(command_line); | |
531 | setup_nr_cpu_ids(); | |
532 | setup_per_cpu_areas(); | |
533 | smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ | |
534 | ||
535 | build_all_zonelists(NULL, NULL); | |
536 | page_alloc_init(); | |
537 | ||
538 | pr_notice("Kernel command line: %s\n", boot_command_line); | |
539 | parse_early_param(); | |
540 | after_dashes = parse_args("Booting kernel", | |
541 | static_command_line, __start___param, | |
542 | __stop___param - __start___param, | |
543 | -1, -1, NULL, &unknown_bootoption); | |
544 | if (!IS_ERR_OR_NULL(after_dashes)) | |
545 | parse_args("Setting init args", after_dashes, NULL, 0, -1, -1, | |
546 | NULL, set_init_arg); | |
547 | ||
548 | jump_label_init(); | |
549 | ||
550 | /* | |
551 | * These use large bootmem allocations and must precede | |
552 | * kmem_cache_init() | |
553 | */ | |
554 | setup_log_buf(0); | |
555 | pidhash_init(); | |
556 | vfs_caches_init_early(); | |
557 | sort_main_extable(); | |
558 | trap_init(); | |
559 | mm_init(); | |
560 | ||
561 | /* | |
562 | * Set up the scheduler prior starting any interrupts (such as the | |
563 | * timer interrupt). Full topology setup happens at smp_init() | |
564 | * time - but meanwhile we still have a functioning scheduler. | |
565 | */ | |
566 | sched_init(); | |
567 | /* | |
568 | * Disable preemption - early bootup scheduling is extremely | |
569 | * fragile until we cpu_idle() for the first time. | |
570 | */ | |
571 | preempt_disable(); | |
572 | if (WARN(!irqs_disabled(), | |
573 | "Interrupts were enabled *very* early, fixing it\n")) | |
574 | local_irq_disable(); | |
575 | idr_init_cache(); | |
576 | rcu_init(); | |
577 | ||
578 | /* trace_printk() and trace points may be used after this */ | |
579 | trace_init(); | |
580 | ||
581 | context_tracking_init(); | |
582 | radix_tree_init(); | |
583 | /* init some links before init_ISA_irqs() */ | |
584 | early_irq_init(); | |
585 | init_IRQ(); | |
586 | tick_init(); | |
587 | rcu_init_nohz(); | |
588 | init_timers(); | |
589 | hrtimers_init(); | |
590 | softirq_init(); | |
591 | timekeeping_init(); | |
592 | time_init(); | |
593 | sched_clock_postinit(); | |
594 | perf_event_init(); | |
595 | profile_init(); | |
596 | call_function_init(); | |
597 | WARN(!irqs_disabled(), "Interrupts were enabled early\n"); | |
598 | early_boot_irqs_disabled = false; | |
599 | local_irq_enable(); | |
600 | ||
601 | kmem_cache_init_late(); | |
602 | ||
603 | /* | |
604 | * HACK ALERT! This is early. We're enabling the console before | |
605 | * we've done PCI setups etc, and console_init() must be aware of | |
606 | * this. But we do want output early, in case something goes wrong. | |
607 | */ | |
608 | console_init(); | |
609 | if (panic_later) | |
610 | panic("Too many boot %s vars at `%s'", panic_later, | |
611 | panic_param); | |
612 | ||
613 | lockdep_info(); | |
614 | ||
615 | /* | |
616 | * Need to run this when irqs are enabled, because it wants | |
617 | * to self-test [hard/soft]-irqs on/off lock inversion bugs | |
618 | * too: | |
619 | */ | |
620 | locking_selftest(); | |
621 | ||
622 | #ifdef CONFIG_BLK_DEV_INITRD | |
623 | if (initrd_start && !initrd_below_start_ok && | |
624 | page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { | |
625 | pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", | |
626 | page_to_pfn(virt_to_page((void *)initrd_start)), | |
627 | min_low_pfn); | |
628 | initrd_start = 0; | |
629 | } | |
630 | #endif | |
631 | page_ext_init(); | |
632 | debug_objects_mem_init(); | |
633 | kmemleak_init(); | |
634 | setup_per_cpu_pageset(); | |
635 | numa_policy_init(); | |
636 | if (late_time_init) | |
637 | late_time_init(); | |
638 | sched_clock_init(); | |
639 | calibrate_delay(); | |
640 | pidmap_init(); | |
641 | anon_vma_init(); | |
642 | acpi_early_init(); | |
643 | #ifdef CONFIG_X86 | |
644 | if (efi_enabled(EFI_RUNTIME_SERVICES)) | |
645 | efi_enter_virtual_mode(); | |
646 | #endif | |
647 | #ifdef CONFIG_X86_ESPFIX64 | |
648 | /* Should be run before the first non-init thread is created */ | |
649 | init_espfix_bsp(); | |
650 | #endif | |
651 | thread_info_cache_init(); | |
652 | cred_init(); | |
653 | fork_init(); | |
654 | proc_caches_init(); | |
655 | buffer_init(); | |
656 | key_init(); | |
657 | security_init(); | |
658 | dbg_late_init(); | |
659 | vfs_caches_init(totalram_pages); | |
660 | signals_init(); | |
661 | /* rootfs populating might need page-writeback */ | |
662 | page_writeback_init(); | |
663 | proc_root_init(); | |
664 | nsfs_init(); | |
665 | cpuset_init(); | |
666 | cgroup_init(); | |
667 | taskstats_init_early(); | |
668 | delayacct_init(); | |
669 | ||
670 | check_bugs(); | |
671 | ||
672 | acpi_subsystem_init(); | |
673 | sfi_init_late(); | |
674 | ||
675 | if (efi_enabled(EFI_RUNTIME_SERVICES)) { | |
676 | efi_late_init(); | |
677 | efi_free_boot_services(); | |
678 | } | |
679 | ||
680 | ftrace_init(); | |
681 | ||
682 | /* Do the rest non-__init'ed, we're now alive */ | |
683 | rest_init(); | |
684 | } | |
685 | ||
686 | /* Call all constructor functions linked into the kernel. */ | |
687 | static void __init do_ctors(void) | |
688 | { | |
689 | #ifdef CONFIG_CONSTRUCTORS | |
690 | ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; | |
691 | ||
692 | for (; fn < (ctor_fn_t *) __ctors_end; fn++) | |
693 | (*fn)(); | |
694 | #endif | |
695 | } | |
696 | ||
697 | bool initcall_debug; | |
698 | core_param(initcall_debug, initcall_debug, bool, 0644); | |
699 | ||
700 | #ifdef CONFIG_KALLSYMS | |
701 | struct blacklist_entry { | |
702 | struct list_head next; | |
703 | char *buf; | |
704 | }; | |
705 | ||
706 | static __initdata_or_module LIST_HEAD(blacklisted_initcalls); | |
707 | ||
708 | static int __init initcall_blacklist(char *str) | |
709 | { | |
710 | char *str_entry; | |
711 | struct blacklist_entry *entry; | |
712 | ||
713 | /* str argument is a comma-separated list of functions */ | |
714 | do { | |
715 | str_entry = strsep(&str, ","); | |
716 | if (str_entry) { | |
717 | pr_debug("blacklisting initcall %s\n", str_entry); | |
718 | entry = alloc_bootmem(sizeof(*entry)); | |
719 | entry->buf = alloc_bootmem(strlen(str_entry) + 1); | |
720 | strcpy(entry->buf, str_entry); | |
721 | list_add(&entry->next, &blacklisted_initcalls); | |
722 | } | |
723 | } while (str_entry); | |
724 | ||
725 | return 0; | |
726 | } | |
727 | ||
728 | static bool __init_or_module initcall_blacklisted(initcall_t fn) | |
729 | { | |
730 | struct list_head *tmp; | |
731 | struct blacklist_entry *entry; | |
732 | char *fn_name; | |
733 | ||
734 | fn_name = kasprintf(GFP_KERNEL, "%pf", fn); | |
735 | if (!fn_name) | |
736 | return false; | |
737 | ||
738 | list_for_each(tmp, &blacklisted_initcalls) { | |
739 | entry = list_entry(tmp, struct blacklist_entry, next); | |
740 | if (!strcmp(fn_name, entry->buf)) { | |
741 | pr_debug("initcall %s blacklisted\n", fn_name); | |
742 | kfree(fn_name); | |
743 | return true; | |
744 | } | |
745 | } | |
746 | ||
747 | kfree(fn_name); | |
748 | return false; | |
749 | } | |
750 | #else | |
751 | static int __init initcall_blacklist(char *str) | |
752 | { | |
753 | pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); | |
754 | return 0; | |
755 | } | |
756 | ||
757 | static bool __init_or_module initcall_blacklisted(initcall_t fn) | |
758 | { | |
759 | return false; | |
760 | } | |
761 | #endif | |
762 | __setup("initcall_blacklist=", initcall_blacklist); | |
763 | ||
764 | static int __init_or_module do_one_initcall_debug(initcall_t fn) | |
765 | { | |
766 | ktime_t calltime, delta, rettime; | |
767 | unsigned long long duration; | |
768 | int ret; | |
769 | ||
770 | printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current)); | |
771 | calltime = ktime_get(); | |
772 | ret = fn(); | |
773 | rettime = ktime_get(); | |
774 | delta = ktime_sub(rettime, calltime); | |
775 | duration = (unsigned long long) ktime_to_ns(delta) >> 10; | |
776 | printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", | |
777 | fn, ret, duration); | |
778 | ||
779 | return ret; | |
780 | } | |
781 | ||
782 | int __init_or_module do_one_initcall(initcall_t fn) | |
783 | { | |
784 | int count = preempt_count(); | |
785 | int ret; | |
786 | char msgbuf[64]; | |
787 | ||
788 | if (initcall_blacklisted(fn)) | |
789 | return -EPERM; | |
790 | ||
791 | if (initcall_debug) | |
792 | ret = do_one_initcall_debug(fn); | |
793 | else | |
794 | ret = fn(); | |
795 | ||
796 | msgbuf[0] = 0; | |
797 | ||
798 | if (preempt_count() != count) { | |
799 | sprintf(msgbuf, "preemption imbalance "); | |
800 | preempt_count_set(count); | |
801 | } | |
802 | if (irqs_disabled()) { | |
803 | strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); | |
804 | local_irq_enable(); | |
805 | } | |
806 | WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf); | |
807 | ||
808 | return ret; | |
809 | } | |
810 | ||
811 | ||
812 | extern initcall_t __initcall_start[]; | |
813 | extern initcall_t __initcall0_start[]; | |
814 | extern initcall_t __initcall1_start[]; | |
815 | extern initcall_t __initcall2_start[]; | |
816 | extern initcall_t __initcall3_start[]; | |
817 | extern initcall_t __initcall4_start[]; | |
818 | extern initcall_t __initcall5_start[]; | |
819 | extern initcall_t __initcall6_start[]; | |
820 | extern initcall_t __initcall7_start[]; | |
821 | extern initcall_t __initcall_end[]; | |
822 | ||
823 | static initcall_t *initcall_levels[] __initdata = { | |
824 | __initcall0_start, | |
825 | __initcall1_start, | |
826 | __initcall2_start, | |
827 | __initcall3_start, | |
828 | __initcall4_start, | |
829 | __initcall5_start, | |
830 | __initcall6_start, | |
831 | __initcall7_start, | |
832 | __initcall_end, | |
833 | }; | |
834 | ||
835 | /* Keep these in sync with initcalls in include/linux/init.h */ | |
836 | static char *initcall_level_names[] __initdata = { | |
837 | "early", | |
838 | "core", | |
839 | "postcore", | |
840 | "arch", | |
841 | "subsys", | |
842 | "fs", | |
843 | "device", | |
844 | "late", | |
845 | }; | |
846 | ||
847 | static void __init do_initcall_level(int level) | |
848 | { | |
849 | initcall_t *fn; | |
850 | ||
851 | strcpy(initcall_command_line, saved_command_line); | |
852 | parse_args(initcall_level_names[level], | |
853 | initcall_command_line, __start___param, | |
854 | __stop___param - __start___param, | |
855 | level, level, | |
856 | NULL, &repair_env_string); | |
857 | ||
858 | for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) | |
859 | do_one_initcall(*fn); | |
860 | } | |
861 | ||
862 | static void __init do_initcalls(void) | |
863 | { | |
864 | int level; | |
865 | ||
866 | for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) | |
867 | do_initcall_level(level); | |
868 | } | |
869 | ||
870 | /* | |
871 | * Ok, the machine is now initialized. None of the devices | |
872 | * have been touched yet, but the CPU subsystem is up and | |
873 | * running, and memory and process management works. | |
874 | * | |
875 | * Now we can finally start doing some real work.. | |
876 | */ | |
877 | static void __init do_basic_setup(void) | |
878 | { | |
879 | cpuset_init_smp(); | |
880 | usermodehelper_init(); | |
881 | shmem_init(); | |
882 | driver_init(); | |
883 | init_irq_proc(); | |
884 | do_ctors(); | |
885 | usermodehelper_enable(); | |
886 | do_initcalls(); | |
887 | random_int_secret_init(); | |
888 | } | |
889 | ||
890 | static void __init do_pre_smp_initcalls(void) | |
891 | { | |
892 | initcall_t *fn; | |
893 | ||
894 | for (fn = __initcall_start; fn < __initcall0_start; fn++) | |
895 | do_one_initcall(*fn); | |
896 | } | |
897 | ||
898 | /* | |
899 | * This function requests modules which should be loaded by default and is | |
900 | * called twice right after initrd is mounted and right before init is | |
901 | * exec'd. If such modules are on either initrd or rootfs, they will be | |
902 | * loaded before control is passed to userland. | |
903 | */ | |
904 | void __init load_default_modules(void) | |
905 | { | |
906 | load_default_elevator_module(); | |
907 | } | |
908 | ||
909 | static int run_init_process(const char *init_filename) | |
910 | { | |
911 | argv_init[0] = init_filename; | |
912 | return do_execve(getname_kernel(init_filename), | |
913 | (const char __user *const __user *)argv_init, | |
914 | (const char __user *const __user *)envp_init); | |
915 | } | |
916 | ||
917 | static int try_to_run_init_process(const char *init_filename) | |
918 | { | |
919 | int ret; | |
920 | ||
921 | ret = run_init_process(init_filename); | |
922 | ||
923 | if (ret && ret != -ENOENT) { | |
924 | pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", | |
925 | init_filename, ret); | |
926 | } | |
927 | ||
928 | return ret; | |
929 | } | |
930 | ||
931 | static noinline void __init kernel_init_freeable(void); | |
932 | ||
933 | static int __ref kernel_init(void *unused) | |
934 | { | |
935 | int ret; | |
936 | ||
937 | kernel_init_freeable(); | |
938 | /* need to finish all async __init code before freeing the memory */ | |
939 | async_synchronize_full(); | |
940 | free_initmem(); | |
941 | mark_rodata_ro(); | |
942 | system_state = SYSTEM_RUNNING; | |
943 | numa_default_policy(); | |
944 | ||
945 | flush_delayed_fput(); | |
946 | ||
947 | if (ramdisk_execute_command) { | |
948 | ret = run_init_process(ramdisk_execute_command); | |
949 | if (!ret) | |
950 | return 0; | |
951 | pr_err("Failed to execute %s (error %d)\n", | |
952 | ramdisk_execute_command, ret); | |
953 | } | |
954 | ||
955 | /* | |
956 | * We try each of these until one succeeds. | |
957 | * | |
958 | * The Bourne shell can be used instead of init if we are | |
959 | * trying to recover a really broken machine. | |
960 | */ | |
961 | if (execute_command) { | |
962 | ret = run_init_process(execute_command); | |
963 | if (!ret) | |
964 | return 0; | |
965 | panic("Requested init %s failed (error %d).", | |
966 | execute_command, ret); | |
967 | } | |
968 | if (!try_to_run_init_process("/sbin/init") || | |
969 | !try_to_run_init_process("/etc/init") || | |
970 | !try_to_run_init_process("/bin/init") || | |
971 | !try_to_run_init_process("/bin/sh")) | |
972 | return 0; | |
973 | ||
974 | panic("No working init found. Try passing init= option to kernel. " | |
975 | "See Linux Documentation/init.txt for guidance."); | |
976 | } | |
977 | ||
978 | static noinline void __init kernel_init_freeable(void) | |
979 | { | |
980 | /* | |
981 | * Wait until kthreadd is all set-up. | |
982 | */ | |
983 | wait_for_completion(&kthreadd_done); | |
984 | ||
985 | /* Now the scheduler is fully set up and can do blocking allocations */ | |
986 | gfp_allowed_mask = __GFP_BITS_MASK; | |
987 | ||
988 | /* | |
989 | * init can allocate pages on any node | |
990 | */ | |
991 | set_mems_allowed(node_states[N_MEMORY]); | |
992 | /* | |
993 | * init can run on any cpu. | |
994 | */ | |
995 | set_cpus_allowed_ptr(current, cpu_all_mask); | |
996 | ||
997 | cad_pid = task_pid(current); | |
998 | ||
999 | smp_prepare_cpus(setup_max_cpus); | |
1000 | ||
1001 | do_pre_smp_initcalls(); | |
1002 | lockup_detector_init(); | |
1003 | ||
1004 | smp_init(); | |
1005 | sched_init_smp(); | |
1006 | ||
1007 | page_alloc_init_late(); | |
1008 | ||
1009 | do_basic_setup(); | |
1010 | ||
1011 | /* Open the /dev/console on the rootfs, this should never fail */ | |
1012 | if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) | |
1013 | pr_err("Warning: unable to open an initial console.\n"); | |
1014 | ||
1015 | (void) sys_dup(0); | |
1016 | (void) sys_dup(0); | |
1017 | /* | |
1018 | * check if there is an early userspace init. If yes, let it do all | |
1019 | * the work | |
1020 | */ | |
1021 | ||
1022 | if (!ramdisk_execute_command) | |
1023 | ramdisk_execute_command = "/init"; | |
1024 | ||
1025 | if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) { | |
1026 | ramdisk_execute_command = NULL; | |
1027 | prepare_namespace(); | |
1028 | } | |
1029 | ||
1030 | /* | |
1031 | * Ok, we have completed the initial bootup, and | |
1032 | * we're essentially up and running. Get rid of the | |
1033 | * initmem segments and start the user-mode stuff.. | |
1034 | * | |
1035 | * rootfs is available now, try loading the public keys | |
1036 | * and default modules | |
1037 | */ | |
1038 | ||
1039 | integrity_load_keys(); | |
1040 | load_default_modules(); | |
1041 | } |