]>
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 LT |
26 | #include <linux/bootmem.h> |
27 | #include <linux/tty.h> | |
28 | #include <linux/gfp.h> | |
29 | #include <linux/percpu.h> | |
30 | #include <linux/kmod.h> | |
db64fe02 | 31 | #include <linux/vmalloc.h> |
1da177e4 | 32 | #include <linux/kernel_stat.h> |
d7cd5611 | 33 | #include <linux/start_kernel.h> |
1da177e4 | 34 | #include <linux/security.h> |
3d442233 | 35 | #include <linux/smp.h> |
1da177e4 LT |
36 | #include <linux/workqueue.h> |
37 | #include <linux/profile.h> | |
38 | #include <linux/rcupdate.h> | |
39 | #include <linux/moduleparam.h> | |
40 | #include <linux/kallsyms.h> | |
41 | #include <linux/writeback.h> | |
42 | #include <linux/cpu.h> | |
43 | #include <linux/cpuset.h> | |
ddbcc7e8 | 44 | #include <linux/cgroup.h> |
1da177e4 | 45 | #include <linux/efi.h> |
906568c9 | 46 | #include <linux/tick.h> |
6168a702 | 47 | #include <linux/interrupt.h> |
c757249a | 48 | #include <linux/taskstats_kern.h> |
ca74e92b | 49 | #include <linux/delayacct.h> |
1da177e4 LT |
50 | #include <linux/unistd.h> |
51 | #include <linux/rmap.h> | |
52 | #include <linux/mempolicy.h> | |
53 | #include <linux/key.h> | |
b6cd0b77 | 54 | #include <linux/buffer_head.h> |
94b6da5a | 55 | #include <linux/page_cgroup.h> |
9a11b49a | 56 | #include <linux/debug_locks.h> |
3ac7fe5a | 57 | #include <linux/debugobjects.h> |
fbb9ce95 | 58 | #include <linux/lockdep.h> |
3c7b4e6b | 59 | #include <linux/kmemleak.h> |
84d73786 | 60 | #include <linux/pid_namespace.h> |
1f21782e | 61 | #include <linux/device.h> |
73c27992 | 62 | #include <linux/kthread.h> |
e6fe6649 | 63 | #include <linux/sched.h> |
a1c9eea9 | 64 | #include <linux/signal.h> |
199f0ca5 | 65 | #include <linux/idr.h> |
68bf21aa | 66 | #include <linux/ftrace.h> |
22a9d645 | 67 | #include <linux/async.h> |
dfec072e | 68 | #include <linux/kmemcheck.h> |
02af61bb | 69 | #include <linux/kmemtrace.h> |
3f5ec136 | 70 | #include <trace/boot.h> |
1da177e4 LT |
71 | |
72 | #include <asm/io.h> | |
73 | #include <asm/bugs.h> | |
74 | #include <asm/setup.h> | |
a940199f | 75 | #include <asm/sections.h> |
37b73c82 | 76 | #include <asm/cacheflush.h> |
1da177e4 | 77 | |
1da177e4 LT |
78 | #ifdef CONFIG_X86_LOCAL_APIC |
79 | #include <asm/smp.h> | |
80 | #endif | |
81 | ||
aae5f662 | 82 | static int kernel_init(void *); |
1da177e4 LT |
83 | |
84 | extern void init_IRQ(void); | |
1da177e4 LT |
85 | extern void fork_init(unsigned long); |
86 | extern void mca_init(void); | |
87 | extern void sbus_init(void); | |
1da177e4 LT |
88 | extern void prio_tree_init(void); |
89 | extern void radix_tree_init(void); | |
90 | extern void free_initmem(void); | |
1da177e4 LT |
91 | #ifdef CONFIG_ACPI |
92 | extern void acpi_early_init(void); | |
93 | #else | |
94 | static inline void acpi_early_init(void) { } | |
95 | #endif | |
37b73c82 AV |
96 | #ifndef CONFIG_DEBUG_RODATA |
97 | static inline void mark_rodata_ro(void) { } | |
98 | #endif | |
1da177e4 LT |
99 | |
100 | #ifdef CONFIG_TC | |
101 | extern void tc_init(void); | |
102 | #endif | |
103 | ||
a6826048 | 104 | enum system_states system_state __read_mostly; |
1da177e4 LT |
105 | EXPORT_SYMBOL(system_state); |
106 | ||
107 | /* | |
108 | * Boot command-line arguments | |
109 | */ | |
110 | #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT | |
111 | #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT | |
112 | ||
113 | extern void time_init(void); | |
114 | /* Default late time init is NULL. archs can override this later. */ | |
d2e3192b | 115 | void (*__initdata late_time_init)(void); |
1da177e4 LT |
116 | extern void softirq_init(void); |
117 | ||
30d7e0d4 ABL |
118 | /* Untouched command line saved by arch-specific code. */ |
119 | char __initdata boot_command_line[COMMAND_LINE_SIZE]; | |
120 | /* Untouched saved command line (eg. for /proc) */ | |
121 | char *saved_command_line; | |
122 | /* Command line for parameter parsing */ | |
123 | static char *static_command_line; | |
1da177e4 LT |
124 | |
125 | static char *execute_command; | |
ffdfc409 | 126 | static char *ramdisk_execute_command; |
1da177e4 | 127 | |
8b3b2955 | 128 | #ifdef CONFIG_SMP |
1da177e4 | 129 | /* Setup configured maximum number of CPUs to activate */ |
ca74a6f8 | 130 | unsigned int __initdata setup_max_cpus = NR_CPUS; |
7e96287d | 131 | |
1da177e4 LT |
132 | /* |
133 | * Setup routine for controlling SMP activation | |
134 | * | |
135 | * Command-line option of "nosmp" or "maxcpus=0" will disable SMP | |
136 | * activation entirely (the MPS table probe still happens, though). | |
137 | * | |
138 | * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer | |
139 | * greater than 0, limits the maximum number of CPUs activated in | |
140 | * SMP mode to <NUM>. | |
141 | */ | |
65a4e574 IM |
142 | |
143 | void __weak arch_disable_smp_support(void) { } | |
61ec7567 | 144 | |
1da177e4 LT |
145 | static int __init nosmp(char *str) |
146 | { | |
ca74a6f8 | 147 | setup_max_cpus = 0; |
65a4e574 IM |
148 | arch_disable_smp_support(); |
149 | ||
8b3b2955 | 150 | return 0; |
1da177e4 LT |
151 | } |
152 | ||
8b3b2955 | 153 | early_param("nosmp", nosmp); |
1da177e4 LT |
154 | |
155 | static int __init maxcpus(char *str) | |
156 | { | |
ca74a6f8 AK |
157 | get_option(&str, &setup_max_cpus); |
158 | if (setup_max_cpus == 0) | |
65a4e574 | 159 | arch_disable_smp_support(); |
61ec7567 LB |
160 | |
161 | return 0; | |
1da177e4 LT |
162 | } |
163 | ||
81340977 | 164 | early_param("maxcpus", maxcpus); |
8b3b2955 | 165 | #else |
65a4e574 | 166 | const unsigned int setup_max_cpus = NR_CPUS; |
8b3b2955 JB |
167 | #endif |
168 | ||
169 | /* | |
170 | * If set, this is an indication to the drivers that reset the underlying | |
171 | * device before going ahead with the initialization otherwise driver might | |
172 | * rely on the BIOS and skip the reset operation. | |
173 | * | |
174 | * This is useful if kernel is booting in an unreliable environment. | |
175 | * For ex. kdump situaiton where previous kernel has crashed, BIOS has been | |
176 | * skipped and devices will be in unknown state. | |
177 | */ | |
178 | unsigned int reset_devices; | |
179 | EXPORT_SYMBOL(reset_devices); | |
1da177e4 | 180 | |
7e96287d VG |
181 | static int __init set_reset_devices(char *str) |
182 | { | |
183 | reset_devices = 1; | |
184 | return 1; | |
185 | } | |
186 | ||
187 | __setup("reset_devices", set_reset_devices); | |
188 | ||
1da177e4 LT |
189 | static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; |
190 | char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; | |
191 | static const char *panic_later, *panic_param; | |
192 | ||
193 | extern struct obs_kernel_param __setup_start[], __setup_end[]; | |
194 | ||
195 | static int __init obsolete_checksetup(char *line) | |
196 | { | |
197 | struct obs_kernel_param *p; | |
33df0d19 | 198 | int had_early_param = 0; |
1da177e4 LT |
199 | |
200 | p = __setup_start; | |
201 | do { | |
202 | int n = strlen(p->str); | |
203 | if (!strncmp(line, p->str, n)) { | |
204 | if (p->early) { | |
33df0d19 RR |
205 | /* Already done in parse_early_param? |
206 | * (Needs exact match on param part). | |
207 | * Keep iterating, as we can have early | |
208 | * params and __setups of same names 8( */ | |
1da177e4 | 209 | if (line[n] == '\0' || line[n] == '=') |
33df0d19 | 210 | had_early_param = 1; |
1da177e4 LT |
211 | } else if (!p->setup_func) { |
212 | printk(KERN_WARNING "Parameter %s is obsolete," | |
213 | " ignored\n", p->str); | |
214 | return 1; | |
215 | } else if (p->setup_func(line + n)) | |
216 | return 1; | |
217 | } | |
218 | p++; | |
219 | } while (p < __setup_end); | |
33df0d19 RR |
220 | |
221 | return had_early_param; | |
1da177e4 LT |
222 | } |
223 | ||
224 | /* | |
225 | * This should be approx 2 Bo*oMips to start (note initial shift), and will | |
226 | * still work even if initially too large, it will just take slightly longer | |
227 | */ | |
228 | unsigned long loops_per_jiffy = (1<<12); | |
229 | ||
230 | EXPORT_SYMBOL(loops_per_jiffy); | |
231 | ||
232 | static int __init debug_kernel(char *str) | |
233 | { | |
1da177e4 | 234 | console_loglevel = 10; |
f6f21c81 | 235 | return 0; |
1da177e4 LT |
236 | } |
237 | ||
238 | static int __init quiet_kernel(char *str) | |
239 | { | |
1da177e4 | 240 | console_loglevel = 4; |
f6f21c81 | 241 | return 0; |
1da177e4 LT |
242 | } |
243 | ||
f6f21c81 YL |
244 | early_param("debug", debug_kernel); |
245 | early_param("quiet", quiet_kernel); | |
1da177e4 LT |
246 | |
247 | static int __init loglevel(char *str) | |
248 | { | |
249 | get_option(&str, &console_loglevel); | |
d9d4fcfe | 250 | return 0; |
1da177e4 LT |
251 | } |
252 | ||
f6f21c81 | 253 | early_param("loglevel", loglevel); |
1da177e4 LT |
254 | |
255 | /* | |
256 | * Unknown boot options get handed to init, unless they look like | |
257 | * failed parameters | |
258 | */ | |
259 | static int __init unknown_bootoption(char *param, char *val) | |
260 | { | |
261 | /* Change NUL term back to "=", to make "param" the whole string. */ | |
262 | if (val) { | |
263 | /* param=val or param="val"? */ | |
264 | if (val == param+strlen(param)+1) | |
265 | val[-1] = '='; | |
266 | else if (val == param+strlen(param)+2) { | |
267 | val[-2] = '='; | |
268 | memmove(val-1, val, strlen(val)+1); | |
269 | val--; | |
270 | } else | |
271 | BUG(); | |
272 | } | |
273 | ||
274 | /* Handle obsolete-style parameters */ | |
275 | if (obsolete_checksetup(param)) | |
276 | return 0; | |
277 | ||
278 | /* | |
211fee8a | 279 | * Preemptive maintenance for "why didn't my misspelled command |
1da177e4 LT |
280 | * line work?" |
281 | */ | |
282 | if (strchr(param, '.') && (!val || strchr(param, '.') < val)) { | |
283 | printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param); | |
284 | return 0; | |
285 | } | |
286 | ||
287 | if (panic_later) | |
288 | return 0; | |
289 | ||
290 | if (val) { | |
291 | /* Environment option */ | |
292 | unsigned int i; | |
293 | for (i = 0; envp_init[i]; i++) { | |
294 | if (i == MAX_INIT_ENVS) { | |
295 | panic_later = "Too many boot env vars at `%s'"; | |
296 | panic_param = param; | |
297 | } | |
298 | if (!strncmp(param, envp_init[i], val - param)) | |
299 | break; | |
300 | } | |
301 | envp_init[i] = param; | |
302 | } else { | |
303 | /* Command line option */ | |
304 | unsigned int i; | |
305 | for (i = 0; argv_init[i]; i++) { | |
306 | if (i == MAX_INIT_ARGS) { | |
307 | panic_later = "Too many boot init vars at `%s'"; | |
308 | panic_param = param; | |
309 | } | |
310 | } | |
311 | argv_init[i] = param; | |
312 | } | |
313 | return 0; | |
314 | } | |
315 | ||
12d6f21e IM |
316 | #ifdef CONFIG_DEBUG_PAGEALLOC |
317 | int __read_mostly debug_pagealloc_enabled = 0; | |
318 | #endif | |
319 | ||
1da177e4 LT |
320 | static int __init init_setup(char *str) |
321 | { | |
322 | unsigned int i; | |
323 | ||
324 | execute_command = str; | |
325 | /* | |
326 | * In case LILO is going to boot us with default command line, | |
327 | * it prepends "auto" before the whole cmdline which makes | |
328 | * the shell think it should execute a script with such name. | |
329 | * So we ignore all arguments entered _before_ init=... [MJ] | |
330 | */ | |
331 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
332 | argv_init[i] = NULL; | |
333 | return 1; | |
334 | } | |
335 | __setup("init=", init_setup); | |
336 | ||
ffdfc409 OJ |
337 | static int __init rdinit_setup(char *str) |
338 | { | |
339 | unsigned int i; | |
340 | ||
341 | ramdisk_execute_command = str; | |
342 | /* See "auto" comment in init_setup */ | |
343 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
344 | argv_init[i] = NULL; | |
345 | return 1; | |
346 | } | |
347 | __setup("rdinit=", rdinit_setup); | |
348 | ||
1da177e4 LT |
349 | #ifndef CONFIG_SMP |
350 | ||
351 | #ifdef CONFIG_X86_LOCAL_APIC | |
352 | static void __init smp_init(void) | |
353 | { | |
354 | APIC_init_uniprocessor(); | |
355 | } | |
356 | #else | |
357 | #define smp_init() do { } while (0) | |
358 | #endif | |
359 | ||
360 | static inline void setup_per_cpu_areas(void) { } | |
e0982e90 | 361 | static inline void setup_nr_cpu_ids(void) { } |
1da177e4 LT |
362 | static inline void smp_prepare_cpus(unsigned int maxcpus) { } |
363 | ||
364 | #else | |
365 | ||
321a8e9d MT |
366 | #if NR_CPUS > BITS_PER_LONG |
367 | cpumask_t cpu_mask_all __read_mostly = CPU_MASK_ALL; | |
368 | EXPORT_SYMBOL(cpu_mask_all); | |
369 | #endif | |
370 | ||
e0982e90 MT |
371 | /* Setup number of possible processor ids */ |
372 | int nr_cpu_ids __read_mostly = NR_CPUS; | |
373 | EXPORT_SYMBOL(nr_cpu_ids); | |
374 | ||
375 | /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */ | |
376 | static void __init setup_nr_cpu_ids(void) | |
377 | { | |
e0c0ba73 | 378 | nr_cpu_ids = find_last_bit(cpumask_bits(cpu_possible_mask),NR_CPUS) + 1; |
e0982e90 MT |
379 | } |
380 | ||
dd5af90a | 381 | #ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA |
b73b459f | 382 | unsigned long __per_cpu_offset[NR_CPUS] __read_mostly; |
1da177e4 LT |
383 | |
384 | EXPORT_SYMBOL(__per_cpu_offset); | |
385 | ||
386 | static void __init setup_per_cpu_areas(void) | |
387 | { | |
388 | unsigned long size, i; | |
389 | char *ptr; | |
63872f87 | 390 | unsigned long nr_possible_cpus = num_possible_cpus(); |
1da177e4 LT |
391 | |
392 | /* Copy section for each CPU (we discard the original) */ | |
b6e3590f JF |
393 | size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE); |
394 | ptr = alloc_bootmem_pages(size * nr_possible_cpus); | |
1da177e4 | 395 | |
0a945022 | 396 | for_each_possible_cpu(i) { |
1da177e4 LT |
397 | __per_cpu_offset[i] = ptr - __per_cpu_start; |
398 | memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start); | |
63872f87 | 399 | ptr += size; |
1da177e4 LT |
400 | } |
401 | } | |
dd5af90a | 402 | #endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */ |
1da177e4 LT |
403 | |
404 | /* Called by boot processor to activate the rest. */ | |
405 | static void __init smp_init(void) | |
406 | { | |
53b8a315 | 407 | unsigned int cpu; |
1da177e4 | 408 | |
e761b772 MK |
409 | /* |
410 | * Set up the current CPU as possible to migrate to. | |
411 | * The other ones will be done by cpu_up/cpu_down() | |
412 | */ | |
2b17fa50 | 413 | set_cpu_active(smp_processor_id(), true); |
e761b772 | 414 | |
1da177e4 | 415 | /* FIXME: This should be done in userspace --RR */ |
53b8a315 | 416 | for_each_present_cpu(cpu) { |
ca74a6f8 | 417 | if (num_online_cpus() >= setup_max_cpus) |
1da177e4 | 418 | break; |
53b8a315 CL |
419 | if (!cpu_online(cpu)) |
420 | cpu_up(cpu); | |
1da177e4 LT |
421 | } |
422 | ||
423 | /* Any cleanup work */ | |
424 | printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus()); | |
ca74a6f8 | 425 | smp_cpus_done(setup_max_cpus); |
1da177e4 LT |
426 | } |
427 | ||
428 | #endif | |
429 | ||
30d7e0d4 ABL |
430 | /* |
431 | * We need to store the untouched command line for future reference. | |
432 | * We also need to store the touched command line since the parameter | |
433 | * parsing is performed in place, and we should allow a component to | |
434 | * store reference of name/value for future reference. | |
435 | */ | |
436 | static void __init setup_command_line(char *command_line) | |
437 | { | |
438 | saved_command_line = alloc_bootmem(strlen (boot_command_line)+1); | |
439 | static_command_line = alloc_bootmem(strlen (command_line)+1); | |
440 | strcpy (saved_command_line, boot_command_line); | |
441 | strcpy (static_command_line, command_line); | |
442 | } | |
443 | ||
1da177e4 LT |
444 | /* |
445 | * We need to finalize in a non-__init function or else race conditions | |
446 | * between the root thread and the init thread may cause start_kernel to | |
447 | * be reaped by free_initmem before the root thread has proceeded to | |
448 | * cpu_idle. | |
449 | * | |
450 | * gcc-3.4 accidentally inlines this function, so use noinline. | |
451 | */ | |
452 | ||
f99ebf0a | 453 | static noinline void __init_refok rest_init(void) |
1da177e4 LT |
454 | __releases(kernel_lock) |
455 | { | |
73c27992 EB |
456 | int pid; |
457 | ||
aae5f662 | 458 | kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND); |
1da177e4 | 459 | numa_default_policy(); |
73c27992 | 460 | pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); |
5cd20455 | 461 | kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); |
1da177e4 | 462 | unlock_kernel(); |
f340c0d1 IM |
463 | |
464 | /* | |
465 | * The boot idle thread must execute schedule() | |
1df21055 | 466 | * at least once to get things moving: |
f340c0d1 | 467 | */ |
1df21055 | 468 | init_idle_bootup_task(current); |
a6826048 | 469 | rcu_scheduler_starting(); |
5bfb5d69 | 470 | preempt_enable_no_resched(); |
f340c0d1 | 471 | schedule(); |
5bfb5d69 | 472 | preempt_disable(); |
f340c0d1 | 473 | |
5bfb5d69 | 474 | /* Call into cpu_idle with preempt disabled */ |
1da177e4 | 475 | cpu_idle(); |
1df21055 | 476 | } |
1da177e4 LT |
477 | |
478 | /* Check for early params. */ | |
479 | static int __init do_early_param(char *param, char *val) | |
480 | { | |
481 | struct obs_kernel_param *p; | |
482 | ||
483 | for (p = __setup_start; p < __setup_end; p++) { | |
18a8bd94 YL |
484 | if ((p->early && strcmp(param, p->str) == 0) || |
485 | (strcmp(param, "console") == 0 && | |
486 | strcmp(p->str, "earlycon") == 0) | |
487 | ) { | |
1da177e4 LT |
488 | if (p->setup_func(val) != 0) |
489 | printk(KERN_WARNING | |
490 | "Malformed early option '%s'\n", param); | |
491 | } | |
492 | } | |
493 | /* We accept everything at this stage. */ | |
494 | return 0; | |
495 | } | |
496 | ||
13977091 MD |
497 | void __init parse_early_options(char *cmdline) |
498 | { | |
499 | parse_args("early options", cmdline, NULL, 0, do_early_param); | |
500 | } | |
501 | ||
1da177e4 LT |
502 | /* Arch code calls this early on, or if not, just before other parsing. */ |
503 | void __init parse_early_param(void) | |
504 | { | |
505 | static __initdata int done = 0; | |
506 | static __initdata char tmp_cmdline[COMMAND_LINE_SIZE]; | |
507 | ||
508 | if (done) | |
509 | return; | |
510 | ||
511 | /* All fall through to do_early_param. */ | |
30d7e0d4 | 512 | strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); |
13977091 | 513 | parse_early_options(tmp_cmdline); |
1da177e4 LT |
514 | done = 1; |
515 | } | |
516 | ||
517 | /* | |
518 | * Activate the first processor. | |
519 | */ | |
520 | ||
44fd2299 SS |
521 | static void __init boot_cpu_init(void) |
522 | { | |
523 | int cpu = smp_processor_id(); | |
524 | /* Mark the boot cpu "present", "online" etc for SMP and UP case */ | |
915441b6 RR |
525 | set_cpu_online(cpu, true); |
526 | set_cpu_present(cpu, true); | |
527 | set_cpu_possible(cpu, true); | |
44fd2299 SS |
528 | } |
529 | ||
839ad62e | 530 | void __init __weak smp_setup_processor_id(void) |
033ab7f8 AM |
531 | { |
532 | } | |
533 | ||
8c9843e5 BH |
534 | void __init __weak thread_info_cache_init(void) |
535 | { | |
536 | } | |
537 | ||
444f478f PE |
538 | /* |
539 | * Set up kernel memory allocators | |
540 | */ | |
541 | static void __init mm_init(void) | |
542 | { | |
ca371c0d KH |
543 | /* |
544 | * page_cgroup requires countinous pages as memmap | |
545 | * and it's bigger than MAX_ORDER unless SPARSEMEM. | |
546 | */ | |
547 | page_cgroup_init_flatmem(); | |
444f478f PE |
548 | mem_init(); |
549 | kmem_cache_init(); | |
c868d550 | 550 | pgtable_cache_init(); |
444f478f PE |
551 | vmalloc_init(); |
552 | } | |
553 | ||
1da177e4 LT |
554 | asmlinkage void __init start_kernel(void) |
555 | { | |
556 | char * command_line; | |
557 | extern struct kernel_param __start___param[], __stop___param[]; | |
033ab7f8 AM |
558 | |
559 | smp_setup_processor_id(); | |
560 | ||
fbb9ce95 IM |
561 | /* |
562 | * Need to run as early as possible, to initialize the | |
563 | * lockdep hash: | |
564 | */ | |
565 | lockdep_init(); | |
3ac7fe5a | 566 | debug_objects_early_init(); |
42059429 IM |
567 | |
568 | /* | |
569 | * Set up the the initial canary ASAP: | |
570 | */ | |
571 | boot_init_stack_canary(); | |
572 | ||
ddbcc7e8 | 573 | cgroup_init_early(); |
fbb9ce95 IM |
574 | |
575 | local_irq_disable(); | |
576 | early_boot_irqs_off(); | |
243c7621 | 577 | early_init_irq_lock_class(); |
fbb9ce95 | 578 | |
1da177e4 LT |
579 | /* |
580 | * Interrupts are still disabled. Do necessary setups, then | |
581 | * enable them | |
582 | */ | |
583 | lock_kernel(); | |
906568c9 | 584 | tick_init(); |
44fd2299 | 585 | boot_cpu_init(); |
1da177e4 | 586 | page_address_init(); |
657cafa6 | 587 | printk(KERN_NOTICE "%s", linux_banner); |
1da177e4 | 588 | setup_arch(&command_line); |
cf475ad2 | 589 | mm_init_owner(&init_mm, &init_task); |
30d7e0d4 | 590 | setup_command_line(command_line); |
1da177e4 | 591 | setup_per_cpu_areas(); |
e0982e90 | 592 | setup_nr_cpu_ids(); |
44fd2299 | 593 | smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ |
1da177e4 | 594 | |
83b519e8 PE |
595 | build_all_zonelists(); |
596 | page_alloc_init(); | |
597 | ||
598 | printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line); | |
599 | parse_early_param(); | |
600 | parse_args("Booting kernel", static_command_line, __start___param, | |
601 | __stop___param - __start___param, | |
602 | &unknown_bootoption); | |
603 | /* | |
604 | * These use large bootmem allocations and must precede | |
605 | * kmem_cache_init() | |
606 | */ | |
607 | pidhash_init(); | |
83b519e8 PE |
608 | vfs_caches_init_early(); |
609 | sort_main_extable(); | |
610 | trap_init(); | |
444f478f | 611 | mm_init(); |
1da177e4 LT |
612 | /* |
613 | * Set up the scheduler prior starting any interrupts (such as the | |
614 | * timer interrupt). Full topology setup happens at smp_init() | |
615 | * time - but meanwhile we still have a functioning scheduler. | |
616 | */ | |
617 | sched_init(); | |
618 | /* | |
619 | * Disable preemption - early bootup scheduling is extremely | |
620 | * fragile until we cpu_idle() for the first time. | |
621 | */ | |
622 | preempt_disable(); | |
c4a68306 AB |
623 | if (!irqs_disabled()) { |
624 | printk(KERN_WARNING "start_kernel(): bug: interrupts were " | |
625 | "enabled *very* early, fixing it\n"); | |
626 | local_irq_disable(); | |
627 | } | |
1da177e4 | 628 | rcu_init(); |
0b8f1efa YL |
629 | /* init some links before init_ISA_irqs() */ |
630 | early_irq_init(); | |
1da177e4 | 631 | init_IRQ(); |
3c7b4e6b | 632 | prio_tree_init(); |
1da177e4 | 633 | init_timers(); |
c0a31329 | 634 | hrtimers_init(); |
1da177e4 | 635 | softirq_init(); |
ad596171 | 636 | timekeeping_init(); |
88fecaa2 | 637 | time_init(); |
3e51f33f | 638 | sched_clock_init(); |
93e02814 HC |
639 | profile_init(); |
640 | if (!irqs_disabled()) | |
24d431d0 RL |
641 | printk(KERN_CRIT "start_kernel(): bug: interrupts were " |
642 | "enabled early\n"); | |
fbb9ce95 | 643 | early_boot_irqs_on(); |
93e02814 | 644 | local_irq_enable(); |
7e85ee0c | 645 | kmem_cache_init_late(); |
1da177e4 LT |
646 | |
647 | /* | |
648 | * HACK ALERT! This is early. We're enabling the console before | |
649 | * we've done PCI setups etc, and console_init() must be aware of | |
650 | * this. But we do want output early, in case something goes wrong. | |
651 | */ | |
652 | console_init(); | |
653 | if (panic_later) | |
654 | panic(panic_later, panic_param); | |
fbb9ce95 IM |
655 | |
656 | lockdep_info(); | |
657 | ||
9a11b49a IM |
658 | /* |
659 | * Need to run this when irqs are enabled, because it wants | |
660 | * to self-test [hard/soft]-irqs on/off lock inversion bugs | |
661 | * too: | |
662 | */ | |
663 | locking_selftest(); | |
664 | ||
1da177e4 LT |
665 | #ifdef CONFIG_BLK_DEV_INITRD |
666 | if (initrd_start && !initrd_below_start_ok && | |
bd673c7c | 667 | page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { |
1da177e4 | 668 | printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - " |
fb6624eb | 669 | "disabling it.\n", |
bd673c7c GU |
670 | page_to_pfn(virt_to_page((void *)initrd_start)), |
671 | min_low_pfn); | |
1da177e4 LT |
672 | initrd_start = 0; |
673 | } | |
674 | #endif | |
94b6da5a | 675 | page_cgroup_init(); |
a03c2a48 | 676 | enable_debug_pagealloc(); |
d221938c | 677 | cpu_hotplug_init(); |
b9ce08c0 | 678 | kmemtrace_init(); |
3c7b4e6b | 679 | kmemleak_init(); |
3ac7fe5a | 680 | debug_objects_mem_init(); |
199f0ca5 | 681 | idr_init_cache(); |
e7c8d5c9 | 682 | setup_per_cpu_pageset(); |
1da177e4 LT |
683 | numa_policy_init(); |
684 | if (late_time_init) | |
685 | late_time_init(); | |
686 | calibrate_delay(); | |
687 | pidmap_init(); | |
1da177e4 LT |
688 | anon_vma_init(); |
689 | #ifdef CONFIG_X86 | |
690 | if (efi_enabled) | |
691 | efi_enter_virtual_mode(); | |
692 | #endif | |
8c9843e5 | 693 | thread_info_cache_init(); |
d84f4f99 | 694 | cred_init(); |
1da177e4 LT |
695 | fork_init(num_physpages); |
696 | proc_caches_init(); | |
697 | buffer_init(); | |
1da177e4 LT |
698 | key_init(); |
699 | security_init(); | |
700 | vfs_caches_init(num_physpages); | |
701 | radix_tree_init(); | |
702 | signals_init(); | |
703 | /* rootfs populating might need page-writeback */ | |
704 | page_writeback_init(); | |
705 | #ifdef CONFIG_PROC_FS | |
706 | proc_root_init(); | |
707 | #endif | |
ddbcc7e8 | 708 | cgroup_init(); |
1da177e4 | 709 | cpuset_init(); |
c757249a | 710 | taskstats_init_early(); |
ca74e92b | 711 | delayacct_init(); |
1da177e4 LT |
712 | |
713 | check_bugs(); | |
714 | ||
715 | acpi_early_init(); /* before LAPIC and SMP init */ | |
716 | ||
68bf21aa SR |
717 | ftrace_init(); |
718 | ||
1da177e4 LT |
719 | /* Do the rest non-__init'ed, we're now alive */ |
720 | rest_init(); | |
721 | } | |
722 | ||
22a9d645 | 723 | int initcall_debug; |
d0ea3d7d | 724 | core_param(initcall_debug, initcall_debug, bool, 0644); |
1da177e4 | 725 | |
59f9415f | 726 | int do_one_initcall(initcall_t fn) |
1da177e4 | 727 | { |
1da177e4 | 728 | int count = preempt_count(); |
74239072 | 729 | ktime_t calltime, delta, rettime; |
a76bfd0d | 730 | char msgbuf[64]; |
74239072 FW |
731 | struct boot_trace_call call; |
732 | struct boot_trace_ret ret; | |
1da177e4 | 733 | |
e0df154f | 734 | if (initcall_debug) { |
74239072 FW |
735 | call.caller = task_pid_nr(current); |
736 | printk("calling %pF @ %i\n", fn, call.caller); | |
737 | calltime = ktime_get(); | |
738 | trace_boot_call(&call, fn); | |
71566a0d | 739 | enable_boot_trace(); |
e0df154f | 740 | } |
1da177e4 | 741 | |
74239072 | 742 | ret.result = fn(); |
1da177e4 | 743 | |
e0df154f | 744 | if (initcall_debug) { |
71566a0d | 745 | disable_boot_trace(); |
74239072 FW |
746 | rettime = ktime_get(); |
747 | delta = ktime_sub(rettime, calltime); | |
1d926f27 | 748 | ret.duration = (unsigned long long) ktime_to_ns(delta) >> 10; |
74239072 | 749 | trace_boot_ret(&ret, fn); |
ca538f6b | 750 | printk("initcall %pF returned %d after %Ld usecs\n", fn, |
74239072 | 751 | ret.result, ret.duration); |
e0df154f | 752 | } |
8f0c45cd | 753 | |
e0df154f | 754 | msgbuf[0] = 0; |
e662e1cf | 755 | |
74239072 FW |
756 | if (ret.result && ret.result != -ENODEV && initcall_debug) |
757 | sprintf(msgbuf, "error code %d ", ret.result); | |
e662e1cf | 758 | |
e0df154f | 759 | if (preempt_count() != count) { |
a76bfd0d | 760 | strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf)); |
e0df154f | 761 | preempt_count() = count; |
1da177e4 | 762 | } |
e0df154f | 763 | if (irqs_disabled()) { |
a76bfd0d | 764 | strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); |
e0df154f LT |
765 | local_irq_enable(); |
766 | } | |
767 | if (msgbuf[0]) { | |
96d746c6 | 768 | printk("initcall %pF returned with %s\n", fn, msgbuf); |
e0df154f | 769 | } |
59f9415f | 770 | |
74239072 | 771 | return ret.result; |
e0df154f LT |
772 | } |
773 | ||
774 | ||
c2147a50 | 775 | extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[]; |
e0df154f LT |
776 | |
777 | static void __init do_initcalls(void) | |
778 | { | |
779 | initcall_t *call; | |
780 | ||
c2147a50 | 781 | for (call = __early_initcall_end; call < __initcall_end; call++) |
e0df154f | 782 | do_one_initcall(*call); |
1da177e4 LT |
783 | |
784 | /* Make sure there is no pending stuff from the initcall sequence */ | |
785 | flush_scheduled_work(); | |
786 | } | |
787 | ||
788 | /* | |
789 | * Ok, the machine is now initialized. None of the devices | |
790 | * have been touched yet, but the CPU subsystem is up and | |
791 | * running, and memory and process management works. | |
792 | * | |
793 | * Now we can finally start doing some real work.. | |
794 | */ | |
795 | static void __init do_basic_setup(void) | |
796 | { | |
4446a36f | 797 | rcu_init_sched(); /* needed by module_init stage. */ |
4403b406 | 798 | init_workqueues(); |
759ee091 | 799 | cpuset_init_smp(); |
1da177e4 LT |
800 | usermodehelper_init(); |
801 | driver_init(); | |
b04c3afb | 802 | init_irq_proc(); |
1da177e4 LT |
803 | do_initcalls(); |
804 | } | |
805 | ||
7babe8db | 806 | static void __init do_pre_smp_initcalls(void) |
c2147a50 EGM |
807 | { |
808 | initcall_t *call; | |
809 | ||
810 | for (call = __initcall_start; call < __early_initcall_end; call++) | |
811 | do_one_initcall(*call); | |
812 | } | |
813 | ||
1da177e4 LT |
814 | static void run_init_process(char *init_filename) |
815 | { | |
816 | argv_init[0] = init_filename; | |
67608567 | 817 | kernel_execve(init_filename, argv_init, envp_init); |
1da177e4 LT |
818 | } |
819 | ||
ee5bfa64 VG |
820 | /* This is a non __init function. Force it to be noinline otherwise gcc |
821 | * makes it inline to init() and it becomes part of init.text section | |
822 | */ | |
f99ebf0a | 823 | static noinline int init_post(void) |
acdd052a | 824 | __releases(kernel_lock) |
ee5bfa64 | 825 | { |
22a9d645 AV |
826 | /* need to finish all async __init code before freeing the memory */ |
827 | async_synchronize_full(); | |
ee5bfa64 VG |
828 | free_initmem(); |
829 | unlock_kernel(); | |
830 | mark_rodata_ro(); | |
831 | system_state = SYSTEM_RUNNING; | |
832 | numa_default_policy(); | |
833 | ||
834 | if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) | |
835 | printk(KERN_WARNING "Warning: unable to open an initial console.\n"); | |
836 | ||
837 | (void) sys_dup(0); | |
838 | (void) sys_dup(0); | |
839 | ||
fae5fa44 ON |
840 | current->signal->flags |= SIGNAL_UNKILLABLE; |
841 | ||
ee5bfa64 VG |
842 | if (ramdisk_execute_command) { |
843 | run_init_process(ramdisk_execute_command); | |
844 | printk(KERN_WARNING "Failed to execute %s\n", | |
845 | ramdisk_execute_command); | |
846 | } | |
847 | ||
848 | /* | |
849 | * We try each of these until one succeeds. | |
850 | * | |
851 | * The Bourne shell can be used instead of init if we are | |
852 | * trying to recover a really broken machine. | |
853 | */ | |
854 | if (execute_command) { | |
855 | run_init_process(execute_command); | |
856 | printk(KERN_WARNING "Failed to execute %s. Attempting " | |
857 | "defaults...\n", execute_command); | |
858 | } | |
859 | run_init_process("/sbin/init"); | |
860 | run_init_process("/etc/init"); | |
861 | run_init_process("/bin/init"); | |
862 | run_init_process("/bin/sh"); | |
863 | ||
864 | panic("No init found. Try passing init= option to kernel."); | |
865 | } | |
866 | ||
aae5f662 | 867 | static int __init kernel_init(void * unused) |
1da177e4 LT |
868 | { |
869 | lock_kernel(); | |
58568d2a MX |
870 | |
871 | /* | |
872 | * init can allocate pages on any node | |
873 | */ | |
874 | set_mems_allowed(node_possible_map); | |
1da177e4 LT |
875 | /* |
876 | * init can run on any cpu. | |
877 | */ | |
1a2142af | 878 | set_cpus_allowed_ptr(current, cpu_all_mask); |
1da177e4 LT |
879 | /* |
880 | * Tell the world that we're going to be the grim | |
881 | * reaper of innocent orphaned children. | |
882 | * | |
883 | * We don't want people to have to make incorrect | |
884 | * assumptions about where in the task array this | |
885 | * can be found. | |
886 | */ | |
84d73786 | 887 | init_pid_ns.child_reaper = current; |
1da177e4 | 888 | |
9ec52099 CLG |
889 | cad_pid = task_pid(current); |
890 | ||
ca74a6f8 | 891 | smp_prepare_cpus(setup_max_cpus); |
1da177e4 LT |
892 | |
893 | do_pre_smp_initcalls(); | |
3bf77af6 | 894 | start_boot_trace(); |
1da177e4 | 895 | |
1da177e4 LT |
896 | smp_init(); |
897 | sched_init_smp(); | |
898 | ||
1da177e4 LT |
899 | do_basic_setup(); |
900 | ||
901 | /* | |
902 | * check if there is an early userspace init. If yes, let it do all | |
903 | * the work | |
904 | */ | |
ffdfc409 OJ |
905 | |
906 | if (!ramdisk_execute_command) | |
907 | ramdisk_execute_command = "/init"; | |
908 | ||
909 | if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) { | |
910 | ramdisk_execute_command = NULL; | |
1da177e4 | 911 | prepare_namespace(); |
ffdfc409 | 912 | } |
1da177e4 LT |
913 | |
914 | /* | |
915 | * Ok, we have completed the initial bootup, and | |
916 | * we're essentially up and running. Get rid of the | |
917 | * initmem segments and start the user-mode stuff.. | |
918 | */ | |
71566a0d | 919 | |
ee5bfa64 VG |
920 | init_post(); |
921 | return 0; | |
1da177e4 | 922 | } |