main.c 20 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835
  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. #include <linux/types.h>
  12. #include <linux/module.h>
  13. #include <linux/proc_fs.h>
  14. #include <linux/kernel.h>
  15. #include <linux/syscalls.h>
  16. #include <linux/stackprotector.h>
  17. #include <linux/string.h>
  18. #include <linux/ctype.h>
  19. #include <linux/delay.h>
  20. #include <linux/ioport.h>
  21. #include <linux/init.h>
  22. #include <linux/initrd.h>
  23. #include <linux/bootmem.h>
  24. #include <linux/acpi.h>
  25. #include <linux/tty.h>
  26. #include <linux/percpu.h>
  27. #include <linux/kmod.h>
  28. #include <linux/vmalloc.h>
  29. #include <linux/kernel_stat.h>
  30. #include <linux/start_kernel.h>
  31. #include <linux/security.h>
  32. #include <linux/smp.h>
  33. #include <linux/profile.h>
  34. #include <linux/rcupdate.h>
  35. #include <linux/moduleparam.h>
  36. #include <linux/kallsyms.h>
  37. #include <linux/writeback.h>
  38. #include <linux/cpu.h>
  39. #include <linux/cpuset.h>
  40. #include <linux/cgroup.h>
  41. #include <linux/efi.h>
  42. #include <linux/tick.h>
  43. #include <linux/interrupt.h>
  44. #include <linux/taskstats_kern.h>
  45. #include <linux/delayacct.h>
  46. #include <linux/unistd.h>
  47. #include <linux/rmap.h>
  48. #include <linux/mempolicy.h>
  49. #include <linux/key.h>
  50. #include <linux/buffer_head.h>
  51. #include <linux/page_cgroup.h>
  52. #include <linux/debug_locks.h>
  53. #include <linux/debugobjects.h>
  54. #include <linux/lockdep.h>
  55. #include <linux/kmemleak.h>
  56. #include <linux/pid_namespace.h>
  57. #include <linux/device.h>
  58. #include <linux/kthread.h>
  59. #include <linux/sched.h>
  60. #include <linux/signal.h>
  61. #include <linux/idr.h>
  62. #include <linux/kgdb.h>
  63. #include <linux/ftrace.h>
  64. #include <linux/async.h>
  65. #include <linux/kmemcheck.h>
  66. #include <linux/sfi.h>
  67. #include <linux/shmem_fs.h>
  68. #include <linux/slab.h>
  69. #include <linux/perf_event.h>
  70. #include <asm/io.h>
  71. #include <asm/bugs.h>
  72. #include <asm/setup.h>
  73. #include <asm/sections.h>
  74. #include <asm/cacheflush.h>
  75. #ifdef CONFIG_X86_LOCAL_APIC
  76. #include <asm/smp.h>
  77. #endif
  78. static int kernel_init(void *);
  79. extern void init_IRQ(void);
  80. extern void fork_init(unsigned long);
  81. extern void mca_init(void);
  82. extern void sbus_init(void);
  83. extern void prio_tree_init(void);
  84. extern void radix_tree_init(void);
  85. extern void free_initmem(void);
  86. #ifndef CONFIG_DEBUG_RODATA
  87. static inline void mark_rodata_ro(void) { }
  88. #endif
  89. #ifdef CONFIG_TC
  90. extern void tc_init(void);
  91. #endif
  92. /*
  93. * Debug helper: via this flag we know that we are in 'early bootup code'
  94. * where only the boot processor is running with IRQ disabled. This means
  95. * two things - IRQ must not be enabled before the flag is cleared and some
  96. * operations which are not allowed with IRQ disabled are allowed while the
  97. * flag is set.
  98. */
  99. bool early_boot_irqs_disabled __read_mostly;
  100. enum system_states system_state __read_mostly;
  101. EXPORT_SYMBOL(system_state);
  102. /*
  103. * Boot command-line arguments
  104. */
  105. #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
  106. #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
  107. extern void time_init(void);
  108. /* Default late time init is NULL. archs can override this later. */
  109. void (*__initdata late_time_init)(void);
  110. extern void softirq_init(void);
  111. /* Untouched command line saved by arch-specific code. */
  112. char __initdata boot_command_line[COMMAND_LINE_SIZE];
  113. /* Untouched saved command line (eg. for /proc) */
  114. char *saved_command_line;
  115. /* Command line for parameter parsing */
  116. static char *static_command_line;
  117. static char *execute_command;
  118. static char *ramdisk_execute_command;
  119. /*
  120. * If set, this is an indication to the drivers that reset the underlying
  121. * device before going ahead with the initialization otherwise driver might
  122. * rely on the BIOS and skip the reset operation.
  123. *
  124. * This is useful if kernel is booting in an unreliable environment.
  125. * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
  126. * skipped and devices will be in unknown state.
  127. */
  128. unsigned int reset_devices;
  129. EXPORT_SYMBOL(reset_devices);
  130. static int __init set_reset_devices(char *str)
  131. {
  132. reset_devices = 1;
  133. return 1;
  134. }
  135. __setup("reset_devices", set_reset_devices);
  136. static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
  137. const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
  138. static const char *panic_later, *panic_param;
  139. extern const struct obs_kernel_param __setup_start[], __setup_end[];
  140. static int __init obsolete_checksetup(char *line)
  141. {
  142. const struct obs_kernel_param *p;
  143. int had_early_param = 0;
  144. p = __setup_start;
  145. do {
  146. int n = strlen(p->str);
  147. if (!strncmp(line, p->str, n)) {
  148. if (p->early) {
  149. /* Already done in parse_early_param?
  150. * (Needs exact match on param part).
  151. * Keep iterating, as we can have early
  152. * params and __setups of same names 8( */
  153. if (line[n] == '\0' || line[n] == '=')
  154. had_early_param = 1;
  155. } else if (!p->setup_func) {
  156. printk(KERN_WARNING "Parameter %s is obsolete,"
  157. " ignored\n", p->str);
  158. return 1;
  159. } else if (p->setup_func(line + n))
  160. return 1;
  161. }
  162. p++;
  163. } while (p < __setup_end);
  164. return had_early_param;
  165. }
  166. /*
  167. * This should be approx 2 Bo*oMips to start (note initial shift), and will
  168. * still work even if initially too large, it will just take slightly longer
  169. */
  170. unsigned long loops_per_jiffy = (1<<12);
  171. EXPORT_SYMBOL(loops_per_jiffy);
  172. static int __init debug_kernel(char *str)
  173. {
  174. console_loglevel = 10;
  175. return 0;
  176. }
  177. static int __init quiet_kernel(char *str)
  178. {
  179. console_loglevel = 4;
  180. return 0;
  181. }
  182. early_param("debug", debug_kernel);
  183. early_param("quiet", quiet_kernel);
  184. static int __init loglevel(char *str)
  185. {
  186. get_option(&str, &console_loglevel);
  187. return 0;
  188. }
  189. early_param("loglevel", loglevel);
  190. /*
  191. * Unknown boot options get handed to init, unless they look like
  192. * unused parameters (modprobe will find them in /proc/cmdline).
  193. */
  194. static int __init unknown_bootoption(char *param, char *val)
  195. {
  196. /* Change NUL term back to "=", to make "param" the whole string. */
  197. if (val) {
  198. /* param=val or param="val"? */
  199. if (val == param+strlen(param)+1)
  200. val[-1] = '=';
  201. else if (val == param+strlen(param)+2) {
  202. val[-2] = '=';
  203. memmove(val-1, val, strlen(val)+1);
  204. val--;
  205. } else
  206. BUG();
  207. }
  208. /* Handle obsolete-style parameters */
  209. if (obsolete_checksetup(param))
  210. return 0;
  211. /* Unused module parameter. */
  212. if (strchr(param, '.') && (!val || strchr(param, '.') < val))
  213. return 0;
  214. if (panic_later)
  215. return 0;
  216. if (val) {
  217. /* Environment option */
  218. unsigned int i;
  219. for (i = 0; envp_init[i]; i++) {
  220. if (i == MAX_INIT_ENVS) {
  221. panic_later = "Too many boot env vars at `%s'";
  222. panic_param = param;
  223. }
  224. if (!strncmp(param, envp_init[i], val - param))
  225. break;
  226. }
  227. envp_init[i] = param;
  228. } else {
  229. /* Command line option */
  230. unsigned int i;
  231. for (i = 0; argv_init[i]; i++) {
  232. if (i == MAX_INIT_ARGS) {
  233. panic_later = "Too many boot init vars at `%s'";
  234. panic_param = param;
  235. }
  236. }
  237. argv_init[i] = param;
  238. }
  239. return 0;
  240. }
  241. #ifdef CONFIG_DEBUG_PAGEALLOC
  242. int __read_mostly debug_pagealloc_enabled = 0;
  243. #endif
  244. static int __init init_setup(char *str)
  245. {
  246. unsigned int i;
  247. execute_command = str;
  248. /*
  249. * In case LILO is going to boot us with default command line,
  250. * it prepends "auto" before the whole cmdline which makes
  251. * the shell think it should execute a script with such name.
  252. * So we ignore all arguments entered _before_ init=... [MJ]
  253. */
  254. for (i = 1; i < MAX_INIT_ARGS; i++)
  255. argv_init[i] = NULL;
  256. return 1;
  257. }
  258. __setup("init=", init_setup);
  259. static int __init rdinit_setup(char *str)
  260. {
  261. unsigned int i;
  262. ramdisk_execute_command = str;
  263. /* See "auto" comment in init_setup */
  264. for (i = 1; i < MAX_INIT_ARGS; i++)
  265. argv_init[i] = NULL;
  266. return 1;
  267. }
  268. __setup("rdinit=", rdinit_setup);
  269. #ifndef CONFIG_SMP
  270. static const unsigned int setup_max_cpus = NR_CPUS;
  271. #ifdef CONFIG_X86_LOCAL_APIC
  272. static void __init smp_init(void)
  273. {
  274. APIC_init_uniprocessor();
  275. }
  276. #else
  277. #define smp_init() do { } while (0)
  278. #endif
  279. static inline void setup_nr_cpu_ids(void) { }
  280. static inline void smp_prepare_cpus(unsigned int maxcpus) { }
  281. #endif
  282. /*
  283. * We need to store the untouched command line for future reference.
  284. * We also need to store the touched command line since the parameter
  285. * parsing is performed in place, and we should allow a component to
  286. * store reference of name/value for future reference.
  287. */
  288. static void __init setup_command_line(char *command_line)
  289. {
  290. saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
  291. static_command_line = alloc_bootmem(strlen (command_line)+1);
  292. strcpy (saved_command_line, boot_command_line);
  293. strcpy (static_command_line, command_line);
  294. }
  295. /*
  296. * We need to finalize in a non-__init function or else race conditions
  297. * between the root thread and the init thread may cause start_kernel to
  298. * be reaped by free_initmem before the root thread has proceeded to
  299. * cpu_idle.
  300. *
  301. * gcc-3.4 accidentally inlines this function, so use noinline.
  302. */
  303. static __initdata DECLARE_COMPLETION(kthreadd_done);
  304. static noinline void __init_refok rest_init(void)
  305. {
  306. int pid;
  307. rcu_scheduler_starting();
  308. /*
  309. * We need to spawn init first so that it obtains pid 1, however
  310. * the init task will end up wanting to create kthreads, which, if
  311. * we schedule it before we create kthreadd, will OOPS.
  312. */
  313. kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
  314. numa_default_policy();
  315. pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
  316. rcu_read_lock();
  317. kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
  318. rcu_read_unlock();
  319. complete(&kthreadd_done);
  320. /*
  321. * The boot idle thread must execute schedule()
  322. * at least once to get things moving:
  323. */
  324. init_idle_bootup_task(current);
  325. preempt_enable_no_resched();
  326. schedule();
  327. preempt_disable();
  328. /* Call into cpu_idle with preempt disabled */
  329. cpu_idle();
  330. }
  331. /* Check for early params. */
  332. static int __init do_early_param(char *param, char *val)
  333. {
  334. const struct obs_kernel_param *p;
  335. for (p = __setup_start; p < __setup_end; p++) {
  336. if ((p->early && strcmp(param, p->str) == 0) ||
  337. (strcmp(param, "console") == 0 &&
  338. strcmp(p->str, "earlycon") == 0)
  339. ) {
  340. if (p->setup_func(val) != 0)
  341. printk(KERN_WARNING
  342. "Malformed early option '%s'\n", param);
  343. }
  344. }
  345. /* We accept everything at this stage. */
  346. return 0;
  347. }
  348. void __init parse_early_options(char *cmdline)
  349. {
  350. parse_args("early options", cmdline, NULL, 0, do_early_param);
  351. }
  352. /* Arch code calls this early on, or if not, just before other parsing. */
  353. void __init parse_early_param(void)
  354. {
  355. static __initdata int done = 0;
  356. static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
  357. if (done)
  358. return;
  359. /* All fall through to do_early_param. */
  360. strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
  361. parse_early_options(tmp_cmdline);
  362. done = 1;
  363. }
  364. /*
  365. * Activate the first processor.
  366. */
  367. static void __init boot_cpu_init(void)
  368. {
  369. int cpu = smp_processor_id();
  370. /* Mark the boot cpu "present", "online" etc for SMP and UP case */
  371. set_cpu_online(cpu, true);
  372. set_cpu_active(cpu, true);
  373. set_cpu_present(cpu, true);
  374. set_cpu_possible(cpu, true);
  375. }
  376. void __init __weak smp_setup_processor_id(void)
  377. {
  378. }
  379. void __init __weak thread_info_cache_init(void)
  380. {
  381. }
  382. /*
  383. * Set up kernel memory allocators
  384. */
  385. static void __init mm_init(void)
  386. {
  387. /*
  388. * page_cgroup requires countinous pages as memmap
  389. * and it's bigger than MAX_ORDER unless SPARSEMEM.
  390. */
  391. page_cgroup_init_flatmem();
  392. mem_init();
  393. kmem_cache_init();
  394. percpu_init_late();
  395. pgtable_cache_init();
  396. vmalloc_init();
  397. }
  398. asmlinkage void __init start_kernel(void)
  399. {
  400. char * command_line;
  401. extern const struct kernel_param __start___param[], __stop___param[];
  402. smp_setup_processor_id();
  403. /*
  404. * Need to run as early as possible, to initialize the
  405. * lockdep hash:
  406. */
  407. lockdep_init();
  408. debug_objects_early_init();
  409. /*
  410. * Set up the the initial canary ASAP:
  411. */
  412. boot_init_stack_canary();
  413. cgroup_init_early();
  414. local_irq_disable();
  415. early_boot_irqs_disabled = true;
  416. /*
  417. * Interrupts are still disabled. Do necessary setups, then
  418. * enable them
  419. */
  420. tick_init();
  421. boot_cpu_init();
  422. page_address_init();
  423. printk(KERN_NOTICE "%s", linux_banner);
  424. setup_arch(&command_line);
  425. mm_init_owner(&init_mm, &init_task);
  426. mm_init_cpumask(&init_mm);
  427. setup_command_line(command_line);
  428. setup_nr_cpu_ids();
  429. setup_per_cpu_areas();
  430. smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
  431. build_all_zonelists(NULL);
  432. page_alloc_init();
  433. printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
  434. parse_early_param();
  435. parse_args("Booting kernel", static_command_line, __start___param,
  436. __stop___param - __start___param,
  437. &unknown_bootoption);
  438. /*
  439. * These use large bootmem allocations and must precede
  440. * kmem_cache_init()
  441. */
  442. setup_log_buf(0);
  443. pidhash_init();
  444. vfs_caches_init_early();
  445. sort_main_extable();
  446. trap_init();
  447. mm_init();
  448. /*
  449. * Set up the scheduler prior starting any interrupts (such as the
  450. * timer interrupt). Full topology setup happens at smp_init()
  451. * time - but meanwhile we still have a functioning scheduler.
  452. */
  453. sched_init();
  454. /*
  455. * Disable preemption - early bootup scheduling is extremely
  456. * fragile until we cpu_idle() for the first time.
  457. */
  458. preempt_disable();
  459. if (!irqs_disabled()) {
  460. printk(KERN_WARNING "start_kernel(): bug: interrupts were "
  461. "enabled *very* early, fixing it\n");
  462. local_irq_disable();
  463. }
  464. idr_init_cache();
  465. perf_event_init();
  466. rcu_init();
  467. radix_tree_init();
  468. /* init some links before init_ISA_irqs() */
  469. early_irq_init();
  470. init_IRQ();
  471. prio_tree_init();
  472. init_timers();
  473. hrtimers_init();
  474. softirq_init();
  475. timekeeping_init();
  476. time_init();
  477. profile_init();
  478. call_function_init();
  479. if (!irqs_disabled())
  480. printk(KERN_CRIT "start_kernel(): bug: interrupts were "
  481. "enabled early\n");
  482. early_boot_irqs_disabled = false;
  483. local_irq_enable();
  484. /* Interrupts are enabled now so all GFP allocations are safe. */
  485. gfp_allowed_mask = __GFP_BITS_MASK;
  486. kmem_cache_init_late();
  487. /*
  488. * HACK ALERT! This is early. We're enabling the console before
  489. * we've done PCI setups etc, and console_init() must be aware of
  490. * this. But we do want output early, in case something goes wrong.
  491. */
  492. console_init();
  493. if (panic_later)
  494. panic(panic_later, panic_param);
  495. lockdep_info();
  496. /*
  497. * Need to run this when irqs are enabled, because it wants
  498. * to self-test [hard/soft]-irqs on/off lock inversion bugs
  499. * too:
  500. */
  501. locking_selftest();
  502. #ifdef CONFIG_BLK_DEV_INITRD
  503. if (initrd_start && !initrd_below_start_ok &&
  504. page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
  505. printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
  506. "disabling it.\n",
  507. page_to_pfn(virt_to_page((void *)initrd_start)),
  508. min_low_pfn);
  509. initrd_start = 0;
  510. }
  511. #endif
  512. page_cgroup_init();
  513. enable_debug_pagealloc();
  514. debug_objects_mem_init();
  515. kmemleak_init();
  516. setup_per_cpu_pageset();
  517. numa_policy_init();
  518. if (late_time_init)
  519. late_time_init();
  520. sched_clock_init();
  521. calibrate_delay();
  522. pidmap_init();
  523. anon_vma_init();
  524. #ifdef CONFIG_X86
  525. if (efi_enabled)
  526. efi_enter_virtual_mode();
  527. #endif
  528. thread_info_cache_init();
  529. cred_init();
  530. fork_init(totalram_pages);
  531. proc_caches_init();
  532. buffer_init();
  533. key_init();
  534. security_init();
  535. dbg_late_init();
  536. vfs_caches_init(totalram_pages);
  537. signals_init();
  538. /* rootfs populating might need page-writeback */
  539. page_writeback_init();
  540. #ifdef CONFIG_PROC_FS
  541. proc_root_init();
  542. #endif
  543. cgroup_init();
  544. cpuset_init();
  545. taskstats_init_early();
  546. delayacct_init();
  547. check_bugs();
  548. acpi_early_init(); /* before LAPIC and SMP init */
  549. sfi_init_late();
  550. ftrace_init();
  551. /* Do the rest non-__init'ed, we're now alive */
  552. rest_init();
  553. }
  554. /* Call all constructor functions linked into the kernel. */
  555. static void __init do_ctors(void)
  556. {
  557. #ifdef CONFIG_CONSTRUCTORS
  558. ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
  559. for (; fn < (ctor_fn_t *) __ctors_end; fn++)
  560. (*fn)();
  561. #endif
  562. }
  563. int initcall_debug;
  564. core_param(initcall_debug, initcall_debug, bool, 0644);
  565. static char msgbuf[64];
  566. static int __init_or_module do_one_initcall_debug(initcall_t fn)
  567. {
  568. ktime_t calltime, delta, rettime;
  569. unsigned long long duration;
  570. int ret;
  571. printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current));
  572. calltime = ktime_get();
  573. ret = fn();
  574. rettime = ktime_get();
  575. delta = ktime_sub(rettime, calltime);
  576. duration = (unsigned long long) ktime_to_ns(delta) >> 10;
  577. printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", fn,
  578. ret, duration);
  579. return ret;
  580. }
  581. int __init_or_module do_one_initcall(initcall_t fn)
  582. {
  583. int count = preempt_count();
  584. int ret;
  585. if (initcall_debug)
  586. ret = do_one_initcall_debug(fn);
  587. else
  588. ret = fn();
  589. msgbuf[0] = 0;
  590. if (ret && ret != -ENODEV && initcall_debug)
  591. sprintf(msgbuf, "error code %d ", ret);
  592. if (preempt_count() != count) {
  593. strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
  594. preempt_count() = count;
  595. }
  596. if (irqs_disabled()) {
  597. strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
  598. local_irq_enable();
  599. }
  600. if (msgbuf[0]) {
  601. printk("initcall %pF returned with %s\n", fn, msgbuf);
  602. }
  603. return ret;
  604. }
  605. extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
  606. static void __init do_initcalls(void)
  607. {
  608. initcall_t *fn;
  609. for (fn = __early_initcall_end; fn < __initcall_end; fn++)
  610. do_one_initcall(*fn);
  611. }
  612. /*
  613. * Ok, the machine is now initialized. None of the devices
  614. * have been touched yet, but the CPU subsystem is up and
  615. * running, and memory and process management works.
  616. *
  617. * Now we can finally start doing some real work..
  618. */
  619. static void __init do_basic_setup(void)
  620. {
  621. cpuset_init_smp();
  622. usermodehelper_init();
  623. init_tmpfs();
  624. driver_init();
  625. init_irq_proc();
  626. do_ctors();
  627. do_initcalls();
  628. }
  629. static void __init do_pre_smp_initcalls(void)
  630. {
  631. initcall_t *fn;
  632. for (fn = __initcall_start; fn < __early_initcall_end; fn++)
  633. do_one_initcall(*fn);
  634. }
  635. static void run_init_process(const char *init_filename)
  636. {
  637. argv_init[0] = init_filename;
  638. kernel_execve(init_filename, argv_init, envp_init);
  639. }
  640. /* This is a non __init function. Force it to be noinline otherwise gcc
  641. * makes it inline to init() and it becomes part of init.text section
  642. */
  643. static noinline int init_post(void)
  644. {
  645. /* need to finish all async __init code before freeing the memory */
  646. async_synchronize_full();
  647. free_initmem();
  648. mark_rodata_ro();
  649. system_state = SYSTEM_RUNNING;
  650. numa_default_policy();
  651. current->signal->flags |= SIGNAL_UNKILLABLE;
  652. if (ramdisk_execute_command) {
  653. run_init_process(ramdisk_execute_command);
  654. printk(KERN_WARNING "Failed to execute %s\n",
  655. ramdisk_execute_command);
  656. }
  657. /*
  658. * We try each of these until one succeeds.
  659. *
  660. * The Bourne shell can be used instead of init if we are
  661. * trying to recover a really broken machine.
  662. */
  663. if (execute_command) {
  664. run_init_process(execute_command);
  665. printk(KERN_WARNING "Failed to execute %s. Attempting "
  666. "defaults...\n", execute_command);
  667. }
  668. run_init_process("/sbin/init");
  669. run_init_process("/etc/init");
  670. run_init_process("/bin/init");
  671. run_init_process("/bin/sh");
  672. panic("No init found. Try passing init= option to kernel. "
  673. "See Linux Documentation/init.txt for guidance.");
  674. }
  675. static int __init kernel_init(void * unused)
  676. {
  677. /*
  678. * Wait until kthreadd is all set-up.
  679. */
  680. wait_for_completion(&kthreadd_done);
  681. /*
  682. * init can allocate pages on any node
  683. */
  684. set_mems_allowed(node_states[N_HIGH_MEMORY]);
  685. /*
  686. * init can run on any cpu.
  687. */
  688. set_cpus_allowed_ptr(current, cpu_all_mask);
  689. cad_pid = task_pid(current);
  690. smp_prepare_cpus(setup_max_cpus);
  691. do_pre_smp_initcalls();
  692. lockup_detector_init();
  693. smp_init();
  694. sched_init_smp();
  695. do_basic_setup();
  696. /* Open the /dev/console on the rootfs, this should never fail */
  697. if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
  698. printk(KERN_WARNING "Warning: unable to open an initial console.\n");
  699. (void) sys_dup(0);
  700. (void) sys_dup(0);
  701. /*
  702. * check if there is an early userspace init. If yes, let it do all
  703. * the work
  704. */
  705. if (!ramdisk_execute_command)
  706. ramdisk_execute_command = "/init";
  707. if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
  708. ramdisk_execute_command = NULL;
  709. prepare_namespace();
  710. }
  711. /*
  712. * Ok, we have completed the initial bootup, and
  713. * we're essentially up and running. Get rid of the
  714. * initmem segments and start the user-mode stuff..
  715. */
  716. init_post();
  717. return 0;
  718. }