main.c 22 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921
  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. #define DEBUG /* Enable initcall_debug */
  12. #include <linux/types.h>
  13. #include <linux/module.h>
  14. #include <linux/proc_fs.h>
  15. #include <linux/kernel.h>
  16. #include <linux/syscalls.h>
  17. #include <linux/stackprotector.h>
  18. #include <linux/string.h>
  19. #include <linux/ctype.h>
  20. #include <linux/delay.h>
  21. #include <linux/ioport.h>
  22. #include <linux/init.h>
  23. #include <linux/initrd.h>
  24. #include <linux/bootmem.h>
  25. #include <linux/acpi.h>
  26. #include <linux/tty.h>
  27. #include <linux/percpu.h>
  28. #include <linux/kmod.h>
  29. #include <linux/vmalloc.h>
  30. #include <linux/kernel_stat.h>
  31. #include <linux/start_kernel.h>
  32. #include <linux/security.h>
  33. #include <linux/smp.h>
  34. #include <linux/profile.h>
  35. #include <linux/rcupdate.h>
  36. #include <linux/moduleparam.h>
  37. #include <linux/kallsyms.h>
  38. #include <linux/writeback.h>
  39. #include <linux/cpu.h>
  40. #include <linux/cpuset.h>
  41. #include <linux/cgroup.h>
  42. #include <linux/efi.h>
  43. #include <linux/tick.h>
  44. #include <linux/interrupt.h>
  45. #include <linux/taskstats_kern.h>
  46. #include <linux/delayacct.h>
  47. #include <linux/unistd.h>
  48. #include <linux/rmap.h>
  49. #include <linux/mempolicy.h>
  50. #include <linux/key.h>
  51. #include <linux/buffer_head.h>
  52. #include <linux/page_cgroup.h>
  53. #include <linux/debug_locks.h>
  54. #include <linux/debugobjects.h>
  55. #include <linux/lockdep.h>
  56. #include <linux/kmemleak.h>
  57. #include <linux/pid_namespace.h>
  58. #include <linux/device.h>
  59. #include <linux/kthread.h>
  60. #include <linux/sched.h>
  61. #include <linux/signal.h>
  62. #include <linux/idr.h>
  63. #include <linux/kgdb.h>
  64. #include <linux/ftrace.h>
  65. #include <linux/async.h>
  66. #include <linux/kmemcheck.h>
  67. #include <linux/sfi.h>
  68. #include <linux/shmem_fs.h>
  69. #include <linux/slab.h>
  70. #include <linux/perf_event.h>
  71. #include <linux/file.h>
  72. #include <linux/ptrace.h>
  73. #include <linux/blkdev.h>
  74. #include <linux/elevator.h>
  75. #include <linux/sched_clock.h>
  76. #include <linux/context_tracking.h>
  77. #include <linux/random.h>
  78. #include <asm/io.h>
  79. #include <asm/bugs.h>
  80. #include <asm/setup.h>
  81. #include <asm/sections.h>
  82. #include <asm/cacheflush.h>
  83. #ifdef CONFIG_X86_LOCAL_APIC
  84. #include <asm/smp.h>
  85. #endif
  86. static int kernel_init(void *);
  87. extern void init_IRQ(void);
  88. extern void fork_init(unsigned long);
  89. extern void mca_init(void);
  90. extern void sbus_init(void);
  91. extern void radix_tree_init(void);
  92. #ifndef CONFIG_DEBUG_RODATA
  93. static inline void mark_rodata_ro(void) { }
  94. #endif
  95. #ifdef CONFIG_TC
  96. extern void tc_init(void);
  97. #endif
  98. /*
  99. * Debug helper: via this flag we know that we are in 'early bootup code'
  100. * where only the boot processor is running with IRQ disabled. This means
  101. * two things - IRQ must not be enabled before the flag is cleared and some
  102. * operations which are not allowed with IRQ disabled are allowed while the
  103. * flag is set.
  104. */
  105. bool early_boot_irqs_disabled __read_mostly;
  106. enum system_states system_state __read_mostly;
  107. EXPORT_SYMBOL(system_state);
  108. /*
  109. * Boot command-line arguments
  110. */
  111. #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
  112. #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
  113. extern void time_init(void);
  114. /* Default late time init is NULL. archs can override this later. */
  115. void (*__initdata late_time_init)(void);
  116. extern void softirq_init(void);
  117. /* Untouched command line saved by arch-specific code. */
  118. char __initdata boot_command_line[COMMAND_LINE_SIZE];
  119. /* Untouched saved command line (eg. for /proc) */
  120. char *saved_command_line;
  121. /* Command line for parameter parsing */
  122. static char *static_command_line;
  123. static char *execute_command;
  124. static char *ramdisk_execute_command;
  125. /*
  126. * Used to generate warnings if static_key manipulation functions are used
  127. * before jump_label_init is called.
  128. */
  129. bool static_key_initialized __read_mostly = false;
  130. EXPORT_SYMBOL_GPL(static_key_initialized);
  131. /*
  132. * If set, this is an indication to the drivers that reset the underlying
  133. * device before going ahead with the initialization otherwise driver might
  134. * rely on the BIOS and skip the reset operation.
  135. *
  136. * This is useful if kernel is booting in an unreliable environment.
  137. * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
  138. * skipped and devices will be in unknown state.
  139. */
  140. unsigned int reset_devices;
  141. EXPORT_SYMBOL(reset_devices);
  142. static int __init set_reset_devices(char *str)
  143. {
  144. reset_devices = 1;
  145. return 1;
  146. }
  147. __setup("reset_devices", set_reset_devices);
  148. static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
  149. const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
  150. static const char *panic_later, *panic_param;
  151. extern const struct obs_kernel_param __setup_start[], __setup_end[];
  152. static int __init obsolete_checksetup(char *line)
  153. {
  154. const struct obs_kernel_param *p;
  155. int had_early_param = 0;
  156. p = __setup_start;
  157. do {
  158. int n = strlen(p->str);
  159. if (parameqn(line, p->str, n)) {
  160. if (p->early) {
  161. /* Already done in parse_early_param?
  162. * (Needs exact match on param part).
  163. * Keep iterating, as we can have early
  164. * params and __setups of same names 8( */
  165. if (line[n] == '\0' || line[n] == '=')
  166. had_early_param = 1;
  167. } else if (!p->setup_func) {
  168. pr_warn("Parameter %s is obsolete, ignored\n",
  169. p->str);
  170. return 1;
  171. } else if (p->setup_func(line + n))
  172. return 1;
  173. }
  174. p++;
  175. } while (p < __setup_end);
  176. return had_early_param;
  177. }
  178. /*
  179. * This should be approx 2 Bo*oMips to start (note initial shift), and will
  180. * still work even if initially too large, it will just take slightly longer
  181. */
  182. unsigned long loops_per_jiffy = (1<<12);
  183. EXPORT_SYMBOL(loops_per_jiffy);
  184. static int __init debug_kernel(char *str)
  185. {
  186. console_loglevel = 10;
  187. return 0;
  188. }
  189. static int __init quiet_kernel(char *str)
  190. {
  191. console_loglevel = 4;
  192. return 0;
  193. }
  194. early_param("debug", debug_kernel);
  195. early_param("quiet", quiet_kernel);
  196. static int __init loglevel(char *str)
  197. {
  198. int newlevel;
  199. /*
  200. * Only update loglevel value when a correct setting was passed,
  201. * to prevent blind crashes (when loglevel being set to 0) that
  202. * are quite hard to debug
  203. */
  204. if (get_option(&str, &newlevel)) {
  205. console_loglevel = newlevel;
  206. return 0;
  207. }
  208. return -EINVAL;
  209. }
  210. early_param("loglevel", loglevel);
  211. /* Change NUL term back to "=", to make "param" the whole string. */
  212. static int __init repair_env_string(char *param, char *val, const char *unused)
  213. {
  214. if (val) {
  215. /* param=val or param="val"? */
  216. if (val == param+strlen(param)+1)
  217. val[-1] = '=';
  218. else if (val == param+strlen(param)+2) {
  219. val[-2] = '=';
  220. memmove(val-1, val, strlen(val)+1);
  221. val--;
  222. } else
  223. BUG();
  224. }
  225. return 0;
  226. }
  227. /*
  228. * Unknown boot options get handed to init, unless they look like
  229. * unused parameters (modprobe will find them in /proc/cmdline).
  230. */
  231. static int __init unknown_bootoption(char *param, char *val, const char *unused)
  232. {
  233. repair_env_string(param, val, unused);
  234. /* Handle obsolete-style parameters */
  235. if (obsolete_checksetup(param))
  236. return 0;
  237. /* Unused module parameter. */
  238. if (strchr(param, '.') && (!val || strchr(param, '.') < val))
  239. return 0;
  240. if (panic_later)
  241. return 0;
  242. if (val) {
  243. /* Environment option */
  244. unsigned int i;
  245. for (i = 0; envp_init[i]; i++) {
  246. if (i == MAX_INIT_ENVS) {
  247. panic_later = "Too many boot env vars at `%s'";
  248. panic_param = param;
  249. }
  250. if (!strncmp(param, envp_init[i], val - param))
  251. break;
  252. }
  253. envp_init[i] = param;
  254. } else {
  255. /* Command line option */
  256. unsigned int i;
  257. for (i = 0; argv_init[i]; i++) {
  258. if (i == MAX_INIT_ARGS) {
  259. panic_later = "Too many boot init vars at `%s'";
  260. panic_param = param;
  261. }
  262. }
  263. argv_init[i] = param;
  264. }
  265. return 0;
  266. }
  267. static int __init init_setup(char *str)
  268. {
  269. unsigned int i;
  270. execute_command = str;
  271. /*
  272. * In case LILO is going to boot us with default command line,
  273. * it prepends "auto" before the whole cmdline which makes
  274. * the shell think it should execute a script with such name.
  275. * So we ignore all arguments entered _before_ init=... [MJ]
  276. */
  277. for (i = 1; i < MAX_INIT_ARGS; i++)
  278. argv_init[i] = NULL;
  279. return 1;
  280. }
  281. __setup("init=", init_setup);
  282. static int __init rdinit_setup(char *str)
  283. {
  284. unsigned int i;
  285. ramdisk_execute_command = str;
  286. /* See "auto" comment in init_setup */
  287. for (i = 1; i < MAX_INIT_ARGS; i++)
  288. argv_init[i] = NULL;
  289. return 1;
  290. }
  291. __setup("rdinit=", rdinit_setup);
  292. #ifndef CONFIG_SMP
  293. static const unsigned int setup_max_cpus = NR_CPUS;
  294. #ifdef CONFIG_X86_LOCAL_APIC
  295. static void __init smp_init(void)
  296. {
  297. APIC_init_uniprocessor();
  298. }
  299. #else
  300. #define smp_init() do { } while (0)
  301. #endif
  302. static inline void setup_nr_cpu_ids(void) { }
  303. static inline void smp_prepare_cpus(unsigned int maxcpus) { }
  304. #endif
  305. /*
  306. * We need to store the untouched command line for future reference.
  307. * We also need to store the touched command line since the parameter
  308. * parsing is performed in place, and we should allow a component to
  309. * store reference of name/value for future reference.
  310. */
  311. static void __init setup_command_line(char *command_line)
  312. {
  313. saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
  314. static_command_line = alloc_bootmem(strlen (command_line)+1);
  315. strcpy (saved_command_line, boot_command_line);
  316. strcpy (static_command_line, command_line);
  317. }
  318. /*
  319. * We need to finalize in a non-__init function or else race conditions
  320. * between the root thread and the init thread may cause start_kernel to
  321. * be reaped by free_initmem before the root thread has proceeded to
  322. * cpu_idle.
  323. *
  324. * gcc-3.4 accidentally inlines this function, so use noinline.
  325. */
  326. static __initdata DECLARE_COMPLETION(kthreadd_done);
  327. static noinline void __init_refok rest_init(void)
  328. {
  329. int pid;
  330. rcu_scheduler_starting();
  331. /*
  332. * We need to spawn init first so that it obtains pid 1, however
  333. * the init task will end up wanting to create kthreads, which, if
  334. * we schedule it before we create kthreadd, will OOPS.
  335. */
  336. kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
  337. numa_default_policy();
  338. pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
  339. rcu_read_lock();
  340. kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
  341. rcu_read_unlock();
  342. complete(&kthreadd_done);
  343. /*
  344. * The boot idle thread must execute schedule()
  345. * at least once to get things moving:
  346. */
  347. init_idle_bootup_task(current);
  348. schedule_preempt_disabled();
  349. /* Call into cpu_idle with preempt disabled */
  350. cpu_startup_entry(CPUHP_ONLINE);
  351. }
  352. /* Check for early params. */
  353. static int __init do_early_param(char *param, char *val, const char *unused)
  354. {
  355. const struct obs_kernel_param *p;
  356. for (p = __setup_start; p < __setup_end; p++) {
  357. if ((p->early && parameq(param, p->str)) ||
  358. (strcmp(param, "console") == 0 &&
  359. strcmp(p->str, "earlycon") == 0)
  360. ) {
  361. if (p->setup_func(val) != 0)
  362. pr_warn("Malformed early option '%s'\n", param);
  363. }
  364. }
  365. /* We accept everything at this stage. */
  366. return 0;
  367. }
  368. void __init parse_early_options(char *cmdline)
  369. {
  370. parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
  371. }
  372. /* Arch code calls this early on, or if not, just before other parsing. */
  373. void __init parse_early_param(void)
  374. {
  375. static __initdata int done = 0;
  376. static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
  377. if (done)
  378. return;
  379. /* All fall through to do_early_param. */
  380. strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
  381. parse_early_options(tmp_cmdline);
  382. done = 1;
  383. }
  384. /*
  385. * Activate the first processor.
  386. */
  387. static void __init boot_cpu_init(void)
  388. {
  389. int cpu = smp_processor_id();
  390. /* Mark the boot cpu "present", "online" etc for SMP and UP case */
  391. set_cpu_online(cpu, true);
  392. set_cpu_active(cpu, true);
  393. set_cpu_present(cpu, true);
  394. set_cpu_possible(cpu, true);
  395. }
  396. void __init __weak smp_setup_processor_id(void)
  397. {
  398. }
  399. # if THREAD_SIZE >= PAGE_SIZE
  400. void __init __weak thread_info_cache_init(void)
  401. {
  402. }
  403. #endif
  404. /*
  405. * Set up kernel memory allocators
  406. */
  407. static void __init mm_init(void)
  408. {
  409. /*
  410. * page_cgroup requires contiguous pages,
  411. * bigger than MAX_ORDER unless SPARSEMEM.
  412. */
  413. page_cgroup_init_flatmem();
  414. mem_init();
  415. kmem_cache_init();
  416. percpu_init_late();
  417. pgtable_cache_init();
  418. vmalloc_init();
  419. }
  420. asmlinkage void __init start_kernel(void)
  421. {
  422. char * command_line;
  423. extern const struct kernel_param __start___param[], __stop___param[];
  424. /*
  425. * Need to run as early as possible, to initialize the
  426. * lockdep hash:
  427. */
  428. lockdep_init();
  429. smp_setup_processor_id();
  430. debug_objects_early_init();
  431. /*
  432. * Set up the the initial canary ASAP:
  433. */
  434. boot_init_stack_canary();
  435. cgroup_init_early();
  436. local_irq_disable();
  437. early_boot_irqs_disabled = true;
  438. /*
  439. * Interrupts are still disabled. Do necessary setups, then
  440. * enable them
  441. */
  442. boot_cpu_init();
  443. page_address_init();
  444. pr_notice("%s", linux_banner);
  445. setup_arch(&command_line);
  446. mm_init_owner(&init_mm, &init_task);
  447. mm_init_cpumask(&init_mm);
  448. setup_command_line(command_line);
  449. setup_nr_cpu_ids();
  450. setup_per_cpu_areas();
  451. smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
  452. build_all_zonelists(NULL, NULL);
  453. page_alloc_init();
  454. pr_notice("Kernel command line: %s\n", boot_command_line);
  455. parse_early_param();
  456. parse_args("Booting kernel", static_command_line, __start___param,
  457. __stop___param - __start___param,
  458. -1, -1, &unknown_bootoption);
  459. jump_label_init();
  460. /*
  461. * These use large bootmem allocations and must precede
  462. * kmem_cache_init()
  463. */
  464. setup_log_buf(0);
  465. pidhash_init();
  466. vfs_caches_init_early();
  467. sort_main_extable();
  468. trap_init();
  469. mm_init();
  470. /*
  471. * Set up the scheduler prior starting any interrupts (such as the
  472. * timer interrupt). Full topology setup happens at smp_init()
  473. * time - but meanwhile we still have a functioning scheduler.
  474. */
  475. sched_init();
  476. /*
  477. * Disable preemption - early bootup scheduling is extremely
  478. * fragile until we cpu_idle() for the first time.
  479. */
  480. preempt_disable();
  481. if (WARN(!irqs_disabled(), "Interrupts were enabled *very* early, fixing it\n"))
  482. local_irq_disable();
  483. idr_init_cache();
  484. rcu_init();
  485. tick_nohz_init();
  486. context_tracking_init();
  487. radix_tree_init();
  488. /* init some links before init_ISA_irqs() */
  489. early_irq_init();
  490. init_IRQ();
  491. tick_init();
  492. init_timers();
  493. hrtimers_init();
  494. softirq_init();
  495. timekeeping_init();
  496. time_init();
  497. sched_clock_postinit();
  498. perf_event_init();
  499. profile_init();
  500. call_function_init();
  501. WARN(!irqs_disabled(), "Interrupts were enabled early\n");
  502. early_boot_irqs_disabled = false;
  503. local_irq_enable();
  504. kmem_cache_init_late();
  505. /*
  506. * HACK ALERT! This is early. We're enabling the console before
  507. * we've done PCI setups etc, and console_init() must be aware of
  508. * this. But we do want output early, in case something goes wrong.
  509. */
  510. console_init();
  511. if (panic_later)
  512. panic(panic_later, panic_param);
  513. lockdep_info();
  514. /*
  515. * Need to run this when irqs are enabled, because it wants
  516. * to self-test [hard/soft]-irqs on/off lock inversion bugs
  517. * too:
  518. */
  519. locking_selftest();
  520. #ifdef CONFIG_BLK_DEV_INITRD
  521. if (initrd_start && !initrd_below_start_ok &&
  522. page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
  523. pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
  524. page_to_pfn(virt_to_page((void *)initrd_start)),
  525. min_low_pfn);
  526. initrd_start = 0;
  527. }
  528. #endif
  529. page_cgroup_init();
  530. debug_objects_mem_init();
  531. kmemleak_init();
  532. setup_per_cpu_pageset();
  533. numa_policy_init();
  534. if (late_time_init)
  535. late_time_init();
  536. sched_clock_init();
  537. calibrate_delay();
  538. pidmap_init();
  539. anon_vma_init();
  540. #ifdef CONFIG_X86
  541. if (efi_enabled(EFI_RUNTIME_SERVICES))
  542. efi_enter_virtual_mode();
  543. #endif
  544. thread_info_cache_init();
  545. cred_init();
  546. fork_init(totalram_pages);
  547. proc_caches_init();
  548. buffer_init();
  549. key_init();
  550. security_init();
  551. dbg_late_init();
  552. vfs_caches_init(totalram_pages);
  553. signals_init();
  554. /* rootfs populating might need page-writeback */
  555. page_writeback_init();
  556. #ifdef CONFIG_PROC_FS
  557. proc_root_init();
  558. #endif
  559. cgroup_init();
  560. cpuset_init();
  561. taskstats_init_early();
  562. delayacct_init();
  563. check_bugs();
  564. acpi_early_init(); /* before LAPIC and SMP init */
  565. sfi_init_late();
  566. if (efi_enabled(EFI_RUNTIME_SERVICES)) {
  567. efi_late_init();
  568. efi_free_boot_services();
  569. }
  570. ftrace_init();
  571. /* Do the rest non-__init'ed, we're now alive */
  572. rest_init();
  573. }
  574. /* Call all constructor functions linked into the kernel. */
  575. static void __init do_ctors(void)
  576. {
  577. #ifdef CONFIG_CONSTRUCTORS
  578. ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
  579. for (; fn < (ctor_fn_t *) __ctors_end; fn++)
  580. (*fn)();
  581. #endif
  582. }
  583. bool initcall_debug;
  584. core_param(initcall_debug, initcall_debug, bool, 0644);
  585. static int __init_or_module do_one_initcall_debug(initcall_t fn)
  586. {
  587. ktime_t calltime, delta, rettime;
  588. unsigned long long duration;
  589. int ret;
  590. pr_debug("calling %pF @ %i\n", fn, task_pid_nr(current));
  591. calltime = ktime_get();
  592. ret = fn();
  593. rettime = ktime_get();
  594. delta = ktime_sub(rettime, calltime);
  595. duration = (unsigned long long) ktime_to_ns(delta) >> 10;
  596. pr_debug("initcall %pF returned %d after %lld usecs\n",
  597. fn, ret, duration);
  598. return ret;
  599. }
  600. int __init_or_module do_one_initcall(initcall_t fn)
  601. {
  602. int count = preempt_count();
  603. int ret;
  604. char msgbuf[64];
  605. if (initcall_debug)
  606. ret = do_one_initcall_debug(fn);
  607. else
  608. ret = fn();
  609. msgbuf[0] = 0;
  610. if (preempt_count() != count) {
  611. sprintf(msgbuf, "preemption imbalance ");
  612. preempt_count() = count;
  613. }
  614. if (irqs_disabled()) {
  615. strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
  616. local_irq_enable();
  617. }
  618. WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
  619. return ret;
  620. }
  621. extern initcall_t __initcall_start[];
  622. extern initcall_t __initcall0_start[];
  623. extern initcall_t __initcall1_start[];
  624. extern initcall_t __initcall2_start[];
  625. extern initcall_t __initcall3_start[];
  626. extern initcall_t __initcall4_start[];
  627. extern initcall_t __initcall5_start[];
  628. extern initcall_t __initcall6_start[];
  629. extern initcall_t __initcall7_start[];
  630. extern initcall_t __initcall_end[];
  631. static initcall_t *initcall_levels[] __initdata = {
  632. __initcall0_start,
  633. __initcall1_start,
  634. __initcall2_start,
  635. __initcall3_start,
  636. __initcall4_start,
  637. __initcall5_start,
  638. __initcall6_start,
  639. __initcall7_start,
  640. __initcall_end,
  641. };
  642. /* Keep these in sync with initcalls in include/linux/init.h */
  643. static char *initcall_level_names[] __initdata = {
  644. "early",
  645. "core",
  646. "postcore",
  647. "arch",
  648. "subsys",
  649. "fs",
  650. "device",
  651. "late",
  652. };
  653. static void __init do_initcall_level(int level)
  654. {
  655. extern const struct kernel_param __start___param[], __stop___param[];
  656. initcall_t *fn;
  657. strcpy(static_command_line, saved_command_line);
  658. parse_args(initcall_level_names[level],
  659. static_command_line, __start___param,
  660. __stop___param - __start___param,
  661. level, level,
  662. &repair_env_string);
  663. for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
  664. do_one_initcall(*fn);
  665. }
  666. static void __init do_initcalls(void)
  667. {
  668. int level;
  669. for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
  670. do_initcall_level(level);
  671. }
  672. /*
  673. * Ok, the machine is now initialized. None of the devices
  674. * have been touched yet, but the CPU subsystem is up and
  675. * running, and memory and process management works.
  676. *
  677. * Now we can finally start doing some real work..
  678. */
  679. static void __init do_basic_setup(void)
  680. {
  681. cpuset_init_smp();
  682. usermodehelper_init();
  683. shmem_init();
  684. driver_init();
  685. init_irq_proc();
  686. do_ctors();
  687. usermodehelper_enable();
  688. do_initcalls();
  689. random_int_secret_init();
  690. }
  691. static void __init do_pre_smp_initcalls(void)
  692. {
  693. initcall_t *fn;
  694. for (fn = __initcall_start; fn < __initcall0_start; fn++)
  695. do_one_initcall(*fn);
  696. }
  697. /*
  698. * This function requests modules which should be loaded by default and is
  699. * called twice right after initrd is mounted and right before init is
  700. * exec'd. If such modules are on either initrd or rootfs, they will be
  701. * loaded before control is passed to userland.
  702. */
  703. void __init load_default_modules(void)
  704. {
  705. load_default_elevator_module();
  706. }
  707. static int run_init_process(const char *init_filename)
  708. {
  709. argv_init[0] = init_filename;
  710. return do_execve(init_filename,
  711. (const char __user *const __user *)argv_init,
  712. (const char __user *const __user *)envp_init);
  713. }
  714. static noinline void __init kernel_init_freeable(void);
  715. static int __ref kernel_init(void *unused)
  716. {
  717. kernel_init_freeable();
  718. /* need to finish all async __init code before freeing the memory */
  719. async_synchronize_full();
  720. free_initmem();
  721. mark_rodata_ro();
  722. system_state = SYSTEM_RUNNING;
  723. numa_default_policy();
  724. flush_delayed_fput();
  725. if (ramdisk_execute_command) {
  726. if (!run_init_process(ramdisk_execute_command))
  727. return 0;
  728. pr_err("Failed to execute %s\n", ramdisk_execute_command);
  729. }
  730. /*
  731. * We try each of these until one succeeds.
  732. *
  733. * The Bourne shell can be used instead of init if we are
  734. * trying to recover a really broken machine.
  735. */
  736. if (execute_command) {
  737. if (!run_init_process(execute_command))
  738. return 0;
  739. pr_err("Failed to execute %s. Attempting defaults...\n",
  740. execute_command);
  741. }
  742. if (!run_init_process("/sbin/init") ||
  743. !run_init_process("/etc/init") ||
  744. !run_init_process("/bin/init") ||
  745. !run_init_process("/bin/sh"))
  746. return 0;
  747. panic("No init found. Try passing init= option to kernel. "
  748. "See Linux Documentation/init.txt for guidance.");
  749. }
  750. static noinline void __init kernel_init_freeable(void)
  751. {
  752. /*
  753. * Wait until kthreadd is all set-up.
  754. */
  755. wait_for_completion(&kthreadd_done);
  756. /* Now the scheduler is fully set up and can do blocking allocations */
  757. gfp_allowed_mask = __GFP_BITS_MASK;
  758. /*
  759. * init can allocate pages on any node
  760. */
  761. set_mems_allowed(node_states[N_MEMORY]);
  762. /*
  763. * init can run on any cpu.
  764. */
  765. set_cpus_allowed_ptr(current, cpu_all_mask);
  766. cad_pid = task_pid(current);
  767. smp_prepare_cpus(setup_max_cpus);
  768. do_pre_smp_initcalls();
  769. lockup_detector_init();
  770. smp_init();
  771. sched_init_smp();
  772. do_basic_setup();
  773. /* Open the /dev/console on the rootfs, this should never fail */
  774. if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
  775. pr_err("Warning: unable to open an initial console.\n");
  776. (void) sys_dup(0);
  777. (void) sys_dup(0);
  778. /*
  779. * check if there is an early userspace init. If yes, let it do all
  780. * the work
  781. */
  782. if (!ramdisk_execute_command)
  783. ramdisk_execute_command = "/init";
  784. if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
  785. ramdisk_execute_command = NULL;
  786. prepare_namespace();
  787. }
  788. /*
  789. * Ok, we have completed the initial bootup, and
  790. * we're essentially up and running. Get rid of the
  791. * initmem segments and start the user-mode stuff..
  792. */
  793. /* rootfs is available now, try loading default modules */
  794. load_default_modules();
  795. }