main.c 23 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942
  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. /* Untouched command line saved by arch-specific code. */
  117. char __initdata boot_command_line[COMMAND_LINE_SIZE];
  118. /* Untouched saved command line (eg. for /proc) */
  119. char *saved_command_line;
  120. /* Command line for parameter parsing */
  121. static char *static_command_line;
  122. /* Command line for per-initcall parameter parsing */
  123. static char *initcall_command_line;
  124. static char *execute_command;
  125. static char *ramdisk_execute_command;
  126. /*
  127. * Used to generate warnings if static_key manipulation functions are used
  128. * before jump_label_init is called.
  129. */
  130. bool static_key_initialized __read_mostly = false;
  131. EXPORT_SYMBOL_GPL(static_key_initialized);
  132. /*
  133. * If set, this is an indication to the drivers that reset the underlying
  134. * device before going ahead with the initialization otherwise driver might
  135. * rely on the BIOS and skip the reset operation.
  136. *
  137. * This is useful if kernel is booting in an unreliable environment.
  138. * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
  139. * skipped and devices will be in unknown state.
  140. */
  141. unsigned int reset_devices;
  142. EXPORT_SYMBOL(reset_devices);
  143. static int __init set_reset_devices(char *str)
  144. {
  145. reset_devices = 1;
  146. return 1;
  147. }
  148. __setup("reset_devices", set_reset_devices);
  149. static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
  150. const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
  151. static const char *panic_later, *panic_param;
  152. extern const struct obs_kernel_param __setup_start[], __setup_end[];
  153. static int __init obsolete_checksetup(char *line)
  154. {
  155. const struct obs_kernel_param *p;
  156. int had_early_param = 0;
  157. p = __setup_start;
  158. do {
  159. int n = strlen(p->str);
  160. if (parameqn(line, p->str, n)) {
  161. if (p->early) {
  162. /* Already done in parse_early_param?
  163. * (Needs exact match on param part).
  164. * Keep iterating, as we can have early
  165. * params and __setups of same names 8( */
  166. if (line[n] == '\0' || line[n] == '=')
  167. had_early_param = 1;
  168. } else if (!p->setup_func) {
  169. pr_warn("Parameter %s is obsolete, ignored\n",
  170. p->str);
  171. return 1;
  172. } else if (p->setup_func(line + n))
  173. return 1;
  174. }
  175. p++;
  176. } while (p < __setup_end);
  177. return had_early_param;
  178. }
  179. /*
  180. * This should be approx 2 Bo*oMips to start (note initial shift), and will
  181. * still work even if initially too large, it will just take slightly longer
  182. */
  183. unsigned long loops_per_jiffy = (1<<12);
  184. EXPORT_SYMBOL(loops_per_jiffy);
  185. static int __init debug_kernel(char *str)
  186. {
  187. console_loglevel = 10;
  188. return 0;
  189. }
  190. static int __init quiet_kernel(char *str)
  191. {
  192. console_loglevel = 4;
  193. return 0;
  194. }
  195. early_param("debug", debug_kernel);
  196. early_param("quiet", quiet_kernel);
  197. static int __init loglevel(char *str)
  198. {
  199. int newlevel;
  200. /*
  201. * Only update loglevel value when a correct setting was passed,
  202. * to prevent blind crashes (when loglevel being set to 0) that
  203. * are quite hard to debug
  204. */
  205. if (get_option(&str, &newlevel)) {
  206. console_loglevel = newlevel;
  207. return 0;
  208. }
  209. return -EINVAL;
  210. }
  211. early_param("loglevel", loglevel);
  212. /* Change NUL term back to "=", to make "param" the whole string. */
  213. static int __init repair_env_string(char *param, char *val, const char *unused)
  214. {
  215. if (val) {
  216. /* param=val or param="val"? */
  217. if (val == param+strlen(param)+1)
  218. val[-1] = '=';
  219. else if (val == param+strlen(param)+2) {
  220. val[-2] = '=';
  221. memmove(val-1, val, strlen(val)+1);
  222. val--;
  223. } else
  224. BUG();
  225. }
  226. return 0;
  227. }
  228. /*
  229. * Unknown boot options get handed to init, unless they look like
  230. * unused parameters (modprobe will find them in /proc/cmdline).
  231. */
  232. static int __init unknown_bootoption(char *param, char *val, const char *unused)
  233. {
  234. repair_env_string(param, val, unused);
  235. /* Handle obsolete-style parameters */
  236. if (obsolete_checksetup(param))
  237. return 0;
  238. /* Unused module parameter. */
  239. if (strchr(param, '.') && (!val || strchr(param, '.') < val))
  240. return 0;
  241. if (panic_later)
  242. return 0;
  243. if (val) {
  244. /* Environment option */
  245. unsigned int i;
  246. for (i = 0; envp_init[i]; i++) {
  247. if (i == MAX_INIT_ENVS) {
  248. panic_later = "Too many boot env vars at `%s'";
  249. panic_param = param;
  250. }
  251. if (!strncmp(param, envp_init[i], val - param))
  252. break;
  253. }
  254. envp_init[i] = param;
  255. } else {
  256. /* Command line option */
  257. unsigned int i;
  258. for (i = 0; argv_init[i]; i++) {
  259. if (i == MAX_INIT_ARGS) {
  260. panic_later = "Too many boot init vars at `%s'";
  261. panic_param = param;
  262. }
  263. }
  264. argv_init[i] = param;
  265. }
  266. return 0;
  267. }
  268. static int __init init_setup(char *str)
  269. {
  270. unsigned int i;
  271. execute_command = str;
  272. /*
  273. * In case LILO is going to boot us with default command line,
  274. * it prepends "auto" before the whole cmdline which makes
  275. * the shell think it should execute a script with such name.
  276. * So we ignore all arguments entered _before_ init=... [MJ]
  277. */
  278. for (i = 1; i < MAX_INIT_ARGS; i++)
  279. argv_init[i] = NULL;
  280. return 1;
  281. }
  282. __setup("init=", init_setup);
  283. static int __init rdinit_setup(char *str)
  284. {
  285. unsigned int i;
  286. ramdisk_execute_command = str;
  287. /* See "auto" comment in init_setup */
  288. for (i = 1; i < MAX_INIT_ARGS; i++)
  289. argv_init[i] = NULL;
  290. return 1;
  291. }
  292. __setup("rdinit=", rdinit_setup);
  293. #ifndef CONFIG_SMP
  294. static const unsigned int setup_max_cpus = NR_CPUS;
  295. #ifdef CONFIG_X86_LOCAL_APIC
  296. static void __init smp_init(void)
  297. {
  298. APIC_init_uniprocessor();
  299. }
  300. #else
  301. #define smp_init() do { } while (0)
  302. #endif
  303. static inline void setup_nr_cpu_ids(void) { }
  304. static inline void smp_prepare_cpus(unsigned int maxcpus) { }
  305. #endif
  306. /*
  307. * We need to store the untouched command line for future reference.
  308. * We also need to store the touched command line since the parameter
  309. * parsing is performed in place, and we should allow a component to
  310. * store reference of name/value for future reference.
  311. */
  312. static void __init setup_command_line(char *command_line)
  313. {
  314. saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
  315. initcall_command_line = alloc_bootmem(strlen (boot_command_line)+1);
  316. static_command_line = alloc_bootmem(strlen (command_line)+1);
  317. strcpy (saved_command_line, boot_command_line);
  318. strcpy (static_command_line, command_line);
  319. }
  320. /*
  321. * We need to finalize in a non-__init function or else race conditions
  322. * between the root thread and the init thread may cause start_kernel to
  323. * be reaped by free_initmem before the root thread has proceeded to
  324. * cpu_idle.
  325. *
  326. * gcc-3.4 accidentally inlines this function, so use noinline.
  327. */
  328. static __initdata DECLARE_COMPLETION(kthreadd_done);
  329. static noinline void __init_refok rest_init(void)
  330. {
  331. int pid;
  332. rcu_scheduler_starting();
  333. /*
  334. * We need to spawn init first so that it obtains pid 1, however
  335. * the init task will end up wanting to create kthreads, which, if
  336. * we schedule it before we create kthreadd, will OOPS.
  337. */
  338. kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
  339. numa_default_policy();
  340. pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
  341. rcu_read_lock();
  342. kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
  343. rcu_read_unlock();
  344. complete(&kthreadd_done);
  345. /*
  346. * The boot idle thread must execute schedule()
  347. * at least once to get things moving:
  348. */
  349. init_idle_bootup_task(current);
  350. schedule_preempt_disabled();
  351. /* Call into cpu_idle with preempt disabled */
  352. cpu_startup_entry(CPUHP_ONLINE);
  353. }
  354. /* Check for early params. */
  355. static int __init do_early_param(char *param, char *val, const char *unused)
  356. {
  357. const struct obs_kernel_param *p;
  358. for (p = __setup_start; p < __setup_end; p++) {
  359. if ((p->early && parameq(param, p->str)) ||
  360. (strcmp(param, "console") == 0 &&
  361. strcmp(p->str, "earlycon") == 0)
  362. ) {
  363. if (p->setup_func(val) != 0)
  364. pr_warn("Malformed early option '%s'\n", param);
  365. }
  366. }
  367. /* We accept everything at this stage. */
  368. return 0;
  369. }
  370. void __init parse_early_options(char *cmdline)
  371. {
  372. parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
  373. }
  374. /* Arch code calls this early on, or if not, just before other parsing. */
  375. void __init parse_early_param(void)
  376. {
  377. static __initdata int done = 0;
  378. static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
  379. if (done)
  380. return;
  381. /* All fall through to do_early_param. */
  382. strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
  383. parse_early_options(tmp_cmdline);
  384. done = 1;
  385. }
  386. /*
  387. * Activate the first processor.
  388. */
  389. static void __init boot_cpu_init(void)
  390. {
  391. int cpu = smp_processor_id();
  392. /* Mark the boot cpu "present", "online" etc for SMP and UP case */
  393. set_cpu_online(cpu, true);
  394. set_cpu_active(cpu, true);
  395. set_cpu_present(cpu, true);
  396. set_cpu_possible(cpu, true);
  397. }
  398. void __init __weak smp_setup_processor_id(void)
  399. {
  400. }
  401. # if THREAD_SIZE >= PAGE_SIZE
  402. void __init __weak thread_info_cache_init(void)
  403. {
  404. }
  405. #endif
  406. /*
  407. * Set up kernel memory allocators
  408. */
  409. static void __init mm_init(void)
  410. {
  411. /*
  412. * page_cgroup requires contiguous pages,
  413. * bigger than MAX_ORDER unless SPARSEMEM.
  414. */
  415. page_cgroup_init_flatmem();
  416. mem_init();
  417. kmem_cache_init();
  418. percpu_init_late();
  419. pgtable_cache_init();
  420. vmalloc_init();
  421. }
  422. asmlinkage void __init start_kernel(void)
  423. {
  424. char * command_line;
  425. extern const struct kernel_param __start___param[], __stop___param[];
  426. /*
  427. * Need to run as early as possible, to initialize the
  428. * lockdep hash:
  429. */
  430. lockdep_init();
  431. smp_setup_processor_id();
  432. debug_objects_early_init();
  433. /*
  434. * Set up the the initial canary ASAP:
  435. */
  436. boot_init_stack_canary();
  437. cgroup_init_early();
  438. local_irq_disable();
  439. early_boot_irqs_disabled = true;
  440. /*
  441. * Interrupts are still disabled. Do necessary setups, then
  442. * enable them
  443. */
  444. boot_cpu_init();
  445. page_address_init();
  446. pr_notice("%s", linux_banner);
  447. setup_arch(&command_line);
  448. mm_init_owner(&init_mm, &init_task);
  449. mm_init_cpumask(&init_mm);
  450. setup_command_line(command_line);
  451. setup_nr_cpu_ids();
  452. setup_per_cpu_areas();
  453. smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
  454. build_all_zonelists(NULL, NULL);
  455. page_alloc_init();
  456. pr_notice("Kernel command line: %s\n", boot_command_line);
  457. parse_early_param();
  458. parse_args("Booting kernel", static_command_line, __start___param,
  459. __stop___param - __start___param,
  460. -1, -1, &unknown_bootoption);
  461. jump_label_init();
  462. /*
  463. * These use large bootmem allocations and must precede
  464. * kmem_cache_init()
  465. */
  466. setup_log_buf(0);
  467. pidhash_init();
  468. vfs_caches_init_early();
  469. sort_main_extable();
  470. trap_init();
  471. mm_init();
  472. /*
  473. * Set up the scheduler prior starting any interrupts (such as the
  474. * timer interrupt). Full topology setup happens at smp_init()
  475. * time - but meanwhile we still have a functioning scheduler.
  476. */
  477. sched_init();
  478. /*
  479. * Disable preemption - early bootup scheduling is extremely
  480. * fragile until we cpu_idle() for the first time.
  481. */
  482. preempt_disable();
  483. if (WARN(!irqs_disabled(), "Interrupts were enabled *very* early, fixing it\n"))
  484. local_irq_disable();
  485. idr_init_cache();
  486. rcu_init();
  487. tick_nohz_init();
  488. context_tracking_init();
  489. radix_tree_init();
  490. /* init some links before init_ISA_irqs() */
  491. early_irq_init();
  492. init_IRQ();
  493. tick_init();
  494. init_timers();
  495. hrtimers_init();
  496. softirq_init();
  497. timekeeping_init();
  498. time_init();
  499. sched_clock_postinit();
  500. perf_event_init();
  501. profile_init();
  502. call_function_init();
  503. WARN(!irqs_disabled(), "Interrupts were enabled early\n");
  504. early_boot_irqs_disabled = false;
  505. local_irq_enable();
  506. kmem_cache_init_late();
  507. /*
  508. * HACK ALERT! This is early. We're enabling the console before
  509. * we've done PCI setups etc, and console_init() must be aware of
  510. * this. But we do want output early, in case something goes wrong.
  511. */
  512. console_init();
  513. if (panic_later)
  514. panic(panic_later, panic_param);
  515. lockdep_info();
  516. /*
  517. * Need to run this when irqs are enabled, because it wants
  518. * to self-test [hard/soft]-irqs on/off lock inversion bugs
  519. * too:
  520. */
  521. locking_selftest();
  522. #ifdef CONFIG_BLK_DEV_INITRD
  523. if (initrd_start && !initrd_below_start_ok &&
  524. page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
  525. pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
  526. page_to_pfn(virt_to_page((void *)initrd_start)),
  527. min_low_pfn);
  528. initrd_start = 0;
  529. }
  530. #endif
  531. page_cgroup_init();
  532. debug_objects_mem_init();
  533. kmemleak_init();
  534. setup_per_cpu_pageset();
  535. numa_policy_init();
  536. if (late_time_init)
  537. late_time_init();
  538. sched_clock_init();
  539. calibrate_delay();
  540. pidmap_init();
  541. anon_vma_init();
  542. #ifdef CONFIG_X86
  543. if (efi_enabled(EFI_RUNTIME_SERVICES))
  544. efi_enter_virtual_mode();
  545. #endif
  546. thread_info_cache_init();
  547. cred_init();
  548. fork_init(totalram_pages);
  549. proc_caches_init();
  550. buffer_init();
  551. key_init();
  552. security_init();
  553. dbg_late_init();
  554. vfs_caches_init(totalram_pages);
  555. signals_init();
  556. /* rootfs populating might need page-writeback */
  557. page_writeback_init();
  558. #ifdef CONFIG_PROC_FS
  559. proc_root_init();
  560. #endif
  561. cgroup_init();
  562. cpuset_init();
  563. taskstats_init_early();
  564. delayacct_init();
  565. check_bugs();
  566. acpi_early_init(); /* before LAPIC and SMP init */
  567. sfi_init_late();
  568. if (efi_enabled(EFI_RUNTIME_SERVICES)) {
  569. efi_late_init();
  570. efi_free_boot_services();
  571. }
  572. ftrace_init();
  573. /* Do the rest non-__init'ed, we're now alive */
  574. rest_init();
  575. }
  576. /* Call all constructor functions linked into the kernel. */
  577. static void __init do_ctors(void)
  578. {
  579. #ifdef CONFIG_CONSTRUCTORS
  580. ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
  581. for (; fn < (ctor_fn_t *) __ctors_end; fn++)
  582. (*fn)();
  583. #endif
  584. }
  585. bool initcall_debug;
  586. core_param(initcall_debug, initcall_debug, bool, 0644);
  587. static int __init_or_module do_one_initcall_debug(initcall_t fn)
  588. {
  589. ktime_t calltime, delta, rettime;
  590. unsigned long long duration;
  591. int ret;
  592. pr_debug("calling %pF @ %i\n", fn, task_pid_nr(current));
  593. calltime = ktime_get();
  594. ret = fn();
  595. rettime = ktime_get();
  596. delta = ktime_sub(rettime, calltime);
  597. duration = (unsigned long long) ktime_to_ns(delta) >> 10;
  598. pr_debug("initcall %pF returned %d after %lld usecs\n",
  599. fn, ret, duration);
  600. return ret;
  601. }
  602. int __init_or_module do_one_initcall(initcall_t fn)
  603. {
  604. int count = preempt_count();
  605. int ret;
  606. char msgbuf[64];
  607. if (initcall_debug)
  608. ret = do_one_initcall_debug(fn);
  609. else
  610. ret = fn();
  611. msgbuf[0] = 0;
  612. if (preempt_count() != count) {
  613. sprintf(msgbuf, "preemption imbalance ");
  614. preempt_count_set(count);
  615. }
  616. if (irqs_disabled()) {
  617. strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
  618. local_irq_enable();
  619. }
  620. WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
  621. return ret;
  622. }
  623. extern initcall_t __initcall_start[];
  624. extern initcall_t __initcall0_start[];
  625. extern initcall_t __initcall1_start[];
  626. extern initcall_t __initcall2_start[];
  627. extern initcall_t __initcall3_start[];
  628. extern initcall_t __initcall4_start[];
  629. extern initcall_t __initcall5_start[];
  630. extern initcall_t __initcall6_start[];
  631. extern initcall_t __initcall7_start[];
  632. extern initcall_t __initcall_end[];
  633. static initcall_t *initcall_levels[] __initdata = {
  634. __initcall0_start,
  635. __initcall1_start,
  636. __initcall2_start,
  637. __initcall3_start,
  638. __initcall4_start,
  639. __initcall5_start,
  640. __initcall6_start,
  641. __initcall7_start,
  642. __initcall_end,
  643. };
  644. /* Keep these in sync with initcalls in include/linux/init.h */
  645. static char *initcall_level_names[] __initdata = {
  646. "early",
  647. "core",
  648. "postcore",
  649. "arch",
  650. "subsys",
  651. "fs",
  652. "device",
  653. "late",
  654. };
  655. static void __init do_initcall_level(int level)
  656. {
  657. extern const struct kernel_param __start___param[], __stop___param[];
  658. initcall_t *fn;
  659. strcpy(initcall_command_line, saved_command_line);
  660. parse_args(initcall_level_names[level],
  661. initcall_command_line, __start___param,
  662. __stop___param - __start___param,
  663. level, level,
  664. &repair_env_string);
  665. for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
  666. do_one_initcall(*fn);
  667. }
  668. static void __init do_initcalls(void)
  669. {
  670. int level;
  671. for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
  672. do_initcall_level(level);
  673. }
  674. /*
  675. * Ok, the machine is now initialized. None of the devices
  676. * have been touched yet, but the CPU subsystem is up and
  677. * running, and memory and process management works.
  678. *
  679. * Now we can finally start doing some real work..
  680. */
  681. static void __init do_basic_setup(void)
  682. {
  683. cpuset_init_smp();
  684. usermodehelper_init();
  685. shmem_init();
  686. driver_init();
  687. init_irq_proc();
  688. do_ctors();
  689. usermodehelper_enable();
  690. do_initcalls();
  691. random_int_secret_init();
  692. }
  693. static void __init do_pre_smp_initcalls(void)
  694. {
  695. initcall_t *fn;
  696. for (fn = __initcall_start; fn < __initcall0_start; fn++)
  697. do_one_initcall(*fn);
  698. }
  699. /*
  700. * This function requests modules which should be loaded by default and is
  701. * called twice right after initrd is mounted and right before init is
  702. * exec'd. If such modules are on either initrd or rootfs, they will be
  703. * loaded before control is passed to userland.
  704. */
  705. void __init load_default_modules(void)
  706. {
  707. load_default_elevator_module();
  708. }
  709. static int run_init_process(const char *init_filename)
  710. {
  711. argv_init[0] = init_filename;
  712. return do_execve(init_filename,
  713. (const char __user *const __user *)argv_init,
  714. (const char __user *const __user *)envp_init);
  715. }
  716. static int try_to_run_init_process(const char *init_filename)
  717. {
  718. int ret;
  719. ret = run_init_process(init_filename);
  720. if (ret && ret != -ENOENT) {
  721. pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
  722. init_filename, ret);
  723. }
  724. return ret;
  725. }
  726. static noinline void __init kernel_init_freeable(void);
  727. static int __ref kernel_init(void *unused)
  728. {
  729. int ret;
  730. kernel_init_freeable();
  731. /* need to finish all async __init code before freeing the memory */
  732. async_synchronize_full();
  733. free_initmem();
  734. mark_rodata_ro();
  735. system_state = SYSTEM_RUNNING;
  736. numa_default_policy();
  737. flush_delayed_fput();
  738. if (ramdisk_execute_command) {
  739. ret = run_init_process(ramdisk_execute_command);
  740. if (!ret)
  741. return 0;
  742. pr_err("Failed to execute %s (error %d)\n",
  743. ramdisk_execute_command, ret);
  744. }
  745. /*
  746. * We try each of these until one succeeds.
  747. *
  748. * The Bourne shell can be used instead of init if we are
  749. * trying to recover a really broken machine.
  750. */
  751. if (execute_command) {
  752. ret = run_init_process(execute_command);
  753. if (!ret)
  754. return 0;
  755. pr_err("Failed to execute %s (error %d). Attempting defaults...\n",
  756. execute_command, ret);
  757. }
  758. if (!try_to_run_init_process("/sbin/init") ||
  759. !try_to_run_init_process("/etc/init") ||
  760. !try_to_run_init_process("/bin/init") ||
  761. !try_to_run_init_process("/bin/sh"))
  762. return 0;
  763. panic("No working init found. Try passing init= option to kernel. "
  764. "See Linux Documentation/init.txt for guidance.");
  765. }
  766. static noinline void __init kernel_init_freeable(void)
  767. {
  768. /*
  769. * Wait until kthreadd is all set-up.
  770. */
  771. wait_for_completion(&kthreadd_done);
  772. /* Now the scheduler is fully set up and can do blocking allocations */
  773. gfp_allowed_mask = __GFP_BITS_MASK;
  774. /*
  775. * init can allocate pages on any node
  776. */
  777. set_mems_allowed(node_states[N_MEMORY]);
  778. /*
  779. * init can run on any cpu.
  780. */
  781. set_cpus_allowed_ptr(current, cpu_all_mask);
  782. cad_pid = task_pid(current);
  783. smp_prepare_cpus(setup_max_cpus);
  784. do_pre_smp_initcalls();
  785. lockup_detector_init();
  786. smp_init();
  787. sched_init_smp();
  788. do_basic_setup();
  789. /* Open the /dev/console on the rootfs, this should never fail */
  790. if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
  791. pr_err("Warning: unable to open an initial console.\n");
  792. (void) sys_dup(0);
  793. (void) sys_dup(0);
  794. /*
  795. * check if there is an early userspace init. If yes, let it do all
  796. * the work
  797. */
  798. if (!ramdisk_execute_command)
  799. ramdisk_execute_command = "/init";
  800. if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
  801. ramdisk_execute_command = NULL;
  802. prepare_namespace();
  803. }
  804. /*
  805. * Ok, we have completed the initial bootup, and
  806. * we're essentially up and running. Get rid of the
  807. * initmem segments and start the user-mode stuff..
  808. */
  809. /* rootfs is available now, try loading default modules */
  810. load_default_modules();
  811. }