main.c 23 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939
  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 radix_tree_init(void);
  90. #ifndef CONFIG_DEBUG_RODATA
  91. static inline void mark_rodata_ro(void) { }
  92. #endif
  93. /*
  94. * Debug helper: via this flag we know that we are in 'early bootup code'
  95. * where only the boot processor is running with IRQ disabled. This means
  96. * two things - IRQ must not be enabled before the flag is cleared and some
  97. * operations which are not allowed with IRQ disabled are allowed while the
  98. * flag is set.
  99. */
  100. bool early_boot_irqs_disabled __read_mostly;
  101. enum system_states system_state __read_mostly;
  102. EXPORT_SYMBOL(system_state);
  103. /*
  104. * Boot command-line arguments
  105. */
  106. #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
  107. #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
  108. extern void time_init(void);
  109. /* Default late time init is NULL. archs can override this later. */
  110. void (*__initdata late_time_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. /* Command line for per-initcall parameter parsing */
  118. static char *initcall_command_line;
  119. static char *execute_command;
  120. static char *ramdisk_execute_command;
  121. /*
  122. * Used to generate warnings if static_key manipulation functions are used
  123. * before jump_label_init is called.
  124. */
  125. bool static_key_initialized __read_mostly = false;
  126. EXPORT_SYMBOL_GPL(static_key_initialized);
  127. /*
  128. * If set, this is an indication to the drivers that reset the underlying
  129. * device before going ahead with the initialization otherwise driver might
  130. * rely on the BIOS and skip the reset operation.
  131. *
  132. * This is useful if kernel is booting in an unreliable environment.
  133. * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
  134. * skipped and devices will be in unknown state.
  135. */
  136. unsigned int reset_devices;
  137. EXPORT_SYMBOL(reset_devices);
  138. static int __init set_reset_devices(char *str)
  139. {
  140. reset_devices = 1;
  141. return 1;
  142. }
  143. __setup("reset_devices", set_reset_devices);
  144. static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
  145. const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
  146. static const char *panic_later, *panic_param;
  147. extern const struct obs_kernel_param __setup_start[], __setup_end[];
  148. static int __init obsolete_checksetup(char *line)
  149. {
  150. const struct obs_kernel_param *p;
  151. int had_early_param = 0;
  152. p = __setup_start;
  153. do {
  154. int n = strlen(p->str);
  155. if (parameqn(line, p->str, n)) {
  156. if (p->early) {
  157. /* Already done in parse_early_param?
  158. * (Needs exact match on param part).
  159. * Keep iterating, as we can have early
  160. * params and __setups of same names 8( */
  161. if (line[n] == '\0' || line[n] == '=')
  162. had_early_param = 1;
  163. } else if (!p->setup_func) {
  164. pr_warn("Parameter %s is obsolete, ignored\n",
  165. p->str);
  166. return 1;
  167. } else if (p->setup_func(line + n))
  168. return 1;
  169. }
  170. p++;
  171. } while (p < __setup_end);
  172. return had_early_param;
  173. }
  174. /*
  175. * This should be approx 2 Bo*oMips to start (note initial shift), and will
  176. * still work even if initially too large, it will just take slightly longer
  177. */
  178. unsigned long loops_per_jiffy = (1<<12);
  179. EXPORT_SYMBOL(loops_per_jiffy);
  180. static int __init debug_kernel(char *str)
  181. {
  182. console_loglevel = 10;
  183. return 0;
  184. }
  185. static int __init quiet_kernel(char *str)
  186. {
  187. console_loglevel = 4;
  188. return 0;
  189. }
  190. early_param("debug", debug_kernel);
  191. early_param("quiet", quiet_kernel);
  192. static int __init loglevel(char *str)
  193. {
  194. int newlevel;
  195. /*
  196. * Only update loglevel value when a correct setting was passed,
  197. * to prevent blind crashes (when loglevel being set to 0) that
  198. * are quite hard to debug
  199. */
  200. if (get_option(&str, &newlevel)) {
  201. console_loglevel = newlevel;
  202. return 0;
  203. }
  204. return -EINVAL;
  205. }
  206. early_param("loglevel", loglevel);
  207. /* Change NUL term back to "=", to make "param" the whole string. */
  208. static int __init repair_env_string(char *param, char *val, const char *unused)
  209. {
  210. if (val) {
  211. /* param=val or param="val"? */
  212. if (val == param+strlen(param)+1)
  213. val[-1] = '=';
  214. else if (val == param+strlen(param)+2) {
  215. val[-2] = '=';
  216. memmove(val-1, val, strlen(val)+1);
  217. val--;
  218. } else
  219. BUG();
  220. }
  221. return 0;
  222. }
  223. /*
  224. * Unknown boot options get handed to init, unless they look like
  225. * unused parameters (modprobe will find them in /proc/cmdline).
  226. */
  227. static int __init unknown_bootoption(char *param, char *val, const char *unused)
  228. {
  229. repair_env_string(param, val, unused);
  230. /* Handle obsolete-style parameters */
  231. if (obsolete_checksetup(param))
  232. return 0;
  233. /* Unused module parameter. */
  234. if (strchr(param, '.') && (!val || strchr(param, '.') < val))
  235. return 0;
  236. if (panic_later)
  237. return 0;
  238. if (val) {
  239. /* Environment option */
  240. unsigned int i;
  241. for (i = 0; envp_init[i]; i++) {
  242. if (i == MAX_INIT_ENVS) {
  243. panic_later = "env";
  244. panic_param = param;
  245. }
  246. if (!strncmp(param, envp_init[i], val - param))
  247. break;
  248. }
  249. envp_init[i] = param;
  250. } else {
  251. /* Command line option */
  252. unsigned int i;
  253. for (i = 0; argv_init[i]; i++) {
  254. if (i == MAX_INIT_ARGS) {
  255. panic_later = "init";
  256. panic_param = param;
  257. }
  258. }
  259. argv_init[i] = param;
  260. }
  261. return 0;
  262. }
  263. static int __init init_setup(char *str)
  264. {
  265. unsigned int i;
  266. execute_command = str;
  267. /*
  268. * In case LILO is going to boot us with default command line,
  269. * it prepends "auto" before the whole cmdline which makes
  270. * the shell think it should execute a script with such name.
  271. * So we ignore all arguments entered _before_ init=... [MJ]
  272. */
  273. for (i = 1; i < MAX_INIT_ARGS; i++)
  274. argv_init[i] = NULL;
  275. return 1;
  276. }
  277. __setup("init=", init_setup);
  278. static int __init rdinit_setup(char *str)
  279. {
  280. unsigned int i;
  281. ramdisk_execute_command = str;
  282. /* See "auto" comment in init_setup */
  283. for (i = 1; i < MAX_INIT_ARGS; i++)
  284. argv_init[i] = NULL;
  285. return 1;
  286. }
  287. __setup("rdinit=", rdinit_setup);
  288. #ifndef CONFIG_SMP
  289. static const unsigned int setup_max_cpus = NR_CPUS;
  290. #ifdef CONFIG_X86_LOCAL_APIC
  291. static void __init smp_init(void)
  292. {
  293. APIC_init_uniprocessor();
  294. }
  295. #else
  296. #define smp_init() do { } while (0)
  297. #endif
  298. static inline void setup_nr_cpu_ids(void) { }
  299. static inline void smp_prepare_cpus(unsigned int maxcpus) { }
  300. #endif
  301. /*
  302. * We need to store the untouched command line for future reference.
  303. * We also need to store the touched command line since the parameter
  304. * parsing is performed in place, and we should allow a component to
  305. * store reference of name/value for future reference.
  306. */
  307. static void __init setup_command_line(char *command_line)
  308. {
  309. saved_command_line =
  310. memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
  311. initcall_command_line =
  312. memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
  313. static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0);
  314. strcpy (saved_command_line, boot_command_line);
  315. strcpy (static_command_line, command_line);
  316. }
  317. /*
  318. * We need to finalize in a non-__init function or else race conditions
  319. * between the root thread and the init thread may cause start_kernel to
  320. * be reaped by free_initmem before the root thread has proceeded to
  321. * cpu_idle.
  322. *
  323. * gcc-3.4 accidentally inlines this function, so use noinline.
  324. */
  325. static __initdata DECLARE_COMPLETION(kthreadd_done);
  326. static noinline void __init_refok rest_init(void)
  327. {
  328. int pid;
  329. rcu_scheduler_starting();
  330. /*
  331. * We need to spawn init first so that it obtains pid 1, however
  332. * the init task will end up wanting to create kthreads, which, if
  333. * we schedule it before we create kthreadd, will OOPS.
  334. */
  335. kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
  336. numa_default_policy();
  337. pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
  338. rcu_read_lock();
  339. kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
  340. rcu_read_unlock();
  341. complete(&kthreadd_done);
  342. /*
  343. * The boot idle thread must execute schedule()
  344. * at least once to get things moving:
  345. */
  346. init_idle_bootup_task(current);
  347. schedule_preempt_disabled();
  348. /* Call into cpu_idle with preempt disabled */
  349. cpu_startup_entry(CPUHP_ONLINE);
  350. }
  351. /* Check for early params. */
  352. static int __init do_early_param(char *param, char *val, const char *unused)
  353. {
  354. const struct obs_kernel_param *p;
  355. for (p = __setup_start; p < __setup_end; p++) {
  356. if ((p->early && parameq(param, p->str)) ||
  357. (strcmp(param, "console") == 0 &&
  358. strcmp(p->str, "earlycon") == 0)
  359. ) {
  360. if (p->setup_func(val) != 0)
  361. pr_warn("Malformed early option '%s'\n", param);
  362. }
  363. }
  364. /* We accept everything at this stage. */
  365. return 0;
  366. }
  367. void __init parse_early_options(char *cmdline)
  368. {
  369. parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
  370. }
  371. /* Arch code calls this early on, or if not, just before other parsing. */
  372. void __init parse_early_param(void)
  373. {
  374. static __initdata int done = 0;
  375. static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
  376. if (done)
  377. return;
  378. /* All fall through to do_early_param. */
  379. strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
  380. parse_early_options(tmp_cmdline);
  381. done = 1;
  382. }
  383. /*
  384. * Activate the first processor.
  385. */
  386. static void __init boot_cpu_init(void)
  387. {
  388. int cpu = smp_processor_id();
  389. /* Mark the boot cpu "present", "online" etc for SMP and UP case */
  390. set_cpu_online(cpu, true);
  391. set_cpu_active(cpu, true);
  392. set_cpu_present(cpu, true);
  393. set_cpu_possible(cpu, true);
  394. }
  395. void __init __weak smp_setup_processor_id(void)
  396. {
  397. }
  398. # if THREAD_SIZE >= PAGE_SIZE
  399. void __init __weak thread_info_cache_init(void)
  400. {
  401. }
  402. #endif
  403. /*
  404. * Set up kernel memory allocators
  405. */
  406. static void __init mm_init(void)
  407. {
  408. /*
  409. * page_cgroup requires contiguous pages,
  410. * bigger than MAX_ORDER unless SPARSEMEM.
  411. */
  412. page_cgroup_init_flatmem();
  413. mem_init();
  414. kmem_cache_init();
  415. percpu_init_late();
  416. pgtable_init();
  417. vmalloc_init();
  418. }
  419. asmlinkage void __init start_kernel(void)
  420. {
  421. char * command_line;
  422. extern const struct kernel_param __start___param[], __stop___param[];
  423. /*
  424. * Need to run as early as possible, to initialize the
  425. * lockdep hash:
  426. */
  427. lockdep_init();
  428. smp_setup_processor_id();
  429. debug_objects_early_init();
  430. /*
  431. * Set up the the initial canary ASAP:
  432. */
  433. boot_init_stack_canary();
  434. cgroup_init_early();
  435. local_irq_disable();
  436. early_boot_irqs_disabled = true;
  437. /*
  438. * Interrupts are still disabled. Do necessary setups, then
  439. * enable them
  440. */
  441. boot_cpu_init();
  442. page_address_init();
  443. pr_notice("%s", linux_banner);
  444. setup_arch(&command_line);
  445. mm_init_owner(&init_mm, &init_task);
  446. mm_init_cpumask(&init_mm);
  447. setup_command_line(command_line);
  448. setup_nr_cpu_ids();
  449. setup_per_cpu_areas();
  450. smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
  451. build_all_zonelists(NULL, NULL);
  452. page_alloc_init();
  453. pr_notice("Kernel command line: %s\n", boot_command_line);
  454. parse_early_param();
  455. parse_args("Booting kernel", static_command_line, __start___param,
  456. __stop___param - __start___param,
  457. -1, -1, &unknown_bootoption);
  458. jump_label_init();
  459. /*
  460. * These use large bootmem allocations and must precede
  461. * kmem_cache_init()
  462. */
  463. setup_log_buf(0);
  464. pidhash_init();
  465. vfs_caches_init_early();
  466. sort_main_extable();
  467. trap_init();
  468. mm_init();
  469. /*
  470. * Set up the scheduler prior starting any interrupts (such as the
  471. * timer interrupt). Full topology setup happens at smp_init()
  472. * time - but meanwhile we still have a functioning scheduler.
  473. */
  474. sched_init();
  475. /*
  476. * Disable preemption - early bootup scheduling is extremely
  477. * fragile until we cpu_idle() for the first time.
  478. */
  479. preempt_disable();
  480. if (WARN(!irqs_disabled(), "Interrupts were enabled *very* early, fixing it\n"))
  481. local_irq_disable();
  482. idr_init_cache();
  483. rcu_init();
  484. tick_nohz_init();
  485. context_tracking_init();
  486. radix_tree_init();
  487. /* init some links before init_ISA_irqs() */
  488. early_irq_init();
  489. init_IRQ();
  490. tick_init();
  491. init_timers();
  492. hrtimers_init();
  493. softirq_init();
  494. timekeeping_init();
  495. time_init();
  496. sched_clock_postinit();
  497. perf_event_init();
  498. profile_init();
  499. call_function_init();
  500. WARN(!irqs_disabled(), "Interrupts were enabled early\n");
  501. early_boot_irqs_disabled = false;
  502. local_irq_enable();
  503. kmem_cache_init_late();
  504. /*
  505. * HACK ALERT! This is early. We're enabling the console before
  506. * we've done PCI setups etc, and console_init() must be aware of
  507. * this. But we do want output early, in case something goes wrong.
  508. */
  509. console_init();
  510. if (panic_later)
  511. panic("Too many boot %s vars at `%s'", panic_later,
  512. 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. acpi_early_init();
  541. #ifdef CONFIG_X86
  542. if (efi_enabled(EFI_RUNTIME_SERVICES))
  543. efi_enter_virtual_mode();
  544. #endif
  545. thread_info_cache_init();
  546. cred_init();
  547. fork_init(totalram_pages);
  548. proc_caches_init();
  549. buffer_init();
  550. key_init();
  551. security_init();
  552. dbg_late_init();
  553. vfs_caches_init(totalram_pages);
  554. signals_init();
  555. /* rootfs populating might need page-writeback */
  556. page_writeback_init();
  557. #ifdef CONFIG_PROC_FS
  558. proc_root_init();
  559. #endif
  560. cgroup_init();
  561. cpuset_init();
  562. taskstats_init_early();
  563. delayacct_init();
  564. check_bugs();
  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_set(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(initcall_command_line, saved_command_line);
  658. parse_args(initcall_level_names[level],
  659. initcall_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(getname_kernel(init_filename),
  711. (const char __user *const __user *)argv_init,
  712. (const char __user *const __user *)envp_init);
  713. }
  714. static int try_to_run_init_process(const char *init_filename)
  715. {
  716. int ret;
  717. ret = run_init_process(init_filename);
  718. if (ret && ret != -ENOENT) {
  719. pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
  720. init_filename, ret);
  721. }
  722. return ret;
  723. }
  724. static noinline void __init kernel_init_freeable(void);
  725. static int __ref kernel_init(void *unused)
  726. {
  727. int ret;
  728. kernel_init_freeable();
  729. /* need to finish all async __init code before freeing the memory */
  730. async_synchronize_full();
  731. free_initmem();
  732. mark_rodata_ro();
  733. system_state = SYSTEM_RUNNING;
  734. numa_default_policy();
  735. flush_delayed_fput();
  736. if (ramdisk_execute_command) {
  737. ret = run_init_process(ramdisk_execute_command);
  738. if (!ret)
  739. return 0;
  740. pr_err("Failed to execute %s (error %d)\n",
  741. ramdisk_execute_command, ret);
  742. }
  743. /*
  744. * We try each of these until one succeeds.
  745. *
  746. * The Bourne shell can be used instead of init if we are
  747. * trying to recover a really broken machine.
  748. */
  749. if (execute_command) {
  750. ret = run_init_process(execute_command);
  751. if (!ret)
  752. return 0;
  753. pr_err("Failed to execute %s (error %d). Attempting defaults...\n",
  754. execute_command, ret);
  755. }
  756. if (!try_to_run_init_process("/sbin/init") ||
  757. !try_to_run_init_process("/etc/init") ||
  758. !try_to_run_init_process("/bin/init") ||
  759. !try_to_run_init_process("/bin/sh"))
  760. return 0;
  761. panic("No working init found. Try passing init= option to kernel. "
  762. "See Linux Documentation/init.txt for guidance.");
  763. }
  764. static noinline void __init kernel_init_freeable(void)
  765. {
  766. /*
  767. * Wait until kthreadd is all set-up.
  768. */
  769. wait_for_completion(&kthreadd_done);
  770. /* Now the scheduler is fully set up and can do blocking allocations */
  771. gfp_allowed_mask = __GFP_BITS_MASK;
  772. /*
  773. * init can allocate pages on any node
  774. */
  775. set_mems_allowed(node_states[N_MEMORY]);
  776. /*
  777. * init can run on any cpu.
  778. */
  779. set_cpus_allowed_ptr(current, cpu_all_mask);
  780. cad_pid = task_pid(current);
  781. smp_prepare_cpus(setup_max_cpus);
  782. do_pre_smp_initcalls();
  783. lockup_detector_init();
  784. smp_init();
  785. sched_init_smp();
  786. do_basic_setup();
  787. /* Open the /dev/console on the rootfs, this should never fail */
  788. if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
  789. pr_err("Warning: unable to open an initial console.\n");
  790. (void) sys_dup(0);
  791. (void) sys_dup(0);
  792. /*
  793. * check if there is an early userspace init. If yes, let it do all
  794. * the work
  795. */
  796. if (!ramdisk_execute_command)
  797. ramdisk_execute_command = "/init";
  798. if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
  799. ramdisk_execute_command = NULL;
  800. prepare_namespace();
  801. }
  802. /*
  803. * Ok, we have completed the initial bootup, and
  804. * we're essentially up and running. Get rid of the
  805. * initmem segments and start the user-mode stuff..
  806. */
  807. /* rootfs is available now, try loading default modules */
  808. load_default_modules();
  809. }