main.c 22 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919
  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 <asm/io.h>
  78. #include <asm/bugs.h>
  79. #include <asm/setup.h>
  80. #include <asm/sections.h>
  81. #include <asm/cacheflush.h>
  82. #ifdef CONFIG_X86_LOCAL_APIC
  83. #include <asm/smp.h>
  84. #endif
  85. static int kernel_init(void *);
  86. extern void init_IRQ(void);
  87. extern void fork_init(unsigned long);
  88. extern void mca_init(void);
  89. extern void sbus_init(void);
  90. extern void radix_tree_init(void);
  91. #ifndef CONFIG_DEBUG_RODATA
  92. static inline void mark_rodata_ro(void) { }
  93. #endif
  94. #ifdef CONFIG_TC
  95. extern void tc_init(void);
  96. #endif
  97. /*
  98. * Debug helper: via this flag we know that we are in 'early bootup code'
  99. * where only the boot processor is running with IRQ disabled. This means
  100. * two things - IRQ must not be enabled before the flag is cleared and some
  101. * operations which are not allowed with IRQ disabled are allowed while the
  102. * flag is set.
  103. */
  104. bool early_boot_irqs_disabled __read_mostly;
  105. enum system_states system_state __read_mostly;
  106. EXPORT_SYMBOL(system_state);
  107. /*
  108. * Boot command-line arguments
  109. */
  110. #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
  111. #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
  112. extern void time_init(void);
  113. /* Default late time init is NULL. archs can override this later. */
  114. void (*__initdata late_time_init)(void);
  115. extern void softirq_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. static char *execute_command;
  123. static char *ramdisk_execute_command;
  124. /*
  125. * Used to generate warnings if static_key manipulation functions are used
  126. * before jump_label_init is called.
  127. */
  128. bool static_key_initialized __read_mostly = false;
  129. EXPORT_SYMBOL_GPL(static_key_initialized);
  130. /*
  131. * If set, this is an indication to the drivers that reset the underlying
  132. * device before going ahead with the initialization otherwise driver might
  133. * rely on the BIOS and skip the reset operation.
  134. *
  135. * This is useful if kernel is booting in an unreliable environment.
  136. * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
  137. * skipped and devices will be in unknown state.
  138. */
  139. unsigned int reset_devices;
  140. EXPORT_SYMBOL(reset_devices);
  141. static int __init set_reset_devices(char *str)
  142. {
  143. reset_devices = 1;
  144. return 1;
  145. }
  146. __setup("reset_devices", set_reset_devices);
  147. static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
  148. const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
  149. static const char *panic_later, *panic_param;
  150. extern const struct obs_kernel_param __setup_start[], __setup_end[];
  151. static int __init obsolete_checksetup(char *line)
  152. {
  153. const struct obs_kernel_param *p;
  154. int had_early_param = 0;
  155. p = __setup_start;
  156. do {
  157. int n = strlen(p->str);
  158. if (parameqn(line, p->str, n)) {
  159. if (p->early) {
  160. /* Already done in parse_early_param?
  161. * (Needs exact match on param part).
  162. * Keep iterating, as we can have early
  163. * params and __setups of same names 8( */
  164. if (line[n] == '\0' || line[n] == '=')
  165. had_early_param = 1;
  166. } else if (!p->setup_func) {
  167. pr_warn("Parameter %s is obsolete, ignored\n",
  168. p->str);
  169. return 1;
  170. } else if (p->setup_func(line + n))
  171. return 1;
  172. }
  173. p++;
  174. } while (p < __setup_end);
  175. return had_early_param;
  176. }
  177. /*
  178. * This should be approx 2 Bo*oMips to start (note initial shift), and will
  179. * still work even if initially too large, it will just take slightly longer
  180. */
  181. unsigned long loops_per_jiffy = (1<<12);
  182. EXPORT_SYMBOL(loops_per_jiffy);
  183. static int __init debug_kernel(char *str)
  184. {
  185. console_loglevel = 10;
  186. return 0;
  187. }
  188. static int __init quiet_kernel(char *str)
  189. {
  190. console_loglevel = 4;
  191. return 0;
  192. }
  193. early_param("debug", debug_kernel);
  194. early_param("quiet", quiet_kernel);
  195. static int __init loglevel(char *str)
  196. {
  197. int newlevel;
  198. /*
  199. * Only update loglevel value when a correct setting was passed,
  200. * to prevent blind crashes (when loglevel being set to 0) that
  201. * are quite hard to debug
  202. */
  203. if (get_option(&str, &newlevel)) {
  204. console_loglevel = newlevel;
  205. return 0;
  206. }
  207. return -EINVAL;
  208. }
  209. early_param("loglevel", loglevel);
  210. /* Change NUL term back to "=", to make "param" the whole string. */
  211. static int __init repair_env_string(char *param, char *val, const char *unused)
  212. {
  213. if (val) {
  214. /* param=val or param="val"? */
  215. if (val == param+strlen(param)+1)
  216. val[-1] = '=';
  217. else if (val == param+strlen(param)+2) {
  218. val[-2] = '=';
  219. memmove(val-1, val, strlen(val)+1);
  220. val--;
  221. } else
  222. BUG();
  223. }
  224. return 0;
  225. }
  226. /*
  227. * Unknown boot options get handed to init, unless they look like
  228. * unused parameters (modprobe will find them in /proc/cmdline).
  229. */
  230. static int __init unknown_bootoption(char *param, char *val, const char *unused)
  231. {
  232. repair_env_string(param, val, unused);
  233. /* Handle obsolete-style parameters */
  234. if (obsolete_checksetup(param))
  235. return 0;
  236. /* Unused module parameter. */
  237. if (strchr(param, '.') && (!val || strchr(param, '.') < val))
  238. return 0;
  239. if (panic_later)
  240. return 0;
  241. if (val) {
  242. /* Environment option */
  243. unsigned int i;
  244. for (i = 0; envp_init[i]; i++) {
  245. if (i == MAX_INIT_ENVS) {
  246. panic_later = "Too many boot env vars at `%s'";
  247. panic_param = param;
  248. }
  249. if (!strncmp(param, envp_init[i], val - param))
  250. break;
  251. }
  252. envp_init[i] = param;
  253. } else {
  254. /* Command line option */
  255. unsigned int i;
  256. for (i = 0; argv_init[i]; i++) {
  257. if (i == MAX_INIT_ARGS) {
  258. panic_later = "Too many boot init vars at `%s'";
  259. panic_param = param;
  260. }
  261. }
  262. argv_init[i] = param;
  263. }
  264. return 0;
  265. }
  266. static int __init init_setup(char *str)
  267. {
  268. unsigned int i;
  269. execute_command = str;
  270. /*
  271. * In case LILO is going to boot us with default command line,
  272. * it prepends "auto" before the whole cmdline which makes
  273. * the shell think it should execute a script with such name.
  274. * So we ignore all arguments entered _before_ init=... [MJ]
  275. */
  276. for (i = 1; i < MAX_INIT_ARGS; i++)
  277. argv_init[i] = NULL;
  278. return 1;
  279. }
  280. __setup("init=", init_setup);
  281. static int __init rdinit_setup(char *str)
  282. {
  283. unsigned int i;
  284. ramdisk_execute_command = str;
  285. /* See "auto" comment in init_setup */
  286. for (i = 1; i < MAX_INIT_ARGS; i++)
  287. argv_init[i] = NULL;
  288. return 1;
  289. }
  290. __setup("rdinit=", rdinit_setup);
  291. #ifndef CONFIG_SMP
  292. static const unsigned int setup_max_cpus = NR_CPUS;
  293. #ifdef CONFIG_X86_LOCAL_APIC
  294. static void __init smp_init(void)
  295. {
  296. APIC_init_uniprocessor();
  297. }
  298. #else
  299. #define smp_init() do { } while (0)
  300. #endif
  301. static inline void setup_nr_cpu_ids(void) { }
  302. static inline void smp_prepare_cpus(unsigned int maxcpus) { }
  303. #endif
  304. /*
  305. * We need to store the untouched command line for future reference.
  306. * We also need to store the touched command line since the parameter
  307. * parsing is performed in place, and we should allow a component to
  308. * store reference of name/value for future reference.
  309. */
  310. static void __init setup_command_line(char *command_line)
  311. {
  312. saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
  313. static_command_line = alloc_bootmem(strlen (command_line)+1);
  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_cache_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(panic_later, panic_param);
  512. lockdep_info();
  513. /*
  514. * Need to run this when irqs are enabled, because it wants
  515. * to self-test [hard/soft]-irqs on/off lock inversion bugs
  516. * too:
  517. */
  518. locking_selftest();
  519. #ifdef CONFIG_BLK_DEV_INITRD
  520. if (initrd_start && !initrd_below_start_ok &&
  521. page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
  522. pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
  523. page_to_pfn(virt_to_page((void *)initrd_start)),
  524. min_low_pfn);
  525. initrd_start = 0;
  526. }
  527. #endif
  528. page_cgroup_init();
  529. debug_objects_mem_init();
  530. kmemleak_init();
  531. setup_per_cpu_pageset();
  532. numa_policy_init();
  533. if (late_time_init)
  534. late_time_init();
  535. sched_clock_init();
  536. calibrate_delay();
  537. pidmap_init();
  538. anon_vma_init();
  539. #ifdef CONFIG_X86
  540. if (efi_enabled(EFI_RUNTIME_SERVICES))
  541. efi_enter_virtual_mode();
  542. #endif
  543. thread_info_cache_init();
  544. cred_init();
  545. fork_init(totalram_pages);
  546. proc_caches_init();
  547. buffer_init();
  548. key_init();
  549. security_init();
  550. dbg_late_init();
  551. vfs_caches_init(totalram_pages);
  552. signals_init();
  553. /* rootfs populating might need page-writeback */
  554. page_writeback_init();
  555. #ifdef CONFIG_PROC_FS
  556. proc_root_init();
  557. #endif
  558. cgroup_init();
  559. cpuset_init();
  560. taskstats_init_early();
  561. delayacct_init();
  562. check_bugs();
  563. acpi_early_init(); /* before LAPIC and SMP init */
  564. sfi_init_late();
  565. if (efi_enabled(EFI_RUNTIME_SERVICES)) {
  566. efi_late_init();
  567. efi_free_boot_services();
  568. }
  569. ftrace_init();
  570. /* Do the rest non-__init'ed, we're now alive */
  571. rest_init();
  572. }
  573. /* Call all constructor functions linked into the kernel. */
  574. static void __init do_ctors(void)
  575. {
  576. #ifdef CONFIG_CONSTRUCTORS
  577. ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
  578. for (; fn < (ctor_fn_t *) __ctors_end; fn++)
  579. (*fn)();
  580. #endif
  581. }
  582. bool initcall_debug;
  583. core_param(initcall_debug, initcall_debug, bool, 0644);
  584. static int __init_or_module do_one_initcall_debug(initcall_t fn)
  585. {
  586. ktime_t calltime, delta, rettime;
  587. unsigned long long duration;
  588. int ret;
  589. pr_debug("calling %pF @ %i\n", fn, task_pid_nr(current));
  590. calltime = ktime_get();
  591. ret = fn();
  592. rettime = ktime_get();
  593. delta = ktime_sub(rettime, calltime);
  594. duration = (unsigned long long) ktime_to_ns(delta) >> 10;
  595. pr_debug("initcall %pF returned %d after %lld usecs\n",
  596. fn, ret, duration);
  597. return ret;
  598. }
  599. int __init_or_module do_one_initcall(initcall_t fn)
  600. {
  601. int count = preempt_count();
  602. int ret;
  603. char msgbuf[64];
  604. if (initcall_debug)
  605. ret = do_one_initcall_debug(fn);
  606. else
  607. ret = fn();
  608. msgbuf[0] = 0;
  609. if (preempt_count() != count) {
  610. sprintf(msgbuf, "preemption imbalance ");
  611. preempt_count() = count;
  612. }
  613. if (irqs_disabled()) {
  614. strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
  615. local_irq_enable();
  616. }
  617. WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
  618. return ret;
  619. }
  620. extern initcall_t __initcall_start[];
  621. extern initcall_t __initcall0_start[];
  622. extern initcall_t __initcall1_start[];
  623. extern initcall_t __initcall2_start[];
  624. extern initcall_t __initcall3_start[];
  625. extern initcall_t __initcall4_start[];
  626. extern initcall_t __initcall5_start[];
  627. extern initcall_t __initcall6_start[];
  628. extern initcall_t __initcall7_start[];
  629. extern initcall_t __initcall_end[];
  630. static initcall_t *initcall_levels[] __initdata = {
  631. __initcall0_start,
  632. __initcall1_start,
  633. __initcall2_start,
  634. __initcall3_start,
  635. __initcall4_start,
  636. __initcall5_start,
  637. __initcall6_start,
  638. __initcall7_start,
  639. __initcall_end,
  640. };
  641. /* Keep these in sync with initcalls in include/linux/init.h */
  642. static char *initcall_level_names[] __initdata = {
  643. "early",
  644. "core",
  645. "postcore",
  646. "arch",
  647. "subsys",
  648. "fs",
  649. "device",
  650. "late",
  651. };
  652. static void __init do_initcall_level(int level)
  653. {
  654. extern const struct kernel_param __start___param[], __stop___param[];
  655. initcall_t *fn;
  656. strcpy(static_command_line, saved_command_line);
  657. parse_args(initcall_level_names[level],
  658. static_command_line, __start___param,
  659. __stop___param - __start___param,
  660. level, level,
  661. &repair_env_string);
  662. for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
  663. do_one_initcall(*fn);
  664. }
  665. static void __init do_initcalls(void)
  666. {
  667. int level;
  668. for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
  669. do_initcall_level(level);
  670. }
  671. /*
  672. * Ok, the machine is now initialized. None of the devices
  673. * have been touched yet, but the CPU subsystem is up and
  674. * running, and memory and process management works.
  675. *
  676. * Now we can finally start doing some real work..
  677. */
  678. static void __init do_basic_setup(void)
  679. {
  680. cpuset_init_smp();
  681. usermodehelper_init();
  682. shmem_init();
  683. driver_init();
  684. init_irq_proc();
  685. do_ctors();
  686. usermodehelper_enable();
  687. do_initcalls();
  688. }
  689. static void __init do_pre_smp_initcalls(void)
  690. {
  691. initcall_t *fn;
  692. for (fn = __initcall_start; fn < __initcall0_start; fn++)
  693. do_one_initcall(*fn);
  694. }
  695. /*
  696. * This function requests modules which should be loaded by default and is
  697. * called twice right after initrd is mounted and right before init is
  698. * exec'd. If such modules are on either initrd or rootfs, they will be
  699. * loaded before control is passed to userland.
  700. */
  701. void __init load_default_modules(void)
  702. {
  703. load_default_elevator_module();
  704. }
  705. static int run_init_process(const char *init_filename)
  706. {
  707. argv_init[0] = init_filename;
  708. return do_execve(init_filename,
  709. (const char __user *const __user *)argv_init,
  710. (const char __user *const __user *)envp_init);
  711. }
  712. static noinline void __init kernel_init_freeable(void);
  713. static int __ref kernel_init(void *unused)
  714. {
  715. kernel_init_freeable();
  716. /* need to finish all async __init code before freeing the memory */
  717. async_synchronize_full();
  718. free_initmem();
  719. mark_rodata_ro();
  720. system_state = SYSTEM_RUNNING;
  721. numa_default_policy();
  722. flush_delayed_fput();
  723. if (ramdisk_execute_command) {
  724. if (!run_init_process(ramdisk_execute_command))
  725. return 0;
  726. pr_err("Failed to execute %s\n", ramdisk_execute_command);
  727. }
  728. /*
  729. * We try each of these until one succeeds.
  730. *
  731. * The Bourne shell can be used instead of init if we are
  732. * trying to recover a really broken machine.
  733. */
  734. if (execute_command) {
  735. if (!run_init_process(execute_command))
  736. return 0;
  737. pr_err("Failed to execute %s. Attempting defaults...\n",
  738. execute_command);
  739. }
  740. if (!run_init_process("/sbin/init") ||
  741. !run_init_process("/etc/init") ||
  742. !run_init_process("/bin/init") ||
  743. !run_init_process("/bin/sh"))
  744. return 0;
  745. panic("No init found. Try passing init= option to kernel. "
  746. "See Linux Documentation/init.txt for guidance.");
  747. }
  748. static noinline void __init kernel_init_freeable(void)
  749. {
  750. /*
  751. * Wait until kthreadd is all set-up.
  752. */
  753. wait_for_completion(&kthreadd_done);
  754. /* Now the scheduler is fully set up and can do blocking allocations */
  755. gfp_allowed_mask = __GFP_BITS_MASK;
  756. /*
  757. * init can allocate pages on any node
  758. */
  759. set_mems_allowed(node_states[N_MEMORY]);
  760. /*
  761. * init can run on any cpu.
  762. */
  763. set_cpus_allowed_ptr(current, cpu_all_mask);
  764. cad_pid = task_pid(current);
  765. smp_prepare_cpus(setup_max_cpus);
  766. do_pre_smp_initcalls();
  767. lockup_detector_init();
  768. smp_init();
  769. sched_init_smp();
  770. do_basic_setup();
  771. /* Open the /dev/console on the rootfs, this should never fail */
  772. if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
  773. pr_err("Warning: unable to open an initial console.\n");
  774. (void) sys_dup(0);
  775. (void) sys_dup(0);
  776. /*
  777. * check if there is an early userspace init. If yes, let it do all
  778. * the work
  779. */
  780. if (!ramdisk_execute_command)
  781. ramdisk_execute_command = "/init";
  782. if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
  783. ramdisk_execute_command = NULL;
  784. prepare_namespace();
  785. }
  786. /*
  787. * Ok, we have completed the initial bootup, and
  788. * we're essentially up and running. Get rid of the
  789. * initmem segments and start the user-mode stuff..
  790. */
  791. /* rootfs is available now, try loading default modules */
  792. load_default_modules();
  793. }