main.c 22 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914
  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. * If set, this is an indication to the drivers that reset the underlying
  127. * device before going ahead with the initialization otherwise driver might
  128. * rely on the BIOS and skip the reset operation.
  129. *
  130. * This is useful if kernel is booting in an unreliable environment.
  131. * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
  132. * skipped and devices will be in unknown state.
  133. */
  134. unsigned int reset_devices;
  135. EXPORT_SYMBOL(reset_devices);
  136. static int __init set_reset_devices(char *str)
  137. {
  138. reset_devices = 1;
  139. return 1;
  140. }
  141. __setup("reset_devices", set_reset_devices);
  142. static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
  143. const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
  144. static const char *panic_later, *panic_param;
  145. extern const struct obs_kernel_param __setup_start[], __setup_end[];
  146. static int __init obsolete_checksetup(char *line)
  147. {
  148. const struct obs_kernel_param *p;
  149. int had_early_param = 0;
  150. p = __setup_start;
  151. do {
  152. int n = strlen(p->str);
  153. if (parameqn(line, p->str, n)) {
  154. if (p->early) {
  155. /* Already done in parse_early_param?
  156. * (Needs exact match on param part).
  157. * Keep iterating, as we can have early
  158. * params and __setups of same names 8( */
  159. if (line[n] == '\0' || line[n] == '=')
  160. had_early_param = 1;
  161. } else if (!p->setup_func) {
  162. pr_warn("Parameter %s is obsolete, ignored\n",
  163. p->str);
  164. return 1;
  165. } else if (p->setup_func(line + n))
  166. return 1;
  167. }
  168. p++;
  169. } while (p < __setup_end);
  170. return had_early_param;
  171. }
  172. /*
  173. * This should be approx 2 Bo*oMips to start (note initial shift), and will
  174. * still work even if initially too large, it will just take slightly longer
  175. */
  176. unsigned long loops_per_jiffy = (1<<12);
  177. EXPORT_SYMBOL(loops_per_jiffy);
  178. static int __init debug_kernel(char *str)
  179. {
  180. console_loglevel = 10;
  181. return 0;
  182. }
  183. static int __init quiet_kernel(char *str)
  184. {
  185. console_loglevel = 4;
  186. return 0;
  187. }
  188. early_param("debug", debug_kernel);
  189. early_param("quiet", quiet_kernel);
  190. static int __init loglevel(char *str)
  191. {
  192. int newlevel;
  193. /*
  194. * Only update loglevel value when a correct setting was passed,
  195. * to prevent blind crashes (when loglevel being set to 0) that
  196. * are quite hard to debug
  197. */
  198. if (get_option(&str, &newlevel)) {
  199. console_loglevel = newlevel;
  200. return 0;
  201. }
  202. return -EINVAL;
  203. }
  204. early_param("loglevel", loglevel);
  205. /* Change NUL term back to "=", to make "param" the whole string. */
  206. static int __init repair_env_string(char *param, char *val, const char *unused)
  207. {
  208. if (val) {
  209. /* param=val or param="val"? */
  210. if (val == param+strlen(param)+1)
  211. val[-1] = '=';
  212. else if (val == param+strlen(param)+2) {
  213. val[-2] = '=';
  214. memmove(val-1, val, strlen(val)+1);
  215. val--;
  216. } else
  217. BUG();
  218. }
  219. return 0;
  220. }
  221. /*
  222. * Unknown boot options get handed to init, unless they look like
  223. * unused parameters (modprobe will find them in /proc/cmdline).
  224. */
  225. static int __init unknown_bootoption(char *param, char *val, const char *unused)
  226. {
  227. repair_env_string(param, val, unused);
  228. /* Handle obsolete-style parameters */
  229. if (obsolete_checksetup(param))
  230. return 0;
  231. /* Unused module parameter. */
  232. if (strchr(param, '.') && (!val || strchr(param, '.') < val))
  233. return 0;
  234. if (panic_later)
  235. return 0;
  236. if (val) {
  237. /* Environment option */
  238. unsigned int i;
  239. for (i = 0; envp_init[i]; i++) {
  240. if (i == MAX_INIT_ENVS) {
  241. panic_later = "Too many boot env vars at `%s'";
  242. panic_param = param;
  243. }
  244. if (!strncmp(param, envp_init[i], val - param))
  245. break;
  246. }
  247. envp_init[i] = param;
  248. } else {
  249. /* Command line option */
  250. unsigned int i;
  251. for (i = 0; argv_init[i]; i++) {
  252. if (i == MAX_INIT_ARGS) {
  253. panic_later = "Too many boot init vars at `%s'";
  254. panic_param = param;
  255. }
  256. }
  257. argv_init[i] = param;
  258. }
  259. return 0;
  260. }
  261. static int __init init_setup(char *str)
  262. {
  263. unsigned int i;
  264. execute_command = str;
  265. /*
  266. * In case LILO is going to boot us with default command line,
  267. * it prepends "auto" before the whole cmdline which makes
  268. * the shell think it should execute a script with such name.
  269. * So we ignore all arguments entered _before_ init=... [MJ]
  270. */
  271. for (i = 1; i < MAX_INIT_ARGS; i++)
  272. argv_init[i] = NULL;
  273. return 1;
  274. }
  275. __setup("init=", init_setup);
  276. static int __init rdinit_setup(char *str)
  277. {
  278. unsigned int i;
  279. ramdisk_execute_command = str;
  280. /* See "auto" comment in init_setup */
  281. for (i = 1; i < MAX_INIT_ARGS; i++)
  282. argv_init[i] = NULL;
  283. return 1;
  284. }
  285. __setup("rdinit=", rdinit_setup);
  286. #ifndef CONFIG_SMP
  287. static const unsigned int setup_max_cpus = NR_CPUS;
  288. #ifdef CONFIG_X86_LOCAL_APIC
  289. static void __init smp_init(void)
  290. {
  291. APIC_init_uniprocessor();
  292. }
  293. #else
  294. #define smp_init() do { } while (0)
  295. #endif
  296. static inline void setup_nr_cpu_ids(void) { }
  297. static inline void smp_prepare_cpus(unsigned int maxcpus) { }
  298. #endif
  299. /*
  300. * We need to store the untouched command line for future reference.
  301. * We also need to store the touched command line since the parameter
  302. * parsing is performed in place, and we should allow a component to
  303. * store reference of name/value for future reference.
  304. */
  305. static void __init setup_command_line(char *command_line)
  306. {
  307. saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
  308. static_command_line = alloc_bootmem(strlen (command_line)+1);
  309. strcpy (saved_command_line, boot_command_line);
  310. strcpy (static_command_line, command_line);
  311. }
  312. /*
  313. * We need to finalize in a non-__init function or else race conditions
  314. * between the root thread and the init thread may cause start_kernel to
  315. * be reaped by free_initmem before the root thread has proceeded to
  316. * cpu_idle.
  317. *
  318. * gcc-3.4 accidentally inlines this function, so use noinline.
  319. */
  320. static __initdata DECLARE_COMPLETION(kthreadd_done);
  321. static noinline void __init_refok rest_init(void)
  322. {
  323. int pid;
  324. rcu_scheduler_starting();
  325. /*
  326. * We need to spawn init first so that it obtains pid 1, however
  327. * the init task will end up wanting to create kthreads, which, if
  328. * we schedule it before we create kthreadd, will OOPS.
  329. */
  330. kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
  331. numa_default_policy();
  332. pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
  333. rcu_read_lock();
  334. kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
  335. rcu_read_unlock();
  336. complete(&kthreadd_done);
  337. /*
  338. * The boot idle thread must execute schedule()
  339. * at least once to get things moving:
  340. */
  341. init_idle_bootup_task(current);
  342. schedule_preempt_disabled();
  343. /* Call into cpu_idle with preempt disabled */
  344. cpu_startup_entry(CPUHP_ONLINE);
  345. }
  346. /* Check for early params. */
  347. static int __init do_early_param(char *param, char *val, const char *unused)
  348. {
  349. const struct obs_kernel_param *p;
  350. for (p = __setup_start; p < __setup_end; p++) {
  351. if ((p->early && parameq(param, p->str)) ||
  352. (strcmp(param, "console") == 0 &&
  353. strcmp(p->str, "earlycon") == 0)
  354. ) {
  355. if (p->setup_func(val) != 0)
  356. pr_warn("Malformed early option '%s'\n", param);
  357. }
  358. }
  359. /* We accept everything at this stage. */
  360. return 0;
  361. }
  362. void __init parse_early_options(char *cmdline)
  363. {
  364. parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
  365. }
  366. /* Arch code calls this early on, or if not, just before other parsing. */
  367. void __init parse_early_param(void)
  368. {
  369. static __initdata int done = 0;
  370. static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
  371. if (done)
  372. return;
  373. /* All fall through to do_early_param. */
  374. strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
  375. parse_early_options(tmp_cmdline);
  376. done = 1;
  377. }
  378. /*
  379. * Activate the first processor.
  380. */
  381. static void __init boot_cpu_init(void)
  382. {
  383. int cpu = smp_processor_id();
  384. /* Mark the boot cpu "present", "online" etc for SMP and UP case */
  385. set_cpu_online(cpu, true);
  386. set_cpu_active(cpu, true);
  387. set_cpu_present(cpu, true);
  388. set_cpu_possible(cpu, true);
  389. }
  390. void __init __weak smp_setup_processor_id(void)
  391. {
  392. }
  393. # if THREAD_SIZE >= PAGE_SIZE
  394. void __init __weak thread_info_cache_init(void)
  395. {
  396. }
  397. #endif
  398. /*
  399. * Set up kernel memory allocators
  400. */
  401. static void __init mm_init(void)
  402. {
  403. /*
  404. * page_cgroup requires contiguous pages,
  405. * bigger than MAX_ORDER unless SPARSEMEM.
  406. */
  407. page_cgroup_init_flatmem();
  408. mem_init();
  409. kmem_cache_init();
  410. percpu_init_late();
  411. pgtable_cache_init();
  412. vmalloc_init();
  413. }
  414. asmlinkage void __init start_kernel(void)
  415. {
  416. char * command_line;
  417. extern const struct kernel_param __start___param[], __stop___param[];
  418. /*
  419. * Need to run as early as possible, to initialize the
  420. * lockdep hash:
  421. */
  422. lockdep_init();
  423. smp_setup_processor_id();
  424. debug_objects_early_init();
  425. /*
  426. * Set up the the initial canary ASAP:
  427. */
  428. boot_init_stack_canary();
  429. cgroup_init_early();
  430. local_irq_disable();
  431. early_boot_irqs_disabled = true;
  432. /*
  433. * Interrupts are still disabled. Do necessary setups, then
  434. * enable them
  435. */
  436. boot_cpu_init();
  437. page_address_init();
  438. pr_notice("%s", linux_banner);
  439. setup_arch(&command_line);
  440. mm_init_owner(&init_mm, &init_task);
  441. mm_init_cpumask(&init_mm);
  442. setup_command_line(command_line);
  443. setup_nr_cpu_ids();
  444. setup_per_cpu_areas();
  445. smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
  446. build_all_zonelists(NULL, NULL);
  447. page_alloc_init();
  448. pr_notice("Kernel command line: %s\n", boot_command_line);
  449. parse_early_param();
  450. parse_args("Booting kernel", static_command_line, __start___param,
  451. __stop___param - __start___param,
  452. -1, -1, &unknown_bootoption);
  453. jump_label_init();
  454. /*
  455. * These use large bootmem allocations and must precede
  456. * kmem_cache_init()
  457. */
  458. setup_log_buf(0);
  459. pidhash_init();
  460. vfs_caches_init_early();
  461. sort_main_extable();
  462. trap_init();
  463. mm_init();
  464. /*
  465. * Set up the scheduler prior starting any interrupts (such as the
  466. * timer interrupt). Full topology setup happens at smp_init()
  467. * time - but meanwhile we still have a functioning scheduler.
  468. */
  469. sched_init();
  470. /*
  471. * Disable preemption - early bootup scheduling is extremely
  472. * fragile until we cpu_idle() for the first time.
  473. */
  474. preempt_disable();
  475. if (WARN(!irqs_disabled(), "Interrupts were enabled *very* early, fixing it\n"))
  476. local_irq_disable();
  477. idr_init_cache();
  478. rcu_init();
  479. tick_nohz_init();
  480. context_tracking_init();
  481. radix_tree_init();
  482. /* init some links before init_ISA_irqs() */
  483. early_irq_init();
  484. init_IRQ();
  485. tick_init();
  486. init_timers();
  487. hrtimers_init();
  488. softirq_init();
  489. timekeeping_init();
  490. time_init();
  491. sched_clock_postinit();
  492. perf_event_init();
  493. profile_init();
  494. call_function_init();
  495. WARN(!irqs_disabled(), "Interrupts were enabled early\n");
  496. early_boot_irqs_disabled = false;
  497. local_irq_enable();
  498. kmem_cache_init_late();
  499. /*
  500. * HACK ALERT! This is early. We're enabling the console before
  501. * we've done PCI setups etc, and console_init() must be aware of
  502. * this. But we do want output early, in case something goes wrong.
  503. */
  504. console_init();
  505. if (panic_later)
  506. panic(panic_later, panic_param);
  507. lockdep_info();
  508. /*
  509. * Need to run this when irqs are enabled, because it wants
  510. * to self-test [hard/soft]-irqs on/off lock inversion bugs
  511. * too:
  512. */
  513. locking_selftest();
  514. #ifdef CONFIG_BLK_DEV_INITRD
  515. if (initrd_start && !initrd_below_start_ok &&
  516. page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
  517. pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
  518. page_to_pfn(virt_to_page((void *)initrd_start)),
  519. min_low_pfn);
  520. initrd_start = 0;
  521. }
  522. #endif
  523. page_cgroup_init();
  524. debug_objects_mem_init();
  525. kmemleak_init();
  526. setup_per_cpu_pageset();
  527. numa_policy_init();
  528. if (late_time_init)
  529. late_time_init();
  530. sched_clock_init();
  531. calibrate_delay();
  532. pidmap_init();
  533. anon_vma_init();
  534. #ifdef CONFIG_X86
  535. if (efi_enabled(EFI_RUNTIME_SERVICES))
  536. efi_enter_virtual_mode();
  537. #endif
  538. thread_info_cache_init();
  539. cred_init();
  540. fork_init(totalram_pages);
  541. proc_caches_init();
  542. buffer_init();
  543. key_init();
  544. security_init();
  545. dbg_late_init();
  546. vfs_caches_init(totalram_pages);
  547. signals_init();
  548. /* rootfs populating might need page-writeback */
  549. page_writeback_init();
  550. #ifdef CONFIG_PROC_FS
  551. proc_root_init();
  552. #endif
  553. cgroup_init();
  554. cpuset_init();
  555. taskstats_init_early();
  556. delayacct_init();
  557. check_bugs();
  558. acpi_early_init(); /* before LAPIC and SMP init */
  559. sfi_init_late();
  560. if (efi_enabled(EFI_RUNTIME_SERVICES)) {
  561. efi_late_init();
  562. efi_free_boot_services();
  563. }
  564. ftrace_init();
  565. /* Do the rest non-__init'ed, we're now alive */
  566. rest_init();
  567. }
  568. /* Call all constructor functions linked into the kernel. */
  569. static void __init do_ctors(void)
  570. {
  571. #ifdef CONFIG_CONSTRUCTORS
  572. ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
  573. for (; fn < (ctor_fn_t *) __ctors_end; fn++)
  574. (*fn)();
  575. #endif
  576. }
  577. bool initcall_debug;
  578. core_param(initcall_debug, initcall_debug, bool, 0644);
  579. static int __init_or_module do_one_initcall_debug(initcall_t fn)
  580. {
  581. ktime_t calltime, delta, rettime;
  582. unsigned long long duration;
  583. int ret;
  584. pr_debug("calling %pF @ %i\n", fn, task_pid_nr(current));
  585. calltime = ktime_get();
  586. ret = fn();
  587. rettime = ktime_get();
  588. delta = ktime_sub(rettime, calltime);
  589. duration = (unsigned long long) ktime_to_ns(delta) >> 10;
  590. pr_debug("initcall %pF returned %d after %lld usecs\n",
  591. fn, ret, duration);
  592. return ret;
  593. }
  594. int __init_or_module do_one_initcall(initcall_t fn)
  595. {
  596. int count = preempt_count();
  597. int ret;
  598. char msgbuf[64];
  599. if (initcall_debug)
  600. ret = do_one_initcall_debug(fn);
  601. else
  602. ret = fn();
  603. msgbuf[0] = 0;
  604. if (preempt_count() != count) {
  605. sprintf(msgbuf, "preemption imbalance ");
  606. preempt_count() = count;
  607. }
  608. if (irqs_disabled()) {
  609. strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
  610. local_irq_enable();
  611. }
  612. WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
  613. return ret;
  614. }
  615. extern initcall_t __initcall_start[];
  616. extern initcall_t __initcall0_start[];
  617. extern initcall_t __initcall1_start[];
  618. extern initcall_t __initcall2_start[];
  619. extern initcall_t __initcall3_start[];
  620. extern initcall_t __initcall4_start[];
  621. extern initcall_t __initcall5_start[];
  622. extern initcall_t __initcall6_start[];
  623. extern initcall_t __initcall7_start[];
  624. extern initcall_t __initcall_end[];
  625. static initcall_t *initcall_levels[] __initdata = {
  626. __initcall0_start,
  627. __initcall1_start,
  628. __initcall2_start,
  629. __initcall3_start,
  630. __initcall4_start,
  631. __initcall5_start,
  632. __initcall6_start,
  633. __initcall7_start,
  634. __initcall_end,
  635. };
  636. /* Keep these in sync with initcalls in include/linux/init.h */
  637. static char *initcall_level_names[] __initdata = {
  638. "early",
  639. "core",
  640. "postcore",
  641. "arch",
  642. "subsys",
  643. "fs",
  644. "device",
  645. "late",
  646. };
  647. static void __init do_initcall_level(int level)
  648. {
  649. extern const struct kernel_param __start___param[], __stop___param[];
  650. initcall_t *fn;
  651. strcpy(static_command_line, saved_command_line);
  652. parse_args(initcall_level_names[level],
  653. static_command_line, __start___param,
  654. __stop___param - __start___param,
  655. level, level,
  656. &repair_env_string);
  657. for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
  658. do_one_initcall(*fn);
  659. }
  660. static void __init do_initcalls(void)
  661. {
  662. int level;
  663. for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
  664. do_initcall_level(level);
  665. }
  666. /*
  667. * Ok, the machine is now initialized. None of the devices
  668. * have been touched yet, but the CPU subsystem is up and
  669. * running, and memory and process management works.
  670. *
  671. * Now we can finally start doing some real work..
  672. */
  673. static void __init do_basic_setup(void)
  674. {
  675. cpuset_init_smp();
  676. usermodehelper_init();
  677. shmem_init();
  678. driver_init();
  679. init_irq_proc();
  680. do_ctors();
  681. usermodehelper_enable();
  682. do_initcalls();
  683. random_int_secret_init();
  684. }
  685. static void __init do_pre_smp_initcalls(void)
  686. {
  687. initcall_t *fn;
  688. for (fn = __initcall_start; fn < __initcall0_start; fn++)
  689. do_one_initcall(*fn);
  690. }
  691. /*
  692. * This function requests modules which should be loaded by default and is
  693. * called twice right after initrd is mounted and right before init is
  694. * exec'd. If such modules are on either initrd or rootfs, they will be
  695. * loaded before control is passed to userland.
  696. */
  697. void __init load_default_modules(void)
  698. {
  699. load_default_elevator_module();
  700. }
  701. static int run_init_process(const char *init_filename)
  702. {
  703. argv_init[0] = init_filename;
  704. return do_execve(init_filename,
  705. (const char __user *const __user *)argv_init,
  706. (const char __user *const __user *)envp_init);
  707. }
  708. static noinline void __init kernel_init_freeable(void);
  709. static int __ref kernel_init(void *unused)
  710. {
  711. kernel_init_freeable();
  712. /* need to finish all async __init code before freeing the memory */
  713. async_synchronize_full();
  714. free_initmem();
  715. mark_rodata_ro();
  716. system_state = SYSTEM_RUNNING;
  717. numa_default_policy();
  718. flush_delayed_fput();
  719. if (ramdisk_execute_command) {
  720. if (!run_init_process(ramdisk_execute_command))
  721. return 0;
  722. pr_err("Failed to execute %s\n", ramdisk_execute_command);
  723. }
  724. /*
  725. * We try each of these until one succeeds.
  726. *
  727. * The Bourne shell can be used instead of init if we are
  728. * trying to recover a really broken machine.
  729. */
  730. if (execute_command) {
  731. if (!run_init_process(execute_command))
  732. return 0;
  733. pr_err("Failed to execute %s. Attempting defaults...\n",
  734. execute_command);
  735. }
  736. if (!run_init_process("/sbin/init") ||
  737. !run_init_process("/etc/init") ||
  738. !run_init_process("/bin/init") ||
  739. !run_init_process("/bin/sh"))
  740. return 0;
  741. panic("No init found. Try passing init= option to kernel. "
  742. "See Linux Documentation/init.txt for guidance.");
  743. }
  744. static noinline void __init kernel_init_freeable(void)
  745. {
  746. /*
  747. * Wait until kthreadd is all set-up.
  748. */
  749. wait_for_completion(&kthreadd_done);
  750. /* Now the scheduler is fully set up and can do blocking allocations */
  751. gfp_allowed_mask = __GFP_BITS_MASK;
  752. /*
  753. * init can allocate pages on any node
  754. */
  755. set_mems_allowed(node_states[N_MEMORY]);
  756. /*
  757. * init can run on any cpu.
  758. */
  759. set_cpus_allowed_ptr(current, cpu_all_mask);
  760. cad_pid = task_pid(current);
  761. smp_prepare_cpus(setup_max_cpus);
  762. do_pre_smp_initcalls();
  763. lockup_detector_init();
  764. smp_init();
  765. sched_init_smp();
  766. do_basic_setup();
  767. /* Open the /dev/console on the rootfs, this should never fail */
  768. if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
  769. pr_err("Warning: unable to open an initial console.\n");
  770. (void) sys_dup(0);
  771. (void) sys_dup(0);
  772. /*
  773. * check if there is an early userspace init. If yes, let it do all
  774. * the work
  775. */
  776. if (!ramdisk_execute_command)
  777. ramdisk_execute_command = "/init";
  778. if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
  779. ramdisk_execute_command = NULL;
  780. prepare_namespace();
  781. }
  782. /*
  783. * Ok, we have completed the initial bootup, and
  784. * we're essentially up and running. Get rid of the
  785. * initmem segments and start the user-mode stuff..
  786. */
  787. /* rootfs is available now, try loading default modules */
  788. load_default_modules();
  789. }