main.c 22 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910
  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 <asm/io.h>
  77. #include <asm/bugs.h>
  78. #include <asm/setup.h>
  79. #include <asm/sections.h>
  80. #include <asm/cacheflush.h>
  81. #ifdef CONFIG_X86_LOCAL_APIC
  82. #include <asm/smp.h>
  83. #endif
  84. static int kernel_init(void *);
  85. extern void init_IRQ(void);
  86. extern void fork_init(unsigned long);
  87. extern void mca_init(void);
  88. extern void sbus_init(void);
  89. extern void radix_tree_init(void);
  90. #ifndef CONFIG_DEBUG_RODATA
  91. static inline void mark_rodata_ro(void) { }
  92. #endif
  93. #ifdef CONFIG_TC
  94. extern void tc_init(void);
  95. #endif
  96. /*
  97. * Debug helper: via this flag we know that we are in 'early bootup code'
  98. * where only the boot processor is running with IRQ disabled. This means
  99. * two things - IRQ must not be enabled before the flag is cleared and some
  100. * operations which are not allowed with IRQ disabled are allowed while the
  101. * flag is set.
  102. */
  103. bool early_boot_irqs_disabled __read_mostly;
  104. enum system_states system_state __read_mostly;
  105. EXPORT_SYMBOL(system_state);
  106. /*
  107. * Boot command-line arguments
  108. */
  109. #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
  110. #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
  111. extern void time_init(void);
  112. /* Default late time init is NULL. archs can override this later. */
  113. void (*__initdata late_time_init)(void);
  114. extern void softirq_init(void);
  115. /* Untouched command line saved by arch-specific code. */
  116. char __initdata boot_command_line[COMMAND_LINE_SIZE];
  117. /* Untouched saved command line (eg. for /proc) */
  118. char *saved_command_line;
  119. /* Command line for parameter parsing */
  120. static char *static_command_line;
  121. static char *execute_command;
  122. static char *ramdisk_execute_command;
  123. /*
  124. * If set, this is an indication to the drivers that reset the underlying
  125. * device before going ahead with the initialization otherwise driver might
  126. * rely on the BIOS and skip the reset operation.
  127. *
  128. * This is useful if kernel is booting in an unreliable environment.
  129. * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
  130. * skipped and devices will be in unknown state.
  131. */
  132. unsigned int reset_devices;
  133. EXPORT_SYMBOL(reset_devices);
  134. static int __init set_reset_devices(char *str)
  135. {
  136. reset_devices = 1;
  137. return 1;
  138. }
  139. __setup("reset_devices", set_reset_devices);
  140. static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
  141. const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
  142. static const char *panic_later, *panic_param;
  143. extern const struct obs_kernel_param __setup_start[], __setup_end[];
  144. static int __init obsolete_checksetup(char *line)
  145. {
  146. const struct obs_kernel_param *p;
  147. int had_early_param = 0;
  148. p = __setup_start;
  149. do {
  150. int n = strlen(p->str);
  151. if (parameqn(line, p->str, n)) {
  152. if (p->early) {
  153. /* Already done in parse_early_param?
  154. * (Needs exact match on param part).
  155. * Keep iterating, as we can have early
  156. * params and __setups of same names 8( */
  157. if (line[n] == '\0' || line[n] == '=')
  158. had_early_param = 1;
  159. } else if (!p->setup_func) {
  160. pr_warn("Parameter %s is obsolete, ignored\n",
  161. p->str);
  162. return 1;
  163. } else if (p->setup_func(line + n))
  164. return 1;
  165. }
  166. p++;
  167. } while (p < __setup_end);
  168. return had_early_param;
  169. }
  170. /*
  171. * This should be approx 2 Bo*oMips to start (note initial shift), and will
  172. * still work even if initially too large, it will just take slightly longer
  173. */
  174. unsigned long loops_per_jiffy = (1<<12);
  175. EXPORT_SYMBOL(loops_per_jiffy);
  176. static int __init debug_kernel(char *str)
  177. {
  178. console_loglevel = 10;
  179. return 0;
  180. }
  181. static int __init quiet_kernel(char *str)
  182. {
  183. console_loglevel = 4;
  184. return 0;
  185. }
  186. early_param("debug", debug_kernel);
  187. early_param("quiet", quiet_kernel);
  188. static int __init loglevel(char *str)
  189. {
  190. int newlevel;
  191. /*
  192. * Only update loglevel value when a correct setting was passed,
  193. * to prevent blind crashes (when loglevel being set to 0) that
  194. * are quite hard to debug
  195. */
  196. if (get_option(&str, &newlevel)) {
  197. console_loglevel = newlevel;
  198. return 0;
  199. }
  200. return -EINVAL;
  201. }
  202. early_param("loglevel", loglevel);
  203. /* Change NUL term back to "=", to make "param" the whole string. */
  204. static int __init repair_env_string(char *param, char *val, const char *unused)
  205. {
  206. if (val) {
  207. /* param=val or param="val"? */
  208. if (val == param+strlen(param)+1)
  209. val[-1] = '=';
  210. else if (val == param+strlen(param)+2) {
  211. val[-2] = '=';
  212. memmove(val-1, val, strlen(val)+1);
  213. val--;
  214. } else
  215. BUG();
  216. }
  217. return 0;
  218. }
  219. /*
  220. * Unknown boot options get handed to init, unless they look like
  221. * unused parameters (modprobe will find them in /proc/cmdline).
  222. */
  223. static int __init unknown_bootoption(char *param, char *val, const char *unused)
  224. {
  225. repair_env_string(param, val, unused);
  226. /* Handle obsolete-style parameters */
  227. if (obsolete_checksetup(param))
  228. return 0;
  229. /* Unused module parameter. */
  230. if (strchr(param, '.') && (!val || strchr(param, '.') < val))
  231. return 0;
  232. if (panic_later)
  233. return 0;
  234. if (val) {
  235. /* Environment option */
  236. unsigned int i;
  237. for (i = 0; envp_init[i]; i++) {
  238. if (i == MAX_INIT_ENVS) {
  239. panic_later = "Too many boot env vars at `%s'";
  240. panic_param = param;
  241. }
  242. if (!strncmp(param, envp_init[i], val - param))
  243. break;
  244. }
  245. envp_init[i] = param;
  246. } else {
  247. /* Command line option */
  248. unsigned int i;
  249. for (i = 0; argv_init[i]; i++) {
  250. if (i == MAX_INIT_ARGS) {
  251. panic_later = "Too many boot init vars at `%s'";
  252. panic_param = param;
  253. }
  254. }
  255. argv_init[i] = param;
  256. }
  257. return 0;
  258. }
  259. static int __init init_setup(char *str)
  260. {
  261. unsigned int i;
  262. execute_command = str;
  263. /*
  264. * In case LILO is going to boot us with default command line,
  265. * it prepends "auto" before the whole cmdline which makes
  266. * the shell think it should execute a script with such name.
  267. * So we ignore all arguments entered _before_ init=... [MJ]
  268. */
  269. for (i = 1; i < MAX_INIT_ARGS; i++)
  270. argv_init[i] = NULL;
  271. return 1;
  272. }
  273. __setup("init=", init_setup);
  274. static int __init rdinit_setup(char *str)
  275. {
  276. unsigned int i;
  277. ramdisk_execute_command = str;
  278. /* See "auto" comment in init_setup */
  279. for (i = 1; i < MAX_INIT_ARGS; i++)
  280. argv_init[i] = NULL;
  281. return 1;
  282. }
  283. __setup("rdinit=", rdinit_setup);
  284. #ifndef CONFIG_SMP
  285. static const unsigned int setup_max_cpus = NR_CPUS;
  286. #ifdef CONFIG_X86_LOCAL_APIC
  287. static void __init smp_init(void)
  288. {
  289. APIC_init_uniprocessor();
  290. }
  291. #else
  292. #define smp_init() do { } while (0)
  293. #endif
  294. static inline void setup_nr_cpu_ids(void) { }
  295. static inline void smp_prepare_cpus(unsigned int maxcpus) { }
  296. #endif
  297. /*
  298. * We need to store the untouched command line for future reference.
  299. * We also need to store the touched command line since the parameter
  300. * parsing is performed in place, and we should allow a component to
  301. * store reference of name/value for future reference.
  302. */
  303. static void __init setup_command_line(char *command_line)
  304. {
  305. saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
  306. static_command_line = alloc_bootmem(strlen (command_line)+1);
  307. strcpy (saved_command_line, boot_command_line);
  308. strcpy (static_command_line, command_line);
  309. }
  310. /*
  311. * We need to finalize in a non-__init function or else race conditions
  312. * between the root thread and the init thread may cause start_kernel to
  313. * be reaped by free_initmem before the root thread has proceeded to
  314. * cpu_idle.
  315. *
  316. * gcc-3.4 accidentally inlines this function, so use noinline.
  317. */
  318. static __initdata DECLARE_COMPLETION(kthreadd_done);
  319. static noinline void __init_refok rest_init(void)
  320. {
  321. int pid;
  322. rcu_scheduler_starting();
  323. /*
  324. * We need to spawn init first so that it obtains pid 1, however
  325. * the init task will end up wanting to create kthreads, which, if
  326. * we schedule it before we create kthreadd, will OOPS.
  327. */
  328. kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
  329. numa_default_policy();
  330. pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
  331. rcu_read_lock();
  332. kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
  333. rcu_read_unlock();
  334. complete(&kthreadd_done);
  335. /*
  336. * The boot idle thread must execute schedule()
  337. * at least once to get things moving:
  338. */
  339. init_idle_bootup_task(current);
  340. schedule_preempt_disabled();
  341. /* Call into cpu_idle with preempt disabled */
  342. cpu_startup_entry(CPUHP_ONLINE);
  343. }
  344. /* Check for early params. */
  345. static int __init do_early_param(char *param, char *val, const char *unused)
  346. {
  347. const struct obs_kernel_param *p;
  348. for (p = __setup_start; p < __setup_end; p++) {
  349. if ((p->early && parameq(param, p->str)) ||
  350. (strcmp(param, "console") == 0 &&
  351. strcmp(p->str, "earlycon") == 0)
  352. ) {
  353. if (p->setup_func(val) != 0)
  354. pr_warn("Malformed early option '%s'\n", param);
  355. }
  356. }
  357. /* We accept everything at this stage. */
  358. return 0;
  359. }
  360. void __init parse_early_options(char *cmdline)
  361. {
  362. parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
  363. }
  364. /* Arch code calls this early on, or if not, just before other parsing. */
  365. void __init parse_early_param(void)
  366. {
  367. static __initdata int done = 0;
  368. static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
  369. if (done)
  370. return;
  371. /* All fall through to do_early_param. */
  372. strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
  373. parse_early_options(tmp_cmdline);
  374. done = 1;
  375. }
  376. /*
  377. * Activate the first processor.
  378. */
  379. static void __init boot_cpu_init(void)
  380. {
  381. int cpu = smp_processor_id();
  382. /* Mark the boot cpu "present", "online" etc for SMP and UP case */
  383. set_cpu_online(cpu, true);
  384. set_cpu_active(cpu, true);
  385. set_cpu_present(cpu, true);
  386. set_cpu_possible(cpu, true);
  387. }
  388. void __init __weak smp_setup_processor_id(void)
  389. {
  390. }
  391. # if THREAD_SIZE >= PAGE_SIZE
  392. void __init __weak thread_info_cache_init(void)
  393. {
  394. }
  395. #endif
  396. /*
  397. * Set up kernel memory allocators
  398. */
  399. static void __init mm_init(void)
  400. {
  401. /*
  402. * page_cgroup requires contiguous pages,
  403. * bigger than MAX_ORDER unless SPARSEMEM.
  404. */
  405. page_cgroup_init_flatmem();
  406. mem_init();
  407. kmem_cache_init();
  408. percpu_init_late();
  409. pgtable_cache_init();
  410. vmalloc_init();
  411. }
  412. asmlinkage void __init start_kernel(void)
  413. {
  414. char * command_line;
  415. extern const struct kernel_param __start___param[], __stop___param[];
  416. /*
  417. * Need to run as early as possible, to initialize the
  418. * lockdep hash:
  419. */
  420. lockdep_init();
  421. smp_setup_processor_id();
  422. debug_objects_early_init();
  423. /*
  424. * Set up the the initial canary ASAP:
  425. */
  426. boot_init_stack_canary();
  427. cgroup_init_early();
  428. local_irq_disable();
  429. early_boot_irqs_disabled = true;
  430. /*
  431. * Interrupts are still disabled. Do necessary setups, then
  432. * enable them
  433. */
  434. boot_cpu_init();
  435. page_address_init();
  436. pr_notice("%s", linux_banner);
  437. setup_arch(&command_line);
  438. mm_init_owner(&init_mm, &init_task);
  439. mm_init_cpumask(&init_mm);
  440. setup_command_line(command_line);
  441. setup_nr_cpu_ids();
  442. setup_per_cpu_areas();
  443. smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
  444. build_all_zonelists(NULL, NULL);
  445. page_alloc_init();
  446. pr_notice("Kernel command line: %s\n", boot_command_line);
  447. parse_early_param();
  448. parse_args("Booting kernel", static_command_line, __start___param,
  449. __stop___param - __start___param,
  450. -1, -1, &unknown_bootoption);
  451. jump_label_init();
  452. /*
  453. * These use large bootmem allocations and must precede
  454. * kmem_cache_init()
  455. */
  456. setup_log_buf(0);
  457. pidhash_init();
  458. vfs_caches_init_early();
  459. sort_main_extable();
  460. trap_init();
  461. mm_init();
  462. /*
  463. * Set up the scheduler prior starting any interrupts (such as the
  464. * timer interrupt). Full topology setup happens at smp_init()
  465. * time - but meanwhile we still have a functioning scheduler.
  466. */
  467. sched_init();
  468. /*
  469. * Disable preemption - early bootup scheduling is extremely
  470. * fragile until we cpu_idle() for the first time.
  471. */
  472. preempt_disable();
  473. if (WARN(!irqs_disabled(), "Interrupts were enabled *very* early, fixing it\n"))
  474. local_irq_disable();
  475. idr_init_cache();
  476. rcu_init();
  477. tick_nohz_init();
  478. radix_tree_init();
  479. /* init some links before init_ISA_irqs() */
  480. early_irq_init();
  481. init_IRQ();
  482. tick_init();
  483. init_timers();
  484. hrtimers_init();
  485. softirq_init();
  486. timekeeping_init();
  487. time_init();
  488. sched_clock_postinit();
  489. perf_event_init();
  490. profile_init();
  491. call_function_init();
  492. WARN(!irqs_disabled(), "Interrupts were enabled early\n");
  493. early_boot_irqs_disabled = false;
  494. local_irq_enable();
  495. kmem_cache_init_late();
  496. /*
  497. * HACK ALERT! This is early. We're enabling the console before
  498. * we've done PCI setups etc, and console_init() must be aware of
  499. * this. But we do want output early, in case something goes wrong.
  500. */
  501. console_init();
  502. if (panic_later)
  503. panic(panic_later, panic_param);
  504. lockdep_info();
  505. /*
  506. * Need to run this when irqs are enabled, because it wants
  507. * to self-test [hard/soft]-irqs on/off lock inversion bugs
  508. * too:
  509. */
  510. locking_selftest();
  511. #ifdef CONFIG_BLK_DEV_INITRD
  512. if (initrd_start && !initrd_below_start_ok &&
  513. page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
  514. pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
  515. page_to_pfn(virt_to_page((void *)initrd_start)),
  516. min_low_pfn);
  517. initrd_start = 0;
  518. }
  519. #endif
  520. page_cgroup_init();
  521. debug_objects_mem_init();
  522. kmemleak_init();
  523. setup_per_cpu_pageset();
  524. numa_policy_init();
  525. if (late_time_init)
  526. late_time_init();
  527. sched_clock_init();
  528. calibrate_delay();
  529. pidmap_init();
  530. anon_vma_init();
  531. #ifdef CONFIG_X86
  532. if (efi_enabled(EFI_RUNTIME_SERVICES))
  533. efi_enter_virtual_mode();
  534. #endif
  535. thread_info_cache_init();
  536. cred_init();
  537. fork_init(totalram_pages);
  538. proc_caches_init();
  539. buffer_init();
  540. key_init();
  541. security_init();
  542. dbg_late_init();
  543. vfs_caches_init(totalram_pages);
  544. signals_init();
  545. /* rootfs populating might need page-writeback */
  546. page_writeback_init();
  547. #ifdef CONFIG_PROC_FS
  548. proc_root_init();
  549. #endif
  550. cgroup_init();
  551. cpuset_init();
  552. taskstats_init_early();
  553. delayacct_init();
  554. check_bugs();
  555. acpi_early_init(); /* before LAPIC and SMP init */
  556. sfi_init_late();
  557. if (efi_enabled(EFI_RUNTIME_SERVICES)) {
  558. efi_late_init();
  559. efi_free_boot_services();
  560. }
  561. ftrace_init();
  562. /* Do the rest non-__init'ed, we're now alive */
  563. rest_init();
  564. }
  565. /* Call all constructor functions linked into the kernel. */
  566. static void __init do_ctors(void)
  567. {
  568. #ifdef CONFIG_CONSTRUCTORS
  569. ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
  570. for (; fn < (ctor_fn_t *) __ctors_end; fn++)
  571. (*fn)();
  572. #endif
  573. }
  574. bool initcall_debug;
  575. core_param(initcall_debug, initcall_debug, bool, 0644);
  576. static int __init_or_module do_one_initcall_debug(initcall_t fn)
  577. {
  578. ktime_t calltime, delta, rettime;
  579. unsigned long long duration;
  580. int ret;
  581. pr_debug("calling %pF @ %i\n", fn, task_pid_nr(current));
  582. calltime = ktime_get();
  583. ret = fn();
  584. rettime = ktime_get();
  585. delta = ktime_sub(rettime, calltime);
  586. duration = (unsigned long long) ktime_to_ns(delta) >> 10;
  587. pr_debug("initcall %pF returned %d after %lld usecs\n",
  588. fn, ret, duration);
  589. return ret;
  590. }
  591. int __init_or_module do_one_initcall(initcall_t fn)
  592. {
  593. int count = preempt_count();
  594. int ret;
  595. char msgbuf[64];
  596. if (initcall_debug)
  597. ret = do_one_initcall_debug(fn);
  598. else
  599. ret = fn();
  600. msgbuf[0] = 0;
  601. if (preempt_count() != count) {
  602. sprintf(msgbuf, "preemption imbalance ");
  603. preempt_count() = count;
  604. }
  605. if (irqs_disabled()) {
  606. strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
  607. local_irq_enable();
  608. }
  609. WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
  610. return ret;
  611. }
  612. extern initcall_t __initcall_start[];
  613. extern initcall_t __initcall0_start[];
  614. extern initcall_t __initcall1_start[];
  615. extern initcall_t __initcall2_start[];
  616. extern initcall_t __initcall3_start[];
  617. extern initcall_t __initcall4_start[];
  618. extern initcall_t __initcall5_start[];
  619. extern initcall_t __initcall6_start[];
  620. extern initcall_t __initcall7_start[];
  621. extern initcall_t __initcall_end[];
  622. static initcall_t *initcall_levels[] __initdata = {
  623. __initcall0_start,
  624. __initcall1_start,
  625. __initcall2_start,
  626. __initcall3_start,
  627. __initcall4_start,
  628. __initcall5_start,
  629. __initcall6_start,
  630. __initcall7_start,
  631. __initcall_end,
  632. };
  633. /* Keep these in sync with initcalls in include/linux/init.h */
  634. static char *initcall_level_names[] __initdata = {
  635. "early",
  636. "core",
  637. "postcore",
  638. "arch",
  639. "subsys",
  640. "fs",
  641. "device",
  642. "late",
  643. };
  644. static void __init do_initcall_level(int level)
  645. {
  646. extern const struct kernel_param __start___param[], __stop___param[];
  647. initcall_t *fn;
  648. strcpy(static_command_line, saved_command_line);
  649. parse_args(initcall_level_names[level],
  650. static_command_line, __start___param,
  651. __stop___param - __start___param,
  652. level, level,
  653. &repair_env_string);
  654. for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
  655. do_one_initcall(*fn);
  656. }
  657. static void __init do_initcalls(void)
  658. {
  659. int level;
  660. for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
  661. do_initcall_level(level);
  662. }
  663. /*
  664. * Ok, the machine is now initialized. None of the devices
  665. * have been touched yet, but the CPU subsystem is up and
  666. * running, and memory and process management works.
  667. *
  668. * Now we can finally start doing some real work..
  669. */
  670. static void __init do_basic_setup(void)
  671. {
  672. cpuset_init_smp();
  673. usermodehelper_init();
  674. shmem_init();
  675. driver_init();
  676. init_irq_proc();
  677. do_ctors();
  678. usermodehelper_enable();
  679. do_initcalls();
  680. }
  681. static void __init do_pre_smp_initcalls(void)
  682. {
  683. initcall_t *fn;
  684. for (fn = __initcall_start; fn < __initcall0_start; fn++)
  685. do_one_initcall(*fn);
  686. }
  687. /*
  688. * This function requests modules which should be loaded by default and is
  689. * called twice right after initrd is mounted and right before init is
  690. * exec'd. If such modules are on either initrd or rootfs, they will be
  691. * loaded before control is passed to userland.
  692. */
  693. void __init load_default_modules(void)
  694. {
  695. load_default_elevator_module();
  696. }
  697. static int run_init_process(const char *init_filename)
  698. {
  699. argv_init[0] = init_filename;
  700. return do_execve(init_filename,
  701. (const char __user *const __user *)argv_init,
  702. (const char __user *const __user *)envp_init);
  703. }
  704. static noinline void __init kernel_init_freeable(void);
  705. static int __ref kernel_init(void *unused)
  706. {
  707. kernel_init_freeable();
  708. /* need to finish all async __init code before freeing the memory */
  709. async_synchronize_full();
  710. free_initmem();
  711. mark_rodata_ro();
  712. system_state = SYSTEM_RUNNING;
  713. numa_default_policy();
  714. flush_delayed_fput();
  715. if (ramdisk_execute_command) {
  716. if (!run_init_process(ramdisk_execute_command))
  717. return 0;
  718. pr_err("Failed to execute %s\n", ramdisk_execute_command);
  719. }
  720. /*
  721. * We try each of these until one succeeds.
  722. *
  723. * The Bourne shell can be used instead of init if we are
  724. * trying to recover a really broken machine.
  725. */
  726. if (execute_command) {
  727. if (!run_init_process(execute_command))
  728. return 0;
  729. pr_err("Failed to execute %s. Attempting defaults...\n",
  730. execute_command);
  731. }
  732. if (!run_init_process("/sbin/init") ||
  733. !run_init_process("/etc/init") ||
  734. !run_init_process("/bin/init") ||
  735. !run_init_process("/bin/sh"))
  736. return 0;
  737. panic("No init found. Try passing init= option to kernel. "
  738. "See Linux Documentation/init.txt for guidance.");
  739. }
  740. static noinline void __init kernel_init_freeable(void)
  741. {
  742. /*
  743. * Wait until kthreadd is all set-up.
  744. */
  745. wait_for_completion(&kthreadd_done);
  746. /* Now the scheduler is fully set up and can do blocking allocations */
  747. gfp_allowed_mask = __GFP_BITS_MASK;
  748. /*
  749. * init can allocate pages on any node
  750. */
  751. set_mems_allowed(node_states[N_MEMORY]);
  752. /*
  753. * init can run on any cpu.
  754. */
  755. set_cpus_allowed_ptr(current, cpu_all_mask);
  756. cad_pid = task_pid(current);
  757. smp_prepare_cpus(setup_max_cpus);
  758. do_pre_smp_initcalls();
  759. lockup_detector_init();
  760. smp_init();
  761. sched_init_smp();
  762. do_basic_setup();
  763. /* Open the /dev/console on the rootfs, this should never fail */
  764. if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
  765. pr_err("Warning: unable to open an initial console.\n");
  766. (void) sys_dup(0);
  767. (void) sys_dup(0);
  768. /*
  769. * check if there is an early userspace init. If yes, let it do all
  770. * the work
  771. */
  772. if (!ramdisk_execute_command)
  773. ramdisk_execute_command = "/init";
  774. if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
  775. ramdisk_execute_command = NULL;
  776. prepare_namespace();
  777. }
  778. /*
  779. * Ok, we have completed the initial bootup, and
  780. * we're essentially up and running. Get rid of the
  781. * initmem segments and start the user-mode stuff..
  782. */
  783. /* rootfs is available now, try loading default modules */
  784. load_default_modules();
  785. }