main.c 22 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912
  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. * If set, this is an indication to the drivers that reset the underlying
  126. * device before going ahead with the initialization otherwise driver might
  127. * rely on the BIOS and skip the reset operation.
  128. *
  129. * This is useful if kernel is booting in an unreliable environment.
  130. * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
  131. * skipped and devices will be in unknown state.
  132. */
  133. unsigned int reset_devices;
  134. EXPORT_SYMBOL(reset_devices);
  135. static int __init set_reset_devices(char *str)
  136. {
  137. reset_devices = 1;
  138. return 1;
  139. }
  140. __setup("reset_devices", set_reset_devices);
  141. static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
  142. const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
  143. static const char *panic_later, *panic_param;
  144. extern const struct obs_kernel_param __setup_start[], __setup_end[];
  145. static int __init obsolete_checksetup(char *line)
  146. {
  147. const struct obs_kernel_param *p;
  148. int had_early_param = 0;
  149. p = __setup_start;
  150. do {
  151. int n = strlen(p->str);
  152. if (parameqn(line, p->str, n)) {
  153. if (p->early) {
  154. /* Already done in parse_early_param?
  155. * (Needs exact match on param part).
  156. * Keep iterating, as we can have early
  157. * params and __setups of same names 8( */
  158. if (line[n] == '\0' || line[n] == '=')
  159. had_early_param = 1;
  160. } else if (!p->setup_func) {
  161. pr_warn("Parameter %s is obsolete, ignored\n",
  162. p->str);
  163. return 1;
  164. } else if (p->setup_func(line + n))
  165. return 1;
  166. }
  167. p++;
  168. } while (p < __setup_end);
  169. return had_early_param;
  170. }
  171. /*
  172. * This should be approx 2 Bo*oMips to start (note initial shift), and will
  173. * still work even if initially too large, it will just take slightly longer
  174. */
  175. unsigned long loops_per_jiffy = (1<<12);
  176. EXPORT_SYMBOL(loops_per_jiffy);
  177. static int __init debug_kernel(char *str)
  178. {
  179. console_loglevel = 10;
  180. return 0;
  181. }
  182. static int __init quiet_kernel(char *str)
  183. {
  184. console_loglevel = 4;
  185. return 0;
  186. }
  187. early_param("debug", debug_kernel);
  188. early_param("quiet", quiet_kernel);
  189. static int __init loglevel(char *str)
  190. {
  191. int newlevel;
  192. /*
  193. * Only update loglevel value when a correct setting was passed,
  194. * to prevent blind crashes (when loglevel being set to 0) that
  195. * are quite hard to debug
  196. */
  197. if (get_option(&str, &newlevel)) {
  198. console_loglevel = newlevel;
  199. return 0;
  200. }
  201. return -EINVAL;
  202. }
  203. early_param("loglevel", loglevel);
  204. /* Change NUL term back to "=", to make "param" the whole string. */
  205. static int __init repair_env_string(char *param, char *val, const char *unused)
  206. {
  207. if (val) {
  208. /* param=val or param="val"? */
  209. if (val == param+strlen(param)+1)
  210. val[-1] = '=';
  211. else if (val == param+strlen(param)+2) {
  212. val[-2] = '=';
  213. memmove(val-1, val, strlen(val)+1);
  214. val--;
  215. } else
  216. BUG();
  217. }
  218. return 0;
  219. }
  220. /*
  221. * Unknown boot options get handed to init, unless they look like
  222. * unused parameters (modprobe will find them in /proc/cmdline).
  223. */
  224. static int __init unknown_bootoption(char *param, char *val, const char *unused)
  225. {
  226. repair_env_string(param, val, unused);
  227. /* Handle obsolete-style parameters */
  228. if (obsolete_checksetup(param))
  229. return 0;
  230. /* Unused module parameter. */
  231. if (strchr(param, '.') && (!val || strchr(param, '.') < val))
  232. return 0;
  233. if (panic_later)
  234. return 0;
  235. if (val) {
  236. /* Environment option */
  237. unsigned int i;
  238. for (i = 0; envp_init[i]; i++) {
  239. if (i == MAX_INIT_ENVS) {
  240. panic_later = "Too many boot env vars at `%s'";
  241. panic_param = param;
  242. }
  243. if (!strncmp(param, envp_init[i], val - param))
  244. break;
  245. }
  246. envp_init[i] = param;
  247. } else {
  248. /* Command line option */
  249. unsigned int i;
  250. for (i = 0; argv_init[i]; i++) {
  251. if (i == MAX_INIT_ARGS) {
  252. panic_later = "Too many boot init vars at `%s'";
  253. panic_param = param;
  254. }
  255. }
  256. argv_init[i] = param;
  257. }
  258. return 0;
  259. }
  260. static int __init init_setup(char *str)
  261. {
  262. unsigned int i;
  263. execute_command = str;
  264. /*
  265. * In case LILO is going to boot us with default command line,
  266. * it prepends "auto" before the whole cmdline which makes
  267. * the shell think it should execute a script with such name.
  268. * So we ignore all arguments entered _before_ init=... [MJ]
  269. */
  270. for (i = 1; i < MAX_INIT_ARGS; i++)
  271. argv_init[i] = NULL;
  272. return 1;
  273. }
  274. __setup("init=", init_setup);
  275. static int __init rdinit_setup(char *str)
  276. {
  277. unsigned int i;
  278. ramdisk_execute_command = str;
  279. /* See "auto" comment in init_setup */
  280. for (i = 1; i < MAX_INIT_ARGS; i++)
  281. argv_init[i] = NULL;
  282. return 1;
  283. }
  284. __setup("rdinit=", rdinit_setup);
  285. #ifndef CONFIG_SMP
  286. static const unsigned int setup_max_cpus = NR_CPUS;
  287. #ifdef CONFIG_X86_LOCAL_APIC
  288. static void __init smp_init(void)
  289. {
  290. APIC_init_uniprocessor();
  291. }
  292. #else
  293. #define smp_init() do { } while (0)
  294. #endif
  295. static inline void setup_nr_cpu_ids(void) { }
  296. static inline void smp_prepare_cpus(unsigned int maxcpus) { }
  297. #endif
  298. /*
  299. * We need to store the untouched command line for future reference.
  300. * We also need to store the touched command line since the parameter
  301. * parsing is performed in place, and we should allow a component to
  302. * store reference of name/value for future reference.
  303. */
  304. static void __init setup_command_line(char *command_line)
  305. {
  306. saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
  307. static_command_line = alloc_bootmem(strlen (command_line)+1);
  308. strcpy (saved_command_line, boot_command_line);
  309. strcpy (static_command_line, command_line);
  310. }
  311. /*
  312. * We need to finalize in a non-__init function or else race conditions
  313. * between the root thread and the init thread may cause start_kernel to
  314. * be reaped by free_initmem before the root thread has proceeded to
  315. * cpu_idle.
  316. *
  317. * gcc-3.4 accidentally inlines this function, so use noinline.
  318. */
  319. static __initdata DECLARE_COMPLETION(kthreadd_done);
  320. static noinline void __init_refok rest_init(void)
  321. {
  322. int pid;
  323. rcu_scheduler_starting();
  324. /*
  325. * We need to spawn init first so that it obtains pid 1, however
  326. * the init task will end up wanting to create kthreads, which, if
  327. * we schedule it before we create kthreadd, will OOPS.
  328. */
  329. kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
  330. numa_default_policy();
  331. pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
  332. rcu_read_lock();
  333. kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
  334. rcu_read_unlock();
  335. complete(&kthreadd_done);
  336. /*
  337. * The boot idle thread must execute schedule()
  338. * at least once to get things moving:
  339. */
  340. init_idle_bootup_task(current);
  341. schedule_preempt_disabled();
  342. /* Call into cpu_idle with preempt disabled */
  343. cpu_startup_entry(CPUHP_ONLINE);
  344. }
  345. /* Check for early params. */
  346. static int __init do_early_param(char *param, char *val, const char *unused)
  347. {
  348. const struct obs_kernel_param *p;
  349. for (p = __setup_start; p < __setup_end; p++) {
  350. if ((p->early && parameq(param, p->str)) ||
  351. (strcmp(param, "console") == 0 &&
  352. strcmp(p->str, "earlycon") == 0)
  353. ) {
  354. if (p->setup_func(val) != 0)
  355. pr_warn("Malformed early option '%s'\n", param);
  356. }
  357. }
  358. /* We accept everything at this stage. */
  359. return 0;
  360. }
  361. void __init parse_early_options(char *cmdline)
  362. {
  363. parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
  364. }
  365. /* Arch code calls this early on, or if not, just before other parsing. */
  366. void __init parse_early_param(void)
  367. {
  368. static __initdata int done = 0;
  369. static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
  370. if (done)
  371. return;
  372. /* All fall through to do_early_param. */
  373. strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
  374. parse_early_options(tmp_cmdline);
  375. done = 1;
  376. }
  377. /*
  378. * Activate the first processor.
  379. */
  380. static void __init boot_cpu_init(void)
  381. {
  382. int cpu = smp_processor_id();
  383. /* Mark the boot cpu "present", "online" etc for SMP and UP case */
  384. set_cpu_online(cpu, true);
  385. set_cpu_active(cpu, true);
  386. set_cpu_present(cpu, true);
  387. set_cpu_possible(cpu, true);
  388. }
  389. void __init __weak smp_setup_processor_id(void)
  390. {
  391. }
  392. # if THREAD_SIZE >= PAGE_SIZE
  393. void __init __weak thread_info_cache_init(void)
  394. {
  395. }
  396. #endif
  397. /*
  398. * Set up kernel memory allocators
  399. */
  400. static void __init mm_init(void)
  401. {
  402. /*
  403. * page_cgroup requires contiguous pages,
  404. * bigger than MAX_ORDER unless SPARSEMEM.
  405. */
  406. page_cgroup_init_flatmem();
  407. mem_init();
  408. kmem_cache_init();
  409. percpu_init_late();
  410. pgtable_cache_init();
  411. vmalloc_init();
  412. }
  413. asmlinkage void __init start_kernel(void)
  414. {
  415. char * command_line;
  416. extern const struct kernel_param __start___param[], __stop___param[];
  417. /*
  418. * Need to run as early as possible, to initialize the
  419. * lockdep hash:
  420. */
  421. lockdep_init();
  422. smp_setup_processor_id();
  423. debug_objects_early_init();
  424. /*
  425. * Set up the the initial canary ASAP:
  426. */
  427. boot_init_stack_canary();
  428. cgroup_init_early();
  429. local_irq_disable();
  430. early_boot_irqs_disabled = true;
  431. /*
  432. * Interrupts are still disabled. Do necessary setups, then
  433. * enable them
  434. */
  435. boot_cpu_init();
  436. page_address_init();
  437. pr_notice("%s", linux_banner);
  438. setup_arch(&command_line);
  439. mm_init_owner(&init_mm, &init_task);
  440. mm_init_cpumask(&init_mm);
  441. setup_command_line(command_line);
  442. setup_nr_cpu_ids();
  443. setup_per_cpu_areas();
  444. smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
  445. build_all_zonelists(NULL, NULL);
  446. page_alloc_init();
  447. pr_notice("Kernel command line: %s\n", boot_command_line);
  448. parse_early_param();
  449. parse_args("Booting kernel", static_command_line, __start___param,
  450. __stop___param - __start___param,
  451. -1, -1, &unknown_bootoption);
  452. jump_label_init();
  453. /*
  454. * These use large bootmem allocations and must precede
  455. * kmem_cache_init()
  456. */
  457. setup_log_buf(0);
  458. pidhash_init();
  459. vfs_caches_init_early();
  460. sort_main_extable();
  461. trap_init();
  462. mm_init();
  463. /*
  464. * Set up the scheduler prior starting any interrupts (such as the
  465. * timer interrupt). Full topology setup happens at smp_init()
  466. * time - but meanwhile we still have a functioning scheduler.
  467. */
  468. sched_init();
  469. /*
  470. * Disable preemption - early bootup scheduling is extremely
  471. * fragile until we cpu_idle() for the first time.
  472. */
  473. preempt_disable();
  474. if (WARN(!irqs_disabled(), "Interrupts were enabled *very* early, fixing it\n"))
  475. local_irq_disable();
  476. idr_init_cache();
  477. rcu_init();
  478. tick_nohz_init();
  479. context_tracking_init();
  480. radix_tree_init();
  481. /* init some links before init_ISA_irqs() */
  482. early_irq_init();
  483. init_IRQ();
  484. tick_init();
  485. init_timers();
  486. hrtimers_init();
  487. softirq_init();
  488. timekeeping_init();
  489. time_init();
  490. sched_clock_postinit();
  491. perf_event_init();
  492. profile_init();
  493. call_function_init();
  494. WARN(!irqs_disabled(), "Interrupts were enabled early\n");
  495. early_boot_irqs_disabled = false;
  496. local_irq_enable();
  497. kmem_cache_init_late();
  498. /*
  499. * HACK ALERT! This is early. We're enabling the console before
  500. * we've done PCI setups etc, and console_init() must be aware of
  501. * this. But we do want output early, in case something goes wrong.
  502. */
  503. console_init();
  504. if (panic_later)
  505. panic(panic_later, panic_param);
  506. lockdep_info();
  507. /*
  508. * Need to run this when irqs are enabled, because it wants
  509. * to self-test [hard/soft]-irqs on/off lock inversion bugs
  510. * too:
  511. */
  512. locking_selftest();
  513. #ifdef CONFIG_BLK_DEV_INITRD
  514. if (initrd_start && !initrd_below_start_ok &&
  515. page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
  516. pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
  517. page_to_pfn(virt_to_page((void *)initrd_start)),
  518. min_low_pfn);
  519. initrd_start = 0;
  520. }
  521. #endif
  522. page_cgroup_init();
  523. debug_objects_mem_init();
  524. kmemleak_init();
  525. setup_per_cpu_pageset();
  526. numa_policy_init();
  527. if (late_time_init)
  528. late_time_init();
  529. sched_clock_init();
  530. calibrate_delay();
  531. pidmap_init();
  532. anon_vma_init();
  533. #ifdef CONFIG_X86
  534. if (efi_enabled(EFI_RUNTIME_SERVICES))
  535. efi_enter_virtual_mode();
  536. #endif
  537. thread_info_cache_init();
  538. cred_init();
  539. fork_init(totalram_pages);
  540. proc_caches_init();
  541. buffer_init();
  542. key_init();
  543. security_init();
  544. dbg_late_init();
  545. vfs_caches_init(totalram_pages);
  546. signals_init();
  547. /* rootfs populating might need page-writeback */
  548. page_writeback_init();
  549. #ifdef CONFIG_PROC_FS
  550. proc_root_init();
  551. #endif
  552. cgroup_init();
  553. cpuset_init();
  554. taskstats_init_early();
  555. delayacct_init();
  556. check_bugs();
  557. acpi_early_init(); /* before LAPIC and SMP init */
  558. sfi_init_late();
  559. if (efi_enabled(EFI_RUNTIME_SERVICES)) {
  560. efi_late_init();
  561. efi_free_boot_services();
  562. }
  563. ftrace_init();
  564. /* Do the rest non-__init'ed, we're now alive */
  565. rest_init();
  566. }
  567. /* Call all constructor functions linked into the kernel. */
  568. static void __init do_ctors(void)
  569. {
  570. #ifdef CONFIG_CONSTRUCTORS
  571. ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
  572. for (; fn < (ctor_fn_t *) __ctors_end; fn++)
  573. (*fn)();
  574. #endif
  575. }
  576. bool initcall_debug;
  577. core_param(initcall_debug, initcall_debug, bool, 0644);
  578. static int __init_or_module do_one_initcall_debug(initcall_t fn)
  579. {
  580. ktime_t calltime, delta, rettime;
  581. unsigned long long duration;
  582. int ret;
  583. pr_debug("calling %pF @ %i\n", fn, task_pid_nr(current));
  584. calltime = ktime_get();
  585. ret = fn();
  586. rettime = ktime_get();
  587. delta = ktime_sub(rettime, calltime);
  588. duration = (unsigned long long) ktime_to_ns(delta) >> 10;
  589. pr_debug("initcall %pF returned %d after %lld usecs\n",
  590. fn, ret, duration);
  591. return ret;
  592. }
  593. int __init_or_module do_one_initcall(initcall_t fn)
  594. {
  595. int count = preempt_count();
  596. int ret;
  597. char msgbuf[64];
  598. if (initcall_debug)
  599. ret = do_one_initcall_debug(fn);
  600. else
  601. ret = fn();
  602. msgbuf[0] = 0;
  603. if (preempt_count() != count) {
  604. sprintf(msgbuf, "preemption imbalance ");
  605. preempt_count() = count;
  606. }
  607. if (irqs_disabled()) {
  608. strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
  609. local_irq_enable();
  610. }
  611. WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
  612. return ret;
  613. }
  614. extern initcall_t __initcall_start[];
  615. extern initcall_t __initcall0_start[];
  616. extern initcall_t __initcall1_start[];
  617. extern initcall_t __initcall2_start[];
  618. extern initcall_t __initcall3_start[];
  619. extern initcall_t __initcall4_start[];
  620. extern initcall_t __initcall5_start[];
  621. extern initcall_t __initcall6_start[];
  622. extern initcall_t __initcall7_start[];
  623. extern initcall_t __initcall_end[];
  624. static initcall_t *initcall_levels[] __initdata = {
  625. __initcall0_start,
  626. __initcall1_start,
  627. __initcall2_start,
  628. __initcall3_start,
  629. __initcall4_start,
  630. __initcall5_start,
  631. __initcall6_start,
  632. __initcall7_start,
  633. __initcall_end,
  634. };
  635. /* Keep these in sync with initcalls in include/linux/init.h */
  636. static char *initcall_level_names[] __initdata = {
  637. "early",
  638. "core",
  639. "postcore",
  640. "arch",
  641. "subsys",
  642. "fs",
  643. "device",
  644. "late",
  645. };
  646. static void __init do_initcall_level(int level)
  647. {
  648. extern const struct kernel_param __start___param[], __stop___param[];
  649. initcall_t *fn;
  650. strcpy(static_command_line, saved_command_line);
  651. parse_args(initcall_level_names[level],
  652. static_command_line, __start___param,
  653. __stop___param - __start___param,
  654. level, level,
  655. &repair_env_string);
  656. for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
  657. do_one_initcall(*fn);
  658. }
  659. static void __init do_initcalls(void)
  660. {
  661. int level;
  662. for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
  663. do_initcall_level(level);
  664. }
  665. /*
  666. * Ok, the machine is now initialized. None of the devices
  667. * have been touched yet, but the CPU subsystem is up and
  668. * running, and memory and process management works.
  669. *
  670. * Now we can finally start doing some real work..
  671. */
  672. static void __init do_basic_setup(void)
  673. {
  674. cpuset_init_smp();
  675. usermodehelper_init();
  676. shmem_init();
  677. driver_init();
  678. init_irq_proc();
  679. do_ctors();
  680. usermodehelper_enable();
  681. do_initcalls();
  682. }
  683. static void __init do_pre_smp_initcalls(void)
  684. {
  685. initcall_t *fn;
  686. for (fn = __initcall_start; fn < __initcall0_start; fn++)
  687. do_one_initcall(*fn);
  688. }
  689. /*
  690. * This function requests modules which should be loaded by default and is
  691. * called twice right after initrd is mounted and right before init is
  692. * exec'd. If such modules are on either initrd or rootfs, they will be
  693. * loaded before control is passed to userland.
  694. */
  695. void __init load_default_modules(void)
  696. {
  697. load_default_elevator_module();
  698. }
  699. static int run_init_process(const char *init_filename)
  700. {
  701. argv_init[0] = init_filename;
  702. return do_execve(init_filename,
  703. (const char __user *const __user *)argv_init,
  704. (const char __user *const __user *)envp_init);
  705. }
  706. static noinline void __init kernel_init_freeable(void);
  707. static int __ref kernel_init(void *unused)
  708. {
  709. kernel_init_freeable();
  710. /* need to finish all async __init code before freeing the memory */
  711. async_synchronize_full();
  712. free_initmem();
  713. mark_rodata_ro();
  714. system_state = SYSTEM_RUNNING;
  715. numa_default_policy();
  716. flush_delayed_fput();
  717. if (ramdisk_execute_command) {
  718. if (!run_init_process(ramdisk_execute_command))
  719. return 0;
  720. pr_err("Failed to execute %s\n", ramdisk_execute_command);
  721. }
  722. /*
  723. * We try each of these until one succeeds.
  724. *
  725. * The Bourne shell can be used instead of init if we are
  726. * trying to recover a really broken machine.
  727. */
  728. if (execute_command) {
  729. if (!run_init_process(execute_command))
  730. return 0;
  731. pr_err("Failed to execute %s. Attempting defaults...\n",
  732. execute_command);
  733. }
  734. if (!run_init_process("/sbin/init") ||
  735. !run_init_process("/etc/init") ||
  736. !run_init_process("/bin/init") ||
  737. !run_init_process("/bin/sh"))
  738. return 0;
  739. panic("No init found. Try passing init= option to kernel. "
  740. "See Linux Documentation/init.txt for guidance.");
  741. }
  742. static noinline void __init kernel_init_freeable(void)
  743. {
  744. /*
  745. * Wait until kthreadd is all set-up.
  746. */
  747. wait_for_completion(&kthreadd_done);
  748. /* Now the scheduler is fully set up and can do blocking allocations */
  749. gfp_allowed_mask = __GFP_BITS_MASK;
  750. /*
  751. * init can allocate pages on any node
  752. */
  753. set_mems_allowed(node_states[N_MEMORY]);
  754. /*
  755. * init can run on any cpu.
  756. */
  757. set_cpus_allowed_ptr(current, cpu_all_mask);
  758. cad_pid = task_pid(current);
  759. smp_prepare_cpus(setup_max_cpus);
  760. do_pre_smp_initcalls();
  761. lockup_detector_init();
  762. smp_init();
  763. sched_init_smp();
  764. do_basic_setup();
  765. /* Open the /dev/console on the rootfs, this should never fail */
  766. if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
  767. pr_err("Warning: unable to open an initial console.\n");
  768. (void) sys_dup(0);
  769. (void) sys_dup(0);
  770. /*
  771. * check if there is an early userspace init. If yes, let it do all
  772. * the work
  773. */
  774. if (!ramdisk_execute_command)
  775. ramdisk_execute_command = "/init";
  776. if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
  777. ramdisk_execute_command = NULL;
  778. prepare_namespace();
  779. }
  780. /*
  781. * Ok, we have completed the initial bootup, and
  782. * we're essentially up and running. Get rid of the
  783. * initmem segments and start the user-mode stuff..
  784. */
  785. /* rootfs is available now, try loading default modules */
  786. load_default_modules();
  787. }