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. #include <linux/types.h>
  12. #include <linux/module.h>
  13. #include <linux/proc_fs.h>
  14. #include <linux/kernel.h>
  15. #include <linux/syscalls.h>
  16. #include <linux/stackprotector.h>
  17. #include <linux/string.h>
  18. #include <linux/ctype.h>
  19. #include <linux/delay.h>
  20. #include <linux/ioport.h>
  21. #include <linux/init.h>
  22. #include <linux/initrd.h>
  23. #include <linux/bootmem.h>
  24. #include <linux/acpi.h>
  25. #include <linux/tty.h>
  26. #include <linux/percpu.h>
  27. #include <linux/kmod.h>
  28. #include <linux/vmalloc.h>
  29. #include <linux/kernel_stat.h>
  30. #include <linux/start_kernel.h>
  31. #include <linux/security.h>
  32. #include <linux/smp.h>
  33. #include <linux/profile.h>
  34. #include <linux/rcupdate.h>
  35. #include <linux/moduleparam.h>
  36. #include <linux/kallsyms.h>
  37. #include <linux/writeback.h>
  38. #include <linux/cpu.h>
  39. #include <linux/cpuset.h>
  40. #include <linux/cgroup.h>
  41. #include <linux/efi.h>
  42. #include <linux/tick.h>
  43. #include <linux/interrupt.h>
  44. #include <linux/taskstats_kern.h>
  45. #include <linux/delayacct.h>
  46. #include <linux/unistd.h>
  47. #include <linux/rmap.h>
  48. #include <linux/mempolicy.h>
  49. #include <linux/key.h>
  50. #include <linux/buffer_head.h>
  51. #include <linux/page_cgroup.h>
  52. #include <linux/debug_locks.h>
  53. #include <linux/debugobjects.h>
  54. #include <linux/lockdep.h>
  55. #include <linux/kmemleak.h>
  56. #include <linux/pid_namespace.h>
  57. #include <linux/device.h>
  58. #include <linux/kthread.h>
  59. #include <linux/sched.h>
  60. #include <linux/signal.h>
  61. #include <linux/idr.h>
  62. #include <linux/kgdb.h>
  63. #include <linux/ftrace.h>
  64. #include <linux/async.h>
  65. #include <linux/kmemcheck.h>
  66. #include <linux/sfi.h>
  67. #include <linux/shmem_fs.h>
  68. #include <linux/slab.h>
  69. #include <linux/perf_event.h>
  70. #include <linux/file.h>
  71. #include <linux/ptrace.h>
  72. #include <linux/blkdev.h>
  73. #include <linux/elevator.h>
  74. #include <asm/io.h>
  75. #include <asm/bugs.h>
  76. #include <asm/setup.h>
  77. #include <asm/sections.h>
  78. #include <asm/cacheflush.h>
  79. #ifdef CONFIG_X86_LOCAL_APIC
  80. #include <asm/smp.h>
  81. #endif
  82. static int kernel_init(void *);
  83. extern void init_IRQ(void);
  84. extern void fork_init(unsigned long);
  85. extern void mca_init(void);
  86. extern void sbus_init(void);
  87. extern void radix_tree_init(void);
  88. #ifndef CONFIG_DEBUG_RODATA
  89. static inline void mark_rodata_ro(void) { }
  90. #endif
  91. #ifdef CONFIG_TC
  92. extern void tc_init(void);
  93. #endif
  94. /*
  95. * Debug helper: via this flag we know that we are in 'early bootup code'
  96. * where only the boot processor is running with IRQ disabled. This means
  97. * two things - IRQ must not be enabled before the flag is cleared and some
  98. * operations which are not allowed with IRQ disabled are allowed while the
  99. * flag is set.
  100. */
  101. bool early_boot_irqs_disabled __read_mostly;
  102. enum system_states system_state __read_mostly;
  103. EXPORT_SYMBOL(system_state);
  104. /*
  105. * Boot command-line arguments
  106. */
  107. #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
  108. #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
  109. extern void time_init(void);
  110. /* Default late time init is NULL. archs can override this later. */
  111. void (*__initdata late_time_init)(void);
  112. extern void softirq_init(void);
  113. /* Untouched command line saved by arch-specific code. */
  114. char __initdata boot_command_line[COMMAND_LINE_SIZE];
  115. /* Untouched saved command line (eg. for /proc) */
  116. char *saved_command_line;
  117. /* Command line for parameter parsing */
  118. static char *static_command_line;
  119. static char *execute_command;
  120. static char *ramdisk_execute_command;
  121. /*
  122. * If set, this is an indication to the drivers that reset the underlying
  123. * device before going ahead with the initialization otherwise driver might
  124. * rely on the BIOS and skip the reset operation.
  125. *
  126. * This is useful if kernel is booting in an unreliable environment.
  127. * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
  128. * skipped and devices will be in unknown state.
  129. */
  130. unsigned int reset_devices;
  131. EXPORT_SYMBOL(reset_devices);
  132. static int __init set_reset_devices(char *str)
  133. {
  134. reset_devices = 1;
  135. return 1;
  136. }
  137. __setup("reset_devices", set_reset_devices);
  138. static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
  139. const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
  140. static const char *panic_later, *panic_param;
  141. extern const struct obs_kernel_param __setup_start[], __setup_end[];
  142. static int __init obsolete_checksetup(char *line)
  143. {
  144. const struct obs_kernel_param *p;
  145. int had_early_param = 0;
  146. p = __setup_start;
  147. do {
  148. int n = strlen(p->str);
  149. if (parameqn(line, p->str, n)) {
  150. if (p->early) {
  151. /* Already done in parse_early_param?
  152. * (Needs exact match on param part).
  153. * Keep iterating, as we can have early
  154. * params and __setups of same names 8( */
  155. if (line[n] == '\0' || line[n] == '=')
  156. had_early_param = 1;
  157. } else if (!p->setup_func) {
  158. printk(KERN_WARNING "Parameter %s is obsolete,"
  159. " ignored\n", p->str);
  160. return 1;
  161. } else if (p->setup_func(line + n))
  162. return 1;
  163. }
  164. p++;
  165. } while (p < __setup_end);
  166. return had_early_param;
  167. }
  168. /*
  169. * This should be approx 2 Bo*oMips to start (note initial shift), and will
  170. * still work even if initially too large, it will just take slightly longer
  171. */
  172. unsigned long loops_per_jiffy = (1<<12);
  173. EXPORT_SYMBOL(loops_per_jiffy);
  174. static int __init debug_kernel(char *str)
  175. {
  176. console_loglevel = 10;
  177. return 0;
  178. }
  179. static int __init quiet_kernel(char *str)
  180. {
  181. console_loglevel = 4;
  182. return 0;
  183. }
  184. early_param("debug", debug_kernel);
  185. early_param("quiet", quiet_kernel);
  186. static int __init loglevel(char *str)
  187. {
  188. int newlevel;
  189. /*
  190. * Only update loglevel value when a correct setting was passed,
  191. * to prevent blind crashes (when loglevel being set to 0) that
  192. * are quite hard to debug
  193. */
  194. if (get_option(&str, &newlevel)) {
  195. console_loglevel = newlevel;
  196. return 0;
  197. }
  198. return -EINVAL;
  199. }
  200. early_param("loglevel", loglevel);
  201. /* Change NUL term back to "=", to make "param" the whole string. */
  202. static int __init repair_env_string(char *param, char *val, const char *unused)
  203. {
  204. if (val) {
  205. /* param=val or param="val"? */
  206. if (val == param+strlen(param)+1)
  207. val[-1] = '=';
  208. else if (val == param+strlen(param)+2) {
  209. val[-2] = '=';
  210. memmove(val-1, val, strlen(val)+1);
  211. val--;
  212. } else
  213. BUG();
  214. }
  215. return 0;
  216. }
  217. /*
  218. * Unknown boot options get handed to init, unless they look like
  219. * unused parameters (modprobe will find them in /proc/cmdline).
  220. */
  221. static int __init unknown_bootoption(char *param, char *val, const char *unused)
  222. {
  223. repair_env_string(param, val, unused);
  224. /* Handle obsolete-style parameters */
  225. if (obsolete_checksetup(param))
  226. return 0;
  227. /* Unused module parameter. */
  228. if (strchr(param, '.') && (!val || strchr(param, '.') < val))
  229. return 0;
  230. if (panic_later)
  231. return 0;
  232. if (val) {
  233. /* Environment option */
  234. unsigned int i;
  235. for (i = 0; envp_init[i]; i++) {
  236. if (i == MAX_INIT_ENVS) {
  237. panic_later = "Too many boot env vars at `%s'";
  238. panic_param = param;
  239. }
  240. if (!strncmp(param, envp_init[i], val - param))
  241. break;
  242. }
  243. envp_init[i] = param;
  244. } else {
  245. /* Command line option */
  246. unsigned int i;
  247. for (i = 0; argv_init[i]; i++) {
  248. if (i == MAX_INIT_ARGS) {
  249. panic_later = "Too many boot init vars at `%s'";
  250. panic_param = param;
  251. }
  252. }
  253. argv_init[i] = param;
  254. }
  255. return 0;
  256. }
  257. static int __init init_setup(char *str)
  258. {
  259. unsigned int i;
  260. execute_command = str;
  261. /*
  262. * In case LILO is going to boot us with default command line,
  263. * it prepends "auto" before the whole cmdline which makes
  264. * the shell think it should execute a script with such name.
  265. * So we ignore all arguments entered _before_ init=... [MJ]
  266. */
  267. for (i = 1; i < MAX_INIT_ARGS; i++)
  268. argv_init[i] = NULL;
  269. return 1;
  270. }
  271. __setup("init=", init_setup);
  272. static int __init rdinit_setup(char *str)
  273. {
  274. unsigned int i;
  275. ramdisk_execute_command = str;
  276. /* See "auto" comment in init_setup */
  277. for (i = 1; i < MAX_INIT_ARGS; i++)
  278. argv_init[i] = NULL;
  279. return 1;
  280. }
  281. __setup("rdinit=", rdinit_setup);
  282. #ifndef CONFIG_SMP
  283. static const unsigned int setup_max_cpus = NR_CPUS;
  284. #ifdef CONFIG_X86_LOCAL_APIC
  285. static void __init smp_init(void)
  286. {
  287. APIC_init_uniprocessor();
  288. }
  289. #else
  290. #define smp_init() do { } while (0)
  291. #endif
  292. static inline void setup_nr_cpu_ids(void) { }
  293. static inline void smp_prepare_cpus(unsigned int maxcpus) { }
  294. #endif
  295. /*
  296. * We need to store the untouched command line for future reference.
  297. * We also need to store the touched command line since the parameter
  298. * parsing is performed in place, and we should allow a component to
  299. * store reference of name/value for future reference.
  300. */
  301. static void __init setup_command_line(char *command_line)
  302. {
  303. saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
  304. static_command_line = alloc_bootmem(strlen (command_line)+1);
  305. strcpy (saved_command_line, boot_command_line);
  306. strcpy (static_command_line, command_line);
  307. }
  308. /*
  309. * We need to finalize in a non-__init function or else race conditions
  310. * between the root thread and the init thread may cause start_kernel to
  311. * be reaped by free_initmem before the root thread has proceeded to
  312. * cpu_idle.
  313. *
  314. * gcc-3.4 accidentally inlines this function, so use noinline.
  315. */
  316. static __initdata DECLARE_COMPLETION(kthreadd_done);
  317. static noinline void __init_refok rest_init(void)
  318. {
  319. int pid;
  320. rcu_scheduler_starting();
  321. /*
  322. * We need to spawn init first so that it obtains pid 1, however
  323. * the init task will end up wanting to create kthreads, which, if
  324. * we schedule it before we create kthreadd, will OOPS.
  325. */
  326. kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
  327. numa_default_policy();
  328. pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
  329. rcu_read_lock();
  330. kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
  331. rcu_read_unlock();
  332. complete(&kthreadd_done);
  333. /*
  334. * The boot idle thread must execute schedule()
  335. * at least once to get things moving:
  336. */
  337. init_idle_bootup_task(current);
  338. schedule_preempt_disabled();
  339. /* Call into cpu_idle with preempt disabled */
  340. cpu_idle();
  341. }
  342. /* Check for early params. */
  343. static int __init do_early_param(char *param, char *val, const char *unused)
  344. {
  345. const struct obs_kernel_param *p;
  346. for (p = __setup_start; p < __setup_end; p++) {
  347. if ((p->early && parameq(param, p->str)) ||
  348. (strcmp(param, "console") == 0 &&
  349. strcmp(p->str, "earlycon") == 0)
  350. ) {
  351. if (p->setup_func(val) != 0)
  352. printk(KERN_WARNING
  353. "Malformed early option '%s'\n", param);
  354. }
  355. }
  356. /* We accept everything at this stage. */
  357. return 0;
  358. }
  359. void __init parse_early_options(char *cmdline)
  360. {
  361. parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
  362. }
  363. /* Arch code calls this early on, or if not, just before other parsing. */
  364. void __init parse_early_param(void)
  365. {
  366. static __initdata int done = 0;
  367. static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
  368. if (done)
  369. return;
  370. /* All fall through to do_early_param. */
  371. strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
  372. parse_early_options(tmp_cmdline);
  373. done = 1;
  374. }
  375. /*
  376. * Activate the first processor.
  377. */
  378. static void __init boot_cpu_init(void)
  379. {
  380. int cpu = smp_processor_id();
  381. /* Mark the boot cpu "present", "online" etc for SMP and UP case */
  382. set_cpu_online(cpu, true);
  383. set_cpu_active(cpu, true);
  384. set_cpu_present(cpu, true);
  385. set_cpu_possible(cpu, true);
  386. }
  387. void __init __weak smp_setup_processor_id(void)
  388. {
  389. }
  390. # if THREAD_SIZE >= PAGE_SIZE
  391. void __init __weak thread_info_cache_init(void)
  392. {
  393. }
  394. #endif
  395. /*
  396. * Set up kernel memory allocators
  397. */
  398. static void __init mm_init(void)
  399. {
  400. /*
  401. * page_cgroup requires contiguous pages,
  402. * bigger than MAX_ORDER unless SPARSEMEM.
  403. */
  404. page_cgroup_init_flatmem();
  405. mem_init();
  406. kmem_cache_init();
  407. percpu_init_late();
  408. pgtable_cache_init();
  409. vmalloc_init();
  410. }
  411. asmlinkage void __init start_kernel(void)
  412. {
  413. char * command_line;
  414. extern const struct kernel_param __start___param[], __stop___param[];
  415. /*
  416. * Need to run as early as possible, to initialize the
  417. * lockdep hash:
  418. */
  419. lockdep_init();
  420. smp_setup_processor_id();
  421. debug_objects_early_init();
  422. /*
  423. * Set up the the initial canary ASAP:
  424. */
  425. boot_init_stack_canary();
  426. cgroup_init_early();
  427. local_irq_disable();
  428. early_boot_irqs_disabled = true;
  429. /*
  430. * Interrupts are still disabled. Do necessary setups, then
  431. * enable them
  432. */
  433. tick_init();
  434. boot_cpu_init();
  435. page_address_init();
  436. printk(KERN_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. printk(KERN_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 (!irqs_disabled()) {
  474. printk(KERN_WARNING "start_kernel(): bug: interrupts were "
  475. "enabled *very* early, fixing it\n");
  476. local_irq_disable();
  477. }
  478. idr_init_cache();
  479. perf_event_init();
  480. rcu_init();
  481. radix_tree_init();
  482. /* init some links before init_ISA_irqs() */
  483. early_irq_init();
  484. init_IRQ();
  485. init_timers();
  486. hrtimers_init();
  487. softirq_init();
  488. timekeeping_init();
  489. time_init();
  490. profile_init();
  491. call_function_init();
  492. if (!irqs_disabled())
  493. printk(KERN_CRIT "start_kernel(): bug: interrupts were "
  494. "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. printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
  517. "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 char msgbuf[64];
  580. static int __init_or_module do_one_initcall_debug(initcall_t fn)
  581. {
  582. ktime_t calltime, delta, rettime;
  583. unsigned long long duration;
  584. int ret;
  585. printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current));
  586. calltime = ktime_get();
  587. ret = fn();
  588. rettime = ktime_get();
  589. delta = ktime_sub(rettime, calltime);
  590. duration = (unsigned long long) ktime_to_ns(delta) >> 10;
  591. printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", fn,
  592. ret, duration);
  593. return ret;
  594. }
  595. int __init_or_module do_one_initcall(initcall_t fn)
  596. {
  597. int count = preempt_count();
  598. int ret;
  599. if (initcall_debug)
  600. ret = do_one_initcall_debug(fn);
  601. else
  602. ret = fn();
  603. msgbuf[0] = 0;
  604. if (ret && ret != -ENODEV && initcall_debug)
  605. sprintf(msgbuf, "error code %d ", ret);
  606. if (preempt_count() != count) {
  607. strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
  608. preempt_count() = count;
  609. }
  610. if (irqs_disabled()) {
  611. strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
  612. local_irq_enable();
  613. }
  614. if (msgbuf[0]) {
  615. printk("initcall %pF returned with %s\n", fn, msgbuf);
  616. }
  617. return ret;
  618. }
  619. extern initcall_t __initcall_start[];
  620. extern initcall_t __initcall0_start[];
  621. extern initcall_t __initcall1_start[];
  622. extern initcall_t __initcall2_start[];
  623. extern initcall_t __initcall3_start[];
  624. extern initcall_t __initcall4_start[];
  625. extern initcall_t __initcall5_start[];
  626. extern initcall_t __initcall6_start[];
  627. extern initcall_t __initcall7_start[];
  628. extern initcall_t __initcall_end[];
  629. static initcall_t *initcall_levels[] __initdata = {
  630. __initcall0_start,
  631. __initcall1_start,
  632. __initcall2_start,
  633. __initcall3_start,
  634. __initcall4_start,
  635. __initcall5_start,
  636. __initcall6_start,
  637. __initcall7_start,
  638. __initcall_end,
  639. };
  640. /* Keep these in sync with initcalls in include/linux/init.h */
  641. static char *initcall_level_names[] __initdata = {
  642. "early",
  643. "core",
  644. "postcore",
  645. "arch",
  646. "subsys",
  647. "fs",
  648. "device",
  649. "late",
  650. };
  651. static void __init do_initcall_level(int level)
  652. {
  653. extern const struct kernel_param __start___param[], __stop___param[];
  654. initcall_t *fn;
  655. strcpy(static_command_line, saved_command_line);
  656. parse_args(initcall_level_names[level],
  657. static_command_line, __start___param,
  658. __stop___param - __start___param,
  659. level, level,
  660. &repair_env_string);
  661. for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
  662. do_one_initcall(*fn);
  663. }
  664. static void __init do_initcalls(void)
  665. {
  666. int level;
  667. for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
  668. do_initcall_level(level);
  669. }
  670. /*
  671. * Ok, the machine is now initialized. None of the devices
  672. * have been touched yet, but the CPU subsystem is up and
  673. * running, and memory and process management works.
  674. *
  675. * Now we can finally start doing some real work..
  676. */
  677. static void __init do_basic_setup(void)
  678. {
  679. cpuset_init_smp();
  680. usermodehelper_init();
  681. shmem_init();
  682. driver_init();
  683. init_irq_proc();
  684. do_ctors();
  685. usermodehelper_enable();
  686. do_initcalls();
  687. }
  688. static void __init do_pre_smp_initcalls(void)
  689. {
  690. initcall_t *fn;
  691. for (fn = __initcall_start; fn < __initcall0_start; fn++)
  692. do_one_initcall(*fn);
  693. }
  694. /*
  695. * This function requests modules which should be loaded by default and is
  696. * called twice right after initrd is mounted and right before init is
  697. * exec'd. If such modules are on either initrd or rootfs, they will be
  698. * loaded before control is passed to userland.
  699. */
  700. void __init load_default_modules(void)
  701. {
  702. load_default_elevator_module();
  703. }
  704. static int run_init_process(const char *init_filename)
  705. {
  706. argv_init[0] = init_filename;
  707. return do_execve(init_filename,
  708. (const char __user *const __user *)argv_init,
  709. (const char __user *const __user *)envp_init);
  710. }
  711. static noinline void __init kernel_init_freeable(void);
  712. static int __ref kernel_init(void *unused)
  713. {
  714. kernel_init_freeable();
  715. /* need to finish all async __init code before freeing the memory */
  716. async_synchronize_full();
  717. free_initmem();
  718. mark_rodata_ro();
  719. system_state = SYSTEM_RUNNING;
  720. numa_default_policy();
  721. flush_delayed_fput();
  722. if (ramdisk_execute_command) {
  723. if (!run_init_process(ramdisk_execute_command))
  724. return 0;
  725. printk(KERN_WARNING "Failed to execute %s\n",
  726. 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. printk(KERN_WARNING "Failed to execute %s. Attempting "
  738. "defaults...\n", 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. printk(KERN_WARNING "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. }