main.c 20 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821
  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/string.h>
  17. #include <linux/ctype.h>
  18. #include <linux/delay.h>
  19. #include <linux/utsname.h>
  20. #include <linux/ioport.h>
  21. #include <linux/init.h>
  22. #include <linux/smp_lock.h>
  23. #include <linux/initrd.h>
  24. #include <linux/hdreg.h>
  25. #include <linux/bootmem.h>
  26. #include <linux/tty.h>
  27. #include <linux/gfp.h>
  28. #include <linux/percpu.h>
  29. #include <linux/kmod.h>
  30. #include <linux/kernel_stat.h>
  31. #include <linux/start_kernel.h>
  32. #include <linux/security.h>
  33. #include <linux/workqueue.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/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/unwind.h>
  51. #include <linux/buffer_head.h>
  52. #include <linux/debug_locks.h>
  53. #include <linux/lockdep.h>
  54. #include <linux/pid_namespace.h>
  55. #include <linux/device.h>
  56. #include <asm/io.h>
  57. #include <asm/bugs.h>
  58. #include <asm/setup.h>
  59. #include <asm/sections.h>
  60. #include <asm/cacheflush.h>
  61. #ifdef CONFIG_X86_LOCAL_APIC
  62. #include <asm/smp.h>
  63. #endif
  64. /*
  65. * This is one of the first .c files built. Error out early if we have compiler
  66. * trouble.
  67. *
  68. * Versions of gcc older than that listed below may actually compile and link
  69. * okay, but the end product can have subtle run time bugs. To avoid associated
  70. * bogus bug reports, we flatly refuse to compile with a gcc that is known to be
  71. * too old from the very beginning.
  72. */
  73. #if (__GNUC__ < 3) || (__GNUC__ == 3 && __GNUC_MINOR__ < 2)
  74. #error Sorry, your GCC is too old. It builds incorrect kernels.
  75. #endif
  76. #if __GNUC__ == 4 && __GNUC_MINOR__ == 1 && __GNUC_PATCHLEVEL__ == 0
  77. #warning gcc-4.1.0 is known to miscompile the kernel. A different compiler version is recommended.
  78. #endif
  79. static int init(void *);
  80. extern void init_IRQ(void);
  81. extern void fork_init(unsigned long);
  82. extern void mca_init(void);
  83. extern void sbus_init(void);
  84. extern void signals_init(void);
  85. extern void pidhash_init(void);
  86. extern void pidmap_init(void);
  87. extern void prio_tree_init(void);
  88. extern void radix_tree_init(void);
  89. extern void free_initmem(void);
  90. extern void prepare_namespace(void);
  91. #ifdef CONFIG_ACPI
  92. extern void acpi_early_init(void);
  93. #else
  94. static inline void acpi_early_init(void) { }
  95. #endif
  96. #ifndef CONFIG_DEBUG_RODATA
  97. static inline void mark_rodata_ro(void) { }
  98. #endif
  99. #ifdef CONFIG_TC
  100. extern void tc_init(void);
  101. #endif
  102. enum system_states system_state;
  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 (*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. /* Setup configured maximum number of CPUs to activate */
  122. static unsigned int max_cpus = NR_CPUS;
  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. /*
  135. * Setup routine for controlling SMP activation
  136. *
  137. * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
  138. * activation entirely (the MPS table probe still happens, though).
  139. *
  140. * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
  141. * greater than 0, limits the maximum number of CPUs activated in
  142. * SMP mode to <NUM>.
  143. */
  144. static int __init nosmp(char *str)
  145. {
  146. max_cpus = 0;
  147. return 1;
  148. }
  149. __setup("nosmp", nosmp);
  150. static int __init maxcpus(char *str)
  151. {
  152. get_option(&str, &max_cpus);
  153. return 1;
  154. }
  155. __setup("maxcpus=", maxcpus);
  156. static int __init set_reset_devices(char *str)
  157. {
  158. reset_devices = 1;
  159. return 1;
  160. }
  161. __setup("reset_devices", set_reset_devices);
  162. static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
  163. char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
  164. static const char *panic_later, *panic_param;
  165. extern struct obs_kernel_param __setup_start[], __setup_end[];
  166. static int __init obsolete_checksetup(char *line)
  167. {
  168. struct obs_kernel_param *p;
  169. int had_early_param = 0;
  170. p = __setup_start;
  171. do {
  172. int n = strlen(p->str);
  173. if (!strncmp(line, p->str, n)) {
  174. if (p->early) {
  175. /* Already done in parse_early_param?
  176. * (Needs exact match on param part).
  177. * Keep iterating, as we can have early
  178. * params and __setups of same names 8( */
  179. if (line[n] == '\0' || line[n] == '=')
  180. had_early_param = 1;
  181. } else if (!p->setup_func) {
  182. printk(KERN_WARNING "Parameter %s is obsolete,"
  183. " ignored\n", p->str);
  184. return 1;
  185. } else if (p->setup_func(line + n))
  186. return 1;
  187. }
  188. p++;
  189. } while (p < __setup_end);
  190. return had_early_param;
  191. }
  192. /*
  193. * This should be approx 2 Bo*oMips to start (note initial shift), and will
  194. * still work even if initially too large, it will just take slightly longer
  195. */
  196. unsigned long loops_per_jiffy = (1<<12);
  197. EXPORT_SYMBOL(loops_per_jiffy);
  198. static int __init debug_kernel(char *str)
  199. {
  200. if (*str)
  201. return 0;
  202. console_loglevel = 10;
  203. return 1;
  204. }
  205. static int __init quiet_kernel(char *str)
  206. {
  207. if (*str)
  208. return 0;
  209. console_loglevel = 4;
  210. return 1;
  211. }
  212. __setup("debug", debug_kernel);
  213. __setup("quiet", quiet_kernel);
  214. static int __init loglevel(char *str)
  215. {
  216. get_option(&str, &console_loglevel);
  217. return 1;
  218. }
  219. __setup("loglevel=", loglevel);
  220. /*
  221. * Unknown boot options get handed to init, unless they look like
  222. * failed parameters
  223. */
  224. static int __init unknown_bootoption(char *param, char *val)
  225. {
  226. /* Change NUL term back to "=", to make "param" the whole string. */
  227. if (val) {
  228. /* param=val or param="val"? */
  229. if (val == param+strlen(param)+1)
  230. val[-1] = '=';
  231. else if (val == param+strlen(param)+2) {
  232. val[-2] = '=';
  233. memmove(val-1, val, strlen(val)+1);
  234. val--;
  235. } else
  236. BUG();
  237. }
  238. /* Handle obsolete-style parameters */
  239. if (obsolete_checksetup(param))
  240. return 0;
  241. /*
  242. * Preemptive maintenance for "why didn't my mispelled command
  243. * line work?"
  244. */
  245. if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
  246. printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
  247. return 0;
  248. }
  249. if (panic_later)
  250. return 0;
  251. if (val) {
  252. /* Environment option */
  253. unsigned int i;
  254. for (i = 0; envp_init[i]; i++) {
  255. if (i == MAX_INIT_ENVS) {
  256. panic_later = "Too many boot env vars at `%s'";
  257. panic_param = param;
  258. }
  259. if (!strncmp(param, envp_init[i], val - param))
  260. break;
  261. }
  262. envp_init[i] = param;
  263. } else {
  264. /* Command line option */
  265. unsigned int i;
  266. for (i = 0; argv_init[i]; i++) {
  267. if (i == MAX_INIT_ARGS) {
  268. panic_later = "Too many boot init vars at `%s'";
  269. panic_param = param;
  270. }
  271. }
  272. argv_init[i] = param;
  273. }
  274. return 0;
  275. }
  276. static int __init init_setup(char *str)
  277. {
  278. unsigned int i;
  279. execute_command = str;
  280. /*
  281. * In case LILO is going to boot us with default command line,
  282. * it prepends "auto" before the whole cmdline which makes
  283. * the shell think it should execute a script with such name.
  284. * So we ignore all arguments entered _before_ init=... [MJ]
  285. */
  286. for (i = 1; i < MAX_INIT_ARGS; i++)
  287. argv_init[i] = NULL;
  288. return 1;
  289. }
  290. __setup("init=", init_setup);
  291. static int __init rdinit_setup(char *str)
  292. {
  293. unsigned int i;
  294. ramdisk_execute_command = str;
  295. /* See "auto" comment in init_setup */
  296. for (i = 1; i < MAX_INIT_ARGS; i++)
  297. argv_init[i] = NULL;
  298. return 1;
  299. }
  300. __setup("rdinit=", rdinit_setup);
  301. #ifndef CONFIG_SMP
  302. #ifdef CONFIG_X86_LOCAL_APIC
  303. static void __init smp_init(void)
  304. {
  305. APIC_init_uniprocessor();
  306. }
  307. #else
  308. #define smp_init() do { } while (0)
  309. #endif
  310. static inline void setup_per_cpu_areas(void) { }
  311. static inline void smp_prepare_cpus(unsigned int maxcpus) { }
  312. #else
  313. #ifdef __GENERIC_PER_CPU
  314. unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
  315. EXPORT_SYMBOL(__per_cpu_offset);
  316. static void __init setup_per_cpu_areas(void)
  317. {
  318. unsigned long size, i;
  319. char *ptr;
  320. unsigned long nr_possible_cpus = num_possible_cpus();
  321. /* Copy section for each CPU (we discard the original) */
  322. size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
  323. #ifdef CONFIG_MODULES
  324. if (size < PERCPU_ENOUGH_ROOM)
  325. size = PERCPU_ENOUGH_ROOM;
  326. #endif
  327. ptr = alloc_bootmem(size * nr_possible_cpus);
  328. for_each_possible_cpu(i) {
  329. __per_cpu_offset[i] = ptr - __per_cpu_start;
  330. memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
  331. ptr += size;
  332. }
  333. }
  334. #endif /* !__GENERIC_PER_CPU */
  335. /* Called by boot processor to activate the rest. */
  336. static void __init smp_init(void)
  337. {
  338. unsigned int cpu;
  339. /* FIXME: This should be done in userspace --RR */
  340. for_each_present_cpu(cpu) {
  341. if (num_online_cpus() >= max_cpus)
  342. break;
  343. if (!cpu_online(cpu))
  344. cpu_up(cpu);
  345. }
  346. /* Any cleanup work */
  347. printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
  348. smp_cpus_done(max_cpus);
  349. }
  350. #endif
  351. /*
  352. * We need to store the untouched command line for future reference.
  353. * We also need to store the touched command line since the parameter
  354. * parsing is performed in place, and we should allow a component to
  355. * store reference of name/value for future reference.
  356. */
  357. static void __init setup_command_line(char *command_line)
  358. {
  359. saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
  360. static_command_line = alloc_bootmem(strlen (command_line)+1);
  361. strcpy (saved_command_line, boot_command_line);
  362. strcpy (static_command_line, command_line);
  363. }
  364. /*
  365. * We need to finalize in a non-__init function or else race conditions
  366. * between the root thread and the init thread may cause start_kernel to
  367. * be reaped by free_initmem before the root thread has proceeded to
  368. * cpu_idle.
  369. *
  370. * gcc-3.4 accidentally inlines this function, so use noinline.
  371. */
  372. static void noinline rest_init(void)
  373. __releases(kernel_lock)
  374. {
  375. kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
  376. numa_default_policy();
  377. unlock_kernel();
  378. /*
  379. * The boot idle thread must execute schedule()
  380. * at least once to get things moving:
  381. */
  382. init_idle_bootup_task(current);
  383. preempt_enable_no_resched();
  384. schedule();
  385. preempt_disable();
  386. /* Call into cpu_idle with preempt disabled */
  387. cpu_idle();
  388. }
  389. /* Check for early params. */
  390. static int __init do_early_param(char *param, char *val)
  391. {
  392. struct obs_kernel_param *p;
  393. for (p = __setup_start; p < __setup_end; p++) {
  394. if (p->early && strcmp(param, p->str) == 0) {
  395. if (p->setup_func(val) != 0)
  396. printk(KERN_WARNING
  397. "Malformed early option '%s'\n", param);
  398. }
  399. }
  400. /* We accept everything at this stage. */
  401. return 0;
  402. }
  403. /* Arch code calls this early on, or if not, just before other parsing. */
  404. void __init parse_early_param(void)
  405. {
  406. static __initdata int done = 0;
  407. static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
  408. if (done)
  409. return;
  410. /* All fall through to do_early_param. */
  411. strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
  412. parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
  413. done = 1;
  414. }
  415. /*
  416. * Activate the first processor.
  417. */
  418. static void __init boot_cpu_init(void)
  419. {
  420. int cpu = smp_processor_id();
  421. /* Mark the boot cpu "present", "online" etc for SMP and UP case */
  422. cpu_set(cpu, cpu_online_map);
  423. cpu_set(cpu, cpu_present_map);
  424. cpu_set(cpu, cpu_possible_map);
  425. }
  426. void __init __attribute__((weak)) smp_setup_processor_id(void)
  427. {
  428. }
  429. asmlinkage void __init start_kernel(void)
  430. {
  431. char * command_line;
  432. extern struct kernel_param __start___param[], __stop___param[];
  433. smp_setup_processor_id();
  434. /*
  435. * Need to run as early as possible, to initialize the
  436. * lockdep hash:
  437. */
  438. unwind_init();
  439. lockdep_init();
  440. local_irq_disable();
  441. early_boot_irqs_off();
  442. early_init_irq_lock_class();
  443. /*
  444. * Interrupts are still disabled. Do necessary setups, then
  445. * enable them
  446. */
  447. lock_kernel();
  448. tick_init();
  449. boot_cpu_init();
  450. page_address_init();
  451. printk(KERN_NOTICE);
  452. printk(linux_banner);
  453. setup_arch(&command_line);
  454. setup_command_line(command_line);
  455. unwind_setup();
  456. setup_per_cpu_areas();
  457. smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
  458. /*
  459. * Set up the scheduler prior starting any interrupts (such as the
  460. * timer interrupt). Full topology setup happens at smp_init()
  461. * time - but meanwhile we still have a functioning scheduler.
  462. */
  463. sched_init();
  464. /*
  465. * Disable preemption - early bootup scheduling is extremely
  466. * fragile until we cpu_idle() for the first time.
  467. */
  468. preempt_disable();
  469. build_all_zonelists();
  470. page_alloc_init();
  471. printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
  472. parse_early_param();
  473. parse_args("Booting kernel", static_command_line, __start___param,
  474. __stop___param - __start___param,
  475. &unknown_bootoption);
  476. if (!irqs_disabled()) {
  477. printk(KERN_WARNING "start_kernel(): bug: interrupts were "
  478. "enabled *very* early, fixing it\n");
  479. local_irq_disable();
  480. }
  481. sort_main_extable();
  482. trap_init();
  483. rcu_init();
  484. init_IRQ();
  485. pidhash_init();
  486. init_timers();
  487. hrtimers_init();
  488. softirq_init();
  489. timekeeping_init();
  490. time_init();
  491. profile_init();
  492. if (!irqs_disabled())
  493. printk("start_kernel(): bug: interrupts were enabled early\n");
  494. early_boot_irqs_on();
  495. local_irq_enable();
  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. initrd_start < min_low_pfn << PAGE_SHIFT) {
  514. printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
  515. "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
  516. initrd_start = 0;
  517. }
  518. #endif
  519. vfs_caches_init_early();
  520. cpuset_init_early();
  521. mem_init();
  522. kmem_cache_init();
  523. setup_per_cpu_pageset();
  524. numa_policy_init();
  525. if (late_time_init)
  526. late_time_init();
  527. calibrate_delay();
  528. pidmap_init();
  529. pgtable_cache_init();
  530. prio_tree_init();
  531. anon_vma_init();
  532. #ifdef CONFIG_X86
  533. if (efi_enabled)
  534. efi_enter_virtual_mode();
  535. #endif
  536. fork_init(num_physpages);
  537. proc_caches_init();
  538. buffer_init();
  539. unnamed_dev_init();
  540. key_init();
  541. security_init();
  542. vfs_caches_init(num_physpages);
  543. radix_tree_init();
  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. cpuset_init();
  551. taskstats_init_early();
  552. delayacct_init();
  553. check_bugs();
  554. acpi_early_init(); /* before LAPIC and SMP init */
  555. /* Do the rest non-__init'ed, we're now alive */
  556. rest_init();
  557. }
  558. static int __initdata initcall_debug;
  559. static int __init initcall_debug_setup(char *str)
  560. {
  561. initcall_debug = 1;
  562. return 1;
  563. }
  564. __setup("initcall_debug", initcall_debug_setup);
  565. extern initcall_t __initcall_start[], __initcall_end[];
  566. static void __init do_initcalls(void)
  567. {
  568. initcall_t *call;
  569. int count = preempt_count();
  570. for (call = __initcall_start; call < __initcall_end; call++) {
  571. char *msg = NULL;
  572. char msgbuf[40];
  573. int result;
  574. if (initcall_debug) {
  575. printk("Calling initcall 0x%p", *call);
  576. print_fn_descriptor_symbol(": %s()",
  577. (unsigned long) *call);
  578. printk("\n");
  579. }
  580. result = (*call)();
  581. if (result && result != -ENODEV && initcall_debug) {
  582. sprintf(msgbuf, "error code %d", result);
  583. msg = msgbuf;
  584. }
  585. if (preempt_count() != count) {
  586. msg = "preemption imbalance";
  587. preempt_count() = count;
  588. }
  589. if (irqs_disabled()) {
  590. msg = "disabled interrupts";
  591. local_irq_enable();
  592. }
  593. if (msg) {
  594. printk(KERN_WARNING "initcall at 0x%p", *call);
  595. print_fn_descriptor_symbol(": %s()",
  596. (unsigned long) *call);
  597. printk(": returned with %s\n", msg);
  598. }
  599. }
  600. /* Make sure there is no pending stuff from the initcall sequence */
  601. flush_scheduled_work();
  602. }
  603. /*
  604. * Ok, the machine is now initialized. None of the devices
  605. * have been touched yet, but the CPU subsystem is up and
  606. * running, and memory and process management works.
  607. *
  608. * Now we can finally start doing some real work..
  609. */
  610. static void __init do_basic_setup(void)
  611. {
  612. /* drivers will send hotplug events */
  613. init_workqueues();
  614. usermodehelper_init();
  615. driver_init();
  616. init_irq_proc();
  617. do_initcalls();
  618. }
  619. static void __init do_pre_smp_initcalls(void)
  620. {
  621. extern int spawn_ksoftirqd(void);
  622. #ifdef CONFIG_SMP
  623. extern int migration_init(void);
  624. migration_init();
  625. #endif
  626. spawn_ksoftirqd();
  627. spawn_softlockup_task();
  628. }
  629. static void run_init_process(char *init_filename)
  630. {
  631. argv_init[0] = init_filename;
  632. kernel_execve(init_filename, argv_init, envp_init);
  633. }
  634. /* This is a non __init function. Force it to be noinline otherwise gcc
  635. * makes it inline to init() and it becomes part of init.text section
  636. */
  637. static int noinline init_post(void)
  638. {
  639. free_initmem();
  640. unlock_kernel();
  641. mark_rodata_ro();
  642. system_state = SYSTEM_RUNNING;
  643. numa_default_policy();
  644. if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
  645. printk(KERN_WARNING "Warning: unable to open an initial console.\n");
  646. (void) sys_dup(0);
  647. (void) sys_dup(0);
  648. if (ramdisk_execute_command) {
  649. run_init_process(ramdisk_execute_command);
  650. printk(KERN_WARNING "Failed to execute %s\n",
  651. ramdisk_execute_command);
  652. }
  653. /*
  654. * We try each of these until one succeeds.
  655. *
  656. * The Bourne shell can be used instead of init if we are
  657. * trying to recover a really broken machine.
  658. */
  659. if (execute_command) {
  660. run_init_process(execute_command);
  661. printk(KERN_WARNING "Failed to execute %s. Attempting "
  662. "defaults...\n", execute_command);
  663. }
  664. run_init_process("/sbin/init");
  665. run_init_process("/etc/init");
  666. run_init_process("/bin/init");
  667. run_init_process("/bin/sh");
  668. panic("No init found. Try passing init= option to kernel.");
  669. }
  670. static int __init init(void * unused)
  671. {
  672. lock_kernel();
  673. /*
  674. * init can run on any cpu.
  675. */
  676. set_cpus_allowed(current, CPU_MASK_ALL);
  677. /*
  678. * Tell the world that we're going to be the grim
  679. * reaper of innocent orphaned children.
  680. *
  681. * We don't want people to have to make incorrect
  682. * assumptions about where in the task array this
  683. * can be found.
  684. */
  685. init_pid_ns.child_reaper = current;
  686. cad_pid = task_pid(current);
  687. smp_prepare_cpus(max_cpus);
  688. do_pre_smp_initcalls();
  689. smp_init();
  690. sched_init_smp();
  691. cpuset_init_smp();
  692. do_basic_setup();
  693. /*
  694. * check if there is an early userspace init. If yes, let it do all
  695. * the work
  696. */
  697. if (!ramdisk_execute_command)
  698. ramdisk_execute_command = "/init";
  699. if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
  700. ramdisk_execute_command = NULL;
  701. prepare_namespace();
  702. }
  703. /*
  704. * Ok, we have completed the initial bootup, and
  705. * we're essentially up and running. Get rid of the
  706. * initmem segments and start the user-mode stuff..
  707. */
  708. init_post();
  709. return 0;
  710. }