main.c 21 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. #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/utsname.h>
  21. #include <linux/ioport.h>
  22. #include <linux/init.h>
  23. #include <linux/smp_lock.h>
  24. #include <linux/initrd.h>
  25. #include <linux/bootmem.h>
  26. #include <linux/acpi.h>
  27. #include <linux/tty.h>
  28. #include <linux/gfp.h>
  29. #include <linux/percpu.h>
  30. #include <linux/kmod.h>
  31. #include <linux/vmalloc.h>
  32. #include <linux/kernel_stat.h>
  33. #include <linux/start_kernel.h>
  34. #include <linux/security.h>
  35. #include <linux/smp.h>
  36. #include <linux/workqueue.h>
  37. #include <linux/profile.h>
  38. #include <linux/rcupdate.h>
  39. #include <linux/moduleparam.h>
  40. #include <linux/kallsyms.h>
  41. #include <linux/writeback.h>
  42. #include <linux/cpu.h>
  43. #include <linux/cpuset.h>
  44. #include <linux/cgroup.h>
  45. #include <linux/efi.h>
  46. #include <linux/tick.h>
  47. #include <linux/interrupt.h>
  48. #include <linux/taskstats_kern.h>
  49. #include <linux/delayacct.h>
  50. #include <linux/unistd.h>
  51. #include <linux/rmap.h>
  52. #include <linux/mempolicy.h>
  53. #include <linux/key.h>
  54. #include <linux/buffer_head.h>
  55. #include <linux/page_cgroup.h>
  56. #include <linux/debug_locks.h>
  57. #include <linux/debugobjects.h>
  58. #include <linux/lockdep.h>
  59. #include <linux/kmemleak.h>
  60. #include <linux/pid_namespace.h>
  61. #include <linux/device.h>
  62. #include <linux/kthread.h>
  63. #include <linux/sched.h>
  64. #include <linux/signal.h>
  65. #include <linux/idr.h>
  66. #include <linux/ftrace.h>
  67. #include <linux/async.h>
  68. #include <linux/kmemcheck.h>
  69. #include <linux/kmemtrace.h>
  70. #include <linux/shmem_fs.h>
  71. #include <trace/boot.h>
  72. #include <asm/io.h>
  73. #include <asm/bugs.h>
  74. #include <asm/setup.h>
  75. #include <asm/sections.h>
  76. #include <asm/cacheflush.h>
  77. #ifdef CONFIG_X86_LOCAL_APIC
  78. #include <asm/smp.h>
  79. #endif
  80. static int kernel_init(void *);
  81. extern void init_IRQ(void);
  82. extern void fork_init(unsigned long);
  83. extern void mca_init(void);
  84. extern void sbus_init(void);
  85. extern void prio_tree_init(void);
  86. extern void radix_tree_init(void);
  87. extern void free_initmem(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. enum system_states system_state __read_mostly;
  95. EXPORT_SYMBOL(system_state);
  96. /*
  97. * Boot command-line arguments
  98. */
  99. #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
  100. #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
  101. extern void time_init(void);
  102. /* Default late time init is NULL. archs can override this later. */
  103. void (*__initdata late_time_init)(void);
  104. extern void softirq_init(void);
  105. /* Untouched command line saved by arch-specific code. */
  106. char __initdata boot_command_line[COMMAND_LINE_SIZE];
  107. /* Untouched saved command line (eg. for /proc) */
  108. char *saved_command_line;
  109. /* Command line for parameter parsing */
  110. static char *static_command_line;
  111. static char *execute_command;
  112. static char *ramdisk_execute_command;
  113. #ifdef CONFIG_SMP
  114. /* Setup configured maximum number of CPUs to activate */
  115. unsigned int __initdata setup_max_cpus = NR_CPUS;
  116. /*
  117. * Setup routine for controlling SMP activation
  118. *
  119. * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
  120. * activation entirely (the MPS table probe still happens, though).
  121. *
  122. * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
  123. * greater than 0, limits the maximum number of CPUs activated in
  124. * SMP mode to <NUM>.
  125. */
  126. void __weak arch_disable_smp_support(void) { }
  127. static int __init nosmp(char *str)
  128. {
  129. setup_max_cpus = 0;
  130. arch_disable_smp_support();
  131. return 0;
  132. }
  133. early_param("nosmp", nosmp);
  134. static int __init maxcpus(char *str)
  135. {
  136. get_option(&str, &setup_max_cpus);
  137. if (setup_max_cpus == 0)
  138. arch_disable_smp_support();
  139. return 0;
  140. }
  141. early_param("maxcpus", maxcpus);
  142. #else
  143. const unsigned int setup_max_cpus = NR_CPUS;
  144. #endif
  145. /*
  146. * If set, this is an indication to the drivers that reset the underlying
  147. * device before going ahead with the initialization otherwise driver might
  148. * rely on the BIOS and skip the reset operation.
  149. *
  150. * This is useful if kernel is booting in an unreliable environment.
  151. * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
  152. * skipped and devices will be in unknown state.
  153. */
  154. unsigned int reset_devices;
  155. EXPORT_SYMBOL(reset_devices);
  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. console_loglevel = 10;
  201. return 0;
  202. }
  203. static int __init quiet_kernel(char *str)
  204. {
  205. console_loglevel = 4;
  206. return 0;
  207. }
  208. early_param("debug", debug_kernel);
  209. early_param("quiet", quiet_kernel);
  210. static int __init loglevel(char *str)
  211. {
  212. get_option(&str, &console_loglevel);
  213. return 0;
  214. }
  215. early_param("loglevel", loglevel);
  216. /*
  217. * Unknown boot options get handed to init, unless they look like
  218. * failed parameters
  219. */
  220. static int __init unknown_bootoption(char *param, char *val)
  221. {
  222. /* Change NUL term back to "=", to make "param" the whole string. */
  223. if (val) {
  224. /* param=val or param="val"? */
  225. if (val == param+strlen(param)+1)
  226. val[-1] = '=';
  227. else if (val == param+strlen(param)+2) {
  228. val[-2] = '=';
  229. memmove(val-1, val, strlen(val)+1);
  230. val--;
  231. } else
  232. BUG();
  233. }
  234. /* Handle obsolete-style parameters */
  235. if (obsolete_checksetup(param))
  236. return 0;
  237. /*
  238. * Preemptive maintenance for "why didn't my misspelled command
  239. * line work?"
  240. */
  241. if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
  242. printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
  243. return 0;
  244. }
  245. if (panic_later)
  246. return 0;
  247. if (val) {
  248. /* Environment option */
  249. unsigned int i;
  250. for (i = 0; envp_init[i]; i++) {
  251. if (i == MAX_INIT_ENVS) {
  252. panic_later = "Too many boot env vars at `%s'";
  253. panic_param = param;
  254. }
  255. if (!strncmp(param, envp_init[i], val - param))
  256. break;
  257. }
  258. envp_init[i] = param;
  259. } else {
  260. /* Command line option */
  261. unsigned int i;
  262. for (i = 0; argv_init[i]; i++) {
  263. if (i == MAX_INIT_ARGS) {
  264. panic_later = "Too many boot init vars at `%s'";
  265. panic_param = param;
  266. }
  267. }
  268. argv_init[i] = param;
  269. }
  270. return 0;
  271. }
  272. #ifdef CONFIG_DEBUG_PAGEALLOC
  273. int __read_mostly debug_pagealloc_enabled = 0;
  274. #endif
  275. static int __init init_setup(char *str)
  276. {
  277. unsigned int i;
  278. execute_command = str;
  279. /*
  280. * In case LILO is going to boot us with default command line,
  281. * it prepends "auto" before the whole cmdline which makes
  282. * the shell think it should execute a script with such name.
  283. * So we ignore all arguments entered _before_ init=... [MJ]
  284. */
  285. for (i = 1; i < MAX_INIT_ARGS; i++)
  286. argv_init[i] = NULL;
  287. return 1;
  288. }
  289. __setup("init=", init_setup);
  290. static int __init rdinit_setup(char *str)
  291. {
  292. unsigned int i;
  293. ramdisk_execute_command = str;
  294. /* See "auto" comment in init_setup */
  295. for (i = 1; i < MAX_INIT_ARGS; i++)
  296. argv_init[i] = NULL;
  297. return 1;
  298. }
  299. __setup("rdinit=", rdinit_setup);
  300. #ifndef CONFIG_SMP
  301. #ifdef CONFIG_X86_LOCAL_APIC
  302. static void __init smp_init(void)
  303. {
  304. APIC_init_uniprocessor();
  305. }
  306. #else
  307. #define smp_init() do { } while (0)
  308. #endif
  309. static inline void setup_nr_cpu_ids(void) { }
  310. static inline void smp_prepare_cpus(unsigned int maxcpus) { }
  311. #else
  312. #if NR_CPUS > BITS_PER_LONG
  313. cpumask_t cpu_mask_all __read_mostly = CPU_MASK_ALL;
  314. EXPORT_SYMBOL(cpu_mask_all);
  315. #endif
  316. /* Setup number of possible processor ids */
  317. int nr_cpu_ids __read_mostly = NR_CPUS;
  318. EXPORT_SYMBOL(nr_cpu_ids);
  319. /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
  320. static void __init setup_nr_cpu_ids(void)
  321. {
  322. nr_cpu_ids = find_last_bit(cpumask_bits(cpu_possible_mask),NR_CPUS) + 1;
  323. }
  324. /* Called by boot processor to activate the rest. */
  325. static void __init smp_init(void)
  326. {
  327. unsigned int cpu;
  328. /*
  329. * Set up the current CPU as possible to migrate to.
  330. * The other ones will be done by cpu_up/cpu_down()
  331. */
  332. set_cpu_active(smp_processor_id(), true);
  333. /* FIXME: This should be done in userspace --RR */
  334. for_each_present_cpu(cpu) {
  335. if (num_online_cpus() >= setup_max_cpus)
  336. break;
  337. if (!cpu_online(cpu))
  338. cpu_up(cpu);
  339. }
  340. /* Any cleanup work */
  341. printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
  342. smp_cpus_done(setup_max_cpus);
  343. }
  344. #endif
  345. /*
  346. * We need to store the untouched command line for future reference.
  347. * We also need to store the touched command line since the parameter
  348. * parsing is performed in place, and we should allow a component to
  349. * store reference of name/value for future reference.
  350. */
  351. static void __init setup_command_line(char *command_line)
  352. {
  353. saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
  354. static_command_line = alloc_bootmem(strlen (command_line)+1);
  355. strcpy (saved_command_line, boot_command_line);
  356. strcpy (static_command_line, command_line);
  357. }
  358. /*
  359. * We need to finalize in a non-__init function or else race conditions
  360. * between the root thread and the init thread may cause start_kernel to
  361. * be reaped by free_initmem before the root thread has proceeded to
  362. * cpu_idle.
  363. *
  364. * gcc-3.4 accidentally inlines this function, so use noinline.
  365. */
  366. static noinline void __init_refok rest_init(void)
  367. __releases(kernel_lock)
  368. {
  369. int pid;
  370. rcu_scheduler_starting();
  371. kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
  372. numa_default_policy();
  373. pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
  374. kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
  375. unlock_kernel();
  376. /*
  377. * The boot idle thread must execute schedule()
  378. * at least once to get things moving:
  379. */
  380. init_idle_bootup_task(current);
  381. preempt_enable_no_resched();
  382. schedule();
  383. preempt_disable();
  384. /* Call into cpu_idle with preempt disabled */
  385. cpu_idle();
  386. }
  387. /* Check for early params. */
  388. static int __init do_early_param(char *param, char *val)
  389. {
  390. struct obs_kernel_param *p;
  391. for (p = __setup_start; p < __setup_end; p++) {
  392. if ((p->early && strcmp(param, p->str) == 0) ||
  393. (strcmp(param, "console") == 0 &&
  394. strcmp(p->str, "earlycon") == 0)
  395. ) {
  396. if (p->setup_func(val) != 0)
  397. printk(KERN_WARNING
  398. "Malformed early option '%s'\n", param);
  399. }
  400. }
  401. /* We accept everything at this stage. */
  402. return 0;
  403. }
  404. void __init parse_early_options(char *cmdline)
  405. {
  406. parse_args("early options", cmdline, NULL, 0, do_early_param);
  407. }
  408. /* Arch code calls this early on, or if not, just before other parsing. */
  409. void __init parse_early_param(void)
  410. {
  411. static __initdata int done = 0;
  412. static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
  413. if (done)
  414. return;
  415. /* All fall through to do_early_param. */
  416. strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
  417. parse_early_options(tmp_cmdline);
  418. done = 1;
  419. }
  420. /*
  421. * Activate the first processor.
  422. */
  423. static void __init boot_cpu_init(void)
  424. {
  425. int cpu = smp_processor_id();
  426. /* Mark the boot cpu "present", "online" etc for SMP and UP case */
  427. set_cpu_online(cpu, true);
  428. set_cpu_present(cpu, true);
  429. set_cpu_possible(cpu, true);
  430. }
  431. void __init __weak smp_setup_processor_id(void)
  432. {
  433. }
  434. void __init __weak thread_info_cache_init(void)
  435. {
  436. }
  437. /*
  438. * Set up kernel memory allocators
  439. */
  440. static void __init mm_init(void)
  441. {
  442. /*
  443. * page_cgroup requires countinous pages as memmap
  444. * and it's bigger than MAX_ORDER unless SPARSEMEM.
  445. */
  446. page_cgroup_init_flatmem();
  447. mem_init();
  448. kmem_cache_init();
  449. pgtable_cache_init();
  450. vmalloc_init();
  451. }
  452. asmlinkage void __init start_kernel(void)
  453. {
  454. char * command_line;
  455. extern struct kernel_param __start___param[], __stop___param[];
  456. smp_setup_processor_id();
  457. /*
  458. * Need to run as early as possible, to initialize the
  459. * lockdep hash:
  460. */
  461. lockdep_init();
  462. debug_objects_early_init();
  463. /*
  464. * Set up the the initial canary ASAP:
  465. */
  466. boot_init_stack_canary();
  467. cgroup_init_early();
  468. local_irq_disable();
  469. early_boot_irqs_off();
  470. early_init_irq_lock_class();
  471. /*
  472. * Interrupts are still disabled. Do necessary setups, then
  473. * enable them
  474. */
  475. lock_kernel();
  476. tick_init();
  477. boot_cpu_init();
  478. page_address_init();
  479. printk(KERN_NOTICE "%s", linux_banner);
  480. setup_arch(&command_line);
  481. mm_init_owner(&init_mm, &init_task);
  482. setup_command_line(command_line);
  483. setup_nr_cpu_ids();
  484. setup_per_cpu_areas();
  485. smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
  486. build_all_zonelists();
  487. page_alloc_init();
  488. printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
  489. parse_early_param();
  490. parse_args("Booting kernel", static_command_line, __start___param,
  491. __stop___param - __start___param,
  492. &unknown_bootoption);
  493. /*
  494. * These use large bootmem allocations and must precede
  495. * kmem_cache_init()
  496. */
  497. pidhash_init();
  498. vfs_caches_init_early();
  499. sort_main_extable();
  500. trap_init();
  501. mm_init();
  502. /*
  503. * Set up the scheduler prior starting any interrupts (such as the
  504. * timer interrupt). Full topology setup happens at smp_init()
  505. * time - but meanwhile we still have a functioning scheduler.
  506. */
  507. sched_init();
  508. /*
  509. * Disable preemption - early bootup scheduling is extremely
  510. * fragile until we cpu_idle() for the first time.
  511. */
  512. preempt_disable();
  513. if (!irqs_disabled()) {
  514. printk(KERN_WARNING "start_kernel(): bug: interrupts were "
  515. "enabled *very* early, fixing it\n");
  516. local_irq_disable();
  517. }
  518. rcu_init();
  519. /* init some links before init_ISA_irqs() */
  520. early_irq_init();
  521. init_IRQ();
  522. prio_tree_init();
  523. init_timers();
  524. hrtimers_init();
  525. softirq_init();
  526. timekeeping_init();
  527. time_init();
  528. profile_init();
  529. if (!irqs_disabled())
  530. printk(KERN_CRIT "start_kernel(): bug: interrupts were "
  531. "enabled early\n");
  532. early_boot_irqs_on();
  533. local_irq_enable();
  534. /* Interrupts are enabled now so all GFP allocations are safe. */
  535. set_gfp_allowed_mask(__GFP_BITS_MASK);
  536. kmem_cache_init_late();
  537. /*
  538. * HACK ALERT! This is early. We're enabling the console before
  539. * we've done PCI setups etc, and console_init() must be aware of
  540. * this. But we do want output early, in case something goes wrong.
  541. */
  542. console_init();
  543. if (panic_later)
  544. panic(panic_later, panic_param);
  545. lockdep_info();
  546. /*
  547. * Need to run this when irqs are enabled, because it wants
  548. * to self-test [hard/soft]-irqs on/off lock inversion bugs
  549. * too:
  550. */
  551. locking_selftest();
  552. #ifdef CONFIG_BLK_DEV_INITRD
  553. if (initrd_start && !initrd_below_start_ok &&
  554. page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
  555. printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
  556. "disabling it.\n",
  557. page_to_pfn(virt_to_page((void *)initrd_start)),
  558. min_low_pfn);
  559. initrd_start = 0;
  560. }
  561. #endif
  562. page_cgroup_init();
  563. enable_debug_pagealloc();
  564. kmemtrace_init();
  565. kmemleak_init();
  566. debug_objects_mem_init();
  567. idr_init_cache();
  568. setup_per_cpu_pageset();
  569. numa_policy_init();
  570. if (late_time_init)
  571. late_time_init();
  572. sched_clock_init();
  573. calibrate_delay();
  574. pidmap_init();
  575. anon_vma_init();
  576. #ifdef CONFIG_X86
  577. if (efi_enabled)
  578. efi_enter_virtual_mode();
  579. #endif
  580. thread_info_cache_init();
  581. cred_init();
  582. fork_init(totalram_pages);
  583. proc_caches_init();
  584. buffer_init();
  585. key_init();
  586. security_init();
  587. vfs_caches_init(totalram_pages);
  588. radix_tree_init();
  589. signals_init();
  590. /* rootfs populating might need page-writeback */
  591. page_writeback_init();
  592. #ifdef CONFIG_PROC_FS
  593. proc_root_init();
  594. #endif
  595. cgroup_init();
  596. cpuset_init();
  597. taskstats_init_early();
  598. delayacct_init();
  599. check_bugs();
  600. acpi_early_init(); /* before LAPIC and SMP init */
  601. ftrace_init();
  602. /* Do the rest non-__init'ed, we're now alive */
  603. rest_init();
  604. }
  605. /* Call all constructor functions linked into the kernel. */
  606. static void __init do_ctors(void)
  607. {
  608. #ifdef CONFIG_CONSTRUCTORS
  609. ctor_fn_t *call = (ctor_fn_t *) __ctors_start;
  610. for (; call < (ctor_fn_t *) __ctors_end; call++)
  611. (*call)();
  612. #endif
  613. }
  614. int initcall_debug;
  615. core_param(initcall_debug, initcall_debug, bool, 0644);
  616. static char msgbuf[64];
  617. static struct boot_trace_call call;
  618. static struct boot_trace_ret ret;
  619. int do_one_initcall(initcall_t fn)
  620. {
  621. int count = preempt_count();
  622. ktime_t calltime, delta, rettime;
  623. if (initcall_debug) {
  624. call.caller = task_pid_nr(current);
  625. printk("calling %pF @ %i\n", fn, call.caller);
  626. calltime = ktime_get();
  627. trace_boot_call(&call, fn);
  628. enable_boot_trace();
  629. }
  630. ret.result = fn();
  631. if (initcall_debug) {
  632. disable_boot_trace();
  633. rettime = ktime_get();
  634. delta = ktime_sub(rettime, calltime);
  635. ret.duration = (unsigned long long) ktime_to_ns(delta) >> 10;
  636. trace_boot_ret(&ret, fn);
  637. printk("initcall %pF returned %d after %Ld usecs\n", fn,
  638. ret.result, ret.duration);
  639. }
  640. msgbuf[0] = 0;
  641. if (ret.result && ret.result != -ENODEV && initcall_debug)
  642. sprintf(msgbuf, "error code %d ", ret.result);
  643. if (preempt_count() != count) {
  644. strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
  645. preempt_count() = count;
  646. }
  647. if (irqs_disabled()) {
  648. strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
  649. local_irq_enable();
  650. }
  651. if (msgbuf[0]) {
  652. printk("initcall %pF returned with %s\n", fn, msgbuf);
  653. }
  654. return ret.result;
  655. }
  656. extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
  657. static void __init do_initcalls(void)
  658. {
  659. initcall_t *call;
  660. for (call = __early_initcall_end; call < __initcall_end; call++)
  661. do_one_initcall(*call);
  662. /* Make sure there is no pending stuff from the initcall sequence */
  663. flush_scheduled_work();
  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. rcu_init_sched(); /* needed by module_init stage. */
  675. init_workqueues();
  676. cpuset_init_smp();
  677. usermodehelper_init();
  678. init_tmpfs();
  679. driver_init();
  680. init_irq_proc();
  681. do_ctors();
  682. do_initcalls();
  683. }
  684. static void __init do_pre_smp_initcalls(void)
  685. {
  686. initcall_t *call;
  687. for (call = __initcall_start; call < __early_initcall_end; call++)
  688. do_one_initcall(*call);
  689. }
  690. static void run_init_process(char *init_filename)
  691. {
  692. argv_init[0] = init_filename;
  693. kernel_execve(init_filename, argv_init, envp_init);
  694. }
  695. /* This is a non __init function. Force it to be noinline otherwise gcc
  696. * makes it inline to init() and it becomes part of init.text section
  697. */
  698. static noinline int init_post(void)
  699. __releases(kernel_lock)
  700. {
  701. /* need to finish all async __init code before freeing the memory */
  702. async_synchronize_full();
  703. free_initmem();
  704. unlock_kernel();
  705. mark_rodata_ro();
  706. system_state = SYSTEM_RUNNING;
  707. numa_default_policy();
  708. if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
  709. printk(KERN_WARNING "Warning: unable to open an initial console.\n");
  710. (void) sys_dup(0);
  711. (void) sys_dup(0);
  712. current->signal->flags |= SIGNAL_UNKILLABLE;
  713. if (ramdisk_execute_command) {
  714. run_init_process(ramdisk_execute_command);
  715. printk(KERN_WARNING "Failed to execute %s\n",
  716. ramdisk_execute_command);
  717. }
  718. /*
  719. * We try each of these until one succeeds.
  720. *
  721. * The Bourne shell can be used instead of init if we are
  722. * trying to recover a really broken machine.
  723. */
  724. if (execute_command) {
  725. run_init_process(execute_command);
  726. printk(KERN_WARNING "Failed to execute %s. Attempting "
  727. "defaults...\n", execute_command);
  728. }
  729. run_init_process("/sbin/init");
  730. run_init_process("/etc/init");
  731. run_init_process("/bin/init");
  732. run_init_process("/bin/sh");
  733. panic("No init found. Try passing init= option to kernel.");
  734. }
  735. static int __init kernel_init(void * unused)
  736. {
  737. lock_kernel();
  738. /*
  739. * init can allocate pages on any node
  740. */
  741. set_mems_allowed(node_possible_map);
  742. /*
  743. * init can run on any cpu.
  744. */
  745. set_cpus_allowed_ptr(current, cpu_all_mask);
  746. /*
  747. * Tell the world that we're going to be the grim
  748. * reaper of innocent orphaned children.
  749. *
  750. * We don't want people to have to make incorrect
  751. * assumptions about where in the task array this
  752. * can be found.
  753. */
  754. init_pid_ns.child_reaper = current;
  755. cad_pid = task_pid(current);
  756. smp_prepare_cpus(setup_max_cpus);
  757. do_pre_smp_initcalls();
  758. start_boot_trace();
  759. smp_init();
  760. sched_init_smp();
  761. do_basic_setup();
  762. /*
  763. * check if there is an early userspace init. If yes, let it do all
  764. * the work
  765. */
  766. if (!ramdisk_execute_command)
  767. ramdisk_execute_command = "/init";
  768. if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
  769. ramdisk_execute_command = NULL;
  770. prepare_namespace();
  771. }
  772. /*
  773. * Ok, we have completed the initial bootup, and
  774. * we're essentially up and running. Get rid of the
  775. * initmem segments and start the user-mode stuff..
  776. */
  777. init_post();
  778. return 0;
  779. }