main.c 25 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037
  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 <linux/random.h>
  78. #include <linux/list.h>
  79. #include <linux/integrity.h>
  80. #include <asm/io.h>
  81. #include <asm/bugs.h>
  82. #include <asm/setup.h>
  83. #include <asm/sections.h>
  84. #include <asm/cacheflush.h>
  85. #ifdef CONFIG_X86_LOCAL_APIC
  86. #include <asm/smp.h>
  87. #endif
  88. static int kernel_init(void *);
  89. extern void init_IRQ(void);
  90. extern void fork_init(unsigned long);
  91. extern void radix_tree_init(void);
  92. #ifndef CONFIG_DEBUG_RODATA
  93. static inline void mark_rodata_ro(void) { }
  94. #endif
  95. /*
  96. * Debug helper: via this flag we know that we are in 'early bootup code'
  97. * where only the boot processor is running with IRQ disabled. This means
  98. * two things - IRQ must not be enabled before the flag is cleared and some
  99. * operations which are not allowed with IRQ disabled are allowed while the
  100. * flag is set.
  101. */
  102. bool early_boot_irqs_disabled __read_mostly;
  103. enum system_states system_state __read_mostly;
  104. EXPORT_SYMBOL(system_state);
  105. /*
  106. * Boot command-line arguments
  107. */
  108. #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
  109. #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
  110. extern void time_init(void);
  111. /* Default late time init is NULL. archs can override this later. */
  112. void (*__initdata late_time_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. /* Command line for per-initcall parameter parsing */
  120. static char *initcall_command_line;
  121. static char *execute_command;
  122. static char *ramdisk_execute_command;
  123. /*
  124. * Used to generate warnings if static_key manipulation functions are used
  125. * before jump_label_init is called.
  126. */
  127. bool static_key_initialized __read_mostly = false;
  128. EXPORT_SYMBOL_GPL(static_key_initialized);
  129. /*
  130. * If set, this is an indication to the drivers that reset the underlying
  131. * device before going ahead with the initialization otherwise driver might
  132. * rely on the BIOS and skip the reset operation.
  133. *
  134. * This is useful if kernel is booting in an unreliable environment.
  135. * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
  136. * skipped and devices will be in unknown state.
  137. */
  138. unsigned int reset_devices;
  139. EXPORT_SYMBOL(reset_devices);
  140. static int __init set_reset_devices(char *str)
  141. {
  142. reset_devices = 1;
  143. return 1;
  144. }
  145. __setup("reset_devices", set_reset_devices);
  146. static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
  147. const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
  148. static const char *panic_later, *panic_param;
  149. extern const struct obs_kernel_param __setup_start[], __setup_end[];
  150. static int __init obsolete_checksetup(char *line)
  151. {
  152. const struct obs_kernel_param *p;
  153. int had_early_param = 0;
  154. p = __setup_start;
  155. do {
  156. int n = strlen(p->str);
  157. if (parameqn(line, p->str, n)) {
  158. if (p->early) {
  159. /* Already done in parse_early_param?
  160. * (Needs exact match on param part).
  161. * Keep iterating, as we can have early
  162. * params and __setups of same names 8( */
  163. if (line[n] == '\0' || line[n] == '=')
  164. had_early_param = 1;
  165. } else if (!p->setup_func) {
  166. pr_warn("Parameter %s is obsolete, ignored\n",
  167. p->str);
  168. return 1;
  169. } else if (p->setup_func(line + n))
  170. return 1;
  171. }
  172. p++;
  173. } while (p < __setup_end);
  174. return had_early_param;
  175. }
  176. /*
  177. * This should be approx 2 Bo*oMips to start (note initial shift), and will
  178. * still work even if initially too large, it will just take slightly longer
  179. */
  180. unsigned long loops_per_jiffy = (1<<12);
  181. EXPORT_SYMBOL(loops_per_jiffy);
  182. static int __init debug_kernel(char *str)
  183. {
  184. console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
  185. return 0;
  186. }
  187. static int __init quiet_kernel(char *str)
  188. {
  189. console_loglevel = CONSOLE_LOGLEVEL_QUIET;
  190. return 0;
  191. }
  192. early_param("debug", debug_kernel);
  193. early_param("quiet", quiet_kernel);
  194. static int __init loglevel(char *str)
  195. {
  196. int newlevel;
  197. /*
  198. * Only update loglevel value when a correct setting was passed,
  199. * to prevent blind crashes (when loglevel being set to 0) that
  200. * are quite hard to debug
  201. */
  202. if (get_option(&str, &newlevel)) {
  203. console_loglevel = newlevel;
  204. return 0;
  205. }
  206. return -EINVAL;
  207. }
  208. early_param("loglevel", loglevel);
  209. /* Change NUL term back to "=", to make "param" the whole string. */
  210. static int __init repair_env_string(char *param, char *val, const char *unused)
  211. {
  212. if (val) {
  213. /* param=val or param="val"? */
  214. if (val == param+strlen(param)+1)
  215. val[-1] = '=';
  216. else if (val == param+strlen(param)+2) {
  217. val[-2] = '=';
  218. memmove(val-1, val, strlen(val)+1);
  219. val--;
  220. } else
  221. BUG();
  222. }
  223. return 0;
  224. }
  225. /* Anything after -- gets handed straight to init. */
  226. static int __init set_init_arg(char *param, char *val, const char *unused)
  227. {
  228. unsigned int i;
  229. if (panic_later)
  230. return 0;
  231. repair_env_string(param, val, unused);
  232. for (i = 0; argv_init[i]; i++) {
  233. if (i == MAX_INIT_ARGS) {
  234. panic_later = "init";
  235. panic_param = param;
  236. return 0;
  237. }
  238. }
  239. argv_init[i] = param;
  240. return 0;
  241. }
  242. /*
  243. * Unknown boot options get handed to init, unless they look like
  244. * unused parameters (modprobe will find them in /proc/cmdline).
  245. */
  246. static int __init unknown_bootoption(char *param, char *val, const char *unused)
  247. {
  248. repair_env_string(param, val, unused);
  249. /* Handle obsolete-style parameters */
  250. if (obsolete_checksetup(param))
  251. return 0;
  252. /* Unused module parameter. */
  253. if (strchr(param, '.') && (!val || strchr(param, '.') < val))
  254. return 0;
  255. if (panic_later)
  256. return 0;
  257. if (val) {
  258. /* Environment option */
  259. unsigned int i;
  260. for (i = 0; envp_init[i]; i++) {
  261. if (i == MAX_INIT_ENVS) {
  262. panic_later = "env";
  263. panic_param = param;
  264. }
  265. if (!strncmp(param, envp_init[i], val - param))
  266. break;
  267. }
  268. envp_init[i] = param;
  269. } else {
  270. /* Command line option */
  271. unsigned int i;
  272. for (i = 0; argv_init[i]; i++) {
  273. if (i == MAX_INIT_ARGS) {
  274. panic_later = "init";
  275. panic_param = param;
  276. }
  277. }
  278. argv_init[i] = param;
  279. }
  280. return 0;
  281. }
  282. static int __init init_setup(char *str)
  283. {
  284. unsigned int i;
  285. execute_command = str;
  286. /*
  287. * In case LILO is going to boot us with default command line,
  288. * it prepends "auto" before the whole cmdline which makes
  289. * the shell think it should execute a script with such name.
  290. * So we ignore all arguments entered _before_ init=... [MJ]
  291. */
  292. for (i = 1; i < MAX_INIT_ARGS; i++)
  293. argv_init[i] = NULL;
  294. return 1;
  295. }
  296. __setup("init=", init_setup);
  297. static int __init rdinit_setup(char *str)
  298. {
  299. unsigned int i;
  300. ramdisk_execute_command = str;
  301. /* See "auto" comment in init_setup */
  302. for (i = 1; i < MAX_INIT_ARGS; i++)
  303. argv_init[i] = NULL;
  304. return 1;
  305. }
  306. __setup("rdinit=", rdinit_setup);
  307. #ifndef CONFIG_SMP
  308. static const unsigned int setup_max_cpus = NR_CPUS;
  309. #ifdef CONFIG_X86_LOCAL_APIC
  310. static void __init smp_init(void)
  311. {
  312. APIC_init_uniprocessor();
  313. }
  314. #else
  315. #define smp_init() do { } while (0)
  316. #endif
  317. static inline void setup_nr_cpu_ids(void) { }
  318. static inline void smp_prepare_cpus(unsigned int maxcpus) { }
  319. #endif
  320. /*
  321. * We need to store the untouched command line for future reference.
  322. * We also need to store the touched command line since the parameter
  323. * parsing is performed in place, and we should allow a component to
  324. * store reference of name/value for future reference.
  325. */
  326. static void __init setup_command_line(char *command_line)
  327. {
  328. saved_command_line =
  329. memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
  330. initcall_command_line =
  331. memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
  332. static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0);
  333. strcpy (saved_command_line, boot_command_line);
  334. strcpy (static_command_line, command_line);
  335. }
  336. /*
  337. * We need to finalize in a non-__init function or else race conditions
  338. * between the root thread and the init thread may cause start_kernel to
  339. * be reaped by free_initmem before the root thread has proceeded to
  340. * cpu_idle.
  341. *
  342. * gcc-3.4 accidentally inlines this function, so use noinline.
  343. */
  344. static __initdata DECLARE_COMPLETION(kthreadd_done);
  345. static noinline void __init_refok rest_init(void)
  346. {
  347. int pid;
  348. rcu_scheduler_starting();
  349. /*
  350. * We need to spawn init first so that it obtains pid 1, however
  351. * the init task will end up wanting to create kthreads, which, if
  352. * we schedule it before we create kthreadd, will OOPS.
  353. */
  354. kernel_thread(kernel_init, NULL, CLONE_FS);
  355. numa_default_policy();
  356. pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
  357. rcu_read_lock();
  358. kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
  359. rcu_read_unlock();
  360. complete(&kthreadd_done);
  361. /*
  362. * The boot idle thread must execute schedule()
  363. * at least once to get things moving:
  364. */
  365. init_idle_bootup_task(current);
  366. schedule_preempt_disabled();
  367. /* Call into cpu_idle with preempt disabled */
  368. cpu_startup_entry(CPUHP_ONLINE);
  369. }
  370. /* Check for early params. */
  371. static int __init do_early_param(char *param, char *val, const char *unused)
  372. {
  373. const struct obs_kernel_param *p;
  374. for (p = __setup_start; p < __setup_end; p++) {
  375. if ((p->early && parameq(param, p->str)) ||
  376. (strcmp(param, "console") == 0 &&
  377. strcmp(p->str, "earlycon") == 0)
  378. ) {
  379. if (p->setup_func(val) != 0)
  380. pr_warn("Malformed early option '%s'\n", param);
  381. }
  382. }
  383. /* We accept everything at this stage. */
  384. return 0;
  385. }
  386. void __init parse_early_options(char *cmdline)
  387. {
  388. parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
  389. }
  390. /* Arch code calls this early on, or if not, just before other parsing. */
  391. void __init parse_early_param(void)
  392. {
  393. static __initdata int done = 0;
  394. static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
  395. if (done)
  396. return;
  397. /* All fall through to do_early_param. */
  398. strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
  399. parse_early_options(tmp_cmdline);
  400. done = 1;
  401. }
  402. /*
  403. * Activate the first processor.
  404. */
  405. static void __init boot_cpu_init(void)
  406. {
  407. int cpu = smp_processor_id();
  408. /* Mark the boot cpu "present", "online" etc for SMP and UP case */
  409. set_cpu_online(cpu, true);
  410. set_cpu_active(cpu, true);
  411. set_cpu_present(cpu, true);
  412. set_cpu_possible(cpu, true);
  413. }
  414. void __init __weak smp_setup_processor_id(void)
  415. {
  416. }
  417. # if THREAD_SIZE >= PAGE_SIZE
  418. void __init __weak thread_info_cache_init(void)
  419. {
  420. }
  421. #endif
  422. /*
  423. * Set up kernel memory allocators
  424. */
  425. static void __init mm_init(void)
  426. {
  427. /*
  428. * page_cgroup requires contiguous pages,
  429. * bigger than MAX_ORDER unless SPARSEMEM.
  430. */
  431. page_cgroup_init_flatmem();
  432. mem_init();
  433. kmem_cache_init();
  434. percpu_init_late();
  435. pgtable_init();
  436. vmalloc_init();
  437. }
  438. asmlinkage __visible void __init start_kernel(void)
  439. {
  440. char * command_line, *after_dashes;
  441. extern const struct kernel_param __start___param[], __stop___param[];
  442. /*
  443. * Need to run as early as possible, to initialize the
  444. * lockdep hash:
  445. */
  446. lockdep_init();
  447. smp_setup_processor_id();
  448. debug_objects_early_init();
  449. /*
  450. * Set up the the initial canary ASAP:
  451. */
  452. boot_init_stack_canary();
  453. cgroup_init_early();
  454. local_irq_disable();
  455. early_boot_irqs_disabled = true;
  456. /*
  457. * Interrupts are still disabled. Do necessary setups, then
  458. * enable them
  459. */
  460. boot_cpu_init();
  461. page_address_init();
  462. pr_notice("%s", linux_banner);
  463. setup_arch(&command_line);
  464. mm_init_cpumask(&init_mm);
  465. setup_command_line(command_line);
  466. setup_nr_cpu_ids();
  467. setup_per_cpu_areas();
  468. smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
  469. build_all_zonelists(NULL, NULL);
  470. page_alloc_init();
  471. pr_notice("Kernel command line: %s\n", boot_command_line);
  472. parse_early_param();
  473. after_dashes = parse_args("Booting kernel",
  474. static_command_line, __start___param,
  475. __stop___param - __start___param,
  476. -1, -1, &unknown_bootoption);
  477. if (after_dashes)
  478. parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
  479. set_init_arg);
  480. jump_label_init();
  481. /*
  482. * These use large bootmem allocations and must precede
  483. * kmem_cache_init()
  484. */
  485. setup_log_buf(0);
  486. pidhash_init();
  487. vfs_caches_init_early();
  488. sort_main_extable();
  489. trap_init();
  490. mm_init();
  491. /*
  492. * Set up the scheduler prior starting any interrupts (such as the
  493. * timer interrupt). Full topology setup happens at smp_init()
  494. * time - but meanwhile we still have a functioning scheduler.
  495. */
  496. sched_init();
  497. /*
  498. * Disable preemption - early bootup scheduling is extremely
  499. * fragile until we cpu_idle() for the first time.
  500. */
  501. preempt_disable();
  502. if (WARN(!irqs_disabled(), "Interrupts were enabled *very* early, fixing it\n"))
  503. local_irq_disable();
  504. idr_init_cache();
  505. rcu_init();
  506. tick_nohz_init();
  507. context_tracking_init();
  508. radix_tree_init();
  509. /* init some links before init_ISA_irqs() */
  510. early_irq_init();
  511. init_IRQ();
  512. tick_init();
  513. init_timers();
  514. hrtimers_init();
  515. softirq_init();
  516. timekeeping_init();
  517. time_init();
  518. sched_clock_postinit();
  519. perf_event_init();
  520. profile_init();
  521. call_function_init();
  522. WARN(!irqs_disabled(), "Interrupts were enabled early\n");
  523. early_boot_irqs_disabled = false;
  524. local_irq_enable();
  525. kmem_cache_init_late();
  526. /*
  527. * HACK ALERT! This is early. We're enabling the console before
  528. * we've done PCI setups etc, and console_init() must be aware of
  529. * this. But we do want output early, in case something goes wrong.
  530. */
  531. console_init();
  532. if (panic_later)
  533. panic("Too many boot %s vars at `%s'", panic_later,
  534. panic_param);
  535. lockdep_info();
  536. /*
  537. * Need to run this when irqs are enabled, because it wants
  538. * to self-test [hard/soft]-irqs on/off lock inversion bugs
  539. * too:
  540. */
  541. locking_selftest();
  542. #ifdef CONFIG_BLK_DEV_INITRD
  543. if (initrd_start && !initrd_below_start_ok &&
  544. page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
  545. pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
  546. page_to_pfn(virt_to_page((void *)initrd_start)),
  547. min_low_pfn);
  548. initrd_start = 0;
  549. }
  550. #endif
  551. page_cgroup_init();
  552. debug_objects_mem_init();
  553. kmemleak_init();
  554. setup_per_cpu_pageset();
  555. numa_policy_init();
  556. if (late_time_init)
  557. late_time_init();
  558. sched_clock_init();
  559. calibrate_delay();
  560. pidmap_init();
  561. anon_vma_init();
  562. acpi_early_init();
  563. #ifdef CONFIG_X86
  564. if (efi_enabled(EFI_RUNTIME_SERVICES))
  565. efi_enter_virtual_mode();
  566. #endif
  567. #ifdef CONFIG_X86_ESPFIX64
  568. /* Should be run before the first non-init thread is created */
  569. init_espfix_bsp();
  570. #endif
  571. thread_info_cache_init();
  572. cred_init();
  573. fork_init(totalram_pages);
  574. proc_caches_init();
  575. buffer_init();
  576. key_init();
  577. security_init();
  578. dbg_late_init();
  579. vfs_caches_init(totalram_pages);
  580. signals_init();
  581. /* rootfs populating might need page-writeback */
  582. page_writeback_init();
  583. proc_root_init();
  584. cgroup_init();
  585. cpuset_init();
  586. taskstats_init_early();
  587. delayacct_init();
  588. check_bugs();
  589. sfi_init_late();
  590. if (efi_enabled(EFI_RUNTIME_SERVICES)) {
  591. efi_late_init();
  592. efi_free_boot_services();
  593. }
  594. ftrace_init();
  595. /* Do the rest non-__init'ed, we're now alive */
  596. rest_init();
  597. }
  598. /* Call all constructor functions linked into the kernel. */
  599. static void __init do_ctors(void)
  600. {
  601. #ifdef CONFIG_CONSTRUCTORS
  602. ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
  603. for (; fn < (ctor_fn_t *) __ctors_end; fn++)
  604. (*fn)();
  605. #endif
  606. }
  607. bool initcall_debug;
  608. core_param(initcall_debug, initcall_debug, bool, 0644);
  609. #ifdef CONFIG_KALLSYMS
  610. struct blacklist_entry {
  611. struct list_head next;
  612. char *buf;
  613. };
  614. static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
  615. static int __init initcall_blacklist(char *str)
  616. {
  617. char *str_entry;
  618. struct blacklist_entry *entry;
  619. /* str argument is a comma-separated list of functions */
  620. do {
  621. str_entry = strsep(&str, ",");
  622. if (str_entry) {
  623. pr_debug("blacklisting initcall %s\n", str_entry);
  624. entry = alloc_bootmem(sizeof(*entry));
  625. entry->buf = alloc_bootmem(strlen(str_entry) + 1);
  626. strcpy(entry->buf, str_entry);
  627. list_add(&entry->next, &blacklisted_initcalls);
  628. }
  629. } while (str_entry);
  630. return 0;
  631. }
  632. static bool __init_or_module initcall_blacklisted(initcall_t fn)
  633. {
  634. struct list_head *tmp;
  635. struct blacklist_entry *entry;
  636. char *fn_name;
  637. fn_name = kasprintf(GFP_KERNEL, "%pf", fn);
  638. if (!fn_name)
  639. return false;
  640. list_for_each(tmp, &blacklisted_initcalls) {
  641. entry = list_entry(tmp, struct blacklist_entry, next);
  642. if (!strcmp(fn_name, entry->buf)) {
  643. pr_debug("initcall %s blacklisted\n", fn_name);
  644. kfree(fn_name);
  645. return true;
  646. }
  647. }
  648. kfree(fn_name);
  649. return false;
  650. }
  651. #else
  652. static int __init initcall_blacklist(char *str)
  653. {
  654. pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
  655. return 0;
  656. }
  657. static bool __init_or_module initcall_blacklisted(initcall_t fn)
  658. {
  659. return false;
  660. }
  661. #endif
  662. __setup("initcall_blacklist=", initcall_blacklist);
  663. static int __init_or_module do_one_initcall_debug(initcall_t fn)
  664. {
  665. ktime_t calltime, delta, rettime;
  666. unsigned long long duration;
  667. int ret;
  668. printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current));
  669. calltime = ktime_get();
  670. ret = fn();
  671. rettime = ktime_get();
  672. delta = ktime_sub(rettime, calltime);
  673. duration = (unsigned long long) ktime_to_ns(delta) >> 10;
  674. printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n",
  675. fn, ret, duration);
  676. return ret;
  677. }
  678. int __init_or_module do_one_initcall(initcall_t fn)
  679. {
  680. int count = preempt_count();
  681. int ret;
  682. char msgbuf[64];
  683. if (initcall_blacklisted(fn))
  684. return -EPERM;
  685. if (initcall_debug)
  686. ret = do_one_initcall_debug(fn);
  687. else
  688. ret = fn();
  689. msgbuf[0] = 0;
  690. if (preempt_count() != count) {
  691. sprintf(msgbuf, "preemption imbalance ");
  692. preempt_count_set(count);
  693. }
  694. if (irqs_disabled()) {
  695. strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
  696. local_irq_enable();
  697. }
  698. WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
  699. return ret;
  700. }
  701. extern initcall_t __initcall_start[];
  702. extern initcall_t __initcall0_start[];
  703. extern initcall_t __initcall1_start[];
  704. extern initcall_t __initcall2_start[];
  705. extern initcall_t __initcall3_start[];
  706. extern initcall_t __initcall4_start[];
  707. extern initcall_t __initcall5_start[];
  708. extern initcall_t __initcall6_start[];
  709. extern initcall_t __initcall7_start[];
  710. extern initcall_t __initcall_end[];
  711. static initcall_t *initcall_levels[] __initdata = {
  712. __initcall0_start,
  713. __initcall1_start,
  714. __initcall2_start,
  715. __initcall3_start,
  716. __initcall4_start,
  717. __initcall5_start,
  718. __initcall6_start,
  719. __initcall7_start,
  720. __initcall_end,
  721. };
  722. /* Keep these in sync with initcalls in include/linux/init.h */
  723. static char *initcall_level_names[] __initdata = {
  724. "early",
  725. "core",
  726. "postcore",
  727. "arch",
  728. "subsys",
  729. "fs",
  730. "device",
  731. "late",
  732. };
  733. static void __init do_initcall_level(int level)
  734. {
  735. extern const struct kernel_param __start___param[], __stop___param[];
  736. initcall_t *fn;
  737. strcpy(initcall_command_line, saved_command_line);
  738. parse_args(initcall_level_names[level],
  739. initcall_command_line, __start___param,
  740. __stop___param - __start___param,
  741. level, level,
  742. &repair_env_string);
  743. for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
  744. do_one_initcall(*fn);
  745. }
  746. static void __init do_initcalls(void)
  747. {
  748. int level;
  749. for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
  750. do_initcall_level(level);
  751. }
  752. /*
  753. * Ok, the machine is now initialized. None of the devices
  754. * have been touched yet, but the CPU subsystem is up and
  755. * running, and memory and process management works.
  756. *
  757. * Now we can finally start doing some real work..
  758. */
  759. static void __init do_basic_setup(void)
  760. {
  761. cpuset_init_smp();
  762. usermodehelper_init();
  763. shmem_init();
  764. driver_init();
  765. init_irq_proc();
  766. do_ctors();
  767. usermodehelper_enable();
  768. do_initcalls();
  769. random_int_secret_init();
  770. }
  771. static void __init do_pre_smp_initcalls(void)
  772. {
  773. initcall_t *fn;
  774. for (fn = __initcall_start; fn < __initcall0_start; fn++)
  775. do_one_initcall(*fn);
  776. }
  777. /*
  778. * This function requests modules which should be loaded by default and is
  779. * called twice right after initrd is mounted and right before init is
  780. * exec'd. If such modules are on either initrd or rootfs, they will be
  781. * loaded before control is passed to userland.
  782. */
  783. void __init load_default_modules(void)
  784. {
  785. load_default_elevator_module();
  786. }
  787. static int run_init_process(const char *init_filename)
  788. {
  789. argv_init[0] = init_filename;
  790. return do_execve(getname_kernel(init_filename),
  791. (const char __user *const __user *)argv_init,
  792. (const char __user *const __user *)envp_init);
  793. }
  794. static int try_to_run_init_process(const char *init_filename)
  795. {
  796. int ret;
  797. ret = run_init_process(init_filename);
  798. if (ret && ret != -ENOENT) {
  799. pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
  800. init_filename, ret);
  801. }
  802. return ret;
  803. }
  804. static noinline void __init kernel_init_freeable(void);
  805. static int __ref kernel_init(void *unused)
  806. {
  807. int ret;
  808. kernel_init_freeable();
  809. /* need to finish all async __init code before freeing the memory */
  810. async_synchronize_full();
  811. free_initmem();
  812. mark_rodata_ro();
  813. system_state = SYSTEM_RUNNING;
  814. numa_default_policy();
  815. flush_delayed_fput();
  816. if (ramdisk_execute_command) {
  817. ret = run_init_process(ramdisk_execute_command);
  818. if (!ret)
  819. return 0;
  820. pr_err("Failed to execute %s (error %d)\n",
  821. ramdisk_execute_command, ret);
  822. }
  823. /*
  824. * We try each of these until one succeeds.
  825. *
  826. * The Bourne shell can be used instead of init if we are
  827. * trying to recover a really broken machine.
  828. */
  829. if (execute_command) {
  830. ret = run_init_process(execute_command);
  831. if (!ret)
  832. return 0;
  833. pr_err("Failed to execute %s (error %d). Attempting defaults...\n",
  834. execute_command, ret);
  835. }
  836. if (!try_to_run_init_process("/sbin/init") ||
  837. !try_to_run_init_process("/etc/init") ||
  838. !try_to_run_init_process("/bin/init") ||
  839. !try_to_run_init_process("/bin/sh"))
  840. return 0;
  841. panic("No working init found. Try passing init= option to kernel. "
  842. "See Linux Documentation/init.txt for guidance.");
  843. }
  844. static noinline void __init kernel_init_freeable(void)
  845. {
  846. /*
  847. * Wait until kthreadd is all set-up.
  848. */
  849. wait_for_completion(&kthreadd_done);
  850. /* Now the scheduler is fully set up and can do blocking allocations */
  851. gfp_allowed_mask = __GFP_BITS_MASK;
  852. /*
  853. * init can allocate pages on any node
  854. */
  855. set_mems_allowed(node_states[N_MEMORY]);
  856. /*
  857. * init can run on any cpu.
  858. */
  859. set_cpus_allowed_ptr(current, cpu_all_mask);
  860. cad_pid = task_pid(current);
  861. smp_prepare_cpus(setup_max_cpus);
  862. do_pre_smp_initcalls();
  863. lockup_detector_init();
  864. smp_init();
  865. sched_init_smp();
  866. do_basic_setup();
  867. /* Open the /dev/console on the rootfs, this should never fail */
  868. if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
  869. pr_err("Warning: unable to open an initial console.\n");
  870. (void) sys_dup(0);
  871. (void) sys_dup(0);
  872. /*
  873. * check if there is an early userspace init. If yes, let it do all
  874. * the work
  875. */
  876. if (!ramdisk_execute_command)
  877. ramdisk_execute_command = "/init";
  878. if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
  879. ramdisk_execute_command = NULL;
  880. prepare_namespace();
  881. }
  882. /*
  883. * Ok, we have completed the initial bootup, and
  884. * we're essentially up and running. Get rid of the
  885. * initmem segments and start the user-mode stuff..
  886. *
  887. * rootfs is available now, try loading the public keys
  888. * and default modules
  889. */
  890. integrity_load_keys();
  891. load_default_modules();
  892. }