4 * Copyright (C) 1991, 1992 Linus Torvalds
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>
12 #define DEBUG /* Enable initcall_debug */
14 #include <linux/types.h>
15 #include <linux/module.h>
16 #include <linux/proc_fs.h>
17 #include <linux/kernel.h>
18 #include <linux/syscalls.h>
19 #include <linux/stackprotector.h>
20 #include <linux/string.h>
21 #include <linux/ctype.h>
22 #include <linux/delay.h>
23 #include <linux/ioport.h>
24 #include <linux/init.h>
25 #include <linux/initrd.h>
26 #include <linux/bootmem.h>
27 #include <linux/acpi.h>
28 #include <linux/tty.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/profile.h>
37 #include <linux/rcupdate.h>
38 #include <linux/moduleparam.h>
39 #include <linux/kallsyms.h>
40 #include <linux/writeback.h>
41 #include <linux/cpu.h>
42 #include <linux/cpuset.h>
43 #include <linux/cgroup.h>
44 #include <linux/efi.h>
45 #include <linux/tick.h>
46 #include <linux/interrupt.h>
47 #include <linux/taskstats_kern.h>
48 #include <linux/delayacct.h>
49 #include <linux/unistd.h>
50 #include <linux/rmap.h>
51 #include <linux/mempolicy.h>
52 #include <linux/key.h>
53 #include <linux/buffer_head.h>
54 #include <linux/page_ext.h>
55 #include <linux/debug_locks.h>
56 #include <linux/debugobjects.h>
57 #include <linux/lockdep.h>
58 #include <linux/kmemleak.h>
59 #include <linux/pid_namespace.h>
60 #include <linux/device.h>
61 #include <linux/kthread.h>
62 #include <linux/sched.h>
63 #include <linux/signal.h>
64 #include <linux/idr.h>
65 #include <linux/kgdb.h>
66 #include <linux/ftrace.h>
67 #include <linux/async.h>
68 #include <linux/kmemcheck.h>
69 #include <linux/sfi.h>
70 #include <linux/shmem_fs.h>
71 #include <linux/slab.h>
72 #include <linux/perf_event.h>
73 #include <linux/file.h>
74 #include <linux/ptrace.h>
75 #include <linux/blkdev.h>
76 #include <linux/elevator.h>
77 #include <linux/sched_clock.h>
78 #include <linux/context_tracking.h>
79 #include <linux/random.h>
80 #include <linux/list.h>
81 #include <linux/integrity.h>
82 #include <linux/proc_ns.h>
86 #include <asm/setup.h>
87 #include <asm/sections.h>
88 #include <asm/cacheflush.h>
90 #ifdef CONFIG_X86_LOCAL_APIC
94 static int kernel_init(void *);
96 extern void init_IRQ(void);
97 extern void fork_init(unsigned long);
98 extern void radix_tree_init(void);
99 #ifndef CONFIG_DEBUG_RODATA
100 static inline void mark_rodata_ro(void) { }
104 * Debug helper: via this flag we know that we are in 'early bootup code'
105 * where only the boot processor is running with IRQ disabled. This means
106 * two things - IRQ must not be enabled before the flag is cleared and some
107 * operations which are not allowed with IRQ disabled are allowed while the
110 bool early_boot_irqs_disabled __read_mostly;
112 enum system_states system_state __read_mostly;
113 EXPORT_SYMBOL(system_state);
116 * Boot command-line arguments
118 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
119 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
121 extern void time_init(void);
122 /* Default late time init is NULL. archs can override this later. */
123 void (*__initdata late_time_init)(void);
125 /* Untouched command line saved by arch-specific code. */
126 char __initdata boot_command_line[COMMAND_LINE_SIZE];
127 /* Untouched saved command line (eg. for /proc) */
128 char *saved_command_line;
129 /* Command line for parameter parsing */
130 static char *static_command_line;
131 /* Command line for per-initcall parameter parsing */
132 static char *initcall_command_line;
134 static char *execute_command;
135 static char *ramdisk_execute_command;
138 * Used to generate warnings if static_key manipulation functions are used
139 * before jump_label_init is called.
141 bool static_key_initialized __read_mostly;
142 EXPORT_SYMBOL_GPL(static_key_initialized);
145 * If set, this is an indication to the drivers that reset the underlying
146 * device before going ahead with the initialization otherwise driver might
147 * rely on the BIOS and skip the reset operation.
149 * This is useful if kernel is booting in an unreliable environment.
150 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
151 * skipped and devices will be in unknown state.
153 unsigned int reset_devices;
154 EXPORT_SYMBOL(reset_devices);
156 static int __init set_reset_devices(char *str)
162 __setup("reset_devices", set_reset_devices);
164 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
165 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
166 static const char *panic_later, *panic_param;
168 extern const struct obs_kernel_param __setup_start[], __setup_end[];
170 static int __init obsolete_checksetup(char *line)
172 const struct obs_kernel_param *p;
173 int had_early_param = 0;
177 int n = strlen(p->str);
178 if (parameqn(line, p->str, n)) {
180 /* Already done in parse_early_param?
181 * (Needs exact match on param part).
182 * Keep iterating, as we can have early
183 * params and __setups of same names 8( */
184 if (line[n] == '\0' || line[n] == '=')
186 } else if (!p->setup_func) {
187 pr_warn("Parameter %s is obsolete, ignored\n",
190 } else if (p->setup_func(line + n))
194 } while (p < __setup_end);
196 return had_early_param;
200 * This should be approx 2 Bo*oMips to start (note initial shift), and will
201 * still work even if initially too large, it will just take slightly longer
203 unsigned long loops_per_jiffy = (1<<12);
204 EXPORT_SYMBOL(loops_per_jiffy);
206 static int __init debug_kernel(char *str)
208 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
212 static int __init quiet_kernel(char *str)
214 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
218 early_param("debug", debug_kernel);
219 early_param("quiet", quiet_kernel);
221 static int __init loglevel(char *str)
226 * Only update loglevel value when a correct setting was passed,
227 * to prevent blind crashes (when loglevel being set to 0) that
228 * are quite hard to debug
230 if (get_option(&str, &newlevel)) {
231 console_loglevel = newlevel;
238 early_param("loglevel", loglevel);
240 /* Change NUL term back to "=", to make "param" the whole string. */
241 static int __init repair_env_string(char *param, char *val, const char *unused)
244 /* param=val or param="val"? */
245 if (val == param+strlen(param)+1)
247 else if (val == param+strlen(param)+2) {
249 memmove(val-1, val, strlen(val)+1);
257 /* Anything after -- gets handed straight to init. */
258 static int __init set_init_arg(char *param, char *val, const char *unused)
265 repair_env_string(param, val, unused);
267 for (i = 0; argv_init[i]; i++) {
268 if (i == MAX_INIT_ARGS) {
269 panic_later = "init";
274 argv_init[i] = param;
279 * Unknown boot options get handed to init, unless they look like
280 * unused parameters (modprobe will find them in /proc/cmdline).
282 static int __init unknown_bootoption(char *param, char *val, const char *unused)
284 repair_env_string(param, val, unused);
286 /* Handle obsolete-style parameters */
287 if (obsolete_checksetup(param))
290 /* Unused module parameter. */
291 if (strchr(param, '.') && (!val || strchr(param, '.') < val))
298 /* Environment option */
300 for (i = 0; envp_init[i]; i++) {
301 if (i == MAX_INIT_ENVS) {
305 if (!strncmp(param, envp_init[i], val - param))
308 envp_init[i] = param;
310 /* Command line option */
312 for (i = 0; argv_init[i]; i++) {
313 if (i == MAX_INIT_ARGS) {
314 panic_later = "init";
318 argv_init[i] = param;
323 static int __init init_setup(char *str)
327 execute_command = str;
329 * In case LILO is going to boot us with default command line,
330 * it prepends "auto" before the whole cmdline which makes
331 * the shell think it should execute a script with such name.
332 * So we ignore all arguments entered _before_ init=... [MJ]
334 for (i = 1; i < MAX_INIT_ARGS; i++)
338 __setup("init=", init_setup);
340 static int __init rdinit_setup(char *str)
344 ramdisk_execute_command = str;
345 /* See "auto" comment in init_setup */
346 for (i = 1; i < MAX_INIT_ARGS; i++)
350 __setup("rdinit=", rdinit_setup);
353 static const unsigned int setup_max_cpus = NR_CPUS;
354 #ifdef CONFIG_X86_LOCAL_APIC
355 static void __init smp_init(void)
357 APIC_init_uniprocessor();
360 #define smp_init() do { } while (0)
363 static inline void setup_nr_cpu_ids(void) { }
364 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
368 * We need to store the untouched command line for future reference.
369 * We also need to store the touched command line since the parameter
370 * parsing is performed in place, and we should allow a component to
371 * store reference of name/value for future reference.
373 static void __init setup_command_line(char *command_line)
376 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
377 initcall_command_line =
378 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
379 static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0);
380 strcpy(saved_command_line, boot_command_line);
381 strcpy(static_command_line, command_line);
385 * We need to finalize in a non-__init function or else race conditions
386 * between the root thread and the init thread may cause start_kernel to
387 * be reaped by free_initmem before the root thread has proceeded to
390 * gcc-3.4 accidentally inlines this function, so use noinline.
393 static __initdata DECLARE_COMPLETION(kthreadd_done);
395 static noinline void __init_refok rest_init(void)
399 rcu_scheduler_starting();
401 * We need to spawn init first so that it obtains pid 1, however
402 * the init task will end up wanting to create kthreads, which, if
403 * we schedule it before we create kthreadd, will OOPS.
405 kernel_thread(kernel_init, NULL, CLONE_FS);
406 numa_default_policy();
407 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
409 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
411 complete(&kthreadd_done);
414 * The boot idle thread must execute schedule()
415 * at least once to get things moving:
417 init_idle_bootup_task(current);
418 schedule_preempt_disabled();
419 /* Call into cpu_idle with preempt disabled */
420 cpu_startup_entry(CPUHP_ONLINE);
423 /* Check for early params. */
424 static int __init do_early_param(char *param, char *val, const char *unused)
426 const struct obs_kernel_param *p;
428 for (p = __setup_start; p < __setup_end; p++) {
429 if ((p->early && parameq(param, p->str)) ||
430 (strcmp(param, "console") == 0 &&
431 strcmp(p->str, "earlycon") == 0)
433 if (p->setup_func(val) != 0)
434 pr_warn("Malformed early option '%s'\n", param);
437 /* We accept everything at this stage. */
441 void __init parse_early_options(char *cmdline)
443 parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
446 /* Arch code calls this early on, or if not, just before other parsing. */
447 void __init parse_early_param(void)
449 static int done __initdata;
450 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
455 /* All fall through to do_early_param. */
456 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
457 parse_early_options(tmp_cmdline);
462 * Activate the first processor.
465 static void __init boot_cpu_init(void)
467 int cpu = smp_processor_id();
468 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
469 set_cpu_online(cpu, true);
470 set_cpu_active(cpu, true);
471 set_cpu_present(cpu, true);
472 set_cpu_possible(cpu, true);
475 void __init __weak smp_setup_processor_id(void)
479 # if THREAD_SIZE >= PAGE_SIZE
480 void __init __weak thread_info_cache_init(void)
486 * Set up kernel memory allocators
488 static void __init mm_init(void)
491 * page_ext requires contiguous pages,
492 * bigger than MAX_ORDER unless SPARSEMEM.
494 page_ext_init_flatmem();
502 asmlinkage __visible void __init start_kernel(void)
508 * Need to run as early as possible, to initialize the
512 set_task_stack_end_magic(&init_task);
513 smp_setup_processor_id();
514 debug_objects_early_init();
517 * Set up the the initial canary ASAP:
519 boot_init_stack_canary();
524 early_boot_irqs_disabled = true;
527 * Interrupts are still disabled. Do necessary setups, then
532 pr_notice("%s", linux_banner);
533 setup_arch(&command_line);
534 mm_init_cpumask(&init_mm);
535 setup_command_line(command_line);
537 setup_per_cpu_areas();
538 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
540 build_all_zonelists(NULL, NULL);
543 pr_notice("Kernel command line: %s\n", boot_command_line);
545 after_dashes = parse_args("Booting kernel",
546 static_command_line, __start___param,
547 __stop___param - __start___param,
548 -1, -1, &unknown_bootoption);
549 if (!IS_ERR_OR_NULL(after_dashes))
550 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
556 * These use large bootmem allocations and must precede
561 vfs_caches_init_early();
567 * Set up the scheduler prior starting any interrupts (such as the
568 * timer interrupt). Full topology setup happens at smp_init()
569 * time - but meanwhile we still have a functioning scheduler.
573 * Disable preemption - early bootup scheduling is extremely
574 * fragile until we cpu_idle() for the first time.
577 if (WARN(!irqs_disabled(),
578 "Interrupts were enabled *very* early, fixing it\n"))
583 /* trace_printk() and trace points may be used after this */
586 context_tracking_init();
588 /* init some links before init_ISA_irqs() */
598 sched_clock_postinit();
601 call_function_init();
602 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
603 early_boot_irqs_disabled = false;
606 kmem_cache_init_late();
609 * HACK ALERT! This is early. We're enabling the console before
610 * we've done PCI setups etc, and console_init() must be aware of
611 * this. But we do want output early, in case something goes wrong.
615 panic("Too many boot %s vars at `%s'", panic_later,
621 * Need to run this when irqs are enabled, because it wants
622 * to self-test [hard/soft]-irqs on/off lock inversion bugs
627 #ifdef CONFIG_BLK_DEV_INITRD
628 if (initrd_start && !initrd_below_start_ok &&
629 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
630 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
631 page_to_pfn(virt_to_page((void *)initrd_start)),
637 debug_objects_mem_init();
639 setup_per_cpu_pageset();
649 if (efi_enabled(EFI_RUNTIME_SERVICES))
650 efi_enter_virtual_mode();
652 #ifdef CONFIG_X86_ESPFIX64
653 /* Should be run before the first non-init thread is created */
656 thread_info_cache_init();
658 fork_init(totalram_pages);
664 vfs_caches_init(totalram_pages);
666 /* rootfs populating might need page-writeback */
667 page_writeback_init();
672 taskstats_init_early();
679 if (efi_enabled(EFI_RUNTIME_SERVICES)) {
681 efi_free_boot_services();
686 /* Do the rest non-__init'ed, we're now alive */
690 /* Call all constructor functions linked into the kernel. */
691 static void __init do_ctors(void)
693 #ifdef CONFIG_CONSTRUCTORS
694 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
696 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
702 core_param(initcall_debug, initcall_debug, bool, 0644);
704 #ifdef CONFIG_KALLSYMS
705 struct blacklist_entry {
706 struct list_head next;
710 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
712 static int __init initcall_blacklist(char *str)
715 struct blacklist_entry *entry;
717 /* str argument is a comma-separated list of functions */
719 str_entry = strsep(&str, ",");
721 pr_debug("blacklisting initcall %s\n", str_entry);
722 entry = alloc_bootmem(sizeof(*entry));
723 entry->buf = alloc_bootmem(strlen(str_entry) + 1);
724 strcpy(entry->buf, str_entry);
725 list_add(&entry->next, &blacklisted_initcalls);
732 static bool __init_or_module initcall_blacklisted(initcall_t fn)
734 struct list_head *tmp;
735 struct blacklist_entry *entry;
738 fn_name = kasprintf(GFP_KERNEL, "%pf", fn);
742 list_for_each(tmp, &blacklisted_initcalls) {
743 entry = list_entry(tmp, struct blacklist_entry, next);
744 if (!strcmp(fn_name, entry->buf)) {
745 pr_debug("initcall %s blacklisted\n", fn_name);
755 static int __init initcall_blacklist(char *str)
757 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
761 static bool __init_or_module initcall_blacklisted(initcall_t fn)
766 __setup("initcall_blacklist=", initcall_blacklist);
768 static int __init_or_module do_one_initcall_debug(initcall_t fn)
770 ktime_t calltime, delta, rettime;
771 unsigned long long duration;
774 printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current));
775 calltime = ktime_get();
777 rettime = ktime_get();
778 delta = ktime_sub(rettime, calltime);
779 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
780 printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n",
786 int __init_or_module do_one_initcall(initcall_t fn)
788 int count = preempt_count();
792 if (initcall_blacklisted(fn))
796 ret = do_one_initcall_debug(fn);
802 if (preempt_count() != count) {
803 sprintf(msgbuf, "preemption imbalance ");
804 preempt_count_set(count);
806 if (irqs_disabled()) {
807 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
810 WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
816 extern initcall_t __initcall_start[];
817 extern initcall_t __initcall0_start[];
818 extern initcall_t __initcall1_start[];
819 extern initcall_t __initcall2_start[];
820 extern initcall_t __initcall3_start[];
821 extern initcall_t __initcall4_start[];
822 extern initcall_t __initcall5_start[];
823 extern initcall_t __initcall6_start[];
824 extern initcall_t __initcall7_start[];
825 extern initcall_t __initcall_end[];
827 static initcall_t *initcall_levels[] __initdata = {
839 /* Keep these in sync with initcalls in include/linux/init.h */
840 static char *initcall_level_names[] __initdata = {
851 static void __init do_initcall_level(int level)
855 strcpy(initcall_command_line, saved_command_line);
856 parse_args(initcall_level_names[level],
857 initcall_command_line, __start___param,
858 __stop___param - __start___param,
862 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
863 do_one_initcall(*fn);
866 static void __init do_initcalls(void)
870 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
871 do_initcall_level(level);
875 * Ok, the machine is now initialized. None of the devices
876 * have been touched yet, but the CPU subsystem is up and
877 * running, and memory and process management works.
879 * Now we can finally start doing some real work..
881 static void __init do_basic_setup(void)
884 usermodehelper_init();
889 usermodehelper_enable();
891 random_int_secret_init();
894 static void __init do_pre_smp_initcalls(void)
898 for (fn = __initcall_start; fn < __initcall0_start; fn++)
899 do_one_initcall(*fn);
903 * This function requests modules which should be loaded by default and is
904 * called twice right after initrd is mounted and right before init is
905 * exec'd. If such modules are on either initrd or rootfs, they will be
906 * loaded before control is passed to userland.
908 void __init load_default_modules(void)
910 load_default_elevator_module();
913 static int run_init_process(const char *init_filename)
915 argv_init[0] = init_filename;
916 return do_execve(getname_kernel(init_filename),
917 (const char __user *const __user *)argv_init,
918 (const char __user *const __user *)envp_init);
921 static int try_to_run_init_process(const char *init_filename)
925 ret = run_init_process(init_filename);
927 if (ret && ret != -ENOENT) {
928 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
935 static noinline void __init kernel_init_freeable(void);
937 static int __ref kernel_init(void *unused)
941 kernel_init_freeable();
942 /* need to finish all async __init code before freeing the memory */
943 async_synchronize_full();
946 system_state = SYSTEM_RUNNING;
947 numa_default_policy();
949 flush_delayed_fput();
951 if (ramdisk_execute_command) {
952 ret = run_init_process(ramdisk_execute_command);
955 pr_err("Failed to execute %s (error %d)\n",
956 ramdisk_execute_command, ret);
960 * We try each of these until one succeeds.
962 * The Bourne shell can be used instead of init if we are
963 * trying to recover a really broken machine.
965 if (execute_command) {
966 ret = run_init_process(execute_command);
969 #ifndef CONFIG_INIT_FALLBACK
970 panic("Requested init %s failed (error %d).",
971 execute_command, ret);
973 pr_err("Failed to execute %s (error %d). Attempting defaults...\n",
974 execute_command, ret);
977 if (!try_to_run_init_process("/sbin/init") ||
978 !try_to_run_init_process("/etc/init") ||
979 !try_to_run_init_process("/bin/init") ||
980 !try_to_run_init_process("/bin/sh"))
983 panic("No working init found. Try passing init= option to kernel. "
984 "See Linux Documentation/init.txt for guidance.");
987 static noinline void __init kernel_init_freeable(void)
990 * Wait until kthreadd is all set-up.
992 wait_for_completion(&kthreadd_done);
994 /* Now the scheduler is fully set up and can do blocking allocations */
995 gfp_allowed_mask = __GFP_BITS_MASK;
998 * init can allocate pages on any node
1000 set_mems_allowed(node_states[N_MEMORY]);
1002 * init can run on any cpu.
1004 set_cpus_allowed_ptr(current, cpu_all_mask);
1006 cad_pid = task_pid(current);
1008 smp_prepare_cpus(setup_max_cpus);
1010 do_pre_smp_initcalls();
1011 lockup_detector_init();
1018 /* Open the /dev/console on the rootfs, this should never fail */
1019 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1020 pr_err("Warning: unable to open an initial console.\n");
1025 * check if there is an early userspace init. If yes, let it do all
1029 if (!ramdisk_execute_command)
1030 ramdisk_execute_command = "/init";
1032 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
1033 ramdisk_execute_command = NULL;
1034 prepare_namespace();
1038 * Ok, we have completed the initial bootup, and
1039 * we're essentially up and running. Get rid of the
1040 * initmem segments and start the user-mode stuff..
1042 * rootfs is available now, try loading the public keys
1043 * and default modules
1046 integrity_load_keys();
1047 load_default_modules();