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/ptrace.h>
74 #include <linux/blkdev.h>
75 #include <linux/elevator.h>
76 #include <linux/sched_clock.h>
77 #include <linux/context_tracking.h>
78 #include <linux/random.h>
79 #include <linux/list.h>
80 #include <linux/integrity.h>
81 #include <linux/proc_ns.h>
83 #include <linux/kaiser.h>
84 #include <linux/cache.h>
88 #include <asm/setup.h>
89 #include <asm/sections.h>
90 #include <asm/cacheflush.h>
92 static int kernel_init(void *);
94 extern void init_IRQ(void);
95 extern void fork_init(void);
96 extern void radix_tree_init(void);
99 * Debug helper: via this flag we know that we are in 'early bootup code'
100 * where only the boot processor is running with IRQ disabled. This means
101 * two things - IRQ must not be enabled before the flag is cleared and some
102 * operations which are not allowed with IRQ disabled are allowed while the
105 bool early_boot_irqs_disabled __read_mostly;
107 enum system_states system_state __read_mostly;
108 EXPORT_SYMBOL(system_state);
111 * Boot command-line arguments
113 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
114 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
116 extern void time_init(void);
117 /* Default late time init is NULL. archs can override this later. */
118 void (*__initdata late_time_init)(void);
120 /* Untouched command line saved by arch-specific code. */
121 char __initdata boot_command_line[COMMAND_LINE_SIZE];
122 /* Untouched saved command line (eg. for /proc) */
123 char *saved_command_line;
124 /* Command line for parameter parsing */
125 static char *static_command_line;
126 /* Command line for per-initcall parameter parsing */
127 static char *initcall_command_line;
129 static char *execute_command;
130 static char *ramdisk_execute_command;
133 * Used to generate warnings if static_key manipulation functions are used
134 * before jump_label_init is called.
136 bool static_key_initialized __read_mostly;
137 EXPORT_SYMBOL_GPL(static_key_initialized);
140 * If set, this is an indication to the drivers that reset the underlying
141 * device before going ahead with the initialization otherwise driver might
142 * rely on the BIOS and skip the reset operation.
144 * This is useful if kernel is booting in an unreliable environment.
145 * For ex. kdump situation where previous kernel has crashed, BIOS has been
146 * skipped and devices will be in unknown state.
148 unsigned int reset_devices;
149 EXPORT_SYMBOL(reset_devices);
151 static int __init set_reset_devices(char *str)
157 __setup("reset_devices", set_reset_devices);
159 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
160 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
161 static const char *panic_later, *panic_param;
163 extern const struct obs_kernel_param __setup_start[], __setup_end[];
165 static bool __init obsolete_checksetup(char *line)
167 const struct obs_kernel_param *p;
168 bool had_early_param = false;
172 int n = strlen(p->str);
173 if (parameqn(line, p->str, n)) {
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 = true;
181 } else if (!p->setup_func) {
182 pr_warn("Parameter %s is obsolete, ignored\n",
185 } else if (p->setup_func(line + n))
189 } while (p < __setup_end);
191 return had_early_param;
195 * This should be approx 2 Bo*oMips to start (note initial shift), and will
196 * still work even if initially too large, it will just take slightly longer
198 unsigned long loops_per_jiffy = (1<<12);
199 EXPORT_SYMBOL(loops_per_jiffy);
201 static int __init debug_kernel(char *str)
203 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
207 static int __init quiet_kernel(char *str)
209 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
213 early_param("debug", debug_kernel);
214 early_param("quiet", quiet_kernel);
216 static int __init loglevel(char *str)
221 * Only update loglevel value when a correct setting was passed,
222 * to prevent blind crashes (when loglevel being set to 0) that
223 * are quite hard to debug
225 if (get_option(&str, &newlevel)) {
226 console_loglevel = newlevel;
233 early_param("loglevel", loglevel);
235 /* Change NUL term back to "=", to make "param" the whole string. */
236 static int __init repair_env_string(char *param, char *val,
237 const char *unused, void *arg)
240 /* param=val or param="val"? */
241 if (val == param+strlen(param)+1)
243 else if (val == param+strlen(param)+2) {
245 memmove(val-1, val, strlen(val)+1);
253 /* Anything after -- gets handed straight to init. */
254 static int __init set_init_arg(char *param, char *val,
255 const char *unused, void *arg)
262 repair_env_string(param, val, unused, NULL);
264 for (i = 0; argv_init[i]; i++) {
265 if (i == MAX_INIT_ARGS) {
266 panic_later = "init";
271 argv_init[i] = param;
276 * Unknown boot options get handed to init, unless they look like
277 * unused parameters (modprobe will find them in /proc/cmdline).
279 static int __init unknown_bootoption(char *param, char *val,
280 const char *unused, void *arg)
282 repair_env_string(param, val, unused, NULL);
284 /* Handle obsolete-style parameters */
285 if (obsolete_checksetup(param))
288 /* Unused module parameter. */
289 if (strchr(param, '.') && (!val || strchr(param, '.') < val))
296 /* Environment option */
298 for (i = 0; envp_init[i]; i++) {
299 if (i == MAX_INIT_ENVS) {
303 if (!strncmp(param, envp_init[i], val - param))
306 envp_init[i] = param;
308 /* Command line option */
310 for (i = 0; argv_init[i]; i++) {
311 if (i == MAX_INIT_ARGS) {
312 panic_later = "init";
316 argv_init[i] = param;
321 static int __init init_setup(char *str)
325 execute_command = str;
327 * In case LILO is going to boot us with default command line,
328 * it prepends "auto" before the whole cmdline which makes
329 * the shell think it should execute a script with such name.
330 * So we ignore all arguments entered _before_ init=... [MJ]
332 for (i = 1; i < MAX_INIT_ARGS; i++)
336 __setup("init=", init_setup);
338 static int __init rdinit_setup(char *str)
342 ramdisk_execute_command = str;
343 /* See "auto" comment in init_setup */
344 for (i = 1; i < MAX_INIT_ARGS; i++)
348 __setup("rdinit=", rdinit_setup);
351 static const unsigned int setup_max_cpus = NR_CPUS;
352 static inline void setup_nr_cpu_ids(void) { }
353 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
357 * We need to store the untouched command line for future reference.
358 * We also need to store the touched command line since the parameter
359 * parsing is performed in place, and we should allow a component to
360 * store reference of name/value for future reference.
362 static void __init setup_command_line(char *command_line)
365 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
366 initcall_command_line =
367 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
368 static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0);
369 strcpy(saved_command_line, boot_command_line);
370 strcpy(static_command_line, command_line);
374 * We need to finalize in a non-__init function or else race conditions
375 * between the root thread and the init thread may cause start_kernel to
376 * be reaped by free_initmem before the root thread has proceeded to
379 * gcc-3.4 accidentally inlines this function, so use noinline.
382 static __initdata DECLARE_COMPLETION(kthreadd_done);
384 static noinline void __ref rest_init(void)
388 rcu_scheduler_starting();
390 * We need to spawn init first so that it obtains pid 1, however
391 * the init task will end up wanting to create kthreads, which, if
392 * we schedule it before we create kthreadd, will OOPS.
394 kernel_thread(kernel_init, NULL, CLONE_FS);
395 numa_default_policy();
396 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
398 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
400 complete(&kthreadd_done);
403 * The boot idle thread must execute schedule()
404 * at least once to get things moving:
406 init_idle_bootup_task(current);
407 schedule_preempt_disabled();
408 /* Call into cpu_idle with preempt disabled */
409 cpu_startup_entry(CPUHP_ONLINE);
412 /* Check for early params. */
413 static int __init do_early_param(char *param, char *val,
414 const char *unused, void *arg)
416 const struct obs_kernel_param *p;
418 for (p = __setup_start; p < __setup_end; p++) {
419 if ((p->early && parameq(param, p->str)) ||
420 (strcmp(param, "console") == 0 &&
421 strcmp(p->str, "earlycon") == 0)
423 if (p->setup_func(val) != 0)
424 pr_warn("Malformed early option '%s'\n", param);
427 /* We accept everything at this stage. */
431 void __init parse_early_options(char *cmdline)
433 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
437 /* Arch code calls this early on, or if not, just before other parsing. */
438 void __init parse_early_param(void)
440 static int done __initdata;
441 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
446 /* All fall through to do_early_param. */
447 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
448 parse_early_options(tmp_cmdline);
452 void __init __weak smp_setup_processor_id(void)
456 # if THREAD_SIZE >= PAGE_SIZE
457 void __init __weak thread_stack_cache_init(void)
463 * Set up kernel memory allocators
465 static void __init mm_init(void)
468 * page_ext requires contiguous pages,
469 * bigger than MAX_ORDER unless SPARSEMEM.
471 page_ext_init_flatmem();
481 asmlinkage __visible void __init start_kernel(void)
486 set_task_stack_end_magic(&init_task);
487 smp_setup_processor_id();
488 debug_objects_early_init();
491 * Set up the the initial canary ASAP:
493 boot_init_stack_canary();
498 early_boot_irqs_disabled = true;
501 * Interrupts are still disabled. Do necessary setups, then
506 pr_notice("%s", linux_banner);
507 setup_arch(&command_line);
508 mm_init_cpumask(&init_mm);
509 setup_command_line(command_line);
511 setup_per_cpu_areas();
512 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
513 boot_cpu_hotplug_init();
515 build_all_zonelists(NULL, NULL);
518 pr_notice("Kernel command line: %s\n", boot_command_line);
519 /* parameters may set static keys */
522 after_dashes = parse_args("Booting kernel",
523 static_command_line, __start___param,
524 __stop___param - __start___param,
525 -1, -1, NULL, &unknown_bootoption);
526 if (!IS_ERR_OR_NULL(after_dashes))
527 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
531 * These use large bootmem allocations and must precede
536 vfs_caches_init_early();
542 * Set up the scheduler prior starting any interrupts (such as the
543 * timer interrupt). Full topology setup happens at smp_init()
544 * time - but meanwhile we still have a functioning scheduler.
548 * Disable preemption - early bootup scheduling is extremely
549 * fragile until we cpu_idle() for the first time.
552 if (WARN(!irqs_disabled(),
553 "Interrupts were enabled *very* early, fixing it\n"))
558 /* trace_printk() and trace points may be used after this */
561 context_tracking_init();
563 /* init some links before init_ISA_irqs() */
573 sched_clock_postinit();
577 call_function_init();
578 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
579 early_boot_irqs_disabled = false;
582 kmem_cache_init_late();
585 * HACK ALERT! This is early. We're enabling the console before
586 * we've done PCI setups etc, and console_init() must be aware of
587 * this. But we do want output early, in case something goes wrong.
591 panic("Too many boot %s vars at `%s'", panic_later,
597 * Need to run this when irqs are enabled, because it wants
598 * to self-test [hard/soft]-irqs on/off lock inversion bugs
603 #ifdef CONFIG_BLK_DEV_INITRD
604 if (initrd_start && !initrd_below_start_ok &&
605 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
606 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
607 page_to_pfn(virt_to_page((void *)initrd_start)),
613 debug_objects_mem_init();
615 setup_per_cpu_pageset();
625 if (efi_enabled(EFI_RUNTIME_SERVICES))
626 efi_enter_virtual_mode();
628 #ifdef CONFIG_X86_ESPFIX64
629 /* Should be run before the first non-init thread is created */
632 thread_stack_cache_init();
642 /* rootfs populating might need page-writeback */
643 page_writeback_init();
648 taskstats_init_early();
653 acpi_subsystem_init();
656 if (efi_enabled(EFI_RUNTIME_SERVICES)) {
658 efi_free_boot_services();
663 /* Do the rest non-__init'ed, we're now alive */
667 /* Call all constructor functions linked into the kernel. */
668 static void __init do_ctors(void)
670 #ifdef CONFIG_CONSTRUCTORS
671 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
673 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
679 core_param(initcall_debug, initcall_debug, bool, 0644);
681 #ifdef CONFIG_KALLSYMS
682 struct blacklist_entry {
683 struct list_head next;
687 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
689 static int __init initcall_blacklist(char *str)
692 struct blacklist_entry *entry;
694 /* str argument is a comma-separated list of functions */
696 str_entry = strsep(&str, ",");
698 pr_debug("blacklisting initcall %s\n", str_entry);
699 entry = alloc_bootmem(sizeof(*entry));
700 entry->buf = alloc_bootmem(strlen(str_entry) + 1);
701 strcpy(entry->buf, str_entry);
702 list_add(&entry->next, &blacklisted_initcalls);
709 static bool __init_or_module initcall_blacklisted(initcall_t fn)
711 struct blacklist_entry *entry;
712 char fn_name[KSYM_SYMBOL_LEN];
715 if (list_empty(&blacklisted_initcalls))
718 addr = (unsigned long) dereference_function_descriptor(fn);
719 sprint_symbol_no_offset(fn_name, addr);
722 * fn will be "function_name [module_name]" where [module_name] is not
723 * displayed for built-in init functions. Strip off the [module_name].
725 strreplace(fn_name, ' ', '\0');
727 list_for_each_entry(entry, &blacklisted_initcalls, next) {
728 if (!strcmp(fn_name, entry->buf)) {
729 pr_debug("initcall %s blacklisted\n", fn_name);
737 static int __init initcall_blacklist(char *str)
739 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
743 static bool __init_or_module initcall_blacklisted(initcall_t fn)
748 __setup("initcall_blacklist=", initcall_blacklist);
750 static int __init_or_module do_one_initcall_debug(initcall_t fn)
752 ktime_t calltime, delta, rettime;
753 unsigned long long duration;
756 printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current));
757 calltime = ktime_get();
759 rettime = ktime_get();
760 delta = ktime_sub(rettime, calltime);
761 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
762 printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n",
768 int __init_or_module do_one_initcall(initcall_t fn)
770 int count = preempt_count();
774 if (initcall_blacklisted(fn))
778 ret = do_one_initcall_debug(fn);
784 if (preempt_count() != count) {
785 sprintf(msgbuf, "preemption imbalance ");
786 preempt_count_set(count);
788 if (irqs_disabled()) {
789 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
792 WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
794 add_latent_entropy();
799 extern initcall_t __initcall_start[];
800 extern initcall_t __initcall0_start[];
801 extern initcall_t __initcall1_start[];
802 extern initcall_t __initcall2_start[];
803 extern initcall_t __initcall3_start[];
804 extern initcall_t __initcall4_start[];
805 extern initcall_t __initcall5_start[];
806 extern initcall_t __initcall6_start[];
807 extern initcall_t __initcall7_start[];
808 extern initcall_t __initcall_end[];
810 static initcall_t *initcall_levels[] __initdata = {
822 /* Keep these in sync with initcalls in include/linux/init.h */
823 static char *initcall_level_names[] __initdata = {
834 static void __init do_initcall_level(int level)
838 strcpy(initcall_command_line, saved_command_line);
839 parse_args(initcall_level_names[level],
840 initcall_command_line, __start___param,
841 __stop___param - __start___param,
843 NULL, &repair_env_string);
845 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
846 do_one_initcall(*fn);
849 static void __init do_initcalls(void)
853 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
854 do_initcall_level(level);
858 * Ok, the machine is now initialized. None of the devices
859 * have been touched yet, but the CPU subsystem is up and
860 * running, and memory and process management works.
862 * Now we can finally start doing some real work..
864 static void __init do_basic_setup(void)
871 usermodehelper_enable();
875 static void __init do_pre_smp_initcalls(void)
879 for (fn = __initcall_start; fn < __initcall0_start; fn++)
880 do_one_initcall(*fn);
884 * This function requests modules which should be loaded by default and is
885 * called twice right after initrd is mounted and right before init is
886 * exec'd. If such modules are on either initrd or rootfs, they will be
887 * loaded before control is passed to userland.
889 void __init load_default_modules(void)
891 load_default_elevator_module();
894 static int run_init_process(const char *init_filename)
896 argv_init[0] = init_filename;
897 return do_execve(getname_kernel(init_filename),
898 (const char __user *const __user *)argv_init,
899 (const char __user *const __user *)envp_init);
902 static int try_to_run_init_process(const char *init_filename)
906 ret = run_init_process(init_filename);
908 if (ret && ret != -ENOENT) {
909 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
916 static noinline void __init kernel_init_freeable(void);
918 #if defined(CONFIG_DEBUG_RODATA) || defined(CONFIG_DEBUG_SET_MODULE_RONX)
919 bool rodata_enabled __ro_after_init = true;
920 static int __init set_debug_rodata(char *str)
922 return strtobool(str, &rodata_enabled);
924 __setup("rodata=", set_debug_rodata);
927 #ifdef CONFIG_DEBUG_RODATA
928 static void mark_readonly(void)
933 pr_info("Kernel memory protection disabled.\n");
936 static inline void mark_readonly(void)
938 pr_warn("This architecture does not have kernel memory protection.\n");
942 static int __ref kernel_init(void *unused)
946 kernel_init_freeable();
947 /* need to finish all async __init code before freeing the memory */
948 async_synchronize_full();
951 system_state = SYSTEM_RUNNING;
952 numa_default_policy();
954 rcu_end_inkernel_boot();
956 if (ramdisk_execute_command) {
957 ret = run_init_process(ramdisk_execute_command);
960 pr_err("Failed to execute %s (error %d)\n",
961 ramdisk_execute_command, ret);
965 * We try each of these until one succeeds.
967 * The Bourne shell can be used instead of init if we are
968 * trying to recover a really broken machine.
970 if (execute_command) {
971 ret = run_init_process(execute_command);
974 panic("Requested init %s failed (error %d).",
975 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();
1016 page_alloc_init_late();
1020 /* Open the /dev/console on the rootfs, this should never fail */
1021 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1022 pr_err("Warning: unable to open an initial console.\n");
1027 * check if there is an early userspace init. If yes, let it do all
1031 if (!ramdisk_execute_command)
1032 ramdisk_execute_command = "/init";
1034 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
1035 ramdisk_execute_command = NULL;
1036 prepare_namespace();
1040 * Ok, we have completed the initial bootup, and
1041 * we're essentially up and running. Get rid of the
1042 * initmem segments and start the user-mode stuff..
1044 * rootfs is available now, try loading the public keys
1045 * and default modules
1048 integrity_load_keys();
1049 load_default_modules();