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>
87 #include <asm/setup.h>
88 #include <asm/sections.h>
89 #include <asm/cacheflush.h>
91 static int kernel_init(void *);
93 extern void init_IRQ(void);
94 extern void fork_init(void);
95 extern void radix_tree_init(void);
98 * Debug helper: via this flag we know that we are in 'early bootup code'
99 * where only the boot processor is running with IRQ disabled. This means
100 * two things - IRQ must not be enabled before the flag is cleared and some
101 * operations which are not allowed with IRQ disabled are allowed while the
104 bool early_boot_irqs_disabled __read_mostly;
106 enum system_states system_state __read_mostly;
107 EXPORT_SYMBOL(system_state);
110 * Boot command-line arguments
112 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
113 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
115 extern void time_init(void);
116 /* Default late time init is NULL. archs can override this later. */
117 void (*__initdata late_time_init)(void);
119 /* Untouched command line saved by arch-specific code. */
120 char __initdata boot_command_line[COMMAND_LINE_SIZE];
121 /* Untouched saved command line (eg. for /proc) */
122 char *saved_command_line;
123 /* Command line for parameter parsing */
124 static char *static_command_line;
125 /* Command line for per-initcall parameter parsing */
126 static char *initcall_command_line;
128 static char *execute_command;
129 static char *ramdisk_execute_command;
132 * Used to generate warnings if static_key manipulation functions are used
133 * before jump_label_init is called.
135 bool static_key_initialized __read_mostly;
136 EXPORT_SYMBOL_GPL(static_key_initialized);
139 * If set, this is an indication to the drivers that reset the underlying
140 * device before going ahead with the initialization otherwise driver might
141 * rely on the BIOS and skip the reset operation.
143 * This is useful if kernel is booting in an unreliable environment.
144 * For ex. kdump situation where previous kernel has crashed, BIOS has been
145 * skipped and devices will be in unknown state.
147 unsigned int reset_devices;
148 EXPORT_SYMBOL(reset_devices);
150 static int __init set_reset_devices(char *str)
156 __setup("reset_devices", set_reset_devices);
158 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
159 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
160 static const char *panic_later, *panic_param;
162 extern const struct obs_kernel_param __setup_start[], __setup_end[];
164 static bool __init obsolete_checksetup(char *line)
166 const struct obs_kernel_param *p;
167 bool had_early_param = false;
171 int n = strlen(p->str);
172 if (parameqn(line, p->str, n)) {
174 /* Already done in parse_early_param?
175 * (Needs exact match on param part).
176 * Keep iterating, as we can have early
177 * params and __setups of same names 8( */
178 if (line[n] == '\0' || line[n] == '=')
179 had_early_param = true;
180 } else if (!p->setup_func) {
181 pr_warn("Parameter %s is obsolete, ignored\n",
184 } else if (p->setup_func(line + n))
188 } while (p < __setup_end);
190 return had_early_param;
194 * This should be approx 2 Bo*oMips to start (note initial shift), and will
195 * still work even if initially too large, it will just take slightly longer
197 unsigned long loops_per_jiffy = (1<<12);
198 EXPORT_SYMBOL(loops_per_jiffy);
200 static int __init debug_kernel(char *str)
202 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
206 static int __init quiet_kernel(char *str)
208 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
212 early_param("debug", debug_kernel);
213 early_param("quiet", quiet_kernel);
215 static int __init loglevel(char *str)
220 * Only update loglevel value when a correct setting was passed,
221 * to prevent blind crashes (when loglevel being set to 0) that
222 * are quite hard to debug
224 if (get_option(&str, &newlevel)) {
225 console_loglevel = newlevel;
232 early_param("loglevel", loglevel);
234 /* Change NUL term back to "=", to make "param" the whole string. */
235 static int __init repair_env_string(char *param, char *val,
236 const char *unused, void *arg)
239 /* param=val or param="val"? */
240 if (val == param+strlen(param)+1)
242 else if (val == param+strlen(param)+2) {
244 memmove(val-1, val, strlen(val)+1);
252 /* Anything after -- gets handed straight to init. */
253 static int __init set_init_arg(char *param, char *val,
254 const char *unused, void *arg)
261 repair_env_string(param, val, unused, NULL);
263 for (i = 0; argv_init[i]; i++) {
264 if (i == MAX_INIT_ARGS) {
265 panic_later = "init";
270 argv_init[i] = param;
275 * Unknown boot options get handed to init, unless they look like
276 * unused parameters (modprobe will find them in /proc/cmdline).
278 static int __init unknown_bootoption(char *param, char *val,
279 const char *unused, void *arg)
281 repair_env_string(param, val, unused, NULL);
283 /* Handle obsolete-style parameters */
284 if (obsolete_checksetup(param))
287 /* Unused module parameter. */
288 if (strchr(param, '.') && (!val || strchr(param, '.') < val))
295 /* Environment option */
297 for (i = 0; envp_init[i]; i++) {
298 if (i == MAX_INIT_ENVS) {
302 if (!strncmp(param, envp_init[i], val - param))
305 envp_init[i] = param;
307 /* Command line option */
309 for (i = 0; argv_init[i]; i++) {
310 if (i == MAX_INIT_ARGS) {
311 panic_later = "init";
315 argv_init[i] = param;
320 static int __init init_setup(char *str)
324 execute_command = str;
326 * In case LILO is going to boot us with default command line,
327 * it prepends "auto" before the whole cmdline which makes
328 * the shell think it should execute a script with such name.
329 * So we ignore all arguments entered _before_ init=... [MJ]
331 for (i = 1; i < MAX_INIT_ARGS; i++)
335 __setup("init=", init_setup);
337 static int __init rdinit_setup(char *str)
341 ramdisk_execute_command = str;
342 /* See "auto" comment in init_setup */
343 for (i = 1; i < MAX_INIT_ARGS; i++)
347 __setup("rdinit=", rdinit_setup);
350 static const unsigned int setup_max_cpus = NR_CPUS;
351 static inline void setup_nr_cpu_ids(void) { }
352 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
356 * We need to store the untouched command line for future reference.
357 * We also need to store the touched command line since the parameter
358 * parsing is performed in place, and we should allow a component to
359 * store reference of name/value for future reference.
361 static void __init setup_command_line(char *command_line)
364 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
365 initcall_command_line =
366 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
367 static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0);
368 strcpy(saved_command_line, boot_command_line);
369 strcpy(static_command_line, command_line);
373 * We need to finalize in a non-__init function or else race conditions
374 * between the root thread and the init thread may cause start_kernel to
375 * be reaped by free_initmem before the root thread has proceeded to
378 * gcc-3.4 accidentally inlines this function, so use noinline.
381 static __initdata DECLARE_COMPLETION(kthreadd_done);
383 static noinline void __init_refok rest_init(void)
387 rcu_scheduler_starting();
389 * We need to spawn init first so that it obtains pid 1, however
390 * the init task will end up wanting to create kthreads, which, if
391 * we schedule it before we create kthreadd, will OOPS.
393 kernel_thread(kernel_init, NULL, CLONE_FS);
394 numa_default_policy();
395 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
397 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
399 complete(&kthreadd_done);
402 * The boot idle thread must execute schedule()
403 * at least once to get things moving:
405 init_idle_bootup_task(current);
406 schedule_preempt_disabled();
407 /* Call into cpu_idle with preempt disabled */
408 cpu_startup_entry(CPUHP_ONLINE);
411 /* Check for early params. */
412 static int __init do_early_param(char *param, char *val,
413 const char *unused, void *arg)
415 const struct obs_kernel_param *p;
417 for (p = __setup_start; p < __setup_end; p++) {
418 if ((p->early && parameq(param, p->str)) ||
419 (strcmp(param, "console") == 0 &&
420 strcmp(p->str, "earlycon") == 0)
422 if (p->setup_func(val) != 0)
423 pr_warn("Malformed early option '%s'\n", param);
426 /* We accept everything at this stage. */
430 void __init parse_early_options(char *cmdline)
432 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
436 /* Arch code calls this early on, or if not, just before other parsing. */
437 void __init parse_early_param(void)
439 static int done __initdata;
440 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
445 /* All fall through to do_early_param. */
446 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
447 parse_early_options(tmp_cmdline);
451 void __init __weak smp_setup_processor_id(void)
455 # if THREAD_SIZE >= PAGE_SIZE
456 void __init __weak thread_stack_cache_init(void)
462 * Set up kernel memory allocators
464 static void __init mm_init(void)
467 * page_ext requires contiguous pages,
468 * bigger than MAX_ORDER unless SPARSEMEM.
470 page_ext_init_flatmem();
479 asmlinkage __visible void __init start_kernel(void)
484 set_task_stack_end_magic(&init_task);
485 smp_setup_processor_id();
486 debug_objects_early_init();
489 * Set up the the initial canary ASAP:
491 boot_init_stack_canary();
496 early_boot_irqs_disabled = true;
499 * Interrupts are still disabled. Do necessary setups, then
504 pr_notice("%s", linux_banner);
505 setup_arch(&command_line);
506 mm_init_cpumask(&init_mm);
507 setup_command_line(command_line);
509 setup_per_cpu_areas();
510 boot_cpu_state_init();
511 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
513 build_all_zonelists(NULL, NULL);
516 pr_notice("Kernel command line: %s\n", boot_command_line);
518 after_dashes = parse_args("Booting kernel",
519 static_command_line, __start___param,
520 __stop___param - __start___param,
521 -1, -1, NULL, &unknown_bootoption);
522 if (!IS_ERR_OR_NULL(after_dashes))
523 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
529 * These use large bootmem allocations and must precede
534 vfs_caches_init_early();
540 * Set up the scheduler prior starting any interrupts (such as the
541 * timer interrupt). Full topology setup happens at smp_init()
542 * time - but meanwhile we still have a functioning scheduler.
546 * Disable preemption - early bootup scheduling is extremely
547 * fragile until we cpu_idle() for the first time.
550 if (WARN(!irqs_disabled(),
551 "Interrupts were enabled *very* early, fixing it\n"))
556 /* trace_printk() and trace points may be used after this */
559 context_tracking_init();
561 /* init some links before init_ISA_irqs() */
571 sched_clock_postinit();
575 call_function_init();
576 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
577 early_boot_irqs_disabled = false;
580 kmem_cache_init_late();
583 * HACK ALERT! This is early. We're enabling the console before
584 * we've done PCI setups etc, and console_init() must be aware of
585 * this. But we do want output early, in case something goes wrong.
589 panic("Too many boot %s vars at `%s'", panic_later,
595 * Need to run this when irqs are enabled, because it wants
596 * to self-test [hard/soft]-irqs on/off lock inversion bugs
601 #ifdef CONFIG_BLK_DEV_INITRD
602 if (initrd_start && !initrd_below_start_ok &&
603 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
604 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
605 page_to_pfn(virt_to_page((void *)initrd_start)),
611 debug_objects_mem_init();
613 setup_per_cpu_pageset();
623 if (efi_enabled(EFI_RUNTIME_SERVICES))
624 efi_enter_virtual_mode();
626 #ifdef CONFIG_X86_ESPFIX64
627 /* Should be run before the first non-init thread is created */
630 thread_stack_cache_init();
640 /* rootfs populating might need page-writeback */
641 page_writeback_init();
646 taskstats_init_early();
651 acpi_subsystem_init();
654 if (efi_enabled(EFI_RUNTIME_SERVICES)) {
656 efi_free_boot_services();
661 /* Do the rest non-__init'ed, we're now alive */
665 /* Call all constructor functions linked into the kernel. */
666 static void __init do_ctors(void)
668 #ifdef CONFIG_CONSTRUCTORS
669 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
671 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
677 core_param(initcall_debug, initcall_debug, bool, 0644);
679 #ifdef CONFIG_KALLSYMS
680 struct blacklist_entry {
681 struct list_head next;
685 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
687 static int __init initcall_blacklist(char *str)
690 struct blacklist_entry *entry;
692 /* str argument is a comma-separated list of functions */
694 str_entry = strsep(&str, ",");
696 pr_debug("blacklisting initcall %s\n", str_entry);
697 entry = alloc_bootmem(sizeof(*entry));
698 entry->buf = alloc_bootmem(strlen(str_entry) + 1);
699 strcpy(entry->buf, str_entry);
700 list_add(&entry->next, &blacklisted_initcalls);
707 static bool __init_or_module initcall_blacklisted(initcall_t fn)
709 struct blacklist_entry *entry;
710 char fn_name[KSYM_SYMBOL_LEN];
713 if (list_empty(&blacklisted_initcalls))
716 addr = (unsigned long) dereference_function_descriptor(fn);
717 sprint_symbol_no_offset(fn_name, addr);
719 list_for_each_entry(entry, &blacklisted_initcalls, next) {
720 if (!strcmp(fn_name, entry->buf)) {
721 pr_debug("initcall %s blacklisted\n", fn_name);
729 static int __init initcall_blacklist(char *str)
731 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
735 static bool __init_or_module initcall_blacklisted(initcall_t fn)
740 __setup("initcall_blacklist=", initcall_blacklist);
742 static int __init_or_module do_one_initcall_debug(initcall_t fn)
744 ktime_t calltime, delta, rettime;
745 unsigned long long duration;
748 printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current));
749 calltime = ktime_get();
751 rettime = ktime_get();
752 delta = ktime_sub(rettime, calltime);
753 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
754 printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n",
760 int __init_or_module do_one_initcall(initcall_t fn)
762 int count = preempt_count();
766 if (initcall_blacklisted(fn))
770 ret = do_one_initcall_debug(fn);
776 if (preempt_count() != count) {
777 sprintf(msgbuf, "preemption imbalance ");
778 preempt_count_set(count);
780 if (irqs_disabled()) {
781 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
784 WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
790 extern initcall_t __initcall_start[];
791 extern initcall_t __initcall0_start[];
792 extern initcall_t __initcall1_start[];
793 extern initcall_t __initcall2_start[];
794 extern initcall_t __initcall3_start[];
795 extern initcall_t __initcall4_start[];
796 extern initcall_t __initcall5_start[];
797 extern initcall_t __initcall6_start[];
798 extern initcall_t __initcall7_start[];
799 extern initcall_t __initcall_end[];
801 static initcall_t *initcall_levels[] __initdata = {
813 /* Keep these in sync with initcalls in include/linux/init.h */
814 static char *initcall_level_names[] __initdata = {
825 static void __init do_initcall_level(int level)
829 strcpy(initcall_command_line, saved_command_line);
830 parse_args(initcall_level_names[level],
831 initcall_command_line, __start___param,
832 __stop___param - __start___param,
834 NULL, &repair_env_string);
836 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
837 do_one_initcall(*fn);
840 static void __init do_initcalls(void)
844 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
845 do_initcall_level(level);
849 * Ok, the machine is now initialized. None of the devices
850 * have been touched yet, but the CPU subsystem is up and
851 * running, and memory and process management works.
853 * Now we can finally start doing some real work..
855 static void __init do_basic_setup(void)
862 usermodehelper_enable();
864 random_int_secret_init();
867 static void __init do_pre_smp_initcalls(void)
871 for (fn = __initcall_start; fn < __initcall0_start; fn++)
872 do_one_initcall(*fn);
876 * This function requests modules which should be loaded by default and is
877 * called twice right after initrd is mounted and right before init is
878 * exec'd. If such modules are on either initrd or rootfs, they will be
879 * loaded before control is passed to userland.
881 void __init load_default_modules(void)
883 load_default_elevator_module();
886 static int run_init_process(const char *init_filename)
888 argv_init[0] = init_filename;
889 return do_execve(getname_kernel(init_filename),
890 (const char __user *const __user *)argv_init,
891 (const char __user *const __user *)envp_init);
894 static int try_to_run_init_process(const char *init_filename)
898 ret = run_init_process(init_filename);
900 if (ret && ret != -ENOENT) {
901 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
908 static noinline void __init kernel_init_freeable(void);
910 #ifdef CONFIG_DEBUG_RODATA
911 static bool rodata_enabled = true;
912 static int __init set_debug_rodata(char *str)
914 return strtobool(str, &rodata_enabled);
916 __setup("rodata=", set_debug_rodata);
918 static void mark_readonly(void)
923 pr_info("Kernel memory protection disabled.\n");
926 static inline void mark_readonly(void)
928 pr_warn("This architecture does not have kernel memory protection.\n");
932 static int __ref kernel_init(void *unused)
936 kernel_init_freeable();
937 /* need to finish all async __init code before freeing the memory */
938 async_synchronize_full();
941 system_state = SYSTEM_RUNNING;
942 numa_default_policy();
944 flush_delayed_fput();
946 rcu_end_inkernel_boot();
948 if (ramdisk_execute_command) {
949 ret = run_init_process(ramdisk_execute_command);
952 pr_err("Failed to execute %s (error %d)\n",
953 ramdisk_execute_command, ret);
957 * We try each of these until one succeeds.
959 * The Bourne shell can be used instead of init if we are
960 * trying to recover a really broken machine.
962 if (execute_command) {
963 ret = run_init_process(execute_command);
966 panic("Requested init %s failed (error %d).",
967 execute_command, ret);
969 if (!try_to_run_init_process("/sbin/init") ||
970 !try_to_run_init_process("/etc/init") ||
971 !try_to_run_init_process("/bin/init") ||
972 !try_to_run_init_process("/bin/sh"))
975 panic("No working init found. Try passing init= option to kernel. "
976 "See Linux Documentation/init.txt for guidance.");
979 static noinline void __init kernel_init_freeable(void)
982 * Wait until kthreadd is all set-up.
984 wait_for_completion(&kthreadd_done);
986 /* Now the scheduler is fully set up and can do blocking allocations */
987 gfp_allowed_mask = __GFP_BITS_MASK;
990 * init can allocate pages on any node
992 set_mems_allowed(node_states[N_MEMORY]);
994 * init can run on any cpu.
996 set_cpus_allowed_ptr(current, cpu_all_mask);
998 cad_pid = task_pid(current);
1000 smp_prepare_cpus(setup_max_cpus);
1002 do_pre_smp_initcalls();
1003 lockup_detector_init();
1008 page_alloc_init_late();
1012 /* Open the /dev/console on the rootfs, this should never fail */
1013 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1014 pr_err("Warning: unable to open an initial console.\n");
1019 * check if there is an early userspace init. If yes, let it do all
1023 if (!ramdisk_execute_command)
1024 ramdisk_execute_command = "/init";
1026 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
1027 ramdisk_execute_command = NULL;
1028 prepare_namespace();
1032 * Ok, we have completed the initial bootup, and
1033 * we're essentially up and running. Get rid of the
1034 * initmem segments and start the user-mode stuff..
1036 * rootfs is available now, try loading the public keys
1037 * and default modules
1040 integrity_load_keys();
1041 load_default_modules();