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_cgroup.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>
80 #include <asm/setup.h>
81 #include <asm/sections.h>
82 #include <asm/cacheflush.h>
84 #ifdef CONFIG_X86_LOCAL_APIC
88 static int kernel_init(void *);
90 extern void init_IRQ(void);
91 extern void fork_init(unsigned long);
92 extern void mca_init(void);
93 extern void sbus_init(void);
94 extern void radix_tree_init(void);
95 #ifndef CONFIG_DEBUG_RODATA
96 static inline void mark_rodata_ro(void) { }
100 extern void tc_init(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);
124 extern void softirq_init(void);
126 /* Untouched command line saved by arch-specific code. */
127 char __initdata boot_command_line[COMMAND_LINE_SIZE];
128 /* Untouched saved command line (eg. for /proc) */
129 char *saved_command_line;
130 /* Command line for parameter parsing */
131 static char *static_command_line;
133 static char *execute_command;
134 static char *ramdisk_execute_command;
137 * If set, this is an indication to the drivers that reset the underlying
138 * device before going ahead with the initialization otherwise driver might
139 * rely on the BIOS and skip the reset operation.
141 * This is useful if kernel is booting in an unreliable environment.
142 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
143 * skipped and devices will be in unknown state.
145 unsigned int reset_devices;
146 EXPORT_SYMBOL(reset_devices);
148 static int __init set_reset_devices(char *str)
154 __setup("reset_devices", set_reset_devices);
156 static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
157 const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
158 static const char *panic_later, *panic_param;
160 extern const struct obs_kernel_param __setup_start[], __setup_end[];
162 static int __init obsolete_checksetup(char *line)
164 const struct obs_kernel_param *p;
165 int had_early_param = 0;
169 int n = strlen(p->str);
170 if (parameqn(line, p->str, n)) {
172 /* Already done in parse_early_param?
173 * (Needs exact match on param part).
174 * Keep iterating, as we can have early
175 * params and __setups of same names 8( */
176 if (line[n] == '\0' || line[n] == '=')
178 } else if (!p->setup_func) {
179 pr_warn("Parameter %s is obsolete, ignored\n",
182 } else if (p->setup_func(line + n))
186 } while (p < __setup_end);
188 return had_early_param;
192 * This should be approx 2 Bo*oMips to start (note initial shift), and will
193 * still work even if initially too large, it will just take slightly longer
195 unsigned long loops_per_jiffy = (1<<12);
197 EXPORT_SYMBOL(loops_per_jiffy);
199 static int __init debug_kernel(char *str)
201 console_loglevel = 10;
205 static int __init quiet_kernel(char *str)
207 console_loglevel = 4;
211 early_param("debug", debug_kernel);
212 early_param("quiet", quiet_kernel);
214 static int __init loglevel(char *str)
219 * Only update loglevel value when a correct setting was passed,
220 * to prevent blind crashes (when loglevel being set to 0) that
221 * are quite hard to debug
223 if (get_option(&str, &newlevel)) {
224 console_loglevel = newlevel;
231 early_param("loglevel", loglevel);
233 /* Change NUL term back to "=", to make "param" the whole string. */
234 static int __init repair_env_string(char *param, char *val, const char *unused)
237 /* param=val or param="val"? */
238 if (val == param+strlen(param)+1)
240 else if (val == param+strlen(param)+2) {
242 memmove(val-1, val, strlen(val)+1);
251 * Unknown boot options get handed to init, unless they look like
252 * unused parameters (modprobe will find them in /proc/cmdline).
254 static int __init unknown_bootoption(char *param, char *val, const char *unused)
256 repair_env_string(param, val, unused);
258 /* Handle obsolete-style parameters */
259 if (obsolete_checksetup(param))
262 /* Unused module parameter. */
263 if (strchr(param, '.') && (!val || strchr(param, '.') < val))
270 /* Environment option */
272 for (i = 0; envp_init[i]; i++) {
273 if (i == MAX_INIT_ENVS) {
274 panic_later = "Too many boot env vars at `%s'";
277 if (!strncmp(param, envp_init[i], val - param))
280 envp_init[i] = param;
282 /* Command line option */
284 for (i = 0; argv_init[i]; i++) {
285 if (i == MAX_INIT_ARGS) {
286 panic_later = "Too many boot init vars at `%s'";
290 argv_init[i] = param;
295 static int __init init_setup(char *str)
299 execute_command = str;
301 * In case LILO is going to boot us with default command line,
302 * it prepends "auto" before the whole cmdline which makes
303 * the shell think it should execute a script with such name.
304 * So we ignore all arguments entered _before_ init=... [MJ]
306 for (i = 1; i < MAX_INIT_ARGS; i++)
310 __setup("init=", init_setup);
312 static int __init rdinit_setup(char *str)
316 ramdisk_execute_command = str;
317 /* See "auto" comment in init_setup */
318 for (i = 1; i < MAX_INIT_ARGS; i++)
322 __setup("rdinit=", rdinit_setup);
325 static const unsigned int setup_max_cpus = NR_CPUS;
326 #ifdef CONFIG_X86_LOCAL_APIC
327 static void __init smp_init(void)
329 APIC_init_uniprocessor();
332 #define smp_init() do { } while (0)
335 static inline void setup_nr_cpu_ids(void) { }
336 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
340 * We need to store the untouched command line for future reference.
341 * We also need to store the touched command line since the parameter
342 * parsing is performed in place, and we should allow a component to
343 * store reference of name/value for future reference.
345 static void __init setup_command_line(char *command_line)
347 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
348 static_command_line = alloc_bootmem(strlen (command_line)+1);
349 strcpy (saved_command_line, boot_command_line);
350 strcpy (static_command_line, command_line);
354 * We need to finalize in a non-__init function or else race conditions
355 * between the root thread and the init thread may cause start_kernel to
356 * be reaped by free_initmem before the root thread has proceeded to
359 * gcc-3.4 accidentally inlines this function, so use noinline.
362 static __initdata DECLARE_COMPLETION(kthreadd_done);
364 static noinline void __init_refok rest_init(void)
368 rcu_scheduler_starting();
370 * We need to spawn init first so that it obtains pid 1, however
371 * the init task will end up wanting to create kthreads, which, if
372 * we schedule it before we create kthreadd, will OOPS.
374 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
375 numa_default_policy();
376 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
378 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
380 complete(&kthreadd_done);
383 * The boot idle thread must execute schedule()
384 * at least once to get things moving:
386 init_idle_bootup_task(current);
387 schedule_preempt_disabled();
388 /* Call into cpu_idle with preempt disabled */
389 cpu_startup_entry(CPUHP_ONLINE);
392 /* Check for early params. */
393 static int __init do_early_param(char *param, char *val, const char *unused)
395 const struct obs_kernel_param *p;
397 for (p = __setup_start; p < __setup_end; p++) {
398 if ((p->early && parameq(param, p->str)) ||
399 (strcmp(param, "console") == 0 &&
400 strcmp(p->str, "earlycon") == 0)
402 if (p->setup_func(val) != 0)
403 pr_warn("Malformed early option '%s'\n", param);
406 /* We accept everything at this stage. */
410 void __init parse_early_options(char *cmdline)
412 parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
415 /* Arch code calls this early on, or if not, just before other parsing. */
416 void __init parse_early_param(void)
418 static __initdata int done = 0;
419 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
424 /* All fall through to do_early_param. */
425 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
426 parse_early_options(tmp_cmdline);
431 * Activate the first processor.
434 static void __init boot_cpu_init(void)
436 int cpu = smp_processor_id();
437 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
438 set_cpu_online(cpu, true);
439 set_cpu_active(cpu, true);
440 set_cpu_present(cpu, true);
441 set_cpu_possible(cpu, true);
444 void __init __weak smp_setup_processor_id(void)
448 # if THREAD_SIZE >= PAGE_SIZE
449 void __init __weak thread_info_cache_init(void)
455 * Set up kernel memory allocators
457 static void __init mm_init(void)
460 * page_cgroup requires contiguous pages,
461 * bigger than MAX_ORDER unless SPARSEMEM.
463 page_cgroup_init_flatmem();
467 pgtable_cache_init();
471 asmlinkage void __init start_kernel(void)
474 extern const struct kernel_param __start___param[], __stop___param[];
477 * Need to run as early as possible, to initialize the
481 smp_setup_processor_id();
482 debug_objects_early_init();
485 * Set up the the initial canary ASAP:
487 boot_init_stack_canary();
492 early_boot_irqs_disabled = true;
495 * Interrupts are still disabled. Do necessary setups, then
500 pr_notice("%s", linux_banner);
501 setup_arch(&command_line);
502 mm_init_owner(&init_mm, &init_task);
503 mm_init_cpumask(&init_mm);
504 setup_command_line(command_line);
506 setup_per_cpu_areas();
507 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
509 build_all_zonelists(NULL, NULL);
512 pr_notice("Kernel command line: %s\n", boot_command_line);
514 parse_args("Booting kernel", static_command_line, __start___param,
515 __stop___param - __start___param,
516 -1, -1, &unknown_bootoption);
521 * These use large bootmem allocations and must precede
526 vfs_caches_init_early();
532 * Set up the scheduler prior starting any interrupts (such as the
533 * timer interrupt). Full topology setup happens at smp_init()
534 * time - but meanwhile we still have a functioning scheduler.
538 * Disable preemption - early bootup scheduling is extremely
539 * fragile until we cpu_idle() for the first time.
542 if (WARN(!irqs_disabled(), "Interrupts were enabled *very* early, fixing it\n"))
548 /* init some links before init_ISA_irqs() */
558 call_function_init();
559 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
560 early_boot_irqs_disabled = false;
563 kmem_cache_init_late();
566 * HACK ALERT! This is early. We're enabling the console before
567 * we've done PCI setups etc, and console_init() must be aware of
568 * this. But we do want output early, in case something goes wrong.
572 panic(panic_later, panic_param);
577 * Need to run this when irqs are enabled, because it wants
578 * to self-test [hard/soft]-irqs on/off lock inversion bugs
583 #ifdef CONFIG_BLK_DEV_INITRD
584 if (initrd_start && !initrd_below_start_ok &&
585 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
586 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
587 page_to_pfn(virt_to_page((void *)initrd_start)),
593 debug_objects_mem_init();
595 setup_per_cpu_pageset();
604 if (efi_enabled(EFI_RUNTIME_SERVICES))
605 efi_enter_virtual_mode();
607 thread_info_cache_init();
609 fork_init(totalram_pages);
615 vfs_caches_init(totalram_pages);
617 /* rootfs populating might need page-writeback */
618 page_writeback_init();
619 #ifdef CONFIG_PROC_FS
624 taskstats_init_early();
629 acpi_early_init(); /* before LAPIC and SMP init */
632 if (efi_enabled(EFI_RUNTIME_SERVICES)) {
634 efi_free_boot_services();
639 /* Do the rest non-__init'ed, we're now alive */
643 /* Call all constructor functions linked into the kernel. */
644 static void __init do_ctors(void)
646 #ifdef CONFIG_CONSTRUCTORS
647 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
649 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
655 core_param(initcall_debug, initcall_debug, bool, 0644);
657 static char msgbuf[64];
659 static int __init_or_module do_one_initcall_debug(initcall_t fn)
661 ktime_t calltime, delta, rettime;
662 unsigned long long duration;
665 pr_debug("calling %pF @ %i\n", fn, task_pid_nr(current));
666 calltime = ktime_get();
668 rettime = ktime_get();
669 delta = ktime_sub(rettime, calltime);
670 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
671 pr_debug("initcall %pF returned %d after %lld usecs\n",
677 int __init_or_module do_one_initcall(initcall_t fn)
679 int count = preempt_count();
683 ret = do_one_initcall_debug(fn);
689 if (ret && ret != -ENODEV && initcall_debug)
690 sprintf(msgbuf, "error code %d ", ret);
692 if (preempt_count() != count) {
693 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
694 preempt_count() = count;
696 if (irqs_disabled()) {
697 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
700 WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
706 extern initcall_t __initcall_start[];
707 extern initcall_t __initcall0_start[];
708 extern initcall_t __initcall1_start[];
709 extern initcall_t __initcall2_start[];
710 extern initcall_t __initcall3_start[];
711 extern initcall_t __initcall4_start[];
712 extern initcall_t __initcall5_start[];
713 extern initcall_t __initcall6_start[];
714 extern initcall_t __initcall7_start[];
715 extern initcall_t __initcall_end[];
717 static initcall_t *initcall_levels[] __initdata = {
729 /* Keep these in sync with initcalls in include/linux/init.h */
730 static char *initcall_level_names[] __initdata = {
741 static void __init do_initcall_level(int level)
743 extern const struct kernel_param __start___param[], __stop___param[];
746 strcpy(static_command_line, saved_command_line);
747 parse_args(initcall_level_names[level],
748 static_command_line, __start___param,
749 __stop___param - __start___param,
753 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
754 do_one_initcall(*fn);
757 static void __init do_initcalls(void)
761 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
762 do_initcall_level(level);
766 * Ok, the machine is now initialized. None of the devices
767 * have been touched yet, but the CPU subsystem is up and
768 * running, and memory and process management works.
770 * Now we can finally start doing some real work..
772 static void __init do_basic_setup(void)
775 usermodehelper_init();
780 usermodehelper_enable();
784 static void __init do_pre_smp_initcalls(void)
788 for (fn = __initcall_start; fn < __initcall0_start; fn++)
789 do_one_initcall(*fn);
793 * This function requests modules which should be loaded by default and is
794 * called twice right after initrd is mounted and right before init is
795 * exec'd. If such modules are on either initrd or rootfs, they will be
796 * loaded before control is passed to userland.
798 void __init load_default_modules(void)
800 load_default_elevator_module();
803 static int run_init_process(const char *init_filename)
805 argv_init[0] = init_filename;
806 return do_execve(init_filename,
807 (const char __user *const __user *)argv_init,
808 (const char __user *const __user *)envp_init);
811 static noinline void __init kernel_init_freeable(void);
813 static int __ref kernel_init(void *unused)
815 kernel_init_freeable();
816 /* need to finish all async __init code before freeing the memory */
817 async_synchronize_full();
820 system_state = SYSTEM_RUNNING;
821 numa_default_policy();
823 flush_delayed_fput();
825 if (ramdisk_execute_command) {
826 if (!run_init_process(ramdisk_execute_command))
828 pr_err("Failed to execute %s\n", ramdisk_execute_command);
832 * We try each of these until one succeeds.
834 * The Bourne shell can be used instead of init if we are
835 * trying to recover a really broken machine.
837 if (execute_command) {
838 if (!run_init_process(execute_command))
840 pr_err("Failed to execute %s. Attempting defaults...\n",
843 if (!run_init_process("/sbin/init") ||
844 !run_init_process("/etc/init") ||
845 !run_init_process("/bin/init") ||
846 !run_init_process("/bin/sh"))
849 panic("No init found. Try passing init= option to kernel. "
850 "See Linux Documentation/init.txt for guidance.");
853 static noinline void __init kernel_init_freeable(void)
856 * Wait until kthreadd is all set-up.
858 wait_for_completion(&kthreadd_done);
860 /* Now the scheduler is fully set up and can do blocking allocations */
861 gfp_allowed_mask = __GFP_BITS_MASK;
864 * init can allocate pages on any node
866 set_mems_allowed(node_states[N_MEMORY]);
868 * init can run on any cpu.
870 set_cpus_allowed_ptr(current, cpu_all_mask);
872 cad_pid = task_pid(current);
874 smp_prepare_cpus(setup_max_cpus);
876 do_pre_smp_initcalls();
877 lockup_detector_init();
884 /* Open the /dev/console on the rootfs, this should never fail */
885 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
886 pr_err("Warning: unable to open an initial console.\n");
891 * check if there is an early userspace init. If yes, let it do all
895 if (!ramdisk_execute_command)
896 ramdisk_execute_command = "/init";
898 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
899 ramdisk_execute_command = NULL;
904 * Ok, we have completed the initial bootup, and
905 * we're essentially up and running. Get rid of the
906 * initmem segments and start the user-mode stuff..
909 /* rootfs is available now, try loading default modules */
910 load_default_modules();