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 #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>
74 #include <asm/setup.h>
75 #include <asm/sections.h>
76 #include <asm/cacheflush.h>
78 #ifdef CONFIG_X86_LOCAL_APIC
82 static int kernel_init(void *);
84 extern void init_IRQ(void);
85 extern void fork_init(unsigned long);
86 extern void mca_init(void);
87 extern void sbus_init(void);
88 extern void prio_tree_init(void);
89 extern void radix_tree_init(void);
90 #ifndef CONFIG_DEBUG_RODATA
91 static inline void mark_rodata_ro(void) { }
95 extern void tc_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);
119 extern void softirq_init(void);
121 /* Untouched command line saved by arch-specific code. */
122 char __initdata boot_command_line[COMMAND_LINE_SIZE];
123 /* Untouched saved command line (eg. for /proc) */
124 char *saved_command_line;
125 /* Command line for parameter parsing */
126 static char *static_command_line;
128 static char *execute_command;
129 static char *ramdisk_execute_command;
132 * If set, this is an indication to the drivers that reset the underlying
133 * device before going ahead with the initialization otherwise driver might
134 * rely on the BIOS and skip the reset operation.
136 * This is useful if kernel is booting in an unreliable environment.
137 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
138 * skipped and devices will be in unknown state.
140 unsigned int reset_devices;
141 EXPORT_SYMBOL(reset_devices);
143 static int __init set_reset_devices(char *str)
149 __setup("reset_devices", set_reset_devices);
151 static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
152 const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
153 static const char *panic_later, *panic_param;
155 extern const struct obs_kernel_param __setup_start[], __setup_end[];
157 static int __init obsolete_checksetup(char *line)
159 const struct obs_kernel_param *p;
160 int had_early_param = 0;
164 int n = strlen(p->str);
165 if (parameqn(line, p->str, n)) {
167 /* Already done in parse_early_param?
168 * (Needs exact match on param part).
169 * Keep iterating, as we can have early
170 * params and __setups of same names 8( */
171 if (line[n] == '\0' || line[n] == '=')
173 } else if (!p->setup_func) {
174 printk(KERN_WARNING "Parameter %s is obsolete,"
175 " ignored\n", p->str);
177 } else if (p->setup_func(line + n))
181 } while (p < __setup_end);
183 return had_early_param;
187 * This should be approx 2 Bo*oMips to start (note initial shift), and will
188 * still work even if initially too large, it will just take slightly longer
190 unsigned long loops_per_jiffy = (1<<12);
192 EXPORT_SYMBOL(loops_per_jiffy);
194 static int __init debug_kernel(char *str)
196 console_loglevel = 10;
200 static int __init quiet_kernel(char *str)
202 console_loglevel = 4;
206 early_param("debug", debug_kernel);
207 early_param("quiet", quiet_kernel);
209 static int __init loglevel(char *str)
214 * Only update loglevel value when a correct setting was passed,
215 * to prevent blind crashes (when loglevel being set to 0) that
216 * are quite hard to debug
218 if (get_option(&str, &newlevel)) {
219 console_loglevel = newlevel;
226 early_param("loglevel", loglevel);
228 /* Change NUL term back to "=", to make "param" the whole string. */
229 static int __init repair_env_string(char *param, char *val, const char *unused)
232 /* param=val or param="val"? */
233 if (val == param+strlen(param)+1)
235 else if (val == param+strlen(param)+2) {
237 memmove(val-1, val, strlen(val)+1);
246 * Unknown boot options get handed to init, unless they look like
247 * unused parameters (modprobe will find them in /proc/cmdline).
249 static int __init unknown_bootoption(char *param, char *val, const char *unused)
251 repair_env_string(param, val, unused);
253 /* Handle obsolete-style parameters */
254 if (obsolete_checksetup(param))
257 /* Unused module parameter. */
258 if (strchr(param, '.') && (!val || strchr(param, '.') < val))
265 /* Environment option */
267 for (i = 0; envp_init[i]; i++) {
268 if (i == MAX_INIT_ENVS) {
269 panic_later = "Too many boot env vars at `%s'";
272 if (!strncmp(param, envp_init[i], val - param))
275 envp_init[i] = param;
277 /* Command line option */
279 for (i = 0; argv_init[i]; i++) {
280 if (i == MAX_INIT_ARGS) {
281 panic_later = "Too many boot init vars at `%s'";
285 argv_init[i] = param;
290 static int __init init_setup(char *str)
294 execute_command = str;
296 * In case LILO is going to boot us with default command line,
297 * it prepends "auto" before the whole cmdline which makes
298 * the shell think it should execute a script with such name.
299 * So we ignore all arguments entered _before_ init=... [MJ]
301 for (i = 1; i < MAX_INIT_ARGS; i++)
305 __setup("init=", init_setup);
307 static int __init rdinit_setup(char *str)
311 ramdisk_execute_command = str;
312 /* See "auto" comment in init_setup */
313 for (i = 1; i < MAX_INIT_ARGS; i++)
317 __setup("rdinit=", rdinit_setup);
320 static const unsigned int setup_max_cpus = NR_CPUS;
321 #ifdef CONFIG_X86_LOCAL_APIC
322 static void __init smp_init(void)
324 APIC_init_uniprocessor();
327 #define smp_init() do { } while (0)
330 static inline void setup_nr_cpu_ids(void) { }
331 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
335 * We need to store the untouched command line for future reference.
336 * We also need to store the touched command line since the parameter
337 * parsing is performed in place, and we should allow a component to
338 * store reference of name/value for future reference.
340 static void __init setup_command_line(char *command_line)
342 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
343 static_command_line = alloc_bootmem(strlen (command_line)+1);
344 strcpy (saved_command_line, boot_command_line);
345 strcpy (static_command_line, command_line);
349 * We need to finalize in a non-__init function or else race conditions
350 * between the root thread and the init thread may cause start_kernel to
351 * be reaped by free_initmem before the root thread has proceeded to
354 * gcc-3.4 accidentally inlines this function, so use noinline.
357 static __initdata DECLARE_COMPLETION(kthreadd_done);
359 static noinline void __init_refok rest_init(void)
363 rcu_scheduler_starting();
365 * We need to spawn init first so that it obtains pid 1, however
366 * the init task will end up wanting to create kthreads, which, if
367 * we schedule it before we create kthreadd, will OOPS.
369 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
370 numa_default_policy();
371 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
373 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
375 complete(&kthreadd_done);
378 * The boot idle thread must execute schedule()
379 * at least once to get things moving:
381 init_idle_bootup_task(current);
382 schedule_preempt_disabled();
383 /* Call into cpu_idle with preempt disabled */
387 /* Check for early params. */
388 static int __init do_early_param(char *param, char *val, const char *unused)
390 const struct obs_kernel_param *p;
392 for (p = __setup_start; p < __setup_end; p++) {
393 if ((p->early && parameq(param, p->str)) ||
394 (strcmp(param, "console") == 0 &&
395 strcmp(p->str, "earlycon") == 0)
397 if (p->setup_func(val) != 0)
399 "Malformed early option '%s'\n", param);
402 /* We accept everything at this stage. */
406 void __init parse_early_options(char *cmdline)
408 parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
411 /* Arch code calls this early on, or if not, just before other parsing. */
412 void __init parse_early_param(void)
414 static __initdata int done = 0;
415 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
420 /* All fall through to do_early_param. */
421 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
422 parse_early_options(tmp_cmdline);
427 * Activate the first processor.
430 static void __init boot_cpu_init(void)
432 int cpu = smp_processor_id();
433 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
434 set_cpu_online(cpu, true);
435 set_cpu_active(cpu, true);
436 set_cpu_present(cpu, true);
437 set_cpu_possible(cpu, true);
440 void __init __weak smp_setup_processor_id(void)
444 void __init __weak thread_info_cache_init(void)
449 * Set up kernel memory allocators
451 static void __init mm_init(void)
454 * page_cgroup requires contiguous pages,
455 * bigger than MAX_ORDER unless SPARSEMEM.
457 page_cgroup_init_flatmem();
461 pgtable_cache_init();
465 asmlinkage void __init start_kernel(void)
468 extern const struct kernel_param __start___param[], __stop___param[];
471 * Need to run as early as possible, to initialize the
475 smp_setup_processor_id();
476 debug_objects_early_init();
479 * Set up the the initial canary ASAP:
481 boot_init_stack_canary();
486 early_boot_irqs_disabled = true;
489 * Interrupts are still disabled. Do necessary setups, then
495 printk(KERN_NOTICE "%s", linux_banner);
496 setup_arch(&command_line);
497 mm_init_owner(&init_mm, &init_task);
498 mm_init_cpumask(&init_mm);
499 setup_command_line(command_line);
501 setup_per_cpu_areas();
502 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
504 build_all_zonelists(NULL);
507 printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
509 parse_args("Booting kernel", static_command_line, __start___param,
510 __stop___param - __start___param,
511 0, 0, &unknown_bootoption);
516 * These use large bootmem allocations and must precede
521 vfs_caches_init_early();
527 * Set up the scheduler prior starting any interrupts (such as the
528 * timer interrupt). Full topology setup happens at smp_init()
529 * time - but meanwhile we still have a functioning scheduler.
533 * Disable preemption - early bootup scheduling is extremely
534 * fragile until we cpu_idle() for the first time.
537 if (!irqs_disabled()) {
538 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
539 "enabled *very* early, fixing it\n");
546 /* init some links before init_ISA_irqs() */
556 call_function_init();
557 if (!irqs_disabled())
558 printk(KERN_CRIT "start_kernel(): bug: interrupts were "
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 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
588 page_to_pfn(virt_to_page((void *)initrd_start)),
594 debug_objects_mem_init();
596 setup_per_cpu_pageset();
606 efi_enter_virtual_mode();
608 thread_info_cache_init();
610 fork_init(totalram_pages);
616 vfs_caches_init(totalram_pages);
618 /* rootfs populating might need page-writeback */
619 page_writeback_init();
620 #ifdef CONFIG_PROC_FS
625 taskstats_init_early();
630 acpi_early_init(); /* before LAPIC and SMP init */
635 /* Do the rest non-__init'ed, we're now alive */
639 /* Call all constructor functions linked into the kernel. */
640 static void __init do_ctors(void)
642 #ifdef CONFIG_CONSTRUCTORS
643 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
645 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
651 core_param(initcall_debug, initcall_debug, bool, 0644);
653 static char msgbuf[64];
655 static int __init_or_module do_one_initcall_debug(initcall_t fn)
657 ktime_t calltime, delta, rettime;
658 unsigned long long duration;
661 printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current));
662 calltime = ktime_get();
664 rettime = ktime_get();
665 delta = ktime_sub(rettime, calltime);
666 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
667 printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", fn,
673 int __init_or_module do_one_initcall(initcall_t fn)
675 int count = preempt_count();
679 ret = do_one_initcall_debug(fn);
685 if (ret && ret != -ENODEV && initcall_debug)
686 sprintf(msgbuf, "error code %d ", ret);
688 if (preempt_count() != count) {
689 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
690 preempt_count() = count;
692 if (irqs_disabled()) {
693 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
697 printk("initcall %pF returned with %s\n", fn, msgbuf);
704 extern initcall_t __initcall_start[];
705 extern initcall_t __initcall0_start[];
706 extern initcall_t __initcall1_start[];
707 extern initcall_t __initcall2_start[];
708 extern initcall_t __initcall3_start[];
709 extern initcall_t __initcall4_start[];
710 extern initcall_t __initcall5_start[];
711 extern initcall_t __initcall6_start[];
712 extern initcall_t __initcall7_start[];
713 extern initcall_t __initcall_end[];
715 static initcall_t *initcall_levels[] __initdata = {
727 static char *initcall_level_names[] __initdata = {
738 static void __init do_initcall_level(int level)
740 extern const struct kernel_param __start___param[], __stop___param[];
743 strcpy(static_command_line, saved_command_line);
744 parse_args(initcall_level_names[level],
745 static_command_line, __start___param,
746 __stop___param - __start___param,
750 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
751 do_one_initcall(*fn);
754 static void __init do_initcalls(void)
758 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
759 pr_info("initlevel:%d=%s, %d registered initcalls\n",
760 level, initcall_level_names[level],
761 (int) (initcall_levels[level+1]
762 - initcall_levels[level]));
763 do_initcall_level(level);
768 * Ok, the machine is now initialized. None of the devices
769 * have been touched yet, but the CPU subsystem is up and
770 * running, and memory and process management works.
772 * Now we can finally start doing some real work..
774 static void __init do_basic_setup(void)
777 usermodehelper_init();
782 usermodehelper_enable();
786 static void __init do_pre_smp_initcalls(void)
790 for (fn = __initcall_start; fn < __initcall0_start; fn++)
791 do_one_initcall(*fn);
794 static void run_init_process(const char *init_filename)
796 argv_init[0] = init_filename;
797 kernel_execve(init_filename, argv_init, envp_init);
800 /* This is a non __init function. Force it to be noinline otherwise gcc
801 * makes it inline to init() and it becomes part of init.text section
803 static noinline int init_post(void)
805 /* need to finish all async __init code before freeing the memory */
806 async_synchronize_full();
809 system_state = SYSTEM_RUNNING;
810 numa_default_policy();
813 current->signal->flags |= SIGNAL_UNKILLABLE;
815 if (ramdisk_execute_command) {
816 run_init_process(ramdisk_execute_command);
817 printk(KERN_WARNING "Failed to execute %s\n",
818 ramdisk_execute_command);
822 * We try each of these until one succeeds.
824 * The Bourne shell can be used instead of init if we are
825 * trying to recover a really broken machine.
827 if (execute_command) {
828 run_init_process(execute_command);
829 printk(KERN_WARNING "Failed to execute %s. Attempting "
830 "defaults...\n", execute_command);
832 run_init_process("/sbin/init");
833 run_init_process("/etc/init");
834 run_init_process("/bin/init");
835 run_init_process("/bin/sh");
837 panic("No init found. Try passing init= option to kernel. "
838 "See Linux Documentation/init.txt for guidance.");
841 static int __init kernel_init(void * unused)
844 * Wait until kthreadd is all set-up.
846 wait_for_completion(&kthreadd_done);
848 /* Now the scheduler is fully set up and can do blocking allocations */
849 gfp_allowed_mask = __GFP_BITS_MASK;
852 * init can allocate pages on any node
854 set_mems_allowed(node_states[N_HIGH_MEMORY]);
856 * init can run on any cpu.
858 set_cpus_allowed_ptr(current, cpu_all_mask);
860 cad_pid = task_pid(current);
862 smp_prepare_cpus(setup_max_cpus);
864 do_pre_smp_initcalls();
865 lockup_detector_init();
872 /* Open the /dev/console on the rootfs, this should never fail */
873 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
874 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
879 * check if there is an early userspace init. If yes, let it do all
883 if (!ramdisk_execute_command)
884 ramdisk_execute_command = "/init";
886 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
887 ramdisk_execute_command = NULL;
892 * Ok, we have completed the initial bootup, and
893 * we're essentially up and running. Get rid of the
894 * initmem segments and start the user-mode stuff..