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/smp_lock.h>
24 #include <linux/initrd.h>
25 #include <linux/bootmem.h>
26 #include <linux/acpi.h>
27 #include <linux/tty.h>
28 #include <linux/gfp.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/workqueue.h>
37 #include <linux/profile.h>
38 #include <linux/rcupdate.h>
39 #include <linux/moduleparam.h>
40 #include <linux/kallsyms.h>
41 #include <linux/writeback.h>
42 #include <linux/cpu.h>
43 #include <linux/cpuset.h>
44 #include <linux/cgroup.h>
45 #include <linux/efi.h>
46 #include <linux/tick.h>
47 #include <linux/interrupt.h>
48 #include <linux/taskstats_kern.h>
49 #include <linux/delayacct.h>
50 #include <linux/unistd.h>
51 #include <linux/rmap.h>
52 #include <linux/mempolicy.h>
53 #include <linux/key.h>
54 #include <linux/buffer_head.h>
55 #include <linux/page_cgroup.h>
56 #include <linux/debug_locks.h>
57 #include <linux/debugobjects.h>
58 #include <linux/lockdep.h>
59 #include <linux/kmemleak.h>
60 #include <linux/pid_namespace.h>
61 #include <linux/device.h>
62 #include <linux/kthread.h>
63 #include <linux/sched.h>
64 #include <linux/signal.h>
65 #include <linux/idr.h>
66 #include <linux/ftrace.h>
67 #include <linux/async.h>
68 #include <linux/kmemcheck.h>
69 #include <linux/kmemtrace.h>
70 #include <linux/sfi.h>
71 #include <linux/shmem_fs.h>
72 #include <trace/boot.h>
76 #include <asm/setup.h>
77 #include <asm/sections.h>
78 #include <asm/cacheflush.h>
80 #ifdef CONFIG_X86_LOCAL_APIC
84 static int kernel_init(void *);
86 extern void init_IRQ(void);
87 extern void fork_init(unsigned long);
88 extern void mca_init(void);
89 extern void sbus_init(void);
90 extern void prio_tree_init(void);
91 extern void radix_tree_init(void);
92 extern void free_initmem(void);
93 #ifndef CONFIG_DEBUG_RODATA
94 static inline void mark_rodata_ro(void) { }
98 extern void tc_init(void);
101 enum system_states system_state __read_mostly;
102 EXPORT_SYMBOL(system_state);
105 * Boot command-line arguments
107 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
108 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
110 extern void time_init(void);
111 /* Default late time init is NULL. archs can override this later. */
112 void (*__initdata late_time_init)(void);
113 extern void softirq_init(void);
115 /* Untouched command line saved by arch-specific code. */
116 char __initdata boot_command_line[COMMAND_LINE_SIZE];
117 /* Untouched saved command line (eg. for /proc) */
118 char *saved_command_line;
119 /* Command line for parameter parsing */
120 static char *static_command_line;
122 static char *execute_command;
123 static char *ramdisk_execute_command;
126 /* Setup configured maximum number of CPUs to activate */
127 unsigned int __initdata setup_max_cpus = NR_CPUS;
130 * Setup routine for controlling SMP activation
132 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
133 * activation entirely (the MPS table probe still happens, though).
135 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
136 * greater than 0, limits the maximum number of CPUs activated in
140 void __weak arch_disable_smp_support(void) { }
142 static int __init nosmp(char *str)
145 arch_disable_smp_support();
150 early_param("nosmp", nosmp);
152 static int __init maxcpus(char *str)
154 get_option(&str, &setup_max_cpus);
155 if (setup_max_cpus == 0)
156 arch_disable_smp_support();
161 early_param("maxcpus", maxcpus);
163 const unsigned int setup_max_cpus = NR_CPUS;
167 * If set, this is an indication to the drivers that reset the underlying
168 * device before going ahead with the initialization otherwise driver might
169 * rely on the BIOS and skip the reset operation.
171 * This is useful if kernel is booting in an unreliable environment.
172 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
173 * skipped and devices will be in unknown state.
175 unsigned int reset_devices;
176 EXPORT_SYMBOL(reset_devices);
178 static int __init set_reset_devices(char *str)
184 __setup("reset_devices", set_reset_devices);
186 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
187 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
188 static const char *panic_later, *panic_param;
190 extern struct obs_kernel_param __setup_start[], __setup_end[];
192 static int __init obsolete_checksetup(char *line)
194 struct obs_kernel_param *p;
195 int had_early_param = 0;
199 int n = strlen(p->str);
200 if (!strncmp(line, p->str, n)) {
202 /* Already done in parse_early_param?
203 * (Needs exact match on param part).
204 * Keep iterating, as we can have early
205 * params and __setups of same names 8( */
206 if (line[n] == '\0' || line[n] == '=')
208 } else if (!p->setup_func) {
209 printk(KERN_WARNING "Parameter %s is obsolete,"
210 " ignored\n", p->str);
212 } else if (p->setup_func(line + n))
216 } while (p < __setup_end);
218 return had_early_param;
222 * This should be approx 2 Bo*oMips to start (note initial shift), and will
223 * still work even if initially too large, it will just take slightly longer
225 unsigned long loops_per_jiffy = (1<<12);
227 EXPORT_SYMBOL(loops_per_jiffy);
229 static int __init debug_kernel(char *str)
231 console_loglevel = 10;
235 static int __init quiet_kernel(char *str)
237 console_loglevel = 4;
241 early_param("debug", debug_kernel);
242 early_param("quiet", quiet_kernel);
244 static int __init loglevel(char *str)
246 get_option(&str, &console_loglevel);
250 early_param("loglevel", loglevel);
253 * Unknown boot options get handed to init, unless they look like
254 * unused parameters (modprobe will find them in /proc/cmdline).
256 static int __init unknown_bootoption(char *param, char *val)
258 /* Change NUL term back to "=", to make "param" the whole string. */
260 /* param=val or param="val"? */
261 if (val == param+strlen(param)+1)
263 else if (val == param+strlen(param)+2) {
265 memmove(val-1, val, strlen(val)+1);
271 /* Handle obsolete-style parameters */
272 if (obsolete_checksetup(param))
275 /* Unused module parameter. */
276 if (strchr(param, '.') && (!val || strchr(param, '.') < val))
283 /* Environment option */
285 for (i = 0; envp_init[i]; i++) {
286 if (i == MAX_INIT_ENVS) {
287 panic_later = "Too many boot env vars at `%s'";
290 if (!strncmp(param, envp_init[i], val - param))
293 envp_init[i] = param;
295 /* Command line option */
297 for (i = 0; argv_init[i]; i++) {
298 if (i == MAX_INIT_ARGS) {
299 panic_later = "Too many boot init vars at `%s'";
303 argv_init[i] = param;
308 #ifdef CONFIG_DEBUG_PAGEALLOC
309 int __read_mostly debug_pagealloc_enabled = 0;
312 static int __init init_setup(char *str)
316 execute_command = str;
318 * In case LILO is going to boot us with default command line,
319 * it prepends "auto" before the whole cmdline which makes
320 * the shell think it should execute a script with such name.
321 * So we ignore all arguments entered _before_ init=... [MJ]
323 for (i = 1; i < MAX_INIT_ARGS; i++)
327 __setup("init=", init_setup);
329 static int __init rdinit_setup(char *str)
333 ramdisk_execute_command = str;
334 /* See "auto" comment in init_setup */
335 for (i = 1; i < MAX_INIT_ARGS; i++)
339 __setup("rdinit=", rdinit_setup);
343 #ifdef CONFIG_X86_LOCAL_APIC
344 static void __init smp_init(void)
346 APIC_init_uniprocessor();
349 #define smp_init() do { } while (0)
352 static inline void setup_nr_cpu_ids(void) { }
353 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
357 /* Setup number of possible processor ids */
358 int nr_cpu_ids __read_mostly = NR_CPUS;
359 EXPORT_SYMBOL(nr_cpu_ids);
361 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
362 static void __init setup_nr_cpu_ids(void)
364 nr_cpu_ids = find_last_bit(cpumask_bits(cpu_possible_mask),NR_CPUS) + 1;
367 /* Called by boot processor to activate the rest. */
368 static void __init smp_init(void)
372 /* FIXME: This should be done in userspace --RR */
373 for_each_present_cpu(cpu) {
374 if (num_online_cpus() >= setup_max_cpus)
376 if (!cpu_online(cpu))
380 /* Any cleanup work */
381 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
382 smp_cpus_done(setup_max_cpus);
388 * We need to store the untouched command line for future reference.
389 * We also need to store the touched command line since the parameter
390 * parsing is performed in place, and we should allow a component to
391 * store reference of name/value for future reference.
393 static void __init setup_command_line(char *command_line)
395 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
396 static_command_line = alloc_bootmem(strlen (command_line)+1);
397 strcpy (saved_command_line, boot_command_line);
398 strcpy (static_command_line, command_line);
402 * We need to finalize in a non-__init function or else race conditions
403 * between the root thread and the init thread may cause start_kernel to
404 * be reaped by free_initmem before the root thread has proceeded to
407 * gcc-3.4 accidentally inlines this function, so use noinline.
410 static noinline void __init_refok rest_init(void)
411 __releases(kernel_lock)
415 rcu_scheduler_starting();
416 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
417 numa_default_policy();
418 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
419 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
423 * The boot idle thread must execute schedule()
424 * at least once to get things moving:
426 init_idle_bootup_task(current);
427 preempt_enable_no_resched();
431 /* Call into cpu_idle with preempt disabled */
435 /* Check for early params. */
436 static int __init do_early_param(char *param, char *val)
438 struct obs_kernel_param *p;
440 for (p = __setup_start; p < __setup_end; p++) {
441 if ((p->early && strcmp(param, p->str) == 0) ||
442 (strcmp(param, "console") == 0 &&
443 strcmp(p->str, "earlycon") == 0)
445 if (p->setup_func(val) != 0)
447 "Malformed early option '%s'\n", param);
450 /* We accept everything at this stage. */
454 void __init parse_early_options(char *cmdline)
456 parse_args("early options", cmdline, NULL, 0, do_early_param);
459 /* Arch code calls this early on, or if not, just before other parsing. */
460 void __init parse_early_param(void)
462 static __initdata int done = 0;
463 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
468 /* All fall through to do_early_param. */
469 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
470 parse_early_options(tmp_cmdline);
475 * Activate the first processor.
478 static void __init boot_cpu_init(void)
480 int cpu = smp_processor_id();
481 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
482 set_cpu_online(cpu, true);
483 set_cpu_active(cpu, true);
484 set_cpu_present(cpu, true);
485 set_cpu_possible(cpu, true);
488 void __init __weak smp_setup_processor_id(void)
492 void __init __weak thread_info_cache_init(void)
497 * Set up kernel memory allocators
499 static void __init mm_init(void)
502 * page_cgroup requires countinous pages as memmap
503 * and it's bigger than MAX_ORDER unless SPARSEMEM.
505 page_cgroup_init_flatmem();
508 pgtable_cache_init();
512 asmlinkage void __init start_kernel(void)
515 extern struct kernel_param __start___param[], __stop___param[];
517 smp_setup_processor_id();
520 * Need to run as early as possible, to initialize the
524 debug_objects_early_init();
527 * Set up the the initial canary ASAP:
529 boot_init_stack_canary();
534 early_boot_irqs_off();
535 early_init_irq_lock_class();
538 * Interrupts are still disabled. Do necessary setups, then
545 printk(KERN_NOTICE "%s", linux_banner);
546 setup_arch(&command_line);
547 mm_init_owner(&init_mm, &init_task);
548 setup_command_line(command_line);
550 setup_per_cpu_areas();
551 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
553 build_all_zonelists();
556 printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
558 parse_args("Booting kernel", static_command_line, __start___param,
559 __stop___param - __start___param,
560 &unknown_bootoption);
562 * These use large bootmem allocations and must precede
566 vfs_caches_init_early();
571 * Set up the scheduler prior starting any interrupts (such as the
572 * timer interrupt). Full topology setup happens at smp_init()
573 * time - but meanwhile we still have a functioning scheduler.
577 * Disable preemption - early bootup scheduling is extremely
578 * fragile until we cpu_idle() for the first time.
581 if (!irqs_disabled()) {
582 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
583 "enabled *very* early, fixing it\n");
587 /* init some links before init_ISA_irqs() */
597 if (!irqs_disabled())
598 printk(KERN_CRIT "start_kernel(): bug: interrupts were "
600 early_boot_irqs_on();
603 /* Interrupts are enabled now so all GFP allocations are safe. */
604 set_gfp_allowed_mask(__GFP_BITS_MASK);
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(panic_later, panic_param);
620 * Need to run this when irqs are enabled, because it wants
621 * to self-test [hard/soft]-irqs on/off lock inversion bugs
626 #ifdef CONFIG_BLK_DEV_INITRD
627 if (initrd_start && !initrd_below_start_ok &&
628 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
629 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
631 page_to_pfn(virt_to_page((void *)initrd_start)),
637 enable_debug_pagealloc();
640 debug_objects_mem_init();
642 setup_per_cpu_pageset();
652 efi_enter_virtual_mode();
654 thread_info_cache_init();
656 fork_init(totalram_pages);
662 vfs_caches_init(totalram_pages);
664 /* rootfs populating might need page-writeback */
665 page_writeback_init();
666 #ifdef CONFIG_PROC_FS
671 taskstats_init_early();
676 acpi_early_init(); /* before LAPIC and SMP init */
681 /* Do the rest non-__init'ed, we're now alive */
685 /* Call all constructor functions linked into the kernel. */
686 static void __init do_ctors(void)
688 #ifdef CONFIG_CONSTRUCTORS
689 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
691 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
697 core_param(initcall_debug, initcall_debug, bool, 0644);
699 static char msgbuf[64];
700 static struct boot_trace_call call;
701 static struct boot_trace_ret ret;
703 int do_one_initcall(initcall_t fn)
705 int count = preempt_count();
706 ktime_t calltime, delta, rettime;
708 if (initcall_debug) {
709 call.caller = task_pid_nr(current);
710 printk("calling %pF @ %i\n", fn, call.caller);
711 calltime = ktime_get();
712 trace_boot_call(&call, fn);
718 if (initcall_debug) {
719 disable_boot_trace();
720 rettime = ktime_get();
721 delta = ktime_sub(rettime, calltime);
722 ret.duration = (unsigned long long) ktime_to_ns(delta) >> 10;
723 trace_boot_ret(&ret, fn);
724 printk("initcall %pF returned %d after %Ld usecs\n", fn,
725 ret.result, ret.duration);
730 if (ret.result && ret.result != -ENODEV && initcall_debug)
731 sprintf(msgbuf, "error code %d ", ret.result);
733 if (preempt_count() != count) {
734 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
735 preempt_count() = count;
737 if (irqs_disabled()) {
738 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
742 printk("initcall %pF returned with %s\n", fn, msgbuf);
749 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
751 static void __init do_initcalls(void)
755 for (fn = __early_initcall_end; fn < __initcall_end; fn++)
756 do_one_initcall(*fn);
758 /* Make sure there is no pending stuff from the initcall sequence */
759 flush_scheduled_work();
763 * Ok, the machine is now initialized. None of the devices
764 * have been touched yet, but the CPU subsystem is up and
765 * running, and memory and process management works.
767 * Now we can finally start doing some real work..
769 static void __init do_basic_setup(void)
773 usermodehelper_init();
781 static void __init do_pre_smp_initcalls(void)
785 for (fn = __initcall_start; fn < __early_initcall_end; fn++)
786 do_one_initcall(*fn);
789 static void run_init_process(char *init_filename)
791 argv_init[0] = init_filename;
792 kernel_execve(init_filename, argv_init, envp_init);
795 /* This is a non __init function. Force it to be noinline otherwise gcc
796 * makes it inline to init() and it becomes part of init.text section
798 static noinline int init_post(void)
799 __releases(kernel_lock)
801 /* need to finish all async __init code before freeing the memory */
802 async_synchronize_full();
806 system_state = SYSTEM_RUNNING;
807 numa_default_policy();
809 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
810 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
815 current->signal->flags |= SIGNAL_UNKILLABLE;
817 if (ramdisk_execute_command) {
818 run_init_process(ramdisk_execute_command);
819 printk(KERN_WARNING "Failed to execute %s\n",
820 ramdisk_execute_command);
824 * We try each of these until one succeeds.
826 * The Bourne shell can be used instead of init if we are
827 * trying to recover a really broken machine.
829 if (execute_command) {
830 run_init_process(execute_command);
831 printk(KERN_WARNING "Failed to execute %s. Attempting "
832 "defaults...\n", execute_command);
834 run_init_process("/sbin/init");
835 run_init_process("/etc/init");
836 run_init_process("/bin/init");
837 run_init_process("/bin/sh");
839 panic("No init found. Try passing init= option to kernel.");
842 static int __init kernel_init(void * unused)
847 * init can allocate pages on any node
849 set_mems_allowed(node_possible_map);
851 * init can run on any cpu.
853 set_cpus_allowed_ptr(current, cpu_all_mask);
855 * Tell the world that we're going to be the grim
856 * reaper of innocent orphaned children.
858 * We don't want people to have to make incorrect
859 * assumptions about where in the task array this
862 init_pid_ns.child_reaper = current;
864 cad_pid = task_pid(current);
866 smp_prepare_cpus(setup_max_cpus);
868 do_pre_smp_initcalls();
877 * check if there is an early userspace init. If yes, let it do all
881 if (!ramdisk_execute_command)
882 ramdisk_execute_command = "/init";
884 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
885 ramdisk_execute_command = NULL;
890 * Ok, we have completed the initial bootup, and
891 * we're essentially up and running. Get rid of the
892 * initmem segments and start the user-mode stuff..