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/string.h>
18 #include <linux/ctype.h>
19 #include <linux/delay.h>
20 #include <linux/utsname.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/tty.h>
27 #include <linux/gfp.h>
28 #include <linux/percpu.h>
29 #include <linux/kmod.h>
30 #include <linux/vmalloc.h>
31 #include <linux/kernel_stat.h>
32 #include <linux/start_kernel.h>
33 #include <linux/security.h>
34 #include <linux/smp.h>
35 #include <linux/workqueue.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/unwind.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/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/ftrace.h>
66 #include <trace/boot.h>
70 #include <asm/setup.h>
71 #include <asm/sections.h>
72 #include <asm/cacheflush.h>
73 #include <trace/kmemtrace.h>
75 #ifdef CONFIG_X86_LOCAL_APIC
80 * This is one of the first .c files built. Error out early if we have compiler
84 #if __GNUC__ == 4 && __GNUC_MINOR__ == 1 && __GNUC_PATCHLEVEL__ == 0
85 #warning gcc-4.1.0 is known to miscompile the kernel. A different compiler version is recommended.
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 prio_tree_init(void);
95 extern void radix_tree_init(void);
96 extern void free_initmem(void);
98 extern void acpi_early_init(void);
100 static inline void acpi_early_init(void) { }
102 #ifndef CONFIG_DEBUG_RODATA
103 static inline void mark_rodata_ro(void) { }
107 extern void tc_init(void);
110 enum system_states system_state;
111 EXPORT_SYMBOL(system_state);
114 * Boot command-line arguments
116 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
117 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
119 extern void time_init(void);
120 /* Default late time init is NULL. archs can override this later. */
121 void (*late_time_init)(void);
122 extern void softirq_init(void);
124 /* Untouched command line saved by arch-specific code. */
125 char __initdata boot_command_line[COMMAND_LINE_SIZE];
126 /* Untouched saved command line (eg. for /proc) */
127 char *saved_command_line;
128 /* Command line for parameter parsing */
129 static char *static_command_line;
131 static char *execute_command;
132 static char *ramdisk_execute_command;
135 /* Setup configured maximum number of CPUs to activate */
136 unsigned int __initdata setup_max_cpus = NR_CPUS;
139 * Setup routine for controlling SMP activation
141 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
142 * activation entirely (the MPS table probe still happens, though).
144 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
145 * greater than 0, limits the maximum number of CPUs activated in
148 #ifndef CONFIG_X86_IO_APIC
149 static inline void disable_ioapic_setup(void) {};
152 static int __init nosmp(char *str)
155 disable_ioapic_setup();
159 early_param("nosmp", nosmp);
161 static int __init maxcpus(char *str)
163 get_option(&str, &setup_max_cpus);
164 if (setup_max_cpus == 0)
165 disable_ioapic_setup();
170 early_param("maxcpus", maxcpus);
172 #define setup_max_cpus NR_CPUS
176 * If set, this is an indication to the drivers that reset the underlying
177 * device before going ahead with the initialization otherwise driver might
178 * rely on the BIOS and skip the reset operation.
180 * This is useful if kernel is booting in an unreliable environment.
181 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
182 * skipped and devices will be in unknown state.
184 unsigned int reset_devices;
185 EXPORT_SYMBOL(reset_devices);
187 static int __init set_reset_devices(char *str)
193 __setup("reset_devices", set_reset_devices);
195 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
196 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
197 static const char *panic_later, *panic_param;
199 extern struct obs_kernel_param __setup_start[], __setup_end[];
201 static int __init obsolete_checksetup(char *line)
203 struct obs_kernel_param *p;
204 int had_early_param = 0;
208 int n = strlen(p->str);
209 if (!strncmp(line, p->str, n)) {
211 /* Already done in parse_early_param?
212 * (Needs exact match on param part).
213 * Keep iterating, as we can have early
214 * params and __setups of same names 8( */
215 if (line[n] == '\0' || line[n] == '=')
217 } else if (!p->setup_func) {
218 printk(KERN_WARNING "Parameter %s is obsolete,"
219 " ignored\n", p->str);
221 } else if (p->setup_func(line + n))
225 } while (p < __setup_end);
227 return had_early_param;
231 * This should be approx 2 Bo*oMips to start (note initial shift), and will
232 * still work even if initially too large, it will just take slightly longer
234 unsigned long loops_per_jiffy = (1<<12);
236 EXPORT_SYMBOL(loops_per_jiffy);
238 static int __init debug_kernel(char *str)
240 console_loglevel = 10;
244 static int __init quiet_kernel(char *str)
246 console_loglevel = 4;
250 early_param("debug", debug_kernel);
251 early_param("quiet", quiet_kernel);
253 static int __init loglevel(char *str)
255 get_option(&str, &console_loglevel);
259 early_param("loglevel", loglevel);
262 * Unknown boot options get handed to init, unless they look like
265 static int __init unknown_bootoption(char *param, char *val)
267 /* Change NUL term back to "=", to make "param" the whole string. */
269 /* param=val or param="val"? */
270 if (val == param+strlen(param)+1)
272 else if (val == param+strlen(param)+2) {
274 memmove(val-1, val, strlen(val)+1);
280 /* Handle obsolete-style parameters */
281 if (obsolete_checksetup(param))
285 * Preemptive maintenance for "why didn't my misspelled command
288 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
289 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
297 /* Environment option */
299 for (i = 0; envp_init[i]; i++) {
300 if (i == MAX_INIT_ENVS) {
301 panic_later = "Too many boot env vars at `%s'";
304 if (!strncmp(param, envp_init[i], val - param))
307 envp_init[i] = param;
309 /* Command line option */
311 for (i = 0; argv_init[i]; i++) {
312 if (i == MAX_INIT_ARGS) {
313 panic_later = "Too many boot init vars at `%s'";
317 argv_init[i] = param;
322 #ifdef CONFIG_DEBUG_PAGEALLOC
323 int __read_mostly debug_pagealloc_enabled = 0;
326 static int __init init_setup(char *str)
330 execute_command = str;
332 * In case LILO is going to boot us with default command line,
333 * it prepends "auto" before the whole cmdline which makes
334 * the shell think it should execute a script with such name.
335 * So we ignore all arguments entered _before_ init=... [MJ]
337 for (i = 1; i < MAX_INIT_ARGS; i++)
341 __setup("init=", init_setup);
343 static int __init rdinit_setup(char *str)
347 ramdisk_execute_command = str;
348 /* See "auto" comment in init_setup */
349 for (i = 1; i < MAX_INIT_ARGS; i++)
353 __setup("rdinit=", rdinit_setup);
357 #ifdef CONFIG_X86_LOCAL_APIC
358 static void __init smp_init(void)
360 APIC_init_uniprocessor();
363 #define smp_init() do { } while (0)
366 static inline void setup_per_cpu_areas(void) { }
367 static inline void setup_nr_cpu_ids(void) { }
368 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
372 #if NR_CPUS > BITS_PER_LONG
373 cpumask_t cpu_mask_all __read_mostly = CPU_MASK_ALL;
374 EXPORT_SYMBOL(cpu_mask_all);
377 /* Setup number of possible processor ids */
378 int nr_cpu_ids __read_mostly = NR_CPUS;
379 EXPORT_SYMBOL(nr_cpu_ids);
381 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
382 static void __init setup_nr_cpu_ids(void)
384 int cpu, highest_cpu = 0;
386 for_each_possible_cpu(cpu)
389 nr_cpu_ids = highest_cpu + 1;
392 #ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA
393 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
395 EXPORT_SYMBOL(__per_cpu_offset);
397 static void __init setup_per_cpu_areas(void)
399 unsigned long size, i;
401 unsigned long nr_possible_cpus = num_possible_cpus();
403 /* Copy section for each CPU (we discard the original) */
404 size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
405 ptr = alloc_bootmem_pages(size * nr_possible_cpus);
407 for_each_possible_cpu(i) {
408 __per_cpu_offset[i] = ptr - __per_cpu_start;
409 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
413 #endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */
415 /* Called by boot processor to activate the rest. */
416 static void __init smp_init(void)
421 * Set up the current CPU as possible to migrate to.
422 * The other ones will be done by cpu_up/cpu_down()
424 cpu = smp_processor_id();
425 cpu_set(cpu, cpu_active_map);
427 /* FIXME: This should be done in userspace --RR */
428 for_each_present_cpu(cpu) {
429 if (num_online_cpus() >= setup_max_cpus)
431 if (!cpu_online(cpu))
435 /* Any cleanup work */
436 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
437 smp_cpus_done(setup_max_cpus);
443 * We need to store the untouched command line for future reference.
444 * We also need to store the touched command line since the parameter
445 * parsing is performed in place, and we should allow a component to
446 * store reference of name/value for future reference.
448 static void __init setup_command_line(char *command_line)
450 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
451 static_command_line = alloc_bootmem(strlen (command_line)+1);
452 strcpy (saved_command_line, boot_command_line);
453 strcpy (static_command_line, command_line);
457 * We need to finalize in a non-__init function or else race conditions
458 * between the root thread and the init thread may cause start_kernel to
459 * be reaped by free_initmem before the root thread has proceeded to
462 * gcc-3.4 accidentally inlines this function, so use noinline.
465 static void noinline __init_refok rest_init(void)
466 __releases(kernel_lock)
470 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
471 numa_default_policy();
472 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
473 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
477 * The boot idle thread must execute schedule()
478 * at least once to get things moving:
480 init_idle_bootup_task(current);
481 preempt_enable_no_resched();
485 /* Call into cpu_idle with preempt disabled */
489 /* Check for early params. */
490 static int __init do_early_param(char *param, char *val)
492 struct obs_kernel_param *p;
494 for (p = __setup_start; p < __setup_end; p++) {
495 if ((p->early && strcmp(param, p->str) == 0) ||
496 (strcmp(param, "console") == 0 &&
497 strcmp(p->str, "earlycon") == 0)
499 if (p->setup_func(val) != 0)
501 "Malformed early option '%s'\n", param);
504 /* We accept everything at this stage. */
508 /* Arch code calls this early on, or if not, just before other parsing. */
509 void __init parse_early_param(void)
511 static __initdata int done = 0;
512 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
517 /* All fall through to do_early_param. */
518 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
519 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
524 * Activate the first processor.
527 static void __init boot_cpu_init(void)
529 int cpu = smp_processor_id();
530 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
531 cpu_set(cpu, cpu_online_map);
532 cpu_set(cpu, cpu_present_map);
533 cpu_set(cpu, cpu_possible_map);
536 void __init __weak smp_setup_processor_id(void)
540 void __init __weak thread_info_cache_init(void)
544 void __init __weak arch_early_irq_init(void)
548 void __init __weak early_irq_init(void)
550 arch_early_irq_init();
553 asmlinkage void __init start_kernel(void)
556 extern struct kernel_param __start___param[], __stop___param[];
558 smp_setup_processor_id();
561 * Need to run as early as possible, to initialize the
566 debug_objects_early_init();
570 early_boot_irqs_off();
571 early_init_irq_lock_class();
574 * Interrupts are still disabled. Do necessary setups, then
582 printk(linux_banner);
583 setup_arch(&command_line);
584 mm_init_owner(&init_mm, &init_task);
585 setup_command_line(command_line);
587 setup_per_cpu_areas();
589 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
592 * Set up the scheduler prior starting any interrupts (such as the
593 * timer interrupt). Full topology setup happens at smp_init()
594 * time - but meanwhile we still have a functioning scheduler.
598 * Disable preemption - early bootup scheduling is extremely
599 * fragile until we cpu_idle() for the first time.
602 build_all_zonelists();
604 printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
606 parse_args("Booting kernel", static_command_line, __start___param,
607 __stop___param - __start___param,
608 &unknown_bootoption);
609 if (!irqs_disabled()) {
610 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
611 "enabled *very* early, fixing it\n");
617 /* init some links before init_ISA_irqs() */
628 if (!irqs_disabled())
629 printk("start_kernel(): bug: interrupts were enabled early\n");
630 early_boot_irqs_on();
634 * HACK ALERT! This is early. We're enabling the console before
635 * we've done PCI setups etc, and console_init() must be aware of
636 * this. But we do want output early, in case something goes wrong.
640 panic(panic_later, panic_param);
645 * Need to run this when irqs are enabled, because it wants
646 * to self-test [hard/soft]-irqs on/off lock inversion bugs
651 #ifdef CONFIG_BLK_DEV_INITRD
652 if (initrd_start && !initrd_below_start_ok &&
653 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
654 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
656 page_to_pfn(virt_to_page((void *)initrd_start)),
662 vfs_caches_init_early();
666 enable_debug_pagealloc();
670 debug_objects_mem_init();
672 setup_per_cpu_pageset();
678 pgtable_cache_init();
683 efi_enter_virtual_mode();
685 thread_info_cache_init();
687 fork_init(num_physpages);
692 vfs_caches_init(num_physpages);
695 /* rootfs populating might need page-writeback */
696 page_writeback_init();
697 #ifdef CONFIG_PROC_FS
702 taskstats_init_early();
707 acpi_early_init(); /* before LAPIC and SMP init */
711 /* Do the rest non-__init'ed, we're now alive */
715 static int initcall_debug;
716 core_param(initcall_debug, initcall_debug, bool, 0644);
718 int do_one_initcall(initcall_t fn)
720 int count = preempt_count();
721 ktime_t calltime, delta, rettime;
723 struct boot_trace_call call;
724 struct boot_trace_ret ret;
726 if (initcall_debug) {
727 call.caller = task_pid_nr(current);
728 printk("calling %pF @ %i\n", fn, call.caller);
729 calltime = ktime_get();
730 trace_boot_call(&call, fn);
736 if (initcall_debug) {
737 disable_boot_trace();
738 rettime = ktime_get();
739 delta = ktime_sub(rettime, calltime);
740 ret.duration = (unsigned long long) ktime_to_ns(delta) >> 10;
741 trace_boot_ret(&ret, fn);
742 printk("initcall %pF returned %d after %Ld usecs\n", fn,
743 ret.result, ret.duration);
748 if (ret.result && ret.result != -ENODEV && initcall_debug)
749 sprintf(msgbuf, "error code %d ", ret.result);
751 if (preempt_count() != count) {
752 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
753 preempt_count() = count;
755 if (irqs_disabled()) {
756 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
760 printk("initcall %pF returned with %s\n", fn, msgbuf);
767 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
769 static void __init do_initcalls(void)
773 for (call = __early_initcall_end; call < __initcall_end; call++)
774 do_one_initcall(*call);
776 /* Make sure there is no pending stuff from the initcall sequence */
777 flush_scheduled_work();
781 * Ok, the machine is now initialized. None of the devices
782 * have been touched yet, but the CPU subsystem is up and
783 * running, and memory and process management works.
785 * Now we can finally start doing some real work..
787 static void __init do_basic_setup(void)
789 rcu_init_sched(); /* needed by module_init stage. */
791 usermodehelper_init();
797 static void __init do_pre_smp_initcalls(void)
801 for (call = __initcall_start; call < __early_initcall_end; call++)
802 do_one_initcall(*call);
805 static void run_init_process(char *init_filename)
807 argv_init[0] = init_filename;
808 kernel_execve(init_filename, argv_init, envp_init);
811 /* This is a non __init function. Force it to be noinline otherwise gcc
812 * makes it inline to init() and it becomes part of init.text section
814 static int noinline init_post(void)
819 system_state = SYSTEM_RUNNING;
820 numa_default_policy();
822 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
823 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
828 current->signal->flags |= SIGNAL_UNKILLABLE;
830 if (ramdisk_execute_command) {
831 run_init_process(ramdisk_execute_command);
832 printk(KERN_WARNING "Failed to execute %s\n",
833 ramdisk_execute_command);
837 * We try each of these until one succeeds.
839 * The Bourne shell can be used instead of init if we are
840 * trying to recover a really broken machine.
842 if (execute_command) {
843 run_init_process(execute_command);
844 printk(KERN_WARNING "Failed to execute %s. Attempting "
845 "defaults...\n", execute_command);
847 run_init_process("/sbin/init");
848 run_init_process("/etc/init");
849 run_init_process("/bin/init");
850 run_init_process("/bin/sh");
852 panic("No init found. Try passing init= option to kernel.");
855 static int __init kernel_init(void * unused)
859 * init can run on any cpu.
861 set_cpus_allowed_ptr(current, CPU_MASK_ALL_PTR);
863 * Tell the world that we're going to be the grim
864 * reaper of innocent orphaned children.
866 * We don't want people to have to make incorrect
867 * assumptions about where in the task array this
870 init_pid_ns.child_reaper = current;
872 cad_pid = task_pid(current);
874 smp_prepare_cpus(setup_max_cpus);
876 do_pre_smp_initcalls();
887 * check if there is an early userspace init. If yes, let it do all
891 if (!ramdisk_execute_command)
892 ramdisk_execute_command = "/init";
894 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
895 ramdisk_execute_command = NULL;
900 * Ok, we have completed the initial bootup, and
901 * we're essentially up and running. Get rid of the
902 * initmem segments and start the user-mode stuff..