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>
74 #ifdef CONFIG_X86_LOCAL_APIC
78 static int kernel_init(void *);
80 extern void init_IRQ(void);
81 extern void fork_init(unsigned long);
82 extern void mca_init(void);
83 extern void sbus_init(void);
84 extern void prio_tree_init(void);
85 extern void radix_tree_init(void);
86 extern void free_initmem(void);
88 extern void acpi_early_init(void);
90 static inline void acpi_early_init(void) { }
92 #ifndef CONFIG_DEBUG_RODATA
93 static inline void mark_rodata_ro(void) { }
97 extern void tc_init(void);
100 enum system_states system_state;
101 EXPORT_SYMBOL(system_state);
104 * Boot command-line arguments
106 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
107 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
109 extern void time_init(void);
110 /* Default late time init is NULL. archs can override this later. */
111 void (*late_time_init)(void);
112 extern void softirq_init(void);
114 /* Untouched command line saved by arch-specific code. */
115 char __initdata boot_command_line[COMMAND_LINE_SIZE];
116 /* Untouched saved command line (eg. for /proc) */
117 char *saved_command_line;
118 /* Command line for parameter parsing */
119 static char *static_command_line;
121 static char *execute_command;
122 static char *ramdisk_execute_command;
125 /* Setup configured maximum number of CPUs to activate */
126 unsigned int __initdata setup_max_cpus = NR_CPUS;
129 * Setup routine for controlling SMP activation
131 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
132 * activation entirely (the MPS table probe still happens, though).
134 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
135 * greater than 0, limits the maximum number of CPUs activated in
138 #ifndef CONFIG_X86_IO_APIC
139 static inline void disable_ioapic_setup(void) {};
142 static int __init nosmp(char *str)
145 disable_ioapic_setup();
149 early_param("nosmp", nosmp);
151 static int __init maxcpus(char *str)
153 get_option(&str, &setup_max_cpus);
154 if (setup_max_cpus == 0)
155 disable_ioapic_setup();
160 early_param("maxcpus", maxcpus);
162 #define setup_max_cpus NR_CPUS
166 * If set, this is an indication to the drivers that reset the underlying
167 * device before going ahead with the initialization otherwise driver might
168 * rely on the BIOS and skip the reset operation.
170 * This is useful if kernel is booting in an unreliable environment.
171 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
172 * skipped and devices will be in unknown state.
174 unsigned int reset_devices;
175 EXPORT_SYMBOL(reset_devices);
177 static int __init set_reset_devices(char *str)
183 __setup("reset_devices", set_reset_devices);
185 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
186 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
187 static const char *panic_later, *panic_param;
189 extern struct obs_kernel_param __setup_start[], __setup_end[];
191 static int __init obsolete_checksetup(char *line)
193 struct obs_kernel_param *p;
194 int had_early_param = 0;
198 int n = strlen(p->str);
199 if (!strncmp(line, p->str, n)) {
201 /* Already done in parse_early_param?
202 * (Needs exact match on param part).
203 * Keep iterating, as we can have early
204 * params and __setups of same names 8( */
205 if (line[n] == '\0' || line[n] == '=')
207 } else if (!p->setup_func) {
208 printk(KERN_WARNING "Parameter %s is obsolete,"
209 " ignored\n", p->str);
211 } else if (p->setup_func(line + n))
215 } while (p < __setup_end);
217 return had_early_param;
221 * This should be approx 2 Bo*oMips to start (note initial shift), and will
222 * still work even if initially too large, it will just take slightly longer
224 unsigned long loops_per_jiffy = (1<<12);
226 EXPORT_SYMBOL(loops_per_jiffy);
228 static int __init debug_kernel(char *str)
230 console_loglevel = 10;
234 static int __init quiet_kernel(char *str)
236 console_loglevel = 4;
240 early_param("debug", debug_kernel);
241 early_param("quiet", quiet_kernel);
243 static int __init loglevel(char *str)
245 get_option(&str, &console_loglevel);
249 early_param("loglevel", loglevel);
252 * Unknown boot options get handed to init, unless they look like
255 static int __init unknown_bootoption(char *param, char *val)
257 /* Change NUL term back to "=", to make "param" the whole string. */
259 /* param=val or param="val"? */
260 if (val == param+strlen(param)+1)
262 else if (val == param+strlen(param)+2) {
264 memmove(val-1, val, strlen(val)+1);
270 /* Handle obsolete-style parameters */
271 if (obsolete_checksetup(param))
275 * Preemptive maintenance for "why didn't my misspelled command
278 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
279 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
287 /* Environment option */
289 for (i = 0; envp_init[i]; i++) {
290 if (i == MAX_INIT_ENVS) {
291 panic_later = "Too many boot env vars at `%s'";
294 if (!strncmp(param, envp_init[i], val - param))
297 envp_init[i] = param;
299 /* Command line option */
301 for (i = 0; argv_init[i]; i++) {
302 if (i == MAX_INIT_ARGS) {
303 panic_later = "Too many boot init vars at `%s'";
307 argv_init[i] = param;
312 #ifdef CONFIG_DEBUG_PAGEALLOC
313 int __read_mostly debug_pagealloc_enabled = 0;
316 static int __init init_setup(char *str)
320 execute_command = str;
322 * In case LILO is going to boot us with default command line,
323 * it prepends "auto" before the whole cmdline which makes
324 * the shell think it should execute a script with such name.
325 * So we ignore all arguments entered _before_ init=... [MJ]
327 for (i = 1; i < MAX_INIT_ARGS; i++)
331 __setup("init=", init_setup);
333 static int __init rdinit_setup(char *str)
337 ramdisk_execute_command = str;
338 /* See "auto" comment in init_setup */
339 for (i = 1; i < MAX_INIT_ARGS; i++)
343 __setup("rdinit=", rdinit_setup);
347 #ifdef CONFIG_X86_LOCAL_APIC
348 static void __init smp_init(void)
350 APIC_init_uniprocessor();
353 #define smp_init() do { } while (0)
356 static inline void setup_per_cpu_areas(void) { }
357 static inline void setup_nr_cpu_ids(void) { }
358 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
362 #if NR_CPUS > BITS_PER_LONG
363 cpumask_t cpu_mask_all __read_mostly = CPU_MASK_ALL;
364 EXPORT_SYMBOL(cpu_mask_all);
367 /* Setup number of possible processor ids */
368 int nr_cpu_ids __read_mostly = NR_CPUS;
369 EXPORT_SYMBOL(nr_cpu_ids);
371 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
372 static void __init setup_nr_cpu_ids(void)
374 int cpu, highest_cpu = 0;
376 for_each_possible_cpu(cpu)
379 nr_cpu_ids = highest_cpu + 1;
382 #ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA
383 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
385 EXPORT_SYMBOL(__per_cpu_offset);
387 static void __init setup_per_cpu_areas(void)
389 unsigned long size, i;
391 unsigned long nr_possible_cpus = num_possible_cpus();
393 /* Copy section for each CPU (we discard the original) */
394 size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
395 ptr = alloc_bootmem_pages(size * nr_possible_cpus);
397 for_each_possible_cpu(i) {
398 __per_cpu_offset[i] = ptr - __per_cpu_start;
399 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
403 #endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */
405 /* Called by boot processor to activate the rest. */
406 static void __init smp_init(void)
411 * Set up the current CPU as possible to migrate to.
412 * The other ones will be done by cpu_up/cpu_down()
414 cpu = smp_processor_id();
415 cpu_set(cpu, cpu_active_map);
417 /* FIXME: This should be done in userspace --RR */
418 for_each_present_cpu(cpu) {
419 if (num_online_cpus() >= setup_max_cpus)
421 if (!cpu_online(cpu))
425 /* Any cleanup work */
426 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
427 smp_cpus_done(setup_max_cpus);
433 * We need to store the untouched command line for future reference.
434 * We also need to store the touched command line since the parameter
435 * parsing is performed in place, and we should allow a component to
436 * store reference of name/value for future reference.
438 static void __init setup_command_line(char *command_line)
440 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
441 static_command_line = alloc_bootmem(strlen (command_line)+1);
442 strcpy (saved_command_line, boot_command_line);
443 strcpy (static_command_line, command_line);
447 * We need to finalize in a non-__init function or else race conditions
448 * between the root thread and the init thread may cause start_kernel to
449 * be reaped by free_initmem before the root thread has proceeded to
452 * gcc-3.4 accidentally inlines this function, so use noinline.
455 static void noinline __init_refok rest_init(void)
456 __releases(kernel_lock)
460 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
461 numa_default_policy();
462 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
463 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
467 * The boot idle thread must execute schedule()
468 * at least once to get things moving:
470 init_idle_bootup_task(current);
471 preempt_enable_no_resched();
475 /* Call into cpu_idle with preempt disabled */
479 /* Check for early params. */
480 static int __init do_early_param(char *param, char *val)
482 struct obs_kernel_param *p;
484 for (p = __setup_start; p < __setup_end; p++) {
485 if ((p->early && strcmp(param, p->str) == 0) ||
486 (strcmp(param, "console") == 0 &&
487 strcmp(p->str, "earlycon") == 0)
489 if (p->setup_func(val) != 0)
491 "Malformed early option '%s'\n", param);
494 /* We accept everything at this stage. */
498 /* Arch code calls this early on, or if not, just before other parsing. */
499 void __init parse_early_param(void)
501 static __initdata int done = 0;
502 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
507 /* All fall through to do_early_param. */
508 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
509 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
514 * Activate the first processor.
517 static void __init boot_cpu_init(void)
519 int cpu = smp_processor_id();
520 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
521 cpu_set(cpu, cpu_online_map);
522 cpu_set(cpu, cpu_present_map);
523 cpu_set(cpu, cpu_possible_map);
526 void __init __weak smp_setup_processor_id(void)
530 void __init __weak thread_info_cache_init(void)
534 asmlinkage void __init start_kernel(void)
537 extern struct kernel_param __start___param[], __stop___param[];
539 smp_setup_processor_id();
542 * Need to run as early as possible, to initialize the
547 debug_objects_early_init();
551 early_boot_irqs_off();
552 early_init_irq_lock_class();
555 * Interrupts are still disabled. Do necessary setups, then
563 printk(linux_banner);
564 setup_arch(&command_line);
565 mm_init_owner(&init_mm, &init_task);
566 setup_command_line(command_line);
568 setup_per_cpu_areas();
570 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
573 * Set up the scheduler prior starting any interrupts (such as the
574 * timer interrupt). Full topology setup happens at smp_init()
575 * time - but meanwhile we still have a functioning scheduler.
579 * Disable preemption - early bootup scheduling is extremely
580 * fragile until we cpu_idle() for the first time.
583 build_all_zonelists();
585 printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
587 parse_args("Booting kernel", static_command_line, __start___param,
588 __stop___param - __start___param,
589 &unknown_bootoption);
590 if (!irqs_disabled()) {
591 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
592 "enabled *very* early, fixing it\n");
598 /* init some links before init_ISA_irqs() */
609 if (!irqs_disabled())
610 printk("start_kernel(): bug: interrupts were enabled early\n");
611 early_boot_irqs_on();
615 * HACK ALERT! This is early. We're enabling the console before
616 * we've done PCI setups etc, and console_init() must be aware of
617 * this. But we do want output early, in case something goes wrong.
621 panic(panic_later, panic_param);
626 * Need to run this when irqs are enabled, because it wants
627 * to self-test [hard/soft]-irqs on/off lock inversion bugs
632 #ifdef CONFIG_BLK_DEV_INITRD
633 if (initrd_start && !initrd_below_start_ok &&
634 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
635 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
637 page_to_pfn(virt_to_page((void *)initrd_start)),
643 vfs_caches_init_early();
647 enable_debug_pagealloc();
650 debug_objects_mem_init();
652 setup_per_cpu_pageset();
658 pgtable_cache_init();
663 efi_enter_virtual_mode();
665 thread_info_cache_init();
667 fork_init(num_physpages);
672 vfs_caches_init(num_physpages);
675 /* rootfs populating might need page-writeback */
676 page_writeback_init();
677 #ifdef CONFIG_PROC_FS
682 taskstats_init_early();
687 acpi_early_init(); /* before LAPIC and SMP init */
691 /* Do the rest non-__init'ed, we're now alive */
695 static int initcall_debug;
696 core_param(initcall_debug, initcall_debug, bool, 0644);
698 int do_one_initcall(initcall_t fn)
700 int count = preempt_count();
701 ktime_t calltime, delta, rettime;
703 struct boot_trace_call call;
704 struct boot_trace_ret ret;
706 if (initcall_debug) {
707 call.caller = task_pid_nr(current);
708 printk("calling %pF @ %i\n", fn, call.caller);
709 calltime = ktime_get();
710 trace_boot_call(&call, fn);
716 if (initcall_debug) {
717 disable_boot_trace();
718 rettime = ktime_get();
719 delta = ktime_sub(rettime, calltime);
720 ret.duration = (unsigned long long) ktime_to_ns(delta) >> 10;
721 trace_boot_ret(&ret, fn);
722 printk("initcall %pF returned %d after %Ld usecs\n", fn,
723 ret.result, ret.duration);
728 if (ret.result && ret.result != -ENODEV && initcall_debug)
729 sprintf(msgbuf, "error code %d ", ret.result);
731 if (preempt_count() != count) {
732 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
733 preempt_count() = count;
735 if (irqs_disabled()) {
736 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
740 printk("initcall %pF returned with %s\n", fn, msgbuf);
747 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
749 static void __init do_initcalls(void)
753 for (call = __early_initcall_end; call < __initcall_end; call++)
754 do_one_initcall(*call);
756 /* Make sure there is no pending stuff from the initcall sequence */
757 flush_scheduled_work();
761 * Ok, the machine is now initialized. None of the devices
762 * have been touched yet, but the CPU subsystem is up and
763 * running, and memory and process management works.
765 * Now we can finally start doing some real work..
767 static void __init do_basic_setup(void)
769 rcu_init_sched(); /* needed by module_init stage. */
771 usermodehelper_init();
777 static void __init do_pre_smp_initcalls(void)
781 for (call = __initcall_start; call < __early_initcall_end; call++)
782 do_one_initcall(*call);
785 static void run_init_process(char *init_filename)
787 argv_init[0] = init_filename;
788 kernel_execve(init_filename, argv_init, envp_init);
791 /* This is a non __init function. Force it to be noinline otherwise gcc
792 * makes it inline to init() and it becomes part of init.text section
794 static int noinline init_post(void)
799 system_state = SYSTEM_RUNNING;
800 numa_default_policy();
802 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
803 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
808 current->signal->flags |= SIGNAL_UNKILLABLE;
810 if (ramdisk_execute_command) {
811 run_init_process(ramdisk_execute_command);
812 printk(KERN_WARNING "Failed to execute %s\n",
813 ramdisk_execute_command);
817 * We try each of these until one succeeds.
819 * The Bourne shell can be used instead of init if we are
820 * trying to recover a really broken machine.
822 if (execute_command) {
823 run_init_process(execute_command);
824 printk(KERN_WARNING "Failed to execute %s. Attempting "
825 "defaults...\n", execute_command);
827 run_init_process("/sbin/init");
828 run_init_process("/etc/init");
829 run_init_process("/bin/init");
830 run_init_process("/bin/sh");
832 panic("No init found. Try passing init= option to kernel.");
835 static int __init kernel_init(void * unused)
839 * init can run on any cpu.
841 set_cpus_allowed_ptr(current, CPU_MASK_ALL_PTR);
843 * Tell the world that we're going to be the grim
844 * reaper of innocent orphaned children.
846 * We don't want people to have to make incorrect
847 * assumptions about where in the task array this
850 init_pid_ns.child_reaper = current;
852 cad_pid = task_pid(current);
854 smp_prepare_cpus(setup_max_cpus);
856 do_pre_smp_initcalls();
867 * check if there is an early userspace init. If yes, let it do all
871 if (!ramdisk_execute_command)
872 ramdisk_execute_command = "/init";
874 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
875 ramdisk_execute_command = NULL;
880 * Ok, we have completed the initial bootup, and
881 * we're essentially up and running. Get rid of the
882 * initmem segments and start the user-mode stuff..