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/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/pid_namespace.h>
59 #include <linux/device.h>
60 #include <linux/kthread.h>
61 #include <linux/sched.h>
62 #include <linux/signal.h>
63 #include <linux/idr.h>
64 #include <linux/ftrace.h>
65 #include <trace/boot.h>
69 #include <asm/setup.h>
70 #include <asm/sections.h>
71 #include <asm/cacheflush.h>
73 #ifdef CONFIG_X86_LOCAL_APIC
77 static int kernel_init(void *);
79 extern void init_IRQ(void);
80 extern void fork_init(unsigned long);
81 extern void mca_init(void);
82 extern void sbus_init(void);
83 extern void prio_tree_init(void);
84 extern void radix_tree_init(void);
85 extern void free_initmem(void);
87 extern void acpi_early_init(void);
89 static inline void acpi_early_init(void) { }
91 #ifndef CONFIG_DEBUG_RODATA
92 static inline void mark_rodata_ro(void) { }
96 extern void tc_init(void);
99 enum system_states system_state;
100 EXPORT_SYMBOL(system_state);
103 * Boot command-line arguments
105 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
106 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
108 extern void time_init(void);
109 /* Default late time init is NULL. archs can override this later. */
110 void (*__initdata late_time_init)(void);
111 extern void softirq_init(void);
113 /* Untouched command line saved by arch-specific code. */
114 char __initdata boot_command_line[COMMAND_LINE_SIZE];
115 /* Untouched saved command line (eg. for /proc) */
116 char *saved_command_line;
117 /* Command line for parameter parsing */
118 static char *static_command_line;
120 static char *execute_command;
121 static char *ramdisk_execute_command;
124 /* Setup configured maximum number of CPUs to activate */
125 unsigned int __initdata setup_max_cpus = NR_CPUS;
128 * Setup routine for controlling SMP activation
130 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
131 * activation entirely (the MPS table probe still happens, though).
133 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
134 * greater than 0, limits the maximum number of CPUs activated in
137 #ifndef CONFIG_X86_IO_APIC
138 static inline void disable_ioapic_setup(void) {};
141 static int __init nosmp(char *str)
144 disable_ioapic_setup();
148 early_param("nosmp", nosmp);
150 static int __init maxcpus(char *str)
152 get_option(&str, &setup_max_cpus);
153 if (setup_max_cpus == 0)
154 disable_ioapic_setup();
159 early_param("maxcpus", maxcpus);
161 #define setup_max_cpus NR_CPUS
165 * If set, this is an indication to the drivers that reset the underlying
166 * device before going ahead with the initialization otherwise driver might
167 * rely on the BIOS and skip the reset operation.
169 * This is useful if kernel is booting in an unreliable environment.
170 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
171 * skipped and devices will be in unknown state.
173 unsigned int reset_devices;
174 EXPORT_SYMBOL(reset_devices);
176 static int __init set_reset_devices(char *str)
182 __setup("reset_devices", set_reset_devices);
184 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
185 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
186 static const char *panic_later, *panic_param;
188 extern struct obs_kernel_param __setup_start[], __setup_end[];
190 static int __init obsolete_checksetup(char *line)
192 struct obs_kernel_param *p;
193 int had_early_param = 0;
197 int n = strlen(p->str);
198 if (!strncmp(line, p->str, n)) {
200 /* Already done in parse_early_param?
201 * (Needs exact match on param part).
202 * Keep iterating, as we can have early
203 * params and __setups of same names 8( */
204 if (line[n] == '\0' || line[n] == '=')
206 } else if (!p->setup_func) {
207 printk(KERN_WARNING "Parameter %s is obsolete,"
208 " ignored\n", p->str);
210 } else if (p->setup_func(line + n))
214 } while (p < __setup_end);
216 return had_early_param;
220 * This should be approx 2 Bo*oMips to start (note initial shift), and will
221 * still work even if initially too large, it will just take slightly longer
223 unsigned long loops_per_jiffy = (1<<12);
225 EXPORT_SYMBOL(loops_per_jiffy);
227 static int __init debug_kernel(char *str)
229 console_loglevel = 10;
233 static int __init quiet_kernel(char *str)
235 console_loglevel = 4;
239 early_param("debug", debug_kernel);
240 early_param("quiet", quiet_kernel);
242 static int __init loglevel(char *str)
244 get_option(&str, &console_loglevel);
248 early_param("loglevel", loglevel);
251 * Unknown boot options get handed to init, unless they look like
254 static int __init unknown_bootoption(char *param, char *val)
256 /* Change NUL term back to "=", to make "param" the whole string. */
258 /* param=val or param="val"? */
259 if (val == param+strlen(param)+1)
261 else if (val == param+strlen(param)+2) {
263 memmove(val-1, val, strlen(val)+1);
269 /* Handle obsolete-style parameters */
270 if (obsolete_checksetup(param))
274 * Preemptive maintenance for "why didn't my misspelled command
277 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
278 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
286 /* Environment option */
288 for (i = 0; envp_init[i]; i++) {
289 if (i == MAX_INIT_ENVS) {
290 panic_later = "Too many boot env vars at `%s'";
293 if (!strncmp(param, envp_init[i], val - param))
296 envp_init[i] = param;
298 /* Command line option */
300 for (i = 0; argv_init[i]; i++) {
301 if (i == MAX_INIT_ARGS) {
302 panic_later = "Too many boot init vars at `%s'";
306 argv_init[i] = param;
311 #ifdef CONFIG_DEBUG_PAGEALLOC
312 int __read_mostly debug_pagealloc_enabled = 0;
315 static int __init init_setup(char *str)
319 execute_command = str;
321 * In case LILO is going to boot us with default command line,
322 * it prepends "auto" before the whole cmdline which makes
323 * the shell think it should execute a script with such name.
324 * So we ignore all arguments entered _before_ init=... [MJ]
326 for (i = 1; i < MAX_INIT_ARGS; i++)
330 __setup("init=", init_setup);
332 static int __init rdinit_setup(char *str)
336 ramdisk_execute_command = str;
337 /* See "auto" comment in init_setup */
338 for (i = 1; i < MAX_INIT_ARGS; i++)
342 __setup("rdinit=", rdinit_setup);
346 #ifdef CONFIG_X86_LOCAL_APIC
347 static void __init smp_init(void)
349 APIC_init_uniprocessor();
352 #define smp_init() do { } while (0)
355 static inline void setup_per_cpu_areas(void) { }
356 static inline void setup_nr_cpu_ids(void) { }
357 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
361 #if NR_CPUS > BITS_PER_LONG
362 cpumask_t cpu_mask_all __read_mostly = CPU_MASK_ALL;
363 EXPORT_SYMBOL(cpu_mask_all);
366 /* Setup number of possible processor ids */
367 int nr_cpu_ids __read_mostly = NR_CPUS;
368 EXPORT_SYMBOL(nr_cpu_ids);
370 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
371 static void __init setup_nr_cpu_ids(void)
373 nr_cpu_ids = find_last_bit(cpumask_bits(cpu_possible_mask),NR_CPUS) + 1;
376 #ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA
377 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
379 EXPORT_SYMBOL(__per_cpu_offset);
381 static void __init setup_per_cpu_areas(void)
383 unsigned long size, i;
385 unsigned long nr_possible_cpus = num_possible_cpus();
387 /* Copy section for each CPU (we discard the original) */
388 size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
389 ptr = alloc_bootmem_pages(size * nr_possible_cpus);
391 for_each_possible_cpu(i) {
392 __per_cpu_offset[i] = ptr - __per_cpu_start;
393 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
397 #endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */
399 /* Called by boot processor to activate the rest. */
400 static void __init smp_init(void)
405 * Set up the current CPU as possible to migrate to.
406 * The other ones will be done by cpu_up/cpu_down()
408 cpu = smp_processor_id();
409 cpu_set(cpu, cpu_active_map);
411 /* FIXME: This should be done in userspace --RR */
412 for_each_present_cpu(cpu) {
413 if (num_online_cpus() >= setup_max_cpus)
415 if (!cpu_online(cpu))
419 /* Any cleanup work */
420 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
421 smp_cpus_done(setup_max_cpus);
427 * We need to store the untouched command line for future reference.
428 * We also need to store the touched command line since the parameter
429 * parsing is performed in place, and we should allow a component to
430 * store reference of name/value for future reference.
432 static void __init setup_command_line(char *command_line)
434 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
435 static_command_line = alloc_bootmem(strlen (command_line)+1);
436 strcpy (saved_command_line, boot_command_line);
437 strcpy (static_command_line, command_line);
441 * We need to finalize in a non-__init function or else race conditions
442 * between the root thread and the init thread may cause start_kernel to
443 * be reaped by free_initmem before the root thread has proceeded to
446 * gcc-3.4 accidentally inlines this function, so use noinline.
449 static noinline void __init_refok rest_init(void)
450 __releases(kernel_lock)
454 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
455 numa_default_policy();
456 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
457 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
461 * The boot idle thread must execute schedule()
462 * at least once to get things moving:
464 init_idle_bootup_task(current);
465 preempt_enable_no_resched();
469 /* Call into cpu_idle with preempt disabled */
473 /* Check for early params. */
474 static int __init do_early_param(char *param, char *val)
476 struct obs_kernel_param *p;
478 for (p = __setup_start; p < __setup_end; p++) {
479 if ((p->early && strcmp(param, p->str) == 0) ||
480 (strcmp(param, "console") == 0 &&
481 strcmp(p->str, "earlycon") == 0)
483 if (p->setup_func(val) != 0)
485 "Malformed early option '%s'\n", param);
488 /* We accept everything at this stage. */
492 /* Arch code calls this early on, or if not, just before other parsing. */
493 void __init parse_early_param(void)
495 static __initdata int done = 0;
496 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
501 /* All fall through to do_early_param. */
502 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
503 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
508 * Activate the first processor.
511 static void __init boot_cpu_init(void)
513 int cpu = smp_processor_id();
514 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
515 set_cpu_online(cpu, true);
516 set_cpu_present(cpu, true);
517 set_cpu_possible(cpu, true);
520 void __init __weak smp_setup_processor_id(void)
524 void __init __weak thread_info_cache_init(void)
528 asmlinkage void __init start_kernel(void)
531 extern struct kernel_param __start___param[], __stop___param[];
533 smp_setup_processor_id();
536 * Need to run as early as possible, to initialize the
540 debug_objects_early_init();
544 early_boot_irqs_off();
545 early_init_irq_lock_class();
548 * Interrupts are still disabled. Do necessary setups, then
556 printk(linux_banner);
557 setup_arch(&command_line);
558 mm_init_owner(&init_mm, &init_task);
559 setup_command_line(command_line);
560 setup_per_cpu_areas();
562 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
565 * Set up the scheduler prior starting any interrupts (such as the
566 * timer interrupt). Full topology setup happens at smp_init()
567 * time - but meanwhile we still have a functioning scheduler.
571 * Disable preemption - early bootup scheduling is extremely
572 * fragile until we cpu_idle() for the first time.
575 build_all_zonelists();
577 printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
579 parse_args("Booting kernel", static_command_line, __start___param,
580 __stop___param - __start___param,
581 &unknown_bootoption);
582 if (!irqs_disabled()) {
583 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
584 "enabled *very* early, fixing it\n");
590 /* init some links before init_ISA_irqs() */
601 if (!irqs_disabled())
602 printk("start_kernel(): bug: interrupts were enabled early\n");
603 early_boot_irqs_on();
607 * HACK ALERT! This is early. We're enabling the console before
608 * we've done PCI setups etc, and console_init() must be aware of
609 * this. But we do want output early, in case something goes wrong.
613 panic(panic_later, panic_param);
618 * Need to run this when irqs are enabled, because it wants
619 * to self-test [hard/soft]-irqs on/off lock inversion bugs
624 #ifdef CONFIG_BLK_DEV_INITRD
625 if (initrd_start && !initrd_below_start_ok &&
626 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
627 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
629 page_to_pfn(virt_to_page((void *)initrd_start)),
635 vfs_caches_init_early();
639 enable_debug_pagealloc();
642 debug_objects_mem_init();
644 setup_per_cpu_pageset();
650 pgtable_cache_init();
655 efi_enter_virtual_mode();
657 thread_info_cache_init();
659 fork_init(num_physpages);
664 vfs_caches_init(num_physpages);
667 /* rootfs populating might need page-writeback */
668 page_writeback_init();
669 #ifdef CONFIG_PROC_FS
674 taskstats_init_early();
679 acpi_early_init(); /* before LAPIC and SMP init */
683 /* Do the rest non-__init'ed, we're now alive */
687 static int initcall_debug;
688 core_param(initcall_debug, initcall_debug, bool, 0644);
690 int do_one_initcall(initcall_t fn)
692 int count = preempt_count();
693 ktime_t calltime, delta, rettime;
695 struct boot_trace_call call;
696 struct boot_trace_ret ret;
698 if (initcall_debug) {
699 call.caller = task_pid_nr(current);
700 printk("calling %pF @ %i\n", fn, call.caller);
701 calltime = ktime_get();
702 trace_boot_call(&call, fn);
708 if (initcall_debug) {
709 disable_boot_trace();
710 rettime = ktime_get();
711 delta = ktime_sub(rettime, calltime);
712 ret.duration = (unsigned long long) ktime_to_ns(delta) >> 10;
713 trace_boot_ret(&ret, fn);
714 printk("initcall %pF returned %d after %Ld usecs\n", fn,
715 ret.result, ret.duration);
720 if (ret.result && ret.result != -ENODEV && initcall_debug)
721 sprintf(msgbuf, "error code %d ", ret.result);
723 if (preempt_count() != count) {
724 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
725 preempt_count() = count;
727 if (irqs_disabled()) {
728 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
732 printk("initcall %pF returned with %s\n", fn, msgbuf);
739 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
741 static void __init do_initcalls(void)
745 for (call = __early_initcall_end; call < __initcall_end; call++)
746 do_one_initcall(*call);
748 /* Make sure there is no pending stuff from the initcall sequence */
749 flush_scheduled_work();
753 * Ok, the machine is now initialized. None of the devices
754 * have been touched yet, but the CPU subsystem is up and
755 * running, and memory and process management works.
757 * Now we can finally start doing some real work..
759 static void __init do_basic_setup(void)
761 rcu_init_sched(); /* needed by module_init stage. */
763 usermodehelper_init();
769 static void __init do_pre_smp_initcalls(void)
773 for (call = __initcall_start; call < __early_initcall_end; call++)
774 do_one_initcall(*call);
777 static void run_init_process(char *init_filename)
779 argv_init[0] = init_filename;
780 kernel_execve(init_filename, argv_init, envp_init);
783 /* This is a non __init function. Force it to be noinline otherwise gcc
784 * makes it inline to init() and it becomes part of init.text section
786 static noinline int init_post(void)
791 system_state = SYSTEM_RUNNING;
792 numa_default_policy();
794 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
795 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
800 current->signal->flags |= SIGNAL_UNKILLABLE;
802 if (ramdisk_execute_command) {
803 run_init_process(ramdisk_execute_command);
804 printk(KERN_WARNING "Failed to execute %s\n",
805 ramdisk_execute_command);
809 * We try each of these until one succeeds.
811 * The Bourne shell can be used instead of init if we are
812 * trying to recover a really broken machine.
814 if (execute_command) {
815 run_init_process(execute_command);
816 printk(KERN_WARNING "Failed to execute %s. Attempting "
817 "defaults...\n", execute_command);
819 run_init_process("/sbin/init");
820 run_init_process("/etc/init");
821 run_init_process("/bin/init");
822 run_init_process("/bin/sh");
824 panic("No init found. Try passing init= option to kernel.");
827 static int __init kernel_init(void * unused)
831 * init can run on any cpu.
833 set_cpus_allowed_ptr(current, CPU_MASK_ALL_PTR);
835 * Tell the world that we're going to be the grim
836 * reaper of innocent orphaned children.
838 * We don't want people to have to make incorrect
839 * assumptions about where in the task array this
842 init_pid_ns.child_reaper = current;
844 cad_pid = task_pid(current);
846 smp_prepare_cpus(setup_max_cpus);
848 do_pre_smp_initcalls();
859 * check if there is an early userspace init. If yes, let it do all
863 if (!ramdisk_execute_command)
864 ramdisk_execute_command = "/init";
866 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
867 ramdisk_execute_command = NULL;
872 * Ok, we have completed the initial bootup, and
873 * we're essentially up and running. Get rid of the
874 * initmem segments and start the user-mode stuff..