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/utsname.h>
22 #include <linux/ioport.h>
23 #include <linux/init.h>
24 #include <linux/smp_lock.h>
25 #include <linux/initrd.h>
26 #include <linux/bootmem.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 <trace/boot.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 extern void free_initmem(void);
92 extern void acpi_early_init(void);
94 static inline void acpi_early_init(void) { }
96 #ifndef CONFIG_DEBUG_RODATA
97 static inline void mark_rodata_ro(void) { }
101 extern void tc_init(void);
104 enum system_states system_state __read_mostly;
105 EXPORT_SYMBOL(system_state);
108 * Boot command-line arguments
110 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
111 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
113 extern void time_init(void);
114 /* Default late time init is NULL. archs can override this later. */
115 void (*__initdata late_time_init)(void);
116 extern void softirq_init(void);
118 /* Untouched command line saved by arch-specific code. */
119 char __initdata boot_command_line[COMMAND_LINE_SIZE];
120 /* Untouched saved command line (eg. for /proc) */
121 char *saved_command_line;
122 /* Command line for parameter parsing */
123 static char *static_command_line;
125 static char *execute_command;
126 static char *ramdisk_execute_command;
129 /* Setup configured maximum number of CPUs to activate */
130 unsigned int __initdata setup_max_cpus = NR_CPUS;
133 * Setup routine for controlling SMP activation
135 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
136 * activation entirely (the MPS table probe still happens, though).
138 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
139 * greater than 0, limits the maximum number of CPUs activated in
143 void __weak arch_disable_smp_support(void) { }
145 static int __init nosmp(char *str)
148 arch_disable_smp_support();
153 early_param("nosmp", nosmp);
155 static int __init maxcpus(char *str)
157 get_option(&str, &setup_max_cpus);
158 if (setup_max_cpus == 0)
159 arch_disable_smp_support();
164 early_param("maxcpus", maxcpus);
166 const unsigned int setup_max_cpus = NR_CPUS;
170 * If set, this is an indication to the drivers that reset the underlying
171 * device before going ahead with the initialization otherwise driver might
172 * rely on the BIOS and skip the reset operation.
174 * This is useful if kernel is booting in an unreliable environment.
175 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
176 * skipped and devices will be in unknown state.
178 unsigned int reset_devices;
179 EXPORT_SYMBOL(reset_devices);
181 static int __init set_reset_devices(char *str)
187 __setup("reset_devices", set_reset_devices);
189 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
190 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
191 static const char *panic_later, *panic_param;
193 extern struct obs_kernel_param __setup_start[], __setup_end[];
195 static int __init obsolete_checksetup(char *line)
197 struct obs_kernel_param *p;
198 int had_early_param = 0;
202 int n = strlen(p->str);
203 if (!strncmp(line, p->str, n)) {
205 /* Already done in parse_early_param?
206 * (Needs exact match on param part).
207 * Keep iterating, as we can have early
208 * params and __setups of same names 8( */
209 if (line[n] == '\0' || line[n] == '=')
211 } else if (!p->setup_func) {
212 printk(KERN_WARNING "Parameter %s is obsolete,"
213 " ignored\n", p->str);
215 } else if (p->setup_func(line + n))
219 } while (p < __setup_end);
221 return had_early_param;
225 * This should be approx 2 Bo*oMips to start (note initial shift), and will
226 * still work even if initially too large, it will just take slightly longer
228 unsigned long loops_per_jiffy = (1<<12);
230 EXPORT_SYMBOL(loops_per_jiffy);
232 static int __init debug_kernel(char *str)
234 console_loglevel = 10;
238 static int __init quiet_kernel(char *str)
240 console_loglevel = 4;
244 early_param("debug", debug_kernel);
245 early_param("quiet", quiet_kernel);
247 static int __init loglevel(char *str)
249 get_option(&str, &console_loglevel);
253 early_param("loglevel", loglevel);
256 * Unknown boot options get handed to init, unless they look like
259 static int __init unknown_bootoption(char *param, char *val)
261 /* Change NUL term back to "=", to make "param" the whole string. */
263 /* param=val or param="val"? */
264 if (val == param+strlen(param)+1)
266 else if (val == param+strlen(param)+2) {
268 memmove(val-1, val, strlen(val)+1);
274 /* Handle obsolete-style parameters */
275 if (obsolete_checksetup(param))
279 * Preemptive maintenance for "why didn't my misspelled command
282 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
283 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
291 /* Environment option */
293 for (i = 0; envp_init[i]; i++) {
294 if (i == MAX_INIT_ENVS) {
295 panic_later = "Too many boot env vars at `%s'";
298 if (!strncmp(param, envp_init[i], val - param))
301 envp_init[i] = param;
303 /* Command line option */
305 for (i = 0; argv_init[i]; i++) {
306 if (i == MAX_INIT_ARGS) {
307 panic_later = "Too many boot init vars at `%s'";
311 argv_init[i] = param;
316 #ifdef CONFIG_DEBUG_PAGEALLOC
317 int __read_mostly debug_pagealloc_enabled = 0;
320 static int __init init_setup(char *str)
324 execute_command = str;
326 * In case LILO is going to boot us with default command line,
327 * it prepends "auto" before the whole cmdline which makes
328 * the shell think it should execute a script with such name.
329 * So we ignore all arguments entered _before_ init=... [MJ]
331 for (i = 1; i < MAX_INIT_ARGS; i++)
335 __setup("init=", init_setup);
337 static int __init rdinit_setup(char *str)
341 ramdisk_execute_command = str;
342 /* See "auto" comment in init_setup */
343 for (i = 1; i < MAX_INIT_ARGS; i++)
347 __setup("rdinit=", rdinit_setup);
351 #ifdef CONFIG_X86_LOCAL_APIC
352 static void __init smp_init(void)
354 APIC_init_uniprocessor();
357 #define smp_init() do { } while (0)
360 static inline void setup_nr_cpu_ids(void) { }
361 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
365 #if NR_CPUS > BITS_PER_LONG
366 cpumask_t cpu_mask_all __read_mostly = CPU_MASK_ALL;
367 EXPORT_SYMBOL(cpu_mask_all);
370 /* Setup number of possible processor ids */
371 int nr_cpu_ids __read_mostly = NR_CPUS;
372 EXPORT_SYMBOL(nr_cpu_ids);
374 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
375 static void __init setup_nr_cpu_ids(void)
377 nr_cpu_ids = find_last_bit(cpumask_bits(cpu_possible_mask),NR_CPUS) + 1;
380 /* Called by boot processor to activate the rest. */
381 static void __init smp_init(void)
386 * Set up the current CPU as possible to migrate to.
387 * The other ones will be done by cpu_up/cpu_down()
389 set_cpu_active(smp_processor_id(), true);
391 /* FIXME: This should be done in userspace --RR */
392 for_each_present_cpu(cpu) {
393 if (num_online_cpus() >= setup_max_cpus)
395 if (!cpu_online(cpu))
399 /* Any cleanup work */
400 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
401 smp_cpus_done(setup_max_cpus);
407 * We need to store the untouched command line for future reference.
408 * We also need to store the touched command line since the parameter
409 * parsing is performed in place, and we should allow a component to
410 * store reference of name/value for future reference.
412 static void __init setup_command_line(char *command_line)
414 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
415 static_command_line = alloc_bootmem(strlen (command_line)+1);
416 strcpy (saved_command_line, boot_command_line);
417 strcpy (static_command_line, command_line);
421 * We need to finalize in a non-__init function or else race conditions
422 * between the root thread and the init thread may cause start_kernel to
423 * be reaped by free_initmem before the root thread has proceeded to
426 * gcc-3.4 accidentally inlines this function, so use noinline.
429 static noinline void __init_refok rest_init(void)
430 __releases(kernel_lock)
434 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
435 numa_default_policy();
436 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
437 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
441 * The boot idle thread must execute schedule()
442 * at least once to get things moving:
444 init_idle_bootup_task(current);
445 rcu_scheduler_starting();
446 preempt_enable_no_resched();
450 /* Call into cpu_idle with preempt disabled */
454 /* Check for early params. */
455 static int __init do_early_param(char *param, char *val)
457 struct obs_kernel_param *p;
459 for (p = __setup_start; p < __setup_end; p++) {
460 if ((p->early && strcmp(param, p->str) == 0) ||
461 (strcmp(param, "console") == 0 &&
462 strcmp(p->str, "earlycon") == 0)
464 if (p->setup_func(val) != 0)
466 "Malformed early option '%s'\n", param);
469 /* We accept everything at this stage. */
473 void __init parse_early_options(char *cmdline)
475 parse_args("early options", cmdline, NULL, 0, do_early_param);
478 /* Arch code calls this early on, or if not, just before other parsing. */
479 void __init parse_early_param(void)
481 static __initdata int done = 0;
482 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
487 /* All fall through to do_early_param. */
488 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
489 parse_early_options(tmp_cmdline);
494 * Activate the first processor.
497 static void __init boot_cpu_init(void)
499 int cpu = smp_processor_id();
500 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
501 set_cpu_online(cpu, true);
502 set_cpu_present(cpu, true);
503 set_cpu_possible(cpu, true);
506 void __init __weak smp_setup_processor_id(void)
510 void __init __weak thread_info_cache_init(void)
515 * Set up kernel memory allocators
517 static void __init mm_init(void)
520 * page_cgroup requires countinous pages as memmap
521 * and it's bigger than MAX_ORDER unless SPARSEMEM.
523 page_cgroup_init_flatmem();
526 pgtable_cache_init();
530 asmlinkage void __init start_kernel(void)
533 extern struct kernel_param __start___param[], __stop___param[];
535 smp_setup_processor_id();
538 * Need to run as early as possible, to initialize the
542 debug_objects_early_init();
545 * Set up the the initial canary ASAP:
547 boot_init_stack_canary();
552 early_boot_irqs_off();
553 early_init_irq_lock_class();
556 * Interrupts are still disabled. Do necessary setups, then
563 printk(KERN_NOTICE "%s", linux_banner);
564 setup_arch(&command_line);
565 mm_init_owner(&init_mm, &init_task);
566 setup_command_line(command_line);
567 setup_per_cpu_areas();
569 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
571 build_all_zonelists();
574 printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
576 parse_args("Booting kernel", static_command_line, __start___param,
577 __stop___param - __start___param,
578 &unknown_bootoption);
580 * These use large bootmem allocations and must precede
584 vfs_caches_init_early();
589 * Set up the scheduler prior starting any interrupts (such as the
590 * timer interrupt). Full topology setup happens at smp_init()
591 * time - but meanwhile we still have a functioning scheduler.
595 * Disable preemption - early bootup scheduling is extremely
596 * fragile until we cpu_idle() for the first time.
599 if (!irqs_disabled()) {
600 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
601 "enabled *very* early, fixing it\n");
605 /* init some links before init_ISA_irqs() */
616 if (!irqs_disabled())
617 printk(KERN_CRIT "start_kernel(): bug: interrupts were "
619 early_boot_irqs_on();
622 /* Interrupts are enabled now so all GFP allocations are safe. */
623 set_gfp_allowed_mask(__GFP_BITS_MASK);
625 kmem_cache_init_late();
628 * HACK ALERT! This is early. We're enabling the console before
629 * we've done PCI setups etc, and console_init() must be aware of
630 * this. But we do want output early, in case something goes wrong.
634 panic(panic_later, panic_param);
639 * Need to run this when irqs are enabled, because it wants
640 * to self-test [hard/soft]-irqs on/off lock inversion bugs
645 #ifdef CONFIG_BLK_DEV_INITRD
646 if (initrd_start && !initrd_below_start_ok &&
647 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
648 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
650 page_to_pfn(virt_to_page((void *)initrd_start)),
656 enable_debug_pagealloc();
660 debug_objects_mem_init();
662 setup_per_cpu_pageset();
671 efi_enter_virtual_mode();
673 thread_info_cache_init();
675 fork_init(num_physpages);
680 vfs_caches_init(num_physpages);
683 /* rootfs populating might need page-writeback */
684 page_writeback_init();
685 #ifdef CONFIG_PROC_FS
690 taskstats_init_early();
695 acpi_early_init(); /* before LAPIC and SMP init */
699 /* Do the rest non-__init'ed, we're now alive */
703 /* Call all constructor functions linked into the kernel. */
704 static void __init do_ctors(void)
706 #ifdef CONFIG_CONSTRUCTORS
707 ctor_fn_t *call = (ctor_fn_t *) __ctors_start;
709 for (; call < (ctor_fn_t *) __ctors_end; call++)
715 core_param(initcall_debug, initcall_debug, bool, 0644);
717 int do_one_initcall(initcall_t fn)
719 int count = preempt_count();
720 ktime_t calltime, delta, rettime;
722 struct boot_trace_call call;
723 struct boot_trace_ret ret;
725 if (initcall_debug) {
726 call.caller = task_pid_nr(current);
727 printk("calling %pF @ %i\n", fn, call.caller);
728 calltime = ktime_get();
729 trace_boot_call(&call, fn);
735 if (initcall_debug) {
736 disable_boot_trace();
737 rettime = ktime_get();
738 delta = ktime_sub(rettime, calltime);
739 ret.duration = (unsigned long long) ktime_to_ns(delta) >> 10;
740 trace_boot_ret(&ret, fn);
741 printk("initcall %pF returned %d after %Ld usecs\n", fn,
742 ret.result, ret.duration);
747 if (ret.result && ret.result != -ENODEV && initcall_debug)
748 sprintf(msgbuf, "error code %d ", ret.result);
750 if (preempt_count() != count) {
751 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
752 preempt_count() = count;
754 if (irqs_disabled()) {
755 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
759 printk("initcall %pF returned with %s\n", fn, msgbuf);
766 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
768 static void __init do_initcalls(void)
772 for (call = __early_initcall_end; call < __initcall_end; call++)
773 do_one_initcall(*call);
775 /* Make sure there is no pending stuff from the initcall sequence */
776 flush_scheduled_work();
780 * Ok, the machine is now initialized. None of the devices
781 * have been touched yet, but the CPU subsystem is up and
782 * running, and memory and process management works.
784 * Now we can finally start doing some real work..
786 static void __init do_basic_setup(void)
788 rcu_init_sched(); /* needed by module_init stage. */
791 usermodehelper_init();
798 static void __init do_pre_smp_initcalls(void)
802 for (call = __initcall_start; call < __early_initcall_end; call++)
803 do_one_initcall(*call);
806 static void run_init_process(char *init_filename)
808 argv_init[0] = init_filename;
809 kernel_execve(init_filename, argv_init, envp_init);
812 /* This is a non __init function. Force it to be noinline otherwise gcc
813 * makes it inline to init() and it becomes part of init.text section
815 static noinline int init_post(void)
816 __releases(kernel_lock)
818 /* need to finish all async __init code before freeing the memory */
819 async_synchronize_full();
823 system_state = SYSTEM_RUNNING;
824 numa_default_policy();
826 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
827 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
832 current->signal->flags |= SIGNAL_UNKILLABLE;
834 if (ramdisk_execute_command) {
835 run_init_process(ramdisk_execute_command);
836 printk(KERN_WARNING "Failed to execute %s\n",
837 ramdisk_execute_command);
841 * We try each of these until one succeeds.
843 * The Bourne shell can be used instead of init if we are
844 * trying to recover a really broken machine.
846 if (execute_command) {
847 run_init_process(execute_command);
848 printk(KERN_WARNING "Failed to execute %s. Attempting "
849 "defaults...\n", execute_command);
851 run_init_process("/sbin/init");
852 run_init_process("/etc/init");
853 run_init_process("/bin/init");
854 run_init_process("/bin/sh");
856 panic("No init found. Try passing init= option to kernel.");
859 static int __init kernel_init(void * unused)
864 * init can allocate pages on any node
866 set_mems_allowed(node_possible_map);
868 * init can run on any cpu.
870 set_cpus_allowed_ptr(current, cpu_all_mask);
872 * Tell the world that we're going to be the grim
873 * reaper of innocent orphaned children.
875 * We don't want people to have to make incorrect
876 * assumptions about where in the task array this
879 init_pid_ns.child_reaper = current;
881 cad_pid = task_pid(current);
883 smp_prepare_cpus(setup_max_cpus);
885 do_pre_smp_initcalls();
894 * check if there is an early userspace init. If yes, let it do all
898 if (!ramdisk_execute_command)
899 ramdisk_execute_command = "/init";
901 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
902 ramdisk_execute_command = NULL;
907 * Ok, we have completed the initial bootup, and
908 * we're essentially up and running. Get rid of the
909 * initmem segments and start the user-mode stuff..