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>
69 #include <asm/setup.h>
70 #include <asm/sections.h>
71 #include <asm/cacheflush.h>
73 #ifdef CONFIG_X86_LOCAL_APIC
78 * This is one of the first .c files built. Error out early if we have compiler
82 #if __GNUC__ == 4 && __GNUC_MINOR__ == 1 && __GNUC_PATCHLEVEL__ == 0
83 #warning gcc-4.1.0 is known to miscompile the kernel. A different compiler version is recommended.
86 static int kernel_init(void *);
88 extern void init_IRQ(void);
89 extern void fork_init(unsigned long);
90 extern void mca_init(void);
91 extern void sbus_init(void);
92 extern void prio_tree_init(void);
93 extern void radix_tree_init(void);
94 extern void free_initmem(void);
96 extern void acpi_early_init(void);
98 static inline void acpi_early_init(void) { }
100 #ifndef CONFIG_DEBUG_RODATA
101 static inline void mark_rodata_ro(void) { }
105 extern void tc_init(void);
108 enum system_states system_state;
109 EXPORT_SYMBOL(system_state);
112 * Boot command-line arguments
114 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
115 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
117 extern void time_init(void);
118 /* Default late time init is NULL. archs can override this later. */
119 void (*late_time_init)(void);
120 extern void softirq_init(void);
122 /* Untouched command line saved by arch-specific code. */
123 char __initdata boot_command_line[COMMAND_LINE_SIZE];
124 /* Untouched saved command line (eg. for /proc) */
125 char *saved_command_line;
126 /* Command line for parameter parsing */
127 static char *static_command_line;
129 static char *execute_command;
130 static char *ramdisk_execute_command;
133 /* Setup configured maximum number of CPUs to activate */
134 unsigned int __initdata setup_max_cpus = NR_CPUS;
137 * Setup routine for controlling SMP activation
139 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
140 * activation entirely (the MPS table probe still happens, though).
142 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
143 * greater than 0, limits the maximum number of CPUs activated in
146 #ifndef CONFIG_X86_IO_APIC
147 static inline void disable_ioapic_setup(void) {};
150 static int __init nosmp(char *str)
153 disable_ioapic_setup();
157 early_param("nosmp", nosmp);
159 static int __init maxcpus(char *str)
161 get_option(&str, &setup_max_cpus);
162 if (setup_max_cpus == 0)
163 disable_ioapic_setup();
168 early_param("maxcpus", maxcpus);
170 #define setup_max_cpus NR_CPUS
174 * If set, this is an indication to the drivers that reset the underlying
175 * device before going ahead with the initialization otherwise driver might
176 * rely on the BIOS and skip the reset operation.
178 * This is useful if kernel is booting in an unreliable environment.
179 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
180 * skipped and devices will be in unknown state.
182 unsigned int reset_devices;
183 EXPORT_SYMBOL(reset_devices);
185 static int __init set_reset_devices(char *str)
191 __setup("reset_devices", set_reset_devices);
193 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
194 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
195 static const char *panic_later, *panic_param;
197 extern struct obs_kernel_param __setup_start[], __setup_end[];
199 static int __init obsolete_checksetup(char *line)
201 struct obs_kernel_param *p;
202 int had_early_param = 0;
206 int n = strlen(p->str);
207 if (!strncmp(line, p->str, n)) {
209 /* Already done in parse_early_param?
210 * (Needs exact match on param part).
211 * Keep iterating, as we can have early
212 * params and __setups of same names 8( */
213 if (line[n] == '\0' || line[n] == '=')
215 } else if (!p->setup_func) {
216 printk(KERN_WARNING "Parameter %s is obsolete,"
217 " ignored\n", p->str);
219 } else if (p->setup_func(line + n))
223 } while (p < __setup_end);
225 return had_early_param;
229 * This should be approx 2 Bo*oMips to start (note initial shift), and will
230 * still work even if initially too large, it will just take slightly longer
232 unsigned long loops_per_jiffy = (1<<12);
234 EXPORT_SYMBOL(loops_per_jiffy);
236 static int __init debug_kernel(char *str)
238 console_loglevel = 10;
242 static int __init quiet_kernel(char *str)
244 console_loglevel = 4;
248 early_param("debug", debug_kernel);
249 early_param("quiet", quiet_kernel);
251 static int __init loglevel(char *str)
253 get_option(&str, &console_loglevel);
257 early_param("loglevel", loglevel);
260 * Unknown boot options get handed to init, unless they look like
263 static int __init unknown_bootoption(char *param, char *val)
265 /* Change NUL term back to "=", to make "param" the whole string. */
267 /* param=val or param="val"? */
268 if (val == param+strlen(param)+1)
270 else if (val == param+strlen(param)+2) {
272 memmove(val-1, val, strlen(val)+1);
278 /* Handle obsolete-style parameters */
279 if (obsolete_checksetup(param))
283 * Preemptive maintenance for "why didn't my misspelled command
286 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
287 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
295 /* Environment option */
297 for (i = 0; envp_init[i]; i++) {
298 if (i == MAX_INIT_ENVS) {
299 panic_later = "Too many boot env vars at `%s'";
302 if (!strncmp(param, envp_init[i], val - param))
305 envp_init[i] = param;
307 /* Command line option */
309 for (i = 0; argv_init[i]; i++) {
310 if (i == MAX_INIT_ARGS) {
311 panic_later = "Too many boot init vars at `%s'";
315 argv_init[i] = param;
320 #ifdef CONFIG_DEBUG_PAGEALLOC
321 int __read_mostly debug_pagealloc_enabled = 0;
324 static int __init init_setup(char *str)
328 execute_command = str;
330 * In case LILO is going to boot us with default command line,
331 * it prepends "auto" before the whole cmdline which makes
332 * the shell think it should execute a script with such name.
333 * So we ignore all arguments entered _before_ init=... [MJ]
335 for (i = 1; i < MAX_INIT_ARGS; i++)
339 __setup("init=", init_setup);
341 static int __init rdinit_setup(char *str)
345 ramdisk_execute_command = str;
346 /* See "auto" comment in init_setup */
347 for (i = 1; i < MAX_INIT_ARGS; i++)
351 __setup("rdinit=", rdinit_setup);
355 #ifdef CONFIG_X86_LOCAL_APIC
356 static void __init smp_init(void)
358 APIC_init_uniprocessor();
361 #define smp_init() do { } while (0)
364 static inline void setup_per_cpu_areas(void) { }
365 static inline void setup_nr_cpu_ids(void) { }
366 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
370 #if NR_CPUS > BITS_PER_LONG
371 cpumask_t cpu_mask_all __read_mostly = CPU_MASK_ALL;
372 EXPORT_SYMBOL(cpu_mask_all);
375 /* Setup number of possible processor ids */
376 int nr_cpu_ids __read_mostly = NR_CPUS;
377 EXPORT_SYMBOL(nr_cpu_ids);
379 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
380 static void __init setup_nr_cpu_ids(void)
382 int cpu, highest_cpu = 0;
384 for_each_possible_cpu(cpu)
387 nr_cpu_ids = highest_cpu + 1;
390 #ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA
391 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
393 EXPORT_SYMBOL(__per_cpu_offset);
395 static void __init setup_per_cpu_areas(void)
397 unsigned long size, i;
399 unsigned long nr_possible_cpus = num_possible_cpus();
401 /* Copy section for each CPU (we discard the original) */
402 size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
403 ptr = alloc_bootmem_pages(size * nr_possible_cpus);
405 for_each_possible_cpu(i) {
406 __per_cpu_offset[i] = ptr - __per_cpu_start;
407 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
411 #endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */
413 /* Called by boot processor to activate the rest. */
414 static void __init smp_init(void)
419 * Set up the current CPU as possible to migrate to.
420 * The other ones will be done by cpu_up/cpu_down()
422 cpu = smp_processor_id();
423 cpu_set(cpu, cpu_active_map);
425 /* FIXME: This should be done in userspace --RR */
426 for_each_present_cpu(cpu) {
427 if (num_online_cpus() >= setup_max_cpus)
429 if (!cpu_online(cpu))
433 /* Any cleanup work */
434 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
435 smp_cpus_done(setup_max_cpus);
441 * We need to store the untouched command line for future reference.
442 * We also need to store the touched command line since the parameter
443 * parsing is performed in place, and we should allow a component to
444 * store reference of name/value for future reference.
446 static void __init setup_command_line(char *command_line)
448 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
449 static_command_line = alloc_bootmem(strlen (command_line)+1);
450 strcpy (saved_command_line, boot_command_line);
451 strcpy (static_command_line, command_line);
455 * We need to finalize in a non-__init function or else race conditions
456 * between the root thread and the init thread may cause start_kernel to
457 * be reaped by free_initmem before the root thread has proceeded to
460 * gcc-3.4 accidentally inlines this function, so use noinline.
463 static void noinline __init_refok rest_init(void)
464 __releases(kernel_lock)
468 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
469 numa_default_policy();
470 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
471 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
475 * The boot idle thread must execute schedule()
476 * at least once to get things moving:
478 init_idle_bootup_task(current);
479 preempt_enable_no_resched();
483 /* Call into cpu_idle with preempt disabled */
487 /* Check for early params. */
488 static int __init do_early_param(char *param, char *val)
490 struct obs_kernel_param *p;
492 for (p = __setup_start; p < __setup_end; p++) {
493 if ((p->early && strcmp(param, p->str) == 0) ||
494 (strcmp(param, "console") == 0 &&
495 strcmp(p->str, "earlycon") == 0)
497 if (p->setup_func(val) != 0)
499 "Malformed early option '%s'\n", param);
502 /* We accept everything at this stage. */
506 /* Arch code calls this early on, or if not, just before other parsing. */
507 void __init parse_early_param(void)
509 static __initdata int done = 0;
510 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
515 /* All fall through to do_early_param. */
516 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
517 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
522 * Activate the first processor.
525 static void __init boot_cpu_init(void)
527 int cpu = smp_processor_id();
528 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
529 cpu_set(cpu, cpu_online_map);
530 cpu_set(cpu, cpu_present_map);
531 cpu_set(cpu, cpu_possible_map);
534 void __init __weak smp_setup_processor_id(void)
538 void __init __weak thread_info_cache_init(void)
542 asmlinkage void __init start_kernel(void)
545 extern struct kernel_param __start___param[], __stop___param[];
547 smp_setup_processor_id();
550 * Need to run as early as possible, to initialize the
555 debug_objects_early_init();
559 early_boot_irqs_off();
560 early_init_irq_lock_class();
563 * Interrupts are still disabled. Do necessary setups, then
571 printk(linux_banner);
572 setup_arch(&command_line);
573 mm_init_owner(&init_mm, &init_task);
574 setup_command_line(command_line);
576 setup_per_cpu_areas();
578 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
581 * Set up the scheduler prior starting any interrupts (such as the
582 * timer interrupt). Full topology setup happens at smp_init()
583 * time - but meanwhile we still have a functioning scheduler.
587 * Disable preemption - early bootup scheduling is extremely
588 * fragile until we cpu_idle() for the first time.
591 build_all_zonelists();
593 printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
595 parse_args("Booting kernel", static_command_line, __start___param,
596 __stop___param - __start___param,
597 &unknown_bootoption);
598 if (!irqs_disabled()) {
599 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
600 "enabled *very* early, fixing it\n");
615 if (!irqs_disabled())
616 printk("start_kernel(): bug: interrupts were enabled early\n");
617 early_boot_irqs_on();
621 * HACK ALERT! This is early. We're enabling the console before
622 * we've done PCI setups etc, and console_init() must be aware of
623 * this. But we do want output early, in case something goes wrong.
627 panic(panic_later, panic_param);
632 * Need to run this when irqs are enabled, because it wants
633 * to self-test [hard/soft]-irqs on/off lock inversion bugs
638 #ifdef CONFIG_BLK_DEV_INITRD
639 if (initrd_start && !initrd_below_start_ok &&
640 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
641 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
643 page_to_pfn(virt_to_page((void *)initrd_start)),
649 vfs_caches_init_early();
653 enable_debug_pagealloc();
656 debug_objects_mem_init();
658 setup_per_cpu_pageset();
664 pgtable_cache_init();
669 efi_enter_virtual_mode();
671 thread_info_cache_init();
672 fork_init(num_physpages);
677 vfs_caches_init(num_physpages);
680 /* rootfs populating might need page-writeback */
681 page_writeback_init();
682 #ifdef CONFIG_PROC_FS
687 taskstats_init_early();
692 acpi_early_init(); /* before LAPIC and SMP init */
696 /* Do the rest non-__init'ed, we're now alive */
700 static int initcall_debug;
701 core_param(initcall_debug, initcall_debug, bool, 0644);
703 int do_one_initcall(initcall_t fn)
705 int count = preempt_count();
708 struct boot_trace it;
710 if (initcall_debug) {
711 it.caller = task_pid_nr(current);
712 printk("calling %pF @ %i\n", fn, it.caller);
713 it.calltime = ktime_get();
718 if (initcall_debug) {
719 it.rettime = ktime_get();
720 delta = ktime_sub(it.rettime, it.calltime);
721 it.duration = (unsigned long long) delta.tv64 >> 10;
722 printk("initcall %pF returned %d after %Ld usecs\n", fn,
723 it.result, it.duration);
729 if (it.result && it.result != -ENODEV && initcall_debug)
730 sprintf(msgbuf, "error code %d ", it.result);
732 if (preempt_count() != count) {
733 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
734 preempt_count() = count;
736 if (irqs_disabled()) {
737 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
741 printk("initcall %pF returned with %s\n", fn, msgbuf);
748 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
750 static void __init do_initcalls(void)
754 for (call = __early_initcall_end; call < __initcall_end; call++)
755 do_one_initcall(*call);
757 /* Make sure there is no pending stuff from the initcall sequence */
758 flush_scheduled_work();
762 * Ok, the machine is now initialized. None of the devices
763 * have been touched yet, but the CPU subsystem is up and
764 * running, and memory and process management works.
766 * Now we can finally start doing some real work..
768 static void __init do_basic_setup(void)
770 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);
856 smp_prepare_cpus(setup_max_cpus);
858 do_pre_smp_initcalls();
869 * check if there is an early userspace init. If yes, let it do all
873 if (!ramdisk_execute_command)
874 ramdisk_execute_command = "/init";
876 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
877 ramdisk_execute_command = NULL;
882 * Ok, we have completed the initial bootup, and
883 * we're essentially up and running. Get rid of the
884 * initmem segments and start the user-mode stuff..