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/hdreg.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/kernel_stat.h>
32 #include <linux/start_kernel.h>
33 #include <linux/security.h>
34 #include <linux/workqueue.h>
35 #include <linux/profile.h>
36 #include <linux/rcupdate.h>
37 #include <linux/moduleparam.h>
38 #include <linux/kallsyms.h>
39 #include <linux/writeback.h>
40 #include <linux/cpu.h>
41 #include <linux/cpuset.h>
42 #include <linux/efi.h>
43 #include <linux/tick.h>
44 #include <linux/taskstats_kern.h>
45 #include <linux/delayacct.h>
46 #include <linux/unistd.h>
47 #include <linux/rmap.h>
48 #include <linux/mempolicy.h>
49 #include <linux/key.h>
50 #include <linux/unwind.h>
51 #include <linux/buffer_head.h>
52 #include <linux/debug_locks.h>
53 #include <linux/lockdep.h>
54 #include <linux/pid_namespace.h>
55 #include <linux/device.h>
59 #include <asm/setup.h>
60 #include <asm/sections.h>
61 #include <asm/cacheflush.h>
63 #ifdef CONFIG_X86_LOCAL_APIC
68 * This is one of the first .c files built. Error out early if we have compiler
71 * Versions of gcc older than that listed below may actually compile and link
72 * okay, but the end product can have subtle run time bugs. To avoid associated
73 * bogus bug reports, we flatly refuse to compile with a gcc that is known to be
74 * too old from the very beginning.
76 #if (__GNUC__ < 3) || (__GNUC__ == 3 && __GNUC_MINOR__ < 2)
77 #error Sorry, your GCC is too old. It builds incorrect kernels.
80 #if __GNUC__ == 4 && __GNUC_MINOR__ == 1 && __GNUC_PATCHLEVEL__ == 0
81 #warning gcc-4.1.0 is known to miscompile the kernel. A different compiler version is recommended.
84 static int init(void *);
86 extern void init_IRQ(void);
87 extern void fork_init(unsigned long);
88 extern void mca_init(void);
89 extern void sbus_init(void);
90 extern void signals_init(void);
91 extern void pidhash_init(void);
92 extern void pidmap_init(void);
93 extern void prio_tree_init(void);
94 extern void radix_tree_init(void);
95 extern void free_initmem(void);
96 extern void prepare_namespace(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;
134 /* Setup configured maximum number of CPUs to activate */
135 static unsigned int max_cpus = NR_CPUS;
138 * If set, this is an indication to the drivers that reset the underlying
139 * device before going ahead with the initialization otherwise driver might
140 * rely on the BIOS and skip the reset operation.
142 * This is useful if kernel is booting in an unreliable environment.
143 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
144 * skipped and devices will be in unknown state.
146 unsigned int reset_devices;
147 EXPORT_SYMBOL(reset_devices);
150 * Setup routine for controlling SMP activation
152 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
153 * activation entirely (the MPS table probe still happens, though).
155 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
156 * greater than 0, limits the maximum number of CPUs activated in
159 static int __init nosmp(char *str)
165 __setup("nosmp", nosmp);
167 static int __init maxcpus(char *str)
169 get_option(&str, &max_cpus);
173 __setup("maxcpus=", maxcpus);
175 static int __init set_reset_devices(char *str)
181 __setup("reset_devices", set_reset_devices);
183 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
184 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
185 static const char *panic_later, *panic_param;
187 extern struct obs_kernel_param __setup_start[], __setup_end[];
189 static int __init obsolete_checksetup(char *line)
191 struct obs_kernel_param *p;
192 int had_early_param = 0;
196 int n = strlen(p->str);
197 if (!strncmp(line, p->str, n)) {
199 /* Already done in parse_early_param?
200 * (Needs exact match on param part).
201 * Keep iterating, as we can have early
202 * params and __setups of same names 8( */
203 if (line[n] == '\0' || line[n] == '=')
205 } else if (!p->setup_func) {
206 printk(KERN_WARNING "Parameter %s is obsolete,"
207 " ignored\n", p->str);
209 } else if (p->setup_func(line + n))
213 } while (p < __setup_end);
215 return had_early_param;
219 * This should be approx 2 Bo*oMips to start (note initial shift), and will
220 * still work even if initially too large, it will just take slightly longer
222 unsigned long loops_per_jiffy = (1<<12);
224 EXPORT_SYMBOL(loops_per_jiffy);
226 static int __init debug_kernel(char *str)
230 console_loglevel = 10;
234 static int __init quiet_kernel(char *str)
238 console_loglevel = 4;
242 __setup("debug", debug_kernel);
243 __setup("quiet", quiet_kernel);
245 static int __init loglevel(char *str)
247 get_option(&str, &console_loglevel);
251 __setup("loglevel=", loglevel);
254 * Unknown boot options get handed to init, unless they look like
257 static int __init unknown_bootoption(char *param, char *val)
259 /* Change NUL term back to "=", to make "param" the whole string. */
261 /* param=val or param="val"? */
262 if (val == param+strlen(param)+1)
264 else if (val == param+strlen(param)+2) {
266 memmove(val-1, val, strlen(val)+1);
272 /* Handle obsolete-style parameters */
273 if (obsolete_checksetup(param))
277 * Preemptive maintenance for "why didn't my mispelled command
280 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
281 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
289 /* Environment option */
291 for (i = 0; envp_init[i]; i++) {
292 if (i == MAX_INIT_ENVS) {
293 panic_later = "Too many boot env vars at `%s'";
296 if (!strncmp(param, envp_init[i], val - param))
299 envp_init[i] = param;
301 /* Command line option */
303 for (i = 0; argv_init[i]; i++) {
304 if (i == MAX_INIT_ARGS) {
305 panic_later = "Too many boot init vars at `%s'";
309 argv_init[i] = param;
314 static int __init init_setup(char *str)
318 execute_command = str;
320 * In case LILO is going to boot us with default command line,
321 * it prepends "auto" before the whole cmdline which makes
322 * the shell think it should execute a script with such name.
323 * So we ignore all arguments entered _before_ init=... [MJ]
325 for (i = 1; i < MAX_INIT_ARGS; i++)
329 __setup("init=", init_setup);
331 static int __init rdinit_setup(char *str)
335 ramdisk_execute_command = str;
336 /* See "auto" comment in init_setup */
337 for (i = 1; i < MAX_INIT_ARGS; i++)
341 __setup("rdinit=", rdinit_setup);
345 #ifdef CONFIG_X86_LOCAL_APIC
346 static void __init smp_init(void)
348 APIC_init_uniprocessor();
351 #define smp_init() do { } while (0)
354 static inline void setup_per_cpu_areas(void) { }
355 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
359 #ifdef __GENERIC_PER_CPU
360 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
362 EXPORT_SYMBOL(__per_cpu_offset);
364 static void __init setup_per_cpu_areas(void)
366 unsigned long size, i;
368 unsigned long nr_possible_cpus = num_possible_cpus();
370 /* Copy section for each CPU (we discard the original) */
371 size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
372 #ifdef CONFIG_MODULES
373 if (size < PERCPU_ENOUGH_ROOM)
374 size = PERCPU_ENOUGH_ROOM;
376 ptr = alloc_bootmem(size * nr_possible_cpus);
378 for_each_possible_cpu(i) {
379 __per_cpu_offset[i] = ptr - __per_cpu_start;
380 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
384 #endif /* !__GENERIC_PER_CPU */
386 /* Called by boot processor to activate the rest. */
387 static void __init smp_init(void)
391 /* FIXME: This should be done in userspace --RR */
392 for_each_present_cpu(i) {
393 if (num_online_cpus() >= max_cpus)
399 /* Any cleanup work */
400 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
401 smp_cpus_done(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 void noinline rest_init(void)
430 __releases(kernel_lock)
432 kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
433 numa_default_policy();
437 * The boot idle thread must execute schedule()
438 * at least one to get things moving:
440 preempt_enable_no_resched();
444 /* Call into cpu_idle with preempt disabled */
448 /* Check for early params. */
449 static int __init do_early_param(char *param, char *val)
451 struct obs_kernel_param *p;
453 for (p = __setup_start; p < __setup_end; p++) {
454 if (p->early && strcmp(param, p->str) == 0) {
455 if (p->setup_func(val) != 0)
457 "Malformed early option '%s'\n", param);
460 /* We accept everything at this stage. */
464 /* Arch code calls this early on, or if not, just before other parsing. */
465 void __init parse_early_param(void)
467 static __initdata int done = 0;
468 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
473 /* All fall through to do_early_param. */
474 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
475 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
480 * Activate the first processor.
483 static void __init boot_cpu_init(void)
485 int cpu = smp_processor_id();
486 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
487 cpu_set(cpu, cpu_online_map);
488 cpu_set(cpu, cpu_present_map);
489 cpu_set(cpu, cpu_possible_map);
492 void __init __attribute__((weak)) smp_setup_processor_id(void)
496 asmlinkage void __init start_kernel(void)
499 extern struct kernel_param __start___param[], __stop___param[];
501 smp_setup_processor_id();
504 * Need to run as early as possible, to initialize the
511 early_boot_irqs_off();
512 early_init_irq_lock_class();
515 * Interrupts are still disabled. Do necessary setups, then
523 printk(linux_banner);
524 setup_arch(&command_line);
525 setup_command_line(command_line);
527 setup_per_cpu_areas();
528 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
531 * Set up the scheduler prior starting any interrupts (such as the
532 * timer interrupt). Full topology setup happens at smp_init()
533 * time - but meanwhile we still have a functioning scheduler.
537 * Disable preemption - early bootup scheduling is extremely
538 * fragile until we cpu_idle() for the first time.
541 build_all_zonelists();
543 printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
545 parse_args("Booting kernel", static_command_line, __start___param,
546 __stop___param - __start___param,
547 &unknown_bootoption);
548 if (!irqs_disabled()) {
549 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
550 "enabled *very* early, fixing it\n");
564 if (!irqs_disabled())
565 printk("start_kernel(): bug: interrupts were enabled early\n");
566 early_boot_irqs_on();
570 * HACK ALERT! This is early. We're enabling the console before
571 * we've done PCI setups etc, and console_init() must be aware of
572 * this. But we do want output early, in case something goes wrong.
576 panic(panic_later, panic_param);
581 * Need to run this when irqs are enabled, because it wants
582 * to self-test [hard/soft]-irqs on/off lock inversion bugs
587 #ifdef CONFIG_BLK_DEV_INITRD
588 if (initrd_start && !initrd_below_start_ok &&
589 initrd_start < min_low_pfn << PAGE_SHIFT) {
590 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
591 "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
595 vfs_caches_init_early();
599 setup_per_cpu_pageset();
605 pgtable_cache_init();
610 efi_enter_virtual_mode();
612 fork_init(num_physpages);
618 vfs_caches_init(num_physpages);
621 /* rootfs populating might need page-writeback */
622 page_writeback_init();
623 #ifdef CONFIG_PROC_FS
627 taskstats_init_early();
632 acpi_early_init(); /* before LAPIC and SMP init */
634 /* Do the rest non-__init'ed, we're now alive */
638 static int __initdata initcall_debug;
640 static int __init initcall_debug_setup(char *str)
645 __setup("initcall_debug", initcall_debug_setup);
647 extern initcall_t __initcall_start[], __initcall_end[];
649 static void __init do_initcalls(void)
652 int count = preempt_count();
654 for (call = __initcall_start; call < __initcall_end; call++) {
659 if (initcall_debug) {
660 printk("Calling initcall 0x%p", *call);
661 print_fn_descriptor_symbol(": %s()",
662 (unsigned long) *call);
668 if (result && result != -ENODEV && initcall_debug) {
669 sprintf(msgbuf, "error code %d", result);
672 if (preempt_count() != count) {
673 msg = "preemption imbalance";
674 preempt_count() = count;
676 if (irqs_disabled()) {
677 msg = "disabled interrupts";
681 printk(KERN_WARNING "initcall at 0x%p", *call);
682 print_fn_descriptor_symbol(": %s()",
683 (unsigned long) *call);
684 printk(": returned with %s\n", msg);
688 /* Make sure there is no pending stuff from the initcall sequence */
689 flush_scheduled_work();
693 * Ok, the machine is now initialized. None of the devices
694 * have been touched yet, but the CPU subsystem is up and
695 * running, and memory and process management works.
697 * Now we can finally start doing some real work..
699 static void __init do_basic_setup(void)
701 /* drivers will send hotplug events */
703 usermodehelper_init();
706 #ifdef CONFIG_PROC_FS
713 static void __init do_pre_smp_initcalls(void)
715 extern int spawn_ksoftirqd(void);
717 extern int migration_init(void);
722 spawn_softlockup_task();
725 static void run_init_process(char *init_filename)
727 argv_init[0] = init_filename;
728 kernel_execve(init_filename, argv_init, envp_init);
731 /* This is a non __init function. Force it to be noinline otherwise gcc
732 * makes it inline to init() and it becomes part of init.text section
734 static int noinline init_post(void)
739 system_state = SYSTEM_RUNNING;
740 numa_default_policy();
742 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
743 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
748 if (ramdisk_execute_command) {
749 run_init_process(ramdisk_execute_command);
750 printk(KERN_WARNING "Failed to execute %s\n",
751 ramdisk_execute_command);
755 * We try each of these until one succeeds.
757 * The Bourne shell can be used instead of init if we are
758 * trying to recover a really broken machine.
760 if (execute_command) {
761 run_init_process(execute_command);
762 printk(KERN_WARNING "Failed to execute %s. Attempting "
763 "defaults...\n", execute_command);
765 run_init_process("/sbin/init");
766 run_init_process("/etc/init");
767 run_init_process("/bin/init");
768 run_init_process("/bin/sh");
770 panic("No init found. Try passing init= option to kernel.");
773 static int __init init(void * unused)
777 * init can run on any cpu.
779 set_cpus_allowed(current, CPU_MASK_ALL);
781 * Tell the world that we're going to be the grim
782 * reaper of innocent orphaned children.
784 * We don't want people to have to make incorrect
785 * assumptions about where in the task array this
788 init_pid_ns.child_reaper = current;
790 cad_pid = task_pid(current);
792 smp_prepare_cpus(max_cpus);
794 do_pre_smp_initcalls();
804 * check if there is an early userspace init. If yes, let it do all
808 if (!ramdisk_execute_command)
809 ramdisk_execute_command = "/init";
811 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
812 ramdisk_execute_command = NULL;
817 * Ok, we have completed the initial bootup, and
818 * we're essentially up and running. Get rid of the
819 * initmem segments and start the user-mode stuff..