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/taskstats_kern.h>
44 #include <linux/delayacct.h>
45 #include <linux/unistd.h>
46 #include <linux/rmap.h>
47 #include <linux/mempolicy.h>
48 #include <linux/key.h>
49 #include <linux/unwind.h>
50 #include <linux/buffer_head.h>
51 #include <linux/debug_locks.h>
52 #include <linux/lockdep.h>
53 #include <linux/utsrelease.h>
54 #include <linux/pid_namespace.h>
55 #include <linux/compile.h>
56 #include <linux/device.h>
60 #include <asm/setup.h>
61 #include <asm/sections.h>
62 #include <asm/cacheflush.h>
64 #ifdef CONFIG_X86_LOCAL_APIC
69 * This is one of the first .c files built. Error out early if we have compiler
72 * Versions of gcc older than that listed below may actually compile and link
73 * okay, but the end product can have subtle run time bugs. To avoid associated
74 * bogus bug reports, we flatly refuse to compile with a gcc that is known to be
75 * too old from the very beginning.
77 #if (__GNUC__ < 3) || (__GNUC__ == 3 && __GNUC_MINOR__ < 2)
78 #error Sorry, your GCC is too old. It builds incorrect kernels.
81 #if __GNUC__ == 4 && __GNUC_MINOR__ == 1 && __GNUC_PATCHLEVEL__ == 0
82 #warning gcc-4.1.0 is known to miscompile the kernel. A different compiler version is recommended.
85 static int init(void *);
87 extern void init_IRQ(void);
88 extern void fork_init(unsigned long);
89 extern void mca_init(void);
90 extern void sbus_init(void);
91 extern void sysctl_init(void);
92 extern void signals_init(void);
93 extern void pidhash_init(void);
94 extern void pidmap_init(void);
95 extern void prio_tree_init(void);
96 extern void radix_tree_init(void);
97 extern void free_initmem(void);
98 extern void prepare_namespace(void);
100 extern void acpi_early_init(void);
102 static inline void acpi_early_init(void) { }
104 #ifndef CONFIG_DEBUG_RODATA
105 static inline void mark_rodata_ro(void) { }
109 extern void tc_init(void);
112 enum system_states system_state;
113 EXPORT_SYMBOL(system_state);
116 * Boot command-line arguments
118 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
119 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
121 extern void time_init(void);
122 /* Default late time init is NULL. archs can override this later. */
123 void (*late_time_init)(void);
124 extern void softirq_init(void);
126 /* Untouched command line (eg. for /proc) saved by arch-specific code. */
127 char saved_command_line[COMMAND_LINE_SIZE];
129 static char *execute_command;
130 static char *ramdisk_execute_command;
132 /* Setup configured maximum number of CPUs to activate */
133 static unsigned int max_cpus = NR_CPUS;
136 * If set, this is an indication to the drivers that reset the underlying
137 * device before going ahead with the initialization otherwise driver might
138 * rely on the BIOS and skip the reset operation.
140 * This is useful if kernel is booting in an unreliable environment.
141 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
142 * skipped and devices will be in unknown state.
144 unsigned int reset_devices;
145 EXPORT_SYMBOL(reset_devices);
148 * Setup routine for controlling SMP activation
150 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
151 * activation entirely (the MPS table probe still happens, though).
153 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
154 * greater than 0, limits the maximum number of CPUs activated in
157 static int __init nosmp(char *str)
163 __setup("nosmp", nosmp);
165 static int __init maxcpus(char *str)
167 get_option(&str, &max_cpus);
171 __setup("maxcpus=", maxcpus);
173 static int __init set_reset_devices(char *str)
179 __setup("reset_devices", set_reset_devices);
181 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
182 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
183 static const char *panic_later, *panic_param;
185 extern struct obs_kernel_param __setup_start[], __setup_end[];
187 static int __init obsolete_checksetup(char *line)
189 struct obs_kernel_param *p;
190 int had_early_param = 0;
194 int n = strlen(p->str);
195 if (!strncmp(line, p->str, n)) {
197 /* Already done in parse_early_param?
198 * (Needs exact match on param part).
199 * Keep iterating, as we can have early
200 * params and __setups of same names 8( */
201 if (line[n] == '\0' || line[n] == '=')
203 } else if (!p->setup_func) {
204 printk(KERN_WARNING "Parameter %s is obsolete,"
205 " ignored\n", p->str);
207 } else if (p->setup_func(line + n))
211 } while (p < __setup_end);
213 return had_early_param;
217 * This should be approx 2 Bo*oMips to start (note initial shift), and will
218 * still work even if initially too large, it will just take slightly longer
220 unsigned long loops_per_jiffy = (1<<12);
222 EXPORT_SYMBOL(loops_per_jiffy);
224 static int __init debug_kernel(char *str)
228 console_loglevel = 10;
232 static int __init quiet_kernel(char *str)
236 console_loglevel = 4;
240 __setup("debug", debug_kernel);
241 __setup("quiet", quiet_kernel);
243 static int __init loglevel(char *str)
245 get_option(&str, &console_loglevel);
249 __setup("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 mispelled 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 static int __init init_setup(char *str)
316 execute_command = str;
318 * In case LILO is going to boot us with default command line,
319 * it prepends "auto" before the whole cmdline which makes
320 * the shell think it should execute a script with such name.
321 * So we ignore all arguments entered _before_ init=... [MJ]
323 for (i = 1; i < MAX_INIT_ARGS; i++)
327 __setup("init=", init_setup);
329 static int __init rdinit_setup(char *str)
333 ramdisk_execute_command = str;
334 /* See "auto" comment in init_setup */
335 for (i = 1; i < MAX_INIT_ARGS; i++)
339 __setup("rdinit=", rdinit_setup);
343 #ifdef CONFIG_X86_LOCAL_APIC
344 static void __init smp_init(void)
346 APIC_init_uniprocessor();
349 #define smp_init() do { } while (0)
352 static inline void setup_per_cpu_areas(void) { }
353 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
357 #ifdef __GENERIC_PER_CPU
358 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
360 EXPORT_SYMBOL(__per_cpu_offset);
362 static void __init setup_per_cpu_areas(void)
364 unsigned long size, i;
366 unsigned long nr_possible_cpus = num_possible_cpus();
368 /* Copy section for each CPU (we discard the original) */
369 size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
370 #ifdef CONFIG_MODULES
371 if (size < PERCPU_ENOUGH_ROOM)
372 size = PERCPU_ENOUGH_ROOM;
374 ptr = alloc_bootmem(size * nr_possible_cpus);
376 for_each_possible_cpu(i) {
377 __per_cpu_offset[i] = ptr - __per_cpu_start;
378 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
382 #endif /* !__GENERIC_PER_CPU */
384 /* Called by boot processor to activate the rest. */
385 static void __init smp_init(void)
389 /* FIXME: This should be done in userspace --RR */
390 for_each_present_cpu(i) {
391 if (num_online_cpus() >= max_cpus)
397 /* Any cleanup work */
398 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
399 smp_cpus_done(max_cpus);
401 /* Get other processors into their bootup holding patterns. */
410 * We need to finalize in a non-__init function or else race conditions
411 * between the root thread and the init thread may cause start_kernel to
412 * be reaped by free_initmem before the root thread has proceeded to
415 * gcc-3.4 accidentally inlines this function, so use noinline.
418 static void noinline rest_init(void)
419 __releases(kernel_lock)
421 kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
422 numa_default_policy();
426 * The boot idle thread must execute schedule()
427 * at least one to get things moving:
429 preempt_enable_no_resched();
433 /* Call into cpu_idle with preempt disabled */
437 /* Check for early params. */
438 static int __init do_early_param(char *param, char *val)
440 struct obs_kernel_param *p;
442 for (p = __setup_start; p < __setup_end; p++) {
443 if (p->early && strcmp(param, p->str) == 0) {
444 if (p->setup_func(val) != 0)
446 "Malformed early option '%s'\n", param);
449 /* We accept everything at this stage. */
453 /* Arch code calls this early on, or if not, just before other parsing. */
454 void __init parse_early_param(void)
456 static __initdata int done = 0;
457 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
462 /* All fall through to do_early_param. */
463 strlcpy(tmp_cmdline, saved_command_line, COMMAND_LINE_SIZE);
464 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
469 * Activate the first processor.
472 static void __init boot_cpu_init(void)
474 int cpu = smp_processor_id();
475 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
476 cpu_set(cpu, cpu_online_map);
477 cpu_set(cpu, cpu_present_map);
478 cpu_set(cpu, cpu_possible_map);
481 void __init __attribute__((weak)) smp_setup_processor_id(void)
485 static const char linux_banner[] =
486 "Linux version " UTS_RELEASE
487 " (" LINUX_COMPILE_BY "@" LINUX_COMPILE_HOST ")"
488 " (" LINUX_COMPILER ")"
489 " " UTS_VERSION "\n";
491 asmlinkage void __init start_kernel(void)
494 extern struct kernel_param __start___param[], __stop___param[];
496 smp_setup_processor_id();
499 * Need to run as early as possible, to initialize the
506 early_boot_irqs_off();
507 early_init_irq_lock_class();
510 * Interrupts are still disabled. Do necessary setups, then
517 printk(linux_banner);
518 setup_arch(&command_line);
520 setup_per_cpu_areas();
521 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
524 * Set up the scheduler prior starting any interrupts (such as the
525 * timer interrupt). Full topology setup happens at smp_init()
526 * time - but meanwhile we still have a functioning scheduler.
530 * Disable preemption - early bootup scheduling is extremely
531 * fragile until we cpu_idle() for the first time.
534 build_all_zonelists();
536 printk(KERN_NOTICE "Kernel command line: %s\n", saved_command_line);
538 parse_args("Booting kernel", command_line, __start___param,
539 __stop___param - __start___param,
540 &unknown_bootoption);
552 if (!irqs_disabled())
553 printk("start_kernel(): bug: interrupts were enabled early\n");
554 early_boot_irqs_on();
558 * HACK ALERT! This is early. We're enabling the console before
559 * we've done PCI setups etc, and console_init() must be aware of
560 * this. But we do want output early, in case something goes wrong.
564 panic(panic_later, panic_param);
569 * Need to run this when irqs are enabled, because it wants
570 * to self-test [hard/soft]-irqs on/off lock inversion bugs
575 #ifdef CONFIG_BLK_DEV_INITRD
576 if (initrd_start && !initrd_below_start_ok &&
577 initrd_start < min_low_pfn << PAGE_SHIFT) {
578 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
579 "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
583 vfs_caches_init_early();
587 setup_per_cpu_pageset();
593 pgtable_cache_init();
598 efi_enter_virtual_mode();
600 fork_init(num_physpages);
606 vfs_caches_init(num_physpages);
609 /* rootfs populating might need page-writeback */
610 page_writeback_init();
611 #ifdef CONFIG_PROC_FS
615 taskstats_init_early();
620 acpi_early_init(); /* before LAPIC and SMP init */
622 /* Do the rest non-__init'ed, we're now alive */
626 static int __initdata initcall_debug;
628 static int __init initcall_debug_setup(char *str)
633 __setup("initcall_debug", initcall_debug_setup);
635 extern initcall_t __initcall_start[], __initcall_end[];
637 static void __init do_initcalls(void)
640 int count = preempt_count();
642 for (call = __initcall_start; call < __initcall_end; call++) {
647 if (initcall_debug) {
648 printk("Calling initcall 0x%p", *call);
649 print_fn_descriptor_symbol(": %s()",
650 (unsigned long) *call);
656 if (result && result != -ENODEV && initcall_debug) {
657 sprintf(msgbuf, "error code %d", result);
660 if (preempt_count() != count) {
661 msg = "preemption imbalance";
662 preempt_count() = count;
664 if (irqs_disabled()) {
665 msg = "disabled interrupts";
669 printk(KERN_WARNING "initcall at 0x%p", *call);
670 print_fn_descriptor_symbol(": %s()",
671 (unsigned long) *call);
672 printk(": returned with %s\n", msg);
676 /* Make sure there is no pending stuff from the initcall sequence */
677 flush_scheduled_work();
681 * Ok, the machine is now initialized. None of the devices
682 * have been touched yet, but the CPU subsystem is up and
683 * running, and memory and process management works.
685 * Now we can finally start doing some real work..
687 static void __init do_basic_setup(void)
689 /* drivers will send hotplug events */
691 usermodehelper_init();
701 static void do_pre_smp_initcalls(void)
703 extern int spawn_ksoftirqd(void);
705 extern int migration_init(void);
710 spawn_softlockup_task();
713 static void run_init_process(char *init_filename)
715 argv_init[0] = init_filename;
716 kernel_execve(init_filename, argv_init, envp_init);
719 static int init(void * unused)
723 * init can run on any cpu.
725 set_cpus_allowed(current, CPU_MASK_ALL);
727 * Tell the world that we're going to be the grim
728 * reaper of innocent orphaned children.
730 * We don't want people to have to make incorrect
731 * assumptions about where in the task array this
734 init_pid_ns.child_reaper = current;
736 cad_pid = task_pid(current);
738 smp_prepare_cpus(max_cpus);
740 do_pre_smp_initcalls();
750 * check if there is an early userspace init. If yes, let it do all
754 if (!ramdisk_execute_command)
755 ramdisk_execute_command = "/init";
757 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
758 ramdisk_execute_command = NULL;
763 * Ok, we have completed the initial bootup, and
764 * we're essentially up and running. Get rid of the
765 * initmem segments and start the user-mode stuff..
770 system_state = SYSTEM_RUNNING;
771 numa_default_policy();
773 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
774 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
779 if (ramdisk_execute_command) {
780 run_init_process(ramdisk_execute_command);
781 printk(KERN_WARNING "Failed to execute %s\n",
782 ramdisk_execute_command);
786 * We try each of these until one succeeds.
788 * The Bourne shell can be used instead of init if we are
789 * trying to recover a really broken machine.
791 if (execute_command) {
792 run_init_process(execute_command);
793 printk(KERN_WARNING "Failed to execute %s. Attempting "
794 "defaults...\n", execute_command);
796 run_init_process("/sbin/init");
797 run_init_process("/etc/init");
798 run_init_process("/bin/init");
799 run_init_process("/bin/sh");
801 panic("No init found. Try passing init= option to kernel.");