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 #define __KERNEL_SYSCALLS__
14 #include <linux/config.h>
15 #include <linux/types.h>
16 #include <linux/module.h>
17 #include <linux/proc_fs.h>
18 #include <linux/devfs_fs_kernel.h>
19 #include <linux/kernel.h>
20 #include <linux/syscalls.h>
21 #include <linux/string.h>
22 #include <linux/ctype.h>
23 #include <linux/delay.h>
24 #include <linux/utsname.h>
25 #include <linux/ioport.h>
26 #include <linux/init.h>
27 #include <linux/smp_lock.h>
28 #include <linux/initrd.h>
29 #include <linux/hdreg.h>
30 #include <linux/bootmem.h>
31 #include <linux/tty.h>
32 #include <linux/gfp.h>
33 #include <linux/percpu.h>
34 #include <linux/kmod.h>
35 #include <linux/kernel_stat.h>
36 #include <linux/security.h>
37 #include <linux/workqueue.h>
38 #include <linux/profile.h>
39 #include <linux/rcupdate.h>
40 #include <linux/moduleparam.h>
41 #include <linux/kallsyms.h>
42 #include <linux/writeback.h>
43 #include <linux/cpu.h>
44 #include <linux/cpuset.h>
45 #include <linux/efi.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>
55 #include <asm/setup.h>
56 #include <asm/sections.h>
57 #include <asm/cacheflush.h>
59 #ifdef CONFIG_X86_LOCAL_APIC
64 * This is one of the first .c files built. Error out early if we have compiler
67 * Versions of gcc older than that listed below may actually compile and link
68 * okay, but the end product can have subtle run time bugs. To avoid associated
69 * bogus bug reports, we flatly refuse to compile with a gcc that is known to be
70 * too old from the very beginning.
72 #if (__GNUC__ < 3) || (__GNUC__ == 3 && __GNUC_MINOR__ < 2)
73 #error Sorry, your GCC is too old. It builds incorrect kernels.
76 static int init(void *);
78 extern void init_IRQ(void);
79 extern void fork_init(unsigned long);
80 extern void mca_init(void);
81 extern void sbus_init(void);
82 extern void sysctl_init(void);
83 extern void signals_init(void);
84 extern void pidhash_init(void);
85 extern void pidmap_init(void);
86 extern void prio_tree_init(void);
87 extern void radix_tree_init(void);
88 extern void free_initmem(void);
89 extern void populate_rootfs(void);
90 extern void driver_init(void);
91 extern void prepare_namespace(void);
93 extern void acpi_early_init(void);
95 static inline void acpi_early_init(void) { }
97 #ifndef CONFIG_DEBUG_RODATA
98 static inline void mark_rodata_ro(void) { }
102 extern void tc_init(void);
105 enum system_states system_state;
106 EXPORT_SYMBOL(system_state);
109 * Boot command-line arguments
111 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
112 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
114 extern void time_init(void);
115 /* Default late time init is NULL. archs can override this later. */
116 void (*late_time_init)(void);
117 extern void softirq_init(void);
119 /* Untouched command line (eg. for /proc) saved by arch-specific code. */
120 char saved_command_line[COMMAND_LINE_SIZE];
122 static char *execute_command;
123 static char *ramdisk_execute_command;
125 /* Setup configured maximum number of CPUs to activate */
126 static unsigned int max_cpus = NR_CPUS;
129 * Setup routine for controlling SMP activation
131 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
132 * activation entirely (the MPS table probe still happens, though).
134 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
135 * greater than 0, limits the maximum number of CPUs activated in
138 static int __init nosmp(char *str)
144 __setup("nosmp", nosmp);
146 static int __init maxcpus(char *str)
148 get_option(&str, &max_cpus);
152 __setup("maxcpus=", maxcpus);
154 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
155 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
156 static const char *panic_later, *panic_param;
158 extern struct obs_kernel_param __setup_start[], __setup_end[];
160 static int __init obsolete_checksetup(char *line)
162 struct obs_kernel_param *p;
166 int n = strlen(p->str);
167 if (!strncmp(line, p->str, n)) {
169 /* Already done in parse_early_param? (Needs
170 * exact match on param part) */
171 if (line[n] == '\0' || line[n] == '=')
173 } else if (!p->setup_func) {
174 printk(KERN_WARNING "Parameter %s is obsolete,"
175 " ignored\n", p->str);
177 } else if (p->setup_func(line + n))
181 } while (p < __setup_end);
186 * This should be approx 2 Bo*oMips to start (note initial shift), and will
187 * still work even if initially too large, it will just take slightly longer
189 unsigned long loops_per_jiffy = (1<<12);
191 EXPORT_SYMBOL(loops_per_jiffy);
193 static int __init debug_kernel(char *str)
197 console_loglevel = 10;
201 static int __init quiet_kernel(char *str)
205 console_loglevel = 4;
209 __setup("debug", debug_kernel);
210 __setup("quiet", quiet_kernel);
212 static int __init loglevel(char *str)
214 get_option(&str, &console_loglevel);
218 __setup("loglevel=", loglevel);
221 * Unknown boot options get handed to init, unless they look like
224 static int __init unknown_bootoption(char *param, char *val)
226 /* Change NUL term back to "=", to make "param" the whole string. */
228 /* param=val or param="val"? */
229 if (val == param+strlen(param)+1)
231 else if (val == param+strlen(param)+2) {
233 memmove(val-1, val, strlen(val)+1);
239 /* Handle obsolete-style parameters */
240 if (obsolete_checksetup(param))
244 * Preemptive maintenance for "why didn't my mispelled command
247 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
248 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
256 /* Environment option */
258 for (i = 0; envp_init[i]; i++) {
259 if (i == MAX_INIT_ENVS) {
260 panic_later = "Too many boot env vars at `%s'";
263 if (!strncmp(param, envp_init[i], val - param))
266 envp_init[i] = param;
268 /* Command line option */
270 for (i = 0; argv_init[i]; i++) {
271 if (i == MAX_INIT_ARGS) {
272 panic_later = "Too many boot init vars at `%s'";
276 argv_init[i] = param;
281 static int __init init_setup(char *str)
285 execute_command = str;
287 * In case LILO is going to boot us with default command line,
288 * it prepends "auto" before the whole cmdline which makes
289 * the shell think it should execute a script with such name.
290 * So we ignore all arguments entered _before_ init=... [MJ]
292 for (i = 1; i < MAX_INIT_ARGS; i++)
296 __setup("init=", init_setup);
298 static int __init rdinit_setup(char *str)
302 ramdisk_execute_command = str;
303 /* See "auto" comment in init_setup */
304 for (i = 1; i < MAX_INIT_ARGS; i++)
308 __setup("rdinit=", rdinit_setup);
312 #ifdef CONFIG_X86_LOCAL_APIC
313 static void __init smp_init(void)
315 APIC_init_uniprocessor();
318 #define smp_init() do { } while (0)
321 static inline void setup_per_cpu_areas(void) { }
322 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
326 #ifdef __GENERIC_PER_CPU
327 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
329 EXPORT_SYMBOL(__per_cpu_offset);
331 static void __init setup_per_cpu_areas(void)
333 unsigned long size, i;
335 unsigned long nr_possible_cpus = num_possible_cpus();
337 /* Copy section for each CPU (we discard the original) */
338 size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
339 #ifdef CONFIG_MODULES
340 if (size < PERCPU_ENOUGH_ROOM)
341 size = PERCPU_ENOUGH_ROOM;
343 ptr = alloc_bootmem(size * nr_possible_cpus);
345 for_each_possible_cpu(i) {
346 __per_cpu_offset[i] = ptr - __per_cpu_start;
347 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
351 #endif /* !__GENERIC_PER_CPU */
353 /* Called by boot processor to activate the rest. */
354 static void __init smp_init(void)
358 /* FIXME: This should be done in userspace --RR */
359 for_each_present_cpu(i) {
360 if (num_online_cpus() >= max_cpus)
366 /* Any cleanup work */
367 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
368 smp_cpus_done(max_cpus);
370 /* Get other processors into their bootup holding patterns. */
379 * We need to finalize in a non-__init function or else race conditions
380 * between the root thread and the init thread may cause start_kernel to
381 * be reaped by free_initmem before the root thread has proceeded to
384 * gcc-3.4 accidentally inlines this function, so use noinline.
387 static void noinline rest_init(void)
388 __releases(kernel_lock)
390 kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
391 numa_default_policy();
395 * The boot idle thread must execute schedule()
396 * at least one to get things moving:
398 preempt_enable_no_resched();
402 /* Call into cpu_idle with preempt disabled */
406 /* Check for early params. */
407 static int __init do_early_param(char *param, char *val)
409 struct obs_kernel_param *p;
411 for (p = __setup_start; p < __setup_end; p++) {
412 if (p->early && strcmp(param, p->str) == 0) {
413 if (p->setup_func(val) != 0)
415 "Malformed early option '%s'\n", param);
418 /* We accept everything at this stage. */
422 /* Arch code calls this early on, or if not, just before other parsing. */
423 void __init parse_early_param(void)
425 static __initdata int done = 0;
426 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
431 /* All fall through to do_early_param. */
432 strlcpy(tmp_cmdline, saved_command_line, COMMAND_LINE_SIZE);
433 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
438 * Activate the first processor.
441 static void __init boot_cpu_init(void)
443 int cpu = smp_processor_id();
444 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
445 cpu_set(cpu, cpu_online_map);
446 cpu_set(cpu, cpu_present_map);
447 cpu_set(cpu, cpu_possible_map);
450 asmlinkage void __init start_kernel(void)
453 extern struct kernel_param __start___param[], __stop___param[];
455 * Interrupts are still disabled. Do necessary setups, then
462 printk(linux_banner);
463 setup_arch(&command_line);
464 setup_per_cpu_areas();
465 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
468 * Set up the scheduler prior starting any interrupts (such as the
469 * timer interrupt). Full topology setup happens at smp_init()
470 * time - but meanwhile we still have a functioning scheduler.
474 * Disable preemption - early bootup scheduling is extremely
475 * fragile until we cpu_idle() for the first time.
478 build_all_zonelists();
480 printk(KERN_NOTICE "Kernel command line: %s\n", saved_command_line);
482 parse_args("Booting kernel", command_line, __start___param,
483 __stop___param - __start___param,
484 &unknown_bootoption);
498 * HACK ALERT! This is early. We're enabling the console before
499 * we've done PCI setups etc, and console_init() must be aware of
500 * this. But we do want output early, in case something goes wrong.
504 panic(panic_later, panic_param);
507 #ifdef CONFIG_BLK_DEV_INITRD
508 if (initrd_start && !initrd_below_start_ok &&
509 initrd_start < min_low_pfn << PAGE_SHIFT) {
510 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
511 "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
515 vfs_caches_init_early();
519 setup_per_cpu_pageset();
525 pgtable_cache_init();
530 efi_enter_virtual_mode();
532 fork_init(num_physpages);
538 vfs_caches_init(num_physpages);
541 /* rootfs populating might need page-writeback */
542 page_writeback_init();
543 #ifdef CONFIG_PROC_FS
550 acpi_early_init(); /* before LAPIC and SMP init */
552 /* Do the rest non-__init'ed, we're now alive */
556 static int __initdata initcall_debug;
558 static int __init initcall_debug_setup(char *str)
563 __setup("initcall_debug", initcall_debug_setup);
565 struct task_struct *child_reaper = &init_task;
567 extern initcall_t __initcall_start[], __initcall_end[];
569 static void __init do_initcalls(void)
572 int count = preempt_count();
574 for (call = __initcall_start; call < __initcall_end; call++) {
579 if (initcall_debug) {
580 printk("Calling initcall 0x%p", *call);
581 print_fn_descriptor_symbol(": %s()",
582 (unsigned long) *call);
588 if (result && result != -ENODEV && initcall_debug) {
589 sprintf(msgbuf, "error code %d", result);
592 if (preempt_count() != count) {
593 msg = "preemption imbalance";
594 preempt_count() = count;
596 if (irqs_disabled()) {
597 msg = "disabled interrupts";
601 printk(KERN_WARNING "initcall at 0x%p", *call);
602 print_fn_descriptor_symbol(": %s()",
603 (unsigned long) *call);
604 printk(": returned with %s\n", msg);
608 /* Make sure there is no pending stuff from the initcall sequence */
609 flush_scheduled_work();
613 * Ok, the machine is now initialized. None of the devices
614 * have been touched yet, but the CPU subsystem is up and
615 * running, and memory and process management works.
617 * Now we can finally start doing some real work..
619 static void __init do_basic_setup(void)
621 /* drivers will send hotplug events */
623 usermodehelper_init();
633 static void do_pre_smp_initcalls(void)
635 extern int spawn_ksoftirqd(void);
637 extern int migration_init(void);
642 spawn_softlockup_task();
645 static void run_init_process(char *init_filename)
647 argv_init[0] = init_filename;
648 execve(init_filename, argv_init, envp_init);
651 static int init(void * unused)
655 * init can run on any cpu.
657 set_cpus_allowed(current, CPU_MASK_ALL);
659 * Tell the world that we're going to be the grim
660 * reaper of innocent orphaned children.
662 * We don't want people to have to make incorrect
663 * assumptions about where in the task array this
666 child_reaper = current;
668 smp_prepare_cpus(max_cpus);
670 do_pre_smp_initcalls();
678 * Do this before initcalls, because some drivers want to access
686 * check if there is an early userspace init. If yes, let it do all
690 if (!ramdisk_execute_command)
691 ramdisk_execute_command = "/init";
693 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
694 ramdisk_execute_command = NULL;
699 * Ok, we have completed the initial bootup, and
700 * we're essentially up and running. Get rid of the
701 * initmem segments and start the user-mode stuff..
706 system_state = SYSTEM_RUNNING;
707 numa_default_policy();
709 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
710 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
715 if (ramdisk_execute_command) {
716 run_init_process(ramdisk_execute_command);
717 printk(KERN_WARNING "Failed to execute %s\n",
718 ramdisk_execute_command);
722 * We try each of these until one succeeds.
724 * The Bourne shell can be used instead of init if we are
725 * trying to recover a really broken machine.
727 if (execute_command) {
728 run_init_process(execute_command);
729 printk(KERN_WARNING "Failed to execute %s. Attempting "
730 "defaults...\n", execute_command);
732 run_init_process("/sbin/init");
733 run_init_process("/etc/init");
734 run_init_process("/bin/init");
735 run_init_process("/bin/sh");
737 panic("No init found. Try passing init= option to kernel.");