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/efi.h>
45 #include <linux/unistd.h>
46 #include <linux/rmap.h>
47 #include <linux/mempolicy.h>
51 #include <asm/setup.h>
54 * This is one of the first .c files built. Error out early
55 * if we have compiler trouble..
57 #if __GNUC__ == 2 && __GNUC_MINOR__ == 96
58 #ifdef CONFIG_FRAME_POINTER
59 #error This compiler cannot compile correctly with frame pointers enabled
63 #ifdef CONFIG_X86_LOCAL_APIC
68 * Versions of gcc older than that listed below may actually compile
69 * and link okay, but the end product can have subtle run time bugs.
70 * To avoid associated bogus bug reports, we flatly refuse to compile
71 * with a gcc that is known to be too old from the very beginning.
73 #if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 95)
74 #error Sorry, your GCC is too old. It builds incorrect kernels.
77 extern char *linux_banner;
79 static int init(void *);
81 extern void init_IRQ(void);
82 extern void sock_init(void);
83 extern void fork_init(unsigned long);
84 extern void mca_init(void);
85 extern void sbus_init(void);
86 extern void sysctl_init(void);
87 extern void signals_init(void);
88 extern void buffer_init(void);
89 extern void pidhash_init(void);
90 extern void pidmap_init(void);
91 extern void prio_tree_init(void);
92 extern void radix_tree_init(void);
93 extern void free_initmem(void);
94 extern void populate_rootfs(void);
95 extern void driver_init(void);
96 extern void prepare_namespace(void);
98 extern void acpi_early_init(void);
100 static inline void acpi_early_init(void) { }
104 extern void tc_init(void);
107 enum system_states system_state;
108 EXPORT_SYMBOL(system_state);
111 * Boot command-line arguments
113 #define MAX_INIT_ARGS 32
114 #define MAX_INIT_ENVS 32
116 extern void time_init(void);
117 /* Default late time init is NULL. archs can override this later. */
118 void (*late_time_init)(void);
119 extern void softirq_init(void);
121 /* Untouched command line (eg. for /proc) saved by arch-specific code. */
122 char saved_command_line[COMMAND_LINE_SIZE];
124 static char *execute_command;
126 /* Setup configured maximum number of CPUs to activate */
127 static unsigned int max_cpus = NR_CPUS;
130 * Setup routine for controlling SMP activation
132 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
133 * activation entirely (the MPS table probe still happens, though).
135 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
136 * greater than 0, limits the maximum number of CPUs activated in
139 static int __init nosmp(char *str)
145 __setup("nosmp", nosmp);
147 static int __init maxcpus(char *str)
149 get_option(&str, &max_cpus);
153 __setup("maxcpus=", maxcpus);
155 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
156 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
157 static const char *panic_later, *panic_param;
159 static int __init obsolete_checksetup(char *line)
161 struct obs_kernel_param *p;
162 extern struct obs_kernel_param __setup_start, __setup_end;
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);
185 static unsigned long preset_lpj;
186 static int __init lpj_setup(char *str)
188 preset_lpj = simple_strtoul(str,NULL,0);
192 __setup("lpj=", lpj_setup);
195 * This should be approx 2 Bo*oMips to start (note initial shift), and will
196 * still work even if initially too large, it will just take slightly longer
198 unsigned long loops_per_jiffy = (1<<12);
200 EXPORT_SYMBOL(loops_per_jiffy);
203 * This is the number of bits of precision for the loops_per_jiffy. Each
204 * bit takes on average 1.5/HZ seconds. This (like the original) is a little
209 void __devinit calibrate_delay(void)
211 unsigned long ticks, loopbit;
212 int lps_precision = LPS_PREC;
215 loops_per_jiffy = preset_lpj;
216 printk("Calibrating delay loop (skipped)... "
217 "%lu.%02lu BogoMIPS preset\n",
218 loops_per_jiffy/(500000/HZ),
219 (loops_per_jiffy/(5000/HZ)) % 100);
221 loops_per_jiffy = (1<<12);
223 printk(KERN_DEBUG "Calibrating delay loop... ");
224 while ((loops_per_jiffy <<= 1) != 0) {
225 /* wait for "start of" clock tick */
227 while (ticks == jiffies)
231 __delay(loops_per_jiffy);
232 ticks = jiffies - ticks;
238 * Do a binary approximation to get loops_per_jiffy set to
239 * equal one clock (up to lps_precision bits)
241 loops_per_jiffy >>= 1;
242 loopbit = loops_per_jiffy;
243 while (lps_precision-- && (loopbit >>= 1)) {
244 loops_per_jiffy |= loopbit;
246 while (ticks == jiffies)
249 __delay(loops_per_jiffy);
250 if (jiffies != ticks) /* longer than 1 tick */
251 loops_per_jiffy &= ~loopbit;
254 /* Round the value and print it */
255 printk("%lu.%02lu BogoMIPS (lpj=%lu)\n",
256 loops_per_jiffy/(500000/HZ),
257 (loops_per_jiffy/(5000/HZ)) % 100,
263 static int __init debug_kernel(char *str)
267 console_loglevel = 10;
271 static int __init quiet_kernel(char *str)
275 console_loglevel = 4;
279 __setup("debug", debug_kernel);
280 __setup("quiet", quiet_kernel);
283 * Unknown boot options get handed to init, unless they look like
286 static int __init unknown_bootoption(char *param, char *val)
288 /* Change NUL term back to "=", to make "param" the whole string. */
290 if (val[-1] == '"') val[-2] = '=';
294 /* Handle obsolete-style parameters */
295 if (obsolete_checksetup(param))
299 * Preemptive maintenance for "why didn't my mispelled command
302 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
303 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
311 /* Environment option */
313 for (i = 0; envp_init[i]; i++) {
314 if (i == MAX_INIT_ENVS) {
315 panic_later = "Too many boot env vars at `%s'";
318 if (!strncmp(param, envp_init[i], val - param))
321 envp_init[i] = param;
323 /* Command line option */
325 for (i = 0; argv_init[i]; i++) {
326 if (i == MAX_INIT_ARGS) {
327 panic_later = "Too many boot init vars at `%s'";
331 argv_init[i] = param;
336 static int __init init_setup(char *str)
340 execute_command = str;
342 * In case LILO is going to boot us with default command line,
343 * it prepends "auto" before the whole cmdline which makes
344 * the shell think it should execute a script with such name.
345 * So we ignore all arguments entered _before_ init=... [MJ]
347 for (i = 1; i < MAX_INIT_ARGS; i++)
351 __setup("init=", init_setup);
353 extern void setup_arch(char **);
354 extern void cpu_idle(void);
358 #ifdef CONFIG_X86_LOCAL_APIC
359 static void __init smp_init(void)
361 APIC_init_uniprocessor();
364 #define smp_init() do { } while (0)
367 static inline void setup_per_cpu_areas(void) { }
368 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
372 #ifdef __GENERIC_PER_CPU
373 unsigned long __per_cpu_offset[NR_CPUS];
375 EXPORT_SYMBOL(__per_cpu_offset);
377 static void __init setup_per_cpu_areas(void)
379 unsigned long size, i;
381 /* Created by linker magic */
382 extern char __per_cpu_start[], __per_cpu_end[];
384 /* Copy section for each CPU (we discard the original) */
385 size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
386 #ifdef CONFIG_MODULES
387 if (size < PERCPU_ENOUGH_ROOM)
388 size = PERCPU_ENOUGH_ROOM;
391 ptr = alloc_bootmem(size * NR_CPUS);
393 for (i = 0; i < NR_CPUS; i++, ptr += size) {
394 __per_cpu_offset[i] = ptr - __per_cpu_start;
395 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
398 #endif /* !__GENERIC_PER_CPU */
400 /* Called by boot processor to activate the rest. */
401 static void __init smp_init(void)
405 /* FIXME: This should be done in userspace --RR */
406 for_each_present_cpu(i) {
407 if (num_online_cpus() >= max_cpus)
413 /* Any cleanup work */
414 printk("Brought up %ld CPUs\n", (long)num_online_cpus());
415 smp_cpus_done(max_cpus);
417 /* Get other processors into their bootup holding patterns. */
427 * We need to finalize in a non-__init function or else race conditions
428 * between the root thread and the init thread may cause start_kernel to
429 * be reaped by free_initmem before the root thread has proceeded to
432 * gcc-3.4 accidentally inlines this function, so use noinline.
435 static void noinline rest_init(void)
437 kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
438 numa_default_policy();
443 /* Check for early params. */
444 static int __init do_early_param(char *param, char *val)
446 struct obs_kernel_param *p;
447 extern struct obs_kernel_param __setup_start, __setup_end;
449 for (p = &__setup_start; p < &__setup_end; p++) {
450 if (p->early && strcmp(param, p->str) == 0) {
451 if (p->setup_func(val) != 0)
453 "Malformed early option '%s'\n", param);
456 /* We accept everything at this stage. */
460 /* Arch code calls this early on, or if not, just before other parsing. */
461 void __init parse_early_param(void)
463 static __initdata int done = 0;
464 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
469 /* All fall through to do_early_param. */
470 strlcpy(tmp_cmdline, saved_command_line, COMMAND_LINE_SIZE);
471 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
476 * Activate the first processor.
479 asmlinkage void __init start_kernel(void)
482 extern struct kernel_param __start___param[], __stop___param[];
484 * Interrupts are still disabled. Do necessary setups, then
489 printk(linux_banner);
490 setup_arch(&command_line);
491 setup_per_cpu_areas();
494 * Mark the boot cpu "online" so that it can call console drivers in
495 * printk() and can access its per-cpu storage.
497 smp_prepare_boot_cpu();
500 * Set up the scheduler prior starting any interrupts (such as the
501 * timer interrupt). Full topology setup happens at smp_init()
502 * time - but meanwhile we still have a functioning scheduler.
505 build_all_zonelists();
507 printk("Kernel command line: %s\n", saved_command_line);
509 parse_args("Booting kernel", command_line, __start___param,
510 __stop___param - __start___param,
511 &unknown_bootoption);
522 * HACK ALERT! This is early. We're enabling the console before
523 * we've done PCI setups etc, and console_init() must be aware of
524 * this. But we do want output early, in case something goes wrong.
528 panic(panic_later, panic_param);
531 #ifdef CONFIG_BLK_DEV_INITRD
532 if (initrd_start && !initrd_below_start_ok &&
533 initrd_start < min_low_pfn << PAGE_SHIFT) {
534 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
535 "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
539 vfs_caches_init_early();
547 pgtable_cache_init();
552 efi_enter_virtual_mode();
554 fork_init(num_physpages);
558 security_scaffolding_startup();
559 vfs_caches_init(num_physpages);
562 /* rootfs populating might need page-writeback */
563 page_writeback_init();
564 #ifdef CONFIG_PROC_FS
569 acpi_early_init(); /* before LAPIC and SMP init */
571 /* Do the rest non-__init'ed, we're now alive */
575 static int __initdata initcall_debug;
577 static int __init initcall_debug_setup(char *str)
582 __setup("initcall_debug", initcall_debug_setup);
584 struct task_struct *child_reaper = &init_task;
586 extern initcall_t __initcall_start, __initcall_end;
588 static void __init do_initcalls(void)
591 int count = preempt_count();
593 for (call = &__initcall_start; call < &__initcall_end; call++) {
596 if (initcall_debug) {
597 printk(KERN_DEBUG "Calling initcall 0x%p", *call);
598 print_symbol(": %s()", (unsigned long) *call);
605 if (preempt_count() != count) {
606 msg = "preemption imbalance";
607 preempt_count() = count;
609 if (irqs_disabled()) {
610 msg = "disabled interrupts";
614 printk("error in initcall at 0x%p: "
615 "returned with %s\n", *call, msg);
619 /* Make sure there is no pending stuff from the initcall sequence */
620 flush_scheduled_work();
624 * Ok, the machine is now initialized. None of the devices
625 * have been touched yet, but the CPU subsystem is up and
626 * running, and memory and process management works.
628 * Now we can finally start doing some real work..
630 static void __init do_basic_setup(void)
632 /* drivers will send hotplug events */
634 usermodehelper_init();
642 /* Networking initialization needs a process context */
648 static void do_pre_smp_initcalls(void)
650 extern int spawn_ksoftirqd(void);
652 extern int migration_init(void);
659 static void run_init_process(char *init_filename)
661 argv_init[0] = init_filename;
662 execve(init_filename, argv_init, envp_init);
665 static inline void fixup_cpu_present_map(void)
671 * If arch is not hotplug ready and did not populate
672 * cpu_present_map, just make cpu_present_map same as cpu_possible_map
673 * for other cpu bringup code to function as normal. e.g smp_init() etc.
675 if (cpus_empty(cpu_present_map)) {
677 cpu_set(i, cpu_present_map);
683 static int init(void * unused)
687 * Tell the world that we're going to be the grim
688 * reaper of innocent orphaned children.
690 * We don't want people to have to make incorrect
691 * assumptions about where in the task array this
694 child_reaper = current;
696 /* Sets up cpus_possible() */
697 smp_prepare_cpus(max_cpus);
699 do_pre_smp_initcalls();
701 fixup_cpu_present_map();
706 * Do this before initcalls, because some drivers want to access
714 * check if there is an early userspace init. If yes, let it do all
717 if (sys_access((const char __user *) "/init", 0) == 0)
718 execute_command = "/init";
723 * Ok, we have completed the initial bootup, and
724 * we're essentially up and running. Get rid of the
725 * initmem segments and start the user-mode stuff..
729 system_state = SYSTEM_RUNNING;
730 numa_default_policy();
732 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
733 printk("Warning: unable to open an initial console.\n");
739 * We try each of these until one succeeds.
741 * The Bourne shell can be used instead of init if we are
742 * trying to recover a really broken machine.
746 run_init_process(execute_command);
748 run_init_process("/sbin/init");
749 run_init_process("/etc/init");
750 run_init_process("/bin/init");
751 run_init_process("/bin/sh");
753 panic("No init found. Try passing init= option to kernel.");
756 static int early_param_test(char *rest)
758 printk("early_parm_test: %s\n", rest ?: "(null)");
759 return rest ? 0 : -EINVAL;
761 early_param("testsetup", early_param_test);
762 static int early_setup_test(char *rest)
764 printk("early_setup_test: %s\n", rest ?: "(null)");
767 __setup("testsetup_long", early_setup_test);