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/kernel_stat.h>
35 #include <linux/security.h>
36 #include <linux/workqueue.h>
37 #include <linux/profile.h>
38 #include <linux/rcupdate.h>
39 #include <linux/moduleparam.h>
40 #include <linux/kallsyms.h>
41 #include <linux/writeback.h>
42 #include <linux/cpu.h>
43 #include <linux/efi.h>
44 #include <linux/unistd.h>
45 #include <linux/rmap.h>
46 #include <linux/mempolicy.h>
50 #include <asm/setup.h>
53 * This is one of the first .c files built. Error out early
54 * if we have compiler trouble..
56 #if __GNUC__ == 2 && __GNUC_MINOR__ == 96
57 #ifdef CONFIG_FRAME_POINTER
58 #error This compiler cannot compile correctly with frame pointers enabled
62 #ifdef CONFIG_X86_LOCAL_APIC
67 * Versions of gcc older than that listed below may actually compile
68 * and link okay, but the end product can have subtle run time bugs.
69 * To avoid associated bogus bug reports, we flatly refuse to compile
70 * with a gcc that is known to be too old from the very beginning.
72 #if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 95)
73 #error Sorry, your GCC is too old. It builds incorrect kernels.
76 extern char *linux_banner;
78 static int init(void *);
80 extern void init_IRQ(void);
81 extern void sock_init(void);
82 extern void fork_init(unsigned long);
83 extern void mca_init(void);
84 extern void sbus_init(void);
85 extern void sysctl_init(void);
86 extern void signals_init(void);
87 extern void buffer_init(void);
88 extern void pidhash_init(void);
89 extern void pidmap_init(void);
90 extern void prio_tree_init(void);
91 extern void radix_tree_init(void);
92 extern void free_initmem(void);
93 extern void populate_rootfs(void);
94 extern void driver_init(void);
95 extern void prepare_namespace(void);
98 extern void tc_init(void);
101 enum system_states system_state;
102 EXPORT_SYMBOL(system_state);
105 * Boot command-line arguments
107 #define MAX_INIT_ARGS 8
108 #define MAX_INIT_ENVS 8
110 extern void time_init(void);
111 /* Default late time init is NULL. archs can override this later. */
112 void (*late_time_init)(void);
113 extern void softirq_init(void);
115 /* Untouched command line (eg. for /proc) saved by arch-specific code. */
116 char saved_command_line[COMMAND_LINE_SIZE];
118 static char *execute_command;
120 /* Setup configured maximum number of CPUs to activate */
121 static unsigned int max_cpus = NR_CPUS;
124 * Setup routine for controlling SMP activation
126 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
127 * activation entirely (the MPS table probe still happens, though).
129 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
130 * greater than 0, limits the maximum number of CPUs activated in
133 static int __init nosmp(char *str)
139 __setup("nosmp", nosmp);
141 static int __init maxcpus(char *str)
143 get_option(&str, &max_cpus);
147 __setup("maxcpus=", maxcpus);
149 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
150 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
151 static const char *panic_later, *panic_param;
153 __setup("profile=", profile_setup);
155 static int __init obsolete_checksetup(char *line)
157 struct obs_kernel_param *p;
158 extern struct obs_kernel_param __setup_start, __setup_end;
162 int n = strlen(p->str);
163 if (!strncmp(line, p->str, n)) {
165 /* Already done in parse_early_param? (Needs
166 * exact match on param part) */
167 if (line[n] == '\0' || line[n] == '=')
169 } else if (!p->setup_func) {
170 printk(KERN_WARNING "Parameter %s is obsolete,"
171 " ignored\n", p->str);
173 } else if (p->setup_func(line + n))
177 } while (p < &__setup_end);
181 /* this should be approx 2 Bo*oMips to start (note initial shift), and will
182 still work even if initially too large, it will just take slightly longer */
183 unsigned long loops_per_jiffy = (1<<12);
185 EXPORT_SYMBOL(loops_per_jiffy);
187 /* This is the number of bits of precision for the loops_per_jiffy. Each
188 bit takes on average 1.5/HZ seconds. This (like the original) is a little
192 void __devinit calibrate_delay(void)
194 unsigned long ticks, loopbit;
195 int lps_precision = LPS_PREC;
197 loops_per_jiffy = (1<<12);
199 printk("Calibrating delay loop... ");
200 while ((loops_per_jiffy <<= 1) != 0) {
201 /* wait for "start of" clock tick */
203 while (ticks == jiffies)
207 __delay(loops_per_jiffy);
208 ticks = jiffies - ticks;
213 /* Do a binary approximation to get loops_per_jiffy set to equal one clock
214 (up to lps_precision bits) */
215 loops_per_jiffy >>= 1;
216 loopbit = loops_per_jiffy;
217 while ( lps_precision-- && (loopbit >>= 1) ) {
218 loops_per_jiffy |= loopbit;
220 while (ticks == jiffies);
222 __delay(loops_per_jiffy);
223 if (jiffies != ticks) /* longer than 1 tick */
224 loops_per_jiffy &= ~loopbit;
227 /* Round the value and print it */
228 printk("%lu.%02lu BogoMIPS\n",
229 loops_per_jiffy/(500000/HZ),
230 (loops_per_jiffy/(5000/HZ)) % 100);
233 static int __init debug_kernel(char *str)
237 console_loglevel = 10;
241 static int __init quiet_kernel(char *str)
245 console_loglevel = 4;
249 __setup("debug", debug_kernel);
250 __setup("quiet", quiet_kernel);
252 /* Unknown boot options get handed to init, unless they look like
254 static int __init unknown_bootoption(char *param, char *val)
256 /* Change NUL term back to "=", to make "param" the whole string. */
260 /* Handle obsolete-style parameters */
261 if (obsolete_checksetup(param))
264 /* Preemptive maintenance for "why didn't my mispelled command
266 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
267 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
275 /* Environment option */
277 for (i = 0; envp_init[i]; i++) {
278 if (i == MAX_INIT_ENVS) {
279 panic_later = "Too many boot env vars at `%s'";
282 if (!strncmp(param, envp_init[i], val - param))
285 envp_init[i] = param;
287 /* Command line option */
289 for (i = 0; argv_init[i]; i++) {
290 if (i == MAX_INIT_ARGS) {
291 panic_later = "Too many boot init vars at `%s'";
295 argv_init[i] = param;
300 static int __init init_setup(char *str)
304 execute_command = str;
305 /* In case LILO is going to boot us with default command line,
306 * it prepends "auto" before the whole cmdline which makes
307 * the shell think it should execute a script with such name.
308 * So we ignore all arguments entered _before_ init=... [MJ]
310 for (i = 1; i < MAX_INIT_ARGS; i++)
314 __setup("init=", init_setup);
316 extern void setup_arch(char **);
317 extern void cpu_idle(void);
321 #ifdef CONFIG_X86_LOCAL_APIC
322 static void __init smp_init(void)
324 APIC_init_uniprocessor();
327 #define smp_init() do { } while (0)
330 static inline void setup_per_cpu_areas(void) { }
331 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
335 #ifdef __GENERIC_PER_CPU
336 unsigned long __per_cpu_offset[NR_CPUS];
338 EXPORT_SYMBOL(__per_cpu_offset);
340 static void __init setup_per_cpu_areas(void)
342 unsigned long size, i;
344 /* Created by linker magic */
345 extern char __per_cpu_start[], __per_cpu_end[];
347 /* Copy section for each CPU (we discard the original) */
348 size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
349 #ifdef CONFIG_MODULES
350 if (size < PERCPU_ENOUGH_ROOM)
351 size = PERCPU_ENOUGH_ROOM;
354 ptr = alloc_bootmem(size * NR_CPUS);
356 for (i = 0; i < NR_CPUS; i++, ptr += size) {
357 __per_cpu_offset[i] = ptr - __per_cpu_start;
358 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
361 #endif /* !__GENERIC_PER_CPU */
363 /* Called by boot processor to activate the rest. */
364 static void __init smp_init(void)
368 /* FIXME: This should be done in userspace --RR */
369 for_each_present_cpu(i) {
370 if (num_online_cpus() >= max_cpus)
376 /* Any cleanup work */
377 printk("Brought up %ld CPUs\n", (long)num_online_cpus());
378 smp_cpus_done(max_cpus);
380 /* Get other processors into their bootup holding patterns. */
390 * We need to finalize in a non-__init function or else race conditions
391 * between the root thread and the init thread may cause start_kernel to
392 * be reaped by free_initmem before the root thread has proceeded to
395 * gcc-3.4 accidentally inlines this function, so use noinline.
398 static void noinline rest_init(void)
400 kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
401 numa_default_policy();
406 /* Check for early params. */
407 static int __init do_early_param(char *param, char *val)
409 struct obs_kernel_param *p;
410 extern struct obs_kernel_param __setup_start, __setup_end;
412 for (p = &__setup_start; p < &__setup_end; p++) {
413 if (p->early && strcmp(param, p->str) == 0) {
414 if (p->setup_func(val) != 0)
416 "Malformed early option '%s'\n", param);
419 /* We accept everything at this stage. */
423 /* Arch code calls this early on, or if not, just before other parsing. */
424 void __init parse_early_param(void)
426 static __initdata int done = 0;
427 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
432 /* All fall through to do_early_param. */
433 strlcpy(tmp_cmdline, saved_command_line, COMMAND_LINE_SIZE);
434 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
439 * Activate the first processor.
442 asmlinkage void __init start_kernel(void)
445 extern struct kernel_param __start___param[], __stop___param[];
447 * Interrupts are still disabled. Do necessary setups, then
452 printk(linux_banner);
453 setup_arch(&command_line);
454 setup_per_cpu_areas();
457 * Mark the boot cpu "online" so that it can call console drivers in
458 * printk() and can access its per-cpu storage.
460 smp_prepare_boot_cpu();
463 * Set up the scheduler prior starting any interrupts (such as the
464 * timer interrupt). Full topology setup happens at smp_init()
465 * time - but meanwhile we still have a functioning scheduler.
469 build_all_zonelists();
471 printk("Kernel command line: %s\n", saved_command_line);
473 parse_args("Booting kernel", command_line, __start___param,
474 __stop___param - __start___param,
475 &unknown_bootoption);
486 * HACK ALERT! This is early. We're enabling the console before
487 * we've done PCI setups etc, and console_init() must be aware of
488 * this. But we do want output early, in case something goes wrong.
492 panic(panic_later, panic_param);
495 #ifdef CONFIG_BLK_DEV_INITRD
496 if (initrd_start && !initrd_below_start_ok &&
497 initrd_start < min_low_pfn << PAGE_SHIFT) {
498 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
499 "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
503 vfs_caches_init_early();
511 pgtable_cache_init();
516 efi_enter_virtual_mode();
518 fork_init(num_physpages);
522 security_scaffolding_startup();
523 vfs_caches_init(num_physpages);
526 /* rootfs populating might need page-writeback */
527 page_writeback_init();
528 #ifdef CONFIG_PROC_FS
534 * We count on the initial thread going ok
535 * Like idlers init is an unlocked kernel thread, which will
536 * make syscalls (and thus be locked).
538 init_idle(current, smp_processor_id());
540 /* Do the rest non-__init'ed, we're now alive */
544 static int __initdata initcall_debug;
546 static int __init initcall_debug_setup(char *str)
551 __setup("initcall_debug", initcall_debug_setup);
553 struct task_struct *child_reaper = &init_task;
555 extern initcall_t __initcall_start, __initcall_end;
557 static void __init do_initcalls(void)
560 int count = preempt_count();
562 for (call = &__initcall_start; call < &__initcall_end; call++) {
565 if (initcall_debug) {
566 printk(KERN_DEBUG "Calling initcall 0x%p", *call);
567 print_symbol(": %s()", (unsigned long) *call);
574 if (preempt_count() != count) {
575 msg = "preemption imbalance";
576 preempt_count() = count;
578 if (irqs_disabled()) {
579 msg = "disabled interrupts";
583 printk("error in initcall at 0x%p: "
584 "returned with %s\n", *call, msg);
588 /* Make sure there is no pending stuff from the initcall sequence */
589 flush_scheduled_work();
593 * Ok, the machine is now initialized. None of the devices
594 * have been touched yet, but the CPU subsystem is up and
595 * running, and memory and process management works.
597 * Now we can finally start doing some real work..
599 static void __init do_basic_setup(void)
607 /* Networking initialization needs a process context */
614 static void do_pre_smp_initcalls(void)
616 extern int spawn_ksoftirqd(void);
618 extern int migration_init(void);
625 static void run_init_process(char *init_filename)
627 argv_init[0] = init_filename;
628 execve(init_filename, argv_init, envp_init);
631 static inline void fixup_cpu_present_map(void)
637 * If arch is not hotplug ready and did not populate
638 * cpu_present_map, just make cpu_present_map same as cpu_possible_map
639 * for other cpu bringup code to function as normal. e.g smp_init() etc.
641 if (cpus_empty(cpu_present_map)) {
643 cpu_set(i, cpu_present_map);
649 static int init(void * unused)
653 * Tell the world that we're going to be the grim
654 * reaper of innocent orphaned children.
656 * We don't want people to have to make incorrect
657 * assumptions about where in the task array this
660 child_reaper = current;
662 /* Sets up cpus_possible() */
663 smp_prepare_cpus(max_cpus);
665 do_pre_smp_initcalls();
667 fixup_cpu_present_map();
672 * Do this before initcalls, because some drivers want to access
680 * check if there is an early userspace init. If yes, let it do all
683 if (sys_access((const char __user *) "/init", 0) == 0)
684 execute_command = "/init";
689 * Ok, we have completed the initial bootup, and
690 * we're essentially up and running. Get rid of the
691 * initmem segments and start the user-mode stuff..
695 system_state = SYSTEM_RUNNING;
696 numa_default_policy();
698 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
699 printk("Warning: unable to open an initial console.\n");
705 * We try each of these until one succeeds.
707 * The Bourne shell can be used instead of init if we are
708 * trying to recover a really broken machine.
712 run_init_process(execute_command);
714 run_init_process("/sbin/init");
715 run_init_process("/etc/init");
716 run_init_process("/bin/init");
717 run_init_process("/bin/sh");
719 panic("No init found. Try passing init= option to kernel.");
722 static int early_param_test(char *rest)
724 printk("early_parm_test: %s\n", rest ?: "(null)");
725 return rest ? 0 : -EINVAL;
727 early_param("testsetup", early_param_test);
728 static int early_setup_test(char *rest)
730 printk("early_setup_test: %s\n", rest ?: "(null)");
733 __setup("testsetup_long", early_setup_test);