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>
48 #include <linux/key.h>
52 #include <asm/setup.h>
55 * This is one of the first .c files built. Error out early
56 * if we have compiler trouble..
58 #if __GNUC__ == 2 && __GNUC_MINOR__ == 96
59 #ifdef CONFIG_FRAME_POINTER
60 #error This compiler cannot compile correctly with frame pointers enabled
64 #ifdef CONFIG_X86_LOCAL_APIC
69 * Versions of gcc older than that listed below may actually compile
70 * and link okay, but the end product can have subtle run time bugs.
71 * To avoid associated bogus bug reports, we flatly refuse to compile
72 * with a gcc that is known to be too old from the very beginning.
74 #if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 95)
75 #error Sorry, your GCC is too old. It builds incorrect kernels.
78 extern char *linux_banner;
80 static int init(void *);
82 extern void init_IRQ(void);
83 extern void sock_init(void);
84 extern void fork_init(unsigned long);
85 extern void mca_init(void);
86 extern void sbus_init(void);
87 extern void sysctl_init(void);
88 extern void signals_init(void);
89 extern void buffer_init(void);
90 extern void pidhash_init(void);
91 extern void pidmap_init(void);
92 extern void prio_tree_init(void);
93 extern void radix_tree_init(void);
94 extern void free_initmem(void);
95 extern void populate_rootfs(void);
96 extern void driver_init(void);
97 extern void prepare_namespace(void);
99 extern void acpi_early_init(void);
101 static inline void acpi_early_init(void) { }
105 extern void tc_init(void);
108 enum system_states system_state;
109 EXPORT_SYMBOL(system_state);
112 * The kernel_magic value represents the address of _end, which allows
113 * namelist tools to "match" each other respectively. That way a tool
114 * that looks at /dev/mem can verify that it is using the right System.map
115 * file -- if kernel_magic doesn't equal the namelist value of _end,
118 extern unsigned long _end;
119 unsigned long *kernel_magic = &_end;
122 * Boot command-line arguments
124 #define MAX_INIT_ARGS 32
125 #define MAX_INIT_ENVS 32
127 extern void time_init(void);
128 /* Default late time init is NULL. archs can override this later. */
129 void (*late_time_init)(void);
130 extern void softirq_init(void);
132 /* Untouched command line (eg. for /proc) saved by arch-specific code. */
133 char saved_command_line[COMMAND_LINE_SIZE];
135 static char *execute_command;
137 /* Setup configured maximum number of CPUs to activate */
138 static unsigned int max_cpus = NR_CPUS;
141 * Setup routine for controlling SMP activation
143 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
144 * activation entirely (the MPS table probe still happens, though).
146 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
147 * greater than 0, limits the maximum number of CPUs activated in
150 static int __init nosmp(char *str)
156 __setup("nosmp", nosmp);
158 static int __init maxcpus(char *str)
160 get_option(&str, &max_cpus);
164 __setup("maxcpus=", maxcpus);
166 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
167 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
168 static const char *panic_later, *panic_param;
170 static int __init obsolete_checksetup(char *line)
172 struct obs_kernel_param *p;
173 extern struct obs_kernel_param __setup_start, __setup_end;
177 int n = strlen(p->str);
178 if (!strncmp(line, p->str, n)) {
180 /* Already done in parse_early_param? (Needs
181 * exact match on param part) */
182 if (line[n] == '\0' || line[n] == '=')
184 } else if (!p->setup_func) {
185 printk(KERN_WARNING "Parameter %s is obsolete,"
186 " ignored\n", p->str);
188 } else if (p->setup_func(line + n))
192 } while (p < &__setup_end);
196 static unsigned long preset_lpj;
197 static int __init lpj_setup(char *str)
199 preset_lpj = simple_strtoul(str,NULL,0);
203 __setup("lpj=", lpj_setup);
206 * This should be approx 2 Bo*oMips to start (note initial shift), and will
207 * still work even if initially too large, it will just take slightly longer
209 unsigned long loops_per_jiffy = (1<<12);
211 EXPORT_SYMBOL(loops_per_jiffy);
214 * This is the number of bits of precision for the loops_per_jiffy. Each
215 * bit takes on average 1.5/HZ seconds. This (like the original) is a little
220 void __devinit calibrate_delay(void)
222 unsigned long ticks, loopbit;
223 int lps_precision = LPS_PREC;
226 loops_per_jiffy = preset_lpj;
227 printk("Calibrating delay loop (skipped)... "
228 "%lu.%02lu BogoMIPS preset\n",
229 loops_per_jiffy/(500000/HZ),
230 (loops_per_jiffy/(5000/HZ)) % 100);
232 loops_per_jiffy = (1<<12);
234 printk(KERN_DEBUG "Calibrating delay loop... ");
235 while ((loops_per_jiffy <<= 1) != 0) {
236 /* wait for "start of" clock tick */
238 while (ticks == jiffies)
242 __delay(loops_per_jiffy);
243 ticks = jiffies - ticks;
249 * Do a binary approximation to get loops_per_jiffy set to
250 * equal one clock (up to lps_precision bits)
252 loops_per_jiffy >>= 1;
253 loopbit = loops_per_jiffy;
254 while (lps_precision-- && (loopbit >>= 1)) {
255 loops_per_jiffy |= loopbit;
257 while (ticks == jiffies)
260 __delay(loops_per_jiffy);
261 if (jiffies != ticks) /* longer than 1 tick */
262 loops_per_jiffy &= ~loopbit;
265 /* Round the value and print it */
266 printk("%lu.%02lu BogoMIPS (lpj=%lu)\n",
267 loops_per_jiffy/(500000/HZ),
268 (loops_per_jiffy/(5000/HZ)) % 100,
274 static int __init debug_kernel(char *str)
278 console_loglevel = 10;
282 static int __init quiet_kernel(char *str)
286 console_loglevel = 4;
290 __setup("debug", debug_kernel);
291 __setup("quiet", quiet_kernel);
294 * Unknown boot options get handed to init, unless they look like
297 static int __init unknown_bootoption(char *param, char *val)
299 /* Change NUL term back to "=", to make "param" the whole string. */
301 /* param=val or param="val"? */
302 if (val == param+strlen(param)+1)
304 else if (val == param+strlen(param)+2) {
306 memmove(val-1, val, strlen(val)+1);
312 /* Handle obsolete-style parameters */
313 if (obsolete_checksetup(param))
317 * Preemptive maintenance for "why didn't my mispelled command
320 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
321 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
329 /* Environment option */
331 for (i = 0; envp_init[i]; i++) {
332 if (i == MAX_INIT_ENVS) {
333 panic_later = "Too many boot env vars at `%s'";
336 if (!strncmp(param, envp_init[i], val - param))
339 envp_init[i] = param;
341 /* Command line option */
343 for (i = 0; argv_init[i]; i++) {
344 if (i == MAX_INIT_ARGS) {
345 panic_later = "Too many boot init vars at `%s'";
349 argv_init[i] = param;
354 static int __init init_setup(char *str)
358 execute_command = str;
360 * In case LILO is going to boot us with default command line,
361 * it prepends "auto" before the whole cmdline which makes
362 * the shell think it should execute a script with such name.
363 * So we ignore all arguments entered _before_ init=... [MJ]
365 for (i = 1; i < MAX_INIT_ARGS; i++)
369 __setup("init=", init_setup);
371 extern void setup_arch(char **);
372 extern void cpu_idle(void);
376 #ifdef CONFIG_X86_LOCAL_APIC
377 static void __init smp_init(void)
379 APIC_init_uniprocessor();
382 #define smp_init() do { } while (0)
385 static inline void setup_per_cpu_areas(void) { }
386 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
390 #ifdef __GENERIC_PER_CPU
391 unsigned long __per_cpu_offset[NR_CPUS];
393 EXPORT_SYMBOL(__per_cpu_offset);
395 static void __init setup_per_cpu_areas(void)
397 unsigned long size, i;
399 /* Created by linker magic */
400 extern char __per_cpu_start[], __per_cpu_end[];
402 /* Copy section for each CPU (we discard the original) */
403 size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
404 #ifdef CONFIG_MODULES
405 if (size < PERCPU_ENOUGH_ROOM)
406 size = PERCPU_ENOUGH_ROOM;
409 ptr = alloc_bootmem(size * NR_CPUS);
411 for (i = 0; i < NR_CPUS; i++, ptr += size) {
412 __per_cpu_offset[i] = ptr - __per_cpu_start;
413 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
416 #endif /* !__GENERIC_PER_CPU */
418 /* Called by boot processor to activate the rest. */
419 static void __init smp_init(void)
423 /* FIXME: This should be done in userspace --RR */
424 for_each_present_cpu(i) {
425 if (num_online_cpus() >= max_cpus)
431 /* Any cleanup work */
432 printk("Brought up %ld CPUs\n", (long)num_online_cpus());
433 smp_cpus_done(max_cpus);
435 /* Get other processors into their bootup holding patterns. */
445 * We need to finalize in a non-__init function or else race conditions
446 * between the root thread and the init thread may cause start_kernel to
447 * be reaped by free_initmem before the root thread has proceeded to
450 * gcc-3.4 accidentally inlines this function, so use noinline.
453 static void noinline rest_init(void)
454 __releases(kernel_lock)
456 kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
457 numa_default_policy();
462 /* Check for early params. */
463 static int __init do_early_param(char *param, char *val)
465 struct obs_kernel_param *p;
466 extern struct obs_kernel_param __setup_start, __setup_end;
468 for (p = &__setup_start; p < &__setup_end; p++) {
469 if (p->early && strcmp(param, p->str) == 0) {
470 if (p->setup_func(val) != 0)
472 "Malformed early option '%s'\n", param);
475 /* We accept everything at this stage. */
479 /* Arch code calls this early on, or if not, just before other parsing. */
480 void __init parse_early_param(void)
482 static __initdata int done = 0;
483 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
488 /* All fall through to do_early_param. */
489 strlcpy(tmp_cmdline, saved_command_line, COMMAND_LINE_SIZE);
490 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
495 * Activate the first processor.
498 asmlinkage void __init start_kernel(void)
501 extern struct kernel_param __start___param[], __stop___param[];
503 * Interrupts are still disabled. Do necessary setups, then
508 printk(linux_banner);
509 setup_arch(&command_line);
510 setup_per_cpu_areas();
513 * Mark the boot cpu "online" so that it can call console drivers in
514 * printk() and can access its per-cpu storage.
516 smp_prepare_boot_cpu();
519 * Set up the scheduler prior starting any interrupts (such as the
520 * timer interrupt). Full topology setup happens at smp_init()
521 * time - but meanwhile we still have a functioning scheduler.
524 build_all_zonelists();
526 printk("Kernel command line: %s\n", saved_command_line);
528 parse_args("Booting kernel", command_line, __start___param,
529 __stop___param - __start___param,
530 &unknown_bootoption);
541 * HACK ALERT! This is early. We're enabling the console before
542 * we've done PCI setups etc, and console_init() must be aware of
543 * this. But we do want output early, in case something goes wrong.
547 panic(panic_later, panic_param);
550 #ifdef CONFIG_BLK_DEV_INITRD
551 if (initrd_start && !initrd_below_start_ok &&
552 initrd_start < min_low_pfn << PAGE_SHIFT) {
553 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
554 "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
558 vfs_caches_init_early();
566 pgtable_cache_init();
571 efi_enter_virtual_mode();
573 fork_init(num_physpages);
578 vfs_caches_init(num_physpages);
581 /* rootfs populating might need page-writeback */
582 page_writeback_init();
583 #ifdef CONFIG_PROC_FS
588 acpi_early_init(); /* before LAPIC and SMP init */
590 /* Do the rest non-__init'ed, we're now alive */
594 static int __initdata initcall_debug;
596 static int __init initcall_debug_setup(char *str)
601 __setup("initcall_debug", initcall_debug_setup);
603 struct task_struct *child_reaper = &init_task;
605 extern initcall_t __initcall_start, __initcall_end;
607 static void __init do_initcalls(void)
610 int count = preempt_count();
612 for (call = &__initcall_start; call < &__initcall_end; call++) {
615 if (initcall_debug) {
616 printk(KERN_DEBUG "Calling initcall 0x%p", *call);
617 print_fn_descriptor_symbol(": %s()", (unsigned long) *call);
624 if (preempt_count() != count) {
625 msg = "preemption imbalance";
626 preempt_count() = count;
628 if (irqs_disabled()) {
629 msg = "disabled interrupts";
633 printk("error in initcall at 0x%p: "
634 "returned with %s\n", *call, msg);
638 /* Make sure there is no pending stuff from the initcall sequence */
639 flush_scheduled_work();
643 * Ok, the machine is now initialized. None of the devices
644 * have been touched yet, but the CPU subsystem is up and
645 * running, and memory and process management works.
647 * Now we can finally start doing some real work..
649 static void __init do_basic_setup(void)
651 /* drivers will send hotplug events */
653 usermodehelper_init();
661 /* Networking initialization needs a process context */
667 static void do_pre_smp_initcalls(void)
669 extern int spawn_ksoftirqd(void);
671 extern int migration_init(void);
678 static void run_init_process(char *init_filename)
680 argv_init[0] = init_filename;
681 execve(init_filename, argv_init, envp_init);
684 static inline void fixup_cpu_present_map(void)
690 * If arch is not hotplug ready and did not populate
691 * cpu_present_map, just make cpu_present_map same as cpu_possible_map
692 * for other cpu bringup code to function as normal. e.g smp_init() etc.
694 if (cpus_empty(cpu_present_map)) {
696 cpu_set(i, cpu_present_map);
702 static int init(void * unused)
706 * Tell the world that we're going to be the grim
707 * reaper of innocent orphaned children.
709 * We don't want people to have to make incorrect
710 * assumptions about where in the task array this
713 child_reaper = current;
715 /* Sets up cpus_possible() */
716 smp_prepare_cpus(max_cpus);
718 do_pre_smp_initcalls();
720 fixup_cpu_present_map();
724 * Do this before initcalls, because some drivers want to access
732 * check if there is an early userspace init. If yes, let it do all
735 if (sys_access((const char __user *) "/init", 0) == 0)
736 execute_command = "/init";
741 * Ok, we have completed the initial bootup, and
742 * we're essentially up and running. Get rid of the
743 * initmem segments and start the user-mode stuff..
747 system_state = SYSTEM_RUNNING;
748 numa_default_policy();
750 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
751 printk("Warning: unable to open an initial console.\n");
757 * We try each of these until one succeeds.
759 * The Bourne shell can be used instead of init if we are
760 * trying to recover a really broken machine.
764 run_init_process(execute_command);
766 run_init_process("/sbin/init");
767 run_init_process("/etc/init");
768 run_init_process("/bin/init");
769 run_init_process("/bin/sh");
771 panic("No init found. Try passing init= option to kernel.");
774 static int early_param_test(char *rest)
776 printk("early_parm_test: %s\n", rest ?: "(null)");
777 return rest ? 0 : -EINVAL;
779 early_param("testsetup", early_param_test);
780 static int early_setup_test(char *rest)
782 printk("early_setup_test: %s\n", rest ?: "(null)");
785 __setup("testsetup_long", early_setup_test);