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>
51 #include <linux/ckrm.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);
99 extern void tc_init(void);
102 enum system_states system_state;
103 EXPORT_SYMBOL(system_state);
106 * Boot command-line arguments
108 #define MAX_INIT_ARGS 8
109 #define MAX_INIT_ENVS 8
111 extern void time_init(void);
112 /* Default late time init is NULL. archs can override this later. */
113 void (*late_time_init)(void);
114 extern void softirq_init(void);
116 static char *execute_command;
118 /* Setup configured maximum number of CPUs to activate */
119 static unsigned int max_cpus = NR_CPUS;
122 * Setup routine for controlling SMP activation
124 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
125 * activation entirely (the MPS table probe still happens, though).
127 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
128 * greater than 0, limits the maximum number of CPUs activated in
131 static int __init nosmp(char *str)
137 __setup("nosmp", nosmp);
139 static int __init maxcpus(char *str)
141 get_option(&str, &max_cpus);
145 __setup("maxcpus=", maxcpus);
147 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
148 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
149 static const char *panic_later, *panic_param;
151 __setup("profile=", profile_setup);
153 static int __init obsolete_checksetup(char *line)
155 struct obs_kernel_param *p;
156 extern struct obs_kernel_param __setup_start, __setup_end;
160 int n = strlen(p->str);
161 if (!strncmp(line, p->str, n)) {
162 if (!p->setup_func) {
163 printk(KERN_WARNING "Parameter %s is obsolete, ignored\n", p->str);
165 } else if (p->setup_func(line + n))
169 } while (p < &__setup_end);
173 /* this should be approx 2 Bo*oMips to start (note initial shift), and will
174 still work even if initially too large, it will just take slightly longer */
175 unsigned long loops_per_jiffy = (1<<12);
177 EXPORT_SYMBOL(loops_per_jiffy);
179 /* This is the number of bits of precision for the loops_per_jiffy. Each
180 bit takes on average 1.5/HZ seconds. This (like the original) is a little
184 void __devinit calibrate_delay(void)
186 unsigned long ticks, loopbit;
187 int lps_precision = LPS_PREC;
189 loops_per_jiffy = (1<<12);
191 printk("Calibrating delay loop... ");
192 while ((loops_per_jiffy <<= 1) != 0) {
193 /* wait for "start of" clock tick */
195 while (ticks == jiffies)
199 __delay(loops_per_jiffy);
200 ticks = jiffies - ticks;
205 /* Do a binary approximation to get loops_per_jiffy set to equal one clock
206 (up to lps_precision bits) */
207 loops_per_jiffy >>= 1;
208 loopbit = loops_per_jiffy;
209 while ( lps_precision-- && (loopbit >>= 1) ) {
210 loops_per_jiffy |= loopbit;
212 while (ticks == jiffies);
214 __delay(loops_per_jiffy);
215 if (jiffies != ticks) /* longer than 1 tick */
216 loops_per_jiffy &= ~loopbit;
219 /* Round the value and print it */
220 printk("%lu.%02lu BogoMIPS\n",
221 loops_per_jiffy/(500000/HZ),
222 (loops_per_jiffy/(5000/HZ)) % 100);
225 static int __init debug_kernel(char *str)
229 console_loglevel = 10;
233 static int __init quiet_kernel(char *str)
237 console_loglevel = 4;
241 __setup("debug", debug_kernel);
242 __setup("quiet", quiet_kernel);
244 /* Unknown boot options get handed to init, unless they look like
246 static int __init unknown_bootoption(char *param, char *val)
248 /* Change NUL term back to "=", to make "param" the whole string. */
252 /* Handle obsolete-style parameters */
253 if (obsolete_checksetup(param))
256 /* Preemptive maintenance for "why didn't my mispelled command
258 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
259 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
267 /* Environment option */
269 for (i = 0; envp_init[i]; i++) {
270 if (i == MAX_INIT_ENVS) {
271 panic_later = "Too many boot env vars at `%s'";
274 if (!strncmp(param, envp_init[i], val - param))
277 envp_init[i] = param;
279 /* Command line option */
281 for (i = 0; argv_init[i]; i++) {
282 if (i == MAX_INIT_ARGS) {
283 panic_later = "Too many boot init vars at `%s'";
287 argv_init[i] = param;
292 static int __init init_setup(char *str)
296 execute_command = str;
297 /* In case LILO is going to boot us with default command line,
298 * it prepends "auto" before the whole cmdline which makes
299 * the shell think it should execute a script with such name.
300 * So we ignore all arguments entered _before_ init=... [MJ]
302 for (i = 1; i < MAX_INIT_ARGS; i++)
306 __setup("init=", init_setup);
308 extern void setup_arch(char **);
309 extern void cpu_idle(void);
313 #ifdef CONFIG_X86_LOCAL_APIC
314 static void __init smp_init(void)
316 APIC_init_uniprocessor();
319 #define smp_init() do { } while (0)
322 static inline void setup_per_cpu_areas(void) { }
323 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
327 #ifdef __GENERIC_PER_CPU
328 unsigned long __per_cpu_offset[NR_CPUS];
330 EXPORT_SYMBOL(__per_cpu_offset);
332 static void __init setup_per_cpu_areas(void)
334 unsigned long size, i;
336 /* Created by linker magic */
337 extern char __per_cpu_start[], __per_cpu_end[];
339 /* Copy section for each CPU (we discard the original) */
340 size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
341 #ifdef CONFIG_MODULES
342 if (size < PERCPU_ENOUGH_ROOM)
343 size = PERCPU_ENOUGH_ROOM;
346 ptr = alloc_bootmem(size * NR_CPUS);
348 for (i = 0; i < NR_CPUS; i++, ptr += size) {
349 __per_cpu_offset[i] = ptr - __per_cpu_start;
350 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
353 #endif /* !__GENERIC_PER_CPU */
355 /* Called by boot processor to activate the rest. */
356 static void __init smp_init(void)
360 /* FIXME: This should be done in userspace --RR */
361 for_each_present_cpu(i) {
362 if (num_online_cpus() >= max_cpus)
368 /* Any cleanup work */
369 printk("Brought up %ld CPUs\n", (long)num_online_cpus());
370 smp_cpus_done(max_cpus);
372 /* Get other processors into their bootup holding patterns. */
382 * We need to finalize in a non-__init function or else race conditions
383 * between the root thread and the init thread may cause start_kernel to
384 * be reaped by free_initmem before the root thread has proceeded to
387 * gcc-3.4 accidentally inlines this function, so use noinline.
390 static void noinline rest_init(void)
392 kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
393 numa_default_policy();
399 * Activate the first processor.
402 asmlinkage void __init start_kernel(void)
405 extern char saved_command_line[];
406 extern struct kernel_param __start___param[], __stop___param[];
408 * Interrupts are still disabled. Do necessary setups, then
413 printk(linux_banner);
414 setup_arch(&command_line);
415 setup_per_cpu_areas();
418 * Mark the boot cpu "online" so that it can call console drivers in
419 * printk() and can access its per-cpu storage.
421 smp_prepare_boot_cpu();
424 * Set up the scheduler prior starting any interrupts (such as the
425 * timer interrupt). Full topology setup happens at smp_init()
426 * time - but meanwhile we still have a functioning scheduler.
430 build_all_zonelists();
432 printk("Kernel command line: %s\n", saved_command_line);
433 parse_args("Booting kernel", command_line, __start___param,
434 __stop___param - __start___param,
435 &unknown_bootoption);
441 /* MEF: In 2.6.5. ckrm_init was right after pidhash_init() but
442 before sched_init(). Will leave it after pidhash_init()
451 * HACK ALERT! This is early. We're enabling the console before
452 * we've done PCI setups etc, and console_init() must be aware of
453 * this. But we do want output early, in case something goes wrong.
457 panic(panic_later, panic_param);
460 #ifdef CONFIG_BLK_DEV_INITRD
461 if (initrd_start && !initrd_below_start_ok &&
462 initrd_start < min_low_pfn << PAGE_SHIFT) {
463 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
464 "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
475 pgtable_cache_init();
480 efi_enter_virtual_mode();
482 fork_init(num_physpages);
486 security_scaffolding_startup();
487 vfs_caches_init(num_physpages);
490 /* rootfs populating might need page-writeback */
491 page_writeback_init();
492 #ifdef CONFIG_PROC_FS
499 * We count on the initial thread going ok
500 * Like idlers init is an unlocked kernel thread, which will
501 * make syscalls (and thus be locked).
503 init_idle(current, smp_processor_id());
505 /* Do the rest non-__init'ed, we're now alive */
509 static int __initdata initcall_debug;
511 static int __init initcall_debug_setup(char *str)
516 __setup("initcall_debug", initcall_debug_setup);
518 struct task_struct *child_reaper = &init_task;
520 extern initcall_t __initcall_start, __initcall_end;
522 static void __init do_initcalls(void)
525 int count = preempt_count();
527 for (call = &__initcall_start; call < &__initcall_end; call++) {
530 if (initcall_debug) {
531 printk(KERN_DEBUG "Calling initcall 0x%p", *call);
532 print_symbol(": %s()", (unsigned long) *call);
539 if (preempt_count() != count) {
540 msg = "preemption imbalance";
541 preempt_count() = count;
543 if (irqs_disabled()) {
544 msg = "disabled interrupts";
548 printk("error in initcall at 0x%p: "
549 "returned with %s\n", *call, msg);
553 /* Make sure there is no pending stuff from the initcall sequence */
554 flush_scheduled_work();
558 * Ok, the machine is now initialized. None of the devices
559 * have been touched yet, but the CPU subsystem is up and
560 * running, and memory and process management works.
562 * Now we can finally start doing some real work..
564 static void __init do_basic_setup(void)
572 /* Networking initialization needs a process context */
579 static void do_pre_smp_initcalls(void)
581 extern int spawn_ksoftirqd(void);
583 extern int migration_init(void);
590 static void run_init_process(char *init_filename)
592 argv_init[0] = init_filename;
593 execve(init_filename, argv_init, envp_init);
596 static inline void fixup_cpu_present_map(void)
602 * If arch is not hotplug ready and did not populate
603 * cpu_present_map, just make cpu_present_map same as cpu_possible_map
604 * for other cpu bringup code to function as normal. e.g smp_init() etc.
606 if (cpus_empty(cpu_present_map)) {
608 cpu_set(i, cpu_present_map);
614 static int init(void * unused)
618 * Tell the world that we're going to be the grim
619 * reaper of innocent orphaned children.
621 * We don't want people to have to make incorrect
622 * assumptions about where in the task array this
625 child_reaper = current;
627 /* Sets up cpus_possible() */
628 smp_prepare_cpus(max_cpus);
630 do_pre_smp_initcalls();
632 fixup_cpu_present_map();
637 * Do this before initcalls, because some drivers want to access
645 * check if there is an early userspace init. If yes, let it do all
648 if (sys_access((const char __user *) "/init", 0) == 0)
649 execute_command = "/init";
654 * Ok, we have completed the initial bootup, and
655 * we're essentially up and running. Get rid of the
656 * initmem segments and start the user-mode stuff..
660 system_state = SYSTEM_RUNNING;
661 numa_default_policy();
663 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
664 printk("Warning: unable to open an initial console.\n");
670 * We try each of these until one succeeds.
672 * The Bourne shell can be used instead of init if we are
673 * trying to recover a really broken machine.
677 run_init_process(execute_command);
679 run_init_process("/sbin/init");
680 run_init_process("/etc/init");
681 run_init_process("/bin/init");
682 run_init_process("/bin/sh");
684 panic("No init found. Try passing init= option to kernel.");