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>
52 #include <linux/ckrm.h>
53 #ifdef CONFIG_CKRM_CPU_SCHEDULE
54 int __init init_ckrm_sched_res(void);
56 #define init_ckrm_sched_res() ((void)0)
60 * This is one of the first .c files built. Error out early
61 * if we have compiler trouble..
63 #if __GNUC__ == 2 && __GNUC_MINOR__ == 96
64 #ifdef CONFIG_FRAME_POINTER
65 #error This compiler cannot compile correctly with frame pointers enabled
69 #ifdef CONFIG_X86_LOCAL_APIC
74 * Versions of gcc older than that listed below may actually compile
75 * and link okay, but the end product can have subtle run time bugs.
76 * To avoid associated bogus bug reports, we flatly refuse to compile
77 * with a gcc that is known to be too old from the very beginning.
79 #if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 95)
80 #error Sorry, your GCC is too old. It builds incorrect kernels.
83 extern char *linux_banner;
85 static int init(void *);
87 extern void init_IRQ(void);
88 extern void sock_init(void);
89 extern void fork_init(unsigned long);
90 extern void mca_init(void);
91 extern void sbus_init(void);
92 extern void sysctl_init(void);
93 extern void signals_init(void);
94 extern void buffer_init(void);
95 extern void pidhash_init(void);
96 extern void pidmap_init(void);
97 extern void prio_tree_init(void);
98 extern void radix_tree_init(void);
99 extern void free_initmem(void);
100 extern void populate_rootfs(void);
101 extern void driver_init(void);
102 extern void prepare_namespace(void);
105 extern void tc_init(void);
108 enum system_states system_state;
109 EXPORT_SYMBOL(system_state);
112 * Boot command-line arguments
114 #define MAX_INIT_ARGS 8
115 #define MAX_INIT_ENVS 8
117 extern void time_init(void);
118 /* Default late time init is NULL. archs can override this later. */
119 void (*late_time_init)(void);
120 extern void softirq_init(void);
122 /* Untouched command line (eg. for /proc) saved by arch-specific code. */
123 char saved_command_line[COMMAND_LINE_SIZE];
125 static char *execute_command;
127 /* Setup configured maximum number of CPUs to activate */
128 static unsigned int max_cpus = NR_CPUS;
131 * Setup routine for controlling SMP activation
133 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
134 * activation entirely (the MPS table probe still happens, though).
136 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
137 * greater than 0, limits the maximum number of CPUs activated in
140 static int __init nosmp(char *str)
146 __setup("nosmp", nosmp);
148 static int __init maxcpus(char *str)
150 get_option(&str, &max_cpus);
154 __setup("maxcpus=", maxcpus);
156 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
157 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
158 static const char *panic_later, *panic_param;
160 __setup("profile=", profile_setup);
162 static int __init obsolete_checksetup(char *line)
164 struct obs_kernel_param *p;
165 extern struct obs_kernel_param __setup_start, __setup_end;
169 int n = strlen(p->str);
170 if (!strncmp(line, p->str, n)) {
172 /* Already done in parse_early_param? (Needs
173 * exact match on param part) */
174 if (line[n] == '\0' || line[n] == '=')
176 } else if (!p->setup_func) {
177 printk(KERN_WARNING "Parameter %s is obsolete,"
178 " ignored\n", p->str);
180 } else if (p->setup_func(line + n))
184 } while (p < &__setup_end);
188 /* this should be approx 2 Bo*oMips to start (note initial shift), and will
189 still work even if initially too large, it will just take slightly longer */
190 unsigned long loops_per_jiffy = (1<<12);
192 EXPORT_SYMBOL(loops_per_jiffy);
194 /* This is the number of bits of precision for the loops_per_jiffy. Each
195 bit takes on average 1.5/HZ seconds. This (like the original) is a little
199 void __devinit calibrate_delay(void)
201 unsigned long ticks, loopbit;
202 int lps_precision = LPS_PREC;
204 loops_per_jiffy = (1<<12);
206 printk("Calibrating delay loop... ");
207 while ((loops_per_jiffy <<= 1) != 0) {
208 /* wait for "start of" clock tick */
210 while (ticks == jiffies)
214 __delay(loops_per_jiffy);
215 ticks = jiffies - ticks;
220 /* Do a binary approximation to get loops_per_jiffy set to equal one clock
221 (up to lps_precision bits) */
222 loops_per_jiffy >>= 1;
223 loopbit = loops_per_jiffy;
224 while ( lps_precision-- && (loopbit >>= 1) ) {
225 loops_per_jiffy |= loopbit;
227 while (ticks == jiffies);
229 __delay(loops_per_jiffy);
230 if (jiffies != ticks) /* longer than 1 tick */
231 loops_per_jiffy &= ~loopbit;
234 /* Round the value and print it */
235 printk("%lu.%02lu BogoMIPS\n",
236 loops_per_jiffy/(500000/HZ),
237 (loops_per_jiffy/(5000/HZ)) % 100);
240 static int __init debug_kernel(char *str)
244 console_loglevel = 10;
248 static int __init quiet_kernel(char *str)
252 console_loglevel = 4;
256 __setup("debug", debug_kernel);
257 __setup("quiet", quiet_kernel);
259 /* Unknown boot options get handed to init, unless they look like
261 static int __init unknown_bootoption(char *param, char *val)
263 /* Change NUL term back to "=", to make "param" the whole string. */
267 /* Handle obsolete-style parameters */
268 if (obsolete_checksetup(param))
271 /* Preemptive maintenance for "why didn't my mispelled command
273 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
274 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
282 /* Environment option */
284 for (i = 0; envp_init[i]; i++) {
285 if (i == MAX_INIT_ENVS) {
286 panic_later = "Too many boot env vars at `%s'";
289 if (!strncmp(param, envp_init[i], val - param))
292 envp_init[i] = param;
294 /* Command line option */
296 for (i = 0; argv_init[i]; i++) {
297 if (i == MAX_INIT_ARGS) {
298 panic_later = "Too many boot init vars at `%s'";
302 argv_init[i] = param;
307 static int __init init_setup(char *str)
311 execute_command = str;
312 /* In case LILO is going to boot us with default command line,
313 * it prepends "auto" before the whole cmdline which makes
314 * the shell think it should execute a script with such name.
315 * So we ignore all arguments entered _before_ init=... [MJ]
317 for (i = 1; i < MAX_INIT_ARGS; i++)
321 __setup("init=", init_setup);
323 extern void setup_arch(char **);
324 extern void cpu_idle(void);
328 #ifdef CONFIG_X86_LOCAL_APIC
329 static void __init smp_init(void)
331 APIC_init_uniprocessor();
334 #define smp_init() do { } while (0)
337 static inline void setup_per_cpu_areas(void) { }
338 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
342 #ifdef __GENERIC_PER_CPU
343 unsigned long __per_cpu_offset[NR_CPUS];
345 EXPORT_SYMBOL(__per_cpu_offset);
347 static void __init setup_per_cpu_areas(void)
349 unsigned long size, i;
351 /* Created by linker magic */
352 extern char __per_cpu_start[], __per_cpu_end[];
354 /* Copy section for each CPU (we discard the original) */
355 size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
356 #ifdef CONFIG_MODULES
357 if (size < PERCPU_ENOUGH_ROOM)
358 size = PERCPU_ENOUGH_ROOM;
361 ptr = alloc_bootmem(size * NR_CPUS);
363 for (i = 0; i < NR_CPUS; i++, ptr += size) {
364 __per_cpu_offset[i] = ptr - __per_cpu_start;
365 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
368 #endif /* !__GENERIC_PER_CPU */
370 /* Called by boot processor to activate the rest. */
371 static void __init smp_init(void)
375 /* FIXME: This should be done in userspace --RR */
376 for_each_present_cpu(i) {
377 if (num_online_cpus() >= max_cpus)
383 /* Any cleanup work */
384 printk("Brought up %ld CPUs\n", (long)num_online_cpus());
385 smp_cpus_done(max_cpus);
387 /* Get other processors into their bootup holding patterns. */
397 * We need to finalize in a non-__init function or else race conditions
398 * between the root thread and the init thread may cause start_kernel to
399 * be reaped by free_initmem before the root thread has proceeded to
402 * gcc-3.4 accidentally inlines this function, so use noinline.
405 static void noinline rest_init(void)
407 kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
408 numa_default_policy();
409 system_state = SYSTEM_BOOTING_SCHEDULER_OK;
414 /* Check for early params. */
415 static int __init do_early_param(char *param, char *val)
417 struct obs_kernel_param *p;
418 extern struct obs_kernel_param __setup_start, __setup_end;
420 for (p = &__setup_start; p < &__setup_end; p++) {
421 if (p->early && strcmp(param, p->str) == 0) {
422 if (p->setup_func(val) != 0)
424 "Malformed early option '%s'\n", param);
427 /* We accept everything at this stage. */
431 /* Arch code calls this early on, or if not, just before other parsing. */
432 void __init parse_early_param(void)
434 static __initdata int done = 0;
435 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
440 /* All fall through to do_early_param. */
441 strlcpy(tmp_cmdline, saved_command_line, COMMAND_LINE_SIZE);
442 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
447 * Activate the first processor.
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
460 printk(linux_banner);
461 setup_arch(&command_line);
462 setup_per_cpu_areas();
465 * Mark the boot cpu "online" so that it can call console drivers in
466 * printk() and can access its per-cpu storage.
468 smp_prepare_boot_cpu();
470 * Set up the scheduler prior starting any interrupts (such as the
471 * timer interrupt). Full topology setup happens at smp_init()
472 * time - but meanwhile we still have a functioning scheduler.
476 build_all_zonelists();
478 printk("Kernel command line: %s\n", saved_command_line);
480 parse_args("Booting kernel", command_line, __start___param,
481 __stop___param - __start___param,
482 &unknown_bootoption);
488 /* MEF: In 2.6.5. ckrm_init was right after pidhash_init() but
489 before sched_init(). Will leave it after pidhash_init()
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();
523 pgtable_cache_init();
528 efi_enter_virtual_mode();
530 fork_init(num_physpages);
534 security_scaffolding_startup();
535 vfs_caches_init(num_physpages);
538 /* rootfs populating might need page-writeback */
539 page_writeback_init();
540 #ifdef CONFIG_PROC_FS
547 * We count on the initial thread going ok
548 * Like idlers init is an unlocked kernel thread, which will
549 * make syscalls (and thus be locked).
551 init_idle(current, smp_processor_id());
553 /* Do the rest non-__init'ed, we're now alive */
557 static int __initdata initcall_debug;
559 static int __init initcall_debug_setup(char *str)
564 __setup("initcall_debug", initcall_debug_setup);
566 struct task_struct *child_reaper = &init_task;
568 extern initcall_t __initcall_start, __initcall_end;
570 static void __init do_initcalls(void)
573 int count = preempt_count();
575 for (call = &__initcall_start; call < &__initcall_end; call++) {
578 if (initcall_debug) {
579 printk(KERN_DEBUG "Calling initcall 0x%p", *call);
580 print_symbol(": %s()", (unsigned long) *call);
587 if (preempt_count() != count) {
588 msg = "preemption imbalance";
589 preempt_count() = count;
591 if (irqs_disabled()) {
592 msg = "disabled interrupts";
596 printk("error in initcall at 0x%p: "
597 "returned with %s\n", *call, msg);
601 /* Make sure there is no pending stuff from the initcall sequence */
602 flush_scheduled_work();
606 * Ok, the machine is now initialized. None of the devices
607 * have been touched yet, but the CPU subsystem is up and
608 * running, and memory and process management works.
610 * Now we can finally start doing some real work..
612 static void __init do_basic_setup(void)
620 /* Networking initialization needs a process context */
627 static void do_pre_smp_initcalls(void)
629 extern int spawn_ksoftirqd(void);
631 extern int migration_init(void);
638 static void run_init_process(char *init_filename)
640 argv_init[0] = init_filename;
641 execve(init_filename, argv_init, envp_init);
644 static inline void fixup_cpu_present_map(void)
650 * If arch is not hotplug ready and did not populate
651 * cpu_present_map, just make cpu_present_map same as cpu_possible_map
652 * for other cpu bringup code to function as normal. e.g smp_init() etc.
654 if (cpus_empty(cpu_present_map)) {
656 cpu_set(i, cpu_present_map);
662 static int init(void * unused)
666 * Tell the world that we're going to be the grim
667 * reaper of innocent orphaned children.
669 * We don't want people to have to make incorrect
670 * assumptions about where in the task array this
673 child_reaper = current;
675 /* Sets up cpus_possible() */
676 smp_prepare_cpus(max_cpus);
678 do_pre_smp_initcalls();
680 fixup_cpu_present_map();
684 * Do this before initcalls, because some drivers want to access
689 init_ckrm_sched_res();
694 * check if there is an early userspace init. If yes, let it do all
697 if (sys_access((const char __user *) "/init", 0) == 0)
698 execute_command = "/init";
703 * Ok, we have completed the initial bootup, and
704 * we're essentially up and running. Get rid of the
705 * initmem segments and start the user-mode stuff..
709 system_state = SYSTEM_RUNNING;
710 numa_default_policy();
712 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
713 printk("Warning: unable to open an initial console.\n");
719 * We try each of these until one succeeds.
721 * The Bourne shell can be used instead of init if we are
722 * trying to recover a really broken machine.
726 run_init_process(execute_command);
728 run_init_process("/sbin/init");
729 run_init_process("/etc/init");
730 run_init_process("/bin/init");
731 run_init_process("/bin/sh");
733 panic("No init found. Try passing init= option to kernel.");
736 static int early_param_test(char *rest)
738 printk("early_parm_test: %s\n", rest ?: "(null)");
739 return rest ? 0 : -EINVAL;
741 early_param("testsetup", early_param_test);
742 static int early_setup_test(char *rest)
744 printk("early_setup_test: %s\n", rest ?: "(null)");
747 __setup("testsetup_long", early_setup_test);