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>
52 int __init init_ckrm_sched_res(void);
56 * This is one of the first .c files built. Error out early
57 * if we have compiler trouble..
59 #if __GNUC__ == 2 && __GNUC_MINOR__ == 96
60 #ifdef CONFIG_FRAME_POINTER
61 #error This compiler cannot compile correctly with frame pointers enabled
65 #ifdef CONFIG_X86_LOCAL_APIC
70 * Versions of gcc older than that listed below may actually compile
71 * and link okay, but the end product can have subtle run time bugs.
72 * To avoid associated bogus bug reports, we flatly refuse to compile
73 * with a gcc that is known to be too old from the very beginning.
75 #if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 95)
76 #error Sorry, your GCC is too old. It builds incorrect kernels.
79 extern char *linux_banner;
81 static int init(void *);
83 extern void init_IRQ(void);
84 extern void sock_init(void);
85 extern void fork_init(unsigned long);
86 extern void mca_init(void);
87 extern void sbus_init(void);
88 extern void sysctl_init(void);
89 extern void signals_init(void);
90 extern void buffer_init(void);
91 extern void pidhash_init(void);
92 extern void pidmap_init(void);
93 extern void prio_tree_init(void);
94 extern void radix_tree_init(void);
95 extern void free_initmem(void);
96 extern void populate_rootfs(void);
97 extern void driver_init(void);
98 extern void prepare_namespace(void);
101 extern void tc_init(void);
104 enum system_states system_state;
105 EXPORT_SYMBOL(system_state);
108 * Boot command-line arguments
110 #define MAX_INIT_ARGS 8
111 #define MAX_INIT_ENVS 8
113 extern void time_init(void);
114 /* Default late time init is NULL. archs can override this later. */
115 void (*late_time_init)(void);
116 extern void softirq_init(void);
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)) {
164 if (!p->setup_func) {
165 printk(KERN_WARNING "Parameter %s is obsolete, ignored\n", p->str);
167 } else if (p->setup_func(line + n))
171 } while (p < &__setup_end);
175 /* this should be approx 2 Bo*oMips to start (note initial shift), and will
176 still work even if initially too large, it will just take slightly longer */
177 unsigned long loops_per_jiffy = (1<<12);
179 EXPORT_SYMBOL(loops_per_jiffy);
181 /* This is the number of bits of precision for the loops_per_jiffy. Each
182 bit takes on average 1.5/HZ seconds. This (like the original) is a little
186 void __devinit calibrate_delay(void)
188 unsigned long ticks, loopbit;
189 int lps_precision = LPS_PREC;
191 loops_per_jiffy = (1<<12);
193 printk("Calibrating delay loop... ");
194 while ((loops_per_jiffy <<= 1) != 0) {
195 /* wait for "start of" clock tick */
197 while (ticks == jiffies)
201 __delay(loops_per_jiffy);
202 ticks = jiffies - ticks;
207 /* Do a binary approximation to get loops_per_jiffy set to equal one clock
208 (up to lps_precision bits) */
209 loops_per_jiffy >>= 1;
210 loopbit = loops_per_jiffy;
211 while ( lps_precision-- && (loopbit >>= 1) ) {
212 loops_per_jiffy |= loopbit;
214 while (ticks == jiffies);
216 __delay(loops_per_jiffy);
217 if (jiffies != ticks) /* longer than 1 tick */
218 loops_per_jiffy &= ~loopbit;
221 /* Round the value and print it */
222 printk("%lu.%02lu BogoMIPS\n",
223 loops_per_jiffy/(500000/HZ),
224 (loops_per_jiffy/(5000/HZ)) % 100);
227 static int __init debug_kernel(char *str)
231 console_loglevel = 10;
235 static int __init quiet_kernel(char *str)
239 console_loglevel = 4;
243 __setup("debug", debug_kernel);
244 __setup("quiet", quiet_kernel);
246 /* Unknown boot options get handed to init, unless they look like
248 static int __init unknown_bootoption(char *param, char *val)
250 /* Change NUL term back to "=", to make "param" the whole string. */
254 /* Handle obsolete-style parameters */
255 if (obsolete_checksetup(param))
258 /* Preemptive maintenance for "why didn't my mispelled command
260 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
261 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
269 /* Environment option */
271 for (i = 0; envp_init[i]; i++) {
272 if (i == MAX_INIT_ENVS) {
273 panic_later = "Too many boot env vars at `%s'";
276 if (!strncmp(param, envp_init[i], val - param))
279 envp_init[i] = param;
281 /* Command line option */
283 for (i = 0; argv_init[i]; i++) {
284 if (i == MAX_INIT_ARGS) {
285 panic_later = "Too many boot init vars at `%s'";
289 argv_init[i] = param;
294 static int __init init_setup(char *str)
298 execute_command = str;
299 /* In case LILO is going to boot us with default command line,
300 * it prepends "auto" before the whole cmdline which makes
301 * the shell think it should execute a script with such name.
302 * So we ignore all arguments entered _before_ init=... [MJ]
304 for (i = 1; i < MAX_INIT_ARGS; i++)
308 __setup("init=", init_setup);
310 extern void setup_arch(char **);
311 extern void cpu_idle(void);
315 #ifdef CONFIG_X86_LOCAL_APIC
316 static void __init smp_init(void)
318 APIC_init_uniprocessor();
321 #define smp_init() do { } while (0)
324 static inline void setup_per_cpu_areas(void) { }
325 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
329 #ifdef __GENERIC_PER_CPU
330 unsigned long __per_cpu_offset[NR_CPUS];
332 EXPORT_SYMBOL(__per_cpu_offset);
334 static void __init setup_per_cpu_areas(void)
336 unsigned long size, i;
338 /* Created by linker magic */
339 extern char __per_cpu_start[], __per_cpu_end[];
341 /* Copy section for each CPU (we discard the original) */
342 size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
343 #ifdef CONFIG_MODULES
344 if (size < PERCPU_ENOUGH_ROOM)
345 size = PERCPU_ENOUGH_ROOM;
348 ptr = alloc_bootmem(size * NR_CPUS);
350 for (i = 0; i < NR_CPUS; i++, ptr += size) {
351 __per_cpu_offset[i] = ptr - __per_cpu_start;
352 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
355 #endif /* !__GENERIC_PER_CPU */
357 /* Called by boot processor to activate the rest. */
358 static void __init smp_init(void)
362 /* FIXME: This should be done in userspace --RR */
363 for_each_present_cpu(i) {
364 if (num_online_cpus() >= max_cpus)
370 /* Any cleanup work */
371 printk("Brought up %ld CPUs\n", (long)num_online_cpus());
372 smp_cpus_done(max_cpus);
374 /* Get other processors into their bootup holding patterns. */
384 * We need to finalize in a non-__init function or else race conditions
385 * between the root thread and the init thread may cause start_kernel to
386 * be reaped by free_initmem before the root thread has proceeded to
389 * gcc-3.4 accidentally inlines this function, so use noinline.
392 static void noinline rest_init(void)
394 kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
395 numa_default_policy();
401 * Activate the first processor.
404 asmlinkage void __init start_kernel(void)
407 extern char saved_command_line[];
408 extern struct kernel_param __start___param[], __stop___param[];
410 * Interrupts are still disabled. Do necessary setups, then
415 printk(linux_banner);
416 setup_arch(&command_line);
417 setup_per_cpu_areas();
420 * Mark the boot cpu "online" so that it can call console drivers in
421 * printk() and can access its per-cpu storage.
423 smp_prepare_boot_cpu();
425 * Set up the scheduler prior starting any interrupts (such as the
426 * timer interrupt). Full topology setup happens at smp_init()
427 * time - but meanwhile we still have a functioning scheduler.
431 build_all_zonelists();
433 printk("Kernel command line: %s\n", saved_command_line);
434 parse_args("Booting kernel", command_line, __start___param,
435 __stop___param - __start___param,
436 &unknown_bootoption);
442 /* MEF: In 2.6.5. ckrm_init was right after pidhash_init() but
443 before sched_init(). Will leave it after pidhash_init()
452 * HACK ALERT! This is early. We're enabling the console before
453 * we've done PCI setups etc, and console_init() must be aware of
454 * this. But we do want output early, in case something goes wrong.
458 panic(panic_later, panic_param);
461 #ifdef CONFIG_BLK_DEV_INITRD
462 if (initrd_start && !initrd_below_start_ok &&
463 initrd_start < min_low_pfn << PAGE_SHIFT) {
464 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
465 "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
476 pgtable_cache_init();
481 efi_enter_virtual_mode();
483 fork_init(num_physpages);
487 security_scaffolding_startup();
488 vfs_caches_init(num_physpages);
491 /* rootfs populating might need page-writeback */
492 page_writeback_init();
493 #ifdef CONFIG_PROC_FS
500 * We count on the initial thread going ok
501 * Like idlers init is an unlocked kernel thread, which will
502 * make syscalls (and thus be locked).
504 init_idle(current, smp_processor_id());
506 /* Do the rest non-__init'ed, we're now alive */
510 static int __initdata initcall_debug;
512 static int __init initcall_debug_setup(char *str)
517 __setup("initcall_debug", initcall_debug_setup);
519 struct task_struct *child_reaper = &init_task;
521 extern initcall_t __initcall_start, __initcall_end;
523 static void __init do_initcalls(void)
526 int count = preempt_count();
528 for (call = &__initcall_start; call < &__initcall_end; call++) {
531 if (initcall_debug) {
532 printk(KERN_DEBUG "Calling initcall 0x%p", *call);
533 print_symbol(": %s()", (unsigned long) *call);
540 if (preempt_count() != count) {
541 msg = "preemption imbalance";
542 preempt_count() = count;
544 if (irqs_disabled()) {
545 msg = "disabled interrupts";
549 printk("error in initcall at 0x%p: "
550 "returned with %s\n", *call, msg);
554 /* Make sure there is no pending stuff from the initcall sequence */
555 flush_scheduled_work();
559 * Ok, the machine is now initialized. None of the devices
560 * have been touched yet, but the CPU subsystem is up and
561 * running, and memory and process management works.
563 * Now we can finally start doing some real work..
565 static void __init do_basic_setup(void)
573 /* Networking initialization needs a process context */
580 static void do_pre_smp_initcalls(void)
582 extern int spawn_ksoftirqd(void);
584 extern int migration_init(void);
591 static void run_init_process(char *init_filename)
593 argv_init[0] = init_filename;
594 execve(init_filename, argv_init, envp_init);
597 static inline void fixup_cpu_present_map(void)
603 * If arch is not hotplug ready and did not populate
604 * cpu_present_map, just make cpu_present_map same as cpu_possible_map
605 * for other cpu bringup code to function as normal. e.g smp_init() etc.
607 if (cpus_empty(cpu_present_map)) {
609 cpu_set(i, cpu_present_map);
615 static int init(void * unused)
619 * Tell the world that we're going to be the grim
620 * reaper of innocent orphaned children.
622 * We don't want people to have to make incorrect
623 * assumptions about where in the task array this
626 child_reaper = current;
628 /* Sets up cpus_possible() */
629 smp_prepare_cpus(max_cpus);
631 do_pre_smp_initcalls();
633 fixup_cpu_present_map();
638 * Do this before initcalls, because some drivers want to access
643 init_ckrm_sched_res();
646 * check if there is an early userspace init. If yes, let it do all
649 if (sys_access((const char __user *) "/init", 0) == 0)
650 execute_command = "/init";
655 * Ok, we have completed the initial bootup, and
656 * we're essentially up and running. Get rid of the
657 * initmem segments and start the user-mode stuff..
661 system_state = SYSTEM_RUNNING;
662 numa_default_policy();
664 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
665 printk("Warning: unable to open an initial console.\n");
671 * We try each of these until one succeeds.
673 * The Bourne shell can be used instead of init if we are
674 * trying to recover a really broken machine.
678 run_init_process(execute_command);
680 run_init_process("/sbin/init");
681 run_init_process("/etc/init");
682 run_init_process("/bin/init");
683 run_init_process("/bin/sh");
685 panic("No init found. Try passing init= option to kernel.");