sched: cfs core code
[usb.git] / init / main.c
blob0eb1c7463fe4d08c3012ee66cb486c72122a26d5
1 /*
2 * linux/init/main.c
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 #include <linux/types.h>
13 #include <linux/module.h>
14 #include <linux/proc_fs.h>
15 #include <linux/kernel.h>
16 #include <linux/syscalls.h>
17 #include <linux/string.h>
18 #include <linux/ctype.h>
19 #include <linux/delay.h>
20 #include <linux/utsname.h>
21 #include <linux/ioport.h>
22 #include <linux/init.h>
23 #include <linux/smp_lock.h>
24 #include <linux/initrd.h>
25 #include <linux/hdreg.h>
26 #include <linux/bootmem.h>
27 #include <linux/tty.h>
28 #include <linux/gfp.h>
29 #include <linux/percpu.h>
30 #include <linux/kmod.h>
31 #include <linux/kernel_stat.h>
32 #include <linux/start_kernel.h>
33 #include <linux/security.h>
34 #include <linux/workqueue.h>
35 #include <linux/profile.h>
36 #include <linux/rcupdate.h>
37 #include <linux/moduleparam.h>
38 #include <linux/kallsyms.h>
39 #include <linux/writeback.h>
40 #include <linux/cpu.h>
41 #include <linux/cpuset.h>
42 #include <linux/efi.h>
43 #include <linux/tick.h>
44 #include <linux/interrupt.h>
45 #include <linux/taskstats_kern.h>
46 #include <linux/delayacct.h>
47 #include <linux/unistd.h>
48 #include <linux/rmap.h>
49 #include <linux/mempolicy.h>
50 #include <linux/key.h>
51 #include <linux/unwind.h>
52 #include <linux/buffer_head.h>
53 #include <linux/debug_locks.h>
54 #include <linux/lockdep.h>
55 #include <linux/pid_namespace.h>
56 #include <linux/device.h>
57 #include <linux/kthread.h>
59 #include <asm/io.h>
60 #include <asm/bugs.h>
61 #include <asm/setup.h>
62 #include <asm/sections.h>
63 #include <asm/cacheflush.h>
65 #ifdef CONFIG_X86_LOCAL_APIC
66 #include <asm/smp.h>
67 #endif
70 * This is one of the first .c files built. Error out early if we have compiler
71 * trouble.
73 * Versions of gcc older than that listed below may actually compile and link
74 * okay, but the end product can have subtle run time bugs. To avoid associated
75 * bogus bug reports, we flatly refuse to compile with a gcc that is known to be
76 * too old from the very beginning.
78 #if (__GNUC__ < 3) || (__GNUC__ == 3 && __GNUC_MINOR__ < 2)
79 #error Sorry, your GCC is too old. It builds incorrect kernels.
80 #endif
82 #if __GNUC__ == 4 && __GNUC_MINOR__ == 1 && __GNUC_PATCHLEVEL__ == 0
83 #warning gcc-4.1.0 is known to miscompile the kernel. A different compiler version is recommended.
84 #endif
86 static int kernel_init(void *);
88 extern void init_IRQ(void);
89 extern void fork_init(unsigned long);
90 extern void mca_init(void);
91 extern void sbus_init(void);
92 extern void signals_init(void);
93 extern void pidhash_init(void);
94 extern void pidmap_init(void);
95 extern void prio_tree_init(void);
96 extern void radix_tree_init(void);
97 extern void free_initmem(void);
98 #ifdef CONFIG_ACPI
99 extern void acpi_early_init(void);
100 #else
101 static inline void acpi_early_init(void) { }
102 #endif
103 #ifndef CONFIG_DEBUG_RODATA
104 static inline void mark_rodata_ro(void) { }
105 #endif
107 #ifdef CONFIG_TC
108 extern void tc_init(void);
109 #endif
111 enum system_states system_state;
112 EXPORT_SYMBOL(system_state);
115 * Boot command-line arguments
117 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
118 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
120 extern void time_init(void);
121 /* Default late time init is NULL. archs can override this later. */
122 void (*late_time_init)(void);
123 extern void softirq_init(void);
125 /* Untouched command line saved by arch-specific code. */
126 char __initdata boot_command_line[COMMAND_LINE_SIZE];
127 /* Untouched saved command line (eg. for /proc) */
128 char *saved_command_line;
129 /* Command line for parameter parsing */
130 static char *static_command_line;
132 static char *execute_command;
133 static char *ramdisk_execute_command;
135 /* Setup configured maximum number of CPUs to activate */
136 static unsigned int max_cpus = NR_CPUS;
139 * If set, this is an indication to the drivers that reset the underlying
140 * device before going ahead with the initialization otherwise driver might
141 * rely on the BIOS and skip the reset operation.
143 * This is useful if kernel is booting in an unreliable environment.
144 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
145 * skipped and devices will be in unknown state.
147 unsigned int reset_devices;
148 EXPORT_SYMBOL(reset_devices);
151 * Setup routine for controlling SMP activation
153 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
154 * activation entirely (the MPS table probe still happens, though).
156 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
157 * greater than 0, limits the maximum number of CPUs activated in
158 * SMP mode to <NUM>.
160 static int __init nosmp(char *str)
162 max_cpus = 0;
163 return 1;
166 __setup("nosmp", nosmp);
168 static int __init maxcpus(char *str)
170 get_option(&str, &max_cpus);
171 return 1;
174 __setup("maxcpus=", maxcpus);
176 static int __init set_reset_devices(char *str)
178 reset_devices = 1;
179 return 1;
182 __setup("reset_devices", set_reset_devices);
184 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
185 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
186 static const char *panic_later, *panic_param;
188 extern struct obs_kernel_param __setup_start[], __setup_end[];
190 static int __init obsolete_checksetup(char *line)
192 struct obs_kernel_param *p;
193 int had_early_param = 0;
195 p = __setup_start;
196 do {
197 int n = strlen(p->str);
198 if (!strncmp(line, p->str, n)) {
199 if (p->early) {
200 /* Already done in parse_early_param?
201 * (Needs exact match on param part).
202 * Keep iterating, as we can have early
203 * params and __setups of same names 8( */
204 if (line[n] == '\0' || line[n] == '=')
205 had_early_param = 1;
206 } else if (!p->setup_func) {
207 printk(KERN_WARNING "Parameter %s is obsolete,"
208 " ignored\n", p->str);
209 return 1;
210 } else if (p->setup_func(line + n))
211 return 1;
213 p++;
214 } while (p < __setup_end);
216 return had_early_param;
220 * This should be approx 2 Bo*oMips to start (note initial shift), and will
221 * still work even if initially too large, it will just take slightly longer
223 unsigned long loops_per_jiffy = (1<<12);
225 EXPORT_SYMBOL(loops_per_jiffy);
227 static int __init debug_kernel(char *str)
229 if (*str)
230 return 0;
231 console_loglevel = 10;
232 return 1;
235 static int __init quiet_kernel(char *str)
237 if (*str)
238 return 0;
239 console_loglevel = 4;
240 return 1;
243 __setup("debug", debug_kernel);
244 __setup("quiet", quiet_kernel);
246 static int __init loglevel(char *str)
248 get_option(&str, &console_loglevel);
249 return 1;
252 __setup("loglevel=", loglevel);
255 * Unknown boot options get handed to init, unless they look like
256 * failed parameters
258 static int __init unknown_bootoption(char *param, char *val)
260 /* Change NUL term back to "=", to make "param" the whole string. */
261 if (val) {
262 /* param=val or param="val"? */
263 if (val == param+strlen(param)+1)
264 val[-1] = '=';
265 else if (val == param+strlen(param)+2) {
266 val[-2] = '=';
267 memmove(val-1, val, strlen(val)+1);
268 val--;
269 } else
270 BUG();
273 /* Handle obsolete-style parameters */
274 if (obsolete_checksetup(param))
275 return 0;
278 * Preemptive maintenance for "why didn't my mispelled command
279 * line work?"
281 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
282 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
283 return 0;
286 if (panic_later)
287 return 0;
289 if (val) {
290 /* Environment option */
291 unsigned int i;
292 for (i = 0; envp_init[i]; i++) {
293 if (i == MAX_INIT_ENVS) {
294 panic_later = "Too many boot env vars at `%s'";
295 panic_param = param;
297 if (!strncmp(param, envp_init[i], val - param))
298 break;
300 envp_init[i] = param;
301 } else {
302 /* Command line option */
303 unsigned int i;
304 for (i = 0; argv_init[i]; i++) {
305 if (i == MAX_INIT_ARGS) {
306 panic_later = "Too many boot init vars at `%s'";
307 panic_param = param;
310 argv_init[i] = param;
312 return 0;
315 static int __init init_setup(char *str)
317 unsigned int i;
319 execute_command = str;
321 * In case LILO is going to boot us with default command line,
322 * it prepends "auto" before the whole cmdline which makes
323 * the shell think it should execute a script with such name.
324 * So we ignore all arguments entered _before_ init=... [MJ]
326 for (i = 1; i < MAX_INIT_ARGS; i++)
327 argv_init[i] = NULL;
328 return 1;
330 __setup("init=", init_setup);
332 static int __init rdinit_setup(char *str)
334 unsigned int i;
336 ramdisk_execute_command = str;
337 /* See "auto" comment in init_setup */
338 for (i = 1; i < MAX_INIT_ARGS; i++)
339 argv_init[i] = NULL;
340 return 1;
342 __setup("rdinit=", rdinit_setup);
344 #ifndef CONFIG_SMP
346 #ifdef CONFIG_X86_LOCAL_APIC
347 static void __init smp_init(void)
349 APIC_init_uniprocessor();
351 #else
352 #define smp_init() do { } while (0)
353 #endif
355 static inline void setup_per_cpu_areas(void) { }
356 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
358 #else
360 #ifdef __GENERIC_PER_CPU
361 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
363 EXPORT_SYMBOL(__per_cpu_offset);
365 static void __init setup_per_cpu_areas(void)
367 unsigned long size, i;
368 char *ptr;
369 unsigned long nr_possible_cpus = num_possible_cpus();
371 /* Copy section for each CPU (we discard the original) */
372 size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
373 ptr = alloc_bootmem_pages(size * nr_possible_cpus);
375 for_each_possible_cpu(i) {
376 __per_cpu_offset[i] = ptr - __per_cpu_start;
377 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
378 ptr += size;
381 #endif /* !__GENERIC_PER_CPU */
383 /* Called by boot processor to activate the rest. */
384 static void __init smp_init(void)
386 unsigned int cpu;
388 /* FIXME: This should be done in userspace --RR */
389 for_each_present_cpu(cpu) {
390 if (num_online_cpus() >= max_cpus)
391 break;
392 if (!cpu_online(cpu))
393 cpu_up(cpu);
396 /* Any cleanup work */
397 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
398 smp_cpus_done(max_cpus);
401 #endif
404 * We need to store the untouched command line for future reference.
405 * We also need to store the touched command line since the parameter
406 * parsing is performed in place, and we should allow a component to
407 * store reference of name/value for future reference.
409 static void __init setup_command_line(char *command_line)
411 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
412 static_command_line = alloc_bootmem(strlen (command_line)+1);
413 strcpy (saved_command_line, boot_command_line);
414 strcpy (static_command_line, command_line);
418 * We need to finalize in a non-__init function or else race conditions
419 * between the root thread and the init thread may cause start_kernel to
420 * be reaped by free_initmem before the root thread has proceeded to
421 * cpu_idle.
423 * gcc-3.4 accidentally inlines this function, so use noinline.
426 static void noinline __init_refok rest_init(void)
427 __releases(kernel_lock)
429 int pid;
431 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
432 numa_default_policy();
433 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
434 kthreadd_task = find_task_by_pid(pid);
435 unlock_kernel();
438 * The boot idle thread must execute schedule()
439 * at least once to get things moving:
441 init_idle_bootup_task(current);
442 preempt_enable_no_resched();
443 schedule();
444 preempt_disable();
446 /* Call into cpu_idle with preempt disabled */
447 cpu_idle();
450 /* Check for early params. */
451 static int __init do_early_param(char *param, char *val)
453 struct obs_kernel_param *p;
455 for (p = __setup_start; p < __setup_end; p++) {
456 if (p->early && strcmp(param, p->str) == 0) {
457 if (p->setup_func(val) != 0)
458 printk(KERN_WARNING
459 "Malformed early option '%s'\n", param);
462 /* We accept everything at this stage. */
463 return 0;
466 /* Arch code calls this early on, or if not, just before other parsing. */
467 void __init parse_early_param(void)
469 static __initdata int done = 0;
470 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
472 if (done)
473 return;
475 /* All fall through to do_early_param. */
476 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
477 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
478 done = 1;
482 * Activate the first processor.
485 static void __init boot_cpu_init(void)
487 int cpu = smp_processor_id();
488 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
489 cpu_set(cpu, cpu_online_map);
490 cpu_set(cpu, cpu_present_map);
491 cpu_set(cpu, cpu_possible_map);
494 void __init __attribute__((weak)) smp_setup_processor_id(void)
498 asmlinkage void __init start_kernel(void)
500 char * command_line;
501 extern struct kernel_param __start___param[], __stop___param[];
503 smp_setup_processor_id();
506 * Need to run as early as possible, to initialize the
507 * lockdep hash:
509 unwind_init();
510 lockdep_init();
512 local_irq_disable();
513 early_boot_irqs_off();
514 early_init_irq_lock_class();
517 * Interrupts are still disabled. Do necessary setups, then
518 * enable them
520 lock_kernel();
521 tick_init();
522 boot_cpu_init();
523 page_address_init();
524 printk(KERN_NOTICE);
525 printk(linux_banner);
526 setup_arch(&command_line);
527 setup_command_line(command_line);
528 unwind_setup();
529 setup_per_cpu_areas();
530 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
533 * Set up the scheduler prior starting any interrupts (such as the
534 * timer interrupt). Full topology setup happens at smp_init()
535 * time - but meanwhile we still have a functioning scheduler.
537 sched_init();
539 * Disable preemption - early bootup scheduling is extremely
540 * fragile until we cpu_idle() for the first time.
542 preempt_disable();
543 build_all_zonelists();
544 page_alloc_init();
545 printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
546 parse_early_param();
547 parse_args("Booting kernel", static_command_line, __start___param,
548 __stop___param - __start___param,
549 &unknown_bootoption);
550 if (!irqs_disabled()) {
551 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
552 "enabled *very* early, fixing it\n");
553 local_irq_disable();
555 sort_main_extable();
556 trap_init();
557 rcu_init();
558 init_IRQ();
559 pidhash_init();
560 init_timers();
561 hrtimers_init();
562 softirq_init();
563 timekeeping_init();
564 time_init();
565 profile_init();
566 if (!irqs_disabled())
567 printk("start_kernel(): bug: interrupts were enabled early\n");
568 early_boot_irqs_on();
569 local_irq_enable();
572 * HACK ALERT! This is early. We're enabling the console before
573 * we've done PCI setups etc, and console_init() must be aware of
574 * this. But we do want output early, in case something goes wrong.
576 console_init();
577 if (panic_later)
578 panic(panic_later, panic_param);
580 lockdep_info();
583 * Need to run this when irqs are enabled, because it wants
584 * to self-test [hard/soft]-irqs on/off lock inversion bugs
585 * too:
587 locking_selftest();
589 #ifdef CONFIG_BLK_DEV_INITRD
590 if (initrd_start && !initrd_below_start_ok &&
591 initrd_start < min_low_pfn << PAGE_SHIFT) {
592 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
593 "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
594 initrd_start = 0;
596 #endif
597 vfs_caches_init_early();
598 cpuset_init_early();
599 mem_init();
600 kmem_cache_init();
601 setup_per_cpu_pageset();
602 numa_policy_init();
603 if (late_time_init)
604 late_time_init();
605 calibrate_delay();
606 pidmap_init();
607 pgtable_cache_init();
608 prio_tree_init();
609 anon_vma_init();
610 #ifdef CONFIG_X86
611 if (efi_enabled)
612 efi_enter_virtual_mode();
613 #endif
614 fork_init(num_physpages);
615 proc_caches_init();
616 buffer_init();
617 unnamed_dev_init();
618 key_init();
619 security_init();
620 vfs_caches_init(num_physpages);
621 radix_tree_init();
622 signals_init();
623 /* rootfs populating might need page-writeback */
624 page_writeback_init();
625 #ifdef CONFIG_PROC_FS
626 proc_root_init();
627 #endif
628 cpuset_init();
629 taskstats_init_early();
630 delayacct_init();
632 check_bugs();
634 acpi_early_init(); /* before LAPIC and SMP init */
636 /* Do the rest non-__init'ed, we're now alive */
637 rest_init();
640 static int __initdata initcall_debug;
642 static int __init initcall_debug_setup(char *str)
644 initcall_debug = 1;
645 return 1;
647 __setup("initcall_debug", initcall_debug_setup);
649 extern initcall_t __initcall_start[], __initcall_end[];
651 static void __init do_initcalls(void)
653 initcall_t *call;
654 int count = preempt_count();
656 for (call = __initcall_start; call < __initcall_end; call++) {
657 ktime_t t0, t1, delta;
658 char *msg = NULL;
659 char msgbuf[40];
660 int result;
662 if (initcall_debug) {
663 printk("Calling initcall 0x%p", *call);
664 print_fn_descriptor_symbol(": %s()",
665 (unsigned long) *call);
666 printk("\n");
667 t0 = ktime_get();
670 result = (*call)();
672 if (initcall_debug) {
673 t1 = ktime_get();
674 delta = ktime_sub(t1, t0);
676 printk("initcall 0x%p", *call);
677 print_fn_descriptor_symbol(": %s()",
678 (unsigned long) *call);
679 printk(" returned %d.\n", result);
681 printk("initcall 0x%p ran for %Ld msecs: ",
682 *call, (unsigned long long)delta.tv64 >> 20);
683 print_fn_descriptor_symbol("%s()\n",
684 (unsigned long) *call);
687 if (result && result != -ENODEV && initcall_debug) {
688 sprintf(msgbuf, "error code %d", result);
689 msg = msgbuf;
691 if (preempt_count() != count) {
692 msg = "preemption imbalance";
693 preempt_count() = count;
695 if (irqs_disabled()) {
696 msg = "disabled interrupts";
697 local_irq_enable();
699 if (msg) {
700 printk(KERN_WARNING "initcall at 0x%p", *call);
701 print_fn_descriptor_symbol(": %s()",
702 (unsigned long) *call);
703 printk(": returned with %s\n", msg);
707 /* Make sure there is no pending stuff from the initcall sequence */
708 flush_scheduled_work();
712 * Ok, the machine is now initialized. None of the devices
713 * have been touched yet, but the CPU subsystem is up and
714 * running, and memory and process management works.
716 * Now we can finally start doing some real work..
718 static void __init do_basic_setup(void)
720 /* drivers will send hotplug events */
721 init_workqueues();
722 usermodehelper_init();
723 driver_init();
724 init_irq_proc();
725 do_initcalls();
728 static void __init do_pre_smp_initcalls(void)
730 extern int spawn_ksoftirqd(void);
731 #ifdef CONFIG_SMP
732 extern int migration_init(void);
734 migration_init();
735 #endif
736 spawn_ksoftirqd();
737 spawn_softlockup_task();
740 static void run_init_process(char *init_filename)
742 argv_init[0] = init_filename;
743 kernel_execve(init_filename, argv_init, envp_init);
746 /* This is a non __init function. Force it to be noinline otherwise gcc
747 * makes it inline to init() and it becomes part of init.text section
749 static int noinline init_post(void)
751 free_initmem();
752 unlock_kernel();
753 mark_rodata_ro();
754 system_state = SYSTEM_RUNNING;
755 numa_default_policy();
757 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
758 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
760 (void) sys_dup(0);
761 (void) sys_dup(0);
763 if (ramdisk_execute_command) {
764 run_init_process(ramdisk_execute_command);
765 printk(KERN_WARNING "Failed to execute %s\n",
766 ramdisk_execute_command);
770 * We try each of these until one succeeds.
772 * The Bourne shell can be used instead of init if we are
773 * trying to recover a really broken machine.
775 if (execute_command) {
776 run_init_process(execute_command);
777 printk(KERN_WARNING "Failed to execute %s. Attempting "
778 "defaults...\n", execute_command);
780 run_init_process("/sbin/init");
781 run_init_process("/etc/init");
782 run_init_process("/bin/init");
783 run_init_process("/bin/sh");
785 panic("No init found. Try passing init= option to kernel.");
788 static int __init kernel_init(void * unused)
790 lock_kernel();
792 * init can run on any cpu.
794 set_cpus_allowed(current, CPU_MASK_ALL);
796 * Tell the world that we're going to be the grim
797 * reaper of innocent orphaned children.
799 * We don't want people to have to make incorrect
800 * assumptions about where in the task array this
801 * can be found.
803 init_pid_ns.child_reaper = current;
805 __set_special_pids(1, 1);
806 cad_pid = task_pid(current);
808 smp_prepare_cpus(max_cpus);
810 do_pre_smp_initcalls();
812 smp_init();
813 sched_init_smp();
815 cpuset_init_smp();
817 do_basic_setup();
820 * check if there is an early userspace init. If yes, let it do all
821 * the work
824 if (!ramdisk_execute_command)
825 ramdisk_execute_command = "/init";
827 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
828 ramdisk_execute_command = NULL;
829 prepare_namespace();
833 * Ok, we have completed the initial bootup, and
834 * we're essentially up and running. Get rid of the
835 * initmem segments and start the user-mode stuff..
837 init_post();
838 return 0;