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/bootmem.h>
26 #include <linux/tty.h>
27 #include <linux/gfp.h>
28 #include <linux/percpu.h>
29 #include <linux/kmod.h>
30 #include <linux/vmalloc.h>
31 #include <linux/kernel_stat.h>
32 #include <linux/start_kernel.h>
33 #include <linux/security.h>
34 #include <linux/smp.h>
35 #include <linux/workqueue.h>
36 #include <linux/profile.h>
37 #include <linux/rcupdate.h>
38 #include <linux/moduleparam.h>
39 #include <linux/kallsyms.h>
40 #include <linux/writeback.h>
41 #include <linux/cpu.h>
42 #include <linux/cpuset.h>
43 #include <linux/cgroup.h>
44 #include <linux/efi.h>
45 #include <linux/tick.h>
46 #include <linux/interrupt.h>
47 #include <linux/taskstats_kern.h>
48 #include <linux/delayacct.h>
49 #include <linux/unistd.h>
50 #include <linux/rmap.h>
51 #include <linux/mempolicy.h>
52 #include <linux/key.h>
53 #include <linux/unwind.h>
54 #include <linux/buffer_head.h>
55 #include <linux/page_cgroup.h>
56 #include <linux/debug_locks.h>
57 #include <linux/debugobjects.h>
58 #include <linux/lockdep.h>
59 #include <linux/pid_namespace.h>
60 #include <linux/device.h>
61 #include <linux/kthread.h>
62 #include <linux/sched.h>
63 #include <linux/signal.h>
64 #include <linux/idr.h>
65 #include <linux/ftrace.h>
66 #include <trace/boot.h>
70 #include <asm/setup.h>
71 #include <asm/sections.h>
72 #include <asm/cacheflush.h>
74 #ifdef CONFIG_X86_LOCAL_APIC
79 * This is one of the first .c files built. Error out early if we have compiler
83 #if __GNUC__ == 4 && __GNUC_MINOR__ == 1 && __GNUC_PATCHLEVEL__ == 0
84 #warning gcc-4.1.0 is known to miscompile the kernel. A different compiler version is recommended.
87 static int kernel_init(void *);
89 extern void init_IRQ(void);
90 extern void fork_init(unsigned long);
91 extern void mca_init(void);
92 extern void sbus_init(void);
93 extern void prio_tree_init(void);
94 extern void radix_tree_init(void);
95 extern void free_initmem(void);
97 extern void acpi_early_init(void);
99 static inline void acpi_early_init(void) { }
101 #ifndef CONFIG_DEBUG_RODATA
102 static inline void mark_rodata_ro(void) { }
106 extern void tc_init(void);
109 enum system_states system_state;
110 EXPORT_SYMBOL(system_state);
113 * Boot command-line arguments
115 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
116 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
118 extern void time_init(void);
119 /* Default late time init is NULL. archs can override this later. */
120 void (*late_time_init)(void);
121 extern void softirq_init(void);
123 /* Untouched command line saved by arch-specific code. */
124 char __initdata boot_command_line[COMMAND_LINE_SIZE];
125 /* Untouched saved command line (eg. for /proc) */
126 char *saved_command_line;
127 /* Command line for parameter parsing */
128 static char *static_command_line;
130 static char *execute_command;
131 static char *ramdisk_execute_command;
134 /* Setup configured maximum number of CPUs to activate */
135 unsigned int __initdata setup_max_cpus = NR_CPUS;
138 * Setup routine for controlling SMP activation
140 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
141 * activation entirely (the MPS table probe still happens, though).
143 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
144 * greater than 0, limits the maximum number of CPUs activated in
147 #ifndef CONFIG_X86_IO_APIC
148 static inline void disable_ioapic_setup(void) {};
151 static int __init nosmp(char *str)
154 disable_ioapic_setup();
158 early_param("nosmp", nosmp);
160 static int __init maxcpus(char *str)
162 get_option(&str, &setup_max_cpus);
163 if (setup_max_cpus == 0)
164 disable_ioapic_setup();
169 early_param("maxcpus", maxcpus);
171 #define setup_max_cpus NR_CPUS
175 * If set, this is an indication to the drivers that reset the underlying
176 * device before going ahead with the initialization otherwise driver might
177 * rely on the BIOS and skip the reset operation.
179 * This is useful if kernel is booting in an unreliable environment.
180 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
181 * skipped and devices will be in unknown state.
183 unsigned int reset_devices;
184 EXPORT_SYMBOL(reset_devices);
186 static int __init set_reset_devices(char *str)
192 __setup("reset_devices", set_reset_devices);
194 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
195 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
196 static const char *panic_later, *panic_param;
198 extern struct obs_kernel_param __setup_start[], __setup_end[];
200 static int __init obsolete_checksetup(char *line)
202 struct obs_kernel_param *p;
203 int had_early_param = 0;
207 int n = strlen(p->str);
208 if (!strncmp(line, p->str, n)) {
210 /* Already done in parse_early_param?
211 * (Needs exact match on param part).
212 * Keep iterating, as we can have early
213 * params and __setups of same names 8( */
214 if (line[n] == '\0' || line[n] == '=')
216 } else if (!p->setup_func) {
217 printk(KERN_WARNING "Parameter %s is obsolete,"
218 " ignored\n", p->str);
220 } else if (p->setup_func(line + n))
224 } while (p < __setup_end);
226 return had_early_param;
230 * This should be approx 2 Bo*oMips to start (note initial shift), and will
231 * still work even if initially too large, it will just take slightly longer
233 unsigned long loops_per_jiffy = (1<<12);
235 EXPORT_SYMBOL(loops_per_jiffy);
237 static int __init debug_kernel(char *str)
239 console_loglevel = 10;
243 static int __init quiet_kernel(char *str)
245 console_loglevel = 4;
249 early_param("debug", debug_kernel);
250 early_param("quiet", quiet_kernel);
252 static int __init loglevel(char *str)
254 get_option(&str, &console_loglevel);
258 early_param("loglevel", loglevel);
261 * Unknown boot options get handed to init, unless they look like
264 static int __init unknown_bootoption(char *param, char *val)
266 /* Change NUL term back to "=", to make "param" the whole string. */
268 /* param=val or param="val"? */
269 if (val == param+strlen(param)+1)
271 else if (val == param+strlen(param)+2) {
273 memmove(val-1, val, strlen(val)+1);
279 /* Handle obsolete-style parameters */
280 if (obsolete_checksetup(param))
284 * Preemptive maintenance for "why didn't my misspelled command
287 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
288 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
296 /* Environment option */
298 for (i = 0; envp_init[i]; i++) {
299 if (i == MAX_INIT_ENVS) {
300 panic_later = "Too many boot env vars at `%s'";
303 if (!strncmp(param, envp_init[i], val - param))
306 envp_init[i] = param;
308 /* Command line option */
310 for (i = 0; argv_init[i]; i++) {
311 if (i == MAX_INIT_ARGS) {
312 panic_later = "Too many boot init vars at `%s'";
316 argv_init[i] = param;
321 #ifdef CONFIG_DEBUG_PAGEALLOC
322 int __read_mostly debug_pagealloc_enabled = 0;
325 static int __init init_setup(char *str)
329 execute_command = str;
331 * In case LILO is going to boot us with default command line,
332 * it prepends "auto" before the whole cmdline which makes
333 * the shell think it should execute a script with such name.
334 * So we ignore all arguments entered _before_ init=... [MJ]
336 for (i = 1; i < MAX_INIT_ARGS; i++)
340 __setup("init=", init_setup);
342 static int __init rdinit_setup(char *str)
346 ramdisk_execute_command = str;
347 /* See "auto" comment in init_setup */
348 for (i = 1; i < MAX_INIT_ARGS; i++)
352 __setup("rdinit=", rdinit_setup);
356 #ifdef CONFIG_X86_LOCAL_APIC
357 static void __init smp_init(void)
359 APIC_init_uniprocessor();
362 #define smp_init() do { } while (0)
365 static inline void setup_per_cpu_areas(void) { }
366 static inline void setup_nr_cpu_ids(void) { }
367 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
371 #if NR_CPUS > BITS_PER_LONG
372 cpumask_t cpu_mask_all __read_mostly = CPU_MASK_ALL;
373 EXPORT_SYMBOL(cpu_mask_all);
376 /* Setup number of possible processor ids */
377 int nr_cpu_ids __read_mostly = NR_CPUS;
378 EXPORT_SYMBOL(nr_cpu_ids);
380 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
381 static void __init setup_nr_cpu_ids(void)
383 nr_cpu_ids = find_last_bit(cpumask_bits(cpu_possible_mask),NR_CPUS) + 1;
386 #ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA
387 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
389 EXPORT_SYMBOL(__per_cpu_offset);
391 static void __init setup_per_cpu_areas(void)
393 unsigned long size, i;
395 unsigned long nr_possible_cpus = num_possible_cpus();
397 /* Copy section for each CPU (we discard the original) */
398 size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
399 ptr = alloc_bootmem_pages(size * nr_possible_cpus);
401 for_each_possible_cpu(i) {
402 __per_cpu_offset[i] = ptr - __per_cpu_start;
403 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
407 #endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */
409 /* Called by boot processor to activate the rest. */
410 static void __init smp_init(void)
415 * Set up the current CPU as possible to migrate to.
416 * The other ones will be done by cpu_up/cpu_down()
418 cpu = smp_processor_id();
419 cpu_set(cpu, cpu_active_map);
421 /* FIXME: This should be done in userspace --RR */
422 for_each_present_cpu(cpu) {
423 if (num_online_cpus() >= setup_max_cpus)
425 if (!cpu_online(cpu))
429 /* Any cleanup work */
430 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
431 smp_cpus_done(setup_max_cpus);
437 * We need to store the untouched command line for future reference.
438 * We also need to store the touched command line since the parameter
439 * parsing is performed in place, and we should allow a component to
440 * store reference of name/value for future reference.
442 static void __init setup_command_line(char *command_line)
444 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
445 static_command_line = alloc_bootmem(strlen (command_line)+1);
446 strcpy (saved_command_line, boot_command_line);
447 strcpy (static_command_line, command_line);
451 * We need to finalize in a non-__init function or else race conditions
452 * between the root thread and the init thread may cause start_kernel to
453 * be reaped by free_initmem before the root thread has proceeded to
456 * gcc-3.4 accidentally inlines this function, so use noinline.
459 static void noinline __init_refok rest_init(void)
460 __releases(kernel_lock)
464 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
465 numa_default_policy();
466 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
467 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
471 * The boot idle thread must execute schedule()
472 * at least once to get things moving:
474 init_idle_bootup_task(current);
475 preempt_enable_no_resched();
479 /* Call into cpu_idle with preempt disabled */
483 /* Check for early params. */
484 static int __init do_early_param(char *param, char *val)
486 struct obs_kernel_param *p;
488 for (p = __setup_start; p < __setup_end; p++) {
489 if ((p->early && strcmp(param, p->str) == 0) ||
490 (strcmp(param, "console") == 0 &&
491 strcmp(p->str, "earlycon") == 0)
493 if (p->setup_func(val) != 0)
495 "Malformed early option '%s'\n", param);
498 /* We accept everything at this stage. */
502 /* Arch code calls this early on, or if not, just before other parsing. */
503 void __init parse_early_param(void)
505 static __initdata int done = 0;
506 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
511 /* All fall through to do_early_param. */
512 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
513 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
518 * Activate the first processor.
521 static void __init boot_cpu_init(void)
523 int cpu = smp_processor_id();
524 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
525 set_cpu_online(cpu, true);
526 set_cpu_present(cpu, true);
527 set_cpu_possible(cpu, true);
530 void __init __weak smp_setup_processor_id(void)
534 void __init __weak thread_info_cache_init(void)
538 void __init __weak arch_early_irq_init(void)
542 void __init __weak early_irq_init(void)
544 arch_early_irq_init();
547 asmlinkage void __init start_kernel(void)
550 extern struct kernel_param __start___param[], __stop___param[];
552 smp_setup_processor_id();
555 * Need to run as early as possible, to initialize the
560 debug_objects_early_init();
564 early_boot_irqs_off();
565 early_init_irq_lock_class();
568 * Interrupts are still disabled. Do necessary setups, then
576 printk(linux_banner);
577 setup_arch(&command_line);
578 mm_init_owner(&init_mm, &init_task);
579 setup_command_line(command_line);
581 setup_per_cpu_areas();
583 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
586 * Set up the scheduler prior starting any interrupts (such as the
587 * timer interrupt). Full topology setup happens at smp_init()
588 * time - but meanwhile we still have a functioning scheduler.
592 * Disable preemption - early bootup scheduling is extremely
593 * fragile until we cpu_idle() for the first time.
596 build_all_zonelists();
598 printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
600 parse_args("Booting kernel", static_command_line, __start___param,
601 __stop___param - __start___param,
602 &unknown_bootoption);
603 if (!irqs_disabled()) {
604 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
605 "enabled *very* early, fixing it\n");
611 /* init some links before init_ISA_irqs() */
622 if (!irqs_disabled())
623 printk("start_kernel(): bug: interrupts were enabled early\n");
624 early_boot_irqs_on();
628 * HACK ALERT! This is early. We're enabling the console before
629 * we've done PCI setups etc, and console_init() must be aware of
630 * this. But we do want output early, in case something goes wrong.
634 panic(panic_later, panic_param);
639 * Need to run this when irqs are enabled, because it wants
640 * to self-test [hard/soft]-irqs on/off lock inversion bugs
645 #ifdef CONFIG_BLK_DEV_INITRD
646 if (initrd_start && !initrd_below_start_ok &&
647 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
648 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
650 page_to_pfn(virt_to_page((void *)initrd_start)),
656 vfs_caches_init_early();
660 enable_debug_pagealloc();
663 debug_objects_mem_init();
665 setup_per_cpu_pageset();
671 pgtable_cache_init();
676 efi_enter_virtual_mode();
678 thread_info_cache_init();
680 fork_init(num_physpages);
685 vfs_caches_init(num_physpages);
688 /* rootfs populating might need page-writeback */
689 page_writeback_init();
690 #ifdef CONFIG_PROC_FS
695 taskstats_init_early();
700 acpi_early_init(); /* before LAPIC and SMP init */
704 /* Do the rest non-__init'ed, we're now alive */
708 static int initcall_debug;
709 core_param(initcall_debug, initcall_debug, bool, 0644);
711 int do_one_initcall(initcall_t fn)
713 int count = preempt_count();
714 ktime_t calltime, delta, rettime;
716 struct boot_trace_call call;
717 struct boot_trace_ret ret;
719 if (initcall_debug) {
720 call.caller = task_pid_nr(current);
721 printk("calling %pF @ %i\n", fn, call.caller);
722 calltime = ktime_get();
723 trace_boot_call(&call, fn);
729 if (initcall_debug) {
730 disable_boot_trace();
731 rettime = ktime_get();
732 delta = ktime_sub(rettime, calltime);
733 ret.duration = (unsigned long long) ktime_to_ns(delta) >> 10;
734 trace_boot_ret(&ret, fn);
735 printk("initcall %pF returned %d after %Ld usecs\n", fn,
736 ret.result, ret.duration);
741 if (ret.result && ret.result != -ENODEV && initcall_debug)
742 sprintf(msgbuf, "error code %d ", ret.result);
744 if (preempt_count() != count) {
745 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
746 preempt_count() = count;
748 if (irqs_disabled()) {
749 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
753 printk("initcall %pF returned with %s\n", fn, msgbuf);
760 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
762 static void __init do_initcalls(void)
766 for (call = __early_initcall_end; call < __initcall_end; call++)
767 do_one_initcall(*call);
769 /* Make sure there is no pending stuff from the initcall sequence */
770 flush_scheduled_work();
774 * Ok, the machine is now initialized. None of the devices
775 * have been touched yet, but the CPU subsystem is up and
776 * running, and memory and process management works.
778 * Now we can finally start doing some real work..
780 static void __init do_basic_setup(void)
782 rcu_init_sched(); /* needed by module_init stage. */
784 usermodehelper_init();
790 static void __init do_pre_smp_initcalls(void)
794 for (call = __initcall_start; call < __early_initcall_end; call++)
795 do_one_initcall(*call);
798 static void run_init_process(char *init_filename)
800 argv_init[0] = init_filename;
801 kernel_execve(init_filename, argv_init, envp_init);
804 /* This is a non __init function. Force it to be noinline otherwise gcc
805 * makes it inline to init() and it becomes part of init.text section
807 static int noinline init_post(void)
812 system_state = SYSTEM_RUNNING;
813 numa_default_policy();
815 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
816 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
821 current->signal->flags |= SIGNAL_UNKILLABLE;
823 if (ramdisk_execute_command) {
824 run_init_process(ramdisk_execute_command);
825 printk(KERN_WARNING "Failed to execute %s\n",
826 ramdisk_execute_command);
830 * We try each of these until one succeeds.
832 * The Bourne shell can be used instead of init if we are
833 * trying to recover a really broken machine.
835 if (execute_command) {
836 run_init_process(execute_command);
837 printk(KERN_WARNING "Failed to execute %s. Attempting "
838 "defaults...\n", execute_command);
840 run_init_process("/sbin/init");
841 run_init_process("/etc/init");
842 run_init_process("/bin/init");
843 run_init_process("/bin/sh");
845 panic("No init found. Try passing init= option to kernel.");
848 static int __init kernel_init(void * unused)
852 * init can run on any cpu.
854 set_cpus_allowed_ptr(current, CPU_MASK_ALL_PTR);
856 * Tell the world that we're going to be the grim
857 * reaper of innocent orphaned children.
859 * We don't want people to have to make incorrect
860 * assumptions about where in the task array this
863 init_pid_ns.child_reaper = current;
865 cad_pid = task_pid(current);
867 smp_prepare_cpus(setup_max_cpus);
869 do_pre_smp_initcalls();
880 * check if there is an early userspace init. If yes, let it do all
884 if (!ramdisk_execute_command)
885 ramdisk_execute_command = "/init";
887 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
888 ramdisk_execute_command = NULL;
893 * Ok, we have completed the initial bootup, and
894 * we're essentially up and running. Get rid of the
895 * initmem segments and start the user-mode stuff..