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/stackprotector.h>
18 #include <linux/string.h>
19 #include <linux/ctype.h>
20 #include <linux/delay.h>
21 #include <linux/utsname.h>
22 #include <linux/ioport.h>
23 #include <linux/init.h>
24 #include <linux/smp_lock.h>
25 #include <linux/initrd.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/vmalloc.h>
32 #include <linux/kernel_stat.h>
33 #include <linux/start_kernel.h>
34 #include <linux/security.h>
35 #include <linux/smp.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/cpuset.h>
44 #include <linux/cgroup.h>
45 #include <linux/efi.h>
46 #include <linux/tick.h>
47 #include <linux/interrupt.h>
48 #include <linux/taskstats_kern.h>
49 #include <linux/delayacct.h>
50 #include <linux/unistd.h>
51 #include <linux/rmap.h>
52 #include <linux/mempolicy.h>
53 #include <linux/key.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/kmemleak.h>
60 #include <linux/pid_namespace.h>
61 #include <linux/device.h>
62 #include <linux/kthread.h>
63 #include <linux/sched.h>
64 #include <linux/signal.h>
65 #include <linux/idr.h>
66 #include <linux/ftrace.h>
67 #include <linux/async.h>
68 #include <linux/kmemtrace.h>
69 #include <trace/boot.h>
73 #include <asm/setup.h>
74 #include <asm/sections.h>
75 #include <asm/cacheflush.h>
77 #ifdef CONFIG_X86_LOCAL_APIC
81 static int kernel_init(void *);
83 extern void init_IRQ(void);
84 extern void fork_init(unsigned long);
85 extern void mca_init(void);
86 extern void sbus_init(void);
87 extern void prio_tree_init(void);
88 extern void radix_tree_init(void);
89 extern void free_initmem(void);
91 extern void acpi_early_init(void);
93 static inline void acpi_early_init(void) { }
95 #ifndef CONFIG_DEBUG_RODATA
96 static inline void mark_rodata_ro(void) { }
100 extern void tc_init(void);
103 enum system_states system_state __read_mostly;
104 EXPORT_SYMBOL(system_state);
107 * Boot command-line arguments
109 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
110 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
112 extern void time_init(void);
113 /* Default late time init is NULL. archs can override this later. */
114 void (*__initdata late_time_init)(void);
115 extern void softirq_init(void);
117 /* Untouched command line saved by arch-specific code. */
118 char __initdata boot_command_line[COMMAND_LINE_SIZE];
119 /* Untouched saved command line (eg. for /proc) */
120 char *saved_command_line;
121 /* Command line for parameter parsing */
122 static char *static_command_line;
124 static char *execute_command;
125 static char *ramdisk_execute_command;
128 /* Setup configured maximum number of CPUs to activate */
129 unsigned int __initdata setup_max_cpus = NR_CPUS;
132 * Setup routine for controlling SMP activation
134 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
135 * activation entirely (the MPS table probe still happens, though).
137 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
138 * greater than 0, limits the maximum number of CPUs activated in
142 void __weak arch_disable_smp_support(void) { }
144 static int __init nosmp(char *str)
147 arch_disable_smp_support();
152 early_param("nosmp", nosmp);
154 static int __init maxcpus(char *str)
156 get_option(&str, &setup_max_cpus);
157 if (setup_max_cpus == 0)
158 arch_disable_smp_support();
163 early_param("maxcpus", maxcpus);
165 const unsigned int setup_max_cpus = NR_CPUS;
169 * If set, this is an indication to the drivers that reset the underlying
170 * device before going ahead with the initialization otherwise driver might
171 * rely on the BIOS and skip the reset operation.
173 * This is useful if kernel is booting in an unreliable environment.
174 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
175 * skipped and devices will be in unknown state.
177 unsigned int reset_devices;
178 EXPORT_SYMBOL(reset_devices);
180 static int __init set_reset_devices(char *str)
186 __setup("reset_devices", set_reset_devices);
188 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
189 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
190 static const char *panic_later, *panic_param;
192 extern struct obs_kernel_param __setup_start[], __setup_end[];
194 static int __init obsolete_checksetup(char *line)
196 struct obs_kernel_param *p;
197 int had_early_param = 0;
201 int n = strlen(p->str);
202 if (!strncmp(line, p->str, n)) {
204 /* Already done in parse_early_param?
205 * (Needs exact match on param part).
206 * Keep iterating, as we can have early
207 * params and __setups of same names 8( */
208 if (line[n] == '\0' || line[n] == '=')
210 } else if (!p->setup_func) {
211 printk(KERN_WARNING "Parameter %s is obsolete,"
212 " ignored\n", p->str);
214 } else if (p->setup_func(line + n))
218 } while (p < __setup_end);
220 return had_early_param;
224 * This should be approx 2 Bo*oMips to start (note initial shift), and will
225 * still work even if initially too large, it will just take slightly longer
227 unsigned long loops_per_jiffy = (1<<12);
229 EXPORT_SYMBOL(loops_per_jiffy);
231 static int __init debug_kernel(char *str)
233 console_loglevel = 10;
237 static int __init quiet_kernel(char *str)
239 console_loglevel = 4;
243 early_param("debug", debug_kernel);
244 early_param("quiet", quiet_kernel);
246 static int __init loglevel(char *str)
248 get_option(&str, &console_loglevel);
252 early_param("loglevel", loglevel);
255 * Unknown boot options get handed to init, unless they look like
258 static int __init unknown_bootoption(char *param, char *val)
260 /* Change NUL term back to "=", to make "param" the whole string. */
262 /* param=val or param="val"? */
263 if (val == param+strlen(param)+1)
265 else if (val == param+strlen(param)+2) {
267 memmove(val-1, val, strlen(val)+1);
273 /* Handle obsolete-style parameters */
274 if (obsolete_checksetup(param))
278 * Preemptive maintenance for "why didn't my misspelled command
281 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
282 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
290 /* Environment option */
292 for (i = 0; envp_init[i]; i++) {
293 if (i == MAX_INIT_ENVS) {
294 panic_later = "Too many boot env vars at `%s'";
297 if (!strncmp(param, envp_init[i], val - param))
300 envp_init[i] = param;
302 /* Command line option */
304 for (i = 0; argv_init[i]; i++) {
305 if (i == MAX_INIT_ARGS) {
306 panic_later = "Too many boot init vars at `%s'";
310 argv_init[i] = param;
315 #ifdef CONFIG_DEBUG_PAGEALLOC
316 int __read_mostly debug_pagealloc_enabled = 0;
319 static int __init init_setup(char *str)
323 execute_command = str;
325 * In case LILO is going to boot us with default command line,
326 * it prepends "auto" before the whole cmdline which makes
327 * the shell think it should execute a script with such name.
328 * So we ignore all arguments entered _before_ init=... [MJ]
330 for (i = 1; i < MAX_INIT_ARGS; i++)
334 __setup("init=", init_setup);
336 static int __init rdinit_setup(char *str)
340 ramdisk_execute_command = str;
341 /* See "auto" comment in init_setup */
342 for (i = 1; i < MAX_INIT_ARGS; i++)
346 __setup("rdinit=", rdinit_setup);
350 #ifdef CONFIG_X86_LOCAL_APIC
351 static void __init smp_init(void)
353 APIC_init_uniprocessor();
356 #define smp_init() do { } while (0)
359 static inline void setup_per_cpu_areas(void) { }
360 static inline void setup_nr_cpu_ids(void) { }
361 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
365 #if NR_CPUS > BITS_PER_LONG
366 cpumask_t cpu_mask_all __read_mostly = CPU_MASK_ALL;
367 EXPORT_SYMBOL(cpu_mask_all);
370 /* Setup number of possible processor ids */
371 int nr_cpu_ids __read_mostly = NR_CPUS;
372 EXPORT_SYMBOL(nr_cpu_ids);
374 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
375 static void __init setup_nr_cpu_ids(void)
377 nr_cpu_ids = find_last_bit(cpumask_bits(cpu_possible_mask),NR_CPUS) + 1;
380 #ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA
381 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
383 EXPORT_SYMBOL(__per_cpu_offset);
385 static void __init setup_per_cpu_areas(void)
387 unsigned long size, i;
389 unsigned long nr_possible_cpus = num_possible_cpus();
391 /* Copy section for each CPU (we discard the original) */
392 size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
393 ptr = alloc_bootmem_pages(size * nr_possible_cpus);
395 for_each_possible_cpu(i) {
396 __per_cpu_offset[i] = ptr - __per_cpu_start;
397 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
401 #endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */
403 /* Called by boot processor to activate the rest. */
404 static void __init smp_init(void)
409 * Set up the current CPU as possible to migrate to.
410 * The other ones will be done by cpu_up/cpu_down()
412 set_cpu_active(smp_processor_id(), true);
414 /* FIXME: This should be done in userspace --RR */
415 for_each_present_cpu(cpu) {
416 if (num_online_cpus() >= setup_max_cpus)
418 if (!cpu_online(cpu))
422 /* Any cleanup work */
423 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
424 smp_cpus_done(setup_max_cpus);
430 * We need to store the untouched command line for future reference.
431 * We also need to store the touched command line since the parameter
432 * parsing is performed in place, and we should allow a component to
433 * store reference of name/value for future reference.
435 static void __init setup_command_line(char *command_line)
437 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
438 static_command_line = alloc_bootmem(strlen (command_line)+1);
439 strcpy (saved_command_line, boot_command_line);
440 strcpy (static_command_line, command_line);
444 * We need to finalize in a non-__init function or else race conditions
445 * between the root thread and the init thread may cause start_kernel to
446 * be reaped by free_initmem before the root thread has proceeded to
449 * gcc-3.4 accidentally inlines this function, so use noinline.
452 static noinline void __init_refok rest_init(void)
453 __releases(kernel_lock)
457 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
458 numa_default_policy();
459 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
460 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
464 * The boot idle thread must execute schedule()
465 * at least once to get things moving:
467 init_idle_bootup_task(current);
468 rcu_scheduler_starting();
469 preempt_enable_no_resched();
473 /* Call into cpu_idle with preempt disabled */
477 /* Check for early params. */
478 static int __init do_early_param(char *param, char *val)
480 struct obs_kernel_param *p;
482 for (p = __setup_start; p < __setup_end; p++) {
483 if ((p->early && strcmp(param, p->str) == 0) ||
484 (strcmp(param, "console") == 0 &&
485 strcmp(p->str, "earlycon") == 0)
487 if (p->setup_func(val) != 0)
489 "Malformed early option '%s'\n", param);
492 /* We accept everything at this stage. */
496 void __init parse_early_options(char *cmdline)
498 parse_args("early options", cmdline, NULL, 0, do_early_param);
501 /* Arch code calls this early on, or if not, just before other parsing. */
502 void __init parse_early_param(void)
504 static __initdata int done = 0;
505 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
510 /* All fall through to do_early_param. */
511 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
512 parse_early_options(tmp_cmdline);
517 * Activate the first processor.
520 static void __init boot_cpu_init(void)
522 int cpu = smp_processor_id();
523 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
524 set_cpu_online(cpu, true);
525 set_cpu_present(cpu, true);
526 set_cpu_possible(cpu, true);
529 void __init __weak smp_setup_processor_id(void)
533 void __init __weak thread_info_cache_init(void)
538 * Set up kernel memory allocators
540 static void __init mm_init(void)
543 * page_cgroup requires countinous pages as memmap
544 * and it's bigger than MAX_ORDER unless SPARSEMEM.
546 page_cgroup_init_flatmem();
552 asmlinkage void __init start_kernel(void)
555 extern struct kernel_param __start___param[], __stop___param[];
557 smp_setup_processor_id();
560 * Need to run as early as possible, to initialize the
564 debug_objects_early_init();
567 * Set up the the initial canary ASAP:
569 boot_init_stack_canary();
574 early_boot_irqs_off();
575 early_init_irq_lock_class();
578 * Interrupts are still disabled. Do necessary setups, then
585 printk(KERN_NOTICE "%s", linux_banner);
586 setup_arch(&command_line);
587 mm_init_owner(&init_mm, &init_task);
588 setup_command_line(command_line);
589 setup_per_cpu_areas();
591 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
593 build_all_zonelists();
596 printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
598 parse_args("Booting kernel", static_command_line, __start___param,
599 __stop___param - __start___param,
600 &unknown_bootoption);
602 * These use large bootmem allocations and must precede
606 vfs_caches_init_early();
611 * Set up the scheduler prior starting any interrupts (such as the
612 * timer interrupt). Full topology setup happens at smp_init()
613 * time - but meanwhile we still have a functioning scheduler.
617 * Disable preemption - early bootup scheduling is extremely
618 * fragile until we cpu_idle() for the first time.
621 if (!irqs_disabled()) {
622 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
623 "enabled *very* early, fixing it\n");
627 /* init some links before init_ISA_irqs() */
638 if (!irqs_disabled())
639 printk(KERN_CRIT "start_kernel(): bug: interrupts were "
641 early_boot_irqs_on();
643 kmem_cache_init_late();
646 * HACK ALERT! This is early. We're enabling the console before
647 * we've done PCI setups etc, and console_init() must be aware of
648 * this. But we do want output early, in case something goes wrong.
652 panic(panic_later, panic_param);
657 * Need to run this when irqs are enabled, because it wants
658 * to self-test [hard/soft]-irqs on/off lock inversion bugs
663 #ifdef CONFIG_BLK_DEV_INITRD
664 if (initrd_start && !initrd_below_start_ok &&
665 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
666 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
668 page_to_pfn(virt_to_page((void *)initrd_start)),
675 enable_debug_pagealloc();
679 debug_objects_mem_init();
681 setup_per_cpu_pageset();
687 pgtable_cache_init();
691 efi_enter_virtual_mode();
693 thread_info_cache_init();
695 fork_init(num_physpages);
700 vfs_caches_init(num_physpages);
703 /* rootfs populating might need page-writeback */
704 page_writeback_init();
705 #ifdef CONFIG_PROC_FS
710 taskstats_init_early();
715 acpi_early_init(); /* before LAPIC and SMP init */
719 /* Do the rest non-__init'ed, we're now alive */
724 core_param(initcall_debug, initcall_debug, bool, 0644);
726 int do_one_initcall(initcall_t fn)
728 int count = preempt_count();
729 ktime_t calltime, delta, rettime;
731 struct boot_trace_call call;
732 struct boot_trace_ret ret;
734 if (initcall_debug) {
735 call.caller = task_pid_nr(current);
736 printk("calling %pF @ %i\n", fn, call.caller);
737 calltime = ktime_get();
738 trace_boot_call(&call, fn);
744 if (initcall_debug) {
745 disable_boot_trace();
746 rettime = ktime_get();
747 delta = ktime_sub(rettime, calltime);
748 ret.duration = (unsigned long long) ktime_to_ns(delta) >> 10;
749 trace_boot_ret(&ret, fn);
750 printk("initcall %pF returned %d after %Ld usecs\n", fn,
751 ret.result, ret.duration);
756 if (ret.result && ret.result != -ENODEV && initcall_debug)
757 sprintf(msgbuf, "error code %d ", ret.result);
759 if (preempt_count() != count) {
760 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
761 preempt_count() = count;
763 if (irqs_disabled()) {
764 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
768 printk("initcall %pF returned with %s\n", fn, msgbuf);
775 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
777 static void __init do_initcalls(void)
781 for (call = __early_initcall_end; call < __initcall_end; call++)
782 do_one_initcall(*call);
784 /* Make sure there is no pending stuff from the initcall sequence */
785 flush_scheduled_work();
789 * Ok, the machine is now initialized. None of the devices
790 * have been touched yet, but the CPU subsystem is up and
791 * running, and memory and process management works.
793 * Now we can finally start doing some real work..
795 static void __init do_basic_setup(void)
797 rcu_init_sched(); /* needed by module_init stage. */
800 usermodehelper_init();
806 static void __init do_pre_smp_initcalls(void)
810 for (call = __initcall_start; call < __early_initcall_end; call++)
811 do_one_initcall(*call);
814 static void run_init_process(char *init_filename)
816 argv_init[0] = init_filename;
817 kernel_execve(init_filename, argv_init, envp_init);
820 /* This is a non __init function. Force it to be noinline otherwise gcc
821 * makes it inline to init() and it becomes part of init.text section
823 static noinline int init_post(void)
824 __releases(kernel_lock)
826 /* need to finish all async __init code before freeing the memory */
827 async_synchronize_full();
831 system_state = SYSTEM_RUNNING;
832 numa_default_policy();
834 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
835 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
840 current->signal->flags |= SIGNAL_UNKILLABLE;
842 if (ramdisk_execute_command) {
843 run_init_process(ramdisk_execute_command);
844 printk(KERN_WARNING "Failed to execute %s\n",
845 ramdisk_execute_command);
849 * We try each of these until one succeeds.
851 * The Bourne shell can be used instead of init if we are
852 * trying to recover a really broken machine.
854 if (execute_command) {
855 run_init_process(execute_command);
856 printk(KERN_WARNING "Failed to execute %s. Attempting "
857 "defaults...\n", execute_command);
859 run_init_process("/sbin/init");
860 run_init_process("/etc/init");
861 run_init_process("/bin/init");
862 run_init_process("/bin/sh");
864 panic("No init found. Try passing init= option to kernel.");
867 static int __init kernel_init(void * unused)
871 * init can run on any cpu.
873 set_cpus_allowed_ptr(current, cpu_all_mask);
875 * Tell the world that we're going to be the grim
876 * reaper of innocent orphaned children.
878 * We don't want people to have to make incorrect
879 * assumptions about where in the task array this
882 init_pid_ns.child_reaper = current;
884 cad_pid = task_pid(current);
886 smp_prepare_cpus(setup_max_cpus);
888 do_pre_smp_initcalls();
897 * check if there is an early userspace init. If yes, let it do all
901 if (!ramdisk_execute_command)
902 ramdisk_execute_command = "/init";
904 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
905 ramdisk_execute_command = NULL;
910 * Ok, we have completed the initial bootup, and
911 * we're essentially up and running. Get rid of the
912 * initmem segments and start the user-mode stuff..