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/kernel_stat.h>
31 #include <linux/start_kernel.h>
32 #include <linux/security.h>
33 #include <linux/smp.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/cgroup.h>
43 #include <linux/efi.h>
44 #include <linux/tick.h>
45 #include <linux/interrupt.h>
46 #include <linux/taskstats_kern.h>
47 #include <linux/delayacct.h>
48 #include <linux/unistd.h>
49 #include <linux/rmap.h>
50 #include <linux/mempolicy.h>
51 #include <linux/key.h>
52 #include <linux/unwind.h>
53 #include <linux/buffer_head.h>
54 #include <linux/debug_locks.h>
55 #include <linux/debugobjects.h>
56 #include <linux/lockdep.h>
57 #include <linux/pid_namespace.h>
58 #include <linux/device.h>
59 #include <linux/kthread.h>
60 #include <linux/sched.h>
61 #include <linux/signal.h>
62 #include <linux/idr.h>
63 #include <linux/ftrace.h>
67 #include <asm/setup.h>
68 #include <asm/sections.h>
69 #include <asm/cacheflush.h>
71 #ifdef CONFIG_X86_LOCAL_APIC
76 * This is one of the first .c files built. Error out early if we have compiler
80 #if __GNUC__ == 4 && __GNUC_MINOR__ == 1 && __GNUC_PATCHLEVEL__ == 0
81 #warning gcc-4.1.0 is known to miscompile the kernel. A different compiler version is recommended.
84 static int kernel_init(void *);
86 extern void init_IRQ(void);
87 extern void fork_init(unsigned long);
88 extern void mca_init(void);
89 extern void sbus_init(void);
90 extern void prio_tree_init(void);
91 extern void radix_tree_init(void);
92 extern void free_initmem(void);
94 extern void acpi_early_init(void);
96 static inline void acpi_early_init(void) { }
98 #ifndef CONFIG_DEBUG_RODATA
99 static inline void mark_rodata_ro(void) { }
103 extern void tc_init(void);
106 enum system_states system_state;
107 EXPORT_SYMBOL(system_state);
110 * Boot command-line arguments
112 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
113 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
115 extern void time_init(void);
116 /* Default late time init is NULL. archs can override this later. */
117 void (*late_time_init)(void);
118 extern void softirq_init(void);
120 /* Untouched command line saved by arch-specific code. */
121 char __initdata boot_command_line[COMMAND_LINE_SIZE];
122 /* Untouched saved command line (eg. for /proc) */
123 char *saved_command_line;
124 /* Command line for parameter parsing */
125 static char *static_command_line;
127 static char *execute_command;
128 static char *ramdisk_execute_command;
131 /* Setup configured maximum number of CPUs to activate */
132 unsigned int __initdata setup_max_cpus = NR_CPUS;
135 * Setup routine for controlling SMP activation
137 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
138 * activation entirely (the MPS table probe still happens, though).
140 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
141 * greater than 0, limits the maximum number of CPUs activated in
144 #ifndef CONFIG_X86_IO_APIC
145 static inline void disable_ioapic_setup(void) {};
148 static int __init nosmp(char *str)
151 disable_ioapic_setup();
155 early_param("nosmp", nosmp);
157 static int __init maxcpus(char *str)
159 get_option(&str, &setup_max_cpus);
160 if (setup_max_cpus == 0)
161 disable_ioapic_setup();
166 early_param("maxcpus", maxcpus);
168 #define setup_max_cpus NR_CPUS
172 * If set, this is an indication to the drivers that reset the underlying
173 * device before going ahead with the initialization otherwise driver might
174 * rely on the BIOS and skip the reset operation.
176 * This is useful if kernel is booting in an unreliable environment.
177 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
178 * skipped and devices will be in unknown state.
180 unsigned int reset_devices;
181 EXPORT_SYMBOL(reset_devices);
183 static int __init set_reset_devices(char *str)
189 __setup("reset_devices", set_reset_devices);
191 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
192 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
193 static const char *panic_later, *panic_param;
195 extern struct obs_kernel_param __setup_start[], __setup_end[];
197 static int __init obsolete_checksetup(char *line)
199 struct obs_kernel_param *p;
200 int had_early_param = 0;
204 int n = strlen(p->str);
205 if (!strncmp(line, p->str, n)) {
207 /* Already done in parse_early_param?
208 * (Needs exact match on param part).
209 * Keep iterating, as we can have early
210 * params and __setups of same names 8( */
211 if (line[n] == '\0' || line[n] == '=')
213 } else if (!p->setup_func) {
214 printk(KERN_WARNING "Parameter %s is obsolete,"
215 " ignored\n", p->str);
217 } else if (p->setup_func(line + n))
221 } while (p < __setup_end);
223 return had_early_param;
227 * This should be approx 2 Bo*oMips to start (note initial shift), and will
228 * still work even if initially too large, it will just take slightly longer
230 unsigned long loops_per_jiffy = (1<<12);
232 EXPORT_SYMBOL(loops_per_jiffy);
234 static int __init debug_kernel(char *str)
236 console_loglevel = 10;
240 static int __init quiet_kernel(char *str)
242 console_loglevel = 4;
246 early_param("debug", debug_kernel);
247 early_param("quiet", quiet_kernel);
249 static int __init loglevel(char *str)
251 get_option(&str, &console_loglevel);
255 early_param("loglevel", loglevel);
258 * 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. */
265 /* param=val or param="val"? */
266 if (val == param+strlen(param)+1)
268 else if (val == param+strlen(param)+2) {
270 memmove(val-1, val, strlen(val)+1);
276 /* Handle obsolete-style parameters */
277 if (obsolete_checksetup(param))
281 * Preemptive maintenance for "why didn't my misspelled command
284 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
285 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
293 /* Environment option */
295 for (i = 0; envp_init[i]; i++) {
296 if (i == MAX_INIT_ENVS) {
297 panic_later = "Too many boot env vars at `%s'";
300 if (!strncmp(param, envp_init[i], val - param))
303 envp_init[i] = param;
305 /* Command line option */
307 for (i = 0; argv_init[i]; i++) {
308 if (i == MAX_INIT_ARGS) {
309 panic_later = "Too many boot init vars at `%s'";
313 argv_init[i] = param;
318 #ifdef CONFIG_DEBUG_PAGEALLOC
319 int __read_mostly debug_pagealloc_enabled = 0;
322 static int __init init_setup(char *str)
326 execute_command = str;
328 * In case LILO is going to boot us with default command line,
329 * it prepends "auto" before the whole cmdline which makes
330 * the shell think it should execute a script with such name.
331 * So we ignore all arguments entered _before_ init=... [MJ]
333 for (i = 1; i < MAX_INIT_ARGS; i++)
337 __setup("init=", init_setup);
339 static int __init rdinit_setup(char *str)
343 ramdisk_execute_command = str;
344 /* See "auto" comment in init_setup */
345 for (i = 1; i < MAX_INIT_ARGS; i++)
349 __setup("rdinit=", rdinit_setup);
353 #ifdef CONFIG_X86_LOCAL_APIC
354 static void __init smp_init(void)
356 APIC_init_uniprocessor();
359 #define smp_init() do { } while (0)
362 static inline void setup_per_cpu_areas(void) { }
363 static inline void setup_nr_cpu_ids(void) { }
364 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
368 #if NR_CPUS > BITS_PER_LONG
369 cpumask_t cpu_mask_all __read_mostly = CPU_MASK_ALL;
370 EXPORT_SYMBOL(cpu_mask_all);
373 /* Setup number of possible processor ids */
374 int nr_cpu_ids __read_mostly = NR_CPUS;
375 EXPORT_SYMBOL(nr_cpu_ids);
377 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
378 static void __init setup_nr_cpu_ids(void)
380 int cpu, highest_cpu = 0;
382 for_each_possible_cpu(cpu)
385 nr_cpu_ids = highest_cpu + 1;
388 #ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA
389 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
391 EXPORT_SYMBOL(__per_cpu_offset);
393 static void __init setup_per_cpu_areas(void)
395 unsigned long size, i;
397 unsigned long nr_possible_cpus = num_possible_cpus();
399 /* Copy section for each CPU (we discard the original) */
400 size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
401 ptr = alloc_bootmem_pages(size * nr_possible_cpus);
403 for_each_possible_cpu(i) {
404 __per_cpu_offset[i] = ptr - __per_cpu_start;
405 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
409 #endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */
411 /* Called by boot processor to activate the rest. */
412 static void __init smp_init(void)
417 * Set up the current CPU as possible to migrate to.
418 * The other ones will be done by cpu_up/cpu_down()
420 cpu = smp_processor_id();
421 cpu_set(cpu, cpu_active_map);
423 /* FIXME: This should be done in userspace --RR */
424 for_each_present_cpu(cpu) {
425 if (num_online_cpus() >= setup_max_cpus)
427 if (!cpu_online(cpu))
431 /* Any cleanup work */
432 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
433 smp_cpus_done(setup_max_cpus);
439 * We need to store the untouched command line for future reference.
440 * We also need to store the touched command line since the parameter
441 * parsing is performed in place, and we should allow a component to
442 * store reference of name/value for future reference.
444 static void __init setup_command_line(char *command_line)
446 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
447 static_command_line = alloc_bootmem(strlen (command_line)+1);
448 strcpy (saved_command_line, boot_command_line);
449 strcpy (static_command_line, command_line);
453 * We need to finalize in a non-__init function or else race conditions
454 * between the root thread and the init thread may cause start_kernel to
455 * be reaped by free_initmem before the root thread has proceeded to
458 * gcc-3.4 accidentally inlines this function, so use noinline.
461 static void noinline __init_refok rest_init(void)
462 __releases(kernel_lock)
466 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
467 numa_default_policy();
468 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
469 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
473 * The boot idle thread must execute schedule()
474 * at least once to get things moving:
476 init_idle_bootup_task(current);
477 preempt_enable_no_resched();
481 /* Call into cpu_idle with preempt disabled */
485 /* Check for early params. */
486 static int __init do_early_param(char *param, char *val)
488 struct obs_kernel_param *p;
490 for (p = __setup_start; p < __setup_end; p++) {
491 if ((p->early && strcmp(param, p->str) == 0) ||
492 (strcmp(param, "console") == 0 &&
493 strcmp(p->str, "earlycon") == 0)
495 if (p->setup_func(val) != 0)
497 "Malformed early option '%s'\n", param);
500 /* We accept everything at this stage. */
504 /* Arch code calls this early on, or if not, just before other parsing. */
505 void __init parse_early_param(void)
507 static __initdata int done = 0;
508 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
513 /* All fall through to do_early_param. */
514 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
515 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
520 * Activate the first processor.
523 static void __init boot_cpu_init(void)
525 int cpu = smp_processor_id();
526 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
527 cpu_set(cpu, cpu_online_map);
528 cpu_set(cpu, cpu_present_map);
529 cpu_set(cpu, cpu_possible_map);
532 void __init __weak smp_setup_processor_id(void)
536 void __init __weak thread_info_cache_init(void)
540 asmlinkage void __init start_kernel(void)
543 extern struct kernel_param __start___param[], __stop___param[];
545 smp_setup_processor_id();
548 * Need to run as early as possible, to initialize the
553 debug_objects_early_init();
557 early_boot_irqs_off();
558 early_init_irq_lock_class();
561 * Interrupts are still disabled. Do necessary setups, then
569 printk(linux_banner);
570 setup_arch(&command_line);
571 mm_init_owner(&init_mm, &init_task);
572 setup_command_line(command_line);
574 setup_per_cpu_areas();
576 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
579 * Set up the scheduler prior starting any interrupts (such as the
580 * timer interrupt). Full topology setup happens at smp_init()
581 * time - but meanwhile we still have a functioning scheduler.
585 * Disable preemption - early bootup scheduling is extremely
586 * fragile until we cpu_idle() for the first time.
589 build_all_zonelists();
591 printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
593 parse_args("Booting kernel", static_command_line, __start___param,
594 __stop___param - __start___param,
595 &unknown_bootoption);
596 if (!irqs_disabled()) {
597 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
598 "enabled *very* early, fixing it\n");
613 if (!irqs_disabled())
614 printk("start_kernel(): bug: interrupts were enabled early\n");
615 early_boot_irqs_on();
619 * HACK ALERT! This is early. We're enabling the console before
620 * we've done PCI setups etc, and console_init() must be aware of
621 * this. But we do want output early, in case something goes wrong.
625 panic(panic_later, panic_param);
630 * Need to run this when irqs are enabled, because it wants
631 * to self-test [hard/soft]-irqs on/off lock inversion bugs
636 #ifdef CONFIG_BLK_DEV_INITRD
637 if (initrd_start && !initrd_below_start_ok &&
638 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
639 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
641 page_to_pfn(virt_to_page((void *)initrd_start)),
646 vfs_caches_init_early();
649 enable_debug_pagealloc();
652 debug_objects_mem_init();
654 setup_per_cpu_pageset();
660 pgtable_cache_init();
665 efi_enter_virtual_mode();
667 thread_info_cache_init();
668 fork_init(num_physpages);
674 vfs_caches_init(num_physpages);
677 /* rootfs populating might need page-writeback */
678 page_writeback_init();
679 #ifdef CONFIG_PROC_FS
684 taskstats_init_early();
689 acpi_early_init(); /* before LAPIC and SMP init */
693 /* Do the rest non-__init'ed, we're now alive */
697 static int initcall_debug;
699 static int __init initcall_debug_setup(char *str)
704 __setup("initcall_debug", initcall_debug_setup);
706 int do_one_initcall(initcall_t fn)
708 int count = preempt_count();
709 ktime_t t0, t1, delta;
713 if (initcall_debug) {
714 printk("calling %pF\n", fn);
720 if (initcall_debug) {
722 delta = ktime_sub(t1, t0);
724 printk("initcall %pF returned %d after %Ld msecs\n",
726 (unsigned long long) delta.tv64 >> 20);
731 if (result && result != -ENODEV && initcall_debug)
732 sprintf(msgbuf, "error code %d ", result);
734 if (preempt_count() != count) {
735 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
736 preempt_count() = count;
738 if (irqs_disabled()) {
739 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
743 printk("initcall %pF returned with %s\n", fn, msgbuf);
750 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
752 static void __init do_initcalls(void)
756 for (call = __early_initcall_end; call < __initcall_end; call++)
757 do_one_initcall(*call);
759 /* Make sure there is no pending stuff from the initcall sequence */
760 flush_scheduled_work();
764 * Ok, the machine is now initialized. None of the devices
765 * have been touched yet, but the CPU subsystem is up and
766 * running, and memory and process management works.
768 * Now we can finally start doing some real work..
770 static void __init do_basic_setup(void)
772 rcu_init_sched(); /* needed by module_init stage. */
773 /* drivers will send hotplug events */
775 usermodehelper_init();
781 static void __init do_pre_smp_initcalls(void)
785 for (call = __initcall_start; call < __early_initcall_end; call++)
786 do_one_initcall(*call);
789 static void run_init_process(char *init_filename)
791 argv_init[0] = init_filename;
792 kernel_execve(init_filename, argv_init, envp_init);
795 /* This is a non __init function. Force it to be noinline otherwise gcc
796 * makes it inline to init() and it becomes part of init.text section
798 static int noinline init_post(void)
803 system_state = SYSTEM_RUNNING;
804 numa_default_policy();
806 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
807 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
812 current->signal->flags |= SIGNAL_UNKILLABLE;
814 if (ramdisk_execute_command) {
815 run_init_process(ramdisk_execute_command);
816 printk(KERN_WARNING "Failed to execute %s\n",
817 ramdisk_execute_command);
821 * We try each of these until one succeeds.
823 * The Bourne shell can be used instead of init if we are
824 * trying to recover a really broken machine.
826 if (execute_command) {
827 run_init_process(execute_command);
828 printk(KERN_WARNING "Failed to execute %s. Attempting "
829 "defaults...\n", execute_command);
831 run_init_process("/sbin/init");
832 run_init_process("/etc/init");
833 run_init_process("/bin/init");
834 run_init_process("/bin/sh");
836 panic("No init found. Try passing init= option to kernel.");
839 static int __init kernel_init(void * unused)
843 * init can run on any cpu.
845 set_cpus_allowed_ptr(current, CPU_MASK_ALL_PTR);
847 * Tell the world that we're going to be the grim
848 * reaper of innocent orphaned children.
850 * We don't want people to have to make incorrect
851 * assumptions about where in the task array this
854 init_pid_ns.child_reaper = current;
856 cad_pid = task_pid(current);
858 smp_prepare_cpus(setup_max_cpus);
860 do_pre_smp_initcalls();
870 * check if there is an early userspace init. If yes, let it do all
874 if (!ramdisk_execute_command)
875 ramdisk_execute_command = "/init";
877 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
878 ramdisk_execute_command = NULL;
883 * Ok, we have completed the initial bootup, and
884 * we're essentially up and running. Get rid of the
885 * initmem segments and start the user-mode stuff..