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/ioport.h>
22 #include <linux/init.h>
23 #include <linux/initrd.h>
24 #include <linux/bootmem.h>
25 #include <linux/acpi.h>
26 #include <linux/tty.h>
27 #include <linux/percpu.h>
28 #include <linux/kmod.h>
29 #include <linux/vmalloc.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/profile.h>
35 #include <linux/rcupdate.h>
36 #include <linux/moduleparam.h>
37 #include <linux/kallsyms.h>
38 #include <linux/writeback.h>
39 #include <linux/cpu.h>
40 #include <linux/cpuset.h>
41 #include <linux/cgroup.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/buffer_head.h>
52 #include <linux/page_cgroup.h>
53 #include <linux/debug_locks.h>
54 #include <linux/debugobjects.h>
55 #include <linux/lockdep.h>
56 #include <linux/kmemleak.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/kgdb.h>
64 #include <linux/ftrace.h>
65 #include <linux/async.h>
66 #include <linux/kmemcheck.h>
67 #include <linux/sfi.h>
68 #include <linux/shmem_fs.h>
69 #include <linux/slab.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);
90 #ifndef CONFIG_DEBUG_RODATA
91 static inline void mark_rodata_ro(void) { }
95 extern void tc_init(void);
98 enum system_states system_state __read_mostly;
99 EXPORT_SYMBOL(system_state);
102 * Boot command-line arguments
104 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
105 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
107 extern void time_init(void);
108 /* Default late time init is NULL. archs can override this later. */
109 void (*__initdata late_time_init)(void);
110 extern void softirq_init(void);
112 /* Untouched command line saved by arch-specific code. */
113 char __initdata boot_command_line[COMMAND_LINE_SIZE];
114 /* Untouched saved command line (eg. for /proc) */
115 char *saved_command_line;
116 /* Command line for parameter parsing */
117 static char *static_command_line;
119 static char *execute_command;
120 static char *ramdisk_execute_command;
123 /* Setup configured maximum number of CPUs to activate */
124 unsigned int setup_max_cpus = NR_CPUS;
125 EXPORT_SYMBOL(setup_max_cpus);
129 * Setup routine for controlling SMP activation
131 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
132 * activation entirely (the MPS table probe still happens, though).
134 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
135 * greater than 0, limits the maximum number of CPUs activated in
139 void __weak arch_disable_smp_support(void) { }
141 static int __init nosmp(char *str)
144 arch_disable_smp_support();
149 early_param("nosmp", nosmp);
151 /* this is hard limit */
152 static int __init nrcpus(char *str)
156 get_option(&str, &nr_cpus);
157 if (nr_cpus > 0 && nr_cpus < nr_cpu_ids)
158 nr_cpu_ids = nr_cpus;
163 early_param("nr_cpus", nrcpus);
165 static int __init maxcpus(char *str)
167 get_option(&str, &setup_max_cpus);
168 if (setup_max_cpus == 0)
169 arch_disable_smp_support();
174 early_param("maxcpus", maxcpus);
176 static const unsigned int setup_max_cpus = NR_CPUS;
180 * If set, this is an indication to the drivers that reset the underlying
181 * device before going ahead with the initialization otherwise driver might
182 * rely on the BIOS and skip the reset operation.
184 * This is useful if kernel is booting in an unreliable environment.
185 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
186 * skipped and devices will be in unknown state.
188 unsigned int reset_devices;
189 EXPORT_SYMBOL(reset_devices);
191 static int __init set_reset_devices(char *str)
197 __setup("reset_devices", set_reset_devices);
199 static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
200 const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
201 static const char *panic_later, *panic_param;
203 extern const struct obs_kernel_param __setup_start[], __setup_end[];
205 static int __init obsolete_checksetup(char *line)
207 const struct obs_kernel_param *p;
208 int had_early_param = 0;
212 int n = strlen(p->str);
213 if (!strncmp(line, p->str, n)) {
215 /* Already done in parse_early_param?
216 * (Needs exact match on param part).
217 * Keep iterating, as we can have early
218 * params and __setups of same names 8( */
219 if (line[n] == '\0' || line[n] == '=')
221 } else if (!p->setup_func) {
222 printk(KERN_WARNING "Parameter %s is obsolete,"
223 " ignored\n", p->str);
225 } else if (p->setup_func(line + n))
229 } while (p < __setup_end);
231 return had_early_param;
235 * This should be approx 2 Bo*oMips to start (note initial shift), and will
236 * still work even if initially too large, it will just take slightly longer
238 unsigned long loops_per_jiffy = (1<<12);
240 EXPORT_SYMBOL(loops_per_jiffy);
242 static int __init debug_kernel(char *str)
244 console_loglevel = 10;
248 static int __init quiet_kernel(char *str)
250 console_loglevel = 4;
254 early_param("debug", debug_kernel);
255 early_param("quiet", quiet_kernel);
257 static int __init loglevel(char *str)
259 get_option(&str, &console_loglevel);
263 early_param("loglevel", loglevel);
266 * Unknown boot options get handed to init, unless they look like
267 * unused parameters (modprobe will find them in /proc/cmdline).
269 static int __init unknown_bootoption(char *param, char *val)
271 /* Change NUL term back to "=", to make "param" the whole string. */
273 /* param=val or param="val"? */
274 if (val == param+strlen(param)+1)
276 else if (val == param+strlen(param)+2) {
278 memmove(val-1, val, strlen(val)+1);
284 /* Handle obsolete-style parameters */
285 if (obsolete_checksetup(param))
288 /* Unused module parameter. */
289 if (strchr(param, '.') && (!val || strchr(param, '.') < val))
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_nr_cpu_ids(void) { }
366 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
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 /* Called by boot processor to activate the rest. */
381 static void __init smp_init(void)
385 /* FIXME: This should be done in userspace --RR */
386 for_each_present_cpu(cpu) {
387 if (num_online_cpus() >= setup_max_cpus)
389 if (!cpu_online(cpu))
393 /* Any cleanup work */
394 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
395 smp_cpus_done(setup_max_cpus);
401 * We need to store the untouched command line for future reference.
402 * We also need to store the touched command line since the parameter
403 * parsing is performed in place, and we should allow a component to
404 * store reference of name/value for future reference.
406 static void __init setup_command_line(char *command_line)
408 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
409 static_command_line = alloc_bootmem(strlen (command_line)+1);
410 strcpy (saved_command_line, boot_command_line);
411 strcpy (static_command_line, command_line);
415 * We need to finalize in a non-__init function or else race conditions
416 * between the root thread and the init thread may cause start_kernel to
417 * be reaped by free_initmem before the root thread has proceeded to
420 * gcc-3.4 accidentally inlines this function, so use noinline.
423 static __initdata DECLARE_COMPLETION(kthreadd_done);
425 static noinline void __init_refok rest_init(void)
429 rcu_scheduler_starting();
431 * We need to spawn init first so that it obtains pid 1, however
432 * the init task will end up wanting to create kthreads, which, if
433 * we schedule it before we create kthreadd, will OOPS.
435 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
436 numa_default_policy();
437 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
439 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
441 complete(&kthreadd_done);
444 * The boot idle thread must execute schedule()
445 * at least once to get things moving:
447 init_idle_bootup_task(current);
448 preempt_enable_no_resched();
452 /* Call into cpu_idle with preempt disabled */
456 /* Check for early params. */
457 static int __init do_early_param(char *param, char *val)
459 const struct obs_kernel_param *p;
461 for (p = __setup_start; p < __setup_end; p++) {
462 if ((p->early && strcmp(param, p->str) == 0) ||
463 (strcmp(param, "console") == 0 &&
464 strcmp(p->str, "earlycon") == 0)
466 if (p->setup_func(val) != 0)
468 "Malformed early option '%s'\n", param);
471 /* We accept everything at this stage. */
475 void __init parse_early_options(char *cmdline)
477 parse_args("early options", cmdline, NULL, 0, do_early_param);
480 /* Arch code calls this early on, or if not, just before other parsing. */
481 void __init parse_early_param(void)
483 static __initdata int done = 0;
484 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
489 /* All fall through to do_early_param. */
490 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
491 parse_early_options(tmp_cmdline);
496 * Activate the first processor.
499 static void __init boot_cpu_init(void)
501 int cpu = smp_processor_id();
502 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
503 set_cpu_online(cpu, true);
504 set_cpu_active(cpu, true);
505 set_cpu_present(cpu, true);
506 set_cpu_possible(cpu, true);
509 void __init __weak smp_setup_processor_id(void)
513 void __init __weak thread_info_cache_init(void)
518 * Set up kernel memory allocators
520 static void __init mm_init(void)
523 * page_cgroup requires countinous pages as memmap
524 * and it's bigger than MAX_ORDER unless SPARSEMEM.
526 page_cgroup_init_flatmem();
530 pgtable_cache_init();
534 asmlinkage void __init start_kernel(void)
537 extern const struct kernel_param __start___param[], __stop___param[];
539 smp_setup_processor_id();
542 * Need to run as early as possible, to initialize the
546 debug_objects_early_init();
549 * Set up the the initial canary ASAP:
551 boot_init_stack_canary();
556 early_boot_irqs_off();
559 * Interrupts are still disabled. Do necessary setups, then
565 printk(KERN_NOTICE "%s", linux_banner);
566 setup_arch(&command_line);
567 mm_init_owner(&init_mm, &init_task);
568 setup_command_line(command_line);
570 setup_per_cpu_areas();
571 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
573 build_all_zonelists(NULL);
576 printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
578 parse_args("Booting kernel", static_command_line, __start___param,
579 __stop___param - __start___param,
580 &unknown_bootoption);
582 * These use large bootmem allocations and must precede
586 vfs_caches_init_early();
591 * Set up the scheduler prior starting any interrupts (such as the
592 * timer interrupt). Full topology setup happens at smp_init()
593 * time - but meanwhile we still have a functioning scheduler.
597 * Disable preemption - early bootup scheduling is extremely
598 * fragile until we cpu_idle() for the first time.
601 if (!irqs_disabled()) {
602 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
603 "enabled *very* early, fixing it\n");
608 /* init some links before init_ISA_irqs() */
618 if (!irqs_disabled())
619 printk(KERN_CRIT "start_kernel(): bug: interrupts were "
621 early_boot_irqs_on();
624 /* Interrupts are enabled now so all GFP allocations are safe. */
625 gfp_allowed_mask = __GFP_BITS_MASK;
627 kmem_cache_init_late();
630 * HACK ALERT! This is early. We're enabling the console before
631 * we've done PCI setups etc, and console_init() must be aware of
632 * this. But we do want output early, in case something goes wrong.
636 panic(panic_later, panic_param);
641 * Need to run this when irqs are enabled, because it wants
642 * to self-test [hard/soft]-irqs on/off lock inversion bugs
647 #ifdef CONFIG_BLK_DEV_INITRD
648 if (initrd_start && !initrd_below_start_ok &&
649 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
650 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
652 page_to_pfn(virt_to_page((void *)initrd_start)),
658 enable_debug_pagealloc();
660 debug_objects_mem_init();
662 setup_per_cpu_pageset();
672 efi_enter_virtual_mode();
674 thread_info_cache_init();
676 fork_init(totalram_pages);
682 vfs_caches_init(totalram_pages);
684 /* rootfs populating might need page-writeback */
685 page_writeback_init();
686 #ifdef CONFIG_PROC_FS
691 taskstats_init_early();
696 acpi_early_init(); /* before LAPIC and SMP init */
701 /* Do the rest non-__init'ed, we're now alive */
705 /* Call all constructor functions linked into the kernel. */
706 static void __init do_ctors(void)
708 #ifdef CONFIG_CONSTRUCTORS
709 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
711 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
717 core_param(initcall_debug, initcall_debug, bool, 0644);
719 static char msgbuf[64];
721 static int __init_or_module do_one_initcall_debug(initcall_t fn)
723 ktime_t calltime, delta, rettime;
724 unsigned long long duration;
727 printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current));
728 calltime = ktime_get();
730 rettime = ktime_get();
731 delta = ktime_sub(rettime, calltime);
732 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
733 printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", fn,
739 int __init_or_module do_one_initcall(initcall_t fn)
741 int count = preempt_count();
745 ret = do_one_initcall_debug(fn);
751 if (ret && ret != -ENODEV && initcall_debug)
752 sprintf(msgbuf, "error code %d ", ret);
754 if (preempt_count() != count) {
755 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
756 preempt_count() = count;
758 if (irqs_disabled()) {
759 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
763 printk("initcall %pF returned with %s\n", fn, msgbuf);
770 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
772 static void __init do_initcalls(void)
776 for (fn = __early_initcall_end; fn < __initcall_end; fn++)
777 do_one_initcall(*fn);
779 /* Make sure there is no pending stuff from the initcall sequence */
780 flush_scheduled_work();
784 * Ok, the machine is now initialized. None of the devices
785 * have been touched yet, but the CPU subsystem is up and
786 * running, and memory and process management works.
788 * Now we can finally start doing some real work..
790 static void __init do_basic_setup(void)
793 usermodehelper_init();
801 static void __init do_pre_smp_initcalls(void)
805 for (fn = __initcall_start; fn < __early_initcall_end; fn++)
806 do_one_initcall(*fn);
809 static void run_init_process(const char *init_filename)
811 argv_init[0] = init_filename;
812 kernel_execve(init_filename, argv_init, envp_init);
815 /* This is a non __init function. Force it to be noinline otherwise gcc
816 * makes it inline to init() and it becomes part of init.text section
818 static noinline int init_post(void)
820 /* need to finish all async __init code before freeing the memory */
821 async_synchronize_full();
824 system_state = SYSTEM_RUNNING;
825 numa_default_policy();
828 current->signal->flags |= SIGNAL_UNKILLABLE;
830 if (ramdisk_execute_command) {
831 run_init_process(ramdisk_execute_command);
832 printk(KERN_WARNING "Failed to execute %s\n",
833 ramdisk_execute_command);
837 * We try each of these until one succeeds.
839 * The Bourne shell can be used instead of init if we are
840 * trying to recover a really broken machine.
842 if (execute_command) {
843 run_init_process(execute_command);
844 printk(KERN_WARNING "Failed to execute %s. Attempting "
845 "defaults...\n", execute_command);
847 run_init_process("/sbin/init");
848 run_init_process("/etc/init");
849 run_init_process("/bin/init");
850 run_init_process("/bin/sh");
852 panic("No init found. Try passing init= option to kernel. "
853 "See Linux Documentation/init.txt for guidance.");
856 static int __init kernel_init(void * unused)
859 * Wait until kthreadd is all set-up.
861 wait_for_completion(&kthreadd_done);
863 * init can allocate pages on any node
865 set_mems_allowed(node_states[N_HIGH_MEMORY]);
867 * init can run on any cpu.
869 set_cpus_allowed_ptr(current, cpu_all_mask);
871 * Tell the world that we're going to be the grim
872 * reaper of innocent orphaned children.
874 * We don't want people to have to make incorrect
875 * assumptions about where in the task array this
878 init_pid_ns.child_reaper = current;
880 cad_pid = task_pid(current);
882 smp_prepare_cpus(setup_max_cpus);
884 do_pre_smp_initcalls();
885 lockup_detector_init();
892 /* Open the /dev/console on the rootfs, this should never fail */
893 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
894 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
899 * check if there is an early userspace init. If yes, let it do all
903 if (!ramdisk_execute_command)
904 ramdisk_execute_command = "/init";
906 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
907 ramdisk_execute_command = NULL;
912 * Ok, we have completed the initial bootup, and
913 * we're essentially up and running. Get rid of the
914 * initmem segments and start the user-mode stuff..