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>
70 #include <linux/perf_event.h>
71 #include <linux/random.h>
72 #include <linux/kaiser.h>
76 #include <asm/setup.h>
77 #include <asm/sections.h>
78 #include <asm/cacheflush.h>
80 #ifdef CONFIG_X86_LOCAL_APIC
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);
93 #ifndef CONFIG_DEBUG_RODATA
94 static inline void mark_rodata_ro(void) { }
98 extern void tc_init(void);
102 * Debug helper: via this flag we know that we are in 'early bootup code'
103 * where only the boot processor is running with IRQ disabled. This means
104 * two things - IRQ must not be enabled before the flag is cleared and some
105 * operations which are not allowed with IRQ disabled are allowed while the
108 bool early_boot_irqs_disabled __read_mostly;
110 enum system_states system_state __read_mostly;
111 EXPORT_SYMBOL(system_state);
114 * Boot command-line arguments
116 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
117 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
119 extern void time_init(void);
120 /* Default late time init is NULL. archs can override this later. */
121 void (*__initdata late_time_init)(void);
122 extern void softirq_init(void);
124 /* Untouched command line saved by arch-specific code. */
125 char __initdata boot_command_line[COMMAND_LINE_SIZE];
126 /* Untouched saved command line (eg. for /proc) */
127 char *saved_command_line;
128 /* Command line for parameter parsing */
129 static char *static_command_line;
131 static char *execute_command;
132 static char *ramdisk_execute_command;
135 * If set, this is an indication to the drivers that reset the underlying
136 * device before going ahead with the initialization otherwise driver might
137 * rely on the BIOS and skip the reset operation.
139 * This is useful if kernel is booting in an unreliable environment.
140 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
141 * skipped and devices will be in unknown state.
143 unsigned int reset_devices;
144 EXPORT_SYMBOL(reset_devices);
146 static int __init set_reset_devices(char *str)
152 __setup("reset_devices", set_reset_devices);
154 static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
155 const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
156 static const char *panic_later, *panic_param;
158 extern const struct obs_kernel_param __setup_start[], __setup_end[];
160 static int __init obsolete_checksetup(char *line)
162 const struct obs_kernel_param *p;
163 int had_early_param = 0;
167 int n = strlen(p->str);
168 if (parameqn(line, p->str, n)) {
170 /* Already done in parse_early_param?
171 * (Needs exact match on param part).
172 * Keep iterating, as we can have early
173 * params and __setups of same names 8( */
174 if (line[n] == '\0' || line[n] == '=')
176 } else if (!p->setup_func) {
177 printk(KERN_WARNING "Parameter %s is obsolete,"
178 " ignored\n", p->str);
180 } else if (p->setup_func(line + n))
184 } while (p < __setup_end);
186 return had_early_param;
190 * This should be approx 2 Bo*oMips to start (note initial shift), and will
191 * still work even if initially too large, it will just take slightly longer
193 unsigned long loops_per_jiffy = (1<<12);
195 EXPORT_SYMBOL(loops_per_jiffy);
197 static int __init debug_kernel(char *str)
199 console_loglevel = 10;
203 static int __init quiet_kernel(char *str)
205 console_loglevel = 4;
209 early_param("debug", debug_kernel);
210 early_param("quiet", quiet_kernel);
212 static int __init loglevel(char *str)
217 * Only update loglevel value when a correct setting was passed,
218 * to prevent blind crashes (when loglevel being set to 0) that
219 * are quite hard to debug
221 if (get_option(&str, &newlevel)) {
222 console_loglevel = newlevel;
229 early_param("loglevel", loglevel);
232 * Unknown boot options get handed to init, unless they look like
233 * unused parameters (modprobe will find them in /proc/cmdline).
235 static int __init unknown_bootoption(char *param, char *val)
237 /* Change NUL term back to "=", to make "param" the whole string. */
239 /* param=val or param="val"? */
240 if (val == param+strlen(param)+1)
242 else if (val == param+strlen(param)+2) {
244 memmove(val-1, val, strlen(val)+1);
250 /* Handle obsolete-style parameters */
251 if (obsolete_checksetup(param))
254 /* Unused module parameter. */
255 if (strchr(param, '.') && (!val || strchr(param, '.') < val))
262 /* Environment option */
264 for (i = 0; envp_init[i]; i++) {
265 if (i == MAX_INIT_ENVS) {
266 panic_later = "Too many boot env vars at `%s'";
269 if (!strncmp(param, envp_init[i], val - param))
272 envp_init[i] = param;
274 /* Command line option */
276 for (i = 0; argv_init[i]; i++) {
277 if (i == MAX_INIT_ARGS) {
278 panic_later = "Too many boot init vars at `%s'";
282 argv_init[i] = param;
287 #ifdef CONFIG_DEBUG_PAGEALLOC
288 int __read_mostly debug_pagealloc_enabled = 0;
291 static int __init init_setup(char *str)
295 execute_command = str;
297 * In case LILO is going to boot us with default command line,
298 * it prepends "auto" before the whole cmdline which makes
299 * the shell think it should execute a script with such name.
300 * So we ignore all arguments entered _before_ init=... [MJ]
302 for (i = 1; i < MAX_INIT_ARGS; i++)
306 __setup("init=", init_setup);
308 static int __init rdinit_setup(char *str)
312 ramdisk_execute_command = str;
313 /* See "auto" comment in init_setup */
314 for (i = 1; i < MAX_INIT_ARGS; i++)
318 __setup("rdinit=", rdinit_setup);
321 static const unsigned int setup_max_cpus = NR_CPUS;
322 #ifdef CONFIG_X86_LOCAL_APIC
323 static void __init smp_init(void)
325 APIC_init_uniprocessor();
328 #define smp_init() do { } while (0)
331 static inline void setup_nr_cpu_ids(void) { }
332 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
336 * We need to store the untouched command line for future reference.
337 * We also need to store the touched command line since the parameter
338 * parsing is performed in place, and we should allow a component to
339 * store reference of name/value for future reference.
341 static void __init setup_command_line(char *command_line)
343 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
344 static_command_line = alloc_bootmem(strlen (command_line)+1);
345 strcpy (saved_command_line, boot_command_line);
346 strcpy (static_command_line, command_line);
350 * We need to finalize in a non-__init function or else race conditions
351 * between the root thread and the init thread may cause start_kernel to
352 * be reaped by free_initmem before the root thread has proceeded to
355 * gcc-3.4 accidentally inlines this function, so use noinline.
358 static __initdata DECLARE_COMPLETION(kthreadd_done);
360 static noinline void __init_refok rest_init(void)
364 rcu_scheduler_starting();
366 * We need to spawn init first so that it obtains pid 1, however
367 * the init task will end up wanting to create kthreads, which, if
368 * we schedule it before we create kthreadd, will OOPS.
370 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
371 numa_default_policy();
372 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
374 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
376 complete(&kthreadd_done);
379 * The boot idle thread must execute schedule()
380 * at least once to get things moving:
382 init_idle_bootup_task(current);
383 preempt_enable_no_resched();
386 /* Call into cpu_idle with preempt disabled */
391 /* Check for early params. */
392 static int __init do_early_param(char *param, char *val)
394 const struct obs_kernel_param *p;
396 for (p = __setup_start; p < __setup_end; p++) {
397 if ((p->early && parameq(param, p->str)) ||
398 (strcmp(param, "console") == 0 &&
399 strcmp(p->str, "earlycon") == 0)
401 if (p->setup_func(val) != 0)
403 "Malformed early option '%s'\n", param);
406 /* We accept everything at this stage. */
410 void __init parse_early_options(char *cmdline)
412 parse_args("early options", cmdline, NULL, 0, do_early_param);
415 /* Arch code calls this early on, or if not, just before other parsing. */
416 void __init parse_early_param(void)
418 static __initdata int done = 0;
419 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
424 /* All fall through to do_early_param. */
425 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
426 parse_early_options(tmp_cmdline);
431 * Activate the first processor.
434 static void __init boot_cpu_init(void)
436 int cpu = smp_processor_id();
437 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
438 set_cpu_online(cpu, true);
439 set_cpu_active(cpu, true);
440 set_cpu_present(cpu, true);
441 set_cpu_possible(cpu, true);
444 void __init __weak smp_setup_processor_id(void)
448 void __init __weak thread_info_cache_init(void)
453 * Set up kernel memory allocators
455 static void __init mm_init(void)
458 * page_cgroup requires countinous pages as memmap
459 * and it's bigger than MAX_ORDER unless SPARSEMEM.
461 page_cgroup_init_flatmem();
465 pgtable_cache_init();
470 asmlinkage void __init start_kernel(void)
473 extern const struct kernel_param __start___param[], __stop___param[];
475 smp_setup_processor_id();
478 * Need to run as early as possible, to initialize the
482 debug_objects_early_init();
485 * Set up the the initial canary ASAP:
487 boot_init_stack_canary();
492 early_boot_irqs_disabled = true;
495 * Interrupts are still disabled. Do necessary setups, then
501 printk(KERN_NOTICE "%s", linux_banner);
502 setup_arch(&command_line);
503 mm_init_owner(&init_mm, &init_task);
504 mm_init_cpumask(&init_mm);
505 setup_command_line(command_line);
507 setup_per_cpu_areas();
508 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
510 build_all_zonelists(NULL);
513 printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
515 parse_args("Booting kernel", static_command_line, __start___param,
516 __stop___param - __start___param,
517 &unknown_bootoption);
522 * These use large bootmem allocations and must precede
527 vfs_caches_init_early();
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.
539 * Disable preemption - early bootup scheduling is extremely
540 * fragile until we cpu_idle() for the first time.
543 if (!irqs_disabled()) {
544 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
545 "enabled *very* early, fixing it\n");
552 /* init some links before init_ISA_irqs() */
562 call_function_init();
563 if (!irqs_disabled())
564 printk(KERN_CRIT "start_kernel(): bug: interrupts were "
566 early_boot_irqs_disabled = false;
569 kmem_cache_init_late();
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.
578 panic(panic_later, panic_param);
583 * Need to run this when irqs are enabled, because it wants
584 * to self-test [hard/soft]-irqs on/off lock inversion bugs
589 #ifdef CONFIG_BLK_DEV_INITRD
590 if (initrd_start && !initrd_below_start_ok &&
591 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
592 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
594 page_to_pfn(virt_to_page((void *)initrd_start)),
600 enable_debug_pagealloc();
601 debug_objects_mem_init();
603 setup_per_cpu_pageset();
612 if (efi_enabled(EFI_RUNTIME_SERVICES))
613 efi_enter_virtual_mode();
615 #ifdef CONFIG_X86_ESPFIX64
616 /* Should be run before the first non-init thread is created */
619 thread_info_cache_init();
621 fork_init(totalram_pages);
627 vfs_caches_init(totalram_pages);
629 /* rootfs populating might need page-writeback */
630 page_writeback_init();
631 #ifdef CONFIG_PROC_FS
636 taskstats_init_early();
641 acpi_early_init(); /* before LAPIC and SMP init */
646 /* Do the rest non-__init'ed, we're now alive */
650 /* Call all constructor functions linked into the kernel. */
651 static void __init do_ctors(void)
653 #ifdef CONFIG_CONSTRUCTORS
654 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
656 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
662 core_param(initcall_debug, initcall_debug, bool, 0644);
664 static char msgbuf[64];
666 static int __init_or_module do_one_initcall_debug(initcall_t fn)
668 ktime_t calltime, delta, rettime;
669 unsigned long long duration;
672 printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current));
673 calltime = ktime_get();
675 rettime = ktime_get();
676 delta = ktime_sub(rettime, calltime);
677 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
678 printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", fn,
684 int __init_or_module do_one_initcall(initcall_t fn)
686 int count = preempt_count();
690 ret = do_one_initcall_debug(fn);
696 if (ret && ret != -ENODEV && initcall_debug)
697 sprintf(msgbuf, "error code %d ", ret);
699 if (preempt_count() != count) {
700 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
701 preempt_count() = count;
703 if (irqs_disabled()) {
704 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
708 printk("initcall %pF returned with %s\n", fn, msgbuf);
715 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
717 static void __init do_initcalls(void)
721 for (fn = __early_initcall_end; fn < __initcall_end; fn++)
722 do_one_initcall(*fn);
726 * Ok, the machine is now initialized. None of the devices
727 * have been touched yet, but the CPU subsystem is up and
728 * running, and memory and process management works.
730 * Now we can finally start doing some real work..
732 static void __init do_basic_setup(void)
735 usermodehelper_init();
740 usermodehelper_enable();
742 random_int_secret_init();
745 static void __init do_pre_smp_initcalls(void)
749 for (fn = __initcall_start; fn < __early_initcall_end; fn++)
750 do_one_initcall(*fn);
753 static void run_init_process(const char *init_filename)
755 argv_init[0] = init_filename;
756 kernel_execve(init_filename, argv_init, envp_init);
759 /* This is a non __init function. Force it to be noinline otherwise gcc
760 * makes it inline to init() and it becomes part of init.text section
762 static noinline int init_post(void)
764 /* need to finish all async __init code before freeing the memory */
765 async_synchronize_full();
768 system_state = SYSTEM_RUNNING;
769 numa_default_policy();
772 current->signal->flags |= SIGNAL_UNKILLABLE;
774 if (ramdisk_execute_command) {
775 run_init_process(ramdisk_execute_command);
776 printk(KERN_WARNING "Failed to execute %s\n",
777 ramdisk_execute_command);
781 * We try each of these until one succeeds.
783 * The Bourne shell can be used instead of init if we are
784 * trying to recover a really broken machine.
786 if (execute_command) {
787 run_init_process(execute_command);
788 printk(KERN_WARNING "Failed to execute %s. Attempting "
789 "defaults...\n", execute_command);
791 run_init_process("/sbin/init");
792 run_init_process("/etc/init");
793 run_init_process("/bin/init");
794 run_init_process("/bin/sh");
796 panic("No init found. Try passing init= option to kernel. "
797 "See Linux Documentation/init.txt for guidance.");
800 static int __init kernel_init(void * unused)
803 * Wait until kthreadd is all set-up.
805 wait_for_completion(&kthreadd_done);
807 /* Now the scheduler is fully set up and can do blocking allocations */
808 gfp_allowed_mask = __GFP_BITS_MASK;
811 * init can allocate pages on any node
813 set_mems_allowed(node_states[N_HIGH_MEMORY]);
815 * init can run on any cpu.
817 set_cpus_allowed_ptr(current, cpu_all_mask);
819 cad_pid = task_pid(current);
821 smp_prepare_cpus(setup_max_cpus);
823 do_pre_smp_initcalls();
824 lockup_detector_init();
831 /* Open the /dev/console on the rootfs, this should never fail */
832 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
833 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
838 * check if there is an early userspace init. If yes, let it do all
842 if (!ramdisk_execute_command)
843 ramdisk_execute_command = "/init";
845 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
846 ramdisk_execute_command = NULL;
851 * Ok, we have completed the initial bootup, and
852 * we're essentially up and running. Get rid of the
853 * initmem segments and start the user-mode stuff..