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/smp_lock.h>
24 #include <linux/initrd.h>
25 #include <linux/bootmem.h>
26 #include <linux/acpi.h>
27 #include <linux/tty.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/buffer_head.h>
54 #include <linux/page_cgroup.h>
55 #include <linux/debug_locks.h>
56 #include <linux/debugobjects.h>
57 #include <linux/lockdep.h>
58 #include <linux/kmemleak.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/kgdb.h>
66 #include <linux/ftrace.h>
67 #include <linux/async.h>
68 #include <linux/kmemcheck.h>
69 #include <linux/kmemtrace.h>
70 #include <linux/sfi.h>
71 #include <linux/shmem_fs.h>
72 #include <linux/slab.h>
73 #include <trace/boot.h>
77 #include <asm/setup.h>
78 #include <asm/sections.h>
79 #include <asm/cacheflush.h>
81 #ifdef CONFIG_X86_LOCAL_APIC
85 static int kernel_init(void *);
87 extern void init_IRQ(void);
88 extern void fork_init(unsigned long);
89 extern void mca_init(void);
90 extern void sbus_init(void);
91 extern void prio_tree_init(void);
92 extern void radix_tree_init(void);
93 extern void free_initmem(void);
94 #ifndef CONFIG_DEBUG_RODATA
95 static inline void mark_rodata_ro(void) { }
99 extern void tc_init(void);
102 enum system_states system_state __read_mostly
;
103 EXPORT_SYMBOL(system_state
);
106 * Boot command-line arguments
108 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
109 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
111 extern void time_init(void);
112 /* Default late time init is NULL. archs can override this later. */
113 void (*__initdata late_time_init
)(void);
114 extern void softirq_init(void);
116 /* Untouched command line saved by arch-specific code. */
117 char __initdata boot_command_line
[COMMAND_LINE_SIZE
];
118 /* Untouched saved command line (eg. for /proc) */
119 char *saved_command_line
;
120 /* Command line for parameter parsing */
121 static char *static_command_line
;
123 static char *execute_command
;
124 static char *ramdisk_execute_command
;
127 /* Setup configured maximum number of CPUs to activate */
128 unsigned int setup_max_cpus
= NR_CPUS
;
129 EXPORT_SYMBOL(setup_max_cpus
);
133 * Setup routine for controlling SMP activation
135 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
136 * activation entirely (the MPS table probe still happens, though).
138 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
139 * greater than 0, limits the maximum number of CPUs activated in
143 void __weak
arch_disable_smp_support(void) { }
145 static int __init
nosmp(char *str
)
148 arch_disable_smp_support();
153 early_param("nosmp", nosmp
);
155 /* this is hard limit */
156 static int __init
nrcpus(char *str
)
160 get_option(&str
, &nr_cpus
);
161 if (nr_cpus
> 0 && nr_cpus
< nr_cpu_ids
)
162 nr_cpu_ids
= nr_cpus
;
167 early_param("nr_cpus", nrcpus
);
169 static int __init
maxcpus(char *str
)
171 get_option(&str
, &setup_max_cpus
);
172 if (setup_max_cpus
== 0)
173 arch_disable_smp_support();
178 early_param("maxcpus", maxcpus
);
180 static const unsigned int setup_max_cpus
= NR_CPUS
;
184 * If set, this is an indication to the drivers that reset the underlying
185 * device before going ahead with the initialization otherwise driver might
186 * rely on the BIOS and skip the reset operation.
188 * This is useful if kernel is booting in an unreliable environment.
189 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
190 * skipped and devices will be in unknown state.
192 unsigned int reset_devices
;
193 EXPORT_SYMBOL(reset_devices
);
195 static int __init
set_reset_devices(char *str
)
201 __setup("reset_devices", set_reset_devices
);
203 static char * argv_init
[MAX_INIT_ARGS
+2] = { "init", NULL
, };
204 char * envp_init
[MAX_INIT_ENVS
+2] = { "HOME=/", "TERM=linux", NULL
, };
205 static const char *panic_later
, *panic_param
;
207 extern struct obs_kernel_param __setup_start
[], __setup_end
[];
209 static int __init
obsolete_checksetup(char *line
)
211 struct obs_kernel_param
*p
;
212 int had_early_param
= 0;
216 int n
= strlen(p
->str
);
217 if (!strncmp(line
, p
->str
, n
)) {
219 /* Already done in parse_early_param?
220 * (Needs exact match on param part).
221 * Keep iterating, as we can have early
222 * params and __setups of same names 8( */
223 if (line
[n
] == '\0' || line
[n
] == '=')
225 } else if (!p
->setup_func
) {
226 printk(KERN_WARNING
"Parameter %s is obsolete,"
227 " ignored\n", p
->str
);
229 } else if (p
->setup_func(line
+ n
))
233 } while (p
< __setup_end
);
235 return had_early_param
;
239 * This should be approx 2 Bo*oMips to start (note initial shift), and will
240 * still work even if initially too large, it will just take slightly longer
242 unsigned long loops_per_jiffy
= (1<<12);
244 EXPORT_SYMBOL(loops_per_jiffy
);
246 static int __init
debug_kernel(char *str
)
248 console_loglevel
= 10;
252 static int __init
quiet_kernel(char *str
)
254 console_loglevel
= 4;
258 early_param("debug", debug_kernel
);
259 early_param("quiet", quiet_kernel
);
261 static int __init
loglevel(char *str
)
263 get_option(&str
, &console_loglevel
);
267 early_param("loglevel", loglevel
);
270 * Unknown boot options get handed to init, unless they look like
271 * unused parameters (modprobe will find them in /proc/cmdline).
273 static int __init
unknown_bootoption(char *param
, char *val
)
275 /* Change NUL term back to "=", to make "param" the whole string. */
277 /* param=val or param="val"? */
278 if (val
== param
+strlen(param
)+1)
280 else if (val
== param
+strlen(param
)+2) {
282 memmove(val
-1, val
, strlen(val
)+1);
288 /* Handle obsolete-style parameters */
289 if (obsolete_checksetup(param
))
292 /* Unused module parameter. */
293 if (strchr(param
, '.') && (!val
|| strchr(param
, '.') < val
))
300 /* Environment option */
302 for (i
= 0; envp_init
[i
]; i
++) {
303 if (i
== MAX_INIT_ENVS
) {
304 panic_later
= "Too many boot env vars at `%s'";
307 if (!strncmp(param
, envp_init
[i
], val
- param
))
310 envp_init
[i
] = param
;
312 /* Command line option */
314 for (i
= 0; argv_init
[i
]; i
++) {
315 if (i
== MAX_INIT_ARGS
) {
316 panic_later
= "Too many boot init vars at `%s'";
320 argv_init
[i
] = param
;
325 #ifdef CONFIG_DEBUG_PAGEALLOC
326 int __read_mostly debug_pagealloc_enabled
= 0;
329 static int __init
init_setup(char *str
)
333 execute_command
= str
;
335 * In case LILO is going to boot us with default command line,
336 * it prepends "auto" before the whole cmdline which makes
337 * the shell think it should execute a script with such name.
338 * So we ignore all arguments entered _before_ init=... [MJ]
340 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
344 __setup("init=", init_setup
);
346 static int __init
rdinit_setup(char *str
)
350 ramdisk_execute_command
= str
;
351 /* See "auto" comment in init_setup */
352 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
356 __setup("rdinit=", rdinit_setup
);
360 #ifdef CONFIG_X86_LOCAL_APIC
361 static void __init
smp_init(void)
363 APIC_init_uniprocessor();
366 #define smp_init() do { } while (0)
369 static inline void setup_nr_cpu_ids(void) { }
370 static inline void smp_prepare_cpus(unsigned int maxcpus
) { }
374 /* Setup number of possible processor ids */
375 int nr_cpu_ids __read_mostly
= NR_CPUS
;
376 EXPORT_SYMBOL(nr_cpu_ids
);
378 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
379 static void __init
setup_nr_cpu_ids(void)
381 nr_cpu_ids
= find_last_bit(cpumask_bits(cpu_possible_mask
),NR_CPUS
) + 1;
384 /* Called by boot processor to activate the rest. */
385 static void __init
smp_init(void)
389 /* FIXME: This should be done in userspace --RR */
390 for_each_present_cpu(cpu
) {
391 if (num_online_cpus() >= setup_max_cpus
)
393 if (!cpu_online(cpu
))
397 /* Any cleanup work */
398 printk(KERN_INFO
"Brought up %ld CPUs\n", (long)num_online_cpus());
399 smp_cpus_done(setup_max_cpus
);
405 * We need to store the untouched command line for future reference.
406 * We also need to store the touched command line since the parameter
407 * parsing is performed in place, and we should allow a component to
408 * store reference of name/value for future reference.
410 static void __init
setup_command_line(char *command_line
)
412 saved_command_line
= alloc_bootmem(strlen (boot_command_line
)+1);
413 static_command_line
= alloc_bootmem(strlen (command_line
)+1);
414 strcpy (saved_command_line
, boot_command_line
);
415 strcpy (static_command_line
, command_line
);
419 * We need to finalize in a non-__init function or else race conditions
420 * between the root thread and the init thread may cause start_kernel to
421 * be reaped by free_initmem before the root thread has proceeded to
424 * gcc-3.4 accidentally inlines this function, so use noinline.
427 static __initdata
DECLARE_COMPLETION(kthreadd_done
);
429 static noinline
void __init_refok
rest_init(void)
430 __releases(kernel_lock
)
434 rcu_scheduler_starting();
436 * We need to spawn init first so that it obtains pid 1, however
437 * the init task will end up wanting to create kthreads, which, if
438 * we schedule it before we create kthreadd, will OOPS.
440 kernel_thread(kernel_init
, NULL
, CLONE_FS
| CLONE_SIGHAND
);
441 numa_default_policy();
442 pid
= kernel_thread(kthreadd
, NULL
, CLONE_FS
| CLONE_FILES
);
444 kthreadd_task
= find_task_by_pid_ns(pid
, &init_pid_ns
);
446 complete(&kthreadd_done
);
450 * The boot idle thread must execute schedule()
451 * at least once to get things moving:
453 init_idle_bootup_task(current
);
454 preempt_enable_no_resched();
458 /* Call into cpu_idle with preempt disabled */
462 /* Check for early params. */
463 static int __init
do_early_param(char *param
, char *val
)
465 struct obs_kernel_param
*p
;
467 for (p
= __setup_start
; p
< __setup_end
; p
++) {
468 if ((p
->early
&& strcmp(param
, p
->str
) == 0) ||
469 (strcmp(param
, "console") == 0 &&
470 strcmp(p
->str
, "earlycon") == 0)
472 if (p
->setup_func(val
) != 0)
474 "Malformed early option '%s'\n", param
);
477 /* We accept everything at this stage. */
481 void __init
parse_early_options(char *cmdline
)
483 parse_args("early options", cmdline
, NULL
, 0, do_early_param
);
486 /* Arch code calls this early on, or if not, just before other parsing. */
487 void __init
parse_early_param(void)
489 static __initdata
int done
= 0;
490 static __initdata
char tmp_cmdline
[COMMAND_LINE_SIZE
];
495 /* All fall through to do_early_param. */
496 strlcpy(tmp_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
497 parse_early_options(tmp_cmdline
);
502 * Activate the first processor.
505 static void __init
boot_cpu_init(void)
507 int cpu
= smp_processor_id();
508 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
509 set_cpu_online(cpu
, true);
510 set_cpu_active(cpu
, true);
511 set_cpu_present(cpu
, true);
512 set_cpu_possible(cpu
, true);
515 void __init __weak
smp_setup_processor_id(void)
519 void __init __weak
thread_info_cache_init(void)
524 * Set up kernel memory allocators
526 static void __init
mm_init(void)
529 * page_cgroup requires countinous pages as memmap
530 * and it's bigger than MAX_ORDER unless SPARSEMEM.
532 page_cgroup_init_flatmem();
536 pgtable_cache_init();
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
552 debug_objects_early_init();
555 * Set up the the initial canary ASAP:
557 boot_init_stack_canary();
562 early_boot_irqs_off();
563 early_init_irq_lock_class();
566 * Interrupts are still disabled. Do necessary setups, then
573 printk(KERN_NOTICE
"%s", linux_banner
);
574 setup_arch(&command_line
);
575 mm_init_owner(&init_mm
, &init_task
);
576 setup_command_line(command_line
);
578 setup_per_cpu_areas();
579 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
581 build_all_zonelists(NULL
);
584 printk(KERN_NOTICE
"Kernel command line: %s\n", boot_command_line
);
586 parse_args("Booting kernel", static_command_line
, __start___param
,
587 __stop___param
- __start___param
,
588 &unknown_bootoption
);
590 * These use large bootmem allocations and must precede
594 vfs_caches_init_early();
599 * Set up the scheduler prior starting any interrupts (such as the
600 * timer interrupt). Full topology setup happens at smp_init()
601 * time - but meanwhile we still have a functioning scheduler.
605 * Disable preemption - early bootup scheduling is extremely
606 * fragile until we cpu_idle() for the first time.
609 if (!irqs_disabled()) {
610 printk(KERN_WARNING
"start_kernel(): bug: interrupts were "
611 "enabled *very* early, fixing it\n");
616 /* init some links before init_ISA_irqs() */
626 if (!irqs_disabled())
627 printk(KERN_CRIT
"start_kernel(): bug: interrupts were "
629 early_boot_irqs_on();
632 /* Interrupts are enabled now so all GFP allocations are safe. */
633 gfp_allowed_mask
= __GFP_BITS_MASK
;
635 kmem_cache_init_late();
638 * HACK ALERT! This is early. We're enabling the console before
639 * we've done PCI setups etc, and console_init() must be aware of
640 * this. But we do want output early, in case something goes wrong.
644 panic(panic_later
, panic_param
);
649 * Need to run this when irqs are enabled, because it wants
650 * to self-test [hard/soft]-irqs on/off lock inversion bugs
655 #ifdef CONFIG_BLK_DEV_INITRD
656 if (initrd_start
&& !initrd_below_start_ok
&&
657 page_to_pfn(virt_to_page((void *)initrd_start
)) < min_low_pfn
) {
658 printk(KERN_CRIT
"initrd overwritten (0x%08lx < 0x%08lx) - "
660 page_to_pfn(virt_to_page((void *)initrd_start
)),
666 enable_debug_pagealloc();
669 debug_objects_mem_init();
671 setup_per_cpu_pageset();
681 efi_enter_virtual_mode();
683 thread_info_cache_init();
685 fork_init(totalram_pages
);
691 vfs_caches_init(totalram_pages
);
693 /* rootfs populating might need page-writeback */
694 page_writeback_init();
695 #ifdef CONFIG_PROC_FS
700 taskstats_init_early();
705 acpi_early_init(); /* before LAPIC and SMP init */
710 /* Do the rest non-__init'ed, we're now alive */
714 /* Call all constructor functions linked into the kernel. */
715 static void __init
do_ctors(void)
717 #ifdef CONFIG_CONSTRUCTORS
718 ctor_fn_t
*fn
= (ctor_fn_t
*) __ctors_start
;
720 for (; fn
< (ctor_fn_t
*) __ctors_end
; fn
++)
726 core_param(initcall_debug
, initcall_debug
, bool, 0644);
728 static char msgbuf
[64];
729 static struct boot_trace_call call
;
730 static struct boot_trace_ret ret
;
732 int do_one_initcall(initcall_t fn
)
734 int count
= preempt_count();
735 ktime_t calltime
, delta
, rettime
;
737 if (initcall_debug
) {
738 call
.caller
= task_pid_nr(current
);
739 printk("calling %pF @ %i\n", fn
, call
.caller
);
740 calltime
= ktime_get();
741 trace_boot_call(&call
, fn
);
747 if (initcall_debug
) {
748 disable_boot_trace();
749 rettime
= ktime_get();
750 delta
= ktime_sub(rettime
, calltime
);
751 ret
.duration
= (unsigned long long) ktime_to_ns(delta
) >> 10;
752 trace_boot_ret(&ret
, fn
);
753 printk("initcall %pF returned %d after %Ld usecs\n", fn
,
754 ret
.result
, ret
.duration
);
759 if (ret
.result
&& ret
.result
!= -ENODEV
&& initcall_debug
)
760 sprintf(msgbuf
, "error code %d ", ret
.result
);
762 if (preempt_count() != count
) {
763 strlcat(msgbuf
, "preemption imbalance ", sizeof(msgbuf
));
764 preempt_count() = count
;
766 if (irqs_disabled()) {
767 strlcat(msgbuf
, "disabled interrupts ", sizeof(msgbuf
));
771 printk("initcall %pF returned with %s\n", fn
, msgbuf
);
778 extern initcall_t __initcall_start
[], __initcall_end
[], __early_initcall_end
[];
780 static void __init
do_initcalls(void)
784 for (fn
= __early_initcall_end
; fn
< __initcall_end
; fn
++)
785 do_one_initcall(*fn
);
787 /* Make sure there is no pending stuff from the initcall sequence */
788 flush_scheduled_work();
792 * Ok, the machine is now initialized. None of the devices
793 * have been touched yet, but the CPU subsystem is up and
794 * running, and memory and process management works.
796 * Now we can finally start doing some real work..
798 static void __init
do_basic_setup(void)
802 usermodehelper_init();
810 static void __init
do_pre_smp_initcalls(void)
814 for (fn
= __initcall_start
; fn
< __early_initcall_end
; fn
++)
815 do_one_initcall(*fn
);
818 static void run_init_process(char *init_filename
)
820 argv_init
[0] = init_filename
;
821 kernel_execve(init_filename
, argv_init
, envp_init
);
824 /* This is a non __init function. Force it to be noinline otherwise gcc
825 * makes it inline to init() and it becomes part of init.text section
827 static noinline
int init_post(void)
828 __releases(kernel_lock
)
830 /* need to finish all async __init code before freeing the memory */
831 async_synchronize_full();
835 system_state
= SYSTEM_RUNNING
;
836 numa_default_policy();
839 current
->signal
->flags
|= SIGNAL_UNKILLABLE
;
841 if (ramdisk_execute_command
) {
842 run_init_process(ramdisk_execute_command
);
843 printk(KERN_WARNING
"Failed to execute %s\n",
844 ramdisk_execute_command
);
848 * We try each of these until one succeeds.
850 * The Bourne shell can be used instead of init if we are
851 * trying to recover a really broken machine.
853 if (execute_command
) {
854 run_init_process(execute_command
);
855 printk(KERN_WARNING
"Failed to execute %s. Attempting "
856 "defaults...\n", execute_command
);
858 run_init_process("/sbin/init");
859 run_init_process("/etc/init");
860 run_init_process("/bin/init");
861 run_init_process("/bin/sh");
863 panic("No init found. Try passing init= option to kernel. "
864 "See Linux Documentation/init.txt for guidance.");
867 static int __init
kernel_init(void * unused
)
870 * Wait until kthreadd is all set-up.
872 wait_for_completion(&kthreadd_done
);
876 * init can allocate pages on any node
878 set_mems_allowed(node_states
[N_HIGH_MEMORY
]);
880 * init can run on any cpu.
882 set_cpus_allowed_ptr(current
, cpu_all_mask
);
884 * Tell the world that we're going to be the grim
885 * reaper of innocent orphaned children.
887 * We don't want people to have to make incorrect
888 * assumptions about where in the task array this
891 init_pid_ns
.child_reaper
= current
;
893 cad_pid
= task_pid(current
);
895 smp_prepare_cpus(setup_max_cpus
);
897 do_pre_smp_initcalls();
905 /* Open the /dev/console on the rootfs, this should never fail */
906 if (sys_open((const char __user
*) "/dev/console", O_RDWR
, 0) < 0)
907 printk(KERN_WARNING
"Warning: unable to open an initial console.\n");
912 * check if there is an early userspace init. If yes, let it do all
916 if (!ramdisk_execute_command
)
917 ramdisk_execute_command
= "/init";
919 if (sys_access((const char __user
*) ramdisk_execute_command
, 0) != 0) {
920 ramdisk_execute_command
= NULL
;
925 * Ok, we have completed the initial bootup, and
926 * we're essentially up and running. Get rid of the
927 * initmem segments and start the user-mode stuff..