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/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/ftrace.h>
66 #include <linux/async.h>
67 #include <linux/kmemtrace.h>
68 #include <trace/boot.h>
72 #include <asm/setup.h>
73 #include <asm/sections.h>
74 #include <asm/cacheflush.h>
76 #ifdef CONFIG_X86_LOCAL_APIC
80 static int kernel_init(void *);
82 extern void init_IRQ(void);
83 extern void fork_init(unsigned long);
84 extern void mca_init(void);
85 extern void sbus_init(void);
86 extern void prio_tree_init(void);
87 extern void radix_tree_init(void);
88 extern void free_initmem(void);
90 extern void acpi_early_init(void);
92 static inline void acpi_early_init(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 __initdata setup_max_cpus
= NR_CPUS
;
131 * Setup routine for controlling SMP activation
133 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
134 * activation entirely (the MPS table probe still happens, though).
136 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
137 * greater than 0, limits the maximum number of CPUs activated in
141 void __weak
arch_disable_smp_support(void) { }
143 static int __init
nosmp(char *str
)
146 arch_disable_smp_support();
151 early_param("nosmp", nosmp
);
153 static int __init
maxcpus(char *str
)
155 get_option(&str
, &setup_max_cpus
);
156 if (setup_max_cpus
== 0)
157 arch_disable_smp_support();
162 early_param("maxcpus", maxcpus
);
164 const unsigned int setup_max_cpus
= NR_CPUS
;
168 * If set, this is an indication to the drivers that reset the underlying
169 * device before going ahead with the initialization otherwise driver might
170 * rely on the BIOS and skip the reset operation.
172 * This is useful if kernel is booting in an unreliable environment.
173 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
174 * skipped and devices will be in unknown state.
176 unsigned int reset_devices
;
177 EXPORT_SYMBOL(reset_devices
);
179 static int __init
set_reset_devices(char *str
)
185 __setup("reset_devices", set_reset_devices
);
187 static char * argv_init
[MAX_INIT_ARGS
+2] = { "init", NULL
, };
188 char * envp_init
[MAX_INIT_ENVS
+2] = { "HOME=/", "TERM=linux", NULL
, };
189 static const char *panic_later
, *panic_param
;
191 extern struct obs_kernel_param __setup_start
[], __setup_end
[];
193 static int __init
obsolete_checksetup(char *line
)
195 struct obs_kernel_param
*p
;
196 int had_early_param
= 0;
200 int n
= strlen(p
->str
);
201 if (!strncmp(line
, p
->str
, n
)) {
203 /* Already done in parse_early_param?
204 * (Needs exact match on param part).
205 * Keep iterating, as we can have early
206 * params and __setups of same names 8( */
207 if (line
[n
] == '\0' || line
[n
] == '=')
209 } else if (!p
->setup_func
) {
210 printk(KERN_WARNING
"Parameter %s is obsolete,"
211 " ignored\n", p
->str
);
213 } else if (p
->setup_func(line
+ n
))
217 } while (p
< __setup_end
);
219 return had_early_param
;
223 * This should be approx 2 Bo*oMips to start (note initial shift), and will
224 * still work even if initially too large, it will just take slightly longer
226 unsigned long loops_per_jiffy
= (1<<12);
228 EXPORT_SYMBOL(loops_per_jiffy
);
230 static int __init
debug_kernel(char *str
)
232 console_loglevel
= 10;
236 static int __init
quiet_kernel(char *str
)
238 console_loglevel
= 4;
242 early_param("debug", debug_kernel
);
243 early_param("quiet", quiet_kernel
);
245 static int __init
loglevel(char *str
)
247 get_option(&str
, &console_loglevel
);
251 early_param("loglevel", loglevel
);
254 * Unknown boot options get handed to init, unless they look like
257 static int __init
unknown_bootoption(char *param
, char *val
)
259 /* Change NUL term back to "=", to make "param" the whole string. */
261 /* param=val or param="val"? */
262 if (val
== param
+strlen(param
)+1)
264 else if (val
== param
+strlen(param
)+2) {
266 memmove(val
-1, val
, strlen(val
)+1);
272 /* Handle obsolete-style parameters */
273 if (obsolete_checksetup(param
))
277 * Preemptive maintenance for "why didn't my misspelled command
280 if (strchr(param
, '.') && (!val
|| strchr(param
, '.') < val
)) {
281 printk(KERN_ERR
"Unknown boot option `%s': ignoring\n", param
);
289 /* Environment option */
291 for (i
= 0; envp_init
[i
]; i
++) {
292 if (i
== MAX_INIT_ENVS
) {
293 panic_later
= "Too many boot env vars at `%s'";
296 if (!strncmp(param
, envp_init
[i
], val
- param
))
299 envp_init
[i
] = param
;
301 /* Command line option */
303 for (i
= 0; argv_init
[i
]; i
++) {
304 if (i
== MAX_INIT_ARGS
) {
305 panic_later
= "Too many boot init vars at `%s'";
309 argv_init
[i
] = param
;
314 #ifdef CONFIG_DEBUG_PAGEALLOC
315 int __read_mostly debug_pagealloc_enabled
= 0;
318 static int __init
init_setup(char *str
)
322 execute_command
= str
;
324 * In case LILO is going to boot us with default command line,
325 * it prepends "auto" before the whole cmdline which makes
326 * the shell think it should execute a script with such name.
327 * So we ignore all arguments entered _before_ init=... [MJ]
329 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
333 __setup("init=", init_setup
);
335 static int __init
rdinit_setup(char *str
)
339 ramdisk_execute_command
= str
;
340 /* See "auto" comment in init_setup */
341 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
345 __setup("rdinit=", rdinit_setup
);
349 #ifdef CONFIG_X86_LOCAL_APIC
350 static void __init
smp_init(void)
352 APIC_init_uniprocessor();
355 #define smp_init() do { } while (0)
358 static inline void setup_per_cpu_areas(void) { }
359 static inline void setup_nr_cpu_ids(void) { }
360 static inline void smp_prepare_cpus(unsigned int maxcpus
) { }
364 #if NR_CPUS > BITS_PER_LONG
365 cpumask_t cpu_mask_all __read_mostly
= CPU_MASK_ALL
;
366 EXPORT_SYMBOL(cpu_mask_all
);
369 /* Setup number of possible processor ids */
370 int nr_cpu_ids __read_mostly
= NR_CPUS
;
371 EXPORT_SYMBOL(nr_cpu_ids
);
373 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
374 static void __init
setup_nr_cpu_ids(void)
376 nr_cpu_ids
= find_last_bit(cpumask_bits(cpu_possible_mask
),NR_CPUS
) + 1;
379 #ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA
380 unsigned long __per_cpu_offset
[NR_CPUS
] __read_mostly
;
382 EXPORT_SYMBOL(__per_cpu_offset
);
384 static void __init
setup_per_cpu_areas(void)
386 unsigned long size
, i
;
388 unsigned long nr_possible_cpus
= num_possible_cpus();
390 /* Copy section for each CPU (we discard the original) */
391 size
= ALIGN(PERCPU_ENOUGH_ROOM
, PAGE_SIZE
);
392 ptr
= alloc_bootmem_pages(size
* nr_possible_cpus
);
394 for_each_possible_cpu(i
) {
395 __per_cpu_offset
[i
] = ptr
- __per_cpu_start
;
396 memcpy(ptr
, __per_cpu_start
, __per_cpu_end
- __per_cpu_start
);
400 #endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */
402 /* Called by boot processor to activate the rest. */
403 static void __init
smp_init(void)
408 * Set up the current CPU as possible to migrate to.
409 * The other ones will be done by cpu_up/cpu_down()
411 set_cpu_active(smp_processor_id(), true);
413 /* FIXME: This should be done in userspace --RR */
414 for_each_present_cpu(cpu
) {
415 if (num_online_cpus() >= setup_max_cpus
)
417 if (!cpu_online(cpu
))
421 /* Any cleanup work */
422 printk(KERN_INFO
"Brought up %ld CPUs\n", (long)num_online_cpus());
423 smp_cpus_done(setup_max_cpus
);
429 * We need to store the untouched command line for future reference.
430 * We also need to store the touched command line since the parameter
431 * parsing is performed in place, and we should allow a component to
432 * store reference of name/value for future reference.
434 static void __init
setup_command_line(char *command_line
)
436 saved_command_line
= alloc_bootmem(strlen (boot_command_line
)+1);
437 static_command_line
= alloc_bootmem(strlen (command_line
)+1);
438 strcpy (saved_command_line
, boot_command_line
);
439 strcpy (static_command_line
, command_line
);
443 * We need to finalize in a non-__init function or else race conditions
444 * between the root thread and the init thread may cause start_kernel to
445 * be reaped by free_initmem before the root thread has proceeded to
448 * gcc-3.4 accidentally inlines this function, so use noinline.
451 static noinline
void __init_refok
rest_init(void)
452 __releases(kernel_lock
)
456 kernel_thread(kernel_init
, NULL
, CLONE_FS
| CLONE_SIGHAND
);
457 numa_default_policy();
458 pid
= kernel_thread(kthreadd
, NULL
, CLONE_FS
| CLONE_FILES
);
459 kthreadd_task
= find_task_by_pid_ns(pid
, &init_pid_ns
);
463 * The boot idle thread must execute schedule()
464 * at least once to get things moving:
466 init_idle_bootup_task(current
);
467 rcu_scheduler_starting();
468 preempt_enable_no_resched();
472 /* Call into cpu_idle with preempt disabled */
476 /* Check for early params. */
477 static int __init
do_early_param(char *param
, char *val
)
479 struct obs_kernel_param
*p
;
481 for (p
= __setup_start
; p
< __setup_end
; p
++) {
482 if ((p
->early
&& strcmp(param
, p
->str
) == 0) ||
483 (strcmp(param
, "console") == 0 &&
484 strcmp(p
->str
, "earlycon") == 0)
486 if (p
->setup_func(val
) != 0)
488 "Malformed early option '%s'\n", param
);
491 /* We accept everything at this stage. */
495 void __init
parse_early_options(char *cmdline
)
497 parse_args("early options", cmdline
, NULL
, 0, do_early_param
);
500 /* Arch code calls this early on, or if not, just before other parsing. */
501 void __init
parse_early_param(void)
503 static __initdata
int done
= 0;
504 static __initdata
char tmp_cmdline
[COMMAND_LINE_SIZE
];
509 /* All fall through to do_early_param. */
510 strlcpy(tmp_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
511 parse_early_options(tmp_cmdline
);
516 * Activate the first processor.
519 static void __init
boot_cpu_init(void)
521 int cpu
= smp_processor_id();
522 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
523 set_cpu_online(cpu
, true);
524 set_cpu_present(cpu
, true);
525 set_cpu_possible(cpu
, true);
528 void __init __weak
smp_setup_processor_id(void)
532 void __init __weak
thread_info_cache_init(void)
536 asmlinkage
void __init
start_kernel(void)
539 extern struct kernel_param __start___param
[], __stop___param
[];
541 smp_setup_processor_id();
544 * Need to run as early as possible, to initialize the
548 debug_objects_early_init();
551 * Set up the the initial canary ASAP:
553 boot_init_stack_canary();
558 early_boot_irqs_off();
559 early_init_irq_lock_class();
562 * Interrupts are still disabled. Do necessary setups, then
569 printk(KERN_NOTICE
"%s", linux_banner
);
570 setup_arch(&command_line
);
571 mm_init_owner(&init_mm
, &init_task
);
572 setup_command_line(command_line
);
573 setup_per_cpu_areas();
575 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
577 build_all_zonelists();
580 printk(KERN_NOTICE
"Kernel command line: %s\n", boot_command_line
);
582 parse_args("Booting kernel", static_command_line
, __start___param
,
583 __stop___param
- __start___param
,
584 &unknown_bootoption
);
586 * These use large bootmem allocations and must precede
591 vfs_caches_init_early();
595 * Set up kernel memory allocators
600 * Set up the scheduler prior starting any interrupts (such as the
601 * timer interrupt). Full topology setup happens at smp_init()
602 * time - but meanwhile we still have a functioning scheduler.
606 * Disable preemption - early bootup scheduling is extremely
607 * fragile until we cpu_idle() for the first time.
610 if (!irqs_disabled()) {
611 printk(KERN_WARNING
"start_kernel(): bug: interrupts were "
612 "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();
633 * HACK ALERT! This is early. We're enabling the console before
634 * we've done PCI setups etc, and console_init() must be aware of
635 * this. But we do want output early, in case something goes wrong.
639 panic(panic_later
, panic_param
);
644 * Need to run this when irqs are enabled, because it wants
645 * to self-test [hard/soft]-irqs on/off lock inversion bugs
650 #ifdef CONFIG_BLK_DEV_INITRD
651 if (initrd_start
&& !initrd_below_start_ok
&&
652 page_to_pfn(virt_to_page((void *)initrd_start
)) < min_low_pfn
) {
653 printk(KERN_CRIT
"initrd overwritten (0x%08lx < 0x%08lx) - "
655 page_to_pfn(virt_to_page((void *)initrd_start
)),
662 enable_debug_pagealloc();
665 debug_objects_mem_init();
667 setup_per_cpu_pageset();
673 pgtable_cache_init();
678 efi_enter_virtual_mode();
680 thread_info_cache_init();
682 fork_init(num_physpages
);
687 vfs_caches_init(num_physpages
);
690 /* rootfs populating might need page-writeback */
691 page_writeback_init();
692 #ifdef CONFIG_PROC_FS
697 taskstats_init_early();
702 acpi_early_init(); /* before LAPIC and SMP init */
706 /* Do the rest non-__init'ed, we're now alive */
711 core_param(initcall_debug
, initcall_debug
, bool, 0644);
713 int do_one_initcall(initcall_t fn
)
715 int count
= preempt_count();
716 ktime_t calltime
, delta
, rettime
;
718 struct boot_trace_call call
;
719 struct boot_trace_ret ret
;
721 if (initcall_debug
) {
722 call
.caller
= task_pid_nr(current
);
723 printk("calling %pF @ %i\n", fn
, call
.caller
);
724 calltime
= ktime_get();
725 trace_boot_call(&call
, fn
);
731 if (initcall_debug
) {
732 disable_boot_trace();
733 rettime
= ktime_get();
734 delta
= ktime_sub(rettime
, calltime
);
735 ret
.duration
= (unsigned long long) ktime_to_ns(delta
) >> 10;
736 trace_boot_ret(&ret
, fn
);
737 printk("initcall %pF returned %d after %Ld usecs\n", fn
,
738 ret
.result
, ret
.duration
);
743 if (ret
.result
&& ret
.result
!= -ENODEV
&& initcall_debug
)
744 sprintf(msgbuf
, "error code %d ", ret
.result
);
746 if (preempt_count() != count
) {
747 strlcat(msgbuf
, "preemption imbalance ", sizeof(msgbuf
));
748 preempt_count() = count
;
750 if (irqs_disabled()) {
751 strlcat(msgbuf
, "disabled interrupts ", sizeof(msgbuf
));
755 printk("initcall %pF returned with %s\n", fn
, msgbuf
);
762 extern initcall_t __initcall_start
[], __initcall_end
[], __early_initcall_end
[];
764 static void __init
do_initcalls(void)
768 for (call
= __early_initcall_end
; call
< __initcall_end
; call
++)
769 do_one_initcall(*call
);
771 /* Make sure there is no pending stuff from the initcall sequence */
772 flush_scheduled_work();
776 * Ok, the machine is now initialized. None of the devices
777 * have been touched yet, but the CPU subsystem is up and
778 * running, and memory and process management works.
780 * Now we can finally start doing some real work..
782 static void __init
do_basic_setup(void)
784 rcu_init_sched(); /* needed by module_init stage. */
787 usermodehelper_init();
793 static void __init
do_pre_smp_initcalls(void)
797 for (call
= __initcall_start
; call
< __early_initcall_end
; call
++)
798 do_one_initcall(*call
);
801 static void run_init_process(char *init_filename
)
803 argv_init
[0] = init_filename
;
804 kernel_execve(init_filename
, argv_init
, envp_init
);
807 /* This is a non __init function. Force it to be noinline otherwise gcc
808 * makes it inline to init() and it becomes part of init.text section
810 static noinline
int init_post(void)
811 __releases(kernel_lock
)
813 /* need to finish all async __init code before freeing the memory */
814 async_synchronize_full();
818 system_state
= SYSTEM_RUNNING
;
819 numa_default_policy();
821 if (sys_open((const char __user
*) "/dev/console", O_RDWR
, 0) < 0)
822 printk(KERN_WARNING
"Warning: unable to open an initial console.\n");
827 current
->signal
->flags
|= SIGNAL_UNKILLABLE
;
829 if (ramdisk_execute_command
) {
830 run_init_process(ramdisk_execute_command
);
831 printk(KERN_WARNING
"Failed to execute %s\n",
832 ramdisk_execute_command
);
836 * We try each of these until one succeeds.
838 * The Bourne shell can be used instead of init if we are
839 * trying to recover a really broken machine.
841 if (execute_command
) {
842 run_init_process(execute_command
);
843 printk(KERN_WARNING
"Failed to execute %s. Attempting "
844 "defaults...\n", execute_command
);
846 run_init_process("/sbin/init");
847 run_init_process("/etc/init");
848 run_init_process("/bin/init");
849 run_init_process("/bin/sh");
851 panic("No init found. Try passing init= option to kernel.");
854 static int __init
kernel_init(void * unused
)
858 * init can run on any cpu.
860 set_cpus_allowed_ptr(current
, cpu_all_mask
);
862 * Tell the world that we're going to be the grim
863 * reaper of innocent orphaned children.
865 * We don't want people to have to make incorrect
866 * assumptions about where in the task array this
869 init_pid_ns
.child_reaper
= current
;
871 cad_pid
= task_pid(current
);
873 smp_prepare_cpus(setup_max_cpus
);
875 do_pre_smp_initcalls();
884 * check if there is an early userspace init. If yes, let it do all
888 if (!ramdisk_execute_command
)
889 ramdisk_execute_command
= "/init";
891 if (sys_access((const char __user
*) ramdisk_execute_command
, 0) != 0) {
892 ramdisk_execute_command
= NULL
;
897 * Ok, we have completed the initial bootup, and
898 * we're essentially up and running. Get rid of the
899 * initmem segments and start the user-mode stuff..