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/hdreg.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/kernel_stat.h>
32 #include <linux/security.h>
33 #include <linux/workqueue.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/efi.h>
42 #include <linux/taskstats_kern.h>
43 #include <linux/delayacct.h>
44 #include <linux/unistd.h>
45 #include <linux/rmap.h>
46 #include <linux/mempolicy.h>
47 #include <linux/key.h>
48 #include <linux/unwind.h>
49 #include <linux/buffer_head.h>
50 #include <linux/debug_locks.h>
51 #include <linux/lockdep.h>
55 #include <asm/setup.h>
56 #include <asm/sections.h>
57 #include <asm/cacheflush.h>
59 #ifdef CONFIG_X86_LOCAL_APIC
64 * This is one of the first .c files built. Error out early if we have compiler
67 * Versions of gcc older than that listed below may actually compile and link
68 * okay, but the end product can have subtle run time bugs. To avoid associated
69 * bogus bug reports, we flatly refuse to compile with a gcc that is known to be
70 * too old from the very beginning.
72 #if (__GNUC__ < 3) || (__GNUC__ == 3 && __GNUC_MINOR__ < 2)
73 #error Sorry, your GCC is too old. It builds incorrect kernels.
76 static int init(void *);
78 extern void init_IRQ(void);
79 extern void fork_init(unsigned long);
80 extern void mca_init(void);
81 extern void sbus_init(void);
82 extern void sysctl_init(void);
83 extern void signals_init(void);
84 extern void pidhash_init(void);
85 extern void pidmap_init(void);
86 extern void prio_tree_init(void);
87 extern void radix_tree_init(void);
88 extern void free_initmem(void);
89 extern void populate_rootfs(void);
90 extern void driver_init(void);
91 extern void prepare_namespace(void);
93 extern void acpi_early_init(void);
95 static inline void acpi_early_init(void) { }
97 #ifndef CONFIG_DEBUG_RODATA
98 static inline void mark_rodata_ro(void) { }
102 extern void tc_init(void);
105 #if defined(CONFIG_LAB)
106 extern void lab_run(void);
109 enum system_states system_state
;
110 EXPORT_SYMBOL(system_state
);
113 * Boot command-line arguments
115 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
116 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
118 extern void time_init(void);
119 /* Default late time init is NULL. archs can override this later. */
120 void (*late_time_init
)(void);
121 extern void softirq_init(void);
123 /* Untouched command line (eg. for /proc) saved by arch-specific code. */
124 char saved_command_line
[COMMAND_LINE_SIZE
];
126 static char *execute_command
;
127 static char *ramdisk_execute_command
;
129 /* Setup configured maximum number of CPUs to activate */
130 static unsigned int max_cpus
= NR_CPUS
;
133 * If set, this is an indication to the drivers that reset the underlying
134 * device before going ahead with the initialization otherwise driver might
135 * rely on the BIOS and skip the reset operation.
137 * This is useful if kernel is booting in an unreliable environment.
138 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
139 * skipped and devices will be in unknown state.
141 unsigned int reset_devices
;
142 EXPORT_SYMBOL(reset_devices
);
145 * Setup routine for controlling SMP activation
147 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
148 * activation entirely (the MPS table probe still happens, though).
150 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
151 * greater than 0, limits the maximum number of CPUs activated in
154 static int __init
nosmp(char *str
)
160 __setup("nosmp", nosmp
);
162 static int __init
maxcpus(char *str
)
164 get_option(&str
, &max_cpus
);
168 __setup("maxcpus=", maxcpus
);
170 static int __init
set_reset_devices(char *str
)
176 __setup("reset_devices", set_reset_devices
);
178 static char * argv_init
[MAX_INIT_ARGS
+2] = { "init", NULL
, };
179 char * envp_init
[MAX_INIT_ENVS
+2] = { "HOME=/", "TERM=linux", NULL
, };
180 static const char *panic_later
, *panic_param
;
182 extern struct obs_kernel_param __setup_start
[], __setup_end
[];
184 static int __init
obsolete_checksetup(char *line
)
186 struct obs_kernel_param
*p
;
187 int had_early_param
= 0;
191 int n
= strlen(p
->str
);
192 if (!strncmp(line
, p
->str
, n
)) {
194 /* Already done in parse_early_param?
195 * (Needs exact match on param part).
196 * Keep iterating, as we can have early
197 * params and __setups of same names 8( */
198 if (line
[n
] == '\0' || line
[n
] == '=')
200 } else if (!p
->setup_func
) {
201 printk(KERN_WARNING
"Parameter %s is obsolete,"
202 " ignored\n", p
->str
);
204 } else if (p
->setup_func(line
+ n
))
208 } while (p
< __setup_end
);
210 return had_early_param
;
214 * This should be approx 2 Bo*oMips to start (note initial shift), and will
215 * still work even if initially too large, it will just take slightly longer
217 unsigned long loops_per_jiffy
= (1<<12);
219 EXPORT_SYMBOL(loops_per_jiffy
);
221 static int __init
debug_kernel(char *str
)
225 console_loglevel
= 10;
229 static int __init
quiet_kernel(char *str
)
233 console_loglevel
= 4;
237 __setup("debug", debug_kernel
);
238 __setup("quiet", quiet_kernel
);
240 static int __init
loglevel(char *str
)
242 get_option(&str
, &console_loglevel
);
246 __setup("loglevel=", loglevel
);
249 * Unknown boot options get handed to init, unless they look like
252 static int __init
unknown_bootoption(char *param
, char *val
)
254 /* Change NUL term back to "=", to make "param" the whole string. */
256 /* param=val or param="val"? */
257 if (val
== param
+strlen(param
)+1)
259 else if (val
== param
+strlen(param
)+2) {
261 memmove(val
-1, val
, strlen(val
)+1);
267 /* Handle obsolete-style parameters */
268 if (obsolete_checksetup(param
))
272 * Preemptive maintenance for "why didn't my mispelled command
275 if (strchr(param
, '.') && (!val
|| strchr(param
, '.') < val
)) {
276 printk(KERN_ERR
"Unknown boot option `%s': ignoring\n", param
);
284 /* Environment option */
286 for (i
= 0; envp_init
[i
]; i
++) {
287 if (i
== MAX_INIT_ENVS
) {
288 panic_later
= "Too many boot env vars at `%s'";
291 if (!strncmp(param
, envp_init
[i
], val
- param
))
294 envp_init
[i
] = param
;
296 /* Command line option */
298 for (i
= 0; argv_init
[i
]; i
++) {
299 if (i
== MAX_INIT_ARGS
) {
300 panic_later
= "Too many boot init vars at `%s'";
304 argv_init
[i
] = param
;
309 static int __init
init_setup(char *str
)
313 execute_command
= str
;
315 * In case LILO is going to boot us with default command line,
316 * it prepends "auto" before the whole cmdline which makes
317 * the shell think it should execute a script with such name.
318 * So we ignore all arguments entered _before_ init=... [MJ]
320 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
324 __setup("init=", init_setup
);
326 static int __init
rdinit_setup(char *str
)
330 ramdisk_execute_command
= str
;
331 /* See "auto" comment in init_setup */
332 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
336 __setup("rdinit=", rdinit_setup
);
340 #ifdef CONFIG_X86_LOCAL_APIC
341 static void __init
smp_init(void)
343 APIC_init_uniprocessor();
346 #define smp_init() do { } while (0)
349 static inline void setup_per_cpu_areas(void) { }
350 static inline void smp_prepare_cpus(unsigned int maxcpus
) { }
354 #ifdef __GENERIC_PER_CPU
355 unsigned long __per_cpu_offset
[NR_CPUS
] __read_mostly
;
357 EXPORT_SYMBOL(__per_cpu_offset
);
359 static void __init
setup_per_cpu_areas(void)
361 unsigned long size
, i
;
363 unsigned long nr_possible_cpus
= num_possible_cpus();
365 /* Copy section for each CPU (we discard the original) */
366 size
= ALIGN(__per_cpu_end
- __per_cpu_start
, SMP_CACHE_BYTES
);
367 #ifdef CONFIG_MODULES
368 if (size
< PERCPU_ENOUGH_ROOM
)
369 size
= PERCPU_ENOUGH_ROOM
;
371 ptr
= alloc_bootmem(size
* nr_possible_cpus
);
373 for_each_possible_cpu(i
) {
374 __per_cpu_offset
[i
] = ptr
- __per_cpu_start
;
375 memcpy(ptr
, __per_cpu_start
, __per_cpu_end
- __per_cpu_start
);
379 #endif /* !__GENERIC_PER_CPU */
381 /* Called by boot processor to activate the rest. */
382 static void __init
smp_init(void)
386 /* FIXME: This should be done in userspace --RR */
387 for_each_present_cpu(i
) {
388 if (num_online_cpus() >= max_cpus
)
394 /* Any cleanup work */
395 printk(KERN_INFO
"Brought up %ld CPUs\n", (long)num_online_cpus());
396 smp_cpus_done(max_cpus
);
398 /* Get other processors into their bootup holding patterns. */
407 * We need to finalize in a non-__init function or else race conditions
408 * between the root thread and the init thread may cause start_kernel to
409 * be reaped by free_initmem before the root thread has proceeded to
412 * gcc-3.4 accidentally inlines this function, so use noinline.
415 static void noinline
rest_init(void)
416 __releases(kernel_lock
)
418 kernel_thread(init
, NULL
, CLONE_FS
| CLONE_SIGHAND
);
419 numa_default_policy();
423 * The boot idle thread must execute schedule()
424 * at least one to get things moving:
426 preempt_enable_no_resched();
430 /* Call into cpu_idle with preempt disabled */
434 /* Check for early params. */
435 static int __init
do_early_param(char *param
, char *val
)
437 struct obs_kernel_param
*p
;
439 for (p
= __setup_start
; p
< __setup_end
; p
++) {
440 if (p
->early
&& strcmp(param
, p
->str
) == 0) {
441 if (p
->setup_func(val
) != 0)
443 "Malformed early option '%s'\n", param
);
446 /* We accept everything at this stage. */
450 /* Arch code calls this early on, or if not, just before other parsing. */
451 void __init
parse_early_param(void)
453 static __initdata
int done
= 0;
454 static __initdata
char tmp_cmdline
[COMMAND_LINE_SIZE
];
459 /* All fall through to do_early_param. */
460 strlcpy(tmp_cmdline
, saved_command_line
, COMMAND_LINE_SIZE
);
461 parse_args("early options", tmp_cmdline
, NULL
, 0, do_early_param
);
466 * Activate the first processor.
469 static void __init
boot_cpu_init(void)
471 int cpu
= smp_processor_id();
472 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
473 cpu_set(cpu
, cpu_online_map
);
474 cpu_set(cpu
, cpu_present_map
);
475 cpu_set(cpu
, cpu_possible_map
);
478 void __init
__attribute__((weak
)) smp_setup_processor_id(void)
482 asmlinkage
void __init
start_kernel(void)
485 extern struct kernel_param __start___param
[], __stop___param
[];
487 smp_setup_processor_id();
490 * Need to run as early as possible, to initialize the
497 early_boot_irqs_off();
498 early_init_irq_lock_class();
501 * Interrupts are still disabled. Do necessary setups, then
508 printk(linux_banner
);
509 setup_arch(&command_line
);
511 setup_per_cpu_areas();
512 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
515 * Set up the scheduler prior starting any interrupts (such as the
516 * timer interrupt). Full topology setup happens at smp_init()
517 * time - but meanwhile we still have a functioning scheduler.
521 * Disable preemption - early bootup scheduling is extremely
522 * fragile until we cpu_idle() for the first time.
525 build_all_zonelists();
527 printk(KERN_NOTICE
"Kernel command line: %s\n", saved_command_line
);
529 parse_args("Booting kernel", command_line
, __start___param
,
530 __stop___param
- __start___param
,
531 &unknown_bootoption
);
543 if (!irqs_disabled())
544 printk("start_kernel(): bug: interrupts were enabled early\n");
545 early_boot_irqs_on();
549 * HACK ALERT! This is early. We're enabling the console before
550 * we've done PCI setups etc, and console_init() must be aware of
551 * this. But we do want output early, in case something goes wrong.
555 panic(panic_later
, panic_param
);
560 * Need to run this when irqs are enabled, because it wants
561 * to self-test [hard/soft]-irqs on/off lock inversion bugs
566 #ifdef CONFIG_BLK_DEV_INITRD
567 if (initrd_start
&& !initrd_below_start_ok
&&
568 initrd_start
< min_low_pfn
<< PAGE_SHIFT
) {
569 printk(KERN_CRIT
"initrd overwritten (0x%08lx < 0x%08lx) - "
570 "disabling it.\n",initrd_start
,min_low_pfn
<< PAGE_SHIFT
);
574 vfs_caches_init_early();
578 setup_per_cpu_pageset();
584 pgtable_cache_init();
589 efi_enter_virtual_mode();
591 fork_init(num_physpages
);
597 vfs_caches_init(num_physpages
);
600 /* rootfs populating might need page-writeback */
601 page_writeback_init();
602 #ifdef CONFIG_PROC_FS
606 taskstats_init_early();
611 acpi_early_init(); /* before LAPIC and SMP init */
613 /* Do the rest non-__init'ed, we're now alive */
617 static int __initdata initcall_debug
;
619 static int __init
initcall_debug_setup(char *str
)
624 __setup("initcall_debug", initcall_debug_setup
);
626 struct task_struct
*child_reaper
= &init_task
;
628 extern initcall_t __initcall_start
[], __initcall_end
[];
630 static void __init
do_initcalls(void)
633 int count
= preempt_count();
635 for (call
= __initcall_start
; call
< __initcall_end
; call
++) {
640 if (initcall_debug
) {
641 printk("Calling initcall 0x%p", *call
);
642 print_fn_descriptor_symbol(": %s()",
643 (unsigned long) *call
);
649 if (result
&& result
!= -ENODEV
&& initcall_debug
) {
650 sprintf(msgbuf
, "error code %d", result
);
653 if (preempt_count() != count
) {
654 msg
= "preemption imbalance";
655 preempt_count() = count
;
657 if (irqs_disabled()) {
658 msg
= "disabled interrupts";
662 printk(KERN_WARNING
"initcall at 0x%p", *call
);
663 print_fn_descriptor_symbol(": %s()",
664 (unsigned long) *call
);
665 printk(": returned with %s\n", msg
);
669 /* Make sure there is no pending stuff from the initcall sequence */
670 flush_scheduled_work();
674 * Ok, the machine is now initialized. None of the devices
675 * have been touched yet, but the CPU subsystem is up and
676 * running, and memory and process management works.
678 * Now we can finally start doing some real work..
680 static void __init
do_basic_setup(void)
682 /* drivers will send hotplug events */
684 usermodehelper_init();
694 static void do_pre_smp_initcalls(void)
696 extern int spawn_ksoftirqd(void);
698 extern int migration_init(void);
703 spawn_softlockup_task();
706 #if !defined(CONFIG_LAB)
707 static void run_init_process(char *init_filename
)
709 argv_init
[0] = init_filename
;
710 kernel_execve(init_filename
, argv_init
, envp_init
);
714 static int init(void * unused
)
718 * init can run on any cpu.
720 set_cpus_allowed(current
, CPU_MASK_ALL
);
722 * Tell the world that we're going to be the grim
723 * reaper of innocent orphaned children.
725 * We don't want people to have to make incorrect
726 * assumptions about where in the task array this
729 child_reaper
= current
;
731 cad_pid
= task_pid(current
);
733 smp_prepare_cpus(max_cpus
);
735 do_pre_smp_initcalls();
743 * Do this before initcalls, because some drivers want to access
751 * check if there is an early userspace init. If yes, let it do all
755 if (!ramdisk_execute_command
)
756 ramdisk_execute_command
= "/init";
758 if (sys_access((const char __user
*) ramdisk_execute_command
, 0) != 0) {
759 ramdisk_execute_command
= NULL
;
764 * Ok, we have completed the initial bootup, and
765 * we're essentially up and running. Get rid of the
766 * initmem segments and start the user-mode stuff..
771 system_state
= SYSTEM_RUNNING
;
772 numa_default_policy();
774 #if !defined(CONFIG_LAB)
775 if (sys_open((const char __user
*) "/dev/console", O_RDWR
, 0) < 0)
776 printk(KERN_WARNING
"Warning: unable to open an initial console.\n");
781 if (ramdisk_execute_command
) {
782 run_init_process(ramdisk_execute_command
);
783 printk(KERN_WARNING
"Failed to execute %s\n",
784 ramdisk_execute_command
);
788 * We try each of these until one succeeds.
790 * The Bourne shell can be used instead of init if we are
791 * trying to recover a really broken machine.
793 if (execute_command
) {
794 run_init_process(execute_command
);
795 printk(KERN_WARNING
"Failed to execute %s. Attempting "
796 "defaults...\n", execute_command
);
798 run_init_process("/sbin/init");
799 run_init_process("/etc/init");
800 run_init_process("/bin/init");
801 run_init_process("/bin/sh");
803 panic("No init found. Try passing init= option to kernel.");