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/start_kernel.h>
33 #include <linux/security.h>
34 #include <linux/workqueue.h>
35 #include <linux/profile.h>
36 #include <linux/rcupdate.h>
37 #include <linux/moduleparam.h>
38 #include <linux/kallsyms.h>
39 #include <linux/writeback.h>
40 #include <linux/cpu.h>
41 #include <linux/cpuset.h>
42 #include <linux/efi.h>
43 #include <linux/taskstats_kern.h>
44 #include <linux/delayacct.h>
45 #include <linux/unistd.h>
46 #include <linux/rmap.h>
47 #include <linux/mempolicy.h>
48 #include <linux/key.h>
49 #include <linux/unwind.h>
50 #include <linux/buffer_head.h>
51 #include <linux/debug_locks.h>
52 #include <linux/lockdep.h>
56 #include <asm/setup.h>
57 #include <asm/sections.h>
58 #include <asm/cacheflush.h>
60 #ifdef CONFIG_X86_LOCAL_APIC
65 * This is one of the first .c files built. Error out early if we have compiler
68 * Versions of gcc older than that listed below may actually compile and link
69 * okay, but the end product can have subtle run time bugs. To avoid associated
70 * bogus bug reports, we flatly refuse to compile with a gcc that is known to be
71 * too old from the very beginning.
73 #if (__GNUC__ < 3) || (__GNUC__ == 3 && __GNUC_MINOR__ < 2)
74 #error Sorry, your GCC is too old. It builds incorrect kernels.
77 static int init(void *);
79 extern void init_IRQ(void);
80 extern void fork_init(unsigned long);
81 extern void mca_init(void);
82 extern void sbus_init(void);
83 extern void sysctl_init(void);
84 extern void signals_init(void);
85 extern void pidhash_init(void);
86 extern void pidmap_init(void);
87 extern void prio_tree_init(void);
88 extern void radix_tree_init(void);
89 extern void free_initmem(void);
90 extern void populate_rootfs(void);
91 extern void driver_init(void);
92 extern void prepare_namespace(void);
94 extern void acpi_early_init(void);
96 static inline void acpi_early_init(void) { }
98 #ifndef CONFIG_DEBUG_RODATA
99 static inline void mark_rodata_ro(void) { }
103 extern void tc_init(void);
106 enum system_states system_state
;
107 EXPORT_SYMBOL(system_state
);
110 * Boot command-line arguments
112 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
113 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
115 extern void time_init(void);
116 /* Default late time init is NULL. archs can override this later. */
117 void (*late_time_init
)(void);
118 extern void softirq_init(void);
120 /* Untouched command line (eg. for /proc) saved by arch-specific code. */
121 char saved_command_line
[COMMAND_LINE_SIZE
];
123 static char *execute_command
;
124 static char *ramdisk_execute_command
;
126 /* Setup configured maximum number of CPUs to activate */
127 static unsigned int max_cpus
= NR_CPUS
;
130 * If set, this is an indication to the drivers that reset the underlying
131 * device before going ahead with the initialization otherwise driver might
132 * rely on the BIOS and skip the reset operation.
134 * This is useful if kernel is booting in an unreliable environment.
135 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
136 * skipped and devices will be in unknown state.
138 unsigned int reset_devices
;
139 EXPORT_SYMBOL(reset_devices
);
142 * Setup routine for controlling SMP activation
144 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
145 * activation entirely (the MPS table probe still happens, though).
147 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
148 * greater than 0, limits the maximum number of CPUs activated in
151 static int __init
nosmp(char *str
)
157 __setup("nosmp", nosmp
);
159 static int __init
maxcpus(char *str
)
161 get_option(&str
, &max_cpus
);
165 __setup("maxcpus=", maxcpus
);
167 static int __init
set_reset_devices(char *str
)
173 __setup("reset_devices", set_reset_devices
);
175 static char * argv_init
[MAX_INIT_ARGS
+2] = { "init", NULL
, };
176 char * envp_init
[MAX_INIT_ENVS
+2] = { "HOME=/", "TERM=linux", NULL
, };
177 static const char *panic_later
, *panic_param
;
179 extern struct obs_kernel_param __setup_start
[], __setup_end
[];
181 static int __init
obsolete_checksetup(char *line
)
183 struct obs_kernel_param
*p
;
184 int had_early_param
= 0;
188 int n
= strlen(p
->str
);
189 if (!strncmp(line
, p
->str
, n
)) {
191 /* Already done in parse_early_param?
192 * (Needs exact match on param part).
193 * Keep iterating, as we can have early
194 * params and __setups of same names 8( */
195 if (line
[n
] == '\0' || line
[n
] == '=')
197 } else if (!p
->setup_func
) {
198 printk(KERN_WARNING
"Parameter %s is obsolete,"
199 " ignored\n", p
->str
);
201 } else if (p
->setup_func(line
+ n
))
205 } while (p
< __setup_end
);
207 return had_early_param
;
211 * This should be approx 2 Bo*oMips to start (note initial shift), and will
212 * still work even if initially too large, it will just take slightly longer
214 unsigned long loops_per_jiffy
= (1<<12);
216 EXPORT_SYMBOL(loops_per_jiffy
);
218 static int __init
debug_kernel(char *str
)
222 console_loglevel
= 10;
226 static int __init
quiet_kernel(char *str
)
230 console_loglevel
= 4;
234 __setup("debug", debug_kernel
);
235 __setup("quiet", quiet_kernel
);
237 static int __init
loglevel(char *str
)
239 get_option(&str
, &console_loglevel
);
243 __setup("loglevel=", loglevel
);
246 * Unknown boot options get handed to init, unless they look like
249 static int __init
unknown_bootoption(char *param
, char *val
)
251 /* Change NUL term back to "=", to make "param" the whole string. */
253 /* param=val or param="val"? */
254 if (val
== param
+strlen(param
)+1)
256 else if (val
== param
+strlen(param
)+2) {
258 memmove(val
-1, val
, strlen(val
)+1);
264 /* Handle obsolete-style parameters */
265 if (obsolete_checksetup(param
))
269 * Preemptive maintenance for "why didn't my mispelled command
272 if (strchr(param
, '.') && (!val
|| strchr(param
, '.') < val
)) {
273 printk(KERN_ERR
"Unknown boot option `%s': ignoring\n", param
);
281 /* Environment option */
283 for (i
= 0; envp_init
[i
]; i
++) {
284 if (i
== MAX_INIT_ENVS
) {
285 panic_later
= "Too many boot env vars at `%s'";
288 if (!strncmp(param
, envp_init
[i
], val
- param
))
291 envp_init
[i
] = param
;
293 /* Command line option */
295 for (i
= 0; argv_init
[i
]; i
++) {
296 if (i
== MAX_INIT_ARGS
) {
297 panic_later
= "Too many boot init vars at `%s'";
301 argv_init
[i
] = param
;
306 static int __init
init_setup(char *str
)
310 execute_command
= str
;
312 * In case LILO is going to boot us with default command line,
313 * it prepends "auto" before the whole cmdline which makes
314 * the shell think it should execute a script with such name.
315 * So we ignore all arguments entered _before_ init=... [MJ]
317 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
321 __setup("init=", init_setup
);
323 static int __init
rdinit_setup(char *str
)
327 ramdisk_execute_command
= str
;
328 /* See "auto" comment in init_setup */
329 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
333 __setup("rdinit=", rdinit_setup
);
337 #ifdef CONFIG_X86_LOCAL_APIC
338 static void __init
smp_init(void)
340 APIC_init_uniprocessor();
343 #define smp_init() do { } while (0)
346 static inline void setup_per_cpu_areas(void) { }
347 static inline void smp_prepare_cpus(unsigned int maxcpus
) { }
351 #ifdef __GENERIC_PER_CPU
352 unsigned long __per_cpu_offset
[NR_CPUS
] __read_mostly
;
354 EXPORT_SYMBOL(__per_cpu_offset
);
356 static void __init
setup_per_cpu_areas(void)
358 unsigned long size
, i
;
360 unsigned long nr_possible_cpus
= num_possible_cpus();
362 /* Copy section for each CPU (we discard the original) */
363 size
= ALIGN(__per_cpu_end
- __per_cpu_start
, SMP_CACHE_BYTES
);
364 #ifdef CONFIG_MODULES
365 if (size
< PERCPU_ENOUGH_ROOM
)
366 size
= PERCPU_ENOUGH_ROOM
;
368 ptr
= alloc_bootmem(size
* nr_possible_cpus
);
370 for_each_possible_cpu(i
) {
371 __per_cpu_offset
[i
] = ptr
- __per_cpu_start
;
372 memcpy(ptr
, __per_cpu_start
, __per_cpu_end
- __per_cpu_start
);
376 #endif /* !__GENERIC_PER_CPU */
378 /* Called by boot processor to activate the rest. */
379 static void __init
smp_init(void)
383 /* FIXME: This should be done in userspace --RR */
384 for_each_present_cpu(i
) {
385 if (num_online_cpus() >= max_cpus
)
391 /* Any cleanup work */
392 printk(KERN_INFO
"Brought up %ld CPUs\n", (long)num_online_cpus());
393 smp_cpus_done(max_cpus
);
395 /* Get other processors into their bootup holding patterns. */
404 * We need to finalize in a non-__init function or else race conditions
405 * between the root thread and the init thread may cause start_kernel to
406 * be reaped by free_initmem before the root thread has proceeded to
409 * gcc-3.4 accidentally inlines this function, so use noinline.
412 static void noinline
rest_init(void)
413 __releases(kernel_lock
)
415 kernel_thread(init
, NULL
, CLONE_FS
| CLONE_SIGHAND
);
416 numa_default_policy();
420 * The boot idle thread must execute schedule()
421 * at least one to get things moving:
423 preempt_enable_no_resched();
427 /* Call into cpu_idle with preempt disabled */
431 /* Check for early params. */
432 static int __init
do_early_param(char *param
, char *val
)
434 struct obs_kernel_param
*p
;
436 for (p
= __setup_start
; p
< __setup_end
; p
++) {
437 if (p
->early
&& strcmp(param
, p
->str
) == 0) {
438 if (p
->setup_func(val
) != 0)
440 "Malformed early option '%s'\n", param
);
443 /* We accept everything at this stage. */
447 /* Arch code calls this early on, or if not, just before other parsing. */
448 void __init
parse_early_param(void)
450 static __initdata
int done
= 0;
451 static __initdata
char tmp_cmdline
[COMMAND_LINE_SIZE
];
456 /* All fall through to do_early_param. */
457 strlcpy(tmp_cmdline
, saved_command_line
, COMMAND_LINE_SIZE
);
458 parse_args("early options", tmp_cmdline
, NULL
, 0, do_early_param
);
463 * Activate the first processor.
466 static void __init
boot_cpu_init(void)
468 int cpu
= smp_processor_id();
469 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
470 cpu_set(cpu
, cpu_online_map
);
471 cpu_set(cpu
, cpu_present_map
);
472 cpu_set(cpu
, cpu_possible_map
);
475 void __init
__attribute__((weak
)) smp_setup_processor_id(void)
479 asmlinkage
void __init
start_kernel(void)
482 extern struct kernel_param __start___param
[], __stop___param
[];
484 smp_setup_processor_id();
487 * Need to run as early as possible, to initialize the
494 early_boot_irqs_off();
495 early_init_irq_lock_class();
498 * Interrupts are still disabled. Do necessary setups, then
505 printk(linux_banner
);
506 setup_arch(&command_line
);
508 setup_per_cpu_areas();
509 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
512 * Set up the scheduler prior starting any interrupts (such as the
513 * timer interrupt). Full topology setup happens at smp_init()
514 * time - but meanwhile we still have a functioning scheduler.
518 * Disable preemption - early bootup scheduling is extremely
519 * fragile until we cpu_idle() for the first time.
522 build_all_zonelists();
524 printk(KERN_NOTICE
"Kernel command line: %s\n", saved_command_line
);
526 parse_args("Booting kernel", command_line
, __start___param
,
527 __stop___param
- __start___param
,
528 &unknown_bootoption
);
540 if (!irqs_disabled())
541 printk("start_kernel(): bug: interrupts were enabled early\n");
542 early_boot_irqs_on();
546 * HACK ALERT! This is early. We're enabling the console before
547 * we've done PCI setups etc, and console_init() must be aware of
548 * this. But we do want output early, in case something goes wrong.
552 panic(panic_later
, panic_param
);
557 * Need to run this when irqs are enabled, because it wants
558 * to self-test [hard/soft]-irqs on/off lock inversion bugs
563 #ifdef CONFIG_BLK_DEV_INITRD
564 if (initrd_start
&& !initrd_below_start_ok
&&
565 initrd_start
< min_low_pfn
<< PAGE_SHIFT
) {
566 printk(KERN_CRIT
"initrd overwritten (0x%08lx < 0x%08lx) - "
567 "disabling it.\n",initrd_start
,min_low_pfn
<< PAGE_SHIFT
);
571 vfs_caches_init_early();
575 setup_per_cpu_pageset();
581 pgtable_cache_init();
586 efi_enter_virtual_mode();
588 fork_init(num_physpages
);
594 vfs_caches_init(num_physpages
);
597 /* rootfs populating might need page-writeback */
598 page_writeback_init();
599 #ifdef CONFIG_PROC_FS
603 taskstats_init_early();
608 acpi_early_init(); /* before LAPIC and SMP init */
610 /* Do the rest non-__init'ed, we're now alive */
614 static int __initdata initcall_debug
;
616 static int __init
initcall_debug_setup(char *str
)
621 __setup("initcall_debug", initcall_debug_setup
);
623 struct task_struct
*child_reaper
= &init_task
;
625 extern initcall_t __initcall_start
[], __initcall_end
[];
627 static void __init
do_initcalls(void)
630 int count
= preempt_count();
632 for (call
= __initcall_start
; call
< __initcall_end
; call
++) {
637 if (initcall_debug
) {
638 printk("Calling initcall 0x%p", *call
);
639 print_fn_descriptor_symbol(": %s()",
640 (unsigned long) *call
);
646 if (result
&& result
!= -ENODEV
&& initcall_debug
) {
647 sprintf(msgbuf
, "error code %d", result
);
650 if (preempt_count() != count
) {
651 msg
= "preemption imbalance";
652 preempt_count() = count
;
654 if (irqs_disabled()) {
655 msg
= "disabled interrupts";
659 printk(KERN_WARNING
"initcall at 0x%p", *call
);
660 print_fn_descriptor_symbol(": %s()",
661 (unsigned long) *call
);
662 printk(": returned with %s\n", msg
);
666 /* Make sure there is no pending stuff from the initcall sequence */
667 flush_scheduled_work();
671 * Ok, the machine is now initialized. None of the devices
672 * have been touched yet, but the CPU subsystem is up and
673 * running, and memory and process management works.
675 * Now we can finally start doing some real work..
677 static void __init
do_basic_setup(void)
679 /* drivers will send hotplug events */
681 usermodehelper_init();
691 static void do_pre_smp_initcalls(void)
693 extern int spawn_ksoftirqd(void);
695 extern int migration_init(void);
700 spawn_softlockup_task();
703 static void run_init_process(char *init_filename
)
705 argv_init
[0] = init_filename
;
706 kernel_execve(init_filename
, argv_init
, envp_init
);
709 static int init(void * unused
)
713 * init can run on any cpu.
715 set_cpus_allowed(current
, CPU_MASK_ALL
);
717 * Tell the world that we're going to be the grim
718 * reaper of innocent orphaned children.
720 * We don't want people to have to make incorrect
721 * assumptions about where in the task array this
724 child_reaper
= current
;
726 cad_pid
= task_pid(current
);
728 smp_prepare_cpus(max_cpus
);
730 do_pre_smp_initcalls();
738 * Do this before initcalls, because some drivers want to access
746 * check if there is an early userspace init. If yes, let it do all
750 if (!ramdisk_execute_command
)
751 ramdisk_execute_command
= "/init";
753 if (sys_access((const char __user
*) ramdisk_execute_command
, 0) != 0) {
754 ramdisk_execute_command
= NULL
;
759 * Ok, we have completed the initial bootup, and
760 * we're essentially up and running. Get rid of the
761 * initmem segments and start the user-mode stuff..
766 system_state
= SYSTEM_RUNNING
;
767 numa_default_policy();
769 if (sys_open((const char __user
*) "/dev/console", O_RDWR
, 0) < 0)
770 printk(KERN_WARNING
"Warning: unable to open an initial console.\n");
775 if (ramdisk_execute_command
) {
776 run_init_process(ramdisk_execute_command
);
777 printk(KERN_WARNING
"Failed to execute %s\n",
778 ramdisk_execute_command
);
782 * We try each of these until one succeeds.
784 * The Bourne shell can be used instead of init if we are
785 * trying to recover a really broken machine.
787 if (execute_command
) {
788 run_init_process(execute_command
);
789 printk(KERN_WARNING
"Failed to execute %s. Attempting "
790 "defaults...\n", execute_command
);
792 run_init_process("/sbin/init");
793 run_init_process("/etc/init");
794 run_init_process("/bin/init");
795 run_init_process("/bin/sh");
797 panic("No init found. Try passing init= option to kernel.");