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 #define __KERNEL_SYSCALLS__
14 #include <linux/config.h>
15 #include <linux/types.h>
16 #include <linux/module.h>
17 #include <linux/proc_fs.h>
18 #include <linux/devfs_fs_kernel.h>
19 #include <linux/kernel.h>
20 #include <linux/syscalls.h>
21 #include <linux/string.h>
22 #include <linux/ctype.h>
23 #include <linux/delay.h>
24 #include <linux/utsname.h>
25 #include <linux/ioport.h>
26 #include <linux/init.h>
27 #include <linux/smp_lock.h>
28 #include <linux/initrd.h>
29 #include <linux/hdreg.h>
30 #include <linux/bootmem.h>
31 #include <linux/tty.h>
32 #include <linux/gfp.h>
33 #include <linux/percpu.h>
34 #include <linux/kmod.h>
35 #include <linux/kernel_stat.h>
36 #include <linux/security.h>
37 #include <linux/workqueue.h>
38 #include <linux/profile.h>
39 #include <linux/rcupdate.h>
40 #include <linux/moduleparam.h>
41 #include <linux/kallsyms.h>
42 #include <linux/writeback.h>
43 #include <linux/cpu.h>
44 #include <linux/cpuset.h>
45 #include <linux/efi.h>
46 #include <linux/unistd.h>
47 #include <linux/rmap.h>
48 #include <linux/mempolicy.h>
49 #include <linux/key.h>
53 #include <asm/setup.h>
54 #include <asm/sections.h>
55 #include <asm/cacheflush.h>
57 #ifdef CONFIG_X86_LOCAL_APIC
62 * This is one of the first .c files built. Error out early if we have compiler
65 * Versions of gcc older than that listed below may actually compile and link
66 * okay, but the end product can have subtle run time bugs. To avoid associated
67 * bogus bug reports, we flatly refuse to compile with a gcc that is known to be
68 * too old from the very beginning.
70 #if (__GNUC__ < 3) || (__GNUC__ == 3 && __GNUC_MINOR__ < 2)
71 #error Sorry, your GCC is too old. It builds incorrect kernels.
74 static int init(void *);
76 extern void init_IRQ(void);
77 extern void fork_init(unsigned long);
78 extern void mca_init(void);
79 extern void sbus_init(void);
80 extern void sysctl_init(void);
81 extern void signals_init(void);
82 extern void buffer_init(void);
83 extern void pidhash_init(void);
84 extern void pidmap_init(void);
85 extern void prio_tree_init(void);
86 extern void radix_tree_init(void);
87 extern void free_initmem(void);
88 extern void populate_rootfs(void);
89 extern void driver_init(void);
90 extern void prepare_namespace(void);
92 extern void acpi_early_init(void);
94 static inline void acpi_early_init(void) { }
96 #ifndef CONFIG_DEBUG_RODATA
97 static inline void mark_rodata_ro(void) { }
101 extern void tc_init(void);
104 enum system_states system_state
;
105 EXPORT_SYMBOL(system_state
);
108 * Boot command-line arguments
110 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
111 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
113 extern void time_init(void);
114 /* Default late time init is NULL. archs can override this later. */
115 void (*late_time_init
)(void);
116 extern void softirq_init(void);
118 /* Untouched command line (eg. for /proc) saved by arch-specific code. */
119 char saved_command_line
[COMMAND_LINE_SIZE
];
121 static char *execute_command
;
122 static char *ramdisk_execute_command
;
124 /* Setup configured maximum number of CPUs to activate */
125 static unsigned int max_cpus
= NR_CPUS
;
128 * Setup routine for controlling SMP activation
130 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
131 * activation entirely (the MPS table probe still happens, though).
133 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
134 * greater than 0, limits the maximum number of CPUs activated in
137 static int __init
nosmp(char *str
)
143 __setup("nosmp", nosmp
);
145 static int __init
maxcpus(char *str
)
147 get_option(&str
, &max_cpus
);
151 __setup("maxcpus=", maxcpus
);
153 static char * argv_init
[MAX_INIT_ARGS
+2] = { "init", NULL
, };
154 char * envp_init
[MAX_INIT_ENVS
+2] = { "HOME=/", "TERM=linux", NULL
, };
155 static const char *panic_later
, *panic_param
;
157 extern struct obs_kernel_param __setup_start
[], __setup_end
[];
159 static int __init
obsolete_checksetup(char *line
)
161 struct obs_kernel_param
*p
;
165 int n
= strlen(p
->str
);
166 if (!strncmp(line
, p
->str
, n
)) {
168 /* Already done in parse_early_param? (Needs
169 * exact match on param part) */
170 if (line
[n
] == '\0' || line
[n
] == '=')
172 } else if (!p
->setup_func
) {
173 printk(KERN_WARNING
"Parameter %s is obsolete,"
174 " ignored\n", p
->str
);
176 } else if (p
->setup_func(line
+ n
))
180 } while (p
< __setup_end
);
185 * This should be approx 2 Bo*oMips to start (note initial shift), and will
186 * still work even if initially too large, it will just take slightly longer
188 unsigned long loops_per_jiffy
= (1<<12);
190 EXPORT_SYMBOL(loops_per_jiffy
);
192 static int __init
debug_kernel(char *str
)
196 console_loglevel
= 10;
200 static int __init
quiet_kernel(char *str
)
204 console_loglevel
= 4;
208 __setup("debug", debug_kernel
);
209 __setup("quiet", quiet_kernel
);
211 static int __init
loglevel(char *str
)
213 get_option(&str
, &console_loglevel
);
217 __setup("loglevel=", loglevel
);
220 * Unknown boot options get handed to init, unless they look like
223 static int __init
unknown_bootoption(char *param
, char *val
)
225 /* Change NUL term back to "=", to make "param" the whole string. */
227 /* param=val or param="val"? */
228 if (val
== param
+strlen(param
)+1)
230 else if (val
== param
+strlen(param
)+2) {
232 memmove(val
-1, val
, strlen(val
)+1);
238 /* Handle obsolete-style parameters */
239 if (obsolete_checksetup(param
))
243 * Preemptive maintenance for "why didn't my mispelled command
246 if (strchr(param
, '.') && (!val
|| strchr(param
, '.') < val
)) {
247 printk(KERN_ERR
"Unknown boot option `%s': ignoring\n", param
);
255 /* Environment option */
257 for (i
= 0; envp_init
[i
]; i
++) {
258 if (i
== MAX_INIT_ENVS
) {
259 panic_later
= "Too many boot env vars at `%s'";
262 if (!strncmp(param
, envp_init
[i
], val
- param
))
265 envp_init
[i
] = param
;
267 /* Command line option */
269 for (i
= 0; argv_init
[i
]; i
++) {
270 if (i
== MAX_INIT_ARGS
) {
271 panic_later
= "Too many boot init vars at `%s'";
275 argv_init
[i
] = param
;
280 static int __init
init_setup(char *str
)
284 execute_command
= str
;
286 * In case LILO is going to boot us with default command line,
287 * it prepends "auto" before the whole cmdline which makes
288 * the shell think it should execute a script with such name.
289 * So we ignore all arguments entered _before_ init=... [MJ]
291 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
295 __setup("init=", init_setup
);
297 static int __init
rdinit_setup(char *str
)
301 ramdisk_execute_command
= str
;
302 /* See "auto" comment in init_setup */
303 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
307 __setup("rdinit=", rdinit_setup
);
311 #ifdef CONFIG_X86_LOCAL_APIC
312 static void __init
smp_init(void)
314 APIC_init_uniprocessor();
317 #define smp_init() do { } while (0)
320 static inline void setup_per_cpu_areas(void) { }
321 static inline void smp_prepare_cpus(unsigned int maxcpus
) { }
325 #ifdef __GENERIC_PER_CPU
326 unsigned long __per_cpu_offset
[NR_CPUS
] __read_mostly
;
328 EXPORT_SYMBOL(__per_cpu_offset
);
330 static void __init
setup_per_cpu_areas(void)
332 unsigned long size
, i
;
334 unsigned long nr_possible_cpus
= num_possible_cpus();
336 /* Copy section for each CPU (we discard the original) */
337 size
= ALIGN(__per_cpu_end
- __per_cpu_start
, SMP_CACHE_BYTES
);
338 #ifdef CONFIG_MODULES
339 if (size
< PERCPU_ENOUGH_ROOM
)
340 size
= PERCPU_ENOUGH_ROOM
;
342 ptr
= alloc_bootmem(size
* nr_possible_cpus
);
345 __per_cpu_offset
[i
] = ptr
- __per_cpu_start
;
346 memcpy(ptr
, __per_cpu_start
, __per_cpu_end
- __per_cpu_start
);
350 #endif /* !__GENERIC_PER_CPU */
352 /* Called by boot processor to activate the rest. */
353 static void __init
smp_init(void)
357 /* FIXME: This should be done in userspace --RR */
358 for_each_present_cpu(i
) {
359 if (num_online_cpus() >= max_cpus
)
365 /* Any cleanup work */
366 printk(KERN_INFO
"Brought up %ld CPUs\n", (long)num_online_cpus());
367 smp_cpus_done(max_cpus
);
369 /* Get other processors into their bootup holding patterns. */
378 * We need to finalize in a non-__init function or else race conditions
379 * between the root thread and the init thread may cause start_kernel to
380 * be reaped by free_initmem before the root thread has proceeded to
383 * gcc-3.4 accidentally inlines this function, so use noinline.
386 static void noinline
rest_init(void)
387 __releases(kernel_lock
)
389 kernel_thread(init
, NULL
, CLONE_FS
| CLONE_SIGHAND
);
390 numa_default_policy();
394 * The boot idle thread must execute schedule()
395 * at least one to get things moving:
397 preempt_enable_no_resched();
401 /* Call into cpu_idle with preempt disabled */
405 /* Check for early params. */
406 static int __init
do_early_param(char *param
, char *val
)
408 struct obs_kernel_param
*p
;
410 for (p
= __setup_start
; p
< __setup_end
; p
++) {
411 if (p
->early
&& strcmp(param
, p
->str
) == 0) {
412 if (p
->setup_func(val
) != 0)
414 "Malformed early option '%s'\n", param
);
417 /* We accept everything at this stage. */
421 /* Arch code calls this early on, or if not, just before other parsing. */
422 void __init
parse_early_param(void)
424 static __initdata
int done
= 0;
425 static __initdata
char tmp_cmdline
[COMMAND_LINE_SIZE
];
430 /* All fall through to do_early_param. */
431 strlcpy(tmp_cmdline
, saved_command_line
, COMMAND_LINE_SIZE
);
432 parse_args("early options", tmp_cmdline
, NULL
, 0, do_early_param
);
437 * Activate the first processor.
440 static void __init
boot_cpu_init(void)
442 int cpu
= smp_processor_id();
443 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
444 cpu_set(cpu
, cpu_online_map
);
445 cpu_set(cpu
, cpu_present_map
);
446 cpu_set(cpu
, cpu_possible_map
);
449 asmlinkage
void __init
start_kernel(void)
452 extern struct kernel_param __start___param
[], __stop___param
[];
454 * Interrupts are still disabled. Do necessary setups, then
461 printk(linux_banner
);
462 setup_arch(&command_line
);
463 setup_per_cpu_areas();
464 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
467 * Set up the scheduler prior starting any interrupts (such as the
468 * timer interrupt). Full topology setup happens at smp_init()
469 * time - but meanwhile we still have a functioning scheduler.
473 * Disable preemption - early bootup scheduling is extremely
474 * fragile until we cpu_idle() for the first time.
477 build_all_zonelists();
479 printk(KERN_NOTICE
"Kernel command line: %s\n", saved_command_line
);
481 parse_args("Booting kernel", command_line
, __start___param
,
482 __stop___param
- __start___param
,
483 &unknown_bootoption
);
495 * HACK ALERT! This is early. We're enabling the console before
496 * we've done PCI setups etc, and console_init() must be aware of
497 * this. But we do want output early, in case something goes wrong.
501 panic(panic_later
, panic_param
);
504 #ifdef CONFIG_BLK_DEV_INITRD
505 if (initrd_start
&& !initrd_below_start_ok
&&
506 initrd_start
< min_low_pfn
<< PAGE_SHIFT
) {
507 printk(KERN_CRIT
"initrd overwritten (0x%08lx < 0x%08lx) - "
508 "disabling it.\n",initrd_start
,min_low_pfn
<< PAGE_SHIFT
);
512 vfs_caches_init_early();
516 setup_per_cpu_pageset();
522 pgtable_cache_init();
527 efi_enter_virtual_mode();
529 fork_init(num_physpages
);
535 vfs_caches_init(num_physpages
);
538 /* rootfs populating might need page-writeback */
539 page_writeback_init();
540 #ifdef CONFIG_PROC_FS
547 acpi_early_init(); /* before LAPIC and SMP init */
549 /* Do the rest non-__init'ed, we're now alive */
553 static int __initdata initcall_debug
;
555 static int __init
initcall_debug_setup(char *str
)
560 __setup("initcall_debug", initcall_debug_setup
);
562 struct task_struct
*child_reaper
= &init_task
;
564 extern initcall_t __initcall_start
[], __initcall_end
[];
566 static void __init
do_initcalls(void)
569 int count
= preempt_count();
571 for (call
= __initcall_start
; call
< __initcall_end
; call
++) {
576 if (initcall_debug
) {
577 printk("Calling initcall 0x%p", *call
);
578 print_fn_descriptor_symbol(": %s()",
579 (unsigned long) *call
);
585 if (result
&& (result
!= -ENODEV
|| initcall_debug
)) {
586 sprintf(msgbuf
, "error code %d", result
);
589 if (preempt_count() != count
) {
590 msg
= "preemption imbalance";
591 preempt_count() = count
;
593 if (irqs_disabled()) {
594 msg
= "disabled interrupts";
598 printk(KERN_WARNING
"initcall at 0x%p", *call
);
599 print_fn_descriptor_symbol(": %s()",
600 (unsigned long) *call
);
601 printk(": returned with %s\n", msg
);
605 /* Make sure there is no pending stuff from the initcall sequence */
606 flush_scheduled_work();
610 * Ok, the machine is now initialized. None of the devices
611 * have been touched yet, but the CPU subsystem is up and
612 * running, and memory and process management works.
614 * Now we can finally start doing some real work..
616 static void __init
do_basic_setup(void)
618 /* drivers will send hotplug events */
620 usermodehelper_init();
630 static void do_pre_smp_initcalls(void)
632 extern int spawn_ksoftirqd(void);
634 extern int migration_init(void);
639 spawn_softlockup_task();
642 static void run_init_process(char *init_filename
)
644 argv_init
[0] = init_filename
;
645 execve(init_filename
, argv_init
, envp_init
);
648 static inline void fixup_cpu_present_map(void)
654 * If arch is not hotplug ready and did not populate
655 * cpu_present_map, just make cpu_present_map same as cpu_possible_map
656 * for other cpu bringup code to function as normal. e.g smp_init() etc.
658 if (cpus_empty(cpu_present_map
)) {
660 cpu_set(i
, cpu_present_map
);
666 static int init(void * unused
)
670 * init can run on any cpu.
672 set_cpus_allowed(current
, CPU_MASK_ALL
);
674 * Tell the world that we're going to be the grim
675 * reaper of innocent orphaned children.
677 * We don't want people to have to make incorrect
678 * assumptions about where in the task array this
681 child_reaper
= current
;
683 smp_prepare_cpus(max_cpus
);
685 do_pre_smp_initcalls();
687 fixup_cpu_present_map();
694 * Do this before initcalls, because some drivers want to access
702 * check if there is an early userspace init. If yes, let it do all
706 if (!ramdisk_execute_command
)
707 ramdisk_execute_command
= "/init";
709 if (sys_access((const char __user
*) ramdisk_execute_command
, 0) != 0) {
710 ramdisk_execute_command
= NULL
;
715 * Ok, we have completed the initial bootup, and
716 * we're essentially up and running. Get rid of the
717 * initmem segments and start the user-mode stuff..
722 system_state
= SYSTEM_RUNNING
;
723 numa_default_policy();
725 if (sys_open((const char __user
*) "/dev/console", O_RDWR
, 0) < 0)
726 printk(KERN_WARNING
"Warning: unable to open an initial console.\n");
731 if (ramdisk_execute_command
) {
732 run_init_process(ramdisk_execute_command
);
733 printk(KERN_WARNING
"Failed to execute %s\n",
734 ramdisk_execute_command
);
738 * We try each of these until one succeeds.
740 * The Bourne shell can be used instead of init if we are
741 * trying to recover a really broken machine.
743 if (execute_command
) {
744 run_init_process(execute_command
);
745 printk(KERN_WARNING
"Failed to execute %s. Attempting "
746 "defaults...\n", execute_command
);
748 run_init_process("/sbin/init");
749 run_init_process("/etc/init");
750 run_init_process("/bin/init");
751 run_init_process("/bin/sh");
753 panic("No init found. Try passing init= option to kernel.");