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>
58 * This is one of the first .c files built. Error out early
59 * if we have compiler trouble..
62 #ifdef CONFIG_X86_LOCAL_APIC
67 * Versions of gcc older than that listed below may actually compile
68 * and link okay, but the end product can have subtle run time bugs.
69 * To avoid associated bogus bug reports, we flatly refuse to compile
70 * with a gcc that is known to be 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 buffer_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 * Setup routine for controlling SMP activation
132 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
133 * activation entirely (the MPS table probe still happens, though).
135 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
136 * greater than 0, limits the maximum number of CPUs activated in
139 static int __init
nosmp(char *str
)
145 __setup("nosmp", nosmp
);
147 static int __init
maxcpus(char *str
)
149 get_option(&str
, &max_cpus
);
153 __setup("maxcpus=", maxcpus
);
155 static char * argv_init
[MAX_INIT_ARGS
+2] = { "init", NULL
, };
156 char * envp_init
[MAX_INIT_ENVS
+2] = { "HOME=/", "TERM=linux", NULL
, };
157 static const char *panic_later
, *panic_param
;
159 extern struct obs_kernel_param __setup_start
[], __setup_end
[];
161 static int __init
obsolete_checksetup(char *line
)
163 struct obs_kernel_param
*p
;
167 int n
= strlen(p
->str
);
168 if (!strncmp(line
, p
->str
, n
)) {
170 /* Already done in parse_early_param? (Needs
171 * exact match on param part) */
172 if (line
[n
] == '\0' || line
[n
] == '=')
174 } else if (!p
->setup_func
) {
175 printk(KERN_WARNING
"Parameter %s is obsolete,"
176 " ignored\n", p
->str
);
178 } else if (p
->setup_func(line
+ n
))
182 } while (p
< __setup_end
);
187 * This should be approx 2 Bo*oMips to start (note initial shift), and will
188 * still work even if initially too large, it will just take slightly longer
190 unsigned long loops_per_jiffy
= (1<<12);
192 EXPORT_SYMBOL(loops_per_jiffy
);
194 static int __init
debug_kernel(char *str
)
198 console_loglevel
= 10;
202 static int __init
quiet_kernel(char *str
)
206 console_loglevel
= 4;
210 __setup("debug", debug_kernel
);
211 __setup("quiet", quiet_kernel
);
213 static int __init
loglevel(char *str
)
215 get_option(&str
, &console_loglevel
);
219 __setup("loglevel=", loglevel
);
222 * Unknown boot options get handed to init, unless they look like
225 static int __init
unknown_bootoption(char *param
, char *val
)
227 /* Change NUL term back to "=", to make "param" the whole string. */
229 /* param=val or param="val"? */
230 if (val
== param
+strlen(param
)+1)
232 else if (val
== param
+strlen(param
)+2) {
234 memmove(val
-1, val
, strlen(val
)+1);
240 /* Handle obsolete-style parameters */
241 if (obsolete_checksetup(param
))
245 * Preemptive maintenance for "why didn't my mispelled command
248 if (strchr(param
, '.') && (!val
|| strchr(param
, '.') < val
)) {
249 printk(KERN_ERR
"Unknown boot option `%s': ignoring\n", param
);
257 /* Environment option */
259 for (i
= 0; envp_init
[i
]; i
++) {
260 if (i
== MAX_INIT_ENVS
) {
261 panic_later
= "Too many boot env vars at `%s'";
264 if (!strncmp(param
, envp_init
[i
], val
- param
))
267 envp_init
[i
] = param
;
269 /* Command line option */
271 for (i
= 0; argv_init
[i
]; i
++) {
272 if (i
== MAX_INIT_ARGS
) {
273 panic_later
= "Too many boot init vars at `%s'";
277 argv_init
[i
] = param
;
282 static int __init
init_setup(char *str
)
286 execute_command
= str
;
288 * In case LILO is going to boot us with default command line,
289 * it prepends "auto" before the whole cmdline which makes
290 * the shell think it should execute a script with such name.
291 * So we ignore all arguments entered _before_ init=... [MJ]
293 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
297 __setup("init=", init_setup
);
299 static int __init
rdinit_setup(char *str
)
303 ramdisk_execute_command
= str
;
304 /* See "auto" comment in init_setup */
305 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
309 __setup("rdinit=", rdinit_setup
);
311 extern void setup_arch(char **);
315 #ifdef CONFIG_X86_LOCAL_APIC
316 static void __init
smp_init(void)
318 APIC_init_uniprocessor();
321 #define smp_init() do { } while (0)
324 static inline void setup_per_cpu_areas(void) { }
325 static inline void smp_prepare_cpus(unsigned int maxcpus
) { }
329 #ifdef __GENERIC_PER_CPU
330 unsigned long __per_cpu_offset
[NR_CPUS
];
332 EXPORT_SYMBOL(__per_cpu_offset
);
334 static void __init
setup_per_cpu_areas(void)
336 unsigned long size
, i
;
339 /* Copy section for each CPU (we discard the original) */
340 size
= ALIGN(__per_cpu_end
- __per_cpu_start
, SMP_CACHE_BYTES
);
341 #ifdef CONFIG_MODULES
342 if (size
< PERCPU_ENOUGH_ROOM
)
343 size
= PERCPU_ENOUGH_ROOM
;
346 ptr
= alloc_bootmem(size
* NR_CPUS
);
348 for (i
= 0; i
< NR_CPUS
; i
++, ptr
+= size
) {
349 __per_cpu_offset
[i
] = ptr
- __per_cpu_start
;
350 memcpy(ptr
, __per_cpu_start
, __per_cpu_end
- __per_cpu_start
);
353 #endif /* !__GENERIC_PER_CPU */
355 /* Called by boot processor to activate the rest. */
356 static void __init
smp_init(void)
360 /* FIXME: This should be done in userspace --RR */
361 for_each_present_cpu(i
) {
362 if (num_online_cpus() >= max_cpus
)
368 /* Any cleanup work */
369 printk(KERN_INFO
"Brought up %ld CPUs\n", (long)num_online_cpus());
370 smp_cpus_done(max_cpus
);
372 /* Get other processors into their bootup holding patterns. */
381 * We need to finalize in a non-__init function or else race conditions
382 * between the root thread and the init thread may cause start_kernel to
383 * be reaped by free_initmem before the root thread has proceeded to
386 * gcc-3.4 accidentally inlines this function, so use noinline.
389 static void noinline
rest_init(void)
390 __releases(kernel_lock
)
392 kernel_thread(init
, NULL
, CLONE_FS
| CLONE_SIGHAND
);
393 numa_default_policy();
397 * The boot idle thread must execute schedule()
398 * at least one to get things moving:
400 preempt_enable_no_resched();
404 /* Call into cpu_idle with preempt disabled */
408 /* Check for early params. */
409 static int __init
do_early_param(char *param
, char *val
)
411 struct obs_kernel_param
*p
;
413 for (p
= __setup_start
; p
< __setup_end
; p
++) {
414 if (p
->early
&& strcmp(param
, p
->str
) == 0) {
415 if (p
->setup_func(val
) != 0)
417 "Malformed early option '%s'\n", param
);
420 /* We accept everything at this stage. */
424 /* Arch code calls this early on, or if not, just before other parsing. */
425 void __init
parse_early_param(void)
427 static __initdata
int done
= 0;
428 static __initdata
char tmp_cmdline
[COMMAND_LINE_SIZE
];
433 /* All fall through to do_early_param. */
434 strlcpy(tmp_cmdline
, saved_command_line
, COMMAND_LINE_SIZE
);
435 parse_args("early options", tmp_cmdline
, NULL
, 0, do_early_param
);
440 * Activate the first processor.
443 asmlinkage
void __init
start_kernel(void)
446 extern struct kernel_param __start___param
[], __stop___param
[];
448 * Interrupts are still disabled. Do necessary setups, then
454 printk(linux_banner
);
455 setup_arch(&command_line
);
456 setup_per_cpu_areas();
459 * Mark the boot cpu "online" so that it can call console drivers in
460 * printk() and can access its per-cpu storage.
462 smp_prepare_boot_cpu();
465 * Set up the scheduler prior starting any interrupts (such as the
466 * timer interrupt). Full topology setup happens at smp_init()
467 * time - but meanwhile we still have a functioning scheduler.
471 * Disable preemption - early bootup scheduling is extremely
472 * fragile until we cpu_idle() for the first time.
475 build_all_zonelists();
477 printk(KERN_NOTICE
"Kernel command line: %s\n", saved_command_line
);
479 parse_args("Booting kernel", command_line
, __start___param
,
480 __stop___param
- __start___param
,
481 &unknown_bootoption
);
492 * HACK ALERT! This is early. We're enabling the console before
493 * we've done PCI setups etc, and console_init() must be aware of
494 * this. But we do want output early, in case something goes wrong.
498 panic(panic_later
, panic_param
);
501 #ifdef CONFIG_BLK_DEV_INITRD
502 if (initrd_start
&& !initrd_below_start_ok
&&
503 initrd_start
< min_low_pfn
<< PAGE_SHIFT
) {
504 printk(KERN_CRIT
"initrd overwritten (0x%08lx < 0x%08lx) - "
505 "disabling it.\n",initrd_start
,min_low_pfn
<< PAGE_SHIFT
);
509 vfs_caches_init_early();
513 setup_per_cpu_pageset();
519 pgtable_cache_init();
524 efi_enter_virtual_mode();
526 fork_init(num_physpages
);
532 vfs_caches_init(num_physpages
);
535 /* rootfs populating might need page-writeback */
536 page_writeback_init();
537 #ifdef CONFIG_PROC_FS
544 acpi_early_init(); /* before LAPIC and SMP init */
546 /* Do the rest non-__init'ed, we're now alive */
550 static int __initdata initcall_debug
;
552 static int __init
initcall_debug_setup(char *str
)
557 __setup("initcall_debug", initcall_debug_setup
);
559 struct task_struct
*child_reaper
= &init_task
;
561 extern initcall_t __initcall_start
[], __initcall_end
[];
563 static void __init
do_initcalls(void)
566 int count
= preempt_count();
568 for (call
= __initcall_start
; call
< __initcall_end
; call
++) {
571 if (initcall_debug
) {
572 printk(KERN_DEBUG
"Calling initcall 0x%p", *call
);
573 print_fn_descriptor_symbol(": %s()", (unsigned long) *call
);
580 if (preempt_count() != count
) {
581 msg
= "preemption imbalance";
582 preempt_count() = count
;
584 if (irqs_disabled()) {
585 msg
= "disabled interrupts";
589 printk(KERN_WARNING
"error in initcall at 0x%p: "
590 "returned with %s\n", *call
, msg
);
594 /* Make sure there is no pending stuff from the initcall sequence */
595 flush_scheduled_work();
599 * Ok, the machine is now initialized. None of the devices
600 * have been touched yet, but the CPU subsystem is up and
601 * running, and memory and process management works.
603 * Now we can finally start doing some real work..
605 static void __init
do_basic_setup(void)
607 /* drivers will send hotplug events */
609 usermodehelper_init();
619 static void do_pre_smp_initcalls(void)
621 extern int spawn_ksoftirqd(void);
623 extern int migration_init(void);
628 spawn_softlockup_task();
631 static void run_init_process(char *init_filename
)
633 argv_init
[0] = init_filename
;
634 execve(init_filename
, argv_init
, envp_init
);
637 static inline void fixup_cpu_present_map(void)
643 * If arch is not hotplug ready and did not populate
644 * cpu_present_map, just make cpu_present_map same as cpu_possible_map
645 * for other cpu bringup code to function as normal. e.g smp_init() etc.
647 if (cpus_empty(cpu_present_map
)) {
649 cpu_set(i
, cpu_present_map
);
655 static int init(void * unused
)
659 * init can run on any cpu.
661 set_cpus_allowed(current
, CPU_MASK_ALL
);
663 * Tell the world that we're going to be the grim
664 * reaper of innocent orphaned children.
666 * We don't want people to have to make incorrect
667 * assumptions about where in the task array this
670 child_reaper
= current
;
672 /* Sets up cpus_possible() */
673 smp_prepare_cpus(max_cpus
);
675 do_pre_smp_initcalls();
677 fixup_cpu_present_map();
684 * Do this before initcalls, because some drivers want to access
692 * check if there is an early userspace init. If yes, let it do all
696 if (!ramdisk_execute_command
)
697 ramdisk_execute_command
= "/init";
699 if (sys_access((const char __user
*) ramdisk_execute_command
, 0) != 0) {
700 ramdisk_execute_command
= NULL
;
705 * Ok, we have completed the initial bootup, and
706 * we're essentially up and running. Get rid of the
707 * initmem segments and start the user-mode stuff..
712 system_state
= SYSTEM_RUNNING
;
713 numa_default_policy();
715 if (sys_open((const char __user
*) "/dev/console", O_RDWR
, 0) < 0)
716 printk(KERN_WARNING
"Warning: unable to open an initial console.\n");
721 if (ramdisk_execute_command
) {
722 run_init_process(ramdisk_execute_command
);
723 printk(KERN_WARNING
"Failed to execute %s\n",
724 ramdisk_execute_command
);
728 * We try each of these until one succeeds.
730 * The Bourne shell can be used instead of init if we are
731 * trying to recover a really broken machine.
733 if (execute_command
) {
734 run_init_process(execute_command
);
735 printk(KERN_WARNING
"Failed to execute %s. Attempting "
736 "defaults...\n", execute_command
);
738 run_init_process("/sbin/init");
739 run_init_process("/etc/init");
740 run_init_process("/bin/init");
741 run_init_process("/bin/sh");
743 panic("No init found. Try passing init= option to kernel.");