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..
61 #if __GNUC__ == 2 && __GNUC_MINOR__ == 96
62 #ifdef CONFIG_FRAME_POINTER
63 #error This compiler cannot compile correctly with frame pointers enabled
67 #ifdef CONFIG_X86_LOCAL_APIC
72 * Versions of gcc older than that listed below may actually compile
73 * and link okay, but the end product can have subtle run time bugs.
74 * To avoid associated bogus bug reports, we flatly refuse to compile
75 * with a gcc that is known to be too old from the very beginning.
77 #if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 95)
78 #error Sorry, your GCC is too old. It builds incorrect kernels.
81 static int init(void *);
83 extern void init_IRQ(void);
84 extern void fork_init(unsigned long);
85 extern void mca_init(void);
86 extern void sbus_init(void);
87 extern void sysctl_init(void);
88 extern void signals_init(void);
89 extern void buffer_init(void);
90 extern void pidhash_init(void);
91 extern void pidmap_init(void);
92 extern void prio_tree_init(void);
93 extern void radix_tree_init(void);
94 extern void free_initmem(void);
95 extern void populate_rootfs(void);
96 extern void driver_init(void);
97 extern void prepare_namespace(void);
99 extern void acpi_early_init(void);
101 static inline void acpi_early_init(void) { }
103 #ifndef CONFIG_DEBUG_RODATA
104 static inline void mark_rodata_ro(void) { }
108 extern void tc_init(void);
111 enum system_states system_state
;
112 EXPORT_SYMBOL(system_state
);
115 * Boot command-line arguments
117 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
118 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
120 extern void time_init(void);
121 /* Default late time init is NULL. archs can override this later. */
122 void (*late_time_init
)(void);
123 extern void softirq_init(void);
125 /* Untouched command line (eg. for /proc) saved by arch-specific code. */
126 char saved_command_line
[COMMAND_LINE_SIZE
];
128 static char *execute_command
;
129 static char *ramdisk_execute_command
;
131 /* Setup configured maximum number of CPUs to activate */
132 static unsigned int max_cpus
= NR_CPUS
;
135 * Setup routine for controlling SMP activation
137 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
138 * activation entirely (the MPS table probe still happens, though).
140 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
141 * greater than 0, limits the maximum number of CPUs activated in
144 static int __init
nosmp(char *str
)
150 __setup("nosmp", nosmp
);
152 static int __init
maxcpus(char *str
)
154 get_option(&str
, &max_cpus
);
158 __setup("maxcpus=", maxcpus
);
160 static char * argv_init
[MAX_INIT_ARGS
+2] = { "init", NULL
, };
161 char * envp_init
[MAX_INIT_ENVS
+2] = { "HOME=/", "TERM=linux", NULL
, };
162 static const char *panic_later
, *panic_param
;
164 extern struct obs_kernel_param __setup_start
[], __setup_end
[];
166 static int __init
obsolete_checksetup(char *line
)
168 struct obs_kernel_param
*p
;
172 int n
= strlen(p
->str
);
173 if (!strncmp(line
, p
->str
, n
)) {
175 /* Already done in parse_early_param? (Needs
176 * exact match on param part) */
177 if (line
[n
] == '\0' || line
[n
] == '=')
179 } else if (!p
->setup_func
) {
180 printk(KERN_WARNING
"Parameter %s is obsolete,"
181 " ignored\n", p
->str
);
183 } else if (p
->setup_func(line
+ n
))
187 } while (p
< __setup_end
);
192 * This should be approx 2 Bo*oMips to start (note initial shift), and will
193 * still work even if initially too large, it will just take slightly longer
195 unsigned long loops_per_jiffy
= (1<<12);
197 EXPORT_SYMBOL(loops_per_jiffy
);
199 static int __init
debug_kernel(char *str
)
203 console_loglevel
= 10;
207 static int __init
quiet_kernel(char *str
)
211 console_loglevel
= 4;
215 __setup("debug", debug_kernel
);
216 __setup("quiet", quiet_kernel
);
218 static int __init
loglevel(char *str
)
220 get_option(&str
, &console_loglevel
);
224 __setup("loglevel=", loglevel
);
227 * Unknown boot options get handed to init, unless they look like
230 static int __init
unknown_bootoption(char *param
, char *val
)
232 /* Change NUL term back to "=", to make "param" the whole string. */
234 /* param=val or param="val"? */
235 if (val
== param
+strlen(param
)+1)
237 else if (val
== param
+strlen(param
)+2) {
239 memmove(val
-1, val
, strlen(val
)+1);
245 /* Handle obsolete-style parameters */
246 if (obsolete_checksetup(param
))
250 * Preemptive maintenance for "why didn't my mispelled command
253 if (strchr(param
, '.') && (!val
|| strchr(param
, '.') < val
)) {
254 printk(KERN_ERR
"Unknown boot option `%s': ignoring\n", param
);
262 /* Environment option */
264 for (i
= 0; envp_init
[i
]; i
++) {
265 if (i
== MAX_INIT_ENVS
) {
266 panic_later
= "Too many boot env vars at `%s'";
269 if (!strncmp(param
, envp_init
[i
], val
- param
))
272 envp_init
[i
] = param
;
274 /* Command line option */
276 for (i
= 0; argv_init
[i
]; i
++) {
277 if (i
== MAX_INIT_ARGS
) {
278 panic_later
= "Too many boot init vars at `%s'";
282 argv_init
[i
] = param
;
287 static int __init
init_setup(char *str
)
291 execute_command
= str
;
293 * In case LILO is going to boot us with default command line,
294 * it prepends "auto" before the whole cmdline which makes
295 * the shell think it should execute a script with such name.
296 * So we ignore all arguments entered _before_ init=... [MJ]
298 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
302 __setup("init=", init_setup
);
304 static int __init
rdinit_setup(char *str
)
308 ramdisk_execute_command
= str
;
309 /* See "auto" comment in init_setup */
310 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
314 __setup("rdinit=", rdinit_setup
);
316 extern void setup_arch(char **);
320 #ifdef CONFIG_X86_LOCAL_APIC
321 static void __init
smp_init(void)
323 APIC_init_uniprocessor();
326 #define smp_init() do { } while (0)
329 static inline void setup_per_cpu_areas(void) { }
330 static inline void smp_prepare_cpus(unsigned int maxcpus
) { }
334 #ifdef __GENERIC_PER_CPU
335 unsigned long __per_cpu_offset
[NR_CPUS
];
337 EXPORT_SYMBOL(__per_cpu_offset
);
339 static void __init
setup_per_cpu_areas(void)
341 unsigned long size
, i
;
344 /* Copy section for each CPU (we discard the original) */
345 size
= ALIGN(__per_cpu_end
- __per_cpu_start
, SMP_CACHE_BYTES
);
346 #ifdef CONFIG_MODULES
347 if (size
< PERCPU_ENOUGH_ROOM
)
348 size
= PERCPU_ENOUGH_ROOM
;
351 ptr
= alloc_bootmem(size
* NR_CPUS
);
353 for (i
= 0; i
< NR_CPUS
; i
++, ptr
+= size
) {
354 __per_cpu_offset
[i
] = ptr
- __per_cpu_start
;
355 memcpy(ptr
, __per_cpu_start
, __per_cpu_end
- __per_cpu_start
);
358 #endif /* !__GENERIC_PER_CPU */
360 /* Called by boot processor to activate the rest. */
361 static void __init
smp_init(void)
365 /* FIXME: This should be done in userspace --RR */
366 for_each_present_cpu(i
) {
367 if (num_online_cpus() >= max_cpus
)
373 /* Any cleanup work */
374 printk(KERN_INFO
"Brought up %ld CPUs\n", (long)num_online_cpus());
375 smp_cpus_done(max_cpus
);
377 /* Get other processors into their bootup holding patterns. */
386 * We need to finalize in a non-__init function or else race conditions
387 * between the root thread and the init thread may cause start_kernel to
388 * be reaped by free_initmem before the root thread has proceeded to
391 * gcc-3.4 accidentally inlines this function, so use noinline.
394 static void noinline
rest_init(void)
395 __releases(kernel_lock
)
397 kernel_thread(init
, NULL
, CLONE_FS
| CLONE_SIGHAND
);
398 numa_default_policy();
402 * The boot idle thread must execute schedule()
403 * at least one to get things moving:
405 preempt_enable_no_resched();
409 /* Call into cpu_idle with preempt disabled */
413 /* Check for early params. */
414 static int __init
do_early_param(char *param
, char *val
)
416 struct obs_kernel_param
*p
;
418 for (p
= __setup_start
; p
< __setup_end
; p
++) {
419 if (p
->early
&& strcmp(param
, p
->str
) == 0) {
420 if (p
->setup_func(val
) != 0)
422 "Malformed early option '%s'\n", param
);
425 /* We accept everything at this stage. */
429 /* Arch code calls this early on, or if not, just before other parsing. */
430 void __init
parse_early_param(void)
432 static __initdata
int done
= 0;
433 static __initdata
char tmp_cmdline
[COMMAND_LINE_SIZE
];
438 /* All fall through to do_early_param. */
439 strlcpy(tmp_cmdline
, saved_command_line
, COMMAND_LINE_SIZE
);
440 parse_args("early options", tmp_cmdline
, NULL
, 0, do_early_param
);
445 * Activate the first processor.
448 asmlinkage
void __init
start_kernel(void)
451 extern struct kernel_param __start___param
[], __stop___param
[];
453 * Interrupts are still disabled. Do necessary setups, then
459 printk(linux_banner
);
460 setup_arch(&command_line
);
461 setup_per_cpu_areas();
464 * Mark the boot cpu "online" so that it can call console drivers in
465 * printk() and can access its per-cpu storage.
467 smp_prepare_boot_cpu();
470 * Set up the scheduler prior starting any interrupts (such as the
471 * timer interrupt). Full topology setup happens at smp_init()
472 * time - but meanwhile we still have a functioning scheduler.
476 * Disable preemption - early bootup scheduling is extremely
477 * fragile until we cpu_idle() for the first time.
480 build_all_zonelists();
482 printk(KERN_NOTICE
"Kernel command line: %s\n", saved_command_line
);
484 parse_args("Booting kernel", command_line
, __start___param
,
485 __stop___param
- __start___param
,
486 &unknown_bootoption
);
497 * HACK ALERT! This is early. We're enabling the console before
498 * we've done PCI setups etc, and console_init() must be aware of
499 * this. But we do want output early, in case something goes wrong.
503 panic(panic_later
, panic_param
);
506 #ifdef CONFIG_BLK_DEV_INITRD
507 if (initrd_start
&& !initrd_below_start_ok
&&
508 initrd_start
< min_low_pfn
<< PAGE_SHIFT
) {
509 printk(KERN_CRIT
"initrd overwritten (0x%08lx < 0x%08lx) - "
510 "disabling it.\n",initrd_start
,min_low_pfn
<< PAGE_SHIFT
);
514 vfs_caches_init_early();
518 setup_per_cpu_pageset();
524 pgtable_cache_init();
529 efi_enter_virtual_mode();
531 fork_init(num_physpages
);
537 vfs_caches_init(num_physpages
);
540 /* rootfs populating might need page-writeback */
541 page_writeback_init();
542 #ifdef CONFIG_PROC_FS
549 acpi_early_init(); /* before LAPIC and SMP init */
551 /* Do the rest non-__init'ed, we're now alive */
555 static int __initdata initcall_debug
;
557 static int __init
initcall_debug_setup(char *str
)
562 __setup("initcall_debug", initcall_debug_setup
);
564 struct task_struct
*child_reaper
= &init_task
;
566 extern initcall_t __initcall_start
[], __initcall_end
[];
568 static void __init
do_initcalls(void)
571 int count
= preempt_count();
573 for (call
= __initcall_start
; call
< __initcall_end
; call
++) {
576 if (initcall_debug
) {
577 printk(KERN_DEBUG
"Calling initcall 0x%p", *call
);
578 print_fn_descriptor_symbol(": %s()", (unsigned long) *call
);
585 if (preempt_count() != count
) {
586 msg
= "preemption imbalance";
587 preempt_count() = count
;
589 if (irqs_disabled()) {
590 msg
= "disabled interrupts";
594 printk(KERN_WARNING
"error in initcall at 0x%p: "
595 "returned with %s\n", *call
, msg
);
599 /* Make sure there is no pending stuff from the initcall sequence */
600 flush_scheduled_work();
604 * Ok, the machine is now initialized. None of the devices
605 * have been touched yet, but the CPU subsystem is up and
606 * running, and memory and process management works.
608 * Now we can finally start doing some real work..
610 static void __init
do_basic_setup(void)
612 /* drivers will send hotplug events */
614 usermodehelper_init();
624 static void do_pre_smp_initcalls(void)
626 extern int spawn_ksoftirqd(void);
628 extern int migration_init(void);
633 spawn_softlockup_task();
636 static void run_init_process(char *init_filename
)
638 argv_init
[0] = init_filename
;
639 execve(init_filename
, argv_init
, envp_init
);
642 static inline void fixup_cpu_present_map(void)
648 * If arch is not hotplug ready and did not populate
649 * cpu_present_map, just make cpu_present_map same as cpu_possible_map
650 * for other cpu bringup code to function as normal. e.g smp_init() etc.
652 if (cpus_empty(cpu_present_map
)) {
654 cpu_set(i
, cpu_present_map
);
660 static int init(void * unused
)
664 * init can run on any cpu.
666 set_cpus_allowed(current
, CPU_MASK_ALL
);
668 * Tell the world that we're going to be the grim
669 * reaper of innocent orphaned children.
671 * We don't want people to have to make incorrect
672 * assumptions about where in the task array this
675 child_reaper
= current
;
677 /* Sets up cpus_possible() */
678 smp_prepare_cpus(max_cpus
);
680 do_pre_smp_initcalls();
682 fixup_cpu_present_map();
689 * Do this before initcalls, because some drivers want to access
697 * check if there is an early userspace init. If yes, let it do all
701 if (!ramdisk_execute_command
)
702 ramdisk_execute_command
= "/init";
704 if (sys_access((const char __user
*) ramdisk_execute_command
, 0) != 0) {
705 ramdisk_execute_command
= NULL
;
710 * Ok, we have completed the initial bootup, and
711 * we're essentially up and running. Get rid of the
712 * initmem segments and start the user-mode stuff..
717 system_state
= SYSTEM_RUNNING
;
718 numa_default_policy();
720 if (sys_open((const char __user
*) "/dev/console", O_RDWR
, 0) < 0)
721 printk(KERN_WARNING
"Warning: unable to open an initial console.\n");
726 if (ramdisk_execute_command
) {
727 run_init_process(ramdisk_execute_command
);
728 printk(KERN_WARNING
"Failed to execute %s\n",
729 ramdisk_execute_command
);
733 * We try each of these until one succeeds.
735 * The Bourne shell can be used instead of init if we are
736 * trying to recover a really broken machine.
738 if (execute_command
) {
739 run_init_process(execute_command
);
740 printk(KERN_WARNING
"Failed to execute %s. Attempting "
741 "defaults...\n", execute_command
);
743 run_init_process("/sbin/init");
744 run_init_process("/etc/init");
745 run_init_process("/bin/init");
746 run_init_process("/bin/sh");
748 panic("No init found. Try passing init= option to kernel.");