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/efi.h>
45 #include <linux/unistd.h>
46 #include <linux/rmap.h>
47 #include <linux/mempolicy.h>
48 #include <linux/key.h>
52 #include <asm/setup.h>
55 * This is one of the first .c files built. Error out early
56 * if we have compiler trouble..
58 #if __GNUC__ == 2 && __GNUC_MINOR__ == 96
59 #ifdef CONFIG_FRAME_POINTER
60 #error This compiler cannot compile correctly with frame pointers enabled
64 #ifdef CONFIG_X86_LOCAL_APIC
69 * Versions of gcc older than that listed below may actually compile
70 * and link okay, but the end product can have subtle run time bugs.
71 * To avoid associated bogus bug reports, we flatly refuse to compile
72 * with a gcc that is known to be too old from the very beginning.
74 #if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 95)
75 #error Sorry, your GCC is too old. It builds incorrect kernels.
78 extern char *linux_banner
;
80 static int init(void *);
82 extern void init_IRQ(void);
83 extern void sock_init(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) { }
105 extern void tc_init(void);
108 enum system_states system_state
;
109 EXPORT_SYMBOL(system_state
);
112 * Boot command-line arguments
114 #define MAX_INIT_ARGS 32
115 #define MAX_INIT_ENVS 32
117 extern void time_init(void);
118 /* Default late time init is NULL. archs can override this later. */
119 void (*late_time_init
)(void);
120 extern void softirq_init(void);
122 /* Untouched command line (eg. for /proc) saved by arch-specific code. */
123 char saved_command_line
[COMMAND_LINE_SIZE
];
125 static char *execute_command
;
127 /* Setup configured maximum number of CPUs to activate */
128 static unsigned int max_cpus
= NR_CPUS
;
131 * Setup routine for controlling SMP activation
133 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
134 * activation entirely (the MPS table probe still happens, though).
136 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
137 * greater than 0, limits the maximum number of CPUs activated in
140 static int __init
nosmp(char *str
)
146 __setup("nosmp", nosmp
);
148 static int __init
maxcpus(char *str
)
150 get_option(&str
, &max_cpus
);
154 __setup("maxcpus=", maxcpus
);
156 static char * argv_init
[MAX_INIT_ARGS
+2] = { "init", NULL
, };
157 char * envp_init
[MAX_INIT_ENVS
+2] = { "HOME=/", "TERM=linux", NULL
, };
158 static const char *panic_later
, *panic_param
;
160 static int __init
obsolete_checksetup(char *line
)
162 struct obs_kernel_param
*p
;
163 extern struct obs_kernel_param __setup_start
, __setup_end
;
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
);
186 static unsigned long preset_lpj
;
187 static int __init
lpj_setup(char *str
)
189 preset_lpj
= simple_strtoul(str
,NULL
,0);
193 __setup("lpj=", lpj_setup
);
196 * This should be approx 2 Bo*oMips to start (note initial shift), and will
197 * still work even if initially too large, it will just take slightly longer
199 unsigned long loops_per_jiffy
= (1<<12);
201 EXPORT_SYMBOL(loops_per_jiffy
);
204 * This is the number of bits of precision for the loops_per_jiffy. Each
205 * bit takes on average 1.5/HZ seconds. This (like the original) is a little
210 void __devinit
calibrate_delay(void)
212 unsigned long ticks
, loopbit
;
213 int lps_precision
= LPS_PREC
;
216 loops_per_jiffy
= preset_lpj
;
217 printk("Calibrating delay loop (skipped)... "
218 "%lu.%02lu BogoMIPS preset\n",
219 loops_per_jiffy
/(500000/HZ
),
220 (loops_per_jiffy
/(5000/HZ
)) % 100);
222 loops_per_jiffy
= (1<<12);
224 printk(KERN_DEBUG
"Calibrating delay loop... ");
225 while ((loops_per_jiffy
<<= 1) != 0) {
226 /* wait for "start of" clock tick */
228 while (ticks
== jiffies
)
232 __delay(loops_per_jiffy
);
233 ticks
= jiffies
- ticks
;
239 * Do a binary approximation to get loops_per_jiffy set to
240 * equal one clock (up to lps_precision bits)
242 loops_per_jiffy
>>= 1;
243 loopbit
= loops_per_jiffy
;
244 while (lps_precision
-- && (loopbit
>>= 1)) {
245 loops_per_jiffy
|= loopbit
;
247 while (ticks
== jiffies
)
250 __delay(loops_per_jiffy
);
251 if (jiffies
!= ticks
) /* longer than 1 tick */
252 loops_per_jiffy
&= ~loopbit
;
255 /* Round the value and print it */
256 printk("%lu.%02lu BogoMIPS (lpj=%lu)\n",
257 loops_per_jiffy
/(500000/HZ
),
258 (loops_per_jiffy
/(5000/HZ
)) % 100,
264 static int __init
debug_kernel(char *str
)
268 console_loglevel
= 10;
272 static int __init
quiet_kernel(char *str
)
276 console_loglevel
= 4;
280 __setup("debug", debug_kernel
);
281 __setup("quiet", quiet_kernel
);
284 * Unknown boot options get handed to init, unless they look like
287 static int __init
unknown_bootoption(char *param
, char *val
)
289 /* Change NUL term back to "=", to make "param" the whole string. */
291 /* param=val or param="val"? */
292 if (val
== param
+strlen(param
)+1)
294 else if (val
== param
+strlen(param
)+2) {
296 memmove(val
-1, val
, strlen(val
)+1);
302 /* Handle obsolete-style parameters */
303 if (obsolete_checksetup(param
))
307 * Preemptive maintenance for "why didn't my mispelled command
310 if (strchr(param
, '.') && (!val
|| strchr(param
, '.') < val
)) {
311 printk(KERN_ERR
"Unknown boot option `%s': ignoring\n", param
);
319 /* Environment option */
321 for (i
= 0; envp_init
[i
]; i
++) {
322 if (i
== MAX_INIT_ENVS
) {
323 panic_later
= "Too many boot env vars at `%s'";
326 if (!strncmp(param
, envp_init
[i
], val
- param
))
329 envp_init
[i
] = param
;
331 /* Command line option */
333 for (i
= 0; argv_init
[i
]; i
++) {
334 if (i
== MAX_INIT_ARGS
) {
335 panic_later
= "Too many boot init vars at `%s'";
339 argv_init
[i
] = param
;
344 static int __init
init_setup(char *str
)
348 execute_command
= str
;
350 * In case LILO is going to boot us with default command line,
351 * it prepends "auto" before the whole cmdline which makes
352 * the shell think it should execute a script with such name.
353 * So we ignore all arguments entered _before_ init=... [MJ]
355 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
359 __setup("init=", init_setup
);
361 extern void setup_arch(char **);
362 extern void cpu_idle(void);
366 #ifdef CONFIG_X86_LOCAL_APIC
367 static void __init
smp_init(void)
369 APIC_init_uniprocessor();
372 #define smp_init() do { } while (0)
375 static inline void setup_per_cpu_areas(void) { }
376 static inline void smp_prepare_cpus(unsigned int maxcpus
) { }
380 #ifdef __GENERIC_PER_CPU
381 unsigned long __per_cpu_offset
[NR_CPUS
];
383 EXPORT_SYMBOL(__per_cpu_offset
);
385 static void __init
setup_per_cpu_areas(void)
387 unsigned long size
, i
;
389 /* Created by linker magic */
390 extern char __per_cpu_start
[], __per_cpu_end
[];
392 /* Copy section for each CPU (we discard the original) */
393 size
= ALIGN(__per_cpu_end
- __per_cpu_start
, SMP_CACHE_BYTES
);
394 #ifdef CONFIG_MODULES
395 if (size
< PERCPU_ENOUGH_ROOM
)
396 size
= PERCPU_ENOUGH_ROOM
;
399 ptr
= alloc_bootmem(size
* NR_CPUS
);
401 for (i
= 0; i
< NR_CPUS
; i
++, ptr
+= size
) {
402 __per_cpu_offset
[i
] = ptr
- __per_cpu_start
;
403 memcpy(ptr
, __per_cpu_start
, __per_cpu_end
- __per_cpu_start
);
406 #endif /* !__GENERIC_PER_CPU */
408 /* Called by boot processor to activate the rest. */
409 static void __init
smp_init(void)
413 /* FIXME: This should be done in userspace --RR */
414 for_each_present_cpu(i
) {
415 if (num_online_cpus() >= max_cpus
)
421 /* Any cleanup work */
422 printk("Brought up %ld CPUs\n", (long)num_online_cpus());
423 smp_cpus_done(max_cpus
);
425 /* Get other processors into their bootup holding patterns. */
435 * We need to finalize in a non-__init function or else race conditions
436 * between the root thread and the init thread may cause start_kernel to
437 * be reaped by free_initmem before the root thread has proceeded to
440 * gcc-3.4 accidentally inlines this function, so use noinline.
443 static void noinline
rest_init(void)
444 __releases(kernel_lock
)
446 kernel_thread(init
, NULL
, CLONE_FS
| CLONE_SIGHAND
);
447 numa_default_policy();
452 /* Check for early params. */
453 static int __init
do_early_param(char *param
, char *val
)
455 struct obs_kernel_param
*p
;
456 extern struct obs_kernel_param __setup_start
, __setup_end
;
458 for (p
= &__setup_start
; p
< &__setup_end
; p
++) {
459 if (p
->early
&& strcmp(param
, p
->str
) == 0) {
460 if (p
->setup_func(val
) != 0)
462 "Malformed early option '%s'\n", param
);
465 /* We accept everything at this stage. */
469 /* Arch code calls this early on, or if not, just before other parsing. */
470 void __init
parse_early_param(void)
472 static __initdata
int done
= 0;
473 static __initdata
char tmp_cmdline
[COMMAND_LINE_SIZE
];
478 /* All fall through to do_early_param. */
479 strlcpy(tmp_cmdline
, saved_command_line
, COMMAND_LINE_SIZE
);
480 parse_args("early options", tmp_cmdline
, NULL
, 0, do_early_param
);
485 * Activate the first processor.
488 asmlinkage
void __init
start_kernel(void)
491 extern struct kernel_param __start___param
[], __stop___param
[];
493 * Interrupts are still disabled. Do necessary setups, then
498 printk(linux_banner
);
499 setup_arch(&command_line
);
500 setup_per_cpu_areas();
503 * Mark the boot cpu "online" so that it can call console drivers in
504 * printk() and can access its per-cpu storage.
506 smp_prepare_boot_cpu();
509 * Set up the scheduler prior starting any interrupts (such as the
510 * timer interrupt). Full topology setup happens at smp_init()
511 * time - but meanwhile we still have a functioning scheduler.
514 build_all_zonelists();
516 printk("Kernel command line: %s\n", saved_command_line
);
518 parse_args("Booting kernel", command_line
, __start___param
,
519 __stop___param
- __start___param
,
520 &unknown_bootoption
);
531 * HACK ALERT! This is early. We're enabling the console before
532 * we've done PCI setups etc, and console_init() must be aware of
533 * this. But we do want output early, in case something goes wrong.
537 panic(panic_later
, panic_param
);
540 #ifdef CONFIG_BLK_DEV_INITRD
541 if (initrd_start
&& !initrd_below_start_ok
&&
542 initrd_start
< min_low_pfn
<< PAGE_SHIFT
) {
543 printk(KERN_CRIT
"initrd overwritten (0x%08lx < 0x%08lx) - "
544 "disabling it.\n",initrd_start
,min_low_pfn
<< PAGE_SHIFT
);
548 vfs_caches_init_early();
556 pgtable_cache_init();
561 efi_enter_virtual_mode();
563 fork_init(num_physpages
);
568 vfs_caches_init(num_physpages
);
571 /* rootfs populating might need page-writeback */
572 page_writeback_init();
573 #ifdef CONFIG_PROC_FS
578 acpi_early_init(); /* before LAPIC and SMP init */
580 /* Do the rest non-__init'ed, we're now alive */
584 static int __initdata initcall_debug
;
586 static int __init
initcall_debug_setup(char *str
)
591 __setup("initcall_debug", initcall_debug_setup
);
593 struct task_struct
*child_reaper
= &init_task
;
595 extern initcall_t __initcall_start
, __initcall_end
;
597 static void __init
do_initcalls(void)
600 int count
= preempt_count();
602 for (call
= &__initcall_start
; call
< &__initcall_end
; call
++) {
605 if (initcall_debug
) {
606 printk(KERN_DEBUG
"Calling initcall 0x%p", *call
);
607 print_fn_descriptor_symbol(": %s()", (unsigned long) *call
);
614 if (preempt_count() != count
) {
615 msg
= "preemption imbalance";
616 preempt_count() = count
;
618 if (irqs_disabled()) {
619 msg
= "disabled interrupts";
623 printk("error in initcall at 0x%p: "
624 "returned with %s\n", *call
, msg
);
628 /* Make sure there is no pending stuff from the initcall sequence */
629 flush_scheduled_work();
633 * Ok, the machine is now initialized. None of the devices
634 * have been touched yet, but the CPU subsystem is up and
635 * running, and memory and process management works.
637 * Now we can finally start doing some real work..
639 static void __init
do_basic_setup(void)
641 /* drivers will send hotplug events */
643 usermodehelper_init();
651 /* Networking initialization needs a process context */
657 static void do_pre_smp_initcalls(void)
659 extern int spawn_ksoftirqd(void);
661 extern int migration_init(void);
668 static void run_init_process(char *init_filename
)
670 argv_init
[0] = init_filename
;
671 execve(init_filename
, argv_init
, envp_init
);
674 static inline void fixup_cpu_present_map(void)
680 * If arch is not hotplug ready and did not populate
681 * cpu_present_map, just make cpu_present_map same as cpu_possible_map
682 * for other cpu bringup code to function as normal. e.g smp_init() etc.
684 if (cpus_empty(cpu_present_map
)) {
686 cpu_set(i
, cpu_present_map
);
692 static int init(void * unused
)
696 * Tell the world that we're going to be the grim
697 * reaper of innocent orphaned children.
699 * We don't want people to have to make incorrect
700 * assumptions about where in the task array this
703 child_reaper
= current
;
705 /* Sets up cpus_possible() */
706 smp_prepare_cpus(max_cpus
);
708 do_pre_smp_initcalls();
710 fixup_cpu_present_map();
715 * Do this before initcalls, because some drivers want to access
723 * check if there is an early userspace init. If yes, let it do all
726 if (sys_access((const char __user
*) "/init", 0) == 0)
727 execute_command
= "/init";
732 * Ok, we have completed the initial bootup, and
733 * we're essentially up and running. Get rid of the
734 * initmem segments and start the user-mode stuff..
738 system_state
= SYSTEM_RUNNING
;
739 numa_default_policy();
741 if (sys_open((const char __user
*) "/dev/console", O_RDWR
, 0) < 0)
742 printk("Warning: unable to open an initial console.\n");
748 * We try each of these until one succeeds.
750 * The Bourne shell can be used instead of init if we are
751 * trying to recover a really broken machine.
755 run_init_process(execute_command
);
757 run_init_process("/sbin/init");
758 run_init_process("/etc/init");
759 run_init_process("/bin/init");
760 run_init_process("/bin/sh");
762 panic("No init found. Try passing init= option to kernel.");
765 static int early_param_test(char *rest
)
767 printk("early_parm_test: %s\n", rest
?: "(null)");
768 return rest
? 0 : -EINVAL
;
770 early_param("testsetup", early_param_test
);
771 static int early_setup_test(char *rest
)
773 printk("early_setup_test: %s\n", rest
?: "(null)");
776 __setup("testsetup_long", early_setup_test
);