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>
53 #include <linux/pid_namespace.h>
54 #include <linux/device.h>
58 #include <asm/setup.h>
59 #include <asm/sections.h>
60 #include <asm/cacheflush.h>
62 #ifdef CONFIG_X86_LOCAL_APIC
67 * This is one of the first .c files built. Error out early if we have compiler
70 * Versions of gcc older than that listed below may actually compile and link
71 * okay, but the end product can have subtle run time bugs. To avoid associated
72 * bogus bug reports, we flatly refuse to compile with a gcc that is known to be
73 * too old from the very beginning.
75 #if (__GNUC__ < 3) || (__GNUC__ == 3 && __GNUC_MINOR__ < 2)
76 #error Sorry, your GCC is too old. It builds incorrect kernels.
79 #if __GNUC__ == 4 && __GNUC_MINOR__ == 1 && __GNUC_PATCHLEVEL__ == 0
80 #warning gcc-4.1.0 is known to miscompile the kernel. A different compiler version is recommended.
83 static int init(void *);
85 extern void init_IRQ(void);
86 extern void fork_init(unsigned long);
87 extern void mca_init(void);
88 extern void sbus_init(void);
89 extern void signals_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 prepare_namespace(void);
97 extern void acpi_early_init(void);
99 static inline void acpi_early_init(void) { }
101 #ifndef CONFIG_DEBUG_RODATA
102 static inline void mark_rodata_ro(void) { }
106 extern void tc_init(void);
109 enum system_states system_state;
110 EXPORT_SYMBOL(system_state);
113 * Boot command-line arguments
115 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
116 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
118 extern void time_init(void);
119 /* Default late time init is NULL. archs can override this later. */
120 void (*late_time_init)(void);
121 extern void softirq_init(void);
123 /* Untouched command line saved by arch-specific code. */
124 char __initdata boot_command_line[COMMAND_LINE_SIZE];
125 /* Untouched saved command line (eg. for /proc) */
126 char *saved_command_line;
127 /* Command line for parameter parsing */
128 static char *static_command_line;
130 static char *execute_command;
131 static char *ramdisk_execute_command;
133 /* Setup configured maximum number of CPUs to activate */
134 static unsigned int max_cpus = NR_CPUS;
137 * If set, this is an indication to the drivers that reset the underlying
138 * device before going ahead with the initialization otherwise driver might
139 * rely on the BIOS and skip the reset operation.
141 * This is useful if kernel is booting in an unreliable environment.
142 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
143 * skipped and devices will be in unknown state.
145 unsigned int reset_devices;
146 EXPORT_SYMBOL(reset_devices);
149 * Setup routine for controlling SMP activation
151 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
152 * activation entirely (the MPS table probe still happens, though).
154 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
155 * greater than 0, limits the maximum number of CPUs activated in
158 static int __init nosmp(char *str)
164 __setup("nosmp", nosmp);
166 static int __init maxcpus(char *str)
168 get_option(&str, &max_cpus);
172 __setup("maxcpus=", maxcpus);
174 static int __init set_reset_devices(char *str)
180 __setup("reset_devices", set_reset_devices);
182 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
183 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
184 static const char *panic_later, *panic_param;
186 extern struct obs_kernel_param __setup_start[], __setup_end[];
188 static int __init obsolete_checksetup(char *line)
190 struct obs_kernel_param *p;
191 int had_early_param = 0;
195 int n = strlen(p->str);
196 if (!strncmp(line, p->str, n)) {
198 /* Already done in parse_early_param?
199 * (Needs exact match on param part).
200 * Keep iterating, as we can have early
201 * params and __setups of same names 8( */
202 if (line[n] == '\0' || line[n] == '=')
204 } else if (!p->setup_func) {
205 printk(KERN_WARNING "Parameter %s is obsolete,"
206 " ignored\n", p->str);
208 } else if (p->setup_func(line + n))
212 } while (p < __setup_end);
214 return had_early_param;
218 * This should be approx 2 Bo*oMips to start (note initial shift), and will
219 * still work even if initially too large, it will just take slightly longer
221 unsigned long loops_per_jiffy = (1<<12);
223 EXPORT_SYMBOL(loops_per_jiffy);
225 static int __init debug_kernel(char *str)
229 console_loglevel = 10;
233 static int __init quiet_kernel(char *str)
237 console_loglevel = 4;
241 __setup("debug", debug_kernel);
242 __setup("quiet", quiet_kernel);
244 static int __init loglevel(char *str)
246 get_option(&str, &console_loglevel);
250 __setup("loglevel=", loglevel);
253 * Unknown boot options get handed to init, unless they look like
256 static int __init unknown_bootoption(char *param, char *val)
258 /* Change NUL term back to "=", to make "param" the whole string. */
260 /* param=val or param="val"? */
261 if (val == param+strlen(param)+1)
263 else if (val == param+strlen(param)+2) {
265 memmove(val-1, val, strlen(val)+1);
271 /* Handle obsolete-style parameters */
272 if (obsolete_checksetup(param))
276 * Preemptive maintenance for "why didn't my mispelled command
279 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
280 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
288 /* Environment option */
290 for (i = 0; envp_init[i]; i++) {
291 if (i == MAX_INIT_ENVS) {
292 panic_later = "Too many boot env vars at `%s'";
295 if (!strncmp(param, envp_init[i], val - param))
298 envp_init[i] = param;
300 /* Command line option */
302 for (i = 0; argv_init[i]; i++) {
303 if (i == MAX_INIT_ARGS) {
304 panic_later = "Too many boot init vars at `%s'";
308 argv_init[i] = param;
313 static int __init init_setup(char *str)
317 execute_command = str;
319 * In case LILO is going to boot us with default command line,
320 * it prepends "auto" before the whole cmdline which makes
321 * the shell think it should execute a script with such name.
322 * So we ignore all arguments entered _before_ init=... [MJ]
324 for (i = 1; i < MAX_INIT_ARGS; i++)
328 __setup("init=", init_setup);
330 static int __init rdinit_setup(char *str)
334 ramdisk_execute_command = str;
335 /* See "auto" comment in init_setup */
336 for (i = 1; i < MAX_INIT_ARGS; i++)
340 __setup("rdinit=", rdinit_setup);
344 #ifdef CONFIG_X86_LOCAL_APIC
345 static void __init smp_init(void)
347 APIC_init_uniprocessor();
350 #define smp_init() do { } while (0)
353 static inline void setup_per_cpu_areas(void) { }
354 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
358 #ifdef __GENERIC_PER_CPU
359 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
361 EXPORT_SYMBOL(__per_cpu_offset);
363 static void __init setup_per_cpu_areas(void)
365 unsigned long size, i;
367 unsigned long nr_possible_cpus = num_possible_cpus();
369 /* Copy section for each CPU (we discard the original) */
370 size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
371 #ifdef CONFIG_MODULES
372 if (size < PERCPU_ENOUGH_ROOM)
373 size = PERCPU_ENOUGH_ROOM;
375 ptr = alloc_bootmem(size * nr_possible_cpus);
377 for_each_possible_cpu(i) {
378 __per_cpu_offset[i] = ptr - __per_cpu_start;
379 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
383 #endif /* !__GENERIC_PER_CPU */
385 /* Called by boot processor to activate the rest. */
386 static void __init smp_init(void)
390 /* FIXME: This should be done in userspace --RR */
391 for_each_present_cpu(i) {
392 if (num_online_cpus() >= max_cpus)
398 /* Any cleanup work */
399 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
400 smp_cpus_done(max_cpus);
406 * We need to store the untouched command line for future reference.
407 * We also need to store the touched command line since the parameter
408 * parsing is performed in place, and we should allow a component to
409 * store reference of name/value for future reference.
411 static void __init setup_command_line(char *command_line)
413 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
414 static_command_line = alloc_bootmem(strlen (command_line)+1);
415 strcpy (saved_command_line, boot_command_line);
416 strcpy (static_command_line, command_line);
420 * We need to finalize in a non-__init function or else race conditions
421 * between the root thread and the init thread may cause start_kernel to
422 * be reaped by free_initmem before the root thread has proceeded to
425 * gcc-3.4 accidentally inlines this function, so use noinline.
428 static void noinline rest_init(void)
429 __releases(kernel_lock)
431 kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
432 numa_default_policy();
436 * The boot idle thread must execute schedule()
437 * at least one to get things moving:
439 preempt_enable_no_resched();
443 /* Call into cpu_idle with preempt disabled */
447 /* Check for early params. */
448 static int __init do_early_param(char *param, char *val)
450 struct obs_kernel_param *p;
452 for (p = __setup_start; p < __setup_end; p++) {
453 if (p->early && strcmp(param, p->str) == 0) {
454 if (p->setup_func(val) != 0)
456 "Malformed early option '%s'\n", param);
459 /* We accept everything at this stage. */
463 /* Arch code calls this early on, or if not, just before other parsing. */
464 void __init parse_early_param(void)
466 static __initdata int done = 0;
467 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
472 /* All fall through to do_early_param. */
473 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
474 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
479 * Activate the first processor.
482 static void __init boot_cpu_init(void)
484 int cpu = smp_processor_id();
485 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
486 cpu_set(cpu, cpu_online_map);
487 cpu_set(cpu, cpu_present_map);
488 cpu_set(cpu, cpu_possible_map);
491 void __init __attribute__((weak)) smp_setup_processor_id(void)
495 asmlinkage void __init start_kernel(void)
498 extern struct kernel_param __start___param[], __stop___param[];
500 smp_setup_processor_id();
503 * Need to run as early as possible, to initialize the
510 early_boot_irqs_off();
511 early_init_irq_lock_class();
514 * Interrupts are still disabled. Do necessary setups, then
521 printk(linux_banner);
522 setup_arch(&command_line);
523 setup_command_line(command_line);
525 setup_per_cpu_areas();
526 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
529 * Set up the scheduler prior starting any interrupts (such as the
530 * timer interrupt). Full topology setup happens at smp_init()
531 * time - but meanwhile we still have a functioning scheduler.
535 * Disable preemption - early bootup scheduling is extremely
536 * fragile until we cpu_idle() for the first time.
539 build_all_zonelists();
541 printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
543 parse_args("Booting kernel", static_command_line, __start___param,
544 __stop___param - __start___param,
545 &unknown_bootoption);
546 if (!irqs_disabled()) {
547 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
548 "enabled *very* early, fixing it\n");
562 if (!irqs_disabled())
563 printk("start_kernel(): bug: interrupts were enabled early\n");
564 early_boot_irqs_on();
568 * HACK ALERT! This is early. We're enabling the console before
569 * we've done PCI setups etc, and console_init() must be aware of
570 * this. But we do want output early, in case something goes wrong.
574 panic(panic_later, panic_param);
579 * Need to run this when irqs are enabled, because it wants
580 * to self-test [hard/soft]-irqs on/off lock inversion bugs
585 #ifdef CONFIG_BLK_DEV_INITRD
586 if (initrd_start && !initrd_below_start_ok &&
587 initrd_start < min_low_pfn << PAGE_SHIFT) {
588 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
589 "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
593 vfs_caches_init_early();
597 setup_per_cpu_pageset();
603 pgtable_cache_init();
608 efi_enter_virtual_mode();
610 fork_init(num_physpages);
616 vfs_caches_init(num_physpages);
619 /* rootfs populating might need page-writeback */
620 page_writeback_init();
621 #ifdef CONFIG_PROC_FS
625 taskstats_init_early();
630 acpi_early_init(); /* before LAPIC and SMP init */
632 /* Do the rest non-__init'ed, we're now alive */
636 static int __initdata initcall_debug;
638 static int __init initcall_debug_setup(char *str)
643 __setup("initcall_debug", initcall_debug_setup);
645 extern initcall_t __initcall_start[], __initcall_end[];
647 static void __init do_initcalls(void)
650 int count = preempt_count();
652 for (call = __initcall_start; call < __initcall_end; call++) {
657 if (initcall_debug) {
658 printk("Calling initcall 0x%p", *call);
659 print_fn_descriptor_symbol(": %s()",
660 (unsigned long) *call);
666 if (result && result != -ENODEV && initcall_debug) {
667 sprintf(msgbuf, "error code %d", result);
670 if (preempt_count() != count) {
671 msg = "preemption imbalance";
672 preempt_count() = count;
674 if (irqs_disabled()) {
675 msg = "disabled interrupts";
679 printk(KERN_WARNING "initcall at 0x%p", *call);
680 print_fn_descriptor_symbol(": %s()",
681 (unsigned long) *call);
682 printk(": returned with %s\n", msg);
686 /* Make sure there is no pending stuff from the initcall sequence */
687 flush_scheduled_work();
691 * Ok, the machine is now initialized. None of the devices
692 * have been touched yet, but the CPU subsystem is up and
693 * running, and memory and process management works.
695 * Now we can finally start doing some real work..
697 static void __init do_basic_setup(void)
699 /* drivers will send hotplug events */
701 usermodehelper_init();
704 #ifdef CONFIG_PROC_FS
711 static void __init do_pre_smp_initcalls(void)
713 extern int spawn_ksoftirqd(void);
715 extern int migration_init(void);
720 spawn_softlockup_task();
723 static void run_init_process(char *init_filename)
725 argv_init[0] = init_filename;
726 kernel_execve(init_filename, argv_init, envp_init);
729 /* This is a non __init function. Force it to be noinline otherwise gcc
730 * makes it inline to init() and it becomes part of init.text section
732 static int noinline init_post(void)
737 system_state = SYSTEM_RUNNING;
738 numa_default_policy();
740 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
741 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
746 if (ramdisk_execute_command) {
747 run_init_process(ramdisk_execute_command);
748 printk(KERN_WARNING "Failed to execute %s\n",
749 ramdisk_execute_command);
753 * We try each of these until one succeeds.
755 * The Bourne shell can be used instead of init if we are
756 * trying to recover a really broken machine.
758 if (execute_command) {
759 run_init_process(execute_command);
760 printk(KERN_WARNING "Failed to execute %s. Attempting "
761 "defaults...\n", execute_command);
763 run_init_process("/sbin/init");
764 run_init_process("/etc/init");
765 run_init_process("/bin/init");
766 run_init_process("/bin/sh");
768 panic("No init found. Try passing init= option to kernel.");
771 static int __init init(void * unused)
775 * init can run on any cpu.
777 set_cpus_allowed(current, CPU_MASK_ALL);
779 * Tell the world that we're going to be the grim
780 * reaper of innocent orphaned children.
782 * We don't want people to have to make incorrect
783 * assumptions about where in the task array this
786 init_pid_ns.child_reaper = current;
788 cad_pid = task_pid(current);
790 smp_prepare_cpus(max_cpus);
792 do_pre_smp_initcalls();
802 * check if there is an early userspace init. If yes, let it do all
806 if (!ramdisk_execute_command)
807 ramdisk_execute_command = "/init";
809 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
810 ramdisk_execute_command = NULL;
815 * Ok, we have completed the initial bootup, and
816 * we're essentially up and running. Get rid of the
817 * initmem segments and start the user-mode stuff..