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>
54 #include <linux/ckrm.h>
55 #ifdef CONFIG_CKRM_CPU_SCHEDULE
56 int __init init_ckrm_sched_res(void);
58 #define init_ckrm_sched_res() ((void)0)
60 //#include <linux/ckrm_sched.h>
63 * This is one of the first .c files built. Error out early
64 * if we have compiler trouble..
66 #if __GNUC__ == 2 && __GNUC_MINOR__ == 96
67 #ifdef CONFIG_FRAME_POINTER
68 #error This compiler cannot compile correctly with frame pointers enabled
72 #ifdef CONFIG_X86_LOCAL_APIC
77 * Versions of gcc older than that listed below may actually compile
78 * and link okay, but the end product can have subtle run time bugs.
79 * To avoid associated bogus bug reports, we flatly refuse to compile
80 * with a gcc that is known to be too old from the very beginning.
82 #if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 95)
83 #error Sorry, your GCC is too old. It builds incorrect kernels.
86 extern char *linux_banner;
88 static int init(void *);
90 extern void init_IRQ(void);
91 extern void sock_init(void);
92 extern void fork_init(unsigned long);
93 extern void mca_init(void);
94 extern void sbus_init(void);
95 extern void sysctl_init(void);
96 extern void signals_init(void);
97 extern void buffer_init(void);
98 extern void pidhash_init(void);
99 extern void pidmap_init(void);
100 extern void prio_tree_init(void);
101 extern void radix_tree_init(void);
102 extern void free_initmem(void);
103 extern void populate_rootfs(void);
104 extern void driver_init(void);
105 extern void prepare_namespace(void);
107 extern void acpi_early_init(void);
109 static inline void acpi_early_init(void) { }
113 extern void tc_init(void);
116 enum system_states system_state;
117 EXPORT_SYMBOL(system_state);
120 * The kernel_magic value represents the address of _end, which allows
121 * namelist tools to "match" each other respectively. That way a tool
122 * that looks at /dev/mem can verify that it is using the right System.map
123 * file -- if kernel_magic doesn't equal the namelist value of _end,
126 extern unsigned long _end;
127 unsigned long *kernel_magic = &_end;
130 * Boot command-line arguments
132 #define MAX_INIT_ARGS 32
133 #define MAX_INIT_ENVS 32
135 extern void time_init(void);
136 /* Default late time init is NULL. archs can override this later. */
137 void (*late_time_init)(void);
138 extern void softirq_init(void);
140 /* Untouched command line (eg. for /proc) saved by arch-specific code. */
141 char saved_command_line[COMMAND_LINE_SIZE];
143 static char *execute_command;
145 /* Setup configured maximum number of CPUs to activate */
146 static unsigned int max_cpus = NR_CPUS;
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 char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
175 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
176 static const char *panic_later, *panic_param;
178 static int __init obsolete_checksetup(char *line)
180 struct obs_kernel_param *p;
181 extern struct obs_kernel_param __setup_start, __setup_end;
185 int n = strlen(p->str);
186 if (!strncmp(line, p->str, n)) {
188 /* Already done in parse_early_param? (Needs
189 * exact match on param part) */
190 if (line[n] == '\0' || line[n] == '=')
192 } else if (!p->setup_func) {
193 printk(KERN_WARNING "Parameter %s is obsolete,"
194 " ignored\n", p->str);
196 } else if (p->setup_func(line + n))
200 } while (p < &__setup_end);
204 static unsigned long preset_lpj;
205 static int __init lpj_setup(char *str)
207 preset_lpj = simple_strtoul(str,NULL,0);
211 __setup("lpj=", lpj_setup);
214 * This should be approx 2 Bo*oMips to start (note initial shift), and will
215 * still work even if initially too large, it will just take slightly longer
217 unsigned long loops_per_jiffy = (1<<12);
219 EXPORT_SYMBOL(loops_per_jiffy);
222 * This is the number of bits of precision for the loops_per_jiffy. Each
223 * bit takes on average 1.5/HZ seconds. This (like the original) is a little
228 void __devinit calibrate_delay(void)
230 unsigned long ticks, loopbit;
231 int lps_precision = LPS_PREC;
234 loops_per_jiffy = preset_lpj;
235 printk("Calibrating delay loop (skipped)... "
236 "%lu.%02lu BogoMIPS preset\n",
237 loops_per_jiffy/(500000/HZ),
238 (loops_per_jiffy/(5000/HZ)) % 100);
240 loops_per_jiffy = (1<<12);
242 printk(KERN_DEBUG "Calibrating delay loop... ");
243 while ((loops_per_jiffy <<= 1) != 0) {
244 /* wait for "start of" clock tick */
246 while (ticks == jiffies)
250 __delay(loops_per_jiffy);
251 ticks = jiffies - ticks;
257 * Do a binary approximation to get loops_per_jiffy set to
258 * equal one clock (up to lps_precision bits)
260 loops_per_jiffy >>= 1;
261 loopbit = loops_per_jiffy;
262 while (lps_precision-- && (loopbit >>= 1)) {
263 loops_per_jiffy |= loopbit;
265 while (ticks == jiffies)
268 __delay(loops_per_jiffy);
269 if (jiffies != ticks) /* longer than 1 tick */
270 loops_per_jiffy &= ~loopbit;
273 /* Round the value and print it */
274 printk("%lu.%02lu BogoMIPS (lpj=%lu)\n",
275 loops_per_jiffy/(500000/HZ),
276 (loops_per_jiffy/(5000/HZ)) % 100,
282 static int __init debug_kernel(char *str)
286 console_loglevel = 10;
290 static int __init quiet_kernel(char *str)
294 console_loglevel = 4;
298 __setup("debug", debug_kernel);
299 __setup("quiet", quiet_kernel);
302 * Unknown boot options get handed to init, unless they look like
305 static int __init unknown_bootoption(char *param, char *val)
307 /* Change NUL term back to "=", to make "param" the whole string. */
309 /* param=val or param="val"? */
310 if (val == param+strlen(param)+1)
312 else if (val == param+strlen(param)+2) {
314 memmove(val-1, val, strlen(val)+1);
320 /* Handle obsolete-style parameters */
321 if (obsolete_checksetup(param))
325 * Preemptive maintenance for "why didn't my mispelled command
328 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
329 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
337 /* Environment option */
339 for (i = 0; envp_init[i]; i++) {
340 if (i == MAX_INIT_ENVS) {
341 panic_later = "Too many boot env vars at `%s'";
344 if (!strncmp(param, envp_init[i], val - param))
347 envp_init[i] = param;
349 /* Command line option */
351 for (i = 0; argv_init[i]; i++) {
352 if (i == MAX_INIT_ARGS) {
353 panic_later = "Too many boot init vars at `%s'";
357 argv_init[i] = param;
362 static int __init init_setup(char *str)
366 execute_command = str;
368 * In case LILO is going to boot us with default command line,
369 * it prepends "auto" before the whole cmdline which makes
370 * the shell think it should execute a script with such name.
371 * So we ignore all arguments entered _before_ init=... [MJ]
373 for (i = 1; i < MAX_INIT_ARGS; i++)
377 __setup("init=", init_setup);
379 extern void setup_arch(char **);
380 extern void cpu_idle(void);
384 #ifdef CONFIG_X86_LOCAL_APIC
385 static void __init smp_init(void)
387 APIC_init_uniprocessor();
390 #define smp_init() do { } while (0)
393 static inline void setup_per_cpu_areas(void) { }
394 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
398 #ifdef __GENERIC_PER_CPU
399 unsigned long __per_cpu_offset[NR_CPUS];
401 EXPORT_SYMBOL(__per_cpu_offset);
403 static void __init setup_per_cpu_areas(void)
405 unsigned long size, i;
407 /* Created by linker magic */
408 extern char __per_cpu_start[], __per_cpu_end[];
410 /* Copy section for each CPU (we discard the original) */
411 size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
412 #ifdef CONFIG_MODULES
413 if (size < PERCPU_ENOUGH_ROOM)
414 size = PERCPU_ENOUGH_ROOM;
417 ptr = alloc_bootmem(size * NR_CPUS);
419 for (i = 0; i < NR_CPUS; i++, ptr += size) {
420 __per_cpu_offset[i] = ptr - __per_cpu_start;
421 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
424 #endif /* !__GENERIC_PER_CPU */
426 /* Called by boot processor to activate the rest. */
427 static void __init smp_init(void)
431 /* FIXME: This should be done in userspace --RR */
432 for_each_present_cpu(i) {
433 if (num_online_cpus() >= max_cpus)
439 /* Any cleanup work */
440 printk("Brought up %ld CPUs\n", (long)num_online_cpus());
441 smp_cpus_done(max_cpus);
443 /* Get other processors into their bootup holding patterns. */
453 * We need to finalize in a non-__init function or else race conditions
454 * between the root thread and the init thread may cause start_kernel to
455 * be reaped by free_initmem before the root thread has proceeded to
458 * gcc-3.4 accidentally inlines this function, so use noinline.
461 static void noinline rest_init(void)
462 __releases(kernel_lock)
464 kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
465 numa_default_policy();
470 /* Check for early params. */
471 static int __init do_early_param(char *param, char *val)
473 struct obs_kernel_param *p;
474 extern struct obs_kernel_param __setup_start, __setup_end;
476 for (p = &__setup_start; p < &__setup_end; p++) {
477 if (p->early && strcmp(param, p->str) == 0) {
478 if (p->setup_func(val) != 0)
480 "Malformed early option '%s'\n", param);
483 /* We accept everything at this stage. */
487 /* Arch code calls this early on, or if not, just before other parsing. */
488 void __init parse_early_param(void)
490 static __initdata int done = 0;
491 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
496 /* All fall through to do_early_param. */
497 strlcpy(tmp_cmdline, saved_command_line, COMMAND_LINE_SIZE);
498 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
503 * Activate the first processor.
506 asmlinkage void __init start_kernel(void)
509 extern struct kernel_param __start___param[], __stop___param[];
511 * Interrupts are still disabled. Do necessary setups, then
516 printk(linux_banner);
517 setup_arch(&command_line);
518 setup_per_cpu_areas();
521 * Mark the boot cpu "online" so that it can call console drivers in
522 * printk() and can access its per-cpu storage.
524 smp_prepare_boot_cpu();
527 * Set up the scheduler prior starting any interrupts (such as the
528 * timer interrupt). Full topology setup happens at smp_init()
529 * time - but meanwhile we still have a functioning scheduler.
532 build_all_zonelists();
534 printk("Kernel command line: %s\n", saved_command_line);
536 parse_args("Booting kernel", command_line, __start___param,
537 __stop___param - __start___param,
538 &unknown_bootoption);
550 * HACK ALERT! This is early. We're enabling the console before
551 * we've done PCI setups etc, and console_init() must be aware of
552 * this. But we do want output early, in case something goes wrong.
556 panic(panic_later, panic_param);
559 #ifdef CONFIG_BLK_DEV_INITRD
560 if (initrd_start && !initrd_below_start_ok &&
561 initrd_start < min_low_pfn << PAGE_SHIFT) {
562 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
563 "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
567 vfs_caches_init_early();
575 pgtable_cache_init();
580 efi_enter_virtual_mode();
582 fork_init(num_physpages);
587 vfs_caches_init(num_physpages);
590 /* rootfs populating might need page-writeback */
591 page_writeback_init();
592 #ifdef CONFIG_PROC_FS
597 acpi_early_init(); /* before LAPIC and SMP init */
599 /* Do the rest non-__init'ed, we're now alive */
603 static int __initdata initcall_debug;
605 static int __init initcall_debug_setup(char *str)
610 __setup("initcall_debug", initcall_debug_setup);
612 struct task_struct *child_reaper = &init_task;
614 extern initcall_t __initcall_start, __initcall_end;
616 static void __init do_initcalls(void)
619 int count = preempt_count();
621 for (call = &__initcall_start; call < &__initcall_end; call++) {
624 if (initcall_debug) {
625 printk(KERN_DEBUG "Calling initcall 0x%p", *call);
626 print_fn_descriptor_symbol(": %s()", (unsigned long) *call);
633 if (preempt_count() != count) {
634 msg = "preemption imbalance";
635 preempt_count() = count;
637 if (irqs_disabled()) {
638 msg = "disabled interrupts";
642 printk("error in initcall at 0x%p: "
643 "returned with %s\n", *call, msg);
647 /* Make sure there is no pending stuff from the initcall sequence */
648 flush_scheduled_work();
652 * Ok, the machine is now initialized. None of the devices
653 * have been touched yet, but the CPU subsystem is up and
654 * running, and memory and process management works.
656 * Now we can finally start doing some real work..
658 static void __init do_basic_setup(void)
660 /* drivers will send hotplug events */
662 usermodehelper_init();
670 /* Networking initialization needs a process context */
676 static void do_pre_smp_initcalls(void)
678 extern int spawn_ksoftirqd(void);
680 extern int migration_init(void);
687 static void run_init_process(char *init_filename)
689 argv_init[0] = init_filename;
690 execve(init_filename, argv_init, envp_init);
693 static inline void fixup_cpu_present_map(void)
699 * If arch is not hotplug ready and did not populate
700 * cpu_present_map, just make cpu_present_map same as cpu_possible_map
701 * for other cpu bringup code to function as normal. e.g smp_init() etc.
703 if (cpus_empty(cpu_present_map)) {
705 cpu_set(i, cpu_present_map);
711 static int init(void * unused)
715 * Tell the world that we're going to be the grim
716 * reaper of innocent orphaned children.
718 * We don't want people to have to make incorrect
719 * assumptions about where in the task array this
722 child_reaper = current;
724 /* Sets up cpus_possible() */
725 smp_prepare_cpus(max_cpus);
727 do_pre_smp_initcalls();
729 fixup_cpu_present_map();
733 * Do this before initcalls, because some drivers want to access
739 init_ckrm_sched_res();
744 * check if there is an early userspace init. If yes, let it do all
747 if (sys_access((const char __user *) "/init", 0) == 0)
748 execute_command = "/init";
753 * Ok, we have completed the initial bootup, and
754 * we're essentially up and running. Get rid of the
755 * initmem segments and start the user-mode stuff..
759 system_state = SYSTEM_RUNNING;
760 numa_default_policy();
762 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
763 printk("Warning: unable to open an initial console.\n");
769 * We try each of these until one succeeds.
771 * The Bourne shell can be used instead of init if we are
772 * trying to recover a really broken machine.
776 run_init_process(execute_command);
778 run_init_process("/sbin/init");
779 run_init_process("/etc/init");
780 run_init_process("/bin/init");
781 run_init_process("/bin/sh");
783 panic("No init found. Try passing init= option to kernel.");
786 static int early_param_test(char *rest)
788 printk("early_parm_test: %s\n", rest ?: "(null)");
789 return rest ? 0 : -EINVAL;
791 early_param("testsetup", early_param_test);
792 static int early_setup_test(char *rest)
794 printk("early_setup_test: %s\n", rest ?: "(null)");
797 __setup("testsetup_long", early_setup_test);