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>
51 #include <asm/setup.h>
53 #include <linux/ckrm.h>
54 #ifdef CONFIG_CKRM_CPU_SCHEDULE
55 int __init init_ckrm_sched_res(void);
57 #define init_ckrm_sched_res() ((void)0)
59 //#include <linux/ckrm_sched.h>
62 * This is one of the first .c files built. Error out early
63 * if we have compiler trouble..
65 #if __GNUC__ == 2 && __GNUC_MINOR__ == 96
66 #ifdef CONFIG_FRAME_POINTER
67 #error This compiler cannot compile correctly with frame pointers enabled
71 #ifdef CONFIG_X86_LOCAL_APIC
76 * Versions of gcc older than that listed below may actually compile
77 * and link okay, but the end product can have subtle run time bugs.
78 * To avoid associated bogus bug reports, we flatly refuse to compile
79 * with a gcc that is known to be too old from the very beginning.
81 #if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 95)
82 #error Sorry, your GCC is too old. It builds incorrect kernels.
85 extern char *linux_banner;
87 static int init(void *);
89 extern void init_IRQ(void);
90 extern void sock_init(void);
91 extern void fork_init(unsigned long);
92 extern void mca_init(void);
93 extern void sbus_init(void);
94 extern void sysctl_init(void);
95 extern void signals_init(void);
96 extern void buffer_init(void);
97 extern void pidhash_init(void);
98 extern void pidmap_init(void);
99 extern void prio_tree_init(void);
100 extern void radix_tree_init(void);
101 extern void free_initmem(void);
102 extern void populate_rootfs(void);
103 extern void driver_init(void);
104 extern void prepare_namespace(void);
106 extern void acpi_early_init(void);
108 static inline void acpi_early_init(void) { }
112 extern void tc_init(void);
115 enum system_states system_state;
116 EXPORT_SYMBOL(system_state);
119 * The kernel_magic value represents the address of _end, which allows
120 * namelist tools to "match" each other respectively. That way a tool
121 * that looks at /dev/mem can verify that it is using the right System.map
122 * file -- if kernel_magic doesn't equal the namelist value of _end,
125 extern unsigned long _end;
126 unsigned long *kernel_magic = &_end;
129 * Boot command-line arguments
131 #define MAX_INIT_ARGS 32
132 #define MAX_INIT_ENVS 32
134 extern void time_init(void);
135 /* Default late time init is NULL. archs can override this later. */
136 void (*late_time_init)(void);
137 extern void softirq_init(void);
139 /* Untouched command line (eg. for /proc) saved by arch-specific code. */
140 char saved_command_line[COMMAND_LINE_SIZE];
142 static char *execute_command;
144 /* Setup configured maximum number of CPUs to activate */
145 static unsigned int max_cpus = NR_CPUS;
148 * Setup routine for controlling SMP activation
150 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
151 * activation entirely (the MPS table probe still happens, though).
153 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
154 * greater than 0, limits the maximum number of CPUs activated in
157 static int __init nosmp(char *str)
163 __setup("nosmp", nosmp);
165 static int __init maxcpus(char *str)
167 get_option(&str, &max_cpus);
171 __setup("maxcpus=", maxcpus);
173 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
174 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
175 static const char *panic_later, *panic_param;
177 static int __init obsolete_checksetup(char *line)
179 struct obs_kernel_param *p;
180 extern struct obs_kernel_param __setup_start, __setup_end;
184 int n = strlen(p->str);
185 if (!strncmp(line, p->str, n)) {
187 /* Already done in parse_early_param? (Needs
188 * exact match on param part) */
189 if (line[n] == '\0' || line[n] == '=')
191 } else if (!p->setup_func) {
192 printk(KERN_WARNING "Parameter %s is obsolete,"
193 " ignored\n", p->str);
195 } else if (p->setup_func(line + n))
199 } while (p < &__setup_end);
203 static unsigned long preset_lpj;
204 static int __init lpj_setup(char *str)
206 preset_lpj = simple_strtoul(str,NULL,0);
210 __setup("lpj=", lpj_setup);
213 * This should be approx 2 Bo*oMips to start (note initial shift), and will
214 * still work even if initially too large, it will just take slightly longer
216 unsigned long loops_per_jiffy = (1<<12);
218 EXPORT_SYMBOL(loops_per_jiffy);
221 * This is the number of bits of precision for the loops_per_jiffy. Each
222 * bit takes on average 1.5/HZ seconds. This (like the original) is a little
227 void __devinit calibrate_delay(void)
229 unsigned long ticks, loopbit;
230 int lps_precision = LPS_PREC;
233 loops_per_jiffy = preset_lpj;
234 printk("Calibrating delay loop (skipped)... "
235 "%lu.%02lu BogoMIPS preset\n",
236 loops_per_jiffy/(500000/HZ),
237 (loops_per_jiffy/(5000/HZ)) % 100);
239 loops_per_jiffy = (1<<12);
241 printk(KERN_DEBUG "Calibrating delay loop... ");
242 while ((loops_per_jiffy <<= 1) != 0) {
243 /* wait for "start of" clock tick */
245 while (ticks == jiffies)
249 __delay(loops_per_jiffy);
250 ticks = jiffies - ticks;
256 * Do a binary approximation to get loops_per_jiffy set to
257 * equal one clock (up to lps_precision bits)
259 loops_per_jiffy >>= 1;
260 loopbit = loops_per_jiffy;
261 while (lps_precision-- && (loopbit >>= 1)) {
262 loops_per_jiffy |= loopbit;
264 while (ticks == jiffies)
267 __delay(loops_per_jiffy);
268 if (jiffies != ticks) /* longer than 1 tick */
269 loops_per_jiffy &= ~loopbit;
272 /* Round the value and print it */
273 printk("%lu.%02lu BogoMIPS (lpj=%lu)\n",
274 loops_per_jiffy/(500000/HZ),
275 (loops_per_jiffy/(5000/HZ)) % 100,
281 static int __init debug_kernel(char *str)
285 console_loglevel = 10;
289 static int __init quiet_kernel(char *str)
293 console_loglevel = 4;
297 __setup("debug", debug_kernel);
298 __setup("quiet", quiet_kernel);
301 * Unknown boot options get handed to init, unless they look like
304 static int __init unknown_bootoption(char *param, char *val)
306 /* Change NUL term back to "=", to make "param" the whole string. */
308 if (val[-1] == '"') val[-2] = '=';
312 /* Handle obsolete-style parameters */
313 if (obsolete_checksetup(param))
317 * Preemptive maintenance for "why didn't my mispelled command
320 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
321 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
329 /* Environment option */
331 for (i = 0; envp_init[i]; i++) {
332 if (i == MAX_INIT_ENVS) {
333 panic_later = "Too many boot env vars at `%s'";
336 if (!strncmp(param, envp_init[i], val - param))
339 envp_init[i] = param;
341 /* Command line option */
343 for (i = 0; argv_init[i]; i++) {
344 if (i == MAX_INIT_ARGS) {
345 panic_later = "Too many boot init vars at `%s'";
349 argv_init[i] = param;
354 static int __init init_setup(char *str)
358 execute_command = str;
360 * In case LILO is going to boot us with default command line,
361 * it prepends "auto" before the whole cmdline which makes
362 * the shell think it should execute a script with such name.
363 * So we ignore all arguments entered _before_ init=... [MJ]
365 for (i = 1; i < MAX_INIT_ARGS; i++)
369 __setup("init=", init_setup);
371 extern void setup_arch(char **);
372 extern void cpu_idle(void);
376 #ifdef CONFIG_X86_LOCAL_APIC
377 static void __init smp_init(void)
379 APIC_init_uniprocessor();
382 #define smp_init() do { } while (0)
385 static inline void setup_per_cpu_areas(void) { }
386 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
390 #ifdef __GENERIC_PER_CPU
391 unsigned long __per_cpu_offset[NR_CPUS];
393 EXPORT_SYMBOL(__per_cpu_offset);
395 static void __init setup_per_cpu_areas(void)
397 unsigned long size, i;
399 /* Created by linker magic */
400 extern char __per_cpu_start[], __per_cpu_end[];
402 /* Copy section for each CPU (we discard the original) */
403 size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
404 #ifdef CONFIG_MODULES
405 if (size < PERCPU_ENOUGH_ROOM)
406 size = PERCPU_ENOUGH_ROOM;
409 ptr = alloc_bootmem(size * NR_CPUS);
411 for (i = 0; i < NR_CPUS; i++, ptr += size) {
412 __per_cpu_offset[i] = ptr - __per_cpu_start;
413 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
416 #endif /* !__GENERIC_PER_CPU */
418 /* Called by boot processor to activate the rest. */
419 static void __init smp_init(void)
423 /* FIXME: This should be done in userspace --RR */
424 for_each_present_cpu(i) {
425 if (num_online_cpus() >= max_cpus)
431 /* Any cleanup work */
432 printk("Brought up %ld CPUs\n", (long)num_online_cpus());
433 smp_cpus_done(max_cpus);
435 /* Get other processors into their bootup holding patterns. */
445 * We need to finalize in a non-__init function or else race conditions
446 * between the root thread and the init thread may cause start_kernel to
447 * be reaped by free_initmem before the root thread has proceeded to
450 * gcc-3.4 accidentally inlines this function, so use noinline.
453 static void noinline rest_init(void)
455 kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
456 numa_default_policy();
457 system_state = SYSTEM_BOOTING_SCHEDULER_OK;
462 /* Check for early params. */
463 static int __init do_early_param(char *param, char *val)
465 struct obs_kernel_param *p;
466 extern struct obs_kernel_param __setup_start, __setup_end;
468 for (p = &__setup_start; p < &__setup_end; p++) {
469 if (p->early && strcmp(param, p->str) == 0) {
470 if (p->setup_func(val) != 0)
472 "Malformed early option '%s'\n", param);
475 /* We accept everything at this stage. */
479 /* Arch code calls this early on, or if not, just before other parsing. */
480 void __init parse_early_param(void)
482 static __initdata int done = 0;
483 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
488 /* All fall through to do_early_param. */
489 strlcpy(tmp_cmdline, saved_command_line, COMMAND_LINE_SIZE);
490 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
495 * Activate the first processor.
498 asmlinkage void __init start_kernel(void)
501 extern struct kernel_param __start___param[], __stop___param[];
503 * Interrupts are still disabled. Do necessary setups, then
508 printk(linux_banner);
509 setup_arch(&command_line);
510 setup_per_cpu_areas();
513 * Mark the boot cpu "online" so that it can call console drivers in
514 * printk() and can access its per-cpu storage.
516 smp_prepare_boot_cpu();
519 * Set up the scheduler prior starting any interrupts (such as the
520 * timer interrupt). Full topology setup happens at smp_init()
521 * time - but meanwhile we still have a functioning scheduler.
524 build_all_zonelists();
526 printk("Kernel command line: %s\n", saved_command_line);
528 parse_args("Booting kernel", command_line, __start___param,
529 __stop___param - __start___param,
530 &unknown_bootoption);
536 /* MEF: In 2.6.5. ckrm_init was right after pidhash_init() but
537 before sched_init(). Will leave it after pidhash_init()
546 * HACK ALERT! This is early. We're enabling the console before
547 * we've done PCI setups etc, and console_init() must be aware of
548 * this. But we do want output early, in case something goes wrong.
552 panic(panic_later, panic_param);
555 #ifdef CONFIG_BLK_DEV_INITRD
556 if (initrd_start && !initrd_below_start_ok &&
557 initrd_start < min_low_pfn << PAGE_SHIFT) {
558 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
559 "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
563 vfs_caches_init_early();
571 pgtable_cache_init();
576 efi_enter_virtual_mode();
578 fork_init(num_physpages);
582 security_scaffolding_startup();
583 vfs_caches_init(num_physpages);
586 /* rootfs populating might need page-writeback */
587 page_writeback_init();
588 #ifdef CONFIG_PROC_FS
594 acpi_early_init(); /* before LAPIC and SMP init */
596 /* Do the rest non-__init'ed, we're now alive */
600 static int __initdata initcall_debug;
602 static int __init initcall_debug_setup(char *str)
607 __setup("initcall_debug", initcall_debug_setup);
609 struct task_struct *child_reaper = &init_task;
611 extern initcall_t __initcall_start, __initcall_end;
613 static void __init do_initcalls(void)
616 int count = preempt_count();
618 for (call = &__initcall_start; call < &__initcall_end; call++) {
621 if (initcall_debug) {
622 printk(KERN_DEBUG "Calling initcall 0x%p", *call);
623 print_symbol(": %s()", (unsigned long) *call);
630 if (preempt_count() != count) {
631 msg = "preemption imbalance";
632 preempt_count() = count;
634 if (irqs_disabled()) {
635 msg = "disabled interrupts";
639 printk("error in initcall at 0x%p: "
640 "returned with %s\n", *call, msg);
644 /* Make sure there is no pending stuff from the initcall sequence */
645 flush_scheduled_work();
649 * Ok, the machine is now initialized. None of the devices
650 * have been touched yet, but the CPU subsystem is up and
651 * running, and memory and process management works.
653 * Now we can finally start doing some real work..
655 static void __init do_basic_setup(void)
657 /* drivers will send hotplug events */
659 usermodehelper_init();
667 /* Networking initialization needs a process context */
673 static void do_pre_smp_initcalls(void)
675 extern int spawn_ksoftirqd(void);
677 extern int migration_init(void);
684 static void run_init_process(char *init_filename)
686 argv_init[0] = init_filename;
687 execve(init_filename, argv_init, envp_init);
690 static inline void fixup_cpu_present_map(void)
696 * If arch is not hotplug ready and did not populate
697 * cpu_present_map, just make cpu_present_map same as cpu_possible_map
698 * for other cpu bringup code to function as normal. e.g smp_init() etc.
700 if (cpus_empty(cpu_present_map)) {
702 cpu_set(i, cpu_present_map);
708 static int init(void * unused)
712 * Tell the world that we're going to be the grim
713 * reaper of innocent orphaned children.
715 * We don't want people to have to make incorrect
716 * assumptions about where in the task array this
719 child_reaper = current;
721 /* Sets up cpus_possible() */
722 smp_prepare_cpus(max_cpus);
724 do_pre_smp_initcalls();
726 fixup_cpu_present_map();
730 * Do this before initcalls, because some drivers want to access
737 init_ckrm_sched_res();
742 * check if there is an early userspace init. If yes, let it do all
745 if (sys_access((const char __user *) "/init", 0) == 0)
746 execute_command = "/init";
751 * Ok, we have completed the initial bootup, and
752 * we're essentially up and running. Get rid of the
753 * initmem segments and start the user-mode stuff..
757 system_state = SYSTEM_RUNNING;
758 numa_default_policy();
760 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
761 printk("Warning: unable to open an initial console.\n");
767 * We try each of these until one succeeds.
769 * The Bourne shell can be used instead of init if we are
770 * trying to recover a really broken machine.
774 run_init_process(execute_command);
776 run_init_process("/sbin/init");
777 run_init_process("/etc/init");
778 run_init_process("/bin/init");
779 run_init_process("/bin/sh");
781 panic("No init found. Try passing init= option to kernel.");
784 static int early_param_test(char *rest)
786 printk("early_parm_test: %s\n", rest ?: "(null)");
787 return rest ? 0 : -EINVAL;
789 early_param("testsetup", early_param_test);
790 static int early_setup_test(char *rest)
792 printk("early_setup_test: %s\n", rest ?: "(null)");
795 __setup("testsetup_long", early_setup_test);