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/kernel_stat.h>
35 #include <linux/security.h>
36 #include <linux/workqueue.h>
37 #include <linux/profile.h>
38 #include <linux/rcupdate.h>
39 #include <linux/moduleparam.h>
40 #include <linux/kallsyms.h>
41 #include <linux/writeback.h>
42 #include <linux/cpu.h>
43 #include <linux/efi.h>
44 #include <linux/unistd.h>
45 #include <linux/rmap.h>
46 #include <linux/mempolicy.h>
50 #include <asm/setup.h>
52 #include <linux/ckrm.h>
53 #ifdef CONFIG_CKRM_CPU_SCHEDULE
54 int __init init_ckrm_sched_res(void);
56 #define init_ckrm_sched_res() ((void)0)
58 //#include <linux/ckrm_sched.h>
61 * This is one of the first .c files built. Error out early
62 * if we have compiler trouble..
64 #if __GNUC__ == 2 && __GNUC_MINOR__ == 96
65 #ifdef CONFIG_FRAME_POINTER
66 #error This compiler cannot compile correctly with frame pointers enabled
70 #ifdef CONFIG_X86_LOCAL_APIC
75 * Versions of gcc older than that listed below may actually compile
76 * and link okay, but the end product can have subtle run time bugs.
77 * To avoid associated bogus bug reports, we flatly refuse to compile
78 * with a gcc that is known to be too old from the very beginning.
80 #if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 95)
81 #error Sorry, your GCC is too old. It builds incorrect kernels.
84 extern char *linux_banner;
86 static int init(void *);
88 extern void init_IRQ(void);
89 extern void sock_init(void);
90 extern void fork_init(unsigned long);
91 extern void mca_init(void);
92 extern void sbus_init(void);
93 extern void sysctl_init(void);
94 extern void signals_init(void);
95 extern void buffer_init(void);
96 extern void pidhash_init(void);
97 extern void pidmap_init(void);
98 extern void prio_tree_init(void);
99 extern void radix_tree_init(void);
100 extern void free_initmem(void);
101 extern void populate_rootfs(void);
102 extern void driver_init(void);
103 extern void prepare_namespace(void);
106 extern void tc_init(void);
109 enum system_states system_state;
110 EXPORT_SYMBOL(system_state);
113 * The kernel_magic value represents the address of _end, which allows
114 * namelist tools to "match" each other respectively. That way a tool
115 * that looks at /dev/mem can verify that it is using the right System.map
116 * file -- if kernel_magic doesn't equal the namelist value of _end,
119 extern unsigned long _end;
120 unsigned long *kernel_magic = &_end;
123 * Boot command-line arguments
125 #define MAX_INIT_ARGS 8
126 #define MAX_INIT_ENVS 8
128 extern void time_init(void);
129 /* Default late time init is NULL. archs can override this later. */
130 void (*late_time_init)(void);
131 extern void softirq_init(void);
133 /* Untouched command line (eg. for /proc) saved by arch-specific code. */
134 char saved_command_line[COMMAND_LINE_SIZE];
136 static char *execute_command;
138 /* Setup configured maximum number of CPUs to activate */
139 static unsigned int max_cpus = NR_CPUS;
142 * Setup routine for controlling SMP activation
144 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
145 * activation entirely (the MPS table probe still happens, though).
147 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
148 * greater than 0, limits the maximum number of CPUs activated in
151 static int __init nosmp(char *str)
157 __setup("nosmp", nosmp);
159 static int __init maxcpus(char *str)
161 get_option(&str, &max_cpus);
165 __setup("maxcpus=", maxcpus);
167 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
168 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
169 static const char *panic_later, *panic_param;
171 __setup("profile=", profile_setup);
173 static int __init obsolete_checksetup(char *line)
175 struct obs_kernel_param *p;
176 extern struct obs_kernel_param __setup_start, __setup_end;
180 int n = strlen(p->str);
181 if (!strncmp(line, p->str, n)) {
183 /* Already done in parse_early_param? (Needs
184 * exact match on param part) */
185 if (line[n] == '\0' || line[n] == '=')
187 } else if (!p->setup_func) {
188 printk(KERN_WARNING "Parameter %s is obsolete,"
189 " ignored\n", p->str);
191 } else if (p->setup_func(line + n))
195 } while (p < &__setup_end);
199 /* this should be approx 2 Bo*oMips to start (note initial shift), and will
200 still work even if initially too large, it will just take slightly longer */
201 unsigned long loops_per_jiffy = (1<<12);
203 EXPORT_SYMBOL(loops_per_jiffy);
205 /* This is the number of bits of precision for the loops_per_jiffy. Each
206 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;
215 loops_per_jiffy = (1<<12);
217 printk("Calibrating delay loop... ");
218 while ((loops_per_jiffy <<= 1) != 0) {
219 /* wait for "start of" clock tick */
221 while (ticks == jiffies)
225 __delay(loops_per_jiffy);
226 ticks = jiffies - ticks;
231 /* Do a binary approximation to get loops_per_jiffy set to equal one clock
232 (up to lps_precision bits) */
233 loops_per_jiffy >>= 1;
234 loopbit = loops_per_jiffy;
235 while ( lps_precision-- && (loopbit >>= 1) ) {
236 loops_per_jiffy |= loopbit;
238 while (ticks == jiffies);
240 __delay(loops_per_jiffy);
241 if (jiffies != ticks) /* longer than 1 tick */
242 loops_per_jiffy &= ~loopbit;
245 /* Round the value and print it */
246 printk("%lu.%02lu BogoMIPS\n",
247 loops_per_jiffy/(500000/HZ),
248 (loops_per_jiffy/(5000/HZ)) % 100);
251 static int __init debug_kernel(char *str)
255 console_loglevel = 10;
259 static int __init quiet_kernel(char *str)
263 console_loglevel = 4;
267 __setup("debug", debug_kernel);
268 __setup("quiet", quiet_kernel);
270 /* Unknown boot options get handed to init, unless they look like
272 static int __init unknown_bootoption(char *param, char *val)
274 /* Change NUL term back to "=", to make "param" the whole string. */
278 /* Handle obsolete-style parameters */
279 if (obsolete_checksetup(param))
282 /* Preemptive maintenance for "why didn't my mispelled command
284 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
285 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
293 /* Environment option */
295 for (i = 0; envp_init[i]; i++) {
296 if (i == MAX_INIT_ENVS) {
297 panic_later = "Too many boot env vars at `%s'";
300 if (!strncmp(param, envp_init[i], val - param))
303 envp_init[i] = param;
305 /* Command line option */
307 for (i = 0; argv_init[i]; i++) {
308 if (i == MAX_INIT_ARGS) {
309 panic_later = "Too many boot init vars at `%s'";
313 argv_init[i] = param;
318 static int __init init_setup(char *str)
322 execute_command = str;
323 /* In case LILO is going to boot us with default command line,
324 * it prepends "auto" before the whole cmdline which makes
325 * the shell think it should execute a script with such name.
326 * So we ignore all arguments entered _before_ init=... [MJ]
328 for (i = 1; i < MAX_INIT_ARGS; i++)
332 __setup("init=", init_setup);
334 extern void setup_arch(char **);
335 extern void cpu_idle(void);
339 #ifdef CONFIG_X86_LOCAL_APIC
340 static void __init smp_init(void)
342 APIC_init_uniprocessor();
345 #define smp_init() do { } while (0)
348 static inline void setup_per_cpu_areas(void) { }
349 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
353 #ifdef __GENERIC_PER_CPU
354 unsigned long __per_cpu_offset[NR_CPUS];
356 EXPORT_SYMBOL(__per_cpu_offset);
358 static void __init setup_per_cpu_areas(void)
360 unsigned long size, i;
362 /* Created by linker magic */
363 extern char __per_cpu_start[], __per_cpu_end[];
365 /* Copy section for each CPU (we discard the original) */
366 size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
367 #ifdef CONFIG_MODULES
368 if (size < PERCPU_ENOUGH_ROOM)
369 size = PERCPU_ENOUGH_ROOM;
372 ptr = alloc_bootmem(size * NR_CPUS);
374 for (i = 0; i < NR_CPUS; i++, ptr += size) {
375 __per_cpu_offset[i] = ptr - __per_cpu_start;
376 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
379 #endif /* !__GENERIC_PER_CPU */
381 /* Called by boot processor to activate the rest. */
382 static void __init smp_init(void)
386 /* FIXME: This should be done in userspace --RR */
387 for_each_present_cpu(i) {
388 if (num_online_cpus() >= max_cpus)
394 /* Any cleanup work */
395 printk("Brought up %ld CPUs\n", (long)num_online_cpus());
396 smp_cpus_done(max_cpus);
398 /* Get other processors into their bootup holding patterns. */
408 * We need to finalize in a non-__init function or else race conditions
409 * between the root thread and the init thread may cause start_kernel to
410 * be reaped by free_initmem before the root thread has proceeded to
413 * gcc-3.4 accidentally inlines this function, so use noinline.
416 static void noinline rest_init(void)
418 kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
419 numa_default_policy();
420 system_state = SYSTEM_BOOTING_SCHEDULER_OK;
425 /* Check for early params. */
426 static int __init do_early_param(char *param, char *val)
428 struct obs_kernel_param *p;
429 extern struct obs_kernel_param __setup_start, __setup_end;
431 for (p = &__setup_start; p < &__setup_end; p++) {
432 if (p->early && strcmp(param, p->str) == 0) {
433 if (p->setup_func(val) != 0)
435 "Malformed early option '%s'\n", param);
438 /* We accept everything at this stage. */
442 /* Arch code calls this early on, or if not, just before other parsing. */
443 void __init parse_early_param(void)
445 static __initdata int done = 0;
446 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
451 /* All fall through to do_early_param. */
452 strlcpy(tmp_cmdline, saved_command_line, COMMAND_LINE_SIZE);
453 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
458 * Activate the first processor.
461 asmlinkage void __init start_kernel(void)
464 extern struct kernel_param __start___param[], __stop___param[];
466 * Interrupts are still disabled. Do necessary setups, then
471 printk(linux_banner);
472 setup_arch(&command_line);
473 setup_per_cpu_areas();
476 * Mark the boot cpu "online" so that it can call console drivers in
477 * printk() and can access its per-cpu storage.
479 smp_prepare_boot_cpu();
482 * Set up the scheduler prior starting any interrupts (such as the
483 * timer interrupt). Full topology setup happens at smp_init()
484 * time - but meanwhile we still have a functioning scheduler.
488 build_all_zonelists();
490 printk("Kernel command line: %s\n", saved_command_line);
492 parse_args("Booting kernel", command_line, __start___param,
493 __stop___param - __start___param,
494 &unknown_bootoption);
500 /* MEF: In 2.6.5. ckrm_init was right after pidhash_init() but
501 before sched_init(). Will leave it after pidhash_init()
510 * HACK ALERT! This is early. We're enabling the console before
511 * we've done PCI setups etc, and console_init() must be aware of
512 * this. But we do want output early, in case something goes wrong.
516 panic(panic_later, panic_param);
519 #ifdef CONFIG_BLK_DEV_INITRD
520 if (initrd_start && !initrd_below_start_ok &&
521 initrd_start < min_low_pfn << PAGE_SHIFT) {
522 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
523 "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
527 vfs_caches_init_early();
535 pgtable_cache_init();
540 efi_enter_virtual_mode();
542 fork_init(num_physpages);
546 security_scaffolding_startup();
547 vfs_caches_init(num_physpages);
550 /* rootfs populating might need page-writeback */
551 page_writeback_init();
552 #ifdef CONFIG_PROC_FS
559 * We count on the initial thread going ok
560 * Like idlers init is an unlocked kernel thread, which will
561 * make syscalls (and thus be locked).
563 init_idle(current, smp_processor_id());
565 /* Do the rest non-__init'ed, we're now alive */
569 static int __initdata initcall_debug;
571 static int __init initcall_debug_setup(char *str)
576 __setup("initcall_debug", initcall_debug_setup);
578 struct task_struct *child_reaper = &init_task;
580 extern initcall_t __initcall_start, __initcall_end;
582 static void __init do_initcalls(void)
585 int count = preempt_count();
587 for (call = &__initcall_start; call < &__initcall_end; call++) {
590 if (initcall_debug) {
591 printk(KERN_DEBUG "Calling initcall 0x%p", *call);
592 print_symbol(": %s()", (unsigned long) *call);
599 if (preempt_count() != count) {
600 msg = "preemption imbalance";
601 preempt_count() = count;
603 if (irqs_disabled()) {
604 msg = "disabled interrupts";
608 printk("error in initcall at 0x%p: "
609 "returned with %s\n", *call, msg);
613 /* Make sure there is no pending stuff from the initcall sequence */
614 flush_scheduled_work();
618 * Ok, the machine is now initialized. None of the devices
619 * have been touched yet, but the CPU subsystem is up and
620 * running, and memory and process management works.
622 * Now we can finally start doing some real work..
624 static void __init do_basic_setup(void)
632 /* Networking initialization needs a process context */
639 static void do_pre_smp_initcalls(void)
641 extern int spawn_ksoftirqd(void);
643 extern int migration_init(void);
650 static void run_init_process(char *init_filename)
652 argv_init[0] = init_filename;
653 execve(init_filename, argv_init, envp_init);
656 static inline void fixup_cpu_present_map(void)
662 * If arch is not hotplug ready and did not populate
663 * cpu_present_map, just make cpu_present_map same as cpu_possible_map
664 * for other cpu bringup code to function as normal. e.g smp_init() etc.
666 if (cpus_empty(cpu_present_map)) {
668 cpu_set(i, cpu_present_map);
674 static int init(void * unused)
678 * Tell the world that we're going to be the grim
679 * reaper of innocent orphaned children.
681 * We don't want people to have to make incorrect
682 * assumptions about where in the task array this
685 child_reaper = current;
687 /* Sets up cpus_possible() */
688 smp_prepare_cpus(max_cpus);
690 do_pre_smp_initcalls();
692 fixup_cpu_present_map();
696 * Do this before initcalls, because some drivers want to access
703 init_ckrm_sched_res();
708 * check if there is an early userspace init. If yes, let it do all
711 if (sys_access((const char __user *) "/init", 0) == 0)
712 execute_command = "/init";
717 * Ok, we have completed the initial bootup, and
718 * we're essentially up and running. Get rid of the
719 * initmem segments and start the user-mode stuff..
723 system_state = SYSTEM_RUNNING;
724 numa_default_policy();
726 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
727 printk("Warning: unable to open an initial console.\n");
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.
740 run_init_process(execute_command);
742 run_init_process("/sbin/init");
743 run_init_process("/etc/init");
744 run_init_process("/bin/init");
745 run_init_process("/bin/sh");
747 panic("No init found. Try passing init= option to kernel.");
750 static int early_param_test(char *rest)
752 printk("early_parm_test: %s\n", rest ?: "(null)");
753 return rest ? 0 : -EINVAL;
755 early_param("testsetup", early_param_test);
756 static int early_setup_test(char *rest)
758 printk("early_setup_test: %s\n", rest ?: "(null)");
761 __setup("testsetup_long", early_setup_test);