[PATCH] hugetlb: fix absurd HugePages_Rsvd
[linux-2.6.22.y-op.git] / init / main.c
blob36f608a7cfbaf0fe1e2d3b1f37ff42e945bc100d
1 /*
2 * linux/init/main.c
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/security.h>
33 #include <linux/workqueue.h>
34 #include <linux/profile.h>
35 #include <linux/rcupdate.h>
36 #include <linux/moduleparam.h>
37 #include <linux/kallsyms.h>
38 #include <linux/writeback.h>
39 #include <linux/cpu.h>
40 #include <linux/cpuset.h>
41 #include <linux/efi.h>
42 #include <linux/taskstats_kern.h>
43 #include <linux/delayacct.h>
44 #include <linux/unistd.h>
45 #include <linux/rmap.h>
46 #include <linux/mempolicy.h>
47 #include <linux/key.h>
48 #include <linux/unwind.h>
49 #include <linux/buffer_head.h>
50 #include <linux/debug_locks.h>
51 #include <linux/lockdep.h>
53 #include <asm/io.h>
54 #include <asm/bugs.h>
55 #include <asm/setup.h>
56 #include <asm/sections.h>
57 #include <asm/cacheflush.h>
59 #ifdef CONFIG_X86_LOCAL_APIC
60 #include <asm/smp.h>
61 #endif
64 * This is one of the first .c files built. Error out early if we have compiler
65 * trouble.
67 * Versions of gcc older than that listed below may actually compile and link
68 * okay, but the end product can have subtle run time bugs. To avoid associated
69 * bogus bug reports, we flatly refuse to compile with a gcc that is known to be
70 * too old from the very beginning.
72 #if (__GNUC__ < 3) || (__GNUC__ == 3 && __GNUC_MINOR__ < 2)
73 #error Sorry, your GCC is too old. It builds incorrect kernels.
74 #endif
76 static int init(void *);
78 extern void init_IRQ(void);
79 extern void fork_init(unsigned long);
80 extern void mca_init(void);
81 extern void sbus_init(void);
82 extern void sysctl_init(void);
83 extern void signals_init(void);
84 extern void pidhash_init(void);
85 extern void pidmap_init(void);
86 extern void prio_tree_init(void);
87 extern void radix_tree_init(void);
88 extern void free_initmem(void);
89 extern void populate_rootfs(void);
90 extern void driver_init(void);
91 extern void prepare_namespace(void);
92 #ifdef CONFIG_ACPI
93 extern void acpi_early_init(void);
94 #else
95 static inline void acpi_early_init(void) { }
96 #endif
97 #ifndef CONFIG_DEBUG_RODATA
98 static inline void mark_rodata_ro(void) { }
99 #endif
101 #ifdef CONFIG_TC
102 extern void tc_init(void);
103 #endif
105 enum system_states system_state;
106 EXPORT_SYMBOL(system_state);
109 * Boot command-line arguments
111 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
112 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
114 extern void time_init(void);
115 /* Default late time init is NULL. archs can override this later. */
116 void (*late_time_init)(void);
117 extern void softirq_init(void);
119 /* Untouched command line (eg. for /proc) saved by arch-specific code. */
120 char saved_command_line[COMMAND_LINE_SIZE];
122 static char *execute_command;
123 static char *ramdisk_execute_command;
125 /* Setup configured maximum number of CPUs to activate */
126 static unsigned int max_cpus = NR_CPUS;
129 * If set, this is an indication to the drivers that reset the underlying
130 * device before going ahead with the initialization otherwise driver might
131 * rely on the BIOS and skip the reset operation.
133 * This is useful if kernel is booting in an unreliable environment.
134 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
135 * skipped and devices will be in unknown state.
137 unsigned int reset_devices;
138 EXPORT_SYMBOL(reset_devices);
141 * Setup routine for controlling SMP activation
143 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
144 * activation entirely (the MPS table probe still happens, though).
146 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
147 * greater than 0, limits the maximum number of CPUs activated in
148 * SMP mode to <NUM>.
150 static int __init nosmp(char *str)
152 max_cpus = 0;
153 return 1;
156 __setup("nosmp", nosmp);
158 static int __init maxcpus(char *str)
160 get_option(&str, &max_cpus);
161 return 1;
164 __setup("maxcpus=", maxcpus);
166 static int __init set_reset_devices(char *str)
168 reset_devices = 1;
169 return 1;
172 __setup("reset_devices", set_reset_devices);
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 extern struct obs_kernel_param __setup_start[], __setup_end[];
180 static int __init obsolete_checksetup(char *line)
182 struct obs_kernel_param *p;
183 int had_early_param = 0;
185 p = __setup_start;
186 do {
187 int n = strlen(p->str);
188 if (!strncmp(line, p->str, n)) {
189 if (p->early) {
190 /* Already done in parse_early_param?
191 * (Needs exact match on param part).
192 * Keep iterating, as we can have early
193 * params and __setups of same names 8( */
194 if (line[n] == '\0' || line[n] == '=')
195 had_early_param = 1;
196 } else if (!p->setup_func) {
197 printk(KERN_WARNING "Parameter %s is obsolete,"
198 " ignored\n", p->str);
199 return 1;
200 } else if (p->setup_func(line + n))
201 return 1;
203 p++;
204 } while (p < __setup_end);
206 return had_early_param;
210 * This should be approx 2 Bo*oMips to start (note initial shift), and will
211 * still work even if initially too large, it will just take slightly longer
213 unsigned long loops_per_jiffy = (1<<12);
215 EXPORT_SYMBOL(loops_per_jiffy);
217 static int __init debug_kernel(char *str)
219 if (*str)
220 return 0;
221 console_loglevel = 10;
222 return 1;
225 static int __init quiet_kernel(char *str)
227 if (*str)
228 return 0;
229 console_loglevel = 4;
230 return 1;
233 __setup("debug", debug_kernel);
234 __setup("quiet", quiet_kernel);
236 static int __init loglevel(char *str)
238 get_option(&str, &console_loglevel);
239 return 1;
242 __setup("loglevel=", loglevel);
245 * Unknown boot options get handed to init, unless they look like
246 * failed parameters
248 static int __init unknown_bootoption(char *param, char *val)
250 /* Change NUL term back to "=", to make "param" the whole string. */
251 if (val) {
252 /* param=val or param="val"? */
253 if (val == param+strlen(param)+1)
254 val[-1] = '=';
255 else if (val == param+strlen(param)+2) {
256 val[-2] = '=';
257 memmove(val-1, val, strlen(val)+1);
258 val--;
259 } else
260 BUG();
263 /* Handle obsolete-style parameters */
264 if (obsolete_checksetup(param))
265 return 0;
268 * Preemptive maintenance for "why didn't my mispelled command
269 * line work?"
271 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
272 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
273 return 0;
276 if (panic_later)
277 return 0;
279 if (val) {
280 /* Environment option */
281 unsigned int i;
282 for (i = 0; envp_init[i]; i++) {
283 if (i == MAX_INIT_ENVS) {
284 panic_later = "Too many boot env vars at `%s'";
285 panic_param = param;
287 if (!strncmp(param, envp_init[i], val - param))
288 break;
290 envp_init[i] = param;
291 } else {
292 /* Command line option */
293 unsigned int i;
294 for (i = 0; argv_init[i]; i++) {
295 if (i == MAX_INIT_ARGS) {
296 panic_later = "Too many boot init vars at `%s'";
297 panic_param = param;
300 argv_init[i] = param;
302 return 0;
305 static int __init init_setup(char *str)
307 unsigned int i;
309 execute_command = str;
311 * In case LILO is going to boot us with default command line,
312 * it prepends "auto" before the whole cmdline which makes
313 * the shell think it should execute a script with such name.
314 * So we ignore all arguments entered _before_ init=... [MJ]
316 for (i = 1; i < MAX_INIT_ARGS; i++)
317 argv_init[i] = NULL;
318 return 1;
320 __setup("init=", init_setup);
322 static int __init rdinit_setup(char *str)
324 unsigned int i;
326 ramdisk_execute_command = str;
327 /* See "auto" comment in init_setup */
328 for (i = 1; i < MAX_INIT_ARGS; i++)
329 argv_init[i] = NULL;
330 return 1;
332 __setup("rdinit=", rdinit_setup);
334 #ifndef CONFIG_SMP
336 #ifdef CONFIG_X86_LOCAL_APIC
337 static void __init smp_init(void)
339 APIC_init_uniprocessor();
341 #else
342 #define smp_init() do { } while (0)
343 #endif
345 static inline void setup_per_cpu_areas(void) { }
346 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
348 #else
350 #ifdef __GENERIC_PER_CPU
351 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
353 EXPORT_SYMBOL(__per_cpu_offset);
355 static void __init setup_per_cpu_areas(void)
357 unsigned long size, i;
358 char *ptr;
359 unsigned long nr_possible_cpus = num_possible_cpus();
361 /* Copy section for each CPU (we discard the original) */
362 size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
363 #ifdef CONFIG_MODULES
364 if (size < PERCPU_ENOUGH_ROOM)
365 size = PERCPU_ENOUGH_ROOM;
366 #endif
367 ptr = alloc_bootmem(size * nr_possible_cpus);
369 for_each_possible_cpu(i) {
370 __per_cpu_offset[i] = ptr - __per_cpu_start;
371 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
372 ptr += size;
375 #endif /* !__GENERIC_PER_CPU */
377 /* Called by boot processor to activate the rest. */
378 static void __init smp_init(void)
380 unsigned int i;
382 /* FIXME: This should be done in userspace --RR */
383 for_each_present_cpu(i) {
384 if (num_online_cpus() >= max_cpus)
385 break;
386 if (!cpu_online(i))
387 cpu_up(i);
390 /* Any cleanup work */
391 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
392 smp_cpus_done(max_cpus);
393 #if 0
394 /* Get other processors into their bootup holding patterns. */
396 smp_commence();
397 #endif
400 #endif
403 * We need to finalize in a non-__init function or else race conditions
404 * between the root thread and the init thread may cause start_kernel to
405 * be reaped by free_initmem before the root thread has proceeded to
406 * cpu_idle.
408 * gcc-3.4 accidentally inlines this function, so use noinline.
411 static void noinline rest_init(void)
412 __releases(kernel_lock)
414 kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
415 numa_default_policy();
416 unlock_kernel();
419 * The boot idle thread must execute schedule()
420 * at least one to get things moving:
422 preempt_enable_no_resched();
423 schedule();
424 preempt_disable();
426 /* Call into cpu_idle with preempt disabled */
427 cpu_idle();
430 /* Check for early params. */
431 static int __init do_early_param(char *param, char *val)
433 struct obs_kernel_param *p;
435 for (p = __setup_start; p < __setup_end; p++) {
436 if (p->early && strcmp(param, p->str) == 0) {
437 if (p->setup_func(val) != 0)
438 printk(KERN_WARNING
439 "Malformed early option '%s'\n", param);
442 /* We accept everything at this stage. */
443 return 0;
446 /* Arch code calls this early on, or if not, just before other parsing. */
447 void __init parse_early_param(void)
449 static __initdata int done = 0;
450 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
452 if (done)
453 return;
455 /* All fall through to do_early_param. */
456 strlcpy(tmp_cmdline, saved_command_line, COMMAND_LINE_SIZE);
457 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
458 done = 1;
462 * Activate the first processor.
465 static void __init boot_cpu_init(void)
467 int cpu = smp_processor_id();
468 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
469 cpu_set(cpu, cpu_online_map);
470 cpu_set(cpu, cpu_present_map);
471 cpu_set(cpu, cpu_possible_map);
474 void __init __attribute__((weak)) smp_setup_processor_id(void)
478 asmlinkage void __init start_kernel(void)
480 char * command_line;
481 extern struct kernel_param __start___param[], __stop___param[];
483 smp_setup_processor_id();
486 * Need to run as early as possible, to initialize the
487 * lockdep hash:
489 unwind_init();
490 lockdep_init();
492 local_irq_disable();
493 early_boot_irqs_off();
494 early_init_irq_lock_class();
497 * Interrupts are still disabled. Do necessary setups, then
498 * enable them
500 lock_kernel();
501 boot_cpu_init();
502 page_address_init();
503 printk(KERN_NOTICE);
504 printk(linux_banner);
505 setup_arch(&command_line);
506 unwind_setup();
507 setup_per_cpu_areas();
508 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
511 * Set up the scheduler prior starting any interrupts (such as the
512 * timer interrupt). Full topology setup happens at smp_init()
513 * time - but meanwhile we still have a functioning scheduler.
515 sched_init();
517 * Disable preemption - early bootup scheduling is extremely
518 * fragile until we cpu_idle() for the first time.
520 preempt_disable();
521 build_all_zonelists();
522 page_alloc_init();
523 printk(KERN_NOTICE "Kernel command line: %s\n", saved_command_line);
524 parse_early_param();
525 parse_args("Booting kernel", command_line, __start___param,
526 __stop___param - __start___param,
527 &unknown_bootoption);
528 sort_main_extable();
529 trap_init();
530 rcu_init();
531 init_IRQ();
532 pidhash_init();
533 init_timers();
534 hrtimers_init();
535 softirq_init();
536 timekeeping_init();
537 time_init();
538 profile_init();
539 if (!irqs_disabled())
540 printk("start_kernel(): bug: interrupts were enabled early\n");
541 early_boot_irqs_on();
542 local_irq_enable();
545 * HACK ALERT! This is early. We're enabling the console before
546 * we've done PCI setups etc, and console_init() must be aware of
547 * this. But we do want output early, in case something goes wrong.
549 console_init();
550 if (panic_later)
551 panic(panic_later, panic_param);
553 lockdep_info();
556 * Need to run this when irqs are enabled, because it wants
557 * to self-test [hard/soft]-irqs on/off lock inversion bugs
558 * too:
560 locking_selftest();
562 #ifdef CONFIG_BLK_DEV_INITRD
563 if (initrd_start && !initrd_below_start_ok &&
564 initrd_start < min_low_pfn << PAGE_SHIFT) {
565 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
566 "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
567 initrd_start = 0;
569 #endif
570 vfs_caches_init_early();
571 cpuset_init_early();
572 mem_init();
573 kmem_cache_init();
574 setup_per_cpu_pageset();
575 numa_policy_init();
576 if (late_time_init)
577 late_time_init();
578 calibrate_delay();
579 pidmap_init();
580 pgtable_cache_init();
581 prio_tree_init();
582 anon_vma_init();
583 #ifdef CONFIG_X86
584 if (efi_enabled)
585 efi_enter_virtual_mode();
586 #endif
587 fork_init(num_physpages);
588 proc_caches_init();
589 buffer_init();
590 unnamed_dev_init();
591 key_init();
592 security_init();
593 vfs_caches_init(num_physpages);
594 radix_tree_init();
595 signals_init();
596 /* rootfs populating might need page-writeback */
597 page_writeback_init();
598 #ifdef CONFIG_PROC_FS
599 proc_root_init();
600 #endif
601 cpuset_init();
602 taskstats_init_early();
603 delayacct_init();
605 check_bugs();
607 acpi_early_init(); /* before LAPIC and SMP init */
609 /* Do the rest non-__init'ed, we're now alive */
610 rest_init();
613 static int __initdata initcall_debug;
615 static int __init initcall_debug_setup(char *str)
617 initcall_debug = 1;
618 return 1;
620 __setup("initcall_debug", initcall_debug_setup);
622 struct task_struct *child_reaper = &init_task;
624 extern initcall_t __initcall_start[], __initcall_end[];
626 static void __init do_initcalls(void)
628 initcall_t *call;
629 int count = preempt_count();
631 for (call = __initcall_start; call < __initcall_end; call++) {
632 char *msg = NULL;
633 char msgbuf[40];
634 int result;
636 if (initcall_debug) {
637 printk("Calling initcall 0x%p", *call);
638 print_fn_descriptor_symbol(": %s()",
639 (unsigned long) *call);
640 printk("\n");
643 result = (*call)();
645 if (result && result != -ENODEV && initcall_debug) {
646 sprintf(msgbuf, "error code %d", result);
647 msg = msgbuf;
649 if (preempt_count() != count) {
650 msg = "preemption imbalance";
651 preempt_count() = count;
653 if (irqs_disabled()) {
654 msg = "disabled interrupts";
655 local_irq_enable();
657 if (msg) {
658 printk(KERN_WARNING "initcall at 0x%p", *call);
659 print_fn_descriptor_symbol(": %s()",
660 (unsigned long) *call);
661 printk(": returned with %s\n", msg);
665 /* Make sure there is no pending stuff from the initcall sequence */
666 flush_scheduled_work();
670 * Ok, the machine is now initialized. None of the devices
671 * have been touched yet, but the CPU subsystem is up and
672 * running, and memory and process management works.
674 * Now we can finally start doing some real work..
676 static void __init do_basic_setup(void)
678 /* drivers will send hotplug events */
679 init_workqueues();
680 usermodehelper_init();
681 driver_init();
683 #ifdef CONFIG_SYSCTL
684 sysctl_init();
685 #endif
687 do_initcalls();
690 static void do_pre_smp_initcalls(void)
692 extern int spawn_ksoftirqd(void);
693 #ifdef CONFIG_SMP
694 extern int migration_init(void);
696 migration_init();
697 #endif
698 spawn_ksoftirqd();
699 spawn_softlockup_task();
702 static void run_init_process(char *init_filename)
704 argv_init[0] = init_filename;
705 kernel_execve(init_filename, argv_init, envp_init);
708 static int init(void * unused)
710 lock_kernel();
712 * init can run on any cpu.
714 set_cpus_allowed(current, CPU_MASK_ALL);
716 * Tell the world that we're going to be the grim
717 * reaper of innocent orphaned children.
719 * We don't want people to have to make incorrect
720 * assumptions about where in the task array this
721 * can be found.
723 child_reaper = current;
725 cad_pid = task_pid(current);
727 smp_prepare_cpus(max_cpus);
729 do_pre_smp_initcalls();
731 smp_init();
732 sched_init_smp();
734 cpuset_init_smp();
737 * Do this before initcalls, because some drivers want to access
738 * firmware files.
740 populate_rootfs();
742 do_basic_setup();
745 * check if there is an early userspace init. If yes, let it do all
746 * the work
749 if (!ramdisk_execute_command)
750 ramdisk_execute_command = "/init";
752 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
753 ramdisk_execute_command = NULL;
754 prepare_namespace();
758 * Ok, we have completed the initial bootup, and
759 * we're essentially up and running. Get rid of the
760 * initmem segments and start the user-mode stuff..
762 free_initmem();
763 unlock_kernel();
764 mark_rodata_ro();
765 system_state = SYSTEM_RUNNING;
766 numa_default_policy();
768 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
769 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
771 (void) sys_dup(0);
772 (void) sys_dup(0);
774 if (ramdisk_execute_command) {
775 run_init_process(ramdisk_execute_command);
776 printk(KERN_WARNING "Failed to execute %s\n",
777 ramdisk_execute_command);
781 * We try each of these until one succeeds.
783 * The Bourne shell can be used instead of init if we are
784 * trying to recover a really broken machine.
786 if (execute_command) {
787 run_init_process(execute_command);
788 printk(KERN_WARNING "Failed to execute %s. Attempting "
789 "defaults...\n", execute_command);
791 run_init_process("/sbin/init");
792 run_init_process("/etc/init");
793 run_init_process("/bin/init");
794 run_init_process("/bin/sh");
796 panic("No init found. Try passing init= option to kernel.");