[MIPS] Simplify dump_stack()
[deliverable/linux.git] / init / main.c
1 /*
2 * linux/init/main.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 *
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>
10 */
11
12 #define __KERNEL_SYSCALLS__
13
14 #include <linux/types.h>
15 #include <linux/module.h>
16 #include <linux/proc_fs.h>
17 #include <linux/kernel.h>
18 #include <linux/syscalls.h>
19 #include <linux/string.h>
20 #include <linux/ctype.h>
21 #include <linux/delay.h>
22 #include <linux/utsname.h>
23 #include <linux/ioport.h>
24 #include <linux/init.h>
25 #include <linux/smp_lock.h>
26 #include <linux/initrd.h>
27 #include <linux/hdreg.h>
28 #include <linux/bootmem.h>
29 #include <linux/tty.h>
30 #include <linux/gfp.h>
31 #include <linux/percpu.h>
32 #include <linux/kmod.h>
33 #include <linux/kernel_stat.h>
34 #include <linux/security.h>
35 #include <linux/workqueue.h>
36 #include <linux/profile.h>
37 #include <linux/rcupdate.h>
38 #include <linux/moduleparam.h>
39 #include <linux/kallsyms.h>
40 #include <linux/writeback.h>
41 #include <linux/cpu.h>
42 #include <linux/cpuset.h>
43 #include <linux/efi.h>
44 #include <linux/taskstats_kern.h>
45 #include <linux/delayacct.h>
46 #include <linux/unistd.h>
47 #include <linux/rmap.h>
48 #include <linux/mempolicy.h>
49 #include <linux/key.h>
50 #include <linux/unwind.h>
51 #include <linux/buffer_head.h>
52 #include <linux/debug_locks.h>
53 #include <linux/lockdep.h>
54
55 #include <asm/io.h>
56 #include <asm/bugs.h>
57 #include <asm/setup.h>
58 #include <asm/sections.h>
59 #include <asm/cacheflush.h>
60
61 #ifdef CONFIG_X86_LOCAL_APIC
62 #include <asm/smp.h>
63 #endif
64
65 /*
66 * This is one of the first .c files built. Error out early if we have compiler
67 * trouble.
68 *
69 * Versions of gcc older than that listed below may actually compile and link
70 * okay, but the end product can have subtle run time bugs. To avoid associated
71 * bogus bug reports, we flatly refuse to compile with a gcc that is known to be
72 * too old from the very beginning.
73 */
74 #if (__GNUC__ < 3) || (__GNUC__ == 3 && __GNUC_MINOR__ < 2)
75 #error Sorry, your GCC is too old. It builds incorrect kernels.
76 #endif
77
78 static int init(void *);
79
80 extern void init_IRQ(void);
81 extern void fork_init(unsigned long);
82 extern void mca_init(void);
83 extern void sbus_init(void);
84 extern void sysctl_init(void);
85 extern void signals_init(void);
86 extern void pidhash_init(void);
87 extern void pidmap_init(void);
88 extern void prio_tree_init(void);
89 extern void radix_tree_init(void);
90 extern void free_initmem(void);
91 extern void populate_rootfs(void);
92 extern void driver_init(void);
93 extern void prepare_namespace(void);
94 #ifdef CONFIG_ACPI
95 extern void acpi_early_init(void);
96 #else
97 static inline void acpi_early_init(void) { }
98 #endif
99 #ifndef CONFIG_DEBUG_RODATA
100 static inline void mark_rodata_ro(void) { }
101 #endif
102
103 #ifdef CONFIG_TC
104 extern void tc_init(void);
105 #endif
106
107 enum system_states system_state;
108 EXPORT_SYMBOL(system_state);
109
110 /*
111 * Boot command-line arguments
112 */
113 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
114 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
115
116 extern void time_init(void);
117 /* Default late time init is NULL. archs can override this later. */
118 void (*late_time_init)(void);
119 extern void softirq_init(void);
120
121 /* Untouched command line (eg. for /proc) saved by arch-specific code. */
122 char saved_command_line[COMMAND_LINE_SIZE];
123
124 static char *execute_command;
125 static char *ramdisk_execute_command;
126
127 /* Setup configured maximum number of CPUs to activate */
128 static unsigned int max_cpus = NR_CPUS;
129
130 /*
131 * Setup routine for controlling SMP activation
132 *
133 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
134 * activation entirely (the MPS table probe still happens, though).
135 *
136 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
137 * greater than 0, limits the maximum number of CPUs activated in
138 * SMP mode to <NUM>.
139 */
140 static int __init nosmp(char *str)
141 {
142 max_cpus = 0;
143 return 1;
144 }
145
146 __setup("nosmp", nosmp);
147
148 static int __init maxcpus(char *str)
149 {
150 get_option(&str, &max_cpus);
151 return 1;
152 }
153
154 __setup("maxcpus=", maxcpus);
155
156 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
157 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
158 static const char *panic_later, *panic_param;
159
160 extern struct obs_kernel_param __setup_start[], __setup_end[];
161
162 static int __init obsolete_checksetup(char *line)
163 {
164 struct obs_kernel_param *p;
165 int had_early_param = 0;
166
167 p = __setup_start;
168 do {
169 int n = strlen(p->str);
170 if (!strncmp(line, p->str, n)) {
171 if (p->early) {
172 /* Already done in parse_early_param?
173 * (Needs exact match on param part).
174 * Keep iterating, as we can have early
175 * params and __setups of same names 8( */
176 if (line[n] == '\0' || line[n] == '=')
177 had_early_param = 1;
178 } else if (!p->setup_func) {
179 printk(KERN_WARNING "Parameter %s is obsolete,"
180 " ignored\n", p->str);
181 return 1;
182 } else if (p->setup_func(line + n))
183 return 1;
184 }
185 p++;
186 } while (p < __setup_end);
187
188 return had_early_param;
189 }
190
191 /*
192 * This should be approx 2 Bo*oMips to start (note initial shift), and will
193 * still work even if initially too large, it will just take slightly longer
194 */
195 unsigned long loops_per_jiffy = (1<<12);
196
197 EXPORT_SYMBOL(loops_per_jiffy);
198
199 static int __init debug_kernel(char *str)
200 {
201 if (*str)
202 return 0;
203 console_loglevel = 10;
204 return 1;
205 }
206
207 static int __init quiet_kernel(char *str)
208 {
209 if (*str)
210 return 0;
211 console_loglevel = 4;
212 return 1;
213 }
214
215 __setup("debug", debug_kernel);
216 __setup("quiet", quiet_kernel);
217
218 static int __init loglevel(char *str)
219 {
220 get_option(&str, &console_loglevel);
221 return 1;
222 }
223
224 __setup("loglevel=", loglevel);
225
226 /*
227 * Unknown boot options get handed to init, unless they look like
228 * failed parameters
229 */
230 static int __init unknown_bootoption(char *param, char *val)
231 {
232 /* Change NUL term back to "=", to make "param" the whole string. */
233 if (val) {
234 /* param=val or param="val"? */
235 if (val == param+strlen(param)+1)
236 val[-1] = '=';
237 else if (val == param+strlen(param)+2) {
238 val[-2] = '=';
239 memmove(val-1, val, strlen(val)+1);
240 val--;
241 } else
242 BUG();
243 }
244
245 /* Handle obsolete-style parameters */
246 if (obsolete_checksetup(param))
247 return 0;
248
249 /*
250 * Preemptive maintenance for "why didn't my mispelled command
251 * line work?"
252 */
253 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
254 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
255 return 0;
256 }
257
258 if (panic_later)
259 return 0;
260
261 if (val) {
262 /* Environment option */
263 unsigned int i;
264 for (i = 0; envp_init[i]; i++) {
265 if (i == MAX_INIT_ENVS) {
266 panic_later = "Too many boot env vars at `%s'";
267 panic_param = param;
268 }
269 if (!strncmp(param, envp_init[i], val - param))
270 break;
271 }
272 envp_init[i] = param;
273 } else {
274 /* Command line option */
275 unsigned int i;
276 for (i = 0; argv_init[i]; i++) {
277 if (i == MAX_INIT_ARGS) {
278 panic_later = "Too many boot init vars at `%s'";
279 panic_param = param;
280 }
281 }
282 argv_init[i] = param;
283 }
284 return 0;
285 }
286
287 static int __init init_setup(char *str)
288 {
289 unsigned int i;
290
291 execute_command = str;
292 /*
293 * In case LILO is going to boot us with default command line,
294 * it prepends "auto" before the whole cmdline which makes
295 * the shell think it should execute a script with such name.
296 * So we ignore all arguments entered _before_ init=... [MJ]
297 */
298 for (i = 1; i < MAX_INIT_ARGS; i++)
299 argv_init[i] = NULL;
300 return 1;
301 }
302 __setup("init=", init_setup);
303
304 static int __init rdinit_setup(char *str)
305 {
306 unsigned int i;
307
308 ramdisk_execute_command = str;
309 /* See "auto" comment in init_setup */
310 for (i = 1; i < MAX_INIT_ARGS; i++)
311 argv_init[i] = NULL;
312 return 1;
313 }
314 __setup("rdinit=", rdinit_setup);
315
316 #ifndef CONFIG_SMP
317
318 #ifdef CONFIG_X86_LOCAL_APIC
319 static void __init smp_init(void)
320 {
321 APIC_init_uniprocessor();
322 }
323 #else
324 #define smp_init() do { } while (0)
325 #endif
326
327 static inline void setup_per_cpu_areas(void) { }
328 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
329
330 #else
331
332 #ifdef __GENERIC_PER_CPU
333 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
334
335 EXPORT_SYMBOL(__per_cpu_offset);
336
337 static void __init setup_per_cpu_areas(void)
338 {
339 unsigned long size, i;
340 char *ptr;
341 unsigned long nr_possible_cpus = num_possible_cpus();
342
343 /* Copy section for each CPU (we discard the original) */
344 size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
345 #ifdef CONFIG_MODULES
346 if (size < PERCPU_ENOUGH_ROOM)
347 size = PERCPU_ENOUGH_ROOM;
348 #endif
349 ptr = alloc_bootmem(size * nr_possible_cpus);
350
351 for_each_possible_cpu(i) {
352 __per_cpu_offset[i] = ptr - __per_cpu_start;
353 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
354 ptr += size;
355 }
356 }
357 #endif /* !__GENERIC_PER_CPU */
358
359 /* Called by boot processor to activate the rest. */
360 static void __init smp_init(void)
361 {
362 unsigned int i;
363
364 /* FIXME: This should be done in userspace --RR */
365 for_each_present_cpu(i) {
366 if (num_online_cpus() >= max_cpus)
367 break;
368 if (!cpu_online(i))
369 cpu_up(i);
370 }
371
372 /* Any cleanup work */
373 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
374 smp_cpus_done(max_cpus);
375 #if 0
376 /* Get other processors into their bootup holding patterns. */
377
378 smp_commence();
379 #endif
380 }
381
382 #endif
383
384 /*
385 * We need to finalize in a non-__init function or else race conditions
386 * between the root thread and the init thread may cause start_kernel to
387 * be reaped by free_initmem before the root thread has proceeded to
388 * cpu_idle.
389 *
390 * gcc-3.4 accidentally inlines this function, so use noinline.
391 */
392
393 static void noinline rest_init(void)
394 __releases(kernel_lock)
395 {
396 kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
397 numa_default_policy();
398 unlock_kernel();
399
400 /*
401 * The boot idle thread must execute schedule()
402 * at least one to get things moving:
403 */
404 preempt_enable_no_resched();
405 schedule();
406 preempt_disable();
407
408 /* Call into cpu_idle with preempt disabled */
409 cpu_idle();
410 }
411
412 /* Check for early params. */
413 static int __init do_early_param(char *param, char *val)
414 {
415 struct obs_kernel_param *p;
416
417 for (p = __setup_start; p < __setup_end; p++) {
418 if (p->early && strcmp(param, p->str) == 0) {
419 if (p->setup_func(val) != 0)
420 printk(KERN_WARNING
421 "Malformed early option '%s'\n", param);
422 }
423 }
424 /* We accept everything at this stage. */
425 return 0;
426 }
427
428 /* Arch code calls this early on, or if not, just before other parsing. */
429 void __init parse_early_param(void)
430 {
431 static __initdata int done = 0;
432 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
433
434 if (done)
435 return;
436
437 /* All fall through to do_early_param. */
438 strlcpy(tmp_cmdline, saved_command_line, COMMAND_LINE_SIZE);
439 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
440 done = 1;
441 }
442
443 /*
444 * Activate the first processor.
445 */
446
447 static void __init boot_cpu_init(void)
448 {
449 int cpu = smp_processor_id();
450 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
451 cpu_set(cpu, cpu_online_map);
452 cpu_set(cpu, cpu_present_map);
453 cpu_set(cpu, cpu_possible_map);
454 }
455
456 void __init __attribute__((weak)) smp_setup_processor_id(void)
457 {
458 }
459
460 asmlinkage void __init start_kernel(void)
461 {
462 char * command_line;
463 extern struct kernel_param __start___param[], __stop___param[];
464
465 smp_setup_processor_id();
466
467 /*
468 * Need to run as early as possible, to initialize the
469 * lockdep hash:
470 */
471 unwind_init();
472 lockdep_init();
473
474 local_irq_disable();
475 early_boot_irqs_off();
476 early_init_irq_lock_class();
477
478 /*
479 * Interrupts are still disabled. Do necessary setups, then
480 * enable them
481 */
482 lock_kernel();
483 boot_cpu_init();
484 page_address_init();
485 printk(KERN_NOTICE);
486 printk(linux_banner);
487 setup_arch(&command_line);
488 setup_per_cpu_areas();
489 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
490
491 /*
492 * Set up the scheduler prior starting any interrupts (such as the
493 * timer interrupt). Full topology setup happens at smp_init()
494 * time - but meanwhile we still have a functioning scheduler.
495 */
496 sched_init();
497 /*
498 * Disable preemption - early bootup scheduling is extremely
499 * fragile until we cpu_idle() for the first time.
500 */
501 preempt_disable();
502 build_all_zonelists();
503 page_alloc_init();
504 printk(KERN_NOTICE "Kernel command line: %s\n", saved_command_line);
505 parse_early_param();
506 parse_args("Booting kernel", command_line, __start___param,
507 __stop___param - __start___param,
508 &unknown_bootoption);
509 sort_main_extable();
510 trap_init();
511 rcu_init();
512 init_IRQ();
513 pidhash_init();
514 init_timers();
515 hrtimers_init();
516 softirq_init();
517 timekeeping_init();
518 time_init();
519 profile_init();
520 if (!irqs_disabled())
521 printk("start_kernel(): bug: interrupts were enabled early\n");
522 early_boot_irqs_on();
523 local_irq_enable();
524
525 /*
526 * HACK ALERT! This is early. We're enabling the console before
527 * we've done PCI setups etc, and console_init() must be aware of
528 * this. But we do want output early, in case something goes wrong.
529 */
530 console_init();
531 if (panic_later)
532 panic(panic_later, panic_param);
533
534 lockdep_info();
535
536 /*
537 * Need to run this when irqs are enabled, because it wants
538 * to self-test [hard/soft]-irqs on/off lock inversion bugs
539 * too:
540 */
541 locking_selftest();
542
543 #ifdef CONFIG_BLK_DEV_INITRD
544 if (initrd_start && !initrd_below_start_ok &&
545 initrd_start < min_low_pfn << PAGE_SHIFT) {
546 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
547 "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
548 initrd_start = 0;
549 }
550 #endif
551 vfs_caches_init_early();
552 cpuset_init_early();
553 mem_init();
554 kmem_cache_init();
555 setup_per_cpu_pageset();
556 numa_policy_init();
557 if (late_time_init)
558 late_time_init();
559 calibrate_delay();
560 pidmap_init();
561 pgtable_cache_init();
562 prio_tree_init();
563 anon_vma_init();
564 #ifdef CONFIG_X86
565 if (efi_enabled)
566 efi_enter_virtual_mode();
567 #endif
568 fork_init(num_physpages);
569 proc_caches_init();
570 buffer_init();
571 unnamed_dev_init();
572 key_init();
573 security_init();
574 vfs_caches_init(num_physpages);
575 radix_tree_init();
576 signals_init();
577 /* rootfs populating might need page-writeback */
578 page_writeback_init();
579 #ifdef CONFIG_PROC_FS
580 proc_root_init();
581 #endif
582 cpuset_init();
583 taskstats_init_early();
584 delayacct_init();
585
586 check_bugs();
587
588 acpi_early_init(); /* before LAPIC and SMP init */
589
590 /* Do the rest non-__init'ed, we're now alive */
591 rest_init();
592 }
593
594 static int __initdata initcall_debug;
595
596 static int __init initcall_debug_setup(char *str)
597 {
598 initcall_debug = 1;
599 return 1;
600 }
601 __setup("initcall_debug", initcall_debug_setup);
602
603 struct task_struct *child_reaper = &init_task;
604
605 extern initcall_t __initcall_start[], __initcall_end[];
606
607 static void __init do_initcalls(void)
608 {
609 initcall_t *call;
610 int count = preempt_count();
611
612 for (call = __initcall_start; call < __initcall_end; call++) {
613 char *msg = NULL;
614 char msgbuf[40];
615 int result;
616
617 if (initcall_debug) {
618 printk("Calling initcall 0x%p", *call);
619 print_fn_descriptor_symbol(": %s()",
620 (unsigned long) *call);
621 printk("\n");
622 }
623
624 result = (*call)();
625
626 if (result && result != -ENODEV && initcall_debug) {
627 sprintf(msgbuf, "error code %d", result);
628 msg = msgbuf;
629 }
630 if (preempt_count() != count) {
631 msg = "preemption imbalance";
632 preempt_count() = count;
633 }
634 if (irqs_disabled()) {
635 msg = "disabled interrupts";
636 local_irq_enable();
637 }
638 if (msg) {
639 printk(KERN_WARNING "initcall at 0x%p", *call);
640 print_fn_descriptor_symbol(": %s()",
641 (unsigned long) *call);
642 printk(": returned with %s\n", msg);
643 }
644 }
645
646 /* Make sure there is no pending stuff from the initcall sequence */
647 flush_scheduled_work();
648 }
649
650 /*
651 * Ok, the machine is now initialized. None of the devices
652 * have been touched yet, but the CPU subsystem is up and
653 * running, and memory and process management works.
654 *
655 * Now we can finally start doing some real work..
656 */
657 static void __init do_basic_setup(void)
658 {
659 /* drivers will send hotplug events */
660 init_workqueues();
661 usermodehelper_init();
662 driver_init();
663
664 #ifdef CONFIG_SYSCTL
665 sysctl_init();
666 #endif
667
668 do_initcalls();
669 }
670
671 static void do_pre_smp_initcalls(void)
672 {
673 extern int spawn_ksoftirqd(void);
674 #ifdef CONFIG_SMP
675 extern int migration_init(void);
676
677 migration_init();
678 #endif
679 spawn_ksoftirqd();
680 spawn_softlockup_task();
681 }
682
683 static void run_init_process(char *init_filename)
684 {
685 argv_init[0] = init_filename;
686 execve(init_filename, argv_init, envp_init);
687 }
688
689 static int init(void * unused)
690 {
691 lock_kernel();
692 /*
693 * init can run on any cpu.
694 */
695 set_cpus_allowed(current, CPU_MASK_ALL);
696 /*
697 * Tell the world that we're going to be the grim
698 * reaper of innocent orphaned children.
699 *
700 * We don't want people to have to make incorrect
701 * assumptions about where in the task array this
702 * can be found.
703 */
704 child_reaper = current;
705
706 smp_prepare_cpus(max_cpus);
707
708 do_pre_smp_initcalls();
709
710 smp_init();
711 sched_init_smp();
712
713 cpuset_init_smp();
714
715 /*
716 * Do this before initcalls, because some drivers want to access
717 * firmware files.
718 */
719 populate_rootfs();
720
721 do_basic_setup();
722
723 /*
724 * check if there is an early userspace init. If yes, let it do all
725 * the work
726 */
727
728 if (!ramdisk_execute_command)
729 ramdisk_execute_command = "/init";
730
731 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
732 ramdisk_execute_command = NULL;
733 prepare_namespace();
734 }
735
736 /*
737 * Ok, we have completed the initial bootup, and
738 * we're essentially up and running. Get rid of the
739 * initmem segments and start the user-mode stuff..
740 */
741 free_initmem();
742 unlock_kernel();
743 mark_rodata_ro();
744 system_state = SYSTEM_RUNNING;
745 numa_default_policy();
746
747 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
748 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
749
750 (void) sys_dup(0);
751 (void) sys_dup(0);
752
753 if (ramdisk_execute_command) {
754 run_init_process(ramdisk_execute_command);
755 printk(KERN_WARNING "Failed to execute %s\n",
756 ramdisk_execute_command);
757 }
758
759 /*
760 * We try each of these until one succeeds.
761 *
762 * The Bourne shell can be used instead of init if we are
763 * trying to recover a really broken machine.
764 */
765 if (execute_command) {
766 run_init_process(execute_command);
767 printk(KERN_WARNING "Failed to execute %s. Attempting "
768 "defaults...\n", execute_command);
769 }
770 run_init_process("/sbin/init");
771 run_init_process("/etc/init");
772 run_init_process("/bin/init");
773 run_init_process("/bin/sh");
774
775 panic("No init found. Try passing init= option to kernel.");
776 }
This page took 0.066189 seconds and 5 git commands to generate.