]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - init/main.c
Merge branch 'devel' of master.kernel.org:/home/rmk/linux-2.6-arm
[mirror_ubuntu-artful-kernel.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 * If set, this is an indication to the drivers that reset the underlying
132 * device before going ahead with the initialization otherwise driver might
133 * rely on the BIOS and skip the reset operation.
134 *
135 * This is useful if kernel is booting in an unreliable environment.
136 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
137 * skipped and devices will be in unknown state.
138 */
139 unsigned int reset_devices;
140 EXPORT_SYMBOL(reset_devices);
141
142 /*
143 * Setup routine for controlling SMP activation
144 *
145 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
146 * activation entirely (the MPS table probe still happens, though).
147 *
148 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
149 * greater than 0, limits the maximum number of CPUs activated in
150 * SMP mode to <NUM>.
151 */
152 static int __init nosmp(char *str)
153 {
154 max_cpus = 0;
155 return 1;
156 }
157
158 __setup("nosmp", nosmp);
159
160 static int __init maxcpus(char *str)
161 {
162 get_option(&str, &max_cpus);
163 return 1;
164 }
165
166 __setup("maxcpus=", maxcpus);
167
168 static int __init set_reset_devices(char *str)
169 {
170 reset_devices = 1;
171 return 1;
172 }
173
174 __setup("reset_devices", set_reset_devices);
175
176 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
177 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
178 static const char *panic_later, *panic_param;
179
180 extern struct obs_kernel_param __setup_start[], __setup_end[];
181
182 static int __init obsolete_checksetup(char *line)
183 {
184 struct obs_kernel_param *p;
185 int had_early_param = 0;
186
187 p = __setup_start;
188 do {
189 int n = strlen(p->str);
190 if (!strncmp(line, p->str, n)) {
191 if (p->early) {
192 /* Already done in parse_early_param?
193 * (Needs exact match on param part).
194 * Keep iterating, as we can have early
195 * params and __setups of same names 8( */
196 if (line[n] == '\0' || line[n] == '=')
197 had_early_param = 1;
198 } else if (!p->setup_func) {
199 printk(KERN_WARNING "Parameter %s is obsolete,"
200 " ignored\n", p->str);
201 return 1;
202 } else if (p->setup_func(line + n))
203 return 1;
204 }
205 p++;
206 } while (p < __setup_end);
207
208 return had_early_param;
209 }
210
211 /*
212 * This should be approx 2 Bo*oMips to start (note initial shift), and will
213 * still work even if initially too large, it will just take slightly longer
214 */
215 unsigned long loops_per_jiffy = (1<<12);
216
217 EXPORT_SYMBOL(loops_per_jiffy);
218
219 static int __init debug_kernel(char *str)
220 {
221 if (*str)
222 return 0;
223 console_loglevel = 10;
224 return 1;
225 }
226
227 static int __init quiet_kernel(char *str)
228 {
229 if (*str)
230 return 0;
231 console_loglevel = 4;
232 return 1;
233 }
234
235 __setup("debug", debug_kernel);
236 __setup("quiet", quiet_kernel);
237
238 static int __init loglevel(char *str)
239 {
240 get_option(&str, &console_loglevel);
241 return 1;
242 }
243
244 __setup("loglevel=", loglevel);
245
246 /*
247 * Unknown boot options get handed to init, unless they look like
248 * failed parameters
249 */
250 static int __init unknown_bootoption(char *param, char *val)
251 {
252 /* Change NUL term back to "=", to make "param" the whole string. */
253 if (val) {
254 /* param=val or param="val"? */
255 if (val == param+strlen(param)+1)
256 val[-1] = '=';
257 else if (val == param+strlen(param)+2) {
258 val[-2] = '=';
259 memmove(val-1, val, strlen(val)+1);
260 val--;
261 } else
262 BUG();
263 }
264
265 /* Handle obsolete-style parameters */
266 if (obsolete_checksetup(param))
267 return 0;
268
269 /*
270 * Preemptive maintenance for "why didn't my mispelled command
271 * line work?"
272 */
273 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
274 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
275 return 0;
276 }
277
278 if (panic_later)
279 return 0;
280
281 if (val) {
282 /* Environment option */
283 unsigned int i;
284 for (i = 0; envp_init[i]; i++) {
285 if (i == MAX_INIT_ENVS) {
286 panic_later = "Too many boot env vars at `%s'";
287 panic_param = param;
288 }
289 if (!strncmp(param, envp_init[i], val - param))
290 break;
291 }
292 envp_init[i] = param;
293 } else {
294 /* Command line option */
295 unsigned int i;
296 for (i = 0; argv_init[i]; i++) {
297 if (i == MAX_INIT_ARGS) {
298 panic_later = "Too many boot init vars at `%s'";
299 panic_param = param;
300 }
301 }
302 argv_init[i] = param;
303 }
304 return 0;
305 }
306
307 static int __init init_setup(char *str)
308 {
309 unsigned int i;
310
311 execute_command = str;
312 /*
313 * In case LILO is going to boot us with default command line,
314 * it prepends "auto" before the whole cmdline which makes
315 * the shell think it should execute a script with such name.
316 * So we ignore all arguments entered _before_ init=... [MJ]
317 */
318 for (i = 1; i < MAX_INIT_ARGS; i++)
319 argv_init[i] = NULL;
320 return 1;
321 }
322 __setup("init=", init_setup);
323
324 static int __init rdinit_setup(char *str)
325 {
326 unsigned int i;
327
328 ramdisk_execute_command = str;
329 /* See "auto" comment in init_setup */
330 for (i = 1; i < MAX_INIT_ARGS; i++)
331 argv_init[i] = NULL;
332 return 1;
333 }
334 __setup("rdinit=", rdinit_setup);
335
336 #ifndef CONFIG_SMP
337
338 #ifdef CONFIG_X86_LOCAL_APIC
339 static void __init smp_init(void)
340 {
341 APIC_init_uniprocessor();
342 }
343 #else
344 #define smp_init() do { } while (0)
345 #endif
346
347 static inline void setup_per_cpu_areas(void) { }
348 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
349
350 #else
351
352 #ifdef __GENERIC_PER_CPU
353 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
354
355 EXPORT_SYMBOL(__per_cpu_offset);
356
357 static void __init setup_per_cpu_areas(void)
358 {
359 unsigned long size, i;
360 char *ptr;
361 unsigned long nr_possible_cpus = num_possible_cpus();
362
363 /* Copy section for each CPU (we discard the original) */
364 size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
365 #ifdef CONFIG_MODULES
366 if (size < PERCPU_ENOUGH_ROOM)
367 size = PERCPU_ENOUGH_ROOM;
368 #endif
369 ptr = alloc_bootmem(size * nr_possible_cpus);
370
371 for_each_possible_cpu(i) {
372 __per_cpu_offset[i] = ptr - __per_cpu_start;
373 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
374 ptr += size;
375 }
376 }
377 #endif /* !__GENERIC_PER_CPU */
378
379 /* Called by boot processor to activate the rest. */
380 static void __init smp_init(void)
381 {
382 unsigned int i;
383
384 /* FIXME: This should be done in userspace --RR */
385 for_each_present_cpu(i) {
386 if (num_online_cpus() >= max_cpus)
387 break;
388 if (!cpu_online(i))
389 cpu_up(i);
390 }
391
392 /* Any cleanup work */
393 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
394 smp_cpus_done(max_cpus);
395 #if 0
396 /* Get other processors into their bootup holding patterns. */
397
398 smp_commence();
399 #endif
400 }
401
402 #endif
403
404 /*
405 * We need to finalize in a non-__init function or else race conditions
406 * between the root thread and the init thread may cause start_kernel to
407 * be reaped by free_initmem before the root thread has proceeded to
408 * cpu_idle.
409 *
410 * gcc-3.4 accidentally inlines this function, so use noinline.
411 */
412
413 static void noinline rest_init(void)
414 __releases(kernel_lock)
415 {
416 kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
417 numa_default_policy();
418 unlock_kernel();
419
420 /*
421 * The boot idle thread must execute schedule()
422 * at least one to get things moving:
423 */
424 preempt_enable_no_resched();
425 schedule();
426 preempt_disable();
427
428 /* Call into cpu_idle with preempt disabled */
429 cpu_idle();
430 }
431
432 /* Check for early params. */
433 static int __init do_early_param(char *param, char *val)
434 {
435 struct obs_kernel_param *p;
436
437 for (p = __setup_start; p < __setup_end; p++) {
438 if (p->early && strcmp(param, p->str) == 0) {
439 if (p->setup_func(val) != 0)
440 printk(KERN_WARNING
441 "Malformed early option '%s'\n", param);
442 }
443 }
444 /* We accept everything at this stage. */
445 return 0;
446 }
447
448 /* Arch code calls this early on, or if not, just before other parsing. */
449 void __init parse_early_param(void)
450 {
451 static __initdata int done = 0;
452 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
453
454 if (done)
455 return;
456
457 /* All fall through to do_early_param. */
458 strlcpy(tmp_cmdline, saved_command_line, COMMAND_LINE_SIZE);
459 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
460 done = 1;
461 }
462
463 /*
464 * Activate the first processor.
465 */
466
467 static void __init boot_cpu_init(void)
468 {
469 int cpu = smp_processor_id();
470 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
471 cpu_set(cpu, cpu_online_map);
472 cpu_set(cpu, cpu_present_map);
473 cpu_set(cpu, cpu_possible_map);
474 }
475
476 void __init __attribute__((weak)) smp_setup_processor_id(void)
477 {
478 }
479
480 asmlinkage void __init start_kernel(void)
481 {
482 char * command_line;
483 extern struct kernel_param __start___param[], __stop___param[];
484
485 smp_setup_processor_id();
486
487 /*
488 * Need to run as early as possible, to initialize the
489 * lockdep hash:
490 */
491 unwind_init();
492 lockdep_init();
493
494 local_irq_disable();
495 early_boot_irqs_off();
496 early_init_irq_lock_class();
497
498 /*
499 * Interrupts are still disabled. Do necessary setups, then
500 * enable them
501 */
502 lock_kernel();
503 boot_cpu_init();
504 page_address_init();
505 printk(KERN_NOTICE);
506 printk(linux_banner);
507 setup_arch(&command_line);
508 setup_per_cpu_areas();
509 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
510
511 /*
512 * Set up the scheduler prior starting any interrupts (such as the
513 * timer interrupt). Full topology setup happens at smp_init()
514 * time - but meanwhile we still have a functioning scheduler.
515 */
516 sched_init();
517 /*
518 * Disable preemption - early bootup scheduling is extremely
519 * fragile until we cpu_idle() for the first time.
520 */
521 preempt_disable();
522 build_all_zonelists();
523 page_alloc_init();
524 printk(KERN_NOTICE "Kernel command line: %s\n", saved_command_line);
525 parse_early_param();
526 parse_args("Booting kernel", command_line, __start___param,
527 __stop___param - __start___param,
528 &unknown_bootoption);
529 sort_main_extable();
530 trap_init();
531 rcu_init();
532 init_IRQ();
533 pidhash_init();
534 init_timers();
535 hrtimers_init();
536 softirq_init();
537 timekeeping_init();
538 time_init();
539 profile_init();
540 if (!irqs_disabled())
541 printk("start_kernel(): bug: interrupts were enabled early\n");
542 early_boot_irqs_on();
543 local_irq_enable();
544
545 /*
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.
549 */
550 console_init();
551 if (panic_later)
552 panic(panic_later, panic_param);
553
554 lockdep_info();
555
556 /*
557 * Need to run this when irqs are enabled, because it wants
558 * to self-test [hard/soft]-irqs on/off lock inversion bugs
559 * too:
560 */
561 locking_selftest();
562
563 #ifdef CONFIG_BLK_DEV_INITRD
564 if (initrd_start && !initrd_below_start_ok &&
565 initrd_start < min_low_pfn << PAGE_SHIFT) {
566 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
567 "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
568 initrd_start = 0;
569 }
570 #endif
571 vfs_caches_init_early();
572 cpuset_init_early();
573 mem_init();
574 kmem_cache_init();
575 setup_per_cpu_pageset();
576 numa_policy_init();
577 if (late_time_init)
578 late_time_init();
579 calibrate_delay();
580 pidmap_init();
581 pgtable_cache_init();
582 prio_tree_init();
583 anon_vma_init();
584 #ifdef CONFIG_X86
585 if (efi_enabled)
586 efi_enter_virtual_mode();
587 #endif
588 fork_init(num_physpages);
589 proc_caches_init();
590 buffer_init();
591 unnamed_dev_init();
592 key_init();
593 security_init();
594 vfs_caches_init(num_physpages);
595 radix_tree_init();
596 signals_init();
597 /* rootfs populating might need page-writeback */
598 page_writeback_init();
599 #ifdef CONFIG_PROC_FS
600 proc_root_init();
601 #endif
602 cpuset_init();
603 taskstats_init_early();
604 delayacct_init();
605
606 check_bugs();
607
608 acpi_early_init(); /* before LAPIC and SMP init */
609
610 /* Do the rest non-__init'ed, we're now alive */
611 rest_init();
612 }
613
614 static int __initdata initcall_debug;
615
616 static int __init initcall_debug_setup(char *str)
617 {
618 initcall_debug = 1;
619 return 1;
620 }
621 __setup("initcall_debug", initcall_debug_setup);
622
623 struct task_struct *child_reaper = &init_task;
624
625 extern initcall_t __initcall_start[], __initcall_end[];
626
627 static void __init do_initcalls(void)
628 {
629 initcall_t *call;
630 int count = preempt_count();
631
632 for (call = __initcall_start; call < __initcall_end; call++) {
633 char *msg = NULL;
634 char msgbuf[40];
635 int result;
636
637 if (initcall_debug) {
638 printk("Calling initcall 0x%p", *call);
639 print_fn_descriptor_symbol(": %s()",
640 (unsigned long) *call);
641 printk("\n");
642 }
643
644 result = (*call)();
645
646 if (result && result != -ENODEV && initcall_debug) {
647 sprintf(msgbuf, "error code %d", result);
648 msg = msgbuf;
649 }
650 if (preempt_count() != count) {
651 msg = "preemption imbalance";
652 preempt_count() = count;
653 }
654 if (irqs_disabled()) {
655 msg = "disabled interrupts";
656 local_irq_enable();
657 }
658 if (msg) {
659 printk(KERN_WARNING "initcall at 0x%p", *call);
660 print_fn_descriptor_symbol(": %s()",
661 (unsigned long) *call);
662 printk(": returned with %s\n", msg);
663 }
664 }
665
666 /* Make sure there is no pending stuff from the initcall sequence */
667 flush_scheduled_work();
668 }
669
670 /*
671 * Ok, the machine is now initialized. None of the devices
672 * have been touched yet, but the CPU subsystem is up and
673 * running, and memory and process management works.
674 *
675 * Now we can finally start doing some real work..
676 */
677 static void __init do_basic_setup(void)
678 {
679 /* drivers will send hotplug events */
680 init_workqueues();
681 usermodehelper_init();
682 driver_init();
683
684 #ifdef CONFIG_SYSCTL
685 sysctl_init();
686 #endif
687
688 do_initcalls();
689 }
690
691 static void do_pre_smp_initcalls(void)
692 {
693 extern int spawn_ksoftirqd(void);
694 #ifdef CONFIG_SMP
695 extern int migration_init(void);
696
697 migration_init();
698 #endif
699 spawn_ksoftirqd();
700 spawn_softlockup_task();
701 }
702
703 static void run_init_process(char *init_filename)
704 {
705 argv_init[0] = init_filename;
706 execve(init_filename, argv_init, envp_init);
707 }
708
709 static int init(void * unused)
710 {
711 lock_kernel();
712 /*
713 * init can run on any cpu.
714 */
715 set_cpus_allowed(current, CPU_MASK_ALL);
716 /*
717 * Tell the world that we're going to be the grim
718 * reaper of innocent orphaned children.
719 *
720 * We don't want people to have to make incorrect
721 * assumptions about where in the task array this
722 * can be found.
723 */
724 child_reaper = current;
725
726 smp_prepare_cpus(max_cpus);
727
728 do_pre_smp_initcalls();
729
730 smp_init();
731 sched_init_smp();
732
733 cpuset_init_smp();
734
735 /*
736 * Do this before initcalls, because some drivers want to access
737 * firmware files.
738 */
739 populate_rootfs();
740
741 do_basic_setup();
742
743 /*
744 * check if there is an early userspace init. If yes, let it do all
745 * the work
746 */
747
748 if (!ramdisk_execute_command)
749 ramdisk_execute_command = "/init";
750
751 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
752 ramdisk_execute_command = NULL;
753 prepare_namespace();
754 }
755
756 /*
757 * Ok, we have completed the initial bootup, and
758 * we're essentially up and running. Get rid of the
759 * initmem segments and start the user-mode stuff..
760 */
761 free_initmem();
762 unlock_kernel();
763 mark_rodata_ro();
764 system_state = SYSTEM_RUNNING;
765 numa_default_policy();
766
767 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
768 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
769
770 (void) sys_dup(0);
771 (void) sys_dup(0);
772
773 if (ramdisk_execute_command) {
774 run_init_process(ramdisk_execute_command);
775 printk(KERN_WARNING "Failed to execute %s\n",
776 ramdisk_execute_command);
777 }
778
779 /*
780 * We try each of these until one succeeds.
781 *
782 * The Bourne shell can be used instead of init if we are
783 * trying to recover a really broken machine.
784 */
785 if (execute_command) {
786 run_init_process(execute_command);
787 printk(KERN_WARNING "Failed to execute %s. Attempting "
788 "defaults...\n", execute_command);
789 }
790 run_init_process("/sbin/init");
791 run_init_process("/etc/init");
792 run_init_process("/bin/init");
793 run_init_process("/bin/sh");
794
795 panic("No init found. Try passing init= option to kernel.");
796 }