[PATCH] flat binary loader doesn't check fd table full
[linux/fpc-iii.git] / init / main.c
blob006dcd547dc2e72bd3280ac31e746d34da1970bf
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 #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/cpuset.h>
45 #include <linux/efi.h>
46 #include <linux/unistd.h>
47 #include <linux/rmap.h>
48 #include <linux/mempolicy.h>
49 #include <linux/key.h>
51 #include <asm/io.h>
52 #include <asm/bugs.h>
53 #include <asm/setup.h>
54 #include <asm/sections.h>
55 #include <asm/cacheflush.h>
57 #ifdef CONFIG_X86_LOCAL_APIC
58 #include <asm/smp.h>
59 #endif
62 * This is one of the first .c files built. Error out early if we have compiler
63 * trouble.
65 * Versions of gcc older than that listed below may actually compile and link
66 * okay, but the end product can have subtle run time bugs. To avoid associated
67 * bogus bug reports, we flatly refuse to compile with a gcc that is known to be
68 * too old from the very beginning.
70 #if (__GNUC__ < 3) || (__GNUC__ == 3 && __GNUC_MINOR__ < 2)
71 #error Sorry, your GCC is too old. It builds incorrect kernels.
72 #endif
74 static int init(void *);
76 extern void init_IRQ(void);
77 extern void fork_init(unsigned long);
78 extern void mca_init(void);
79 extern void sbus_init(void);
80 extern void sysctl_init(void);
81 extern void signals_init(void);
82 extern void buffer_init(void);
83 extern void pidhash_init(void);
84 extern void pidmap_init(void);
85 extern void prio_tree_init(void);
86 extern void radix_tree_init(void);
87 extern void free_initmem(void);
88 extern void populate_rootfs(void);
89 extern void driver_init(void);
90 extern void prepare_namespace(void);
91 #ifdef CONFIG_ACPI
92 extern void acpi_early_init(void);
93 #else
94 static inline void acpi_early_init(void) { }
95 #endif
96 #ifndef CONFIG_DEBUG_RODATA
97 static inline void mark_rodata_ro(void) { }
98 #endif
100 #ifdef CONFIG_TC
101 extern void tc_init(void);
102 #endif
104 enum system_states system_state;
105 EXPORT_SYMBOL(system_state);
108 * Boot command-line arguments
110 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
111 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
113 extern void time_init(void);
114 /* Default late time init is NULL. archs can override this later. */
115 void (*late_time_init)(void);
116 extern void softirq_init(void);
118 /* Untouched command line (eg. for /proc) saved by arch-specific code. */
119 char saved_command_line[COMMAND_LINE_SIZE];
121 static char *execute_command;
122 static char *ramdisk_execute_command;
124 /* Setup configured maximum number of CPUs to activate */
125 static unsigned int max_cpus = NR_CPUS;
128 * Setup routine for controlling SMP activation
130 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
131 * activation entirely (the MPS table probe still happens, though).
133 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
134 * greater than 0, limits the maximum number of CPUs activated in
135 * SMP mode to <NUM>.
137 static int __init nosmp(char *str)
139 max_cpus = 0;
140 return 1;
143 __setup("nosmp", nosmp);
145 static int __init maxcpus(char *str)
147 get_option(&str, &max_cpus);
148 return 1;
151 __setup("maxcpus=", maxcpus);
153 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
154 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
155 static const char *panic_later, *panic_param;
157 extern struct obs_kernel_param __setup_start[], __setup_end[];
159 static int __init obsolete_checksetup(char *line)
161 struct obs_kernel_param *p;
163 p = __setup_start;
164 do {
165 int n = strlen(p->str);
166 if (!strncmp(line, p->str, n)) {
167 if (p->early) {
168 /* Already done in parse_early_param? (Needs
169 * exact match on param part) */
170 if (line[n] == '\0' || line[n] == '=')
171 return 1;
172 } else if (!p->setup_func) {
173 printk(KERN_WARNING "Parameter %s is obsolete,"
174 " ignored\n", p->str);
175 return 1;
176 } else if (p->setup_func(line + n))
177 return 1;
179 p++;
180 } while (p < __setup_end);
181 return 0;
185 * This should be approx 2 Bo*oMips to start (note initial shift), and will
186 * still work even if initially too large, it will just take slightly longer
188 unsigned long loops_per_jiffy = (1<<12);
190 EXPORT_SYMBOL(loops_per_jiffy);
192 static int __init debug_kernel(char *str)
194 if (*str)
195 return 0;
196 console_loglevel = 10;
197 return 1;
200 static int __init quiet_kernel(char *str)
202 if (*str)
203 return 0;
204 console_loglevel = 4;
205 return 1;
208 __setup("debug", debug_kernel);
209 __setup("quiet", quiet_kernel);
211 static int __init loglevel(char *str)
213 get_option(&str, &console_loglevel);
214 return 1;
217 __setup("loglevel=", loglevel);
220 * Unknown boot options get handed to init, unless they look like
221 * failed parameters
223 static int __init unknown_bootoption(char *param, char *val)
225 /* Change NUL term back to "=", to make "param" the whole string. */
226 if (val) {
227 /* param=val or param="val"? */
228 if (val == param+strlen(param)+1)
229 val[-1] = '=';
230 else if (val == param+strlen(param)+2) {
231 val[-2] = '=';
232 memmove(val-1, val, strlen(val)+1);
233 val--;
234 } else
235 BUG();
238 /* Handle obsolete-style parameters */
239 if (obsolete_checksetup(param))
240 return 0;
243 * Preemptive maintenance for "why didn't my mispelled command
244 * line work?"
246 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
247 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
248 return 0;
251 if (panic_later)
252 return 0;
254 if (val) {
255 /* Environment option */
256 unsigned int i;
257 for (i = 0; envp_init[i]; i++) {
258 if (i == MAX_INIT_ENVS) {
259 panic_later = "Too many boot env vars at `%s'";
260 panic_param = param;
262 if (!strncmp(param, envp_init[i], val - param))
263 break;
265 envp_init[i] = param;
266 } else {
267 /* Command line option */
268 unsigned int i;
269 for (i = 0; argv_init[i]; i++) {
270 if (i == MAX_INIT_ARGS) {
271 panic_later = "Too many boot init vars at `%s'";
272 panic_param = param;
275 argv_init[i] = param;
277 return 0;
280 static int __init init_setup(char *str)
282 unsigned int i;
284 execute_command = str;
286 * In case LILO is going to boot us with default command line,
287 * it prepends "auto" before the whole cmdline which makes
288 * the shell think it should execute a script with such name.
289 * So we ignore all arguments entered _before_ init=... [MJ]
291 for (i = 1; i < MAX_INIT_ARGS; i++)
292 argv_init[i] = NULL;
293 return 1;
295 __setup("init=", init_setup);
297 static int __init rdinit_setup(char *str)
299 unsigned int i;
301 ramdisk_execute_command = str;
302 /* See "auto" comment in init_setup */
303 for (i = 1; i < MAX_INIT_ARGS; i++)
304 argv_init[i] = NULL;
305 return 1;
307 __setup("rdinit=", rdinit_setup);
309 #ifndef CONFIG_SMP
311 #ifdef CONFIG_X86_LOCAL_APIC
312 static void __init smp_init(void)
314 APIC_init_uniprocessor();
316 #else
317 #define smp_init() do { } while (0)
318 #endif
320 static inline void setup_per_cpu_areas(void) { }
321 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
323 #else
325 #ifdef __GENERIC_PER_CPU
326 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
328 EXPORT_SYMBOL(__per_cpu_offset);
330 static void __init setup_per_cpu_areas(void)
332 unsigned long size, i;
333 char *ptr;
334 unsigned long nr_possible_cpus = num_possible_cpus();
336 /* Copy section for each CPU (we discard the original) */
337 size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
338 #ifdef CONFIG_MODULES
339 if (size < PERCPU_ENOUGH_ROOM)
340 size = PERCPU_ENOUGH_ROOM;
341 #endif
342 ptr = alloc_bootmem(size * nr_possible_cpus);
344 for_each_cpu(i) {
345 __per_cpu_offset[i] = ptr - __per_cpu_start;
346 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
347 ptr += size;
350 #endif /* !__GENERIC_PER_CPU */
352 /* Called by boot processor to activate the rest. */
353 static void __init smp_init(void)
355 unsigned int i;
357 /* FIXME: This should be done in userspace --RR */
358 for_each_present_cpu(i) {
359 if (num_online_cpus() >= max_cpus)
360 break;
361 if (!cpu_online(i))
362 cpu_up(i);
365 /* Any cleanup work */
366 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
367 smp_cpus_done(max_cpus);
368 #if 0
369 /* Get other processors into their bootup holding patterns. */
371 smp_commence();
372 #endif
375 #endif
378 * We need to finalize in a non-__init function or else race conditions
379 * between the root thread and the init thread may cause start_kernel to
380 * be reaped by free_initmem before the root thread has proceeded to
381 * cpu_idle.
383 * gcc-3.4 accidentally inlines this function, so use noinline.
386 static void noinline rest_init(void)
387 __releases(kernel_lock)
389 kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
390 numa_default_policy();
391 unlock_kernel();
394 * The boot idle thread must execute schedule()
395 * at least one to get things moving:
397 preempt_enable_no_resched();
398 schedule();
399 preempt_disable();
401 /* Call into cpu_idle with preempt disabled */
402 cpu_idle();
405 /* Check for early params. */
406 static int __init do_early_param(char *param, char *val)
408 struct obs_kernel_param *p;
410 for (p = __setup_start; p < __setup_end; p++) {
411 if (p->early && strcmp(param, p->str) == 0) {
412 if (p->setup_func(val) != 0)
413 printk(KERN_WARNING
414 "Malformed early option '%s'\n", param);
417 /* We accept everything at this stage. */
418 return 0;
421 /* Arch code calls this early on, or if not, just before other parsing. */
422 void __init parse_early_param(void)
424 static __initdata int done = 0;
425 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
427 if (done)
428 return;
430 /* All fall through to do_early_param. */
431 strlcpy(tmp_cmdline, saved_command_line, COMMAND_LINE_SIZE);
432 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
433 done = 1;
437 * Activate the first processor.
440 static void __init boot_cpu_init(void)
442 int cpu = smp_processor_id();
443 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
444 cpu_set(cpu, cpu_online_map);
445 cpu_set(cpu, cpu_present_map);
446 cpu_set(cpu, cpu_possible_map);
449 asmlinkage void __init start_kernel(void)
451 char * command_line;
452 extern struct kernel_param __start___param[], __stop___param[];
454 * Interrupts are still disabled. Do necessary setups, then
455 * enable them
457 lock_kernel();
458 boot_cpu_init();
459 page_address_init();
460 printk(KERN_NOTICE);
461 printk(linux_banner);
462 setup_arch(&command_line);
463 setup_per_cpu_areas();
464 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
467 * Set up the scheduler prior starting any interrupts (such as the
468 * timer interrupt). Full topology setup happens at smp_init()
469 * time - but meanwhile we still have a functioning scheduler.
471 sched_init();
473 * Disable preemption - early bootup scheduling is extremely
474 * fragile until we cpu_idle() for the first time.
476 preempt_disable();
477 build_all_zonelists();
478 page_alloc_init();
479 printk(KERN_NOTICE "Kernel command line: %s\n", saved_command_line);
480 parse_early_param();
481 parse_args("Booting kernel", command_line, __start___param,
482 __stop___param - __start___param,
483 &unknown_bootoption);
484 sort_main_extable();
485 trap_init();
486 rcu_init();
487 init_IRQ();
488 pidhash_init();
489 init_timers();
490 hrtimers_init();
491 softirq_init();
492 time_init();
495 * HACK ALERT! This is early. We're enabling the console before
496 * we've done PCI setups etc, and console_init() must be aware of
497 * this. But we do want output early, in case something goes wrong.
499 console_init();
500 if (panic_later)
501 panic(panic_later, panic_param);
502 profile_init();
503 local_irq_enable();
504 #ifdef CONFIG_BLK_DEV_INITRD
505 if (initrd_start && !initrd_below_start_ok &&
506 initrd_start < min_low_pfn << PAGE_SHIFT) {
507 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
508 "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
509 initrd_start = 0;
511 #endif
512 vfs_caches_init_early();
513 cpuset_init_early();
514 mem_init();
515 kmem_cache_init();
516 setup_per_cpu_pageset();
517 numa_policy_init();
518 if (late_time_init)
519 late_time_init();
520 calibrate_delay();
521 pidmap_init();
522 pgtable_cache_init();
523 prio_tree_init();
524 anon_vma_init();
525 #ifdef CONFIG_X86
526 if (efi_enabled)
527 efi_enter_virtual_mode();
528 #endif
529 fork_init(num_physpages);
530 proc_caches_init();
531 buffer_init();
532 unnamed_dev_init();
533 key_init();
534 security_init();
535 vfs_caches_init(num_physpages);
536 radix_tree_init();
537 signals_init();
538 /* rootfs populating might need page-writeback */
539 page_writeback_init();
540 #ifdef CONFIG_PROC_FS
541 proc_root_init();
542 #endif
543 cpuset_init();
545 check_bugs();
547 acpi_early_init(); /* before LAPIC and SMP init */
549 /* Do the rest non-__init'ed, we're now alive */
550 rest_init();
553 static int __initdata initcall_debug;
555 static int __init initcall_debug_setup(char *str)
557 initcall_debug = 1;
558 return 1;
560 __setup("initcall_debug", initcall_debug_setup);
562 struct task_struct *child_reaper = &init_task;
564 extern initcall_t __initcall_start[], __initcall_end[];
566 static void __init do_initcalls(void)
568 initcall_t *call;
569 int count = preempt_count();
571 for (call = __initcall_start; call < __initcall_end; call++) {
572 char *msg = NULL;
573 char msgbuf[40];
574 int result;
576 if (initcall_debug) {
577 printk("Calling initcall 0x%p", *call);
578 print_fn_descriptor_symbol(": %s()",
579 (unsigned long) *call);
580 printk("\n");
583 result = (*call)();
585 if (result && (result != -ENODEV || initcall_debug)) {
586 sprintf(msgbuf, "error code %d", result);
587 msg = msgbuf;
589 if (preempt_count() != count) {
590 msg = "preemption imbalance";
591 preempt_count() = count;
593 if (irqs_disabled()) {
594 msg = "disabled interrupts";
595 local_irq_enable();
597 if (msg) {
598 printk(KERN_WARNING "initcall at 0x%p", *call);
599 print_fn_descriptor_symbol(": %s()",
600 (unsigned long) *call);
601 printk(": returned with %s\n", msg);
605 /* Make sure there is no pending stuff from the initcall sequence */
606 flush_scheduled_work();
610 * Ok, the machine is now initialized. None of the devices
611 * have been touched yet, but the CPU subsystem is up and
612 * running, and memory and process management works.
614 * Now we can finally start doing some real work..
616 static void __init do_basic_setup(void)
618 /* drivers will send hotplug events */
619 init_workqueues();
620 usermodehelper_init();
621 driver_init();
623 #ifdef CONFIG_SYSCTL
624 sysctl_init();
625 #endif
627 do_initcalls();
630 static void do_pre_smp_initcalls(void)
632 extern int spawn_ksoftirqd(void);
633 #ifdef CONFIG_SMP
634 extern int migration_init(void);
636 migration_init();
637 #endif
638 spawn_ksoftirqd();
639 spawn_softlockup_task();
642 static void run_init_process(char *init_filename)
644 argv_init[0] = init_filename;
645 execve(init_filename, argv_init, envp_init);
648 static inline void fixup_cpu_present_map(void)
650 #ifdef CONFIG_SMP
651 int i;
654 * If arch is not hotplug ready and did not populate
655 * cpu_present_map, just make cpu_present_map same as cpu_possible_map
656 * for other cpu bringup code to function as normal. e.g smp_init() etc.
658 if (cpus_empty(cpu_present_map)) {
659 for_each_cpu(i) {
660 cpu_set(i, cpu_present_map);
663 #endif
666 static int init(void * unused)
668 lock_kernel();
670 * init can run on any cpu.
672 set_cpus_allowed(current, CPU_MASK_ALL);
674 * Tell the world that we're going to be the grim
675 * reaper of innocent orphaned children.
677 * We don't want people to have to make incorrect
678 * assumptions about where in the task array this
679 * can be found.
681 child_reaper = current;
683 smp_prepare_cpus(max_cpus);
685 do_pre_smp_initcalls();
687 fixup_cpu_present_map();
688 smp_init();
689 sched_init_smp();
691 cpuset_init_smp();
694 * Do this before initcalls, because some drivers want to access
695 * firmware files.
697 populate_rootfs();
699 do_basic_setup();
702 * check if there is an early userspace init. If yes, let it do all
703 * the work
706 if (!ramdisk_execute_command)
707 ramdisk_execute_command = "/init";
709 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
710 ramdisk_execute_command = NULL;
711 prepare_namespace();
715 * Ok, we have completed the initial bootup, and
716 * we're essentially up and running. Get rid of the
717 * initmem segments and start the user-mode stuff..
719 free_initmem();
720 unlock_kernel();
721 mark_rodata_ro();
722 system_state = SYSTEM_RUNNING;
723 numa_default_policy();
725 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
726 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
728 (void) sys_dup(0);
729 (void) sys_dup(0);
731 if (ramdisk_execute_command) {
732 run_init_process(ramdisk_execute_command);
733 printk(KERN_WARNING "Failed to execute %s\n",
734 ramdisk_execute_command);
738 * We try each of these until one succeeds.
740 * The Bourne shell can be used instead of init if we are
741 * trying to recover a really broken machine.
743 if (execute_command) {
744 run_init_process(execute_command);
745 printk(KERN_WARNING "Failed to execute %s. Attempting "
746 "defaults...\n", execute_command);
748 run_init_process("/sbin/init");
749 run_init_process("/etc/init");
750 run_init_process("/bin/init");
751 run_init_process("/bin/sh");
753 panic("No init found. Try passing init= option to kernel.");