Make SLES9 "get_kernel_version" work on the kernel binary again
[linux-2.6/cjktty.git] / init / main.c
blobfcd9ddc3ccf5c02e15f7169b2264c9604b52eac1
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/start_kernel.h>
33 #include <linux/security.h>
34 #include <linux/workqueue.h>
35 #include <linux/profile.h>
36 #include <linux/rcupdate.h>
37 #include <linux/moduleparam.h>
38 #include <linux/kallsyms.h>
39 #include <linux/writeback.h>
40 #include <linux/cpu.h>
41 #include <linux/cpuset.h>
42 #include <linux/efi.h>
43 #include <linux/taskstats_kern.h>
44 #include <linux/delayacct.h>
45 #include <linux/unistd.h>
46 #include <linux/rmap.h>
47 #include <linux/mempolicy.h>
48 #include <linux/key.h>
49 #include <linux/unwind.h>
50 #include <linux/buffer_head.h>
51 #include <linux/debug_locks.h>
52 #include <linux/lockdep.h>
53 #include <linux/utsrelease.h>
54 #include <linux/pid_namespace.h>
55 #include <linux/compile.h>
57 #include <asm/io.h>
58 #include <asm/bugs.h>
59 #include <asm/setup.h>
60 #include <asm/sections.h>
61 #include <asm/cacheflush.h>
63 #ifdef CONFIG_X86_LOCAL_APIC
64 #include <asm/smp.h>
65 #endif
68 * This is one of the first .c files built. Error out early if we have compiler
69 * trouble.
71 * Versions of gcc older than that listed below may actually compile and link
72 * okay, but the end product can have subtle run time bugs. To avoid associated
73 * bogus bug reports, we flatly refuse to compile with a gcc that is known to be
74 * too old from the very beginning.
76 #if (__GNUC__ < 3) || (__GNUC__ == 3 && __GNUC_MINOR__ < 2)
77 #error Sorry, your GCC is too old. It builds incorrect kernels.
78 #endif
80 #if __GNUC__ == 4 && __GNUC_MINOR__ == 1 && __GNUC_PATCHLEVEL__ == 0
81 #warning gcc-4.1.0 is known to miscompile the kernel. A different compiler version is recommended.
82 #endif
84 static int init(void *);
86 extern void init_IRQ(void);
87 extern void fork_init(unsigned long);
88 extern void mca_init(void);
89 extern void sbus_init(void);
90 extern void sysctl_init(void);
91 extern void signals_init(void);
92 extern void pidhash_init(void);
93 extern void pidmap_init(void);
94 extern void prio_tree_init(void);
95 extern void radix_tree_init(void);
96 extern void free_initmem(void);
97 extern void populate_rootfs(void);
98 extern void driver_init(void);
99 extern void prepare_namespace(void);
100 #ifdef CONFIG_ACPI
101 extern void acpi_early_init(void);
102 #else
103 static inline void acpi_early_init(void) { }
104 #endif
105 #ifndef CONFIG_DEBUG_RODATA
106 static inline void mark_rodata_ro(void) { }
107 #endif
109 #ifdef CONFIG_TC
110 extern void tc_init(void);
111 #endif
113 enum system_states system_state;
114 EXPORT_SYMBOL(system_state);
117 * Boot command-line arguments
119 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
120 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
122 extern void time_init(void);
123 /* Default late time init is NULL. archs can override this later. */
124 void (*late_time_init)(void);
125 extern void softirq_init(void);
127 /* Untouched command line (eg. for /proc) saved by arch-specific code. */
128 char saved_command_line[COMMAND_LINE_SIZE];
130 static char *execute_command;
131 static char *ramdisk_execute_command;
133 /* Setup configured maximum number of CPUs to activate */
134 static unsigned int max_cpus = NR_CPUS;
137 * If set, this is an indication to the drivers that reset the underlying
138 * device before going ahead with the initialization otherwise driver might
139 * rely on the BIOS and skip the reset operation.
141 * This is useful if kernel is booting in an unreliable environment.
142 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
143 * skipped and devices will be in unknown state.
145 unsigned int reset_devices;
146 EXPORT_SYMBOL(reset_devices);
149 * Setup routine for controlling SMP activation
151 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
152 * activation entirely (the MPS table probe still happens, though).
154 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
155 * greater than 0, limits the maximum number of CPUs activated in
156 * SMP mode to <NUM>.
158 static int __init nosmp(char *str)
160 max_cpus = 0;
161 return 1;
164 __setup("nosmp", nosmp);
166 static int __init maxcpus(char *str)
168 get_option(&str, &max_cpus);
169 return 1;
172 __setup("maxcpus=", maxcpus);
174 static int __init set_reset_devices(char *str)
176 reset_devices = 1;
177 return 1;
180 __setup("reset_devices", set_reset_devices);
182 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
183 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
184 static const char *panic_later, *panic_param;
186 extern struct obs_kernel_param __setup_start[], __setup_end[];
188 static int __init obsolete_checksetup(char *line)
190 struct obs_kernel_param *p;
191 int had_early_param = 0;
193 p = __setup_start;
194 do {
195 int n = strlen(p->str);
196 if (!strncmp(line, p->str, n)) {
197 if (p->early) {
198 /* Already done in parse_early_param?
199 * (Needs exact match on param part).
200 * Keep iterating, as we can have early
201 * params and __setups of same names 8( */
202 if (line[n] == '\0' || line[n] == '=')
203 had_early_param = 1;
204 } else if (!p->setup_func) {
205 printk(KERN_WARNING "Parameter %s is obsolete,"
206 " ignored\n", p->str);
207 return 1;
208 } else if (p->setup_func(line + n))
209 return 1;
211 p++;
212 } while (p < __setup_end);
214 return had_early_param;
218 * This should be approx 2 Bo*oMips to start (note initial shift), and will
219 * still work even if initially too large, it will just take slightly longer
221 unsigned long loops_per_jiffy = (1<<12);
223 EXPORT_SYMBOL(loops_per_jiffy);
225 static int __init debug_kernel(char *str)
227 if (*str)
228 return 0;
229 console_loglevel = 10;
230 return 1;
233 static int __init quiet_kernel(char *str)
235 if (*str)
236 return 0;
237 console_loglevel = 4;
238 return 1;
241 __setup("debug", debug_kernel);
242 __setup("quiet", quiet_kernel);
244 static int __init loglevel(char *str)
246 get_option(&str, &console_loglevel);
247 return 1;
250 __setup("loglevel=", loglevel);
253 * Unknown boot options get handed to init, unless they look like
254 * failed parameters
256 static int __init unknown_bootoption(char *param, char *val)
258 /* Change NUL term back to "=", to make "param" the whole string. */
259 if (val) {
260 /* param=val or param="val"? */
261 if (val == param+strlen(param)+1)
262 val[-1] = '=';
263 else if (val == param+strlen(param)+2) {
264 val[-2] = '=';
265 memmove(val-1, val, strlen(val)+1);
266 val--;
267 } else
268 BUG();
271 /* Handle obsolete-style parameters */
272 if (obsolete_checksetup(param))
273 return 0;
276 * Preemptive maintenance for "why didn't my mispelled command
277 * line work?"
279 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
280 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
281 return 0;
284 if (panic_later)
285 return 0;
287 if (val) {
288 /* Environment option */
289 unsigned int i;
290 for (i = 0; envp_init[i]; i++) {
291 if (i == MAX_INIT_ENVS) {
292 panic_later = "Too many boot env vars at `%s'";
293 panic_param = param;
295 if (!strncmp(param, envp_init[i], val - param))
296 break;
298 envp_init[i] = param;
299 } else {
300 /* Command line option */
301 unsigned int i;
302 for (i = 0; argv_init[i]; i++) {
303 if (i == MAX_INIT_ARGS) {
304 panic_later = "Too many boot init vars at `%s'";
305 panic_param = param;
308 argv_init[i] = param;
310 return 0;
313 static int __init init_setup(char *str)
315 unsigned int i;
317 execute_command = str;
319 * In case LILO is going to boot us with default command line,
320 * it prepends "auto" before the whole cmdline which makes
321 * the shell think it should execute a script with such name.
322 * So we ignore all arguments entered _before_ init=... [MJ]
324 for (i = 1; i < MAX_INIT_ARGS; i++)
325 argv_init[i] = NULL;
326 return 1;
328 __setup("init=", init_setup);
330 static int __init rdinit_setup(char *str)
332 unsigned int i;
334 ramdisk_execute_command = str;
335 /* See "auto" comment in init_setup */
336 for (i = 1; i < MAX_INIT_ARGS; i++)
337 argv_init[i] = NULL;
338 return 1;
340 __setup("rdinit=", rdinit_setup);
342 #ifndef CONFIG_SMP
344 #ifdef CONFIG_X86_LOCAL_APIC
345 static void __init smp_init(void)
347 APIC_init_uniprocessor();
349 #else
350 #define smp_init() do { } while (0)
351 #endif
353 static inline void setup_per_cpu_areas(void) { }
354 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
356 #else
358 #ifdef __GENERIC_PER_CPU
359 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
361 EXPORT_SYMBOL(__per_cpu_offset);
363 static void __init setup_per_cpu_areas(void)
365 unsigned long size, i;
366 char *ptr;
367 unsigned long nr_possible_cpus = num_possible_cpus();
369 /* Copy section for each CPU (we discard the original) */
370 size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
371 #ifdef CONFIG_MODULES
372 if (size < PERCPU_ENOUGH_ROOM)
373 size = PERCPU_ENOUGH_ROOM;
374 #endif
375 ptr = alloc_bootmem(size * nr_possible_cpus);
377 for_each_possible_cpu(i) {
378 __per_cpu_offset[i] = ptr - __per_cpu_start;
379 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
380 ptr += size;
383 #endif /* !__GENERIC_PER_CPU */
385 /* Called by boot processor to activate the rest. */
386 static void __init smp_init(void)
388 unsigned int i;
390 /* FIXME: This should be done in userspace --RR */
391 for_each_present_cpu(i) {
392 if (num_online_cpus() >= max_cpus)
393 break;
394 if (!cpu_online(i))
395 cpu_up(i);
398 /* Any cleanup work */
399 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
400 smp_cpus_done(max_cpus);
401 #if 0
402 /* Get other processors into their bootup holding patterns. */
404 smp_commence();
405 #endif
408 #endif
411 * We need to finalize in a non-__init function or else race conditions
412 * between the root thread and the init thread may cause start_kernel to
413 * be reaped by free_initmem before the root thread has proceeded to
414 * cpu_idle.
416 * gcc-3.4 accidentally inlines this function, so use noinline.
419 static void noinline rest_init(void)
420 __releases(kernel_lock)
422 kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
423 numa_default_policy();
424 unlock_kernel();
427 * The boot idle thread must execute schedule()
428 * at least one to get things moving:
430 preempt_enable_no_resched();
431 schedule();
432 preempt_disable();
434 /* Call into cpu_idle with preempt disabled */
435 cpu_idle();
438 /* Check for early params. */
439 static int __init do_early_param(char *param, char *val)
441 struct obs_kernel_param *p;
443 for (p = __setup_start; p < __setup_end; p++) {
444 if (p->early && strcmp(param, p->str) == 0) {
445 if (p->setup_func(val) != 0)
446 printk(KERN_WARNING
447 "Malformed early option '%s'\n", param);
450 /* We accept everything at this stage. */
451 return 0;
454 /* Arch code calls this early on, or if not, just before other parsing. */
455 void __init parse_early_param(void)
457 static __initdata int done = 0;
458 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
460 if (done)
461 return;
463 /* All fall through to do_early_param. */
464 strlcpy(tmp_cmdline, saved_command_line, COMMAND_LINE_SIZE);
465 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
466 done = 1;
470 * Activate the first processor.
473 static void __init boot_cpu_init(void)
475 int cpu = smp_processor_id();
476 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
477 cpu_set(cpu, cpu_online_map);
478 cpu_set(cpu, cpu_present_map);
479 cpu_set(cpu, cpu_possible_map);
482 void __init __attribute__((weak)) smp_setup_processor_id(void)
486 static const char linux_banner[] =
487 "Linux version " UTS_RELEASE
488 " (" LINUX_COMPILE_BY "@" LINUX_COMPILE_HOST ")"
489 " (" LINUX_COMPILER ")"
490 " " UTS_VERSION "\n";
492 asmlinkage void __init start_kernel(void)
494 char * command_line;
495 extern struct kernel_param __start___param[], __stop___param[];
497 smp_setup_processor_id();
500 * Need to run as early as possible, to initialize the
501 * lockdep hash:
503 unwind_init();
504 lockdep_init();
506 local_irq_disable();
507 early_boot_irqs_off();
508 early_init_irq_lock_class();
511 * Interrupts are still disabled. Do necessary setups, then
512 * enable them
514 lock_kernel();
515 boot_cpu_init();
516 page_address_init();
517 printk(KERN_NOTICE);
518 printk(linux_banner);
519 setup_arch(&command_line);
520 unwind_setup();
521 setup_per_cpu_areas();
522 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
525 * Set up the scheduler prior starting any interrupts (such as the
526 * timer interrupt). Full topology setup happens at smp_init()
527 * time - but meanwhile we still have a functioning scheduler.
529 sched_init();
531 * Disable preemption - early bootup scheduling is extremely
532 * fragile until we cpu_idle() for the first time.
534 preempt_disable();
535 build_all_zonelists();
536 page_alloc_init();
537 printk(KERN_NOTICE "Kernel command line: %s\n", saved_command_line);
538 parse_early_param();
539 parse_args("Booting kernel", command_line, __start___param,
540 __stop___param - __start___param,
541 &unknown_bootoption);
542 sort_main_extable();
543 trap_init();
544 rcu_init();
545 init_IRQ();
546 pidhash_init();
547 init_timers();
548 hrtimers_init();
549 softirq_init();
550 timekeeping_init();
551 time_init();
552 profile_init();
553 if (!irqs_disabled())
554 printk("start_kernel(): bug: interrupts were enabled early\n");
555 early_boot_irqs_on();
556 local_irq_enable();
559 * HACK ALERT! This is early. We're enabling the console before
560 * we've done PCI setups etc, and console_init() must be aware of
561 * this. But we do want output early, in case something goes wrong.
563 console_init();
564 if (panic_later)
565 panic(panic_later, panic_param);
567 lockdep_info();
570 * Need to run this when irqs are enabled, because it wants
571 * to self-test [hard/soft]-irqs on/off lock inversion bugs
572 * too:
574 locking_selftest();
576 #ifdef CONFIG_BLK_DEV_INITRD
577 if (initrd_start && !initrd_below_start_ok &&
578 initrd_start < min_low_pfn << PAGE_SHIFT) {
579 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
580 "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
581 initrd_start = 0;
583 #endif
584 vfs_caches_init_early();
585 cpuset_init_early();
586 mem_init();
587 kmem_cache_init();
588 setup_per_cpu_pageset();
589 numa_policy_init();
590 if (late_time_init)
591 late_time_init();
592 calibrate_delay();
593 pidmap_init();
594 pgtable_cache_init();
595 prio_tree_init();
596 anon_vma_init();
597 #ifdef CONFIG_X86
598 if (efi_enabled)
599 efi_enter_virtual_mode();
600 #endif
601 fork_init(num_physpages);
602 proc_caches_init();
603 buffer_init();
604 unnamed_dev_init();
605 key_init();
606 security_init();
607 vfs_caches_init(num_physpages);
608 radix_tree_init();
609 signals_init();
610 /* rootfs populating might need page-writeback */
611 page_writeback_init();
612 #ifdef CONFIG_PROC_FS
613 proc_root_init();
614 #endif
615 cpuset_init();
616 taskstats_init_early();
617 delayacct_init();
619 check_bugs();
621 acpi_early_init(); /* before LAPIC and SMP init */
623 /* Do the rest non-__init'ed, we're now alive */
624 rest_init();
627 static int __initdata initcall_debug;
629 static int __init initcall_debug_setup(char *str)
631 initcall_debug = 1;
632 return 1;
634 __setup("initcall_debug", initcall_debug_setup);
636 extern initcall_t __initcall_start[], __initcall_end[];
638 static void __init do_initcalls(void)
640 initcall_t *call;
641 int count = preempt_count();
643 for (call = __initcall_start; call < __initcall_end; call++) {
644 char *msg = NULL;
645 char msgbuf[40];
646 int result;
648 if (initcall_debug) {
649 printk("Calling initcall 0x%p", *call);
650 print_fn_descriptor_symbol(": %s()",
651 (unsigned long) *call);
652 printk("\n");
655 result = (*call)();
657 if (result && result != -ENODEV && initcall_debug) {
658 sprintf(msgbuf, "error code %d", result);
659 msg = msgbuf;
661 if (preempt_count() != count) {
662 msg = "preemption imbalance";
663 preempt_count() = count;
665 if (irqs_disabled()) {
666 msg = "disabled interrupts";
667 local_irq_enable();
669 if (msg) {
670 printk(KERN_WARNING "initcall at 0x%p", *call);
671 print_fn_descriptor_symbol(": %s()",
672 (unsigned long) *call);
673 printk(": returned with %s\n", msg);
677 /* Make sure there is no pending stuff from the initcall sequence */
678 flush_scheduled_work();
682 * Ok, the machine is now initialized. None of the devices
683 * have been touched yet, but the CPU subsystem is up and
684 * running, and memory and process management works.
686 * Now we can finally start doing some real work..
688 static void __init do_basic_setup(void)
690 /* drivers will send hotplug events */
691 init_workqueues();
692 usermodehelper_init();
693 driver_init();
695 #ifdef CONFIG_SYSCTL
696 sysctl_init();
697 #endif
699 do_initcalls();
702 static void do_pre_smp_initcalls(void)
704 extern int spawn_ksoftirqd(void);
705 #ifdef CONFIG_SMP
706 extern int migration_init(void);
708 migration_init();
709 #endif
710 spawn_ksoftirqd();
711 spawn_softlockup_task();
714 static void run_init_process(char *init_filename)
716 argv_init[0] = init_filename;
717 kernel_execve(init_filename, argv_init, envp_init);
720 static int init(void * unused)
722 lock_kernel();
724 * init can run on any cpu.
726 set_cpus_allowed(current, CPU_MASK_ALL);
728 * Tell the world that we're going to be the grim
729 * reaper of innocent orphaned children.
731 * We don't want people to have to make incorrect
732 * assumptions about where in the task array this
733 * can be found.
735 init_pid_ns.child_reaper = current;
737 cad_pid = task_pid(current);
739 smp_prepare_cpus(max_cpus);
741 do_pre_smp_initcalls();
743 smp_init();
744 sched_init_smp();
746 cpuset_init_smp();
749 * Do this before initcalls, because some drivers want to access
750 * firmware files.
752 populate_rootfs();
754 do_basic_setup();
757 * check if there is an early userspace init. If yes, let it do all
758 * the work
761 if (!ramdisk_execute_command)
762 ramdisk_execute_command = "/init";
764 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
765 ramdisk_execute_command = NULL;
766 prepare_namespace();
770 * Ok, we have completed the initial bootup, and
771 * we're essentially up and running. Get rid of the
772 * initmem segments and start the user-mode stuff..
774 free_initmem();
775 unlock_kernel();
776 mark_rodata_ro();
777 system_state = SYSTEM_RUNNING;
778 numa_default_policy();
780 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
781 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
783 (void) sys_dup(0);
784 (void) sys_dup(0);
786 if (ramdisk_execute_command) {
787 run_init_process(ramdisk_execute_command);
788 printk(KERN_WARNING "Failed to execute %s\n",
789 ramdisk_execute_command);
793 * We try each of these until one succeeds.
795 * The Bourne shell can be used instead of init if we are
796 * trying to recover a really broken machine.
798 if (execute_command) {
799 run_init_process(execute_command);
800 printk(KERN_WARNING "Failed to execute %s. Attempting "
801 "defaults...\n", execute_command);
803 run_init_process("/sbin/init");
804 run_init_process("/etc/init");
805 run_init_process("/bin/init");
806 run_init_process("/bin/sh");
808 panic("No init found. Try passing init= option to kernel.");