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 DEBUG /* Enable initcall_debug */
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/stackprotector.h>
20 #include <linux/string.h>
21 #include <linux/ctype.h>
22 #include <linux/delay.h>
23 #include <linux/ioport.h>
24 #include <linux/init.h>
25 #include <linux/initrd.h>
26 #include <linux/bootmem.h>
27 #include <linux/acpi.h>
28 #include <linux/tty.h>
29 #include <linux/percpu.h>
30 #include <linux/kmod.h>
31 #include <linux/vmalloc.h>
32 #include <linux/kernel_stat.h>
33 #include <linux/start_kernel.h>
34 #include <linux/security.h>
35 #include <linux/smp.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/cgroup.h>
44 #include <linux/efi.h>
45 #include <linux/tick.h>
46 #include <linux/interrupt.h>
47 #include <linux/taskstats_kern.h>
48 #include <linux/delayacct.h>
49 #include <linux/unistd.h>
50 #include <linux/rmap.h>
51 #include <linux/mempolicy.h>
52 #include <linux/key.h>
53 #include <linux/buffer_head.h>
54 #include <linux/page_ext.h>
55 #include <linux/debug_locks.h>
56 #include <linux/debugobjects.h>
57 #include <linux/lockdep.h>
58 #include <linux/kmemleak.h>
59 #include <linux/pid_namespace.h>
60 #include <linux/device.h>
61 #include <linux/kthread.h>
62 #include <linux/sched.h>
63 #include <linux/signal.h>
64 #include <linux/idr.h>
65 #include <linux/kgdb.h>
66 #include <linux/ftrace.h>
67 #include <linux/async.h>
68 #include <linux/kmemcheck.h>
69 #include <linux/sfi.h>
70 #include <linux/shmem_fs.h>
71 #include <linux/slab.h>
72 #include <linux/perf_event.h>
73 #include <linux/file.h>
74 #include <linux/ptrace.h>
75 #include <linux/blkdev.h>
76 #include <linux/elevator.h>
77 #include <linux/sched_clock.h>
78 #include <linux/context_tracking.h>
79 #include <linux/random.h>
80 #include <linux/list.h>
81 #include <linux/integrity.h>
82 #include <linux/proc_ns.h>
83 #include <linux/io.h>
84 #include <linux/kaiser.h>
85
86 #include <asm/io.h>
87 #include <asm/bugs.h>
88 #include <asm/setup.h>
89 #include <asm/sections.h>
90 #include <asm/cacheflush.h>
91
92 static int kernel_init(void *);
93
94 extern void init_IRQ(void);
95 extern void fork_init(void);
96 extern void radix_tree_init(void);
97
98 /*
99 * Debug helper: via this flag we know that we are in 'early bootup code'
100 * where only the boot processor is running with IRQ disabled. This means
101 * two things - IRQ must not be enabled before the flag is cleared and some
102 * operations which are not allowed with IRQ disabled are allowed while the
103 * flag is set.
104 */
105 bool early_boot_irqs_disabled __read_mostly;
106
107 enum system_states system_state __read_mostly;
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 (*__initdata late_time_init)(void);
119
120 /* Untouched command line saved by arch-specific code. */
121 char __initdata boot_command_line[COMMAND_LINE_SIZE];
122 /* Untouched saved command line (eg. for /proc) */
123 char *saved_command_line;
124 /* Command line for parameter parsing */
125 static char *static_command_line;
126 /* Command line for per-initcall parameter parsing */
127 static char *initcall_command_line;
128
129 static char *execute_command;
130 static char *ramdisk_execute_command;
131
132 /*
133 * Used to generate warnings if static_key manipulation functions are used
134 * before jump_label_init is called.
135 */
136 bool static_key_initialized __read_mostly;
137 EXPORT_SYMBOL_GPL(static_key_initialized);
138
139 /*
140 * If set, this is an indication to the drivers that reset the underlying
141 * device before going ahead with the initialization otherwise driver might
142 * rely on the BIOS and skip the reset operation.
143 *
144 * This is useful if kernel is booting in an unreliable environment.
145 * For ex. kdump situation where previous kernel has crashed, BIOS has been
146 * skipped and devices will be in unknown state.
147 */
148 unsigned int reset_devices;
149 EXPORT_SYMBOL(reset_devices);
150
set_reset_devices(char * str)151 static int __init set_reset_devices(char *str)
152 {
153 reset_devices = 1;
154 return 1;
155 }
156
157 __setup("reset_devices", set_reset_devices);
158
159 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
160 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
161 static const char *panic_later, *panic_param;
162
163 extern const struct obs_kernel_param __setup_start[], __setup_end[];
164
obsolete_checksetup(char * line)165 static int __init obsolete_checksetup(char *line)
166 {
167 const struct obs_kernel_param *p;
168 int had_early_param = 0;
169
170 p = __setup_start;
171 do {
172 int n = strlen(p->str);
173 if (parameqn(line, p->str, n)) {
174 if (p->early) {
175 /* Already done in parse_early_param?
176 * (Needs exact match on param part).
177 * Keep iterating, as we can have early
178 * params and __setups of same names 8( */
179 if (line[n] == '\0' || line[n] == '=')
180 had_early_param = 1;
181 } else if (!p->setup_func) {
182 pr_warn("Parameter %s is obsolete, ignored\n",
183 p->str);
184 return 1;
185 } else if (p->setup_func(line + n))
186 return 1;
187 }
188 p++;
189 } while (p < __setup_end);
190
191 return had_early_param;
192 }
193
194 /*
195 * This should be approx 2 Bo*oMips to start (note initial shift), and will
196 * still work even if initially too large, it will just take slightly longer
197 */
198 unsigned long loops_per_jiffy = (1<<12);
199 EXPORT_SYMBOL(loops_per_jiffy);
200
debug_kernel(char * str)201 static int __init debug_kernel(char *str)
202 {
203 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
204 return 0;
205 }
206
quiet_kernel(char * str)207 static int __init quiet_kernel(char *str)
208 {
209 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
210 return 0;
211 }
212
213 early_param("debug", debug_kernel);
214 early_param("quiet", quiet_kernel);
215
loglevel(char * str)216 static int __init loglevel(char *str)
217 {
218 int newlevel;
219
220 /*
221 * Only update loglevel value when a correct setting was passed,
222 * to prevent blind crashes (when loglevel being set to 0) that
223 * are quite hard to debug
224 */
225 if (get_option(&str, &newlevel)) {
226 console_loglevel = newlevel;
227 return 0;
228 }
229
230 return -EINVAL;
231 }
232
233 early_param("loglevel", loglevel);
234
235 /* Change NUL term back to "=", to make "param" the whole string. */
repair_env_string(char * param,char * val,const char * unused,void * arg)236 static int __init repair_env_string(char *param, char *val,
237 const char *unused, void *arg)
238 {
239 if (val) {
240 /* param=val or param="val"? */
241 if (val == param+strlen(param)+1)
242 val[-1] = '=';
243 else if (val == param+strlen(param)+2) {
244 val[-2] = '=';
245 memmove(val-1, val, strlen(val)+1);
246 val--;
247 } else
248 BUG();
249 }
250 return 0;
251 }
252
253 /* Anything after -- gets handed straight to init. */
set_init_arg(char * param,char * val,const char * unused,void * arg)254 static int __init set_init_arg(char *param, char *val,
255 const char *unused, void *arg)
256 {
257 unsigned int i;
258
259 if (panic_later)
260 return 0;
261
262 repair_env_string(param, val, unused, NULL);
263
264 for (i = 0; argv_init[i]; i++) {
265 if (i == MAX_INIT_ARGS) {
266 panic_later = "init";
267 panic_param = param;
268 return 0;
269 }
270 }
271 argv_init[i] = param;
272 return 0;
273 }
274
275 /*
276 * Unknown boot options get handed to init, unless they look like
277 * unused parameters (modprobe will find them in /proc/cmdline).
278 */
unknown_bootoption(char * param,char * val,const char * unused,void * arg)279 static int __init unknown_bootoption(char *param, char *val,
280 const char *unused, void *arg)
281 {
282 repair_env_string(param, val, unused, NULL);
283
284 /* Handle obsolete-style parameters */
285 if (obsolete_checksetup(param))
286 return 0;
287
288 /* Unused module parameter. */
289 if (strchr(param, '.') && (!val || strchr(param, '.') < val))
290 return 0;
291
292 if (panic_later)
293 return 0;
294
295 if (val) {
296 /* Environment option */
297 unsigned int i;
298 for (i = 0; envp_init[i]; i++) {
299 if (i == MAX_INIT_ENVS) {
300 panic_later = "env";
301 panic_param = param;
302 }
303 if (!strncmp(param, envp_init[i], val - param))
304 break;
305 }
306 envp_init[i] = param;
307 } else {
308 /* Command line option */
309 unsigned int i;
310 for (i = 0; argv_init[i]; i++) {
311 if (i == MAX_INIT_ARGS) {
312 panic_later = "init";
313 panic_param = param;
314 }
315 }
316 argv_init[i] = param;
317 }
318 return 0;
319 }
320
init_setup(char * str)321 static int __init init_setup(char *str)
322 {
323 unsigned int i;
324
325 execute_command = str;
326 /*
327 * In case LILO is going to boot us with default command line,
328 * it prepends "auto" before the whole cmdline which makes
329 * the shell think it should execute a script with such name.
330 * So we ignore all arguments entered _before_ init=... [MJ]
331 */
332 for (i = 1; i < MAX_INIT_ARGS; i++)
333 argv_init[i] = NULL;
334 return 1;
335 }
336 __setup("init=", init_setup);
337
rdinit_setup(char * str)338 static int __init rdinit_setup(char *str)
339 {
340 unsigned int i;
341
342 ramdisk_execute_command = str;
343 /* See "auto" comment in init_setup */
344 for (i = 1; i < MAX_INIT_ARGS; i++)
345 argv_init[i] = NULL;
346 return 1;
347 }
348 __setup("rdinit=", rdinit_setup);
349
350 #ifndef CONFIG_SMP
351 static const unsigned int setup_max_cpus = NR_CPUS;
setup_nr_cpu_ids(void)352 static inline void setup_nr_cpu_ids(void) { }
smp_prepare_cpus(unsigned int maxcpus)353 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
354 #endif
355
356 /*
357 * We need to store the untouched command line for future reference.
358 * We also need to store the touched command line since the parameter
359 * parsing is performed in place, and we should allow a component to
360 * store reference of name/value for future reference.
361 */
setup_command_line(char * command_line)362 static void __init setup_command_line(char *command_line)
363 {
364 saved_command_line =
365 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
366 initcall_command_line =
367 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
368 static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0);
369 strcpy(saved_command_line, boot_command_line);
370 strcpy(static_command_line, command_line);
371 }
372
373 /*
374 * We need to finalize in a non-__init function or else race conditions
375 * between the root thread and the init thread may cause start_kernel to
376 * be reaped by free_initmem before the root thread has proceeded to
377 * cpu_idle.
378 *
379 * gcc-3.4 accidentally inlines this function, so use noinline.
380 */
381
382 static __initdata DECLARE_COMPLETION(kthreadd_done);
383
rest_init(void)384 static noinline void __init_refok rest_init(void)
385 {
386 int pid;
387
388 rcu_scheduler_starting();
389 smpboot_thread_init();
390 /*
391 * We need to spawn init first so that it obtains pid 1, however
392 * the init task will end up wanting to create kthreads, which, if
393 * we schedule it before we create kthreadd, will OOPS.
394 */
395 kernel_thread(kernel_init, NULL, CLONE_FS);
396 numa_default_policy();
397 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
398 rcu_read_lock();
399 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
400 rcu_read_unlock();
401 complete(&kthreadd_done);
402
403 /*
404 * The boot idle thread must execute schedule()
405 * at least once to get things moving:
406 */
407 init_idle_bootup_task(current);
408 schedule_preempt_disabled();
409 /* Call into cpu_idle with preempt disabled */
410 cpu_startup_entry(CPUHP_ONLINE);
411 }
412
413 /* Check for early params. */
do_early_param(char * param,char * val,const char * unused,void * arg)414 static int __init do_early_param(char *param, char *val,
415 const char *unused, void *arg)
416 {
417 const struct obs_kernel_param *p;
418
419 for (p = __setup_start; p < __setup_end; p++) {
420 if ((p->early && parameq(param, p->str)) ||
421 (strcmp(param, "console") == 0 &&
422 strcmp(p->str, "earlycon") == 0)
423 ) {
424 if (p->setup_func(val) != 0)
425 pr_warn("Malformed early option '%s'\n", param);
426 }
427 }
428 /* We accept everything at this stage. */
429 return 0;
430 }
431
parse_early_options(char * cmdline)432 void __init parse_early_options(char *cmdline)
433 {
434 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
435 do_early_param);
436 }
437
438 /* Arch code calls this early on, or if not, just before other parsing. */
parse_early_param(void)439 void __init parse_early_param(void)
440 {
441 static int done __initdata;
442 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
443
444 if (done)
445 return;
446
447 /* All fall through to do_early_param. */
448 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
449 parse_early_options(tmp_cmdline);
450 done = 1;
451 }
452
453 /*
454 * Activate the first processor.
455 */
456
boot_cpu_init(void)457 static void __init boot_cpu_init(void)
458 {
459 int cpu = smp_processor_id();
460 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
461 set_cpu_online(cpu, true);
462 set_cpu_active(cpu, true);
463 set_cpu_present(cpu, true);
464 set_cpu_possible(cpu, true);
465 }
466
smp_setup_processor_id(void)467 void __init __weak smp_setup_processor_id(void)
468 {
469 }
470
471 # if THREAD_SIZE >= PAGE_SIZE
thread_stack_cache_init(void)472 void __init __weak thread_stack_cache_init(void)
473 {
474 }
475 #endif
476
477 /*
478 * Set up kernel memory allocators
479 */
mm_init(void)480 static void __init mm_init(void)
481 {
482 /*
483 * page_ext requires contiguous pages,
484 * bigger than MAX_ORDER unless SPARSEMEM.
485 */
486 page_ext_init_flatmem();
487 mem_init();
488 kmem_cache_init();
489 percpu_init_late();
490 pgtable_init();
491 vmalloc_init();
492 ioremap_huge_init();
493 kaiser_init();
494 }
495
start_kernel(void)496 asmlinkage __visible void __init start_kernel(void)
497 {
498 char *command_line;
499 char *after_dashes;
500
501 /*
502 * Need to run as early as possible, to initialize the
503 * lockdep hash:
504 */
505 lockdep_init();
506 set_task_stack_end_magic(&init_task);
507 smp_setup_processor_id();
508 debug_objects_early_init();
509
510 /*
511 * Set up the the initial canary ASAP:
512 */
513 boot_init_stack_canary();
514
515 cgroup_init_early();
516
517 local_irq_disable();
518 early_boot_irqs_disabled = true;
519
520 /*
521 * Interrupts are still disabled. Do necessary setups, then
522 * enable them
523 */
524 boot_cpu_init();
525 page_address_init();
526 pr_notice("%s", linux_banner);
527 setup_arch(&command_line);
528 mm_init_cpumask(&init_mm);
529 setup_command_line(command_line);
530 setup_nr_cpu_ids();
531 setup_per_cpu_areas();
532 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
533
534 build_all_zonelists(NULL, NULL);
535 page_alloc_init();
536
537 pr_notice("Kernel command line: %s\n", boot_command_line);
538 /* parameters may set static keys */
539 jump_label_init();
540 parse_early_param();
541 after_dashes = parse_args("Booting kernel",
542 static_command_line, __start___param,
543 __stop___param - __start___param,
544 -1, -1, NULL, &unknown_bootoption);
545 if (!IS_ERR_OR_NULL(after_dashes))
546 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
547 NULL, set_init_arg);
548
549 /*
550 * These use large bootmem allocations and must precede
551 * kmem_cache_init()
552 */
553 setup_log_buf(0);
554 pidhash_init();
555 vfs_caches_init_early();
556 sort_main_extable();
557 trap_init();
558 mm_init();
559
560 /*
561 * Set up the scheduler prior starting any interrupts (such as the
562 * timer interrupt). Full topology setup happens at smp_init()
563 * time - but meanwhile we still have a functioning scheduler.
564 */
565 sched_init();
566 /*
567 * Disable preemption - early bootup scheduling is extremely
568 * fragile until we cpu_idle() for the first time.
569 */
570 preempt_disable();
571 if (WARN(!irqs_disabled(),
572 "Interrupts were enabled *very* early, fixing it\n"))
573 local_irq_disable();
574 idr_init_cache();
575 rcu_init();
576
577 /* trace_printk() and trace points may be used after this */
578 trace_init();
579
580 context_tracking_init();
581 radix_tree_init();
582 /* init some links before init_ISA_irqs() */
583 early_irq_init();
584 init_IRQ();
585 tick_init();
586 rcu_init_nohz();
587 init_timers();
588 hrtimers_init();
589 softirq_init();
590 timekeeping_init();
591 time_init();
592 sched_clock_postinit();
593 perf_event_init();
594 profile_init();
595 call_function_init();
596 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
597 early_boot_irqs_disabled = false;
598 local_irq_enable();
599
600 kmem_cache_init_late();
601
602 /*
603 * HACK ALERT! This is early. We're enabling the console before
604 * we've done PCI setups etc, and console_init() must be aware of
605 * this. But we do want output early, in case something goes wrong.
606 */
607 console_init();
608 if (panic_later)
609 panic("Too many boot %s vars at `%s'", panic_later,
610 panic_param);
611
612 lockdep_info();
613
614 /*
615 * Need to run this when irqs are enabled, because it wants
616 * to self-test [hard/soft]-irqs on/off lock inversion bugs
617 * too:
618 */
619 locking_selftest();
620
621 #ifdef CONFIG_BLK_DEV_INITRD
622 if (initrd_start && !initrd_below_start_ok &&
623 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
624 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
625 page_to_pfn(virt_to_page((void *)initrd_start)),
626 min_low_pfn);
627 initrd_start = 0;
628 }
629 #endif
630 page_ext_init();
631 debug_objects_mem_init();
632 kmemleak_init();
633 setup_per_cpu_pageset();
634 numa_policy_init();
635 if (late_time_init)
636 late_time_init();
637 sched_clock_init();
638 calibrate_delay();
639 pidmap_init();
640 anon_vma_init();
641 acpi_early_init();
642 #ifdef CONFIG_X86
643 if (efi_enabled(EFI_RUNTIME_SERVICES))
644 efi_enter_virtual_mode();
645 #endif
646 #ifdef CONFIG_X86_ESPFIX64
647 /* Should be run before the first non-init thread is created */
648 init_espfix_bsp();
649 #endif
650 thread_stack_cache_init();
651 cred_init();
652 fork_init();
653 proc_caches_init();
654 buffer_init();
655 key_init();
656 security_init();
657 dbg_late_init();
658 vfs_caches_init();
659 signals_init();
660 /* rootfs populating might need page-writeback */
661 page_writeback_init();
662 proc_root_init();
663 nsfs_init();
664 cpuset_init();
665 cgroup_init();
666 taskstats_init_early();
667 delayacct_init();
668
669 check_bugs();
670
671 acpi_subsystem_init();
672 sfi_init_late();
673
674 if (efi_enabled(EFI_RUNTIME_SERVICES)) {
675 efi_late_init();
676 efi_free_boot_services();
677 }
678
679 ftrace_init();
680
681 /* Do the rest non-__init'ed, we're now alive */
682 rest_init();
683
684 prevent_tail_call_optimization();
685 }
686
687 /* Call all constructor functions linked into the kernel. */
do_ctors(void)688 static void __init do_ctors(void)
689 {
690 #ifdef CONFIG_CONSTRUCTORS
691 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
692
693 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
694 (*fn)();
695 #endif
696 }
697
698 bool initcall_debug;
699 core_param(initcall_debug, initcall_debug, bool, 0644);
700
701 #ifdef CONFIG_KALLSYMS
702 struct blacklist_entry {
703 struct list_head next;
704 char *buf;
705 };
706
707 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
708
initcall_blacklist(char * str)709 static int __init initcall_blacklist(char *str)
710 {
711 char *str_entry;
712 struct blacklist_entry *entry;
713
714 /* str argument is a comma-separated list of functions */
715 do {
716 str_entry = strsep(&str, ",");
717 if (str_entry) {
718 pr_debug("blacklisting initcall %s\n", str_entry);
719 entry = alloc_bootmem(sizeof(*entry));
720 entry->buf = alloc_bootmem(strlen(str_entry) + 1);
721 strcpy(entry->buf, str_entry);
722 list_add(&entry->next, &blacklisted_initcalls);
723 }
724 } while (str_entry);
725
726 return 0;
727 }
728
initcall_blacklisted(initcall_t fn)729 static bool __init_or_module initcall_blacklisted(initcall_t fn)
730 {
731 struct list_head *tmp;
732 struct blacklist_entry *entry;
733 char *fn_name;
734
735 fn_name = kasprintf(GFP_KERNEL, "%pf", fn);
736 if (!fn_name)
737 return false;
738
739 list_for_each(tmp, &blacklisted_initcalls) {
740 entry = list_entry(tmp, struct blacklist_entry, next);
741 if (!strcmp(fn_name, entry->buf)) {
742 pr_debug("initcall %s blacklisted\n", fn_name);
743 kfree(fn_name);
744 return true;
745 }
746 }
747
748 kfree(fn_name);
749 return false;
750 }
751 #else
initcall_blacklist(char * str)752 static int __init initcall_blacklist(char *str)
753 {
754 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
755 return 0;
756 }
757
initcall_blacklisted(initcall_t fn)758 static bool __init_or_module initcall_blacklisted(initcall_t fn)
759 {
760 return false;
761 }
762 #endif
763 __setup("initcall_blacklist=", initcall_blacklist);
764
do_one_initcall_debug(initcall_t fn)765 static int __init_or_module do_one_initcall_debug(initcall_t fn)
766 {
767 ktime_t calltime, delta, rettime;
768 unsigned long long duration;
769 int ret;
770
771 printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current));
772 calltime = ktime_get();
773 ret = fn();
774 rettime = ktime_get();
775 delta = ktime_sub(rettime, calltime);
776 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
777 printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n",
778 fn, ret, duration);
779
780 return ret;
781 }
782
do_one_initcall(initcall_t fn)783 int __init_or_module do_one_initcall(initcall_t fn)
784 {
785 int count = preempt_count();
786 int ret;
787 char msgbuf[64];
788
789 if (initcall_blacklisted(fn))
790 return -EPERM;
791
792 if (initcall_debug)
793 ret = do_one_initcall_debug(fn);
794 else
795 ret = fn();
796
797 msgbuf[0] = 0;
798
799 if (preempt_count() != count) {
800 sprintf(msgbuf, "preemption imbalance ");
801 preempt_count_set(count);
802 }
803 if (irqs_disabled()) {
804 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
805 local_irq_enable();
806 }
807 WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
808
809 return ret;
810 }
811
812
813 extern initcall_t __initcall_start[];
814 extern initcall_t __initcall0_start[];
815 extern initcall_t __initcall1_start[];
816 extern initcall_t __initcall2_start[];
817 extern initcall_t __initcall3_start[];
818 extern initcall_t __initcall4_start[];
819 extern initcall_t __initcall5_start[];
820 extern initcall_t __initcall6_start[];
821 extern initcall_t __initcall7_start[];
822 extern initcall_t __initcall_end[];
823
824 static initcall_t *initcall_levels[] __initdata = {
825 __initcall0_start,
826 __initcall1_start,
827 __initcall2_start,
828 __initcall3_start,
829 __initcall4_start,
830 __initcall5_start,
831 __initcall6_start,
832 __initcall7_start,
833 __initcall_end,
834 };
835
836 /* Keep these in sync with initcalls in include/linux/init.h */
837 static char *initcall_level_names[] __initdata = {
838 "early",
839 "core",
840 "postcore",
841 "arch",
842 "subsys",
843 "fs",
844 "device",
845 "late",
846 };
847
do_initcall_level(int level)848 static void __init do_initcall_level(int level)
849 {
850 initcall_t *fn;
851
852 strcpy(initcall_command_line, saved_command_line);
853 parse_args(initcall_level_names[level],
854 initcall_command_line, __start___param,
855 __stop___param - __start___param,
856 level, level,
857 NULL, &repair_env_string);
858
859 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
860 do_one_initcall(*fn);
861 }
862
do_initcalls(void)863 static void __init do_initcalls(void)
864 {
865 int level;
866
867 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
868 do_initcall_level(level);
869 }
870
871 /*
872 * Ok, the machine is now initialized. None of the devices
873 * have been touched yet, but the CPU subsystem is up and
874 * running, and memory and process management works.
875 *
876 * Now we can finally start doing some real work..
877 */
do_basic_setup(void)878 static void __init do_basic_setup(void)
879 {
880 cpuset_init_smp();
881 shmem_init();
882 driver_init();
883 init_irq_proc();
884 do_ctors();
885 usermodehelper_enable();
886 do_initcalls();
887 random_int_secret_init();
888 }
889
do_pre_smp_initcalls(void)890 static void __init do_pre_smp_initcalls(void)
891 {
892 initcall_t *fn;
893
894 for (fn = __initcall_start; fn < __initcall0_start; fn++)
895 do_one_initcall(*fn);
896 }
897
898 /*
899 * This function requests modules which should be loaded by default and is
900 * called twice right after initrd is mounted and right before init is
901 * exec'd. If such modules are on either initrd or rootfs, they will be
902 * loaded before control is passed to userland.
903 */
load_default_modules(void)904 void __init load_default_modules(void)
905 {
906 load_default_elevator_module();
907 }
908
run_init_process(const char * init_filename)909 static int run_init_process(const char *init_filename)
910 {
911 argv_init[0] = init_filename;
912 return do_execve(getname_kernel(init_filename),
913 (const char __user *const __user *)argv_init,
914 (const char __user *const __user *)envp_init);
915 }
916
try_to_run_init_process(const char * init_filename)917 static int try_to_run_init_process(const char *init_filename)
918 {
919 int ret;
920
921 ret = run_init_process(init_filename);
922
923 if (ret && ret != -ENOENT) {
924 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
925 init_filename, ret);
926 }
927
928 return ret;
929 }
930
931 static noinline void __init kernel_init_freeable(void);
932
933 #ifdef CONFIG_DEBUG_RODATA
934 static bool rodata_enabled = true;
set_debug_rodata(char * str)935 static int __init set_debug_rodata(char *str)
936 {
937 return strtobool(str, &rodata_enabled);
938 }
939 __setup("rodata=", set_debug_rodata);
940
mark_readonly(void)941 static void mark_readonly(void)
942 {
943 if (rodata_enabled)
944 mark_rodata_ro();
945 else
946 pr_info("Kernel memory protection disabled.\n");
947 }
948 #else
mark_readonly(void)949 static inline void mark_readonly(void)
950 {
951 pr_warn("This architecture does not have kernel memory protection.\n");
952 }
953 #endif
954
kernel_init(void * unused)955 static int __ref kernel_init(void *unused)
956 {
957 int ret;
958
959 kernel_init_freeable();
960 /* need to finish all async __init code before freeing the memory */
961 async_synchronize_full();
962 free_initmem();
963 mark_readonly();
964 system_state = SYSTEM_RUNNING;
965 numa_default_policy();
966
967 flush_delayed_fput();
968
969 if (ramdisk_execute_command) {
970 ret = run_init_process(ramdisk_execute_command);
971 if (!ret)
972 return 0;
973 pr_err("Failed to execute %s (error %d)\n",
974 ramdisk_execute_command, ret);
975 }
976
977 /*
978 * We try each of these until one succeeds.
979 *
980 * The Bourne shell can be used instead of init if we are
981 * trying to recover a really broken machine.
982 */
983 if (execute_command) {
984 ret = run_init_process(execute_command);
985 if (!ret)
986 return 0;
987 panic("Requested init %s failed (error %d).",
988 execute_command, ret);
989 }
990 if (!try_to_run_init_process("/sbin/init") ||
991 !try_to_run_init_process("/etc/init") ||
992 !try_to_run_init_process("/bin/init") ||
993 !try_to_run_init_process("/bin/sh"))
994 return 0;
995
996 panic("No working init found. Try passing init= option to kernel. "
997 "See Linux Documentation/init.txt for guidance.");
998 }
999
kernel_init_freeable(void)1000 static noinline void __init kernel_init_freeable(void)
1001 {
1002 /*
1003 * Wait until kthreadd is all set-up.
1004 */
1005 wait_for_completion(&kthreadd_done);
1006
1007 /* Now the scheduler is fully set up and can do blocking allocations */
1008 gfp_allowed_mask = __GFP_BITS_MASK;
1009
1010 /*
1011 * init can allocate pages on any node
1012 */
1013 set_mems_allowed(node_states[N_MEMORY]);
1014 /*
1015 * init can run on any cpu.
1016 */
1017 set_cpus_allowed_ptr(current, cpu_all_mask);
1018
1019 cad_pid = get_pid(task_pid(current));
1020
1021 smp_prepare_cpus(setup_max_cpus);
1022
1023 do_pre_smp_initcalls();
1024 lockup_detector_init();
1025
1026 smp_init();
1027 sched_init_smp();
1028
1029 page_alloc_init_late();
1030
1031 do_basic_setup();
1032
1033 /* Open the /dev/console on the rootfs, this should never fail */
1034 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1035 pr_err("Warning: unable to open an initial console.\n");
1036
1037 (void) sys_dup(0);
1038 (void) sys_dup(0);
1039 /*
1040 * check if there is an early userspace init. If yes, let it do all
1041 * the work
1042 */
1043
1044 if (!ramdisk_execute_command)
1045 ramdisk_execute_command = "/init";
1046
1047 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
1048 ramdisk_execute_command = NULL;
1049 prepare_namespace();
1050 }
1051
1052 /*
1053 * Ok, we have completed the initial bootup, and
1054 * we're essentially up and running. Get rid of the
1055 * initmem segments and start the user-mode stuff..
1056 *
1057 * rootfs is available now, try loading the public keys
1058 * and default modules
1059 */
1060
1061 integrity_load_keys();
1062 load_default_modules();
1063 }
1064