Version:  2.0.40 2.2.26 2.4.37 3.4 3.5 3.6 3.7 3.8 3.9 3.10 3.11 3.12 3.13 3.14 3.15 3.16 3.17 3.18 3.19 4.0

Linux/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 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 
 84 #include <asm/io.h>
 85 #include <asm/bugs.h>
 86 #include <asm/setup.h>
 87 #include <asm/sections.h>
 88 #include <asm/cacheflush.h>
 89 
 90 static int kernel_init(void *);
 91 
 92 extern void init_IRQ(void);
 93 extern void fork_init(unsigned long);
 94 extern void radix_tree_init(void);
 95 #ifndef CONFIG_DEBUG_RODATA
 96 static inline void mark_rodata_ro(void) { }
 97 #endif
 98 
 99 /*
100  * Debug helper: via this flag we know that we are in 'early bootup code'
101  * where only the boot processor is running with IRQ disabled.  This means
102  * two things - IRQ must not be enabled before the flag is cleared and some
103  * operations which are not allowed with IRQ disabled are allowed while the
104  * flag is set.
105  */
106 bool early_boot_irqs_disabled __read_mostly;
107 
108 enum system_states system_state __read_mostly;
109 EXPORT_SYMBOL(system_state);
110 
111 /*
112  * Boot command-line arguments
113  */
114 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
115 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
116 
117 extern void time_init(void);
118 /* Default late time init is NULL. archs can override this later. */
119 void (*__initdata late_time_init)(void);
120 
121 /* Untouched command line saved by arch-specific code. */
122 char __initdata boot_command_line[COMMAND_LINE_SIZE];
123 /* Untouched saved command line (eg. for /proc) */
124 char *saved_command_line;
125 /* Command line for parameter parsing */
126 static char *static_command_line;
127 /* Command line for per-initcall parameter parsing */
128 static char *initcall_command_line;
129 
130 static char *execute_command;
131 static char *ramdisk_execute_command;
132 
133 /*
134  * Used to generate warnings if static_key manipulation functions are used
135  * before jump_label_init is called.
136  */
137 bool static_key_initialized __read_mostly;
138 EXPORT_SYMBOL_GPL(static_key_initialized);
139 
140 /*
141  * If set, this is an indication to the drivers that reset the underlying
142  * device before going ahead with the initialization otherwise driver might
143  * rely on the BIOS and skip the reset operation.
144  *
145  * This is useful if kernel is booting in an unreliable environment.
146  * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
147  * skipped and devices will be in unknown state.
148  */
149 unsigned int reset_devices;
150 EXPORT_SYMBOL(reset_devices);
151 
152 static int __init set_reset_devices(char *str)
153 {
154         reset_devices = 1;
155         return 1;
156 }
157 
158 __setup("reset_devices", set_reset_devices);
159 
160 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
161 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
162 static const char *panic_later, *panic_param;
163 
164 extern const struct obs_kernel_param __setup_start[], __setup_end[];
165 
166 static int __init obsolete_checksetup(char *line)
167 {
168         const struct obs_kernel_param *p;
169         int had_early_param = 0;
170 
171         p = __setup_start;
172         do {
173                 int n = strlen(p->str);
174                 if (parameqn(line, p->str, n)) {
175                         if (p->early) {
176                                 /* Already done in parse_early_param?
177                                  * (Needs exact match on param part).
178                                  * Keep iterating, as we can have early
179                                  * params and __setups of same names 8( */
180                                 if (line[n] == '\0' || line[n] == '=')
181                                         had_early_param = 1;
182                         } else if (!p->setup_func) {
183                                 pr_warn("Parameter %s is obsolete, ignored\n",
184                                         p->str);
185                                 return 1;
186                         } else if (p->setup_func(line + n))
187                                 return 1;
188                 }
189                 p++;
190         } while (p < __setup_end);
191 
192         return had_early_param;
193 }
194 
195 /*
196  * This should be approx 2 Bo*oMips to start (note initial shift), and will
197  * still work even if initially too large, it will just take slightly longer
198  */
199 unsigned long loops_per_jiffy = (1<<12);
200 EXPORT_SYMBOL(loops_per_jiffy);
201 
202 static int __init debug_kernel(char *str)
203 {
204         console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
205         return 0;
206 }
207 
208 static int __init quiet_kernel(char *str)
209 {
210         console_loglevel = CONSOLE_LOGLEVEL_QUIET;
211         return 0;
212 }
213 
214 early_param("debug", debug_kernel);
215 early_param("quiet", quiet_kernel);
216 
217 static int __init loglevel(char *str)
218 {
219         int newlevel;
220 
221         /*
222          * Only update loglevel value when a correct setting was passed,
223          * to prevent blind crashes (when loglevel being set to 0) that
224          * are quite hard to debug
225          */
226         if (get_option(&str, &newlevel)) {
227                 console_loglevel = newlevel;
228                 return 0;
229         }
230 
231         return -EINVAL;
232 }
233 
234 early_param("loglevel", loglevel);
235 
236 /* Change NUL term back to "=", to make "param" the whole string. */
237 static int __init repair_env_string(char *param, char *val, const char *unused)
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. */
254 static int __init set_init_arg(char *param, char *val, const char *unused)
255 {
256         unsigned int i;
257 
258         if (panic_later)
259                 return 0;
260 
261         repair_env_string(param, val, unused);
262 
263         for (i = 0; argv_init[i]; i++) {
264                 if (i == MAX_INIT_ARGS) {
265                         panic_later = "init";
266                         panic_param = param;
267                         return 0;
268                 }
269         }
270         argv_init[i] = param;
271         return 0;
272 }
273 
274 /*
275  * Unknown boot options get handed to init, unless they look like
276  * unused parameters (modprobe will find them in /proc/cmdline).
277  */
278 static int __init unknown_bootoption(char *param, char *val, const char *unused)
279 {
280         repair_env_string(param, val, unused);
281 
282         /* Handle obsolete-style parameters */
283         if (obsolete_checksetup(param))
284                 return 0;
285 
286         /* Unused module parameter. */
287         if (strchr(param, '.') && (!val || strchr(param, '.') < val))
288                 return 0;
289 
290         if (panic_later)
291                 return 0;
292 
293         if (val) {
294                 /* Environment option */
295                 unsigned int i;
296                 for (i = 0; envp_init[i]; i++) {
297                         if (i == MAX_INIT_ENVS) {
298                                 panic_later = "env";
299                                 panic_param = param;
300                         }
301                         if (!strncmp(param, envp_init[i], val - param))
302                                 break;
303                 }
304                 envp_init[i] = param;
305         } else {
306                 /* Command line option */
307                 unsigned int i;
308                 for (i = 0; argv_init[i]; i++) {
309                         if (i == MAX_INIT_ARGS) {
310                                 panic_later = "init";
311                                 panic_param = param;
312                         }
313                 }
314                 argv_init[i] = param;
315         }
316         return 0;
317 }
318 
319 static int __init init_setup(char *str)
320 {
321         unsigned int i;
322 
323         execute_command = str;
324         /*
325          * In case LILO is going to boot us with default command line,
326          * it prepends "auto" before the whole cmdline which makes
327          * the shell think it should execute a script with such name.
328          * So we ignore all arguments entered _before_ init=... [MJ]
329          */
330         for (i = 1; i < MAX_INIT_ARGS; i++)
331                 argv_init[i] = NULL;
332         return 1;
333 }
334 __setup("init=", init_setup);
335 
336 static int __init rdinit_setup(char *str)
337 {
338         unsigned int i;
339 
340         ramdisk_execute_command = str;
341         /* See "auto" comment in init_setup */
342         for (i = 1; i < MAX_INIT_ARGS; i++)
343                 argv_init[i] = NULL;
344         return 1;
345 }
346 __setup("rdinit=", rdinit_setup);
347 
348 #ifndef CONFIG_SMP
349 static const unsigned int setup_max_cpus = NR_CPUS;
350 static inline void setup_nr_cpu_ids(void) { }
351 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
352 #endif
353 
354 /*
355  * We need to store the untouched command line for future reference.
356  * We also need to store the touched command line since the parameter
357  * parsing is performed in place, and we should allow a component to
358  * store reference of name/value for future reference.
359  */
360 static void __init setup_command_line(char *command_line)
361 {
362         saved_command_line =
363                 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
364         initcall_command_line =
365                 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
366         static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0);
367         strcpy(saved_command_line, boot_command_line);
368         strcpy(static_command_line, command_line);
369 }
370 
371 /*
372  * We need to finalize in a non-__init function or else race conditions
373  * between the root thread and the init thread may cause start_kernel to
374  * be reaped by free_initmem before the root thread has proceeded to
375  * cpu_idle.
376  *
377  * gcc-3.4 accidentally inlines this function, so use noinline.
378  */
379 
380 static __initdata DECLARE_COMPLETION(kthreadd_done);
381 
382 static noinline void __init_refok rest_init(void)
383 {
384         int pid;
385 
386         rcu_scheduler_starting();
387         /*
388          * We need to spawn init first so that it obtains pid 1, however
389          * the init task will end up wanting to create kthreads, which, if
390          * we schedule it before we create kthreadd, will OOPS.
391          */
392         kernel_thread(kernel_init, NULL, CLONE_FS);
393         numa_default_policy();
394         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
395         rcu_read_lock();
396         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
397         rcu_read_unlock();
398         complete(&kthreadd_done);
399 
400         /*
401          * The boot idle thread must execute schedule()
402          * at least once to get things moving:
403          */
404         init_idle_bootup_task(current);
405         schedule_preempt_disabled();
406         /* Call into cpu_idle with preempt disabled */
407         cpu_startup_entry(CPUHP_ONLINE);
408 }
409 
410 /* Check for early params. */
411 static int __init do_early_param(char *param, char *val, const char *unused)
412 {
413         const struct obs_kernel_param *p;
414 
415         for (p = __setup_start; p < __setup_end; p++) {
416                 if ((p->early && parameq(param, p->str)) ||
417                     (strcmp(param, "console") == 0 &&
418                      strcmp(p->str, "earlycon") == 0)
419                 ) {
420                         if (p->setup_func(val) != 0)
421                                 pr_warn("Malformed early option '%s'\n", param);
422                 }
423         }
424         /* We accept everything at this stage. */
425         return 0;
426 }
427 
428 void __init parse_early_options(char *cmdline)
429 {
430         parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
431 }
432 
433 /* Arch code calls this early on, or if not, just before other parsing. */
434 void __init parse_early_param(void)
435 {
436         static int done __initdata;
437         static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
438 
439         if (done)
440                 return;
441 
442         /* All fall through to do_early_param. */
443         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
444         parse_early_options(tmp_cmdline);
445         done = 1;
446 }
447 
448 /*
449  *      Activate the first processor.
450  */
451 
452 static void __init boot_cpu_init(void)
453 {
454         int cpu = smp_processor_id();
455         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
456         set_cpu_online(cpu, true);
457         set_cpu_active(cpu, true);
458         set_cpu_present(cpu, true);
459         set_cpu_possible(cpu, true);
460 }
461 
462 void __init __weak smp_setup_processor_id(void)
463 {
464 }
465 
466 # if THREAD_SIZE >= PAGE_SIZE
467 void __init __weak thread_info_cache_init(void)
468 {
469 }
470 #endif
471 
472 /*
473  * Set up kernel memory allocators
474  */
475 static void __init mm_init(void)
476 {
477         /*
478          * page_ext requires contiguous pages,
479          * bigger than MAX_ORDER unless SPARSEMEM.
480          */
481         page_ext_init_flatmem();
482         mem_init();
483         kmem_cache_init();
484         percpu_init_late();
485         pgtable_init();
486         vmalloc_init();
487 }
488 
489 asmlinkage __visible void __init start_kernel(void)
490 {
491         char *command_line;
492         char *after_dashes;
493 
494         /*
495          * Need to run as early as possible, to initialize the
496          * lockdep hash:
497          */
498         lockdep_init();
499         set_task_stack_end_magic(&init_task);
500         smp_setup_processor_id();
501         debug_objects_early_init();
502 
503         /*
504          * Set up the the initial canary ASAP:
505          */
506         boot_init_stack_canary();
507 
508         cgroup_init_early();
509 
510         local_irq_disable();
511         early_boot_irqs_disabled = true;
512 
513 /*
514  * Interrupts are still disabled. Do necessary setups, then
515  * enable them
516  */
517         boot_cpu_init();
518         page_address_init();
519         pr_notice("%s", linux_banner);
520         setup_arch(&command_line);
521         mm_init_cpumask(&init_mm);
522         setup_command_line(command_line);
523         setup_nr_cpu_ids();
524         setup_per_cpu_areas();
525         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
526 
527         build_all_zonelists(NULL, NULL);
528         page_alloc_init();
529 
530         pr_notice("Kernel command line: %s\n", boot_command_line);
531         parse_early_param();
532         after_dashes = parse_args("Booting kernel",
533                                   static_command_line, __start___param,
534                                   __stop___param - __start___param,
535                                   -1, -1, &unknown_bootoption);
536         if (!IS_ERR_OR_NULL(after_dashes))
537                 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
538                            set_init_arg);
539 
540         jump_label_init();
541 
542         /*
543          * These use large bootmem allocations and must precede
544          * kmem_cache_init()
545          */
546         setup_log_buf(0);
547         pidhash_init();
548         vfs_caches_init_early();
549         sort_main_extable();
550         trap_init();
551         mm_init();
552 
553         /*
554          * Set up the scheduler prior starting any interrupts (such as the
555          * timer interrupt). Full topology setup happens at smp_init()
556          * time - but meanwhile we still have a functioning scheduler.
557          */
558         sched_init();
559         /*
560          * Disable preemption - early bootup scheduling is extremely
561          * fragile until we cpu_idle() for the first time.
562          */
563         preempt_disable();
564         if (WARN(!irqs_disabled(),
565                  "Interrupts were enabled *very* early, fixing it\n"))
566                 local_irq_disable();
567         idr_init_cache();
568         rcu_init();
569 
570         /* trace_printk() and trace points may be used after this */
571         trace_init();
572 
573         context_tracking_init();
574         radix_tree_init();
575         /* init some links before init_ISA_irqs() */
576         early_irq_init();
577         init_IRQ();
578         tick_init();
579         rcu_init_nohz();
580         init_timers();
581         hrtimers_init();
582         softirq_init();
583         timekeeping_init();
584         time_init();
585         sched_clock_postinit();
586         perf_event_init();
587         profile_init();
588         call_function_init();
589         WARN(!irqs_disabled(), "Interrupts were enabled early\n");
590         early_boot_irqs_disabled = false;
591         local_irq_enable();
592 
593         kmem_cache_init_late();
594 
595         /*
596          * HACK ALERT! This is early. We're enabling the console before
597          * we've done PCI setups etc, and console_init() must be aware of
598          * this. But we do want output early, in case something goes wrong.
599          */
600         console_init();
601         if (panic_later)
602                 panic("Too many boot %s vars at `%s'", panic_later,
603                       panic_param);
604 
605         lockdep_info();
606 
607         /*
608          * Need to run this when irqs are enabled, because it wants
609          * to self-test [hard/soft]-irqs on/off lock inversion bugs
610          * too:
611          */
612         locking_selftest();
613 
614 #ifdef CONFIG_BLK_DEV_INITRD
615         if (initrd_start && !initrd_below_start_ok &&
616             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
617                 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
618                     page_to_pfn(virt_to_page((void *)initrd_start)),
619                     min_low_pfn);
620                 initrd_start = 0;
621         }
622 #endif
623         page_ext_init();
624         debug_objects_mem_init();
625         kmemleak_init();
626         setup_per_cpu_pageset();
627         numa_policy_init();
628         if (late_time_init)
629                 late_time_init();
630         sched_clock_init();
631         calibrate_delay();
632         pidmap_init();
633         anon_vma_init();
634         acpi_early_init();
635 #ifdef CONFIG_X86
636         if (efi_enabled(EFI_RUNTIME_SERVICES))
637                 efi_enter_virtual_mode();
638 #endif
639 #ifdef CONFIG_X86_ESPFIX64
640         /* Should be run before the first non-init thread is created */
641         init_espfix_bsp();
642 #endif
643         thread_info_cache_init();
644         cred_init();
645         fork_init(totalram_pages);
646         proc_caches_init();
647         buffer_init();
648         key_init();
649         security_init();
650         dbg_late_init();
651         vfs_caches_init(totalram_pages);
652         signals_init();
653         /* rootfs populating might need page-writeback */
654         page_writeback_init();
655         proc_root_init();
656         nsfs_init();
657         cgroup_init();
658         cpuset_init();
659         taskstats_init_early();
660         delayacct_init();
661 
662         check_bugs();
663 
664         sfi_init_late();
665 
666         if (efi_enabled(EFI_RUNTIME_SERVICES)) {
667                 efi_late_init();
668                 efi_free_boot_services();
669         }
670 
671         ftrace_init();
672 
673         /* Do the rest non-__init'ed, we're now alive */
674         rest_init();
675 }
676 
677 /* Call all constructor functions linked into the kernel. */
678 static void __init do_ctors(void)
679 {
680 #ifdef CONFIG_CONSTRUCTORS
681         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
682 
683         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
684                 (*fn)();
685 #endif
686 }
687 
688 bool initcall_debug;
689 core_param(initcall_debug, initcall_debug, bool, 0644);
690 
691 #ifdef CONFIG_KALLSYMS
692 struct blacklist_entry {
693         struct list_head next;
694         char *buf;
695 };
696 
697 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
698 
699 static int __init initcall_blacklist(char *str)
700 {
701         char *str_entry;
702         struct blacklist_entry *entry;
703 
704         /* str argument is a comma-separated list of functions */
705         do {
706                 str_entry = strsep(&str, ",");
707                 if (str_entry) {
708                         pr_debug("blacklisting initcall %s\n", str_entry);
709                         entry = alloc_bootmem(sizeof(*entry));
710                         entry->buf = alloc_bootmem(strlen(str_entry) + 1);
711                         strcpy(entry->buf, str_entry);
712                         list_add(&entry->next, &blacklisted_initcalls);
713                 }
714         } while (str_entry);
715 
716         return 0;
717 }
718 
719 static bool __init_or_module initcall_blacklisted(initcall_t fn)
720 {
721         struct list_head *tmp;
722         struct blacklist_entry *entry;
723         char *fn_name;
724 
725         fn_name = kasprintf(GFP_KERNEL, "%pf", fn);
726         if (!fn_name)
727                 return false;
728 
729         list_for_each(tmp, &blacklisted_initcalls) {
730                 entry = list_entry(tmp, struct blacklist_entry, next);
731                 if (!strcmp(fn_name, entry->buf)) {
732                         pr_debug("initcall %s blacklisted\n", fn_name);
733                         kfree(fn_name);
734                         return true;
735                 }
736         }
737 
738         kfree(fn_name);
739         return false;
740 }
741 #else
742 static int __init initcall_blacklist(char *str)
743 {
744         pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
745         return 0;
746 }
747 
748 static bool __init_or_module initcall_blacklisted(initcall_t fn)
749 {
750         return false;
751 }
752 #endif
753 __setup("initcall_blacklist=", initcall_blacklist);
754 
755 static int __init_or_module do_one_initcall_debug(initcall_t fn)
756 {
757         ktime_t calltime, delta, rettime;
758         unsigned long long duration;
759         int ret;
760 
761         printk(KERN_DEBUG "calling  %pF @ %i\n", fn, task_pid_nr(current));
762         calltime = ktime_get();
763         ret = fn();
764         rettime = ktime_get();
765         delta = ktime_sub(rettime, calltime);
766         duration = (unsigned long long) ktime_to_ns(delta) >> 10;
767         printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n",
768                  fn, ret, duration);
769 
770         return ret;
771 }
772 
773 int __init_or_module do_one_initcall(initcall_t fn)
774 {
775         int count = preempt_count();
776         int ret;
777         char msgbuf[64];
778 
779         if (initcall_blacklisted(fn))
780                 return -EPERM;
781 
782         if (initcall_debug)
783                 ret = do_one_initcall_debug(fn);
784         else
785                 ret = fn();
786 
787         msgbuf[0] = 0;
788 
789         if (preempt_count() != count) {
790                 sprintf(msgbuf, "preemption imbalance ");
791                 preempt_count_set(count);
792         }
793         if (irqs_disabled()) {
794                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
795                 local_irq_enable();
796         }
797         WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
798 
799         return ret;
800 }
801 
802 
803 extern initcall_t __initcall_start[];
804 extern initcall_t __initcall0_start[];
805 extern initcall_t __initcall1_start[];
806 extern initcall_t __initcall2_start[];
807 extern initcall_t __initcall3_start[];
808 extern initcall_t __initcall4_start[];
809 extern initcall_t __initcall5_start[];
810 extern initcall_t __initcall6_start[];
811 extern initcall_t __initcall7_start[];
812 extern initcall_t __initcall_end[];
813 
814 static initcall_t *initcall_levels[] __initdata = {
815         __initcall0_start,
816         __initcall1_start,
817         __initcall2_start,
818         __initcall3_start,
819         __initcall4_start,
820         __initcall5_start,
821         __initcall6_start,
822         __initcall7_start,
823         __initcall_end,
824 };
825 
826 /* Keep these in sync with initcalls in include/linux/init.h */
827 static char *initcall_level_names[] __initdata = {
828         "early",
829         "core",
830         "postcore",
831         "arch",
832         "subsys",
833         "fs",
834         "device",
835         "late",
836 };
837 
838 static void __init do_initcall_level(int level)
839 {
840         initcall_t *fn;
841 
842         strcpy(initcall_command_line, saved_command_line);
843         parse_args(initcall_level_names[level],
844                    initcall_command_line, __start___param,
845                    __stop___param - __start___param,
846                    level, level,
847                    &repair_env_string);
848 
849         for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
850                 do_one_initcall(*fn);
851 }
852 
853 static void __init do_initcalls(void)
854 {
855         int level;
856 
857         for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
858                 do_initcall_level(level);
859 }
860 
861 /*
862  * Ok, the machine is now initialized. None of the devices
863  * have been touched yet, but the CPU subsystem is up and
864  * running, and memory and process management works.
865  *
866  * Now we can finally start doing some real work..
867  */
868 static void __init do_basic_setup(void)
869 {
870         cpuset_init_smp();
871         usermodehelper_init();
872         shmem_init();
873         driver_init();
874         init_irq_proc();
875         do_ctors();
876         usermodehelper_enable();
877         do_initcalls();
878         random_int_secret_init();
879 }
880 
881 static void __init do_pre_smp_initcalls(void)
882 {
883         initcall_t *fn;
884 
885         for (fn = __initcall_start; fn < __initcall0_start; fn++)
886                 do_one_initcall(*fn);
887 }
888 
889 /*
890  * This function requests modules which should be loaded by default and is
891  * called twice right after initrd is mounted and right before init is
892  * exec'd.  If such modules are on either initrd or rootfs, they will be
893  * loaded before control is passed to userland.
894  */
895 void __init load_default_modules(void)
896 {
897         load_default_elevator_module();
898 }
899 
900 static int run_init_process(const char *init_filename)
901 {
902         argv_init[0] = init_filename;
903         return do_execve(getname_kernel(init_filename),
904                 (const char __user *const __user *)argv_init,
905                 (const char __user *const __user *)envp_init);
906 }
907 
908 static int try_to_run_init_process(const char *init_filename)
909 {
910         int ret;
911 
912         ret = run_init_process(init_filename);
913 
914         if (ret && ret != -ENOENT) {
915                 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
916                        init_filename, ret);
917         }
918 
919         return ret;
920 }
921 
922 static noinline void __init kernel_init_freeable(void);
923 
924 static int __ref kernel_init(void *unused)
925 {
926         int ret;
927 
928         kernel_init_freeable();
929         /* need to finish all async __init code before freeing the memory */
930         async_synchronize_full();
931         free_initmem();
932         mark_rodata_ro();
933         system_state = SYSTEM_RUNNING;
934         numa_default_policy();
935 
936         flush_delayed_fput();
937 
938         if (ramdisk_execute_command) {
939                 ret = run_init_process(ramdisk_execute_command);
940                 if (!ret)
941                         return 0;
942                 pr_err("Failed to execute %s (error %d)\n",
943                        ramdisk_execute_command, ret);
944         }
945 
946         /*
947          * We try each of these until one succeeds.
948          *
949          * The Bourne shell can be used instead of init if we are
950          * trying to recover a really broken machine.
951          */
952         if (execute_command) {
953                 ret = run_init_process(execute_command);
954                 if (!ret)
955                         return 0;
956                 panic("Requested init %s failed (error %d).",
957                       execute_command, ret);
958         }
959         if (!try_to_run_init_process("/sbin/init") ||
960             !try_to_run_init_process("/etc/init") ||
961             !try_to_run_init_process("/bin/init") ||
962             !try_to_run_init_process("/bin/sh"))
963                 return 0;
964 
965         panic("No working init found.  Try passing init= option to kernel. "
966               "See Linux Documentation/init.txt for guidance.");
967 }
968 
969 static noinline void __init kernel_init_freeable(void)
970 {
971         /*
972          * Wait until kthreadd is all set-up.
973          */
974         wait_for_completion(&kthreadd_done);
975 
976         /* Now the scheduler is fully set up and can do blocking allocations */
977         gfp_allowed_mask = __GFP_BITS_MASK;
978 
979         /*
980          * init can allocate pages on any node
981          */
982         set_mems_allowed(node_states[N_MEMORY]);
983         /*
984          * init can run on any cpu.
985          */
986         set_cpus_allowed_ptr(current, cpu_all_mask);
987 
988         cad_pid = task_pid(current);
989 
990         smp_prepare_cpus(setup_max_cpus);
991 
992         do_pre_smp_initcalls();
993         lockup_detector_init();
994 
995         smp_init();
996         sched_init_smp();
997 
998         do_basic_setup();
999 
1000         /* Open the /dev/console on the rootfs, this should never fail */
1001         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1002                 pr_err("Warning: unable to open an initial console.\n");
1003 
1004         (void) sys_dup(0);
1005         (void) sys_dup(0);
1006         /*
1007          * check if there is an early userspace init.  If yes, let it do all
1008          * the work
1009          */
1010 
1011         if (!ramdisk_execute_command)
1012                 ramdisk_execute_command = "/init";
1013 
1014         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
1015                 ramdisk_execute_command = NULL;
1016                 prepare_namespace();
1017         }
1018 
1019         /*
1020          * Ok, we have completed the initial bootup, and
1021          * we're essentially up and running. Get rid of the
1022          * initmem segments and start the user-mode stuff..
1023          *
1024          * rootfs is available now, try loading the public keys
1025          * and default modules
1026          */
1027 
1028         integrity_load_keys();
1029         load_default_modules();
1030 }
1031 

This page was automatically generated by LXR 0.3.1 (source).  •  Linux is a registered trademark of Linus Torvalds  •  Contact us