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/extable.h>
  16 #include <linux/module.h>
  17 #include <linux/proc_fs.h>
  18 #include <linux/binfmts.h>
  19 #include <linux/kernel.h>
  20 #include <linux/syscalls.h>
  21 #include <linux/stackprotector.h>
  22 #include <linux/string.h>
  23 #include <linux/ctype.h>
  24 #include <linux/delay.h>
  25 #include <linux/ioport.h>
  26 #include <linux/init.h>
  27 #include <linux/initrd.h>
  28 #include <linux/bootmem.h>
  29 #include <linux/acpi.h>
  30 #include <linux/console.h>
  31 #include <linux/nmi.h>
  32 #include <linux/percpu.h>
  33 #include <linux/kmod.h>
  34 #include <linux/vmalloc.h>
  35 #include <linux/kernel_stat.h>
  36 #include <linux/start_kernel.h>
  37 #include <linux/security.h>
  38 #include <linux/smp.h>
  39 #include <linux/profile.h>
  40 #include <linux/rcupdate.h>
  41 #include <linux/moduleparam.h>
  42 #include <linux/kallsyms.h>
  43 #include <linux/writeback.h>
  44 #include <linux/cpu.h>
  45 #include <linux/cpuset.h>
  46 #include <linux/cgroup.h>
  47 #include <linux/efi.h>
  48 #include <linux/tick.h>
  49 #include <linux/sched/isolation.h>
  50 #include <linux/interrupt.h>
  51 #include <linux/taskstats_kern.h>
  52 #include <linux/delayacct.h>
  53 #include <linux/unistd.h>
  54 #include <linux/rmap.h>
  55 #include <linux/mempolicy.h>
  56 #include <linux/key.h>
  57 #include <linux/buffer_head.h>
  58 #include <linux/page_ext.h>
  59 #include <linux/debug_locks.h>
  60 #include <linux/debugobjects.h>
  61 #include <linux/lockdep.h>
  62 #include <linux/kmemleak.h>
  63 #include <linux/pid_namespace.h>
  64 #include <linux/device.h>
  65 #include <linux/kthread.h>
  66 #include <linux/sched.h>
  67 #include <linux/sched/init.h>
  68 #include <linux/signal.h>
  69 #include <linux/idr.h>
  70 #include <linux/kgdb.h>
  71 #include <linux/ftrace.h>
  72 #include <linux/async.h>
  73 #include <linux/sfi.h>
  74 #include <linux/shmem_fs.h>
  75 #include <linux/slab.h>
  76 #include <linux/perf_event.h>
  77 #include <linux/ptrace.h>
  78 #include <linux/pti.h>
  79 #include <linux/blkdev.h>
  80 #include <linux/elevator.h>
  81 #include <linux/sched_clock.h>
  82 #include <linux/sched/task.h>
  83 #include <linux/sched/task_stack.h>
  84 #include <linux/context_tracking.h>
  85 #include <linux/random.h>
  86 #include <linux/list.h>
  87 #include <linux/integrity.h>
  88 #include <linux/proc_ns.h>
  89 #include <linux/io.h>
  90 #include <linux/cache.h>
  91 #include <linux/rodata_test.h>
  92 #include <linux/jump_label.h>
  93
  94 #include <asm/io.h>
  95 #include <asm/bugs.h>
  96 #include <asm/setup.h>
  97 #include <asm/sections.h>
  98 #include <asm/cacheflush.h>
  99
 100 static int kernel_init(void *);
 101
 102 extern void init_IRQ(void);
 103 extern void fork_init(void);
 104 extern void radix_tree_init(void);
 105
 106 /*
 107  * Debug helper: via this flag we know that we are in 'early bootup code'
 108  * where only the boot processor is running with IRQ disabled.  This means
 109  * two things - IRQ must not be enabled before the flag is cleared and some
 110  * operations which are not allowed with IRQ disabled are allowed while the
 111  * flag is set.
 112  */
 113 bool early_boot_irqs_disabled __read_mostly;
 114
 115 enum system_states system_state __read_mostly;
 116 EXPORT_SYMBOL(system_state);
 117
 118 /*
 119  * Boot command-line arguments
 120  */
 121 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 122 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 123
 124 extern void time_init(void);
 125 /* Default late time init is NULL. archs can override this later. */
 126 void (*__initdata late_time_init)(void);
 127
 128 /* Untouched command line saved by arch-specific code. */
 129 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 130 /* Untouched saved command line (eg. for /proc) */
 131 char *saved_command_line;
 132 /* Command line for parameter parsing */
 133 static char *static_command_line;
 134 /* Command line for per-initcall parameter parsing */
 135 static char *initcall_command_line;
 136
 137 static char *execute_command;
 138 static char *ramdisk_execute_command;
 139
 140 /*
 141  * Used to generate warnings if static_key manipulation functions are used
 142  * before jump_label_init is called.
 143  */
 144 bool static_key_initialized __read_mostly;
 145 EXPORT_SYMBOL_GPL(static_key_initialized);
 146
 147 /*
 148  * If set, this is an indication to the drivers that reset the underlying
 149  * device before going ahead with the initialization otherwise driver might
 150  * rely on the BIOS and skip the reset operation.
 151  *
 152  * This is useful if kernel is booting in an unreliable environment.
 153  * For ex. kdump situation where previous kernel has crashed, BIOS has been
 154  * skipped and devices will be in unknown state.
 155  */
 156 unsigned int reset_devices;
 157 EXPORT_SYMBOL(reset_devices);
 158
 159 static int __init set_reset_devices(char *str)
 160 {
 161         reset_devices = 1;
 162         return 1;
 163 }
 164
 165 __setup("reset_devices", set_reset_devices);
 166
 167 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 168 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 169 static const char *panic_later, *panic_param;
 170
 171 extern const struct obs_kernel_param __setup_start[], __setup_end[];
 172
 173 static bool __init obsolete_checksetup(char *line)
 174 {
 175         const struct obs_kernel_param *p;
 176         bool had_early_param = false;
 177
 178         p = __setup_start;
 179         do {
 180                 int n = strlen(p->str);
 181                 if (parameqn(line, p->str, n)) {
 182                         if (p->early) {
 183                                 /* Already done in parse_early_param?
 184                                  * (Needs exact match on param part).
 185                                  * Keep iterating, as we can have early
 186                                  * params and __setups of same names 8( */
 187                                 if (line[n] == '\0' || line[n] == '=')
 188                                         had_early_param = true;
 189                         } else if (!p->setup_func) {
 190                                 pr_warn("Parameter %s is obsolete, ignored\n",
 191                                         p->str);
 192                                 return true;
 193                         } else if (p->setup_func(line + n))
 194                                 return true;
 195                 }
 196                 p++;
 197         } while (p < __setup_end);
 198
 199         return had_early_param;
 200 }
 201
 202 /*
 203  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 204  * still work even if initially too large, it will just take slightly longer
 205  */
 206 unsigned long loops_per_jiffy = (1<<12);
 207 EXPORT_SYMBOL(loops_per_jiffy);
 208
 209 static int __init debug_kernel(char *str)
 210 {
 211         console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
 212         return 0;
 213 }
 214
 215 static int __init quiet_kernel(char *str)
 216 {
 217         console_loglevel = CONSOLE_LOGLEVEL_QUIET;
 218         return 0;
 219 }
 220
 221 early_param("debug", debug_kernel);
 222 early_param("quiet", quiet_kernel);
 223
 224 static int __init loglevel(char *str)
 225 {
 226         int newlevel;
 227
 228         /*
 229          * Only update loglevel value when a correct setting was passed,
 230          * to prevent blind crashes (when loglevel being set to 0) that
 231          * are quite hard to debug
 232          */
 233         if (get_option(&str, &newlevel)) {
 234                 console_loglevel = newlevel;
 235                 return 0;
 236         }
 237
 238         return -EINVAL;
 239 }
 240
 241 early_param("loglevel", loglevel);
 242
 243 /* Change NUL term back to "=", to make "param" the whole string. */
 244 static int __init repair_env_string(char *param, char *val,
 245                                     const char *unused, void *arg)
 246 {
 247         if (val) {
 248                 /* param=val or param="val"? */
 249                 if (val == param+strlen(param)+1)
 250                         val[-1] = '=';
 251                 else if (val == param+strlen(param)+2) {
 252                         val[-2] = '=';
 253                         memmove(val-1, val, strlen(val)+1);
 254                         val--;
 255                 } else
 256                         BUG();
 257         }
 258         return 0;
 259 }
 260
 261 /* Anything after -- gets handed straight to init. */
 262 static int __init set_init_arg(char *param, char *val,
 263                                const char *unused, void *arg)
 264 {
 265         unsigned int i;
 266
 267         if (panic_later)
 268                 return 0;
 269
 270         repair_env_string(param, val, unused, NULL);
 271
 272         for (i = 0; argv_init[i]; i++) {
 273                 if (i == MAX_INIT_ARGS) {
 274                         panic_later = "init";
 275                         panic_param = param;
 276                         return 0;
 277                 }
 278         }
 279         argv_init[i] = param;
 280         return 0;
 281 }
 282
 283 /*
 284  * Unknown boot options get handed to init, unless they look like
 285  * unused parameters (modprobe will find them in /proc/cmdline).
 286  */
 287 static int __init unknown_bootoption(char *param, char *val,
 288                                      const char *unused, void *arg)
 289 {
 290         repair_env_string(param, val, unused, NULL);
 291
 292         /* Handle obsolete-style parameters */
 293         if (obsolete_checksetup(param))
 294                 return 0;
 295
 296         /* Unused module parameter. */
 297         if (strchr(param, '.') && (!val || strchr(param, '.') < val))
 298                 return 0;
 299
 300         if (panic_later)
 301                 return 0;
 302
 303         if (val) {
 304                 /* Environment option */
 305                 unsigned int i;
 306                 for (i = 0; envp_init[i]; i++) {
 307                         if (i == MAX_INIT_ENVS) {
 308                                 panic_later = "env";
 309                                 panic_param = param;
 310                         }
 311                         if (!strncmp(param, envp_init[i], val - param))
 312                                 break;
 313                 }
 314                 envp_init[i] = param;
 315         } else {
 316                 /* Command line option */
 317                 unsigned int i;
 318                 for (i = 0; argv_init[i]; i++) {
 319                         if (i == MAX_INIT_ARGS) {
 320                                 panic_later = "init";
 321                                 panic_param = param;
 322                         }
 323                 }
 324                 argv_init[i] = param;
 325         }
 326         return 0;
 327 }
 328
 329 static int __init init_setup(char *str)
 330 {
 331         unsigned int i;
 332
 333         execute_command = str;
 334         /*
 335          * In case LILO is going to boot us with default command line,
 336          * it prepends "auto" before the whole cmdline which makes
 337          * the shell think it should execute a script with such name.
 338          * So we ignore all arguments entered _before_ init=... [MJ]
 339          */
 340         for (i = 1; i < MAX_INIT_ARGS; i++)
 341                 argv_init[i] = NULL;
 342         return 1;
 343 }
 344 __setup("init=", init_setup);
 345
 346 static int __init rdinit_setup(char *str)
 347 {
 348         unsigned int i;
 349
 350         ramdisk_execute_command = str;
 351         /* See "auto" comment in init_setup */
 352         for (i = 1; i < MAX_INIT_ARGS; i++)
 353                 argv_init[i] = NULL;
 354         return 1;
 355 }
 356 __setup("rdinit=", rdinit_setup);
 357
 358 #ifndef CONFIG_SMP
 359 static const unsigned int setup_max_cpus = NR_CPUS;
 360 static inline void setup_nr_cpu_ids(void) { }
 361 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 362 #endif
 363
 364 /*
 365  * We need to store the untouched command line for future reference.
 366  * We also need to store the touched command line since the parameter
 367  * parsing is performed in place, and we should allow a component to
 368  * store reference of name/value for future reference.
 369  */
 370 static void __init setup_command_line(char *command_line)
 371 {
 372         saved_command_line =
 373                 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
 374         initcall_command_line =
 375                 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
 376         static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0);
 377         strcpy(saved_command_line, boot_command_line);
 378         strcpy(static_command_line, command_line);
 379 }
 380
 381 /*
 382  * We need to finalize in a non-__init function or else race conditions
 383  * between the root thread and the init thread may cause start_kernel to
 384  * be reaped by free_initmem before the root thread has proceeded to
 385  * cpu_idle.
 386  *
 387  * gcc-3.4 accidentally inlines this function, so use noinline.
 388  */
 389
 390 static __initdata DECLARE_COMPLETION(kthreadd_done);
 391
 392 static noinline void __ref rest_init(void)
 393 {
 394         struct task_struct *tsk;
 395         int pid;
 396
 397         rcu_scheduler_starting();
 398         /*
 399          * We need to spawn init first so that it obtains pid 1, however
 400          * the init task will end up wanting to create kthreads, which, if
 401          * we schedule it before we create kthreadd, will OOPS.
 402          */
 403         pid = kernel_thread(kernel_init, NULL, CLONE_FS);
 404         /*
 405          * Pin init on the boot CPU. Task migration is not properly working
 406          * until sched_init_smp() has been run. It will set the allowed
 407          * CPUs for init to the non isolated CPUs.
 408          */
 409         rcu_read_lock();
 410         tsk = find_task_by_pid_ns(pid, &init_pid_ns);
 411         set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
 412         rcu_read_unlock();
 413
 414         numa_default_policy();
 415         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 416         rcu_read_lock();
 417         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 418         rcu_read_unlock();
 419
 420         /*
 421          * Enable might_sleep() and smp_processor_id() checks.
 422          * They cannot be enabled earlier because with CONFIG_PRREMPT=y
 423          * kernel_thread() would trigger might_sleep() splats. With
 424          * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
 425          * already, but it's stuck on the kthreadd_done completion.
 426          */
 427         system_state = SYSTEM_SCHEDULING;
 428
 429         complete(&kthreadd_done);
 430
 431         /*
 432          * The boot idle thread must execute schedule()
 433          * at least once to get things moving:
 434          */
 435         schedule_preempt_disabled();
 436         /* Call into cpu_idle with preempt disabled */
 437         cpu_startup_entry(CPUHP_ONLINE);
 438 }
 439
 440 /* Check for early params. */
 441 static int __init do_early_param(char *param, char *val,
 442                                  const char *unused, void *arg)
 443 {
 444         const struct obs_kernel_param *p;
 445
 446         for (p = __setup_start; p < __setup_end; p++) {
 447                 if ((p->early && parameq(param, p->str)) ||
 448                     (strcmp(param, "console") == 0 &&
 449                      strcmp(p->str, "earlycon") == 0)
 450                 ) {
 451                         if (p->setup_func(val) != 0)
 452                                 pr_warn("Malformed early option '%s'\n", param);
 453                 }
 454         }
 455         /* We accept everything at this stage. */
 456         return 0;
 457 }
 458
 459 void __init parse_early_options(char *cmdline)
 460 {
 461         parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
 462                    do_early_param);
 463 }
 464
 465 /* Arch code calls this early on, or if not, just before other parsing. */
 466 void __init parse_early_param(void)
 467 {
 468         static int done __initdata;
 469         static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
 470
 471         if (done)
 472                 return;
 473
 474         /* All fall through to do_early_param. */
 475         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 476         parse_early_options(tmp_cmdline);
 477         done = 1;
 478 }
 479
 480 void __init __weak arch_post_acpi_subsys_init(void) { }
 481
 482 void __init __weak smp_setup_processor_id(void)
 483 {
 484 }
 485
 486 # if THREAD_SIZE >= PAGE_SIZE
 487 void __init __weak thread_stack_cache_init(void)
 488 {
 489 }
 490 #endif
 491
 492 void __init __weak mem_encrypt_init(void) { }
 493
 494 /*
 495  * Set up kernel memory allocators
 496  */
 497 static void __init mm_init(void)
 498 {
 499         /*
 500          * page_ext requires contiguous pages,
 501          * bigger than MAX_ORDER unless SPARSEMEM.
 502          */
 503         page_ext_init_flatmem();
 504         mem_init();
 505         kmem_cache_init();
 506         pgtable_init();
 507         vmalloc_init();
 508         ioremap_huge_init();
 509         /* Should be run before the first non-init thread is created */
 510         init_espfix_bsp();
 511         /* Should be run after espfix64 is set up. */
 512         pti_init();
 513 }
 514
 515 asmlinkage __visible void __init start_kernel(void)
 516 {
 517         char *command_line;
 518         char *after_dashes;
 519
 520         set_task_stack_end_magic(&init_task);
 521         smp_setup_processor_id();
 522         debug_objects_early_init();
 523
 524         cgroup_init_early();
 525
 526         local_irq_disable();
 527         early_boot_irqs_disabled = true;
 528
 529         /*
 530          * Interrupts are still disabled. Do necessary setups, then
 531          * enable them.
 532          */
 533         boot_cpu_init();
 534         page_address_init();
 535         pr_notice("%s", linux_banner);
 536         setup_arch(&command_line);
 537         /*
 538          * Set up the the initial canary and entropy after arch
 539          * and after adding latent and command line entropy.
 540          */
 541         add_latent_entropy();
 542         add_device_randomness(command_line, strlen(command_line));
 543         boot_init_stack_canary();
 544         mm_init_cpumask(&init_mm);
 545         setup_command_line(command_line);
 546         setup_nr_cpu_ids();
 547         setup_per_cpu_areas();
 548         boot_cpu_state_init();
 549         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 550
 551         build_all_zonelists(NULL);
 552         page_alloc_init();
 553
 554         pr_notice("Kernel command line: %s\n", boot_command_line);
 555         parse_early_param();
 556         after_dashes = parse_args("Booting kernel",
 557                                   static_command_line, __start___param,
 558                                   __stop___param - __start___param,
 559                                   -1, -1, NULL, &unknown_bootoption);
 560         if (!IS_ERR_OR_NULL(after_dashes))
 561                 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
 562                            NULL, set_init_arg);
 563
 564         jump_label_init();
 565
 566         /*
 567          * These use large bootmem allocations and must precede
 568          * kmem_cache_init()
 569          */
 570         setup_log_buf(0);
 571         vfs_caches_init_early();
 572         sort_main_extable();
 573         trap_init();
 574         mm_init();
 575
 576         ftrace_init();
 577
 578         /* trace_printk can be enabled here */
 579         early_trace_init();
 580
 581         /*
 582          * Set up the scheduler prior starting any interrupts (such as the
 583          * timer interrupt). Full topology setup happens at smp_init()
 584          * time - but meanwhile we still have a functioning scheduler.
 585          */
 586         sched_init();
 587         /*
 588          * Disable preemption - early bootup scheduling is extremely
 589          * fragile until we cpu_idle() for the first time.
 590          */
 591         preempt_disable();
 592         if (WARN(!irqs_disabled(),
 593                  "Interrupts were enabled *very* early, fixing it\n"))
 594                 local_irq_disable();
 595         radix_tree_init();
 596
 597         /*
 598          * Set up housekeeping before setting up workqueues to allow the unbound
 599          * workqueue to take non-housekeeping into account.
 600          */
 601         housekeeping_init();
 602
 603         /*
 604          * Allow workqueue creation and work item queueing/cancelling
 605          * early.  Work item execution depends on kthreads and starts after
 606          * workqueue_init().
 607          */
 608         workqueue_init_early();
 609
 610         rcu_init();
 611
 612         /* Trace events are available after this */
 613         trace_init();
 614
 615         context_tracking_init();
 616         /* init some links before init_ISA_irqs() */
 617         early_irq_init();
 618         init_IRQ();
 619         tick_init();
 620         rcu_init_nohz();
 621         init_timers();
 622         hrtimers_init();
 623         softirq_init();
 624         timekeeping_init();
 625         time_init();
 626         sched_clock_postinit();
 627         printk_safe_init();
 628         perf_event_init();
 629         profile_init();
 630         call_function_init();
 631         WARN(!irqs_disabled(), "Interrupts were enabled early\n");
 632         early_boot_irqs_disabled = false;
 633         local_irq_enable();
 634
 635         kmem_cache_init_late();
 636
 637         /*
 638          * HACK ALERT! This is early. We're enabling the console before
 639          * we've done PCI setups etc, and console_init() must be aware of
 640          * this. But we do want output early, in case something goes wrong.
 641          */
 642         console_init();
 643         if (panic_later)
 644                 panic("Too many boot %s vars at `%s'", panic_later,
 645                       panic_param);
 646
 647         lockdep_info();
 648
 649         /*
 650          * Need to run this when irqs are enabled, because it wants
 651          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 652          * too:
 653          */
 654         locking_selftest();
 655
 656         /*
 657          * This needs to be called before any devices perform DMA
 658          * operations that might use the SWIOTLB bounce buffers. It will
 659          * mark the bounce buffers as decrypted so that their usage will
 660          * not cause "plain-text" data to be decrypted when accessed.
 661          */
 662         mem_encrypt_init();
 663
 664 #ifdef CONFIG_BLK_DEV_INITRD
 665         if (initrd_start && !initrd_below_start_ok &&
 666             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 667                 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
 668                     page_to_pfn(virt_to_page((void *)initrd_start)),
 669                     min_low_pfn);
 670                 initrd_start = 0;
 671         }
 672 #endif
 673         page_ext_init();
 674         kmemleak_init();
 675         debug_objects_mem_init();
 676         setup_per_cpu_pageset();
 677         numa_policy_init();
 678         acpi_early_init();
 679         if (late_time_init)
 680                 late_time_init();
 681         calibrate_delay();
 682         pid_idr_init();
 683         anon_vma_init();
 684 #ifdef CONFIG_X86
 685         if (efi_enabled(EFI_RUNTIME_SERVICES))
 686                 efi_enter_virtual_mode();
 687 #endif
 688         thread_stack_cache_init();
 689         cred_init();
 690         fork_init();
 691         proc_caches_init();
 692         buffer_init();
 693         key_init();
 694         security_init();
 695         dbg_late_init();
 696         vfs_caches_init();
 697         pagecache_init();
 698         signals_init();
 699         proc_root_init();
 700         nsfs_init();
 701         cpuset_init();
 702         cgroup_init();
 703         taskstats_init_early();
 704         delayacct_init();
 705
 706         check_bugs();
 707
 708         acpi_subsystem_init();
 709         arch_post_acpi_subsys_init();
 710         sfi_init_late();
 711
 712         if (efi_enabled(EFI_RUNTIME_SERVICES)) {
 713                 efi_free_boot_services();
 714         }
 715
 716         /* Do the rest non-__init'ed, we're now alive */
 717         rest_init();
 718 }
 719
 720 /* Call all constructor functions linked into the kernel. */
 721 static void __init do_ctors(void)
 722 {
 723 #ifdef CONFIG_CONSTRUCTORS
 724         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
 725
 726         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
 727                 (*fn)();
 728 #endif
 729 }
 730
 731 bool initcall_debug;
 732 core_param(initcall_debug, initcall_debug, bool, 0644);
 733
 734 #ifdef CONFIG_KALLSYMS
 735 struct blacklist_entry {
 736         struct list_head next;
 737         char *buf;
 738 };
 739
 740 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
 741
 742 static int __init initcall_blacklist(char *str)
 743 {
 744         char *str_entry;
 745         struct blacklist_entry *entry;
 746
 747         /* str argument is a comma-separated list of functions */
 748         do {
 749                 str_entry = strsep(&str, ",");
 750                 if (str_entry) {
 751                         pr_debug("blacklisting initcall %s\n", str_entry);
 752                         entry = alloc_bootmem(sizeof(*entry));
 753                         entry->buf = alloc_bootmem(strlen(str_entry) + 1);
 754                         strcpy(entry->buf, str_entry);
 755                         list_add(&entry->next, &blacklisted_initcalls);
 756                 }
 757         } while (str_entry);
 758
 759         return 0;
 760 }
 761
 762 static bool __init_or_module initcall_blacklisted(initcall_t fn)
 763 {
 764         struct blacklist_entry *entry;
 765         char fn_name[KSYM_SYMBOL_LEN];
 766         unsigned long addr;
 767
 768         if (list_empty(&blacklisted_initcalls))
 769                 return false;
 770
 771         addr = (unsigned long) dereference_function_descriptor(fn);
 772         sprint_symbol_no_offset(fn_name, addr);
 773
 774         /*
 775          * fn will be "function_name [module_name]" where [module_name] is not
 776          * displayed for built-in init functions.  Strip off the [module_name].
 777          */
 778         strreplace(fn_name, ' ', '\0');
 779
 780         list_for_each_entry(entry, &blacklisted_initcalls, next) {
 781                 if (!strcmp(fn_name, entry->buf)) {
 782                         pr_debug("initcall %s blacklisted\n", fn_name);
 783                         return true;
 784                 }
 785         }
 786
 787         return false;
 788 }
 789 #else
 790 static int __init initcall_blacklist(char *str)
 791 {
 792         pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
 793         return 0;
 794 }
 795
 796 static bool __init_or_module initcall_blacklisted(initcall_t fn)
 797 {
 798         return false;
 799 }
 800 #endif
 801 __setup("initcall_blacklist=", initcall_blacklist);
 802
 803 static int __init_or_module do_one_initcall_debug(initcall_t fn)
 804 {
 805         ktime_t calltime, delta, rettime;
 806         unsigned long long duration;
 807         int ret;
 808
 809         printk(KERN_DEBUG "calling  %pF @ %i\n", fn, task_pid_nr(current));
 810         calltime = ktime_get();
 811         ret = fn();
 812         rettime = ktime_get();
 813         delta = ktime_sub(rettime, calltime);
 814         duration = (unsigned long long) ktime_to_ns(delta) >> 10;
 815         printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n",
 816                  fn, ret, duration);
 817
 818         return ret;
 819 }
 820
 821 int __init_or_module do_one_initcall(initcall_t fn)
 822 {
 823         int count = preempt_count();
 824         int ret;
 825         char msgbuf[64];
 826
 827         if (initcall_blacklisted(fn))
 828                 return -EPERM;
 829
 830         if (initcall_debug)
 831                 ret = do_one_initcall_debug(fn);
 832         else
 833                 ret = fn();
 834
 835         msgbuf[0] = 0;
 836
 837         if (preempt_count() != count) {
 838                 sprintf(msgbuf, "preemption imbalance ");
 839                 preempt_count_set(count);
 840         }
 841         if (irqs_disabled()) {
 842                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 843                 local_irq_enable();
 844         }
 845         WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
 846
 847         add_latent_entropy();
 848         return ret;
 849 }
 850
 851
 852 extern initcall_t __initcall_start[];
 853 extern initcall_t __initcall0_start[];
 854 extern initcall_t __initcall1_start[];
 855 extern initcall_t __initcall2_start[];
 856 extern initcall_t __initcall3_start[];
 857 extern initcall_t __initcall4_start[];
 858 extern initcall_t __initcall5_start[];
 859 extern initcall_t __initcall6_start[];
 860 extern initcall_t __initcall7_start[];
 861 extern initcall_t __initcall_end[];
 862
 863 static initcall_t *initcall_levels[] __initdata = {
 864         __initcall0_start,
 865         __initcall1_start,
 866         __initcall2_start,
 867         __initcall3_start,
 868         __initcall4_start,
 869         __initcall5_start,
 870         __initcall6_start,
 871         __initcall7_start,
 872         __initcall_end,
 873 };
 874
 875 /* Keep these in sync with initcalls in include/linux/init.h */
 876 static char *initcall_level_names[] __initdata = {
 877         "early",
 878         "core",
 879         "postcore",
 880         "arch",
 881         "subsys",
 882         "fs",
 883         "device",
 884         "late",
 885 };
 886
 887 static void __init do_initcall_level(int level)
 888 {
 889         initcall_t *fn;
 890
 891         strcpy(initcall_command_line, saved_command_line);
 892         parse_args(initcall_level_names[level],
 893                    initcall_command_line, __start___param,
 894                    __stop___param - __start___param,
 895                    level, level,
 896                    NULL, &repair_env_string);
 897
 898         for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
 899                 do_one_initcall(*fn);
 900 }
 901
 902 static void __init do_initcalls(void)
 903 {
 904         int level;
 905
 906         for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
 907                 do_initcall_level(level);
 908 }
 909
 910 /*
 911  * Ok, the machine is now initialized. None of the devices
 912  * have been touched yet, but the CPU subsystem is up and
 913  * running, and memory and process management works.
 914  *
 915  * Now we can finally start doing some real work..
 916  */
 917 static void __init do_basic_setup(void)
 918 {
 919         cpuset_init_smp();
 920         shmem_init();
 921         driver_init();
 922         init_irq_proc();
 923         do_ctors();
 924         usermodehelper_enable();
 925         do_initcalls();
 926 }
 927
 928 static void __init do_pre_smp_initcalls(void)
 929 {
 930         initcall_t *fn;
 931
 932         for (fn = __initcall_start; fn < __initcall0_start; fn++)
 933                 do_one_initcall(*fn);
 934 }
 935
 936 /*
 937  * This function requests modules which should be loaded by default and is
 938  * called twice right after initrd is mounted and right before init is
 939  * exec'd.  If such modules are on either initrd or rootfs, they will be
 940  * loaded before control is passed to userland.
 941  */
 942 void __init load_default_modules(void)
 943 {
 944         load_default_elevator_module();
 945 }
 946
 947 static int run_init_process(const char *init_filename)
 948 {
 949         argv_init[0] = init_filename;
 950         return do_execve(getname_kernel(init_filename),
 951                 (const char __user *const __user *)argv_init,
 952                 (const char __user *const __user *)envp_init);
 953 }
 954
 955 static int try_to_run_init_process(const char *init_filename)
 956 {
 957         int ret;
 958
 959         ret = run_init_process(init_filename);
 960
 961         if (ret && ret != -ENOENT) {
 962                 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
 963                        init_filename, ret);
 964         }
 965
 966         return ret;
 967 }
 968
 969 static noinline void __init kernel_init_freeable(void);
 970
 971 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
 972 bool rodata_enabled __ro_after_init = true;
 973 static int __init set_debug_rodata(char *str)
 974 {
 975         return strtobool(str, &rodata_enabled);
 976 }
 977 __setup("rodata=", set_debug_rodata);
 978 #endif
 979
 980 #ifdef CONFIG_STRICT_KERNEL_RWX
 981 static void mark_readonly(void)
 982 {
 983         if (rodata_enabled) {
 984                 mark_rodata_ro();
 985                 rodata_test();
 986         } else
 987                 pr_info("Kernel memory protection disabled.\n");
 988 }
 989 #else
 990 static inline void mark_readonly(void)
 991 {
 992         pr_warn("This architecture does not have kernel memory protection.\n");
 993 }
 994 #endif
 995
 996 static int __ref kernel_init(void *unused)
 997 {
 998         int ret;
 999
1000         kernel_init_freeable();
1001         /* need to finish all async __init code before freeing the memory */
1002         async_synchronize_full();
1003         ftrace_free_init_mem();
1004         jump_label_invalidate_init();
1005         free_initmem();
1006         mark_readonly();
1007         system_state = SYSTEM_RUNNING;
1008         numa_default_policy();
1009
1010         rcu_end_inkernel_boot();
1011
1012         if (ramdisk_execute_command) {
1013                 ret = run_init_process(ramdisk_execute_command);
1014                 if (!ret)
1015                         return 0;
1016                 pr_err("Failed to execute %s (error %d)\n",
1017                        ramdisk_execute_command, ret);
1018         }
1019
1020         /*
1021          * We try each of these until one succeeds.
1022          *
1023          * The Bourne shell can be used instead of init if we are
1024          * trying to recover a really broken machine.
1025          */
1026         if (execute_command) {
1027                 ret = run_init_process(execute_command);
1028                 if (!ret)
1029                         return 0;
1030                 panic("Requested init %s failed (error %d).",
1031                       execute_command, ret);
1032         }
1033         if (!try_to_run_init_process("/sbin/init") ||
1034             !try_to_run_init_process("/etc/init") ||
1035             !try_to_run_init_process("/bin/init") ||
1036             !try_to_run_init_process("/bin/sh"))
1037                 return 0;
1038
1039         panic("No working init found.  Try passing init= option to kernel. "
1040               "See Linux Documentation/admin-guide/init.rst for guidance.");
1041 }
1042
1043 static noinline void __init kernel_init_freeable(void)
1044 {
1045         /*
1046          * Wait until kthreadd is all set-up.
1047          */
1048         wait_for_completion(&kthreadd_done);
1049
1050         /* Now the scheduler is fully set up and can do blocking allocations */
1051         gfp_allowed_mask = __GFP_BITS_MASK;
1052
1053         /*
1054          * init can allocate pages on any node
1055          */
1056         set_mems_allowed(node_states[N_MEMORY]);
1057
1058         cad_pid = task_pid(current);
1059
1060         smp_prepare_cpus(setup_max_cpus);
1061
1062         workqueue_init();
1063
1064         init_mm_internals();
1065
1066         do_pre_smp_initcalls();
1067         lockup_detector_init();
1068
1069         smp_init();
1070         sched_init_smp();
1071
1072         page_alloc_init_late();
1073
1074         do_basic_setup();
1075
1076         /* Open the /dev/console on the rootfs, this should never fail */
1077         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1078                 pr_err("Warning: unable to open an initial console.\n");
1079
1080         (void) sys_dup(0);
1081         (void) sys_dup(0);
1082         /*
1083          * check if there is an early userspace init.  If yes, let it do all
1084          * the work
1085          */
1086
1087         if (!ramdisk_execute_command)
1088                 ramdisk_execute_command = "/init";
1089
1090         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
1091                 ramdisk_execute_command = NULL;
1092                 prepare_namespace();
1093         }
1094
1095         /*
1096          * Ok, we have completed the initial bootup, and
1097          * we're essentially up and running. Get rid of the
1098          * initmem segments and start the user-mode stuff..
1099          *
1100          * rootfs is available now, try loading the public keys
1101          * and default modules
1102          */
1103
1104         integrity_load_keys();
1105         load_default_modules();
1106 }