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