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/interrupt.h>
  50 #include <linux/taskstats_kern.h>
  51 #include <linux/delayacct.h>
  52 #include <linux/unistd.h>
  53 #include <linux/rmap.h>
  54 #include <linux/mempolicy.h>
  55 #include <linux/key.h>
  56 #include <linux/buffer_head.h>
  57 #include <linux/page_ext.h>
  58 #include <linux/debug_locks.h>
  59 #include <linux/debugobjects.h>
  60 #include <linux/lockdep.h>
  61 #include <linux/kmemleak.h>
  62 #include <linux/pid_namespace.h>
  63 #include <linux/device.h>
  64 #include <linux/kthread.h>
  65 #include <linux/sched.h>
  66 #include <linux/sched/init.h>
  67 #include <linux/signal.h>
  68 #include <linux/idr.h>
  69 #include <linux/kgdb.h>
  70 #include <linux/ftrace.h>
  71 #include <linux/async.h>
  72 #include <linux/sfi.h>
  73 #include <linux/shmem_fs.h>
  74 #include <linux/slab.h>
  75 #include <linux/perf_event.h>
  76 #include <linux/ptrace.h>
  77 #include <linux/pti.h>
  78 #include <linux/blkdev.h>
  79 #include <linux/elevator.h>
  80 #include <linux/sched_clock.h>
  81 #include <linux/sched/task.h>
  82 #include <linux/sched/task_stack.h>
  83 #include <linux/context_tracking.h>
  84 #include <linux/random.h>
  85 #include <linux/list.h>
  86 #include <linux/integrity.h>
  87 #include <linux/proc_ns.h>
  88 #include <linux/io.h>
  89 #include <linux/cache.h>
  90 #include <linux/rodata_test.h>
  91
  92 #include <asm/io.h>
  93 #include <asm/bugs.h>
  94 #include <asm/setup.h>
  95 #include <asm/sections.h>
  96 #include <asm/cacheflush.h>
  97
  98 static int kernel_init(void *);
  99
 100 extern void init_IRQ(void);
 101 extern void radix_tree_init(void);
 102
 103 /*
 104  * Debug helper: via this flag we know that we are in 'early bootup code'
 105  * where only the boot processor is running with IRQ disabled.  This means
 106  * two things - IRQ must not be enabled before the flag is cleared and some
 107  * operations which are not allowed with IRQ disabled are allowed while the
 108  * flag is set.
 109  */
 110 bool early_boot_irqs_disabled __read_mostly;
 111
 112 enum system_states system_state __read_mostly;
 113 EXPORT_SYMBOL(system_state);
 114
 115 /*
 116  * Boot command-line arguments
 117  */
 118 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 119 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 120
 121 extern void time_init(void);
 122 /* Default late time init is NULL. archs can override this later. */
 123 void (*__initdata late_time_init)(void);
 124
 125 /* Untouched command line saved by arch-specific code. */
 126 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 127 /* Untouched saved command line (eg. for /proc) */
 128 char *saved_command_line;
 129 /* Command line for parameter parsing */
 130 static char *static_command_line;
 131 /* Command line for per-initcall parameter parsing */
 132 static char *initcall_command_line;
 133
 134 static char *execute_command;
 135 static char *ramdisk_execute_command;
 136
 137 /*
 138  * Used to generate warnings if static_key manipulation functions are used
 139  * before jump_label_init is called.
 140  */
 141 bool static_key_initialized __read_mostly;
 142 EXPORT_SYMBOL_GPL(static_key_initialized);
 143
 144 /*
 145  * If set, this is an indication to the drivers that reset the underlying
 146  * device before going ahead with the initialization otherwise driver might
 147  * rely on the BIOS and skip the reset operation.
 148  *
 149  * This is useful if kernel is booting in an unreliable environment.
 150  * For ex. kdump situation where previous kernel has crashed, BIOS has been
 151  * skipped and devices will be in unknown state.
 152  */
 153 unsigned int reset_devices;
 154 EXPORT_SYMBOL(reset_devices);
 155
 156 static int __init set_reset_devices(char *str)
 157 {
 158         reset_devices = 1;
 159         return 1;
 160 }
 161
 162 __setup("reset_devices", set_reset_devices);
 163
 164 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 165 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 166 static const char *panic_later, *panic_param;
 167
 168 extern const struct obs_kernel_param __setup_start[], __setup_end[];
 169
 170 static bool __init obsolete_checksetup(char *line)
 171 {
 172         const struct obs_kernel_param *p;
 173         bool had_early_param = false;
 174
 175         p = __setup_start;
 176         do {
 177                 int n = strlen(p->str);
 178                 if (parameqn(line, p->str, n)) {
 179                         if (p->early) {
 180                                 /* Already done in parse_early_param?
 181                                  * (Needs exact match on param part).
 182                                  * Keep iterating, as we can have early
 183                                  * params and __setups of same names 8( */
 184                                 if (line[n] == '\0' || line[n] == '=')
 185                                         had_early_param = true;
 186                         } else if (!p->setup_func) {
 187                                 pr_warn("Parameter %s is obsolete, ignored\n",
 188                                         p->str);
 189                                 return true;
 190                         } else if (p->setup_func(line + n))
 191                                 return true;
 192                 }
 193                 p++;
 194         } while (p < __setup_end);
 195
 196         return had_early_param;
 197 }
 198
 199 /*
 200  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 201  * still work even if initially too large, it will just take slightly longer
 202  */
 203 unsigned long loops_per_jiffy = (1<<12);
 204 EXPORT_SYMBOL(loops_per_jiffy);
 205
 206 static int __init debug_kernel(char *str)
 207 {
 208         console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
 209         return 0;
 210 }
 211
 212 static int __init quiet_kernel(char *str)
 213 {
 214         console_loglevel = CONSOLE_LOGLEVEL_QUIET;
 215         return 0;
 216 }
 217
 218 early_param("debug", debug_kernel);
 219 early_param("quiet", quiet_kernel);
 220
 221 static int __init loglevel(char *str)
 222 {
 223         int newlevel;
 224
 225         /*
 226          * Only update loglevel value when a correct setting was passed,
 227          * to prevent blind crashes (when loglevel being set to 0) that
 228          * are quite hard to debug
 229          */
 230         if (get_option(&str, &newlevel)) {
 231                 console_loglevel = newlevel;
 232                 return 0;
 233         }
 234
 235         return -EINVAL;
 236 }
 237
 238 early_param("loglevel", loglevel);
 239
 240 /* Change NUL term back to "=", to make "param" the whole string. */
 241 static int __init repair_env_string(char *param, char *val,
 242                                     const char *unused, void *arg)
 243 {
 244         if (val) {
 245                 /* param=val or param="val"? */
 246                 if (val == param+strlen(param)+1)
 247                         val[-1] = '=';
 248                 else if (val == param+strlen(param)+2) {
 249                         val[-2] = '=';
 250                         memmove(val-1, val, strlen(val)+1);
 251                         val--;
 252                 } else
 253                         BUG();
 254         }
 255         return 0;
 256 }
 257
 258 /* Anything after -- gets handed straight to init. */
 259 static int __init set_init_arg(char *param, char *val,
 260                                const char *unused, void *arg)
 261 {
 262         unsigned int i;
 263
 264         if (panic_later)
 265                 return 0;
 266
 267         repair_env_string(param, val, unused, NULL);
 268
 269         for (i = 0; argv_init[i]; i++) {
 270                 if (i == MAX_INIT_ARGS) {
 271                         panic_later = "init";
 272                         panic_param = param;
 273                         return 0;
 274                 }
 275         }
 276         argv_init[i] = param;
 277         return 0;
 278 }
 279
 280 /*
 281  * Unknown boot options get handed to init, unless they look like
 282  * unused parameters (modprobe will find them in /proc/cmdline).
 283  */
 284 static int __init unknown_bootoption(char *param, char *val,
 285                                      const char *unused, void *arg)
 286 {
 287         repair_env_string(param, val, unused, NULL);
 288
 289         /* Handle obsolete-style parameters */
 290         if (obsolete_checksetup(param))
 291                 return 0;
 292
 293         /* Unused module parameter. */
 294         if (strchr(param, '.') && (!val || strchr(param, '.') < val))
 295                 return 0;
 296
 297         if (panic_later)
 298                 return 0;
 299
 300         if (val) {
 301                 /* Environment option */
 302                 unsigned int i;
 303                 for (i = 0; envp_init[i]; i++) {
 304                         if (i == MAX_INIT_ENVS) {
 305                                 panic_later = "env";
 306                                 panic_param = param;
 307                         }
 308                         if (!strncmp(param, envp_init[i], val - param))
 309                                 break;
 310                 }
 311                 envp_init[i] = param;
 312         } else {
 313                 /* Command line option */
 314                 unsigned int i;
 315                 for (i = 0; argv_init[i]; i++) {
 316                         if (i == MAX_INIT_ARGS) {
 317                                 panic_later = "init";
 318                                 panic_param = param;
 319                         }
 320                 }
 321                 argv_init[i] = param;
 322         }
 323         return 0;
 324 }
 325
 326 static int __init init_setup(char *str)
 327 {
 328         unsigned int i;
 329
 330         execute_command = str;
 331         /*
 332          * In case LILO is going to boot us with default command line,
 333          * it prepends "auto" before the whole cmdline which makes
 334          * the shell think it should execute a script with such name.
 335          * So we ignore all arguments entered _before_ init=... [MJ]
 336          */
 337         for (i = 1; i < MAX_INIT_ARGS; i++)
 338                 argv_init[i] = NULL;
 339         return 1;
 340 }
 341 __setup("init=", init_setup);
 342
 343 static int __init rdinit_setup(char *str)
 344 {
 345         unsigned int i;
 346
 347         ramdisk_execute_command = str;
 348         /* See "auto" comment in init_setup */
 349         for (i = 1; i < MAX_INIT_ARGS; i++)
 350                 argv_init[i] = NULL;
 351         return 1;
 352 }
 353 __setup("rdinit=", rdinit_setup);
 354
 355 #ifndef CONFIG_SMP
 356 static const unsigned int setup_max_cpus = NR_CPUS;
 357 static inline void setup_nr_cpu_ids(void) { }
 358 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 359 #endif
 360
 361 /*
 362  * We need to store the untouched command line for future reference.
 363  * We also need to store the touched command line since the parameter
 364  * parsing is performed in place, and we should allow a component to
 365  * store reference of name/value for future reference.
 366  */
 367 static void __init setup_command_line(char *command_line)
 368 {
 369         saved_command_line =
 370                 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
 371         initcall_command_line =
 372                 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
 373         static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0);
 374         strcpy(saved_command_line, boot_command_line);
 375         strcpy(static_command_line, command_line);
 376 }
 377
 378 /*
 379  * We need to finalize in a non-__init function or else race conditions
 380  * between the root thread and the init thread may cause start_kernel to
 381  * be reaped by free_initmem before the root thread has proceeded to
 382  * cpu_idle.
 383  *
 384  * gcc-3.4 accidentally inlines this function, so use noinline.
 385  */
 386
 387 static __initdata DECLARE_COMPLETION(kthreadd_done);
 388
 389 static noinline void __ref rest_init(void)
 390 {
 391         struct task_struct *tsk;
 392         int pid;
 393
 394         rcu_scheduler_starting();
 395         /*
 396          * We need to spawn init first so that it obtains pid 1, however
 397          * the init task will end up wanting to create kthreads, which, if
 398          * we schedule it before we create kthreadd, will OOPS.
 399          */
 400         pid = kernel_thread(kernel_init, NULL, CLONE_FS);
 401         /*
 402          * Pin init on the boot CPU. Task migration is not properly working
 403          * until sched_init_smp() has been run. It will set the allowed
 404          * CPUs for init to the non isolated CPUs.
 405          */
 406         rcu_read_lock();
 407         tsk = find_task_by_pid_ns(pid, &init_pid_ns);
 408         set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
 409         rcu_read_unlock();
 410
 411         numa_default_policy();
 412         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 413         rcu_read_lock();
 414         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 415         rcu_read_unlock();
 416
 417         /*
 418          * Enable might_sleep() and smp_processor_id() checks.
 419          * They cannot be enabled earlier because with CONFIG_PRREMPT=y
 420          * kernel_thread() would trigger might_sleep() splats. With
 421          * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
 422          * already, but it's stuck on the kthreadd_done completion.
 423          */
 424         system_state = SYSTEM_SCHEDULING;
 425
 426         complete(&kthreadd_done);
 427
 428         /*
 429          * The boot idle thread must execute schedule()
 430          * at least once to get things moving:
 431          */
 432         schedule_preempt_disabled();
 433         /* Call into cpu_idle with preempt disabled */
 434         cpu_startup_entry(CPUHP_ONLINE);
 435 }
 436
 437 /* Check for early params. */
 438 static int __init do_early_param(char *param, char *val,
 439                                  const char *unused, void *arg)
 440 {
 441         const struct obs_kernel_param *p;
 442
 443         for (p = __setup_start; p < __setup_end; p++) {
 444                 if ((p->early && parameq(param, p->str)) ||
 445                     (strcmp(param, "console") == 0 &&
 446                      strcmp(p->str, "earlycon") == 0)
 447                 ) {
 448                         if (p->setup_func(val) != 0)
 449                                 pr_warn("Malformed early option '%s'\n", param);
 450                 }
 451         }
 452         /* We accept everything at this stage. */
 453         return 0;
 454 }
 455
 456 void __init parse_early_options(char *cmdline)
 457 {
 458         parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
 459                    do_early_param);
 460 }
 461
 462 /* Arch code calls this early on, or if not, just before other parsing. */
 463 void __init parse_early_param(void)
 464 {
 465         static int done __initdata;
 466         static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
 467
 468         if (done)
 469                 return;
 470
 471         /* All fall through to do_early_param. */
 472         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 473         parse_early_options(tmp_cmdline);
 474         done = 1;
 475 }
 476
 477 void __init __weak arch_post_acpi_subsys_init(void) { }
 478
 479 void __init __weak smp_setup_processor_id(void)
 480 {
 481 }
 482
 483 # if THREAD_SIZE >= PAGE_SIZE
 484 void __init __weak thread_stack_cache_init(void)
 485 {
 486 }
 487 #endif
 488
 489 void __init __weak mem_encrypt_init(void) { }
 490
 491 /*
 492  * Set up kernel memory allocators
 493  */
 494 static void __init mm_init(void)
 495 {
 496         /*
 497          * page_ext requires contiguous pages,
 498          * bigger than MAX_ORDER unless SPARSEMEM.
 499          */
 500         page_ext_init_flatmem();
 501         mem_init();
 502         kmem_cache_init();
 503         pgtable_init();
 504         vmalloc_init();
 505         ioremap_huge_init();
 506         /* Should be run before the first non-init thread is created */
 507         init_espfix_bsp();
 508         /* Should be run after espfix64 is set up. */
 509         pti_init();
 510 }
 511
 512 asmlinkage __visible void __init start_kernel(void)
 513 {
 514         char *command_line;
 515         char *after_dashes;
 516
 517         set_task_stack_end_magic(&init_task);
 518         smp_setup_processor_id();
 519         debug_objects_early_init();
 520
 521         cgroup_init_early();
 522
 523         local_irq_disable();
 524         early_boot_irqs_disabled = true;
 525
 526         /*
 527          * Interrupts are still disabled. Do necessary setups, then
 528          * enable them.
 529          */
 530         boot_cpu_init();
 531         page_address_init();
 532         pr_notice("%s", linux_banner);
 533         setup_arch(&command_line);
 534         /*
 535          * Set up the the initial canary and entropy after arch
 536          * and after adding latent and command line entropy.
 537          */
 538         add_latent_entropy();
 539         add_device_randomness(command_line, strlen(command_line));
 540         boot_init_stack_canary();
 541         mm_init_cpumask(&init_mm);
 542         setup_command_line(command_line);
 543         setup_nr_cpu_ids();
 544         setup_per_cpu_areas();
 545         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 546         boot_cpu_hotplug_init();
 547
 548         build_all_zonelists(NULL);
 549         page_alloc_init();
 550
 551         pr_notice("Kernel command line: %s\n", boot_command_line);
 552         /* parameters may set static keys */
 553         jump_label_init();
 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         /*
 564          * These use large bootmem allocations and must precede
 565          * kmem_cache_init()
 566          */
 567         setup_log_buf(0);
 568         pidhash_init();
 569         vfs_caches_init_early();
 570         sort_main_extable();
 571         trap_init();
 572         mm_init();
 573
 574         ftrace_init();
 575
 576         /* trace_printk can be enabled here */
 577         early_trace_init();
 578
 579         /*
 580          * Set up the scheduler prior starting any interrupts (such as the
 581          * timer interrupt). Full topology setup happens at smp_init()
 582          * time - but meanwhile we still have a functioning scheduler.
 583          */
 584         sched_init();
 585         /*
 586          * Disable preemption - early bootup scheduling is extremely
 587          * fragile until we cpu_idle() for the first time.
 588          */
 589         preempt_disable();
 590         if (WARN(!irqs_disabled(),
 591                  "Interrupts were enabled *very* early, fixing it\n"))
 592                 local_irq_disable();
 593         radix_tree_init();
 594
 595         /*
 596          * Allow workqueue creation and work item queueing/cancelling
 597          * early.  Work item execution depends on kthreads and starts after
 598          * workqueue_init().
 599          */
 600         workqueue_init_early();
 601
 602         rcu_init();
 603
 604         /* Trace events are available after this */
 605         trace_init();
 606
 607         context_tracking_init();
 608         /* init some links before init_ISA_irqs() */
 609         early_irq_init();
 610         init_IRQ();
 611         tick_init();
 612         rcu_init_nohz();
 613         init_timers();
 614         hrtimers_init();
 615         softirq_init();
 616         timekeeping_init();
 617         time_init();
 618         sched_clock_postinit();
 619         printk_safe_init();
 620         perf_event_init();
 621         profile_init();
 622         call_function_init();
 623         WARN(!irqs_disabled(), "Interrupts were enabled early\n");
 624         early_boot_irqs_disabled = false;
 625         local_irq_enable();
 626
 627         kmem_cache_init_late();
 628
 629         /*
 630          * HACK ALERT! This is early. We're enabling the console before
 631          * we've done PCI setups etc, and console_init() must be aware of
 632          * this. But we do want output early, in case something goes wrong.
 633          */
 634         console_init();
 635         if (panic_later)
 636                 panic("Too many boot %s vars at `%s'", panic_later,
 637                       panic_param);
 638
 639         lockdep_info();
 640
 641         /*
 642          * Need to run this when irqs are enabled, because it wants
 643          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 644          * too:
 645          */
 646         locking_selftest();
 647
 648         /*
 649          * This needs to be called before any devices perform DMA
 650          * operations that might use the SWIOTLB bounce buffers. It will
 651          * mark the bounce buffers as decrypted so that their usage will
 652          * not cause "plain-text" data to be decrypted when accessed.
 653          */
 654         mem_encrypt_init();
 655
 656 #ifdef CONFIG_BLK_DEV_INITRD
 657         if (initrd_start && !initrd_below_start_ok &&
 658             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 659                 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
 660                     page_to_pfn(virt_to_page((void *)initrd_start)),
 661                     min_low_pfn);
 662                 initrd_start = 0;
 663         }
 664 #endif
 665         kmemleak_init();
 666         debug_objects_mem_init();
 667         setup_per_cpu_pageset();
 668         numa_policy_init();
 669         if (late_time_init)
 670                 late_time_init();
 671         calibrate_delay();
 672         pidmap_init();
 673         anon_vma_init();
 674         acpi_early_init();
 675 #ifdef CONFIG_X86
 676         if (efi_enabled(EFI_RUNTIME_SERVICES))
 677                 efi_enter_virtual_mode();
 678 #endif
 679         thread_stack_cache_init();
 680         cred_init();
 681         fork_init();
 682         proc_caches_init();
 683         buffer_init();
 684         key_init();
 685         security_init();
 686         dbg_late_init();
 687         vfs_caches_init();
 688         pagecache_init();
 689         signals_init();
 690         proc_root_init();
 691         nsfs_init();
 692         cpuset_init();
 693         cgroup_init();
 694         taskstats_init_early();
 695         delayacct_init();
 696
 697         check_bugs();
 698
 699         acpi_subsystem_init();
 700         arch_post_acpi_subsys_init();
 701         sfi_init_late();
 702
 703         if (efi_enabled(EFI_RUNTIME_SERVICES)) {
 704                 efi_free_boot_services();
 705         }
 706
 707         /* Do the rest non-__init'ed, we're now alive */
 708         rest_init();
 709
 710         prevent_tail_call_optimization();
 711 }
 712
 713 /* Call all constructor functions linked into the kernel. */
 714 static void __init do_ctors(void)
 715 {
 716 #ifdef CONFIG_CONSTRUCTORS
 717         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
 718
 719         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
 720                 (*fn)();
 721 #endif
 722 }
 723
 724 bool initcall_debug;
 725 core_param(initcall_debug, initcall_debug, bool, 0644);
 726
 727 #ifdef CONFIG_KALLSYMS
 728 struct blacklist_entry {
 729         struct list_head next;
 730         char *buf;
 731 };
 732
 733 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
 734
 735 static int __init initcall_blacklist(char *str)
 736 {
 737         char *str_entry;
 738         struct blacklist_entry *entry;
 739
 740         /* str argument is a comma-separated list of functions */
 741         do {
 742                 str_entry = strsep(&str, ",");
 743                 if (str_entry) {
 744                         pr_debug("blacklisting initcall %s\n", str_entry);
 745                         entry = alloc_bootmem(sizeof(*entry));
 746                         entry->buf = alloc_bootmem(strlen(str_entry) + 1);
 747                         strcpy(entry->buf, str_entry);
 748                         list_add(&entry->next, &blacklisted_initcalls);
 749                 }
 750         } while (str_entry);
 751
 752         return 0;
 753 }
 754
 755 static bool __init_or_module initcall_blacklisted(initcall_t fn)
 756 {
 757         struct blacklist_entry *entry;
 758         char fn_name[KSYM_SYMBOL_LEN];
 759         unsigned long addr;
 760
 761         if (list_empty(&blacklisted_initcalls))
 762                 return false;
 763
 764         addr = (unsigned long) dereference_function_descriptor(fn);
 765         sprint_symbol_no_offset(fn_name, addr);
 766
 767         /*
 768          * fn will be "function_name [module_name]" where [module_name] is not
 769          * displayed for built-in init functions.  Strip off the [module_name].
 770          */
 771         strreplace(fn_name, ' ', '\0');
 772
 773         list_for_each_entry(entry, &blacklisted_initcalls, next) {
 774                 if (!strcmp(fn_name, entry->buf)) {
 775                         pr_debug("initcall %s blacklisted\n", fn_name);
 776                         return true;
 777                 }
 778         }
 779
 780         return false;
 781 }
 782 #else
 783 static int __init initcall_blacklist(char *str)
 784 {
 785         pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
 786         return 0;
 787 }
 788
 789 static bool __init_or_module initcall_blacklisted(initcall_t fn)
 790 {
 791         return false;
 792 }
 793 #endif
 794 __setup("initcall_blacklist=", initcall_blacklist);
 795
 796 static int __init_or_module do_one_initcall_debug(initcall_t fn)
 797 {
 798         ktime_t calltime, delta, rettime;
 799         unsigned long long duration;
 800         int ret;
 801
 802         printk(KERN_DEBUG "calling  %pF @ %i\n", fn, task_pid_nr(current));
 803         calltime = ktime_get();
 804         ret = fn();
 805         rettime = ktime_get();
 806         delta = ktime_sub(rettime, calltime);
 807         duration = (unsigned long long) ktime_to_ns(delta) >> 10;
 808         printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n",
 809                  fn, ret, duration);
 810
 811         return ret;
 812 }
 813
 814 int __init_or_module do_one_initcall(initcall_t fn)
 815 {
 816         int count = preempt_count();
 817         int ret;
 818         char msgbuf[64];
 819
 820         if (initcall_blacklisted(fn))
 821                 return -EPERM;
 822
 823         if (initcall_debug)
 824                 ret = do_one_initcall_debug(fn);
 825         else
 826                 ret = fn();
 827
 828         msgbuf[0] = 0;
 829
 830         if (preempt_count() != count) {
 831                 sprintf(msgbuf, "preemption imbalance ");
 832                 preempt_count_set(count);
 833         }
 834         if (irqs_disabled()) {
 835                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 836                 local_irq_enable();
 837         }
 838         WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
 839
 840         add_latent_entropy();
 841         return ret;
 842 }
 843
 844
 845 extern initcall_t __initcall_start[];
 846 extern initcall_t __initcall0_start[];
 847 extern initcall_t __initcall1_start[];
 848 extern initcall_t __initcall2_start[];
 849 extern initcall_t __initcall3_start[];
 850 extern initcall_t __initcall4_start[];
 851 extern initcall_t __initcall5_start[];
 852 extern initcall_t __initcall6_start[];
 853 extern initcall_t __initcall7_start[];
 854 extern initcall_t __initcall_end[];
 855
 856 static initcall_t *initcall_levels[] __initdata = {
 857         __initcall0_start,
 858         __initcall1_start,
 859         __initcall2_start,
 860         __initcall3_start,
 861         __initcall4_start,
 862         __initcall5_start,
 863         __initcall6_start,
 864         __initcall7_start,
 865         __initcall_end,
 866 };
 867
 868 /* Keep these in sync with initcalls in include/linux/init.h */
 869 static char *initcall_level_names[] __initdata = {
 870         "early",
 871         "core",
 872         "postcore",
 873         "arch",
 874         "subsys",
 875         "fs",
 876         "device",
 877         "late",
 878 };
 879
 880 static void __init do_initcall_level(int level)
 881 {
 882         initcall_t *fn;
 883
 884         strcpy(initcall_command_line, saved_command_line);
 885         parse_args(initcall_level_names[level],
 886                    initcall_command_line, __start___param,
 887                    __stop___param - __start___param,
 888                    level, level,
 889                    NULL, &repair_env_string);
 890
 891         for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
 892                 do_one_initcall(*fn);
 893 }
 894
 895 static void __init do_initcalls(void)
 896 {
 897         int level;
 898
 899         for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
 900                 do_initcall_level(level);
 901 }
 902
 903 /*
 904  * Ok, the machine is now initialized. None of the devices
 905  * have been touched yet, but the CPU subsystem is up and
 906  * running, and memory and process management works.
 907  *
 908  * Now we can finally start doing some real work..
 909  */
 910 static void __init do_basic_setup(void)
 911 {
 912         cpuset_init_smp();
 913         shmem_init();
 914         driver_init();
 915         init_irq_proc();
 916         do_ctors();
 917         usermodehelper_enable();
 918         do_initcalls();
 919 }
 920
 921 static void __init do_pre_smp_initcalls(void)
 922 {
 923         initcall_t *fn;
 924
 925         for (fn = __initcall_start; fn < __initcall0_start; fn++)
 926                 do_one_initcall(*fn);
 927 }
 928
 929 /*
 930  * This function requests modules which should be loaded by default and is
 931  * called twice right after initrd is mounted and right before init is
 932  * exec'd.  If such modules are on either initrd or rootfs, they will be
 933  * loaded before control is passed to userland.
 934  */
 935 void __init load_default_modules(void)
 936 {
 937         load_default_elevator_module();
 938 }
 939
 940 static int run_init_process(const char *init_filename)
 941 {
 942         argv_init[0] = init_filename;
 943         return do_execve(getname_kernel(init_filename),
 944                 (const char __user *const __user *)argv_init,
 945                 (const char __user *const __user *)envp_init);
 946 }
 947
 948 static int try_to_run_init_process(const char *init_filename)
 949 {
 950         int ret;
 951
 952         ret = run_init_process(init_filename);
 953
 954         if (ret && ret != -ENOENT) {
 955                 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
 956                        init_filename, ret);
 957         }
 958
 959         return ret;
 960 }
 961
 962 static noinline void __init kernel_init_freeable(void);
 963
 964 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
 965 bool rodata_enabled __ro_after_init = true;
 966 static int __init set_debug_rodata(char *str)
 967 {
 968         return strtobool(str, &rodata_enabled);
 969 }
 970 __setup("rodata=", set_debug_rodata);
 971 #endif
 972
 973 #ifdef CONFIG_STRICT_KERNEL_RWX
 974 static void mark_readonly(void)
 975 {
 976         if (rodata_enabled) {
 977                 /*
 978                  * load_module() results in W+X mappings, which are cleaned up
 979                  * with call_rcu_sched().  Let's make sure that queued work is
 980                  * flushed so that we don't hit false positives looking for
 981                  * insecure pages which are W+X.
 982                  */
 983                 rcu_barrier_sched();
 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         free_initmem();
1005         mark_readonly();
1006         system_state = SYSTEM_RUNNING;
1007         numa_default_policy();
1008
1009         rcu_end_inkernel_boot();
1010
1011         if (ramdisk_execute_command) {
1012                 ret = run_init_process(ramdisk_execute_command);
1013                 if (!ret)
1014                         return 0;
1015                 pr_err("Failed to execute %s (error %d)\n",
1016                        ramdisk_execute_command, ret);
1017         }
1018
1019         /*
1020          * We try each of these until one succeeds.
1021          *
1022          * The Bourne shell can be used instead of init if we are
1023          * trying to recover a really broken machine.
1024          */
1025         if (execute_command) {
1026                 ret = run_init_process(execute_command);
1027                 if (!ret)
1028                         return 0;
1029                 panic("Requested init %s failed (error %d).",
1030                       execute_command, ret);
1031         }
1032         if (!try_to_run_init_process("/sbin/init") ||
1033             !try_to_run_init_process("/etc/init") ||
1034             !try_to_run_init_process("/bin/init") ||
1035             !try_to_run_init_process("/bin/sh"))
1036                 return 0;
1037
1038         panic("No working init found.  Try passing init= option to kernel. "
1039               "See Linux Documentation/admin-guide/init.rst for guidance.");
1040 }
1041
1042 static noinline void __init kernel_init_freeable(void)
1043 {
1044         /*
1045          * Wait until kthreadd is all set-up.
1046          */
1047         wait_for_completion(&kthreadd_done);
1048
1049         /* Now the scheduler is fully set up and can do blocking allocations */
1050         gfp_allowed_mask = __GFP_BITS_MASK;
1051
1052         /*
1053          * init can allocate pages on any node
1054          */
1055         set_mems_allowed(node_states[N_MEMORY]);
1056
1057         cad_pid = task_pid(current);
1058
1059         smp_prepare_cpus(setup_max_cpus);
1060
1061         workqueue_init();
1062
1063         init_mm_internals();
1064
1065         do_pre_smp_initcalls();
1066         lockup_detector_init();
1067
1068         smp_init();
1069         sched_init_smp();
1070
1071         page_alloc_init_late();
1072         /* Initialize page ext after all struct pages are initialized. */
1073         page_ext_init();
1074
1075         do_basic_setup();
1076
1077         /* Open the /dev/console on the rootfs, this should never fail */
1078         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1079                 pr_err("Warning: unable to open an initial console.\n");
1080
1081         (void) sys_dup(0);
1082         (void) sys_dup(0);
1083         /*
1084          * check if there is an early userspace init.  If yes, let it do all
1085          * the work
1086          */
1087
1088         if (!ramdisk_execute_command)
1089                 ramdisk_execute_command = "/init";
1090
1091         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
1092                 ramdisk_execute_command = NULL;
1093                 prepare_namespace();
1094         }
1095
1096         /*
1097          * Ok, we have completed the initial bootup, and
1098          * we're essentially up and running. Get rid of the
1099          * initmem segments and start the user-mode stuff..
1100          *
1101          * rootfs is available now, try loading the public keys
1102          * and default modules
1103          */
1104
1105         integrity_load_keys();
1106         load_default_modules();
1107 }