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/utsname.h>
  55 #include <linux/rmap.h>
  56 #include <linux/mempolicy.h>
  57 #include <linux/key.h>
  58 #include <linux/buffer_head.h>
  59 #include <linux/page_ext.h>
  60 #include <linux/debug_locks.h>
  61 #include <linux/debugobjects.h>
  62 #include <linux/lockdep.h>
  63 #include <linux/kmemleak.h>
  64 #include <linux/pid_namespace.h>
  65 #include <linux/device.h>
  66 #include <linux/kthread.h>
  67 #include <linux/sched.h>
  68 #include <linux/sched/init.h>
  69 #include <linux/signal.h>
  70 #include <linux/idr.h>
  71 #include <linux/kgdb.h>
  72 #include <linux/ftrace.h>
  73 #include <linux/async.h>
  74 #include <linux/sfi.h>
  75 #include <linux/shmem_fs.h>
  76 #include <linux/slab.h>
  77 #include <linux/perf_event.h>
  78 #include <linux/ptrace.h>
  79 #include <linux/pti.h>
  80 #include <linux/blkdev.h>
  81 #include <linux/elevator.h>
  82 #include <linux/sched/clock.h>
  83 #include <linux/sched/task.h>
  84 #include <linux/sched/task_stack.h>
  85 #include <linux/context_tracking.h>
  86 #include <linux/random.h>
  87 #include <linux/list.h>
  88 #include <linux/integrity.h>
  89 #include <linux/proc_ns.h>
  90 #include <linux/io.h>
  91 #include <linux/cache.h>
  92 #include <linux/rodata_test.h>
  93 #include <linux/jump_label.h>
  94 #include <linux/mem_encrypt.h>
  95
  96 #include <asm/io.h>
  97 #include <asm/bugs.h>
  98 #include <asm/setup.h>
  99 #include <asm/sections.h>
 100 #include <asm/cacheflush.h>
 101
 102 #define CREATE_TRACE_POINTS
 103 #include <trace/events/initcall.h>
 104
 105 static int kernel_init(void *);
 106
 107 extern void init_IRQ(void);
 108 extern void radix_tree_init(void);
 109
 110 /*
 111  * Debug helper: via this flag we know that we are in 'early bootup code'
 112  * where only the boot processor is running with IRQ disabled.  This means
 113  * two things - IRQ must not be enabled before the flag is cleared and some
 114  * operations which are not allowed with IRQ disabled are allowed while the
 115  * flag is set.
 116  */
 117 bool early_boot_irqs_disabled __read_mostly;
 118
 119 enum system_states system_state __read_mostly;
 120 EXPORT_SYMBOL(system_state);
 121
 122 /*
 123  * Boot command-line arguments
 124  */
 125 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 126 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 127
 128 extern void time_init(void);
 129 /* Default late time init is NULL. archs can override this later. */
 130 void (*__initdata late_time_init)(void);
 131
 132 /* Untouched command line saved by arch-specific code. */
 133 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 134 /* Untouched saved command line (eg. for /proc) */
 135 char *saved_command_line;
 136 /* Command line for parameter parsing */
 137 static char *static_command_line;
 138 /* Command line for per-initcall parameter parsing */
 139 static char *initcall_command_line;
 140
 141 static char *execute_command;
 142 static char *ramdisk_execute_command;
 143
 144 /*
 145  * Used to generate warnings if static_key manipulation functions are used
 146  * before jump_label_init is called.
 147  */
 148 bool static_key_initialized __read_mostly;
 149 EXPORT_SYMBOL_GPL(static_key_initialized);
 150
 151 /*
 152  * If set, this is an indication to the drivers that reset the underlying
 153  * device before going ahead with the initialization otherwise driver might
 154  * rely on the BIOS and skip the reset operation.
 155  *
 156  * This is useful if kernel is booting in an unreliable environment.
 157  * For ex. kdump situation where previous kernel has crashed, BIOS has been
 158  * skipped and devices will be in unknown state.
 159  */
 160 unsigned int reset_devices;
 161 EXPORT_SYMBOL(reset_devices);
 162
 163 static int __init set_reset_devices(char *str)
 164 {
 165         reset_devices = 1;
 166         return 1;
 167 }
 168
 169 __setup("reset_devices", set_reset_devices);
 170
 171 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 172 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 173 static const char *panic_later, *panic_param;
 174
 175 extern const struct obs_kernel_param __setup_start[], __setup_end[];
 176
 177 static bool __init obsolete_checksetup(char *line)
 178 {
 179         const struct obs_kernel_param *p;
 180         bool had_early_param = false;
 181
 182         p = __setup_start;
 183         do {
 184                 int n = strlen(p->str);
 185                 if (parameqn(line, p->str, n)) {
 186                         if (p->early) {
 187                                 /* Already done in parse_early_param?
 188                                  * (Needs exact match on param part).
 189                                  * Keep iterating, as we can have early
 190                                  * params and __setups of same names 8( */
 191                                 if (line[n] == '\0' || line[n] == '=')
 192                                         had_early_param = true;
 193                         } else if (!p->setup_func) {
 194                                 pr_warn("Parameter %s is obsolete, ignored\n",
 195                                         p->str);
 196                                 return true;
 197                         } else if (p->setup_func(line + n))
 198                                 return true;
 199                 }
 200                 p++;
 201         } while (p < __setup_end);
 202
 203         return had_early_param;
 204 }
 205
 206 /*
 207  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 208  * still work even if initially too large, it will just take slightly longer
 209  */
 210 unsigned long loops_per_jiffy = (1<<12);
 211 EXPORT_SYMBOL(loops_per_jiffy);
 212
 213 static int __init debug_kernel(char *str)
 214 {
 215         console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
 216         return 0;
 217 }
 218
 219 static int __init quiet_kernel(char *str)
 220 {
 221         console_loglevel = CONSOLE_LOGLEVEL_QUIET;
 222         return 0;
 223 }
 224
 225 early_param("debug", debug_kernel);
 226 early_param("quiet", quiet_kernel);
 227
 228 static int __init loglevel(char *str)
 229 {
 230         int newlevel;
 231
 232         /*
 233          * Only update loglevel value when a correct setting was passed,
 234          * to prevent blind crashes (when loglevel being set to 0) that
 235          * are quite hard to debug
 236          */
 237         if (get_option(&str, &newlevel)) {
 238                 console_loglevel = newlevel;
 239                 return 0;
 240         }
 241
 242         return -EINVAL;
 243 }
 244
 245 early_param("loglevel", loglevel);
 246
 247 /* Change NUL term back to "=", to make "param" the whole string. */
 248 static int __init repair_env_string(char *param, char *val,
 249                                     const char *unused, void *arg)
 250 {
 251         if (val) {
 252                 /* param=val or param="val"? */
 253                 if (val == param+strlen(param)+1)
 254                         val[-1] = '=';
 255                 else if (val == param+strlen(param)+2) {
 256                         val[-2] = '=';
 257                         memmove(val-1, val, strlen(val)+1);
 258                         val--;
 259                 } else
 260                         BUG();
 261         }
 262         return 0;
 263 }
 264
 265 /* Anything after -- gets handed straight to init. */
 266 static int __init set_init_arg(char *param, char *val,
 267                                const char *unused, void *arg)
 268 {
 269         unsigned int i;
 270
 271         if (panic_later)
 272                 return 0;
 273
 274         repair_env_string(param, val, unused, NULL);
 275
 276         for (i = 0; argv_init[i]; i++) {
 277                 if (i == MAX_INIT_ARGS) {
 278                         panic_later = "init";
 279                         panic_param = param;
 280                         return 0;
 281                 }
 282         }
 283         argv_init[i] = param;
 284         return 0;
 285 }
 286
 287 /*
 288  * Unknown boot options get handed to init, unless they look like
 289  * unused parameters (modprobe will find them in /proc/cmdline).
 290  */
 291 static int __init unknown_bootoption(char *param, char *val,
 292                                      const char *unused, void *arg)
 293 {
 294         repair_env_string(param, val, unused, NULL);
 295
 296         /* Handle obsolete-style parameters */
 297         if (obsolete_checksetup(param))
 298                 return 0;
 299
 300         /* Unused module parameter. */
 301         if (strchr(param, '.') && (!val || strchr(param, '.') < val))
 302                 return 0;
 303
 304         if (panic_later)
 305                 return 0;
 306
 307         if (val) {
 308                 /* Environment option */
 309                 unsigned int i;
 310                 for (i = 0; envp_init[i]; i++) {
 311                         if (i == MAX_INIT_ENVS) {
 312                                 panic_later = "env";
 313                                 panic_param = param;
 314                         }
 315                         if (!strncmp(param, envp_init[i], val - param))
 316                                 break;
 317                 }
 318                 envp_init[i] = param;
 319         } else {
 320                 /* Command line option */
 321                 unsigned int i;
 322                 for (i = 0; argv_init[i]; i++) {
 323                         if (i == MAX_INIT_ARGS) {
 324                                 panic_later = "init";
 325                                 panic_param = param;
 326                         }
 327                 }
 328                 argv_init[i] = param;
 329         }
 330         return 0;
 331 }
 332
 333 static int __init init_setup(char *str)
 334 {
 335         unsigned int i;
 336
 337         execute_command = str;
 338         /*
 339          * In case LILO is going to boot us with default command line,
 340          * it prepends "auto" before the whole cmdline which makes
 341          * the shell think it should execute a script with such name.
 342          * So we ignore all arguments entered _before_ init=... [MJ]
 343          */
 344         for (i = 1; i < MAX_INIT_ARGS; i++)
 345                 argv_init[i] = NULL;
 346         return 1;
 347 }
 348 __setup("init=", init_setup);
 349
 350 static int __init rdinit_setup(char *str)
 351 {
 352         unsigned int i;
 353
 354         ramdisk_execute_command = str;
 355         /* See "auto" comment in init_setup */
 356         for (i = 1; i < MAX_INIT_ARGS; i++)
 357                 argv_init[i] = NULL;
 358         return 1;
 359 }
 360 __setup("rdinit=", rdinit_setup);
 361
 362 #ifndef CONFIG_SMP
 363 static const unsigned int setup_max_cpus = NR_CPUS;
 364 static inline void setup_nr_cpu_ids(void) { }
 365 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 366 #endif
 367
 368 /*
 369  * We need to store the untouched command line for future reference.
 370  * We also need to store the touched command line since the parameter
 371  * parsing is performed in place, and we should allow a component to
 372  * store reference of name/value for future reference.
 373  */
 374 static void __init setup_command_line(char *command_line)
 375 {
 376         saved_command_line =
 377                 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
 378         initcall_command_line =
 379                 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
 380         static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0);
 381         strcpy(saved_command_line, boot_command_line);
 382         strcpy(static_command_line, command_line);
 383 }
 384
 385 /*
 386  * We need to finalize in a non-__init function or else race conditions
 387  * between the root thread and the init thread may cause start_kernel to
 388  * be reaped by free_initmem before the root thread has proceeded to
 389  * cpu_idle.
 390  *
 391  * gcc-3.4 accidentally inlines this function, so use noinline.
 392  */
 393
 394 static __initdata DECLARE_COMPLETION(kthreadd_done);
 395
 396 static noinline void __ref rest_init(void)
 397 {
 398         struct task_struct *tsk;
 399         int pid;
 400
 401         rcu_scheduler_starting();
 402         /*
 403          * We need to spawn init first so that it obtains pid 1, however
 404          * the init task will end up wanting to create kthreads, which, if
 405          * we schedule it before we create kthreadd, will OOPS.
 406          */
 407         pid = kernel_thread(kernel_init, NULL, CLONE_FS);
 408         /*
 409          * Pin init on the boot CPU. Task migration is not properly working
 410          * until sched_init_smp() has been run. It will set the allowed
 411          * CPUs for init to the non isolated CPUs.
 412          */
 413         rcu_read_lock();
 414         tsk = find_task_by_pid_ns(pid, &init_pid_ns);
 415         set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
 416         rcu_read_unlock();
 417
 418         numa_default_policy();
 419         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 420         rcu_read_lock();
 421         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 422         rcu_read_unlock();
 423
 424         /*
 425          * Enable might_sleep() and smp_processor_id() checks.
 426          * They cannot be enabled earlier because with CONFIG_PREEMPT=y
 427          * kernel_thread() would trigger might_sleep() splats. With
 428          * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
 429          * already, but it's stuck on the kthreadd_done completion.
 430          */
 431         system_state = SYSTEM_SCHEDULING;
 432
 433         complete(&kthreadd_done);
 434
 435         /*
 436          * The boot idle thread must execute schedule()
 437          * at least once to get things moving:
 438          */
 439         schedule_preempt_disabled();
 440         /* Call into cpu_idle with preempt disabled */
 441         cpu_startup_entry(CPUHP_ONLINE);
 442 }
 443
 444 /* Check for early params. */
 445 static int __init do_early_param(char *param, char *val,
 446                                  const char *unused, void *arg)
 447 {
 448         const struct obs_kernel_param *p;
 449
 450         for (p = __setup_start; p < __setup_end; p++) {
 451                 if ((p->early && parameq(param, p->str)) ||
 452                     (strcmp(param, "console") == 0 &&
 453                      strcmp(p->str, "earlycon") == 0)
 454                 ) {
 455                         if (p->setup_func(val) != 0)
 456                                 pr_warn("Malformed early option '%s'\n", param);
 457                 }
 458         }
 459         /* We accept everything at this stage. */
 460         return 0;
 461 }
 462
 463 void __init parse_early_options(char *cmdline)
 464 {
 465         parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
 466                    do_early_param);
 467 }
 468
 469 /* Arch code calls this early on, or if not, just before other parsing. */
 470 void __init parse_early_param(void)
 471 {
 472         static int done __initdata;
 473         static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
 474
 475         if (done)
 476                 return;
 477
 478         /* All fall through to do_early_param. */
 479         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 480         parse_early_options(tmp_cmdline);
 481         done = 1;
 482 }
 483
 484 void __init __weak arch_post_acpi_subsys_init(void) { }
 485
 486 void __init __weak smp_setup_processor_id(void)
 487 {
 488 }
 489
 490 # if THREAD_SIZE >= PAGE_SIZE
 491 void __init __weak thread_stack_cache_init(void)
 492 {
 493 }
 494 #endif
 495
 496 void __init __weak mem_encrypt_init(void) { }
 497
 498 bool initcall_debug;
 499 core_param(initcall_debug, initcall_debug, bool, 0644);
 500
 501 #ifdef TRACEPOINTS_ENABLED
 502 static void __init initcall_debug_enable(void);
 503 #else
 504 static inline void initcall_debug_enable(void)
 505 {
 506 }
 507 #endif
 508
 509 /*
 510  * Set up kernel memory allocators
 511  */
 512 static void __init mm_init(void)
 513 {
 514         /*
 515          * page_ext requires contiguous pages,
 516          * bigger than MAX_ORDER unless SPARSEMEM.
 517          */
 518         page_ext_init_flatmem();
 519         mem_init();
 520         kmem_cache_init();
 521         pgtable_init();
 522         vmalloc_init();
 523         ioremap_huge_init();
 524         /* Should be run before the first non-init thread is created */
 525         init_espfix_bsp();
 526         /* Should be run after espfix64 is set up. */
 527         pti_init();
 528 }
 529
 530 asmlinkage __visible void __init start_kernel(void)
 531 {
 532         char *command_line;
 533         char *after_dashes;
 534
 535         set_task_stack_end_magic(&init_task);
 536         smp_setup_processor_id();
 537         debug_objects_early_init();
 538
 539         cgroup_init_early();
 540
 541         local_irq_disable();
 542         early_boot_irqs_disabled = true;
 543
 544         /*
 545          * Interrupts are still disabled. Do necessary setups, then
 546          * enable them.
 547          */
 548         boot_cpu_init();
 549         page_address_init();
 550         pr_notice("%s", linux_banner);
 551         setup_arch(&command_line);
 552         /*
 553          * Set up the the initial canary and entropy after arch
 554          * and after adding latent and command line entropy.
 555          */
 556         add_latent_entropy();
 557         add_device_randomness(command_line, strlen(command_line));
 558         boot_init_stack_canary();
 559         mm_init_cpumask(&init_mm);
 560         setup_command_line(command_line);
 561         setup_nr_cpu_ids();
 562         setup_per_cpu_areas();
 563         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 564         boot_cpu_hotplug_init();
 565
 566         build_all_zonelists(NULL);
 567         page_alloc_init();
 568
 569         pr_notice("Kernel command line: %s\n", boot_command_line);
 570         /* parameters may set static keys */
 571         jump_label_init();
 572         parse_early_param();
 573         after_dashes = parse_args("Booting kernel",
 574                                   static_command_line, __start___param,
 575                                   __stop___param - __start___param,
 576                                   -1, -1, NULL, &unknown_bootoption);
 577         if (!IS_ERR_OR_NULL(after_dashes))
 578                 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
 579                            NULL, set_init_arg);
 580
 581         /*
 582          * These use large bootmem allocations and must precede
 583          * kmem_cache_init()
 584          */
 585         setup_log_buf(0);
 586         vfs_caches_init_early();
 587         sort_main_extable();
 588         trap_init();
 589         mm_init();
 590
 591         ftrace_init();
 592
 593         /* trace_printk can be enabled here */
 594         early_trace_init();
 595
 596         /*
 597          * Set up the scheduler prior starting any interrupts (such as the
 598          * timer interrupt). Full topology setup happens at smp_init()
 599          * time - but meanwhile we still have a functioning scheduler.
 600          */
 601         sched_init();
 602         /*
 603          * Disable preemption - early bootup scheduling is extremely
 604          * fragile until we cpu_idle() for the first time.
 605          */
 606         preempt_disable();
 607         if (WARN(!irqs_disabled(),
 608                  "Interrupts were enabled *very* early, fixing it\n"))
 609                 local_irq_disable();
 610         radix_tree_init();
 611
 612         /*
 613          * Set up housekeeping before setting up workqueues to allow the unbound
 614          * workqueue to take non-housekeeping into account.
 615          */
 616         housekeeping_init();
 617
 618         /*
 619          * Allow workqueue creation and work item queueing/cancelling
 620          * early.  Work item execution depends on kthreads and starts after
 621          * workqueue_init().
 622          */
 623         workqueue_init_early();
 624
 625         rcu_init();
 626
 627         /* Trace events are available after this */
 628         trace_init();
 629
 630         if (initcall_debug)
 631                 initcall_debug_enable();
 632
 633         context_tracking_init();
 634         /* init some links before init_ISA_irqs() */
 635         early_irq_init();
 636         init_IRQ();
 637         tick_init();
 638         rcu_init_nohz();
 639         init_timers();
 640         hrtimers_init();
 641         softirq_init();
 642         timekeeping_init();
 643         time_init();
 644         printk_safe_init();
 645         perf_event_init();
 646         profile_init();
 647         call_function_init();
 648         WARN(!irqs_disabled(), "Interrupts were enabled early\n");
 649
 650         early_boot_irqs_disabled = false;
 651         local_irq_enable();
 652
 653         kmem_cache_init_late();
 654
 655         /*
 656          * HACK ALERT! This is early. We're enabling the console before
 657          * we've done PCI setups etc, and console_init() must be aware of
 658          * this. But we do want output early, in case something goes wrong.
 659          */
 660         console_init();
 661         if (panic_later)
 662                 panic("Too many boot %s vars at `%s'", panic_later,
 663                       panic_param);
 664
 665         lockdep_init();
 666
 667         /*
 668          * Need to run this when irqs are enabled, because it wants
 669          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 670          * too:
 671          */
 672         locking_selftest();
 673
 674         /*
 675          * This needs to be called before any devices perform DMA
 676          * operations that might use the SWIOTLB bounce buffers. It will
 677          * mark the bounce buffers as decrypted so that their usage will
 678          * not cause "plain-text" data to be decrypted when accessed.
 679          */
 680         mem_encrypt_init();
 681
 682 #ifdef CONFIG_BLK_DEV_INITRD
 683         if (initrd_start && !initrd_below_start_ok &&
 684             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 685                 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
 686                     page_to_pfn(virt_to_page((void *)initrd_start)),
 687                     min_low_pfn);
 688                 initrd_start = 0;
 689         }
 690 #endif
 691         kmemleak_init();
 692         debug_objects_mem_init();
 693         setup_per_cpu_pageset();
 694         numa_policy_init();
 695         acpi_early_init();
 696         if (late_time_init)
 697                 late_time_init();
 698         sched_clock_init();
 699         calibrate_delay();
 700         pid_idr_init();
 701         anon_vma_init();
 702 #ifdef CONFIG_X86
 703         if (efi_enabled(EFI_RUNTIME_SERVICES))
 704                 efi_enter_virtual_mode();
 705 #endif
 706         thread_stack_cache_init();
 707         cred_init();
 708         fork_init();
 709         proc_caches_init();
 710         uts_ns_init();
 711         buffer_init();
 712         key_init();
 713         security_init();
 714         dbg_late_init();
 715         vfs_caches_init();
 716         pagecache_init();
 717         signals_init();
 718         seq_file_init();
 719         proc_root_init();
 720         nsfs_init();
 721         cpuset_init();
 722         cgroup_init();
 723         taskstats_init_early();
 724         delayacct_init();
 725
 726         check_bugs();
 727
 728         acpi_subsystem_init();
 729         arch_post_acpi_subsys_init();
 730         sfi_init_late();
 731
 732         if (efi_enabled(EFI_RUNTIME_SERVICES)) {
 733                 efi_free_boot_services();
 734         }
 735
 736         /* Do the rest non-__init'ed, we're now alive */
 737         rest_init();
 738 }
 739
 740 /* Call all constructor functions linked into the kernel. */
 741 static void __init do_ctors(void)
 742 {
 743 #ifdef CONFIG_CONSTRUCTORS
 744         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
 745
 746         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
 747                 (*fn)();
 748 #endif
 749 }
 750
 751 #ifdef CONFIG_KALLSYMS
 752 struct blacklist_entry {
 753         struct list_head next;
 754         char *buf;
 755 };
 756
 757 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
 758
 759 static int __init initcall_blacklist(char *str)
 760 {
 761         char *str_entry;
 762         struct blacklist_entry *entry;
 763
 764         /* str argument is a comma-separated list of functions */
 765         do {
 766                 str_entry = strsep(&str, ",");
 767                 if (str_entry) {
 768                         pr_debug("blacklisting initcall %s\n", str_entry);
 769                         entry = alloc_bootmem(sizeof(*entry));
 770                         entry->buf = alloc_bootmem(strlen(str_entry) + 1);
 771                         strcpy(entry->buf, str_entry);
 772                         list_add(&entry->next, &blacklisted_initcalls);
 773                 }
 774         } while (str_entry);
 775
 776         return 0;
 777 }
 778
 779 static bool __init_or_module initcall_blacklisted(initcall_t fn)
 780 {
 781         struct blacklist_entry *entry;
 782         char fn_name[KSYM_SYMBOL_LEN];
 783         unsigned long addr;
 784
 785         if (list_empty(&blacklisted_initcalls))
 786                 return false;
 787
 788         addr = (unsigned long) dereference_function_descriptor(fn);
 789         sprint_symbol_no_offset(fn_name, addr);
 790
 791         /*
 792          * fn will be "function_name [module_name]" where [module_name] is not
 793          * displayed for built-in init functions.  Strip off the [module_name].
 794          */
 795         strreplace(fn_name, ' ', '\0');
 796
 797         list_for_each_entry(entry, &blacklisted_initcalls, next) {
 798                 if (!strcmp(fn_name, entry->buf)) {
 799                         pr_debug("initcall %s blacklisted\n", fn_name);
 800                         return true;
 801                 }
 802         }
 803
 804         return false;
 805 }
 806 #else
 807 static int __init initcall_blacklist(char *str)
 808 {
 809         pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
 810         return 0;
 811 }
 812
 813 static bool __init_or_module initcall_blacklisted(initcall_t fn)
 814 {
 815         return false;
 816 }
 817 #endif
 818 __setup("initcall_blacklist=", initcall_blacklist);
 819
 820 static __init_or_module void
 821 trace_initcall_start_cb(void *data, initcall_t fn)
 822 {
 823         ktime_t *calltime = (ktime_t *)data;
 824
 825         printk(KERN_DEBUG "calling  %pF @ %i\n", fn, task_pid_nr(current));
 826         *calltime = ktime_get();
 827 }
 828
 829 static __init_or_module void
 830 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
 831 {
 832         ktime_t *calltime = (ktime_t *)data;
 833         ktime_t delta, rettime;
 834         unsigned long long duration;
 835
 836         rettime = ktime_get();
 837         delta = ktime_sub(rettime, *calltime);
 838         duration = (unsigned long long) ktime_to_ns(delta) >> 10;
 839         printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n",
 840                  fn, ret, duration);
 841 }
 842
 843 static ktime_t initcall_calltime;
 844
 845 #ifdef TRACEPOINTS_ENABLED
 846 static void __init initcall_debug_enable(void)
 847 {
 848         int ret;
 849
 850         ret = register_trace_initcall_start(trace_initcall_start_cb,
 851                                             &initcall_calltime);
 852         ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
 853                                               &initcall_calltime);
 854         WARN(ret, "Failed to register initcall tracepoints\n");
 855 }
 856 # define do_trace_initcall_start        trace_initcall_start
 857 # define do_trace_initcall_finish       trace_initcall_finish
 858 #else
 859 static inline void do_trace_initcall_start(initcall_t fn)
 860 {
 861         if (!initcall_debug)
 862                 return;
 863         trace_initcall_start_cb(&initcall_calltime, fn);
 864 }
 865 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
 866 {
 867         if (!initcall_debug)
 868                 return;
 869         trace_initcall_finish_cb(&initcall_calltime, fn, ret);
 870 }
 871 #endif /* !TRACEPOINTS_ENABLED */
 872
 873 int __init_or_module do_one_initcall(initcall_t fn)
 874 {
 875         int count = preempt_count();
 876         char msgbuf[64];
 877         int ret;
 878
 879         if (initcall_blacklisted(fn))
 880                 return -EPERM;
 881
 882         do_trace_initcall_start(fn);
 883         ret = fn();
 884         do_trace_initcall_finish(fn, ret);
 885
 886         msgbuf[0] = 0;
 887
 888         if (preempt_count() != count) {
 889                 sprintf(msgbuf, "preemption imbalance ");
 890                 preempt_count_set(count);
 891         }
 892         if (irqs_disabled()) {
 893                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 894                 local_irq_enable();
 895         }
 896         WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
 897
 898         add_latent_entropy();
 899         return ret;
 900 }
 901
 902
 903 extern initcall_entry_t __initcall_start[];
 904 extern initcall_entry_t __initcall0_start[];
 905 extern initcall_entry_t __initcall1_start[];
 906 extern initcall_entry_t __initcall2_start[];
 907 extern initcall_entry_t __initcall3_start[];
 908 extern initcall_entry_t __initcall4_start[];
 909 extern initcall_entry_t __initcall5_start[];
 910 extern initcall_entry_t __initcall6_start[];
 911 extern initcall_entry_t __initcall7_start[];
 912 extern initcall_entry_t __initcall_end[];
 913
 914 static initcall_entry_t *initcall_levels[] __initdata = {
 915         __initcall0_start,
 916         __initcall1_start,
 917         __initcall2_start,
 918         __initcall3_start,
 919         __initcall4_start,
 920         __initcall5_start,
 921         __initcall6_start,
 922         __initcall7_start,
 923         __initcall_end,
 924 };
 925
 926 /* Keep these in sync with initcalls in include/linux/init.h */
 927 static char *initcall_level_names[] __initdata = {
 928         "pure",
 929         "core",
 930         "postcore",
 931         "arch",
 932         "subsys",
 933         "fs",
 934         "device",
 935         "late",
 936 };
 937
 938 static void __init do_initcall_level(int level)
 939 {
 940         initcall_entry_t *fn;
 941
 942         strcpy(initcall_command_line, saved_command_line);
 943         parse_args(initcall_level_names[level],
 944                    initcall_command_line, __start___param,
 945                    __stop___param - __start___param,
 946                    level, level,
 947                    NULL, &repair_env_string);
 948
 949         trace_initcall_level(initcall_level_names[level]);
 950         for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
 951                 do_one_initcall(initcall_from_entry(fn));
 952 }
 953
 954 static void __init do_initcalls(void)
 955 {
 956         int level;
 957
 958         for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
 959                 do_initcall_level(level);
 960 }
 961
 962 /*
 963  * Ok, the machine is now initialized. None of the devices
 964  * have been touched yet, but the CPU subsystem is up and
 965  * running, and memory and process management works.
 966  *
 967  * Now we can finally start doing some real work..
 968  */
 969 static void __init do_basic_setup(void)
 970 {
 971         cpuset_init_smp();
 972         shmem_init();
 973         driver_init();
 974         init_irq_proc();
 975         do_ctors();
 976         usermodehelper_enable();
 977         do_initcalls();
 978 }
 979
 980 static void __init do_pre_smp_initcalls(void)
 981 {
 982         initcall_entry_t *fn;
 983
 984         trace_initcall_level("early");
 985         for (fn = __initcall_start; fn < __initcall0_start; fn++)
 986                 do_one_initcall(initcall_from_entry(fn));
 987 }
 988
 989 /*
 990  * This function requests modules which should be loaded by default and is
 991  * called twice right after initrd is mounted and right before init is
 992  * exec'd.  If such modules are on either initrd or rootfs, they will be
 993  * loaded before control is passed to userland.
 994  */
 995 void __init load_default_modules(void)
 996 {
 997         load_default_elevator_module();
 998 }
 999
1000 static int run_init_process(const char *init_filename)
1001 {
1002         argv_init[0] = init_filename;
1003         pr_info("Run %s as init process\n", init_filename);
1004         return do_execve(getname_kernel(init_filename),
1005                 (const char __user *const __user *)argv_init,
1006                 (const char __user *const __user *)envp_init);
1007 }
1008
1009 static int try_to_run_init_process(const char *init_filename)
1010 {
1011         int ret;
1012
1013         ret = run_init_process(init_filename);
1014
1015         if (ret && ret != -ENOENT) {
1016                 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1017                        init_filename, ret);
1018         }
1019
1020         return ret;
1021 }
1022
1023 static noinline void __init kernel_init_freeable(void);
1024
1025 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1026 bool rodata_enabled __ro_after_init = true;
1027 static int __init set_debug_rodata(char *str)
1028 {
1029         return strtobool(str, &rodata_enabled);
1030 }
1031 __setup("rodata=", set_debug_rodata);
1032 #endif
1033
1034 #ifdef CONFIG_STRICT_KERNEL_RWX
1035 static void mark_readonly(void)
1036 {
1037         if (rodata_enabled) {
1038                 /*
1039                  * load_module() results in W+X mappings, which are cleaned up
1040                  * with call_rcu_sched().  Let's make sure that queued work is
1041                  * flushed so that we don't hit false positives looking for
1042                  * insecure pages which are W+X.
1043                  */
1044                 rcu_barrier_sched();
1045                 mark_rodata_ro();
1046                 rodata_test();
1047         } else
1048                 pr_info("Kernel memory protection disabled.\n");
1049 }
1050 #else
1051 static inline void mark_readonly(void)
1052 {
1053         pr_warn("This architecture does not have kernel memory protection.\n");
1054 }
1055 #endif
1056
1057 static int __ref kernel_init(void *unused)
1058 {
1059         int ret;
1060
1061         kernel_init_freeable();
1062         /* need to finish all async __init code before freeing the memory */
1063         async_synchronize_full();
1064         ftrace_free_init_mem();
1065         jump_label_invalidate_initmem();
1066         free_initmem();
1067         mark_readonly();
1068
1069         /*
1070          * Kernel mappings are now finalized - update the userspace page-table
1071          * to finalize PTI.
1072          */
1073         pti_finalize();
1074
1075         system_state = SYSTEM_RUNNING;
1076         numa_default_policy();
1077
1078         rcu_end_inkernel_boot();
1079
1080         if (ramdisk_execute_command) {
1081                 ret = run_init_process(ramdisk_execute_command);
1082                 if (!ret)
1083                         return 0;
1084                 pr_err("Failed to execute %s (error %d)\n",
1085                        ramdisk_execute_command, ret);
1086         }
1087
1088         /*
1089          * We try each of these until one succeeds.
1090          *
1091          * The Bourne shell can be used instead of init if we are
1092          * trying to recover a really broken machine.
1093          */
1094         if (execute_command) {
1095                 ret = run_init_process(execute_command);
1096                 if (!ret)
1097                         return 0;
1098                 panic("Requested init %s failed (error %d).",
1099                       execute_command, ret);
1100         }
1101         if (!try_to_run_init_process("/sbin/init") ||
1102             !try_to_run_init_process("/etc/init") ||
1103             !try_to_run_init_process("/bin/init") ||
1104             !try_to_run_init_process("/bin/sh"))
1105                 return 0;
1106
1107         panic("No working init found.  Try passing init= option to kernel. "
1108               "See Linux Documentation/admin-guide/init.rst for guidance.");
1109 }
1110
1111 static noinline void __init kernel_init_freeable(void)
1112 {
1113         /*
1114          * Wait until kthreadd is all set-up.
1115          */
1116         wait_for_completion(&kthreadd_done);
1117
1118         /* Now the scheduler is fully set up and can do blocking allocations */
1119         gfp_allowed_mask = __GFP_BITS_MASK;
1120
1121         /*
1122          * init can allocate pages on any node
1123          */
1124         set_mems_allowed(node_states[N_MEMORY]);
1125
1126         cad_pid = task_pid(current);
1127
1128         smp_prepare_cpus(setup_max_cpus);
1129
1130         workqueue_init();
1131
1132         init_mm_internals();
1133
1134         do_pre_smp_initcalls();
1135         lockup_detector_init();
1136
1137         smp_init();
1138         sched_init_smp();
1139
1140         page_alloc_init_late();
1141         /* Initialize page ext after all struct pages are initialized. */
1142         page_ext_init();
1143
1144         do_basic_setup();
1145
1146         /* Open the /dev/console on the rootfs, this should never fail */
1147         if (ksys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1148                 pr_err("Warning: unable to open an initial console.\n");
1149
1150         (void) ksys_dup(0);
1151         (void) ksys_dup(0);
1152         /*
1153          * check if there is an early userspace init.  If yes, let it do all
1154          * the work
1155          */
1156
1157         if (!ramdisk_execute_command)
1158                 ramdisk_execute_command = "/init";
1159
1160         if (ksys_access((const char __user *)
1161                         ramdisk_execute_command, 0) != 0) {
1162                 ramdisk_execute_command = NULL;
1163                 prepare_namespace();
1164         }
1165
1166         /*
1167          * Ok, we have completed the initial bootup, and
1168          * we're essentially up and running. Get rid of the
1169          * initmem segments and start the user-mode stuff..
1170          *
1171          * rootfs is available now, try loading the public keys
1172          * and default modules
1173          */
1174
1175         integrity_load_keys();
1176         load_default_modules();
1177 }