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/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 fork_init(void);
 102 extern void radix_tree_init(void);
 103
 104 /*
 105  * Debug helper: via this flag we know that we are in 'early bootup code'
 106  * where only the boot processor is running with IRQ disabled.  This means
 107  * two things - IRQ must not be enabled before the flag is cleared and some
 108  * operations which are not allowed with IRQ disabled are allowed while the
 109  * flag is set.
 110  */
 111 bool early_boot_irqs_disabled __read_mostly;
 112
 113 enum system_states system_state __read_mostly;
 114 EXPORT_SYMBOL(system_state);
 115
 116 /*
 117  * Boot command-line arguments
 118  */
 119 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 120 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 121
 122 extern void time_init(void);
 123 /* Default late time init is NULL. archs can override this later. */
 124 void (*__initdata late_time_init)(void);
 125
 126 /* Untouched command line saved by arch-specific code. */
 127 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 128 /* Untouched saved command line (eg. for /proc) */
 129 char *saved_command_line;
 130 /* Command line for parameter parsing */
 131 static char *static_command_line;
 132 /* Command line for per-initcall parameter parsing */
 133 static char *initcall_command_line;
 134
 135 static char *execute_command;
 136 static char *ramdisk_execute_command;
 137
 138 /*
 139  * Used to generate warnings if static_key manipulation functions are used
 140  * before jump_label_init is called.
 141  */
 142 bool static_key_initialized __read_mostly;
 143 EXPORT_SYMBOL_GPL(static_key_initialized);
 144
 145 /*
 146  * If set, this is an indication to the drivers that reset the underlying
 147  * device before going ahead with the initialization otherwise driver might
 148  * rely on the BIOS and skip the reset operation.
 149  *
 150  * This is useful if kernel is booting in an unreliable environment.
 151  * For ex. kdump situation where previous kernel has crashed, BIOS has been
 152  * skipped and devices will be in unknown state.
 153  */
 154 unsigned int reset_devices;
 155 EXPORT_SYMBOL(reset_devices);
 156
 157 static int __init set_reset_devices(char *str)
 158 {
 159         reset_devices = 1;
 160         return 1;
 161 }
 162
 163 __setup("reset_devices", set_reset_devices);
 164
 165 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 166 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 167 static const char *panic_later, *panic_param;
 168
 169 extern const struct obs_kernel_param __setup_start[], __setup_end[];
 170
 171 static bool __init obsolete_checksetup(char *line)
 172 {
 173         const struct obs_kernel_param *p;
 174         bool had_early_param = false;
 175
 176         p = __setup_start;
 177         do {
 178                 int n = strlen(p->str);
 179                 if (parameqn(line, p->str, n)) {
 180                         if (p->early) {
 181                                 /* Already done in parse_early_param?
 182                                  * (Needs exact match on param part).
 183                                  * Keep iterating, as we can have early
 184                                  * params and __setups of same names 8( */
 185                                 if (line[n] == '\0' || line[n] == '=')
 186                                         had_early_param = true;
 187                         } else if (!p->setup_func) {
 188                                 pr_warn("Parameter %s is obsolete, ignored\n",
 189                                         p->str);
 190                                 return true;
 191                         } else if (p->setup_func(line + n))
 192                                 return true;
 193                 }
 194                 p++;
 195         } while (p < __setup_end);
 196
 197         return had_early_param;
 198 }
 199
 200 /*
 201  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 202  * still work even if initially too large, it will just take slightly longer
 203  */
 204 unsigned long loops_per_jiffy = (1<<12);
 205 EXPORT_SYMBOL(loops_per_jiffy);
 206
 207 static int __init debug_kernel(char *str)
 208 {
 209         console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
 210         return 0;
 211 }
 212
 213 static int __init quiet_kernel(char *str)
 214 {
 215         console_loglevel = CONSOLE_LOGLEVEL_QUIET;
 216         return 0;
 217 }
 218
 219 early_param("debug", debug_kernel);
 220 early_param("quiet", quiet_kernel);
 221
 222 static int __init loglevel(char *str)
 223 {
 224         int newlevel;
 225
 226         /*
 227          * Only update loglevel value when a correct setting was passed,
 228          * to prevent blind crashes (when loglevel being set to 0) that
 229          * are quite hard to debug
 230          */
 231         if (get_option(&str, &newlevel)) {
 232                 console_loglevel = newlevel;
 233                 return 0;
 234         }
 235
 236         return -EINVAL;
 237 }
 238
 239 early_param("loglevel", loglevel);
 240
 241 /* Change NUL term back to "=", to make "param" the whole string. */
 242 static int __init repair_env_string(char *param, char *val,
 243                                     const char *unused, void *arg)
 244 {
 245         if (val) {
 246                 /* param=val or param="val"? */
 247                 if (val == param+strlen(param)+1)
 248                         val[-1] = '=';
 249                 else if (val == param+strlen(param)+2) {
 250                         val[-2] = '=';
 251                         memmove(val-1, val, strlen(val)+1);
 252                         val--;
 253                 } else
 254                         BUG();
 255         }
 256         return 0;
 257 }
 258
 259 /* Anything after -- gets handed straight to init. */
 260 static int __init set_init_arg(char *param, char *val,
 261                                const char *unused, void *arg)
 262 {
 263         unsigned int i;
 264
 265         if (panic_later)
 266                 return 0;
 267
 268         repair_env_string(param, val, unused, NULL);
 269
 270         for (i = 0; argv_init[i]; i++) {
 271                 if (i == MAX_INIT_ARGS) {
 272                         panic_later = "init";
 273                         panic_param = param;
 274                         return 0;
 275                 }
 276         }
 277         argv_init[i] = param;
 278         return 0;
 279 }
 280
 281 /*
 282  * Unknown boot options get handed to init, unless they look like
 283  * unused parameters (modprobe will find them in /proc/cmdline).
 284  */
 285 static int __init unknown_bootoption(char *param, char *val,
 286                                      const char *unused, void *arg)
 287 {
 288         repair_env_string(param, val, unused, NULL);
 289
 290         /* Handle obsolete-style parameters */
 291         if (obsolete_checksetup(param))
 292                 return 0;
 293
 294         /* Unused module parameter. */
 295         if (strchr(param, '.') && (!val || strchr(param, '.') < val))
 296                 return 0;
 297
 298         if (panic_later)
 299                 return 0;
 300
 301         if (val) {
 302                 /* Environment option */
 303                 unsigned int i;
 304                 for (i = 0; envp_init[i]; i++) {
 305                         if (i == MAX_INIT_ENVS) {
 306                                 panic_later = "env";
 307                                 panic_param = param;
 308                         }
 309                         if (!strncmp(param, envp_init[i], val - param))
 310                                 break;
 311                 }
 312                 envp_init[i] = param;
 313         } else {
 314                 /* Command line option */
 315                 unsigned int i;
 316                 for (i = 0; argv_init[i]; i++) {
 317                         if (i == MAX_INIT_ARGS) {
 318                                 panic_later = "init";
 319                                 panic_param = param;
 320                         }
 321                 }
 322                 argv_init[i] = param;
 323         }
 324         return 0;
 325 }
 326
 327 static int __init init_setup(char *str)
 328 {
 329         unsigned int i;
 330
 331         execute_command = str;
 332         /*
 333          * In case LILO is going to boot us with default command line,
 334          * it prepends "auto" before the whole cmdline which makes
 335          * the shell think it should execute a script with such name.
 336          * So we ignore all arguments entered _before_ init=... [MJ]
 337          */
 338         for (i = 1; i < MAX_INIT_ARGS; i++)
 339                 argv_init[i] = NULL;
 340         return 1;
 341 }
 342 __setup("init=", init_setup);
 343
 344 static int __init rdinit_setup(char *str)
 345 {
 346         unsigned int i;
 347
 348         ramdisk_execute_command = str;
 349         /* See "auto" comment in init_setup */
 350         for (i = 1; i < MAX_INIT_ARGS; i++)
 351                 argv_init[i] = NULL;
 352         return 1;
 353 }
 354 __setup("rdinit=", rdinit_setup);
 355
 356 #ifndef CONFIG_SMP
 357 static const unsigned int setup_max_cpus = NR_CPUS;
 358 static inline void setup_nr_cpu_ids(void) { }
 359 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 360 #endif
 361
 362 /*
 363  * We need to store the untouched command line for future reference.
 364  * We also need to store the touched command line since the parameter
 365  * parsing is performed in place, and we should allow a component to
 366  * store reference of name/value for future reference.
 367  */
 368 static void __init setup_command_line(char *command_line)
 369 {
 370         saved_command_line =
 371                 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
 372         initcall_command_line =
 373                 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
 374         static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0);
 375         strcpy(saved_command_line, boot_command_line);
 376         strcpy(static_command_line, command_line);
 377 }
 378
 379 /*
 380  * We need to finalize in a non-__init function or else race conditions
 381  * between the root thread and the init thread may cause start_kernel to
 382  * be reaped by free_initmem before the root thread has proceeded to
 383  * cpu_idle.
 384  *
 385  * gcc-3.4 accidentally inlines this function, so use noinline.
 386  */
 387
 388 static __initdata DECLARE_COMPLETION(kthreadd_done);
 389
 390 static noinline void __ref rest_init(void)
 391 {
 392         struct task_struct *tsk;
 393         int pid;
 394
 395         rcu_scheduler_starting();
 396         /*
 397          * We need to spawn init first so that it obtains pid 1, however
 398          * the init task will end up wanting to create kthreads, which, if
 399          * we schedule it before we create kthreadd, will OOPS.
 400          */
 401         pid = kernel_thread(kernel_init, NULL, CLONE_FS);
 402         /*
 403          * Pin init on the boot CPU. Task migration is not properly working
 404          * until sched_init_smp() has been run. It will set the allowed
 405          * CPUs for init to the non isolated CPUs.
 406          */
 407         rcu_read_lock();
 408         tsk = find_task_by_pid_ns(pid, &init_pid_ns);
 409         set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
 410         rcu_read_unlock();
 411
 412         numa_default_policy();
 413         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 414         rcu_read_lock();
 415         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 416         rcu_read_unlock();
 417
 418         /*
 419          * Enable might_sleep() and smp_processor_id() checks.
 420          * They cannot be enabled earlier because with CONFIG_PRREMPT=y
 421          * kernel_thread() would trigger might_sleep() splats. With
 422          * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
 423          * already, but it's stuck on the kthreadd_done completion.
 424          */
 425         system_state = SYSTEM_SCHEDULING;
 426
 427         complete(&kthreadd_done);
 428
 429         /*
 430          * The boot idle thread must execute schedule()
 431          * at least once to get things moving:
 432          */
 433         schedule_preempt_disabled();
 434         /* Call into cpu_idle with preempt disabled */
 435         cpu_startup_entry(CPUHP_ONLINE);
 436 }
 437
 438 /* Check for early params. */
 439 static int __init do_early_param(char *param, char *val,
 440                                  const char *unused, void *arg)
 441 {
 442         const struct obs_kernel_param *p;
 443
 444         for (p = __setup_start; p < __setup_end; p++) {
 445                 if ((p->early && parameq(param, p->str)) ||
 446                     (strcmp(param, "console") == 0 &&
 447                      strcmp(p->str, "earlycon") == 0)
 448                 ) {
 449                         if (p->setup_func(val) != 0)
 450                                 pr_warn("Malformed early option '%s'\n", param);
 451                 }
 452         }
 453         /* We accept everything at this stage. */
 454         return 0;
 455 }
 456
 457 void __init parse_early_options(char *cmdline)
 458 {
 459         parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
 460                    do_early_param);
 461 }
 462
 463 /* Arch code calls this early on, or if not, just before other parsing. */
 464 void __init parse_early_param(void)
 465 {
 466         static int done __initdata;
 467         static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
 468
 469         if (done)
 470                 return;
 471
 472         /* All fall through to do_early_param. */
 473         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 474         parse_early_options(tmp_cmdline);
 475         done = 1;
 476 }
 477
 478 void __init __weak arch_post_acpi_subsys_init(void) { }
 479
 480 void __init __weak smp_setup_processor_id(void)
 481 {
 482 }
 483
 484 # if THREAD_SIZE >= PAGE_SIZE
 485 void __init __weak thread_stack_cache_init(void)
 486 {
 487 }
 488 #endif
 489
 490 void __init __weak mem_encrypt_init(void) { }
 491
 492 /*
 493  * Set up kernel memory allocators
 494  */
 495 static void __init mm_init(void)
 496 {
 497         /*
 498          * page_ext requires contiguous pages,
 499          * bigger than MAX_ORDER unless SPARSEMEM.
 500          */
 501         page_ext_init_flatmem();
 502         mem_init();
 503         kmem_cache_init();
 504         pgtable_init();
 505         vmalloc_init();
 506         ioremap_huge_init();
 507 }
 508
 509 asmlinkage __visible void __init start_kernel(void)
 510 {
 511         char *command_line;
 512         char *after_dashes;
 513
 514         set_task_stack_end_magic(&init_task);
 515         smp_setup_processor_id();
 516         debug_objects_early_init();
 517
 518         cgroup_init_early();
 519
 520         local_irq_disable();
 521         early_boot_irqs_disabled = true;
 522
 523         /*
 524          * Interrupts are still disabled. Do necessary setups, then
 525          * enable them.
 526          */
 527         boot_cpu_init();
 528         page_address_init();
 529         pr_notice("%s", linux_banner);
 530         setup_arch(&command_line);
 531         /*
 532          * Set up the the initial canary and entropy after arch
 533          * and after adding latent and command line entropy.
 534          */
 535         add_latent_entropy();
 536         add_device_randomness(command_line, strlen(command_line));
 537         boot_init_stack_canary();
 538         mm_init_cpumask(&init_mm);
 539         setup_command_line(command_line);
 540         setup_nr_cpu_ids();
 541         setup_per_cpu_areas();
 542         boot_cpu_state_init();
 543         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 544
 545         build_all_zonelists(NULL);
 546         page_alloc_init();
 547
 548         pr_notice("Kernel command line: %s\n", boot_command_line);
 549         parse_early_param();
 550         after_dashes = parse_args("Booting kernel",
 551                                   static_command_line, __start___param,
 552                                   __stop___param - __start___param,
 553                                   -1, -1, NULL, &unknown_bootoption);
 554         if (!IS_ERR_OR_NULL(after_dashes))
 555                 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
 556                            NULL, set_init_arg);
 557
 558         jump_label_init();
 559
 560         /*
 561          * These use large bootmem allocations and must precede
 562          * kmem_cache_init()
 563          */
 564         setup_log_buf(0);
 565         vfs_caches_init_early();
 566         sort_main_extable();
 567         trap_init();
 568         mm_init();
 569
 570         ftrace_init();
 571
 572         /* trace_printk can be enabled here */
 573         early_trace_init();
 574
 575         /*
 576          * Set up the scheduler prior starting any interrupts (such as the
 577          * timer interrupt). Full topology setup happens at smp_init()
 578          * time - but meanwhile we still have a functioning scheduler.
 579          */
 580         sched_init();
 581         /*
 582          * Disable preemption - early bootup scheduling is extremely
 583          * fragile until we cpu_idle() for the first time.
 584          */
 585         preempt_disable();
 586         if (WARN(!irqs_disabled(),
 587                  "Interrupts were enabled *very* early, fixing it\n"))
 588                 local_irq_disable();
 589         radix_tree_init();
 590
 591         /*
 592          * Allow workqueue creation and work item queueing/cancelling
 593          * early.  Work item execution depends on kthreads and starts after
 594          * workqueue_init().
 595          */
 596         workqueue_init_early();
 597
 598         rcu_init();
 599
 600         /* Trace events are available after this */
 601         trace_init();
 602
 603         context_tracking_init();
 604         /* init some links before init_ISA_irqs() */
 605         early_irq_init();
 606         init_IRQ();
 607         tick_init();
 608         housekeeping_init();
 609         rcu_init_nohz();
 610         init_timers();
 611         hrtimers_init();
 612         softirq_init();
 613         timekeeping_init();
 614         time_init();
 615         sched_clock_postinit();
 616         printk_safe_init();
 617         perf_event_init();
 618         profile_init();
 619         call_function_init();
 620         WARN(!irqs_disabled(), "Interrupts were enabled early\n");
 621         early_boot_irqs_disabled = false;
 622         local_irq_enable();
 623
 624         kmem_cache_init_late();
 625
 626         /*
 627          * HACK ALERT! This is early. We're enabling the console before
 628          * we've done PCI setups etc, and console_init() must be aware of
 629          * this. But we do want output early, in case something goes wrong.
 630          */
 631         console_init();
 632         if (panic_later)
 633                 panic("Too many boot %s vars at `%s'", panic_later,
 634                       panic_param);
 635
 636         lockdep_info();
 637
 638         /*
 639          * Need to run this when irqs are enabled, because it wants
 640          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 641          * too:
 642          */
 643         locking_selftest();
 644
 645         /*
 646          * This needs to be called before any devices perform DMA
 647          * operations that might use the SWIOTLB bounce buffers. It will
 648          * mark the bounce buffers as decrypted so that their usage will
 649          * not cause "plain-text" data to be decrypted when accessed.
 650          */
 651         mem_encrypt_init();
 652
 653 #ifdef CONFIG_BLK_DEV_INITRD
 654         if (initrd_start && !initrd_below_start_ok &&
 655             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 656                 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
 657                     page_to_pfn(virt_to_page((void *)initrd_start)),
 658                     min_low_pfn);
 659                 initrd_start = 0;
 660         }
 661 #endif
 662         page_ext_init();
 663         kmemleak_init();
 664         debug_objects_mem_init();
 665         setup_per_cpu_pageset();
 666         numa_policy_init();
 667         acpi_early_init();
 668         if (late_time_init)
 669                 late_time_init();
 670         calibrate_delay();
 671         pid_idr_init();
 672         anon_vma_init();
 673 #ifdef CONFIG_X86
 674         if (efi_enabled(EFI_RUNTIME_SERVICES))
 675                 efi_enter_virtual_mode();
 676 #endif
 677 #ifdef CONFIG_X86_ESPFIX64
 678         /* Should be run before the first non-init thread is created */
 679         init_espfix_bsp();
 680 #endif
 681         thread_stack_cache_init();
 682         cred_init();
 683         fork_init();
 684         proc_caches_init();
 685         buffer_init();
 686         key_init();
 687         security_init();
 688         dbg_late_init();
 689         vfs_caches_init();
 690         pagecache_init();
 691         signals_init();
 692         proc_root_init();
 693         nsfs_init();
 694         cpuset_init();
 695         cgroup_init();
 696         taskstats_init_early();
 697         delayacct_init();
 698
 699         check_bugs();
 700
 701         acpi_subsystem_init();
 702         arch_post_acpi_subsys_init();
 703         sfi_init_late();
 704
 705         if (efi_enabled(EFI_RUNTIME_SERVICES)) {
 706                 efi_free_boot_services();
 707         }
 708
 709         /* Do the rest non-__init'ed, we're now alive */
 710         rest_init();
 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                 mark_rodata_ro();
 978                 rodata_test();
 979         } else
 980                 pr_info("Kernel memory protection disabled.\n");
 981 }
 982 #else
 983 static inline void mark_readonly(void)
 984 {
 985         pr_warn("This architecture does not have kernel memory protection.\n");
 986 }
 987 #endif
 988
 989 static int __ref kernel_init(void *unused)
 990 {
 991         int ret;
 992
 993         kernel_init_freeable();
 994         /* need to finish all async __init code before freeing the memory */
 995         async_synchronize_full();
 996         ftrace_free_init_mem();
 997         free_initmem();
 998         mark_readonly();
 999         system_state = SYSTEM_RUNNING;
1000         numa_default_policy();
1001
1002         rcu_end_inkernel_boot();
1003
1004         if (ramdisk_execute_command) {
1005                 ret = run_init_process(ramdisk_execute_command);
1006                 if (!ret)
1007                         return 0;
1008                 pr_err("Failed to execute %s (error %d)\n",
1009                        ramdisk_execute_command, ret);
1010         }
1011
1012         /*
1013          * We try each of these until one succeeds.
1014          *
1015          * The Bourne shell can be used instead of init if we are
1016          * trying to recover a really broken machine.
1017          */
1018         if (execute_command) {
1019                 ret = run_init_process(execute_command);
1020                 if (!ret)
1021                         return 0;
1022                 panic("Requested init %s failed (error %d).",
1023                       execute_command, ret);
1024         }
1025         if (!try_to_run_init_process("/sbin/init") ||
1026             !try_to_run_init_process("/etc/init") ||
1027             !try_to_run_init_process("/bin/init") ||
1028             !try_to_run_init_process("/bin/sh"))
1029                 return 0;
1030
1031         panic("No working init found.  Try passing init= option to kernel. "
1032               "See Linux Documentation/admin-guide/init.rst for guidance.");
1033 }
1034
1035 static noinline void __init kernel_init_freeable(void)
1036 {
1037         /*
1038          * Wait until kthreadd is all set-up.
1039          */
1040         wait_for_completion(&kthreadd_done);
1041
1042         /* Now the scheduler is fully set up and can do blocking allocations */
1043         gfp_allowed_mask = __GFP_BITS_MASK;
1044
1045         /*
1046          * init can allocate pages on any node
1047          */
1048         set_mems_allowed(node_states[N_MEMORY]);
1049
1050         cad_pid = task_pid(current);
1051
1052         smp_prepare_cpus(setup_max_cpus);
1053
1054         workqueue_init();
1055
1056         init_mm_internals();
1057
1058         do_pre_smp_initcalls();
1059         lockup_detector_init();
1060
1061         smp_init();
1062         sched_init_smp();
1063
1064         page_alloc_init_late();
1065
1066         do_basic_setup();
1067
1068         /* Open the /dev/console on the rootfs, this should never fail */
1069         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1070                 pr_err("Warning: unable to open an initial console.\n");
1071
1072         (void) sys_dup(0);
1073         (void) sys_dup(0);
1074         /*
1075          * check if there is an early userspace init.  If yes, let it do all
1076          * the work
1077          */
1078
1079         if (!ramdisk_execute_command)
1080                 ramdisk_execute_command = "/init";
1081
1082         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
1083                 ramdisk_execute_command = NULL;
1084                 prepare_namespace();
1085         }
1086
1087         /*
1088          * Ok, we have completed the initial bootup, and
1089          * we're essentially up and running. Get rid of the
1090          * initmem segments and start the user-mode stuff..
1091          *
1092          * rootfs is available now, try loading the public keys
1093          * and default modules
1094          */
1095
1096         integrity_load_keys();
1097         load_default_modules();
1098 }