1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * linux/init/main.c
4 *
5 * Copyright (C) 1991, 1992 Linus Torvalds
6 *
7 * GK 2/5/95 - Changed to support mounting root fs via NFS
8 * Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
9 * Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
10 * Simplified starting of init: Michael A. Griffith <grif@acm.org>
11 */
12
13 #define DEBUG /* Enable initcall_debug */
14
15 #include <linux/types.h>
16 #include <linux/extable.h>
17 #include <linux/module.h>
18 #include <linux/proc_fs.h>
19 #include <linux/binfmts.h>
20 #include <linux/kernel.h>
21 #include <linux/syscalls.h>
22 #include <linux/stackprotector.h>
23 #include <linux/string.h>
24 #include <linux/ctype.h>
25 #include <linux/delay.h>
26 #include <linux/ioport.h>
27 #include <linux/init.h>
28 #include <linux/initrd.h>
29 #include <linux/memblock.h>
30 #include <linux/acpi.h>
31 #include <linux/console.h>
32 #include <linux/nmi.h>
33 #include <linux/percpu.h>
34 #include <linux/kmod.h>
35 #include <linux/vmalloc.h>
36 #include <linux/kernel_stat.h>
37 #include <linux/start_kernel.h>
38 #include <linux/security.h>
39 #include <linux/smp.h>
40 #include <linux/profile.h>
41 #include <linux/rcupdate.h>
42 #include <linux/moduleparam.h>
43 #include <linux/kallsyms.h>
44 #include <linux/writeback.h>
45 #include <linux/cpu.h>
46 #include <linux/cpuset.h>
47 #include <linux/cgroup.h>
48 #include <linux/efi.h>
49 #include <linux/tick.h>
50 #include <linux/sched/isolation.h>
51 #include <linux/interrupt.h>
52 #include <linux/taskstats_kern.h>
53 #include <linux/delayacct.h>
54 #include <linux/unistd.h>
55 #include <linux/utsname.h>
56 #include <linux/rmap.h>
57 #include <linux/mempolicy.h>
58 #include <linux/key.h>
59 #include <linux/buffer_head.h>
60 #include <linux/page_ext.h>
61 #include <linux/debug_locks.h>
62 #include <linux/debugobjects.h>
63 #include <linux/lockdep.h>
64 #include <linux/kmemleak.h>
65 #include <linux/pid_namespace.h>
66 #include <linux/device.h>
67 #include <linux/kthread.h>
68 #include <linux/sched.h>
69 #include <linux/sched/init.h>
70 #include <linux/signal.h>
71 #include <linux/idr.h>
72 #include <linux/kgdb.h>
73 #include <linux/ftrace.h>
74 #include <linux/async.h>
75 #include <linux/sfi.h>
76 #include <linux/shmem_fs.h>
77 #include <linux/slab.h>
78 #include <linux/perf_event.h>
79 #include <linux/ptrace.h>
80 #include <linux/pti.h>
81 #include <linux/blkdev.h>
82 #include <linux/elevator.h>
83 #include <linux/sched/clock.h>
84 #include <linux/sched/task.h>
85 #include <linux/sched/task_stack.h>
86 #include <linux/context_tracking.h>
87 #include <linux/random.h>
88 #include <linux/list.h>
89 #include <linux/integrity.h>
90 #include <linux/proc_ns.h>
91 #include <linux/io.h>
92 #include <linux/cache.h>
93 #include <linux/rodata_test.h>
94 #include <linux/jump_label.h>
95 #include <linux/mem_encrypt.h>
96
97 #include <asm/io.h>
98 #include <asm/bugs.h>
99 #include <asm/setup.h>
100 #include <asm/sections.h>
101 #include <asm/cacheflush.h>
102
103 #define CREATE_TRACE_POINTS
104 #include <trace/events/initcall.h>
105
106 static int kernel_init(void *);
107
108 extern void init_IRQ(void);
109 extern void radix_tree_init(void);
110
111 /*
112 * Debug helper: via this flag we know that we are in 'early bootup code'
113 * where only the boot processor is running with IRQ disabled. This means
114 * two things - IRQ must not be enabled before the flag is cleared and some
115 * operations which are not allowed with IRQ disabled are allowed while the
116 * flag is set.
117 */
118 bool early_boot_irqs_disabled __read_mostly;
119
120 enum system_states system_state __read_mostly;
121 EXPORT_SYMBOL(system_state);
122
123 /*
124 * Boot command-line arguments
125 */
126 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
127 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
128
129 extern void time_init(void);
130 /* Default late time init is NULL. archs can override this later. */
131 void (*__initdata late_time_init)(void);
132
133 /* Untouched command line saved by arch-specific code. */
134 char __initdata boot_command_line[COMMAND_LINE_SIZE];
135 /* Untouched saved command line (eg. for /proc) */
136 char *saved_command_line;
137 /* Command line for parameter parsing */
138 static char *static_command_line;
139 /* Command line for per-initcall parameter parsing */
140 static char *initcall_command_line;
141
142 static char *execute_command;
143 static char *ramdisk_execute_command;
144
145 /*
146 * Used to generate warnings if static_key manipulation functions are used
147 * before jump_label_init is called.
148 */
149 bool static_key_initialized __read_mostly;
150 EXPORT_SYMBOL_GPL(static_key_initialized);
151
152 /*
153 * If set, this is an indication to the drivers that reset the underlying
154 * device before going ahead with the initialization otherwise driver might
155 * rely on the BIOS and skip the reset operation.
156 *
157 * This is useful if kernel is booting in an unreliable environment.
158 * For ex. kdump situation where previous kernel has crashed, BIOS has been
159 * skipped and devices will be in unknown state.
160 */
161 unsigned int reset_devices;
162 EXPORT_SYMBOL(reset_devices);
163
set_reset_devices(char * str)164 static int __init set_reset_devices(char *str)
165 {
166 reset_devices = 1;
167 return 1;
168 }
169
170 __setup("reset_devices", set_reset_devices);
171
172 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
173 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
174 static const char *panic_later, *panic_param;
175
176 extern const struct obs_kernel_param __setup_start[], __setup_end[];
177
obsolete_checksetup(char * line)178 static bool __init obsolete_checksetup(char *line)
179 {
180 const struct obs_kernel_param *p;
181 bool had_early_param = false;
182
183 p = __setup_start;
184 do {
185 int n = strlen(p->str);
186 if (parameqn(line, p->str, n)) {
187 if (p->early) {
188 /* Already done in parse_early_param?
189 * (Needs exact match on param part).
190 * Keep iterating, as we can have early
191 * params and __setups of same names 8( */
192 if (line[n] == '\0' || line[n] == '=')
193 had_early_param = true;
194 } else if (!p->setup_func) {
195 pr_warn("Parameter %s is obsolete, ignored\n",
196 p->str);
197 return true;
198 } else if (p->setup_func(line + n))
199 return true;
200 }
201 p++;
202 } while (p < __setup_end);
203
204 return had_early_param;
205 }
206
207 /*
208 * This should be approx 2 Bo*oMips to start (note initial shift), and will
209 * still work even if initially too large, it will just take slightly longer
210 */
211 unsigned long loops_per_jiffy = (1<<12);
212 EXPORT_SYMBOL(loops_per_jiffy);
213
debug_kernel(char * str)214 static int __init debug_kernel(char *str)
215 {
216 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
217 return 0;
218 }
219
quiet_kernel(char * str)220 static int __init quiet_kernel(char *str)
221 {
222 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
223 return 0;
224 }
225
226 early_param("debug", debug_kernel);
227 early_param("quiet", quiet_kernel);
228
loglevel(char * str)229 static int __init loglevel(char *str)
230 {
231 int newlevel;
232
233 /*
234 * Only update loglevel value when a correct setting was passed,
235 * to prevent blind crashes (when loglevel being set to 0) that
236 * are quite hard to debug
237 */
238 if (get_option(&str, &newlevel)) {
239 console_loglevel = newlevel;
240 return 0;
241 }
242
243 return -EINVAL;
244 }
245
246 early_param("loglevel", loglevel);
247
248 /* Change NUL term back to "=", to make "param" the whole string. */
repair_env_string(char * param,char * val,const char * unused,void * arg)249 static int __init repair_env_string(char *param, char *val,
250 const char *unused, void *arg)
251 {
252 if (val) {
253 /* param=val or param="val"? */
254 if (val == param+strlen(param)+1)
255 val[-1] = '=';
256 else if (val == param+strlen(param)+2) {
257 val[-2] = '=';
258 memmove(val-1, val, strlen(val)+1);
259 val--;
260 } else
261 BUG();
262 }
263 return 0;
264 }
265
266 /* Anything after -- gets handed straight to init. */
set_init_arg(char * param,char * val,const char * unused,void * arg)267 static int __init set_init_arg(char *param, char *val,
268 const char *unused, void *arg)
269 {
270 unsigned int i;
271
272 if (panic_later)
273 return 0;
274
275 repair_env_string(param, val, unused, NULL);
276
277 for (i = 0; argv_init[i]; i++) {
278 if (i == MAX_INIT_ARGS) {
279 panic_later = "init";
280 panic_param = param;
281 return 0;
282 }
283 }
284 argv_init[i] = param;
285 return 0;
286 }
287
288 /*
289 * Unknown boot options get handed to init, unless they look like
290 * unused parameters (modprobe will find them in /proc/cmdline).
291 */
unknown_bootoption(char * param,char * val,const char * unused,void * arg)292 static int __init unknown_bootoption(char *param, char *val,
293 const char *unused, void *arg)
294 {
295 repair_env_string(param, val, unused, NULL);
296
297 /* Handle obsolete-style parameters */
298 if (obsolete_checksetup(param))
299 return 0;
300
301 /* Unused module parameter. */
302 if (strchr(param, '.') && (!val || strchr(param, '.') < val))
303 return 0;
304
305 if (panic_later)
306 return 0;
307
308 if (val) {
309 /* Environment option */
310 unsigned int i;
311 for (i = 0; envp_init[i]; i++) {
312 if (i == MAX_INIT_ENVS) {
313 panic_later = "env";
314 panic_param = param;
315 }
316 if (!strncmp(param, envp_init[i], val - param))
317 break;
318 }
319 envp_init[i] = param;
320 } else {
321 /* Command line option */
322 unsigned int i;
323 for (i = 0; argv_init[i]; i++) {
324 if (i == MAX_INIT_ARGS) {
325 panic_later = "init";
326 panic_param = param;
327 }
328 }
329 argv_init[i] = param;
330 }
331 return 0;
332 }
333
init_setup(char * str)334 static int __init init_setup(char *str)
335 {
336 unsigned int i;
337
338 execute_command = str;
339 /*
340 * In case LILO is going to boot us with default command line,
341 * it prepends "auto" before the whole cmdline which makes
342 * the shell think it should execute a script with such name.
343 * So we ignore all arguments entered _before_ init=... [MJ]
344 */
345 for (i = 1; i < MAX_INIT_ARGS; i++)
346 argv_init[i] = NULL;
347 return 1;
348 }
349 __setup("init=", init_setup);
350
rdinit_setup(char * str)351 static int __init rdinit_setup(char *str)
352 {
353 unsigned int i;
354
355 ramdisk_execute_command = str;
356 /* See "auto" comment in init_setup */
357 for (i = 1; i < MAX_INIT_ARGS; i++)
358 argv_init[i] = NULL;
359 return 1;
360 }
361 __setup("rdinit=", rdinit_setup);
362
363 #ifndef CONFIG_SMP
364 static const unsigned int setup_max_cpus = NR_CPUS;
setup_nr_cpu_ids(void)365 static inline void setup_nr_cpu_ids(void) { }
smp_prepare_cpus(unsigned int maxcpus)366 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
367 #endif
368
369 /*
370 * We need to store the untouched command line for future reference.
371 * We also need to store the touched command line since the parameter
372 * parsing is performed in place, and we should allow a component to
373 * store reference of name/value for future reference.
374 */
setup_command_line(char * command_line)375 static void __init setup_command_line(char *command_line)
376 {
377 size_t len = strlen(boot_command_line) + 1;
378
379 saved_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
380 if (!saved_command_line)
381 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
382
383 initcall_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
384 if (!initcall_command_line)
385 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
386
387 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
388 if (!static_command_line)
389 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
390
391 strcpy(saved_command_line, boot_command_line);
392 strcpy(static_command_line, command_line);
393 }
394
395 /*
396 * We need to finalize in a non-__init function or else race conditions
397 * between the root thread and the init thread may cause start_kernel to
398 * be reaped by free_initmem before the root thread has proceeded to
399 * cpu_idle.
400 *
401 * gcc-3.4 accidentally inlines this function, so use noinline.
402 */
403
404 static __initdata DECLARE_COMPLETION(kthreadd_done);
405
rest_init(void)406 noinline void __ref rest_init(void)
407 {
408 struct task_struct *tsk;
409 int pid;
410
411 rcu_scheduler_starting();
412 /*
413 * We need to spawn init first so that it obtains pid 1, however
414 * the init task will end up wanting to create kthreads, which, if
415 * we schedule it before we create kthreadd, will OOPS.
416 */
417 pid = kernel_thread(kernel_init, NULL, CLONE_FS);
418 /*
419 * Pin init on the boot CPU. Task migration is not properly working
420 * until sched_init_smp() has been run. It will set the allowed
421 * CPUs for init to the non isolated CPUs.
422 */
423 rcu_read_lock();
424 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
425 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
426 rcu_read_unlock();
427
428 numa_default_policy();
429 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
430 rcu_read_lock();
431 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
432 rcu_read_unlock();
433
434 /*
435 * Enable might_sleep() and smp_processor_id() checks.
436 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
437 * kernel_thread() would trigger might_sleep() splats. With
438 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
439 * already, but it's stuck on the kthreadd_done completion.
440 */
441 system_state = SYSTEM_SCHEDULING;
442
443 complete(&kthreadd_done);
444
445 /*
446 * The boot idle thread must execute schedule()
447 * at least once to get things moving:
448 */
449 schedule_preempt_disabled();
450 /* Call into cpu_idle with preempt disabled */
451 cpu_startup_entry(CPUHP_ONLINE);
452 }
453
454 /* Check for early params. */
do_early_param(char * param,char * val,const char * unused,void * arg)455 static int __init do_early_param(char *param, char *val,
456 const char *unused, void *arg)
457 {
458 const struct obs_kernel_param *p;
459
460 for (p = __setup_start; p < __setup_end; p++) {
461 if ((p->early && parameq(param, p->str)) ||
462 (strcmp(param, "console") == 0 &&
463 strcmp(p->str, "earlycon") == 0)
464 ) {
465 if (p->setup_func(val) != 0)
466 pr_warn("Malformed early option '%s'\n", param);
467 }
468 }
469 /* We accept everything at this stage. */
470 return 0;
471 }
472
parse_early_options(char * cmdline)473 void __init parse_early_options(char *cmdline)
474 {
475 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
476 do_early_param);
477 }
478
479 /* Arch code calls this early on, or if not, just before other parsing. */
parse_early_param(void)480 void __init parse_early_param(void)
481 {
482 static int done __initdata;
483 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
484
485 if (done)
486 return;
487
488 /* All fall through to do_early_param. */
489 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
490 parse_early_options(tmp_cmdline);
491 done = 1;
492 }
493
arch_post_acpi_subsys_init(void)494 void __init __weak arch_post_acpi_subsys_init(void) { }
495
smp_setup_processor_id(void)496 void __init __weak smp_setup_processor_id(void)
497 {
498 }
499
500 # if THREAD_SIZE >= PAGE_SIZE
thread_stack_cache_init(void)501 void __init __weak thread_stack_cache_init(void)
502 {
503 }
504 #endif
505
mem_encrypt_init(void)506 void __init __weak mem_encrypt_init(void) { }
507
poking_init(void)508 void __init __weak poking_init(void) { }
509
pgtable_cache_init(void)510 void __init __weak pgtable_cache_init(void) { }
511
512 bool initcall_debug;
513 core_param(initcall_debug, initcall_debug, bool, 0644);
514
515 #ifdef TRACEPOINTS_ENABLED
516 static void __init initcall_debug_enable(void);
517 #else
initcall_debug_enable(void)518 static inline void initcall_debug_enable(void)
519 {
520 }
521 #endif
522
523 /* Report memory auto-initialization states for this boot. */
report_meminit(void)524 static void __init report_meminit(void)
525 {
526 const char *stack;
527
528 if (IS_ENABLED(CONFIG_INIT_STACK_ALL))
529 stack = "all";
530 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
531 stack = "byref_all";
532 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
533 stack = "byref";
534 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
535 stack = "__user";
536 else
537 stack = "off";
538
539 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
540 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
541 want_init_on_free() ? "on" : "off");
542 if (want_init_on_free())
543 pr_info("mem auto-init: clearing system memory may take some time...\n");
544 }
545
546 /*
547 * Set up kernel memory allocators
548 */
mm_init(void)549 static void __init mm_init(void)
550 {
551 /*
552 * page_ext requires contiguous pages,
553 * bigger than MAX_ORDER unless SPARSEMEM.
554 */
555 page_ext_init_flatmem();
556 init_debug_pagealloc();
557 report_meminit();
558 mem_init();
559 kmem_cache_init();
560 kmemleak_init();
561 pgtable_init();
562 debug_objects_mem_init();
563 vmalloc_init();
564 ioremap_huge_init();
565 /* Should be run before the first non-init thread is created */
566 init_espfix_bsp();
567 /* Should be run after espfix64 is set up. */
568 pti_init();
569 }
570
arch_call_rest_init(void)571 void __init __weak arch_call_rest_init(void)
572 {
573 rest_init();
574 }
575
start_kernel(void)576 asmlinkage __visible void __init start_kernel(void)
577 {
578 char *command_line;
579 char *after_dashes;
580
581 set_task_stack_end_magic(&init_task);
582 smp_setup_processor_id();
583 debug_objects_early_init();
584
585 cgroup_init_early();
586
587 local_irq_disable();
588 early_boot_irqs_disabled = true;
589
590 /*
591 * Interrupts are still disabled. Do necessary setups, then
592 * enable them.
593 */
594 boot_cpu_init();
595 page_address_init();
596 pr_notice("%s", linux_banner);
597 early_security_init();
598 setup_arch(&command_line);
599 setup_command_line(command_line);
600 setup_nr_cpu_ids();
601 setup_per_cpu_areas();
602 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
603 boot_cpu_hotplug_init();
604
605 build_all_zonelists(NULL);
606 page_alloc_init();
607
608 pr_notice("Kernel command line: %s\n", boot_command_line);
609 /* parameters may set static keys */
610 jump_label_init();
611 parse_early_param();
612 after_dashes = parse_args("Booting kernel",
613 static_command_line, __start___param,
614 __stop___param - __start___param,
615 -1, -1, NULL, &unknown_bootoption);
616 if (!IS_ERR_OR_NULL(after_dashes))
617 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
618 NULL, set_init_arg);
619
620 /*
621 * These use large bootmem allocations and must precede
622 * kmem_cache_init()
623 */
624 setup_log_buf(0);
625 vfs_caches_init_early();
626 sort_main_extable();
627 trap_init();
628 mm_init();
629
630 ftrace_init();
631
632 /* trace_printk can be enabled here */
633 early_trace_init();
634
635 /*
636 * Set up the scheduler prior starting any interrupts (such as the
637 * timer interrupt). Full topology setup happens at smp_init()
638 * time - but meanwhile we still have a functioning scheduler.
639 */
640 sched_init();
641 /*
642 * Disable preemption - early bootup scheduling is extremely
643 * fragile until we cpu_idle() for the first time.
644 */
645 preempt_disable();
646 if (WARN(!irqs_disabled(),
647 "Interrupts were enabled *very* early, fixing it\n"))
648 local_irq_disable();
649 radix_tree_init();
650
651 /*
652 * Set up housekeeping before setting up workqueues to allow the unbound
653 * workqueue to take non-housekeeping into account.
654 */
655 housekeeping_init();
656
657 /*
658 * Allow workqueue creation and work item queueing/cancelling
659 * early. Work item execution depends on kthreads and starts after
660 * workqueue_init().
661 */
662 workqueue_init_early();
663
664 rcu_init();
665
666 /* Trace events are available after this */
667 trace_init();
668
669 if (initcall_debug)
670 initcall_debug_enable();
671
672 context_tracking_init();
673 /* init some links before init_ISA_irqs() */
674 early_irq_init();
675 init_IRQ();
676 tick_init();
677 rcu_init_nohz();
678 init_timers();
679 hrtimers_init();
680 softirq_init();
681 timekeeping_init();
682
683 /*
684 * For best initial stack canary entropy, prepare it after:
685 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
686 * - timekeeping_init() for ktime entropy used in rand_initialize()
687 * - rand_initialize() to get any arch-specific entropy like RDRAND
688 * - add_latent_entropy() to get any latent entropy
689 * - adding command line entropy
690 */
691 rand_initialize();
692 add_latent_entropy();
693 add_device_randomness(command_line, strlen(command_line));
694 boot_init_stack_canary();
695
696 time_init();
697 printk_safe_init();
698 perf_event_init();
699 profile_init();
700 call_function_init();
701 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
702
703 early_boot_irqs_disabled = false;
704 local_irq_enable();
705
706 kmem_cache_init_late();
707
708 /*
709 * HACK ALERT! This is early. We're enabling the console before
710 * we've done PCI setups etc, and console_init() must be aware of
711 * this. But we do want output early, in case something goes wrong.
712 */
713 console_init();
714 if (panic_later)
715 panic("Too many boot %s vars at `%s'", panic_later,
716 panic_param);
717
718 lockdep_init();
719
720 /*
721 * Need to run this when irqs are enabled, because it wants
722 * to self-test [hard/soft]-irqs on/off lock inversion bugs
723 * too:
724 */
725 locking_selftest();
726
727 /*
728 * This needs to be called before any devices perform DMA
729 * operations that might use the SWIOTLB bounce buffers. It will
730 * mark the bounce buffers as decrypted so that their usage will
731 * not cause "plain-text" data to be decrypted when accessed.
732 */
733 mem_encrypt_init();
734
735 #ifdef CONFIG_BLK_DEV_INITRD
736 if (initrd_start && !initrd_below_start_ok &&
737 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
738 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
739 page_to_pfn(virt_to_page((void *)initrd_start)),
740 min_low_pfn);
741 initrd_start = 0;
742 }
743 #endif
744 setup_per_cpu_pageset();
745 numa_policy_init();
746 acpi_early_init();
747 if (late_time_init)
748 late_time_init();
749 sched_clock_init();
750 calibrate_delay();
751 pid_idr_init();
752 anon_vma_init();
753 #ifdef CONFIG_X86
754 if (efi_enabled(EFI_RUNTIME_SERVICES))
755 efi_enter_virtual_mode();
756 #endif
757 thread_stack_cache_init();
758 cred_init();
759 fork_init();
760 proc_caches_init();
761 uts_ns_init();
762 buffer_init();
763 key_init();
764 security_init();
765 dbg_late_init();
766 vfs_caches_init();
767 pagecache_init();
768 signals_init();
769 seq_file_init();
770 proc_root_init();
771 nsfs_init();
772 cpuset_init();
773 cgroup_init();
774 taskstats_init_early();
775 delayacct_init();
776
777 poking_init();
778 check_bugs();
779
780 acpi_subsystem_init();
781 arch_post_acpi_subsys_init();
782 sfi_init_late();
783
784 /* Do the rest non-__init'ed, we're now alive */
785 arch_call_rest_init();
786 }
787
788 /* Call all constructor functions linked into the kernel. */
do_ctors(void)789 static void __init do_ctors(void)
790 {
791 #ifdef CONFIG_CONSTRUCTORS
792 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
793
794 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
795 (*fn)();
796 #endif
797 }
798
799 #ifdef CONFIG_KALLSYMS
800 struct blacklist_entry {
801 struct list_head next;
802 char *buf;
803 };
804
805 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
806
initcall_blacklist(char * str)807 static int __init initcall_blacklist(char *str)
808 {
809 char *str_entry;
810 struct blacklist_entry *entry;
811
812 /* str argument is a comma-separated list of functions */
813 do {
814 str_entry = strsep(&str, ",");
815 if (str_entry) {
816 pr_debug("blacklisting initcall %s\n", str_entry);
817 entry = memblock_alloc(sizeof(*entry),
818 SMP_CACHE_BYTES);
819 if (!entry)
820 panic("%s: Failed to allocate %zu bytes\n",
821 __func__, sizeof(*entry));
822 entry->buf = memblock_alloc(strlen(str_entry) + 1,
823 SMP_CACHE_BYTES);
824 if (!entry->buf)
825 panic("%s: Failed to allocate %zu bytes\n",
826 __func__, strlen(str_entry) + 1);
827 strcpy(entry->buf, str_entry);
828 list_add(&entry->next, &blacklisted_initcalls);
829 }
830 } while (str_entry);
831
832 return 0;
833 }
834
initcall_blacklisted(initcall_t fn)835 static bool __init_or_module initcall_blacklisted(initcall_t fn)
836 {
837 struct blacklist_entry *entry;
838 char fn_name[KSYM_SYMBOL_LEN];
839 unsigned long addr;
840
841 if (list_empty(&blacklisted_initcalls))
842 return false;
843
844 addr = (unsigned long) dereference_function_descriptor(fn);
845 sprint_symbol_no_offset(fn_name, addr);
846
847 /*
848 * fn will be "function_name [module_name]" where [module_name] is not
849 * displayed for built-in init functions. Strip off the [module_name].
850 */
851 strreplace(fn_name, ' ', '\0');
852
853 list_for_each_entry(entry, &blacklisted_initcalls, next) {
854 if (!strcmp(fn_name, entry->buf)) {
855 pr_debug("initcall %s blacklisted\n", fn_name);
856 return true;
857 }
858 }
859
860 return false;
861 }
862 #else
initcall_blacklist(char * str)863 static int __init initcall_blacklist(char *str)
864 {
865 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
866 return 0;
867 }
868
initcall_blacklisted(initcall_t fn)869 static bool __init_or_module initcall_blacklisted(initcall_t fn)
870 {
871 return false;
872 }
873 #endif
874 __setup("initcall_blacklist=", initcall_blacklist);
875
876 static __init_or_module void
trace_initcall_start_cb(void * data,initcall_t fn)877 trace_initcall_start_cb(void *data, initcall_t fn)
878 {
879 ktime_t *calltime = (ktime_t *)data;
880
881 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
882 *calltime = ktime_get();
883 }
884
885 static __init_or_module void
trace_initcall_finish_cb(void * data,initcall_t fn,int ret)886 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
887 {
888 ktime_t *calltime = (ktime_t *)data;
889 ktime_t delta, rettime;
890 unsigned long long duration;
891
892 rettime = ktime_get();
893 delta = ktime_sub(rettime, *calltime);
894 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
895 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
896 fn, ret, duration);
897 }
898
899 static ktime_t initcall_calltime;
900
901 #ifdef TRACEPOINTS_ENABLED
initcall_debug_enable(void)902 static void __init initcall_debug_enable(void)
903 {
904 int ret;
905
906 ret = register_trace_initcall_start(trace_initcall_start_cb,
907 &initcall_calltime);
908 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
909 &initcall_calltime);
910 WARN(ret, "Failed to register initcall tracepoints\n");
911 }
912 # define do_trace_initcall_start trace_initcall_start
913 # define do_trace_initcall_finish trace_initcall_finish
914 #else
do_trace_initcall_start(initcall_t fn)915 static inline void do_trace_initcall_start(initcall_t fn)
916 {
917 if (!initcall_debug)
918 return;
919 trace_initcall_start_cb(&initcall_calltime, fn);
920 }
do_trace_initcall_finish(initcall_t fn,int ret)921 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
922 {
923 if (!initcall_debug)
924 return;
925 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
926 }
927 #endif /* !TRACEPOINTS_ENABLED */
928
do_one_initcall(initcall_t fn)929 int __init_or_module do_one_initcall(initcall_t fn)
930 {
931 int count = preempt_count();
932 char msgbuf[64];
933 int ret;
934
935 if (initcall_blacklisted(fn))
936 return -EPERM;
937
938 do_trace_initcall_start(fn);
939 ret = fn();
940 do_trace_initcall_finish(fn, ret);
941
942 msgbuf[0] = 0;
943
944 if (preempt_count() != count) {
945 sprintf(msgbuf, "preemption imbalance ");
946 preempt_count_set(count);
947 }
948 if (irqs_disabled()) {
949 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
950 local_irq_enable();
951 }
952 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
953
954 add_latent_entropy();
955 return ret;
956 }
957
958
959 extern initcall_entry_t __initcall_start[];
960 extern initcall_entry_t __initcall0_start[];
961 extern initcall_entry_t __initcall1_start[];
962 extern initcall_entry_t __initcall2_start[];
963 extern initcall_entry_t __initcall3_start[];
964 extern initcall_entry_t __initcall4_start[];
965 extern initcall_entry_t __initcall5_start[];
966 extern initcall_entry_t __initcall6_start[];
967 extern initcall_entry_t __initcall7_start[];
968 extern initcall_entry_t __initcall_end[];
969
970 static initcall_entry_t *initcall_levels[] __initdata = {
971 __initcall0_start,
972 __initcall1_start,
973 __initcall2_start,
974 __initcall3_start,
975 __initcall4_start,
976 __initcall5_start,
977 __initcall6_start,
978 __initcall7_start,
979 __initcall_end,
980 };
981
982 /* Keep these in sync with initcalls in include/linux/init.h */
983 static const char *initcall_level_names[] __initdata = {
984 "pure",
985 "core",
986 "postcore",
987 "arch",
988 "subsys",
989 "fs",
990 "device",
991 "late",
992 };
993
do_initcall_level(int level)994 static void __init do_initcall_level(int level)
995 {
996 initcall_entry_t *fn;
997
998 strcpy(initcall_command_line, saved_command_line);
999 parse_args(initcall_level_names[level],
1000 initcall_command_line, __start___param,
1001 __stop___param - __start___param,
1002 level, level,
1003 NULL, &repair_env_string);
1004
1005 trace_initcall_level(initcall_level_names[level]);
1006 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1007 do_one_initcall(initcall_from_entry(fn));
1008 }
1009
do_initcalls(void)1010 static void __init do_initcalls(void)
1011 {
1012 int level;
1013
1014 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
1015 do_initcall_level(level);
1016 }
1017
1018 /*
1019 * Ok, the machine is now initialized. None of the devices
1020 * have been touched yet, but the CPU subsystem is up and
1021 * running, and memory and process management works.
1022 *
1023 * Now we can finally start doing some real work..
1024 */
do_basic_setup(void)1025 static void __init do_basic_setup(void)
1026 {
1027 cpuset_init_smp();
1028 driver_init();
1029 init_irq_proc();
1030 do_ctors();
1031 usermodehelper_enable();
1032 do_initcalls();
1033 }
1034
do_pre_smp_initcalls(void)1035 static void __init do_pre_smp_initcalls(void)
1036 {
1037 initcall_entry_t *fn;
1038
1039 trace_initcall_level("early");
1040 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1041 do_one_initcall(initcall_from_entry(fn));
1042 }
1043
run_init_process(const char * init_filename)1044 static int run_init_process(const char *init_filename)
1045 {
1046 argv_init[0] = init_filename;
1047 pr_info("Run %s as init process\n", init_filename);
1048 return do_execve(getname_kernel(init_filename),
1049 (const char __user *const __user *)argv_init,
1050 (const char __user *const __user *)envp_init);
1051 }
1052
try_to_run_init_process(const char * init_filename)1053 static int try_to_run_init_process(const char *init_filename)
1054 {
1055 int ret;
1056
1057 ret = run_init_process(init_filename);
1058
1059 if (ret && ret != -ENOENT) {
1060 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1061 init_filename, ret);
1062 }
1063
1064 return ret;
1065 }
1066
1067 static noinline void __init kernel_init_freeable(void);
1068
1069 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1070 bool rodata_enabled __ro_after_init = true;
set_debug_rodata(char * str)1071 static int __init set_debug_rodata(char *str)
1072 {
1073 return strtobool(str, &rodata_enabled);
1074 }
1075 __setup("rodata=", set_debug_rodata);
1076 #endif
1077
1078 #ifdef CONFIG_STRICT_KERNEL_RWX
mark_readonly(void)1079 static void mark_readonly(void)
1080 {
1081 if (rodata_enabled) {
1082 /*
1083 * load_module() results in W+X mappings, which are cleaned
1084 * up with call_rcu(). Let's make sure that queued work is
1085 * flushed so that we don't hit false positives looking for
1086 * insecure pages which are W+X.
1087 */
1088 rcu_barrier();
1089 mark_rodata_ro();
1090 rodata_test();
1091 } else
1092 pr_info("Kernel memory protection disabled.\n");
1093 }
1094 #else
mark_readonly(void)1095 static inline void mark_readonly(void)
1096 {
1097 pr_warn("This architecture does not have kernel memory protection.\n");
1098 }
1099 #endif
1100
free_initmem(void)1101 void __weak free_initmem(void)
1102 {
1103 free_initmem_default(POISON_FREE_INITMEM);
1104 }
1105
kernel_init(void * unused)1106 static int __ref kernel_init(void *unused)
1107 {
1108 int ret;
1109
1110 kernel_init_freeable();
1111 /* need to finish all async __init code before freeing the memory */
1112 async_synchronize_full();
1113 ftrace_free_init_mem();
1114 free_initmem();
1115 mark_readonly();
1116
1117 /*
1118 * Kernel mappings are now finalized - update the userspace page-table
1119 * to finalize PTI.
1120 */
1121 pti_finalize();
1122
1123 system_state = SYSTEM_RUNNING;
1124 numa_default_policy();
1125
1126 rcu_end_inkernel_boot();
1127
1128 if (ramdisk_execute_command) {
1129 ret = run_init_process(ramdisk_execute_command);
1130 if (!ret)
1131 return 0;
1132 pr_err("Failed to execute %s (error %d)\n",
1133 ramdisk_execute_command, ret);
1134 }
1135
1136 /*
1137 * We try each of these until one succeeds.
1138 *
1139 * The Bourne shell can be used instead of init if we are
1140 * trying to recover a really broken machine.
1141 */
1142 if (execute_command) {
1143 ret = run_init_process(execute_command);
1144 if (!ret)
1145 return 0;
1146 panic("Requested init %s failed (error %d).",
1147 execute_command, ret);
1148 }
1149 if (!try_to_run_init_process("/sbin/init") ||
1150 !try_to_run_init_process("/etc/init") ||
1151 !try_to_run_init_process("/bin/init") ||
1152 !try_to_run_init_process("/bin/sh"))
1153 return 0;
1154
1155 panic("No working init found. Try passing init= option to kernel. "
1156 "See Linux Documentation/admin-guide/init.rst for guidance.");
1157 }
1158
kernel_init_freeable(void)1159 static noinline void __init kernel_init_freeable(void)
1160 {
1161 /*
1162 * Wait until kthreadd is all set-up.
1163 */
1164 wait_for_completion(&kthreadd_done);
1165
1166 /* Now the scheduler is fully set up and can do blocking allocations */
1167 gfp_allowed_mask = __GFP_BITS_MASK;
1168
1169 /*
1170 * init can allocate pages on any node
1171 */
1172 set_mems_allowed(node_states[N_MEMORY]);
1173
1174 cad_pid = task_pid(current);
1175
1176 smp_prepare_cpus(setup_max_cpus);
1177
1178 workqueue_init();
1179
1180 init_mm_internals();
1181
1182 do_pre_smp_initcalls();
1183 lockup_detector_init();
1184
1185 smp_init();
1186 sched_init_smp();
1187
1188 page_alloc_init_late();
1189 /* Initialize page ext after all struct pages are initialized. */
1190 page_ext_init();
1191
1192 do_basic_setup();
1193
1194 /* Open the /dev/console on the rootfs, this should never fail */
1195 if (ksys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1196 pr_err("Warning: unable to open an initial console.\n");
1197
1198 (void) ksys_dup(0);
1199 (void) ksys_dup(0);
1200 /*
1201 * check if there is an early userspace init. If yes, let it do all
1202 * the work
1203 */
1204
1205 if (!ramdisk_execute_command)
1206 ramdisk_execute_command = "/init";
1207
1208 if (ksys_access((const char __user *)
1209 ramdisk_execute_command, 0) != 0) {
1210 ramdisk_execute_command = NULL;
1211 prepare_namespace();
1212 }
1213
1214 /*
1215 * Ok, we have completed the initial bootup, and
1216 * we're essentially up and running. Get rid of the
1217 * initmem segments and start the user-mode stuff..
1218 *
1219 * rootfs is available now, try loading the public keys
1220 * and default modules
1221 */
1222
1223 integrity_load_keys();
1224 }
1225