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 fork_init(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 saved_command_line =
378 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
379 initcall_command_line =
380 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
381 static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0);
382 strcpy(saved_command_line, boot_command_line);
383 strcpy(static_command_line, command_line);
384 }
385
386 /*
387 * We need to finalize in a non-__init function or else race conditions
388 * between the root thread and the init thread may cause start_kernel to
389 * be reaped by free_initmem before the root thread has proceeded to
390 * cpu_idle.
391 *
392 * gcc-3.4 accidentally inlines this function, so use noinline.
393 */
394
395 static __initdata DECLARE_COMPLETION(kthreadd_done);
396
rest_init(void)397 static noinline void __ref rest_init(void)
398 {
399 struct task_struct *tsk;
400 int pid;
401
402 rcu_scheduler_starting();
403 /*
404 * We need to spawn init first so that it obtains pid 1, however
405 * the init task will end up wanting to create kthreads, which, if
406 * we schedule it before we create kthreadd, will OOPS.
407 */
408 pid = kernel_thread(kernel_init, NULL, CLONE_FS);
409 /*
410 * Pin init on the boot CPU. Task migration is not properly working
411 * until sched_init_smp() has been run. It will set the allowed
412 * CPUs for init to the non isolated CPUs.
413 */
414 rcu_read_lock();
415 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
416 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
417 rcu_read_unlock();
418
419 numa_default_policy();
420 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
421 rcu_read_lock();
422 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
423 rcu_read_unlock();
424
425 /*
426 * Enable might_sleep() and smp_processor_id() checks.
427 * They cannot be enabled earlier because with CONFIG_PREEMPT=y
428 * kernel_thread() would trigger might_sleep() splats. With
429 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
430 * already, but it's stuck on the kthreadd_done completion.
431 */
432 system_state = SYSTEM_SCHEDULING;
433
434 complete(&kthreadd_done);
435
436 /*
437 * The boot idle thread must execute schedule()
438 * at least once to get things moving:
439 */
440 schedule_preempt_disabled();
441 /* Call into cpu_idle with preempt disabled */
442 cpu_startup_entry(CPUHP_ONLINE);
443 }
444
445 /* Check for early params. */
do_early_param(char * param,char * val,const char * unused,void * arg)446 static int __init do_early_param(char *param, char *val,
447 const char *unused, void *arg)
448 {
449 const struct obs_kernel_param *p;
450
451 for (p = __setup_start; p < __setup_end; p++) {
452 if ((p->early && parameq(param, p->str)) ||
453 (strcmp(param, "console") == 0 &&
454 strcmp(p->str, "earlycon") == 0)
455 ) {
456 if (p->setup_func(val) != 0)
457 pr_warn("Malformed early option '%s'\n", param);
458 }
459 }
460 /* We accept everything at this stage. */
461 return 0;
462 }
463
parse_early_options(char * cmdline)464 void __init parse_early_options(char *cmdline)
465 {
466 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
467 do_early_param);
468 }
469
470 /* Arch code calls this early on, or if not, just before other parsing. */
parse_early_param(void)471 void __init parse_early_param(void)
472 {
473 static int done __initdata;
474 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
475
476 if (done)
477 return;
478
479 /* All fall through to do_early_param. */
480 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
481 parse_early_options(tmp_cmdline);
482 done = 1;
483 }
484
arch_post_acpi_subsys_init(void)485 void __init __weak arch_post_acpi_subsys_init(void) { }
486
smp_setup_processor_id(void)487 void __init __weak smp_setup_processor_id(void)
488 {
489 }
490
491 # if THREAD_SIZE >= PAGE_SIZE
thread_stack_cache_init(void)492 void __init __weak thread_stack_cache_init(void)
493 {
494 }
495 #endif
496
mem_encrypt_init(void)497 void __init __weak mem_encrypt_init(void) { }
498
499 bool initcall_debug;
500 core_param(initcall_debug, initcall_debug, bool, 0644);
501
502 #ifdef TRACEPOINTS_ENABLED
503 static void __init initcall_debug_enable(void);
504 #else
initcall_debug_enable(void)505 static inline void initcall_debug_enable(void)
506 {
507 }
508 #endif
509
510 /*
511 * Set up kernel memory allocators
512 */
mm_init(void)513 static void __init mm_init(void)
514 {
515 /*
516 * page_ext requires contiguous pages,
517 * bigger than MAX_ORDER unless SPARSEMEM.
518 */
519 page_ext_init_flatmem();
520 mem_init();
521 kmem_cache_init();
522 pgtable_init();
523 vmalloc_init();
524 ioremap_huge_init();
525 /* Should be run before the first non-init thread is created */
526 init_espfix_bsp();
527 /* Should be run after espfix64 is set up. */
528 pti_init();
529 }
530
start_kernel(void)531 asmlinkage __visible void __init start_kernel(void)
532 {
533 char *command_line;
534 char *after_dashes;
535
536 set_task_stack_end_magic(&init_task);
537 smp_setup_processor_id();
538 debug_objects_early_init();
539
540 cgroup_init_early();
541
542 local_irq_disable();
543 early_boot_irqs_disabled = true;
544
545 /*
546 * Interrupts are still disabled. Do necessary setups, then
547 * enable them.
548 */
549 boot_cpu_init();
550 page_address_init();
551 pr_notice("%s", linux_banner);
552 setup_arch(&command_line);
553 /*
554 * Set up the the initial canary and entropy after arch
555 * and after adding latent and command line entropy.
556 */
557 add_latent_entropy();
558 add_device_randomness(command_line, strlen(command_line));
559 boot_init_stack_canary();
560 mm_init_cpumask(&init_mm);
561 setup_command_line(command_line);
562 setup_nr_cpu_ids();
563 setup_per_cpu_areas();
564 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
565 boot_cpu_hotplug_init();
566
567 build_all_zonelists(NULL);
568 page_alloc_init();
569
570 pr_notice("Kernel command line: %s\n", boot_command_line);
571 parse_early_param();
572 after_dashes = parse_args("Booting kernel",
573 static_command_line, __start___param,
574 __stop___param - __start___param,
575 -1, -1, NULL, &unknown_bootoption);
576 if (!IS_ERR_OR_NULL(after_dashes))
577 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
578 NULL, set_init_arg);
579
580 jump_label_init();
581
582 /*
583 * These use large bootmem allocations and must precede
584 * kmem_cache_init()
585 */
586 setup_log_buf(0);
587 vfs_caches_init_early();
588 sort_main_extable();
589 trap_init();
590 mm_init();
591
592 ftrace_init();
593
594 /* trace_printk can be enabled here */
595 early_trace_init();
596
597 /*
598 * Set up the scheduler prior starting any interrupts (such as the
599 * timer interrupt). Full topology setup happens at smp_init()
600 * time - but meanwhile we still have a functioning scheduler.
601 */
602 sched_init();
603 /*
604 * Disable preemption - early bootup scheduling is extremely
605 * fragile until we cpu_idle() for the first time.
606 */
607 preempt_disable();
608 if (WARN(!irqs_disabled(),
609 "Interrupts were enabled *very* early, fixing it\n"))
610 local_irq_disable();
611 radix_tree_init();
612
613 /*
614 * Set up housekeeping before setting up workqueues to allow the unbound
615 * workqueue to take non-housekeeping into account.
616 */
617 housekeeping_init();
618
619 /*
620 * Allow workqueue creation and work item queueing/cancelling
621 * early. Work item execution depends on kthreads and starts after
622 * workqueue_init().
623 */
624 workqueue_init_early();
625
626 rcu_init();
627
628 /* Trace events are available after this */
629 trace_init();
630
631 if (initcall_debug)
632 initcall_debug_enable();
633
634 context_tracking_init();
635 /* init some links before init_ISA_irqs() */
636 early_irq_init();
637 init_IRQ();
638 tick_init();
639 rcu_init_nohz();
640 init_timers();
641 hrtimers_init();
642 softirq_init();
643 timekeeping_init();
644 time_init();
645 printk_safe_init();
646 perf_event_init();
647 profile_init();
648 call_function_init();
649 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
650
651 early_boot_irqs_disabled = false;
652 local_irq_enable();
653
654 kmem_cache_init_late();
655
656 /*
657 * HACK ALERT! This is early. We're enabling the console before
658 * we've done PCI setups etc, and console_init() must be aware of
659 * this. But we do want output early, in case something goes wrong.
660 */
661 console_init();
662 if (panic_later)
663 panic("Too many boot %s vars at `%s'", panic_later,
664 panic_param);
665
666 lockdep_init();
667
668 /*
669 * Need to run this when irqs are enabled, because it wants
670 * to self-test [hard/soft]-irqs on/off lock inversion bugs
671 * too:
672 */
673 locking_selftest();
674
675 /*
676 * This needs to be called before any devices perform DMA
677 * operations that might use the SWIOTLB bounce buffers. It will
678 * mark the bounce buffers as decrypted so that their usage will
679 * not cause "plain-text" data to be decrypted when accessed.
680 */
681 mem_encrypt_init();
682
683 #ifdef CONFIG_BLK_DEV_INITRD
684 if (initrd_start && !initrd_below_start_ok &&
685 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
686 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
687 page_to_pfn(virt_to_page((void *)initrd_start)),
688 min_low_pfn);
689 initrd_start = 0;
690 }
691 #endif
692 page_ext_init();
693 kmemleak_init();
694 debug_objects_mem_init();
695 setup_per_cpu_pageset();
696 numa_policy_init();
697 acpi_early_init();
698 if (late_time_init)
699 late_time_init();
700 sched_clock_init();
701 calibrate_delay();
702 pid_idr_init();
703 anon_vma_init();
704 #ifdef CONFIG_X86
705 if (efi_enabled(EFI_RUNTIME_SERVICES))
706 efi_enter_virtual_mode();
707 #endif
708 thread_stack_cache_init();
709 cred_init();
710 fork_init();
711 proc_caches_init();
712 uts_ns_init();
713 buffer_init();
714 key_init();
715 security_init();
716 dbg_late_init();
717 vfs_caches_init();
718 pagecache_init();
719 signals_init();
720 seq_file_init();
721 proc_root_init();
722 nsfs_init();
723 cpuset_init();
724 cgroup_init();
725 taskstats_init_early();
726 delayacct_init();
727
728 check_bugs();
729
730 acpi_subsystem_init();
731 arch_post_acpi_subsys_init();
732 sfi_init_late();
733
734 if (efi_enabled(EFI_RUNTIME_SERVICES)) {
735 efi_free_boot_services();
736 }
737
738 /* Do the rest non-__init'ed, we're now alive */
739 rest_init();
740 }
741
742 /* Call all constructor functions linked into the kernel. */
do_ctors(void)743 static void __init do_ctors(void)
744 {
745 #ifdef CONFIG_CONSTRUCTORS
746 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
747
748 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
749 (*fn)();
750 #endif
751 }
752
753 #ifdef CONFIG_KALLSYMS
754 struct blacklist_entry {
755 struct list_head next;
756 char *buf;
757 };
758
759 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
760
initcall_blacklist(char * str)761 static int __init initcall_blacklist(char *str)
762 {
763 char *str_entry;
764 struct blacklist_entry *entry;
765
766 /* str argument is a comma-separated list of functions */
767 do {
768 str_entry = strsep(&str, ",");
769 if (str_entry) {
770 pr_debug("blacklisting initcall %s\n", str_entry);
771 entry = alloc_bootmem(sizeof(*entry));
772 entry->buf = alloc_bootmem(strlen(str_entry) + 1);
773 strcpy(entry->buf, str_entry);
774 list_add(&entry->next, &blacklisted_initcalls);
775 }
776 } while (str_entry);
777
778 return 0;
779 }
780
initcall_blacklisted(initcall_t fn)781 static bool __init_or_module initcall_blacklisted(initcall_t fn)
782 {
783 struct blacklist_entry *entry;
784 char fn_name[KSYM_SYMBOL_LEN];
785 unsigned long addr;
786
787 if (list_empty(&blacklisted_initcalls))
788 return false;
789
790 addr = (unsigned long) dereference_function_descriptor(fn);
791 sprint_symbol_no_offset(fn_name, addr);
792
793 /*
794 * fn will be "function_name [module_name]" where [module_name] is not
795 * displayed for built-in init functions. Strip off the [module_name].
796 */
797 strreplace(fn_name, ' ', '\0');
798
799 list_for_each_entry(entry, &blacklisted_initcalls, next) {
800 if (!strcmp(fn_name, entry->buf)) {
801 pr_debug("initcall %s blacklisted\n", fn_name);
802 return true;
803 }
804 }
805
806 return false;
807 }
808 #else
initcall_blacklist(char * str)809 static int __init initcall_blacklist(char *str)
810 {
811 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
812 return 0;
813 }
814
initcall_blacklisted(initcall_t fn)815 static bool __init_or_module initcall_blacklisted(initcall_t fn)
816 {
817 return false;
818 }
819 #endif
820 __setup("initcall_blacklist=", initcall_blacklist);
821
822 static __init_or_module void
trace_initcall_start_cb(void * data,initcall_t fn)823 trace_initcall_start_cb(void *data, initcall_t fn)
824 {
825 ktime_t *calltime = (ktime_t *)data;
826
827 printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current));
828 *calltime = ktime_get();
829 }
830
831 static __init_or_module void
trace_initcall_finish_cb(void * data,initcall_t fn,int ret)832 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
833 {
834 ktime_t *calltime = (ktime_t *)data;
835 ktime_t delta, rettime;
836 unsigned long long duration;
837
838 rettime = ktime_get();
839 delta = ktime_sub(rettime, *calltime);
840 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
841 printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n",
842 fn, ret, duration);
843 }
844
845 static ktime_t initcall_calltime;
846
847 #ifdef TRACEPOINTS_ENABLED
initcall_debug_enable(void)848 static void __init initcall_debug_enable(void)
849 {
850 int ret;
851
852 ret = register_trace_initcall_start(trace_initcall_start_cb,
853 &initcall_calltime);
854 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
855 &initcall_calltime);
856 WARN(ret, "Failed to register initcall tracepoints\n");
857 }
858 # define do_trace_initcall_start trace_initcall_start
859 # define do_trace_initcall_finish trace_initcall_finish
860 #else
do_trace_initcall_start(initcall_t fn)861 static inline void do_trace_initcall_start(initcall_t fn)
862 {
863 if (!initcall_debug)
864 return;
865 trace_initcall_start_cb(&initcall_calltime, fn);
866 }
do_trace_initcall_finish(initcall_t fn,int ret)867 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
868 {
869 if (!initcall_debug)
870 return;
871 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
872 }
873 #endif /* !TRACEPOINTS_ENABLED */
874
do_one_initcall(initcall_t fn)875 int __init_or_module do_one_initcall(initcall_t fn)
876 {
877 int count = preempt_count();
878 char msgbuf[64];
879 int ret;
880
881 if (initcall_blacklisted(fn))
882 return -EPERM;
883
884 do_trace_initcall_start(fn);
885 ret = fn();
886 do_trace_initcall_finish(fn, ret);
887
888 msgbuf[0] = 0;
889
890 if (preempt_count() != count) {
891 sprintf(msgbuf, "preemption imbalance ");
892 preempt_count_set(count);
893 }
894 if (irqs_disabled()) {
895 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
896 local_irq_enable();
897 }
898 WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
899
900 add_latent_entropy();
901 return ret;
902 }
903
904
905 extern initcall_entry_t __initcall_start[];
906 extern initcall_entry_t __initcall0_start[];
907 extern initcall_entry_t __initcall1_start[];
908 extern initcall_entry_t __initcall2_start[];
909 extern initcall_entry_t __initcall3_start[];
910 extern initcall_entry_t __initcall4_start[];
911 extern initcall_entry_t __initcall5_start[];
912 extern initcall_entry_t __initcall6_start[];
913 extern initcall_entry_t __initcall7_start[];
914 extern initcall_entry_t __initcall_end[];
915
916 static initcall_entry_t *initcall_levels[] __initdata = {
917 __initcall0_start,
918 __initcall1_start,
919 __initcall2_start,
920 __initcall3_start,
921 __initcall4_start,
922 __initcall5_start,
923 __initcall6_start,
924 __initcall7_start,
925 __initcall_end,
926 };
927
928 /* Keep these in sync with initcalls in include/linux/init.h */
929 static char *initcall_level_names[] __initdata = {
930 "pure",
931 "core",
932 "postcore",
933 "arch",
934 "subsys",
935 "fs",
936 "device",
937 "late",
938 };
939
do_initcall_level(int level)940 static void __init do_initcall_level(int level)
941 {
942 initcall_entry_t *fn;
943
944 strcpy(initcall_command_line, saved_command_line);
945 parse_args(initcall_level_names[level],
946 initcall_command_line, __start___param,
947 __stop___param - __start___param,
948 level, level,
949 NULL, &repair_env_string);
950
951 trace_initcall_level(initcall_level_names[level]);
952 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
953 do_one_initcall(initcall_from_entry(fn));
954 }
955
do_initcalls(void)956 static void __init do_initcalls(void)
957 {
958 int level;
959
960 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
961 do_initcall_level(level);
962 }
963
964 /*
965 * Ok, the machine is now initialized. None of the devices
966 * have been touched yet, but the CPU subsystem is up and
967 * running, and memory and process management works.
968 *
969 * Now we can finally start doing some real work..
970 */
do_basic_setup(void)971 static void __init do_basic_setup(void)
972 {
973 cpuset_init_smp();
974 shmem_init();
975 driver_init();
976 init_irq_proc();
977 do_ctors();
978 usermodehelper_enable();
979 do_initcalls();
980 }
981
do_pre_smp_initcalls(void)982 static void __init do_pre_smp_initcalls(void)
983 {
984 initcall_entry_t *fn;
985
986 trace_initcall_level("early");
987 for (fn = __initcall_start; fn < __initcall0_start; fn++)
988 do_one_initcall(initcall_from_entry(fn));
989 }
990
991 /*
992 * This function requests modules which should be loaded by default and is
993 * called twice right after initrd is mounted and right before init is
994 * exec'd. If such modules are on either initrd or rootfs, they will be
995 * loaded before control is passed to userland.
996 */
load_default_modules(void)997 void __init load_default_modules(void)
998 {
999 load_default_elevator_module();
1000 }
1001
run_init_process(const char * init_filename)1002 static int run_init_process(const char *init_filename)
1003 {
1004 argv_init[0] = init_filename;
1005 pr_info("Run %s as init process\n", init_filename);
1006 return do_execve(getname_kernel(init_filename),
1007 (const char __user *const __user *)argv_init,
1008 (const char __user *const __user *)envp_init);
1009 }
1010
try_to_run_init_process(const char * init_filename)1011 static int try_to_run_init_process(const char *init_filename)
1012 {
1013 int ret;
1014
1015 ret = run_init_process(init_filename);
1016
1017 if (ret && ret != -ENOENT) {
1018 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1019 init_filename, ret);
1020 }
1021
1022 return ret;
1023 }
1024
1025 static noinline void __init kernel_init_freeable(void);
1026
1027 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1028 bool rodata_enabled __ro_after_init = true;
set_debug_rodata(char * str)1029 static int __init set_debug_rodata(char *str)
1030 {
1031 return strtobool(str, &rodata_enabled);
1032 }
1033 __setup("rodata=", set_debug_rodata);
1034 #endif
1035
1036 #ifdef CONFIG_STRICT_KERNEL_RWX
mark_readonly(void)1037 static void mark_readonly(void)
1038 {
1039 if (rodata_enabled) {
1040 /*
1041 * load_module() results in W+X mappings, which are cleaned up
1042 * with call_rcu_sched(). Let's make sure that queued work is
1043 * flushed so that we don't hit false positives looking for
1044 * insecure pages which are W+X.
1045 */
1046 rcu_barrier_sched();
1047 mark_rodata_ro();
1048 rodata_test();
1049 } else
1050 pr_info("Kernel memory protection disabled.\n");
1051 }
1052 #else
mark_readonly(void)1053 static inline void mark_readonly(void)
1054 {
1055 pr_warn("This architecture does not have kernel memory protection.\n");
1056 }
1057 #endif
1058
kernel_init(void * unused)1059 static int __ref kernel_init(void *unused)
1060 {
1061 int ret;
1062
1063 kernel_init_freeable();
1064 /* need to finish all async __init code before freeing the memory */
1065 async_synchronize_full();
1066 ftrace_free_init_mem();
1067 jump_label_invalidate_initmem();
1068 free_initmem();
1069 mark_readonly();
1070
1071 /*
1072 * Kernel mappings are now finalized - update the userspace page-table
1073 * to finalize PTI.
1074 */
1075 pti_finalize();
1076
1077 system_state = SYSTEM_RUNNING;
1078 numa_default_policy();
1079
1080 rcu_end_inkernel_boot();
1081
1082 if (ramdisk_execute_command) {
1083 ret = run_init_process(ramdisk_execute_command);
1084 if (!ret)
1085 return 0;
1086 pr_err("Failed to execute %s (error %d)\n",
1087 ramdisk_execute_command, ret);
1088 }
1089
1090 /*
1091 * We try each of these until one succeeds.
1092 *
1093 * The Bourne shell can be used instead of init if we are
1094 * trying to recover a really broken machine.
1095 */
1096 if (execute_command) {
1097 ret = run_init_process(execute_command);
1098 if (!ret)
1099 return 0;
1100 panic("Requested init %s failed (error %d).",
1101 execute_command, ret);
1102 }
1103 if (!try_to_run_init_process("/sbin/init") ||
1104 !try_to_run_init_process("/etc/init") ||
1105 !try_to_run_init_process("/bin/init") ||
1106 !try_to_run_init_process("/bin/sh"))
1107 return 0;
1108
1109 panic("No working init found. Try passing init= option to kernel. "
1110 "See Linux Documentation/admin-guide/init.rst for guidance.");
1111 }
1112
kernel_init_freeable(void)1113 static noinline void __init kernel_init_freeable(void)
1114 {
1115 /*
1116 * Wait until kthreadd is all set-up.
1117 */
1118 wait_for_completion(&kthreadd_done);
1119
1120 /* Now the scheduler is fully set up and can do blocking allocations */
1121 gfp_allowed_mask = __GFP_BITS_MASK;
1122
1123 /*
1124 * init can allocate pages on any node
1125 */
1126 set_mems_allowed(node_states[N_MEMORY]);
1127
1128 cad_pid = task_pid(current);
1129
1130 smp_prepare_cpus(setup_max_cpus);
1131
1132 workqueue_init();
1133
1134 init_mm_internals();
1135
1136 do_pre_smp_initcalls();
1137 lockup_detector_init();
1138
1139 smp_init();
1140 sched_init_smp();
1141
1142 page_alloc_init_late();
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 }
1178
