1 // SPDX-License-Identifier: GPL-2.0-only
5 * Copyright (C) 1991, 1992 Linus Torvalds
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>
13 #define DEBUG /* Enable initcall_debug */
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/bootconfig.h>
32 #include <linux/console.h>
33 #include <linux/nmi.h>
34 #include <linux/percpu.h>
35 #include <linux/kmod.h>
36 #include <linux/vmalloc.h>
37 #include <linux/kernel_stat.h>
38 #include <linux/start_kernel.h>
39 #include <linux/security.h>
40 #include <linux/smp.h>
41 #include <linux/profile.h>
42 #include <linux/rcupdate.h>
43 #include <linux/moduleparam.h>
44 #include <linux/kallsyms.h>
45 #include <linux/writeback.h>
46 #include <linux/cpu.h>
47 #include <linux/cpuset.h>
48 #include <linux/cgroup.h>
49 #include <linux/efi.h>
50 #include <linux/tick.h>
51 #include <linux/sched/isolation.h>
52 #include <linux/interrupt.h>
53 #include <linux/taskstats_kern.h>
54 #include <linux/delayacct.h>
55 #include <linux/unistd.h>
56 #include <linux/utsname.h>
57 #include <linux/rmap.h>
58 #include <linux/mempolicy.h>
59 #include <linux/key.h>
60 #include <linux/buffer_head.h>
61 #include <linux/page_ext.h>
62 #include <linux/debug_locks.h>
63 #include <linux/debugobjects.h>
64 #include <linux/lockdep.h>
65 #include <linux/kmemleak.h>
66 #include <linux/pid_namespace.h>
67 #include <linux/device/driver.h>
68 #include <linux/kthread.h>
69 #include <linux/sched.h>
70 #include <linux/sched/init.h>
71 #include <linux/signal.h>
72 #include <linux/idr.h>
73 #include <linux/kgdb.h>
74 #include <linux/ftrace.h>
75 #include <linux/async.h>
76 #include <linux/sfi.h>
77 #include <linux/shmem_fs.h>
78 #include <linux/slab.h>
79 #include <linux/perf_event.h>
80 #include <linux/ptrace.h>
81 #include <linux/pti.h>
82 #include <linux/blkdev.h>
83 #include <linux/elevator.h>
84 #include <linux/sched/clock.h>
85 #include <linux/sched/task.h>
86 #include <linux/sched/task_stack.h>
87 #include <linux/context_tracking.h>
88 #include <linux/random.h>
89 #include <linux/list.h>
90 #include <linux/integrity.h>
91 #include <linux/proc_ns.h>
93 #include <linux/cache.h>
94 #include <linux/rodata_test.h>
95 #include <linux/jump_label.h>
96 #include <linux/mem_encrypt.h>
100 #include <asm/setup.h>
101 #include <asm/sections.h>
102 #include <asm/cacheflush.h>
104 #define CREATE_TRACE_POINTS
105 #include <trace/events/initcall.h>
107 static int kernel_init(void *);
109 extern void init_IRQ(void);
110 extern void radix_tree_init(void);
113 * Debug helper: via this flag we know that we are in 'early bootup code'
114 * where only the boot processor is running with IRQ disabled. This means
115 * two things - IRQ must not be enabled before the flag is cleared and some
116 * operations which are not allowed with IRQ disabled are allowed while the
119 bool early_boot_irqs_disabled __read_mostly;
121 enum system_states system_state __read_mostly;
122 EXPORT_SYMBOL(system_state);
125 * Boot command-line arguments
127 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
128 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
130 extern void time_init(void);
131 /* Default late time init is NULL. archs can override this later. */
132 void (*__initdata late_time_init)(void);
134 /* Untouched command line saved by arch-specific code. */
135 char __initdata boot_command_line[COMMAND_LINE_SIZE];
136 /* Untouched saved command line (eg. for /proc) */
137 char *saved_command_line;
138 /* Command line for parameter parsing */
139 static char *static_command_line;
140 /* Untouched extra command line */
141 static char *extra_command_line;
142 /* Extra init arguments */
143 static char *extra_init_args;
145 #ifdef CONFIG_BOOT_CONFIG
146 /* Is bootconfig on command line? */
147 static bool bootconfig_found;
148 static bool initargs_found;
150 # define bootconfig_found false
151 # define initargs_found false
154 static char *execute_command;
155 static char *ramdisk_execute_command;
158 * Used to generate warnings if static_key manipulation functions are used
159 * before jump_label_init is called.
161 bool static_key_initialized __read_mostly;
162 EXPORT_SYMBOL_GPL(static_key_initialized);
165 * If set, this is an indication to the drivers that reset the underlying
166 * device before going ahead with the initialization otherwise driver might
167 * rely on the BIOS and skip the reset operation.
169 * This is useful if kernel is booting in an unreliable environment.
170 * For ex. kdump situation where previous kernel has crashed, BIOS has been
171 * skipped and devices will be in unknown state.
173 unsigned int reset_devices;
174 EXPORT_SYMBOL(reset_devices);
176 static int __init set_reset_devices(char *str)
182 __setup("reset_devices", set_reset_devices);
184 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
185 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
186 static const char *panic_later, *panic_param;
188 extern const struct obs_kernel_param __setup_start[], __setup_end[];
190 static bool __init obsolete_checksetup(char *line)
192 const struct obs_kernel_param *p;
193 bool had_early_param = false;
197 int n = strlen(p->str);
198 if (parameqn(line, p->str, n)) {
200 /* Already done in parse_early_param?
201 * (Needs exact match on param part).
202 * Keep iterating, as we can have early
203 * params and __setups of same names 8( */
204 if (line[n] == '\0' || line[n] == '=')
205 had_early_param = true;
206 } else if (!p->setup_func) {
207 pr_warn("Parameter %s is obsolete, ignored\n",
210 } else if (p->setup_func(line + n))
214 } while (p < __setup_end);
216 return had_early_param;
220 * This should be approx 2 Bo*oMips to start (note initial shift), and will
221 * still work even if initially too large, it will just take slightly longer
223 unsigned long loops_per_jiffy = (1<<12);
224 EXPORT_SYMBOL(loops_per_jiffy);
226 static int __init debug_kernel(char *str)
228 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
232 static int __init quiet_kernel(char *str)
234 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
238 early_param("debug", debug_kernel);
239 early_param("quiet", quiet_kernel);
241 static int __init loglevel(char *str)
246 * Only update loglevel value when a correct setting was passed,
247 * to prevent blind crashes (when loglevel being set to 0) that
248 * are quite hard to debug
250 if (get_option(&str, &newlevel)) {
251 console_loglevel = newlevel;
258 early_param("loglevel", loglevel);
260 #ifdef CONFIG_BOOT_CONFIG
262 char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
264 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
266 static int __init xbc_snprint_cmdline(char *buf, size_t size,
267 struct xbc_node *root)
269 struct xbc_node *knode, *vnode;
270 char *end = buf + size;
274 xbc_node_for_each_key_value(root, knode, val) {
275 ret = xbc_node_compose_key_after(root, knode,
276 xbc_namebuf, XBC_KEYLEN_MAX);
280 vnode = xbc_node_get_child(knode);
282 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
288 xbc_array_for_each_value(vnode, val) {
289 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
297 return buf - (end - size);
301 /* Make an extra command line under given key word */
302 static char * __init xbc_make_cmdline(const char *key)
304 struct xbc_node *root;
308 root = xbc_find_node(key);
312 /* Count required buffer size */
313 len = xbc_snprint_cmdline(NULL, 0, root);
317 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
319 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
323 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
324 if (ret < 0 || ret > len) {
325 pr_err("Failed to print extra kernel cmdline.\n");
332 static u32 boot_config_checksum(unsigned char *p, u32 size)
342 static int __init bootconfig_params(char *param, char *val,
343 const char *unused, void *arg)
345 if (strcmp(param, "bootconfig") == 0) {
346 bootconfig_found = true;
347 } else if (strcmp(param, "--") == 0) {
348 initargs_found = true;
353 static void __init setup_boot_config(const char *cmdline)
355 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
363 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
364 parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
367 if (!bootconfig_found)
373 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
374 if (memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
377 hdr = (u32 *)(data - 8);
381 if (size >= XBC_DATA_MAX) {
382 pr_err("bootconfig size %d greater than max size %d\n",
387 data = ((void *)hdr) - size;
388 if ((unsigned long)data < initrd_start)
391 if (boot_config_checksum((unsigned char *)data, size) != csum) {
392 pr_err("bootconfig checksum failed\n");
396 copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
398 pr_err("Failed to allocate memory for bootconfig\n");
402 memcpy(copy, data, size);
405 ret = xbc_init(copy, &msg, &pos);
408 pr_err("Failed to init bootconfig: %s.\n", msg);
410 pr_err("Failed to parse bootconfig: %s at %d.\n",
413 pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
414 /* keys starting with "kernel." are passed via cmdline */
415 extra_command_line = xbc_make_cmdline("kernel");
416 /* Also, "init." keys are init arguments */
417 extra_init_args = xbc_make_cmdline("init");
421 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
424 #define setup_boot_config(cmdline) do { } while (0)
426 static int __init warn_bootconfig(char *str)
428 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOTCONFIG is not set.\n");
431 early_param("bootconfig", warn_bootconfig);
435 /* Change NUL term back to "=", to make "param" the whole string. */
436 static void __init repair_env_string(char *param, char *val)
439 /* param=val or param="val"? */
440 if (val == param+strlen(param)+1)
442 else if (val == param+strlen(param)+2) {
444 memmove(val-1, val, strlen(val)+1);
450 /* Anything after -- gets handed straight to init. */
451 static int __init set_init_arg(char *param, char *val,
452 const char *unused, void *arg)
459 repair_env_string(param, val);
461 for (i = 0; argv_init[i]; i++) {
462 if (i == MAX_INIT_ARGS) {
463 panic_later = "init";
468 argv_init[i] = param;
473 * Unknown boot options get handed to init, unless they look like
474 * unused parameters (modprobe will find them in /proc/cmdline).
476 static int __init unknown_bootoption(char *param, char *val,
477 const char *unused, void *arg)
479 size_t len = strlen(param);
481 repair_env_string(param, val);
483 /* Handle obsolete-style parameters */
484 if (obsolete_checksetup(param))
487 /* Unused module parameter. */
488 if (strnchr(param, len, '.'))
495 /* Environment option */
497 for (i = 0; envp_init[i]; i++) {
498 if (i == MAX_INIT_ENVS) {
502 if (!strncmp(param, envp_init[i], len+1))
505 envp_init[i] = param;
507 /* Command line option */
509 for (i = 0; argv_init[i]; i++) {
510 if (i == MAX_INIT_ARGS) {
511 panic_later = "init";
515 argv_init[i] = param;
520 static int __init init_setup(char *str)
524 execute_command = str;
526 * In case LILO is going to boot us with default command line,
527 * it prepends "auto" before the whole cmdline which makes
528 * the shell think it should execute a script with such name.
529 * So we ignore all arguments entered _before_ init=... [MJ]
531 for (i = 1; i < MAX_INIT_ARGS; i++)
535 __setup("init=", init_setup);
537 static int __init rdinit_setup(char *str)
541 ramdisk_execute_command = str;
542 /* See "auto" comment in init_setup */
543 for (i = 1; i < MAX_INIT_ARGS; i++)
547 __setup("rdinit=", rdinit_setup);
550 static const unsigned int setup_max_cpus = NR_CPUS;
551 static inline void setup_nr_cpu_ids(void) { }
552 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
556 * We need to store the untouched command line for future reference.
557 * We also need to store the touched command line since the parameter
558 * parsing is performed in place, and we should allow a component to
559 * store reference of name/value for future reference.
561 static void __init setup_command_line(char *command_line)
563 size_t len, xlen = 0, ilen = 0;
565 if (extra_command_line)
566 xlen = strlen(extra_command_line);
568 ilen = strlen(extra_init_args) + 4; /* for " -- " */
570 len = xlen + strlen(boot_command_line) + 1;
572 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
573 if (!saved_command_line)
574 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
576 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
577 if (!static_command_line)
578 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
582 * We have to put extra_command_line before boot command
583 * lines because there could be dashes (separator of init
584 * command line) in the command lines.
586 strcpy(saved_command_line, extra_command_line);
587 strcpy(static_command_line, extra_command_line);
589 strcpy(saved_command_line + xlen, boot_command_line);
590 strcpy(static_command_line + xlen, command_line);
594 * Append supplemental init boot args to saved_command_line
595 * so that user can check what command line options passed
598 len = strlen(saved_command_line);
599 if (initargs_found) {
600 saved_command_line[len++] = ' ';
602 strcpy(saved_command_line + len, " -- ");
606 strcpy(saved_command_line + len, extra_init_args);
611 * We need to finalize in a non-__init function or else race conditions
612 * between the root thread and the init thread may cause start_kernel to
613 * be reaped by free_initmem before the root thread has proceeded to
616 * gcc-3.4 accidentally inlines this function, so use noinline.
619 static __initdata DECLARE_COMPLETION(kthreadd_done);
621 noinline void __ref rest_init(void)
623 struct task_struct *tsk;
626 rcu_scheduler_starting();
628 * We need to spawn init first so that it obtains pid 1, however
629 * the init task will end up wanting to create kthreads, which, if
630 * we schedule it before we create kthreadd, will OOPS.
632 pid = kernel_thread(kernel_init, NULL, CLONE_FS);
634 * Pin init on the boot CPU. Task migration is not properly working
635 * until sched_init_smp() has been run. It will set the allowed
636 * CPUs for init to the non isolated CPUs.
639 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
640 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
643 numa_default_policy();
644 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
646 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
650 * Enable might_sleep() and smp_processor_id() checks.
651 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
652 * kernel_thread() would trigger might_sleep() splats. With
653 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
654 * already, but it's stuck on the kthreadd_done completion.
656 system_state = SYSTEM_SCHEDULING;
658 complete(&kthreadd_done);
661 * The boot idle thread must execute schedule()
662 * at least once to get things moving:
664 schedule_preempt_disabled();
665 /* Call into cpu_idle with preempt disabled */
666 cpu_startup_entry(CPUHP_ONLINE);
669 /* Check for early params. */
670 static int __init do_early_param(char *param, char *val,
671 const char *unused, void *arg)
673 const struct obs_kernel_param *p;
675 for (p = __setup_start; p < __setup_end; p++) {
676 if ((p->early && parameq(param, p->str)) ||
677 (strcmp(param, "console") == 0 &&
678 strcmp(p->str, "earlycon") == 0)
680 if (p->setup_func(val) != 0)
681 pr_warn("Malformed early option '%s'\n", param);
684 /* We accept everything at this stage. */
688 void __init parse_early_options(char *cmdline)
690 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
694 /* Arch code calls this early on, or if not, just before other parsing. */
695 void __init parse_early_param(void)
697 static int done __initdata;
698 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
703 /* All fall through to do_early_param. */
704 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
705 parse_early_options(tmp_cmdline);
709 void __init __weak arch_post_acpi_subsys_init(void) { }
711 void __init __weak smp_setup_processor_id(void)
715 # if THREAD_SIZE >= PAGE_SIZE
716 void __init __weak thread_stack_cache_init(void)
721 void __init __weak mem_encrypt_init(void) { }
723 void __init __weak poking_init(void) { }
725 void __init __weak pgtable_cache_init(void) { }
728 core_param(initcall_debug, initcall_debug, bool, 0644);
730 #ifdef TRACEPOINTS_ENABLED
731 static void __init initcall_debug_enable(void);
733 static inline void initcall_debug_enable(void)
738 /* Report memory auto-initialization states for this boot. */
739 static void __init report_meminit(void)
743 if (IS_ENABLED(CONFIG_INIT_STACK_ALL))
745 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
747 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
749 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
754 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
755 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
756 want_init_on_free() ? "on" : "off");
757 if (want_init_on_free())
758 pr_info("mem auto-init: clearing system memory may take some time...\n");
762 * Set up kernel memory allocators
764 static void __init mm_init(void)
767 * page_ext requires contiguous pages,
768 * bigger than MAX_ORDER unless SPARSEMEM.
770 page_ext_init_flatmem();
771 init_debug_pagealloc();
777 debug_objects_mem_init();
780 /* Should be run before the first non-init thread is created */
782 /* Should be run after espfix64 is set up. */
786 void __init __weak arch_call_rest_init(void)
791 asmlinkage __visible void __init start_kernel(void)
796 set_task_stack_end_magic(&init_task);
797 smp_setup_processor_id();
798 debug_objects_early_init();
803 early_boot_irqs_disabled = true;
806 * Interrupts are still disabled. Do necessary setups, then
811 pr_notice("%s", linux_banner);
812 early_security_init();
813 setup_arch(&command_line);
814 setup_boot_config(command_line);
815 setup_command_line(command_line);
817 setup_per_cpu_areas();
818 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
819 boot_cpu_hotplug_init();
821 build_all_zonelists(NULL);
824 pr_notice("Kernel command line: %s\n", saved_command_line);
825 /* parameters may set static keys */
828 after_dashes = parse_args("Booting kernel",
829 static_command_line, __start___param,
830 __stop___param - __start___param,
831 -1, -1, NULL, &unknown_bootoption);
832 if (!IS_ERR_OR_NULL(after_dashes))
833 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
836 parse_args("Setting extra init args", extra_init_args,
837 NULL, 0, -1, -1, NULL, set_init_arg);
840 * These use large bootmem allocations and must precede
844 vfs_caches_init_early();
851 /* trace_printk can be enabled here */
855 * Set up the scheduler prior starting any interrupts (such as the
856 * timer interrupt). Full topology setup happens at smp_init()
857 * time - but meanwhile we still have a functioning scheduler.
861 * Disable preemption - early bootup scheduling is extremely
862 * fragile until we cpu_idle() for the first time.
865 if (WARN(!irqs_disabled(),
866 "Interrupts were enabled *very* early, fixing it\n"))
871 * Set up housekeeping before setting up workqueues to allow the unbound
872 * workqueue to take non-housekeeping into account.
877 * Allow workqueue creation and work item queueing/cancelling
878 * early. Work item execution depends on kthreads and starts after
881 workqueue_init_early();
885 /* Trace events are available after this */
889 initcall_debug_enable();
891 context_tracking_init();
892 /* init some links before init_ISA_irqs() */
903 * For best initial stack canary entropy, prepare it after:
904 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
905 * - timekeeping_init() for ktime entropy used in rand_initialize()
906 * - rand_initialize() to get any arch-specific entropy like RDRAND
907 * - add_latent_entropy() to get any latent entropy
908 * - adding command line entropy
911 add_latent_entropy();
912 add_device_randomness(command_line, strlen(command_line));
913 boot_init_stack_canary();
918 call_function_init();
919 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
921 early_boot_irqs_disabled = false;
924 kmem_cache_init_late();
927 * HACK ALERT! This is early. We're enabling the console before
928 * we've done PCI setups etc, and console_init() must be aware of
929 * this. But we do want output early, in case something goes wrong.
933 panic("Too many boot %s vars at `%s'", panic_later,
939 * Need to run this when irqs are enabled, because it wants
940 * to self-test [hard/soft]-irqs on/off lock inversion bugs
946 * This needs to be called before any devices perform DMA
947 * operations that might use the SWIOTLB bounce buffers. It will
948 * mark the bounce buffers as decrypted so that their usage will
949 * not cause "plain-text" data to be decrypted when accessed.
953 #ifdef CONFIG_BLK_DEV_INITRD
954 if (initrd_start && !initrd_below_start_ok &&
955 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
956 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
957 page_to_pfn(virt_to_page((void *)initrd_start)),
962 setup_per_cpu_pageset();
972 if (efi_enabled(EFI_RUNTIME_SERVICES))
973 efi_enter_virtual_mode();
975 thread_stack_cache_init();
992 taskstats_init_early();
998 acpi_subsystem_init();
999 arch_post_acpi_subsys_init();
1002 /* Do the rest non-__init'ed, we're now alive */
1003 arch_call_rest_init();
1006 /* Call all constructor functions linked into the kernel. */
1007 static void __init do_ctors(void)
1009 #ifdef CONFIG_CONSTRUCTORS
1010 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1012 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1017 #ifdef CONFIG_KALLSYMS
1018 struct blacklist_entry {
1019 struct list_head next;
1023 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1025 static int __init initcall_blacklist(char *str)
1028 struct blacklist_entry *entry;
1030 /* str argument is a comma-separated list of functions */
1032 str_entry = strsep(&str, ",");
1034 pr_debug("blacklisting initcall %s\n", str_entry);
1035 entry = memblock_alloc(sizeof(*entry),
1038 panic("%s: Failed to allocate %zu bytes\n",
1039 __func__, sizeof(*entry));
1040 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1043 panic("%s: Failed to allocate %zu bytes\n",
1044 __func__, strlen(str_entry) + 1);
1045 strcpy(entry->buf, str_entry);
1046 list_add(&entry->next, &blacklisted_initcalls);
1048 } while (str_entry);
1053 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1055 struct blacklist_entry *entry;
1056 char fn_name[KSYM_SYMBOL_LEN];
1059 if (list_empty(&blacklisted_initcalls))
1062 addr = (unsigned long) dereference_function_descriptor(fn);
1063 sprint_symbol_no_offset(fn_name, addr);
1066 * fn will be "function_name [module_name]" where [module_name] is not
1067 * displayed for built-in init functions. Strip off the [module_name].
1069 strreplace(fn_name, ' ', '\0');
1071 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1072 if (!strcmp(fn_name, entry->buf)) {
1073 pr_debug("initcall %s blacklisted\n", fn_name);
1081 static int __init initcall_blacklist(char *str)
1083 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1087 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1092 __setup("initcall_blacklist=", initcall_blacklist);
1094 static __init_or_module void
1095 trace_initcall_start_cb(void *data, initcall_t fn)
1097 ktime_t *calltime = (ktime_t *)data;
1099 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1100 *calltime = ktime_get();
1103 static __init_or_module void
1104 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1106 ktime_t *calltime = (ktime_t *)data;
1107 ktime_t delta, rettime;
1108 unsigned long long duration;
1110 rettime = ktime_get();
1111 delta = ktime_sub(rettime, *calltime);
1112 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1113 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1117 static ktime_t initcall_calltime;
1119 #ifdef TRACEPOINTS_ENABLED
1120 static void __init initcall_debug_enable(void)
1124 ret = register_trace_initcall_start(trace_initcall_start_cb,
1125 &initcall_calltime);
1126 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1127 &initcall_calltime);
1128 WARN(ret, "Failed to register initcall tracepoints\n");
1130 # define do_trace_initcall_start trace_initcall_start
1131 # define do_trace_initcall_finish trace_initcall_finish
1133 static inline void do_trace_initcall_start(initcall_t fn)
1135 if (!initcall_debug)
1137 trace_initcall_start_cb(&initcall_calltime, fn);
1139 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1141 if (!initcall_debug)
1143 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1145 #endif /* !TRACEPOINTS_ENABLED */
1147 int __init_or_module do_one_initcall(initcall_t fn)
1149 int count = preempt_count();
1153 if (initcall_blacklisted(fn))
1156 do_trace_initcall_start(fn);
1158 do_trace_initcall_finish(fn, ret);
1162 if (preempt_count() != count) {
1163 sprintf(msgbuf, "preemption imbalance ");
1164 preempt_count_set(count);
1166 if (irqs_disabled()) {
1167 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1170 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1172 add_latent_entropy();
1177 extern initcall_entry_t __initcall_start[];
1178 extern initcall_entry_t __initcall0_start[];
1179 extern initcall_entry_t __initcall1_start[];
1180 extern initcall_entry_t __initcall2_start[];
1181 extern initcall_entry_t __initcall3_start[];
1182 extern initcall_entry_t __initcall4_start[];
1183 extern initcall_entry_t __initcall5_start[];
1184 extern initcall_entry_t __initcall6_start[];
1185 extern initcall_entry_t __initcall7_start[];
1186 extern initcall_entry_t __initcall_end[];
1188 static initcall_entry_t *initcall_levels[] __initdata = {
1200 /* Keep these in sync with initcalls in include/linux/init.h */
1201 static const char *initcall_level_names[] __initdata = {
1212 static int __init ignore_unknown_bootoption(char *param, char *val,
1213 const char *unused, void *arg)
1218 static void __init do_initcall_level(int level, char *command_line)
1220 initcall_entry_t *fn;
1222 parse_args(initcall_level_names[level],
1223 command_line, __start___param,
1224 __stop___param - __start___param,
1226 NULL, ignore_unknown_bootoption);
1228 trace_initcall_level(initcall_level_names[level]);
1229 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1230 do_one_initcall(initcall_from_entry(fn));
1233 static void __init do_initcalls(void)
1236 size_t len = strlen(saved_command_line) + 1;
1239 command_line = kzalloc(len, GFP_KERNEL);
1241 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1243 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1244 /* Parser modifies command_line, restore it each time */
1245 strcpy(command_line, saved_command_line);
1246 do_initcall_level(level, command_line);
1249 kfree(command_line);
1253 * Ok, the machine is now initialized. None of the devices
1254 * have been touched yet, but the CPU subsystem is up and
1255 * running, and memory and process management works.
1257 * Now we can finally start doing some real work..
1259 static void __init do_basic_setup(void)
1265 usermodehelper_enable();
1269 static void __init do_pre_smp_initcalls(void)
1271 initcall_entry_t *fn;
1273 trace_initcall_level("early");
1274 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1275 do_one_initcall(initcall_from_entry(fn));
1278 static int run_init_process(const char *init_filename)
1280 const char *const *p;
1282 argv_init[0] = init_filename;
1283 pr_info("Run %s as init process\n", init_filename);
1284 pr_debug(" with arguments:\n");
1285 for (p = argv_init; *p; p++)
1286 pr_debug(" %s\n", *p);
1287 pr_debug(" with environment:\n");
1288 for (p = envp_init; *p; p++)
1289 pr_debug(" %s\n", *p);
1290 return do_execve(getname_kernel(init_filename),
1291 (const char __user *const __user *)argv_init,
1292 (const char __user *const __user *)envp_init);
1295 static int try_to_run_init_process(const char *init_filename)
1299 ret = run_init_process(init_filename);
1301 if (ret && ret != -ENOENT) {
1302 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1303 init_filename, ret);
1309 static noinline void __init kernel_init_freeable(void);
1311 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1312 bool rodata_enabled __ro_after_init = true;
1313 static int __init set_debug_rodata(char *str)
1315 return strtobool(str, &rodata_enabled);
1317 __setup("rodata=", set_debug_rodata);
1320 #ifdef CONFIG_STRICT_KERNEL_RWX
1321 static void mark_readonly(void)
1323 if (rodata_enabled) {
1325 * load_module() results in W+X mappings, which are cleaned
1326 * up with call_rcu(). Let's make sure that queued work is
1327 * flushed so that we don't hit false positives looking for
1328 * insecure pages which are W+X.
1334 pr_info("Kernel memory protection disabled.\n");
1336 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1337 static inline void mark_readonly(void)
1339 pr_warn("Kernel memory protection not selected by kernel config.\n");
1342 static inline void mark_readonly(void)
1344 pr_warn("This architecture does not have kernel memory protection.\n");
1348 void __weak free_initmem(void)
1350 free_initmem_default(POISON_FREE_INITMEM);
1353 static int __ref kernel_init(void *unused)
1357 kernel_init_freeable();
1358 /* need to finish all async __init code before freeing the memory */
1359 async_synchronize_full();
1360 ftrace_free_init_mem();
1365 * Kernel mappings are now finalized - update the userspace page-table
1370 system_state = SYSTEM_RUNNING;
1371 numa_default_policy();
1373 rcu_end_inkernel_boot();
1375 if (ramdisk_execute_command) {
1376 ret = run_init_process(ramdisk_execute_command);
1379 pr_err("Failed to execute %s (error %d)\n",
1380 ramdisk_execute_command, ret);
1384 * We try each of these until one succeeds.
1386 * The Bourne shell can be used instead of init if we are
1387 * trying to recover a really broken machine.
1389 if (execute_command) {
1390 ret = run_init_process(execute_command);
1393 panic("Requested init %s failed (error %d).",
1394 execute_command, ret);
1396 if (!try_to_run_init_process("/sbin/init") ||
1397 !try_to_run_init_process("/etc/init") ||
1398 !try_to_run_init_process("/bin/init") ||
1399 !try_to_run_init_process("/bin/sh"))
1402 panic("No working init found. Try passing init= option to kernel. "
1403 "See Linux Documentation/admin-guide/init.rst for guidance.");
1406 void console_on_rootfs(void)
1408 /* Open the /dev/console as stdin, this should never fail */
1409 if (ksys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1410 pr_err("Warning: unable to open an initial console.\n");
1412 /* create stdout/stderr */
1417 static noinline void __init kernel_init_freeable(void)
1420 * Wait until kthreadd is all set-up.
1422 wait_for_completion(&kthreadd_done);
1424 /* Now the scheduler is fully set up and can do blocking allocations */
1425 gfp_allowed_mask = __GFP_BITS_MASK;
1428 * init can allocate pages on any node
1430 set_mems_allowed(node_states[N_MEMORY]);
1432 cad_pid = task_pid(current);
1434 smp_prepare_cpus(setup_max_cpus);
1438 init_mm_internals();
1440 do_pre_smp_initcalls();
1441 lockup_detector_init();
1446 page_alloc_init_late();
1447 /* Initialize page ext after all struct pages are initialized. */
1452 console_on_rootfs();
1455 * check if there is an early userspace init. If yes, let it do all
1459 if (!ramdisk_execute_command)
1460 ramdisk_execute_command = "/init";
1462 if (ksys_access((const char __user *)
1463 ramdisk_execute_command, 0) != 0) {
1464 ramdisk_execute_command = NULL;
1465 prepare_namespace();
1469 * Ok, we have completed the initial bootup, and
1470 * we're essentially up and running. Get rid of the
1471 * initmem segments and start the user-mode stuff..
1473 * rootfs is available now, try loading the public keys
1474 * and default modules
1477 integrity_load_keys();