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/kprobes.h>
37 #include <linux/vmalloc.h>
38 #include <linux/kernel_stat.h>
39 #include <linux/start_kernel.h>
40 #include <linux/security.h>
41 #include <linux/smp.h>
42 #include <linux/profile.h>
43 #include <linux/kfence.h>
44 #include <linux/rcupdate.h>
45 #include <linux/moduleparam.h>
46 #include <linux/kallsyms.h>
47 #include <linux/writeback.h>
48 #include <linux/cpu.h>
49 #include <linux/cpuset.h>
50 #include <linux/cgroup.h>
51 #include <linux/efi.h>
52 #include <linux/tick.h>
53 #include <linux/sched/isolation.h>
54 #include <linux/interrupt.h>
55 #include <linux/taskstats_kern.h>
56 #include <linux/delayacct.h>
57 #include <linux/unistd.h>
58 #include <linux/utsname.h>
59 #include <linux/rmap.h>
60 #include <linux/mempolicy.h>
61 #include <linux/key.h>
62 #include <linux/page_ext.h>
63 #include <linux/debug_locks.h>
64 #include <linux/debugobjects.h>
65 #include <linux/lockdep.h>
66 #include <linux/kmemleak.h>
67 #include <linux/padata.h>
68 #include <linux/pid_namespace.h>
69 #include <linux/device/driver.h>
70 #include <linux/kthread.h>
71 #include <linux/sched.h>
72 #include <linux/sched/init.h>
73 #include <linux/signal.h>
74 #include <linux/idr.h>
75 #include <linux/kgdb.h>
76 #include <linux/ftrace.h>
77 #include <linux/async.h>
78 #include <linux/shmem_fs.h>
79 #include <linux/slab.h>
80 #include <linux/perf_event.h>
81 #include <linux/ptrace.h>
82 #include <linux/pti.h>
83 #include <linux/blkdev.h>
84 #include <linux/elevator.h>
85 #include <linux/sched/clock.h>
86 #include <linux/sched/task.h>
87 #include <linux/sched/task_stack.h>
88 #include <linux/context_tracking.h>
89 #include <linux/random.h>
90 #include <linux/list.h>
91 #include <linux/integrity.h>
92 #include <linux/proc_ns.h>
94 #include <linux/cache.h>
95 #include <linux/rodata_test.h>
96 #include <linux/jump_label.h>
97 #include <linux/mem_encrypt.h>
98 #include <linux/kcsan.h>
99 #include <linux/init_syscalls.h>
102 #include <asm/bugs.h>
103 #include <asm/setup.h>
104 #include <asm/sections.h>
105 #include <asm/cacheflush.h>
107 #define CREATE_TRACE_POINTS
108 #include <trace/events/initcall.h>
110 #include <kunit/test.h>
112 static int kernel_init(void *);
114 extern void init_IRQ(void);
115 extern void radix_tree_init(void);
118 * Debug helper: via this flag we know that we are in 'early bootup code'
119 * where only the boot processor is running with IRQ disabled. This means
120 * two things - IRQ must not be enabled before the flag is cleared and some
121 * operations which are not allowed with IRQ disabled are allowed while the
124 bool early_boot_irqs_disabled __read_mostly;
126 enum system_states system_state __read_mostly;
127 EXPORT_SYMBOL(system_state);
130 * Boot command-line arguments
132 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
133 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
135 extern void time_init(void);
136 /* Default late time init is NULL. archs can override this later. */
137 void (*__initdata late_time_init)(void);
139 /* Untouched command line saved by arch-specific code. */
140 char __initdata boot_command_line[COMMAND_LINE_SIZE];
141 /* Untouched saved command line (eg. for /proc) */
142 char *saved_command_line;
143 /* Command line for parameter parsing */
144 static char *static_command_line;
145 /* Untouched extra command line */
146 static char *extra_command_line;
147 /* Extra init arguments */
148 static char *extra_init_args;
150 #ifdef CONFIG_BOOT_CONFIG
151 /* Is bootconfig on command line? */
152 static bool bootconfig_found;
153 static bool initargs_found;
155 # define bootconfig_found false
156 # define initargs_found false
159 static char *execute_command;
160 static char *ramdisk_execute_command = "/init";
163 * Used to generate warnings if static_key manipulation functions are used
164 * before jump_label_init is called.
166 bool static_key_initialized __read_mostly;
167 EXPORT_SYMBOL_GPL(static_key_initialized);
170 * If set, this is an indication to the drivers that reset the underlying
171 * device before going ahead with the initialization otherwise driver might
172 * rely on the BIOS and skip the reset operation.
174 * This is useful if kernel is booting in an unreliable environment.
175 * For ex. kdump situation where previous kernel has crashed, BIOS has been
176 * skipped and devices will be in unknown state.
178 unsigned int reset_devices;
179 EXPORT_SYMBOL(reset_devices);
181 static int __init set_reset_devices(char *str)
187 __setup("reset_devices", set_reset_devices);
189 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
190 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
191 static const char *panic_later, *panic_param;
193 extern const struct obs_kernel_param __setup_start[], __setup_end[];
195 static bool __init obsolete_checksetup(char *line)
197 const struct obs_kernel_param *p;
198 bool had_early_param = false;
202 int n = strlen(p->str);
203 if (parameqn(line, p->str, n)) {
205 /* Already done in parse_early_param?
206 * (Needs exact match on param part).
207 * Keep iterating, as we can have early
208 * params and __setups of same names 8( */
209 if (line[n] == '\0' || line[n] == '=')
210 had_early_param = true;
211 } else if (!p->setup_func) {
212 pr_warn("Parameter %s is obsolete, ignored\n",
215 } else if (p->setup_func(line + n))
219 } while (p < __setup_end);
221 return had_early_param;
225 * This should be approx 2 Bo*oMips to start (note initial shift), and will
226 * still work even if initially too large, it will just take slightly longer
228 unsigned long loops_per_jiffy = (1<<12);
229 EXPORT_SYMBOL(loops_per_jiffy);
231 static int __init debug_kernel(char *str)
233 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
237 static int __init quiet_kernel(char *str)
239 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
243 early_param("debug", debug_kernel);
244 early_param("quiet", quiet_kernel);
246 static int __init loglevel(char *str)
251 * Only update loglevel value when a correct setting was passed,
252 * to prevent blind crashes (when loglevel being set to 0) that
253 * are quite hard to debug
255 if (get_option(&str, &newlevel)) {
256 console_loglevel = newlevel;
263 early_param("loglevel", loglevel);
265 #ifdef CONFIG_BLK_DEV_INITRD
266 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
276 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
278 * Since Grub may align the size of initrd to 4, we must
279 * check the preceding 3 bytes as well.
281 for (i = 0; i < 4; i++) {
282 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
289 hdr = (u32 *)(data - 8);
290 size = le32_to_cpu(hdr[0]);
291 csum = le32_to_cpu(hdr[1]);
293 data = ((void *)hdr) - size;
294 if ((unsigned long)data < initrd_start) {
295 pr_err("bootconfig size %d is greater than initrd size %ld\n",
296 size, initrd_end - initrd_start);
300 /* Remove bootconfig from initramfs/initrd */
301 initrd_end = (unsigned long)data;
310 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
316 #ifdef CONFIG_BOOT_CONFIG
318 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
320 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
322 static int __init xbc_snprint_cmdline(char *buf, size_t size,
323 struct xbc_node *root)
325 struct xbc_node *knode, *vnode;
326 char *end = buf + size;
330 xbc_node_for_each_key_value(root, knode, val) {
331 ret = xbc_node_compose_key_after(root, knode,
332 xbc_namebuf, XBC_KEYLEN_MAX);
336 vnode = xbc_node_get_child(knode);
338 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
344 xbc_array_for_each_value(vnode, val) {
345 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
353 return buf - (end - size);
357 /* Make an extra command line under given key word */
358 static char * __init xbc_make_cmdline(const char *key)
360 struct xbc_node *root;
364 root = xbc_find_node(key);
368 /* Count required buffer size */
369 len = xbc_snprint_cmdline(NULL, 0, root);
373 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
375 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
379 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
380 if (ret < 0 || ret > len) {
381 pr_err("Failed to print extra kernel cmdline.\n");
388 static u32 boot_config_checksum(unsigned char *p, u32 size)
398 static int __init bootconfig_params(char *param, char *val,
399 const char *unused, void *arg)
401 if (strcmp(param, "bootconfig") == 0) {
402 bootconfig_found = true;
407 static void __init setup_boot_config(const char *cmdline)
409 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
413 char *data, *copy, *err;
416 /* Cut out the bootconfig data even if we have no bootconfig option */
417 data = get_boot_config_from_initrd(&size, &csum);
419 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
420 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
423 if (IS_ERR(err) || !bootconfig_found)
426 /* parse_args() stops at '--' and returns an address */
428 initargs_found = true;
431 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
435 if (size >= XBC_DATA_MAX) {
436 pr_err("bootconfig size %d greater than max size %d\n",
441 if (boot_config_checksum((unsigned char *)data, size) != csum) {
442 pr_err("bootconfig checksum failed\n");
446 copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
448 pr_err("Failed to allocate memory for bootconfig\n");
452 memcpy(copy, data, size);
455 ret = xbc_init(copy, &msg, &pos);
458 pr_err("Failed to init bootconfig: %s.\n", msg);
460 pr_err("Failed to parse bootconfig: %s at %d.\n",
463 pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
464 /* keys starting with "kernel." are passed via cmdline */
465 extra_command_line = xbc_make_cmdline("kernel");
466 /* Also, "init." keys are init arguments */
467 extra_init_args = xbc_make_cmdline("init");
474 static void __init setup_boot_config(const char *cmdline)
476 /* Remove bootconfig data from initrd */
477 get_boot_config_from_initrd(NULL, NULL);
480 static int __init warn_bootconfig(char *str)
482 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
485 early_param("bootconfig", warn_bootconfig);
489 /* Change NUL term back to "=", to make "param" the whole string. */
490 static void __init repair_env_string(char *param, char *val)
493 /* param=val or param="val"? */
494 if (val == param+strlen(param)+1)
496 else if (val == param+strlen(param)+2) {
498 memmove(val-1, val, strlen(val)+1);
504 /* Anything after -- gets handed straight to init. */
505 static int __init set_init_arg(char *param, char *val,
506 const char *unused, void *arg)
513 repair_env_string(param, val);
515 for (i = 0; argv_init[i]; i++) {
516 if (i == MAX_INIT_ARGS) {
517 panic_later = "init";
522 argv_init[i] = param;
527 * Unknown boot options get handed to init, unless they look like
528 * unused parameters (modprobe will find them in /proc/cmdline).
530 static int __init unknown_bootoption(char *param, char *val,
531 const char *unused, void *arg)
533 size_t len = strlen(param);
535 repair_env_string(param, val);
537 /* Handle obsolete-style parameters */
538 if (obsolete_checksetup(param))
541 /* Unused module parameter. */
542 if (strnchr(param, len, '.'))
549 /* Environment option */
551 for (i = 0; envp_init[i]; i++) {
552 if (i == MAX_INIT_ENVS) {
556 if (!strncmp(param, envp_init[i], len+1))
559 envp_init[i] = param;
561 /* Command line option */
563 for (i = 0; argv_init[i]; i++) {
564 if (i == MAX_INIT_ARGS) {
565 panic_later = "init";
569 argv_init[i] = param;
574 static int __init init_setup(char *str)
578 execute_command = str;
580 * In case LILO is going to boot us with default command line,
581 * it prepends "auto" before the whole cmdline which makes
582 * the shell think it should execute a script with such name.
583 * So we ignore all arguments entered _before_ init=... [MJ]
585 for (i = 1; i < MAX_INIT_ARGS; i++)
589 __setup("init=", init_setup);
591 static int __init rdinit_setup(char *str)
595 ramdisk_execute_command = str;
596 /* See "auto" comment in init_setup */
597 for (i = 1; i < MAX_INIT_ARGS; i++)
601 __setup("rdinit=", rdinit_setup);
604 static const unsigned int setup_max_cpus = NR_CPUS;
605 static inline void setup_nr_cpu_ids(void) { }
606 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
610 * We need to store the untouched command line for future reference.
611 * We also need to store the touched command line since the parameter
612 * parsing is performed in place, and we should allow a component to
613 * store reference of name/value for future reference.
615 static void __init setup_command_line(char *command_line)
617 size_t len, xlen = 0, ilen = 0;
619 if (extra_command_line)
620 xlen = strlen(extra_command_line);
622 ilen = strlen(extra_init_args) + 4; /* for " -- " */
624 len = xlen + strlen(boot_command_line) + 1;
626 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
627 if (!saved_command_line)
628 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
630 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
631 if (!static_command_line)
632 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
636 * We have to put extra_command_line before boot command
637 * lines because there could be dashes (separator of init
638 * command line) in the command lines.
640 strcpy(saved_command_line, extra_command_line);
641 strcpy(static_command_line, extra_command_line);
643 strcpy(saved_command_line + xlen, boot_command_line);
644 strcpy(static_command_line + xlen, command_line);
648 * Append supplemental init boot args to saved_command_line
649 * so that user can check what command line options passed
652 len = strlen(saved_command_line);
653 if (initargs_found) {
654 saved_command_line[len++] = ' ';
656 strcpy(saved_command_line + len, " -- ");
660 strcpy(saved_command_line + len, extra_init_args);
665 * We need to finalize in a non-__init function or else race conditions
666 * between the root thread and the init thread may cause start_kernel to
667 * be reaped by free_initmem before the root thread has proceeded to
670 * gcc-3.4 accidentally inlines this function, so use noinline.
673 static __initdata DECLARE_COMPLETION(kthreadd_done);
675 noinline void __ref rest_init(void)
677 struct task_struct *tsk;
680 rcu_scheduler_starting();
682 * We need to spawn init first so that it obtains pid 1, however
683 * the init task will end up wanting to create kthreads, which, if
684 * we schedule it before we create kthreadd, will OOPS.
686 pid = kernel_thread(kernel_init, NULL, CLONE_FS);
688 * Pin init on the boot CPU. Task migration is not properly working
689 * until sched_init_smp() has been run. It will set the allowed
690 * CPUs for init to the non isolated CPUs.
693 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
694 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
697 numa_default_policy();
698 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
700 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
704 * Enable might_sleep() and smp_processor_id() checks.
705 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
706 * kernel_thread() would trigger might_sleep() splats. With
707 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
708 * already, but it's stuck on the kthreadd_done completion.
710 system_state = SYSTEM_SCHEDULING;
712 complete(&kthreadd_done);
715 * The boot idle thread must execute schedule()
716 * at least once to get things moving:
718 schedule_preempt_disabled();
719 /* Call into cpu_idle with preempt disabled */
720 cpu_startup_entry(CPUHP_ONLINE);
723 /* Check for early params. */
724 static int __init do_early_param(char *param, char *val,
725 const char *unused, void *arg)
727 const struct obs_kernel_param *p;
729 for (p = __setup_start; p < __setup_end; p++) {
730 if ((p->early && parameq(param, p->str)) ||
731 (strcmp(param, "console") == 0 &&
732 strcmp(p->str, "earlycon") == 0)
734 if (p->setup_func(val) != 0)
735 pr_warn("Malformed early option '%s'\n", param);
738 /* We accept everything at this stage. */
742 void __init parse_early_options(char *cmdline)
744 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
748 /* Arch code calls this early on, or if not, just before other parsing. */
749 void __init parse_early_param(void)
751 static int done __initdata;
752 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
757 /* All fall through to do_early_param. */
758 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
759 parse_early_options(tmp_cmdline);
763 void __init __weak arch_post_acpi_subsys_init(void) { }
765 void __init __weak smp_setup_processor_id(void)
769 # if THREAD_SIZE >= PAGE_SIZE
770 void __init __weak thread_stack_cache_init(void)
775 void __init __weak mem_encrypt_init(void) { }
777 void __init __weak poking_init(void) { }
779 void __init __weak pgtable_cache_init(void) { }
782 core_param(initcall_debug, initcall_debug, bool, 0644);
784 #ifdef TRACEPOINTS_ENABLED
785 static void __init initcall_debug_enable(void);
787 static inline void initcall_debug_enable(void)
792 /* Report memory auto-initialization states for this boot. */
793 static void __init report_meminit(void)
797 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
798 stack = "all(pattern)";
799 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
801 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
802 stack = "byref_all(zero)";
803 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
804 stack = "byref(zero)";
805 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
806 stack = "__user(zero)";
810 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
811 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
812 want_init_on_free() ? "on" : "off");
813 if (want_init_on_free())
814 pr_info("mem auto-init: clearing system memory may take some time...\n");
818 * Set up kernel memory allocators
820 static void __init mm_init(void)
823 * page_ext requires contiguous pages,
824 * bigger than MAX_ORDER unless SPARSEMEM.
826 page_ext_init_flatmem();
827 init_mem_debugging_and_hardening();
831 /* page_owner must be initialized after buddy is ready */
832 page_ext_init_flatmem_late();
836 debug_objects_mem_init();
839 /* Should be run before the first non-init thread is created */
841 /* Should be run after espfix64 is set up. */
845 void __init __weak arch_call_rest_init(void)
850 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
855 set_task_stack_end_magic(&init_task);
856 smp_setup_processor_id();
857 debug_objects_early_init();
862 early_boot_irqs_disabled = true;
865 * Interrupts are still disabled. Do necessary setups, then
870 pr_notice("%s", linux_banner);
871 early_security_init();
872 setup_arch(&command_line);
873 setup_boot_config(command_line);
874 setup_command_line(command_line);
876 setup_per_cpu_areas();
877 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
878 boot_cpu_hotplug_init();
880 build_all_zonelists(NULL);
883 pr_notice("Kernel command line: %s\n", saved_command_line);
884 /* parameters may set static keys */
887 after_dashes = parse_args("Booting kernel",
888 static_command_line, __start___param,
889 __stop___param - __start___param,
890 -1, -1, NULL, &unknown_bootoption);
891 if (!IS_ERR_OR_NULL(after_dashes))
892 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
895 parse_args("Setting extra init args", extra_init_args,
896 NULL, 0, -1, -1, NULL, set_init_arg);
899 * These use large bootmem allocations and must precede
903 vfs_caches_init_early();
910 /* trace_printk can be enabled here */
914 * Set up the scheduler prior starting any interrupts (such as the
915 * timer interrupt). Full topology setup happens at smp_init()
916 * time - but meanwhile we still have a functioning scheduler.
920 * Disable preemption - early bootup scheduling is extremely
921 * fragile until we cpu_idle() for the first time.
924 if (WARN(!irqs_disabled(),
925 "Interrupts were enabled *very* early, fixing it\n"))
930 * Set up housekeeping before setting up workqueues to allow the unbound
931 * workqueue to take non-housekeeping into account.
936 * Allow workqueue creation and work item queueing/cancelling
937 * early. Work item execution depends on kthreads and starts after
940 workqueue_init_early();
944 /* Trace events are available after this */
948 initcall_debug_enable();
950 context_tracking_init();
951 /* init some links before init_ISA_irqs() */
963 * For best initial stack canary entropy, prepare it after:
964 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
965 * - timekeeping_init() for ktime entropy used in rand_initialize()
966 * - rand_initialize() to get any arch-specific entropy like RDRAND
967 * - add_latent_entropy() to get any latent entropy
968 * - adding command line entropy
971 add_latent_entropy();
972 add_device_randomness(command_line, strlen(command_line));
973 boot_init_stack_canary();
978 call_function_init();
979 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
981 early_boot_irqs_disabled = false;
984 kmem_cache_init_late();
987 * HACK ALERT! This is early. We're enabling the console before
988 * we've done PCI setups etc, and console_init() must be aware of
989 * this. But we do want output early, in case something goes wrong.
993 panic("Too many boot %s vars at `%s'", panic_later,
999 * Need to run this when irqs are enabled, because it wants
1000 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1006 * This needs to be called before any devices perform DMA
1007 * operations that might use the SWIOTLB bounce buffers. It will
1008 * mark the bounce buffers as decrypted so that their usage will
1009 * not cause "plain-text" data to be decrypted when accessed.
1013 #ifdef CONFIG_BLK_DEV_INITRD
1014 if (initrd_start && !initrd_below_start_ok &&
1015 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1016 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1017 page_to_pfn(virt_to_page((void *)initrd_start)),
1022 setup_per_cpu_pageset();
1032 if (efi_enabled(EFI_RUNTIME_SERVICES))
1033 efi_enter_virtual_mode();
1035 thread_stack_cache_init();
1051 taskstats_init_early();
1057 acpi_subsystem_init();
1058 arch_post_acpi_subsys_init();
1061 /* Do the rest non-__init'ed, we're now alive */
1062 arch_call_rest_init();
1064 prevent_tail_call_optimization();
1067 /* Call all constructor functions linked into the kernel. */
1068 static void __init do_ctors(void)
1071 * For UML, the constructors have already been called by the
1072 * normal setup code as it's just a normal ELF binary, so we
1073 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1074 * even on UML for modules.
1076 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1077 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1079 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1084 #ifdef CONFIG_KALLSYMS
1085 struct blacklist_entry {
1086 struct list_head next;
1090 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1092 static int __init initcall_blacklist(char *str)
1095 struct blacklist_entry *entry;
1097 /* str argument is a comma-separated list of functions */
1099 str_entry = strsep(&str, ",");
1101 pr_debug("blacklisting initcall %s\n", str_entry);
1102 entry = memblock_alloc(sizeof(*entry),
1105 panic("%s: Failed to allocate %zu bytes\n",
1106 __func__, sizeof(*entry));
1107 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1110 panic("%s: Failed to allocate %zu bytes\n",
1111 __func__, strlen(str_entry) + 1);
1112 strcpy(entry->buf, str_entry);
1113 list_add(&entry->next, &blacklisted_initcalls);
1115 } while (str_entry);
1120 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1122 struct blacklist_entry *entry;
1123 char fn_name[KSYM_SYMBOL_LEN];
1126 if (list_empty(&blacklisted_initcalls))
1129 addr = (unsigned long) dereference_function_descriptor(fn);
1130 sprint_symbol_no_offset(fn_name, addr);
1133 * fn will be "function_name [module_name]" where [module_name] is not
1134 * displayed for built-in init functions. Strip off the [module_name].
1136 strreplace(fn_name, ' ', '\0');
1138 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1139 if (!strcmp(fn_name, entry->buf)) {
1140 pr_debug("initcall %s blacklisted\n", fn_name);
1148 static int __init initcall_blacklist(char *str)
1150 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1154 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1159 __setup("initcall_blacklist=", initcall_blacklist);
1161 static __init_or_module void
1162 trace_initcall_start_cb(void *data, initcall_t fn)
1164 ktime_t *calltime = (ktime_t *)data;
1166 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1167 *calltime = ktime_get();
1170 static __init_or_module void
1171 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1173 ktime_t *calltime = (ktime_t *)data;
1174 ktime_t delta, rettime;
1175 unsigned long long duration;
1177 rettime = ktime_get();
1178 delta = ktime_sub(rettime, *calltime);
1179 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1180 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1184 static ktime_t initcall_calltime;
1186 #ifdef TRACEPOINTS_ENABLED
1187 static void __init initcall_debug_enable(void)
1191 ret = register_trace_initcall_start(trace_initcall_start_cb,
1192 &initcall_calltime);
1193 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1194 &initcall_calltime);
1195 WARN(ret, "Failed to register initcall tracepoints\n");
1197 # define do_trace_initcall_start trace_initcall_start
1198 # define do_trace_initcall_finish trace_initcall_finish
1200 static inline void do_trace_initcall_start(initcall_t fn)
1202 if (!initcall_debug)
1204 trace_initcall_start_cb(&initcall_calltime, fn);
1206 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1208 if (!initcall_debug)
1210 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1212 #endif /* !TRACEPOINTS_ENABLED */
1214 int __init_or_module do_one_initcall(initcall_t fn)
1216 int count = preempt_count();
1220 if (initcall_blacklisted(fn))
1223 do_trace_initcall_start(fn);
1225 do_trace_initcall_finish(fn, ret);
1229 if (preempt_count() != count) {
1230 sprintf(msgbuf, "preemption imbalance ");
1231 preempt_count_set(count);
1233 if (irqs_disabled()) {
1234 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1237 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1239 add_latent_entropy();
1244 extern initcall_entry_t __initcall_start[];
1245 extern initcall_entry_t __initcall0_start[];
1246 extern initcall_entry_t __initcall1_start[];
1247 extern initcall_entry_t __initcall2_start[];
1248 extern initcall_entry_t __initcall3_start[];
1249 extern initcall_entry_t __initcall4_start[];
1250 extern initcall_entry_t __initcall5_start[];
1251 extern initcall_entry_t __initcall6_start[];
1252 extern initcall_entry_t __initcall7_start[];
1253 extern initcall_entry_t __initcall_end[];
1255 static initcall_entry_t *initcall_levels[] __initdata = {
1267 /* Keep these in sync with initcalls in include/linux/init.h */
1268 static const char *initcall_level_names[] __initdata = {
1279 static int __init ignore_unknown_bootoption(char *param, char *val,
1280 const char *unused, void *arg)
1285 static void __init do_initcall_level(int level, char *command_line)
1287 initcall_entry_t *fn;
1289 parse_args(initcall_level_names[level],
1290 command_line, __start___param,
1291 __stop___param - __start___param,
1293 NULL, ignore_unknown_bootoption);
1295 trace_initcall_level(initcall_level_names[level]);
1296 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1297 do_one_initcall(initcall_from_entry(fn));
1300 static void __init do_initcalls(void)
1303 size_t len = strlen(saved_command_line) + 1;
1306 command_line = kzalloc(len, GFP_KERNEL);
1308 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1310 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1311 /* Parser modifies command_line, restore it each time */
1312 strcpy(command_line, saved_command_line);
1313 do_initcall_level(level, command_line);
1316 kfree(command_line);
1320 * Ok, the machine is now initialized. None of the devices
1321 * have been touched yet, but the CPU subsystem is up and
1322 * running, and memory and process management works.
1324 * Now we can finally start doing some real work..
1326 static void __init do_basic_setup(void)
1332 usermodehelper_enable();
1336 static void __init do_pre_smp_initcalls(void)
1338 initcall_entry_t *fn;
1340 trace_initcall_level("early");
1341 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1342 do_one_initcall(initcall_from_entry(fn));
1345 static int run_init_process(const char *init_filename)
1347 const char *const *p;
1349 argv_init[0] = init_filename;
1350 pr_info("Run %s as init process\n", init_filename);
1351 pr_debug(" with arguments:\n");
1352 for (p = argv_init; *p; p++)
1353 pr_debug(" %s\n", *p);
1354 pr_debug(" with environment:\n");
1355 for (p = envp_init; *p; p++)
1356 pr_debug(" %s\n", *p);
1357 return kernel_execve(init_filename, argv_init, envp_init);
1360 static int try_to_run_init_process(const char *init_filename)
1364 ret = run_init_process(init_filename);
1366 if (ret && ret != -ENOENT) {
1367 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1368 init_filename, ret);
1374 static noinline void __init kernel_init_freeable(void);
1376 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1377 bool rodata_enabled __ro_after_init = true;
1378 static int __init set_debug_rodata(char *str)
1380 return strtobool(str, &rodata_enabled);
1382 __setup("rodata=", set_debug_rodata);
1385 #ifdef CONFIG_STRICT_KERNEL_RWX
1386 static void mark_readonly(void)
1388 if (rodata_enabled) {
1390 * load_module() results in W+X mappings, which are cleaned
1391 * up with call_rcu(). Let's make sure that queued work is
1392 * flushed so that we don't hit false positives looking for
1393 * insecure pages which are W+X.
1399 pr_info("Kernel memory protection disabled.\n");
1401 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1402 static inline void mark_readonly(void)
1404 pr_warn("Kernel memory protection not selected by kernel config.\n");
1407 static inline void mark_readonly(void)
1409 pr_warn("This architecture does not have kernel memory protection.\n");
1413 void __weak free_initmem(void)
1415 free_initmem_default(POISON_FREE_INITMEM);
1418 static int __ref kernel_init(void *unused)
1422 kernel_init_freeable();
1423 /* need to finish all async __init code before freeing the memory */
1424 async_synchronize_full();
1425 kprobe_free_init_mem();
1426 ftrace_free_init_mem();
1431 * Kernel mappings are now finalized - update the userspace page-table
1436 system_state = SYSTEM_RUNNING;
1437 numa_default_policy();
1439 rcu_end_inkernel_boot();
1443 if (ramdisk_execute_command) {
1444 ret = run_init_process(ramdisk_execute_command);
1447 pr_err("Failed to execute %s (error %d)\n",
1448 ramdisk_execute_command, ret);
1452 * We try each of these until one succeeds.
1454 * The Bourne shell can be used instead of init if we are
1455 * trying to recover a really broken machine.
1457 if (execute_command) {
1458 ret = run_init_process(execute_command);
1461 panic("Requested init %s failed (error %d).",
1462 execute_command, ret);
1465 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1466 ret = run_init_process(CONFIG_DEFAULT_INIT);
1468 pr_err("Default init %s failed (error %d)\n",
1469 CONFIG_DEFAULT_INIT, ret);
1474 if (!try_to_run_init_process("/sbin/init") ||
1475 !try_to_run_init_process("/etc/init") ||
1476 !try_to_run_init_process("/bin/init") ||
1477 !try_to_run_init_process("/bin/sh"))
1480 panic("No working init found. Try passing init= option to kernel. "
1481 "See Linux Documentation/admin-guide/init.rst for guidance.");
1484 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1485 void __init console_on_rootfs(void)
1487 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1490 pr_err("Warning: unable to open an initial console.\n");
1499 static noinline void __init kernel_init_freeable(void)
1502 * Wait until kthreadd is all set-up.
1504 wait_for_completion(&kthreadd_done);
1506 /* Now the scheduler is fully set up and can do blocking allocations */
1507 gfp_allowed_mask = __GFP_BITS_MASK;
1510 * init can allocate pages on any node
1512 set_mems_allowed(node_states[N_MEMORY]);
1514 cad_pid = task_pid(current);
1516 smp_prepare_cpus(setup_max_cpus);
1520 init_mm_internals();
1522 rcu_init_tasks_generic();
1523 do_pre_smp_initcalls();
1524 lockup_detector_init();
1530 page_alloc_init_late();
1531 /* Initialize page ext after all struct pages are initialized. */
1536 kunit_run_all_tests();
1538 console_on_rootfs();
1541 * check if there is an early userspace init. If yes, let it do all
1544 if (init_eaccess(ramdisk_execute_command) != 0) {
1545 ramdisk_execute_command = NULL;
1546 prepare_namespace();
1550 * Ok, we have completed the initial bootup, and
1551 * we're essentially up and running. Get rid of the
1552 * initmem segments and start the user-mode stuff..
1554 * rootfs is available now, try loading the public keys
1555 * and default modules
1558 integrity_load_keys();