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/rcupdate.h>
44 #include <linux/moduleparam.h>
45 #include <linux/kallsyms.h>
46 #include <linux/writeback.h>
47 #include <linux/cpu.h>
48 #include <linux/cpuset.h>
49 #include <linux/cgroup.h>
50 #include <linux/efi.h>
51 #include <linux/tick.h>
52 #include <linux/sched/isolation.h>
53 #include <linux/interrupt.h>
54 #include <linux/taskstats_kern.h>
55 #include <linux/delayacct.h>
56 #include <linux/unistd.h>
57 #include <linux/utsname.h>
58 #include <linux/rmap.h>
59 #include <linux/mempolicy.h>
60 #include <linux/key.h>
61 #include <linux/buffer_head.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/sfi.h>
79 #include <linux/shmem_fs.h>
80 #include <linux/slab.h>
81 #include <linux/perf_event.h>
82 #include <linux/ptrace.h>
83 #include <linux/pti.h>
84 #include <linux/blkdev.h>
85 #include <linux/elevator.h>
86 #include <linux/sched/clock.h>
87 #include <linux/sched/task.h>
88 #include <linux/sched/task_stack.h>
89 #include <linux/context_tracking.h>
90 #include <linux/random.h>
91 #include <linux/list.h>
92 #include <linux/integrity.h>
93 #include <linux/proc_ns.h>
95 #include <linux/cache.h>
96 #include <linux/rodata_test.h>
97 #include <linux/jump_label.h>
98 #include <linux/mem_encrypt.h>
99 #include <linux/kcsan.h>
100 #include <linux/init_syscalls.h>
103 #include <asm/bugs.h>
104 #include <asm/setup.h>
105 #include <asm/sections.h>
106 #include <asm/cacheflush.h>
108 #define CREATE_TRACE_POINTS
109 #include <trace/events/initcall.h>
111 #include <kunit/test.h>
113 static int kernel_init(void *);
115 extern void init_IRQ(void);
116 extern void radix_tree_init(void);
119 * Debug helper: via this flag we know that we are in 'early bootup code'
120 * where only the boot processor is running with IRQ disabled. This means
121 * two things - IRQ must not be enabled before the flag is cleared and some
122 * operations which are not allowed with IRQ disabled are allowed while the
125 bool early_boot_irqs_disabled __read_mostly;
127 enum system_states system_state __read_mostly;
128 EXPORT_SYMBOL(system_state);
131 * Boot command-line arguments
133 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
134 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
136 extern void time_init(void);
137 /* Default late time init is NULL. archs can override this later. */
138 void (*__initdata late_time_init)(void);
140 /* Untouched command line saved by arch-specific code. */
141 char __initdata boot_command_line[COMMAND_LINE_SIZE];
142 /* Untouched saved command line (eg. for /proc) */
143 char *saved_command_line;
144 /* Command line for parameter parsing */
145 static char *static_command_line;
146 /* Untouched extra command line */
147 static char *extra_command_line;
148 /* Extra init arguments */
149 static char *extra_init_args;
151 #ifdef CONFIG_BOOT_CONFIG
152 /* Is bootconfig on command line? */
153 static bool bootconfig_found;
154 static bool initargs_found;
156 # define bootconfig_found false
157 # define initargs_found false
160 static char *execute_command;
161 static char *ramdisk_execute_command = "/init";
164 * Used to generate warnings if static_key manipulation functions are used
165 * before jump_label_init is called.
167 bool static_key_initialized __read_mostly;
168 EXPORT_SYMBOL_GPL(static_key_initialized);
171 * If set, this is an indication to the drivers that reset the underlying
172 * device before going ahead with the initialization otherwise driver might
173 * rely on the BIOS and skip the reset operation.
175 * This is useful if kernel is booting in an unreliable environment.
176 * For ex. kdump situation where previous kernel has crashed, BIOS has been
177 * skipped and devices will be in unknown state.
179 unsigned int reset_devices;
180 EXPORT_SYMBOL(reset_devices);
182 static int __init set_reset_devices(char *str)
188 __setup("reset_devices", set_reset_devices);
190 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
191 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
192 static const char *panic_later, *panic_param;
194 extern const struct obs_kernel_param __setup_start[], __setup_end[];
196 static bool __init obsolete_checksetup(char *line)
198 const struct obs_kernel_param *p;
199 bool had_early_param = false;
203 int n = strlen(p->str);
204 if (parameqn(line, p->str, n)) {
206 /* Already done in parse_early_param?
207 * (Needs exact match on param part).
208 * Keep iterating, as we can have early
209 * params and __setups of same names 8( */
210 if (line[n] == '\0' || line[n] == '=')
211 had_early_param = true;
212 } else if (!p->setup_func) {
213 pr_warn("Parameter %s is obsolete, ignored\n",
216 } else if (p->setup_func(line + n))
220 } while (p < __setup_end);
222 return had_early_param;
226 * This should be approx 2 Bo*oMips to start (note initial shift), and will
227 * still work even if initially too large, it will just take slightly longer
229 unsigned long loops_per_jiffy = (1<<12);
230 EXPORT_SYMBOL(loops_per_jiffy);
232 static int __init debug_kernel(char *str)
234 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
238 static int __init quiet_kernel(char *str)
240 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
244 early_param("debug", debug_kernel);
245 early_param("quiet", quiet_kernel);
247 static int __init loglevel(char *str)
252 * Only update loglevel value when a correct setting was passed,
253 * to prevent blind crashes (when loglevel being set to 0) that
254 * are quite hard to debug
256 if (get_option(&str, &newlevel)) {
257 console_loglevel = newlevel;
264 early_param("loglevel", loglevel);
266 #ifdef CONFIG_BLK_DEV_INITRD
267 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
276 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
277 if (memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
280 hdr = (u32 *)(data - 8);
284 data = ((void *)hdr) - size;
285 if ((unsigned long)data < initrd_start) {
286 pr_err("bootconfig size %d is greater than initrd size %ld\n",
287 size, initrd_end - initrd_start);
291 /* Remove bootconfig from initramfs/initrd */
292 initrd_end = (unsigned long)data;
301 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
307 #ifdef CONFIG_BOOT_CONFIG
309 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
311 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
313 static int __init xbc_snprint_cmdline(char *buf, size_t size,
314 struct xbc_node *root)
316 struct xbc_node *knode, *vnode;
317 char *end = buf + size;
321 xbc_node_for_each_key_value(root, knode, val) {
322 ret = xbc_node_compose_key_after(root, knode,
323 xbc_namebuf, XBC_KEYLEN_MAX);
327 vnode = xbc_node_get_child(knode);
329 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
335 xbc_array_for_each_value(vnode, val) {
336 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
344 return buf - (end - size);
348 /* Make an extra command line under given key word */
349 static char * __init xbc_make_cmdline(const char *key)
351 struct xbc_node *root;
355 root = xbc_find_node(key);
359 /* Count required buffer size */
360 len = xbc_snprint_cmdline(NULL, 0, root);
364 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
366 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
370 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
371 if (ret < 0 || ret > len) {
372 pr_err("Failed to print extra kernel cmdline.\n");
379 static u32 boot_config_checksum(unsigned char *p, u32 size)
389 static int __init bootconfig_params(char *param, char *val,
390 const char *unused, void *arg)
392 if (strcmp(param, "bootconfig") == 0) {
393 bootconfig_found = true;
398 static void __init setup_boot_config(const char *cmdline)
400 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
404 char *data, *copy, *err;
407 /* Cut out the bootconfig data even if we have no bootconfig option */
408 data = get_boot_config_from_initrd(&size, &csum);
410 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
411 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
414 if (IS_ERR(err) || !bootconfig_found)
417 /* parse_args() stops at '--' and returns an address */
419 initargs_found = true;
422 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
426 if (size >= XBC_DATA_MAX) {
427 pr_err("bootconfig size %d greater than max size %d\n",
432 if (boot_config_checksum((unsigned char *)data, size) != csum) {
433 pr_err("bootconfig checksum failed\n");
437 copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
439 pr_err("Failed to allocate memory for bootconfig\n");
443 memcpy(copy, data, size);
446 ret = xbc_init(copy, &msg, &pos);
449 pr_err("Failed to init bootconfig: %s.\n", msg);
451 pr_err("Failed to parse bootconfig: %s at %d.\n",
454 pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
455 /* keys starting with "kernel." are passed via cmdline */
456 extra_command_line = xbc_make_cmdline("kernel");
457 /* Also, "init." keys are init arguments */
458 extra_init_args = xbc_make_cmdline("init");
465 static void __init setup_boot_config(const char *cmdline)
467 /* Remove bootconfig data from initrd */
468 get_boot_config_from_initrd(NULL, NULL);
471 static int __init warn_bootconfig(char *str)
473 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
476 early_param("bootconfig", warn_bootconfig);
480 /* Change NUL term back to "=", to make "param" the whole string. */
481 static void __init repair_env_string(char *param, char *val)
484 /* param=val or param="val"? */
485 if (val == param+strlen(param)+1)
487 else if (val == param+strlen(param)+2) {
489 memmove(val-1, val, strlen(val)+1);
495 /* Anything after -- gets handed straight to init. */
496 static int __init set_init_arg(char *param, char *val,
497 const char *unused, void *arg)
504 repair_env_string(param, val);
506 for (i = 0; argv_init[i]; i++) {
507 if (i == MAX_INIT_ARGS) {
508 panic_later = "init";
513 argv_init[i] = param;
518 * Unknown boot options get handed to init, unless they look like
519 * unused parameters (modprobe will find them in /proc/cmdline).
521 static int __init unknown_bootoption(char *param, char *val,
522 const char *unused, void *arg)
524 size_t len = strlen(param);
526 repair_env_string(param, val);
528 /* Handle obsolete-style parameters */
529 if (obsolete_checksetup(param))
532 /* Unused module parameter. */
533 if (strnchr(param, len, '.'))
540 /* Environment option */
542 for (i = 0; envp_init[i]; i++) {
543 if (i == MAX_INIT_ENVS) {
547 if (!strncmp(param, envp_init[i], len+1))
550 envp_init[i] = param;
552 /* Command line option */
554 for (i = 0; argv_init[i]; i++) {
555 if (i == MAX_INIT_ARGS) {
556 panic_later = "init";
560 argv_init[i] = param;
565 static int __init init_setup(char *str)
569 execute_command = str;
571 * In case LILO is going to boot us with default command line,
572 * it prepends "auto" before the whole cmdline which makes
573 * the shell think it should execute a script with such name.
574 * So we ignore all arguments entered _before_ init=... [MJ]
576 for (i = 1; i < MAX_INIT_ARGS; i++)
580 __setup("init=", init_setup);
582 static int __init rdinit_setup(char *str)
586 ramdisk_execute_command = str;
587 /* See "auto" comment in init_setup */
588 for (i = 1; i < MAX_INIT_ARGS; i++)
592 __setup("rdinit=", rdinit_setup);
595 static const unsigned int setup_max_cpus = NR_CPUS;
596 static inline void setup_nr_cpu_ids(void) { }
597 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
601 * We need to store the untouched command line for future reference.
602 * We also need to store the touched command line since the parameter
603 * parsing is performed in place, and we should allow a component to
604 * store reference of name/value for future reference.
606 static void __init setup_command_line(char *command_line)
608 size_t len, xlen = 0, ilen = 0;
610 if (extra_command_line)
611 xlen = strlen(extra_command_line);
613 ilen = strlen(extra_init_args) + 4; /* for " -- " */
615 len = xlen + strlen(boot_command_line) + 1;
617 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
618 if (!saved_command_line)
619 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
621 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
622 if (!static_command_line)
623 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
627 * We have to put extra_command_line before boot command
628 * lines because there could be dashes (separator of init
629 * command line) in the command lines.
631 strcpy(saved_command_line, extra_command_line);
632 strcpy(static_command_line, extra_command_line);
634 strcpy(saved_command_line + xlen, boot_command_line);
635 strcpy(static_command_line + xlen, command_line);
639 * Append supplemental init boot args to saved_command_line
640 * so that user can check what command line options passed
643 len = strlen(saved_command_line);
644 if (initargs_found) {
645 saved_command_line[len++] = ' ';
647 strcpy(saved_command_line + len, " -- ");
651 strcpy(saved_command_line + len, extra_init_args);
656 * We need to finalize in a non-__init function or else race conditions
657 * between the root thread and the init thread may cause start_kernel to
658 * be reaped by free_initmem before the root thread has proceeded to
661 * gcc-3.4 accidentally inlines this function, so use noinline.
664 static __initdata DECLARE_COMPLETION(kthreadd_done);
666 noinline void __ref rest_init(void)
668 struct task_struct *tsk;
671 rcu_scheduler_starting();
673 * We need to spawn init first so that it obtains pid 1, however
674 * the init task will end up wanting to create kthreads, which, if
675 * we schedule it before we create kthreadd, will OOPS.
677 pid = kernel_thread(kernel_init, NULL, CLONE_FS);
679 * Pin init on the boot CPU. Task migration is not properly working
680 * until sched_init_smp() has been run. It will set the allowed
681 * CPUs for init to the non isolated CPUs.
684 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
685 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
688 numa_default_policy();
689 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
691 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
695 * Enable might_sleep() and smp_processor_id() checks.
696 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
697 * kernel_thread() would trigger might_sleep() splats. With
698 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
699 * already, but it's stuck on the kthreadd_done completion.
701 system_state = SYSTEM_SCHEDULING;
703 complete(&kthreadd_done);
706 * The boot idle thread must execute schedule()
707 * at least once to get things moving:
709 schedule_preempt_disabled();
710 /* Call into cpu_idle with preempt disabled */
711 cpu_startup_entry(CPUHP_ONLINE);
714 /* Check for early params. */
715 static int __init do_early_param(char *param, char *val,
716 const char *unused, void *arg)
718 const struct obs_kernel_param *p;
720 for (p = __setup_start; p < __setup_end; p++) {
721 if ((p->early && parameq(param, p->str)) ||
722 (strcmp(param, "console") == 0 &&
723 strcmp(p->str, "earlycon") == 0)
725 if (p->setup_func(val) != 0)
726 pr_warn("Malformed early option '%s'\n", param);
729 /* We accept everything at this stage. */
733 void __init parse_early_options(char *cmdline)
735 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
739 /* Arch code calls this early on, or if not, just before other parsing. */
740 void __init parse_early_param(void)
742 static int done __initdata;
743 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
748 /* All fall through to do_early_param. */
749 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
750 parse_early_options(tmp_cmdline);
754 void __init __weak arch_post_acpi_subsys_init(void) { }
756 void __init __weak smp_setup_processor_id(void)
760 # if THREAD_SIZE >= PAGE_SIZE
761 void __init __weak thread_stack_cache_init(void)
766 void __init __weak mem_encrypt_init(void) { }
768 void __init __weak poking_init(void) { }
770 void __init __weak pgtable_cache_init(void) { }
773 core_param(initcall_debug, initcall_debug, bool, 0644);
775 #ifdef TRACEPOINTS_ENABLED
776 static void __init initcall_debug_enable(void);
778 static inline void initcall_debug_enable(void)
783 /* Report memory auto-initialization states for this boot. */
784 static void __init report_meminit(void)
788 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
789 stack = "all(pattern)";
790 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
792 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
793 stack = "byref_all(zero)";
794 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
795 stack = "byref(zero)";
796 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
797 stack = "__user(zero)";
801 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
802 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
803 want_init_on_free() ? "on" : "off");
804 if (want_init_on_free())
805 pr_info("mem auto-init: clearing system memory may take some time...\n");
809 * Set up kernel memory allocators
811 static void __init mm_init(void)
814 * page_ext requires contiguous pages,
815 * bigger than MAX_ORDER unless SPARSEMEM.
817 page_ext_init_flatmem();
818 init_debug_pagealloc();
824 debug_objects_mem_init();
827 /* Should be run before the first non-init thread is created */
829 /* Should be run after espfix64 is set up. */
833 void __init __weak arch_call_rest_init(void)
838 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
843 set_task_stack_end_magic(&init_task);
844 smp_setup_processor_id();
845 debug_objects_early_init();
850 early_boot_irqs_disabled = true;
853 * Interrupts are still disabled. Do necessary setups, then
858 pr_notice("%s", linux_banner);
859 early_security_init();
860 setup_arch(&command_line);
861 setup_boot_config(command_line);
862 setup_command_line(command_line);
864 setup_per_cpu_areas();
865 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
866 boot_cpu_hotplug_init();
868 build_all_zonelists(NULL);
871 pr_notice("Kernel command line: %s\n", saved_command_line);
872 /* parameters may set static keys */
875 after_dashes = parse_args("Booting kernel",
876 static_command_line, __start___param,
877 __stop___param - __start___param,
878 -1, -1, NULL, &unknown_bootoption);
879 if (!IS_ERR_OR_NULL(after_dashes))
880 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
883 parse_args("Setting extra init args", extra_init_args,
884 NULL, 0, -1, -1, NULL, set_init_arg);
887 * These use large bootmem allocations and must precede
891 vfs_caches_init_early();
898 /* trace_printk can be enabled here */
902 * Set up the scheduler prior starting any interrupts (such as the
903 * timer interrupt). Full topology setup happens at smp_init()
904 * time - but meanwhile we still have a functioning scheduler.
908 * Disable preemption - early bootup scheduling is extremely
909 * fragile until we cpu_idle() for the first time.
912 if (WARN(!irqs_disabled(),
913 "Interrupts were enabled *very* early, fixing it\n"))
918 * Set up housekeeping before setting up workqueues to allow the unbound
919 * workqueue to take non-housekeeping into account.
924 * Allow workqueue creation and work item queueing/cancelling
925 * early. Work item execution depends on kthreads and starts after
928 workqueue_init_early();
932 /* Trace events are available after this */
936 initcall_debug_enable();
938 context_tracking_init();
939 /* init some links before init_ISA_irqs() */
950 * For best initial stack canary entropy, prepare it after:
951 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
952 * - timekeeping_init() for ktime entropy used in rand_initialize()
953 * - rand_initialize() to get any arch-specific entropy like RDRAND
954 * - add_latent_entropy() to get any latent entropy
955 * - adding command line entropy
958 add_latent_entropy();
959 add_device_randomness(command_line, strlen(command_line));
960 boot_init_stack_canary();
965 call_function_init();
966 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
968 early_boot_irqs_disabled = false;
971 kmem_cache_init_late();
974 * HACK ALERT! This is early. We're enabling the console before
975 * we've done PCI setups etc, and console_init() must be aware of
976 * this. But we do want output early, in case something goes wrong.
980 panic("Too many boot %s vars at `%s'", panic_later,
986 * Need to run this when irqs are enabled, because it wants
987 * to self-test [hard/soft]-irqs on/off lock inversion bugs
993 * This needs to be called before any devices perform DMA
994 * operations that might use the SWIOTLB bounce buffers. It will
995 * mark the bounce buffers as decrypted so that their usage will
996 * not cause "plain-text" data to be decrypted when accessed.
1000 #ifdef CONFIG_BLK_DEV_INITRD
1001 if (initrd_start && !initrd_below_start_ok &&
1002 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1003 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1004 page_to_pfn(virt_to_page((void *)initrd_start)),
1009 setup_per_cpu_pageset();
1019 if (efi_enabled(EFI_RUNTIME_SERVICES))
1020 efi_enter_virtual_mode();
1022 thread_stack_cache_init();
1039 taskstats_init_early();
1045 acpi_subsystem_init();
1046 arch_post_acpi_subsys_init();
1050 /* Do the rest non-__init'ed, we're now alive */
1051 arch_call_rest_init();
1053 prevent_tail_call_optimization();
1056 /* Call all constructor functions linked into the kernel. */
1057 static void __init do_ctors(void)
1059 #ifdef CONFIG_CONSTRUCTORS
1060 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1062 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1067 #ifdef CONFIG_KALLSYMS
1068 struct blacklist_entry {
1069 struct list_head next;
1073 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1075 static int __init initcall_blacklist(char *str)
1078 struct blacklist_entry *entry;
1080 /* str argument is a comma-separated list of functions */
1082 str_entry = strsep(&str, ",");
1084 pr_debug("blacklisting initcall %s\n", str_entry);
1085 entry = memblock_alloc(sizeof(*entry),
1088 panic("%s: Failed to allocate %zu bytes\n",
1089 __func__, sizeof(*entry));
1090 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1093 panic("%s: Failed to allocate %zu bytes\n",
1094 __func__, strlen(str_entry) + 1);
1095 strcpy(entry->buf, str_entry);
1096 list_add(&entry->next, &blacklisted_initcalls);
1098 } while (str_entry);
1103 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1105 struct blacklist_entry *entry;
1106 char fn_name[KSYM_SYMBOL_LEN];
1109 if (list_empty(&blacklisted_initcalls))
1112 addr = (unsigned long) dereference_function_descriptor(fn);
1113 sprint_symbol_no_offset(fn_name, addr);
1116 * fn will be "function_name [module_name]" where [module_name] is not
1117 * displayed for built-in init functions. Strip off the [module_name].
1119 strreplace(fn_name, ' ', '\0');
1121 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1122 if (!strcmp(fn_name, entry->buf)) {
1123 pr_debug("initcall %s blacklisted\n", fn_name);
1131 static int __init initcall_blacklist(char *str)
1133 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1137 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1142 __setup("initcall_blacklist=", initcall_blacklist);
1144 static __init_or_module void
1145 trace_initcall_start_cb(void *data, initcall_t fn)
1147 ktime_t *calltime = (ktime_t *)data;
1149 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1150 *calltime = ktime_get();
1153 static __init_or_module void
1154 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1156 ktime_t *calltime = (ktime_t *)data;
1157 ktime_t delta, rettime;
1158 unsigned long long duration;
1160 rettime = ktime_get();
1161 delta = ktime_sub(rettime, *calltime);
1162 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1163 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1167 static ktime_t initcall_calltime;
1169 #ifdef TRACEPOINTS_ENABLED
1170 static void __init initcall_debug_enable(void)
1174 ret = register_trace_initcall_start(trace_initcall_start_cb,
1175 &initcall_calltime);
1176 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1177 &initcall_calltime);
1178 WARN(ret, "Failed to register initcall tracepoints\n");
1180 # define do_trace_initcall_start trace_initcall_start
1181 # define do_trace_initcall_finish trace_initcall_finish
1183 static inline void do_trace_initcall_start(initcall_t fn)
1185 if (!initcall_debug)
1187 trace_initcall_start_cb(&initcall_calltime, fn);
1189 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1191 if (!initcall_debug)
1193 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1195 #endif /* !TRACEPOINTS_ENABLED */
1197 int __init_or_module do_one_initcall(initcall_t fn)
1199 int count = preempt_count();
1203 if (initcall_blacklisted(fn))
1206 do_trace_initcall_start(fn);
1208 do_trace_initcall_finish(fn, ret);
1212 if (preempt_count() != count) {
1213 sprintf(msgbuf, "preemption imbalance ");
1214 preempt_count_set(count);
1216 if (irqs_disabled()) {
1217 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1220 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1222 add_latent_entropy();
1227 extern initcall_entry_t __initcall_start[];
1228 extern initcall_entry_t __initcall0_start[];
1229 extern initcall_entry_t __initcall1_start[];
1230 extern initcall_entry_t __initcall2_start[];
1231 extern initcall_entry_t __initcall3_start[];
1232 extern initcall_entry_t __initcall4_start[];
1233 extern initcall_entry_t __initcall5_start[];
1234 extern initcall_entry_t __initcall6_start[];
1235 extern initcall_entry_t __initcall7_start[];
1236 extern initcall_entry_t __initcall_end[];
1238 static initcall_entry_t *initcall_levels[] __initdata = {
1250 /* Keep these in sync with initcalls in include/linux/init.h */
1251 static const char *initcall_level_names[] __initdata = {
1262 static int __init ignore_unknown_bootoption(char *param, char *val,
1263 const char *unused, void *arg)
1268 static void __init do_initcall_level(int level, char *command_line)
1270 initcall_entry_t *fn;
1272 parse_args(initcall_level_names[level],
1273 command_line, __start___param,
1274 __stop___param - __start___param,
1276 NULL, ignore_unknown_bootoption);
1278 trace_initcall_level(initcall_level_names[level]);
1279 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1280 do_one_initcall(initcall_from_entry(fn));
1283 static void __init do_initcalls(void)
1286 size_t len = strlen(saved_command_line) + 1;
1289 command_line = kzalloc(len, GFP_KERNEL);
1291 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1293 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1294 /* Parser modifies command_line, restore it each time */
1295 strcpy(command_line, saved_command_line);
1296 do_initcall_level(level, command_line);
1299 kfree(command_line);
1303 * Ok, the machine is now initialized. None of the devices
1304 * have been touched yet, but the CPU subsystem is up and
1305 * running, and memory and process management works.
1307 * Now we can finally start doing some real work..
1309 static void __init do_basic_setup(void)
1315 usermodehelper_enable();
1319 static void __init do_pre_smp_initcalls(void)
1321 initcall_entry_t *fn;
1323 trace_initcall_level("early");
1324 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1325 do_one_initcall(initcall_from_entry(fn));
1328 static int run_init_process(const char *init_filename)
1330 const char *const *p;
1332 argv_init[0] = init_filename;
1333 pr_info("Run %s as init process\n", init_filename);
1334 pr_debug(" with arguments:\n");
1335 for (p = argv_init; *p; p++)
1336 pr_debug(" %s\n", *p);
1337 pr_debug(" with environment:\n");
1338 for (p = envp_init; *p; p++)
1339 pr_debug(" %s\n", *p);
1340 return kernel_execve(init_filename, argv_init, envp_init);
1343 static int try_to_run_init_process(const char *init_filename)
1347 ret = run_init_process(init_filename);
1349 if (ret && ret != -ENOENT) {
1350 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1351 init_filename, ret);
1357 static noinline void __init kernel_init_freeable(void);
1359 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1360 bool rodata_enabled __ro_after_init = true;
1361 static int __init set_debug_rodata(char *str)
1363 return strtobool(str, &rodata_enabled);
1365 __setup("rodata=", set_debug_rodata);
1368 #ifdef CONFIG_STRICT_KERNEL_RWX
1369 static void mark_readonly(void)
1371 if (rodata_enabled) {
1373 * load_module() results in W+X mappings, which are cleaned
1374 * up with call_rcu(). Let's make sure that queued work is
1375 * flushed so that we don't hit false positives looking for
1376 * insecure pages which are W+X.
1382 pr_info("Kernel memory protection disabled.\n");
1384 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1385 static inline void mark_readonly(void)
1387 pr_warn("Kernel memory protection not selected by kernel config.\n");
1390 static inline void mark_readonly(void)
1392 pr_warn("This architecture does not have kernel memory protection.\n");
1396 void __weak free_initmem(void)
1398 free_initmem_default(POISON_FREE_INITMEM);
1401 static int __ref kernel_init(void *unused)
1405 kernel_init_freeable();
1406 /* need to finish all async __init code before freeing the memory */
1407 async_synchronize_full();
1408 kprobe_free_init_mem();
1409 ftrace_free_init_mem();
1414 * Kernel mappings are now finalized - update the userspace page-table
1419 system_state = SYSTEM_RUNNING;
1420 numa_default_policy();
1422 rcu_end_inkernel_boot();
1426 if (ramdisk_execute_command) {
1427 ret = run_init_process(ramdisk_execute_command);
1430 pr_err("Failed to execute %s (error %d)\n",
1431 ramdisk_execute_command, ret);
1435 * We try each of these until one succeeds.
1437 * The Bourne shell can be used instead of init if we are
1438 * trying to recover a really broken machine.
1440 if (execute_command) {
1441 ret = run_init_process(execute_command);
1444 panic("Requested init %s failed (error %d).",
1445 execute_command, ret);
1448 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1449 ret = run_init_process(CONFIG_DEFAULT_INIT);
1451 pr_err("Default init %s failed (error %d)\n",
1452 CONFIG_DEFAULT_INIT, ret);
1457 if (!try_to_run_init_process("/sbin/init") ||
1458 !try_to_run_init_process("/etc/init") ||
1459 !try_to_run_init_process("/bin/init") ||
1460 !try_to_run_init_process("/bin/sh"))
1463 panic("No working init found. Try passing init= option to kernel. "
1464 "See Linux Documentation/admin-guide/init.rst for guidance.");
1467 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1468 void __init console_on_rootfs(void)
1470 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1473 pr_err("Warning: unable to open an initial console.\n");
1482 static noinline void __init kernel_init_freeable(void)
1485 * Wait until kthreadd is all set-up.
1487 wait_for_completion(&kthreadd_done);
1489 /* Now the scheduler is fully set up and can do blocking allocations */
1490 gfp_allowed_mask = __GFP_BITS_MASK;
1493 * init can allocate pages on any node
1495 set_mems_allowed(node_states[N_MEMORY]);
1497 cad_pid = task_pid(current);
1499 smp_prepare_cpus(setup_max_cpus);
1503 init_mm_internals();
1505 do_pre_smp_initcalls();
1506 lockup_detector_init();
1512 page_alloc_init_late();
1513 /* Initialize page ext after all struct pages are initialized. */
1518 kunit_run_all_tests();
1520 console_on_rootfs();
1523 * check if there is an early userspace init. If yes, let it do all
1526 if (init_eaccess(ramdisk_execute_command) != 0) {
1527 ramdisk_execute_command = NULL;
1528 prepare_namespace();
1532 * Ok, we have completed the initial bootup, and
1533 * we're essentially up and running. Get rid of the
1534 * initmem segments and start the user-mode stuff..
1536 * rootfs is available now, try loading the public keys
1537 * and default modules
1540 integrity_load_keys();