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/padata.h>
67 #include <linux/pid_namespace.h>
68 #include <linux/device/driver.h>
69 #include <linux/kthread.h>
70 #include <linux/sched.h>
71 #include <linux/sched/init.h>
72 #include <linux/signal.h>
73 #include <linux/idr.h>
74 #include <linux/kgdb.h>
75 #include <linux/ftrace.h>
76 #include <linux/async.h>
77 #include <linux/sfi.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 static int kernel_init(void *);
112 extern void init_IRQ(void);
113 extern void radix_tree_init(void);
116 * Debug helper: via this flag we know that we are in 'early bootup code'
117 * where only the boot processor is running with IRQ disabled. This means
118 * two things - IRQ must not be enabled before the flag is cleared and some
119 * operations which are not allowed with IRQ disabled are allowed while the
122 bool early_boot_irqs_disabled __read_mostly;
124 enum system_states system_state __read_mostly;
125 EXPORT_SYMBOL(system_state);
128 * Boot command-line arguments
130 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
131 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
133 extern void time_init(void);
134 /* Default late time init is NULL. archs can override this later. */
135 void (*__initdata late_time_init)(void);
137 /* Untouched command line saved by arch-specific code. */
138 char __initdata boot_command_line[COMMAND_LINE_SIZE];
139 /* Untouched saved command line (eg. for /proc) */
140 char *saved_command_line;
141 /* Command line for parameter parsing */
142 static char *static_command_line;
143 /* Untouched extra command line */
144 static char *extra_command_line;
145 /* Extra init arguments */
146 static char *extra_init_args;
148 #ifdef CONFIG_BOOT_CONFIG
149 /* Is bootconfig on command line? */
150 static bool bootconfig_found;
151 static bool initargs_found;
153 # define bootconfig_found false
154 # define initargs_found false
157 static char *execute_command;
158 static char *ramdisk_execute_command = "/init";
161 * Used to generate warnings if static_key manipulation functions are used
162 * before jump_label_init is called.
164 bool static_key_initialized __read_mostly;
165 EXPORT_SYMBOL_GPL(static_key_initialized);
168 * If set, this is an indication to the drivers that reset the underlying
169 * device before going ahead with the initialization otherwise driver might
170 * rely on the BIOS and skip the reset operation.
172 * This is useful if kernel is booting in an unreliable environment.
173 * For ex. kdump situation where previous kernel has crashed, BIOS has been
174 * skipped and devices will be in unknown state.
176 unsigned int reset_devices;
177 EXPORT_SYMBOL(reset_devices);
179 static int __init set_reset_devices(char *str)
185 __setup("reset_devices", set_reset_devices);
187 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
188 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
189 static const char *panic_later, *panic_param;
191 extern const struct obs_kernel_param __setup_start[], __setup_end[];
193 static bool __init obsolete_checksetup(char *line)
195 const struct obs_kernel_param *p;
196 bool had_early_param = false;
200 int n = strlen(p->str);
201 if (parameqn(line, p->str, n)) {
203 /* Already done in parse_early_param?
204 * (Needs exact match on param part).
205 * Keep iterating, as we can have early
206 * params and __setups of same names 8( */
207 if (line[n] == '\0' || line[n] == '=')
208 had_early_param = true;
209 } else if (!p->setup_func) {
210 pr_warn("Parameter %s is obsolete, ignored\n",
213 } else if (p->setup_func(line + n))
217 } while (p < __setup_end);
219 return had_early_param;
223 * This should be approx 2 Bo*oMips to start (note initial shift), and will
224 * still work even if initially too large, it will just take slightly longer
226 unsigned long loops_per_jiffy = (1<<12);
227 EXPORT_SYMBOL(loops_per_jiffy);
229 static int __init debug_kernel(char *str)
231 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
235 static int __init quiet_kernel(char *str)
237 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
241 early_param("debug", debug_kernel);
242 early_param("quiet", quiet_kernel);
244 static int __init loglevel(char *str)
249 * Only update loglevel value when a correct setting was passed,
250 * to prevent blind crashes (when loglevel being set to 0) that
251 * are quite hard to debug
253 if (get_option(&str, &newlevel)) {
254 console_loglevel = newlevel;
261 early_param("loglevel", loglevel);
263 #ifdef CONFIG_BLK_DEV_INITRD
264 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
273 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
274 if (memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
277 hdr = (u32 *)(data - 8);
281 data = ((void *)hdr) - size;
282 if ((unsigned long)data < initrd_start) {
283 pr_err("bootconfig size %d is greater than initrd size %ld\n",
284 size, initrd_end - initrd_start);
288 /* Remove bootconfig from initramfs/initrd */
289 initrd_end = (unsigned long)data;
298 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
304 #ifdef CONFIG_BOOT_CONFIG
306 char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
308 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
310 static int __init xbc_snprint_cmdline(char *buf, size_t size,
311 struct xbc_node *root)
313 struct xbc_node *knode, *vnode;
314 char *end = buf + size;
318 xbc_node_for_each_key_value(root, knode, val) {
319 ret = xbc_node_compose_key_after(root, knode,
320 xbc_namebuf, XBC_KEYLEN_MAX);
324 vnode = xbc_node_get_child(knode);
326 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
332 xbc_array_for_each_value(vnode, val) {
333 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
341 return buf - (end - size);
345 /* Make an extra command line under given key word */
346 static char * __init xbc_make_cmdline(const char *key)
348 struct xbc_node *root;
352 root = xbc_find_node(key);
356 /* Count required buffer size */
357 len = xbc_snprint_cmdline(NULL, 0, root);
361 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
363 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
367 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
368 if (ret < 0 || ret > len) {
369 pr_err("Failed to print extra kernel cmdline.\n");
376 static u32 boot_config_checksum(unsigned char *p, u32 size)
386 static int __init bootconfig_params(char *param, char *val,
387 const char *unused, void *arg)
389 if (strcmp(param, "bootconfig") == 0) {
390 bootconfig_found = true;
391 } else if (strcmp(param, "--") == 0) {
392 initargs_found = true;
397 static void __init setup_boot_config(const char *cmdline)
399 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
406 /* Cut out the bootconfig data even if we have no bootconfig option */
407 data = get_boot_config_from_initrd(&size, &csum);
409 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
410 parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
413 if (!bootconfig_found)
417 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
421 if (size >= XBC_DATA_MAX) {
422 pr_err("bootconfig size %d greater than max size %d\n",
427 if (boot_config_checksum((unsigned char *)data, size) != csum) {
428 pr_err("bootconfig checksum failed\n");
432 copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
434 pr_err("Failed to allocate memory for bootconfig\n");
438 memcpy(copy, data, size);
441 ret = xbc_init(copy, &msg, &pos);
444 pr_err("Failed to init bootconfig: %s.\n", msg);
446 pr_err("Failed to parse bootconfig: %s at %d.\n",
449 pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
450 /* keys starting with "kernel." are passed via cmdline */
451 extra_command_line = xbc_make_cmdline("kernel");
452 /* Also, "init." keys are init arguments */
453 extra_init_args = xbc_make_cmdline("init");
460 static void __init setup_boot_config(const char *cmdline)
462 /* Remove bootconfig data from initrd */
463 get_boot_config_from_initrd(NULL, NULL);
466 static int __init warn_bootconfig(char *str)
468 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOTCONFIG is not set.\n");
471 early_param("bootconfig", warn_bootconfig);
475 /* Change NUL term back to "=", to make "param" the whole string. */
476 static void __init repair_env_string(char *param, char *val)
479 /* param=val or param="val"? */
480 if (val == param+strlen(param)+1)
482 else if (val == param+strlen(param)+2) {
484 memmove(val-1, val, strlen(val)+1);
490 /* Anything after -- gets handed straight to init. */
491 static int __init set_init_arg(char *param, char *val,
492 const char *unused, void *arg)
499 repair_env_string(param, val);
501 for (i = 0; argv_init[i]; i++) {
502 if (i == MAX_INIT_ARGS) {
503 panic_later = "init";
508 argv_init[i] = param;
513 * Unknown boot options get handed to init, unless they look like
514 * unused parameters (modprobe will find them in /proc/cmdline).
516 static int __init unknown_bootoption(char *param, char *val,
517 const char *unused, void *arg)
519 size_t len = strlen(param);
521 repair_env_string(param, val);
523 /* Handle obsolete-style parameters */
524 if (obsolete_checksetup(param))
527 /* Unused module parameter. */
528 if (strnchr(param, len, '.'))
535 /* Environment option */
537 for (i = 0; envp_init[i]; i++) {
538 if (i == MAX_INIT_ENVS) {
542 if (!strncmp(param, envp_init[i], len+1))
545 envp_init[i] = param;
547 /* Command line option */
549 for (i = 0; argv_init[i]; i++) {
550 if (i == MAX_INIT_ARGS) {
551 panic_later = "init";
555 argv_init[i] = param;
560 static int __init init_setup(char *str)
564 execute_command = str;
566 * In case LILO is going to boot us with default command line,
567 * it prepends "auto" before the whole cmdline which makes
568 * the shell think it should execute a script with such name.
569 * So we ignore all arguments entered _before_ init=... [MJ]
571 for (i = 1; i < MAX_INIT_ARGS; i++)
575 __setup("init=", init_setup);
577 static int __init rdinit_setup(char *str)
581 ramdisk_execute_command = str;
582 /* See "auto" comment in init_setup */
583 for (i = 1; i < MAX_INIT_ARGS; i++)
587 __setup("rdinit=", rdinit_setup);
590 static const unsigned int setup_max_cpus = NR_CPUS;
591 static inline void setup_nr_cpu_ids(void) { }
592 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
596 * We need to store the untouched command line for future reference.
597 * We also need to store the touched command line since the parameter
598 * parsing is performed in place, and we should allow a component to
599 * store reference of name/value for future reference.
601 static void __init setup_command_line(char *command_line)
603 size_t len, xlen = 0, ilen = 0;
605 if (extra_command_line)
606 xlen = strlen(extra_command_line);
608 ilen = strlen(extra_init_args) + 4; /* for " -- " */
610 len = xlen + strlen(boot_command_line) + 1;
612 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
613 if (!saved_command_line)
614 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
616 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
617 if (!static_command_line)
618 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
622 * We have to put extra_command_line before boot command
623 * lines because there could be dashes (separator of init
624 * command line) in the command lines.
626 strcpy(saved_command_line, extra_command_line);
627 strcpy(static_command_line, extra_command_line);
629 strcpy(saved_command_line + xlen, boot_command_line);
630 strcpy(static_command_line + xlen, command_line);
634 * Append supplemental init boot args to saved_command_line
635 * so that user can check what command line options passed
638 len = strlen(saved_command_line);
639 if (initargs_found) {
640 saved_command_line[len++] = ' ';
642 strcpy(saved_command_line + len, " -- ");
646 strcpy(saved_command_line + len, extra_init_args);
651 * We need to finalize in a non-__init function or else race conditions
652 * between the root thread and the init thread may cause start_kernel to
653 * be reaped by free_initmem before the root thread has proceeded to
656 * gcc-3.4 accidentally inlines this function, so use noinline.
659 static __initdata DECLARE_COMPLETION(kthreadd_done);
661 noinline void __ref rest_init(void)
663 struct task_struct *tsk;
666 rcu_scheduler_starting();
668 * We need to spawn init first so that it obtains pid 1, however
669 * the init task will end up wanting to create kthreads, which, if
670 * we schedule it before we create kthreadd, will OOPS.
672 pid = kernel_thread(kernel_init, NULL, CLONE_FS);
674 * Pin init on the boot CPU. Task migration is not properly working
675 * until sched_init_smp() has been run. It will set the allowed
676 * CPUs for init to the non isolated CPUs.
679 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
680 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
683 numa_default_policy();
684 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
686 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
690 * Enable might_sleep() and smp_processor_id() checks.
691 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
692 * kernel_thread() would trigger might_sleep() splats. With
693 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
694 * already, but it's stuck on the kthreadd_done completion.
696 system_state = SYSTEM_SCHEDULING;
698 complete(&kthreadd_done);
701 * The boot idle thread must execute schedule()
702 * at least once to get things moving:
704 schedule_preempt_disabled();
705 /* Call into cpu_idle with preempt disabled */
706 cpu_startup_entry(CPUHP_ONLINE);
709 /* Check for early params. */
710 static int __init do_early_param(char *param, char *val,
711 const char *unused, void *arg)
713 const struct obs_kernel_param *p;
715 for (p = __setup_start; p < __setup_end; p++) {
716 if ((p->early && parameq(param, p->str)) ||
717 (strcmp(param, "console") == 0 &&
718 strcmp(p->str, "earlycon") == 0)
720 if (p->setup_func(val) != 0)
721 pr_warn("Malformed early option '%s'\n", param);
724 /* We accept everything at this stage. */
728 void __init parse_early_options(char *cmdline)
730 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
734 /* Arch code calls this early on, or if not, just before other parsing. */
735 void __init parse_early_param(void)
737 static int done __initdata;
738 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
743 /* All fall through to do_early_param. */
744 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
745 parse_early_options(tmp_cmdline);
749 void __init __weak arch_post_acpi_subsys_init(void) { }
751 void __init __weak smp_setup_processor_id(void)
755 # if THREAD_SIZE >= PAGE_SIZE
756 void __init __weak thread_stack_cache_init(void)
761 void __init __weak mem_encrypt_init(void) { }
763 void __init __weak poking_init(void) { }
765 void __init __weak pgtable_cache_init(void) { }
768 core_param(initcall_debug, initcall_debug, bool, 0644);
770 #ifdef TRACEPOINTS_ENABLED
771 static void __init initcall_debug_enable(void);
773 static inline void initcall_debug_enable(void)
778 /* Report memory auto-initialization states for this boot. */
779 static void __init report_meminit(void)
783 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
784 stack = "all(pattern)";
785 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
787 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
788 stack = "byref_all(zero)";
789 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
790 stack = "byref(zero)";
791 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
792 stack = "__user(zero)";
796 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
797 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
798 want_init_on_free() ? "on" : "off");
799 if (want_init_on_free())
800 pr_info("mem auto-init: clearing system memory may take some time...\n");
804 * Set up kernel memory allocators
806 static void __init mm_init(void)
809 * page_ext requires contiguous pages,
810 * bigger than MAX_ORDER unless SPARSEMEM.
812 page_ext_init_flatmem();
813 init_debug_pagealloc();
819 debug_objects_mem_init();
822 /* Should be run before the first non-init thread is created */
824 /* Should be run after espfix64 is set up. */
828 void __init __weak arch_call_rest_init(void)
833 asmlinkage __visible void __init start_kernel(void)
838 set_task_stack_end_magic(&init_task);
839 smp_setup_processor_id();
840 debug_objects_early_init();
845 early_boot_irqs_disabled = true;
848 * Interrupts are still disabled. Do necessary setups, then
853 pr_notice("%s", linux_banner);
854 early_security_init();
855 setup_arch(&command_line);
856 setup_boot_config(command_line);
857 setup_command_line(command_line);
859 setup_per_cpu_areas();
860 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
861 boot_cpu_hotplug_init();
863 build_all_zonelists(NULL);
866 pr_notice("Kernel command line: %s\n", saved_command_line);
867 /* parameters may set static keys */
870 after_dashes = parse_args("Booting kernel",
871 static_command_line, __start___param,
872 __stop___param - __start___param,
873 -1, -1, NULL, &unknown_bootoption);
874 if (!IS_ERR_OR_NULL(after_dashes))
875 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
878 parse_args("Setting extra init args", extra_init_args,
879 NULL, 0, -1, -1, NULL, set_init_arg);
882 * These use large bootmem allocations and must precede
886 vfs_caches_init_early();
893 /* trace_printk can be enabled here */
897 * Set up the scheduler prior starting any interrupts (such as the
898 * timer interrupt). Full topology setup happens at smp_init()
899 * time - but meanwhile we still have a functioning scheduler.
903 * Disable preemption - early bootup scheduling is extremely
904 * fragile until we cpu_idle() for the first time.
907 if (WARN(!irqs_disabled(),
908 "Interrupts were enabled *very* early, fixing it\n"))
913 * Set up housekeeping before setting up workqueues to allow the unbound
914 * workqueue to take non-housekeeping into account.
919 * Allow workqueue creation and work item queueing/cancelling
920 * early. Work item execution depends on kthreads and starts after
923 workqueue_init_early();
927 /* Trace events are available after this */
931 initcall_debug_enable();
933 context_tracking_init();
934 /* init some links before init_ISA_irqs() */
945 * For best initial stack canary entropy, prepare it after:
946 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
947 * - timekeeping_init() for ktime entropy used in rand_initialize()
948 * - rand_initialize() to get any arch-specific entropy like RDRAND
949 * - add_latent_entropy() to get any latent entropy
950 * - adding command line entropy
953 add_latent_entropy();
954 add_device_randomness(command_line, strlen(command_line));
955 boot_init_stack_canary();
960 call_function_init();
961 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
963 early_boot_irqs_disabled = false;
966 kmem_cache_init_late();
969 * HACK ALERT! This is early. We're enabling the console before
970 * we've done PCI setups etc, and console_init() must be aware of
971 * this. But we do want output early, in case something goes wrong.
975 panic("Too many boot %s vars at `%s'", panic_later,
981 * Need to run this when irqs are enabled, because it wants
982 * to self-test [hard/soft]-irqs on/off lock inversion bugs
988 * This needs to be called before any devices perform DMA
989 * operations that might use the SWIOTLB bounce buffers. It will
990 * mark the bounce buffers as decrypted so that their usage will
991 * not cause "plain-text" data to be decrypted when accessed.
995 #ifdef CONFIG_BLK_DEV_INITRD
996 if (initrd_start && !initrd_below_start_ok &&
997 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
998 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
999 page_to_pfn(virt_to_page((void *)initrd_start)),
1004 setup_per_cpu_pageset();
1014 if (efi_enabled(EFI_RUNTIME_SERVICES))
1015 efi_enter_virtual_mode();
1017 thread_stack_cache_init();
1034 taskstats_init_early();
1040 acpi_subsystem_init();
1041 arch_post_acpi_subsys_init();
1045 /* Do the rest non-__init'ed, we're now alive */
1046 arch_call_rest_init();
1048 prevent_tail_call_optimization();
1051 /* Call all constructor functions linked into the kernel. */
1052 static void __init do_ctors(void)
1054 #ifdef CONFIG_CONSTRUCTORS
1055 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1057 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1062 #ifdef CONFIG_KALLSYMS
1063 struct blacklist_entry {
1064 struct list_head next;
1068 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1070 static int __init initcall_blacklist(char *str)
1073 struct blacklist_entry *entry;
1075 /* str argument is a comma-separated list of functions */
1077 str_entry = strsep(&str, ",");
1079 pr_debug("blacklisting initcall %s\n", str_entry);
1080 entry = memblock_alloc(sizeof(*entry),
1083 panic("%s: Failed to allocate %zu bytes\n",
1084 __func__, sizeof(*entry));
1085 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1088 panic("%s: Failed to allocate %zu bytes\n",
1089 __func__, strlen(str_entry) + 1);
1090 strcpy(entry->buf, str_entry);
1091 list_add(&entry->next, &blacklisted_initcalls);
1093 } while (str_entry);
1098 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1100 struct blacklist_entry *entry;
1101 char fn_name[KSYM_SYMBOL_LEN];
1104 if (list_empty(&blacklisted_initcalls))
1107 addr = (unsigned long) dereference_function_descriptor(fn);
1108 sprint_symbol_no_offset(fn_name, addr);
1111 * fn will be "function_name [module_name]" where [module_name] is not
1112 * displayed for built-in init functions. Strip off the [module_name].
1114 strreplace(fn_name, ' ', '\0');
1116 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1117 if (!strcmp(fn_name, entry->buf)) {
1118 pr_debug("initcall %s blacklisted\n", fn_name);
1126 static int __init initcall_blacklist(char *str)
1128 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1132 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1137 __setup("initcall_blacklist=", initcall_blacklist);
1139 static __init_or_module void
1140 trace_initcall_start_cb(void *data, initcall_t fn)
1142 ktime_t *calltime = (ktime_t *)data;
1144 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1145 *calltime = ktime_get();
1148 static __init_or_module void
1149 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1151 ktime_t *calltime = (ktime_t *)data;
1152 ktime_t delta, rettime;
1153 unsigned long long duration;
1155 rettime = ktime_get();
1156 delta = ktime_sub(rettime, *calltime);
1157 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1158 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1162 static ktime_t initcall_calltime;
1164 #ifdef TRACEPOINTS_ENABLED
1165 static void __init initcall_debug_enable(void)
1169 ret = register_trace_initcall_start(trace_initcall_start_cb,
1170 &initcall_calltime);
1171 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1172 &initcall_calltime);
1173 WARN(ret, "Failed to register initcall tracepoints\n");
1175 # define do_trace_initcall_start trace_initcall_start
1176 # define do_trace_initcall_finish trace_initcall_finish
1178 static inline void do_trace_initcall_start(initcall_t fn)
1180 if (!initcall_debug)
1182 trace_initcall_start_cb(&initcall_calltime, fn);
1184 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1186 if (!initcall_debug)
1188 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1190 #endif /* !TRACEPOINTS_ENABLED */
1192 int __init_or_module do_one_initcall(initcall_t fn)
1194 int count = preempt_count();
1198 if (initcall_blacklisted(fn))
1201 do_trace_initcall_start(fn);
1203 do_trace_initcall_finish(fn, ret);
1207 if (preempt_count() != count) {
1208 sprintf(msgbuf, "preemption imbalance ");
1209 preempt_count_set(count);
1211 if (irqs_disabled()) {
1212 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1215 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1217 add_latent_entropy();
1222 extern initcall_entry_t __initcall_start[];
1223 extern initcall_entry_t __initcall0_start[];
1224 extern initcall_entry_t __initcall1_start[];
1225 extern initcall_entry_t __initcall2_start[];
1226 extern initcall_entry_t __initcall3_start[];
1227 extern initcall_entry_t __initcall4_start[];
1228 extern initcall_entry_t __initcall5_start[];
1229 extern initcall_entry_t __initcall6_start[];
1230 extern initcall_entry_t __initcall7_start[];
1231 extern initcall_entry_t __initcall_end[];
1233 static initcall_entry_t *initcall_levels[] __initdata = {
1245 /* Keep these in sync with initcalls in include/linux/init.h */
1246 static const char *initcall_level_names[] __initdata = {
1257 static int __init ignore_unknown_bootoption(char *param, char *val,
1258 const char *unused, void *arg)
1263 static void __init do_initcall_level(int level, char *command_line)
1265 initcall_entry_t *fn;
1267 parse_args(initcall_level_names[level],
1268 command_line, __start___param,
1269 __stop___param - __start___param,
1271 NULL, ignore_unknown_bootoption);
1273 trace_initcall_level(initcall_level_names[level]);
1274 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1275 do_one_initcall(initcall_from_entry(fn));
1278 static void __init do_initcalls(void)
1281 size_t len = strlen(saved_command_line) + 1;
1284 command_line = kzalloc(len, GFP_KERNEL);
1286 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1288 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1289 /* Parser modifies command_line, restore it each time */
1290 strcpy(command_line, saved_command_line);
1291 do_initcall_level(level, command_line);
1294 kfree(command_line);
1298 * Ok, the machine is now initialized. None of the devices
1299 * have been touched yet, but the CPU subsystem is up and
1300 * running, and memory and process management works.
1302 * Now we can finally start doing some real work..
1304 static void __init do_basic_setup(void)
1310 usermodehelper_enable();
1314 static void __init do_pre_smp_initcalls(void)
1316 initcall_entry_t *fn;
1318 trace_initcall_level("early");
1319 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1320 do_one_initcall(initcall_from_entry(fn));
1323 static int run_init_process(const char *init_filename)
1325 const char *const *p;
1327 argv_init[0] = init_filename;
1328 pr_info("Run %s as init process\n", init_filename);
1329 pr_debug(" with arguments:\n");
1330 for (p = argv_init; *p; p++)
1331 pr_debug(" %s\n", *p);
1332 pr_debug(" with environment:\n");
1333 for (p = envp_init; *p; p++)
1334 pr_debug(" %s\n", *p);
1335 return kernel_execve(init_filename, argv_init, envp_init);
1338 static int try_to_run_init_process(const char *init_filename)
1342 ret = run_init_process(init_filename);
1344 if (ret && ret != -ENOENT) {
1345 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1346 init_filename, ret);
1352 static noinline void __init kernel_init_freeable(void);
1354 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1355 bool rodata_enabled __ro_after_init = true;
1356 static int __init set_debug_rodata(char *str)
1358 return strtobool(str, &rodata_enabled);
1360 __setup("rodata=", set_debug_rodata);
1363 #ifdef CONFIG_STRICT_KERNEL_RWX
1364 static void mark_readonly(void)
1366 if (rodata_enabled) {
1368 * load_module() results in W+X mappings, which are cleaned
1369 * up with call_rcu(). Let's make sure that queued work is
1370 * flushed so that we don't hit false positives looking for
1371 * insecure pages which are W+X.
1377 pr_info("Kernel memory protection disabled.\n");
1379 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1380 static inline void mark_readonly(void)
1382 pr_warn("Kernel memory protection not selected by kernel config.\n");
1385 static inline void mark_readonly(void)
1387 pr_warn("This architecture does not have kernel memory protection.\n");
1391 void __weak free_initmem(void)
1393 free_initmem_default(POISON_FREE_INITMEM);
1396 static int __ref kernel_init(void *unused)
1400 kernel_init_freeable();
1401 /* need to finish all async __init code before freeing the memory */
1402 async_synchronize_full();
1403 ftrace_free_init_mem();
1408 * Kernel mappings are now finalized - update the userspace page-table
1413 system_state = SYSTEM_RUNNING;
1414 numa_default_policy();
1416 rcu_end_inkernel_boot();
1420 if (ramdisk_execute_command) {
1421 ret = run_init_process(ramdisk_execute_command);
1424 pr_err("Failed to execute %s (error %d)\n",
1425 ramdisk_execute_command, ret);
1429 * We try each of these until one succeeds.
1431 * The Bourne shell can be used instead of init if we are
1432 * trying to recover a really broken machine.
1434 if (execute_command) {
1435 ret = run_init_process(execute_command);
1438 panic("Requested init %s failed (error %d).",
1439 execute_command, ret);
1442 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1443 ret = run_init_process(CONFIG_DEFAULT_INIT);
1445 pr_err("Default init %s failed (error %d)\n",
1446 CONFIG_DEFAULT_INIT, ret);
1451 if (!try_to_run_init_process("/sbin/init") ||
1452 !try_to_run_init_process("/etc/init") ||
1453 !try_to_run_init_process("/bin/init") ||
1454 !try_to_run_init_process("/bin/sh"))
1457 panic("No working init found. Try passing init= option to kernel. "
1458 "See Linux Documentation/admin-guide/init.rst for guidance.");
1461 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1462 void __init console_on_rootfs(void)
1464 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1467 pr_err("Warning: unable to open an initial console.\n");
1476 static noinline void __init kernel_init_freeable(void)
1479 * Wait until kthreadd is all set-up.
1481 wait_for_completion(&kthreadd_done);
1483 /* Now the scheduler is fully set up and can do blocking allocations */
1484 gfp_allowed_mask = __GFP_BITS_MASK;
1487 * init can allocate pages on any node
1489 set_mems_allowed(node_states[N_MEMORY]);
1491 cad_pid = task_pid(current);
1493 smp_prepare_cpus(setup_max_cpus);
1497 init_mm_internals();
1499 do_pre_smp_initcalls();
1500 lockup_detector_init();
1506 page_alloc_init_late();
1507 /* Initialize page ext after all struct pages are initialized. */
1512 console_on_rootfs();
1515 * check if there is an early userspace init. If yes, let it do all
1518 if (init_eaccess(ramdisk_execute_command) != 0) {
1519 ramdisk_execute_command = NULL;
1520 prepare_namespace();
1524 * Ok, we have completed the initial bootup, and
1525 * we're essentially up and running. Get rid of the
1526 * initmem segments and start the user-mode stuff..
1528 * rootfs is available now, try loading the public keys
1529 * and default modules
1532 integrity_load_keys();