1 // SPDX-License-Identifier: GPL-2.0-only
5 * Copyright (C) 1991, 1992 Linus Torvalds
7 * GK 2/5/95 - Changed to support mounting root fs via NFS
8 * Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
9 * Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
10 * Simplified starting of init: Michael A. Griffith <grif@acm.org>
13 #define DEBUG /* Enable initcall_debug */
15 #include <linux/types.h>
16 #include <linux/extable.h>
17 #include <linux/module.h>
18 #include <linux/proc_fs.h>
19 #include <linux/binfmts.h>
20 #include <linux/kernel.h>
21 #include <linux/syscalls.h>
22 #include <linux/stackprotector.h>
23 #include <linux/string.h>
24 #include <linux/ctype.h>
25 #include <linux/delay.h>
26 #include <linux/ioport.h>
27 #include <linux/init.h>
28 #include <linux/initrd.h>
29 #include <linux/memblock.h>
30 #include <linux/acpi.h>
31 #include <linux/bootconfig.h>
32 #include <linux/console.h>
33 #include <linux/nmi.h>
34 #include <linux/percpu.h>
35 #include <linux/kmod.h>
36 #include <linux/vmalloc.h>
37 #include <linux/kernel_stat.h>
38 #include <linux/start_kernel.h>
39 #include <linux/security.h>
40 #include <linux/smp.h>
41 #include <linux/profile.h>
42 #include <linux/rcupdate.h>
43 #include <linux/moduleparam.h>
44 #include <linux/kallsyms.h>
45 #include <linux/writeback.h>
46 #include <linux/cpu.h>
47 #include <linux/cpuset.h>
48 #include <linux/cgroup.h>
49 #include <linux/efi.h>
50 #include <linux/tick.h>
51 #include <linux/sched/isolation.h>
52 #include <linux/interrupt.h>
53 #include <linux/taskstats_kern.h>
54 #include <linux/delayacct.h>
55 #include <linux/unistd.h>
56 #include <linux/utsname.h>
57 #include <linux/rmap.h>
58 #include <linux/mempolicy.h>
59 #include <linux/key.h>
60 #include <linux/buffer_head.h>
61 #include <linux/page_ext.h>
62 #include <linux/debug_locks.h>
63 #include <linux/debugobjects.h>
64 #include <linux/lockdep.h>
65 #include <linux/kmemleak.h>
66 #include <linux/pid_namespace.h>
67 #include <linux/device/driver.h>
68 #include <linux/kthread.h>
69 #include <linux/sched.h>
70 #include <linux/sched/init.h>
71 #include <linux/signal.h>
72 #include <linux/idr.h>
73 #include <linux/kgdb.h>
74 #include <linux/ftrace.h>
75 #include <linux/async.h>
76 #include <linux/sfi.h>
77 #include <linux/shmem_fs.h>
78 #include <linux/slab.h>
79 #include <linux/perf_event.h>
80 #include <linux/ptrace.h>
81 #include <linux/pti.h>
82 #include <linux/blkdev.h>
83 #include <linux/elevator.h>
84 #include <linux/sched/clock.h>
85 #include <linux/sched/task.h>
86 #include <linux/sched/task_stack.h>
87 #include <linux/context_tracking.h>
88 #include <linux/random.h>
89 #include <linux/list.h>
90 #include <linux/integrity.h>
91 #include <linux/proc_ns.h>
93 #include <linux/cache.h>
94 #include <linux/rodata_test.h>
95 #include <linux/jump_label.h>
96 #include <linux/mem_encrypt.h>
97 #include <linux/kcsan.h>
100 #include <asm/bugs.h>
101 #include <asm/setup.h>
102 #include <asm/sections.h>
103 #include <asm/cacheflush.h>
105 #define CREATE_TRACE_POINTS
106 #include <trace/events/initcall.h>
108 static int kernel_init(void *);
110 extern void init_IRQ(void);
111 extern void radix_tree_init(void);
114 * Debug helper: via this flag we know that we are in 'early bootup code'
115 * where only the boot processor is running with IRQ disabled. This means
116 * two things - IRQ must not be enabled before the flag is cleared and some
117 * operations which are not allowed with IRQ disabled are allowed while the
120 bool early_boot_irqs_disabled __read_mostly;
122 enum system_states system_state __read_mostly;
123 EXPORT_SYMBOL(system_state);
126 * Boot command-line arguments
128 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
129 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
131 extern void time_init(void);
132 /* Default late time init is NULL. archs can override this later. */
133 void (*__initdata late_time_init)(void);
135 /* Untouched command line saved by arch-specific code. */
136 char __initdata boot_command_line[COMMAND_LINE_SIZE];
137 /* Untouched saved command line (eg. for /proc) */
138 char *saved_command_line;
139 /* Command line for parameter parsing */
140 static char *static_command_line;
141 /* Untouched extra command line */
142 static char *extra_command_line;
143 /* Extra init arguments */
144 static char *extra_init_args;
146 #ifdef CONFIG_BOOT_CONFIG
147 /* Is bootconfig on command line? */
148 static bool bootconfig_found;
149 static bool initargs_found;
151 # define bootconfig_found false
152 # define initargs_found false
155 static char *execute_command;
156 static char *ramdisk_execute_command;
159 * Used to generate warnings if static_key manipulation functions are used
160 * before jump_label_init is called.
162 bool static_key_initialized __read_mostly;
163 EXPORT_SYMBOL_GPL(static_key_initialized);
166 * If set, this is an indication to the drivers that reset the underlying
167 * device before going ahead with the initialization otherwise driver might
168 * rely on the BIOS and skip the reset operation.
170 * This is useful if kernel is booting in an unreliable environment.
171 * For ex. kdump situation where previous kernel has crashed, BIOS has been
172 * skipped and devices will be in unknown state.
174 unsigned int reset_devices;
175 EXPORT_SYMBOL(reset_devices);
177 static int __init set_reset_devices(char *str)
183 __setup("reset_devices", set_reset_devices);
185 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
186 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
187 static const char *panic_later, *panic_param;
189 extern const struct obs_kernel_param __setup_start[], __setup_end[];
191 static bool __init obsolete_checksetup(char *line)
193 const struct obs_kernel_param *p;
194 bool had_early_param = false;
198 int n = strlen(p->str);
199 if (parameqn(line, p->str, n)) {
201 /* Already done in parse_early_param?
202 * (Needs exact match on param part).
203 * Keep iterating, as we can have early
204 * params and __setups of same names 8( */
205 if (line[n] == '\0' || line[n] == '=')
206 had_early_param = true;
207 } else if (!p->setup_func) {
208 pr_warn("Parameter %s is obsolete, ignored\n",
211 } else if (p->setup_func(line + n))
215 } while (p < __setup_end);
217 return had_early_param;
221 * This should be approx 2 Bo*oMips to start (note initial shift), and will
222 * still work even if initially too large, it will just take slightly longer
224 unsigned long loops_per_jiffy = (1<<12);
225 EXPORT_SYMBOL(loops_per_jiffy);
227 static int __init debug_kernel(char *str)
229 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
233 static int __init quiet_kernel(char *str)
235 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
239 early_param("debug", debug_kernel);
240 early_param("quiet", quiet_kernel);
242 static int __init loglevel(char *str)
247 * Only update loglevel value when a correct setting was passed,
248 * to prevent blind crashes (when loglevel being set to 0) that
249 * are quite hard to debug
251 if (get_option(&str, &newlevel)) {
252 console_loglevel = newlevel;
259 early_param("loglevel", loglevel);
261 #ifdef CONFIG_BOOT_CONFIG
263 char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
265 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
267 static int __init xbc_snprint_cmdline(char *buf, size_t size,
268 struct xbc_node *root)
270 struct xbc_node *knode, *vnode;
271 char *end = buf + size;
275 xbc_node_for_each_key_value(root, knode, val) {
276 ret = xbc_node_compose_key_after(root, knode,
277 xbc_namebuf, XBC_KEYLEN_MAX);
281 vnode = xbc_node_get_child(knode);
283 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
289 xbc_array_for_each_value(vnode, val) {
290 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
298 return buf - (end - size);
302 /* Make an extra command line under given key word */
303 static char * __init xbc_make_cmdline(const char *key)
305 struct xbc_node *root;
309 root = xbc_find_node(key);
313 /* Count required buffer size */
314 len = xbc_snprint_cmdline(NULL, 0, root);
318 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
320 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
324 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
325 if (ret < 0 || ret > len) {
326 pr_err("Failed to print extra kernel cmdline.\n");
333 static u32 boot_config_checksum(unsigned char *p, u32 size)
343 static int __init bootconfig_params(char *param, char *val,
344 const char *unused, void *arg)
346 if (strcmp(param, "bootconfig") == 0) {
347 bootconfig_found = true;
348 } else if (strcmp(param, "--") == 0) {
349 initargs_found = true;
354 static void __init setup_boot_config(const char *cmdline)
356 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
364 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
365 parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
368 if (!bootconfig_found)
374 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
375 if (memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
378 hdr = (u32 *)(data - 8);
382 if (size >= XBC_DATA_MAX) {
383 pr_err("bootconfig size %d greater than max size %d\n",
388 data = ((void *)hdr) - size;
389 if ((unsigned long)data < initrd_start)
392 if (boot_config_checksum((unsigned char *)data, size) != csum) {
393 pr_err("bootconfig checksum failed\n");
397 copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
399 pr_err("Failed to allocate memory for bootconfig\n");
403 memcpy(copy, data, size);
406 ret = xbc_init(copy, &msg, &pos);
409 pr_err("Failed to init bootconfig: %s.\n", msg);
411 pr_err("Failed to parse bootconfig: %s at %d.\n",
414 pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
415 /* keys starting with "kernel." are passed via cmdline */
416 extra_command_line = xbc_make_cmdline("kernel");
417 /* Also, "init." keys are init arguments */
418 extra_init_args = xbc_make_cmdline("init");
422 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
425 #define setup_boot_config(cmdline) do { } while (0)
427 static int __init warn_bootconfig(char *str)
429 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOTCONFIG is not set.\n");
432 early_param("bootconfig", warn_bootconfig);
436 /* Change NUL term back to "=", to make "param" the whole string. */
437 static void __init repair_env_string(char *param, char *val)
440 /* param=val or param="val"? */
441 if (val == param+strlen(param)+1)
443 else if (val == param+strlen(param)+2) {
445 memmove(val-1, val, strlen(val)+1);
451 /* Anything after -- gets handed straight to init. */
452 static int __init set_init_arg(char *param, char *val,
453 const char *unused, void *arg)
460 repair_env_string(param, val);
462 for (i = 0; argv_init[i]; i++) {
463 if (i == MAX_INIT_ARGS) {
464 panic_later = "init";
469 argv_init[i] = param;
474 * Unknown boot options get handed to init, unless they look like
475 * unused parameters (modprobe will find them in /proc/cmdline).
477 static int __init unknown_bootoption(char *param, char *val,
478 const char *unused, void *arg)
480 size_t len = strlen(param);
482 repair_env_string(param, val);
484 /* Handle obsolete-style parameters */
485 if (obsolete_checksetup(param))
488 /* Unused module parameter. */
489 if (strnchr(param, len, '.'))
496 /* Environment option */
498 for (i = 0; envp_init[i]; i++) {
499 if (i == MAX_INIT_ENVS) {
503 if (!strncmp(param, envp_init[i], len+1))
506 envp_init[i] = param;
508 /* Command line option */
510 for (i = 0; argv_init[i]; i++) {
511 if (i == MAX_INIT_ARGS) {
512 panic_later = "init";
516 argv_init[i] = param;
521 static int __init init_setup(char *str)
525 execute_command = str;
527 * In case LILO is going to boot us with default command line,
528 * it prepends "auto" before the whole cmdline which makes
529 * the shell think it should execute a script with such name.
530 * So we ignore all arguments entered _before_ init=... [MJ]
532 for (i = 1; i < MAX_INIT_ARGS; i++)
536 __setup("init=", init_setup);
538 static int __init rdinit_setup(char *str)
542 ramdisk_execute_command = str;
543 /* See "auto" comment in init_setup */
544 for (i = 1; i < MAX_INIT_ARGS; i++)
548 __setup("rdinit=", rdinit_setup);
551 static const unsigned int setup_max_cpus = NR_CPUS;
552 static inline void setup_nr_cpu_ids(void) { }
553 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
557 * We need to store the untouched command line for future reference.
558 * We also need to store the touched command line since the parameter
559 * parsing is performed in place, and we should allow a component to
560 * store reference of name/value for future reference.
562 static void __init setup_command_line(char *command_line)
564 size_t len, xlen = 0, ilen = 0;
566 if (extra_command_line)
567 xlen = strlen(extra_command_line);
569 ilen = strlen(extra_init_args) + 4; /* for " -- " */
571 len = xlen + strlen(boot_command_line) + 1;
573 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
574 if (!saved_command_line)
575 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
577 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
578 if (!static_command_line)
579 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
583 * We have to put extra_command_line before boot command
584 * lines because there could be dashes (separator of init
585 * command line) in the command lines.
587 strcpy(saved_command_line, extra_command_line);
588 strcpy(static_command_line, extra_command_line);
590 strcpy(saved_command_line + xlen, boot_command_line);
591 strcpy(static_command_line + xlen, command_line);
595 * Append supplemental init boot args to saved_command_line
596 * so that user can check what command line options passed
599 len = strlen(saved_command_line);
600 if (initargs_found) {
601 saved_command_line[len++] = ' ';
603 strcpy(saved_command_line + len, " -- ");
607 strcpy(saved_command_line + len, extra_init_args);
612 * We need to finalize in a non-__init function or else race conditions
613 * between the root thread and the init thread may cause start_kernel to
614 * be reaped by free_initmem before the root thread has proceeded to
617 * gcc-3.4 accidentally inlines this function, so use noinline.
620 static __initdata DECLARE_COMPLETION(kthreadd_done);
622 noinline void __ref rest_init(void)
624 struct task_struct *tsk;
627 rcu_scheduler_starting();
629 * We need to spawn init first so that it obtains pid 1, however
630 * the init task will end up wanting to create kthreads, which, if
631 * we schedule it before we create kthreadd, will OOPS.
633 pid = kernel_thread(kernel_init, NULL, CLONE_FS);
635 * Pin init on the boot CPU. Task migration is not properly working
636 * until sched_init_smp() has been run. It will set the allowed
637 * CPUs for init to the non isolated CPUs.
640 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
641 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
644 numa_default_policy();
645 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
647 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
651 * Enable might_sleep() and smp_processor_id() checks.
652 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
653 * kernel_thread() would trigger might_sleep() splats. With
654 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
655 * already, but it's stuck on the kthreadd_done completion.
657 system_state = SYSTEM_SCHEDULING;
659 complete(&kthreadd_done);
662 * The boot idle thread must execute schedule()
663 * at least once to get things moving:
665 schedule_preempt_disabled();
666 /* Call into cpu_idle with preempt disabled */
667 cpu_startup_entry(CPUHP_ONLINE);
670 /* Check for early params. */
671 static int __init do_early_param(char *param, char *val,
672 const char *unused, void *arg)
674 const struct obs_kernel_param *p;
676 for (p = __setup_start; p < __setup_end; p++) {
677 if ((p->early && parameq(param, p->str)) ||
678 (strcmp(param, "console") == 0 &&
679 strcmp(p->str, "earlycon") == 0)
681 if (p->setup_func(val) != 0)
682 pr_warn("Malformed early option '%s'\n", param);
685 /* We accept everything at this stage. */
689 void __init parse_early_options(char *cmdline)
691 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
695 /* Arch code calls this early on, or if not, just before other parsing. */
696 void __init parse_early_param(void)
698 static int done __initdata;
699 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
704 /* All fall through to do_early_param. */
705 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
706 parse_early_options(tmp_cmdline);
710 void __init __weak arch_post_acpi_subsys_init(void) { }
712 void __init __weak smp_setup_processor_id(void)
716 # if THREAD_SIZE >= PAGE_SIZE
717 void __init __weak thread_stack_cache_init(void)
722 void __init __weak mem_encrypt_init(void) { }
724 void __init __weak poking_init(void) { }
726 void __init __weak pgtable_cache_init(void) { }
729 core_param(initcall_debug, initcall_debug, bool, 0644);
731 #ifdef TRACEPOINTS_ENABLED
732 static void __init initcall_debug_enable(void);
734 static inline void initcall_debug_enable(void)
739 /* Report memory auto-initialization states for this boot. */
740 static void __init report_meminit(void)
744 if (IS_ENABLED(CONFIG_INIT_STACK_ALL))
746 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
748 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
750 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
755 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
756 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
757 want_init_on_free() ? "on" : "off");
758 if (want_init_on_free())
759 pr_info("mem auto-init: clearing system memory may take some time...\n");
763 * Set up kernel memory allocators
765 static void __init mm_init(void)
768 * page_ext requires contiguous pages,
769 * bigger than MAX_ORDER unless SPARSEMEM.
771 page_ext_init_flatmem();
772 init_debug_pagealloc();
778 debug_objects_mem_init();
781 /* Should be run before the first non-init thread is created */
783 /* Should be run after espfix64 is set up. */
787 void __init __weak arch_call_rest_init(void)
792 asmlinkage __visible void __init start_kernel(void)
797 set_task_stack_end_magic(&init_task);
798 smp_setup_processor_id();
799 debug_objects_early_init();
804 early_boot_irqs_disabled = true;
807 * Interrupts are still disabled. Do necessary setups, then
812 pr_notice("%s", linux_banner);
813 early_security_init();
814 setup_arch(&command_line);
815 setup_boot_config(command_line);
816 setup_command_line(command_line);
818 setup_per_cpu_areas();
819 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
820 boot_cpu_hotplug_init();
822 build_all_zonelists(NULL);
825 pr_notice("Kernel command line: %s\n", saved_command_line);
826 /* parameters may set static keys */
829 after_dashes = parse_args("Booting kernel",
830 static_command_line, __start___param,
831 __stop___param - __start___param,
832 -1, -1, NULL, &unknown_bootoption);
833 if (!IS_ERR_OR_NULL(after_dashes))
834 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
837 parse_args("Setting extra init args", extra_init_args,
838 NULL, 0, -1, -1, NULL, set_init_arg);
841 * These use large bootmem allocations and must precede
845 vfs_caches_init_early();
852 /* trace_printk can be enabled here */
856 * Set up the scheduler prior starting any interrupts (such as the
857 * timer interrupt). Full topology setup happens at smp_init()
858 * time - but meanwhile we still have a functioning scheduler.
862 * Disable preemption - early bootup scheduling is extremely
863 * fragile until we cpu_idle() for the first time.
866 if (WARN(!irqs_disabled(),
867 "Interrupts were enabled *very* early, fixing it\n"))
872 * Set up housekeeping before setting up workqueues to allow the unbound
873 * workqueue to take non-housekeeping into account.
878 * Allow workqueue creation and work item queueing/cancelling
879 * early. Work item execution depends on kthreads and starts after
882 workqueue_init_early();
886 /* Trace events are available after this */
890 initcall_debug_enable();
892 context_tracking_init();
893 /* init some links before init_ISA_irqs() */
904 * For best initial stack canary entropy, prepare it after:
905 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
906 * - timekeeping_init() for ktime entropy used in rand_initialize()
907 * - rand_initialize() to get any arch-specific entropy like RDRAND
908 * - add_latent_entropy() to get any latent entropy
909 * - adding command line entropy
912 add_latent_entropy();
913 add_device_randomness(command_line, strlen(command_line));
914 boot_init_stack_canary();
919 call_function_init();
920 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
922 early_boot_irqs_disabled = false;
925 kmem_cache_init_late();
928 * HACK ALERT! This is early. We're enabling the console before
929 * we've done PCI setups etc, and console_init() must be aware of
930 * this. But we do want output early, in case something goes wrong.
934 panic("Too many boot %s vars at `%s'", panic_later,
940 * Need to run this when irqs are enabled, because it wants
941 * to self-test [hard/soft]-irqs on/off lock inversion bugs
947 * This needs to be called before any devices perform DMA
948 * operations that might use the SWIOTLB bounce buffers. It will
949 * mark the bounce buffers as decrypted so that their usage will
950 * not cause "plain-text" data to be decrypted when accessed.
954 #ifdef CONFIG_BLK_DEV_INITRD
955 if (initrd_start && !initrd_below_start_ok &&
956 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
957 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
958 page_to_pfn(virt_to_page((void *)initrd_start)),
963 setup_per_cpu_pageset();
973 if (efi_enabled(EFI_RUNTIME_SERVICES))
974 efi_enter_virtual_mode();
976 thread_stack_cache_init();
993 taskstats_init_early();
999 acpi_subsystem_init();
1000 arch_post_acpi_subsys_init();
1004 /* Do the rest non-__init'ed, we're now alive */
1005 arch_call_rest_init();
1008 /* Call all constructor functions linked into the kernel. */
1009 static void __init do_ctors(void)
1011 #ifdef CONFIG_CONSTRUCTORS
1012 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1014 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1019 #ifdef CONFIG_KALLSYMS
1020 struct blacklist_entry {
1021 struct list_head next;
1025 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1027 static int __init initcall_blacklist(char *str)
1030 struct blacklist_entry *entry;
1032 /* str argument is a comma-separated list of functions */
1034 str_entry = strsep(&str, ",");
1036 pr_debug("blacklisting initcall %s\n", str_entry);
1037 entry = memblock_alloc(sizeof(*entry),
1040 panic("%s: Failed to allocate %zu bytes\n",
1041 __func__, sizeof(*entry));
1042 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1045 panic("%s: Failed to allocate %zu bytes\n",
1046 __func__, strlen(str_entry) + 1);
1047 strcpy(entry->buf, str_entry);
1048 list_add(&entry->next, &blacklisted_initcalls);
1050 } while (str_entry);
1055 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1057 struct blacklist_entry *entry;
1058 char fn_name[KSYM_SYMBOL_LEN];
1061 if (list_empty(&blacklisted_initcalls))
1064 addr = (unsigned long) dereference_function_descriptor(fn);
1065 sprint_symbol_no_offset(fn_name, addr);
1068 * fn will be "function_name [module_name]" where [module_name] is not
1069 * displayed for built-in init functions. Strip off the [module_name].
1071 strreplace(fn_name, ' ', '\0');
1073 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1074 if (!strcmp(fn_name, entry->buf)) {
1075 pr_debug("initcall %s blacklisted\n", fn_name);
1083 static int __init initcall_blacklist(char *str)
1085 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1089 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1094 __setup("initcall_blacklist=", initcall_blacklist);
1096 static __init_or_module void
1097 trace_initcall_start_cb(void *data, initcall_t fn)
1099 ktime_t *calltime = (ktime_t *)data;
1101 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1102 *calltime = ktime_get();
1105 static __init_or_module void
1106 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1108 ktime_t *calltime = (ktime_t *)data;
1109 ktime_t delta, rettime;
1110 unsigned long long duration;
1112 rettime = ktime_get();
1113 delta = ktime_sub(rettime, *calltime);
1114 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1115 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1119 static ktime_t initcall_calltime;
1121 #ifdef TRACEPOINTS_ENABLED
1122 static void __init initcall_debug_enable(void)
1126 ret = register_trace_initcall_start(trace_initcall_start_cb,
1127 &initcall_calltime);
1128 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1129 &initcall_calltime);
1130 WARN(ret, "Failed to register initcall tracepoints\n");
1132 # define do_trace_initcall_start trace_initcall_start
1133 # define do_trace_initcall_finish trace_initcall_finish
1135 static inline void do_trace_initcall_start(initcall_t fn)
1137 if (!initcall_debug)
1139 trace_initcall_start_cb(&initcall_calltime, fn);
1141 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1143 if (!initcall_debug)
1145 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1147 #endif /* !TRACEPOINTS_ENABLED */
1149 int __init_or_module do_one_initcall(initcall_t fn)
1151 int count = preempt_count();
1155 if (initcall_blacklisted(fn))
1158 do_trace_initcall_start(fn);
1160 do_trace_initcall_finish(fn, ret);
1164 if (preempt_count() != count) {
1165 sprintf(msgbuf, "preemption imbalance ");
1166 preempt_count_set(count);
1168 if (irqs_disabled()) {
1169 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1172 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1174 add_latent_entropy();
1179 extern initcall_entry_t __initcall_start[];
1180 extern initcall_entry_t __initcall0_start[];
1181 extern initcall_entry_t __initcall1_start[];
1182 extern initcall_entry_t __initcall2_start[];
1183 extern initcall_entry_t __initcall3_start[];
1184 extern initcall_entry_t __initcall4_start[];
1185 extern initcall_entry_t __initcall5_start[];
1186 extern initcall_entry_t __initcall6_start[];
1187 extern initcall_entry_t __initcall7_start[];
1188 extern initcall_entry_t __initcall_end[];
1190 static initcall_entry_t *initcall_levels[] __initdata = {
1202 /* Keep these in sync with initcalls in include/linux/init.h */
1203 static const char *initcall_level_names[] __initdata = {
1214 static int __init ignore_unknown_bootoption(char *param, char *val,
1215 const char *unused, void *arg)
1220 static void __init do_initcall_level(int level, char *command_line)
1222 initcall_entry_t *fn;
1224 parse_args(initcall_level_names[level],
1225 command_line, __start___param,
1226 __stop___param - __start___param,
1228 NULL, ignore_unknown_bootoption);
1230 trace_initcall_level(initcall_level_names[level]);
1231 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1232 do_one_initcall(initcall_from_entry(fn));
1235 static void __init do_initcalls(void)
1238 size_t len = strlen(saved_command_line) + 1;
1241 command_line = kzalloc(len, GFP_KERNEL);
1243 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1245 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1246 /* Parser modifies command_line, restore it each time */
1247 strcpy(command_line, saved_command_line);
1248 do_initcall_level(level, command_line);
1251 kfree(command_line);
1255 * Ok, the machine is now initialized. None of the devices
1256 * have been touched yet, but the CPU subsystem is up and
1257 * running, and memory and process management works.
1259 * Now we can finally start doing some real work..
1261 static void __init do_basic_setup(void)
1267 usermodehelper_enable();
1271 static void __init do_pre_smp_initcalls(void)
1273 initcall_entry_t *fn;
1275 trace_initcall_level("early");
1276 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1277 do_one_initcall(initcall_from_entry(fn));
1280 static int run_init_process(const char *init_filename)
1282 const char *const *p;
1284 argv_init[0] = init_filename;
1285 pr_info("Run %s as init process\n", init_filename);
1286 pr_debug(" with arguments:\n");
1287 for (p = argv_init; *p; p++)
1288 pr_debug(" %s\n", *p);
1289 pr_debug(" with environment:\n");
1290 for (p = envp_init; *p; p++)
1291 pr_debug(" %s\n", *p);
1292 return do_execve(getname_kernel(init_filename),
1293 (const char __user *const __user *)argv_init,
1294 (const char __user *const __user *)envp_init);
1297 static int try_to_run_init_process(const char *init_filename)
1301 ret = run_init_process(init_filename);
1303 if (ret && ret != -ENOENT) {
1304 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1305 init_filename, ret);
1311 static noinline void __init kernel_init_freeable(void);
1313 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1314 bool rodata_enabled __ro_after_init = true;
1315 static int __init set_debug_rodata(char *str)
1317 return strtobool(str, &rodata_enabled);
1319 __setup("rodata=", set_debug_rodata);
1322 #ifdef CONFIG_STRICT_KERNEL_RWX
1323 static void mark_readonly(void)
1325 if (rodata_enabled) {
1327 * load_module() results in W+X mappings, which are cleaned
1328 * up with call_rcu(). Let's make sure that queued work is
1329 * flushed so that we don't hit false positives looking for
1330 * insecure pages which are W+X.
1336 pr_info("Kernel memory protection disabled.\n");
1338 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1339 static inline void mark_readonly(void)
1341 pr_warn("Kernel memory protection not selected by kernel config.\n");
1344 static inline void mark_readonly(void)
1346 pr_warn("This architecture does not have kernel memory protection.\n");
1350 void __weak free_initmem(void)
1352 free_initmem_default(POISON_FREE_INITMEM);
1355 static int __ref kernel_init(void *unused)
1359 kernel_init_freeable();
1360 /* need to finish all async __init code before freeing the memory */
1361 async_synchronize_full();
1362 ftrace_free_init_mem();
1367 * Kernel mappings are now finalized - update the userspace page-table
1372 system_state = SYSTEM_RUNNING;
1373 numa_default_policy();
1375 rcu_end_inkernel_boot();
1377 if (ramdisk_execute_command) {
1378 ret = run_init_process(ramdisk_execute_command);
1381 pr_err("Failed to execute %s (error %d)\n",
1382 ramdisk_execute_command, ret);
1386 * We try each of these until one succeeds.
1388 * The Bourne shell can be used instead of init if we are
1389 * trying to recover a really broken machine.
1391 if (execute_command) {
1392 ret = run_init_process(execute_command);
1395 panic("Requested init %s failed (error %d).",
1396 execute_command, ret);
1398 if (!try_to_run_init_process("/sbin/init") ||
1399 !try_to_run_init_process("/etc/init") ||
1400 !try_to_run_init_process("/bin/init") ||
1401 !try_to_run_init_process("/bin/sh"))
1404 panic("No working init found. Try passing init= option to kernel. "
1405 "See Linux Documentation/admin-guide/init.rst for guidance.");
1408 void console_on_rootfs(void)
1410 /* Open the /dev/console as stdin, this should never fail */
1411 if (ksys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1412 pr_err("Warning: unable to open an initial console.\n");
1414 /* create stdout/stderr */
1419 static noinline void __init kernel_init_freeable(void)
1422 * Wait until kthreadd is all set-up.
1424 wait_for_completion(&kthreadd_done);
1426 /* Now the scheduler is fully set up and can do blocking allocations */
1427 gfp_allowed_mask = __GFP_BITS_MASK;
1430 * init can allocate pages on any node
1432 set_mems_allowed(node_states[N_MEMORY]);
1434 cad_pid = task_pid(current);
1436 smp_prepare_cpus(setup_max_cpus);
1440 init_mm_internals();
1442 do_pre_smp_initcalls();
1443 lockup_detector_init();
1448 page_alloc_init_late();
1449 /* Initialize page ext after all struct pages are initialized. */
1454 console_on_rootfs();
1457 * check if there is an early userspace init. If yes, let it do all
1461 if (!ramdisk_execute_command)
1462 ramdisk_execute_command = "/init";
1464 if (ksys_access((const char __user *)
1465 ramdisk_execute_command, 0) != 0) {
1466 ramdisk_execute_command = NULL;
1467 prepare_namespace();
1471 * Ok, we have completed the initial bootup, and
1472 * we're essentially up and running. Get rid of the
1473 * initmem segments and start the user-mode stuff..
1475 * rootfs is available now, try loading the public keys
1476 * and default modules
1479 integrity_load_keys();