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)
277 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
279 * Since Grub may align the size of initrd to 4, we must
280 * check the preceding 3 bytes as well.
282 for (i = 0; i < 4; i++) {
283 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
290 hdr = (u32 *)(data - 8);
294 data = ((void *)hdr) - size;
295 if ((unsigned long)data < initrd_start) {
296 pr_err("bootconfig size %d is greater than initrd size %ld\n",
297 size, initrd_end - initrd_start);
301 /* Remove bootconfig from initramfs/initrd */
302 initrd_end = (unsigned long)data;
311 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
317 #ifdef CONFIG_BOOT_CONFIG
319 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
321 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
323 static int __init xbc_snprint_cmdline(char *buf, size_t size,
324 struct xbc_node *root)
326 struct xbc_node *knode, *vnode;
327 char *end = buf + size;
331 xbc_node_for_each_key_value(root, knode, val) {
332 ret = xbc_node_compose_key_after(root, knode,
333 xbc_namebuf, XBC_KEYLEN_MAX);
337 vnode = xbc_node_get_child(knode);
339 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
345 xbc_array_for_each_value(vnode, val) {
346 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
354 return buf - (end - size);
358 /* Make an extra command line under given key word */
359 static char * __init xbc_make_cmdline(const char *key)
361 struct xbc_node *root;
365 root = xbc_find_node(key);
369 /* Count required buffer size */
370 len = xbc_snprint_cmdline(NULL, 0, root);
374 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
376 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
380 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
381 if (ret < 0 || ret > len) {
382 pr_err("Failed to print extra kernel cmdline.\n");
389 static u32 boot_config_checksum(unsigned char *p, u32 size)
399 static int __init bootconfig_params(char *param, char *val,
400 const char *unused, void *arg)
402 if (strcmp(param, "bootconfig") == 0) {
403 bootconfig_found = true;
408 static void __init setup_boot_config(const char *cmdline)
410 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
414 char *data, *copy, *err;
417 /* Cut out the bootconfig data even if we have no bootconfig option */
418 data = get_boot_config_from_initrd(&size, &csum);
420 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
421 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
424 if (IS_ERR(err) || !bootconfig_found)
427 /* parse_args() stops at '--' and returns an address */
429 initargs_found = true;
432 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
436 if (size >= XBC_DATA_MAX) {
437 pr_err("bootconfig size %d greater than max size %d\n",
442 if (boot_config_checksum((unsigned char *)data, size) != csum) {
443 pr_err("bootconfig checksum failed\n");
447 copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
449 pr_err("Failed to allocate memory for bootconfig\n");
453 memcpy(copy, data, size);
456 ret = xbc_init(copy, &msg, &pos);
459 pr_err("Failed to init bootconfig: %s.\n", msg);
461 pr_err("Failed to parse bootconfig: %s at %d.\n",
464 pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
465 /* keys starting with "kernel." are passed via cmdline */
466 extra_command_line = xbc_make_cmdline("kernel");
467 /* Also, "init." keys are init arguments */
468 extra_init_args = xbc_make_cmdline("init");
475 static void __init setup_boot_config(const char *cmdline)
477 /* Remove bootconfig data from initrd */
478 get_boot_config_from_initrd(NULL, NULL);
481 static int __init warn_bootconfig(char *str)
483 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
486 early_param("bootconfig", warn_bootconfig);
490 /* Change NUL term back to "=", to make "param" the whole string. */
491 static void __init repair_env_string(char *param, char *val)
494 /* param=val or param="val"? */
495 if (val == param+strlen(param)+1)
497 else if (val == param+strlen(param)+2) {
499 memmove(val-1, val, strlen(val)+1);
505 /* Anything after -- gets handed straight to init. */
506 static int __init set_init_arg(char *param, char *val,
507 const char *unused, void *arg)
514 repair_env_string(param, val);
516 for (i = 0; argv_init[i]; i++) {
517 if (i == MAX_INIT_ARGS) {
518 panic_later = "init";
523 argv_init[i] = param;
528 * Unknown boot options get handed to init, unless they look like
529 * unused parameters (modprobe will find them in /proc/cmdline).
531 static int __init unknown_bootoption(char *param, char *val,
532 const char *unused, void *arg)
534 size_t len = strlen(param);
536 repair_env_string(param, val);
538 /* Handle obsolete-style parameters */
539 if (obsolete_checksetup(param))
542 /* Unused module parameter. */
543 if (strnchr(param, len, '.'))
550 /* Environment option */
552 for (i = 0; envp_init[i]; i++) {
553 if (i == MAX_INIT_ENVS) {
557 if (!strncmp(param, envp_init[i], len+1))
560 envp_init[i] = param;
562 /* Command line option */
564 for (i = 0; argv_init[i]; i++) {
565 if (i == MAX_INIT_ARGS) {
566 panic_later = "init";
570 argv_init[i] = param;
575 static int __init init_setup(char *str)
579 execute_command = str;
581 * In case LILO is going to boot us with default command line,
582 * it prepends "auto" before the whole cmdline which makes
583 * the shell think it should execute a script with such name.
584 * So we ignore all arguments entered _before_ init=... [MJ]
586 for (i = 1; i < MAX_INIT_ARGS; i++)
590 __setup("init=", init_setup);
592 static int __init rdinit_setup(char *str)
596 ramdisk_execute_command = str;
597 /* See "auto" comment in init_setup */
598 for (i = 1; i < MAX_INIT_ARGS; i++)
602 __setup("rdinit=", rdinit_setup);
605 static const unsigned int setup_max_cpus = NR_CPUS;
606 static inline void setup_nr_cpu_ids(void) { }
607 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
611 * We need to store the untouched command line for future reference.
612 * We also need to store the touched command line since the parameter
613 * parsing is performed in place, and we should allow a component to
614 * store reference of name/value for future reference.
616 static void __init setup_command_line(char *command_line)
618 size_t len, xlen = 0, ilen = 0;
620 if (extra_command_line)
621 xlen = strlen(extra_command_line);
623 ilen = strlen(extra_init_args) + 4; /* for " -- " */
625 len = xlen + strlen(boot_command_line) + 1;
627 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
628 if (!saved_command_line)
629 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
631 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
632 if (!static_command_line)
633 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
637 * We have to put extra_command_line before boot command
638 * lines because there could be dashes (separator of init
639 * command line) in the command lines.
641 strcpy(saved_command_line, extra_command_line);
642 strcpy(static_command_line, extra_command_line);
644 strcpy(saved_command_line + xlen, boot_command_line);
645 strcpy(static_command_line + xlen, command_line);
649 * Append supplemental init boot args to saved_command_line
650 * so that user can check what command line options passed
653 len = strlen(saved_command_line);
654 if (initargs_found) {
655 saved_command_line[len++] = ' ';
657 strcpy(saved_command_line + len, " -- ");
661 strcpy(saved_command_line + len, extra_init_args);
666 * We need to finalize in a non-__init function or else race conditions
667 * between the root thread and the init thread may cause start_kernel to
668 * be reaped by free_initmem before the root thread has proceeded to
671 * gcc-3.4 accidentally inlines this function, so use noinline.
674 static __initdata DECLARE_COMPLETION(kthreadd_done);
676 noinline void __ref rest_init(void)
678 struct task_struct *tsk;
681 rcu_scheduler_starting();
683 * We need to spawn init first so that it obtains pid 1, however
684 * the init task will end up wanting to create kthreads, which, if
685 * we schedule it before we create kthreadd, will OOPS.
687 pid = kernel_thread(kernel_init, NULL, CLONE_FS);
689 * Pin init on the boot CPU. Task migration is not properly working
690 * until sched_init_smp() has been run. It will set the allowed
691 * CPUs for init to the non isolated CPUs.
694 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
695 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
698 numa_default_policy();
699 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
701 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
705 * Enable might_sleep() and smp_processor_id() checks.
706 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
707 * kernel_thread() would trigger might_sleep() splats. With
708 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
709 * already, but it's stuck on the kthreadd_done completion.
711 system_state = SYSTEM_SCHEDULING;
713 complete(&kthreadd_done);
716 * The boot idle thread must execute schedule()
717 * at least once to get things moving:
719 schedule_preempt_disabled();
720 /* Call into cpu_idle with preempt disabled */
721 cpu_startup_entry(CPUHP_ONLINE);
724 /* Check for early params. */
725 static int __init do_early_param(char *param, char *val,
726 const char *unused, void *arg)
728 const struct obs_kernel_param *p;
730 for (p = __setup_start; p < __setup_end; p++) {
731 if ((p->early && parameq(param, p->str)) ||
732 (strcmp(param, "console") == 0 &&
733 strcmp(p->str, "earlycon") == 0)
735 if (p->setup_func(val) != 0)
736 pr_warn("Malformed early option '%s'\n", param);
739 /* We accept everything at this stage. */
743 void __init parse_early_options(char *cmdline)
745 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
749 /* Arch code calls this early on, or if not, just before other parsing. */
750 void __init parse_early_param(void)
752 static int done __initdata;
753 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
758 /* All fall through to do_early_param. */
759 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
760 parse_early_options(tmp_cmdline);
764 void __init __weak arch_post_acpi_subsys_init(void) { }
766 void __init __weak smp_setup_processor_id(void)
770 # if THREAD_SIZE >= PAGE_SIZE
771 void __init __weak thread_stack_cache_init(void)
776 void __init __weak mem_encrypt_init(void) { }
778 void __init __weak poking_init(void) { }
780 void __init __weak pgtable_cache_init(void) { }
783 core_param(initcall_debug, initcall_debug, bool, 0644);
785 #ifdef TRACEPOINTS_ENABLED
786 static void __init initcall_debug_enable(void);
788 static inline void initcall_debug_enable(void)
793 /* Report memory auto-initialization states for this boot. */
794 static void __init report_meminit(void)
798 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
799 stack = "all(pattern)";
800 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
802 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
803 stack = "byref_all(zero)";
804 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
805 stack = "byref(zero)";
806 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
807 stack = "__user(zero)";
811 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
812 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
813 want_init_on_free() ? "on" : "off");
814 if (want_init_on_free())
815 pr_info("mem auto-init: clearing system memory may take some time...\n");
819 * Set up kernel memory allocators
821 static void __init mm_init(void)
824 * page_ext requires contiguous pages,
825 * bigger than MAX_ORDER unless SPARSEMEM.
827 page_ext_init_flatmem();
828 init_debug_pagealloc();
834 debug_objects_mem_init();
837 /* Should be run before the first non-init thread is created */
839 /* Should be run after espfix64 is set up. */
843 void __init __weak arch_call_rest_init(void)
848 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
853 set_task_stack_end_magic(&init_task);
854 smp_setup_processor_id();
855 debug_objects_early_init();
860 early_boot_irqs_disabled = true;
863 * Interrupts are still disabled. Do necessary setups, then
868 pr_notice("%s", linux_banner);
869 early_security_init();
870 setup_arch(&command_line);
871 setup_boot_config(command_line);
872 setup_command_line(command_line);
874 setup_per_cpu_areas();
875 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
876 boot_cpu_hotplug_init();
878 build_all_zonelists(NULL);
881 pr_notice("Kernel command line: %s\n", saved_command_line);
882 /* parameters may set static keys */
885 after_dashes = parse_args("Booting kernel",
886 static_command_line, __start___param,
887 __stop___param - __start___param,
888 -1, -1, NULL, &unknown_bootoption);
889 if (!IS_ERR_OR_NULL(after_dashes))
890 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
893 parse_args("Setting extra init args", extra_init_args,
894 NULL, 0, -1, -1, NULL, set_init_arg);
897 * These use large bootmem allocations and must precede
901 vfs_caches_init_early();
908 /* trace_printk can be enabled here */
912 * Set up the scheduler prior starting any interrupts (such as the
913 * timer interrupt). Full topology setup happens at smp_init()
914 * time - but meanwhile we still have a functioning scheduler.
918 * Disable preemption - early bootup scheduling is extremely
919 * fragile until we cpu_idle() for the first time.
922 if (WARN(!irqs_disabled(),
923 "Interrupts were enabled *very* early, fixing it\n"))
928 * Set up housekeeping before setting up workqueues to allow the unbound
929 * workqueue to take non-housekeeping into account.
934 * Allow workqueue creation and work item queueing/cancelling
935 * early. Work item execution depends on kthreads and starts after
938 workqueue_init_early();
942 /* Trace events are available after this */
946 initcall_debug_enable();
948 context_tracking_init();
949 /* init some links before init_ISA_irqs() */
960 * For best initial stack canary entropy, prepare it after:
961 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
962 * - timekeeping_init() for ktime entropy used in rand_initialize()
963 * - rand_initialize() to get any arch-specific entropy like RDRAND
964 * - add_latent_entropy() to get any latent entropy
965 * - adding command line entropy
968 add_latent_entropy();
969 add_device_randomness(command_line, strlen(command_line));
970 boot_init_stack_canary();
975 call_function_init();
976 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
978 early_boot_irqs_disabled = false;
981 kmem_cache_init_late();
984 * HACK ALERT! This is early. We're enabling the console before
985 * we've done PCI setups etc, and console_init() must be aware of
986 * this. But we do want output early, in case something goes wrong.
990 panic("Too many boot %s vars at `%s'", panic_later,
996 * Need to run this when irqs are enabled, because it wants
997 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1003 * This needs to be called before any devices perform DMA
1004 * operations that might use the SWIOTLB bounce buffers. It will
1005 * mark the bounce buffers as decrypted so that their usage will
1006 * not cause "plain-text" data to be decrypted when accessed.
1010 #ifdef CONFIG_BLK_DEV_INITRD
1011 if (initrd_start && !initrd_below_start_ok &&
1012 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1013 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1014 page_to_pfn(virt_to_page((void *)initrd_start)),
1019 setup_per_cpu_pageset();
1029 if (efi_enabled(EFI_RUNTIME_SERVICES))
1030 efi_enter_virtual_mode();
1032 thread_stack_cache_init();
1049 taskstats_init_early();
1055 acpi_subsystem_init();
1056 arch_post_acpi_subsys_init();
1060 /* Do the rest non-__init'ed, we're now alive */
1061 arch_call_rest_init();
1063 prevent_tail_call_optimization();
1066 /* Call all constructor functions linked into the kernel. */
1067 static void __init do_ctors(void)
1069 #ifdef CONFIG_CONSTRUCTORS
1070 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1072 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1077 #ifdef CONFIG_KALLSYMS
1078 struct blacklist_entry {
1079 struct list_head next;
1083 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1085 static int __init initcall_blacklist(char *str)
1088 struct blacklist_entry *entry;
1090 /* str argument is a comma-separated list of functions */
1092 str_entry = strsep(&str, ",");
1094 pr_debug("blacklisting initcall %s\n", str_entry);
1095 entry = memblock_alloc(sizeof(*entry),
1098 panic("%s: Failed to allocate %zu bytes\n",
1099 __func__, sizeof(*entry));
1100 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1103 panic("%s: Failed to allocate %zu bytes\n",
1104 __func__, strlen(str_entry) + 1);
1105 strcpy(entry->buf, str_entry);
1106 list_add(&entry->next, &blacklisted_initcalls);
1108 } while (str_entry);
1113 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1115 struct blacklist_entry *entry;
1116 char fn_name[KSYM_SYMBOL_LEN];
1119 if (list_empty(&blacklisted_initcalls))
1122 addr = (unsigned long) dereference_function_descriptor(fn);
1123 sprint_symbol_no_offset(fn_name, addr);
1126 * fn will be "function_name [module_name]" where [module_name] is not
1127 * displayed for built-in init functions. Strip off the [module_name].
1129 strreplace(fn_name, ' ', '\0');
1131 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1132 if (!strcmp(fn_name, entry->buf)) {
1133 pr_debug("initcall %s blacklisted\n", fn_name);
1141 static int __init initcall_blacklist(char *str)
1143 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1147 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1152 __setup("initcall_blacklist=", initcall_blacklist);
1154 static __init_or_module void
1155 trace_initcall_start_cb(void *data, initcall_t fn)
1157 ktime_t *calltime = (ktime_t *)data;
1159 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1160 *calltime = ktime_get();
1163 static __init_or_module void
1164 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1166 ktime_t *calltime = (ktime_t *)data;
1167 ktime_t delta, rettime;
1168 unsigned long long duration;
1170 rettime = ktime_get();
1171 delta = ktime_sub(rettime, *calltime);
1172 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1173 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1177 static ktime_t initcall_calltime;
1179 #ifdef TRACEPOINTS_ENABLED
1180 static void __init initcall_debug_enable(void)
1184 ret = register_trace_initcall_start(trace_initcall_start_cb,
1185 &initcall_calltime);
1186 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1187 &initcall_calltime);
1188 WARN(ret, "Failed to register initcall tracepoints\n");
1190 # define do_trace_initcall_start trace_initcall_start
1191 # define do_trace_initcall_finish trace_initcall_finish
1193 static inline void do_trace_initcall_start(initcall_t fn)
1195 if (!initcall_debug)
1197 trace_initcall_start_cb(&initcall_calltime, fn);
1199 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1201 if (!initcall_debug)
1203 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1205 #endif /* !TRACEPOINTS_ENABLED */
1207 int __init_or_module do_one_initcall(initcall_t fn)
1209 int count = preempt_count();
1213 if (initcall_blacklisted(fn))
1216 do_trace_initcall_start(fn);
1218 do_trace_initcall_finish(fn, ret);
1222 if (preempt_count() != count) {
1223 sprintf(msgbuf, "preemption imbalance ");
1224 preempt_count_set(count);
1226 if (irqs_disabled()) {
1227 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1230 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1232 add_latent_entropy();
1237 extern initcall_entry_t __initcall_start[];
1238 extern initcall_entry_t __initcall0_start[];
1239 extern initcall_entry_t __initcall1_start[];
1240 extern initcall_entry_t __initcall2_start[];
1241 extern initcall_entry_t __initcall3_start[];
1242 extern initcall_entry_t __initcall4_start[];
1243 extern initcall_entry_t __initcall5_start[];
1244 extern initcall_entry_t __initcall6_start[];
1245 extern initcall_entry_t __initcall7_start[];
1246 extern initcall_entry_t __initcall_end[];
1248 static initcall_entry_t *initcall_levels[] __initdata = {
1260 /* Keep these in sync with initcalls in include/linux/init.h */
1261 static const char *initcall_level_names[] __initdata = {
1272 static int __init ignore_unknown_bootoption(char *param, char *val,
1273 const char *unused, void *arg)
1278 static void __init do_initcall_level(int level, char *command_line)
1280 initcall_entry_t *fn;
1282 parse_args(initcall_level_names[level],
1283 command_line, __start___param,
1284 __stop___param - __start___param,
1286 NULL, ignore_unknown_bootoption);
1288 trace_initcall_level(initcall_level_names[level]);
1289 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1290 do_one_initcall(initcall_from_entry(fn));
1293 static void __init do_initcalls(void)
1296 size_t len = strlen(saved_command_line) + 1;
1299 command_line = kzalloc(len, GFP_KERNEL);
1301 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1303 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1304 /* Parser modifies command_line, restore it each time */
1305 strcpy(command_line, saved_command_line);
1306 do_initcall_level(level, command_line);
1309 kfree(command_line);
1313 * Ok, the machine is now initialized. None of the devices
1314 * have been touched yet, but the CPU subsystem is up and
1315 * running, and memory and process management works.
1317 * Now we can finally start doing some real work..
1319 static void __init do_basic_setup(void)
1325 usermodehelper_enable();
1329 static void __init do_pre_smp_initcalls(void)
1331 initcall_entry_t *fn;
1333 trace_initcall_level("early");
1334 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1335 do_one_initcall(initcall_from_entry(fn));
1338 static int run_init_process(const char *init_filename)
1340 const char *const *p;
1342 argv_init[0] = init_filename;
1343 pr_info("Run %s as init process\n", init_filename);
1344 pr_debug(" with arguments:\n");
1345 for (p = argv_init; *p; p++)
1346 pr_debug(" %s\n", *p);
1347 pr_debug(" with environment:\n");
1348 for (p = envp_init; *p; p++)
1349 pr_debug(" %s\n", *p);
1350 return kernel_execve(init_filename, argv_init, envp_init);
1353 static int try_to_run_init_process(const char *init_filename)
1357 ret = run_init_process(init_filename);
1359 if (ret && ret != -ENOENT) {
1360 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1361 init_filename, ret);
1367 static noinline void __init kernel_init_freeable(void);
1369 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1370 bool rodata_enabled __ro_after_init = true;
1371 static int __init set_debug_rodata(char *str)
1373 return strtobool(str, &rodata_enabled);
1375 __setup("rodata=", set_debug_rodata);
1378 #ifdef CONFIG_STRICT_KERNEL_RWX
1379 static void mark_readonly(void)
1381 if (rodata_enabled) {
1383 * load_module() results in W+X mappings, which are cleaned
1384 * up with call_rcu(). Let's make sure that queued work is
1385 * flushed so that we don't hit false positives looking for
1386 * insecure pages which are W+X.
1392 pr_info("Kernel memory protection disabled.\n");
1394 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1395 static inline void mark_readonly(void)
1397 pr_warn("Kernel memory protection not selected by kernel config.\n");
1400 static inline void mark_readonly(void)
1402 pr_warn("This architecture does not have kernel memory protection.\n");
1406 void __weak free_initmem(void)
1408 free_initmem_default(POISON_FREE_INITMEM);
1411 static int __ref kernel_init(void *unused)
1415 kernel_init_freeable();
1416 /* need to finish all async __init code before freeing the memory */
1417 async_synchronize_full();
1418 kprobe_free_init_mem();
1419 ftrace_free_init_mem();
1424 * Kernel mappings are now finalized - update the userspace page-table
1429 system_state = SYSTEM_RUNNING;
1430 numa_default_policy();
1432 rcu_end_inkernel_boot();
1436 if (ramdisk_execute_command) {
1437 ret = run_init_process(ramdisk_execute_command);
1440 pr_err("Failed to execute %s (error %d)\n",
1441 ramdisk_execute_command, ret);
1445 * We try each of these until one succeeds.
1447 * The Bourne shell can be used instead of init if we are
1448 * trying to recover a really broken machine.
1450 if (execute_command) {
1451 ret = run_init_process(execute_command);
1454 panic("Requested init %s failed (error %d).",
1455 execute_command, ret);
1458 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1459 ret = run_init_process(CONFIG_DEFAULT_INIT);
1461 pr_err("Default init %s failed (error %d)\n",
1462 CONFIG_DEFAULT_INIT, ret);
1467 if (!try_to_run_init_process("/sbin/init") ||
1468 !try_to_run_init_process("/etc/init") ||
1469 !try_to_run_init_process("/bin/init") ||
1470 !try_to_run_init_process("/bin/sh"))
1473 panic("No working init found. Try passing init= option to kernel. "
1474 "See Linux Documentation/admin-guide/init.rst for guidance.");
1477 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1478 void __init console_on_rootfs(void)
1480 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1483 pr_err("Warning: unable to open an initial console.\n");
1492 static noinline void __init kernel_init_freeable(void)
1495 * Wait until kthreadd is all set-up.
1497 wait_for_completion(&kthreadd_done);
1499 /* Now the scheduler is fully set up and can do blocking allocations */
1500 gfp_allowed_mask = __GFP_BITS_MASK;
1503 * init can allocate pages on any node
1505 set_mems_allowed(node_states[N_MEMORY]);
1507 cad_pid = task_pid(current);
1509 smp_prepare_cpus(setup_max_cpus);
1513 init_mm_internals();
1515 do_pre_smp_initcalls();
1516 lockup_detector_init();
1522 page_alloc_init_late();
1523 /* Initialize page ext after all struct pages are initialized. */
1528 kunit_run_all_tests();
1530 console_on_rootfs();
1533 * check if there is an early userspace init. If yes, let it do all
1536 if (init_eaccess(ramdisk_execute_command) != 0) {
1537 ramdisk_execute_command = NULL;
1538 prepare_namespace();
1542 * Ok, we have completed the initial bootup, and
1543 * we're essentially up and running. Get rid of the
1544 * initmem segments and start the user-mode stuff..
1546 * rootfs is available now, try loading the public keys
1547 * and default modules
1550 integrity_load_keys();