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/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/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>
98 #include <linux/init_syscalls.h>
101 #include <asm/bugs.h>
102 #include <asm/setup.h>
103 #include <asm/sections.h>
104 #include <asm/cacheflush.h>
106 #define CREATE_TRACE_POINTS
107 #include <trace/events/initcall.h>
109 #include <kunit/test.h>
111 static int kernel_init(void *);
113 extern void init_IRQ(void);
114 extern void radix_tree_init(void);
117 * Debug helper: via this flag we know that we are in 'early bootup code'
118 * where only the boot processor is running with IRQ disabled. This means
119 * two things - IRQ must not be enabled before the flag is cleared and some
120 * operations which are not allowed with IRQ disabled are allowed while the
123 bool early_boot_irqs_disabled __read_mostly
;
125 enum system_states system_state __read_mostly
;
126 EXPORT_SYMBOL(system_state
);
129 * Boot command-line arguments
131 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
132 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
134 extern void time_init(void);
135 /* Default late time init is NULL. archs can override this later. */
136 void (*__initdata late_time_init
)(void);
138 /* Untouched command line saved by arch-specific code. */
139 char __initdata boot_command_line
[COMMAND_LINE_SIZE
];
140 /* Untouched saved command line (eg. for /proc) */
141 char *saved_command_line
;
142 /* Command line for parameter parsing */
143 static char *static_command_line
;
144 /* Untouched extra command line */
145 static char *extra_command_line
;
146 /* Extra init arguments */
147 static char *extra_init_args
;
149 #ifdef CONFIG_BOOT_CONFIG
150 /* Is bootconfig on command line? */
151 static bool bootconfig_found
;
152 static bool initargs_found
;
154 # define bootconfig_found false
155 # define initargs_found false
158 static char *execute_command
;
159 static char *ramdisk_execute_command
= "/init";
162 * Used to generate warnings if static_key manipulation functions are used
163 * before jump_label_init is called.
165 bool static_key_initialized __read_mostly
;
166 EXPORT_SYMBOL_GPL(static_key_initialized
);
169 * If set, this is an indication to the drivers that reset the underlying
170 * device before going ahead with the initialization otherwise driver might
171 * rely on the BIOS and skip the reset operation.
173 * This is useful if kernel is booting in an unreliable environment.
174 * For ex. kdump situation where previous kernel has crashed, BIOS has been
175 * skipped and devices will be in unknown state.
177 unsigned int reset_devices
;
178 EXPORT_SYMBOL(reset_devices
);
180 static int __init
set_reset_devices(char *str
)
186 __setup("reset_devices", set_reset_devices
);
188 static const char *argv_init
[MAX_INIT_ARGS
+2] = { "init", NULL
, };
189 const char *envp_init
[MAX_INIT_ENVS
+2] = { "HOME=/", "TERM=linux", NULL
, };
190 static const char *panic_later
, *panic_param
;
192 extern const struct obs_kernel_param __setup_start
[], __setup_end
[];
194 static bool __init
obsolete_checksetup(char *line
)
196 const struct obs_kernel_param
*p
;
197 bool had_early_param
= false;
201 int n
= strlen(p
->str
);
202 if (parameqn(line
, p
->str
, n
)) {
204 /* Already done in parse_early_param?
205 * (Needs exact match on param part).
206 * Keep iterating, as we can have early
207 * params and __setups of same names 8( */
208 if (line
[n
] == '\0' || line
[n
] == '=')
209 had_early_param
= true;
210 } else if (!p
->setup_func
) {
211 pr_warn("Parameter %s is obsolete, ignored\n",
214 } else if (p
->setup_func(line
+ n
))
218 } while (p
< __setup_end
);
220 return had_early_param
;
224 * This should be approx 2 Bo*oMips to start (note initial shift), and will
225 * still work even if initially too large, it will just take slightly longer
227 unsigned long loops_per_jiffy
= (1<<12);
228 EXPORT_SYMBOL(loops_per_jiffy
);
230 static int __init
debug_kernel(char *str
)
232 console_loglevel
= CONSOLE_LOGLEVEL_DEBUG
;
236 static int __init
quiet_kernel(char *str
)
238 console_loglevel
= CONSOLE_LOGLEVEL_QUIET
;
242 early_param("debug", debug_kernel
);
243 early_param("quiet", quiet_kernel
);
245 static int __init
loglevel(char *str
)
250 * Only update loglevel value when a correct setting was passed,
251 * to prevent blind crashes (when loglevel being set to 0) that
252 * are quite hard to debug
254 if (get_option(&str
, &newlevel
)) {
255 console_loglevel
= newlevel
;
262 early_param("loglevel", loglevel
);
264 #ifdef CONFIG_BLK_DEV_INITRD
265 static void * __init
get_boot_config_from_initrd(u32
*_size
, u32
*_csum
)
275 data
= (char *)initrd_end
- BOOTCONFIG_MAGIC_LEN
;
277 * Since Grub may align the size of initrd to 4, we must
278 * check the preceding 3 bytes as well.
280 for (i
= 0; i
< 4; i
++) {
281 if (!memcmp(data
, BOOTCONFIG_MAGIC
, BOOTCONFIG_MAGIC_LEN
))
288 hdr
= (u32
*)(data
- 8);
289 size
= le32_to_cpu(hdr
[0]);
290 csum
= le32_to_cpu(hdr
[1]);
292 data
= ((void *)hdr
) - size
;
293 if ((unsigned long)data
< initrd_start
) {
294 pr_err("bootconfig size %d is greater than initrd size %ld\n",
295 size
, initrd_end
- initrd_start
);
299 /* Remove bootconfig from initramfs/initrd */
300 initrd_end
= (unsigned long)data
;
309 static void * __init
get_boot_config_from_initrd(u32
*_size
, u32
*_csum
)
315 #ifdef CONFIG_BOOT_CONFIG
317 static char xbc_namebuf
[XBC_KEYLEN_MAX
] __initdata
;
319 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
321 static int __init
xbc_snprint_cmdline(char *buf
, size_t size
,
322 struct xbc_node
*root
)
324 struct xbc_node
*knode
, *vnode
;
325 char *end
= buf
+ size
;
329 xbc_node_for_each_key_value(root
, knode
, val
) {
330 ret
= xbc_node_compose_key_after(root
, knode
,
331 xbc_namebuf
, XBC_KEYLEN_MAX
);
335 vnode
= xbc_node_get_child(knode
);
337 ret
= snprintf(buf
, rest(buf
, end
), "%s ", xbc_namebuf
);
343 xbc_array_for_each_value(vnode
, val
) {
344 ret
= snprintf(buf
, rest(buf
, end
), "%s=\"%s\" ",
352 return buf
- (end
- size
);
356 /* Make an extra command line under given key word */
357 static char * __init
xbc_make_cmdline(const char *key
)
359 struct xbc_node
*root
;
363 root
= xbc_find_node(key
);
367 /* Count required buffer size */
368 len
= xbc_snprint_cmdline(NULL
, 0, root
);
372 new_cmdline
= memblock_alloc(len
+ 1, SMP_CACHE_BYTES
);
374 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
378 ret
= xbc_snprint_cmdline(new_cmdline
, len
+ 1, root
);
379 if (ret
< 0 || ret
> len
) {
380 pr_err("Failed to print extra kernel cmdline.\n");
387 static u32
boot_config_checksum(unsigned char *p
, u32 size
)
397 static int __init
bootconfig_params(char *param
, char *val
,
398 const char *unused
, void *arg
)
400 if (strcmp(param
, "bootconfig") == 0) {
401 bootconfig_found
= true;
406 static void __init
setup_boot_config(const char *cmdline
)
408 static char tmp_cmdline
[COMMAND_LINE_SIZE
] __initdata
;
412 char *data
, *copy
, *err
;
415 /* Cut out the bootconfig data even if we have no bootconfig option */
416 data
= get_boot_config_from_initrd(&size
, &csum
);
418 strlcpy(tmp_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
419 err
= parse_args("bootconfig", tmp_cmdline
, NULL
, 0, 0, 0, NULL
,
422 if (IS_ERR(err
) || !bootconfig_found
)
425 /* parse_args() stops at '--' and returns an address */
427 initargs_found
= true;
430 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
434 if (size
>= XBC_DATA_MAX
) {
435 pr_err("bootconfig size %d greater than max size %d\n",
440 if (boot_config_checksum((unsigned char *)data
, size
) != csum
) {
441 pr_err("bootconfig checksum failed\n");
445 copy
= memblock_alloc(size
+ 1, SMP_CACHE_BYTES
);
447 pr_err("Failed to allocate memory for bootconfig\n");
451 memcpy(copy
, data
, size
);
454 ret
= xbc_init(copy
, &msg
, &pos
);
457 pr_err("Failed to init bootconfig: %s.\n", msg
);
459 pr_err("Failed to parse bootconfig: %s at %d.\n",
462 pr_info("Load bootconfig: %d bytes %d nodes\n", size
, ret
);
463 /* keys starting with "kernel." are passed via cmdline */
464 extra_command_line
= xbc_make_cmdline("kernel");
465 /* Also, "init." keys are init arguments */
466 extra_init_args
= xbc_make_cmdline("init");
473 static void __init
setup_boot_config(const char *cmdline
)
475 /* Remove bootconfig data from initrd */
476 get_boot_config_from_initrd(NULL
, NULL
);
479 static int __init
warn_bootconfig(char *str
)
481 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
484 early_param("bootconfig", warn_bootconfig
);
488 /* Change NUL term back to "=", to make "param" the whole string. */
489 static void __init
repair_env_string(char *param
, char *val
)
492 /* param=val or param="val"? */
493 if (val
== param
+strlen(param
)+1)
495 else if (val
== param
+strlen(param
)+2) {
497 memmove(val
-1, val
, strlen(val
)+1);
503 /* Anything after -- gets handed straight to init. */
504 static int __init
set_init_arg(char *param
, char *val
,
505 const char *unused
, void *arg
)
512 repair_env_string(param
, val
);
514 for (i
= 0; argv_init
[i
]; i
++) {
515 if (i
== MAX_INIT_ARGS
) {
516 panic_later
= "init";
521 argv_init
[i
] = param
;
526 * Unknown boot options get handed to init, unless they look like
527 * unused parameters (modprobe will find them in /proc/cmdline).
529 static int __init
unknown_bootoption(char *param
, char *val
,
530 const char *unused
, void *arg
)
532 size_t len
= strlen(param
);
534 repair_env_string(param
, val
);
536 /* Handle obsolete-style parameters */
537 if (obsolete_checksetup(param
))
540 /* Unused module parameter. */
541 if (strnchr(param
, len
, '.'))
548 /* Environment option */
550 for (i
= 0; envp_init
[i
]; i
++) {
551 if (i
== MAX_INIT_ENVS
) {
555 if (!strncmp(param
, envp_init
[i
], len
+1))
558 envp_init
[i
] = param
;
560 /* Command line option */
562 for (i
= 0; argv_init
[i
]; i
++) {
563 if (i
== MAX_INIT_ARGS
) {
564 panic_later
= "init";
568 argv_init
[i
] = param
;
573 static int __init
init_setup(char *str
)
577 execute_command
= str
;
579 * In case LILO is going to boot us with default command line,
580 * it prepends "auto" before the whole cmdline which makes
581 * the shell think it should execute a script with such name.
582 * So we ignore all arguments entered _before_ init=... [MJ]
584 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
588 __setup("init=", init_setup
);
590 static int __init
rdinit_setup(char *str
)
594 ramdisk_execute_command
= str
;
595 /* See "auto" comment in init_setup */
596 for (i
= 1; i
< MAX_INIT_ARGS
; i
++)
600 __setup("rdinit=", rdinit_setup
);
603 static const unsigned int setup_max_cpus
= NR_CPUS
;
604 static inline void setup_nr_cpu_ids(void) { }
605 static inline void smp_prepare_cpus(unsigned int maxcpus
) { }
609 * We need to store the untouched command line for future reference.
610 * We also need to store the touched command line since the parameter
611 * parsing is performed in place, and we should allow a component to
612 * store reference of name/value for future reference.
614 static void __init
setup_command_line(char *command_line
)
616 size_t len
, xlen
= 0, ilen
= 0;
618 if (extra_command_line
)
619 xlen
= strlen(extra_command_line
);
621 ilen
= strlen(extra_init_args
) + 4; /* for " -- " */
623 len
= xlen
+ strlen(boot_command_line
) + 1;
625 saved_command_line
= memblock_alloc(len
+ ilen
, SMP_CACHE_BYTES
);
626 if (!saved_command_line
)
627 panic("%s: Failed to allocate %zu bytes\n", __func__
, len
+ ilen
);
629 static_command_line
= memblock_alloc(len
, SMP_CACHE_BYTES
);
630 if (!static_command_line
)
631 panic("%s: Failed to allocate %zu bytes\n", __func__
, len
);
635 * We have to put extra_command_line before boot command
636 * lines because there could be dashes (separator of init
637 * command line) in the command lines.
639 strcpy(saved_command_line
, extra_command_line
);
640 strcpy(static_command_line
, extra_command_line
);
642 strcpy(saved_command_line
+ xlen
, boot_command_line
);
643 strcpy(static_command_line
+ xlen
, command_line
);
647 * Append supplemental init boot args to saved_command_line
648 * so that user can check what command line options passed
651 len
= strlen(saved_command_line
);
652 if (initargs_found
) {
653 saved_command_line
[len
++] = ' ';
655 strcpy(saved_command_line
+ len
, " -- ");
659 strcpy(saved_command_line
+ len
, extra_init_args
);
664 * We need to finalize in a non-__init function or else race conditions
665 * between the root thread and the init thread may cause start_kernel to
666 * be reaped by free_initmem before the root thread has proceeded to
669 * gcc-3.4 accidentally inlines this function, so use noinline.
672 static __initdata
DECLARE_COMPLETION(kthreadd_done
);
674 noinline
void __ref
rest_init(void)
676 struct task_struct
*tsk
;
679 rcu_scheduler_starting();
681 * We need to spawn init first so that it obtains pid 1, however
682 * the init task will end up wanting to create kthreads, which, if
683 * we schedule it before we create kthreadd, will OOPS.
685 pid
= kernel_thread(kernel_init
, NULL
, CLONE_FS
);
687 * Pin init on the boot CPU. Task migration is not properly working
688 * until sched_init_smp() has been run. It will set the allowed
689 * CPUs for init to the non isolated CPUs.
692 tsk
= find_task_by_pid_ns(pid
, &init_pid_ns
);
693 set_cpus_allowed_ptr(tsk
, cpumask_of(smp_processor_id()));
696 numa_default_policy();
697 pid
= kernel_thread(kthreadd
, NULL
, CLONE_FS
| CLONE_FILES
);
699 kthreadd_task
= find_task_by_pid_ns(pid
, &init_pid_ns
);
703 * Enable might_sleep() and smp_processor_id() checks.
704 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
705 * kernel_thread() would trigger might_sleep() splats. With
706 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
707 * already, but it's stuck on the kthreadd_done completion.
709 system_state
= SYSTEM_SCHEDULING
;
711 complete(&kthreadd_done
);
714 * The boot idle thread must execute schedule()
715 * at least once to get things moving:
717 schedule_preempt_disabled();
718 /* Call into cpu_idle with preempt disabled */
719 cpu_startup_entry(CPUHP_ONLINE
);
722 /* Check for early params. */
723 static int __init
do_early_param(char *param
, char *val
,
724 const char *unused
, void *arg
)
726 const struct obs_kernel_param
*p
;
728 for (p
= __setup_start
; p
< __setup_end
; p
++) {
729 if ((p
->early
&& parameq(param
, p
->str
)) ||
730 (strcmp(param
, "console") == 0 &&
731 strcmp(p
->str
, "earlycon") == 0)
733 if (p
->setup_func(val
) != 0)
734 pr_warn("Malformed early option '%s'\n", param
);
737 /* We accept everything at this stage. */
741 void __init
parse_early_options(char *cmdline
)
743 parse_args("early options", cmdline
, NULL
, 0, 0, 0, NULL
,
747 /* Arch code calls this early on, or if not, just before other parsing. */
748 void __init
parse_early_param(void)
750 static int done __initdata
;
751 static char tmp_cmdline
[COMMAND_LINE_SIZE
] __initdata
;
756 /* All fall through to do_early_param. */
757 strlcpy(tmp_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
758 parse_early_options(tmp_cmdline
);
762 void __init __weak
arch_post_acpi_subsys_init(void) { }
764 void __init __weak
smp_setup_processor_id(void)
768 # if THREAD_SIZE >= PAGE_SIZE
769 void __init __weak
thread_stack_cache_init(void)
774 void __init __weak
mem_encrypt_init(void) { }
776 void __init __weak
poking_init(void) { }
778 void __init __weak
pgtable_cache_init(void) { }
781 core_param(initcall_debug
, initcall_debug
, bool, 0644);
783 #ifdef TRACEPOINTS_ENABLED
784 static void __init
initcall_debug_enable(void);
786 static inline void initcall_debug_enable(void)
791 /* Report memory auto-initialization states for this boot. */
792 static void __init
report_meminit(void)
796 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN
))
797 stack
= "all(pattern)";
798 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO
))
800 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL
))
801 stack
= "byref_all(zero)";
802 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF
))
803 stack
= "byref(zero)";
804 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER
))
805 stack
= "__user(zero)";
809 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
810 stack
, want_init_on_alloc(GFP_KERNEL
) ? "on" : "off",
811 want_init_on_free() ? "on" : "off");
812 if (want_init_on_free())
813 pr_info("mem auto-init: clearing system memory may take some time...\n");
817 * Set up kernel memory allocators
819 static void __init
mm_init(void)
822 * page_ext requires contiguous pages,
823 * bigger than MAX_ORDER unless SPARSEMEM.
825 page_ext_init_flatmem();
826 init_mem_debugging_and_hardening();
829 /* page_owner must be initialized after buddy is ready */
830 page_ext_init_flatmem_late();
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();
1048 taskstats_init_early();
1054 acpi_subsystem_init();
1055 arch_post_acpi_subsys_init();
1058 /* Do the rest non-__init'ed, we're now alive */
1059 arch_call_rest_init();
1061 prevent_tail_call_optimization();
1064 /* Call all constructor functions linked into the kernel. */
1065 static void __init
do_ctors(void)
1068 * For UML, the constructors have already been called by the
1069 * normal setup code as it's just a normal ELF binary, so we
1070 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1071 * even on UML for modules.
1073 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1074 ctor_fn_t
*fn
= (ctor_fn_t
*) __ctors_start
;
1076 for (; fn
< (ctor_fn_t
*) __ctors_end
; fn
++)
1081 #ifdef CONFIG_KALLSYMS
1082 struct blacklist_entry
{
1083 struct list_head next
;
1087 static __initdata_or_module
LIST_HEAD(blacklisted_initcalls
);
1089 static int __init
initcall_blacklist(char *str
)
1092 struct blacklist_entry
*entry
;
1094 /* str argument is a comma-separated list of functions */
1096 str_entry
= strsep(&str
, ",");
1098 pr_debug("blacklisting initcall %s\n", str_entry
);
1099 entry
= memblock_alloc(sizeof(*entry
),
1102 panic("%s: Failed to allocate %zu bytes\n",
1103 __func__
, sizeof(*entry
));
1104 entry
->buf
= memblock_alloc(strlen(str_entry
) + 1,
1107 panic("%s: Failed to allocate %zu bytes\n",
1108 __func__
, strlen(str_entry
) + 1);
1109 strcpy(entry
->buf
, str_entry
);
1110 list_add(&entry
->next
, &blacklisted_initcalls
);
1112 } while (str_entry
);
1117 static bool __init_or_module
initcall_blacklisted(initcall_t fn
)
1119 struct blacklist_entry
*entry
;
1120 char fn_name
[KSYM_SYMBOL_LEN
];
1123 if (list_empty(&blacklisted_initcalls
))
1126 addr
= (unsigned long) dereference_function_descriptor(fn
);
1127 sprint_symbol_no_offset(fn_name
, addr
);
1130 * fn will be "function_name [module_name]" where [module_name] is not
1131 * displayed for built-in init functions. Strip off the [module_name].
1133 strreplace(fn_name
, ' ', '\0');
1135 list_for_each_entry(entry
, &blacklisted_initcalls
, next
) {
1136 if (!strcmp(fn_name
, entry
->buf
)) {
1137 pr_debug("initcall %s blacklisted\n", fn_name
);
1145 static int __init
initcall_blacklist(char *str
)
1147 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1151 static bool __init_or_module
initcall_blacklisted(initcall_t fn
)
1156 __setup("initcall_blacklist=", initcall_blacklist
);
1158 static __init_or_module
void
1159 trace_initcall_start_cb(void *data
, initcall_t fn
)
1161 ktime_t
*calltime
= (ktime_t
*)data
;
1163 printk(KERN_DEBUG
"calling %pS @ %i\n", fn
, task_pid_nr(current
));
1164 *calltime
= ktime_get();
1167 static __init_or_module
void
1168 trace_initcall_finish_cb(void *data
, initcall_t fn
, int ret
)
1170 ktime_t
*calltime
= (ktime_t
*)data
;
1171 ktime_t delta
, rettime
;
1172 unsigned long long duration
;
1174 rettime
= ktime_get();
1175 delta
= ktime_sub(rettime
, *calltime
);
1176 duration
= (unsigned long long) ktime_to_ns(delta
) >> 10;
1177 printk(KERN_DEBUG
"initcall %pS returned %d after %lld usecs\n",
1181 static ktime_t initcall_calltime
;
1183 #ifdef TRACEPOINTS_ENABLED
1184 static void __init
initcall_debug_enable(void)
1188 ret
= register_trace_initcall_start(trace_initcall_start_cb
,
1189 &initcall_calltime
);
1190 ret
|= register_trace_initcall_finish(trace_initcall_finish_cb
,
1191 &initcall_calltime
);
1192 WARN(ret
, "Failed to register initcall tracepoints\n");
1194 # define do_trace_initcall_start trace_initcall_start
1195 # define do_trace_initcall_finish trace_initcall_finish
1197 static inline void do_trace_initcall_start(initcall_t fn
)
1199 if (!initcall_debug
)
1201 trace_initcall_start_cb(&initcall_calltime
, fn
);
1203 static inline void do_trace_initcall_finish(initcall_t fn
, int ret
)
1205 if (!initcall_debug
)
1207 trace_initcall_finish_cb(&initcall_calltime
, fn
, ret
);
1209 #endif /* !TRACEPOINTS_ENABLED */
1211 int __init_or_module
do_one_initcall(initcall_t fn
)
1213 int count
= preempt_count();
1217 if (initcall_blacklisted(fn
))
1220 do_trace_initcall_start(fn
);
1222 do_trace_initcall_finish(fn
, ret
);
1226 if (preempt_count() != count
) {
1227 sprintf(msgbuf
, "preemption imbalance ");
1228 preempt_count_set(count
);
1230 if (irqs_disabled()) {
1231 strlcat(msgbuf
, "disabled interrupts ", sizeof(msgbuf
));
1234 WARN(msgbuf
[0], "initcall %pS returned with %s\n", fn
, msgbuf
);
1236 add_latent_entropy();
1241 extern initcall_entry_t __initcall_start
[];
1242 extern initcall_entry_t __initcall0_start
[];
1243 extern initcall_entry_t __initcall1_start
[];
1244 extern initcall_entry_t __initcall2_start
[];
1245 extern initcall_entry_t __initcall3_start
[];
1246 extern initcall_entry_t __initcall4_start
[];
1247 extern initcall_entry_t __initcall5_start
[];
1248 extern initcall_entry_t __initcall6_start
[];
1249 extern initcall_entry_t __initcall7_start
[];
1250 extern initcall_entry_t __initcall_end
[];
1252 static initcall_entry_t
*initcall_levels
[] __initdata
= {
1264 /* Keep these in sync with initcalls in include/linux/init.h */
1265 static const char *initcall_level_names
[] __initdata
= {
1276 static int __init
ignore_unknown_bootoption(char *param
, char *val
,
1277 const char *unused
, void *arg
)
1282 static void __init
do_initcall_level(int level
, char *command_line
)
1284 initcall_entry_t
*fn
;
1286 parse_args(initcall_level_names
[level
],
1287 command_line
, __start___param
,
1288 __stop___param
- __start___param
,
1290 NULL
, ignore_unknown_bootoption
);
1292 trace_initcall_level(initcall_level_names
[level
]);
1293 for (fn
= initcall_levels
[level
]; fn
< initcall_levels
[level
+1]; fn
++)
1294 do_one_initcall(initcall_from_entry(fn
));
1297 static void __init
do_initcalls(void)
1300 size_t len
= strlen(saved_command_line
) + 1;
1303 command_line
= kzalloc(len
, GFP_KERNEL
);
1305 panic("%s: Failed to allocate %zu bytes\n", __func__
, len
);
1307 for (level
= 0; level
< ARRAY_SIZE(initcall_levels
) - 1; level
++) {
1308 /* Parser modifies command_line, restore it each time */
1309 strcpy(command_line
, saved_command_line
);
1310 do_initcall_level(level
, command_line
);
1313 kfree(command_line
);
1317 * Ok, the machine is now initialized. None of the devices
1318 * have been touched yet, but the CPU subsystem is up and
1319 * running, and memory and process management works.
1321 * Now we can finally start doing some real work..
1323 static void __init
do_basic_setup(void)
1329 usermodehelper_enable();
1333 static void __init
do_pre_smp_initcalls(void)
1335 initcall_entry_t
*fn
;
1337 trace_initcall_level("early");
1338 for (fn
= __initcall_start
; fn
< __initcall0_start
; fn
++)
1339 do_one_initcall(initcall_from_entry(fn
));
1342 static int run_init_process(const char *init_filename
)
1344 const char *const *p
;
1346 argv_init
[0] = init_filename
;
1347 pr_info("Run %s as init process\n", init_filename
);
1348 pr_debug(" with arguments:\n");
1349 for (p
= argv_init
; *p
; p
++)
1350 pr_debug(" %s\n", *p
);
1351 pr_debug(" with environment:\n");
1352 for (p
= envp_init
; *p
; p
++)
1353 pr_debug(" %s\n", *p
);
1354 return kernel_execve(init_filename
, argv_init
, envp_init
);
1357 static int try_to_run_init_process(const char *init_filename
)
1361 ret
= run_init_process(init_filename
);
1363 if (ret
&& ret
!= -ENOENT
) {
1364 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1365 init_filename
, ret
);
1371 static noinline
void __init
kernel_init_freeable(void);
1373 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1374 bool rodata_enabled __ro_after_init
= true;
1375 static int __init
set_debug_rodata(char *str
)
1377 return strtobool(str
, &rodata_enabled
);
1379 __setup("rodata=", set_debug_rodata
);
1382 #ifdef CONFIG_STRICT_KERNEL_RWX
1383 static void mark_readonly(void)
1385 if (rodata_enabled
) {
1387 * load_module() results in W+X mappings, which are cleaned
1388 * up with call_rcu(). Let's make sure that queued work is
1389 * flushed so that we don't hit false positives looking for
1390 * insecure pages which are W+X.
1396 pr_info("Kernel memory protection disabled.\n");
1398 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1399 static inline void mark_readonly(void)
1401 pr_warn("Kernel memory protection not selected by kernel config.\n");
1404 static inline void mark_readonly(void)
1406 pr_warn("This architecture does not have kernel memory protection.\n");
1410 void __weak
free_initmem(void)
1412 free_initmem_default(POISON_FREE_INITMEM
);
1415 static int __ref
kernel_init(void *unused
)
1419 kernel_init_freeable();
1420 /* need to finish all async __init code before freeing the memory */
1421 async_synchronize_full();
1422 kprobe_free_init_mem();
1423 ftrace_free_init_mem();
1428 * Kernel mappings are now finalized - update the userspace page-table
1433 system_state
= SYSTEM_RUNNING
;
1434 numa_default_policy();
1436 rcu_end_inkernel_boot();
1440 if (ramdisk_execute_command
) {
1441 ret
= run_init_process(ramdisk_execute_command
);
1444 pr_err("Failed to execute %s (error %d)\n",
1445 ramdisk_execute_command
, ret
);
1449 * We try each of these until one succeeds.
1451 * The Bourne shell can be used instead of init if we are
1452 * trying to recover a really broken machine.
1454 if (execute_command
) {
1455 ret
= run_init_process(execute_command
);
1458 panic("Requested init %s failed (error %d).",
1459 execute_command
, ret
);
1462 if (CONFIG_DEFAULT_INIT
[0] != '\0') {
1463 ret
= run_init_process(CONFIG_DEFAULT_INIT
);
1465 pr_err("Default init %s failed (error %d)\n",
1466 CONFIG_DEFAULT_INIT
, ret
);
1471 if (!try_to_run_init_process("/sbin/init") ||
1472 !try_to_run_init_process("/etc/init") ||
1473 !try_to_run_init_process("/bin/init") ||
1474 !try_to_run_init_process("/bin/sh"))
1477 panic("No working init found. Try passing init= option to kernel. "
1478 "See Linux Documentation/admin-guide/init.rst for guidance.");
1481 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1482 void __init
console_on_rootfs(void)
1484 struct file
*file
= filp_open("/dev/console", O_RDWR
, 0);
1487 pr_err("Warning: unable to open an initial console.\n");
1496 static noinline
void __init
kernel_init_freeable(void)
1499 * Wait until kthreadd is all set-up.
1501 wait_for_completion(&kthreadd_done
);
1503 /* Now the scheduler is fully set up and can do blocking allocations */
1504 gfp_allowed_mask
= __GFP_BITS_MASK
;
1507 * init can allocate pages on any node
1509 set_mems_allowed(node_states
[N_MEMORY
]);
1511 cad_pid
= task_pid(current
);
1513 smp_prepare_cpus(setup_max_cpus
);
1517 init_mm_internals();
1519 rcu_init_tasks_generic();
1520 do_pre_smp_initcalls();
1521 lockup_detector_init();
1527 page_alloc_init_late();
1528 /* Initialize page ext after all struct pages are initialized. */
1533 kunit_run_all_tests();
1535 console_on_rootfs();
1538 * check if there is an early userspace init. If yes, let it do all
1541 if (init_eaccess(ramdisk_execute_command
) != 0) {
1542 ramdisk_execute_command
= NULL
;
1543 prepare_namespace();
1547 * Ok, we have completed the initial bootup, and
1548 * we're essentially up and running. Get rid of the
1549 * initmem segments and start the user-mode stuff..
1551 * rootfs is available now, try loading the public keys
1552 * and default modules
1555 integrity_load_keys();