linux/include/linux/kernel.h
<<
>>
Prefs
   1/* SPDX-License-Identifier: GPL-2.0 */
   2/*
   3 * NOTE:
   4 *
   5 * This header has combined a lot of unrelated to each other stuff.
   6 * The process of splitting its content is in progress while keeping
   7 * backward compatibility. That's why it's highly recommended NOT to
   8 * include this header inside another header file, especially under
   9 * generic or architectural include/ directory.
  10 */
  11#ifndef _LINUX_KERNEL_H
  12#define _LINUX_KERNEL_H
  13
  14#include <linux/stdarg.h>
  15#include <linux/align.h>
  16#include <linux/limits.h>
  17#include <linux/linkage.h>
  18#include <linux/stddef.h>
  19#include <linux/types.h>
  20#include <linux/compiler.h>
  21#include <linux/container_of.h>
  22#include <linux/bitops.h>
  23#include <linux/hex.h>
  24#include <linux/kstrtox.h>
  25#include <linux/log2.h>
  26#include <linux/math.h>
  27#include <linux/minmax.h>
  28#include <linux/typecheck.h>
  29#include <linux/panic.h>
  30#include <linux/printk.h>
  31#include <linux/build_bug.h>
  32#include <linux/sprintf.h>
  33#include <linux/static_call_types.h>
  34#include <linux/instruction_pointer.h>
  35#include <asm/byteorder.h>
  36
  37#include <uapi/linux/kernel.h>
  38
  39#define STACK_MAGIC     0xdeadbeef
  40
  41/**
  42 * REPEAT_BYTE - repeat the value @x multiple times as an unsigned long value
  43 * @x: value to repeat
  44 *
  45 * NOTE: @x is not checked for > 0xff; larger values produce odd results.
  46 */
  47#define REPEAT_BYTE(x)  ((~0ul / 0xff) * (x))
  48
  49/* generic data direction definitions */
  50#define READ                    0
  51#define WRITE                   1
  52
  53/**
  54 * ARRAY_SIZE - get the number of elements in array @arr
  55 * @arr: array to be sized
  56 */
  57#define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]) + __must_be_array(arr))
  58
  59#define PTR_IF(cond, ptr)       ((cond) ? (ptr) : NULL)
  60
  61#define u64_to_user_ptr(x) (            \
  62{                                       \
  63        typecheck(u64, (x));            \
  64        (void __user *)(uintptr_t)(x);  \
  65}                                       \
  66)
  67
  68/**
  69 * upper_32_bits - return bits 32-63 of a number
  70 * @n: the number we're accessing
  71 *
  72 * A basic shift-right of a 64- or 32-bit quantity.  Use this to suppress
  73 * the "right shift count >= width of type" warning when that quantity is
  74 * 32-bits.
  75 */
  76#define upper_32_bits(n) ((u32)(((n) >> 16) >> 16))
  77
  78/**
  79 * lower_32_bits - return bits 0-31 of a number
  80 * @n: the number we're accessing
  81 */
  82#define lower_32_bits(n) ((u32)((n) & 0xffffffff))
  83
  84/**
  85 * upper_16_bits - return bits 16-31 of a number
  86 * @n: the number we're accessing
  87 */
  88#define upper_16_bits(n) ((u16)((n) >> 16))
  89
  90/**
  91 * lower_16_bits - return bits 0-15 of a number
  92 * @n: the number we're accessing
  93 */
  94#define lower_16_bits(n) ((u16)((n) & 0xffff))
  95
  96struct completion;
  97struct user;
  98
  99#ifdef CONFIG_PREEMPT_VOLUNTARY_BUILD
 100
 101extern int __cond_resched(void);
 102# define might_resched() __cond_resched()
 103
 104#elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
 105
 106extern int __cond_resched(void);
 107
 108DECLARE_STATIC_CALL(might_resched, __cond_resched);
 109
 110static __always_inline void might_resched(void)
 111{
 112        static_call_mod(might_resched)();
 113}
 114
 115#elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
 116
 117extern int dynamic_might_resched(void);
 118# define might_resched() dynamic_might_resched()
 119
 120#else
 121
 122# define might_resched() do { } while (0)
 123
 124#endif /* CONFIG_PREEMPT_* */
 125
 126#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
 127extern void __might_resched(const char *file, int line, unsigned int offsets);
 128extern void __might_sleep(const char *file, int line);
 129extern void __cant_sleep(const char *file, int line, int preempt_offset);
 130extern void __cant_migrate(const char *file, int line);
 131
 132/**
 133 * might_sleep - annotation for functions that can sleep
 134 *
 135 * this macro will print a stack trace if it is executed in an atomic
 136 * context (spinlock, irq-handler, ...). Additional sections where blocking is
 137 * not allowed can be annotated with non_block_start() and non_block_end()
 138 * pairs.
 139 *
 140 * This is a useful debugging help to be able to catch problems early and not
 141 * be bitten later when the calling function happens to sleep when it is not
 142 * supposed to.
 143 */
 144# define might_sleep() \
 145        do { __might_sleep(__FILE__, __LINE__); might_resched(); } while (0)
 146/**
 147 * cant_sleep - annotation for functions that cannot sleep
 148 *
 149 * this macro will print a stack trace if it is executed with preemption enabled
 150 */
 151# define cant_sleep() \
 152        do { __cant_sleep(__FILE__, __LINE__, 0); } while (0)
 153# define sched_annotate_sleep() (current->task_state_change = 0)
 154
 155/**
 156 * cant_migrate - annotation for functions that cannot migrate
 157 *
 158 * Will print a stack trace if executed in code which is migratable
 159 */
 160# define cant_migrate()                                                 \
 161        do {                                                            \
 162                if (IS_ENABLED(CONFIG_SMP))                             \
 163                        __cant_migrate(__FILE__, __LINE__);             \
 164        } while (0)
 165
 166/**
 167 * non_block_start - annotate the start of section where sleeping is prohibited
 168 *
 169 * This is on behalf of the oom reaper, specifically when it is calling the mmu
 170 * notifiers. The problem is that if the notifier were to block on, for example,
 171 * mutex_lock() and if the process which holds that mutex were to perform a
 172 * sleeping memory allocation, the oom reaper is now blocked on completion of
 173 * that memory allocation. Other blocking calls like wait_event() pose similar
 174 * issues.
 175 */
 176# define non_block_start() (current->non_block_count++)
 177/**
 178 * non_block_end - annotate the end of section where sleeping is prohibited
 179 *
 180 * Closes a section opened by non_block_start().
 181 */
 182# define non_block_end() WARN_ON(current->non_block_count-- == 0)
 183#else
 184  static inline void __might_resched(const char *file, int line,
 185                                     unsigned int offsets) { }
 186static inline void __might_sleep(const char *file, int line) { }
 187# define might_sleep() do { might_resched(); } while (0)
 188# define cant_sleep() do { } while (0)
 189# define cant_migrate()         do { } while (0)
 190# define sched_annotate_sleep() do { } while (0)
 191# define non_block_start() do { } while (0)
 192# define non_block_end() do { } while (0)
 193#endif
 194
 195#define might_sleep_if(cond) do { if (cond) might_sleep(); } while (0)
 196
 197#if defined(CONFIG_MMU) && \
 198        (defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_DEBUG_ATOMIC_SLEEP))
 199#define might_fault() __might_fault(__FILE__, __LINE__)
 200void __might_fault(const char *file, int line);
 201#else
 202static inline void might_fault(void) { }
 203#endif
 204
 205void do_exit(long error_code) __noreturn;
 206
 207extern int get_option(char **str, int *pint);
 208extern char *get_options(const char *str, int nints, int *ints);
 209extern unsigned long long memparse(const char *ptr, char **retptr);
 210extern bool parse_option_str(const char *str, const char *option);
 211extern char *next_arg(char *args, char **param, char **val);
 212
 213extern int core_kernel_text(unsigned long addr);
 214extern int __kernel_text_address(unsigned long addr);
 215extern int kernel_text_address(unsigned long addr);
 216extern int func_ptr_is_kernel_text(void *ptr);
 217
 218extern void bust_spinlocks(int yes);
 219
 220extern int root_mountflags;
 221
 222extern bool early_boot_irqs_disabled;
 223
 224/*
 225 * Values used for system_state. Ordering of the states must not be changed
 226 * as code checks for <, <=, >, >= STATE.
 227 */
 228extern enum system_states {
 229        SYSTEM_BOOTING,
 230        SYSTEM_SCHEDULING,
 231        SYSTEM_FREEING_INITMEM,
 232        SYSTEM_RUNNING,
 233        SYSTEM_HALT,
 234        SYSTEM_POWER_OFF,
 235        SYSTEM_RESTART,
 236        SYSTEM_SUSPEND,
 237} system_state;
 238
 239/*
 240 * General tracing related utility functions - trace_printk(),
 241 * tracing_on/tracing_off and tracing_start()/tracing_stop
 242 *
 243 * Use tracing_on/tracing_off when you want to quickly turn on or off
 244 * tracing. It simply enables or disables the recording of the trace events.
 245 * This also corresponds to the user space /sys/kernel/tracing/tracing_on
 246 * file, which gives a means for the kernel and userspace to interact.
 247 * Place a tracing_off() in the kernel where you want tracing to end.
 248 * From user space, examine the trace, and then echo 1 > tracing_on
 249 * to continue tracing.
 250 *
 251 * tracing_stop/tracing_start has slightly more overhead. It is used
 252 * by things like suspend to ram where disabling the recording of the
 253 * trace is not enough, but tracing must actually stop because things
 254 * like calling smp_processor_id() may crash the system.
 255 *
 256 * Most likely, you want to use tracing_on/tracing_off.
 257 */
 258
 259enum ftrace_dump_mode {
 260        DUMP_NONE,
 261        DUMP_ALL,
 262        DUMP_ORIG,
 263};
 264
 265#ifdef CONFIG_TRACING
 266void tracing_on(void);
 267void tracing_off(void);
 268int tracing_is_on(void);
 269void tracing_snapshot(void);
 270void tracing_snapshot_alloc(void);
 271
 272extern void tracing_start(void);
 273extern void tracing_stop(void);
 274
 275static inline __printf(1, 2)
 276void ____trace_printk_check_format(const char *fmt, ...)
 277{
 278}
 279#define __trace_printk_check_format(fmt, args...)                       \
 280do {                                                                    \
 281        if (0)                                                          \
 282                ____trace_printk_check_format(fmt, ##args);             \
 283} while (0)
 284
 285/**
 286 * trace_printk - printf formatting in the ftrace buffer
 287 * @fmt: the printf format for printing
 288 *
 289 * Note: __trace_printk is an internal function for trace_printk() and
 290 *       the @ip is passed in via the trace_printk() macro.
 291 *
 292 * This function allows a kernel developer to debug fast path sections
 293 * that printk is not appropriate for. By scattering in various
 294 * printk like tracing in the code, a developer can quickly see
 295 * where problems are occurring.
 296 *
 297 * This is intended as a debugging tool for the developer only.
 298 * Please refrain from leaving trace_printks scattered around in
 299 * your code. (Extra memory is used for special buffers that are
 300 * allocated when trace_printk() is used.)
 301 *
 302 * A little optimization trick is done here. If there's only one
 303 * argument, there's no need to scan the string for printf formats.
 304 * The trace_puts() will suffice. But how can we take advantage of
 305 * using trace_puts() when trace_printk() has only one argument?
 306 * By stringifying the args and checking the size we can tell
 307 * whether or not there are args. __stringify((__VA_ARGS__)) will
 308 * turn into "()\0" with a size of 3 when there are no args, anything
 309 * else will be bigger. All we need to do is define a string to this,
 310 * and then take its size and compare to 3. If it's bigger, use
 311 * do_trace_printk() otherwise, optimize it to trace_puts(). Then just
 312 * let gcc optimize the rest.
 313 */
 314
 315#define trace_printk(fmt, ...)                          \
 316do {                                                    \
 317        char _______STR[] = __stringify((__VA_ARGS__)); \
 318        if (sizeof(_______STR) > 3)                     \
 319                do_trace_printk(fmt, ##__VA_ARGS__);    \
 320        else                                            \
 321                trace_puts(fmt);                        \
 322} while (0)
 323
 324#define do_trace_printk(fmt, args...)                                   \
 325do {                                                                    \
 326        static const char *trace_printk_fmt __used                      \
 327                __section("__trace_printk_fmt") =                       \
 328                __builtin_constant_p(fmt) ? fmt : NULL;                 \
 329                                                                        \
 330        __trace_printk_check_format(fmt, ##args);                       \
 331                                                                        \
 332        if (__builtin_constant_p(fmt))                                  \
 333                __trace_bprintk(_THIS_IP_, trace_printk_fmt, ##args);   \
 334        else                                                            \
 335                __trace_printk(_THIS_IP_, fmt, ##args);                 \
 336} while (0)
 337
 338extern __printf(2, 3)
 339int __trace_bprintk(unsigned long ip, const char *fmt, ...);
 340
 341extern __printf(2, 3)
 342int __trace_printk(unsigned long ip, const char *fmt, ...);
 343
 344/**
 345 * trace_puts - write a string into the ftrace buffer
 346 * @str: the string to record
 347 *
 348 * Note: __trace_bputs is an internal function for trace_puts and
 349 *       the @ip is passed in via the trace_puts macro.
 350 *
 351 * This is similar to trace_printk() but is made for those really fast
 352 * paths that a developer wants the least amount of "Heisenbug" effects,
 353 * where the processing of the print format is still too much.
 354 *
 355 * This function allows a kernel developer to debug fast path sections
 356 * that printk is not appropriate for. By scattering in various
 357 * printk like tracing in the code, a developer can quickly see
 358 * where problems are occurring.
 359 *
 360 * This is intended as a debugging tool for the developer only.
 361 * Please refrain from leaving trace_puts scattered around in
 362 * your code. (Extra memory is used for special buffers that are
 363 * allocated when trace_puts() is used.)
 364 *
 365 * Returns: 0 if nothing was written, positive # if string was.
 366 *  (1 when __trace_bputs is used, strlen(str) when __trace_puts is used)
 367 */
 368
 369#define trace_puts(str) ({                                              \
 370        static const char *trace_printk_fmt __used                      \
 371                __section("__trace_printk_fmt") =                       \
 372                __builtin_constant_p(str) ? str : NULL;                 \
 373                                                                        \
 374        if (__builtin_constant_p(str))                                  \
 375                __trace_bputs(_THIS_IP_, trace_printk_fmt);             \
 376        else                                                            \
 377                __trace_puts(_THIS_IP_, str, strlen(str));              \
 378})
 379extern int __trace_bputs(unsigned long ip, const char *str);
 380extern int __trace_puts(unsigned long ip, const char *str, int size);
 381
 382extern void trace_dump_stack(int skip);
 383
 384/*
 385 * The double __builtin_constant_p is because gcc will give us an error
 386 * if we try to allocate the static variable to fmt if it is not a
 387 * constant. Even with the outer if statement.
 388 */
 389#define ftrace_vprintk(fmt, vargs)                                      \
 390do {                                                                    \
 391        if (__builtin_constant_p(fmt)) {                                \
 392                static const char *trace_printk_fmt __used              \
 393                  __section("__trace_printk_fmt") =                     \
 394                        __builtin_constant_p(fmt) ? fmt : NULL;         \
 395                                                                        \
 396                __ftrace_vbprintk(_THIS_IP_, trace_printk_fmt, vargs);  \
 397        } else                                                          \
 398                __ftrace_vprintk(_THIS_IP_, fmt, vargs);                \
 399} while (0)
 400
 401extern __printf(2, 0) int
 402__ftrace_vbprintk(unsigned long ip, const char *fmt, va_list ap);
 403
 404extern __printf(2, 0) int
 405__ftrace_vprintk(unsigned long ip, const char *fmt, va_list ap);
 406
 407extern void ftrace_dump(enum ftrace_dump_mode oops_dump_mode);
 408#else
 409static inline void tracing_start(void) { }
 410static inline void tracing_stop(void) { }
 411static inline void trace_dump_stack(int skip) { }
 412
 413static inline void tracing_on(void) { }
 414static inline void tracing_off(void) { }
 415static inline int tracing_is_on(void) { return 0; }
 416static inline void tracing_snapshot(void) { }
 417static inline void tracing_snapshot_alloc(void) { }
 418
 419static inline __printf(1, 2)
 420int trace_printk(const char *fmt, ...)
 421{
 422        return 0;
 423}
 424static __printf(1, 0) inline int
 425ftrace_vprintk(const char *fmt, va_list ap)
 426{
 427        return 0;
 428}
 429static inline void ftrace_dump(enum ftrace_dump_mode oops_dump_mode) { }
 430#endif /* CONFIG_TRACING */
 431
 432/* Rebuild everything on CONFIG_FTRACE_MCOUNT_RECORD */
 433#ifdef CONFIG_FTRACE_MCOUNT_RECORD
 434# define REBUILD_DUE_TO_FTRACE_MCOUNT_RECORD
 435#endif
 436
 437/* Permissions on a sysfs file: you didn't miss the 0 prefix did you? */
 438#define VERIFY_OCTAL_PERMISSIONS(perms)                                         \
 439        (BUILD_BUG_ON_ZERO((perms) < 0) +                                       \
 440         BUILD_BUG_ON_ZERO((perms) > 0777) +                                    \
 441         /* USER_READABLE >= GROUP_READABLE >= OTHER_READABLE */                \
 442         BUILD_BUG_ON_ZERO((((perms) >> 6) & 4) < (((perms) >> 3) & 4)) +       \
 443         BUILD_BUG_ON_ZERO((((perms) >> 3) & 4) < ((perms) & 4)) +              \
 444         /* USER_WRITABLE >= GROUP_WRITABLE */                                  \
 445         BUILD_BUG_ON_ZERO((((perms) >> 6) & 2) < (((perms) >> 3) & 2)) +       \
 446         /* OTHER_WRITABLE?  Generally considered a bad idea. */                \
 447         BUILD_BUG_ON_ZERO((perms) & 2) +                                       \
 448         (perms))
 449#endif
 450