xref: /linux/include/linux/kernel.h (revision 7d4e49a77d9930c69751b9192448fda6ff9100f1)
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/array_size.h>
17 #include <linux/limits.h>
18 #include <linux/linkage.h>
19 #include <linux/stddef.h>
20 #include <linux/types.h>
21 #include <linux/compiler.h>
22 #include <linux/container_of.h>
23 #include <linux/bitops.h>
24 #include <linux/hex.h>
25 #include <linux/kstrtox.h>
26 #include <linux/log2.h>
27 #include <linux/math.h>
28 #include <linux/minmax.h>
29 #include <linux/typecheck.h>
30 #include <linux/panic.h>
31 #include <linux/printk.h>
32 #include <linux/build_bug.h>
33 #include <linux/sprintf.h>
34 #include <linux/static_call_types.h>
35 #include <linux/instruction_pointer.h>
36 #include <linux/util_macros.h>
37 #include <linux/wordpart.h>
38 
39 #include <asm/byteorder.h>
40 
41 #include <uapi/linux/kernel.h>
42 
43 #define STACK_MAGIC	0xdeadbeef
44 
45 struct completion;
46 struct user;
47 
48 #ifdef CONFIG_PREEMPT_VOLUNTARY_BUILD
49 
50 extern int __cond_resched(void);
51 # define might_resched() __cond_resched()
52 
53 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
54 
55 extern int __cond_resched(void);
56 
57 DECLARE_STATIC_CALL(might_resched, __cond_resched);
58 
might_resched(void)59 static __always_inline void might_resched(void)
60 {
61 	static_call_mod(might_resched)();
62 }
63 
64 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
65 
66 extern int dynamic_might_resched(void);
67 # define might_resched() dynamic_might_resched()
68 
69 #else
70 
71 # define might_resched() do { } while (0)
72 
73 #endif /* CONFIG_PREEMPT_* */
74 
75 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
76 extern void __might_resched(const char *file, int line, unsigned int offsets);
77 extern void __might_sleep(const char *file, int line);
78 extern void __cant_sleep(const char *file, int line, int preempt_offset);
79 extern void __cant_migrate(const char *file, int line);
80 
81 /**
82  * might_sleep - annotation for functions that can sleep
83  *
84  * this macro will print a stack trace if it is executed in an atomic
85  * context (spinlock, irq-handler, ...). Additional sections where blocking is
86  * not allowed can be annotated with non_block_start() and non_block_end()
87  * pairs.
88  *
89  * This is a useful debugging help to be able to catch problems early and not
90  * be bitten later when the calling function happens to sleep when it is not
91  * supposed to.
92  */
93 # define might_sleep() \
94 	do { __might_sleep(__FILE__, __LINE__); might_resched(); } while (0)
95 /**
96  * cant_sleep - annotation for functions that cannot sleep
97  *
98  * this macro will print a stack trace if it is executed with preemption enabled
99  */
100 # define cant_sleep() \
101 	do { __cant_sleep(__FILE__, __LINE__, 0); } while (0)
102 # define sched_annotate_sleep()	(current->task_state_change = 0)
103 
104 /**
105  * cant_migrate - annotation for functions that cannot migrate
106  *
107  * Will print a stack trace if executed in code which is migratable
108  */
109 # define cant_migrate()							\
110 	do {								\
111 		if (IS_ENABLED(CONFIG_SMP))				\
112 			__cant_migrate(__FILE__, __LINE__);		\
113 	} while (0)
114 
115 /**
116  * non_block_start - annotate the start of section where sleeping is prohibited
117  *
118  * This is on behalf of the oom reaper, specifically when it is calling the mmu
119  * notifiers. The problem is that if the notifier were to block on, for example,
120  * mutex_lock() and if the process which holds that mutex were to perform a
121  * sleeping memory allocation, the oom reaper is now blocked on completion of
122  * that memory allocation. Other blocking calls like wait_event() pose similar
123  * issues.
124  */
125 # define non_block_start() (current->non_block_count++)
126 /**
127  * non_block_end - annotate the end of section where sleeping is prohibited
128  *
129  * Closes a section opened by non_block_start().
130  */
131 # define non_block_end() WARN_ON(current->non_block_count-- == 0)
132 #else
__might_resched(const char * file,int line,unsigned int offsets)133   static inline void __might_resched(const char *file, int line,
134 				     unsigned int offsets) { }
__might_sleep(const char * file,int line)135 static inline void __might_sleep(const char *file, int line) { }
136 # define might_sleep() do { might_resched(); } while (0)
137 # define cant_sleep() do { } while (0)
138 # define cant_migrate()		do { } while (0)
139 # define sched_annotate_sleep() do { } while (0)
140 # define non_block_start() do { } while (0)
141 # define non_block_end() do { } while (0)
142 #endif
143 
144 #define might_sleep_if(cond) do { if (cond) might_sleep(); } while (0)
145 
146 #if defined(CONFIG_MMU) && \
147 	(defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_DEBUG_ATOMIC_SLEEP))
148 #define might_fault() __might_fault(__FILE__, __LINE__)
149 void __might_fault(const char *file, int line);
150 #else
might_fault(void)151 static inline void might_fault(void) { }
152 #endif
153 
154 void do_exit(long error_code) __noreturn;
155 
156 extern int core_kernel_text(unsigned long addr);
157 extern int __kernel_text_address(unsigned long addr);
158 extern int kernel_text_address(unsigned long addr);
159 extern int func_ptr_is_kernel_text(void *ptr);
160 
161 extern void bust_spinlocks(int yes);
162 
163 extern int root_mountflags;
164 
165 extern bool early_boot_irqs_disabled;
166 
167 /*
168  * Values used for system_state. Ordering of the states must not be changed
169  * as code checks for <, <=, >, >= STATE.
170  */
171 extern enum system_states {
172 	SYSTEM_BOOTING,
173 	SYSTEM_SCHEDULING,
174 	SYSTEM_FREEING_INITMEM,
175 	SYSTEM_RUNNING,
176 	SYSTEM_HALT,
177 	SYSTEM_POWER_OFF,
178 	SYSTEM_RESTART,
179 	SYSTEM_SUSPEND,
180 } system_state;
181 
182 /*
183  * General tracing related utility functions - trace_printk(),
184  * tracing_on/tracing_off and tracing_start()/tracing_stop
185  *
186  * Use tracing_on/tracing_off when you want to quickly turn on or off
187  * tracing. It simply enables or disables the recording of the trace events.
188  * This also corresponds to the user space /sys/kernel/tracing/tracing_on
189  * file, which gives a means for the kernel and userspace to interact.
190  * Place a tracing_off() in the kernel where you want tracing to end.
191  * From user space, examine the trace, and then echo 1 > tracing_on
192  * to continue tracing.
193  *
194  * tracing_stop/tracing_start has slightly more overhead. It is used
195  * by things like suspend to ram where disabling the recording of the
196  * trace is not enough, but tracing must actually stop because things
197  * like calling smp_processor_id() may crash the system.
198  *
199  * Most likely, you want to use tracing_on/tracing_off.
200  */
201 
202 enum ftrace_dump_mode {
203 	DUMP_NONE,
204 	DUMP_ALL,
205 	DUMP_ORIG,
206 	DUMP_PARAM,
207 };
208 
209 #ifdef CONFIG_TRACING
210 void tracing_on(void);
211 void tracing_off(void);
212 int tracing_is_on(void);
213 void tracing_snapshot(void);
214 void tracing_snapshot_alloc(void);
215 
216 extern void tracing_start(void);
217 extern void tracing_stop(void);
218 
219 static inline __printf(1, 2)
____trace_printk_check_format(const char * fmt,...)220 void ____trace_printk_check_format(const char *fmt, ...)
221 {
222 }
223 #define __trace_printk_check_format(fmt, args...)			\
224 do {									\
225 	if (0)								\
226 		____trace_printk_check_format(fmt, ##args);		\
227 } while (0)
228 
229 /**
230  * trace_printk - printf formatting in the ftrace buffer
231  * @fmt: the printf format for printing
232  *
233  * Note: __trace_printk is an internal function for trace_printk() and
234  *       the @ip is passed in via the trace_printk() macro.
235  *
236  * This function allows a kernel developer to debug fast path sections
237  * that printk is not appropriate for. By scattering in various
238  * printk like tracing in the code, a developer can quickly see
239  * where problems are occurring.
240  *
241  * This is intended as a debugging tool for the developer only.
242  * Please refrain from leaving trace_printks scattered around in
243  * your code. (Extra memory is used for special buffers that are
244  * allocated when trace_printk() is used.)
245  *
246  * A little optimization trick is done here. If there's only one
247  * argument, there's no need to scan the string for printf formats.
248  * The trace_puts() will suffice. But how can we take advantage of
249  * using trace_puts() when trace_printk() has only one argument?
250  * By stringifying the args and checking the size we can tell
251  * whether or not there are args. __stringify((__VA_ARGS__)) will
252  * turn into "()\0" with a size of 3 when there are no args, anything
253  * else will be bigger. All we need to do is define a string to this,
254  * and then take its size and compare to 3. If it's bigger, use
255  * do_trace_printk() otherwise, optimize it to trace_puts(). Then just
256  * let gcc optimize the rest.
257  */
258 
259 #define trace_printk(fmt, ...)				\
260 do {							\
261 	char _______STR[] = __stringify((__VA_ARGS__));	\
262 	if (sizeof(_______STR) > 3)			\
263 		do_trace_printk(fmt, ##__VA_ARGS__);	\
264 	else						\
265 		trace_puts(fmt);			\
266 } while (0)
267 
268 #define do_trace_printk(fmt, args...)					\
269 do {									\
270 	static const char *trace_printk_fmt __used			\
271 		__section("__trace_printk_fmt") =			\
272 		__builtin_constant_p(fmt) ? fmt : NULL;			\
273 									\
274 	__trace_printk_check_format(fmt, ##args);			\
275 									\
276 	if (__builtin_constant_p(fmt))					\
277 		__trace_bprintk(_THIS_IP_, trace_printk_fmt, ##args);	\
278 	else								\
279 		__trace_printk(_THIS_IP_, fmt, ##args);			\
280 } while (0)
281 
282 extern __printf(2, 3)
283 int __trace_bprintk(unsigned long ip, const char *fmt, ...);
284 
285 extern __printf(2, 3)
286 int __trace_printk(unsigned long ip, const char *fmt, ...);
287 
288 /**
289  * trace_puts - write a string into the ftrace buffer
290  * @str: the string to record
291  *
292  * Note: __trace_bputs is an internal function for trace_puts and
293  *       the @ip is passed in via the trace_puts macro.
294  *
295  * This is similar to trace_printk() but is made for those really fast
296  * paths that a developer wants the least amount of "Heisenbug" effects,
297  * where the processing of the print format is still too much.
298  *
299  * This function allows a kernel developer to debug fast path sections
300  * that printk is not appropriate for. By scattering in various
301  * printk like tracing in the code, a developer can quickly see
302  * where problems are occurring.
303  *
304  * This is intended as a debugging tool for the developer only.
305  * Please refrain from leaving trace_puts scattered around in
306  * your code. (Extra memory is used for special buffers that are
307  * allocated when trace_puts() is used.)
308  *
309  * Returns: 0 if nothing was written, positive # if string was.
310  *  (1 when __trace_bputs is used, strlen(str) when __trace_puts is used)
311  */
312 
313 #define trace_puts(str) ({						\
314 	static const char *trace_printk_fmt __used			\
315 		__section("__trace_printk_fmt") =			\
316 		__builtin_constant_p(str) ? str : NULL;			\
317 									\
318 	if (__builtin_constant_p(str))					\
319 		__trace_bputs(_THIS_IP_, trace_printk_fmt);		\
320 	else								\
321 		__trace_puts(_THIS_IP_, str, strlen(str));		\
322 })
323 extern int __trace_bputs(unsigned long ip, const char *str);
324 extern int __trace_puts(unsigned long ip, const char *str, int size);
325 
326 extern void trace_dump_stack(int skip);
327 
328 /*
329  * The double __builtin_constant_p is because gcc will give us an error
330  * if we try to allocate the static variable to fmt if it is not a
331  * constant. Even with the outer if statement.
332  */
333 #define ftrace_vprintk(fmt, vargs)					\
334 do {									\
335 	if (__builtin_constant_p(fmt)) {				\
336 		static const char *trace_printk_fmt __used		\
337 		  __section("__trace_printk_fmt") =			\
338 			__builtin_constant_p(fmt) ? fmt : NULL;		\
339 									\
340 		__ftrace_vbprintk(_THIS_IP_, trace_printk_fmt, vargs);	\
341 	} else								\
342 		__ftrace_vprintk(_THIS_IP_, fmt, vargs);		\
343 } while (0)
344 
345 extern __printf(2, 0) int
346 __ftrace_vbprintk(unsigned long ip, const char *fmt, va_list ap);
347 
348 extern __printf(2, 0) int
349 __ftrace_vprintk(unsigned long ip, const char *fmt, va_list ap);
350 
351 extern void ftrace_dump(enum ftrace_dump_mode oops_dump_mode);
352 #else
tracing_start(void)353 static inline void tracing_start(void) { }
tracing_stop(void)354 static inline void tracing_stop(void) { }
trace_dump_stack(int skip)355 static inline void trace_dump_stack(int skip) { }
356 
tracing_on(void)357 static inline void tracing_on(void) { }
tracing_off(void)358 static inline void tracing_off(void) { }
tracing_is_on(void)359 static inline int tracing_is_on(void) { return 0; }
tracing_snapshot(void)360 static inline void tracing_snapshot(void) { }
tracing_snapshot_alloc(void)361 static inline void tracing_snapshot_alloc(void) { }
362 
363 static inline __printf(1, 2)
trace_printk(const char * fmt,...)364 int trace_printk(const char *fmt, ...)
365 {
366 	return 0;
367 }
368 static __printf(1, 0) inline int
ftrace_vprintk(const char * fmt,va_list ap)369 ftrace_vprintk(const char *fmt, va_list ap)
370 {
371 	return 0;
372 }
ftrace_dump(enum ftrace_dump_mode oops_dump_mode)373 static inline void ftrace_dump(enum ftrace_dump_mode oops_dump_mode) { }
374 #endif /* CONFIG_TRACING */
375 
376 /* Rebuild everything on CONFIG_FTRACE_MCOUNT_RECORD */
377 #ifdef CONFIG_FTRACE_MCOUNT_RECORD
378 # define REBUILD_DUE_TO_FTRACE_MCOUNT_RECORD
379 #endif
380 
381 /* Permissions on a sysfs file: you didn't miss the 0 prefix did you? */
382 #define VERIFY_OCTAL_PERMISSIONS(perms)						\
383 	(BUILD_BUG_ON_ZERO((perms) < 0) +					\
384 	 BUILD_BUG_ON_ZERO((perms) > 0777) +					\
385 	 /* USER_READABLE >= GROUP_READABLE >= OTHER_READABLE */		\
386 	 BUILD_BUG_ON_ZERO((((perms) >> 6) & 4) < (((perms) >> 3) & 4)) +	\
387 	 BUILD_BUG_ON_ZERO((((perms) >> 3) & 4) < ((perms) & 4)) +		\
388 	 /* USER_WRITABLE >= GROUP_WRITABLE */					\
389 	 BUILD_BUG_ON_ZERO((((perms) >> 6) & 2) < (((perms) >> 3) & 2)) +	\
390 	 /* OTHER_WRITABLE?  Generally considered a bad idea. */		\
391 	 BUILD_BUG_ON_ZERO((perms) & 2) +					\
392 	 (perms))
393 #endif
394