]>
Commit | Line | Data |
---|---|---|
1 | #ifndef _LINUX_SCHED_H | |
2 | #define _LINUX_SCHED_H | |
3 | ||
4 | #include <uapi/linux/sched.h> | |
5 | ||
6 | #include <linux/sched/prio.h> | |
7 | ||
8 | ||
9 | struct sched_param { | |
10 | int sched_priority; | |
11 | }; | |
12 | ||
13 | #include <asm/param.h> /* for HZ */ | |
14 | ||
15 | #include <linux/capability.h> | |
16 | #include <linux/threads.h> | |
17 | #include <linux/kernel.h> | |
18 | #include <linux/types.h> | |
19 | #include <linux/timex.h> | |
20 | #include <linux/jiffies.h> | |
21 | #include <linux/plist.h> | |
22 | #include <linux/rbtree.h> | |
23 | #include <linux/thread_info.h> | |
24 | #include <linux/cpumask.h> | |
25 | #include <linux/errno.h> | |
26 | #include <linux/nodemask.h> | |
27 | #include <linux/mm_types.h> | |
28 | #include <linux/preempt.h> | |
29 | ||
30 | #include <asm/page.h> | |
31 | #include <asm/ptrace.h> | |
32 | #include <linux/cputime.h> | |
33 | ||
34 | #include <linux/smp.h> | |
35 | #include <linux/sem.h> | |
36 | #include <linux/shm.h> | |
37 | #include <linux/signal.h> | |
38 | #include <linux/compiler.h> | |
39 | #include <linux/completion.h> | |
40 | #include <linux/pid.h> | |
41 | #include <linux/percpu.h> | |
42 | #include <linux/topology.h> | |
43 | #include <linux/seccomp.h> | |
44 | #include <linux/rcupdate.h> | |
45 | #include <linux/rculist.h> | |
46 | #include <linux/rtmutex.h> | |
47 | ||
48 | #include <linux/time.h> | |
49 | #include <linux/param.h> | |
50 | #include <linux/resource.h> | |
51 | #include <linux/timer.h> | |
52 | #include <linux/hrtimer.h> | |
53 | #include <linux/kcov.h> | |
54 | #include <linux/task_io_accounting.h> | |
55 | #include <linux/latencytop.h> | |
56 | #include <linux/cred.h> | |
57 | #include <linux/llist.h> | |
58 | #include <linux/uidgid.h> | |
59 | #include <linux/gfp.h> | |
60 | #include <linux/magic.h> | |
61 | #include <linux/cgroup-defs.h> | |
62 | ||
63 | #include <asm/processor.h> | |
64 | ||
65 | #define SCHED_ATTR_SIZE_VER0 48 /* sizeof first published struct */ | |
66 | ||
67 | /* | |
68 | * Extended scheduling parameters data structure. | |
69 | * | |
70 | * This is needed because the original struct sched_param can not be | |
71 | * altered without introducing ABI issues with legacy applications | |
72 | * (e.g., in sched_getparam()). | |
73 | * | |
74 | * However, the possibility of specifying more than just a priority for | |
75 | * the tasks may be useful for a wide variety of application fields, e.g., | |
76 | * multimedia, streaming, automation and control, and many others. | |
77 | * | |
78 | * This variant (sched_attr) is meant at describing a so-called | |
79 | * sporadic time-constrained task. In such model a task is specified by: | |
80 | * - the activation period or minimum instance inter-arrival time; | |
81 | * - the maximum (or average, depending on the actual scheduling | |
82 | * discipline) computation time of all instances, a.k.a. runtime; | |
83 | * - the deadline (relative to the actual activation time) of each | |
84 | * instance. | |
85 | * Very briefly, a periodic (sporadic) task asks for the execution of | |
86 | * some specific computation --which is typically called an instance-- | |
87 | * (at most) every period. Moreover, each instance typically lasts no more | |
88 | * than the runtime and must be completed by time instant t equal to | |
89 | * the instance activation time + the deadline. | |
90 | * | |
91 | * This is reflected by the actual fields of the sched_attr structure: | |
92 | * | |
93 | * @size size of the structure, for fwd/bwd compat. | |
94 | * | |
95 | * @sched_policy task's scheduling policy | |
96 | * @sched_flags for customizing the scheduler behaviour | |
97 | * @sched_nice task's nice value (SCHED_NORMAL/BATCH) | |
98 | * @sched_priority task's static priority (SCHED_FIFO/RR) | |
99 | * @sched_deadline representative of the task's deadline | |
100 | * @sched_runtime representative of the task's runtime | |
101 | * @sched_period representative of the task's period | |
102 | * | |
103 | * Given this task model, there are a multiplicity of scheduling algorithms | |
104 | * and policies, that can be used to ensure all the tasks will make their | |
105 | * timing constraints. | |
106 | * | |
107 | * As of now, the SCHED_DEADLINE policy (sched_dl scheduling class) is the | |
108 | * only user of this new interface. More information about the algorithm | |
109 | * available in the scheduling class file or in Documentation/. | |
110 | */ | |
111 | struct sched_attr { | |
112 | u32 size; | |
113 | ||
114 | u32 sched_policy; | |
115 | u64 sched_flags; | |
116 | ||
117 | /* SCHED_NORMAL, SCHED_BATCH */ | |
118 | s32 sched_nice; | |
119 | ||
120 | /* SCHED_FIFO, SCHED_RR */ | |
121 | u32 sched_priority; | |
122 | ||
123 | /* SCHED_DEADLINE */ | |
124 | u64 sched_runtime; | |
125 | u64 sched_deadline; | |
126 | u64 sched_period; | |
127 | }; | |
128 | ||
129 | struct futex_pi_state; | |
130 | struct robust_list_head; | |
131 | struct bio_list; | |
132 | struct fs_struct; | |
133 | struct perf_event_context; | |
134 | struct blk_plug; | |
135 | struct filename; | |
136 | struct nameidata; | |
137 | ||
138 | #define VMACACHE_BITS 2 | |
139 | #define VMACACHE_SIZE (1U << VMACACHE_BITS) | |
140 | #define VMACACHE_MASK (VMACACHE_SIZE - 1) | |
141 | ||
142 | /* | |
143 | * These are the constant used to fake the fixed-point load-average | |
144 | * counting. Some notes: | |
145 | * - 11 bit fractions expand to 22 bits by the multiplies: this gives | |
146 | * a load-average precision of 10 bits integer + 11 bits fractional | |
147 | * - if you want to count load-averages more often, you need more | |
148 | * precision, or rounding will get you. With 2-second counting freq, | |
149 | * the EXP_n values would be 1981, 2034 and 2043 if still using only | |
150 | * 11 bit fractions. | |
151 | */ | |
152 | extern unsigned long avenrun[]; /* Load averages */ | |
153 | extern void get_avenrun(unsigned long *loads, unsigned long offset, int shift); | |
154 | ||
155 | #define FSHIFT 11 /* nr of bits of precision */ | |
156 | #define FIXED_1 (1<<FSHIFT) /* 1.0 as fixed-point */ | |
157 | #define LOAD_FREQ (5*HZ+1) /* 5 sec intervals */ | |
158 | #define EXP_1 1884 /* 1/exp(5sec/1min) as fixed-point */ | |
159 | #define EXP_5 2014 /* 1/exp(5sec/5min) */ | |
160 | #define EXP_15 2037 /* 1/exp(5sec/15min) */ | |
161 | ||
162 | #define CALC_LOAD(load,exp,n) \ | |
163 | load *= exp; \ | |
164 | load += n*(FIXED_1-exp); \ | |
165 | load >>= FSHIFT; | |
166 | ||
167 | extern unsigned long total_forks; | |
168 | extern int nr_threads; | |
169 | DECLARE_PER_CPU(unsigned long, process_counts); | |
170 | extern int nr_processes(void); | |
171 | extern unsigned long nr_running(void); | |
172 | extern bool single_task_running(void); | |
173 | extern unsigned long nr_iowait(void); | |
174 | extern unsigned long nr_iowait_cpu(int cpu); | |
175 | extern void get_iowait_load(unsigned long *nr_waiters, unsigned long *load); | |
176 | ||
177 | extern void calc_global_load(unsigned long ticks); | |
178 | ||
179 | #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON) | |
180 | extern void cpu_load_update_nohz_start(void); | |
181 | extern void cpu_load_update_nohz_stop(void); | |
182 | #else | |
183 | static inline void cpu_load_update_nohz_start(void) { } | |
184 | static inline void cpu_load_update_nohz_stop(void) { } | |
185 | #endif | |
186 | ||
187 | extern void dump_cpu_task(int cpu); | |
188 | ||
189 | struct seq_file; | |
190 | struct cfs_rq; | |
191 | struct task_group; | |
192 | #ifdef CONFIG_SCHED_DEBUG | |
193 | extern void proc_sched_show_task(struct task_struct *p, struct seq_file *m); | |
194 | extern void proc_sched_set_task(struct task_struct *p); | |
195 | #endif | |
196 | ||
197 | /* | |
198 | * Task state bitmask. NOTE! These bits are also | |
199 | * encoded in fs/proc/array.c: get_task_state(). | |
200 | * | |
201 | * We have two separate sets of flags: task->state | |
202 | * is about runnability, while task->exit_state are | |
203 | * about the task exiting. Confusing, but this way | |
204 | * modifying one set can't modify the other one by | |
205 | * mistake. | |
206 | */ | |
207 | #define TASK_RUNNING 0 | |
208 | #define TASK_INTERRUPTIBLE 1 | |
209 | #define TASK_UNINTERRUPTIBLE 2 | |
210 | #define __TASK_STOPPED 4 | |
211 | #define __TASK_TRACED 8 | |
212 | /* in tsk->exit_state */ | |
213 | #define EXIT_DEAD 16 | |
214 | #define EXIT_ZOMBIE 32 | |
215 | #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD) | |
216 | /* in tsk->state again */ | |
217 | #define TASK_DEAD 64 | |
218 | #define TASK_WAKEKILL 128 | |
219 | #define TASK_WAKING 256 | |
220 | #define TASK_PARKED 512 | |
221 | #define TASK_NOLOAD 1024 | |
222 | #define TASK_NEW 2048 | |
223 | #define TASK_STATE_MAX 4096 | |
224 | ||
225 | #define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn" | |
226 | ||
227 | extern char ___assert_task_state[1 - 2*!!( | |
228 | sizeof(TASK_STATE_TO_CHAR_STR)-1 != ilog2(TASK_STATE_MAX)+1)]; | |
229 | ||
230 | /* Convenience macros for the sake of set_current_state */ | |
231 | #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE) | |
232 | #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED) | |
233 | #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED) | |
234 | ||
235 | #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD) | |
236 | ||
237 | /* Convenience macros for the sake of wake_up */ | |
238 | #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE) | |
239 | #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED) | |
240 | ||
241 | /* get_task_state() */ | |
242 | #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ | |
243 | TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ | |
244 | __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD) | |
245 | ||
246 | #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0) | |
247 | #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0) | |
248 | #define task_is_stopped_or_traced(task) \ | |
249 | ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0) | |
250 | #define task_contributes_to_load(task) \ | |
251 | ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \ | |
252 | (task->flags & PF_FROZEN) == 0 && \ | |
253 | (task->state & TASK_NOLOAD) == 0) | |
254 | ||
255 | #ifdef CONFIG_DEBUG_ATOMIC_SLEEP | |
256 | ||
257 | #define __set_current_state(state_value) \ | |
258 | do { \ | |
259 | current->task_state_change = _THIS_IP_; \ | |
260 | current->state = (state_value); \ | |
261 | } while (0) | |
262 | #define set_current_state(state_value) \ | |
263 | do { \ | |
264 | current->task_state_change = _THIS_IP_; \ | |
265 | smp_store_mb(current->state, (state_value)); \ | |
266 | } while (0) | |
267 | ||
268 | #else | |
269 | /* | |
270 | * set_current_state() includes a barrier so that the write of current->state | |
271 | * is correctly serialised wrt the caller's subsequent test of whether to | |
272 | * actually sleep: | |
273 | * | |
274 | * for (;;) { | |
275 | * set_current_state(TASK_UNINTERRUPTIBLE); | |
276 | * if (!need_sleep) | |
277 | * break; | |
278 | * | |
279 | * schedule(); | |
280 | * } | |
281 | * __set_current_state(TASK_RUNNING); | |
282 | * | |
283 | * If the caller does not need such serialisation (because, for instance, the | |
284 | * condition test and condition change and wakeup are under the same lock) then | |
285 | * use __set_current_state(). | |
286 | * | |
287 | * The above is typically ordered against the wakeup, which does: | |
288 | * | |
289 | * need_sleep = false; | |
290 | * wake_up_state(p, TASK_UNINTERRUPTIBLE); | |
291 | * | |
292 | * Where wake_up_state() (and all other wakeup primitives) imply enough | |
293 | * barriers to order the store of the variable against wakeup. | |
294 | * | |
295 | * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is, | |
296 | * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a | |
297 | * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING). | |
298 | * | |
299 | * This is obviously fine, since they both store the exact same value. | |
300 | * | |
301 | * Also see the comments of try_to_wake_up(). | |
302 | */ | |
303 | #define __set_current_state(state_value) \ | |
304 | do { current->state = (state_value); } while (0) | |
305 | #define set_current_state(state_value) \ | |
306 | smp_store_mb(current->state, (state_value)) | |
307 | ||
308 | #endif | |
309 | ||
310 | /* Task command name length */ | |
311 | #define TASK_COMM_LEN 16 | |
312 | ||
313 | #include <linux/spinlock.h> | |
314 | ||
315 | /* | |
316 | * This serializes "schedule()" and also protects | |
317 | * the run-queue from deletions/modifications (but | |
318 | * _adding_ to the beginning of the run-queue has | |
319 | * a separate lock). | |
320 | */ | |
321 | extern rwlock_t tasklist_lock; | |
322 | extern spinlock_t mmlist_lock; | |
323 | ||
324 | struct task_struct; | |
325 | ||
326 | #ifdef CONFIG_PROVE_RCU | |
327 | extern int lockdep_tasklist_lock_is_held(void); | |
328 | #endif /* #ifdef CONFIG_PROVE_RCU */ | |
329 | ||
330 | extern void sched_init(void); | |
331 | extern void sched_init_smp(void); | |
332 | extern asmlinkage void schedule_tail(struct task_struct *prev); | |
333 | extern void init_idle(struct task_struct *idle, int cpu); | |
334 | extern void init_idle_bootup_task(struct task_struct *idle); | |
335 | ||
336 | extern cpumask_var_t cpu_isolated_map; | |
337 | ||
338 | extern int runqueue_is_locked(int cpu); | |
339 | ||
340 | #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON) | |
341 | extern void nohz_balance_enter_idle(int cpu); | |
342 | extern void set_cpu_sd_state_idle(void); | |
343 | extern int get_nohz_timer_target(void); | |
344 | #else | |
345 | static inline void nohz_balance_enter_idle(int cpu) { } | |
346 | static inline void set_cpu_sd_state_idle(void) { } | |
347 | #endif | |
348 | ||
349 | /* | |
350 | * Only dump TASK_* tasks. (0 for all tasks) | |
351 | */ | |
352 | extern void show_state_filter(unsigned long state_filter); | |
353 | ||
354 | static inline void show_state(void) | |
355 | { | |
356 | show_state_filter(0); | |
357 | } | |
358 | ||
359 | extern void show_regs(struct pt_regs *); | |
360 | ||
361 | /* | |
362 | * TASK is a pointer to the task whose backtrace we want to see (or NULL for current | |
363 | * task), SP is the stack pointer of the first frame that should be shown in the back | |
364 | * trace (or NULL if the entire call-chain of the task should be shown). | |
365 | */ | |
366 | extern void show_stack(struct task_struct *task, unsigned long *sp); | |
367 | ||
368 | extern void cpu_init (void); | |
369 | extern void trap_init(void); | |
370 | extern void update_process_times(int user); | |
371 | extern void scheduler_tick(void); | |
372 | extern int sched_cpu_starting(unsigned int cpu); | |
373 | extern int sched_cpu_activate(unsigned int cpu); | |
374 | extern int sched_cpu_deactivate(unsigned int cpu); | |
375 | ||
376 | #ifdef CONFIG_HOTPLUG_CPU | |
377 | extern int sched_cpu_dying(unsigned int cpu); | |
378 | #else | |
379 | # define sched_cpu_dying NULL | |
380 | #endif | |
381 | ||
382 | extern void sched_show_task(struct task_struct *p); | |
383 | ||
384 | #ifdef CONFIG_LOCKUP_DETECTOR | |
385 | extern void touch_softlockup_watchdog_sched(void); | |
386 | extern void touch_softlockup_watchdog(void); | |
387 | extern void touch_softlockup_watchdog_sync(void); | |
388 | extern void touch_all_softlockup_watchdogs(void); | |
389 | extern int proc_dowatchdog_thresh(struct ctl_table *table, int write, | |
390 | void __user *buffer, | |
391 | size_t *lenp, loff_t *ppos); | |
392 | extern unsigned int softlockup_panic; | |
393 | extern unsigned int hardlockup_panic; | |
394 | void lockup_detector_init(void); | |
395 | #else | |
396 | static inline void touch_softlockup_watchdog_sched(void) | |
397 | { | |
398 | } | |
399 | static inline void touch_softlockup_watchdog(void) | |
400 | { | |
401 | } | |
402 | static inline void touch_softlockup_watchdog_sync(void) | |
403 | { | |
404 | } | |
405 | static inline void touch_all_softlockup_watchdogs(void) | |
406 | { | |
407 | } | |
408 | static inline void lockup_detector_init(void) | |
409 | { | |
410 | } | |
411 | #endif | |
412 | ||
413 | #ifdef CONFIG_DETECT_HUNG_TASK | |
414 | void reset_hung_task_detector(void); | |
415 | #else | |
416 | static inline void reset_hung_task_detector(void) | |
417 | { | |
418 | } | |
419 | #endif | |
420 | ||
421 | /* Attach to any functions which should be ignored in wchan output. */ | |
422 | #define __sched __attribute__((__section__(".sched.text"))) | |
423 | ||
424 | /* Linker adds these: start and end of __sched functions */ | |
425 | extern char __sched_text_start[], __sched_text_end[]; | |
426 | ||
427 | /* Is this address in the __sched functions? */ | |
428 | extern int in_sched_functions(unsigned long addr); | |
429 | ||
430 | #define MAX_SCHEDULE_TIMEOUT LONG_MAX | |
431 | extern signed long schedule_timeout(signed long timeout); | |
432 | extern signed long schedule_timeout_interruptible(signed long timeout); | |
433 | extern signed long schedule_timeout_killable(signed long timeout); | |
434 | extern signed long schedule_timeout_uninterruptible(signed long timeout); | |
435 | extern signed long schedule_timeout_idle(signed long timeout); | |
436 | asmlinkage void schedule(void); | |
437 | extern void schedule_preempt_disabled(void); | |
438 | ||
439 | extern long io_schedule_timeout(long timeout); | |
440 | ||
441 | static inline void io_schedule(void) | |
442 | { | |
443 | io_schedule_timeout(MAX_SCHEDULE_TIMEOUT); | |
444 | } | |
445 | ||
446 | void __noreturn do_task_dead(void); | |
447 | ||
448 | struct nsproxy; | |
449 | struct user_namespace; | |
450 | ||
451 | #ifdef CONFIG_MMU | |
452 | extern void arch_pick_mmap_layout(struct mm_struct *mm); | |
453 | extern unsigned long | |
454 | arch_get_unmapped_area(struct file *, unsigned long, unsigned long, | |
455 | unsigned long, unsigned long); | |
456 | extern unsigned long | |
457 | arch_get_unmapped_area_topdown(struct file *filp, unsigned long addr, | |
458 | unsigned long len, unsigned long pgoff, | |
459 | unsigned long flags); | |
460 | #else | |
461 | static inline void arch_pick_mmap_layout(struct mm_struct *mm) {} | |
462 | #endif | |
463 | ||
464 | #define SUID_DUMP_DISABLE 0 /* No setuid dumping */ | |
465 | #define SUID_DUMP_USER 1 /* Dump as user of process */ | |
466 | #define SUID_DUMP_ROOT 2 /* Dump as root */ | |
467 | ||
468 | /* mm flags */ | |
469 | ||
470 | /* for SUID_DUMP_* above */ | |
471 | #define MMF_DUMPABLE_BITS 2 | |
472 | #define MMF_DUMPABLE_MASK ((1 << MMF_DUMPABLE_BITS) - 1) | |
473 | ||
474 | extern void set_dumpable(struct mm_struct *mm, int value); | |
475 | /* | |
476 | * This returns the actual value of the suid_dumpable flag. For things | |
477 | * that are using this for checking for privilege transitions, it must | |
478 | * test against SUID_DUMP_USER rather than treating it as a boolean | |
479 | * value. | |
480 | */ | |
481 | static inline int __get_dumpable(unsigned long mm_flags) | |
482 | { | |
483 | return mm_flags & MMF_DUMPABLE_MASK; | |
484 | } | |
485 | ||
486 | static inline int get_dumpable(struct mm_struct *mm) | |
487 | { | |
488 | return __get_dumpable(mm->flags); | |
489 | } | |
490 | ||
491 | /* coredump filter bits */ | |
492 | #define MMF_DUMP_ANON_PRIVATE 2 | |
493 | #define MMF_DUMP_ANON_SHARED 3 | |
494 | #define MMF_DUMP_MAPPED_PRIVATE 4 | |
495 | #define MMF_DUMP_MAPPED_SHARED 5 | |
496 | #define MMF_DUMP_ELF_HEADERS 6 | |
497 | #define MMF_DUMP_HUGETLB_PRIVATE 7 | |
498 | #define MMF_DUMP_HUGETLB_SHARED 8 | |
499 | #define MMF_DUMP_DAX_PRIVATE 9 | |
500 | #define MMF_DUMP_DAX_SHARED 10 | |
501 | ||
502 | #define MMF_DUMP_FILTER_SHIFT MMF_DUMPABLE_BITS | |
503 | #define MMF_DUMP_FILTER_BITS 9 | |
504 | #define MMF_DUMP_FILTER_MASK \ | |
505 | (((1 << MMF_DUMP_FILTER_BITS) - 1) << MMF_DUMP_FILTER_SHIFT) | |
506 | #define MMF_DUMP_FILTER_DEFAULT \ | |
507 | ((1 << MMF_DUMP_ANON_PRIVATE) | (1 << MMF_DUMP_ANON_SHARED) |\ | |
508 | (1 << MMF_DUMP_HUGETLB_PRIVATE) | MMF_DUMP_MASK_DEFAULT_ELF) | |
509 | ||
510 | #ifdef CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS | |
511 | # define MMF_DUMP_MASK_DEFAULT_ELF (1 << MMF_DUMP_ELF_HEADERS) | |
512 | #else | |
513 | # define MMF_DUMP_MASK_DEFAULT_ELF 0 | |
514 | #endif | |
515 | /* leave room for more dump flags */ | |
516 | #define MMF_VM_MERGEABLE 16 /* KSM may merge identical pages */ | |
517 | #define MMF_VM_HUGEPAGE 17 /* set when VM_HUGEPAGE is set on vma */ | |
518 | /* | |
519 | * This one-shot flag is dropped due to necessity of changing exe once again | |
520 | * on NFS restore | |
521 | */ | |
522 | //#define MMF_EXE_FILE_CHANGED 18 /* see prctl_set_mm_exe_file() */ | |
523 | ||
524 | #define MMF_HAS_UPROBES 19 /* has uprobes */ | |
525 | #define MMF_RECALC_UPROBES 20 /* MMF_HAS_UPROBES can be wrong */ | |
526 | #define MMF_OOM_SKIP 21 /* mm is of no interest for the OOM killer */ | |
527 | #define MMF_UNSTABLE 22 /* mm is unstable for copy_from_user */ | |
528 | #define MMF_HUGE_ZERO_PAGE 23 /* mm has ever used the global huge zero page */ | |
529 | ||
530 | #define MMF_INIT_MASK (MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK) | |
531 | ||
532 | struct sighand_struct { | |
533 | atomic_t count; | |
534 | struct k_sigaction action[_NSIG]; | |
535 | spinlock_t siglock; | |
536 | wait_queue_head_t signalfd_wqh; | |
537 | }; | |
538 | ||
539 | struct pacct_struct { | |
540 | int ac_flag; | |
541 | long ac_exitcode; | |
542 | unsigned long ac_mem; | |
543 | cputime_t ac_utime, ac_stime; | |
544 | unsigned long ac_minflt, ac_majflt; | |
545 | }; | |
546 | ||
547 | struct cpu_itimer { | |
548 | cputime_t expires; | |
549 | cputime_t incr; | |
550 | u32 error; | |
551 | u32 incr_error; | |
552 | }; | |
553 | ||
554 | /** | |
555 | * struct prev_cputime - snaphsot of system and user cputime | |
556 | * @utime: time spent in user mode | |
557 | * @stime: time spent in system mode | |
558 | * @lock: protects the above two fields | |
559 | * | |
560 | * Stores previous user/system time values such that we can guarantee | |
561 | * monotonicity. | |
562 | */ | |
563 | struct prev_cputime { | |
564 | #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE | |
565 | cputime_t utime; | |
566 | cputime_t stime; | |
567 | raw_spinlock_t lock; | |
568 | #endif | |
569 | }; | |
570 | ||
571 | static inline void prev_cputime_init(struct prev_cputime *prev) | |
572 | { | |
573 | #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE | |
574 | prev->utime = prev->stime = 0; | |
575 | raw_spin_lock_init(&prev->lock); | |
576 | #endif | |
577 | } | |
578 | ||
579 | /** | |
580 | * struct task_cputime - collected CPU time counts | |
581 | * @utime: time spent in user mode, in &cputime_t units | |
582 | * @stime: time spent in kernel mode, in &cputime_t units | |
583 | * @sum_exec_runtime: total time spent on the CPU, in nanoseconds | |
584 | * | |
585 | * This structure groups together three kinds of CPU time that are tracked for | |
586 | * threads and thread groups. Most things considering CPU time want to group | |
587 | * these counts together and treat all three of them in parallel. | |
588 | */ | |
589 | struct task_cputime { | |
590 | cputime_t utime; | |
591 | cputime_t stime; | |
592 | unsigned long long sum_exec_runtime; | |
593 | }; | |
594 | ||
595 | /* Alternate field names when used to cache expirations. */ | |
596 | #define virt_exp utime | |
597 | #define prof_exp stime | |
598 | #define sched_exp sum_exec_runtime | |
599 | ||
600 | #define INIT_CPUTIME \ | |
601 | (struct task_cputime) { \ | |
602 | .utime = 0, \ | |
603 | .stime = 0, \ | |
604 | .sum_exec_runtime = 0, \ | |
605 | } | |
606 | ||
607 | /* | |
608 | * This is the atomic variant of task_cputime, which can be used for | |
609 | * storing and updating task_cputime statistics without locking. | |
610 | */ | |
611 | struct task_cputime_atomic { | |
612 | atomic64_t utime; | |
613 | atomic64_t stime; | |
614 | atomic64_t sum_exec_runtime; | |
615 | }; | |
616 | ||
617 | #define INIT_CPUTIME_ATOMIC \ | |
618 | (struct task_cputime_atomic) { \ | |
619 | .utime = ATOMIC64_INIT(0), \ | |
620 | .stime = ATOMIC64_INIT(0), \ | |
621 | .sum_exec_runtime = ATOMIC64_INIT(0), \ | |
622 | } | |
623 | ||
624 | #define PREEMPT_DISABLED (PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED) | |
625 | ||
626 | /* | |
627 | * Disable preemption until the scheduler is running -- use an unconditional | |
628 | * value so that it also works on !PREEMPT_COUNT kernels. | |
629 | * | |
630 | * Reset by start_kernel()->sched_init()->init_idle()->init_idle_preempt_count(). | |
631 | */ | |
632 | #define INIT_PREEMPT_COUNT PREEMPT_OFFSET | |
633 | ||
634 | /* | |
635 | * Initial preempt_count value; reflects the preempt_count schedule invariant | |
636 | * which states that during context switches: | |
637 | * | |
638 | * preempt_count() == 2*PREEMPT_DISABLE_OFFSET | |
639 | * | |
640 | * Note: PREEMPT_DISABLE_OFFSET is 0 for !PREEMPT_COUNT kernels. | |
641 | * Note: See finish_task_switch(). | |
642 | */ | |
643 | #define FORK_PREEMPT_COUNT (2*PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED) | |
644 | ||
645 | /** | |
646 | * struct thread_group_cputimer - thread group interval timer counts | |
647 | * @cputime_atomic: atomic thread group interval timers. | |
648 | * @running: true when there are timers running and | |
649 | * @cputime_atomic receives updates. | |
650 | * @checking_timer: true when a thread in the group is in the | |
651 | * process of checking for thread group timers. | |
652 | * | |
653 | * This structure contains the version of task_cputime, above, that is | |
654 | * used for thread group CPU timer calculations. | |
655 | */ | |
656 | struct thread_group_cputimer { | |
657 | struct task_cputime_atomic cputime_atomic; | |
658 | bool running; | |
659 | bool checking_timer; | |
660 | }; | |
661 | ||
662 | #include <linux/rwsem.h> | |
663 | struct autogroup; | |
664 | ||
665 | /* | |
666 | * NOTE! "signal_struct" does not have its own | |
667 | * locking, because a shared signal_struct always | |
668 | * implies a shared sighand_struct, so locking | |
669 | * sighand_struct is always a proper superset of | |
670 | * the locking of signal_struct. | |
671 | */ | |
672 | struct signal_struct { | |
673 | atomic_t sigcnt; | |
674 | atomic_t live; | |
675 | int nr_threads; | |
676 | struct list_head thread_head; | |
677 | ||
678 | wait_queue_head_t wait_chldexit; /* for wait4() */ | |
679 | ||
680 | /* current thread group signal load-balancing target: */ | |
681 | struct task_struct *curr_target; | |
682 | ||
683 | /* shared signal handling: */ | |
684 | struct sigpending shared_pending; | |
685 | ||
686 | /* thread group exit support */ | |
687 | int group_exit_code; | |
688 | /* overloaded: | |
689 | * - notify group_exit_task when ->count is equal to notify_count | |
690 | * - everyone except group_exit_task is stopped during signal delivery | |
691 | * of fatal signals, group_exit_task processes the signal. | |
692 | */ | |
693 | int notify_count; | |
694 | struct task_struct *group_exit_task; | |
695 | ||
696 | /* thread group stop support, overloads group_exit_code too */ | |
697 | int group_stop_count; | |
698 | unsigned int flags; /* see SIGNAL_* flags below */ | |
699 | ||
700 | /* | |
701 | * PR_SET_CHILD_SUBREAPER marks a process, like a service | |
702 | * manager, to re-parent orphan (double-forking) child processes | |
703 | * to this process instead of 'init'. The service manager is | |
704 | * able to receive SIGCHLD signals and is able to investigate | |
705 | * the process until it calls wait(). All children of this | |
706 | * process will inherit a flag if they should look for a | |
707 | * child_subreaper process at exit. | |
708 | */ | |
709 | unsigned int is_child_subreaper:1; | |
710 | unsigned int has_child_subreaper:1; | |
711 | ||
712 | /* POSIX.1b Interval Timers */ | |
713 | int posix_timer_id; | |
714 | struct list_head posix_timers; | |
715 | ||
716 | /* ITIMER_REAL timer for the process */ | |
717 | struct hrtimer real_timer; | |
718 | struct pid *leader_pid; | |
719 | ktime_t it_real_incr; | |
720 | ||
721 | /* | |
722 | * ITIMER_PROF and ITIMER_VIRTUAL timers for the process, we use | |
723 | * CPUCLOCK_PROF and CPUCLOCK_VIRT for indexing array as these | |
724 | * values are defined to 0 and 1 respectively | |
725 | */ | |
726 | struct cpu_itimer it[2]; | |
727 | ||
728 | /* | |
729 | * Thread group totals for process CPU timers. | |
730 | * See thread_group_cputimer(), et al, for details. | |
731 | */ | |
732 | struct thread_group_cputimer cputimer; | |
733 | ||
734 | /* Earliest-expiration cache. */ | |
735 | struct task_cputime cputime_expires; | |
736 | ||
737 | #ifdef CONFIG_NO_HZ_FULL | |
738 | atomic_t tick_dep_mask; | |
739 | #endif | |
740 | ||
741 | struct list_head cpu_timers[3]; | |
742 | ||
743 | struct pid *tty_old_pgrp; | |
744 | ||
745 | /* boolean value for session group leader */ | |
746 | int leader; | |
747 | ||
748 | struct tty_struct *tty; /* NULL if no tty */ | |
749 | ||
750 | #ifdef CONFIG_SCHED_AUTOGROUP | |
751 | struct autogroup *autogroup; | |
752 | #endif | |
753 | /* | |
754 | * Cumulative resource counters for dead threads in the group, | |
755 | * and for reaped dead child processes forked by this group. | |
756 | * Live threads maintain their own counters and add to these | |
757 | * in __exit_signal, except for the group leader. | |
758 | */ | |
759 | seqlock_t stats_lock; | |
760 | cputime_t utime, stime, cutime, cstime; | |
761 | cputime_t gtime; | |
762 | cputime_t cgtime; | |
763 | struct prev_cputime prev_cputime; | |
764 | unsigned long nvcsw, nivcsw, cnvcsw, cnivcsw; | |
765 | unsigned long min_flt, maj_flt, cmin_flt, cmaj_flt; | |
766 | unsigned long inblock, oublock, cinblock, coublock; | |
767 | unsigned long maxrss, cmaxrss; | |
768 | struct task_io_accounting ioac; | |
769 | ||
770 | /* | |
771 | * Cumulative ns of schedule CPU time fo dead threads in the | |
772 | * group, not including a zombie group leader, (This only differs | |
773 | * from jiffies_to_ns(utime + stime) if sched_clock uses something | |
774 | * other than jiffies.) | |
775 | */ | |
776 | unsigned long long sum_sched_runtime; | |
777 | ||
778 | /* | |
779 | * We don't bother to synchronize most readers of this at all, | |
780 | * because there is no reader checking a limit that actually needs | |
781 | * to get both rlim_cur and rlim_max atomically, and either one | |
782 | * alone is a single word that can safely be read normally. | |
783 | * getrlimit/setrlimit use task_lock(current->group_leader) to | |
784 | * protect this instead of the siglock, because they really | |
785 | * have no need to disable irqs. | |
786 | */ | |
787 | struct rlimit rlim[RLIM_NLIMITS]; | |
788 | ||
789 | #ifdef CONFIG_BSD_PROCESS_ACCT | |
790 | struct pacct_struct pacct; /* per-process accounting information */ | |
791 | #endif | |
792 | #ifdef CONFIG_TASKSTATS | |
793 | struct taskstats *stats; | |
794 | #endif | |
795 | #ifdef CONFIG_AUDIT | |
796 | unsigned audit_tty; | |
797 | struct tty_audit_buf *tty_audit_buf; | |
798 | #endif | |
799 | ||
800 | /* | |
801 | * Thread is the potential origin of an oom condition; kill first on | |
802 | * oom | |
803 | */ | |
804 | bool oom_flag_origin; | |
805 | short oom_score_adj; /* OOM kill score adjustment */ | |
806 | short oom_score_adj_min; /* OOM kill score adjustment min value. | |
807 | * Only settable by CAP_SYS_RESOURCE. */ | |
808 | struct mm_struct *oom_mm; /* recorded mm when the thread group got | |
809 | * killed by the oom killer */ | |
810 | ||
811 | struct mutex cred_guard_mutex; /* guard against foreign influences on | |
812 | * credential calculations | |
813 | * (notably. ptrace) */ | |
814 | }; | |
815 | ||
816 | /* | |
817 | * Bits in flags field of signal_struct. | |
818 | */ | |
819 | #define SIGNAL_STOP_STOPPED 0x00000001 /* job control stop in effect */ | |
820 | #define SIGNAL_STOP_CONTINUED 0x00000002 /* SIGCONT since WCONTINUED reap */ | |
821 | #define SIGNAL_GROUP_EXIT 0x00000004 /* group exit in progress */ | |
822 | #define SIGNAL_GROUP_COREDUMP 0x00000008 /* coredump in progress */ | |
823 | /* | |
824 | * Pending notifications to parent. | |
825 | */ | |
826 | #define SIGNAL_CLD_STOPPED 0x00000010 | |
827 | #define SIGNAL_CLD_CONTINUED 0x00000020 | |
828 | #define SIGNAL_CLD_MASK (SIGNAL_CLD_STOPPED|SIGNAL_CLD_CONTINUED) | |
829 | ||
830 | #define SIGNAL_UNKILLABLE 0x00000040 /* for init: ignore fatal signals */ | |
831 | ||
832 | #define SIGNAL_STOP_MASK (SIGNAL_CLD_MASK | SIGNAL_STOP_STOPPED | \ | |
833 | SIGNAL_STOP_CONTINUED) | |
834 | ||
835 | static inline void signal_set_stop_flags(struct signal_struct *sig, | |
836 | unsigned int flags) | |
837 | { | |
838 | WARN_ON(sig->flags & (SIGNAL_GROUP_EXIT|SIGNAL_GROUP_COREDUMP)); | |
839 | sig->flags = (sig->flags & ~SIGNAL_STOP_MASK) | flags; | |
840 | } | |
841 | ||
842 | /* If true, all threads except ->group_exit_task have pending SIGKILL */ | |
843 | static inline int signal_group_exit(const struct signal_struct *sig) | |
844 | { | |
845 | return (sig->flags & SIGNAL_GROUP_EXIT) || | |
846 | (sig->group_exit_task != NULL); | |
847 | } | |
848 | ||
849 | /* | |
850 | * Some day this will be a full-fledged user tracking system.. | |
851 | */ | |
852 | struct user_struct { | |
853 | atomic_t __count; /* reference count */ | |
854 | atomic_t processes; /* How many processes does this user have? */ | |
855 | atomic_t sigpending; /* How many pending signals does this user have? */ | |
856 | #ifdef CONFIG_INOTIFY_USER | |
857 | atomic_t inotify_watches; /* How many inotify watches does this user have? */ | |
858 | atomic_t inotify_devs; /* How many inotify devs does this user have opened? */ | |
859 | #endif | |
860 | #ifdef CONFIG_FANOTIFY | |
861 | atomic_t fanotify_listeners; | |
862 | #endif | |
863 | #ifdef CONFIG_EPOLL | |
864 | atomic_long_t epoll_watches; /* The number of file descriptors currently watched */ | |
865 | #endif | |
866 | #ifdef CONFIG_POSIX_MQUEUE | |
867 | /* protected by mq_lock */ | |
868 | unsigned long mq_bytes; /* How many bytes can be allocated to mqueue? */ | |
869 | #endif | |
870 | unsigned long locked_shm; /* How many pages of mlocked shm ? */ | |
871 | unsigned long unix_inflight; /* How many files in flight in unix sockets */ | |
872 | atomic_long_t pipe_bufs; /* how many pages are allocated in pipe buffers */ | |
873 | ||
874 | #ifdef CONFIG_KEYS | |
875 | struct key *uid_keyring; /* UID specific keyring */ | |
876 | struct key *session_keyring; /* UID's default session keyring */ | |
877 | #endif | |
878 | ||
879 | /* Hash table maintenance information */ | |
880 | struct hlist_node uidhash_node; | |
881 | kuid_t uid; | |
882 | ||
883 | #if defined(CONFIG_PERF_EVENTS) || defined(CONFIG_BPF_SYSCALL) | |
884 | atomic_long_t locked_vm; | |
885 | #endif | |
886 | }; | |
887 | ||
888 | extern int uids_sysfs_init(void); | |
889 | ||
890 | extern struct user_struct *find_user(kuid_t); | |
891 | ||
892 | extern struct user_struct root_user; | |
893 | #define INIT_USER (&root_user) | |
894 | ||
895 | ||
896 | struct backing_dev_info; | |
897 | struct reclaim_state; | |
898 | ||
899 | #ifdef CONFIG_SCHED_INFO | |
900 | struct sched_info { | |
901 | /* cumulative counters */ | |
902 | unsigned long pcount; /* # of times run on this cpu */ | |
903 | unsigned long long run_delay; /* time spent waiting on a runqueue */ | |
904 | ||
905 | /* timestamps */ | |
906 | unsigned long long last_arrival,/* when we last ran on a cpu */ | |
907 | last_queued; /* when we were last queued to run */ | |
908 | }; | |
909 | #endif /* CONFIG_SCHED_INFO */ | |
910 | ||
911 | #ifdef CONFIG_TASK_DELAY_ACCT | |
912 | struct task_delay_info { | |
913 | spinlock_t lock; | |
914 | unsigned int flags; /* Private per-task flags */ | |
915 | ||
916 | /* For each stat XXX, add following, aligned appropriately | |
917 | * | |
918 | * struct timespec XXX_start, XXX_end; | |
919 | * u64 XXX_delay; | |
920 | * u32 XXX_count; | |
921 | * | |
922 | * Atomicity of updates to XXX_delay, XXX_count protected by | |
923 | * single lock above (split into XXX_lock if contention is an issue). | |
924 | */ | |
925 | ||
926 | /* | |
927 | * XXX_count is incremented on every XXX operation, the delay | |
928 | * associated with the operation is added to XXX_delay. | |
929 | * XXX_delay contains the accumulated delay time in nanoseconds. | |
930 | */ | |
931 | u64 blkio_start; /* Shared by blkio, swapin */ | |
932 | u64 blkio_delay; /* wait for sync block io completion */ | |
933 | u64 swapin_delay; /* wait for swapin block io completion */ | |
934 | u32 blkio_count; /* total count of the number of sync block */ | |
935 | /* io operations performed */ | |
936 | u32 swapin_count; /* total count of the number of swapin block */ | |
937 | /* io operations performed */ | |
938 | ||
939 | u64 freepages_start; | |
940 | u64 freepages_delay; /* wait for memory reclaim */ | |
941 | u32 freepages_count; /* total count of memory reclaim */ | |
942 | }; | |
943 | #endif /* CONFIG_TASK_DELAY_ACCT */ | |
944 | ||
945 | static inline int sched_info_on(void) | |
946 | { | |
947 | #ifdef CONFIG_SCHEDSTATS | |
948 | return 1; | |
949 | #elif defined(CONFIG_TASK_DELAY_ACCT) | |
950 | extern int delayacct_on; | |
951 | return delayacct_on; | |
952 | #else | |
953 | return 0; | |
954 | #endif | |
955 | } | |
956 | ||
957 | #ifdef CONFIG_SCHEDSTATS | |
958 | void force_schedstat_enabled(void); | |
959 | #endif | |
960 | ||
961 | enum cpu_idle_type { | |
962 | CPU_IDLE, | |
963 | CPU_NOT_IDLE, | |
964 | CPU_NEWLY_IDLE, | |
965 | CPU_MAX_IDLE_TYPES | |
966 | }; | |
967 | ||
968 | /* | |
969 | * Integer metrics need fixed point arithmetic, e.g., sched/fair | |
970 | * has a few: load, load_avg, util_avg, freq, and capacity. | |
971 | * | |
972 | * We define a basic fixed point arithmetic range, and then formalize | |
973 | * all these metrics based on that basic range. | |
974 | */ | |
975 | # define SCHED_FIXEDPOINT_SHIFT 10 | |
976 | # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) | |
977 | ||
978 | /* | |
979 | * Increase resolution of cpu_capacity calculations | |
980 | */ | |
981 | #define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT | |
982 | #define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT) | |
983 | ||
984 | /* | |
985 | * Wake-queues are lists of tasks with a pending wakeup, whose | |
986 | * callers have already marked the task as woken internally, | |
987 | * and can thus carry on. A common use case is being able to | |
988 | * do the wakeups once the corresponding user lock as been | |
989 | * released. | |
990 | * | |
991 | * We hold reference to each task in the list across the wakeup, | |
992 | * thus guaranteeing that the memory is still valid by the time | |
993 | * the actual wakeups are performed in wake_up_q(). | |
994 | * | |
995 | * One per task suffices, because there's never a need for a task to be | |
996 | * in two wake queues simultaneously; it is forbidden to abandon a task | |
997 | * in a wake queue (a call to wake_up_q() _must_ follow), so if a task is | |
998 | * already in a wake queue, the wakeup will happen soon and the second | |
999 | * waker can just skip it. | |
1000 | * | |
1001 | * The DEFINE_WAKE_Q macro declares and initializes the list head. | |
1002 | * wake_up_q() does NOT reinitialize the list; it's expected to be | |
1003 | * called near the end of a function. Otherwise, the list can be | |
1004 | * re-initialized for later re-use by wake_q_init(). | |
1005 | * | |
1006 | * Note that this can cause spurious wakeups. schedule() callers | |
1007 | * must ensure the call is done inside a loop, confirming that the | |
1008 | * wakeup condition has in fact occurred. | |
1009 | */ | |
1010 | struct wake_q_node { | |
1011 | struct wake_q_node *next; | |
1012 | }; | |
1013 | ||
1014 | struct wake_q_head { | |
1015 | struct wake_q_node *first; | |
1016 | struct wake_q_node **lastp; | |
1017 | }; | |
1018 | ||
1019 | #define WAKE_Q_TAIL ((struct wake_q_node *) 0x01) | |
1020 | ||
1021 | #define DEFINE_WAKE_Q(name) \ | |
1022 | struct wake_q_head name = { WAKE_Q_TAIL, &name.first } | |
1023 | ||
1024 | static inline void wake_q_init(struct wake_q_head *head) | |
1025 | { | |
1026 | head->first = WAKE_Q_TAIL; | |
1027 | head->lastp = &head->first; | |
1028 | } | |
1029 | ||
1030 | extern void wake_q_add(struct wake_q_head *head, | |
1031 | struct task_struct *task); | |
1032 | extern void wake_up_q(struct wake_q_head *head); | |
1033 | ||
1034 | /* | |
1035 | * sched-domains (multiprocessor balancing) declarations: | |
1036 | */ | |
1037 | #ifdef CONFIG_SMP | |
1038 | #define SD_LOAD_BALANCE 0x0001 /* Do load balancing on this domain. */ | |
1039 | #define SD_BALANCE_NEWIDLE 0x0002 /* Balance when about to become idle */ | |
1040 | #define SD_BALANCE_EXEC 0x0004 /* Balance on exec */ | |
1041 | #define SD_BALANCE_FORK 0x0008 /* Balance on fork, clone */ | |
1042 | #define SD_BALANCE_WAKE 0x0010 /* Balance on wakeup */ | |
1043 | #define SD_WAKE_AFFINE 0x0020 /* Wake task to waking CPU */ | |
1044 | #define SD_ASYM_CPUCAPACITY 0x0040 /* Groups have different max cpu capacities */ | |
1045 | #define SD_SHARE_CPUCAPACITY 0x0080 /* Domain members share cpu capacity */ | |
1046 | #define SD_SHARE_POWERDOMAIN 0x0100 /* Domain members share power domain */ | |
1047 | #define SD_SHARE_PKG_RESOURCES 0x0200 /* Domain members share cpu pkg resources */ | |
1048 | #define SD_SERIALIZE 0x0400 /* Only a single load balancing instance */ | |
1049 | #define SD_ASYM_PACKING 0x0800 /* Place busy groups earlier in the domain */ | |
1050 | #define SD_PREFER_SIBLING 0x1000 /* Prefer to place tasks in a sibling domain */ | |
1051 | #define SD_OVERLAP 0x2000 /* sched_domains of this level overlap */ | |
1052 | #define SD_NUMA 0x4000 /* cross-node balancing */ | |
1053 | ||
1054 | #ifdef CONFIG_SCHED_SMT | |
1055 | static inline int cpu_smt_flags(void) | |
1056 | { | |
1057 | return SD_SHARE_CPUCAPACITY | SD_SHARE_PKG_RESOURCES; | |
1058 | } | |
1059 | #endif | |
1060 | ||
1061 | #ifdef CONFIG_SCHED_MC | |
1062 | static inline int cpu_core_flags(void) | |
1063 | { | |
1064 | return SD_SHARE_PKG_RESOURCES; | |
1065 | } | |
1066 | #endif | |
1067 | ||
1068 | #ifdef CONFIG_NUMA | |
1069 | static inline int cpu_numa_flags(void) | |
1070 | { | |
1071 | return SD_NUMA; | |
1072 | } | |
1073 | #endif | |
1074 | ||
1075 | extern int arch_asym_cpu_priority(int cpu); | |
1076 | ||
1077 | struct sched_domain_attr { | |
1078 | int relax_domain_level; | |
1079 | }; | |
1080 | ||
1081 | #define SD_ATTR_INIT (struct sched_domain_attr) { \ | |
1082 | .relax_domain_level = -1, \ | |
1083 | } | |
1084 | ||
1085 | extern int sched_domain_level_max; | |
1086 | ||
1087 | struct sched_group; | |
1088 | ||
1089 | struct sched_domain_shared { | |
1090 | atomic_t ref; | |
1091 | atomic_t nr_busy_cpus; | |
1092 | int has_idle_cores; | |
1093 | }; | |
1094 | ||
1095 | struct sched_domain { | |
1096 | /* These fields must be setup */ | |
1097 | struct sched_domain *parent; /* top domain must be null terminated */ | |
1098 | struct sched_domain *child; /* bottom domain must be null terminated */ | |
1099 | struct sched_group *groups; /* the balancing groups of the domain */ | |
1100 | unsigned long min_interval; /* Minimum balance interval ms */ | |
1101 | unsigned long max_interval; /* Maximum balance interval ms */ | |
1102 | unsigned int busy_factor; /* less balancing by factor if busy */ | |
1103 | unsigned int imbalance_pct; /* No balance until over watermark */ | |
1104 | unsigned int cache_nice_tries; /* Leave cache hot tasks for # tries */ | |
1105 | unsigned int busy_idx; | |
1106 | unsigned int idle_idx; | |
1107 | unsigned int newidle_idx; | |
1108 | unsigned int wake_idx; | |
1109 | unsigned int forkexec_idx; | |
1110 | unsigned int smt_gain; | |
1111 | ||
1112 | int nohz_idle; /* NOHZ IDLE status */ | |
1113 | int flags; /* See SD_* */ | |
1114 | int level; | |
1115 | ||
1116 | /* Runtime fields. */ | |
1117 | unsigned long last_balance; /* init to jiffies. units in jiffies */ | |
1118 | unsigned int balance_interval; /* initialise to 1. units in ms. */ | |
1119 | unsigned int nr_balance_failed; /* initialise to 0 */ | |
1120 | ||
1121 | /* idle_balance() stats */ | |
1122 | u64 max_newidle_lb_cost; | |
1123 | unsigned long next_decay_max_lb_cost; | |
1124 | ||
1125 | u64 avg_scan_cost; /* select_idle_sibling */ | |
1126 | ||
1127 | #ifdef CONFIG_SCHEDSTATS | |
1128 | /* load_balance() stats */ | |
1129 | unsigned int lb_count[CPU_MAX_IDLE_TYPES]; | |
1130 | unsigned int lb_failed[CPU_MAX_IDLE_TYPES]; | |
1131 | unsigned int lb_balanced[CPU_MAX_IDLE_TYPES]; | |
1132 | unsigned int lb_imbalance[CPU_MAX_IDLE_TYPES]; | |
1133 | unsigned int lb_gained[CPU_MAX_IDLE_TYPES]; | |
1134 | unsigned int lb_hot_gained[CPU_MAX_IDLE_TYPES]; | |
1135 | unsigned int lb_nobusyg[CPU_MAX_IDLE_TYPES]; | |
1136 | unsigned int lb_nobusyq[CPU_MAX_IDLE_TYPES]; | |
1137 | ||
1138 | /* Active load balancing */ | |
1139 | unsigned int alb_count; | |
1140 | unsigned int alb_failed; | |
1141 | unsigned int alb_pushed; | |
1142 | ||
1143 | /* SD_BALANCE_EXEC stats */ | |
1144 | unsigned int sbe_count; | |
1145 | unsigned int sbe_balanced; | |
1146 | unsigned int sbe_pushed; | |
1147 | ||
1148 | /* SD_BALANCE_FORK stats */ | |
1149 | unsigned int sbf_count; | |
1150 | unsigned int sbf_balanced; | |
1151 | unsigned int sbf_pushed; | |
1152 | ||
1153 | /* try_to_wake_up() stats */ | |
1154 | unsigned int ttwu_wake_remote; | |
1155 | unsigned int ttwu_move_affine; | |
1156 | unsigned int ttwu_move_balance; | |
1157 | #endif | |
1158 | #ifdef CONFIG_SCHED_DEBUG | |
1159 | char *name; | |
1160 | #endif | |
1161 | union { | |
1162 | void *private; /* used during construction */ | |
1163 | struct rcu_head rcu; /* used during destruction */ | |
1164 | }; | |
1165 | struct sched_domain_shared *shared; | |
1166 | ||
1167 | unsigned int span_weight; | |
1168 | /* | |
1169 | * Span of all CPUs in this domain. | |
1170 | * | |
1171 | * NOTE: this field is variable length. (Allocated dynamically | |
1172 | * by attaching extra space to the end of the structure, | |
1173 | * depending on how many CPUs the kernel has booted up with) | |
1174 | */ | |
1175 | unsigned long span[0]; | |
1176 | }; | |
1177 | ||
1178 | static inline struct cpumask *sched_domain_span(struct sched_domain *sd) | |
1179 | { | |
1180 | return to_cpumask(sd->span); | |
1181 | } | |
1182 | ||
1183 | extern void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[], | |
1184 | struct sched_domain_attr *dattr_new); | |
1185 | ||
1186 | /* Allocate an array of sched domains, for partition_sched_domains(). */ | |
1187 | cpumask_var_t *alloc_sched_domains(unsigned int ndoms); | |
1188 | void free_sched_domains(cpumask_var_t doms[], unsigned int ndoms); | |
1189 | ||
1190 | bool cpus_share_cache(int this_cpu, int that_cpu); | |
1191 | ||
1192 | typedef const struct cpumask *(*sched_domain_mask_f)(int cpu); | |
1193 | typedef int (*sched_domain_flags_f)(void); | |
1194 | ||
1195 | #define SDTL_OVERLAP 0x01 | |
1196 | ||
1197 | struct sd_data { | |
1198 | struct sched_domain **__percpu sd; | |
1199 | struct sched_domain_shared **__percpu sds; | |
1200 | struct sched_group **__percpu sg; | |
1201 | struct sched_group_capacity **__percpu sgc; | |
1202 | }; | |
1203 | ||
1204 | struct sched_domain_topology_level { | |
1205 | sched_domain_mask_f mask; | |
1206 | sched_domain_flags_f sd_flags; | |
1207 | int flags; | |
1208 | int numa_level; | |
1209 | struct sd_data data; | |
1210 | #ifdef CONFIG_SCHED_DEBUG | |
1211 | char *name; | |
1212 | #endif | |
1213 | }; | |
1214 | ||
1215 | extern void set_sched_topology(struct sched_domain_topology_level *tl); | |
1216 | extern void wake_up_if_idle(int cpu); | |
1217 | ||
1218 | #ifdef CONFIG_SCHED_DEBUG | |
1219 | # define SD_INIT_NAME(type) .name = #type | |
1220 | #else | |
1221 | # define SD_INIT_NAME(type) | |
1222 | #endif | |
1223 | ||
1224 | #else /* CONFIG_SMP */ | |
1225 | ||
1226 | struct sched_domain_attr; | |
1227 | ||
1228 | static inline void | |
1229 | partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[], | |
1230 | struct sched_domain_attr *dattr_new) | |
1231 | { | |
1232 | } | |
1233 | ||
1234 | static inline bool cpus_share_cache(int this_cpu, int that_cpu) | |
1235 | { | |
1236 | return true; | |
1237 | } | |
1238 | ||
1239 | #endif /* !CONFIG_SMP */ | |
1240 | ||
1241 | ||
1242 | struct io_context; /* See blkdev.h */ | |
1243 | ||
1244 | ||
1245 | #ifdef ARCH_HAS_PREFETCH_SWITCH_STACK | |
1246 | extern void prefetch_stack(struct task_struct *t); | |
1247 | #else | |
1248 | static inline void prefetch_stack(struct task_struct *t) { } | |
1249 | #endif | |
1250 | ||
1251 | struct audit_context; /* See audit.c */ | |
1252 | struct mempolicy; | |
1253 | struct pipe_inode_info; | |
1254 | struct uts_namespace; | |
1255 | ||
1256 | struct load_weight { | |
1257 | unsigned long weight; | |
1258 | u32 inv_weight; | |
1259 | }; | |
1260 | ||
1261 | /* | |
1262 | * The load_avg/util_avg accumulates an infinite geometric series | |
1263 | * (see __update_load_avg() in kernel/sched/fair.c). | |
1264 | * | |
1265 | * [load_avg definition] | |
1266 | * | |
1267 | * load_avg = runnable% * scale_load_down(load) | |
1268 | * | |
1269 | * where runnable% is the time ratio that a sched_entity is runnable. | |
1270 | * For cfs_rq, it is the aggregated load_avg of all runnable and | |
1271 | * blocked sched_entities. | |
1272 | * | |
1273 | * load_avg may also take frequency scaling into account: | |
1274 | * | |
1275 | * load_avg = runnable% * scale_load_down(load) * freq% | |
1276 | * | |
1277 | * where freq% is the CPU frequency normalized to the highest frequency. | |
1278 | * | |
1279 | * [util_avg definition] | |
1280 | * | |
1281 | * util_avg = running% * SCHED_CAPACITY_SCALE | |
1282 | * | |
1283 | * where running% is the time ratio that a sched_entity is running on | |
1284 | * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable | |
1285 | * and blocked sched_entities. | |
1286 | * | |
1287 | * util_avg may also factor frequency scaling and CPU capacity scaling: | |
1288 | * | |
1289 | * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity% | |
1290 | * | |
1291 | * where freq% is the same as above, and capacity% is the CPU capacity | |
1292 | * normalized to the greatest capacity (due to uarch differences, etc). | |
1293 | * | |
1294 | * N.B., the above ratios (runnable%, running%, freq%, and capacity%) | |
1295 | * themselves are in the range of [0, 1]. To do fixed point arithmetics, | |
1296 | * we therefore scale them to as large a range as necessary. This is for | |
1297 | * example reflected by util_avg's SCHED_CAPACITY_SCALE. | |
1298 | * | |
1299 | * [Overflow issue] | |
1300 | * | |
1301 | * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities | |
1302 | * with the highest load (=88761), always runnable on a single cfs_rq, | |
1303 | * and should not overflow as the number already hits PID_MAX_LIMIT. | |
1304 | * | |
1305 | * For all other cases (including 32-bit kernels), struct load_weight's | |
1306 | * weight will overflow first before we do, because: | |
1307 | * | |
1308 | * Max(load_avg) <= Max(load.weight) | |
1309 | * | |
1310 | * Then it is the load_weight's responsibility to consider overflow | |
1311 | * issues. | |
1312 | */ | |
1313 | struct sched_avg { | |
1314 | u64 last_update_time, load_sum; | |
1315 | u32 util_sum, period_contrib; | |
1316 | unsigned long load_avg, util_avg; | |
1317 | }; | |
1318 | ||
1319 | #ifdef CONFIG_SCHEDSTATS | |
1320 | struct sched_statistics { | |
1321 | u64 wait_start; | |
1322 | u64 wait_max; | |
1323 | u64 wait_count; | |
1324 | u64 wait_sum; | |
1325 | u64 iowait_count; | |
1326 | u64 iowait_sum; | |
1327 | ||
1328 | u64 sleep_start; | |
1329 | u64 sleep_max; | |
1330 | s64 sum_sleep_runtime; | |
1331 | ||
1332 | u64 block_start; | |
1333 | u64 block_max; | |
1334 | u64 exec_max; | |
1335 | u64 slice_max; | |
1336 | ||
1337 | u64 nr_migrations_cold; | |
1338 | u64 nr_failed_migrations_affine; | |
1339 | u64 nr_failed_migrations_running; | |
1340 | u64 nr_failed_migrations_hot; | |
1341 | u64 nr_forced_migrations; | |
1342 | ||
1343 | u64 nr_wakeups; | |
1344 | u64 nr_wakeups_sync; | |
1345 | u64 nr_wakeups_migrate; | |
1346 | u64 nr_wakeups_local; | |
1347 | u64 nr_wakeups_remote; | |
1348 | u64 nr_wakeups_affine; | |
1349 | u64 nr_wakeups_affine_attempts; | |
1350 | u64 nr_wakeups_passive; | |
1351 | u64 nr_wakeups_idle; | |
1352 | }; | |
1353 | #endif | |
1354 | ||
1355 | struct sched_entity { | |
1356 | struct load_weight load; /* for load-balancing */ | |
1357 | struct rb_node run_node; | |
1358 | struct list_head group_node; | |
1359 | unsigned int on_rq; | |
1360 | ||
1361 | u64 exec_start; | |
1362 | u64 sum_exec_runtime; | |
1363 | u64 vruntime; | |
1364 | u64 prev_sum_exec_runtime; | |
1365 | ||
1366 | u64 nr_migrations; | |
1367 | ||
1368 | #ifdef CONFIG_SCHEDSTATS | |
1369 | struct sched_statistics statistics; | |
1370 | #endif | |
1371 | ||
1372 | #ifdef CONFIG_FAIR_GROUP_SCHED | |
1373 | int depth; | |
1374 | struct sched_entity *parent; | |
1375 | /* rq on which this entity is (to be) queued: */ | |
1376 | struct cfs_rq *cfs_rq; | |
1377 | /* rq "owned" by this entity/group: */ | |
1378 | struct cfs_rq *my_q; | |
1379 | #endif | |
1380 | ||
1381 | #ifdef CONFIG_SMP | |
1382 | /* | |
1383 | * Per entity load average tracking. | |
1384 | * | |
1385 | * Put into separate cache line so it does not | |
1386 | * collide with read-mostly values above. | |
1387 | */ | |
1388 | struct sched_avg avg ____cacheline_aligned_in_smp; | |
1389 | #endif | |
1390 | }; | |
1391 | ||
1392 | struct sched_rt_entity { | |
1393 | struct list_head run_list; | |
1394 | unsigned long timeout; | |
1395 | unsigned long watchdog_stamp; | |
1396 | unsigned int time_slice; | |
1397 | unsigned short on_rq; | |
1398 | unsigned short on_list; | |
1399 | ||
1400 | struct sched_rt_entity *back; | |
1401 | #ifdef CONFIG_RT_GROUP_SCHED | |
1402 | struct sched_rt_entity *parent; | |
1403 | /* rq on which this entity is (to be) queued: */ | |
1404 | struct rt_rq *rt_rq; | |
1405 | /* rq "owned" by this entity/group: */ | |
1406 | struct rt_rq *my_q; | |
1407 | #endif | |
1408 | }; | |
1409 | ||
1410 | struct sched_dl_entity { | |
1411 | struct rb_node rb_node; | |
1412 | ||
1413 | /* | |
1414 | * Original scheduling parameters. Copied here from sched_attr | |
1415 | * during sched_setattr(), they will remain the same until | |
1416 | * the next sched_setattr(). | |
1417 | */ | |
1418 | u64 dl_runtime; /* maximum runtime for each instance */ | |
1419 | u64 dl_deadline; /* relative deadline of each instance */ | |
1420 | u64 dl_period; /* separation of two instances (period) */ | |
1421 | u64 dl_bw; /* dl_runtime / dl_deadline */ | |
1422 | ||
1423 | /* | |
1424 | * Actual scheduling parameters. Initialized with the values above, | |
1425 | * they are continously updated during task execution. Note that | |
1426 | * the remaining runtime could be < 0 in case we are in overrun. | |
1427 | */ | |
1428 | s64 runtime; /* remaining runtime for this instance */ | |
1429 | u64 deadline; /* absolute deadline for this instance */ | |
1430 | unsigned int flags; /* specifying the scheduler behaviour */ | |
1431 | ||
1432 | /* | |
1433 | * Some bool flags: | |
1434 | * | |
1435 | * @dl_throttled tells if we exhausted the runtime. If so, the | |
1436 | * task has to wait for a replenishment to be performed at the | |
1437 | * next firing of dl_timer. | |
1438 | * | |
1439 | * @dl_boosted tells if we are boosted due to DI. If so we are | |
1440 | * outside bandwidth enforcement mechanism (but only until we | |
1441 | * exit the critical section); | |
1442 | * | |
1443 | * @dl_yielded tells if task gave up the cpu before consuming | |
1444 | * all its available runtime during the last job. | |
1445 | */ | |
1446 | int dl_throttled, dl_boosted, dl_yielded; | |
1447 | ||
1448 | /* | |
1449 | * Bandwidth enforcement timer. Each -deadline task has its | |
1450 | * own bandwidth to be enforced, thus we need one timer per task. | |
1451 | */ | |
1452 | struct hrtimer dl_timer; | |
1453 | }; | |
1454 | ||
1455 | union rcu_special { | |
1456 | struct { | |
1457 | u8 blocked; | |
1458 | u8 need_qs; | |
1459 | u8 exp_need_qs; | |
1460 | u8 pad; /* Otherwise the compiler can store garbage here. */ | |
1461 | } b; /* Bits. */ | |
1462 | u32 s; /* Set of bits. */ | |
1463 | }; | |
1464 | struct rcu_node; | |
1465 | ||
1466 | enum perf_event_task_context { | |
1467 | perf_invalid_context = -1, | |
1468 | perf_hw_context = 0, | |
1469 | perf_sw_context, | |
1470 | perf_nr_task_contexts, | |
1471 | }; | |
1472 | ||
1473 | /* Track pages that require TLB flushes */ | |
1474 | struct tlbflush_unmap_batch { | |
1475 | /* | |
1476 | * Each bit set is a CPU that potentially has a TLB entry for one of | |
1477 | * the PFNs being flushed. See set_tlb_ubc_flush_pending(). | |
1478 | */ | |
1479 | struct cpumask cpumask; | |
1480 | ||
1481 | /* True if any bit in cpumask is set */ | |
1482 | bool flush_required; | |
1483 | ||
1484 | /* | |
1485 | * If true then the PTE was dirty when unmapped. The entry must be | |
1486 | * flushed before IO is initiated or a stale TLB entry potentially | |
1487 | * allows an update without redirtying the page. | |
1488 | */ | |
1489 | bool writable; | |
1490 | }; | |
1491 | ||
1492 | struct task_struct { | |
1493 | #ifdef CONFIG_THREAD_INFO_IN_TASK | |
1494 | /* | |
1495 | * For reasons of header soup (see current_thread_info()), this | |
1496 | * must be the first element of task_struct. | |
1497 | */ | |
1498 | struct thread_info thread_info; | |
1499 | #endif | |
1500 | volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */ | |
1501 | void *stack; | |
1502 | atomic_t usage; | |
1503 | unsigned int flags; /* per process flags, defined below */ | |
1504 | unsigned int ptrace; | |
1505 | ||
1506 | #ifdef CONFIG_SMP | |
1507 | struct llist_node wake_entry; | |
1508 | int on_cpu; | |
1509 | #ifdef CONFIG_THREAD_INFO_IN_TASK | |
1510 | unsigned int cpu; /* current CPU */ | |
1511 | #endif | |
1512 | unsigned int wakee_flips; | |
1513 | unsigned long wakee_flip_decay_ts; | |
1514 | struct task_struct *last_wakee; | |
1515 | ||
1516 | int wake_cpu; | |
1517 | #endif | |
1518 | int on_rq; | |
1519 | ||
1520 | int prio, static_prio, normal_prio; | |
1521 | unsigned int rt_priority; | |
1522 | const struct sched_class *sched_class; | |
1523 | struct sched_entity se; | |
1524 | struct sched_rt_entity rt; | |
1525 | #ifdef CONFIG_CGROUP_SCHED | |
1526 | struct task_group *sched_task_group; | |
1527 | #endif | |
1528 | struct sched_dl_entity dl; | |
1529 | ||
1530 | #ifdef CONFIG_PREEMPT_NOTIFIERS | |
1531 | /* list of struct preempt_notifier: */ | |
1532 | struct hlist_head preempt_notifiers; | |
1533 | #endif | |
1534 | ||
1535 | #ifdef CONFIG_BLK_DEV_IO_TRACE | |
1536 | unsigned int btrace_seq; | |
1537 | #endif | |
1538 | ||
1539 | unsigned int policy; | |
1540 | int nr_cpus_allowed; | |
1541 | cpumask_t cpus_allowed; | |
1542 | ||
1543 | #ifdef CONFIG_PREEMPT_RCU | |
1544 | int rcu_read_lock_nesting; | |
1545 | union rcu_special rcu_read_unlock_special; | |
1546 | struct list_head rcu_node_entry; | |
1547 | struct rcu_node *rcu_blocked_node; | |
1548 | #endif /* #ifdef CONFIG_PREEMPT_RCU */ | |
1549 | #ifdef CONFIG_TASKS_RCU | |
1550 | unsigned long rcu_tasks_nvcsw; | |
1551 | bool rcu_tasks_holdout; | |
1552 | struct list_head rcu_tasks_holdout_list; | |
1553 | int rcu_tasks_idle_cpu; | |
1554 | #endif /* #ifdef CONFIG_TASKS_RCU */ | |
1555 | ||
1556 | #ifdef CONFIG_SCHED_INFO | |
1557 | struct sched_info sched_info; | |
1558 | #endif | |
1559 | ||
1560 | struct list_head tasks; | |
1561 | #ifdef CONFIG_SMP | |
1562 | struct plist_node pushable_tasks; | |
1563 | struct rb_node pushable_dl_tasks; | |
1564 | #endif | |
1565 | ||
1566 | struct mm_struct *mm, *active_mm; | |
1567 | /* per-thread vma caching */ | |
1568 | u32 vmacache_seqnum; | |
1569 | struct vm_area_struct *vmacache[VMACACHE_SIZE]; | |
1570 | #if defined(SPLIT_RSS_COUNTING) | |
1571 | struct task_rss_stat rss_stat; | |
1572 | #endif | |
1573 | /* task state */ | |
1574 | int exit_state; | |
1575 | int exit_code, exit_signal; | |
1576 | int pdeath_signal; /* The signal sent when the parent dies */ | |
1577 | unsigned long jobctl; /* JOBCTL_*, siglock protected */ | |
1578 | ||
1579 | /* Used for emulating ABI behavior of previous Linux versions */ | |
1580 | unsigned int personality; | |
1581 | ||
1582 | /* scheduler bits, serialized by scheduler locks */ | |
1583 | unsigned sched_reset_on_fork:1; | |
1584 | unsigned sched_contributes_to_load:1; | |
1585 | unsigned sched_migrated:1; | |
1586 | unsigned sched_remote_wakeup:1; | |
1587 | unsigned :0; /* force alignment to the next boundary */ | |
1588 | ||
1589 | /* unserialized, strictly 'current' */ | |
1590 | unsigned in_execve:1; /* bit to tell LSMs we're in execve */ | |
1591 | unsigned in_iowait:1; | |
1592 | #if !defined(TIF_RESTORE_SIGMASK) | |
1593 | unsigned restore_sigmask:1; | |
1594 | #endif | |
1595 | #ifdef CONFIG_MEMCG | |
1596 | unsigned memcg_may_oom:1; | |
1597 | #ifndef CONFIG_SLOB | |
1598 | unsigned memcg_kmem_skip_account:1; | |
1599 | #endif | |
1600 | #endif | |
1601 | #ifdef CONFIG_COMPAT_BRK | |
1602 | unsigned brk_randomized:1; | |
1603 | #endif | |
1604 | ||
1605 | unsigned long atomic_flags; /* Flags needing atomic access. */ | |
1606 | ||
1607 | struct restart_block restart_block; | |
1608 | ||
1609 | pid_t pid; | |
1610 | pid_t tgid; | |
1611 | ||
1612 | #ifdef CONFIG_CC_STACKPROTECTOR | |
1613 | /* Canary value for the -fstack-protector gcc feature */ | |
1614 | unsigned long stack_canary; | |
1615 | #endif | |
1616 | /* | |
1617 | * pointers to (original) parent process, youngest child, younger sibling, | |
1618 | * older sibling, respectively. (p->father can be replaced with | |
1619 | * p->real_parent->pid) | |
1620 | */ | |
1621 | struct task_struct __rcu *real_parent; /* real parent process */ | |
1622 | struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */ | |
1623 | /* | |
1624 | * children/sibling forms the list of my natural children | |
1625 | */ | |
1626 | struct list_head children; /* list of my children */ | |
1627 | struct list_head sibling; /* linkage in my parent's children list */ | |
1628 | struct task_struct *group_leader; /* threadgroup leader */ | |
1629 | ||
1630 | /* | |
1631 | * ptraced is the list of tasks this task is using ptrace on. | |
1632 | * This includes both natural children and PTRACE_ATTACH targets. | |
1633 | * p->ptrace_entry is p's link on the p->parent->ptraced list. | |
1634 | */ | |
1635 | struct list_head ptraced; | |
1636 | struct list_head ptrace_entry; | |
1637 | ||
1638 | /* PID/PID hash table linkage. */ | |
1639 | struct pid_link pids[PIDTYPE_MAX]; | |
1640 | struct list_head thread_group; | |
1641 | struct list_head thread_node; | |
1642 | ||
1643 | struct completion *vfork_done; /* for vfork() */ | |
1644 | int __user *set_child_tid; /* CLONE_CHILD_SETTID */ | |
1645 | int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */ | |
1646 | ||
1647 | cputime_t utime, stime; | |
1648 | #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME | |
1649 | cputime_t utimescaled, stimescaled; | |
1650 | #endif | |
1651 | cputime_t gtime; | |
1652 | struct prev_cputime prev_cputime; | |
1653 | #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN | |
1654 | seqcount_t vtime_seqcount; | |
1655 | unsigned long long vtime_snap; | |
1656 | enum { | |
1657 | /* Task is sleeping or running in a CPU with VTIME inactive */ | |
1658 | VTIME_INACTIVE = 0, | |
1659 | /* Task runs in userspace in a CPU with VTIME active */ | |
1660 | VTIME_USER, | |
1661 | /* Task runs in kernelspace in a CPU with VTIME active */ | |
1662 | VTIME_SYS, | |
1663 | } vtime_snap_whence; | |
1664 | #endif | |
1665 | ||
1666 | #ifdef CONFIG_NO_HZ_FULL | |
1667 | atomic_t tick_dep_mask; | |
1668 | #endif | |
1669 | unsigned long nvcsw, nivcsw; /* context switch counts */ | |
1670 | u64 start_time; /* monotonic time in nsec */ | |
1671 | u64 real_start_time; /* boot based time in nsec */ | |
1672 | /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */ | |
1673 | unsigned long min_flt, maj_flt; | |
1674 | ||
1675 | struct task_cputime cputime_expires; | |
1676 | struct list_head cpu_timers[3]; | |
1677 | ||
1678 | /* process credentials */ | |
1679 | const struct cred __rcu *ptracer_cred; /* Tracer's credentials at attach */ | |
1680 | const struct cred __rcu *real_cred; /* objective and real subjective task | |
1681 | * credentials (COW) */ | |
1682 | const struct cred __rcu *cred; /* effective (overridable) subjective task | |
1683 | * credentials (COW) */ | |
1684 | char comm[TASK_COMM_LEN]; /* executable name excluding path | |
1685 | - access with [gs]et_task_comm (which lock | |
1686 | it with task_lock()) | |
1687 | - initialized normally by setup_new_exec */ | |
1688 | /* file system info */ | |
1689 | struct nameidata *nameidata; | |
1690 | #ifdef CONFIG_SYSVIPC | |
1691 | /* ipc stuff */ | |
1692 | struct sysv_sem sysvsem; | |
1693 | struct sysv_shm sysvshm; | |
1694 | #endif | |
1695 | #ifdef CONFIG_DETECT_HUNG_TASK | |
1696 | /* hung task detection */ | |
1697 | unsigned long last_switch_count; | |
1698 | #endif | |
1699 | /* filesystem information */ | |
1700 | struct fs_struct *fs; | |
1701 | /* open file information */ | |
1702 | struct files_struct *files; | |
1703 | /* namespaces */ | |
1704 | struct nsproxy *nsproxy; | |
1705 | /* signal handlers */ | |
1706 | struct signal_struct *signal; | |
1707 | struct sighand_struct *sighand; | |
1708 | ||
1709 | sigset_t blocked, real_blocked; | |
1710 | sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */ | |
1711 | struct sigpending pending; | |
1712 | ||
1713 | unsigned long sas_ss_sp; | |
1714 | size_t sas_ss_size; | |
1715 | unsigned sas_ss_flags; | |
1716 | ||
1717 | struct callback_head *task_works; | |
1718 | ||
1719 | struct audit_context *audit_context; | |
1720 | #ifdef CONFIG_AUDITSYSCALL | |
1721 | kuid_t loginuid; | |
1722 | unsigned int sessionid; | |
1723 | #endif | |
1724 | struct seccomp seccomp; | |
1725 | ||
1726 | /* Thread group tracking */ | |
1727 | u32 parent_exec_id; | |
1728 | u32 self_exec_id; | |
1729 | /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, | |
1730 | * mempolicy */ | |
1731 | spinlock_t alloc_lock; | |
1732 | ||
1733 | /* Protection of the PI data structures: */ | |
1734 | raw_spinlock_t pi_lock; | |
1735 | ||
1736 | struct wake_q_node wake_q; | |
1737 | ||
1738 | #ifdef CONFIG_RT_MUTEXES | |
1739 | /* PI waiters blocked on a rt_mutex held by this task */ | |
1740 | struct rb_root pi_waiters; | |
1741 | struct rb_node *pi_waiters_leftmost; | |
1742 | /* Deadlock detection and priority inheritance handling */ | |
1743 | struct rt_mutex_waiter *pi_blocked_on; | |
1744 | #endif | |
1745 | ||
1746 | #ifdef CONFIG_DEBUG_MUTEXES | |
1747 | /* mutex deadlock detection */ | |
1748 | struct mutex_waiter *blocked_on; | |
1749 | #endif | |
1750 | #ifdef CONFIG_TRACE_IRQFLAGS | |
1751 | unsigned int irq_events; | |
1752 | unsigned long hardirq_enable_ip; | |
1753 | unsigned long hardirq_disable_ip; | |
1754 | unsigned int hardirq_enable_event; | |
1755 | unsigned int hardirq_disable_event; | |
1756 | int hardirqs_enabled; | |
1757 | int hardirq_context; | |
1758 | unsigned long softirq_disable_ip; | |
1759 | unsigned long softirq_enable_ip; | |
1760 | unsigned int softirq_disable_event; | |
1761 | unsigned int softirq_enable_event; | |
1762 | int softirqs_enabled; | |
1763 | int softirq_context; | |
1764 | #endif | |
1765 | #ifdef CONFIG_LOCKDEP | |
1766 | # define MAX_LOCK_DEPTH 48UL | |
1767 | u64 curr_chain_key; | |
1768 | int lockdep_depth; | |
1769 | unsigned int lockdep_recursion; | |
1770 | struct held_lock held_locks[MAX_LOCK_DEPTH]; | |
1771 | gfp_t lockdep_reclaim_gfp; | |
1772 | #endif | |
1773 | #ifdef CONFIG_UBSAN | |
1774 | unsigned int in_ubsan; | |
1775 | #endif | |
1776 | ||
1777 | /* journalling filesystem info */ | |
1778 | void *journal_info; | |
1779 | ||
1780 | /* stacked block device info */ | |
1781 | struct bio_list *bio_list; | |
1782 | ||
1783 | #ifdef CONFIG_BLOCK | |
1784 | /* stack plugging */ | |
1785 | struct blk_plug *plug; | |
1786 | #endif | |
1787 | ||
1788 | /* VM state */ | |
1789 | struct reclaim_state *reclaim_state; | |
1790 | ||
1791 | struct backing_dev_info *backing_dev_info; | |
1792 | ||
1793 | struct io_context *io_context; | |
1794 | ||
1795 | unsigned long ptrace_message; | |
1796 | siginfo_t *last_siginfo; /* For ptrace use. */ | |
1797 | struct task_io_accounting ioac; | |
1798 | #if defined(CONFIG_TASK_XACCT) | |
1799 | u64 acct_rss_mem1; /* accumulated rss usage */ | |
1800 | u64 acct_vm_mem1; /* accumulated virtual memory usage */ | |
1801 | cputime_t acct_timexpd; /* stime + utime since last update */ | |
1802 | #endif | |
1803 | #ifdef CONFIG_CPUSETS | |
1804 | nodemask_t mems_allowed; /* Protected by alloc_lock */ | |
1805 | seqcount_t mems_allowed_seq; /* Seqence no to catch updates */ | |
1806 | int cpuset_mem_spread_rotor; | |
1807 | int cpuset_slab_spread_rotor; | |
1808 | #endif | |
1809 | #ifdef CONFIG_CGROUPS | |
1810 | /* Control Group info protected by css_set_lock */ | |
1811 | struct css_set __rcu *cgroups; | |
1812 | /* cg_list protected by css_set_lock and tsk->alloc_lock */ | |
1813 | struct list_head cg_list; | |
1814 | #endif | |
1815 | #ifdef CONFIG_INTEL_RDT_A | |
1816 | int closid; | |
1817 | #endif | |
1818 | #ifdef CONFIG_FUTEX | |
1819 | struct robust_list_head __user *robust_list; | |
1820 | #ifdef CONFIG_COMPAT | |
1821 | struct compat_robust_list_head __user *compat_robust_list; | |
1822 | #endif | |
1823 | struct list_head pi_state_list; | |
1824 | struct futex_pi_state *pi_state_cache; | |
1825 | #endif | |
1826 | #ifdef CONFIG_PERF_EVENTS | |
1827 | struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts]; | |
1828 | struct mutex perf_event_mutex; | |
1829 | struct list_head perf_event_list; | |
1830 | #endif | |
1831 | #ifdef CONFIG_DEBUG_PREEMPT | |
1832 | unsigned long preempt_disable_ip; | |
1833 | #endif | |
1834 | #ifdef CONFIG_NUMA | |
1835 | struct mempolicy *mempolicy; /* Protected by alloc_lock */ | |
1836 | short il_next; | |
1837 | short pref_node_fork; | |
1838 | #endif | |
1839 | #ifdef CONFIG_NUMA_BALANCING | |
1840 | int numa_scan_seq; | |
1841 | unsigned int numa_scan_period; | |
1842 | unsigned int numa_scan_period_max; | |
1843 | int numa_preferred_nid; | |
1844 | unsigned long numa_migrate_retry; | |
1845 | u64 node_stamp; /* migration stamp */ | |
1846 | u64 last_task_numa_placement; | |
1847 | u64 last_sum_exec_runtime; | |
1848 | struct callback_head numa_work; | |
1849 | ||
1850 | struct list_head numa_entry; | |
1851 | struct numa_group *numa_group; | |
1852 | ||
1853 | /* | |
1854 | * numa_faults is an array split into four regions: | |
1855 | * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer | |
1856 | * in this precise order. | |
1857 | * | |
1858 | * faults_memory: Exponential decaying average of faults on a per-node | |
1859 | * basis. Scheduling placement decisions are made based on these | |
1860 | * counts. The values remain static for the duration of a PTE scan. | |
1861 | * faults_cpu: Track the nodes the process was running on when a NUMA | |
1862 | * hinting fault was incurred. | |
1863 | * faults_memory_buffer and faults_cpu_buffer: Record faults per node | |
1864 | * during the current scan window. When the scan completes, the counts | |
1865 | * in faults_memory and faults_cpu decay and these values are copied. | |
1866 | */ | |
1867 | unsigned long *numa_faults; | |
1868 | unsigned long total_numa_faults; | |
1869 | ||
1870 | /* | |
1871 | * numa_faults_locality tracks if faults recorded during the last | |
1872 | * scan window were remote/local or failed to migrate. The task scan | |
1873 | * period is adapted based on the locality of the faults with different | |
1874 | * weights depending on whether they were shared or private faults | |
1875 | */ | |
1876 | unsigned long numa_faults_locality[3]; | |
1877 | ||
1878 | unsigned long numa_pages_migrated; | |
1879 | #endif /* CONFIG_NUMA_BALANCING */ | |
1880 | ||
1881 | #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH | |
1882 | struct tlbflush_unmap_batch tlb_ubc; | |
1883 | #endif | |
1884 | ||
1885 | struct rcu_head rcu; | |
1886 | ||
1887 | /* | |
1888 | * cache last used pipe for splice | |
1889 | */ | |
1890 | struct pipe_inode_info *splice_pipe; | |
1891 | ||
1892 | struct page_frag task_frag; | |
1893 | ||
1894 | #ifdef CONFIG_TASK_DELAY_ACCT | |
1895 | struct task_delay_info *delays; | |
1896 | #endif | |
1897 | #ifdef CONFIG_FAULT_INJECTION | |
1898 | int make_it_fail; | |
1899 | #endif | |
1900 | /* | |
1901 | * when (nr_dirtied >= nr_dirtied_pause), it's time to call | |
1902 | * balance_dirty_pages() for some dirty throttling pause | |
1903 | */ | |
1904 | int nr_dirtied; | |
1905 | int nr_dirtied_pause; | |
1906 | unsigned long dirty_paused_when; /* start of a write-and-pause period */ | |
1907 | ||
1908 | #ifdef CONFIG_LATENCYTOP | |
1909 | int latency_record_count; | |
1910 | struct latency_record latency_record[LT_SAVECOUNT]; | |
1911 | #endif | |
1912 | /* | |
1913 | * time slack values; these are used to round up poll() and | |
1914 | * select() etc timeout values. These are in nanoseconds. | |
1915 | */ | |
1916 | u64 timer_slack_ns; | |
1917 | u64 default_timer_slack_ns; | |
1918 | ||
1919 | #ifdef CONFIG_KASAN | |
1920 | unsigned int kasan_depth; | |
1921 | #endif | |
1922 | #ifdef CONFIG_FUNCTION_GRAPH_TRACER | |
1923 | /* Index of current stored address in ret_stack */ | |
1924 | int curr_ret_stack; | |
1925 | /* Stack of return addresses for return function tracing */ | |
1926 | struct ftrace_ret_stack *ret_stack; | |
1927 | /* time stamp for last schedule */ | |
1928 | unsigned long long ftrace_timestamp; | |
1929 | /* | |
1930 | * Number of functions that haven't been traced | |
1931 | * because of depth overrun. | |
1932 | */ | |
1933 | atomic_t trace_overrun; | |
1934 | /* Pause for the tracing */ | |
1935 | atomic_t tracing_graph_pause; | |
1936 | #endif | |
1937 | #ifdef CONFIG_TRACING | |
1938 | /* state flags for use by tracers */ | |
1939 | unsigned long trace; | |
1940 | /* bitmask and counter of trace recursion */ | |
1941 | unsigned long trace_recursion; | |
1942 | #endif /* CONFIG_TRACING */ | |
1943 | #ifdef CONFIG_KCOV | |
1944 | /* Coverage collection mode enabled for this task (0 if disabled). */ | |
1945 | enum kcov_mode kcov_mode; | |
1946 | /* Size of the kcov_area. */ | |
1947 | unsigned kcov_size; | |
1948 | /* Buffer for coverage collection. */ | |
1949 | void *kcov_area; | |
1950 | /* kcov desciptor wired with this task or NULL. */ | |
1951 | struct kcov *kcov; | |
1952 | #endif | |
1953 | #ifdef CONFIG_MEMCG | |
1954 | struct mem_cgroup *memcg_in_oom; | |
1955 | gfp_t memcg_oom_gfp_mask; | |
1956 | int memcg_oom_order; | |
1957 | ||
1958 | /* number of pages to reclaim on returning to userland */ | |
1959 | unsigned int memcg_nr_pages_over_high; | |
1960 | #endif | |
1961 | #ifdef CONFIG_UPROBES | |
1962 | struct uprobe_task *utask; | |
1963 | #endif | |
1964 | #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) | |
1965 | unsigned int sequential_io; | |
1966 | unsigned int sequential_io_avg; | |
1967 | #endif | |
1968 | #ifdef CONFIG_DEBUG_ATOMIC_SLEEP | |
1969 | unsigned long task_state_change; | |
1970 | #endif | |
1971 | int pagefault_disabled; | |
1972 | #ifdef CONFIG_MMU | |
1973 | struct task_struct *oom_reaper_list; | |
1974 | #endif | |
1975 | #ifdef CONFIG_VMAP_STACK | |
1976 | struct vm_struct *stack_vm_area; | |
1977 | #endif | |
1978 | #ifdef CONFIG_THREAD_INFO_IN_TASK | |
1979 | /* A live task holds one reference. */ | |
1980 | atomic_t stack_refcount; | |
1981 | #endif | |
1982 | /* CPU-specific state of this task */ | |
1983 | struct thread_struct thread; | |
1984 | /* | |
1985 | * WARNING: on x86, 'thread_struct' contains a variable-sized | |
1986 | * structure. It *MUST* be at the end of 'task_struct'. | |
1987 | * | |
1988 | * Do not put anything below here! | |
1989 | */ | |
1990 | }; | |
1991 | ||
1992 | #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT | |
1993 | extern int arch_task_struct_size __read_mostly; | |
1994 | #else | |
1995 | # define arch_task_struct_size (sizeof(struct task_struct)) | |
1996 | #endif | |
1997 | ||
1998 | #ifdef CONFIG_VMAP_STACK | |
1999 | static inline struct vm_struct *task_stack_vm_area(const struct task_struct *t) | |
2000 | { | |
2001 | return t->stack_vm_area; | |
2002 | } | |
2003 | #else | |
2004 | static inline struct vm_struct *task_stack_vm_area(const struct task_struct *t) | |
2005 | { | |
2006 | return NULL; | |
2007 | } | |
2008 | #endif | |
2009 | ||
2010 | /* Future-safe accessor for struct task_struct's cpus_allowed. */ | |
2011 | #define tsk_cpus_allowed(tsk) (&(tsk)->cpus_allowed) | |
2012 | ||
2013 | static inline int tsk_nr_cpus_allowed(struct task_struct *p) | |
2014 | { | |
2015 | return p->nr_cpus_allowed; | |
2016 | } | |
2017 | ||
2018 | #define TNF_MIGRATED 0x01 | |
2019 | #define TNF_NO_GROUP 0x02 | |
2020 | #define TNF_SHARED 0x04 | |
2021 | #define TNF_FAULT_LOCAL 0x08 | |
2022 | #define TNF_MIGRATE_FAIL 0x10 | |
2023 | ||
2024 | static inline bool in_vfork(struct task_struct *tsk) | |
2025 | { | |
2026 | bool ret; | |
2027 | ||
2028 | /* | |
2029 | * need RCU to access ->real_parent if CLONE_VM was used along with | |
2030 | * CLONE_PARENT. | |
2031 | * | |
2032 | * We check real_parent->mm == tsk->mm because CLONE_VFORK does not | |
2033 | * imply CLONE_VM | |
2034 | * | |
2035 | * CLONE_VFORK can be used with CLONE_PARENT/CLONE_THREAD and thus | |
2036 | * ->real_parent is not necessarily the task doing vfork(), so in | |
2037 | * theory we can't rely on task_lock() if we want to dereference it. | |
2038 | * | |
2039 | * And in this case we can't trust the real_parent->mm == tsk->mm | |
2040 | * check, it can be false negative. But we do not care, if init or | |
2041 | * another oom-unkillable task does this it should blame itself. | |
2042 | */ | |
2043 | rcu_read_lock(); | |
2044 | ret = tsk->vfork_done && tsk->real_parent->mm == tsk->mm; | |
2045 | rcu_read_unlock(); | |
2046 | ||
2047 | return ret; | |
2048 | } | |
2049 | ||
2050 | #ifdef CONFIG_NUMA_BALANCING | |
2051 | extern void task_numa_fault(int last_node, int node, int pages, int flags); | |
2052 | extern pid_t task_numa_group_id(struct task_struct *p); | |
2053 | extern void set_numabalancing_state(bool enabled); | |
2054 | extern void task_numa_free(struct task_struct *p); | |
2055 | extern bool should_numa_migrate_memory(struct task_struct *p, struct page *page, | |
2056 | int src_nid, int dst_cpu); | |
2057 | #else | |
2058 | static inline void task_numa_fault(int last_node, int node, int pages, | |
2059 | int flags) | |
2060 | { | |
2061 | } | |
2062 | static inline pid_t task_numa_group_id(struct task_struct *p) | |
2063 | { | |
2064 | return 0; | |
2065 | } | |
2066 | static inline void set_numabalancing_state(bool enabled) | |
2067 | { | |
2068 | } | |
2069 | static inline void task_numa_free(struct task_struct *p) | |
2070 | { | |
2071 | } | |
2072 | static inline bool should_numa_migrate_memory(struct task_struct *p, | |
2073 | struct page *page, int src_nid, int dst_cpu) | |
2074 | { | |
2075 | return true; | |
2076 | } | |
2077 | #endif | |
2078 | ||
2079 | static inline struct pid *task_pid(struct task_struct *task) | |
2080 | { | |
2081 | return task->pids[PIDTYPE_PID].pid; | |
2082 | } | |
2083 | ||
2084 | static inline struct pid *task_tgid(struct task_struct *task) | |
2085 | { | |
2086 | return task->group_leader->pids[PIDTYPE_PID].pid; | |
2087 | } | |
2088 | ||
2089 | /* | |
2090 | * Without tasklist or rcu lock it is not safe to dereference | |
2091 | * the result of task_pgrp/task_session even if task == current, | |
2092 | * we can race with another thread doing sys_setsid/sys_setpgid. | |
2093 | */ | |
2094 | static inline struct pid *task_pgrp(struct task_struct *task) | |
2095 | { | |
2096 | return task->group_leader->pids[PIDTYPE_PGID].pid; | |
2097 | } | |
2098 | ||
2099 | static inline struct pid *task_session(struct task_struct *task) | |
2100 | { | |
2101 | return task->group_leader->pids[PIDTYPE_SID].pid; | |
2102 | } | |
2103 | ||
2104 | struct pid_namespace; | |
2105 | ||
2106 | /* | |
2107 | * the helpers to get the task's different pids as they are seen | |
2108 | * from various namespaces | |
2109 | * | |
2110 | * task_xid_nr() : global id, i.e. the id seen from the init namespace; | |
2111 | * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of | |
2112 | * current. | |
2113 | * task_xid_nr_ns() : id seen from the ns specified; | |
2114 | * | |
2115 | * set_task_vxid() : assigns a virtual id to a task; | |
2116 | * | |
2117 | * see also pid_nr() etc in include/linux/pid.h | |
2118 | */ | |
2119 | pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, | |
2120 | struct pid_namespace *ns); | |
2121 | ||
2122 | static inline pid_t task_pid_nr(struct task_struct *tsk) | |
2123 | { | |
2124 | return tsk->pid; | |
2125 | } | |
2126 | ||
2127 | static inline pid_t task_pid_nr_ns(struct task_struct *tsk, | |
2128 | struct pid_namespace *ns) | |
2129 | { | |
2130 | return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns); | |
2131 | } | |
2132 | ||
2133 | static inline pid_t task_pid_vnr(struct task_struct *tsk) | |
2134 | { | |
2135 | return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL); | |
2136 | } | |
2137 | ||
2138 | ||
2139 | static inline pid_t task_tgid_nr(struct task_struct *tsk) | |
2140 | { | |
2141 | return tsk->tgid; | |
2142 | } | |
2143 | ||
2144 | pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns); | |
2145 | ||
2146 | static inline pid_t task_tgid_vnr(struct task_struct *tsk) | |
2147 | { | |
2148 | return pid_vnr(task_tgid(tsk)); | |
2149 | } | |
2150 | ||
2151 | ||
2152 | static inline int pid_alive(const struct task_struct *p); | |
2153 | static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns) | |
2154 | { | |
2155 | pid_t pid = 0; | |
2156 | ||
2157 | rcu_read_lock(); | |
2158 | if (pid_alive(tsk)) | |
2159 | pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns); | |
2160 | rcu_read_unlock(); | |
2161 | ||
2162 | return pid; | |
2163 | } | |
2164 | ||
2165 | static inline pid_t task_ppid_nr(const struct task_struct *tsk) | |
2166 | { | |
2167 | return task_ppid_nr_ns(tsk, &init_pid_ns); | |
2168 | } | |
2169 | ||
2170 | static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, | |
2171 | struct pid_namespace *ns) | |
2172 | { | |
2173 | return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns); | |
2174 | } | |
2175 | ||
2176 | static inline pid_t task_pgrp_vnr(struct task_struct *tsk) | |
2177 | { | |
2178 | return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL); | |
2179 | } | |
2180 | ||
2181 | ||
2182 | static inline pid_t task_session_nr_ns(struct task_struct *tsk, | |
2183 | struct pid_namespace *ns) | |
2184 | { | |
2185 | return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns); | |
2186 | } | |
2187 | ||
2188 | static inline pid_t task_session_vnr(struct task_struct *tsk) | |
2189 | { | |
2190 | return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL); | |
2191 | } | |
2192 | ||
2193 | /* obsolete, do not use */ | |
2194 | static inline pid_t task_pgrp_nr(struct task_struct *tsk) | |
2195 | { | |
2196 | return task_pgrp_nr_ns(tsk, &init_pid_ns); | |
2197 | } | |
2198 | ||
2199 | /** | |
2200 | * pid_alive - check that a task structure is not stale | |
2201 | * @p: Task structure to be checked. | |
2202 | * | |
2203 | * Test if a process is not yet dead (at most zombie state) | |
2204 | * If pid_alive fails, then pointers within the task structure | |
2205 | * can be stale and must not be dereferenced. | |
2206 | * | |
2207 | * Return: 1 if the process is alive. 0 otherwise. | |
2208 | */ | |
2209 | static inline int pid_alive(const struct task_struct *p) | |
2210 | { | |
2211 | return p->pids[PIDTYPE_PID].pid != NULL; | |
2212 | } | |
2213 | ||
2214 | /** | |
2215 | * is_global_init - check if a task structure is init. Since init | |
2216 | * is free to have sub-threads we need to check tgid. | |
2217 | * @tsk: Task structure to be checked. | |
2218 | * | |
2219 | * Check if a task structure is the first user space task the kernel created. | |
2220 | * | |
2221 | * Return: 1 if the task structure is init. 0 otherwise. | |
2222 | */ | |
2223 | static inline int is_global_init(struct task_struct *tsk) | |
2224 | { | |
2225 | return task_tgid_nr(tsk) == 1; | |
2226 | } | |
2227 | ||
2228 | extern struct pid *cad_pid; | |
2229 | ||
2230 | extern void free_task(struct task_struct *tsk); | |
2231 | #define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0) | |
2232 | ||
2233 | extern void __put_task_struct(struct task_struct *t); | |
2234 | ||
2235 | static inline void put_task_struct(struct task_struct *t) | |
2236 | { | |
2237 | if (atomic_dec_and_test(&t->usage)) | |
2238 | __put_task_struct(t); | |
2239 | } | |
2240 | ||
2241 | struct task_struct *task_rcu_dereference(struct task_struct **ptask); | |
2242 | struct task_struct *try_get_task_struct(struct task_struct **ptask); | |
2243 | ||
2244 | #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN | |
2245 | extern void task_cputime(struct task_struct *t, | |
2246 | cputime_t *utime, cputime_t *stime); | |
2247 | extern cputime_t task_gtime(struct task_struct *t); | |
2248 | #else | |
2249 | static inline void task_cputime(struct task_struct *t, | |
2250 | cputime_t *utime, cputime_t *stime) | |
2251 | { | |
2252 | *utime = t->utime; | |
2253 | *stime = t->stime; | |
2254 | } | |
2255 | ||
2256 | static inline cputime_t task_gtime(struct task_struct *t) | |
2257 | { | |
2258 | return t->gtime; | |
2259 | } | |
2260 | #endif | |
2261 | ||
2262 | #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME | |
2263 | static inline void task_cputime_scaled(struct task_struct *t, | |
2264 | cputime_t *utimescaled, | |
2265 | cputime_t *stimescaled) | |
2266 | { | |
2267 | *utimescaled = t->utimescaled; | |
2268 | *stimescaled = t->stimescaled; | |
2269 | } | |
2270 | #else | |
2271 | static inline void task_cputime_scaled(struct task_struct *t, | |
2272 | cputime_t *utimescaled, | |
2273 | cputime_t *stimescaled) | |
2274 | { | |
2275 | task_cputime(t, utimescaled, stimescaled); | |
2276 | } | |
2277 | #endif | |
2278 | ||
2279 | extern void task_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st); | |
2280 | extern void thread_group_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st); | |
2281 | ||
2282 | /* | |
2283 | * Per process flags | |
2284 | */ | |
2285 | #define PF_IDLE 0x00000002 /* I am an IDLE thread */ | |
2286 | #define PF_EXITING 0x00000004 /* getting shut down */ | |
2287 | #define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */ | |
2288 | #define PF_VCPU 0x00000010 /* I'm a virtual CPU */ | |
2289 | #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ | |
2290 | #define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */ | |
2291 | #define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */ | |
2292 | #define PF_SUPERPRIV 0x00000100 /* used super-user privileges */ | |
2293 | #define PF_DUMPCORE 0x00000200 /* dumped core */ | |
2294 | #define PF_SIGNALED 0x00000400 /* killed by a signal */ | |
2295 | #define PF_MEMALLOC 0x00000800 /* Allocating memory */ | |
2296 | #define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */ | |
2297 | #define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */ | |
2298 | #define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */ | |
2299 | #define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */ | |
2300 | #define PF_FROZEN 0x00010000 /* frozen for system suspend */ | |
2301 | #define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */ | |
2302 | #define PF_KSWAPD 0x00040000 /* I am kswapd */ | |
2303 | #define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */ | |
2304 | #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */ | |
2305 | #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ | |
2306 | #define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */ | |
2307 | #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */ | |
2308 | #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */ | |
2309 | #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ | |
2310 | #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */ | |
2311 | #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */ | |
2312 | #define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */ | |
2313 | ||
2314 | /* | |
2315 | * Only the _current_ task can read/write to tsk->flags, but other | |
2316 | * tasks can access tsk->flags in readonly mode for example | |
2317 | * with tsk_used_math (like during threaded core dumping). | |
2318 | * There is however an exception to this rule during ptrace | |
2319 | * or during fork: the ptracer task is allowed to write to the | |
2320 | * child->flags of its traced child (same goes for fork, the parent | |
2321 | * can write to the child->flags), because we're guaranteed the | |
2322 | * child is not running and in turn not changing child->flags | |
2323 | * at the same time the parent does it. | |
2324 | */ | |
2325 | #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) | |
2326 | #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) | |
2327 | #define clear_used_math() clear_stopped_child_used_math(current) | |
2328 | #define set_used_math() set_stopped_child_used_math(current) | |
2329 | #define conditional_stopped_child_used_math(condition, child) \ | |
2330 | do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) | |
2331 | #define conditional_used_math(condition) \ | |
2332 | conditional_stopped_child_used_math(condition, current) | |
2333 | #define copy_to_stopped_child_used_math(child) \ | |
2334 | do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) | |
2335 | /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ | |
2336 | #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) | |
2337 | #define used_math() tsk_used_math(current) | |
2338 | ||
2339 | /* __GFP_IO isn't allowed if PF_MEMALLOC_NOIO is set in current->flags | |
2340 | * __GFP_FS is also cleared as it implies __GFP_IO. | |
2341 | */ | |
2342 | static inline gfp_t memalloc_noio_flags(gfp_t flags) | |
2343 | { | |
2344 | if (unlikely(current->flags & PF_MEMALLOC_NOIO)) | |
2345 | flags &= ~(__GFP_IO | __GFP_FS); | |
2346 | return flags; | |
2347 | } | |
2348 | ||
2349 | static inline unsigned int memalloc_noio_save(void) | |
2350 | { | |
2351 | unsigned int flags = current->flags & PF_MEMALLOC_NOIO; | |
2352 | current->flags |= PF_MEMALLOC_NOIO; | |
2353 | return flags; | |
2354 | } | |
2355 | ||
2356 | static inline void memalloc_noio_restore(unsigned int flags) | |
2357 | { | |
2358 | current->flags = (current->flags & ~PF_MEMALLOC_NOIO) | flags; | |
2359 | } | |
2360 | ||
2361 | /* Per-process atomic flags. */ | |
2362 | #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ | |
2363 | #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ | |
2364 | #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ | |
2365 | #define PFA_LMK_WAITING 3 /* Lowmemorykiller is waiting */ | |
2366 | ||
2367 | ||
2368 | #define TASK_PFA_TEST(name, func) \ | |
2369 | static inline bool task_##func(struct task_struct *p) \ | |
2370 | { return test_bit(PFA_##name, &p->atomic_flags); } | |
2371 | #define TASK_PFA_SET(name, func) \ | |
2372 | static inline void task_set_##func(struct task_struct *p) \ | |
2373 | { set_bit(PFA_##name, &p->atomic_flags); } | |
2374 | #define TASK_PFA_CLEAR(name, func) \ | |
2375 | static inline void task_clear_##func(struct task_struct *p) \ | |
2376 | { clear_bit(PFA_##name, &p->atomic_flags); } | |
2377 | ||
2378 | TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) | |
2379 | TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) | |
2380 | ||
2381 | TASK_PFA_TEST(SPREAD_PAGE, spread_page) | |
2382 | TASK_PFA_SET(SPREAD_PAGE, spread_page) | |
2383 | TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) | |
2384 | ||
2385 | TASK_PFA_TEST(SPREAD_SLAB, spread_slab) | |
2386 | TASK_PFA_SET(SPREAD_SLAB, spread_slab) | |
2387 | TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) | |
2388 | ||
2389 | TASK_PFA_TEST(LMK_WAITING, lmk_waiting) | |
2390 | TASK_PFA_SET(LMK_WAITING, lmk_waiting) | |
2391 | ||
2392 | /* | |
2393 | * task->jobctl flags | |
2394 | */ | |
2395 | #define JOBCTL_STOP_SIGMASK 0xffff /* signr of the last group stop */ | |
2396 | ||
2397 | #define JOBCTL_STOP_DEQUEUED_BIT 16 /* stop signal dequeued */ | |
2398 | #define JOBCTL_STOP_PENDING_BIT 17 /* task should stop for group stop */ | |
2399 | #define JOBCTL_STOP_CONSUME_BIT 18 /* consume group stop count */ | |
2400 | #define JOBCTL_TRAP_STOP_BIT 19 /* trap for STOP */ | |
2401 | #define JOBCTL_TRAP_NOTIFY_BIT 20 /* trap for NOTIFY */ | |
2402 | #define JOBCTL_TRAPPING_BIT 21 /* switching to TRACED */ | |
2403 | #define JOBCTL_LISTENING_BIT 22 /* ptracer is listening for events */ | |
2404 | ||
2405 | #define JOBCTL_STOP_DEQUEUED (1UL << JOBCTL_STOP_DEQUEUED_BIT) | |
2406 | #define JOBCTL_STOP_PENDING (1UL << JOBCTL_STOP_PENDING_BIT) | |
2407 | #define JOBCTL_STOP_CONSUME (1UL << JOBCTL_STOP_CONSUME_BIT) | |
2408 | #define JOBCTL_TRAP_STOP (1UL << JOBCTL_TRAP_STOP_BIT) | |
2409 | #define JOBCTL_TRAP_NOTIFY (1UL << JOBCTL_TRAP_NOTIFY_BIT) | |
2410 | #define JOBCTL_TRAPPING (1UL << JOBCTL_TRAPPING_BIT) | |
2411 | #define JOBCTL_LISTENING (1UL << JOBCTL_LISTENING_BIT) | |
2412 | ||
2413 | #define JOBCTL_TRAP_MASK (JOBCTL_TRAP_STOP | JOBCTL_TRAP_NOTIFY) | |
2414 | #define JOBCTL_PENDING_MASK (JOBCTL_STOP_PENDING | JOBCTL_TRAP_MASK) | |
2415 | ||
2416 | extern bool task_set_jobctl_pending(struct task_struct *task, | |
2417 | unsigned long mask); | |
2418 | extern void task_clear_jobctl_trapping(struct task_struct *task); | |
2419 | extern void task_clear_jobctl_pending(struct task_struct *task, | |
2420 | unsigned long mask); | |
2421 | ||
2422 | static inline void rcu_copy_process(struct task_struct *p) | |
2423 | { | |
2424 | #ifdef CONFIG_PREEMPT_RCU | |
2425 | p->rcu_read_lock_nesting = 0; | |
2426 | p->rcu_read_unlock_special.s = 0; | |
2427 | p->rcu_blocked_node = NULL; | |
2428 | INIT_LIST_HEAD(&p->rcu_node_entry); | |
2429 | #endif /* #ifdef CONFIG_PREEMPT_RCU */ | |
2430 | #ifdef CONFIG_TASKS_RCU | |
2431 | p->rcu_tasks_holdout = false; | |
2432 | INIT_LIST_HEAD(&p->rcu_tasks_holdout_list); | |
2433 | p->rcu_tasks_idle_cpu = -1; | |
2434 | #endif /* #ifdef CONFIG_TASKS_RCU */ | |
2435 | } | |
2436 | ||
2437 | static inline void tsk_restore_flags(struct task_struct *task, | |
2438 | unsigned long orig_flags, unsigned long flags) | |
2439 | { | |
2440 | task->flags &= ~flags; | |
2441 | task->flags |= orig_flags & flags; | |
2442 | } | |
2443 | ||
2444 | extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, | |
2445 | const struct cpumask *trial); | |
2446 | extern int task_can_attach(struct task_struct *p, | |
2447 | const struct cpumask *cs_cpus_allowed); | |
2448 | #ifdef CONFIG_SMP | |
2449 | extern void do_set_cpus_allowed(struct task_struct *p, | |
2450 | const struct cpumask *new_mask); | |
2451 | ||
2452 | extern int set_cpus_allowed_ptr(struct task_struct *p, | |
2453 | const struct cpumask *new_mask); | |
2454 | #else | |
2455 | static inline void do_set_cpus_allowed(struct task_struct *p, | |
2456 | const struct cpumask *new_mask) | |
2457 | { | |
2458 | } | |
2459 | static inline int set_cpus_allowed_ptr(struct task_struct *p, | |
2460 | const struct cpumask *new_mask) | |
2461 | { | |
2462 | if (!cpumask_test_cpu(0, new_mask)) | |
2463 | return -EINVAL; | |
2464 | return 0; | |
2465 | } | |
2466 | #endif | |
2467 | ||
2468 | #ifdef CONFIG_NO_HZ_COMMON | |
2469 | void calc_load_enter_idle(void); | |
2470 | void calc_load_exit_idle(void); | |
2471 | #else | |
2472 | static inline void calc_load_enter_idle(void) { } | |
2473 | static inline void calc_load_exit_idle(void) { } | |
2474 | #endif /* CONFIG_NO_HZ_COMMON */ | |
2475 | ||
2476 | #ifndef cpu_relax_yield | |
2477 | #define cpu_relax_yield() cpu_relax() | |
2478 | #endif | |
2479 | ||
2480 | /* | |
2481 | * Do not use outside of architecture code which knows its limitations. | |
2482 | * | |
2483 | * sched_clock() has no promise of monotonicity or bounded drift between | |
2484 | * CPUs, use (which you should not) requires disabling IRQs. | |
2485 | * | |
2486 | * Please use one of the three interfaces below. | |
2487 | */ | |
2488 | extern unsigned long long notrace sched_clock(void); | |
2489 | /* | |
2490 | * See the comment in kernel/sched/clock.c | |
2491 | */ | |
2492 | extern u64 running_clock(void); | |
2493 | extern u64 sched_clock_cpu(int cpu); | |
2494 | ||
2495 | ||
2496 | extern void sched_clock_init(void); | |
2497 | ||
2498 | #ifndef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK | |
2499 | static inline void sched_clock_tick(void) | |
2500 | { | |
2501 | } | |
2502 | ||
2503 | static inline void sched_clock_idle_sleep_event(void) | |
2504 | { | |
2505 | } | |
2506 | ||
2507 | static inline void sched_clock_idle_wakeup_event(u64 delta_ns) | |
2508 | { | |
2509 | } | |
2510 | ||
2511 | static inline u64 cpu_clock(int cpu) | |
2512 | { | |
2513 | return sched_clock(); | |
2514 | } | |
2515 | ||
2516 | static inline u64 local_clock(void) | |
2517 | { | |
2518 | return sched_clock(); | |
2519 | } | |
2520 | #else | |
2521 | /* | |
2522 | * Architectures can set this to 1 if they have specified | |
2523 | * CONFIG_HAVE_UNSTABLE_SCHED_CLOCK in their arch Kconfig, | |
2524 | * but then during bootup it turns out that sched_clock() | |
2525 | * is reliable after all: | |
2526 | */ | |
2527 | extern int sched_clock_stable(void); | |
2528 | extern void set_sched_clock_stable(void); | |
2529 | extern void clear_sched_clock_stable(void); | |
2530 | ||
2531 | extern void sched_clock_tick(void); | |
2532 | extern void sched_clock_idle_sleep_event(void); | |
2533 | extern void sched_clock_idle_wakeup_event(u64 delta_ns); | |
2534 | ||
2535 | /* | |
2536 | * As outlined in clock.c, provides a fast, high resolution, nanosecond | |
2537 | * time source that is monotonic per cpu argument and has bounded drift | |
2538 | * between cpus. | |
2539 | * | |
2540 | * ######################### BIG FAT WARNING ########################## | |
2541 | * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can # | |
2542 | * # go backwards !! # | |
2543 | * #################################################################### | |
2544 | */ | |
2545 | static inline u64 cpu_clock(int cpu) | |
2546 | { | |
2547 | return sched_clock_cpu(cpu); | |
2548 | } | |
2549 | ||
2550 | static inline u64 local_clock(void) | |
2551 | { | |
2552 | return sched_clock_cpu(raw_smp_processor_id()); | |
2553 | } | |
2554 | #endif | |
2555 | ||
2556 | #ifdef CONFIG_IRQ_TIME_ACCOUNTING | |
2557 | /* | |
2558 | * An i/f to runtime opt-in for irq time accounting based off of sched_clock. | |
2559 | * The reason for this explicit opt-in is not to have perf penalty with | |
2560 | * slow sched_clocks. | |
2561 | */ | |
2562 | extern void enable_sched_clock_irqtime(void); | |
2563 | extern void disable_sched_clock_irqtime(void); | |
2564 | #else | |
2565 | static inline void enable_sched_clock_irqtime(void) {} | |
2566 | static inline void disable_sched_clock_irqtime(void) {} | |
2567 | #endif | |
2568 | ||
2569 | extern unsigned long long | |
2570 | task_sched_runtime(struct task_struct *task); | |
2571 | ||
2572 | /* sched_exec is called by processes performing an exec */ | |
2573 | #ifdef CONFIG_SMP | |
2574 | extern void sched_exec(void); | |
2575 | #else | |
2576 | #define sched_exec() {} | |
2577 | #endif | |
2578 | ||
2579 | extern void sched_clock_idle_sleep_event(void); | |
2580 | extern void sched_clock_idle_wakeup_event(u64 delta_ns); | |
2581 | ||
2582 | #ifdef CONFIG_HOTPLUG_CPU | |
2583 | extern void idle_task_exit(void); | |
2584 | #else | |
2585 | static inline void idle_task_exit(void) {} | |
2586 | #endif | |
2587 | ||
2588 | #if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP) | |
2589 | extern void wake_up_nohz_cpu(int cpu); | |
2590 | #else | |
2591 | static inline void wake_up_nohz_cpu(int cpu) { } | |
2592 | #endif | |
2593 | ||
2594 | #ifdef CONFIG_NO_HZ_FULL | |
2595 | extern u64 scheduler_tick_max_deferment(void); | |
2596 | #endif | |
2597 | ||
2598 | #ifdef CONFIG_SCHED_AUTOGROUP | |
2599 | extern void sched_autogroup_create_attach(struct task_struct *p); | |
2600 | extern void sched_autogroup_detach(struct task_struct *p); | |
2601 | extern void sched_autogroup_fork(struct signal_struct *sig); | |
2602 | extern void sched_autogroup_exit(struct signal_struct *sig); | |
2603 | extern void sched_autogroup_exit_task(struct task_struct *p); | |
2604 | #ifdef CONFIG_PROC_FS | |
2605 | extern void proc_sched_autogroup_show_task(struct task_struct *p, struct seq_file *m); | |
2606 | extern int proc_sched_autogroup_set_nice(struct task_struct *p, int nice); | |
2607 | #endif | |
2608 | #else | |
2609 | static inline void sched_autogroup_create_attach(struct task_struct *p) { } | |
2610 | static inline void sched_autogroup_detach(struct task_struct *p) { } | |
2611 | static inline void sched_autogroup_fork(struct signal_struct *sig) { } | |
2612 | static inline void sched_autogroup_exit(struct signal_struct *sig) { } | |
2613 | static inline void sched_autogroup_exit_task(struct task_struct *p) { } | |
2614 | #endif | |
2615 | ||
2616 | extern int yield_to(struct task_struct *p, bool preempt); | |
2617 | extern void set_user_nice(struct task_struct *p, long nice); | |
2618 | extern int task_prio(const struct task_struct *p); | |
2619 | /** | |
2620 | * task_nice - return the nice value of a given task. | |
2621 | * @p: the task in question. | |
2622 | * | |
2623 | * Return: The nice value [ -20 ... 0 ... 19 ]. | |
2624 | */ | |
2625 | static inline int task_nice(const struct task_struct *p) | |
2626 | { | |
2627 | return PRIO_TO_NICE((p)->static_prio); | |
2628 | } | |
2629 | extern int can_nice(const struct task_struct *p, const int nice); | |
2630 | extern int task_curr(const struct task_struct *p); | |
2631 | extern int idle_cpu(int cpu); | |
2632 | extern int sched_setscheduler(struct task_struct *, int, | |
2633 | const struct sched_param *); | |
2634 | extern int sched_setscheduler_nocheck(struct task_struct *, int, | |
2635 | const struct sched_param *); | |
2636 | extern int sched_setattr(struct task_struct *, | |
2637 | const struct sched_attr *); | |
2638 | extern struct task_struct *idle_task(int cpu); | |
2639 | /** | |
2640 | * is_idle_task - is the specified task an idle task? | |
2641 | * @p: the task in question. | |
2642 | * | |
2643 | * Return: 1 if @p is an idle task. 0 otherwise. | |
2644 | */ | |
2645 | static inline bool is_idle_task(const struct task_struct *p) | |
2646 | { | |
2647 | return !!(p->flags & PF_IDLE); | |
2648 | } | |
2649 | extern struct task_struct *curr_task(int cpu); | |
2650 | extern void ia64_set_curr_task(int cpu, struct task_struct *p); | |
2651 | ||
2652 | void yield(void); | |
2653 | ||
2654 | union thread_union { | |
2655 | #ifndef CONFIG_THREAD_INFO_IN_TASK | |
2656 | struct thread_info thread_info; | |
2657 | #endif | |
2658 | unsigned long stack[THREAD_SIZE/sizeof(long)]; | |
2659 | }; | |
2660 | ||
2661 | #ifndef __HAVE_ARCH_KSTACK_END | |
2662 | static inline int kstack_end(void *addr) | |
2663 | { | |
2664 | /* Reliable end of stack detection: | |
2665 | * Some APM bios versions misalign the stack | |
2666 | */ | |
2667 | return !(((unsigned long)addr+sizeof(void*)-1) & (THREAD_SIZE-sizeof(void*))); | |
2668 | } | |
2669 | #endif | |
2670 | ||
2671 | extern union thread_union init_thread_union; | |
2672 | extern struct task_struct init_task; | |
2673 | ||
2674 | extern struct mm_struct init_mm; | |
2675 | ||
2676 | extern struct pid_namespace init_pid_ns; | |
2677 | ||
2678 | /* | |
2679 | * find a task by one of its numerical ids | |
2680 | * | |
2681 | * find_task_by_pid_ns(): | |
2682 | * finds a task by its pid in the specified namespace | |
2683 | * find_task_by_vpid(): | |
2684 | * finds a task by its virtual pid | |
2685 | * | |
2686 | * see also find_vpid() etc in include/linux/pid.h | |
2687 | */ | |
2688 | ||
2689 | extern struct task_struct *find_task_by_vpid(pid_t nr); | |
2690 | extern struct task_struct *find_task_by_pid_ns(pid_t nr, | |
2691 | struct pid_namespace *ns); | |
2692 | ||
2693 | /* per-UID process charging. */ | |
2694 | extern struct user_struct * alloc_uid(kuid_t); | |
2695 | static inline struct user_struct *get_uid(struct user_struct *u) | |
2696 | { | |
2697 | atomic_inc(&u->__count); | |
2698 | return u; | |
2699 | } | |
2700 | extern void free_uid(struct user_struct *); | |
2701 | ||
2702 | #include <asm/current.h> | |
2703 | ||
2704 | extern void xtime_update(unsigned long ticks); | |
2705 | ||
2706 | extern int wake_up_state(struct task_struct *tsk, unsigned int state); | |
2707 | extern int wake_up_process(struct task_struct *tsk); | |
2708 | extern void wake_up_new_task(struct task_struct *tsk); | |
2709 | #ifdef CONFIG_SMP | |
2710 | extern void kick_process(struct task_struct *tsk); | |
2711 | #else | |
2712 | static inline void kick_process(struct task_struct *tsk) { } | |
2713 | #endif | |
2714 | extern int sched_fork(unsigned long clone_flags, struct task_struct *p); | |
2715 | extern void sched_dead(struct task_struct *p); | |
2716 | ||
2717 | extern void proc_caches_init(void); | |
2718 | extern void flush_signals(struct task_struct *); | |
2719 | extern void ignore_signals(struct task_struct *); | |
2720 | extern void flush_signal_handlers(struct task_struct *, int force_default); | |
2721 | extern int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info); | |
2722 | ||
2723 | static inline int kernel_dequeue_signal(siginfo_t *info) | |
2724 | { | |
2725 | struct task_struct *tsk = current; | |
2726 | siginfo_t __info; | |
2727 | int ret; | |
2728 | ||
2729 | spin_lock_irq(&tsk->sighand->siglock); | |
2730 | ret = dequeue_signal(tsk, &tsk->blocked, info ?: &__info); | |
2731 | spin_unlock_irq(&tsk->sighand->siglock); | |
2732 | ||
2733 | return ret; | |
2734 | } | |
2735 | ||
2736 | static inline void kernel_signal_stop(void) | |
2737 | { | |
2738 | spin_lock_irq(¤t->sighand->siglock); | |
2739 | if (current->jobctl & JOBCTL_STOP_DEQUEUED) | |
2740 | __set_current_state(TASK_STOPPED); | |
2741 | spin_unlock_irq(¤t->sighand->siglock); | |
2742 | ||
2743 | schedule(); | |
2744 | } | |
2745 | ||
2746 | extern void release_task(struct task_struct * p); | |
2747 | extern int send_sig_info(int, struct siginfo *, struct task_struct *); | |
2748 | extern int force_sigsegv(int, struct task_struct *); | |
2749 | extern int force_sig_info(int, struct siginfo *, struct task_struct *); | |
2750 | extern int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp); | |
2751 | extern int kill_pid_info(int sig, struct siginfo *info, struct pid *pid); | |
2752 | extern int kill_pid_info_as_cred(int, struct siginfo *, struct pid *, | |
2753 | const struct cred *, u32); | |
2754 | extern int kill_pgrp(struct pid *pid, int sig, int priv); | |
2755 | extern int kill_pid(struct pid *pid, int sig, int priv); | |
2756 | extern int kill_proc_info(int, struct siginfo *, pid_t); | |
2757 | extern __must_check bool do_notify_parent(struct task_struct *, int); | |
2758 | extern void __wake_up_parent(struct task_struct *p, struct task_struct *parent); | |
2759 | extern void force_sig(int, struct task_struct *); | |
2760 | extern int send_sig(int, struct task_struct *, int); | |
2761 | extern int zap_other_threads(struct task_struct *p); | |
2762 | extern struct sigqueue *sigqueue_alloc(void); | |
2763 | extern void sigqueue_free(struct sigqueue *); | |
2764 | extern int send_sigqueue(struct sigqueue *, struct task_struct *, int group); | |
2765 | extern int do_sigaction(int, struct k_sigaction *, struct k_sigaction *); | |
2766 | ||
2767 | #ifdef TIF_RESTORE_SIGMASK | |
2768 | /* | |
2769 | * Legacy restore_sigmask accessors. These are inefficient on | |
2770 | * SMP architectures because they require atomic operations. | |
2771 | */ | |
2772 | ||
2773 | /** | |
2774 | * set_restore_sigmask() - make sure saved_sigmask processing gets done | |
2775 | * | |
2776 | * This sets TIF_RESTORE_SIGMASK and ensures that the arch signal code | |
2777 | * will run before returning to user mode, to process the flag. For | |
2778 | * all callers, TIF_SIGPENDING is already set or it's no harm to set | |
2779 | * it. TIF_RESTORE_SIGMASK need not be in the set of bits that the | |
2780 | * arch code will notice on return to user mode, in case those bits | |
2781 | * are scarce. We set TIF_SIGPENDING here to ensure that the arch | |
2782 | * signal code always gets run when TIF_RESTORE_SIGMASK is set. | |
2783 | */ | |
2784 | static inline void set_restore_sigmask(void) | |
2785 | { | |
2786 | set_thread_flag(TIF_RESTORE_SIGMASK); | |
2787 | WARN_ON(!test_thread_flag(TIF_SIGPENDING)); | |
2788 | } | |
2789 | static inline void clear_restore_sigmask(void) | |
2790 | { | |
2791 | clear_thread_flag(TIF_RESTORE_SIGMASK); | |
2792 | } | |
2793 | static inline bool test_restore_sigmask(void) | |
2794 | { | |
2795 | return test_thread_flag(TIF_RESTORE_SIGMASK); | |
2796 | } | |
2797 | static inline bool test_and_clear_restore_sigmask(void) | |
2798 | { | |
2799 | return test_and_clear_thread_flag(TIF_RESTORE_SIGMASK); | |
2800 | } | |
2801 | ||
2802 | #else /* TIF_RESTORE_SIGMASK */ | |
2803 | ||
2804 | /* Higher-quality implementation, used if TIF_RESTORE_SIGMASK doesn't exist. */ | |
2805 | static inline void set_restore_sigmask(void) | |
2806 | { | |
2807 | current->restore_sigmask = true; | |
2808 | WARN_ON(!test_thread_flag(TIF_SIGPENDING)); | |
2809 | } | |
2810 | static inline void clear_restore_sigmask(void) | |
2811 | { | |
2812 | current->restore_sigmask = false; | |
2813 | } | |
2814 | static inline bool test_restore_sigmask(void) | |
2815 | { | |
2816 | return current->restore_sigmask; | |
2817 | } | |
2818 | static inline bool test_and_clear_restore_sigmask(void) | |
2819 | { | |
2820 | if (!current->restore_sigmask) | |
2821 | return false; | |
2822 | current->restore_sigmask = false; | |
2823 | return true; | |
2824 | } | |
2825 | #endif | |
2826 | ||
2827 | static inline void restore_saved_sigmask(void) | |
2828 | { | |
2829 | if (test_and_clear_restore_sigmask()) | |
2830 | __set_current_blocked(¤t->saved_sigmask); | |
2831 | } | |
2832 | ||
2833 | static inline sigset_t *sigmask_to_save(void) | |
2834 | { | |
2835 | sigset_t *res = ¤t->blocked; | |
2836 | if (unlikely(test_restore_sigmask())) | |
2837 | res = ¤t->saved_sigmask; | |
2838 | return res; | |
2839 | } | |
2840 | ||
2841 | static inline int kill_cad_pid(int sig, int priv) | |
2842 | { | |
2843 | return kill_pid(cad_pid, sig, priv); | |
2844 | } | |
2845 | ||
2846 | /* These can be the second arg to send_sig_info/send_group_sig_info. */ | |
2847 | #define SEND_SIG_NOINFO ((struct siginfo *) 0) | |
2848 | #define SEND_SIG_PRIV ((struct siginfo *) 1) | |
2849 | #define SEND_SIG_FORCED ((struct siginfo *) 2) | |
2850 | ||
2851 | /* | |
2852 | * True if we are on the alternate signal stack. | |
2853 | */ | |
2854 | static inline int on_sig_stack(unsigned long sp) | |
2855 | { | |
2856 | /* | |
2857 | * If the signal stack is SS_AUTODISARM then, by construction, we | |
2858 | * can't be on the signal stack unless user code deliberately set | |
2859 | * SS_AUTODISARM when we were already on it. | |
2860 | * | |
2861 | * This improves reliability: if user state gets corrupted such that | |
2862 | * the stack pointer points very close to the end of the signal stack, | |
2863 | * then this check will enable the signal to be handled anyway. | |
2864 | */ | |
2865 | if (current->sas_ss_flags & SS_AUTODISARM) | |
2866 | return 0; | |
2867 | ||
2868 | #ifdef CONFIG_STACK_GROWSUP | |
2869 | return sp >= current->sas_ss_sp && | |
2870 | sp - current->sas_ss_sp < current->sas_ss_size; | |
2871 | #else | |
2872 | return sp > current->sas_ss_sp && | |
2873 | sp - current->sas_ss_sp <= current->sas_ss_size; | |
2874 | #endif | |
2875 | } | |
2876 | ||
2877 | static inline int sas_ss_flags(unsigned long sp) | |
2878 | { | |
2879 | if (!current->sas_ss_size) | |
2880 | return SS_DISABLE; | |
2881 | ||
2882 | return on_sig_stack(sp) ? SS_ONSTACK : 0; | |
2883 | } | |
2884 | ||
2885 | static inline void sas_ss_reset(struct task_struct *p) | |
2886 | { | |
2887 | p->sas_ss_sp = 0; | |
2888 | p->sas_ss_size = 0; | |
2889 | p->sas_ss_flags = SS_DISABLE; | |
2890 | } | |
2891 | ||
2892 | static inline unsigned long sigsp(unsigned long sp, struct ksignal *ksig) | |
2893 | { | |
2894 | if (unlikely((ksig->ka.sa.sa_flags & SA_ONSTACK)) && ! sas_ss_flags(sp)) | |
2895 | #ifdef CONFIG_STACK_GROWSUP | |
2896 | return current->sas_ss_sp; | |
2897 | #else | |
2898 | return current->sas_ss_sp + current->sas_ss_size; | |
2899 | #endif | |
2900 | return sp; | |
2901 | } | |
2902 | ||
2903 | /* | |
2904 | * Routines for handling mm_structs | |
2905 | */ | |
2906 | extern struct mm_struct * mm_alloc(void); | |
2907 | ||
2908 | /* mmdrop drops the mm and the page tables */ | |
2909 | extern void __mmdrop(struct mm_struct *); | |
2910 | static inline void mmdrop(struct mm_struct *mm) | |
2911 | { | |
2912 | if (unlikely(atomic_dec_and_test(&mm->mm_count))) | |
2913 | __mmdrop(mm); | |
2914 | } | |
2915 | ||
2916 | static inline void mmdrop_async_fn(struct work_struct *work) | |
2917 | { | |
2918 | struct mm_struct *mm = container_of(work, struct mm_struct, async_put_work); | |
2919 | __mmdrop(mm); | |
2920 | } | |
2921 | ||
2922 | static inline void mmdrop_async(struct mm_struct *mm) | |
2923 | { | |
2924 | if (unlikely(atomic_dec_and_test(&mm->mm_count))) { | |
2925 | INIT_WORK(&mm->async_put_work, mmdrop_async_fn); | |
2926 | schedule_work(&mm->async_put_work); | |
2927 | } | |
2928 | } | |
2929 | ||
2930 | static inline bool mmget_not_zero(struct mm_struct *mm) | |
2931 | { | |
2932 | return atomic_inc_not_zero(&mm->mm_users); | |
2933 | } | |
2934 | ||
2935 | /* mmput gets rid of the mappings and all user-space */ | |
2936 | extern void mmput(struct mm_struct *); | |
2937 | #ifdef CONFIG_MMU | |
2938 | /* same as above but performs the slow path from the async context. Can | |
2939 | * be called from the atomic context as well | |
2940 | */ | |
2941 | extern void mmput_async(struct mm_struct *); | |
2942 | #endif | |
2943 | ||
2944 | /* Grab a reference to a task's mm, if it is not already going away */ | |
2945 | extern struct mm_struct *get_task_mm(struct task_struct *task); | |
2946 | /* | |
2947 | * Grab a reference to a task's mm, if it is not already going away | |
2948 | * and ptrace_may_access with the mode parameter passed to it | |
2949 | * succeeds. | |
2950 | */ | |
2951 | extern struct mm_struct *mm_access(struct task_struct *task, unsigned int mode); | |
2952 | /* Remove the current tasks stale references to the old mm_struct */ | |
2953 | extern void mm_release(struct task_struct *, struct mm_struct *); | |
2954 | ||
2955 | #ifdef CONFIG_HAVE_COPY_THREAD_TLS | |
2956 | extern int copy_thread_tls(unsigned long, unsigned long, unsigned long, | |
2957 | struct task_struct *, unsigned long); | |
2958 | #else | |
2959 | extern int copy_thread(unsigned long, unsigned long, unsigned long, | |
2960 | struct task_struct *); | |
2961 | ||
2962 | /* Architectures that haven't opted into copy_thread_tls get the tls argument | |
2963 | * via pt_regs, so ignore the tls argument passed via C. */ | |
2964 | static inline int copy_thread_tls( | |
2965 | unsigned long clone_flags, unsigned long sp, unsigned long arg, | |
2966 | struct task_struct *p, unsigned long tls) | |
2967 | { | |
2968 | return copy_thread(clone_flags, sp, arg, p); | |
2969 | } | |
2970 | #endif | |
2971 | extern void flush_thread(void); | |
2972 | ||
2973 | #ifdef CONFIG_HAVE_EXIT_THREAD | |
2974 | extern void exit_thread(struct task_struct *tsk); | |
2975 | #else | |
2976 | static inline void exit_thread(struct task_struct *tsk) | |
2977 | { | |
2978 | } | |
2979 | #endif | |
2980 | ||
2981 | extern void exit_files(struct task_struct *); | |
2982 | extern void __cleanup_sighand(struct sighand_struct *); | |
2983 | ||
2984 | extern void exit_itimers(struct signal_struct *); | |
2985 | extern void flush_itimer_signals(void); | |
2986 | ||
2987 | extern void do_group_exit(int); | |
2988 | ||
2989 | extern int do_execve(struct filename *, | |
2990 | const char __user * const __user *, | |
2991 | const char __user * const __user *); | |
2992 | extern int do_execveat(int, struct filename *, | |
2993 | const char __user * const __user *, | |
2994 | const char __user * const __user *, | |
2995 | int); | |
2996 | extern long _do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *, unsigned long); | |
2997 | extern long do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *); | |
2998 | struct task_struct *fork_idle(int); | |
2999 | extern pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags); | |
3000 | ||
3001 | extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); | |
3002 | static inline void set_task_comm(struct task_struct *tsk, const char *from) | |
3003 | { | |
3004 | __set_task_comm(tsk, from, false); | |
3005 | } | |
3006 | extern char *get_task_comm(char *to, struct task_struct *tsk); | |
3007 | ||
3008 | #ifdef CONFIG_SMP | |
3009 | void scheduler_ipi(void); | |
3010 | extern unsigned long wait_task_inactive(struct task_struct *, long match_state); | |
3011 | #else | |
3012 | static inline void scheduler_ipi(void) { } | |
3013 | static inline unsigned long wait_task_inactive(struct task_struct *p, | |
3014 | long match_state) | |
3015 | { | |
3016 | return 1; | |
3017 | } | |
3018 | #endif | |
3019 | ||
3020 | #define tasklist_empty() \ | |
3021 | list_empty(&init_task.tasks) | |
3022 | ||
3023 | #define next_task(p) \ | |
3024 | list_entry_rcu((p)->tasks.next, struct task_struct, tasks) | |
3025 | ||
3026 | #define for_each_process(p) \ | |
3027 | for (p = &init_task ; (p = next_task(p)) != &init_task ; ) | |
3028 | ||
3029 | extern bool current_is_single_threaded(void); | |
3030 | ||
3031 | /* | |
3032 | * Careful: do_each_thread/while_each_thread is a double loop so | |
3033 | * 'break' will not work as expected - use goto instead. | |
3034 | */ | |
3035 | #define do_each_thread(g, t) \ | |
3036 | for (g = t = &init_task ; (g = t = next_task(g)) != &init_task ; ) do | |
3037 | ||
3038 | #define while_each_thread(g, t) \ | |
3039 | while ((t = next_thread(t)) != g) | |
3040 | ||
3041 | #define __for_each_thread(signal, t) \ | |
3042 | list_for_each_entry_rcu(t, &(signal)->thread_head, thread_node) | |
3043 | ||
3044 | #define for_each_thread(p, t) \ | |
3045 | __for_each_thread((p)->signal, t) | |
3046 | ||
3047 | /* Careful: this is a double loop, 'break' won't work as expected. */ | |
3048 | #define for_each_process_thread(p, t) \ | |
3049 | for_each_process(p) for_each_thread(p, t) | |
3050 | ||
3051 | static inline int get_nr_threads(struct task_struct *tsk) | |
3052 | { | |
3053 | return tsk->signal->nr_threads; | |
3054 | } | |
3055 | ||
3056 | static inline bool thread_group_leader(struct task_struct *p) | |
3057 | { | |
3058 | return p->exit_signal >= 0; | |
3059 | } | |
3060 | ||
3061 | /* Do to the insanities of de_thread it is possible for a process | |
3062 | * to have the pid of the thread group leader without actually being | |
3063 | * the thread group leader. For iteration through the pids in proc | |
3064 | * all we care about is that we have a task with the appropriate | |
3065 | * pid, we don't actually care if we have the right task. | |
3066 | */ | |
3067 | static inline bool has_group_leader_pid(struct task_struct *p) | |
3068 | { | |
3069 | return task_pid(p) == p->signal->leader_pid; | |
3070 | } | |
3071 | ||
3072 | static inline | |
3073 | bool same_thread_group(struct task_struct *p1, struct task_struct *p2) | |
3074 | { | |
3075 | return p1->signal == p2->signal; | |
3076 | } | |
3077 | ||
3078 | static inline struct task_struct *next_thread(const struct task_struct *p) | |
3079 | { | |
3080 | return list_entry_rcu(p->thread_group.next, | |
3081 | struct task_struct, thread_group); | |
3082 | } | |
3083 | ||
3084 | static inline int thread_group_empty(struct task_struct *p) | |
3085 | { | |
3086 | return list_empty(&p->thread_group); | |
3087 | } | |
3088 | ||
3089 | #define delay_group_leader(p) \ | |
3090 | (thread_group_leader(p) && !thread_group_empty(p)) | |
3091 | ||
3092 | /* | |
3093 | * Protects ->fs, ->files, ->mm, ->group_info, ->comm, keyring | |
3094 | * subscriptions and synchronises with wait4(). Also used in procfs. Also | |
3095 | * pins the final release of task.io_context. Also protects ->cpuset and | |
3096 | * ->cgroup.subsys[]. And ->vfork_done. | |
3097 | * | |
3098 | * Nests both inside and outside of read_lock(&tasklist_lock). | |
3099 | * It must not be nested with write_lock_irq(&tasklist_lock), | |
3100 | * neither inside nor outside. | |
3101 | */ | |
3102 | static inline void task_lock(struct task_struct *p) | |
3103 | { | |
3104 | spin_lock(&p->alloc_lock); | |
3105 | } | |
3106 | ||
3107 | static inline void task_unlock(struct task_struct *p) | |
3108 | { | |
3109 | spin_unlock(&p->alloc_lock); | |
3110 | } | |
3111 | ||
3112 | extern struct sighand_struct *__lock_task_sighand(struct task_struct *tsk, | |
3113 | unsigned long *flags); | |
3114 | ||
3115 | static inline struct sighand_struct *lock_task_sighand(struct task_struct *tsk, | |
3116 | unsigned long *flags) | |
3117 | { | |
3118 | struct sighand_struct *ret; | |
3119 | ||
3120 | ret = __lock_task_sighand(tsk, flags); | |
3121 | (void)__cond_lock(&tsk->sighand->siglock, ret); | |
3122 | return ret; | |
3123 | } | |
3124 | ||
3125 | static inline void unlock_task_sighand(struct task_struct *tsk, | |
3126 | unsigned long *flags) | |
3127 | { | |
3128 | spin_unlock_irqrestore(&tsk->sighand->siglock, *flags); | |
3129 | } | |
3130 | ||
3131 | /** | |
3132 | * threadgroup_change_begin - mark the beginning of changes to a threadgroup | |
3133 | * @tsk: task causing the changes | |
3134 | * | |
3135 | * All operations which modify a threadgroup - a new thread joining the | |
3136 | * group, death of a member thread (the assertion of PF_EXITING) and | |
3137 | * exec(2) dethreading the process and replacing the leader - are wrapped | |
3138 | * by threadgroup_change_{begin|end}(). This is to provide a place which | |
3139 | * subsystems needing threadgroup stability can hook into for | |
3140 | * synchronization. | |
3141 | */ | |
3142 | static inline void threadgroup_change_begin(struct task_struct *tsk) | |
3143 | { | |
3144 | might_sleep(); | |
3145 | cgroup_threadgroup_change_begin(tsk); | |
3146 | } | |
3147 | ||
3148 | /** | |
3149 | * threadgroup_change_end - mark the end of changes to a threadgroup | |
3150 | * @tsk: task causing the changes | |
3151 | * | |
3152 | * See threadgroup_change_begin(). | |
3153 | */ | |
3154 | static inline void threadgroup_change_end(struct task_struct *tsk) | |
3155 | { | |
3156 | cgroup_threadgroup_change_end(tsk); | |
3157 | } | |
3158 | ||
3159 | #ifdef CONFIG_THREAD_INFO_IN_TASK | |
3160 | ||
3161 | static inline struct thread_info *task_thread_info(struct task_struct *task) | |
3162 | { | |
3163 | return &task->thread_info; | |
3164 | } | |
3165 | ||
3166 | /* | |
3167 | * When accessing the stack of a non-current task that might exit, use | |
3168 | * try_get_task_stack() instead. task_stack_page will return a pointer | |
3169 | * that could get freed out from under you. | |
3170 | */ | |
3171 | static inline void *task_stack_page(const struct task_struct *task) | |
3172 | { | |
3173 | return task->stack; | |
3174 | } | |
3175 | ||
3176 | #define setup_thread_stack(new,old) do { } while(0) | |
3177 | ||
3178 | static inline unsigned long *end_of_stack(const struct task_struct *task) | |
3179 | { | |
3180 | return task->stack; | |
3181 | } | |
3182 | ||
3183 | #elif !defined(__HAVE_THREAD_FUNCTIONS) | |
3184 | ||
3185 | #define task_thread_info(task) ((struct thread_info *)(task)->stack) | |
3186 | #define task_stack_page(task) ((void *)(task)->stack) | |
3187 | ||
3188 | static inline void setup_thread_stack(struct task_struct *p, struct task_struct *org) | |
3189 | { | |
3190 | *task_thread_info(p) = *task_thread_info(org); | |
3191 | task_thread_info(p)->task = p; | |
3192 | } | |
3193 | ||
3194 | /* | |
3195 | * Return the address of the last usable long on the stack. | |
3196 | * | |
3197 | * When the stack grows down, this is just above the thread | |
3198 | * info struct. Going any lower will corrupt the threadinfo. | |
3199 | * | |
3200 | * When the stack grows up, this is the highest address. | |
3201 | * Beyond that position, we corrupt data on the next page. | |
3202 | */ | |
3203 | static inline unsigned long *end_of_stack(struct task_struct *p) | |
3204 | { | |
3205 | #ifdef CONFIG_STACK_GROWSUP | |
3206 | return (unsigned long *)((unsigned long)task_thread_info(p) + THREAD_SIZE) - 1; | |
3207 | #else | |
3208 | return (unsigned long *)(task_thread_info(p) + 1); | |
3209 | #endif | |
3210 | } | |
3211 | ||
3212 | #endif | |
3213 | ||
3214 | #ifdef CONFIG_THREAD_INFO_IN_TASK | |
3215 | static inline void *try_get_task_stack(struct task_struct *tsk) | |
3216 | { | |
3217 | return atomic_inc_not_zero(&tsk->stack_refcount) ? | |
3218 | task_stack_page(tsk) : NULL; | |
3219 | } | |
3220 | ||
3221 | extern void put_task_stack(struct task_struct *tsk); | |
3222 | #else | |
3223 | static inline void *try_get_task_stack(struct task_struct *tsk) | |
3224 | { | |
3225 | return task_stack_page(tsk); | |
3226 | } | |
3227 | ||
3228 | static inline void put_task_stack(struct task_struct *tsk) {} | |
3229 | #endif | |
3230 | ||
3231 | #define task_stack_end_corrupted(task) \ | |
3232 | (*(end_of_stack(task)) != STACK_END_MAGIC) | |
3233 | ||
3234 | static inline int object_is_on_stack(void *obj) | |
3235 | { | |
3236 | void *stack = task_stack_page(current); | |
3237 | ||
3238 | return (obj >= stack) && (obj < (stack + THREAD_SIZE)); | |
3239 | } | |
3240 | ||
3241 | extern void thread_stack_cache_init(void); | |
3242 | ||
3243 | #ifdef CONFIG_DEBUG_STACK_USAGE | |
3244 | static inline unsigned long stack_not_used(struct task_struct *p) | |
3245 | { | |
3246 | unsigned long *n = end_of_stack(p); | |
3247 | ||
3248 | do { /* Skip over canary */ | |
3249 | # ifdef CONFIG_STACK_GROWSUP | |
3250 | n--; | |
3251 | # else | |
3252 | n++; | |
3253 | # endif | |
3254 | } while (!*n); | |
3255 | ||
3256 | # ifdef CONFIG_STACK_GROWSUP | |
3257 | return (unsigned long)end_of_stack(p) - (unsigned long)n; | |
3258 | # else | |
3259 | return (unsigned long)n - (unsigned long)end_of_stack(p); | |
3260 | # endif | |
3261 | } | |
3262 | #endif | |
3263 | extern void set_task_stack_end_magic(struct task_struct *tsk); | |
3264 | ||
3265 | /* set thread flags in other task's structures | |
3266 | * - see asm/thread_info.h for TIF_xxxx flags available | |
3267 | */ | |
3268 | static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) | |
3269 | { | |
3270 | set_ti_thread_flag(task_thread_info(tsk), flag); | |
3271 | } | |
3272 | ||
3273 | static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) | |
3274 | { | |
3275 | clear_ti_thread_flag(task_thread_info(tsk), flag); | |
3276 | } | |
3277 | ||
3278 | static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) | |
3279 | { | |
3280 | return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); | |
3281 | } | |
3282 | ||
3283 | static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) | |
3284 | { | |
3285 | return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); | |
3286 | } | |
3287 | ||
3288 | static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) | |
3289 | { | |
3290 | return test_ti_thread_flag(task_thread_info(tsk), flag); | |
3291 | } | |
3292 | ||
3293 | static inline void set_tsk_need_resched(struct task_struct *tsk) | |
3294 | { | |
3295 | set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); | |
3296 | } | |
3297 | ||
3298 | static inline void clear_tsk_need_resched(struct task_struct *tsk) | |
3299 | { | |
3300 | clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED); | |
3301 | } | |
3302 | ||
3303 | static inline int test_tsk_need_resched(struct task_struct *tsk) | |
3304 | { | |
3305 | return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); | |
3306 | } | |
3307 | ||
3308 | static inline int restart_syscall(void) | |
3309 | { | |
3310 | set_tsk_thread_flag(current, TIF_SIGPENDING); | |
3311 | return -ERESTARTNOINTR; | |
3312 | } | |
3313 | ||
3314 | static inline int signal_pending(struct task_struct *p) | |
3315 | { | |
3316 | return unlikely(test_tsk_thread_flag(p,TIF_SIGPENDING)); | |
3317 | } | |
3318 | ||
3319 | static inline int __fatal_signal_pending(struct task_struct *p) | |
3320 | { | |
3321 | return unlikely(sigismember(&p->pending.signal, SIGKILL)); | |
3322 | } | |
3323 | ||
3324 | static inline int fatal_signal_pending(struct task_struct *p) | |
3325 | { | |
3326 | return signal_pending(p) && __fatal_signal_pending(p); | |
3327 | } | |
3328 | ||
3329 | static inline int signal_pending_state(long state, struct task_struct *p) | |
3330 | { | |
3331 | if (!(state & (TASK_INTERRUPTIBLE | TASK_WAKEKILL))) | |
3332 | return 0; | |
3333 | if (!signal_pending(p)) | |
3334 | return 0; | |
3335 | ||
3336 | return (state & TASK_INTERRUPTIBLE) || __fatal_signal_pending(p); | |
3337 | } | |
3338 | ||
3339 | /* | |
3340 | * cond_resched() and cond_resched_lock(): latency reduction via | |
3341 | * explicit rescheduling in places that are safe. The return | |
3342 | * value indicates whether a reschedule was done in fact. | |
3343 | * cond_resched_lock() will drop the spinlock before scheduling, | |
3344 | * cond_resched_softirq() will enable bhs before scheduling. | |
3345 | */ | |
3346 | #ifndef CONFIG_PREEMPT | |
3347 | extern int _cond_resched(void); | |
3348 | #else | |
3349 | static inline int _cond_resched(void) { return 0; } | |
3350 | #endif | |
3351 | ||
3352 | #define cond_resched() ({ \ | |
3353 | ___might_sleep(__FILE__, __LINE__, 0); \ | |
3354 | _cond_resched(); \ | |
3355 | }) | |
3356 | ||
3357 | extern int __cond_resched_lock(spinlock_t *lock); | |
3358 | ||
3359 | #define cond_resched_lock(lock) ({ \ | |
3360 | ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\ | |
3361 | __cond_resched_lock(lock); \ | |
3362 | }) | |
3363 | ||
3364 | extern int __cond_resched_softirq(void); | |
3365 | ||
3366 | #define cond_resched_softirq() ({ \ | |
3367 | ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \ | |
3368 | __cond_resched_softirq(); \ | |
3369 | }) | |
3370 | ||
3371 | static inline void cond_resched_rcu(void) | |
3372 | { | |
3373 | #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU) | |
3374 | rcu_read_unlock(); | |
3375 | cond_resched(); | |
3376 | rcu_read_lock(); | |
3377 | #endif | |
3378 | } | |
3379 | ||
3380 | static inline unsigned long get_preempt_disable_ip(struct task_struct *p) | |
3381 | { | |
3382 | #ifdef CONFIG_DEBUG_PREEMPT | |
3383 | return p->preempt_disable_ip; | |
3384 | #else | |
3385 | return 0; | |
3386 | #endif | |
3387 | } | |
3388 | ||
3389 | /* | |
3390 | * Does a critical section need to be broken due to another | |
3391 | * task waiting?: (technically does not depend on CONFIG_PREEMPT, | |
3392 | * but a general need for low latency) | |
3393 | */ | |
3394 | static inline int spin_needbreak(spinlock_t *lock) | |
3395 | { | |
3396 | #ifdef CONFIG_PREEMPT | |
3397 | return spin_is_contended(lock); | |
3398 | #else | |
3399 | return 0; | |
3400 | #endif | |
3401 | } | |
3402 | ||
3403 | /* | |
3404 | * Idle thread specific functions to determine the need_resched | |
3405 | * polling state. | |
3406 | */ | |
3407 | #ifdef TIF_POLLING_NRFLAG | |
3408 | static inline int tsk_is_polling(struct task_struct *p) | |
3409 | { | |
3410 | return test_tsk_thread_flag(p, TIF_POLLING_NRFLAG); | |
3411 | } | |
3412 | ||
3413 | static inline void __current_set_polling(void) | |
3414 | { | |
3415 | set_thread_flag(TIF_POLLING_NRFLAG); | |
3416 | } | |
3417 | ||
3418 | static inline bool __must_check current_set_polling_and_test(void) | |
3419 | { | |
3420 | __current_set_polling(); | |
3421 | ||
3422 | /* | |
3423 | * Polling state must be visible before we test NEED_RESCHED, | |
3424 | * paired by resched_curr() | |
3425 | */ | |
3426 | smp_mb__after_atomic(); | |
3427 | ||
3428 | return unlikely(tif_need_resched()); | |
3429 | } | |
3430 | ||
3431 | static inline void __current_clr_polling(void) | |
3432 | { | |
3433 | clear_thread_flag(TIF_POLLING_NRFLAG); | |
3434 | } | |
3435 | ||
3436 | static inline bool __must_check current_clr_polling_and_test(void) | |
3437 | { | |
3438 | __current_clr_polling(); | |
3439 | ||
3440 | /* | |
3441 | * Polling state must be visible before we test NEED_RESCHED, | |
3442 | * paired by resched_curr() | |
3443 | */ | |
3444 | smp_mb__after_atomic(); | |
3445 | ||
3446 | return unlikely(tif_need_resched()); | |
3447 | } | |
3448 | ||
3449 | #else | |
3450 | static inline int tsk_is_polling(struct task_struct *p) { return 0; } | |
3451 | static inline void __current_set_polling(void) { } | |
3452 | static inline void __current_clr_polling(void) { } | |
3453 | ||
3454 | static inline bool __must_check current_set_polling_and_test(void) | |
3455 | { | |
3456 | return unlikely(tif_need_resched()); | |
3457 | } | |
3458 | static inline bool __must_check current_clr_polling_and_test(void) | |
3459 | { | |
3460 | return unlikely(tif_need_resched()); | |
3461 | } | |
3462 | #endif | |
3463 | ||
3464 | static inline void current_clr_polling(void) | |
3465 | { | |
3466 | __current_clr_polling(); | |
3467 | ||
3468 | /* | |
3469 | * Ensure we check TIF_NEED_RESCHED after we clear the polling bit. | |
3470 | * Once the bit is cleared, we'll get IPIs with every new | |
3471 | * TIF_NEED_RESCHED and the IPI handler, scheduler_ipi(), will also | |
3472 | * fold. | |
3473 | */ | |
3474 | smp_mb(); /* paired with resched_curr() */ | |
3475 | ||
3476 | preempt_fold_need_resched(); | |
3477 | } | |
3478 | ||
3479 | static __always_inline bool need_resched(void) | |
3480 | { | |
3481 | return unlikely(tif_need_resched()); | |
3482 | } | |
3483 | ||
3484 | /* | |
3485 | * Thread group CPU time accounting. | |
3486 | */ | |
3487 | void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times); | |
3488 | void thread_group_cputimer(struct task_struct *tsk, struct task_cputime *times); | |
3489 | ||
3490 | /* | |
3491 | * Reevaluate whether the task has signals pending delivery. | |
3492 | * Wake the task if so. | |
3493 | * This is required every time the blocked sigset_t changes. | |
3494 | * callers must hold sighand->siglock. | |
3495 | */ | |
3496 | extern void recalc_sigpending_and_wake(struct task_struct *t); | |
3497 | extern void recalc_sigpending(void); | |
3498 | ||
3499 | extern void signal_wake_up_state(struct task_struct *t, unsigned int state); | |
3500 | ||
3501 | static inline void signal_wake_up(struct task_struct *t, bool resume) | |
3502 | { | |
3503 | signal_wake_up_state(t, resume ? TASK_WAKEKILL : 0); | |
3504 | } | |
3505 | static inline void ptrace_signal_wake_up(struct task_struct *t, bool resume) | |
3506 | { | |
3507 | signal_wake_up_state(t, resume ? __TASK_TRACED : 0); | |
3508 | } | |
3509 | ||
3510 | /* | |
3511 | * Wrappers for p->thread_info->cpu access. No-op on UP. | |
3512 | */ | |
3513 | #ifdef CONFIG_SMP | |
3514 | ||
3515 | static inline unsigned int task_cpu(const struct task_struct *p) | |
3516 | { | |
3517 | #ifdef CONFIG_THREAD_INFO_IN_TASK | |
3518 | return p->cpu; | |
3519 | #else | |
3520 | return task_thread_info(p)->cpu; | |
3521 | #endif | |
3522 | } | |
3523 | ||
3524 | static inline int task_node(const struct task_struct *p) | |
3525 | { | |
3526 | return cpu_to_node(task_cpu(p)); | |
3527 | } | |
3528 | ||
3529 | extern void set_task_cpu(struct task_struct *p, unsigned int cpu); | |
3530 | ||
3531 | #else | |
3532 | ||
3533 | static inline unsigned int task_cpu(const struct task_struct *p) | |
3534 | { | |
3535 | return 0; | |
3536 | } | |
3537 | ||
3538 | static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) | |
3539 | { | |
3540 | } | |
3541 | ||
3542 | #endif /* CONFIG_SMP */ | |
3543 | ||
3544 | /* | |
3545 | * In order to reduce various lock holder preemption latencies provide an | |
3546 | * interface to see if a vCPU is currently running or not. | |
3547 | * | |
3548 | * This allows us to terminate optimistic spin loops and block, analogous to | |
3549 | * the native optimistic spin heuristic of testing if the lock owner task is | |
3550 | * running or not. | |
3551 | */ | |
3552 | #ifndef vcpu_is_preempted | |
3553 | # define vcpu_is_preempted(cpu) false | |
3554 | #endif | |
3555 | ||
3556 | extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); | |
3557 | extern long sched_getaffinity(pid_t pid, struct cpumask *mask); | |
3558 | ||
3559 | #ifdef CONFIG_CGROUP_SCHED | |
3560 | extern struct task_group root_task_group; | |
3561 | #endif /* CONFIG_CGROUP_SCHED */ | |
3562 | ||
3563 | extern int task_can_switch_user(struct user_struct *up, | |
3564 | struct task_struct *tsk); | |
3565 | ||
3566 | #ifdef CONFIG_TASK_XACCT | |
3567 | static inline void add_rchar(struct task_struct *tsk, ssize_t amt) | |
3568 | { | |
3569 | tsk->ioac.rchar += amt; | |
3570 | } | |
3571 | ||
3572 | static inline void add_wchar(struct task_struct *tsk, ssize_t amt) | |
3573 | { | |
3574 | tsk->ioac.wchar += amt; | |
3575 | } | |
3576 | ||
3577 | static inline void inc_syscr(struct task_struct *tsk) | |
3578 | { | |
3579 | tsk->ioac.syscr++; | |
3580 | } | |
3581 | ||
3582 | static inline void inc_syscw(struct task_struct *tsk) | |
3583 | { | |
3584 | tsk->ioac.syscw++; | |
3585 | } | |
3586 | #else | |
3587 | static inline void add_rchar(struct task_struct *tsk, ssize_t amt) | |
3588 | { | |
3589 | } | |
3590 | ||
3591 | static inline void add_wchar(struct task_struct *tsk, ssize_t amt) | |
3592 | { | |
3593 | } | |
3594 | ||
3595 | static inline void inc_syscr(struct task_struct *tsk) | |
3596 | { | |
3597 | } | |
3598 | ||
3599 | static inline void inc_syscw(struct task_struct *tsk) | |
3600 | { | |
3601 | } | |
3602 | #endif | |
3603 | ||
3604 | #ifndef TASK_SIZE_OF | |
3605 | #define TASK_SIZE_OF(tsk) TASK_SIZE | |
3606 | #endif | |
3607 | ||
3608 | #ifdef CONFIG_MEMCG | |
3609 | extern void mm_update_next_owner(struct mm_struct *mm); | |
3610 | #else | |
3611 | static inline void mm_update_next_owner(struct mm_struct *mm) | |
3612 | { | |
3613 | } | |
3614 | #endif /* CONFIG_MEMCG */ | |
3615 | ||
3616 | static inline unsigned long task_rlimit(const struct task_struct *tsk, | |
3617 | unsigned int limit) | |
3618 | { | |
3619 | return READ_ONCE(tsk->signal->rlim[limit].rlim_cur); | |
3620 | } | |
3621 | ||
3622 | static inline unsigned long task_rlimit_max(const struct task_struct *tsk, | |
3623 | unsigned int limit) | |
3624 | { | |
3625 | return READ_ONCE(tsk->signal->rlim[limit].rlim_max); | |
3626 | } | |
3627 | ||
3628 | static inline unsigned long rlimit(unsigned int limit) | |
3629 | { | |
3630 | return task_rlimit(current, limit); | |
3631 | } | |
3632 | ||
3633 | static inline unsigned long rlimit_max(unsigned int limit) | |
3634 | { | |
3635 | return task_rlimit_max(current, limit); | |
3636 | } | |
3637 | ||
3638 | #define SCHED_CPUFREQ_RT (1U << 0) | |
3639 | #define SCHED_CPUFREQ_DL (1U << 1) | |
3640 | #define SCHED_CPUFREQ_IOWAIT (1U << 2) | |
3641 | ||
3642 | #define SCHED_CPUFREQ_RT_DL (SCHED_CPUFREQ_RT | SCHED_CPUFREQ_DL) | |
3643 | ||
3644 | #ifdef CONFIG_CPU_FREQ | |
3645 | struct update_util_data { | |
3646 | void (*func)(struct update_util_data *data, u64 time, unsigned int flags); | |
3647 | }; | |
3648 | ||
3649 | void cpufreq_add_update_util_hook(int cpu, struct update_util_data *data, | |
3650 | void (*func)(struct update_util_data *data, u64 time, | |
3651 | unsigned int flags)); | |
3652 | void cpufreq_remove_update_util_hook(int cpu); | |
3653 | #endif /* CONFIG_CPU_FREQ */ | |
3654 | ||
3655 | #endif |