Commit | Line | Data |
---|---|---|
457c8996 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
1da177e4 LT |
2 | /* |
3 | * linux/kernel/exit.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | */ | |
7 | ||
1da177e4 LT |
8 | #include <linux/mm.h> |
9 | #include <linux/slab.h> | |
4eb5aaa3 | 10 | #include <linux/sched/autogroup.h> |
6e84f315 | 11 | #include <linux/sched/mm.h> |
03441a34 | 12 | #include <linux/sched/stat.h> |
29930025 | 13 | #include <linux/sched/task.h> |
68db0cf1 | 14 | #include <linux/sched/task_stack.h> |
32ef5517 | 15 | #include <linux/sched/cputime.h> |
1da177e4 | 16 | #include <linux/interrupt.h> |
1da177e4 | 17 | #include <linux/module.h> |
c59ede7b | 18 | #include <linux/capability.h> |
1da177e4 LT |
19 | #include <linux/completion.h> |
20 | #include <linux/personality.h> | |
21 | #include <linux/tty.h> | |
da9cbc87 | 22 | #include <linux/iocontext.h> |
1da177e4 | 23 | #include <linux/key.h> |
1da177e4 LT |
24 | #include <linux/cpu.h> |
25 | #include <linux/acct.h> | |
8f0ab514 | 26 | #include <linux/tsacct_kern.h> |
1da177e4 | 27 | #include <linux/file.h> |
9f3acc31 | 28 | #include <linux/fdtable.h> |
80d26af8 | 29 | #include <linux/freezer.h> |
1da177e4 | 30 | #include <linux/binfmts.h> |
ab516013 | 31 | #include <linux/nsproxy.h> |
84d73786 | 32 | #include <linux/pid_namespace.h> |
1da177e4 LT |
33 | #include <linux/ptrace.h> |
34 | #include <linux/profile.h> | |
35 | #include <linux/mount.h> | |
36 | #include <linux/proc_fs.h> | |
49d769d5 | 37 | #include <linux/kthread.h> |
1da177e4 | 38 | #include <linux/mempolicy.h> |
c757249a | 39 | #include <linux/taskstats_kern.h> |
ca74e92b | 40 | #include <linux/delayacct.h> |
b4f48b63 | 41 | #include <linux/cgroup.h> |
1da177e4 | 42 | #include <linux/syscalls.h> |
7ed20e1a | 43 | #include <linux/signal.h> |
6a14c5c9 | 44 | #include <linux/posix-timers.h> |
9f46080c | 45 | #include <linux/cn_proc.h> |
de5097c2 | 46 | #include <linux/mutex.h> |
0771dfef | 47 | #include <linux/futex.h> |
b92ce558 | 48 | #include <linux/pipe_fs_i.h> |
fa84cb93 | 49 | #include <linux/audit.h> /* for audit_free() */ |
83cc5ed3 | 50 | #include <linux/resource.h> |
6eaeeaba | 51 | #include <linux/task_io_accounting_ops.h> |
355f841a EB |
52 | #include <linux/blkdev.h> |
53 | #include <linux/task_work.h> | |
5ad4e53b | 54 | #include <linux/fs_struct.h> |
d84f4f99 | 55 | #include <linux/init_task.h> |
cdd6c482 | 56 | #include <linux/perf_event.h> |
ad8d75ff | 57 | #include <trace/events/sched.h> |
24f1e32c | 58 | #include <linux/hw_breakpoint.h> |
3d5992d2 | 59 | #include <linux/oom.h> |
54848d73 | 60 | #include <linux/writeback.h> |
40401530 | 61 | #include <linux/shm.h> |
5c9a8750 | 62 | #include <linux/kcov.h> |
50b5e49c | 63 | #include <linux/kmsan.h> |
53d3eaa3 | 64 | #include <linux/random.h> |
8f95c90c | 65 | #include <linux/rcuwait.h> |
7e95a225 | 66 | #include <linux/compat.h> |
b1b6b5a3 | 67 | #include <linux/io_uring.h> |
670721c7 | 68 | #include <linux/kprobes.h> |
54ecbe6f | 69 | #include <linux/rethook.h> |
9db89b41 | 70 | #include <linux/sysfs.h> |
1da177e4 | 71 | |
7c0f6ba6 | 72 | #include <linux/uaccess.h> |
1da177e4 | 73 | #include <asm/unistd.h> |
1da177e4 LT |
74 | #include <asm/mmu_context.h> |
75 | ||
d4ccd54d JH |
76 | /* |
77 | * The default value should be high enough to not crash a system that randomly | |
78 | * crashes its kernel from time to time, but low enough to at least not permit | |
79 | * overflowing 32-bit refcounts or the ldsem writer count. | |
80 | */ | |
81 | static unsigned int oops_limit = 10000; | |
82 | ||
83 | #ifdef CONFIG_SYSCTL | |
84 | static struct ctl_table kern_exit_table[] = { | |
85 | { | |
86 | .procname = "oops_limit", | |
87 | .data = &oops_limit, | |
88 | .maxlen = sizeof(oops_limit), | |
89 | .mode = 0644, | |
90 | .proc_handler = proc_douintvec, | |
91 | }, | |
92 | { } | |
93 | }; | |
94 | ||
95 | static __init int kernel_exit_sysctls_init(void) | |
96 | { | |
97 | register_sysctl_init("kernel", kern_exit_table); | |
98 | return 0; | |
99 | } | |
100 | late_initcall(kernel_exit_sysctls_init); | |
101 | #endif | |
102 | ||
9db89b41 KC |
103 | static atomic_t oops_count = ATOMIC_INIT(0); |
104 | ||
105 | #ifdef CONFIG_SYSFS | |
106 | static ssize_t oops_count_show(struct kobject *kobj, struct kobj_attribute *attr, | |
107 | char *page) | |
108 | { | |
109 | return sysfs_emit(page, "%d\n", atomic_read(&oops_count)); | |
110 | } | |
111 | ||
112 | static struct kobj_attribute oops_count_attr = __ATTR_RO(oops_count); | |
113 | ||
114 | static __init int kernel_exit_sysfs_init(void) | |
115 | { | |
116 | sysfs_add_file_to_group(kernel_kobj, &oops_count_attr.attr, NULL); | |
117 | return 0; | |
118 | } | |
119 | late_initcall(kernel_exit_sysfs_init); | |
120 | #endif | |
121 | ||
d40e48e0 | 122 | static void __unhash_process(struct task_struct *p, bool group_dead) |
1da177e4 LT |
123 | { |
124 | nr_threads--; | |
50d75f8d | 125 | detach_pid(p, PIDTYPE_PID); |
d40e48e0 | 126 | if (group_dead) { |
6883f81a | 127 | detach_pid(p, PIDTYPE_TGID); |
1da177e4 LT |
128 | detach_pid(p, PIDTYPE_PGID); |
129 | detach_pid(p, PIDTYPE_SID); | |
c97d9893 | 130 | |
5e85d4ab | 131 | list_del_rcu(&p->tasks); |
9cd80bbb | 132 | list_del_init(&p->sibling); |
909ea964 | 133 | __this_cpu_dec(process_counts); |
1da177e4 | 134 | } |
47e65328 | 135 | list_del_rcu(&p->thread_group); |
0c740d0a | 136 | list_del_rcu(&p->thread_node); |
1da177e4 LT |
137 | } |
138 | ||
6a14c5c9 ON |
139 | /* |
140 | * This function expects the tasklist_lock write-locked. | |
141 | */ | |
142 | static void __exit_signal(struct task_struct *tsk) | |
143 | { | |
144 | struct signal_struct *sig = tsk->signal; | |
d40e48e0 | 145 | bool group_dead = thread_group_leader(tsk); |
6a14c5c9 | 146 | struct sighand_struct *sighand; |
3f649ab7 | 147 | struct tty_struct *tty; |
5613fda9 | 148 | u64 utime, stime; |
6a14c5c9 | 149 | |
d11c563d | 150 | sighand = rcu_dereference_check(tsk->sighand, |
db1466b3 | 151 | lockdep_tasklist_lock_is_held()); |
6a14c5c9 ON |
152 | spin_lock(&sighand->siglock); |
153 | ||
baa73d9e | 154 | #ifdef CONFIG_POSIX_TIMERS |
6a14c5c9 | 155 | posix_cpu_timers_exit(tsk); |
b95e31c0 | 156 | if (group_dead) |
6a14c5c9 | 157 | posix_cpu_timers_exit_group(tsk); |
baa73d9e | 158 | #endif |
e0a70217 | 159 | |
baa73d9e NP |
160 | if (group_dead) { |
161 | tty = sig->tty; | |
162 | sig->tty = NULL; | |
163 | } else { | |
6a14c5c9 ON |
164 | /* |
165 | * If there is any task waiting for the group exit | |
166 | * then notify it: | |
167 | */ | |
d344193a | 168 | if (sig->notify_count > 0 && !--sig->notify_count) |
60700e38 | 169 | wake_up_process(sig->group_exec_task); |
6db840fa | 170 | |
6a14c5c9 ON |
171 | if (tsk == sig->curr_target) |
172 | sig->curr_target = next_thread(tsk); | |
6a14c5c9 ON |
173 | } |
174 | ||
53d3eaa3 NP |
175 | add_device_randomness((const void*) &tsk->se.sum_exec_runtime, |
176 | sizeof(unsigned long long)); | |
177 | ||
90ed9cbe | 178 | /* |
26e75b5c ON |
179 | * Accumulate here the counters for all threads as they die. We could |
180 | * skip the group leader because it is the last user of signal_struct, | |
181 | * but we want to avoid the race with thread_group_cputime() which can | |
182 | * see the empty ->thread_head list. | |
90ed9cbe RR |
183 | */ |
184 | task_cputime(tsk, &utime, &stime); | |
e78c3496 | 185 | write_seqlock(&sig->stats_lock); |
90ed9cbe RR |
186 | sig->utime += utime; |
187 | sig->stime += stime; | |
188 | sig->gtime += task_gtime(tsk); | |
189 | sig->min_flt += tsk->min_flt; | |
190 | sig->maj_flt += tsk->maj_flt; | |
191 | sig->nvcsw += tsk->nvcsw; | |
192 | sig->nivcsw += tsk->nivcsw; | |
193 | sig->inblock += task_io_get_inblock(tsk); | |
194 | sig->oublock += task_io_get_oublock(tsk); | |
195 | task_io_accounting_add(&sig->ioac, &tsk->ioac); | |
196 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; | |
b3ac022c | 197 | sig->nr_threads--; |
d40e48e0 | 198 | __unhash_process(tsk, group_dead); |
e78c3496 | 199 | write_sequnlock(&sig->stats_lock); |
5876700c | 200 | |
da7978b0 ON |
201 | /* |
202 | * Do this under ->siglock, we can race with another thread | |
203 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. | |
204 | */ | |
205 | flush_sigqueue(&tsk->pending); | |
a7e5328a | 206 | tsk->sighand = NULL; |
6a14c5c9 | 207 | spin_unlock(&sighand->siglock); |
6a14c5c9 | 208 | |
a7e5328a | 209 | __cleanup_sighand(sighand); |
a0be55de | 210 | clear_tsk_thread_flag(tsk, TIF_SIGPENDING); |
d40e48e0 | 211 | if (group_dead) { |
6a14c5c9 | 212 | flush_sigqueue(&sig->shared_pending); |
4ada856f | 213 | tty_kref_put(tty); |
6a14c5c9 ON |
214 | } |
215 | } | |
216 | ||
8c7904a0 EB |
217 | static void delayed_put_task_struct(struct rcu_head *rhp) |
218 | { | |
0a16b607 MD |
219 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
220 | ||
670721c7 | 221 | kprobe_flush_task(tsk); |
54ecbe6f | 222 | rethook_flush_task(tsk); |
4e231c79 | 223 | perf_event_delayed_put(tsk); |
0a16b607 MD |
224 | trace_sched_process_free(tsk); |
225 | put_task_struct(tsk); | |
8c7904a0 EB |
226 | } |
227 | ||
3fbd7ee2 EB |
228 | void put_task_struct_rcu_user(struct task_struct *task) |
229 | { | |
230 | if (refcount_dec_and_test(&task->rcu_users)) | |
231 | call_rcu(&task->rcu, delayed_put_task_struct); | |
232 | } | |
f470021a | 233 | |
2be9880d KW |
234 | void __weak release_thread(struct task_struct *dead_task) |
235 | { | |
236 | } | |
237 | ||
a0be55de | 238 | void release_task(struct task_struct *p) |
1da177e4 | 239 | { |
36c8b586 | 240 | struct task_struct *leader; |
7bc3e6e5 | 241 | struct pid *thread_pid; |
1da177e4 | 242 | int zap_leader; |
1f09f974 | 243 | repeat: |
c69e8d9c | 244 | /* don't need to get the RCU readlock here - the process is dead and |
d11c563d PM |
245 | * can't be modifying its own credentials. But shut RCU-lockdep up */ |
246 | rcu_read_lock(); | |
21d1c5e3 | 247 | dec_rlimit_ucounts(task_ucounts(p), UCOUNT_RLIMIT_NPROC, 1); |
d11c563d | 248 | rcu_read_unlock(); |
c69e8d9c | 249 | |
6b115bf5 | 250 | cgroup_release(p); |
0203026b | 251 | |
1da177e4 | 252 | write_lock_irq(&tasklist_lock); |
a288eecc | 253 | ptrace_release_task(p); |
7bc3e6e5 | 254 | thread_pid = get_pid(p->thread_pid); |
1da177e4 | 255 | __exit_signal(p); |
35f5cad8 | 256 | |
1da177e4 LT |
257 | /* |
258 | * If we are the last non-leader member of the thread | |
259 | * group, and the leader is zombie, then notify the | |
260 | * group leader's parent process. (if it wants notification.) | |
261 | */ | |
262 | zap_leader = 0; | |
263 | leader = p->group_leader; | |
a0be55de IA |
264 | if (leader != p && thread_group_empty(leader) |
265 | && leader->exit_state == EXIT_ZOMBIE) { | |
1da177e4 LT |
266 | /* |
267 | * If we were the last child thread and the leader has | |
268 | * exited already, and the leader's parent ignores SIGCHLD, | |
269 | * then we are the one who should release the leader. | |
dae33574 | 270 | */ |
86773473 | 271 | zap_leader = do_notify_parent(leader, leader->exit_signal); |
dae33574 RM |
272 | if (zap_leader) |
273 | leader->exit_state = EXIT_DEAD; | |
1da177e4 LT |
274 | } |
275 | ||
1da177e4 | 276 | write_unlock_irq(&tasklist_lock); |
3a15fb6e | 277 | seccomp_filter_release(p); |
7bc3e6e5 | 278 | proc_flush_pid(thread_pid); |
6ade99ec | 279 | put_pid(thread_pid); |
1da177e4 | 280 | release_thread(p); |
3fbd7ee2 | 281 | put_task_struct_rcu_user(p); |
1da177e4 LT |
282 | |
283 | p = leader; | |
284 | if (unlikely(zap_leader)) | |
285 | goto repeat; | |
286 | } | |
287 | ||
9d9a6ebf | 288 | int rcuwait_wake_up(struct rcuwait *w) |
8f95c90c | 289 | { |
9d9a6ebf | 290 | int ret = 0; |
8f95c90c DB |
291 | struct task_struct *task; |
292 | ||
293 | rcu_read_lock(); | |
294 | ||
295 | /* | |
296 | * Order condition vs @task, such that everything prior to the load | |
297 | * of @task is visible. This is the condition as to why the user called | |
c9d64a1b | 298 | * rcuwait_wake() in the first place. Pairs with set_current_state() |
8f95c90c DB |
299 | * barrier (A) in rcuwait_wait_event(). |
300 | * | |
301 | * WAIT WAKE | |
302 | * [S] tsk = current [S] cond = true | |
303 | * MB (A) MB (B) | |
304 | * [L] cond [L] tsk | |
305 | */ | |
6dc080ee | 306 | smp_mb(); /* (B) */ |
8f95c90c | 307 | |
8f95c90c DB |
308 | task = rcu_dereference(w->task); |
309 | if (task) | |
9d9a6ebf | 310 | ret = wake_up_process(task); |
8f95c90c | 311 | rcu_read_unlock(); |
9d9a6ebf DB |
312 | |
313 | return ret; | |
8f95c90c | 314 | } |
ac8dec42 | 315 | EXPORT_SYMBOL_GPL(rcuwait_wake_up); |
8f95c90c | 316 | |
1da177e4 LT |
317 | /* |
318 | * Determine if a process group is "orphaned", according to the POSIX | |
319 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
320 | * by terminal-generated stop signals. Newly orphaned process groups are | |
321 | * to receive a SIGHUP and a SIGCONT. | |
322 | * | |
323 | * "I ask you, have you ever known what it is to be an orphan?" | |
324 | */ | |
a0be55de IA |
325 | static int will_become_orphaned_pgrp(struct pid *pgrp, |
326 | struct task_struct *ignored_task) | |
1da177e4 LT |
327 | { |
328 | struct task_struct *p; | |
1da177e4 | 329 | |
0475ac08 | 330 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
331 | if ((p == ignored_task) || |
332 | (p->exit_state && thread_group_empty(p)) || | |
333 | is_global_init(p->real_parent)) | |
1da177e4 | 334 | continue; |
05e83df6 | 335 | |
0475ac08 | 336 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
337 | task_session(p->real_parent) == task_session(p)) |
338 | return 0; | |
0475ac08 | 339 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
340 | |
341 | return 1; | |
1da177e4 LT |
342 | } |
343 | ||
3e7cd6c4 | 344 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
345 | { |
346 | int retval; | |
347 | ||
348 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 349 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
350 | read_unlock(&tasklist_lock); |
351 | ||
352 | return retval; | |
353 | } | |
354 | ||
961c4675 | 355 | static bool has_stopped_jobs(struct pid *pgrp) |
1da177e4 | 356 | { |
1da177e4 LT |
357 | struct task_struct *p; |
358 | ||
0475ac08 | 359 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
961c4675 ON |
360 | if (p->signal->flags & SIGNAL_STOP_STOPPED) |
361 | return true; | |
0475ac08 | 362 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
961c4675 ON |
363 | |
364 | return false; | |
1da177e4 LT |
365 | } |
366 | ||
f49ee505 ON |
367 | /* |
368 | * Check to see if any process groups have become orphaned as | |
369 | * a result of our exiting, and if they have any stopped jobs, | |
370 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
371 | */ | |
372 | static void | |
373 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
374 | { | |
375 | struct pid *pgrp = task_pgrp(tsk); | |
376 | struct task_struct *ignored_task = tsk; | |
377 | ||
378 | if (!parent) | |
a0be55de IA |
379 | /* exit: our father is in a different pgrp than |
380 | * we are and we were the only connection outside. | |
381 | */ | |
f49ee505 ON |
382 | parent = tsk->real_parent; |
383 | else | |
384 | /* reparent: our child is in a different pgrp than | |
385 | * we are, and it was the only connection outside. | |
386 | */ | |
387 | ignored_task = NULL; | |
388 | ||
389 | if (task_pgrp(parent) != pgrp && | |
390 | task_session(parent) == task_session(tsk) && | |
391 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
392 | has_stopped_jobs(pgrp)) { | |
393 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
394 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
395 | } | |
396 | } | |
397 | ||
92307383 | 398 | static void coredump_task_exit(struct task_struct *tsk) |
d67e03e3 EB |
399 | { |
400 | struct core_state *core_state; | |
401 | ||
402 | /* | |
403 | * Serialize with any possible pending coredump. | |
0258b5fd | 404 | * We must hold siglock around checking core_state |
92307383 | 405 | * and setting PF_POSTCOREDUMP. The core-inducing thread |
d67e03e3 | 406 | * will increment ->nr_threads for each thread in the |
92307383 | 407 | * group without PF_POSTCOREDUMP set. |
d67e03e3 | 408 | */ |
0258b5fd | 409 | spin_lock_irq(&tsk->sighand->siglock); |
92307383 | 410 | tsk->flags |= PF_POSTCOREDUMP; |
0258b5fd EB |
411 | core_state = tsk->signal->core_state; |
412 | spin_unlock_irq(&tsk->sighand->siglock); | |
d67e03e3 EB |
413 | if (core_state) { |
414 | struct core_thread self; | |
415 | ||
d67e03e3 EB |
416 | self.task = current; |
417 | if (self.task->flags & PF_SIGNALED) | |
418 | self.next = xchg(&core_state->dumper.next, &self); | |
419 | else | |
420 | self.task = NULL; | |
421 | /* | |
422 | * Implies mb(), the result of xchg() must be visible | |
423 | * to core_state->dumper. | |
424 | */ | |
425 | if (atomic_dec_and_test(&core_state->nr_threads)) | |
426 | complete(&core_state->startup); | |
427 | ||
428 | for (;;) { | |
f5d39b02 | 429 | set_current_state(TASK_UNINTERRUPTIBLE|TASK_FREEZABLE); |
d67e03e3 EB |
430 | if (!self.task) /* see coredump_finish() */ |
431 | break; | |
f5d39b02 | 432 | schedule(); |
d67e03e3 EB |
433 | } |
434 | __set_current_state(TASK_RUNNING); | |
d67e03e3 EB |
435 | } |
436 | } | |
437 | ||
f98bafa0 | 438 | #ifdef CONFIG_MEMCG |
cf475ad2 | 439 | /* |
733eda7a | 440 | * A task is exiting. If it owned this mm, find a new owner for the mm. |
cf475ad2 | 441 | */ |
cf475ad2 BS |
442 | void mm_update_next_owner(struct mm_struct *mm) |
443 | { | |
444 | struct task_struct *c, *g, *p = current; | |
445 | ||
446 | retry: | |
733eda7a KH |
447 | /* |
448 | * If the exiting or execing task is not the owner, it's | |
449 | * someone else's problem. | |
450 | */ | |
451 | if (mm->owner != p) | |
cf475ad2 | 452 | return; |
733eda7a KH |
453 | /* |
454 | * The current owner is exiting/execing and there are no other | |
455 | * candidates. Do not leave the mm pointing to a possibly | |
456 | * freed task structure. | |
457 | */ | |
458 | if (atomic_read(&mm->mm_users) <= 1) { | |
987717e5 | 459 | WRITE_ONCE(mm->owner, NULL); |
733eda7a KH |
460 | return; |
461 | } | |
cf475ad2 BS |
462 | |
463 | read_lock(&tasklist_lock); | |
464 | /* | |
465 | * Search in the children | |
466 | */ | |
467 | list_for_each_entry(c, &p->children, sibling) { | |
468 | if (c->mm == mm) | |
469 | goto assign_new_owner; | |
470 | } | |
471 | ||
472 | /* | |
473 | * Search in the siblings | |
474 | */ | |
dea33cfd | 475 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
cf475ad2 BS |
476 | if (c->mm == mm) |
477 | goto assign_new_owner; | |
478 | } | |
479 | ||
480 | /* | |
f87fb599 | 481 | * Search through everything else, we should not get here often. |
cf475ad2 | 482 | */ |
39af1765 ON |
483 | for_each_process(g) { |
484 | if (g->flags & PF_KTHREAD) | |
485 | continue; | |
486 | for_each_thread(g, c) { | |
487 | if (c->mm == mm) | |
488 | goto assign_new_owner; | |
489 | if (c->mm) | |
490 | break; | |
491 | } | |
f87fb599 | 492 | } |
cf475ad2 | 493 | read_unlock(&tasklist_lock); |
31a78f23 BS |
494 | /* |
495 | * We found no owner yet mm_users > 1: this implies that we are | |
496 | * most likely racing with swapoff (try_to_unuse()) or /proc or | |
e5991371 | 497 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
31a78f23 | 498 | */ |
987717e5 | 499 | WRITE_ONCE(mm->owner, NULL); |
cf475ad2 BS |
500 | return; |
501 | ||
502 | assign_new_owner: | |
503 | BUG_ON(c == p); | |
504 | get_task_struct(c); | |
505 | /* | |
506 | * The task_lock protects c->mm from changing. | |
507 | * We always want mm->owner->mm == mm | |
508 | */ | |
509 | task_lock(c); | |
e5991371 HD |
510 | /* |
511 | * Delay read_unlock() till we have the task_lock() | |
512 | * to ensure that c does not slip away underneath us | |
513 | */ | |
514 | read_unlock(&tasklist_lock); | |
cf475ad2 BS |
515 | if (c->mm != mm) { |
516 | task_unlock(c); | |
517 | put_task_struct(c); | |
518 | goto retry; | |
519 | } | |
987717e5 | 520 | WRITE_ONCE(mm->owner, c); |
bd74fdae | 521 | lru_gen_migrate_mm(mm); |
cf475ad2 BS |
522 | task_unlock(c); |
523 | put_task_struct(c); | |
524 | } | |
f98bafa0 | 525 | #endif /* CONFIG_MEMCG */ |
cf475ad2 | 526 | |
1da177e4 LT |
527 | /* |
528 | * Turn us into a lazy TLB process if we | |
529 | * aren't already.. | |
530 | */ | |
0039962a | 531 | static void exit_mm(void) |
1da177e4 | 532 | { |
0039962a | 533 | struct mm_struct *mm = current->mm; |
1da177e4 | 534 | |
4610ba7a | 535 | exit_mm_release(current, mm); |
1da177e4 LT |
536 | if (!mm) |
537 | return; | |
4fe7efdb | 538 | sync_mm_rss(mm); |
d8ed45c5 | 539 | mmap_read_lock(mm); |
f1f10076 | 540 | mmgrab(mm); |
0039962a | 541 | BUG_ON(mm != current->active_mm); |
1da177e4 | 542 | /* more a memory barrier than a real lock */ |
0039962a | 543 | task_lock(current); |
5bc78502 MD |
544 | /* |
545 | * When a thread stops operating on an address space, the loop | |
546 | * in membarrier_private_expedited() may not observe that | |
547 | * tsk->mm, and the loop in membarrier_global_expedited() may | |
548 | * not observe a MEMBARRIER_STATE_GLOBAL_EXPEDITED | |
549 | * rq->membarrier_state, so those would not issue an IPI. | |
550 | * Membarrier requires a memory barrier after accessing | |
551 | * user-space memory, before clearing tsk->mm or the | |
552 | * rq->membarrier_state. | |
553 | */ | |
554 | smp_mb__after_spinlock(); | |
555 | local_irq_disable(); | |
0039962a | 556 | current->mm = NULL; |
5bc78502 | 557 | membarrier_update_current_mm(NULL); |
1da177e4 | 558 | enter_lazy_tlb(mm, current); |
5bc78502 | 559 | local_irq_enable(); |
0039962a | 560 | task_unlock(current); |
5bc78502 | 561 | mmap_read_unlock(mm); |
cf475ad2 | 562 | mm_update_next_owner(mm); |
1da177e4 | 563 | mmput(mm); |
c32b3cbe | 564 | if (test_thread_flag(TIF_MEMDIE)) |
38531201 | 565 | exit_oom_victim(); |
1da177e4 LT |
566 | } |
567 | ||
c9dc05bf ON |
568 | static struct task_struct *find_alive_thread(struct task_struct *p) |
569 | { | |
570 | struct task_struct *t; | |
571 | ||
572 | for_each_thread(p, t) { | |
573 | if (!(t->flags & PF_EXITING)) | |
574 | return t; | |
575 | } | |
576 | return NULL; | |
577 | } | |
578 | ||
8fb335e0 AV |
579 | static struct task_struct *find_child_reaper(struct task_struct *father, |
580 | struct list_head *dead) | |
1109909c ON |
581 | __releases(&tasklist_lock) |
582 | __acquires(&tasklist_lock) | |
583 | { | |
584 | struct pid_namespace *pid_ns = task_active_pid_ns(father); | |
585 | struct task_struct *reaper = pid_ns->child_reaper; | |
8fb335e0 | 586 | struct task_struct *p, *n; |
1109909c ON |
587 | |
588 | if (likely(reaper != father)) | |
589 | return reaper; | |
590 | ||
c9dc05bf ON |
591 | reaper = find_alive_thread(father); |
592 | if (reaper) { | |
1109909c ON |
593 | pid_ns->child_reaper = reaper; |
594 | return reaper; | |
595 | } | |
596 | ||
597 | write_unlock_irq(&tasklist_lock); | |
8fb335e0 AV |
598 | |
599 | list_for_each_entry_safe(p, n, dead, ptrace_entry) { | |
600 | list_del_init(&p->ptrace_entry); | |
601 | release_task(p); | |
602 | } | |
603 | ||
1109909c ON |
604 | zap_pid_ns_processes(pid_ns); |
605 | write_lock_irq(&tasklist_lock); | |
606 | ||
607 | return father; | |
608 | } | |
609 | ||
1da177e4 | 610 | /* |
ebec18a6 LP |
611 | * When we die, we re-parent all our children, and try to: |
612 | * 1. give them to another thread in our thread group, if such a member exists | |
613 | * 2. give it to the first ancestor process which prctl'd itself as a | |
614 | * child_subreaper for its children (like a service manager) | |
615 | * 3. give it to the init process (PID 1) in our pid namespace | |
1da177e4 | 616 | */ |
1109909c ON |
617 | static struct task_struct *find_new_reaper(struct task_struct *father, |
618 | struct task_struct *child_reaper) | |
1da177e4 | 619 | { |
c9dc05bf | 620 | struct task_struct *thread, *reaper; |
1da177e4 | 621 | |
c9dc05bf ON |
622 | thread = find_alive_thread(father); |
623 | if (thread) | |
950bbabb | 624 | return thread; |
1da177e4 | 625 | |
7d24e2df | 626 | if (father->signal->has_child_subreaper) { |
c6c70f44 | 627 | unsigned int ns_level = task_pid(father)->level; |
ebec18a6 | 628 | /* |
175aed3f | 629 | * Find the first ->is_child_subreaper ancestor in our pid_ns. |
c6c70f44 ON |
630 | * We can't check reaper != child_reaper to ensure we do not |
631 | * cross the namespaces, the exiting parent could be injected | |
632 | * by setns() + fork(). | |
633 | * We check pid->level, this is slightly more efficient than | |
634 | * task_active_pid_ns(reaper) != task_active_pid_ns(father). | |
ebec18a6 | 635 | */ |
c6c70f44 ON |
636 | for (reaper = father->real_parent; |
637 | task_pid(reaper)->level == ns_level; | |
ebec18a6 | 638 | reaper = reaper->real_parent) { |
175aed3f | 639 | if (reaper == &init_task) |
ebec18a6 LP |
640 | break; |
641 | if (!reaper->signal->is_child_subreaper) | |
642 | continue; | |
c9dc05bf ON |
643 | thread = find_alive_thread(reaper); |
644 | if (thread) | |
645 | return thread; | |
ebec18a6 | 646 | } |
1da177e4 | 647 | } |
762a24be | 648 | |
1109909c | 649 | return child_reaper; |
950bbabb ON |
650 | } |
651 | ||
5dfc80be ON |
652 | /* |
653 | * Any that need to be release_task'd are put on the @dead list. | |
654 | */ | |
9cd80bbb | 655 | static void reparent_leader(struct task_struct *father, struct task_struct *p, |
5dfc80be ON |
656 | struct list_head *dead) |
657 | { | |
2831096e | 658 | if (unlikely(p->exit_state == EXIT_DEAD)) |
5dfc80be ON |
659 | return; |
660 | ||
abd50b39 | 661 | /* We don't want people slaying init. */ |
5dfc80be ON |
662 | p->exit_signal = SIGCHLD; |
663 | ||
664 | /* If it has exited notify the new parent about this child's death. */ | |
d21142ec | 665 | if (!p->ptrace && |
5dfc80be | 666 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
86773473 | 667 | if (do_notify_parent(p, p->exit_signal)) { |
5dfc80be | 668 | p->exit_state = EXIT_DEAD; |
dc2fd4b0 | 669 | list_add(&p->ptrace_entry, dead); |
5dfc80be ON |
670 | } |
671 | } | |
672 | ||
673 | kill_orphaned_pgrp(p, father); | |
674 | } | |
675 | ||
482a3767 ON |
676 | /* |
677 | * This does two things: | |
678 | * | |
679 | * A. Make init inherit all the child processes | |
680 | * B. Check to see if any process groups have become orphaned | |
681 | * as a result of our exiting, and if they have any stopped | |
682 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
683 | */ | |
684 | static void forget_original_parent(struct task_struct *father, | |
685 | struct list_head *dead) | |
1da177e4 | 686 | { |
482a3767 | 687 | struct task_struct *p, *t, *reaper; |
762a24be | 688 | |
7c8bd232 | 689 | if (unlikely(!list_empty(&father->ptraced))) |
482a3767 | 690 | exit_ptrace(father, dead); |
f470021a | 691 | |
7c8bd232 | 692 | /* Can drop and reacquire tasklist_lock */ |
8fb335e0 | 693 | reaper = find_child_reaper(father, dead); |
ad9e206a | 694 | if (list_empty(&father->children)) |
482a3767 | 695 | return; |
1109909c ON |
696 | |
697 | reaper = find_new_reaper(father, reaper); | |
2831096e | 698 | list_for_each_entry(p, &father->children, sibling) { |
57a05918 | 699 | for_each_thread(p, t) { |
22a34c6f MB |
700 | RCU_INIT_POINTER(t->real_parent, reaper); |
701 | BUG_ON((!t->ptrace) != (rcu_access_pointer(t->parent) == father)); | |
57a05918 | 702 | if (likely(!t->ptrace)) |
9cd80bbb | 703 | t->parent = t->real_parent; |
9cd80bbb ON |
704 | if (t->pdeath_signal) |
705 | group_send_sig_info(t->pdeath_signal, | |
01024980 EB |
706 | SEND_SIG_NOINFO, t, |
707 | PIDTYPE_TGID); | |
57a05918 | 708 | } |
2831096e ON |
709 | /* |
710 | * If this is a threaded reparent there is no need to | |
711 | * notify anyone anything has happened. | |
712 | */ | |
713 | if (!same_thread_group(reaper, father)) | |
482a3767 | 714 | reparent_leader(father, p, dead); |
1da177e4 | 715 | } |
2831096e | 716 | list_splice_tail_init(&father->children, &reaper->children); |
1da177e4 LT |
717 | } |
718 | ||
719 | /* | |
720 | * Send signals to all our closest relatives so that they know | |
721 | * to properly mourn us.. | |
722 | */ | |
821c7de7 | 723 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 | 724 | { |
53c8f9f1 | 725 | bool autoreap; |
482a3767 ON |
726 | struct task_struct *p, *n; |
727 | LIST_HEAD(dead); | |
1da177e4 | 728 | |
762a24be | 729 | write_lock_irq(&tasklist_lock); |
482a3767 ON |
730 | forget_original_parent(tsk, &dead); |
731 | ||
821c7de7 ON |
732 | if (group_dead) |
733 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 734 | |
b191d649 | 735 | tsk->exit_state = EXIT_ZOMBIE; |
45cdf5cc ON |
736 | if (unlikely(tsk->ptrace)) { |
737 | int sig = thread_group_leader(tsk) && | |
738 | thread_group_empty(tsk) && | |
739 | !ptrace_reparented(tsk) ? | |
740 | tsk->exit_signal : SIGCHLD; | |
741 | autoreap = do_notify_parent(tsk, sig); | |
742 | } else if (thread_group_leader(tsk)) { | |
743 | autoreap = thread_group_empty(tsk) && | |
744 | do_notify_parent(tsk, tsk->exit_signal); | |
745 | } else { | |
746 | autoreap = true; | |
747 | } | |
1da177e4 | 748 | |
30b692d3 CB |
749 | if (autoreap) { |
750 | tsk->exit_state = EXIT_DEAD; | |
6c66e7db | 751 | list_add(&tsk->ptrace_entry, &dead); |
30b692d3 | 752 | } |
1da177e4 | 753 | |
9c339168 ON |
754 | /* mt-exec, de_thread() is waiting for group leader */ |
755 | if (unlikely(tsk->signal->notify_count < 0)) | |
60700e38 | 756 | wake_up_process(tsk->signal->group_exec_task); |
1da177e4 LT |
757 | write_unlock_irq(&tasklist_lock); |
758 | ||
482a3767 ON |
759 | list_for_each_entry_safe(p, n, &dead, ptrace_entry) { |
760 | list_del_init(&p->ptrace_entry); | |
761 | release_task(p); | |
762 | } | |
1da177e4 LT |
763 | } |
764 | ||
e18eecb8 JD |
765 | #ifdef CONFIG_DEBUG_STACK_USAGE |
766 | static void check_stack_usage(void) | |
767 | { | |
768 | static DEFINE_SPINLOCK(low_water_lock); | |
769 | static int lowest_to_date = THREAD_SIZE; | |
e18eecb8 JD |
770 | unsigned long free; |
771 | ||
7c9f8861 | 772 | free = stack_not_used(current); |
e18eecb8 JD |
773 | |
774 | if (free >= lowest_to_date) | |
775 | return; | |
776 | ||
777 | spin_lock(&low_water_lock); | |
778 | if (free < lowest_to_date) { | |
627393d4 | 779 | pr_info("%s (%d) used greatest stack depth: %lu bytes left\n", |
a0be55de | 780 | current->comm, task_pid_nr(current), free); |
e18eecb8 JD |
781 | lowest_to_date = free; |
782 | } | |
783 | spin_unlock(&low_water_lock); | |
784 | } | |
785 | #else | |
786 | static inline void check_stack_usage(void) {} | |
787 | #endif | |
788 | ||
d80f7d7b EB |
789 | static void synchronize_group_exit(struct task_struct *tsk, long code) |
790 | { | |
791 | struct sighand_struct *sighand = tsk->sighand; | |
792 | struct signal_struct *signal = tsk->signal; | |
793 | ||
794 | spin_lock_irq(&sighand->siglock); | |
795 | signal->quick_threads--; | |
796 | if ((signal->quick_threads == 0) && | |
797 | !(signal->flags & SIGNAL_GROUP_EXIT)) { | |
798 | signal->flags = SIGNAL_GROUP_EXIT; | |
799 | signal->group_exit_code = code; | |
800 | signal->group_stop_count = 0; | |
801 | } | |
802 | spin_unlock_irq(&sighand->siglock); | |
803 | } | |
804 | ||
9af6528e | 805 | void __noreturn do_exit(long code) |
1da177e4 LT |
806 | { |
807 | struct task_struct *tsk = current; | |
808 | int group_dead; | |
809 | ||
001c28e5 NP |
810 | WARN_ON(irqs_disabled()); |
811 | ||
d80f7d7b EB |
812 | synchronize_group_exit(tsk, code); |
813 | ||
b1f866b0 | 814 | WARN_ON(tsk->plug); |
22e2c507 | 815 | |
586b58ca | 816 | kcov_task_exit(tsk); |
50b5e49c | 817 | kmsan_task_exit(tsk); |
586b58ca | 818 | |
92307383 | 819 | coredump_task_exit(tsk); |
a288eecc | 820 | ptrace_event(PTRACE_EVENT_EXIT, code); |
1da177e4 | 821 | |
e0e81739 DH |
822 | validate_creds_for_do_exit(tsk); |
823 | ||
f552a27a | 824 | io_uring_files_cancel(); |
d12619b5 | 825 | exit_signals(tsk); /* sets PF_EXITING */ |
1da177e4 | 826 | |
48d212a2 LT |
827 | /* sync mm's RSS info before statistics gathering */ |
828 | if (tsk->mm) | |
829 | sync_mm_rss(tsk->mm); | |
51229b49 | 830 | acct_update_integrals(tsk); |
1da177e4 | 831 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 832 | if (group_dead) { |
43cf75d9 | 833 | /* |
834 | * If the last thread of global init has exited, panic | |
835 | * immediately to get a useable coredump. | |
836 | */ | |
837 | if (unlikely(is_global_init(tsk))) | |
838 | panic("Attempted to kill init! exitcode=0x%08x\n", | |
839 | tsk->signal->group_exit_code ?: (int)code); | |
840 | ||
baa73d9e | 841 | #ifdef CONFIG_POSIX_TIMERS |
778e9a9c | 842 | hrtimer_cancel(&tsk->signal->real_timer); |
d5b36a4d | 843 | exit_itimers(tsk); |
baa73d9e | 844 | #endif |
1f10206c JP |
845 | if (tsk->mm) |
846 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); | |
c3068951 | 847 | } |
f6ec29a4 | 848 | acct_collect(code, group_dead); |
522ed776 MT |
849 | if (group_dead) |
850 | tty_audit_exit(); | |
a4ff8dba | 851 | audit_free(tsk); |
115085ea | 852 | |
48d212a2 | 853 | tsk->exit_code = code; |
115085ea | 854 | taskstats_exit(tsk, group_dead); |
c757249a | 855 | |
0039962a | 856 | exit_mm(); |
1da177e4 | 857 | |
0e464814 | 858 | if (group_dead) |
f6ec29a4 | 859 | acct_process(); |
0a16b607 MD |
860 | trace_sched_process_exit(tsk); |
861 | ||
1da177e4 | 862 | exit_sem(tsk); |
b34a6b1d | 863 | exit_shm(tsk); |
1ec7f1dd AV |
864 | exit_files(tsk); |
865 | exit_fs(tsk); | |
c39df5fa ON |
866 | if (group_dead) |
867 | disassociate_ctty(1); | |
8aac6270 | 868 | exit_task_namespaces(tsk); |
ed3e694d | 869 | exit_task_work(tsk); |
e6464694 | 870 | exit_thread(tsk); |
0b3fcf17 SE |
871 | |
872 | /* | |
873 | * Flush inherited counters to the parent - before the parent | |
874 | * gets woken up by child-exit notifications. | |
875 | * | |
876 | * because of cgroup mode, must be called before cgroup_exit() | |
877 | */ | |
878 | perf_event_exit_task(tsk); | |
879 | ||
8e5bfa8c | 880 | sched_autogroup_exit_task(tsk); |
1ec41830 | 881 | cgroup_exit(tsk); |
1da177e4 | 882 | |
24f1e32c FW |
883 | /* |
884 | * FIXME: do that only when needed, using sched_exit tracepoint | |
885 | */ | |
7c8df286 | 886 | flush_ptrace_hw_breakpoint(tsk); |
33b2fb30 | 887 | |
ccdd29ff | 888 | exit_tasks_rcu_start(); |
821c7de7 | 889 | exit_notify(tsk, group_dead); |
ef982393 | 890 | proc_exit_connector(tsk); |
c11600e4 | 891 | mpol_put_task_policy(tsk); |
42b2dd0a | 892 | #ifdef CONFIG_FUTEX |
c87e2837 IM |
893 | if (unlikely(current->pi_state_cache)) |
894 | kfree(current->pi_state_cache); | |
42b2dd0a | 895 | #endif |
de5097c2 | 896 | /* |
9a11b49a | 897 | * Make sure we are holding no locks: |
de5097c2 | 898 | */ |
1b1d2fb4 | 899 | debug_check_no_locks_held(); |
1da177e4 | 900 | |
afc847b7 | 901 | if (tsk->io_context) |
b69f2292 | 902 | exit_io_context(tsk); |
afc847b7 | 903 | |
b92ce558 | 904 | if (tsk->splice_pipe) |
4b8a8f1e | 905 | free_pipe_info(tsk->splice_pipe); |
b92ce558 | 906 | |
5640f768 ED |
907 | if (tsk->task_frag.page) |
908 | put_page(tsk->task_frag.page); | |
909 | ||
e0e81739 | 910 | validate_creds_for_do_exit(tsk); |
1a03d3f1 | 911 | exit_task_stack_account(tsk); |
e0e81739 | 912 | |
4bcb8232 | 913 | check_stack_usage(); |
7407251a | 914 | preempt_disable(); |
54848d73 WF |
915 | if (tsk->nr_dirtied) |
916 | __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied); | |
f41d911f | 917 | exit_rcu(); |
ccdd29ff | 918 | exit_tasks_rcu_finish(); |
b5740f4b | 919 | |
b09be676 | 920 | lockdep_free_task(tsk); |
9af6528e | 921 | do_task_dead(); |
1da177e4 | 922 | } |
012914da | 923 | |
0e25498f EB |
924 | void __noreturn make_task_dead(int signr) |
925 | { | |
926 | /* | |
927 | * Take the task off the cpu after something catastrophic has | |
928 | * happened. | |
05ea0424 EB |
929 | * |
930 | * We can get here from a kernel oops, sometimes with preemption off. | |
931 | * Start by checking for critical errors. | |
932 | * Then fix up important state like USER_DS and preemption. | |
933 | * Then do everything else. | |
0e25498f | 934 | */ |
05ea0424 | 935 | struct task_struct *tsk = current; |
7535b832 | 936 | unsigned int limit; |
05ea0424 EB |
937 | |
938 | if (unlikely(in_interrupt())) | |
939 | panic("Aiee, killing interrupt handler!"); | |
940 | if (unlikely(!tsk->pid)) | |
941 | panic("Attempted to kill the idle task!"); | |
942 | ||
001c28e5 NP |
943 | if (unlikely(irqs_disabled())) { |
944 | pr_info("note: %s[%d] exited with irqs disabled\n", | |
945 | current->comm, task_pid_nr(current)); | |
946 | local_irq_enable(); | |
947 | } | |
05ea0424 EB |
948 | if (unlikely(in_atomic())) { |
949 | pr_info("note: %s[%d] exited with preempt_count %d\n", | |
950 | current->comm, task_pid_nr(current), | |
951 | preempt_count()); | |
952 | preempt_count_set(PREEMPT_ENABLED); | |
953 | } | |
954 | ||
d4ccd54d JH |
955 | /* |
956 | * Every time the system oopses, if the oops happens while a reference | |
957 | * to an object was held, the reference leaks. | |
958 | * If the oops doesn't also leak memory, repeated oopsing can cause | |
959 | * reference counters to wrap around (if they're not using refcount_t). | |
960 | * This means that repeated oopsing can make unexploitable-looking bugs | |
961 | * exploitable through repeated oopsing. | |
962 | * To make sure this can't happen, place an upper bound on how often the | |
963 | * kernel may oops without panic(). | |
964 | */ | |
7535b832 KC |
965 | limit = READ_ONCE(oops_limit); |
966 | if (atomic_inc_return(&oops_count) >= limit && limit) | |
967 | panic("Oopsed too often (kernel.oops_limit is %d)", limit); | |
d4ccd54d | 968 | |
05ea0424 EB |
969 | /* |
970 | * We're taking recursive faults here in make_task_dead. Safest is to just | |
971 | * leave this task alone and wait for reboot. | |
972 | */ | |
973 | if (unlikely(tsk->flags & PF_EXITING)) { | |
974 | pr_alert("Fixing recursive fault but reboot is needed!\n"); | |
975 | futex_exit_recursive(tsk); | |
912616f1 EB |
976 | tsk->exit_state = EXIT_DEAD; |
977 | refcount_inc(&tsk->rcu_users); | |
7f80a2fd | 978 | do_task_dead(); |
05ea0424 EB |
979 | } |
980 | ||
0e25498f EB |
981 | do_exit(signr); |
982 | } | |
983 | ||
754fe8d2 | 984 | SYSCALL_DEFINE1(exit, int, error_code) |
1da177e4 LT |
985 | { |
986 | do_exit((error_code&0xff)<<8); | |
987 | } | |
988 | ||
1da177e4 LT |
989 | /* |
990 | * Take down every thread in the group. This is called by fatal signals | |
991 | * as well as by sys_exit_group (below). | |
992 | */ | |
eae654f1 | 993 | void __noreturn |
1da177e4 LT |
994 | do_group_exit(int exit_code) |
995 | { | |
bfc4b089 ON |
996 | struct signal_struct *sig = current->signal; |
997 | ||
49697335 | 998 | if (sig->flags & SIGNAL_GROUP_EXIT) |
bfc4b089 | 999 | exit_code = sig->group_exit_code; |
49697335 EB |
1000 | else if (sig->group_exec_task) |
1001 | exit_code = 0; | |
cbe9dac3 | 1002 | else { |
1da177e4 | 1003 | struct sighand_struct *const sighand = current->sighand; |
a0be55de | 1004 | |
1da177e4 | 1005 | spin_lock_irq(&sighand->siglock); |
49697335 | 1006 | if (sig->flags & SIGNAL_GROUP_EXIT) |
1da177e4 LT |
1007 | /* Another thread got here before we took the lock. */ |
1008 | exit_code = sig->group_exit_code; | |
49697335 EB |
1009 | else if (sig->group_exec_task) |
1010 | exit_code = 0; | |
1da177e4 | 1011 | else { |
1da177e4 | 1012 | sig->group_exit_code = exit_code; |
ed5d2cac | 1013 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
1014 | zap_other_threads(current); |
1015 | } | |
1016 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
1017 | } |
1018 | ||
1019 | do_exit(exit_code); | |
1020 | /* NOTREACHED */ | |
1021 | } | |
1022 | ||
1023 | /* | |
1024 | * this kills every thread in the thread group. Note that any externally | |
1025 | * wait4()-ing process will get the correct exit code - even if this | |
1026 | * thread is not the thread group leader. | |
1027 | */ | |
754fe8d2 | 1028 | SYSCALL_DEFINE1(exit_group, int, error_code) |
1da177e4 LT |
1029 | { |
1030 | do_group_exit((error_code & 0xff) << 8); | |
2ed7c03e HC |
1031 | /* NOTREACHED */ |
1032 | return 0; | |
1da177e4 LT |
1033 | } |
1034 | ||
67d7ddde AV |
1035 | struct waitid_info { |
1036 | pid_t pid; | |
1037 | uid_t uid; | |
1038 | int status; | |
1039 | int cause; | |
1040 | }; | |
1041 | ||
9e8ae01d ON |
1042 | struct wait_opts { |
1043 | enum pid_type wo_type; | |
9e8ae01d | 1044 | int wo_flags; |
e1eb1ebc | 1045 | struct pid *wo_pid; |
9e8ae01d | 1046 | |
67d7ddde | 1047 | struct waitid_info *wo_info; |
359566fa | 1048 | int wo_stat; |
ce72a16f | 1049 | struct rusage *wo_rusage; |
9e8ae01d | 1050 | |
ac6424b9 | 1051 | wait_queue_entry_t child_wait; |
9e8ae01d ON |
1052 | int notask_error; |
1053 | }; | |
1054 | ||
989264f4 | 1055 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1056 | { |
5c01ba49 ON |
1057 | return wo->wo_type == PIDTYPE_MAX || |
1058 | task_pid_type(p, wo->wo_type) == wo->wo_pid; | |
1059 | } | |
1da177e4 | 1060 | |
bf959931 ON |
1061 | static int |
1062 | eligible_child(struct wait_opts *wo, bool ptrace, struct task_struct *p) | |
5c01ba49 ON |
1063 | { |
1064 | if (!eligible_pid(wo, p)) | |
1065 | return 0; | |
bf959931 ON |
1066 | |
1067 | /* | |
1068 | * Wait for all children (clone and not) if __WALL is set or | |
1069 | * if it is traced by us. | |
1070 | */ | |
1071 | if (ptrace || (wo->wo_flags & __WALL)) | |
1072 | return 1; | |
1073 | ||
1074 | /* | |
1075 | * Otherwise, wait for clone children *only* if __WCLONE is set; | |
1076 | * otherwise, wait for non-clone children *only*. | |
1077 | * | |
1078 | * Note: a "clone" child here is one that reports to its parent | |
1079 | * using a signal other than SIGCHLD, or a non-leader thread which | |
1080 | * we can only see if it is traced by us. | |
1081 | */ | |
1082 | if ((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) | |
1da177e4 | 1083 | return 0; |
1da177e4 | 1084 | |
14dd0b81 | 1085 | return 1; |
1da177e4 LT |
1086 | } |
1087 | ||
1da177e4 LT |
1088 | /* |
1089 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
1090 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1091 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1092 | * released the lock and the system call should return. | |
1093 | */ | |
9e8ae01d | 1094 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1095 | { |
67d7ddde | 1096 | int state, status; |
6c5f3e7b | 1097 | pid_t pid = task_pid_vnr(p); |
43e13cc1 | 1098 | uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
67d7ddde | 1099 | struct waitid_info *infop; |
1da177e4 | 1100 | |
9e8ae01d | 1101 | if (!likely(wo->wo_flags & WEXITED)) |
98abed02 RM |
1102 | return 0; |
1103 | ||
9e8ae01d | 1104 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
907c311f EB |
1105 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1106 | ? p->signal->group_exit_code : p->exit_code; | |
1da177e4 LT |
1107 | get_task_struct(p); |
1108 | read_unlock(&tasklist_lock); | |
1029a2b5 | 1109 | sched_annotate_sleep(); |
e61a2502 AV |
1110 | if (wo->wo_rusage) |
1111 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1112 | put_task_struct(p); |
76d9871e | 1113 | goto out_info; |
1da177e4 | 1114 | } |
1da177e4 | 1115 | /* |
abd50b39 | 1116 | * Move the task's state to DEAD/TRACE, only one thread can do this. |
1da177e4 | 1117 | */ |
f6507f83 ON |
1118 | state = (ptrace_reparented(p) && thread_group_leader(p)) ? |
1119 | EXIT_TRACE : EXIT_DEAD; | |
abd50b39 | 1120 | if (cmpxchg(&p->exit_state, EXIT_ZOMBIE, state) != EXIT_ZOMBIE) |
1da177e4 | 1121 | return 0; |
986094df ON |
1122 | /* |
1123 | * We own this thread, nobody else can reap it. | |
1124 | */ | |
1125 | read_unlock(&tasklist_lock); | |
1126 | sched_annotate_sleep(); | |
f6507f83 | 1127 | |
befca967 | 1128 | /* |
f6507f83 | 1129 | * Check thread_group_leader() to exclude the traced sub-threads. |
befca967 | 1130 | */ |
f6507f83 | 1131 | if (state == EXIT_DEAD && thread_group_leader(p)) { |
f953ccd0 ON |
1132 | struct signal_struct *sig = p->signal; |
1133 | struct signal_struct *psig = current->signal; | |
1f10206c | 1134 | unsigned long maxrss; |
5613fda9 | 1135 | u64 tgutime, tgstime; |
3795e161 | 1136 | |
1da177e4 LT |
1137 | /* |
1138 | * The resource counters for the group leader are in its | |
1139 | * own task_struct. Those for dead threads in the group | |
1140 | * are in its signal_struct, as are those for the child | |
1141 | * processes it has previously reaped. All these | |
1142 | * accumulate in the parent's signal_struct c* fields. | |
1143 | * | |
1144 | * We don't bother to take a lock here to protect these | |
f953ccd0 ON |
1145 | * p->signal fields because the whole thread group is dead |
1146 | * and nobody can change them. | |
1147 | * | |
dcca3475 | 1148 | * psig->stats_lock also protects us from our sub-threads |
f953ccd0 ON |
1149 | * which can reap other children at the same time. Until |
1150 | * we change k_getrusage()-like users to rely on this lock | |
1151 | * we have to take ->siglock as well. | |
0cf55e1e | 1152 | * |
a0be55de IA |
1153 | * We use thread_group_cputime_adjusted() to get times for |
1154 | * the thread group, which consolidates times for all threads | |
1155 | * in the group including the group leader. | |
1da177e4 | 1156 | */ |
e80d0a1a | 1157 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); |
f953ccd0 | 1158 | spin_lock_irq(¤t->sighand->siglock); |
e78c3496 | 1159 | write_seqlock(&psig->stats_lock); |
64861634 MS |
1160 | psig->cutime += tgutime + sig->cutime; |
1161 | psig->cstime += tgstime + sig->cstime; | |
6fac4829 | 1162 | psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime; |
3795e161 JJ |
1163 | psig->cmin_flt += |
1164 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1165 | psig->cmaj_flt += | |
1166 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1167 | psig->cnvcsw += | |
1168 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1169 | psig->cnivcsw += | |
1170 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1171 | psig->cinblock += |
1172 | task_io_get_inblock(p) + | |
1173 | sig->inblock + sig->cinblock; | |
1174 | psig->coublock += | |
1175 | task_io_get_oublock(p) + | |
1176 | sig->oublock + sig->coublock; | |
1f10206c JP |
1177 | maxrss = max(sig->maxrss, sig->cmaxrss); |
1178 | if (psig->cmaxrss < maxrss) | |
1179 | psig->cmaxrss = maxrss; | |
5995477a AR |
1180 | task_io_accounting_add(&psig->ioac, &p->ioac); |
1181 | task_io_accounting_add(&psig->ioac, &sig->ioac); | |
e78c3496 | 1182 | write_sequnlock(&psig->stats_lock); |
f953ccd0 | 1183 | spin_unlock_irq(¤t->sighand->siglock); |
1da177e4 LT |
1184 | } |
1185 | ||
ce72a16f AV |
1186 | if (wo->wo_rusage) |
1187 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
1da177e4 LT |
1188 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1189 | ? p->signal->group_exit_code : p->exit_code; | |
359566fa | 1190 | wo->wo_stat = status; |
2f4e6e2a | 1191 | |
b4360690 | 1192 | if (state == EXIT_TRACE) { |
1da177e4 | 1193 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1194 | /* We dropped tasklist, ptracer could die and untrace */ |
1195 | ptrace_unlink(p); | |
b4360690 ON |
1196 | |
1197 | /* If parent wants a zombie, don't release it now */ | |
1198 | state = EXIT_ZOMBIE; | |
1199 | if (do_notify_parent(p, p->exit_signal)) | |
1200 | state = EXIT_DEAD; | |
abd50b39 | 1201 | p->exit_state = state; |
1da177e4 LT |
1202 | write_unlock_irq(&tasklist_lock); |
1203 | } | |
abd50b39 | 1204 | if (state == EXIT_DEAD) |
1da177e4 | 1205 | release_task(p); |
2f4e6e2a | 1206 | |
76d9871e AV |
1207 | out_info: |
1208 | infop = wo->wo_info; | |
1209 | if (infop) { | |
1210 | if ((status & 0x7f) == 0) { | |
1211 | infop->cause = CLD_EXITED; | |
1212 | infop->status = status >> 8; | |
1213 | } else { | |
1214 | infop->cause = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1215 | infop->status = status & 0x7f; | |
1216 | } | |
1217 | infop->pid = pid; | |
1218 | infop->uid = uid; | |
1219 | } | |
1220 | ||
67d7ddde | 1221 | return pid; |
1da177e4 LT |
1222 | } |
1223 | ||
90bc8d8b ON |
1224 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
1225 | { | |
1226 | if (ptrace) { | |
570ac933 | 1227 | if (task_is_traced(p) && !(p->jobctl & JOBCTL_LISTENING)) |
90bc8d8b ON |
1228 | return &p->exit_code; |
1229 | } else { | |
1230 | if (p->signal->flags & SIGNAL_STOP_STOPPED) | |
1231 | return &p->signal->group_exit_code; | |
1232 | } | |
1233 | return NULL; | |
1234 | } | |
1235 | ||
19e27463 TH |
1236 | /** |
1237 | * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED | |
1238 | * @wo: wait options | |
1239 | * @ptrace: is the wait for ptrace | |
1240 | * @p: task to wait for | |
1241 | * | |
1242 | * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED. | |
1243 | * | |
1244 | * CONTEXT: | |
1245 | * read_lock(&tasklist_lock), which is released if return value is | |
1246 | * non-zero. Also, grabs and releases @p->sighand->siglock. | |
1247 | * | |
1248 | * RETURNS: | |
1249 | * 0 if wait condition didn't exist and search for other wait conditions | |
1250 | * should continue. Non-zero return, -errno on failure and @p's pid on | |
1251 | * success, implies that tasklist_lock is released and wait condition | |
1252 | * search should terminate. | |
1da177e4 | 1253 | */ |
9e8ae01d ON |
1254 | static int wait_task_stopped(struct wait_opts *wo, |
1255 | int ptrace, struct task_struct *p) | |
1da177e4 | 1256 | { |
67d7ddde AV |
1257 | struct waitid_info *infop; |
1258 | int exit_code, *p_code, why; | |
ee7c82da | 1259 | uid_t uid = 0; /* unneeded, required by compiler */ |
c8950783 | 1260 | pid_t pid; |
1da177e4 | 1261 | |
47918025 ON |
1262 | /* |
1263 | * Traditionally we see ptrace'd stopped tasks regardless of options. | |
1264 | */ | |
9e8ae01d | 1265 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
98abed02 RM |
1266 | return 0; |
1267 | ||
19e27463 TH |
1268 | if (!task_stopped_code(p, ptrace)) |
1269 | return 0; | |
1270 | ||
ee7c82da ON |
1271 | exit_code = 0; |
1272 | spin_lock_irq(&p->sighand->siglock); | |
1273 | ||
90bc8d8b ON |
1274 | p_code = task_stopped_code(p, ptrace); |
1275 | if (unlikely(!p_code)) | |
ee7c82da ON |
1276 | goto unlock_sig; |
1277 | ||
90bc8d8b | 1278 | exit_code = *p_code; |
ee7c82da ON |
1279 | if (!exit_code) |
1280 | goto unlock_sig; | |
1281 | ||
9e8ae01d | 1282 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
90bc8d8b | 1283 | *p_code = 0; |
ee7c82da | 1284 | |
8ca937a6 | 1285 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
ee7c82da ON |
1286 | unlock_sig: |
1287 | spin_unlock_irq(&p->sighand->siglock); | |
1288 | if (!exit_code) | |
1da177e4 LT |
1289 | return 0; |
1290 | ||
1291 | /* | |
1292 | * Now we are pretty sure this task is interesting. | |
1293 | * Make sure it doesn't get reaped out from under us while we | |
1294 | * give up the lock and then examine it below. We don't want to | |
1295 | * keep holding onto the tasklist_lock while we call getrusage and | |
1296 | * possibly take page faults for user memory. | |
1297 | */ | |
1298 | get_task_struct(p); | |
6c5f3e7b | 1299 | pid = task_pid_vnr(p); |
f470021a | 1300 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
1da177e4 | 1301 | read_unlock(&tasklist_lock); |
1029a2b5 | 1302 | sched_annotate_sleep(); |
e61a2502 AV |
1303 | if (wo->wo_rusage) |
1304 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1305 | put_task_struct(p); |
1da177e4 | 1306 | |
bb380ec3 AV |
1307 | if (likely(!(wo->wo_flags & WNOWAIT))) |
1308 | wo->wo_stat = (exit_code << 8) | 0x7f; | |
1da177e4 | 1309 | |
9e8ae01d | 1310 | infop = wo->wo_info; |
67d7ddde AV |
1311 | if (infop) { |
1312 | infop->cause = why; | |
1313 | infop->status = exit_code; | |
1314 | infop->pid = pid; | |
1315 | infop->uid = uid; | |
1316 | } | |
67d7ddde | 1317 | return pid; |
1da177e4 LT |
1318 | } |
1319 | ||
1320 | /* | |
1321 | * Handle do_wait work for one task in a live, non-stopped state. | |
1322 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1323 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1324 | * released the lock and the system call should return. | |
1325 | */ | |
9e8ae01d | 1326 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1327 | { |
bb380ec3 | 1328 | struct waitid_info *infop; |
1da177e4 LT |
1329 | pid_t pid; |
1330 | uid_t uid; | |
1331 | ||
9e8ae01d | 1332 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
98abed02 RM |
1333 | return 0; |
1334 | ||
1da177e4 LT |
1335 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1336 | return 0; | |
1337 | ||
1338 | spin_lock_irq(&p->sighand->siglock); | |
1339 | /* Re-check with the lock held. */ | |
1340 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1341 | spin_unlock_irq(&p->sighand->siglock); | |
1342 | return 0; | |
1343 | } | |
9e8ae01d | 1344 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
1da177e4 | 1345 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
8ca937a6 | 1346 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
1da177e4 LT |
1347 | spin_unlock_irq(&p->sighand->siglock); |
1348 | ||
6c5f3e7b | 1349 | pid = task_pid_vnr(p); |
1da177e4 LT |
1350 | get_task_struct(p); |
1351 | read_unlock(&tasklist_lock); | |
1029a2b5 | 1352 | sched_annotate_sleep(); |
e61a2502 AV |
1353 | if (wo->wo_rusage) |
1354 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1355 | put_task_struct(p); |
1da177e4 | 1356 | |
bb380ec3 AV |
1357 | infop = wo->wo_info; |
1358 | if (!infop) { | |
359566fa | 1359 | wo->wo_stat = 0xffff; |
1da177e4 | 1360 | } else { |
bb380ec3 AV |
1361 | infop->cause = CLD_CONTINUED; |
1362 | infop->pid = pid; | |
1363 | infop->uid = uid; | |
1364 | infop->status = SIGCONT; | |
1da177e4 | 1365 | } |
bb380ec3 | 1366 | return pid; |
1da177e4 LT |
1367 | } |
1368 | ||
98abed02 RM |
1369 | /* |
1370 | * Consider @p for a wait by @parent. | |
1371 | * | |
9e8ae01d | 1372 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1373 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1374 | * Returns zero if the search for a child should continue; | |
9e8ae01d | 1375 | * then ->notask_error is 0 if @p is an eligible child, |
3a2f5a59 | 1376 | * or still -ECHILD. |
98abed02 | 1377 | */ |
b6e763f0 ON |
1378 | static int wait_consider_task(struct wait_opts *wo, int ptrace, |
1379 | struct task_struct *p) | |
98abed02 | 1380 | { |
3245d6ac ON |
1381 | /* |
1382 | * We can race with wait_task_zombie() from another thread. | |
1383 | * Ensure that EXIT_ZOMBIE -> EXIT_DEAD/EXIT_TRACE transition | |
1384 | * can't confuse the checks below. | |
1385 | */ | |
6aa7de05 | 1386 | int exit_state = READ_ONCE(p->exit_state); |
b3ab0316 ON |
1387 | int ret; |
1388 | ||
3245d6ac | 1389 | if (unlikely(exit_state == EXIT_DEAD)) |
b3ab0316 ON |
1390 | return 0; |
1391 | ||
bf959931 | 1392 | ret = eligible_child(wo, ptrace, p); |
14dd0b81 | 1393 | if (!ret) |
98abed02 RM |
1394 | return ret; |
1395 | ||
3245d6ac | 1396 | if (unlikely(exit_state == EXIT_TRACE)) { |
50b8d257 | 1397 | /* |
abd50b39 ON |
1398 | * ptrace == 0 means we are the natural parent. In this case |
1399 | * we should clear notask_error, debugger will notify us. | |
50b8d257 | 1400 | */ |
abd50b39 | 1401 | if (likely(!ptrace)) |
50b8d257 | 1402 | wo->notask_error = 0; |
823b018e | 1403 | return 0; |
50b8d257 | 1404 | } |
823b018e | 1405 | |
377d75da ON |
1406 | if (likely(!ptrace) && unlikely(p->ptrace)) { |
1407 | /* | |
1408 | * If it is traced by its real parent's group, just pretend | |
1409 | * the caller is ptrace_do_wait() and reap this child if it | |
1410 | * is zombie. | |
1411 | * | |
1412 | * This also hides group stop state from real parent; otherwise | |
1413 | * a single stop can be reported twice as group and ptrace stop. | |
1414 | * If a ptracer wants to distinguish these two events for its | |
1415 | * own children it should create a separate process which takes | |
1416 | * the role of real parent. | |
1417 | */ | |
1418 | if (!ptrace_reparented(p)) | |
1419 | ptrace = 1; | |
1420 | } | |
1421 | ||
45cb24a1 | 1422 | /* slay zombie? */ |
3245d6ac | 1423 | if (exit_state == EXIT_ZOMBIE) { |
9b84cca2 | 1424 | /* we don't reap group leaders with subthreads */ |
7c733eb3 ON |
1425 | if (!delay_group_leader(p)) { |
1426 | /* | |
1427 | * A zombie ptracee is only visible to its ptracer. | |
1428 | * Notification and reaping will be cascaded to the | |
1429 | * real parent when the ptracer detaches. | |
1430 | */ | |
1431 | if (unlikely(ptrace) || likely(!p->ptrace)) | |
1432 | return wait_task_zombie(wo, p); | |
1433 | } | |
98abed02 | 1434 | |
f470021a | 1435 | /* |
9b84cca2 TH |
1436 | * Allow access to stopped/continued state via zombie by |
1437 | * falling through. Clearing of notask_error is complex. | |
1438 | * | |
1439 | * When !@ptrace: | |
1440 | * | |
1441 | * If WEXITED is set, notask_error should naturally be | |
1442 | * cleared. If not, subset of WSTOPPED|WCONTINUED is set, | |
1443 | * so, if there are live subthreads, there are events to | |
1444 | * wait for. If all subthreads are dead, it's still safe | |
1445 | * to clear - this function will be called again in finite | |
1446 | * amount time once all the subthreads are released and | |
1447 | * will then return without clearing. | |
1448 | * | |
1449 | * When @ptrace: | |
1450 | * | |
1451 | * Stopped state is per-task and thus can't change once the | |
1452 | * target task dies. Only continued and exited can happen. | |
1453 | * Clear notask_error if WCONTINUED | WEXITED. | |
1454 | */ | |
1455 | if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED))) | |
1456 | wo->notask_error = 0; | |
1457 | } else { | |
1458 | /* | |
1459 | * @p is alive and it's gonna stop, continue or exit, so | |
1460 | * there always is something to wait for. | |
f470021a | 1461 | */ |
9e8ae01d | 1462 | wo->notask_error = 0; |
f470021a RM |
1463 | } |
1464 | ||
98abed02 | 1465 | /* |
45cb24a1 TH |
1466 | * Wait for stopped. Depending on @ptrace, different stopped state |
1467 | * is used and the two don't interact with each other. | |
98abed02 | 1468 | */ |
19e27463 TH |
1469 | ret = wait_task_stopped(wo, ptrace, p); |
1470 | if (ret) | |
1471 | return ret; | |
98abed02 RM |
1472 | |
1473 | /* | |
45cb24a1 TH |
1474 | * Wait for continued. There's only one continued state and the |
1475 | * ptracer can consume it which can confuse the real parent. Don't | |
1476 | * use WCONTINUED from ptracer. You don't need or want it. | |
98abed02 | 1477 | */ |
9e8ae01d | 1478 | return wait_task_continued(wo, p); |
98abed02 RM |
1479 | } |
1480 | ||
1481 | /* | |
1482 | * Do the work of do_wait() for one thread in the group, @tsk. | |
1483 | * | |
9e8ae01d | 1484 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1485 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1486 | * Returns zero if the search for a child should continue; then | |
9e8ae01d | 1487 | * ->notask_error is 0 if there were any eligible children, |
3a2f5a59 | 1488 | * or still -ECHILD. |
98abed02 | 1489 | */ |
9e8ae01d | 1490 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1491 | { |
1492 | struct task_struct *p; | |
1493 | ||
1494 | list_for_each_entry(p, &tsk->children, sibling) { | |
9cd80bbb | 1495 | int ret = wait_consider_task(wo, 0, p); |
a0be55de | 1496 | |
9cd80bbb ON |
1497 | if (ret) |
1498 | return ret; | |
98abed02 RM |
1499 | } |
1500 | ||
1501 | return 0; | |
1502 | } | |
1503 | ||
9e8ae01d | 1504 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1505 | { |
1506 | struct task_struct *p; | |
1507 | ||
f470021a | 1508 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
b6e763f0 | 1509 | int ret = wait_consider_task(wo, 1, p); |
a0be55de | 1510 | |
f470021a | 1511 | if (ret) |
98abed02 | 1512 | return ret; |
98abed02 RM |
1513 | } |
1514 | ||
1515 | return 0; | |
1516 | } | |
1517 | ||
ac6424b9 | 1518 | static int child_wait_callback(wait_queue_entry_t *wait, unsigned mode, |
0b7570e7 ON |
1519 | int sync, void *key) |
1520 | { | |
1521 | struct wait_opts *wo = container_of(wait, struct wait_opts, | |
1522 | child_wait); | |
1523 | struct task_struct *p = key; | |
1524 | ||
5c01ba49 | 1525 | if (!eligible_pid(wo, p)) |
0b7570e7 ON |
1526 | return 0; |
1527 | ||
b4fe5182 ON |
1528 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) |
1529 | return 0; | |
1530 | ||
0b7570e7 ON |
1531 | return default_wake_function(wait, mode, sync, key); |
1532 | } | |
1533 | ||
a7f0765e ON |
1534 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) |
1535 | { | |
0b7570e7 | 1536 | __wake_up_sync_key(&parent->signal->wait_chldexit, |
ce4dd442 | 1537 | TASK_INTERRUPTIBLE, p); |
a7f0765e ON |
1538 | } |
1539 | ||
5449162a JN |
1540 | static bool is_effectively_child(struct wait_opts *wo, bool ptrace, |
1541 | struct task_struct *target) | |
1542 | { | |
1543 | struct task_struct *parent = | |
1544 | !ptrace ? target->real_parent : target->parent; | |
1545 | ||
1546 | return current == parent || (!(wo->wo_flags & __WNOTHREAD) && | |
1547 | same_thread_group(current, parent)); | |
1548 | } | |
1549 | ||
1550 | /* | |
1551 | * Optimization for waiting on PIDTYPE_PID. No need to iterate through child | |
1552 | * and tracee lists to find the target task. | |
1553 | */ | |
1554 | static int do_wait_pid(struct wait_opts *wo) | |
1555 | { | |
1556 | bool ptrace; | |
1557 | struct task_struct *target; | |
1558 | int retval; | |
1559 | ||
1560 | ptrace = false; | |
1561 | target = pid_task(wo->wo_pid, PIDTYPE_TGID); | |
1562 | if (target && is_effectively_child(wo, ptrace, target)) { | |
1563 | retval = wait_consider_task(wo, ptrace, target); | |
1564 | if (retval) | |
1565 | return retval; | |
1566 | } | |
1567 | ||
1568 | ptrace = true; | |
1569 | target = pid_task(wo->wo_pid, PIDTYPE_PID); | |
1570 | if (target && target->ptrace && | |
1571 | is_effectively_child(wo, ptrace, target)) { | |
1572 | retval = wait_consider_task(wo, ptrace, target); | |
1573 | if (retval) | |
1574 | return retval; | |
1575 | } | |
1576 | ||
1577 | return 0; | |
1578 | } | |
1579 | ||
9e8ae01d | 1580 | static long do_wait(struct wait_opts *wo) |
1da177e4 | 1581 | { |
98abed02 | 1582 | int retval; |
1da177e4 | 1583 | |
9e8ae01d | 1584 | trace_sched_process_wait(wo->wo_pid); |
0a16b607 | 1585 | |
0b7570e7 ON |
1586 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); |
1587 | wo->child_wait.private = current; | |
1588 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); | |
1da177e4 | 1589 | repeat: |
98abed02 | 1590 | /* |
3da56d16 | 1591 | * If there is nothing that can match our criteria, just get out. |
9e8ae01d ON |
1592 | * We will clear ->notask_error to zero if we see any child that |
1593 | * might later match our criteria, even if we are not able to reap | |
1594 | * it yet. | |
98abed02 | 1595 | */ |
64a16caf | 1596 | wo->notask_error = -ECHILD; |
9e8ae01d | 1597 | if ((wo->wo_type < PIDTYPE_MAX) && |
1722c14a | 1598 | (!wo->wo_pid || !pid_has_task(wo->wo_pid, wo->wo_type))) |
64a16caf | 1599 | goto notask; |
161550d7 | 1600 | |
f95d39d1 | 1601 | set_current_state(TASK_INTERRUPTIBLE); |
1da177e4 | 1602 | read_lock(&tasklist_lock); |
9e8ae01d | 1603 | |
5449162a JN |
1604 | if (wo->wo_type == PIDTYPE_PID) { |
1605 | retval = do_wait_pid(wo); | |
64a16caf | 1606 | if (retval) |
98abed02 | 1607 | goto end; |
5449162a JN |
1608 | } else { |
1609 | struct task_struct *tsk = current; | |
1610 | ||
1611 | do { | |
1612 | retval = do_wait_thread(wo, tsk); | |
1613 | if (retval) | |
1614 | goto end; | |
98abed02 | 1615 | |
5449162a JN |
1616 | retval = ptrace_do_wait(wo, tsk); |
1617 | if (retval) | |
1618 | goto end; | |
1619 | ||
1620 | if (wo->wo_flags & __WNOTHREAD) | |
1621 | break; | |
1622 | } while_each_thread(current, tsk); | |
1623 | } | |
1da177e4 | 1624 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1625 | |
64a16caf | 1626 | notask: |
9e8ae01d ON |
1627 | retval = wo->notask_error; |
1628 | if (!retval && !(wo->wo_flags & WNOHANG)) { | |
1da177e4 | 1629 | retval = -ERESTARTSYS; |
98abed02 RM |
1630 | if (!signal_pending(current)) { |
1631 | schedule(); | |
1632 | goto repeat; | |
1633 | } | |
1da177e4 | 1634 | } |
1da177e4 | 1635 | end: |
f95d39d1 | 1636 | __set_current_state(TASK_RUNNING); |
0b7570e7 | 1637 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
1da177e4 LT |
1638 | return retval; |
1639 | } | |
1640 | ||
67d7ddde | 1641 | static long kernel_waitid(int which, pid_t upid, struct waitid_info *infop, |
ce72a16f | 1642 | int options, struct rusage *ru) |
1da177e4 | 1643 | { |
9e8ae01d | 1644 | struct wait_opts wo; |
161550d7 EB |
1645 | struct pid *pid = NULL; |
1646 | enum pid_type type; | |
1da177e4 | 1647 | long ret; |
ba7d25f3 | 1648 | unsigned int f_flags = 0; |
1da177e4 | 1649 | |
91c4e8ea ON |
1650 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED| |
1651 | __WNOTHREAD|__WCLONE|__WALL)) | |
1da177e4 LT |
1652 | return -EINVAL; |
1653 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1654 | return -EINVAL; | |
1655 | ||
1656 | switch (which) { | |
1657 | case P_ALL: | |
161550d7 | 1658 | type = PIDTYPE_MAX; |
1da177e4 LT |
1659 | break; |
1660 | case P_PID: | |
161550d7 EB |
1661 | type = PIDTYPE_PID; |
1662 | if (upid <= 0) | |
1da177e4 | 1663 | return -EINVAL; |
3695eae5 CB |
1664 | |
1665 | pid = find_get_pid(upid); | |
1da177e4 LT |
1666 | break; |
1667 | case P_PGID: | |
161550d7 | 1668 | type = PIDTYPE_PGID; |
821cc7b0 | 1669 | if (upid < 0) |
1da177e4 | 1670 | return -EINVAL; |
3695eae5 | 1671 | |
821cc7b0 EB |
1672 | if (upid) |
1673 | pid = find_get_pid(upid); | |
1674 | else | |
1675 | pid = get_task_pid(current, PIDTYPE_PGID); | |
3695eae5 CB |
1676 | break; |
1677 | case P_PIDFD: | |
1678 | type = PIDTYPE_PID; | |
1679 | if (upid < 0) | |
1da177e4 | 1680 | return -EINVAL; |
3695eae5 | 1681 | |
ba7d25f3 | 1682 | pid = pidfd_get_pid(upid, &f_flags); |
3695eae5 CB |
1683 | if (IS_ERR(pid)) |
1684 | return PTR_ERR(pid); | |
ba7d25f3 | 1685 | |
1da177e4 LT |
1686 | break; |
1687 | default: | |
1688 | return -EINVAL; | |
1689 | } | |
1690 | ||
9e8ae01d ON |
1691 | wo.wo_type = type; |
1692 | wo.wo_pid = pid; | |
1693 | wo.wo_flags = options; | |
1694 | wo.wo_info = infop; | |
9e8ae01d | 1695 | wo.wo_rusage = ru; |
ba7d25f3 CB |
1696 | if (f_flags & O_NONBLOCK) |
1697 | wo.wo_flags |= WNOHANG; | |
1698 | ||
9e8ae01d | 1699 | ret = do_wait(&wo); |
ba7d25f3 CB |
1700 | if (!ret && !(options & WNOHANG) && (f_flags & O_NONBLOCK)) |
1701 | ret = -EAGAIN; | |
dfe16dfa | 1702 | |
161550d7 | 1703 | put_pid(pid); |
1da177e4 LT |
1704 | return ret; |
1705 | } | |
1706 | ||
ce72a16f AV |
1707 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
1708 | infop, int, options, struct rusage __user *, ru) | |
1709 | { | |
1710 | struct rusage r; | |
67d7ddde AV |
1711 | struct waitid_info info = {.status = 0}; |
1712 | long err = kernel_waitid(which, upid, &info, options, ru ? &r : NULL); | |
634a8160 | 1713 | int signo = 0; |
6c85501f | 1714 | |
634a8160 AV |
1715 | if (err > 0) { |
1716 | signo = SIGCHLD; | |
1717 | err = 0; | |
ce72a16f AV |
1718 | if (ru && copy_to_user(ru, &r, sizeof(struct rusage))) |
1719 | return -EFAULT; | |
1720 | } | |
67d7ddde AV |
1721 | if (!infop) |
1722 | return err; | |
1723 | ||
41cd7805 | 1724 | if (!user_write_access_begin(infop, sizeof(*infop))) |
1c9fec47 | 1725 | return -EFAULT; |
96ca579a | 1726 | |
634a8160 | 1727 | unsafe_put_user(signo, &infop->si_signo, Efault); |
4c48abe9 | 1728 | unsafe_put_user(0, &infop->si_errno, Efault); |
cc731525 | 1729 | unsafe_put_user(info.cause, &infop->si_code, Efault); |
4c48abe9 AV |
1730 | unsafe_put_user(info.pid, &infop->si_pid, Efault); |
1731 | unsafe_put_user(info.uid, &infop->si_uid, Efault); | |
1732 | unsafe_put_user(info.status, &infop->si_status, Efault); | |
41cd7805 | 1733 | user_write_access_end(); |
ce72a16f | 1734 | return err; |
4c48abe9 | 1735 | Efault: |
41cd7805 | 1736 | user_write_access_end(); |
4c48abe9 | 1737 | return -EFAULT; |
ce72a16f AV |
1738 | } |
1739 | ||
92ebce5a AV |
1740 | long kernel_wait4(pid_t upid, int __user *stat_addr, int options, |
1741 | struct rusage *ru) | |
1da177e4 | 1742 | { |
9e8ae01d | 1743 | struct wait_opts wo; |
161550d7 EB |
1744 | struct pid *pid = NULL; |
1745 | enum pid_type type; | |
1da177e4 LT |
1746 | long ret; |
1747 | ||
1748 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1749 | __WNOTHREAD|__WCLONE|__WALL)) | |
1750 | return -EINVAL; | |
161550d7 | 1751 | |
dd83c161 | 1752 | /* -INT_MIN is not defined */ |
1753 | if (upid == INT_MIN) | |
1754 | return -ESRCH; | |
1755 | ||
161550d7 EB |
1756 | if (upid == -1) |
1757 | type = PIDTYPE_MAX; | |
1758 | else if (upid < 0) { | |
1759 | type = PIDTYPE_PGID; | |
1760 | pid = find_get_pid(-upid); | |
1761 | } else if (upid == 0) { | |
1762 | type = PIDTYPE_PGID; | |
2ae448ef | 1763 | pid = get_task_pid(current, PIDTYPE_PGID); |
161550d7 EB |
1764 | } else /* upid > 0 */ { |
1765 | type = PIDTYPE_PID; | |
1766 | pid = find_get_pid(upid); | |
1767 | } | |
1768 | ||
9e8ae01d ON |
1769 | wo.wo_type = type; |
1770 | wo.wo_pid = pid; | |
1771 | wo.wo_flags = options | WEXITED; | |
1772 | wo.wo_info = NULL; | |
359566fa | 1773 | wo.wo_stat = 0; |
9e8ae01d ON |
1774 | wo.wo_rusage = ru; |
1775 | ret = do_wait(&wo); | |
161550d7 | 1776 | put_pid(pid); |
359566fa AV |
1777 | if (ret > 0 && stat_addr && put_user(wo.wo_stat, stat_addr)) |
1778 | ret = -EFAULT; | |
1da177e4 | 1779 | |
1da177e4 LT |
1780 | return ret; |
1781 | } | |
1782 | ||
8043fc14 CH |
1783 | int kernel_wait(pid_t pid, int *stat) |
1784 | { | |
1785 | struct wait_opts wo = { | |
1786 | .wo_type = PIDTYPE_PID, | |
1787 | .wo_pid = find_get_pid(pid), | |
1788 | .wo_flags = WEXITED, | |
1789 | }; | |
1790 | int ret; | |
1791 | ||
1792 | ret = do_wait(&wo); | |
1793 | if (ret > 0 && wo.wo_stat) | |
1794 | *stat = wo.wo_stat; | |
1795 | put_pid(wo.wo_pid); | |
1796 | return ret; | |
1797 | } | |
1798 | ||
ce72a16f AV |
1799 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
1800 | int, options, struct rusage __user *, ru) | |
1801 | { | |
1802 | struct rusage r; | |
1803 | long err = kernel_wait4(upid, stat_addr, options, ru ? &r : NULL); | |
1804 | ||
1805 | if (err > 0) { | |
1806 | if (ru && copy_to_user(ru, &r, sizeof(struct rusage))) | |
1807 | return -EFAULT; | |
1808 | } | |
1809 | return err; | |
1810 | } | |
1811 | ||
1da177e4 LT |
1812 | #ifdef __ARCH_WANT_SYS_WAITPID |
1813 | ||
1814 | /* | |
1815 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1816 | * implemented by calling sys_wait4() from libc.a. | |
1817 | */ | |
17da2bd9 | 1818 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
1da177e4 | 1819 | { |
d300b610 | 1820 | return kernel_wait4(pid, stat_addr, options, NULL); |
1da177e4 LT |
1821 | } |
1822 | ||
1823 | #endif | |
7e95a225 AV |
1824 | |
1825 | #ifdef CONFIG_COMPAT | |
1826 | COMPAT_SYSCALL_DEFINE4(wait4, | |
1827 | compat_pid_t, pid, | |
1828 | compat_uint_t __user *, stat_addr, | |
1829 | int, options, | |
1830 | struct compat_rusage __user *, ru) | |
1831 | { | |
ce72a16f AV |
1832 | struct rusage r; |
1833 | long err = kernel_wait4(pid, stat_addr, options, ru ? &r : NULL); | |
1834 | if (err > 0) { | |
1835 | if (ru && put_compat_rusage(&r, ru)) | |
1836 | return -EFAULT; | |
7e95a225 | 1837 | } |
ce72a16f | 1838 | return err; |
7e95a225 AV |
1839 | } |
1840 | ||
1841 | COMPAT_SYSCALL_DEFINE5(waitid, | |
1842 | int, which, compat_pid_t, pid, | |
1843 | struct compat_siginfo __user *, infop, int, options, | |
1844 | struct compat_rusage __user *, uru) | |
1845 | { | |
7e95a225 | 1846 | struct rusage ru; |
67d7ddde AV |
1847 | struct waitid_info info = {.status = 0}; |
1848 | long err = kernel_waitid(which, pid, &info, options, uru ? &ru : NULL); | |
634a8160 AV |
1849 | int signo = 0; |
1850 | if (err > 0) { | |
1851 | signo = SIGCHLD; | |
1852 | err = 0; | |
6c85501f AV |
1853 | if (uru) { |
1854 | /* kernel_waitid() overwrites everything in ru */ | |
1855 | if (COMPAT_USE_64BIT_TIME) | |
1856 | err = copy_to_user(uru, &ru, sizeof(ru)); | |
1857 | else | |
1858 | err = put_compat_rusage(&ru, uru); | |
1859 | if (err) | |
1860 | return -EFAULT; | |
1861 | } | |
7e95a225 AV |
1862 | } |
1863 | ||
4c48abe9 AV |
1864 | if (!infop) |
1865 | return err; | |
1866 | ||
41cd7805 | 1867 | if (!user_write_access_begin(infop, sizeof(*infop))) |
1c9fec47 | 1868 | return -EFAULT; |
96ca579a | 1869 | |
634a8160 | 1870 | unsafe_put_user(signo, &infop->si_signo, Efault); |
4c48abe9 | 1871 | unsafe_put_user(0, &infop->si_errno, Efault); |
cc731525 | 1872 | unsafe_put_user(info.cause, &infop->si_code, Efault); |
4c48abe9 AV |
1873 | unsafe_put_user(info.pid, &infop->si_pid, Efault); |
1874 | unsafe_put_user(info.uid, &infop->si_uid, Efault); | |
1875 | unsafe_put_user(info.status, &infop->si_status, Efault); | |
41cd7805 | 1876 | user_write_access_end(); |
67d7ddde | 1877 | return err; |
4c48abe9 | 1878 | Efault: |
41cd7805 | 1879 | user_write_access_end(); |
4c48abe9 | 1880 | return -EFAULT; |
7e95a225 AV |
1881 | } |
1882 | #endif | |
7c2c11b2 | 1883 | |
38fd525a EB |
1884 | /** |
1885 | * thread_group_exited - check that a thread group has exited | |
1886 | * @pid: tgid of thread group to be checked. | |
1887 | * | |
1888 | * Test if the thread group represented by tgid has exited (all | |
1889 | * threads are zombies, dead or completely gone). | |
1890 | * | |
1891 | * Return: true if the thread group has exited. false otherwise. | |
1892 | */ | |
1893 | bool thread_group_exited(struct pid *pid) | |
1894 | { | |
1895 | struct task_struct *task; | |
1896 | bool exited; | |
1897 | ||
1898 | rcu_read_lock(); | |
1899 | task = pid_task(pid, PIDTYPE_PID); | |
1900 | exited = !task || | |
1901 | (READ_ONCE(task->exit_state) && thread_group_empty(task)); | |
1902 | rcu_read_unlock(); | |
1903 | ||
1904 | return exited; | |
1905 | } | |
1906 | EXPORT_SYMBOL(thread_group_exited); | |
1907 | ||
c27cd083 MR |
1908 | /* |
1909 | * This needs to be __function_aligned as GCC implicitly makes any | |
1910 | * implementation of abort() cold and drops alignment specified by | |
1911 | * -falign-functions=N. | |
1912 | * | |
1913 | * See https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88345#c11 | |
1914 | */ | |
1915 | __weak __function_aligned void abort(void) | |
7c2c11b2 SM |
1916 | { |
1917 | BUG(); | |
1918 | ||
1919 | /* if that doesn't kill us, halt */ | |
1920 | panic("Oops failed to kill thread"); | |
1921 | } | |
dc8635b7 | 1922 | EXPORT_SYMBOL(abort); |