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> |
fd593511 | 71 | #include <linux/user_events.h> |
1da177e4 | 72 | |
7c0f6ba6 | 73 | #include <linux/uaccess.h> |
1da177e4 | 74 | #include <asm/unistd.h> |
1da177e4 LT |
75 | #include <asm/mmu_context.h> |
76 | ||
d4ccd54d JH |
77 | /* |
78 | * The default value should be high enough to not crash a system that randomly | |
79 | * crashes its kernel from time to time, but low enough to at least not permit | |
80 | * overflowing 32-bit refcounts or the ldsem writer count. | |
81 | */ | |
82 | static unsigned int oops_limit = 10000; | |
83 | ||
84 | #ifdef CONFIG_SYSCTL | |
85 | static struct ctl_table kern_exit_table[] = { | |
86 | { | |
87 | .procname = "oops_limit", | |
88 | .data = &oops_limit, | |
89 | .maxlen = sizeof(oops_limit), | |
90 | .mode = 0644, | |
91 | .proc_handler = proc_douintvec, | |
92 | }, | |
93 | { } | |
94 | }; | |
95 | ||
96 | static __init int kernel_exit_sysctls_init(void) | |
97 | { | |
98 | register_sysctl_init("kernel", kern_exit_table); | |
99 | return 0; | |
100 | } | |
101 | late_initcall(kernel_exit_sysctls_init); | |
102 | #endif | |
103 | ||
9db89b41 KC |
104 | static atomic_t oops_count = ATOMIC_INIT(0); |
105 | ||
106 | #ifdef CONFIG_SYSFS | |
107 | static ssize_t oops_count_show(struct kobject *kobj, struct kobj_attribute *attr, | |
108 | char *page) | |
109 | { | |
110 | return sysfs_emit(page, "%d\n", atomic_read(&oops_count)); | |
111 | } | |
112 | ||
113 | static struct kobj_attribute oops_count_attr = __ATTR_RO(oops_count); | |
114 | ||
115 | static __init int kernel_exit_sysfs_init(void) | |
116 | { | |
117 | sysfs_add_file_to_group(kernel_kobj, &oops_count_attr.attr, NULL); | |
118 | return 0; | |
119 | } | |
120 | late_initcall(kernel_exit_sysfs_init); | |
121 | #endif | |
122 | ||
d40e48e0 | 123 | static void __unhash_process(struct task_struct *p, bool group_dead) |
1da177e4 LT |
124 | { |
125 | nr_threads--; | |
50d75f8d | 126 | detach_pid(p, PIDTYPE_PID); |
d40e48e0 | 127 | if (group_dead) { |
6883f81a | 128 | detach_pid(p, PIDTYPE_TGID); |
1da177e4 LT |
129 | detach_pid(p, PIDTYPE_PGID); |
130 | detach_pid(p, PIDTYPE_SID); | |
c97d9893 | 131 | |
5e85d4ab | 132 | list_del_rcu(&p->tasks); |
9cd80bbb | 133 | list_del_init(&p->sibling); |
909ea964 | 134 | __this_cpu_dec(process_counts); |
1da177e4 | 135 | } |
47e65328 | 136 | list_del_rcu(&p->thread_group); |
0c740d0a | 137 | list_del_rcu(&p->thread_node); |
1da177e4 LT |
138 | } |
139 | ||
6a14c5c9 ON |
140 | /* |
141 | * This function expects the tasklist_lock write-locked. | |
142 | */ | |
143 | static void __exit_signal(struct task_struct *tsk) | |
144 | { | |
145 | struct signal_struct *sig = tsk->signal; | |
d40e48e0 | 146 | bool group_dead = thread_group_leader(tsk); |
6a14c5c9 | 147 | struct sighand_struct *sighand; |
3f649ab7 | 148 | struct tty_struct *tty; |
5613fda9 | 149 | u64 utime, stime; |
6a14c5c9 | 150 | |
d11c563d | 151 | sighand = rcu_dereference_check(tsk->sighand, |
db1466b3 | 152 | lockdep_tasklist_lock_is_held()); |
6a14c5c9 ON |
153 | spin_lock(&sighand->siglock); |
154 | ||
baa73d9e | 155 | #ifdef CONFIG_POSIX_TIMERS |
6a14c5c9 | 156 | posix_cpu_timers_exit(tsk); |
b95e31c0 | 157 | if (group_dead) |
6a14c5c9 | 158 | posix_cpu_timers_exit_group(tsk); |
baa73d9e | 159 | #endif |
e0a70217 | 160 | |
baa73d9e NP |
161 | if (group_dead) { |
162 | tty = sig->tty; | |
163 | sig->tty = NULL; | |
164 | } else { | |
6a14c5c9 ON |
165 | /* |
166 | * If there is any task waiting for the group exit | |
167 | * then notify it: | |
168 | */ | |
d344193a | 169 | if (sig->notify_count > 0 && !--sig->notify_count) |
60700e38 | 170 | wake_up_process(sig->group_exec_task); |
6db840fa | 171 | |
6a14c5c9 ON |
172 | if (tsk == sig->curr_target) |
173 | sig->curr_target = next_thread(tsk); | |
6a14c5c9 ON |
174 | } |
175 | ||
53d3eaa3 NP |
176 | add_device_randomness((const void*) &tsk->se.sum_exec_runtime, |
177 | sizeof(unsigned long long)); | |
178 | ||
90ed9cbe | 179 | /* |
26e75b5c ON |
180 | * Accumulate here the counters for all threads as they die. We could |
181 | * skip the group leader because it is the last user of signal_struct, | |
182 | * but we want to avoid the race with thread_group_cputime() which can | |
183 | * see the empty ->thread_head list. | |
90ed9cbe RR |
184 | */ |
185 | task_cputime(tsk, &utime, &stime); | |
e78c3496 | 186 | write_seqlock(&sig->stats_lock); |
90ed9cbe RR |
187 | sig->utime += utime; |
188 | sig->stime += stime; | |
189 | sig->gtime += task_gtime(tsk); | |
190 | sig->min_flt += tsk->min_flt; | |
191 | sig->maj_flt += tsk->maj_flt; | |
192 | sig->nvcsw += tsk->nvcsw; | |
193 | sig->nivcsw += tsk->nivcsw; | |
194 | sig->inblock += task_io_get_inblock(tsk); | |
195 | sig->oublock += task_io_get_oublock(tsk); | |
196 | task_io_accounting_add(&sig->ioac, &tsk->ioac); | |
197 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; | |
b3ac022c | 198 | sig->nr_threads--; |
d40e48e0 | 199 | __unhash_process(tsk, group_dead); |
e78c3496 | 200 | write_sequnlock(&sig->stats_lock); |
5876700c | 201 | |
da7978b0 ON |
202 | /* |
203 | * Do this under ->siglock, we can race with another thread | |
204 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. | |
205 | */ | |
206 | flush_sigqueue(&tsk->pending); | |
a7e5328a | 207 | tsk->sighand = NULL; |
6a14c5c9 | 208 | spin_unlock(&sighand->siglock); |
6a14c5c9 | 209 | |
a7e5328a | 210 | __cleanup_sighand(sighand); |
a0be55de | 211 | clear_tsk_thread_flag(tsk, TIF_SIGPENDING); |
d40e48e0 | 212 | if (group_dead) { |
6a14c5c9 | 213 | flush_sigqueue(&sig->shared_pending); |
4ada856f | 214 | tty_kref_put(tty); |
6a14c5c9 ON |
215 | } |
216 | } | |
217 | ||
8c7904a0 EB |
218 | static void delayed_put_task_struct(struct rcu_head *rhp) |
219 | { | |
0a16b607 MD |
220 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
221 | ||
670721c7 | 222 | kprobe_flush_task(tsk); |
54ecbe6f | 223 | rethook_flush_task(tsk); |
4e231c79 | 224 | perf_event_delayed_put(tsk); |
0a16b607 MD |
225 | trace_sched_process_free(tsk); |
226 | put_task_struct(tsk); | |
8c7904a0 EB |
227 | } |
228 | ||
3fbd7ee2 EB |
229 | void put_task_struct_rcu_user(struct task_struct *task) |
230 | { | |
231 | if (refcount_dec_and_test(&task->rcu_users)) | |
232 | call_rcu(&task->rcu, delayed_put_task_struct); | |
233 | } | |
f470021a | 234 | |
2be9880d KW |
235 | void __weak release_thread(struct task_struct *dead_task) |
236 | { | |
237 | } | |
238 | ||
a0be55de | 239 | void release_task(struct task_struct *p) |
1da177e4 | 240 | { |
36c8b586 | 241 | struct task_struct *leader; |
7bc3e6e5 | 242 | struct pid *thread_pid; |
1da177e4 | 243 | int zap_leader; |
1f09f974 | 244 | repeat: |
c69e8d9c | 245 | /* don't need to get the RCU readlock here - the process is dead and |
d11c563d PM |
246 | * can't be modifying its own credentials. But shut RCU-lockdep up */ |
247 | rcu_read_lock(); | |
21d1c5e3 | 248 | dec_rlimit_ucounts(task_ucounts(p), UCOUNT_RLIMIT_NPROC, 1); |
d11c563d | 249 | rcu_read_unlock(); |
c69e8d9c | 250 | |
6b115bf5 | 251 | cgroup_release(p); |
0203026b | 252 | |
1da177e4 | 253 | write_lock_irq(&tasklist_lock); |
a288eecc | 254 | ptrace_release_task(p); |
7bc3e6e5 | 255 | thread_pid = get_pid(p->thread_pid); |
1da177e4 | 256 | __exit_signal(p); |
35f5cad8 | 257 | |
1da177e4 LT |
258 | /* |
259 | * If we are the last non-leader member of the thread | |
260 | * group, and the leader is zombie, then notify the | |
261 | * group leader's parent process. (if it wants notification.) | |
262 | */ | |
263 | zap_leader = 0; | |
264 | leader = p->group_leader; | |
a0be55de IA |
265 | if (leader != p && thread_group_empty(leader) |
266 | && leader->exit_state == EXIT_ZOMBIE) { | |
1da177e4 LT |
267 | /* |
268 | * If we were the last child thread and the leader has | |
269 | * exited already, and the leader's parent ignores SIGCHLD, | |
270 | * then we are the one who should release the leader. | |
dae33574 | 271 | */ |
86773473 | 272 | zap_leader = do_notify_parent(leader, leader->exit_signal); |
dae33574 RM |
273 | if (zap_leader) |
274 | leader->exit_state = EXIT_DEAD; | |
1da177e4 LT |
275 | } |
276 | ||
1da177e4 | 277 | write_unlock_irq(&tasklist_lock); |
3a15fb6e | 278 | seccomp_filter_release(p); |
7bc3e6e5 | 279 | proc_flush_pid(thread_pid); |
6ade99ec | 280 | put_pid(thread_pid); |
1da177e4 | 281 | release_thread(p); |
3fbd7ee2 | 282 | put_task_struct_rcu_user(p); |
1da177e4 LT |
283 | |
284 | p = leader; | |
285 | if (unlikely(zap_leader)) | |
286 | goto repeat; | |
287 | } | |
288 | ||
9d9a6ebf | 289 | int rcuwait_wake_up(struct rcuwait *w) |
8f95c90c | 290 | { |
9d9a6ebf | 291 | int ret = 0; |
8f95c90c DB |
292 | struct task_struct *task; |
293 | ||
294 | rcu_read_lock(); | |
295 | ||
296 | /* | |
297 | * Order condition vs @task, such that everything prior to the load | |
298 | * of @task is visible. This is the condition as to why the user called | |
c9d64a1b | 299 | * rcuwait_wake() in the first place. Pairs with set_current_state() |
8f95c90c DB |
300 | * barrier (A) in rcuwait_wait_event(). |
301 | * | |
302 | * WAIT WAKE | |
303 | * [S] tsk = current [S] cond = true | |
304 | * MB (A) MB (B) | |
305 | * [L] cond [L] tsk | |
306 | */ | |
6dc080ee | 307 | smp_mb(); /* (B) */ |
8f95c90c | 308 | |
8f95c90c DB |
309 | task = rcu_dereference(w->task); |
310 | if (task) | |
9d9a6ebf | 311 | ret = wake_up_process(task); |
8f95c90c | 312 | rcu_read_unlock(); |
9d9a6ebf DB |
313 | |
314 | return ret; | |
8f95c90c | 315 | } |
ac8dec42 | 316 | EXPORT_SYMBOL_GPL(rcuwait_wake_up); |
8f95c90c | 317 | |
1da177e4 LT |
318 | /* |
319 | * Determine if a process group is "orphaned", according to the POSIX | |
320 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
321 | * by terminal-generated stop signals. Newly orphaned process groups are | |
322 | * to receive a SIGHUP and a SIGCONT. | |
323 | * | |
324 | * "I ask you, have you ever known what it is to be an orphan?" | |
325 | */ | |
a0be55de IA |
326 | static int will_become_orphaned_pgrp(struct pid *pgrp, |
327 | struct task_struct *ignored_task) | |
1da177e4 LT |
328 | { |
329 | struct task_struct *p; | |
1da177e4 | 330 | |
0475ac08 | 331 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
332 | if ((p == ignored_task) || |
333 | (p->exit_state && thread_group_empty(p)) || | |
334 | is_global_init(p->real_parent)) | |
1da177e4 | 335 | continue; |
05e83df6 | 336 | |
0475ac08 | 337 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
338 | task_session(p->real_parent) == task_session(p)) |
339 | return 0; | |
0475ac08 | 340 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
341 | |
342 | return 1; | |
1da177e4 LT |
343 | } |
344 | ||
3e7cd6c4 | 345 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
346 | { |
347 | int retval; | |
348 | ||
349 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 350 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
351 | read_unlock(&tasklist_lock); |
352 | ||
353 | return retval; | |
354 | } | |
355 | ||
961c4675 | 356 | static bool has_stopped_jobs(struct pid *pgrp) |
1da177e4 | 357 | { |
1da177e4 LT |
358 | struct task_struct *p; |
359 | ||
0475ac08 | 360 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
961c4675 ON |
361 | if (p->signal->flags & SIGNAL_STOP_STOPPED) |
362 | return true; | |
0475ac08 | 363 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
961c4675 ON |
364 | |
365 | return false; | |
1da177e4 LT |
366 | } |
367 | ||
f49ee505 ON |
368 | /* |
369 | * Check to see if any process groups have become orphaned as | |
370 | * a result of our exiting, and if they have any stopped jobs, | |
371 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
372 | */ | |
373 | static void | |
374 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
375 | { | |
376 | struct pid *pgrp = task_pgrp(tsk); | |
377 | struct task_struct *ignored_task = tsk; | |
378 | ||
379 | if (!parent) | |
a0be55de IA |
380 | /* exit: our father is in a different pgrp than |
381 | * we are and we were the only connection outside. | |
382 | */ | |
f49ee505 ON |
383 | parent = tsk->real_parent; |
384 | else | |
385 | /* reparent: our child is in a different pgrp than | |
386 | * we are, and it was the only connection outside. | |
387 | */ | |
388 | ignored_task = NULL; | |
389 | ||
390 | if (task_pgrp(parent) != pgrp && | |
391 | task_session(parent) == task_session(tsk) && | |
392 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
393 | has_stopped_jobs(pgrp)) { | |
394 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
395 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
396 | } | |
397 | } | |
398 | ||
92307383 | 399 | static void coredump_task_exit(struct task_struct *tsk) |
d67e03e3 EB |
400 | { |
401 | struct core_state *core_state; | |
402 | ||
403 | /* | |
404 | * Serialize with any possible pending coredump. | |
0258b5fd | 405 | * We must hold siglock around checking core_state |
92307383 | 406 | * and setting PF_POSTCOREDUMP. The core-inducing thread |
d67e03e3 | 407 | * will increment ->nr_threads for each thread in the |
92307383 | 408 | * group without PF_POSTCOREDUMP set. |
d67e03e3 | 409 | */ |
0258b5fd | 410 | spin_lock_irq(&tsk->sighand->siglock); |
92307383 | 411 | tsk->flags |= PF_POSTCOREDUMP; |
0258b5fd EB |
412 | core_state = tsk->signal->core_state; |
413 | spin_unlock_irq(&tsk->sighand->siglock); | |
f9010dbd MC |
414 | |
415 | /* The vhost_worker does not particpate in coredumps */ | |
416 | if (core_state && | |
417 | ((tsk->flags & (PF_IO_WORKER | PF_USER_WORKER)) != PF_USER_WORKER)) { | |
d67e03e3 EB |
418 | struct core_thread self; |
419 | ||
d67e03e3 EB |
420 | self.task = current; |
421 | if (self.task->flags & PF_SIGNALED) | |
422 | self.next = xchg(&core_state->dumper.next, &self); | |
423 | else | |
424 | self.task = NULL; | |
425 | /* | |
426 | * Implies mb(), the result of xchg() must be visible | |
427 | * to core_state->dumper. | |
428 | */ | |
429 | if (atomic_dec_and_test(&core_state->nr_threads)) | |
430 | complete(&core_state->startup); | |
431 | ||
432 | for (;;) { | |
f5d39b02 | 433 | set_current_state(TASK_UNINTERRUPTIBLE|TASK_FREEZABLE); |
d67e03e3 EB |
434 | if (!self.task) /* see coredump_finish() */ |
435 | break; | |
f5d39b02 | 436 | schedule(); |
d67e03e3 EB |
437 | } |
438 | __set_current_state(TASK_RUNNING); | |
d67e03e3 EB |
439 | } |
440 | } | |
441 | ||
f98bafa0 | 442 | #ifdef CONFIG_MEMCG |
cf475ad2 | 443 | /* |
733eda7a | 444 | * A task is exiting. If it owned this mm, find a new owner for the mm. |
cf475ad2 | 445 | */ |
cf475ad2 BS |
446 | void mm_update_next_owner(struct mm_struct *mm) |
447 | { | |
448 | struct task_struct *c, *g, *p = current; | |
449 | ||
450 | retry: | |
733eda7a KH |
451 | /* |
452 | * If the exiting or execing task is not the owner, it's | |
453 | * someone else's problem. | |
454 | */ | |
455 | if (mm->owner != p) | |
cf475ad2 | 456 | return; |
733eda7a KH |
457 | /* |
458 | * The current owner is exiting/execing and there are no other | |
459 | * candidates. Do not leave the mm pointing to a possibly | |
460 | * freed task structure. | |
461 | */ | |
462 | if (atomic_read(&mm->mm_users) <= 1) { | |
987717e5 | 463 | WRITE_ONCE(mm->owner, NULL); |
733eda7a KH |
464 | return; |
465 | } | |
cf475ad2 BS |
466 | |
467 | read_lock(&tasklist_lock); | |
468 | /* | |
469 | * Search in the children | |
470 | */ | |
471 | list_for_each_entry(c, &p->children, sibling) { | |
472 | if (c->mm == mm) | |
473 | goto assign_new_owner; | |
474 | } | |
475 | ||
476 | /* | |
477 | * Search in the siblings | |
478 | */ | |
dea33cfd | 479 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
cf475ad2 BS |
480 | if (c->mm == mm) |
481 | goto assign_new_owner; | |
482 | } | |
483 | ||
484 | /* | |
f87fb599 | 485 | * Search through everything else, we should not get here often. |
cf475ad2 | 486 | */ |
39af1765 ON |
487 | for_each_process(g) { |
488 | if (g->flags & PF_KTHREAD) | |
489 | continue; | |
490 | for_each_thread(g, c) { | |
491 | if (c->mm == mm) | |
492 | goto assign_new_owner; | |
493 | if (c->mm) | |
494 | break; | |
495 | } | |
f87fb599 | 496 | } |
cf475ad2 | 497 | read_unlock(&tasklist_lock); |
31a78f23 BS |
498 | /* |
499 | * We found no owner yet mm_users > 1: this implies that we are | |
500 | * most likely racing with swapoff (try_to_unuse()) or /proc or | |
e5991371 | 501 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
31a78f23 | 502 | */ |
987717e5 | 503 | WRITE_ONCE(mm->owner, NULL); |
cf475ad2 BS |
504 | return; |
505 | ||
506 | assign_new_owner: | |
507 | BUG_ON(c == p); | |
508 | get_task_struct(c); | |
509 | /* | |
510 | * The task_lock protects c->mm from changing. | |
511 | * We always want mm->owner->mm == mm | |
512 | */ | |
513 | task_lock(c); | |
e5991371 HD |
514 | /* |
515 | * Delay read_unlock() till we have the task_lock() | |
516 | * to ensure that c does not slip away underneath us | |
517 | */ | |
518 | read_unlock(&tasklist_lock); | |
cf475ad2 BS |
519 | if (c->mm != mm) { |
520 | task_unlock(c); | |
521 | put_task_struct(c); | |
522 | goto retry; | |
523 | } | |
987717e5 | 524 | WRITE_ONCE(mm->owner, c); |
bd74fdae | 525 | lru_gen_migrate_mm(mm); |
cf475ad2 BS |
526 | task_unlock(c); |
527 | put_task_struct(c); | |
528 | } | |
f98bafa0 | 529 | #endif /* CONFIG_MEMCG */ |
cf475ad2 | 530 | |
1da177e4 LT |
531 | /* |
532 | * Turn us into a lazy TLB process if we | |
533 | * aren't already.. | |
534 | */ | |
0039962a | 535 | static void exit_mm(void) |
1da177e4 | 536 | { |
0039962a | 537 | struct mm_struct *mm = current->mm; |
1da177e4 | 538 | |
4610ba7a | 539 | exit_mm_release(current, mm); |
1da177e4 LT |
540 | if (!mm) |
541 | return; | |
4fe7efdb | 542 | sync_mm_rss(mm); |
d8ed45c5 | 543 | mmap_read_lock(mm); |
aa464ba9 | 544 | mmgrab_lazy_tlb(mm); |
0039962a | 545 | BUG_ON(mm != current->active_mm); |
1da177e4 | 546 | /* more a memory barrier than a real lock */ |
0039962a | 547 | task_lock(current); |
5bc78502 MD |
548 | /* |
549 | * When a thread stops operating on an address space, the loop | |
550 | * in membarrier_private_expedited() may not observe that | |
551 | * tsk->mm, and the loop in membarrier_global_expedited() may | |
552 | * not observe a MEMBARRIER_STATE_GLOBAL_EXPEDITED | |
553 | * rq->membarrier_state, so those would not issue an IPI. | |
554 | * Membarrier requires a memory barrier after accessing | |
555 | * user-space memory, before clearing tsk->mm or the | |
556 | * rq->membarrier_state. | |
557 | */ | |
558 | smp_mb__after_spinlock(); | |
559 | local_irq_disable(); | |
0039962a | 560 | current->mm = NULL; |
5bc78502 | 561 | membarrier_update_current_mm(NULL); |
1da177e4 | 562 | enter_lazy_tlb(mm, current); |
5bc78502 | 563 | local_irq_enable(); |
0039962a | 564 | task_unlock(current); |
5bc78502 | 565 | mmap_read_unlock(mm); |
cf475ad2 | 566 | mm_update_next_owner(mm); |
1da177e4 | 567 | mmput(mm); |
c32b3cbe | 568 | if (test_thread_flag(TIF_MEMDIE)) |
38531201 | 569 | exit_oom_victim(); |
1da177e4 LT |
570 | } |
571 | ||
c9dc05bf ON |
572 | static struct task_struct *find_alive_thread(struct task_struct *p) |
573 | { | |
574 | struct task_struct *t; | |
575 | ||
576 | for_each_thread(p, t) { | |
577 | if (!(t->flags & PF_EXITING)) | |
578 | return t; | |
579 | } | |
580 | return NULL; | |
581 | } | |
582 | ||
8fb335e0 AV |
583 | static struct task_struct *find_child_reaper(struct task_struct *father, |
584 | struct list_head *dead) | |
1109909c ON |
585 | __releases(&tasklist_lock) |
586 | __acquires(&tasklist_lock) | |
587 | { | |
588 | struct pid_namespace *pid_ns = task_active_pid_ns(father); | |
589 | struct task_struct *reaper = pid_ns->child_reaper; | |
8fb335e0 | 590 | struct task_struct *p, *n; |
1109909c ON |
591 | |
592 | if (likely(reaper != father)) | |
593 | return reaper; | |
594 | ||
c9dc05bf ON |
595 | reaper = find_alive_thread(father); |
596 | if (reaper) { | |
1109909c ON |
597 | pid_ns->child_reaper = reaper; |
598 | return reaper; | |
599 | } | |
600 | ||
601 | write_unlock_irq(&tasklist_lock); | |
8fb335e0 AV |
602 | |
603 | list_for_each_entry_safe(p, n, dead, ptrace_entry) { | |
604 | list_del_init(&p->ptrace_entry); | |
605 | release_task(p); | |
606 | } | |
607 | ||
1109909c ON |
608 | zap_pid_ns_processes(pid_ns); |
609 | write_lock_irq(&tasklist_lock); | |
610 | ||
611 | return father; | |
612 | } | |
613 | ||
1da177e4 | 614 | /* |
ebec18a6 LP |
615 | * When we die, we re-parent all our children, and try to: |
616 | * 1. give them to another thread in our thread group, if such a member exists | |
617 | * 2. give it to the first ancestor process which prctl'd itself as a | |
618 | * child_subreaper for its children (like a service manager) | |
619 | * 3. give it to the init process (PID 1) in our pid namespace | |
1da177e4 | 620 | */ |
1109909c ON |
621 | static struct task_struct *find_new_reaper(struct task_struct *father, |
622 | struct task_struct *child_reaper) | |
1da177e4 | 623 | { |
c9dc05bf | 624 | struct task_struct *thread, *reaper; |
1da177e4 | 625 | |
c9dc05bf ON |
626 | thread = find_alive_thread(father); |
627 | if (thread) | |
950bbabb | 628 | return thread; |
1da177e4 | 629 | |
7d24e2df | 630 | if (father->signal->has_child_subreaper) { |
c6c70f44 | 631 | unsigned int ns_level = task_pid(father)->level; |
ebec18a6 | 632 | /* |
175aed3f | 633 | * Find the first ->is_child_subreaper ancestor in our pid_ns. |
c6c70f44 ON |
634 | * We can't check reaper != child_reaper to ensure we do not |
635 | * cross the namespaces, the exiting parent could be injected | |
636 | * by setns() + fork(). | |
637 | * We check pid->level, this is slightly more efficient than | |
638 | * task_active_pid_ns(reaper) != task_active_pid_ns(father). | |
ebec18a6 | 639 | */ |
c6c70f44 ON |
640 | for (reaper = father->real_parent; |
641 | task_pid(reaper)->level == ns_level; | |
ebec18a6 | 642 | reaper = reaper->real_parent) { |
175aed3f | 643 | if (reaper == &init_task) |
ebec18a6 LP |
644 | break; |
645 | if (!reaper->signal->is_child_subreaper) | |
646 | continue; | |
c9dc05bf ON |
647 | thread = find_alive_thread(reaper); |
648 | if (thread) | |
649 | return thread; | |
ebec18a6 | 650 | } |
1da177e4 | 651 | } |
762a24be | 652 | |
1109909c | 653 | return child_reaper; |
950bbabb ON |
654 | } |
655 | ||
5dfc80be ON |
656 | /* |
657 | * Any that need to be release_task'd are put on the @dead list. | |
658 | */ | |
9cd80bbb | 659 | static void reparent_leader(struct task_struct *father, struct task_struct *p, |
5dfc80be ON |
660 | struct list_head *dead) |
661 | { | |
2831096e | 662 | if (unlikely(p->exit_state == EXIT_DEAD)) |
5dfc80be ON |
663 | return; |
664 | ||
abd50b39 | 665 | /* We don't want people slaying init. */ |
5dfc80be ON |
666 | p->exit_signal = SIGCHLD; |
667 | ||
668 | /* If it has exited notify the new parent about this child's death. */ | |
d21142ec | 669 | if (!p->ptrace && |
5dfc80be | 670 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
86773473 | 671 | if (do_notify_parent(p, p->exit_signal)) { |
5dfc80be | 672 | p->exit_state = EXIT_DEAD; |
dc2fd4b0 | 673 | list_add(&p->ptrace_entry, dead); |
5dfc80be ON |
674 | } |
675 | } | |
676 | ||
677 | kill_orphaned_pgrp(p, father); | |
678 | } | |
679 | ||
482a3767 ON |
680 | /* |
681 | * This does two things: | |
682 | * | |
683 | * A. Make init inherit all the child processes | |
684 | * B. Check to see if any process groups have become orphaned | |
685 | * as a result of our exiting, and if they have any stopped | |
686 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
687 | */ | |
688 | static void forget_original_parent(struct task_struct *father, | |
689 | struct list_head *dead) | |
1da177e4 | 690 | { |
482a3767 | 691 | struct task_struct *p, *t, *reaper; |
762a24be | 692 | |
7c8bd232 | 693 | if (unlikely(!list_empty(&father->ptraced))) |
482a3767 | 694 | exit_ptrace(father, dead); |
f470021a | 695 | |
7c8bd232 | 696 | /* Can drop and reacquire tasklist_lock */ |
8fb335e0 | 697 | reaper = find_child_reaper(father, dead); |
ad9e206a | 698 | if (list_empty(&father->children)) |
482a3767 | 699 | return; |
1109909c ON |
700 | |
701 | reaper = find_new_reaper(father, reaper); | |
2831096e | 702 | list_for_each_entry(p, &father->children, sibling) { |
57a05918 | 703 | for_each_thread(p, t) { |
22a34c6f MB |
704 | RCU_INIT_POINTER(t->real_parent, reaper); |
705 | BUG_ON((!t->ptrace) != (rcu_access_pointer(t->parent) == father)); | |
57a05918 | 706 | if (likely(!t->ptrace)) |
9cd80bbb | 707 | t->parent = t->real_parent; |
9cd80bbb ON |
708 | if (t->pdeath_signal) |
709 | group_send_sig_info(t->pdeath_signal, | |
01024980 EB |
710 | SEND_SIG_NOINFO, t, |
711 | PIDTYPE_TGID); | |
57a05918 | 712 | } |
2831096e ON |
713 | /* |
714 | * If this is a threaded reparent there is no need to | |
715 | * notify anyone anything has happened. | |
716 | */ | |
717 | if (!same_thread_group(reaper, father)) | |
482a3767 | 718 | reparent_leader(father, p, dead); |
1da177e4 | 719 | } |
2831096e | 720 | list_splice_tail_init(&father->children, &reaper->children); |
1da177e4 LT |
721 | } |
722 | ||
723 | /* | |
724 | * Send signals to all our closest relatives so that they know | |
725 | * to properly mourn us.. | |
726 | */ | |
821c7de7 | 727 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 | 728 | { |
53c8f9f1 | 729 | bool autoreap; |
482a3767 ON |
730 | struct task_struct *p, *n; |
731 | LIST_HEAD(dead); | |
1da177e4 | 732 | |
762a24be | 733 | write_lock_irq(&tasklist_lock); |
482a3767 ON |
734 | forget_original_parent(tsk, &dead); |
735 | ||
821c7de7 ON |
736 | if (group_dead) |
737 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 738 | |
b191d649 | 739 | tsk->exit_state = EXIT_ZOMBIE; |
45cdf5cc ON |
740 | if (unlikely(tsk->ptrace)) { |
741 | int sig = thread_group_leader(tsk) && | |
742 | thread_group_empty(tsk) && | |
743 | !ptrace_reparented(tsk) ? | |
744 | tsk->exit_signal : SIGCHLD; | |
745 | autoreap = do_notify_parent(tsk, sig); | |
746 | } else if (thread_group_leader(tsk)) { | |
747 | autoreap = thread_group_empty(tsk) && | |
748 | do_notify_parent(tsk, tsk->exit_signal); | |
749 | } else { | |
750 | autoreap = true; | |
751 | } | |
1da177e4 | 752 | |
30b692d3 CB |
753 | if (autoreap) { |
754 | tsk->exit_state = EXIT_DEAD; | |
6c66e7db | 755 | list_add(&tsk->ptrace_entry, &dead); |
30b692d3 | 756 | } |
1da177e4 | 757 | |
9c339168 ON |
758 | /* mt-exec, de_thread() is waiting for group leader */ |
759 | if (unlikely(tsk->signal->notify_count < 0)) | |
60700e38 | 760 | wake_up_process(tsk->signal->group_exec_task); |
1da177e4 LT |
761 | write_unlock_irq(&tasklist_lock); |
762 | ||
482a3767 ON |
763 | list_for_each_entry_safe(p, n, &dead, ptrace_entry) { |
764 | list_del_init(&p->ptrace_entry); | |
765 | release_task(p); | |
766 | } | |
1da177e4 LT |
767 | } |
768 | ||
e18eecb8 JD |
769 | #ifdef CONFIG_DEBUG_STACK_USAGE |
770 | static void check_stack_usage(void) | |
771 | { | |
772 | static DEFINE_SPINLOCK(low_water_lock); | |
773 | static int lowest_to_date = THREAD_SIZE; | |
e18eecb8 JD |
774 | unsigned long free; |
775 | ||
7c9f8861 | 776 | free = stack_not_used(current); |
e18eecb8 JD |
777 | |
778 | if (free >= lowest_to_date) | |
779 | return; | |
780 | ||
781 | spin_lock(&low_water_lock); | |
782 | if (free < lowest_to_date) { | |
627393d4 | 783 | pr_info("%s (%d) used greatest stack depth: %lu bytes left\n", |
a0be55de | 784 | current->comm, task_pid_nr(current), free); |
e18eecb8 JD |
785 | lowest_to_date = free; |
786 | } | |
787 | spin_unlock(&low_water_lock); | |
788 | } | |
789 | #else | |
790 | static inline void check_stack_usage(void) {} | |
791 | #endif | |
792 | ||
d80f7d7b EB |
793 | static void synchronize_group_exit(struct task_struct *tsk, long code) |
794 | { | |
795 | struct sighand_struct *sighand = tsk->sighand; | |
796 | struct signal_struct *signal = tsk->signal; | |
797 | ||
798 | spin_lock_irq(&sighand->siglock); | |
799 | signal->quick_threads--; | |
800 | if ((signal->quick_threads == 0) && | |
801 | !(signal->flags & SIGNAL_GROUP_EXIT)) { | |
802 | signal->flags = SIGNAL_GROUP_EXIT; | |
803 | signal->group_exit_code = code; | |
804 | signal->group_stop_count = 0; | |
805 | } | |
806 | spin_unlock_irq(&sighand->siglock); | |
807 | } | |
808 | ||
9af6528e | 809 | void __noreturn do_exit(long code) |
1da177e4 LT |
810 | { |
811 | struct task_struct *tsk = current; | |
812 | int group_dead; | |
813 | ||
001c28e5 NP |
814 | WARN_ON(irqs_disabled()); |
815 | ||
d80f7d7b EB |
816 | synchronize_group_exit(tsk, code); |
817 | ||
b1f866b0 | 818 | WARN_ON(tsk->plug); |
22e2c507 | 819 | |
586b58ca | 820 | kcov_task_exit(tsk); |
50b5e49c | 821 | kmsan_task_exit(tsk); |
586b58ca | 822 | |
92307383 | 823 | coredump_task_exit(tsk); |
a288eecc | 824 | ptrace_event(PTRACE_EVENT_EXIT, code); |
fd593511 | 825 | user_events_exit(tsk); |
1da177e4 | 826 | |
e0e81739 DH |
827 | validate_creds_for_do_exit(tsk); |
828 | ||
f552a27a | 829 | io_uring_files_cancel(); |
d12619b5 | 830 | exit_signals(tsk); /* sets PF_EXITING */ |
1da177e4 | 831 | |
48d212a2 LT |
832 | /* sync mm's RSS info before statistics gathering */ |
833 | if (tsk->mm) | |
834 | sync_mm_rss(tsk->mm); | |
51229b49 | 835 | acct_update_integrals(tsk); |
1da177e4 | 836 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 837 | if (group_dead) { |
43cf75d9 | 838 | /* |
839 | * If the last thread of global init has exited, panic | |
840 | * immediately to get a useable coredump. | |
841 | */ | |
842 | if (unlikely(is_global_init(tsk))) | |
843 | panic("Attempted to kill init! exitcode=0x%08x\n", | |
844 | tsk->signal->group_exit_code ?: (int)code); | |
845 | ||
baa73d9e | 846 | #ifdef CONFIG_POSIX_TIMERS |
778e9a9c | 847 | hrtimer_cancel(&tsk->signal->real_timer); |
d5b36a4d | 848 | exit_itimers(tsk); |
baa73d9e | 849 | #endif |
1f10206c JP |
850 | if (tsk->mm) |
851 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); | |
c3068951 | 852 | } |
f6ec29a4 | 853 | acct_collect(code, group_dead); |
522ed776 MT |
854 | if (group_dead) |
855 | tty_audit_exit(); | |
a4ff8dba | 856 | audit_free(tsk); |
115085ea | 857 | |
48d212a2 | 858 | tsk->exit_code = code; |
115085ea | 859 | taskstats_exit(tsk, group_dead); |
c757249a | 860 | |
0039962a | 861 | exit_mm(); |
1da177e4 | 862 | |
0e464814 | 863 | if (group_dead) |
f6ec29a4 | 864 | acct_process(); |
0a16b607 MD |
865 | trace_sched_process_exit(tsk); |
866 | ||
1da177e4 | 867 | exit_sem(tsk); |
b34a6b1d | 868 | exit_shm(tsk); |
1ec7f1dd AV |
869 | exit_files(tsk); |
870 | exit_fs(tsk); | |
c39df5fa ON |
871 | if (group_dead) |
872 | disassociate_ctty(1); | |
8aac6270 | 873 | exit_task_namespaces(tsk); |
ed3e694d | 874 | exit_task_work(tsk); |
e6464694 | 875 | exit_thread(tsk); |
0b3fcf17 SE |
876 | |
877 | /* | |
878 | * Flush inherited counters to the parent - before the parent | |
879 | * gets woken up by child-exit notifications. | |
880 | * | |
881 | * because of cgroup mode, must be called before cgroup_exit() | |
882 | */ | |
883 | perf_event_exit_task(tsk); | |
884 | ||
8e5bfa8c | 885 | sched_autogroup_exit_task(tsk); |
1ec41830 | 886 | cgroup_exit(tsk); |
1da177e4 | 887 | |
24f1e32c FW |
888 | /* |
889 | * FIXME: do that only when needed, using sched_exit tracepoint | |
890 | */ | |
7c8df286 | 891 | flush_ptrace_hw_breakpoint(tsk); |
33b2fb30 | 892 | |
ccdd29ff | 893 | exit_tasks_rcu_start(); |
821c7de7 | 894 | exit_notify(tsk, group_dead); |
ef982393 | 895 | proc_exit_connector(tsk); |
c11600e4 | 896 | mpol_put_task_policy(tsk); |
42b2dd0a | 897 | #ifdef CONFIG_FUTEX |
c87e2837 IM |
898 | if (unlikely(current->pi_state_cache)) |
899 | kfree(current->pi_state_cache); | |
42b2dd0a | 900 | #endif |
de5097c2 | 901 | /* |
9a11b49a | 902 | * Make sure we are holding no locks: |
de5097c2 | 903 | */ |
1b1d2fb4 | 904 | debug_check_no_locks_held(); |
1da177e4 | 905 | |
afc847b7 | 906 | if (tsk->io_context) |
b69f2292 | 907 | exit_io_context(tsk); |
afc847b7 | 908 | |
b92ce558 | 909 | if (tsk->splice_pipe) |
4b8a8f1e | 910 | free_pipe_info(tsk->splice_pipe); |
b92ce558 | 911 | |
5640f768 ED |
912 | if (tsk->task_frag.page) |
913 | put_page(tsk->task_frag.page); | |
914 | ||
e0e81739 | 915 | validate_creds_for_do_exit(tsk); |
1a03d3f1 | 916 | exit_task_stack_account(tsk); |
e0e81739 | 917 | |
4bcb8232 | 918 | check_stack_usage(); |
7407251a | 919 | preempt_disable(); |
54848d73 WF |
920 | if (tsk->nr_dirtied) |
921 | __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied); | |
f41d911f | 922 | exit_rcu(); |
ccdd29ff | 923 | exit_tasks_rcu_finish(); |
b5740f4b | 924 | |
b09be676 | 925 | lockdep_free_task(tsk); |
9af6528e | 926 | do_task_dead(); |
1da177e4 | 927 | } |
012914da | 928 | |
0e25498f EB |
929 | void __noreturn make_task_dead(int signr) |
930 | { | |
931 | /* | |
932 | * Take the task off the cpu after something catastrophic has | |
933 | * happened. | |
05ea0424 EB |
934 | * |
935 | * We can get here from a kernel oops, sometimes with preemption off. | |
936 | * Start by checking for critical errors. | |
937 | * Then fix up important state like USER_DS and preemption. | |
938 | * Then do everything else. | |
0e25498f | 939 | */ |
05ea0424 | 940 | struct task_struct *tsk = current; |
7535b832 | 941 | unsigned int limit; |
05ea0424 EB |
942 | |
943 | if (unlikely(in_interrupt())) | |
944 | panic("Aiee, killing interrupt handler!"); | |
945 | if (unlikely(!tsk->pid)) | |
946 | panic("Attempted to kill the idle task!"); | |
947 | ||
001c28e5 NP |
948 | if (unlikely(irqs_disabled())) { |
949 | pr_info("note: %s[%d] exited with irqs disabled\n", | |
950 | current->comm, task_pid_nr(current)); | |
951 | local_irq_enable(); | |
952 | } | |
05ea0424 EB |
953 | if (unlikely(in_atomic())) { |
954 | pr_info("note: %s[%d] exited with preempt_count %d\n", | |
955 | current->comm, task_pid_nr(current), | |
956 | preempt_count()); | |
957 | preempt_count_set(PREEMPT_ENABLED); | |
958 | } | |
959 | ||
d4ccd54d JH |
960 | /* |
961 | * Every time the system oopses, if the oops happens while a reference | |
962 | * to an object was held, the reference leaks. | |
963 | * If the oops doesn't also leak memory, repeated oopsing can cause | |
964 | * reference counters to wrap around (if they're not using refcount_t). | |
965 | * This means that repeated oopsing can make unexploitable-looking bugs | |
966 | * exploitable through repeated oopsing. | |
967 | * To make sure this can't happen, place an upper bound on how often the | |
968 | * kernel may oops without panic(). | |
969 | */ | |
7535b832 KC |
970 | limit = READ_ONCE(oops_limit); |
971 | if (atomic_inc_return(&oops_count) >= limit && limit) | |
972 | panic("Oopsed too often (kernel.oops_limit is %d)", limit); | |
d4ccd54d | 973 | |
05ea0424 EB |
974 | /* |
975 | * We're taking recursive faults here in make_task_dead. Safest is to just | |
976 | * leave this task alone and wait for reboot. | |
977 | */ | |
978 | if (unlikely(tsk->flags & PF_EXITING)) { | |
979 | pr_alert("Fixing recursive fault but reboot is needed!\n"); | |
980 | futex_exit_recursive(tsk); | |
912616f1 EB |
981 | tsk->exit_state = EXIT_DEAD; |
982 | refcount_inc(&tsk->rcu_users); | |
7f80a2fd | 983 | do_task_dead(); |
05ea0424 EB |
984 | } |
985 | ||
0e25498f EB |
986 | do_exit(signr); |
987 | } | |
988 | ||
754fe8d2 | 989 | SYSCALL_DEFINE1(exit, int, error_code) |
1da177e4 LT |
990 | { |
991 | do_exit((error_code&0xff)<<8); | |
992 | } | |
993 | ||
1da177e4 LT |
994 | /* |
995 | * Take down every thread in the group. This is called by fatal signals | |
996 | * as well as by sys_exit_group (below). | |
997 | */ | |
eae654f1 | 998 | void __noreturn |
1da177e4 LT |
999 | do_group_exit(int exit_code) |
1000 | { | |
bfc4b089 ON |
1001 | struct signal_struct *sig = current->signal; |
1002 | ||
49697335 | 1003 | if (sig->flags & SIGNAL_GROUP_EXIT) |
bfc4b089 | 1004 | exit_code = sig->group_exit_code; |
49697335 EB |
1005 | else if (sig->group_exec_task) |
1006 | exit_code = 0; | |
cbe9dac3 | 1007 | else { |
1da177e4 | 1008 | struct sighand_struct *const sighand = current->sighand; |
a0be55de | 1009 | |
1da177e4 | 1010 | spin_lock_irq(&sighand->siglock); |
49697335 | 1011 | if (sig->flags & SIGNAL_GROUP_EXIT) |
1da177e4 LT |
1012 | /* Another thread got here before we took the lock. */ |
1013 | exit_code = sig->group_exit_code; | |
49697335 EB |
1014 | else if (sig->group_exec_task) |
1015 | exit_code = 0; | |
1da177e4 | 1016 | else { |
1da177e4 | 1017 | sig->group_exit_code = exit_code; |
ed5d2cac | 1018 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
1019 | zap_other_threads(current); |
1020 | } | |
1021 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
1022 | } |
1023 | ||
1024 | do_exit(exit_code); | |
1025 | /* NOTREACHED */ | |
1026 | } | |
1027 | ||
1028 | /* | |
1029 | * this kills every thread in the thread group. Note that any externally | |
1030 | * wait4()-ing process will get the correct exit code - even if this | |
1031 | * thread is not the thread group leader. | |
1032 | */ | |
754fe8d2 | 1033 | SYSCALL_DEFINE1(exit_group, int, error_code) |
1da177e4 LT |
1034 | { |
1035 | do_group_exit((error_code & 0xff) << 8); | |
2ed7c03e HC |
1036 | /* NOTREACHED */ |
1037 | return 0; | |
1da177e4 LT |
1038 | } |
1039 | ||
67d7ddde AV |
1040 | struct waitid_info { |
1041 | pid_t pid; | |
1042 | uid_t uid; | |
1043 | int status; | |
1044 | int cause; | |
1045 | }; | |
1046 | ||
9e8ae01d ON |
1047 | struct wait_opts { |
1048 | enum pid_type wo_type; | |
9e8ae01d | 1049 | int wo_flags; |
e1eb1ebc | 1050 | struct pid *wo_pid; |
9e8ae01d | 1051 | |
67d7ddde | 1052 | struct waitid_info *wo_info; |
359566fa | 1053 | int wo_stat; |
ce72a16f | 1054 | struct rusage *wo_rusage; |
9e8ae01d | 1055 | |
ac6424b9 | 1056 | wait_queue_entry_t child_wait; |
9e8ae01d ON |
1057 | int notask_error; |
1058 | }; | |
1059 | ||
989264f4 | 1060 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1061 | { |
5c01ba49 ON |
1062 | return wo->wo_type == PIDTYPE_MAX || |
1063 | task_pid_type(p, wo->wo_type) == wo->wo_pid; | |
1064 | } | |
1da177e4 | 1065 | |
bf959931 ON |
1066 | static int |
1067 | eligible_child(struct wait_opts *wo, bool ptrace, struct task_struct *p) | |
5c01ba49 ON |
1068 | { |
1069 | if (!eligible_pid(wo, p)) | |
1070 | return 0; | |
bf959931 ON |
1071 | |
1072 | /* | |
1073 | * Wait for all children (clone and not) if __WALL is set or | |
1074 | * if it is traced by us. | |
1075 | */ | |
1076 | if (ptrace || (wo->wo_flags & __WALL)) | |
1077 | return 1; | |
1078 | ||
1079 | /* | |
1080 | * Otherwise, wait for clone children *only* if __WCLONE is set; | |
1081 | * otherwise, wait for non-clone children *only*. | |
1082 | * | |
1083 | * Note: a "clone" child here is one that reports to its parent | |
1084 | * using a signal other than SIGCHLD, or a non-leader thread which | |
1085 | * we can only see if it is traced by us. | |
1086 | */ | |
1087 | if ((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) | |
1da177e4 | 1088 | return 0; |
1da177e4 | 1089 | |
14dd0b81 | 1090 | return 1; |
1da177e4 LT |
1091 | } |
1092 | ||
1da177e4 LT |
1093 | /* |
1094 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
1095 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1096 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1097 | * released the lock and the system call should return. | |
1098 | */ | |
9e8ae01d | 1099 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1100 | { |
67d7ddde | 1101 | int state, status; |
6c5f3e7b | 1102 | pid_t pid = task_pid_vnr(p); |
43e13cc1 | 1103 | uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
67d7ddde | 1104 | struct waitid_info *infop; |
1da177e4 | 1105 | |
9e8ae01d | 1106 | if (!likely(wo->wo_flags & WEXITED)) |
98abed02 RM |
1107 | return 0; |
1108 | ||
9e8ae01d | 1109 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
907c311f EB |
1110 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1111 | ? p->signal->group_exit_code : p->exit_code; | |
1da177e4 LT |
1112 | get_task_struct(p); |
1113 | read_unlock(&tasklist_lock); | |
1029a2b5 | 1114 | sched_annotate_sleep(); |
e61a2502 AV |
1115 | if (wo->wo_rusage) |
1116 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1117 | put_task_struct(p); |
76d9871e | 1118 | goto out_info; |
1da177e4 | 1119 | } |
1da177e4 | 1120 | /* |
abd50b39 | 1121 | * Move the task's state to DEAD/TRACE, only one thread can do this. |
1da177e4 | 1122 | */ |
f6507f83 ON |
1123 | state = (ptrace_reparented(p) && thread_group_leader(p)) ? |
1124 | EXIT_TRACE : EXIT_DEAD; | |
abd50b39 | 1125 | if (cmpxchg(&p->exit_state, EXIT_ZOMBIE, state) != EXIT_ZOMBIE) |
1da177e4 | 1126 | return 0; |
986094df ON |
1127 | /* |
1128 | * We own this thread, nobody else can reap it. | |
1129 | */ | |
1130 | read_unlock(&tasklist_lock); | |
1131 | sched_annotate_sleep(); | |
f6507f83 | 1132 | |
befca967 | 1133 | /* |
f6507f83 | 1134 | * Check thread_group_leader() to exclude the traced sub-threads. |
befca967 | 1135 | */ |
f6507f83 | 1136 | if (state == EXIT_DEAD && thread_group_leader(p)) { |
f953ccd0 ON |
1137 | struct signal_struct *sig = p->signal; |
1138 | struct signal_struct *psig = current->signal; | |
1f10206c | 1139 | unsigned long maxrss; |
5613fda9 | 1140 | u64 tgutime, tgstime; |
3795e161 | 1141 | |
1da177e4 LT |
1142 | /* |
1143 | * The resource counters for the group leader are in its | |
1144 | * own task_struct. Those for dead threads in the group | |
1145 | * are in its signal_struct, as are those for the child | |
1146 | * processes it has previously reaped. All these | |
1147 | * accumulate in the parent's signal_struct c* fields. | |
1148 | * | |
1149 | * We don't bother to take a lock here to protect these | |
f953ccd0 ON |
1150 | * p->signal fields because the whole thread group is dead |
1151 | * and nobody can change them. | |
1152 | * | |
dcca3475 | 1153 | * psig->stats_lock also protects us from our sub-threads |
f953ccd0 ON |
1154 | * which can reap other children at the same time. Until |
1155 | * we change k_getrusage()-like users to rely on this lock | |
1156 | * we have to take ->siglock as well. | |
0cf55e1e | 1157 | * |
a0be55de IA |
1158 | * We use thread_group_cputime_adjusted() to get times for |
1159 | * the thread group, which consolidates times for all threads | |
1160 | * in the group including the group leader. | |
1da177e4 | 1161 | */ |
e80d0a1a | 1162 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); |
f953ccd0 | 1163 | spin_lock_irq(¤t->sighand->siglock); |
e78c3496 | 1164 | write_seqlock(&psig->stats_lock); |
64861634 MS |
1165 | psig->cutime += tgutime + sig->cutime; |
1166 | psig->cstime += tgstime + sig->cstime; | |
6fac4829 | 1167 | psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime; |
3795e161 JJ |
1168 | psig->cmin_flt += |
1169 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1170 | psig->cmaj_flt += | |
1171 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1172 | psig->cnvcsw += | |
1173 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1174 | psig->cnivcsw += | |
1175 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1176 | psig->cinblock += |
1177 | task_io_get_inblock(p) + | |
1178 | sig->inblock + sig->cinblock; | |
1179 | psig->coublock += | |
1180 | task_io_get_oublock(p) + | |
1181 | sig->oublock + sig->coublock; | |
1f10206c JP |
1182 | maxrss = max(sig->maxrss, sig->cmaxrss); |
1183 | if (psig->cmaxrss < maxrss) | |
1184 | psig->cmaxrss = maxrss; | |
5995477a AR |
1185 | task_io_accounting_add(&psig->ioac, &p->ioac); |
1186 | task_io_accounting_add(&psig->ioac, &sig->ioac); | |
e78c3496 | 1187 | write_sequnlock(&psig->stats_lock); |
f953ccd0 | 1188 | spin_unlock_irq(¤t->sighand->siglock); |
1da177e4 LT |
1189 | } |
1190 | ||
ce72a16f AV |
1191 | if (wo->wo_rusage) |
1192 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
1da177e4 LT |
1193 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1194 | ? p->signal->group_exit_code : p->exit_code; | |
359566fa | 1195 | wo->wo_stat = status; |
2f4e6e2a | 1196 | |
b4360690 | 1197 | if (state == EXIT_TRACE) { |
1da177e4 | 1198 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1199 | /* We dropped tasklist, ptracer could die and untrace */ |
1200 | ptrace_unlink(p); | |
b4360690 ON |
1201 | |
1202 | /* If parent wants a zombie, don't release it now */ | |
1203 | state = EXIT_ZOMBIE; | |
1204 | if (do_notify_parent(p, p->exit_signal)) | |
1205 | state = EXIT_DEAD; | |
abd50b39 | 1206 | p->exit_state = state; |
1da177e4 LT |
1207 | write_unlock_irq(&tasklist_lock); |
1208 | } | |
abd50b39 | 1209 | if (state == EXIT_DEAD) |
1da177e4 | 1210 | release_task(p); |
2f4e6e2a | 1211 | |
76d9871e AV |
1212 | out_info: |
1213 | infop = wo->wo_info; | |
1214 | if (infop) { | |
1215 | if ((status & 0x7f) == 0) { | |
1216 | infop->cause = CLD_EXITED; | |
1217 | infop->status = status >> 8; | |
1218 | } else { | |
1219 | infop->cause = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1220 | infop->status = status & 0x7f; | |
1221 | } | |
1222 | infop->pid = pid; | |
1223 | infop->uid = uid; | |
1224 | } | |
1225 | ||
67d7ddde | 1226 | return pid; |
1da177e4 LT |
1227 | } |
1228 | ||
90bc8d8b ON |
1229 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
1230 | { | |
1231 | if (ptrace) { | |
570ac933 | 1232 | if (task_is_traced(p) && !(p->jobctl & JOBCTL_LISTENING)) |
90bc8d8b ON |
1233 | return &p->exit_code; |
1234 | } else { | |
1235 | if (p->signal->flags & SIGNAL_STOP_STOPPED) | |
1236 | return &p->signal->group_exit_code; | |
1237 | } | |
1238 | return NULL; | |
1239 | } | |
1240 | ||
19e27463 TH |
1241 | /** |
1242 | * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED | |
1243 | * @wo: wait options | |
1244 | * @ptrace: is the wait for ptrace | |
1245 | * @p: task to wait for | |
1246 | * | |
1247 | * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED. | |
1248 | * | |
1249 | * CONTEXT: | |
1250 | * read_lock(&tasklist_lock), which is released if return value is | |
1251 | * non-zero. Also, grabs and releases @p->sighand->siglock. | |
1252 | * | |
1253 | * RETURNS: | |
1254 | * 0 if wait condition didn't exist and search for other wait conditions | |
1255 | * should continue. Non-zero return, -errno on failure and @p's pid on | |
1256 | * success, implies that tasklist_lock is released and wait condition | |
1257 | * search should terminate. | |
1da177e4 | 1258 | */ |
9e8ae01d ON |
1259 | static int wait_task_stopped(struct wait_opts *wo, |
1260 | int ptrace, struct task_struct *p) | |
1da177e4 | 1261 | { |
67d7ddde AV |
1262 | struct waitid_info *infop; |
1263 | int exit_code, *p_code, why; | |
ee7c82da | 1264 | uid_t uid = 0; /* unneeded, required by compiler */ |
c8950783 | 1265 | pid_t pid; |
1da177e4 | 1266 | |
47918025 ON |
1267 | /* |
1268 | * Traditionally we see ptrace'd stopped tasks regardless of options. | |
1269 | */ | |
9e8ae01d | 1270 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
98abed02 RM |
1271 | return 0; |
1272 | ||
19e27463 TH |
1273 | if (!task_stopped_code(p, ptrace)) |
1274 | return 0; | |
1275 | ||
ee7c82da ON |
1276 | exit_code = 0; |
1277 | spin_lock_irq(&p->sighand->siglock); | |
1278 | ||
90bc8d8b ON |
1279 | p_code = task_stopped_code(p, ptrace); |
1280 | if (unlikely(!p_code)) | |
ee7c82da ON |
1281 | goto unlock_sig; |
1282 | ||
90bc8d8b | 1283 | exit_code = *p_code; |
ee7c82da ON |
1284 | if (!exit_code) |
1285 | goto unlock_sig; | |
1286 | ||
9e8ae01d | 1287 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
90bc8d8b | 1288 | *p_code = 0; |
ee7c82da | 1289 | |
8ca937a6 | 1290 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
ee7c82da ON |
1291 | unlock_sig: |
1292 | spin_unlock_irq(&p->sighand->siglock); | |
1293 | if (!exit_code) | |
1da177e4 LT |
1294 | return 0; |
1295 | ||
1296 | /* | |
1297 | * Now we are pretty sure this task is interesting. | |
1298 | * Make sure it doesn't get reaped out from under us while we | |
1299 | * give up the lock and then examine it below. We don't want to | |
1300 | * keep holding onto the tasklist_lock while we call getrusage and | |
1301 | * possibly take page faults for user memory. | |
1302 | */ | |
1303 | get_task_struct(p); | |
6c5f3e7b | 1304 | pid = task_pid_vnr(p); |
f470021a | 1305 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
1da177e4 | 1306 | read_unlock(&tasklist_lock); |
1029a2b5 | 1307 | sched_annotate_sleep(); |
e61a2502 AV |
1308 | if (wo->wo_rusage) |
1309 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1310 | put_task_struct(p); |
1da177e4 | 1311 | |
bb380ec3 AV |
1312 | if (likely(!(wo->wo_flags & WNOWAIT))) |
1313 | wo->wo_stat = (exit_code << 8) | 0x7f; | |
1da177e4 | 1314 | |
9e8ae01d | 1315 | infop = wo->wo_info; |
67d7ddde AV |
1316 | if (infop) { |
1317 | infop->cause = why; | |
1318 | infop->status = exit_code; | |
1319 | infop->pid = pid; | |
1320 | infop->uid = uid; | |
1321 | } | |
67d7ddde | 1322 | return pid; |
1da177e4 LT |
1323 | } |
1324 | ||
1325 | /* | |
1326 | * Handle do_wait work for one task in a live, non-stopped state. | |
1327 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1328 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1329 | * released the lock and the system call should return. | |
1330 | */ | |
9e8ae01d | 1331 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1332 | { |
bb380ec3 | 1333 | struct waitid_info *infop; |
1da177e4 LT |
1334 | pid_t pid; |
1335 | uid_t uid; | |
1336 | ||
9e8ae01d | 1337 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
98abed02 RM |
1338 | return 0; |
1339 | ||
1da177e4 LT |
1340 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1341 | return 0; | |
1342 | ||
1343 | spin_lock_irq(&p->sighand->siglock); | |
1344 | /* Re-check with the lock held. */ | |
1345 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1346 | spin_unlock_irq(&p->sighand->siglock); | |
1347 | return 0; | |
1348 | } | |
9e8ae01d | 1349 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
1da177e4 | 1350 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
8ca937a6 | 1351 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
1da177e4 LT |
1352 | spin_unlock_irq(&p->sighand->siglock); |
1353 | ||
6c5f3e7b | 1354 | pid = task_pid_vnr(p); |
1da177e4 LT |
1355 | get_task_struct(p); |
1356 | read_unlock(&tasklist_lock); | |
1029a2b5 | 1357 | sched_annotate_sleep(); |
e61a2502 AV |
1358 | if (wo->wo_rusage) |
1359 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1360 | put_task_struct(p); |
1da177e4 | 1361 | |
bb380ec3 AV |
1362 | infop = wo->wo_info; |
1363 | if (!infop) { | |
359566fa | 1364 | wo->wo_stat = 0xffff; |
1da177e4 | 1365 | } else { |
bb380ec3 AV |
1366 | infop->cause = CLD_CONTINUED; |
1367 | infop->pid = pid; | |
1368 | infop->uid = uid; | |
1369 | infop->status = SIGCONT; | |
1da177e4 | 1370 | } |
bb380ec3 | 1371 | return pid; |
1da177e4 LT |
1372 | } |
1373 | ||
98abed02 RM |
1374 | /* |
1375 | * Consider @p for a wait by @parent. | |
1376 | * | |
9e8ae01d | 1377 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1378 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1379 | * Returns zero if the search for a child should continue; | |
9e8ae01d | 1380 | * then ->notask_error is 0 if @p is an eligible child, |
3a2f5a59 | 1381 | * or still -ECHILD. |
98abed02 | 1382 | */ |
b6e763f0 ON |
1383 | static int wait_consider_task(struct wait_opts *wo, int ptrace, |
1384 | struct task_struct *p) | |
98abed02 | 1385 | { |
3245d6ac ON |
1386 | /* |
1387 | * We can race with wait_task_zombie() from another thread. | |
1388 | * Ensure that EXIT_ZOMBIE -> EXIT_DEAD/EXIT_TRACE transition | |
1389 | * can't confuse the checks below. | |
1390 | */ | |
6aa7de05 | 1391 | int exit_state = READ_ONCE(p->exit_state); |
b3ab0316 ON |
1392 | int ret; |
1393 | ||
3245d6ac | 1394 | if (unlikely(exit_state == EXIT_DEAD)) |
b3ab0316 ON |
1395 | return 0; |
1396 | ||
bf959931 | 1397 | ret = eligible_child(wo, ptrace, p); |
14dd0b81 | 1398 | if (!ret) |
98abed02 RM |
1399 | return ret; |
1400 | ||
3245d6ac | 1401 | if (unlikely(exit_state == EXIT_TRACE)) { |
50b8d257 | 1402 | /* |
abd50b39 ON |
1403 | * ptrace == 0 means we are the natural parent. In this case |
1404 | * we should clear notask_error, debugger will notify us. | |
50b8d257 | 1405 | */ |
abd50b39 | 1406 | if (likely(!ptrace)) |
50b8d257 | 1407 | wo->notask_error = 0; |
823b018e | 1408 | return 0; |
50b8d257 | 1409 | } |
823b018e | 1410 | |
377d75da ON |
1411 | if (likely(!ptrace) && unlikely(p->ptrace)) { |
1412 | /* | |
1413 | * If it is traced by its real parent's group, just pretend | |
1414 | * the caller is ptrace_do_wait() and reap this child if it | |
1415 | * is zombie. | |
1416 | * | |
1417 | * This also hides group stop state from real parent; otherwise | |
1418 | * a single stop can be reported twice as group and ptrace stop. | |
1419 | * If a ptracer wants to distinguish these two events for its | |
1420 | * own children it should create a separate process which takes | |
1421 | * the role of real parent. | |
1422 | */ | |
1423 | if (!ptrace_reparented(p)) | |
1424 | ptrace = 1; | |
1425 | } | |
1426 | ||
45cb24a1 | 1427 | /* slay zombie? */ |
3245d6ac | 1428 | if (exit_state == EXIT_ZOMBIE) { |
9b84cca2 | 1429 | /* we don't reap group leaders with subthreads */ |
7c733eb3 ON |
1430 | if (!delay_group_leader(p)) { |
1431 | /* | |
1432 | * A zombie ptracee is only visible to its ptracer. | |
1433 | * Notification and reaping will be cascaded to the | |
1434 | * real parent when the ptracer detaches. | |
1435 | */ | |
1436 | if (unlikely(ptrace) || likely(!p->ptrace)) | |
1437 | return wait_task_zombie(wo, p); | |
1438 | } | |
98abed02 | 1439 | |
f470021a | 1440 | /* |
9b84cca2 TH |
1441 | * Allow access to stopped/continued state via zombie by |
1442 | * falling through. Clearing of notask_error is complex. | |
1443 | * | |
1444 | * When !@ptrace: | |
1445 | * | |
1446 | * If WEXITED is set, notask_error should naturally be | |
1447 | * cleared. If not, subset of WSTOPPED|WCONTINUED is set, | |
1448 | * so, if there are live subthreads, there are events to | |
1449 | * wait for. If all subthreads are dead, it's still safe | |
1450 | * to clear - this function will be called again in finite | |
1451 | * amount time once all the subthreads are released and | |
1452 | * will then return without clearing. | |
1453 | * | |
1454 | * When @ptrace: | |
1455 | * | |
1456 | * Stopped state is per-task and thus can't change once the | |
1457 | * target task dies. Only continued and exited can happen. | |
1458 | * Clear notask_error if WCONTINUED | WEXITED. | |
1459 | */ | |
1460 | if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED))) | |
1461 | wo->notask_error = 0; | |
1462 | } else { | |
1463 | /* | |
1464 | * @p is alive and it's gonna stop, continue or exit, so | |
1465 | * there always is something to wait for. | |
f470021a | 1466 | */ |
9e8ae01d | 1467 | wo->notask_error = 0; |
f470021a RM |
1468 | } |
1469 | ||
98abed02 | 1470 | /* |
45cb24a1 TH |
1471 | * Wait for stopped. Depending on @ptrace, different stopped state |
1472 | * is used and the two don't interact with each other. | |
98abed02 | 1473 | */ |
19e27463 TH |
1474 | ret = wait_task_stopped(wo, ptrace, p); |
1475 | if (ret) | |
1476 | return ret; | |
98abed02 RM |
1477 | |
1478 | /* | |
45cb24a1 TH |
1479 | * Wait for continued. There's only one continued state and the |
1480 | * ptracer can consume it which can confuse the real parent. Don't | |
1481 | * use WCONTINUED from ptracer. You don't need or want it. | |
98abed02 | 1482 | */ |
9e8ae01d | 1483 | return wait_task_continued(wo, p); |
98abed02 RM |
1484 | } |
1485 | ||
1486 | /* | |
1487 | * Do the work of do_wait() for one thread in the group, @tsk. | |
1488 | * | |
9e8ae01d | 1489 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1490 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1491 | * Returns zero if the search for a child should continue; then | |
9e8ae01d | 1492 | * ->notask_error is 0 if there were any eligible children, |
3a2f5a59 | 1493 | * or still -ECHILD. |
98abed02 | 1494 | */ |
9e8ae01d | 1495 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1496 | { |
1497 | struct task_struct *p; | |
1498 | ||
1499 | list_for_each_entry(p, &tsk->children, sibling) { | |
9cd80bbb | 1500 | int ret = wait_consider_task(wo, 0, p); |
a0be55de | 1501 | |
9cd80bbb ON |
1502 | if (ret) |
1503 | return ret; | |
98abed02 RM |
1504 | } |
1505 | ||
1506 | return 0; | |
1507 | } | |
1508 | ||
9e8ae01d | 1509 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1510 | { |
1511 | struct task_struct *p; | |
1512 | ||
f470021a | 1513 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
b6e763f0 | 1514 | int ret = wait_consider_task(wo, 1, p); |
a0be55de | 1515 | |
f470021a | 1516 | if (ret) |
98abed02 | 1517 | return ret; |
98abed02 RM |
1518 | } |
1519 | ||
1520 | return 0; | |
1521 | } | |
1522 | ||
ac6424b9 | 1523 | static int child_wait_callback(wait_queue_entry_t *wait, unsigned mode, |
0b7570e7 ON |
1524 | int sync, void *key) |
1525 | { | |
1526 | struct wait_opts *wo = container_of(wait, struct wait_opts, | |
1527 | child_wait); | |
1528 | struct task_struct *p = key; | |
1529 | ||
5c01ba49 | 1530 | if (!eligible_pid(wo, p)) |
0b7570e7 ON |
1531 | return 0; |
1532 | ||
b4fe5182 ON |
1533 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) |
1534 | return 0; | |
1535 | ||
0b7570e7 ON |
1536 | return default_wake_function(wait, mode, sync, key); |
1537 | } | |
1538 | ||
a7f0765e ON |
1539 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) |
1540 | { | |
0b7570e7 | 1541 | __wake_up_sync_key(&parent->signal->wait_chldexit, |
ce4dd442 | 1542 | TASK_INTERRUPTIBLE, p); |
a7f0765e ON |
1543 | } |
1544 | ||
5449162a JN |
1545 | static bool is_effectively_child(struct wait_opts *wo, bool ptrace, |
1546 | struct task_struct *target) | |
1547 | { | |
1548 | struct task_struct *parent = | |
1549 | !ptrace ? target->real_parent : target->parent; | |
1550 | ||
1551 | return current == parent || (!(wo->wo_flags & __WNOTHREAD) && | |
1552 | same_thread_group(current, parent)); | |
1553 | } | |
1554 | ||
1555 | /* | |
1556 | * Optimization for waiting on PIDTYPE_PID. No need to iterate through child | |
1557 | * and tracee lists to find the target task. | |
1558 | */ | |
1559 | static int do_wait_pid(struct wait_opts *wo) | |
1560 | { | |
1561 | bool ptrace; | |
1562 | struct task_struct *target; | |
1563 | int retval; | |
1564 | ||
1565 | ptrace = false; | |
1566 | target = pid_task(wo->wo_pid, PIDTYPE_TGID); | |
1567 | if (target && is_effectively_child(wo, ptrace, target)) { | |
1568 | retval = wait_consider_task(wo, ptrace, target); | |
1569 | if (retval) | |
1570 | return retval; | |
1571 | } | |
1572 | ||
1573 | ptrace = true; | |
1574 | target = pid_task(wo->wo_pid, PIDTYPE_PID); | |
1575 | if (target && target->ptrace && | |
1576 | is_effectively_child(wo, ptrace, target)) { | |
1577 | retval = wait_consider_task(wo, ptrace, target); | |
1578 | if (retval) | |
1579 | return retval; | |
1580 | } | |
1581 | ||
1582 | return 0; | |
1583 | } | |
1584 | ||
9e8ae01d | 1585 | static long do_wait(struct wait_opts *wo) |
1da177e4 | 1586 | { |
98abed02 | 1587 | int retval; |
1da177e4 | 1588 | |
9e8ae01d | 1589 | trace_sched_process_wait(wo->wo_pid); |
0a16b607 | 1590 | |
0b7570e7 ON |
1591 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); |
1592 | wo->child_wait.private = current; | |
1593 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); | |
1da177e4 | 1594 | repeat: |
98abed02 | 1595 | /* |
3da56d16 | 1596 | * If there is nothing that can match our criteria, just get out. |
9e8ae01d ON |
1597 | * We will clear ->notask_error to zero if we see any child that |
1598 | * might later match our criteria, even if we are not able to reap | |
1599 | * it yet. | |
98abed02 | 1600 | */ |
64a16caf | 1601 | wo->notask_error = -ECHILD; |
9e8ae01d | 1602 | if ((wo->wo_type < PIDTYPE_MAX) && |
1722c14a | 1603 | (!wo->wo_pid || !pid_has_task(wo->wo_pid, wo->wo_type))) |
64a16caf | 1604 | goto notask; |
161550d7 | 1605 | |
f95d39d1 | 1606 | set_current_state(TASK_INTERRUPTIBLE); |
1da177e4 | 1607 | read_lock(&tasklist_lock); |
9e8ae01d | 1608 | |
5449162a JN |
1609 | if (wo->wo_type == PIDTYPE_PID) { |
1610 | retval = do_wait_pid(wo); | |
64a16caf | 1611 | if (retval) |
98abed02 | 1612 | goto end; |
5449162a JN |
1613 | } else { |
1614 | struct task_struct *tsk = current; | |
1615 | ||
1616 | do { | |
1617 | retval = do_wait_thread(wo, tsk); | |
1618 | if (retval) | |
1619 | goto end; | |
98abed02 | 1620 | |
5449162a JN |
1621 | retval = ptrace_do_wait(wo, tsk); |
1622 | if (retval) | |
1623 | goto end; | |
1624 | ||
1625 | if (wo->wo_flags & __WNOTHREAD) | |
1626 | break; | |
1627 | } while_each_thread(current, tsk); | |
1628 | } | |
1da177e4 | 1629 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1630 | |
64a16caf | 1631 | notask: |
9e8ae01d ON |
1632 | retval = wo->notask_error; |
1633 | if (!retval && !(wo->wo_flags & WNOHANG)) { | |
1da177e4 | 1634 | retval = -ERESTARTSYS; |
98abed02 RM |
1635 | if (!signal_pending(current)) { |
1636 | schedule(); | |
1637 | goto repeat; | |
1638 | } | |
1da177e4 | 1639 | } |
1da177e4 | 1640 | end: |
f95d39d1 | 1641 | __set_current_state(TASK_RUNNING); |
0b7570e7 | 1642 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
1da177e4 LT |
1643 | return retval; |
1644 | } | |
1645 | ||
67d7ddde | 1646 | static long kernel_waitid(int which, pid_t upid, struct waitid_info *infop, |
ce72a16f | 1647 | int options, struct rusage *ru) |
1da177e4 | 1648 | { |
9e8ae01d | 1649 | struct wait_opts wo; |
161550d7 EB |
1650 | struct pid *pid = NULL; |
1651 | enum pid_type type; | |
1da177e4 | 1652 | long ret; |
ba7d25f3 | 1653 | unsigned int f_flags = 0; |
1da177e4 | 1654 | |
91c4e8ea ON |
1655 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED| |
1656 | __WNOTHREAD|__WCLONE|__WALL)) | |
1da177e4 LT |
1657 | return -EINVAL; |
1658 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1659 | return -EINVAL; | |
1660 | ||
1661 | switch (which) { | |
1662 | case P_ALL: | |
161550d7 | 1663 | type = PIDTYPE_MAX; |
1da177e4 LT |
1664 | break; |
1665 | case P_PID: | |
161550d7 EB |
1666 | type = PIDTYPE_PID; |
1667 | if (upid <= 0) | |
1da177e4 | 1668 | return -EINVAL; |
3695eae5 CB |
1669 | |
1670 | pid = find_get_pid(upid); | |
1da177e4 LT |
1671 | break; |
1672 | case P_PGID: | |
161550d7 | 1673 | type = PIDTYPE_PGID; |
821cc7b0 | 1674 | if (upid < 0) |
1da177e4 | 1675 | return -EINVAL; |
3695eae5 | 1676 | |
821cc7b0 EB |
1677 | if (upid) |
1678 | pid = find_get_pid(upid); | |
1679 | else | |
1680 | pid = get_task_pid(current, PIDTYPE_PGID); | |
3695eae5 CB |
1681 | break; |
1682 | case P_PIDFD: | |
1683 | type = PIDTYPE_PID; | |
1684 | if (upid < 0) | |
1da177e4 | 1685 | return -EINVAL; |
3695eae5 | 1686 | |
ba7d25f3 | 1687 | pid = pidfd_get_pid(upid, &f_flags); |
3695eae5 CB |
1688 | if (IS_ERR(pid)) |
1689 | return PTR_ERR(pid); | |
ba7d25f3 | 1690 | |
1da177e4 LT |
1691 | break; |
1692 | default: | |
1693 | return -EINVAL; | |
1694 | } | |
1695 | ||
9e8ae01d ON |
1696 | wo.wo_type = type; |
1697 | wo.wo_pid = pid; | |
1698 | wo.wo_flags = options; | |
1699 | wo.wo_info = infop; | |
9e8ae01d | 1700 | wo.wo_rusage = ru; |
ba7d25f3 CB |
1701 | if (f_flags & O_NONBLOCK) |
1702 | wo.wo_flags |= WNOHANG; | |
1703 | ||
9e8ae01d | 1704 | ret = do_wait(&wo); |
ba7d25f3 CB |
1705 | if (!ret && !(options & WNOHANG) && (f_flags & O_NONBLOCK)) |
1706 | ret = -EAGAIN; | |
dfe16dfa | 1707 | |
161550d7 | 1708 | put_pid(pid); |
1da177e4 LT |
1709 | return ret; |
1710 | } | |
1711 | ||
ce72a16f AV |
1712 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
1713 | infop, int, options, struct rusage __user *, ru) | |
1714 | { | |
1715 | struct rusage r; | |
67d7ddde AV |
1716 | struct waitid_info info = {.status = 0}; |
1717 | long err = kernel_waitid(which, upid, &info, options, ru ? &r : NULL); | |
634a8160 | 1718 | int signo = 0; |
6c85501f | 1719 | |
634a8160 AV |
1720 | if (err > 0) { |
1721 | signo = SIGCHLD; | |
1722 | err = 0; | |
ce72a16f AV |
1723 | if (ru && copy_to_user(ru, &r, sizeof(struct rusage))) |
1724 | return -EFAULT; | |
1725 | } | |
67d7ddde AV |
1726 | if (!infop) |
1727 | return err; | |
1728 | ||
41cd7805 | 1729 | if (!user_write_access_begin(infop, sizeof(*infop))) |
1c9fec47 | 1730 | return -EFAULT; |
96ca579a | 1731 | |
634a8160 | 1732 | unsafe_put_user(signo, &infop->si_signo, Efault); |
4c48abe9 | 1733 | unsafe_put_user(0, &infop->si_errno, Efault); |
cc731525 | 1734 | unsafe_put_user(info.cause, &infop->si_code, Efault); |
4c48abe9 AV |
1735 | unsafe_put_user(info.pid, &infop->si_pid, Efault); |
1736 | unsafe_put_user(info.uid, &infop->si_uid, Efault); | |
1737 | unsafe_put_user(info.status, &infop->si_status, Efault); | |
41cd7805 | 1738 | user_write_access_end(); |
ce72a16f | 1739 | return err; |
4c48abe9 | 1740 | Efault: |
41cd7805 | 1741 | user_write_access_end(); |
4c48abe9 | 1742 | return -EFAULT; |
ce72a16f AV |
1743 | } |
1744 | ||
92ebce5a AV |
1745 | long kernel_wait4(pid_t upid, int __user *stat_addr, int options, |
1746 | struct rusage *ru) | |
1da177e4 | 1747 | { |
9e8ae01d | 1748 | struct wait_opts wo; |
161550d7 EB |
1749 | struct pid *pid = NULL; |
1750 | enum pid_type type; | |
1da177e4 LT |
1751 | long ret; |
1752 | ||
1753 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1754 | __WNOTHREAD|__WCLONE|__WALL)) | |
1755 | return -EINVAL; | |
161550d7 | 1756 | |
dd83c161 | 1757 | /* -INT_MIN is not defined */ |
1758 | if (upid == INT_MIN) | |
1759 | return -ESRCH; | |
1760 | ||
161550d7 EB |
1761 | if (upid == -1) |
1762 | type = PIDTYPE_MAX; | |
1763 | else if (upid < 0) { | |
1764 | type = PIDTYPE_PGID; | |
1765 | pid = find_get_pid(-upid); | |
1766 | } else if (upid == 0) { | |
1767 | type = PIDTYPE_PGID; | |
2ae448ef | 1768 | pid = get_task_pid(current, PIDTYPE_PGID); |
161550d7 EB |
1769 | } else /* upid > 0 */ { |
1770 | type = PIDTYPE_PID; | |
1771 | pid = find_get_pid(upid); | |
1772 | } | |
1773 | ||
9e8ae01d ON |
1774 | wo.wo_type = type; |
1775 | wo.wo_pid = pid; | |
1776 | wo.wo_flags = options | WEXITED; | |
1777 | wo.wo_info = NULL; | |
359566fa | 1778 | wo.wo_stat = 0; |
9e8ae01d ON |
1779 | wo.wo_rusage = ru; |
1780 | ret = do_wait(&wo); | |
161550d7 | 1781 | put_pid(pid); |
359566fa AV |
1782 | if (ret > 0 && stat_addr && put_user(wo.wo_stat, stat_addr)) |
1783 | ret = -EFAULT; | |
1da177e4 | 1784 | |
1da177e4 LT |
1785 | return ret; |
1786 | } | |
1787 | ||
8043fc14 CH |
1788 | int kernel_wait(pid_t pid, int *stat) |
1789 | { | |
1790 | struct wait_opts wo = { | |
1791 | .wo_type = PIDTYPE_PID, | |
1792 | .wo_pid = find_get_pid(pid), | |
1793 | .wo_flags = WEXITED, | |
1794 | }; | |
1795 | int ret; | |
1796 | ||
1797 | ret = do_wait(&wo); | |
1798 | if (ret > 0 && wo.wo_stat) | |
1799 | *stat = wo.wo_stat; | |
1800 | put_pid(wo.wo_pid); | |
1801 | return ret; | |
1802 | } | |
1803 | ||
ce72a16f AV |
1804 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
1805 | int, options, struct rusage __user *, ru) | |
1806 | { | |
1807 | struct rusage r; | |
1808 | long err = kernel_wait4(upid, stat_addr, options, ru ? &r : NULL); | |
1809 | ||
1810 | if (err > 0) { | |
1811 | if (ru && copy_to_user(ru, &r, sizeof(struct rusage))) | |
1812 | return -EFAULT; | |
1813 | } | |
1814 | return err; | |
1815 | } | |
1816 | ||
1da177e4 LT |
1817 | #ifdef __ARCH_WANT_SYS_WAITPID |
1818 | ||
1819 | /* | |
1820 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1821 | * implemented by calling sys_wait4() from libc.a. | |
1822 | */ | |
17da2bd9 | 1823 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
1da177e4 | 1824 | { |
d300b610 | 1825 | return kernel_wait4(pid, stat_addr, options, NULL); |
1da177e4 LT |
1826 | } |
1827 | ||
1828 | #endif | |
7e95a225 AV |
1829 | |
1830 | #ifdef CONFIG_COMPAT | |
1831 | COMPAT_SYSCALL_DEFINE4(wait4, | |
1832 | compat_pid_t, pid, | |
1833 | compat_uint_t __user *, stat_addr, | |
1834 | int, options, | |
1835 | struct compat_rusage __user *, ru) | |
1836 | { | |
ce72a16f AV |
1837 | struct rusage r; |
1838 | long err = kernel_wait4(pid, stat_addr, options, ru ? &r : NULL); | |
1839 | if (err > 0) { | |
1840 | if (ru && put_compat_rusage(&r, ru)) | |
1841 | return -EFAULT; | |
7e95a225 | 1842 | } |
ce72a16f | 1843 | return err; |
7e95a225 AV |
1844 | } |
1845 | ||
1846 | COMPAT_SYSCALL_DEFINE5(waitid, | |
1847 | int, which, compat_pid_t, pid, | |
1848 | struct compat_siginfo __user *, infop, int, options, | |
1849 | struct compat_rusage __user *, uru) | |
1850 | { | |
7e95a225 | 1851 | struct rusage ru; |
67d7ddde AV |
1852 | struct waitid_info info = {.status = 0}; |
1853 | long err = kernel_waitid(which, pid, &info, options, uru ? &ru : NULL); | |
634a8160 AV |
1854 | int signo = 0; |
1855 | if (err > 0) { | |
1856 | signo = SIGCHLD; | |
1857 | err = 0; | |
6c85501f AV |
1858 | if (uru) { |
1859 | /* kernel_waitid() overwrites everything in ru */ | |
1860 | if (COMPAT_USE_64BIT_TIME) | |
1861 | err = copy_to_user(uru, &ru, sizeof(ru)); | |
1862 | else | |
1863 | err = put_compat_rusage(&ru, uru); | |
1864 | if (err) | |
1865 | return -EFAULT; | |
1866 | } | |
7e95a225 AV |
1867 | } |
1868 | ||
4c48abe9 AV |
1869 | if (!infop) |
1870 | return err; | |
1871 | ||
41cd7805 | 1872 | if (!user_write_access_begin(infop, sizeof(*infop))) |
1c9fec47 | 1873 | return -EFAULT; |
96ca579a | 1874 | |
634a8160 | 1875 | unsafe_put_user(signo, &infop->si_signo, Efault); |
4c48abe9 | 1876 | unsafe_put_user(0, &infop->si_errno, Efault); |
cc731525 | 1877 | unsafe_put_user(info.cause, &infop->si_code, Efault); |
4c48abe9 AV |
1878 | unsafe_put_user(info.pid, &infop->si_pid, Efault); |
1879 | unsafe_put_user(info.uid, &infop->si_uid, Efault); | |
1880 | unsafe_put_user(info.status, &infop->si_status, Efault); | |
41cd7805 | 1881 | user_write_access_end(); |
67d7ddde | 1882 | return err; |
4c48abe9 | 1883 | Efault: |
41cd7805 | 1884 | user_write_access_end(); |
4c48abe9 | 1885 | return -EFAULT; |
7e95a225 AV |
1886 | } |
1887 | #endif | |
7c2c11b2 | 1888 | |
38fd525a EB |
1889 | /** |
1890 | * thread_group_exited - check that a thread group has exited | |
1891 | * @pid: tgid of thread group to be checked. | |
1892 | * | |
1893 | * Test if the thread group represented by tgid has exited (all | |
1894 | * threads are zombies, dead or completely gone). | |
1895 | * | |
1896 | * Return: true if the thread group has exited. false otherwise. | |
1897 | */ | |
1898 | bool thread_group_exited(struct pid *pid) | |
1899 | { | |
1900 | struct task_struct *task; | |
1901 | bool exited; | |
1902 | ||
1903 | rcu_read_lock(); | |
1904 | task = pid_task(pid, PIDTYPE_PID); | |
1905 | exited = !task || | |
1906 | (READ_ONCE(task->exit_state) && thread_group_empty(task)); | |
1907 | rcu_read_unlock(); | |
1908 | ||
1909 | return exited; | |
1910 | } | |
1911 | EXPORT_SYMBOL(thread_group_exited); | |
1912 | ||
c27cd083 MR |
1913 | /* |
1914 | * This needs to be __function_aligned as GCC implicitly makes any | |
1915 | * implementation of abort() cold and drops alignment specified by | |
1916 | * -falign-functions=N. | |
1917 | * | |
1918 | * See https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88345#c11 | |
1919 | */ | |
1920 | __weak __function_aligned void abort(void) | |
7c2c11b2 SM |
1921 | { |
1922 | BUG(); | |
1923 | ||
1924 | /* if that doesn't kill us, halt */ | |
1925 | panic("Oops failed to kill thread"); | |
1926 | } | |
dc8635b7 | 1927 | EXPORT_SYMBOL(abort); |