1 // SPDX-License-Identifier: GPL-2.0-only 1 // SPDX-License-Identifier: GPL-2.0-only
2 /* 2 /*
3 * linux/kernel/exit.c 3 * linux/kernel/exit.c
4 * 4 *
5 * Copyright (C) 1991, 1992 Linus Torvalds 5 * Copyright (C) 1991, 1992 Linus Torvalds
6 */ 6 */
7 7
8 #include <linux/mm.h> 8 #include <linux/mm.h>
9 #include <linux/slab.h> 9 #include <linux/slab.h>
10 #include <linux/sched/autogroup.h> 10 #include <linux/sched/autogroup.h>
11 #include <linux/sched/mm.h> 11 #include <linux/sched/mm.h>
12 #include <linux/sched/stat.h> 12 #include <linux/sched/stat.h>
13 #include <linux/sched/task.h> 13 #include <linux/sched/task.h>
14 #include <linux/sched/task_stack.h> 14 #include <linux/sched/task_stack.h>
15 #include <linux/sched/cputime.h> 15 #include <linux/sched/cputime.h>
16 #include <linux/interrupt.h> 16 #include <linux/interrupt.h>
17 #include <linux/module.h> 17 #include <linux/module.h>
18 #include <linux/capability.h> 18 #include <linux/capability.h>
19 #include <linux/completion.h> 19 #include <linux/completion.h>
20 #include <linux/personality.h> 20 #include <linux/personality.h>
21 #include <linux/tty.h> 21 #include <linux/tty.h>
22 #include <linux/iocontext.h> 22 #include <linux/iocontext.h>
23 #include <linux/key.h> 23 #include <linux/key.h>
24 #include <linux/cpu.h> 24 #include <linux/cpu.h>
25 #include <linux/acct.h> 25 #include <linux/acct.h>
26 #include <linux/tsacct_kern.h> 26 #include <linux/tsacct_kern.h>
27 #include <linux/file.h> 27 #include <linux/file.h>
28 #include <linux/fdtable.h> 28 #include <linux/fdtable.h>
29 #include <linux/freezer.h> 29 #include <linux/freezer.h>
30 #include <linux/binfmts.h> 30 #include <linux/binfmts.h>
31 #include <linux/nsproxy.h> 31 #include <linux/nsproxy.h>
32 #include <linux/pid_namespace.h> 32 #include <linux/pid_namespace.h>
33 #include <linux/ptrace.h> 33 #include <linux/ptrace.h>
34 #include <linux/profile.h> 34 #include <linux/profile.h>
35 #include <linux/mount.h> 35 #include <linux/mount.h>
36 #include <linux/proc_fs.h> 36 #include <linux/proc_fs.h>
37 #include <linux/kthread.h> 37 #include <linux/kthread.h>
38 #include <linux/mempolicy.h> 38 #include <linux/mempolicy.h>
39 #include <linux/taskstats_kern.h> 39 #include <linux/taskstats_kern.h>
40 #include <linux/delayacct.h> 40 #include <linux/delayacct.h>
41 #include <linux/cgroup.h> 41 #include <linux/cgroup.h>
42 #include <linux/syscalls.h> 42 #include <linux/syscalls.h>
43 #include <linux/signal.h> 43 #include <linux/signal.h>
44 #include <linux/posix-timers.h> 44 #include <linux/posix-timers.h>
45 #include <linux/cn_proc.h> 45 #include <linux/cn_proc.h>
46 #include <linux/mutex.h> 46 #include <linux/mutex.h>
47 #include <linux/futex.h> 47 #include <linux/futex.h>
48 #include <linux/pipe_fs_i.h> 48 #include <linux/pipe_fs_i.h>
49 #include <linux/audit.h> /* for audit_free() * 49 #include <linux/audit.h> /* for audit_free() */
50 #include <linux/resource.h> 50 #include <linux/resource.h>
51 #include <linux/task_io_accounting_ops.h> 51 #include <linux/task_io_accounting_ops.h>
52 #include <linux/blkdev.h> 52 #include <linux/blkdev.h>
53 #include <linux/task_work.h> 53 #include <linux/task_work.h>
54 #include <linux/fs_struct.h> 54 #include <linux/fs_struct.h>
55 #include <linux/init_task.h> 55 #include <linux/init_task.h>
56 #include <linux/perf_event.h> 56 #include <linux/perf_event.h>
57 #include <trace/events/sched.h> 57 #include <trace/events/sched.h>
58 #include <linux/hw_breakpoint.h> 58 #include <linux/hw_breakpoint.h>
59 #include <linux/oom.h> 59 #include <linux/oom.h>
60 #include <linux/writeback.h> 60 #include <linux/writeback.h>
61 #include <linux/shm.h> 61 #include <linux/shm.h>
62 #include <linux/kcov.h> 62 #include <linux/kcov.h>
63 #include <linux/kmsan.h> <<
64 #include <linux/random.h> 63 #include <linux/random.h>
65 #include <linux/rcuwait.h> 64 #include <linux/rcuwait.h>
66 #include <linux/compat.h> 65 #include <linux/compat.h>
67 #include <linux/io_uring.h> 66 #include <linux/io_uring.h>
68 #include <linux/kprobes.h> 67 #include <linux/kprobes.h>
69 #include <linux/rethook.h> 68 #include <linux/rethook.h>
70 #include <linux/sysfs.h> <<
71 #include <linux/user_events.h> <<
72 #include <linux/uaccess.h> <<
73 <<
74 #include <uapi/linux/wait.h> <<
75 69
>> 70 #include <linux/uaccess.h>
76 #include <asm/unistd.h> 71 #include <asm/unistd.h>
77 #include <asm/mmu_context.h> 72 #include <asm/mmu_context.h>
78 73
79 #include "exit.h" <<
80 <<
81 /* <<
82 * The default value should be high enough to <<
83 * crashes its kernel from time to time, but l <<
84 * overflowing 32-bit refcounts or the ldsem w <<
85 */ <<
86 static unsigned int oops_limit = 10000; <<
87 <<
88 #ifdef CONFIG_SYSCTL <<
89 static struct ctl_table kern_exit_table[] = { <<
90 { <<
91 .procname = "oops_limit" <<
92 .data = &oops_limit, <<
93 .maxlen = sizeof(oops_ <<
94 .mode = 0644, <<
95 .proc_handler = proc_douintv <<
96 }, <<
97 }; <<
98 <<
99 static __init int kernel_exit_sysctls_init(voi <<
100 { <<
101 register_sysctl_init("kernel", kern_ex <<
102 return 0; <<
103 } <<
104 late_initcall(kernel_exit_sysctls_init); <<
105 #endif <<
106 <<
107 static atomic_t oops_count = ATOMIC_INIT(0); <<
108 <<
109 #ifdef CONFIG_SYSFS <<
110 static ssize_t oops_count_show(struct kobject <<
111 char *page) <<
112 { <<
113 return sysfs_emit(page, "%d\n", atomic <<
114 } <<
115 <<
116 static struct kobj_attribute oops_count_attr = <<
117 <<
118 static __init int kernel_exit_sysfs_init(void) <<
119 { <<
120 sysfs_add_file_to_group(kernel_kobj, & <<
121 return 0; <<
122 } <<
123 late_initcall(kernel_exit_sysfs_init); <<
124 #endif <<
125 <<
126 static void __unhash_process(struct task_struc 74 static void __unhash_process(struct task_struct *p, bool group_dead)
127 { 75 {
128 nr_threads--; 76 nr_threads--;
129 detach_pid(p, PIDTYPE_PID); 77 detach_pid(p, PIDTYPE_PID);
130 if (group_dead) { 78 if (group_dead) {
131 detach_pid(p, PIDTYPE_TGID); 79 detach_pid(p, PIDTYPE_TGID);
132 detach_pid(p, PIDTYPE_PGID); 80 detach_pid(p, PIDTYPE_PGID);
133 detach_pid(p, PIDTYPE_SID); 81 detach_pid(p, PIDTYPE_SID);
134 82
135 list_del_rcu(&p->tasks); 83 list_del_rcu(&p->tasks);
136 list_del_init(&p->sibling); 84 list_del_init(&p->sibling);
137 __this_cpu_dec(process_counts) 85 __this_cpu_dec(process_counts);
138 } 86 }
>> 87 list_del_rcu(&p->thread_group);
139 list_del_rcu(&p->thread_node); 88 list_del_rcu(&p->thread_node);
140 } 89 }
141 90
142 /* 91 /*
143 * This function expects the tasklist_lock wri 92 * This function expects the tasklist_lock write-locked.
144 */ 93 */
145 static void __exit_signal(struct task_struct * 94 static void __exit_signal(struct task_struct *tsk)
146 { 95 {
147 struct signal_struct *sig = tsk->signa 96 struct signal_struct *sig = tsk->signal;
148 bool group_dead = thread_group_leader( 97 bool group_dead = thread_group_leader(tsk);
149 struct sighand_struct *sighand; 98 struct sighand_struct *sighand;
150 struct tty_struct *tty; 99 struct tty_struct *tty;
151 u64 utime, stime; 100 u64 utime, stime;
152 101
153 sighand = rcu_dereference_check(tsk->s 102 sighand = rcu_dereference_check(tsk->sighand,
154 lockde 103 lockdep_tasklist_lock_is_held());
155 spin_lock(&sighand->siglock); 104 spin_lock(&sighand->siglock);
156 105
157 #ifdef CONFIG_POSIX_TIMERS 106 #ifdef CONFIG_POSIX_TIMERS
158 posix_cpu_timers_exit(tsk); 107 posix_cpu_timers_exit(tsk);
159 if (group_dead) 108 if (group_dead)
160 posix_cpu_timers_exit_group(ts 109 posix_cpu_timers_exit_group(tsk);
161 #endif 110 #endif
162 111
163 if (group_dead) { 112 if (group_dead) {
164 tty = sig->tty; 113 tty = sig->tty;
165 sig->tty = NULL; 114 sig->tty = NULL;
166 } else { 115 } else {
167 /* 116 /*
168 * If there is any task waitin 117 * If there is any task waiting for the group exit
169 * then notify it: 118 * then notify it:
170 */ 119 */
171 if (sig->notify_count > 0 && ! 120 if (sig->notify_count > 0 && !--sig->notify_count)
172 wake_up_process(sig->g 121 wake_up_process(sig->group_exec_task);
173 122
174 if (tsk == sig->curr_target) 123 if (tsk == sig->curr_target)
175 sig->curr_target = nex 124 sig->curr_target = next_thread(tsk);
176 } 125 }
177 126
178 add_device_randomness((const void*) &t 127 add_device_randomness((const void*) &tsk->se.sum_exec_runtime,
179 sizeof(unsigned 128 sizeof(unsigned long long));
180 129
181 /* 130 /*
182 * Accumulate here the counters for al 131 * Accumulate here the counters for all threads as they die. We could
183 * skip the group leader because it is 132 * skip the group leader because it is the last user of signal_struct,
184 * but we want to avoid the race with 133 * but we want to avoid the race with thread_group_cputime() which can
185 * see the empty ->thread_head list. 134 * see the empty ->thread_head list.
186 */ 135 */
187 task_cputime(tsk, &utime, &stime); 136 task_cputime(tsk, &utime, &stime);
188 write_seqlock(&sig->stats_lock); 137 write_seqlock(&sig->stats_lock);
189 sig->utime += utime; 138 sig->utime += utime;
190 sig->stime += stime; 139 sig->stime += stime;
191 sig->gtime += task_gtime(tsk); 140 sig->gtime += task_gtime(tsk);
192 sig->min_flt += tsk->min_flt; 141 sig->min_flt += tsk->min_flt;
193 sig->maj_flt += tsk->maj_flt; 142 sig->maj_flt += tsk->maj_flt;
194 sig->nvcsw += tsk->nvcsw; 143 sig->nvcsw += tsk->nvcsw;
195 sig->nivcsw += tsk->nivcsw; 144 sig->nivcsw += tsk->nivcsw;
196 sig->inblock += task_io_get_inblock(ts 145 sig->inblock += task_io_get_inblock(tsk);
197 sig->oublock += task_io_get_oublock(ts 146 sig->oublock += task_io_get_oublock(tsk);
198 task_io_accounting_add(&sig->ioac, &ts 147 task_io_accounting_add(&sig->ioac, &tsk->ioac);
199 sig->sum_sched_runtime += tsk->se.sum_ 148 sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
200 sig->nr_threads--; 149 sig->nr_threads--;
201 __unhash_process(tsk, group_dead); 150 __unhash_process(tsk, group_dead);
202 write_sequnlock(&sig->stats_lock); 151 write_sequnlock(&sig->stats_lock);
203 152
204 /* 153 /*
205 * Do this under ->siglock, we can rac 154 * Do this under ->siglock, we can race with another thread
206 * doing sigqueue_free() if we have SI 155 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
207 */ 156 */
208 flush_sigqueue(&tsk->pending); 157 flush_sigqueue(&tsk->pending);
209 tsk->sighand = NULL; 158 tsk->sighand = NULL;
210 spin_unlock(&sighand->siglock); 159 spin_unlock(&sighand->siglock);
211 160
212 __cleanup_sighand(sighand); 161 __cleanup_sighand(sighand);
213 clear_tsk_thread_flag(tsk, TIF_SIGPEND 162 clear_tsk_thread_flag(tsk, TIF_SIGPENDING);
214 if (group_dead) { 163 if (group_dead) {
215 flush_sigqueue(&sig->shared_pe 164 flush_sigqueue(&sig->shared_pending);
216 tty_kref_put(tty); 165 tty_kref_put(tty);
217 } 166 }
218 } 167 }
219 168
220 static void delayed_put_task_struct(struct rcu 169 static void delayed_put_task_struct(struct rcu_head *rhp)
221 { 170 {
222 struct task_struct *tsk = container_of 171 struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
223 172
224 kprobe_flush_task(tsk); 173 kprobe_flush_task(tsk);
225 rethook_flush_task(tsk); 174 rethook_flush_task(tsk);
226 perf_event_delayed_put(tsk); 175 perf_event_delayed_put(tsk);
227 trace_sched_process_free(tsk); 176 trace_sched_process_free(tsk);
228 put_task_struct(tsk); 177 put_task_struct(tsk);
229 } 178 }
230 179
231 void put_task_struct_rcu_user(struct task_stru 180 void put_task_struct_rcu_user(struct task_struct *task)
232 { 181 {
233 if (refcount_dec_and_test(&task->rcu_u 182 if (refcount_dec_and_test(&task->rcu_users))
234 call_rcu(&task->rcu, delayed_p 183 call_rcu(&task->rcu, delayed_put_task_struct);
235 } 184 }
236 185
237 void __weak release_thread(struct task_struct <<
238 { <<
239 } <<
240 <<
241 void release_task(struct task_struct *p) 186 void release_task(struct task_struct *p)
242 { 187 {
243 struct task_struct *leader; 188 struct task_struct *leader;
244 struct pid *thread_pid; 189 struct pid *thread_pid;
245 int zap_leader; 190 int zap_leader;
246 repeat: 191 repeat:
247 /* don't need to get the RCU readlock 192 /* don't need to get the RCU readlock here - the process is dead and
248 * can't be modifying its own credenti 193 * can't be modifying its own credentials. But shut RCU-lockdep up */
249 rcu_read_lock(); 194 rcu_read_lock();
250 dec_rlimit_ucounts(task_ucounts(p), UC 195 dec_rlimit_ucounts(task_ucounts(p), UCOUNT_RLIMIT_NPROC, 1);
251 rcu_read_unlock(); 196 rcu_read_unlock();
252 197
253 cgroup_release(p); 198 cgroup_release(p);
254 199
255 write_lock_irq(&tasklist_lock); 200 write_lock_irq(&tasklist_lock);
256 ptrace_release_task(p); 201 ptrace_release_task(p);
257 thread_pid = get_pid(p->thread_pid); 202 thread_pid = get_pid(p->thread_pid);
258 __exit_signal(p); 203 __exit_signal(p);
259 204
260 /* 205 /*
261 * If we are the last non-leader membe 206 * If we are the last non-leader member of the thread
262 * group, and the leader is zombie, th 207 * group, and the leader is zombie, then notify the
263 * group leader's parent process. (if 208 * group leader's parent process. (if it wants notification.)
264 */ 209 */
265 zap_leader = 0; 210 zap_leader = 0;
266 leader = p->group_leader; 211 leader = p->group_leader;
267 if (leader != p && thread_group_empty( 212 if (leader != p && thread_group_empty(leader)
268 && leader->exit_state 213 && leader->exit_state == EXIT_ZOMBIE) {
269 /* 214 /*
270 * If we were the last child t 215 * If we were the last child thread and the leader has
271 * exited already, and the lea 216 * exited already, and the leader's parent ignores SIGCHLD,
272 * then we are the one who sho 217 * then we are the one who should release the leader.
273 */ 218 */
274 zap_leader = do_notify_parent( 219 zap_leader = do_notify_parent(leader, leader->exit_signal);
275 if (zap_leader) 220 if (zap_leader)
276 leader->exit_state = E 221 leader->exit_state = EXIT_DEAD;
277 } 222 }
278 223
279 write_unlock_irq(&tasklist_lock); 224 write_unlock_irq(&tasklist_lock);
>> 225 seccomp_filter_release(p);
280 proc_flush_pid(thread_pid); 226 proc_flush_pid(thread_pid);
281 put_pid(thread_pid); 227 put_pid(thread_pid);
282 release_thread(p); 228 release_thread(p);
283 put_task_struct_rcu_user(p); 229 put_task_struct_rcu_user(p);
284 230
285 p = leader; 231 p = leader;
286 if (unlikely(zap_leader)) 232 if (unlikely(zap_leader))
287 goto repeat; 233 goto repeat;
288 } 234 }
289 235
290 int rcuwait_wake_up(struct rcuwait *w) 236 int rcuwait_wake_up(struct rcuwait *w)
291 { 237 {
292 int ret = 0; 238 int ret = 0;
293 struct task_struct *task; 239 struct task_struct *task;
294 240
295 rcu_read_lock(); 241 rcu_read_lock();
296 242
297 /* 243 /*
298 * Order condition vs @task, such that 244 * Order condition vs @task, such that everything prior to the load
299 * of @task is visible. This is the co 245 * of @task is visible. This is the condition as to why the user called
300 * rcuwait_wake() in the first place. 246 * rcuwait_wake() in the first place. Pairs with set_current_state()
301 * barrier (A) in rcuwait_wait_event() 247 * barrier (A) in rcuwait_wait_event().
302 * 248 *
303 * WAIT WAKE 249 * WAIT WAKE
304 * [S] tsk = current [S] cond = t 250 * [S] tsk = current [S] cond = true
305 * MB (A) MB (B) 251 * MB (A) MB (B)
306 * [L] cond [L] tsk 252 * [L] cond [L] tsk
307 */ 253 */
308 smp_mb(); /* (B) */ 254 smp_mb(); /* (B) */
309 255
310 task = rcu_dereference(w->task); 256 task = rcu_dereference(w->task);
311 if (task) 257 if (task)
312 ret = wake_up_process(task); 258 ret = wake_up_process(task);
313 rcu_read_unlock(); 259 rcu_read_unlock();
314 260
315 return ret; 261 return ret;
316 } 262 }
317 EXPORT_SYMBOL_GPL(rcuwait_wake_up); 263 EXPORT_SYMBOL_GPL(rcuwait_wake_up);
318 264
319 /* 265 /*
320 * Determine if a process group is "orphaned", 266 * Determine if a process group is "orphaned", according to the POSIX
321 * definition in 2.2.2.52. Orphaned process g 267 * definition in 2.2.2.52. Orphaned process groups are not to be affected
322 * by terminal-generated stop signals. Newly 268 * by terminal-generated stop signals. Newly orphaned process groups are
323 * to receive a SIGHUP and a SIGCONT. 269 * to receive a SIGHUP and a SIGCONT.
324 * 270 *
325 * "I ask you, have you ever known what it is 271 * "I ask you, have you ever known what it is to be an orphan?"
326 */ 272 */
327 static int will_become_orphaned_pgrp(struct pi 273 static int will_become_orphaned_pgrp(struct pid *pgrp,
328 struct 274 struct task_struct *ignored_task)
329 { 275 {
330 struct task_struct *p; 276 struct task_struct *p;
331 277
332 do_each_pid_task(pgrp, PIDTYPE_PGID, p 278 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
333 if ((p == ignored_task) || 279 if ((p == ignored_task) ||
334 (p->exit_state && thread_g 280 (p->exit_state && thread_group_empty(p)) ||
335 is_global_init(p->real_par 281 is_global_init(p->real_parent))
336 continue; 282 continue;
337 283
338 if (task_pgrp(p->real_parent) 284 if (task_pgrp(p->real_parent) != pgrp &&
339 task_session(p->real_paren 285 task_session(p->real_parent) == task_session(p))
340 return 0; 286 return 0;
341 } while_each_pid_task(pgrp, PIDTYPE_PG 287 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
342 288
343 return 1; 289 return 1;
344 } 290 }
345 291
346 int is_current_pgrp_orphaned(void) 292 int is_current_pgrp_orphaned(void)
347 { 293 {
348 int retval; 294 int retval;
349 295
350 read_lock(&tasklist_lock); 296 read_lock(&tasklist_lock);
351 retval = will_become_orphaned_pgrp(tas 297 retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
352 read_unlock(&tasklist_lock); 298 read_unlock(&tasklist_lock);
353 299
354 return retval; 300 return retval;
355 } 301 }
356 302
357 static bool has_stopped_jobs(struct pid *pgrp) 303 static bool has_stopped_jobs(struct pid *pgrp)
358 { 304 {
359 struct task_struct *p; 305 struct task_struct *p;
360 306
361 do_each_pid_task(pgrp, PIDTYPE_PGID, p 307 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
362 if (p->signal->flags & SIGNAL_ 308 if (p->signal->flags & SIGNAL_STOP_STOPPED)
363 return true; 309 return true;
364 } while_each_pid_task(pgrp, PIDTYPE_PG 310 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
365 311
366 return false; 312 return false;
367 } 313 }
368 314
369 /* 315 /*
370 * Check to see if any process groups have bec 316 * Check to see if any process groups have become orphaned as
371 * a result of our exiting, and if they have a 317 * a result of our exiting, and if they have any stopped jobs,
372 * send them a SIGHUP and then a SIGCONT. (POS 318 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
373 */ 319 */
374 static void 320 static void
375 kill_orphaned_pgrp(struct task_struct *tsk, st 321 kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
376 { 322 {
377 struct pid *pgrp = task_pgrp(tsk); 323 struct pid *pgrp = task_pgrp(tsk);
378 struct task_struct *ignored_task = tsk 324 struct task_struct *ignored_task = tsk;
379 325
380 if (!parent) 326 if (!parent)
381 /* exit: our father is in a di 327 /* exit: our father is in a different pgrp than
382 * we are and we were the only 328 * we are and we were the only connection outside.
383 */ 329 */
384 parent = tsk->real_parent; 330 parent = tsk->real_parent;
385 else 331 else
386 /* reparent: our child is in a 332 /* reparent: our child is in a different pgrp than
387 * we are, and it was the only 333 * we are, and it was the only connection outside.
388 */ 334 */
389 ignored_task = NULL; 335 ignored_task = NULL;
390 336
391 if (task_pgrp(parent) != pgrp && 337 if (task_pgrp(parent) != pgrp &&
392 task_session(parent) == task_sessi 338 task_session(parent) == task_session(tsk) &&
393 will_become_orphaned_pgrp(pgrp, ig 339 will_become_orphaned_pgrp(pgrp, ignored_task) &&
394 has_stopped_jobs(pgrp)) { 340 has_stopped_jobs(pgrp)) {
395 __kill_pgrp_info(SIGHUP, SEND_ 341 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
396 __kill_pgrp_info(SIGCONT, SEND 342 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
397 } 343 }
398 } 344 }
399 345
400 static void coredump_task_exit(struct task_str 346 static void coredump_task_exit(struct task_struct *tsk)
401 { 347 {
402 struct core_state *core_state; 348 struct core_state *core_state;
403 349
404 /* 350 /*
405 * Serialize with any possible pending 351 * Serialize with any possible pending coredump.
406 * We must hold siglock around checkin 352 * We must hold siglock around checking core_state
407 * and setting PF_POSTCOREDUMP. The c 353 * and setting PF_POSTCOREDUMP. The core-inducing thread
408 * will increment ->nr_threads for eac 354 * will increment ->nr_threads for each thread in the
409 * group without PF_POSTCOREDUMP set. 355 * group without PF_POSTCOREDUMP set.
410 */ 356 */
411 spin_lock_irq(&tsk->sighand->siglock); 357 spin_lock_irq(&tsk->sighand->siglock);
412 tsk->flags |= PF_POSTCOREDUMP; 358 tsk->flags |= PF_POSTCOREDUMP;
413 core_state = tsk->signal->core_state; 359 core_state = tsk->signal->core_state;
414 spin_unlock_irq(&tsk->sighand->siglock 360 spin_unlock_irq(&tsk->sighand->siglock);
415 if (core_state) { 361 if (core_state) {
416 struct core_thread self; 362 struct core_thread self;
417 363
418 self.task = current; 364 self.task = current;
419 if (self.task->flags & PF_SIGN 365 if (self.task->flags & PF_SIGNALED)
420 self.next = xchg(&core 366 self.next = xchg(&core_state->dumper.next, &self);
421 else 367 else
422 self.task = NULL; 368 self.task = NULL;
423 /* 369 /*
424 * Implies mb(), the result of 370 * Implies mb(), the result of xchg() must be visible
425 * to core_state->dumper. 371 * to core_state->dumper.
426 */ 372 */
427 if (atomic_dec_and_test(&core_ 373 if (atomic_dec_and_test(&core_state->nr_threads))
428 complete(&core_state-> 374 complete(&core_state->startup);
429 375
430 for (;;) { 376 for (;;) {
431 set_current_state(TASK !! 377 set_current_state(TASK_UNINTERRUPTIBLE);
432 if (!self.task) /* see 378 if (!self.task) /* see coredump_finish() */
433 break; 379 break;
434 schedule(); !! 380 freezable_schedule();
435 } 381 }
436 __set_current_state(TASK_RUNNI 382 __set_current_state(TASK_RUNNING);
437 } 383 }
438 } 384 }
439 385
440 #ifdef CONFIG_MEMCG 386 #ifdef CONFIG_MEMCG
441 /* drops tasklist_lock if succeeds */ <<
442 static bool __try_to_set_owner(struct task_str <<
443 { <<
444 bool ret = false; <<
445 <<
446 task_lock(tsk); <<
447 if (likely(tsk->mm == mm)) { <<
448 /* tsk can't pass exit_mm/exec <<
449 read_unlock(&tasklist_lock); <<
450 WRITE_ONCE(mm->owner, tsk); <<
451 lru_gen_migrate_mm(mm); <<
452 ret = true; <<
453 } <<
454 task_unlock(tsk); <<
455 return ret; <<
456 } <<
457 <<
458 static bool try_to_set_owner(struct task_struc <<
459 { <<
460 struct task_struct *t; <<
461 <<
462 for_each_thread(g, t) { <<
463 struct mm_struct *t_mm = READ_ <<
464 if (t_mm == mm) { <<
465 if (__try_to_set_owner <<
466 return true; <<
467 } else if (t_mm) <<
468 break; <<
469 } <<
470 <<
471 return false; <<
472 } <<
473 <<
474 /* 387 /*
475 * A task is exiting. If it owned this mm, f 388 * A task is exiting. If it owned this mm, find a new owner for the mm.
476 */ 389 */
477 void mm_update_next_owner(struct mm_struct *mm 390 void mm_update_next_owner(struct mm_struct *mm)
478 { 391 {
479 struct task_struct *g, *p = current; !! 392 struct task_struct *c, *g, *p = current;
480 393
>> 394 retry:
481 /* 395 /*
482 * If the exiting or execing task is n 396 * If the exiting or execing task is not the owner, it's
483 * someone else's problem. 397 * someone else's problem.
484 */ 398 */
485 if (mm->owner != p) 399 if (mm->owner != p)
486 return; 400 return;
487 /* 401 /*
488 * The current owner is exiting/execin 402 * The current owner is exiting/execing and there are no other
489 * candidates. Do not leave the mm po 403 * candidates. Do not leave the mm pointing to a possibly
490 * freed task structure. 404 * freed task structure.
491 */ 405 */
492 if (atomic_read(&mm->mm_users) <= 1) { 406 if (atomic_read(&mm->mm_users) <= 1) {
493 WRITE_ONCE(mm->owner, NULL); 407 WRITE_ONCE(mm->owner, NULL);
494 return; 408 return;
495 } 409 }
496 410
497 read_lock(&tasklist_lock); 411 read_lock(&tasklist_lock);
498 /* 412 /*
499 * Search in the children 413 * Search in the children
500 */ 414 */
501 list_for_each_entry(g, &p->children, s !! 415 list_for_each_entry(c, &p->children, sibling) {
502 if (try_to_set_owner(g, mm)) !! 416 if (c->mm == mm)
503 goto ret; !! 417 goto assign_new_owner;
504 } 418 }
>> 419
505 /* 420 /*
506 * Search in the siblings 421 * Search in the siblings
507 */ 422 */
508 list_for_each_entry(g, &p->real_parent !! 423 list_for_each_entry(c, &p->real_parent->children, sibling) {
509 if (try_to_set_owner(g, mm)) !! 424 if (c->mm == mm)
510 goto ret; !! 425 goto assign_new_owner;
511 } 426 }
>> 427
512 /* 428 /*
513 * Search through everything else, we 429 * Search through everything else, we should not get here often.
514 */ 430 */
515 for_each_process(g) { 431 for_each_process(g) {
516 if (atomic_read(&mm->mm_users) <<
517 break; <<
518 if (g->flags & PF_KTHREAD) 432 if (g->flags & PF_KTHREAD)
519 continue; 433 continue;
520 if (try_to_set_owner(g, mm)) !! 434 for_each_thread(g, c) {
521 goto ret; !! 435 if (c->mm == mm)
>> 436 goto assign_new_owner;
>> 437 if (c->mm)
>> 438 break;
>> 439 }
522 } 440 }
523 read_unlock(&tasklist_lock); 441 read_unlock(&tasklist_lock);
524 /* 442 /*
525 * We found no owner yet mm_users > 1: 443 * We found no owner yet mm_users > 1: this implies that we are
526 * most likely racing with swapoff (tr 444 * most likely racing with swapoff (try_to_unuse()) or /proc or
527 * ptrace or page migration (get_task_ 445 * ptrace or page migration (get_task_mm()). Mark owner as NULL.
528 */ 446 */
529 WRITE_ONCE(mm->owner, NULL); 447 WRITE_ONCE(mm->owner, NULL);
530 ret: <<
531 return; 448 return;
532 449
>> 450 assign_new_owner:
>> 451 BUG_ON(c == p);
>> 452 get_task_struct(c);
>> 453 /*
>> 454 * The task_lock protects c->mm from changing.
>> 455 * We always want mm->owner->mm == mm
>> 456 */
>> 457 task_lock(c);
>> 458 /*
>> 459 * Delay read_unlock() till we have the task_lock()
>> 460 * to ensure that c does not slip away underneath us
>> 461 */
>> 462 read_unlock(&tasklist_lock);
>> 463 if (c->mm != mm) {
>> 464 task_unlock(c);
>> 465 put_task_struct(c);
>> 466 goto retry;
>> 467 }
>> 468 WRITE_ONCE(mm->owner, c);
>> 469 task_unlock(c);
>> 470 put_task_struct(c);
533 } 471 }
534 #endif /* CONFIG_MEMCG */ 472 #endif /* CONFIG_MEMCG */
535 473
536 /* 474 /*
537 * Turn us into a lazy TLB process if we 475 * Turn us into a lazy TLB process if we
538 * aren't already.. 476 * aren't already..
539 */ 477 */
540 static void exit_mm(void) 478 static void exit_mm(void)
541 { 479 {
542 struct mm_struct *mm = current->mm; 480 struct mm_struct *mm = current->mm;
543 481
544 exit_mm_release(current, mm); 482 exit_mm_release(current, mm);
545 if (!mm) 483 if (!mm)
546 return; 484 return;
>> 485 sync_mm_rss(mm);
547 mmap_read_lock(mm); 486 mmap_read_lock(mm);
548 mmgrab_lazy_tlb(mm); !! 487 mmgrab(mm);
549 BUG_ON(mm != current->active_mm); 488 BUG_ON(mm != current->active_mm);
550 /* more a memory barrier than a real l 489 /* more a memory barrier than a real lock */
551 task_lock(current); 490 task_lock(current);
552 /* 491 /*
553 * When a thread stops operating on an 492 * When a thread stops operating on an address space, the loop
554 * in membarrier_private_expedited() m 493 * in membarrier_private_expedited() may not observe that
555 * tsk->mm, and the loop in membarrier 494 * tsk->mm, and the loop in membarrier_global_expedited() may
556 * not observe a MEMBARRIER_STATE_GLOB 495 * not observe a MEMBARRIER_STATE_GLOBAL_EXPEDITED
557 * rq->membarrier_state, so those woul 496 * rq->membarrier_state, so those would not issue an IPI.
558 * Membarrier requires a memory barrie 497 * Membarrier requires a memory barrier after accessing
559 * user-space memory, before clearing 498 * user-space memory, before clearing tsk->mm or the
560 * rq->membarrier_state. 499 * rq->membarrier_state.
561 */ 500 */
562 smp_mb__after_spinlock(); 501 smp_mb__after_spinlock();
563 local_irq_disable(); 502 local_irq_disable();
564 current->mm = NULL; 503 current->mm = NULL;
565 membarrier_update_current_mm(NULL); 504 membarrier_update_current_mm(NULL);
566 enter_lazy_tlb(mm, current); 505 enter_lazy_tlb(mm, current);
567 local_irq_enable(); 506 local_irq_enable();
568 task_unlock(current); 507 task_unlock(current);
569 mmap_read_unlock(mm); 508 mmap_read_unlock(mm);
570 mm_update_next_owner(mm); 509 mm_update_next_owner(mm);
571 mmput(mm); 510 mmput(mm);
572 if (test_thread_flag(TIF_MEMDIE)) 511 if (test_thread_flag(TIF_MEMDIE))
573 exit_oom_victim(); 512 exit_oom_victim();
574 } 513 }
575 514
576 static struct task_struct *find_alive_thread(s 515 static struct task_struct *find_alive_thread(struct task_struct *p)
577 { 516 {
578 struct task_struct *t; 517 struct task_struct *t;
579 518
580 for_each_thread(p, t) { 519 for_each_thread(p, t) {
581 if (!(t->flags & PF_EXITING)) 520 if (!(t->flags & PF_EXITING))
582 return t; 521 return t;
583 } 522 }
584 return NULL; 523 return NULL;
585 } 524 }
586 525
587 static struct task_struct *find_child_reaper(s 526 static struct task_struct *find_child_reaper(struct task_struct *father,
588 527 struct list_head *dead)
589 __releases(&tasklist_lock) 528 __releases(&tasklist_lock)
590 __acquires(&tasklist_lock) 529 __acquires(&tasklist_lock)
591 { 530 {
592 struct pid_namespace *pid_ns = task_ac 531 struct pid_namespace *pid_ns = task_active_pid_ns(father);
593 struct task_struct *reaper = pid_ns->c 532 struct task_struct *reaper = pid_ns->child_reaper;
594 struct task_struct *p, *n; 533 struct task_struct *p, *n;
595 534
596 if (likely(reaper != father)) 535 if (likely(reaper != father))
597 return reaper; 536 return reaper;
598 537
599 reaper = find_alive_thread(father); 538 reaper = find_alive_thread(father);
600 if (reaper) { 539 if (reaper) {
601 pid_ns->child_reaper = reaper; 540 pid_ns->child_reaper = reaper;
602 return reaper; 541 return reaper;
603 } 542 }
604 543
605 write_unlock_irq(&tasklist_lock); 544 write_unlock_irq(&tasklist_lock);
606 545
607 list_for_each_entry_safe(p, n, dead, p 546 list_for_each_entry_safe(p, n, dead, ptrace_entry) {
608 list_del_init(&p->ptrace_entry 547 list_del_init(&p->ptrace_entry);
609 release_task(p); 548 release_task(p);
610 } 549 }
611 550
612 zap_pid_ns_processes(pid_ns); 551 zap_pid_ns_processes(pid_ns);
613 write_lock_irq(&tasklist_lock); 552 write_lock_irq(&tasklist_lock);
614 553
615 return father; 554 return father;
616 } 555 }
617 556
618 /* 557 /*
619 * When we die, we re-parent all our children, 558 * When we die, we re-parent all our children, and try to:
620 * 1. give them to another thread in our threa 559 * 1. give them to another thread in our thread group, if such a member exists
621 * 2. give it to the first ancestor process wh 560 * 2. give it to the first ancestor process which prctl'd itself as a
622 * child_subreaper for its children (like a 561 * child_subreaper for its children (like a service manager)
623 * 3. give it to the init process (PID 1) in o 562 * 3. give it to the init process (PID 1) in our pid namespace
624 */ 563 */
625 static struct task_struct *find_new_reaper(str 564 static struct task_struct *find_new_reaper(struct task_struct *father,
626 str 565 struct task_struct *child_reaper)
627 { 566 {
628 struct task_struct *thread, *reaper; 567 struct task_struct *thread, *reaper;
629 568
630 thread = find_alive_thread(father); 569 thread = find_alive_thread(father);
631 if (thread) 570 if (thread)
632 return thread; 571 return thread;
633 572
634 if (father->signal->has_child_subreape 573 if (father->signal->has_child_subreaper) {
635 unsigned int ns_level = task_p 574 unsigned int ns_level = task_pid(father)->level;
636 /* 575 /*
637 * Find the first ->is_child_s 576 * Find the first ->is_child_subreaper ancestor in our pid_ns.
638 * We can't check reaper != ch 577 * We can't check reaper != child_reaper to ensure we do not
639 * cross the namespaces, the e 578 * cross the namespaces, the exiting parent could be injected
640 * by setns() + fork(). 579 * by setns() + fork().
641 * We check pid->level, this i 580 * We check pid->level, this is slightly more efficient than
642 * task_active_pid_ns(reaper) 581 * task_active_pid_ns(reaper) != task_active_pid_ns(father).
643 */ 582 */
644 for (reaper = father->real_par 583 for (reaper = father->real_parent;
645 task_pid(reaper)->level = 584 task_pid(reaper)->level == ns_level;
646 reaper = reaper->real_par 585 reaper = reaper->real_parent) {
647 if (reaper == &init_ta 586 if (reaper == &init_task)
648 break; 587 break;
649 if (!reaper->signal->i 588 if (!reaper->signal->is_child_subreaper)
650 continue; 589 continue;
651 thread = find_alive_th 590 thread = find_alive_thread(reaper);
652 if (thread) 591 if (thread)
653 return thread; 592 return thread;
654 } 593 }
655 } 594 }
656 595
657 return child_reaper; 596 return child_reaper;
658 } 597 }
659 598
660 /* 599 /*
661 * Any that need to be release_task'd are put o 600 * Any that need to be release_task'd are put on the @dead list.
662 */ 601 */
663 static void reparent_leader(struct task_struct 602 static void reparent_leader(struct task_struct *father, struct task_struct *p,
664 struct list_he 603 struct list_head *dead)
665 { 604 {
666 if (unlikely(p->exit_state == EXIT_DEA 605 if (unlikely(p->exit_state == EXIT_DEAD))
667 return; 606 return;
668 607
669 /* We don't want people slaying init. 608 /* We don't want people slaying init. */
670 p->exit_signal = SIGCHLD; 609 p->exit_signal = SIGCHLD;
671 610
672 /* If it has exited notify the new par 611 /* If it has exited notify the new parent about this child's death. */
673 if (!p->ptrace && 612 if (!p->ptrace &&
674 p->exit_state == EXIT_ZOMBIE && th 613 p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) {
675 if (do_notify_parent(p, p->exi 614 if (do_notify_parent(p, p->exit_signal)) {
676 p->exit_state = EXIT_D 615 p->exit_state = EXIT_DEAD;
677 list_add(&p->ptrace_en 616 list_add(&p->ptrace_entry, dead);
678 } 617 }
679 } 618 }
680 619
681 kill_orphaned_pgrp(p, father); 620 kill_orphaned_pgrp(p, father);
682 } 621 }
683 622
684 /* 623 /*
685 * This does two things: 624 * This does two things:
686 * 625 *
687 * A. Make init inherit all the child process 626 * A. Make init inherit all the child processes
688 * B. Check to see if any process groups have 627 * B. Check to see if any process groups have become orphaned
689 * as a result of our exiting, and if the 628 * as a result of our exiting, and if they have any stopped
690 * jobs, send them a SIGHUP and then a SI 629 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
691 */ 630 */
692 static void forget_original_parent(struct task 631 static void forget_original_parent(struct task_struct *father,
693 struct 632 struct list_head *dead)
694 { 633 {
695 struct task_struct *p, *t, *reaper; 634 struct task_struct *p, *t, *reaper;
696 635
697 if (unlikely(!list_empty(&father->ptra 636 if (unlikely(!list_empty(&father->ptraced)))
698 exit_ptrace(father, dead); 637 exit_ptrace(father, dead);
699 638
700 /* Can drop and reacquire tasklist_loc 639 /* Can drop and reacquire tasklist_lock */
701 reaper = find_child_reaper(father, dea 640 reaper = find_child_reaper(father, dead);
702 if (list_empty(&father->children)) 641 if (list_empty(&father->children))
703 return; 642 return;
704 643
705 reaper = find_new_reaper(father, reape 644 reaper = find_new_reaper(father, reaper);
706 list_for_each_entry(p, &father->childr 645 list_for_each_entry(p, &father->children, sibling) {
707 for_each_thread(p, t) { 646 for_each_thread(p, t) {
708 RCU_INIT_POINTER(t->re 647 RCU_INIT_POINTER(t->real_parent, reaper);
709 BUG_ON((!t->ptrace) != 648 BUG_ON((!t->ptrace) != (rcu_access_pointer(t->parent) == father));
710 if (likely(!t->ptrace) 649 if (likely(!t->ptrace))
711 t->parent = t- 650 t->parent = t->real_parent;
712 if (t->pdeath_signal) 651 if (t->pdeath_signal)
713 group_send_sig 652 group_send_sig_info(t->pdeath_signal,
714 653 SEND_SIG_NOINFO, t,
715 654 PIDTYPE_TGID);
716 } 655 }
717 /* 656 /*
718 * If this is a threaded repar 657 * If this is a threaded reparent there is no need to
719 * notify anyone anything has 658 * notify anyone anything has happened.
720 */ 659 */
721 if (!same_thread_group(reaper, 660 if (!same_thread_group(reaper, father))
722 reparent_leader(father 661 reparent_leader(father, p, dead);
723 } 662 }
724 list_splice_tail_init(&father->childre 663 list_splice_tail_init(&father->children, &reaper->children);
725 } 664 }
726 665
727 /* 666 /*
728 * Send signals to all our closest relatives s 667 * Send signals to all our closest relatives so that they know
729 * to properly mourn us.. 668 * to properly mourn us..
730 */ 669 */
731 static void exit_notify(struct task_struct *ts 670 static void exit_notify(struct task_struct *tsk, int group_dead)
732 { 671 {
733 bool autoreap; 672 bool autoreap;
734 struct task_struct *p, *n; 673 struct task_struct *p, *n;
735 LIST_HEAD(dead); 674 LIST_HEAD(dead);
736 675
737 write_lock_irq(&tasklist_lock); 676 write_lock_irq(&tasklist_lock);
738 forget_original_parent(tsk, &dead); 677 forget_original_parent(tsk, &dead);
739 678
740 if (group_dead) 679 if (group_dead)
741 kill_orphaned_pgrp(tsk->group_ 680 kill_orphaned_pgrp(tsk->group_leader, NULL);
742 681
743 tsk->exit_state = EXIT_ZOMBIE; 682 tsk->exit_state = EXIT_ZOMBIE;
744 /* <<
745 * sub-thread or delay_group_leader(), <<
746 * PIDFD_THREAD waiters. <<
747 */ <<
748 if (!thread_group_empty(tsk)) <<
749 do_notify_pidfd(tsk); <<
750 <<
751 if (unlikely(tsk->ptrace)) { 683 if (unlikely(tsk->ptrace)) {
752 int sig = thread_group_leader( 684 int sig = thread_group_leader(tsk) &&
753 thread_group_e 685 thread_group_empty(tsk) &&
754 !ptrace_repare 686 !ptrace_reparented(tsk) ?
755 tsk->exit_signal : SIG 687 tsk->exit_signal : SIGCHLD;
756 autoreap = do_notify_parent(ts 688 autoreap = do_notify_parent(tsk, sig);
757 } else if (thread_group_leader(tsk)) { 689 } else if (thread_group_leader(tsk)) {
758 autoreap = thread_group_empty( 690 autoreap = thread_group_empty(tsk) &&
759 do_notify_parent(tsk, 691 do_notify_parent(tsk, tsk->exit_signal);
760 } else { 692 } else {
761 autoreap = true; 693 autoreap = true;
762 } 694 }
763 695
764 if (autoreap) { 696 if (autoreap) {
765 tsk->exit_state = EXIT_DEAD; 697 tsk->exit_state = EXIT_DEAD;
766 list_add(&tsk->ptrace_entry, & 698 list_add(&tsk->ptrace_entry, &dead);
767 } 699 }
768 700
769 /* mt-exec, de_thread() is waiting for 701 /* mt-exec, de_thread() is waiting for group leader */
770 if (unlikely(tsk->signal->notify_count 702 if (unlikely(tsk->signal->notify_count < 0))
771 wake_up_process(tsk->signal->g 703 wake_up_process(tsk->signal->group_exec_task);
772 write_unlock_irq(&tasklist_lock); 704 write_unlock_irq(&tasklist_lock);
773 705
774 list_for_each_entry_safe(p, n, &dead, 706 list_for_each_entry_safe(p, n, &dead, ptrace_entry) {
775 list_del_init(&p->ptrace_entry 707 list_del_init(&p->ptrace_entry);
776 release_task(p); 708 release_task(p);
777 } 709 }
778 } 710 }
779 711
780 #ifdef CONFIG_DEBUG_STACK_USAGE 712 #ifdef CONFIG_DEBUG_STACK_USAGE
781 unsigned long stack_not_used(struct task_struc <<
782 { <<
783 unsigned long *n = end_of_stack(p); <<
784 <<
785 do { /* Skip over canary */ <<
786 # ifdef CONFIG_STACK_GROWSUP <<
787 n--; <<
788 # else <<
789 n++; <<
790 # endif <<
791 } while (!*n); <<
792 <<
793 # ifdef CONFIG_STACK_GROWSUP <<
794 return (unsigned long)end_of_stack(p) <<
795 # else <<
796 return (unsigned long)n - (unsigned lo <<
797 # endif <<
798 } <<
799 <<
800 /* Count the maximum pages reached in kernel s <<
801 static inline void kstack_histogram(unsigned l <<
802 { <<
803 #ifdef CONFIG_VM_EVENT_COUNTERS <<
804 if (used_stack <= 1024) <<
805 count_vm_event(KSTACK_1K); <<
806 #if THREAD_SIZE > 1024 <<
807 else if (used_stack <= 2048) <<
808 count_vm_event(KSTACK_2K); <<
809 #endif <<
810 #if THREAD_SIZE > 2048 <<
811 else if (used_stack <= 4096) <<
812 count_vm_event(KSTACK_4K); <<
813 #endif <<
814 #if THREAD_SIZE > 4096 <<
815 else if (used_stack <= 8192) <<
816 count_vm_event(KSTACK_8K); <<
817 #endif <<
818 #if THREAD_SIZE > 8192 <<
819 else if (used_stack <= 16384) <<
820 count_vm_event(KSTACK_16K); <<
821 #endif <<
822 #if THREAD_SIZE > 16384 <<
823 else if (used_stack <= 32768) <<
824 count_vm_event(KSTACK_32K); <<
825 #endif <<
826 #if THREAD_SIZE > 32768 <<
827 else if (used_stack <= 65536) <<
828 count_vm_event(KSTACK_64K); <<
829 #endif <<
830 #if THREAD_SIZE > 65536 <<
831 else <<
832 count_vm_event(KSTACK_REST); <<
833 #endif <<
834 #endif /* CONFIG_VM_EVENT_COUNTERS */ <<
835 } <<
836 <<
837 static void check_stack_usage(void) 713 static void check_stack_usage(void)
838 { 714 {
839 static DEFINE_SPINLOCK(low_water_lock) 715 static DEFINE_SPINLOCK(low_water_lock);
840 static int lowest_to_date = THREAD_SIZ 716 static int lowest_to_date = THREAD_SIZE;
841 unsigned long free; 717 unsigned long free;
842 718
843 free = stack_not_used(current); 719 free = stack_not_used(current);
844 kstack_histogram(THREAD_SIZE - free); <<
845 720
846 if (free >= lowest_to_date) 721 if (free >= lowest_to_date)
847 return; 722 return;
848 723
849 spin_lock(&low_water_lock); 724 spin_lock(&low_water_lock);
850 if (free < lowest_to_date) { 725 if (free < lowest_to_date) {
851 pr_info("%s (%d) used greatest 726 pr_info("%s (%d) used greatest stack depth: %lu bytes left\n",
852 current->comm, task_pi 727 current->comm, task_pid_nr(current), free);
853 lowest_to_date = free; 728 lowest_to_date = free;
854 } 729 }
855 spin_unlock(&low_water_lock); 730 spin_unlock(&low_water_lock);
856 } 731 }
857 #else 732 #else
858 static inline void check_stack_usage(void) {} 733 static inline void check_stack_usage(void) {}
859 #endif 734 #endif
860 735
861 static void synchronize_group_exit(struct task <<
862 { <<
863 struct sighand_struct *sighand = tsk-> <<
864 struct signal_struct *signal = tsk->si <<
865 <<
866 spin_lock_irq(&sighand->siglock); <<
867 signal->quick_threads--; <<
868 if ((signal->quick_threads == 0) && <<
869 !(signal->flags & SIGNAL_GROUP_EXI <<
870 signal->flags = SIGNAL_GROUP_E <<
871 signal->group_exit_code = code <<
872 signal->group_stop_count = 0; <<
873 } <<
874 spin_unlock_irq(&sighand->siglock); <<
875 } <<
876 <<
877 void __noreturn do_exit(long code) 736 void __noreturn do_exit(long code)
878 { 737 {
879 struct task_struct *tsk = current; 738 struct task_struct *tsk = current;
880 int group_dead; 739 int group_dead;
881 740
882 WARN_ON(irqs_disabled()); <<
883 <<
884 synchronize_group_exit(tsk, code); <<
885 <<
886 WARN_ON(tsk->plug); 741 WARN_ON(tsk->plug);
887 742
888 kcov_task_exit(tsk); 743 kcov_task_exit(tsk);
889 kmsan_task_exit(tsk); <<
890 744
891 coredump_task_exit(tsk); 745 coredump_task_exit(tsk);
892 ptrace_event(PTRACE_EVENT_EXIT, code); 746 ptrace_event(PTRACE_EVENT_EXIT, code);
893 user_events_exit(tsk); !! 747
>> 748 validate_creds_for_do_exit(tsk);
894 749
895 io_uring_files_cancel(); 750 io_uring_files_cancel();
896 exit_signals(tsk); /* sets PF_EXITING 751 exit_signals(tsk); /* sets PF_EXITING */
897 752
898 seccomp_filter_release(tsk); !! 753 /* sync mm's RSS info before statistics gathering */
899 !! 754 if (tsk->mm)
>> 755 sync_mm_rss(tsk->mm);
900 acct_update_integrals(tsk); 756 acct_update_integrals(tsk);
901 group_dead = atomic_dec_and_test(&tsk- 757 group_dead = atomic_dec_and_test(&tsk->signal->live);
902 if (group_dead) { 758 if (group_dead) {
903 /* 759 /*
904 * If the last thread of globa 760 * If the last thread of global init has exited, panic
905 * immediately to get a useabl 761 * immediately to get a useable coredump.
906 */ 762 */
907 if (unlikely(is_global_init(ts 763 if (unlikely(is_global_init(tsk)))
908 panic("Attempted to ki 764 panic("Attempted to kill init! exitcode=0x%08x\n",
909 tsk->signal->g 765 tsk->signal->group_exit_code ?: (int)code);
910 766
911 #ifdef CONFIG_POSIX_TIMERS 767 #ifdef CONFIG_POSIX_TIMERS
912 hrtimer_cancel(&tsk->signal->r 768 hrtimer_cancel(&tsk->signal->real_timer);
913 exit_itimers(tsk); 769 exit_itimers(tsk);
914 #endif 770 #endif
915 if (tsk->mm) 771 if (tsk->mm)
916 setmax_mm_hiwater_rss( 772 setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm);
917 } 773 }
918 acct_collect(code, group_dead); 774 acct_collect(code, group_dead);
919 if (group_dead) 775 if (group_dead)
920 tty_audit_exit(); 776 tty_audit_exit();
921 audit_free(tsk); 777 audit_free(tsk);
922 778
923 tsk->exit_code = code; 779 tsk->exit_code = code;
924 taskstats_exit(tsk, group_dead); 780 taskstats_exit(tsk, group_dead);
925 781
926 exit_mm(); 782 exit_mm();
927 783
928 if (group_dead) 784 if (group_dead)
929 acct_process(); 785 acct_process();
930 trace_sched_process_exit(tsk); 786 trace_sched_process_exit(tsk);
931 787
932 exit_sem(tsk); 788 exit_sem(tsk);
933 exit_shm(tsk); 789 exit_shm(tsk);
934 exit_files(tsk); 790 exit_files(tsk);
935 exit_fs(tsk); 791 exit_fs(tsk);
936 if (group_dead) 792 if (group_dead)
937 disassociate_ctty(1); 793 disassociate_ctty(1);
938 exit_task_namespaces(tsk); 794 exit_task_namespaces(tsk);
939 exit_task_work(tsk); 795 exit_task_work(tsk);
940 exit_thread(tsk); 796 exit_thread(tsk);
941 797
942 /* 798 /*
943 * Flush inherited counters to the par 799 * Flush inherited counters to the parent - before the parent
944 * gets woken up by child-exit notific 800 * gets woken up by child-exit notifications.
945 * 801 *
946 * because of cgroup mode, must be cal 802 * because of cgroup mode, must be called before cgroup_exit()
947 */ 803 */
948 perf_event_exit_task(tsk); 804 perf_event_exit_task(tsk);
949 805
950 sched_autogroup_exit_task(tsk); 806 sched_autogroup_exit_task(tsk);
951 cgroup_exit(tsk); 807 cgroup_exit(tsk);
952 808
953 /* 809 /*
954 * FIXME: do that only when needed, us 810 * FIXME: do that only when needed, using sched_exit tracepoint
955 */ 811 */
956 flush_ptrace_hw_breakpoint(tsk); 812 flush_ptrace_hw_breakpoint(tsk);
957 813
958 exit_tasks_rcu_start(); 814 exit_tasks_rcu_start();
959 exit_notify(tsk, group_dead); 815 exit_notify(tsk, group_dead);
960 proc_exit_connector(tsk); 816 proc_exit_connector(tsk);
961 mpol_put_task_policy(tsk); 817 mpol_put_task_policy(tsk);
962 #ifdef CONFIG_FUTEX 818 #ifdef CONFIG_FUTEX
963 if (unlikely(current->pi_state_cache)) 819 if (unlikely(current->pi_state_cache))
964 kfree(current->pi_state_cache) 820 kfree(current->pi_state_cache);
965 #endif 821 #endif
966 /* 822 /*
967 * Make sure we are holding no locks: 823 * Make sure we are holding no locks:
968 */ 824 */
969 debug_check_no_locks_held(); 825 debug_check_no_locks_held();
970 826
971 if (tsk->io_context) 827 if (tsk->io_context)
972 exit_io_context(tsk); 828 exit_io_context(tsk);
973 829
974 if (tsk->splice_pipe) 830 if (tsk->splice_pipe)
975 free_pipe_info(tsk->splice_pip 831 free_pipe_info(tsk->splice_pipe);
976 832
977 if (tsk->task_frag.page) 833 if (tsk->task_frag.page)
978 put_page(tsk->task_frag.page); 834 put_page(tsk->task_frag.page);
979 835
>> 836 validate_creds_for_do_exit(tsk);
980 exit_task_stack_account(tsk); 837 exit_task_stack_account(tsk);
981 838
982 check_stack_usage(); 839 check_stack_usage();
983 preempt_disable(); 840 preempt_disable();
984 if (tsk->nr_dirtied) 841 if (tsk->nr_dirtied)
985 __this_cpu_add(dirty_throttle_ 842 __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied);
986 exit_rcu(); 843 exit_rcu();
987 exit_tasks_rcu_finish(); 844 exit_tasks_rcu_finish();
988 845
989 lockdep_free_task(tsk); 846 lockdep_free_task(tsk);
990 do_task_dead(); 847 do_task_dead();
991 } 848 }
992 849
993 void __noreturn make_task_dead(int signr) 850 void __noreturn make_task_dead(int signr)
994 { 851 {
995 /* 852 /*
996 * Take the task off the cpu after som 853 * Take the task off the cpu after something catastrophic has
997 * happened. 854 * happened.
998 * 855 *
999 * We can get here from a kernel oops, 856 * We can get here from a kernel oops, sometimes with preemption off.
1000 * Start by checking for critical err 857 * Start by checking for critical errors.
1001 * Then fix up important state like U 858 * Then fix up important state like USER_DS and preemption.
1002 * Then do everything else. 859 * Then do everything else.
1003 */ 860 */
1004 struct task_struct *tsk = current; 861 struct task_struct *tsk = current;
1005 unsigned int limit; <<
1006 862
1007 if (unlikely(in_interrupt())) 863 if (unlikely(in_interrupt()))
1008 panic("Aiee, killing interrup 864 panic("Aiee, killing interrupt handler!");
1009 if (unlikely(!tsk->pid)) 865 if (unlikely(!tsk->pid))
1010 panic("Attempted to kill the 866 panic("Attempted to kill the idle task!");
1011 867
1012 if (unlikely(irqs_disabled())) { <<
1013 pr_info("note: %s[%d] exited <<
1014 current->comm, task_p <<
1015 local_irq_enable(); <<
1016 } <<
1017 if (unlikely(in_atomic())) { 868 if (unlikely(in_atomic())) {
1018 pr_info("note: %s[%d] exited 869 pr_info("note: %s[%d] exited with preempt_count %d\n",
1019 current->comm, task_p 870 current->comm, task_pid_nr(current),
1020 preempt_count()); 871 preempt_count());
1021 preempt_count_set(PREEMPT_ENA 872 preempt_count_set(PREEMPT_ENABLED);
1022 } 873 }
1023 874
1024 /* 875 /*
1025 * Every time the system oopses, if t <<
1026 * to an object was held, the referen <<
1027 * If the oops doesn't also leak memo <<
1028 * reference counters to wrap around <<
1029 * This means that repeated oopsing c <<
1030 * exploitable through repeated oopsi <<
1031 * To make sure this can't happen, pl <<
1032 * kernel may oops without panic(). <<
1033 */ <<
1034 limit = READ_ONCE(oops_limit); <<
1035 if (atomic_inc_return(&oops_count) >= <<
1036 panic("Oopsed too often (kern <<
1037 <<
1038 /* <<
1039 * We're taking recursive faults here 876 * We're taking recursive faults here in make_task_dead. Safest is to just
1040 * leave this task alone and wait for 877 * leave this task alone and wait for reboot.
1041 */ 878 */
1042 if (unlikely(tsk->flags & PF_EXITING) 879 if (unlikely(tsk->flags & PF_EXITING)) {
1043 pr_alert("Fixing recursive fa 880 pr_alert("Fixing recursive fault but reboot is needed!\n");
1044 futex_exit_recursive(tsk); 881 futex_exit_recursive(tsk);
1045 tsk->exit_state = EXIT_DEAD; 882 tsk->exit_state = EXIT_DEAD;
1046 refcount_inc(&tsk->rcu_users) 883 refcount_inc(&tsk->rcu_users);
1047 do_task_dead(); 884 do_task_dead();
1048 } 885 }
1049 886
1050 do_exit(signr); 887 do_exit(signr);
1051 } 888 }
1052 889
1053 SYSCALL_DEFINE1(exit, int, error_code) 890 SYSCALL_DEFINE1(exit, int, error_code)
1054 { 891 {
1055 do_exit((error_code&0xff)<<8); 892 do_exit((error_code&0xff)<<8);
1056 } 893 }
1057 894
1058 /* 895 /*
1059 * Take down every thread in the group. This 896 * Take down every thread in the group. This is called by fatal signals
1060 * as well as by sys_exit_group (below). 897 * as well as by sys_exit_group (below).
1061 */ 898 */
1062 void __noreturn 899 void __noreturn
1063 do_group_exit(int exit_code) 900 do_group_exit(int exit_code)
1064 { 901 {
1065 struct signal_struct *sig = current-> 902 struct signal_struct *sig = current->signal;
1066 903
1067 if (sig->flags & SIGNAL_GROUP_EXIT) 904 if (sig->flags & SIGNAL_GROUP_EXIT)
1068 exit_code = sig->group_exit_c 905 exit_code = sig->group_exit_code;
1069 else if (sig->group_exec_task) 906 else if (sig->group_exec_task)
1070 exit_code = 0; 907 exit_code = 0;
1071 else { !! 908 else if (!thread_group_empty(current)) {
1072 struct sighand_struct *const 909 struct sighand_struct *const sighand = current->sighand;
1073 910
1074 spin_lock_irq(&sighand->siglo 911 spin_lock_irq(&sighand->siglock);
1075 if (sig->flags & SIGNAL_GROUP 912 if (sig->flags & SIGNAL_GROUP_EXIT)
1076 /* Another thread got 913 /* Another thread got here before we took the lock. */
1077 exit_code = sig->grou 914 exit_code = sig->group_exit_code;
1078 else if (sig->group_exec_task 915 else if (sig->group_exec_task)
1079 exit_code = 0; 916 exit_code = 0;
1080 else { 917 else {
1081 sig->group_exit_code 918 sig->group_exit_code = exit_code;
1082 sig->flags = SIGNAL_G 919 sig->flags = SIGNAL_GROUP_EXIT;
1083 zap_other_threads(cur 920 zap_other_threads(current);
1084 } 921 }
1085 spin_unlock_irq(&sighand->sig 922 spin_unlock_irq(&sighand->siglock);
1086 } 923 }
1087 924
1088 do_exit(exit_code); 925 do_exit(exit_code);
1089 /* NOTREACHED */ 926 /* NOTREACHED */
1090 } 927 }
1091 928
1092 /* 929 /*
1093 * this kills every thread in the thread grou 930 * this kills every thread in the thread group. Note that any externally
1094 * wait4()-ing process will get the correct e 931 * wait4()-ing process will get the correct exit code - even if this
1095 * thread is not the thread group leader. 932 * thread is not the thread group leader.
1096 */ 933 */
1097 SYSCALL_DEFINE1(exit_group, int, error_code) 934 SYSCALL_DEFINE1(exit_group, int, error_code)
1098 { 935 {
1099 do_group_exit((error_code & 0xff) << 936 do_group_exit((error_code & 0xff) << 8);
1100 /* NOTREACHED */ 937 /* NOTREACHED */
1101 return 0; 938 return 0;
1102 } 939 }
1103 940
>> 941 struct waitid_info {
>> 942 pid_t pid;
>> 943 uid_t uid;
>> 944 int status;
>> 945 int cause;
>> 946 };
>> 947
>> 948 struct wait_opts {
>> 949 enum pid_type wo_type;
>> 950 int wo_flags;
>> 951 struct pid *wo_pid;
>> 952
>> 953 struct waitid_info *wo_info;
>> 954 int wo_stat;
>> 955 struct rusage *wo_rusage;
>> 956
>> 957 wait_queue_entry_t child_wait;
>> 958 int notask_error;
>> 959 };
>> 960
1104 static int eligible_pid(struct wait_opts *wo, 961 static int eligible_pid(struct wait_opts *wo, struct task_struct *p)
1105 { 962 {
1106 return wo->wo_type == PIDTYPE_MAX || 963 return wo->wo_type == PIDTYPE_MAX ||
1107 task_pid_type(p, wo->wo_type) 964 task_pid_type(p, wo->wo_type) == wo->wo_pid;
1108 } 965 }
1109 966
1110 static int 967 static int
1111 eligible_child(struct wait_opts *wo, bool ptr 968 eligible_child(struct wait_opts *wo, bool ptrace, struct task_struct *p)
1112 { 969 {
1113 if (!eligible_pid(wo, p)) 970 if (!eligible_pid(wo, p))
1114 return 0; 971 return 0;
1115 972
1116 /* 973 /*
1117 * Wait for all children (clone and n 974 * Wait for all children (clone and not) if __WALL is set or
1118 * if it is traced by us. 975 * if it is traced by us.
1119 */ 976 */
1120 if (ptrace || (wo->wo_flags & __WALL) 977 if (ptrace || (wo->wo_flags & __WALL))
1121 return 1; 978 return 1;
1122 979
1123 /* 980 /*
1124 * Otherwise, wait for clone children 981 * Otherwise, wait for clone children *only* if __WCLONE is set;
1125 * otherwise, wait for non-clone chil 982 * otherwise, wait for non-clone children *only*.
1126 * 983 *
1127 * Note: a "clone" child here is one 984 * Note: a "clone" child here is one that reports to its parent
1128 * using a signal other than SIGCHLD, 985 * using a signal other than SIGCHLD, or a non-leader thread which
1129 * we can only see if it is traced by 986 * we can only see if it is traced by us.
1130 */ 987 */
1131 if ((p->exit_signal != SIGCHLD) ^ !!( 988 if ((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE))
1132 return 0; 989 return 0;
1133 990
1134 return 1; 991 return 1;
1135 } 992 }
1136 993
1137 /* 994 /*
1138 * Handle sys_wait4 work for one task in stat 995 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1139 * read_lock(&tasklist_lock) on entry. If we 996 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1140 * the lock and this task is uninteresting. 997 * the lock and this task is uninteresting. If we return nonzero, we have
1141 * released the lock and the system call shou 998 * released the lock and the system call should return.
1142 */ 999 */
1143 static int wait_task_zombie(struct wait_opts 1000 static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p)
1144 { 1001 {
1145 int state, status; 1002 int state, status;
1146 pid_t pid = task_pid_vnr(p); 1003 pid_t pid = task_pid_vnr(p);
1147 uid_t uid = from_kuid_munged(current_ 1004 uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p));
1148 struct waitid_info *infop; 1005 struct waitid_info *infop;
1149 1006
1150 if (!likely(wo->wo_flags & WEXITED)) 1007 if (!likely(wo->wo_flags & WEXITED))
1151 return 0; 1008 return 0;
1152 1009
1153 if (unlikely(wo->wo_flags & WNOWAIT)) 1010 if (unlikely(wo->wo_flags & WNOWAIT)) {
1154 status = (p->signal->flags & 1011 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1155 ? p->signal->group_ex 1012 ? p->signal->group_exit_code : p->exit_code;
1156 get_task_struct(p); 1013 get_task_struct(p);
1157 read_unlock(&tasklist_lock); 1014 read_unlock(&tasklist_lock);
1158 sched_annotate_sleep(); 1015 sched_annotate_sleep();
1159 if (wo->wo_rusage) 1016 if (wo->wo_rusage)
1160 getrusage(p, RUSAGE_B 1017 getrusage(p, RUSAGE_BOTH, wo->wo_rusage);
1161 put_task_struct(p); 1018 put_task_struct(p);
1162 goto out_info; 1019 goto out_info;
1163 } 1020 }
1164 /* 1021 /*
1165 * Move the task's state to DEAD/TRAC 1022 * Move the task's state to DEAD/TRACE, only one thread can do this.
1166 */ 1023 */
1167 state = (ptrace_reparented(p) && thre 1024 state = (ptrace_reparented(p) && thread_group_leader(p)) ?
1168 EXIT_TRACE : EXIT_DEAD; 1025 EXIT_TRACE : EXIT_DEAD;
1169 if (cmpxchg(&p->exit_state, EXIT_ZOMB 1026 if (cmpxchg(&p->exit_state, EXIT_ZOMBIE, state) != EXIT_ZOMBIE)
1170 return 0; 1027 return 0;
1171 /* 1028 /*
1172 * We own this thread, nobody else ca 1029 * We own this thread, nobody else can reap it.
1173 */ 1030 */
1174 read_unlock(&tasklist_lock); 1031 read_unlock(&tasklist_lock);
1175 sched_annotate_sleep(); 1032 sched_annotate_sleep();
1176 1033
1177 /* 1034 /*
1178 * Check thread_group_leader() to exc 1035 * Check thread_group_leader() to exclude the traced sub-threads.
1179 */ 1036 */
1180 if (state == EXIT_DEAD && thread_grou 1037 if (state == EXIT_DEAD && thread_group_leader(p)) {
1181 struct signal_struct *sig = p 1038 struct signal_struct *sig = p->signal;
1182 struct signal_struct *psig = 1039 struct signal_struct *psig = current->signal;
1183 unsigned long maxrss; 1040 unsigned long maxrss;
1184 u64 tgutime, tgstime; 1041 u64 tgutime, tgstime;
1185 1042
1186 /* 1043 /*
1187 * The resource counters for 1044 * The resource counters for the group leader are in its
1188 * own task_struct. Those fo 1045 * own task_struct. Those for dead threads in the group
1189 * are in its signal_struct, 1046 * are in its signal_struct, as are those for the child
1190 * processes it has previousl 1047 * processes it has previously reaped. All these
1191 * accumulate in the parent's 1048 * accumulate in the parent's signal_struct c* fields.
1192 * 1049 *
1193 * We don't bother to take a 1050 * We don't bother to take a lock here to protect these
1194 * p->signal fields because t 1051 * p->signal fields because the whole thread group is dead
1195 * and nobody can change them 1052 * and nobody can change them.
1196 * 1053 *
1197 * psig->stats_lock also prot !! 1054 * psig->stats_lock also protects us from our sub-theads
1198 * which can reap other child !! 1055 * which can reap other children at the same time. Until
>> 1056 * we change k_getrusage()-like users to rely on this lock
>> 1057 * we have to take ->siglock as well.
1199 * 1058 *
1200 * We use thread_group_cputim 1059 * We use thread_group_cputime_adjusted() to get times for
1201 * the thread group, which co 1060 * the thread group, which consolidates times for all threads
1202 * in the group including the 1061 * in the group including the group leader.
1203 */ 1062 */
1204 thread_group_cputime_adjusted 1063 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1205 write_seqlock_irq(&psig->stat !! 1064 spin_lock_irq(¤t->sighand->siglock);
>> 1065 write_seqlock(&psig->stats_lock);
1206 psig->cutime += tgutime + sig 1066 psig->cutime += tgutime + sig->cutime;
1207 psig->cstime += tgstime + sig 1067 psig->cstime += tgstime + sig->cstime;
1208 psig->cgtime += task_gtime(p) 1068 psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime;
1209 psig->cmin_flt += 1069 psig->cmin_flt +=
1210 p->min_flt + sig->min 1070 p->min_flt + sig->min_flt + sig->cmin_flt;
1211 psig->cmaj_flt += 1071 psig->cmaj_flt +=
1212 p->maj_flt + sig->maj 1072 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1213 psig->cnvcsw += 1073 psig->cnvcsw +=
1214 p->nvcsw + sig->nvcsw 1074 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1215 psig->cnivcsw += 1075 psig->cnivcsw +=
1216 p->nivcsw + sig->nivc 1076 p->nivcsw + sig->nivcsw + sig->cnivcsw;
1217 psig->cinblock += 1077 psig->cinblock +=
1218 task_io_get_inblock(p 1078 task_io_get_inblock(p) +
1219 sig->inblock + sig->c 1079 sig->inblock + sig->cinblock;
1220 psig->coublock += 1080 psig->coublock +=
1221 task_io_get_oublock(p 1081 task_io_get_oublock(p) +
1222 sig->oublock + sig->c 1082 sig->oublock + sig->coublock;
1223 maxrss = max(sig->maxrss, sig 1083 maxrss = max(sig->maxrss, sig->cmaxrss);
1224 if (psig->cmaxrss < maxrss) 1084 if (psig->cmaxrss < maxrss)
1225 psig->cmaxrss = maxrs 1085 psig->cmaxrss = maxrss;
1226 task_io_accounting_add(&psig- 1086 task_io_accounting_add(&psig->ioac, &p->ioac);
1227 task_io_accounting_add(&psig- 1087 task_io_accounting_add(&psig->ioac, &sig->ioac);
1228 write_sequnlock_irq(&psig->st !! 1088 write_sequnlock(&psig->stats_lock);
>> 1089 spin_unlock_irq(¤t->sighand->siglock);
1229 } 1090 }
1230 1091
1231 if (wo->wo_rusage) 1092 if (wo->wo_rusage)
1232 getrusage(p, RUSAGE_BOTH, wo- 1093 getrusage(p, RUSAGE_BOTH, wo->wo_rusage);
1233 status = (p->signal->flags & SIGNAL_G 1094 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1234 ? p->signal->group_exit_code 1095 ? p->signal->group_exit_code : p->exit_code;
1235 wo->wo_stat = status; 1096 wo->wo_stat = status;
1236 1097
1237 if (state == EXIT_TRACE) { 1098 if (state == EXIT_TRACE) {
1238 write_lock_irq(&tasklist_lock 1099 write_lock_irq(&tasklist_lock);
1239 /* We dropped tasklist, ptrac 1100 /* We dropped tasklist, ptracer could die and untrace */
1240 ptrace_unlink(p); 1101 ptrace_unlink(p);
1241 1102
1242 /* If parent wants a zombie, 1103 /* If parent wants a zombie, don't release it now */
1243 state = EXIT_ZOMBIE; 1104 state = EXIT_ZOMBIE;
1244 if (do_notify_parent(p, p->ex 1105 if (do_notify_parent(p, p->exit_signal))
1245 state = EXIT_DEAD; 1106 state = EXIT_DEAD;
1246 p->exit_state = state; 1107 p->exit_state = state;
1247 write_unlock_irq(&tasklist_lo 1108 write_unlock_irq(&tasklist_lock);
1248 } 1109 }
1249 if (state == EXIT_DEAD) 1110 if (state == EXIT_DEAD)
1250 release_task(p); 1111 release_task(p);
1251 1112
1252 out_info: 1113 out_info:
1253 infop = wo->wo_info; 1114 infop = wo->wo_info;
1254 if (infop) { 1115 if (infop) {
1255 if ((status & 0x7f) == 0) { 1116 if ((status & 0x7f) == 0) {
1256 infop->cause = CLD_EX 1117 infop->cause = CLD_EXITED;
1257 infop->status = statu 1118 infop->status = status >> 8;
1258 } else { 1119 } else {
1259 infop->cause = (statu 1120 infop->cause = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1260 infop->status = statu 1121 infop->status = status & 0x7f;
1261 } 1122 }
1262 infop->pid = pid; 1123 infop->pid = pid;
1263 infop->uid = uid; 1124 infop->uid = uid;
1264 } 1125 }
1265 1126
1266 return pid; 1127 return pid;
1267 } 1128 }
1268 1129
1269 static int *task_stopped_code(struct task_str 1130 static int *task_stopped_code(struct task_struct *p, bool ptrace)
1270 { 1131 {
1271 if (ptrace) { 1132 if (ptrace) {
1272 if (task_is_traced(p) && !(p- 1133 if (task_is_traced(p) && !(p->jobctl & JOBCTL_LISTENING))
1273 return &p->exit_code; 1134 return &p->exit_code;
1274 } else { 1135 } else {
1275 if (p->signal->flags & SIGNAL 1136 if (p->signal->flags & SIGNAL_STOP_STOPPED)
1276 return &p->signal->gr 1137 return &p->signal->group_exit_code;
1277 } 1138 }
1278 return NULL; 1139 return NULL;
1279 } 1140 }
1280 1141
1281 /** 1142 /**
1282 * wait_task_stopped - Wait for %TASK_STOPPED 1143 * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED
1283 * @wo: wait options 1144 * @wo: wait options
1284 * @ptrace: is the wait for ptrace 1145 * @ptrace: is the wait for ptrace
1285 * @p: task to wait for 1146 * @p: task to wait for
1286 * 1147 *
1287 * Handle sys_wait4() work for %p in state %T 1148 * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED.
1288 * 1149 *
1289 * CONTEXT: 1150 * CONTEXT:
1290 * read_lock(&tasklist_lock), which is releas 1151 * read_lock(&tasklist_lock), which is released if return value is
1291 * non-zero. Also, grabs and releases @p->si 1152 * non-zero. Also, grabs and releases @p->sighand->siglock.
1292 * 1153 *
1293 * RETURNS: 1154 * RETURNS:
1294 * 0 if wait condition didn't exist and searc 1155 * 0 if wait condition didn't exist and search for other wait conditions
1295 * should continue. Non-zero return, -errno 1156 * should continue. Non-zero return, -errno on failure and @p's pid on
1296 * success, implies that tasklist_lock is rel 1157 * success, implies that tasklist_lock is released and wait condition
1297 * search should terminate. 1158 * search should terminate.
1298 */ 1159 */
1299 static int wait_task_stopped(struct wait_opts 1160 static int wait_task_stopped(struct wait_opts *wo,
1300 int ptrace, s 1161 int ptrace, struct task_struct *p)
1301 { 1162 {
1302 struct waitid_info *infop; 1163 struct waitid_info *infop;
1303 int exit_code, *p_code, why; 1164 int exit_code, *p_code, why;
1304 uid_t uid = 0; /* unneeded, required 1165 uid_t uid = 0; /* unneeded, required by compiler */
1305 pid_t pid; 1166 pid_t pid;
1306 1167
1307 /* 1168 /*
1308 * Traditionally we see ptrace'd stop 1169 * Traditionally we see ptrace'd stopped tasks regardless of options.
1309 */ 1170 */
1310 if (!ptrace && !(wo->wo_flags & WUNTR 1171 if (!ptrace && !(wo->wo_flags & WUNTRACED))
1311 return 0; 1172 return 0;
1312 1173
1313 if (!task_stopped_code(p, ptrace)) 1174 if (!task_stopped_code(p, ptrace))
1314 return 0; 1175 return 0;
1315 1176
1316 exit_code = 0; 1177 exit_code = 0;
1317 spin_lock_irq(&p->sighand->siglock); 1178 spin_lock_irq(&p->sighand->siglock);
1318 1179
1319 p_code = task_stopped_code(p, ptrace) 1180 p_code = task_stopped_code(p, ptrace);
1320 if (unlikely(!p_code)) 1181 if (unlikely(!p_code))
1321 goto unlock_sig; 1182 goto unlock_sig;
1322 1183
1323 exit_code = *p_code; 1184 exit_code = *p_code;
1324 if (!exit_code) 1185 if (!exit_code)
1325 goto unlock_sig; 1186 goto unlock_sig;
1326 1187
1327 if (!unlikely(wo->wo_flags & WNOWAIT) 1188 if (!unlikely(wo->wo_flags & WNOWAIT))
1328 *p_code = 0; 1189 *p_code = 0;
1329 1190
1330 uid = from_kuid_munged(current_user_n 1191 uid = from_kuid_munged(current_user_ns(), task_uid(p));
1331 unlock_sig: 1192 unlock_sig:
1332 spin_unlock_irq(&p->sighand->siglock) 1193 spin_unlock_irq(&p->sighand->siglock);
1333 if (!exit_code) 1194 if (!exit_code)
1334 return 0; 1195 return 0;
1335 1196
1336 /* 1197 /*
1337 * Now we are pretty sure this task i 1198 * Now we are pretty sure this task is interesting.
1338 * Make sure it doesn't get reaped ou 1199 * Make sure it doesn't get reaped out from under us while we
1339 * give up the lock and then examine 1200 * give up the lock and then examine it below. We don't want to
1340 * keep holding onto the tasklist_loc 1201 * keep holding onto the tasklist_lock while we call getrusage and
1341 * possibly take page faults for user 1202 * possibly take page faults for user memory.
1342 */ 1203 */
1343 get_task_struct(p); 1204 get_task_struct(p);
1344 pid = task_pid_vnr(p); 1205 pid = task_pid_vnr(p);
1345 why = ptrace ? CLD_TRAPPED : CLD_STOP 1206 why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
1346 read_unlock(&tasklist_lock); 1207 read_unlock(&tasklist_lock);
1347 sched_annotate_sleep(); 1208 sched_annotate_sleep();
1348 if (wo->wo_rusage) 1209 if (wo->wo_rusage)
1349 getrusage(p, RUSAGE_BOTH, wo- 1210 getrusage(p, RUSAGE_BOTH, wo->wo_rusage);
1350 put_task_struct(p); 1211 put_task_struct(p);
1351 1212
1352 if (likely(!(wo->wo_flags & WNOWAIT)) 1213 if (likely(!(wo->wo_flags & WNOWAIT)))
1353 wo->wo_stat = (exit_code << 8 1214 wo->wo_stat = (exit_code << 8) | 0x7f;
1354 1215
1355 infop = wo->wo_info; 1216 infop = wo->wo_info;
1356 if (infop) { 1217 if (infop) {
1357 infop->cause = why; 1218 infop->cause = why;
1358 infop->status = exit_code; 1219 infop->status = exit_code;
1359 infop->pid = pid; 1220 infop->pid = pid;
1360 infop->uid = uid; 1221 infop->uid = uid;
1361 } 1222 }
1362 return pid; 1223 return pid;
1363 } 1224 }
1364 1225
1365 /* 1226 /*
1366 * Handle do_wait work for one task in a live 1227 * Handle do_wait work for one task in a live, non-stopped state.
1367 * read_lock(&tasklist_lock) on entry. If we 1228 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1368 * the lock and this task is uninteresting. 1229 * the lock and this task is uninteresting. If we return nonzero, we have
1369 * released the lock and the system call shou 1230 * released the lock and the system call should return.
1370 */ 1231 */
1371 static int wait_task_continued(struct wait_op 1232 static int wait_task_continued(struct wait_opts *wo, struct task_struct *p)
1372 { 1233 {
1373 struct waitid_info *infop; 1234 struct waitid_info *infop;
1374 pid_t pid; 1235 pid_t pid;
1375 uid_t uid; 1236 uid_t uid;
1376 1237
1377 if (!unlikely(wo->wo_flags & WCONTINU 1238 if (!unlikely(wo->wo_flags & WCONTINUED))
1378 return 0; 1239 return 0;
1379 1240
1380 if (!(p->signal->flags & SIGNAL_STOP_ 1241 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1381 return 0; 1242 return 0;
1382 1243
1383 spin_lock_irq(&p->sighand->siglock); 1244 spin_lock_irq(&p->sighand->siglock);
1384 /* Re-check with the lock held. */ 1245 /* Re-check with the lock held. */
1385 if (!(p->signal->flags & SIGNAL_STOP_ 1246 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1386 spin_unlock_irq(&p->sighand-> 1247 spin_unlock_irq(&p->sighand->siglock);
1387 return 0; 1248 return 0;
1388 } 1249 }
1389 if (!unlikely(wo->wo_flags & WNOWAIT) 1250 if (!unlikely(wo->wo_flags & WNOWAIT))
1390 p->signal->flags &= ~SIGNAL_S 1251 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1391 uid = from_kuid_munged(current_user_n 1252 uid = from_kuid_munged(current_user_ns(), task_uid(p));
1392 spin_unlock_irq(&p->sighand->siglock) 1253 spin_unlock_irq(&p->sighand->siglock);
1393 1254
1394 pid = task_pid_vnr(p); 1255 pid = task_pid_vnr(p);
1395 get_task_struct(p); 1256 get_task_struct(p);
1396 read_unlock(&tasklist_lock); 1257 read_unlock(&tasklist_lock);
1397 sched_annotate_sleep(); 1258 sched_annotate_sleep();
1398 if (wo->wo_rusage) 1259 if (wo->wo_rusage)
1399 getrusage(p, RUSAGE_BOTH, wo- 1260 getrusage(p, RUSAGE_BOTH, wo->wo_rusage);
1400 put_task_struct(p); 1261 put_task_struct(p);
1401 1262
1402 infop = wo->wo_info; 1263 infop = wo->wo_info;
1403 if (!infop) { 1264 if (!infop) {
1404 wo->wo_stat = 0xffff; 1265 wo->wo_stat = 0xffff;
1405 } else { 1266 } else {
1406 infop->cause = CLD_CONTINUED; 1267 infop->cause = CLD_CONTINUED;
1407 infop->pid = pid; 1268 infop->pid = pid;
1408 infop->uid = uid; 1269 infop->uid = uid;
1409 infop->status = SIGCONT; 1270 infop->status = SIGCONT;
1410 } 1271 }
1411 return pid; 1272 return pid;
1412 } 1273 }
1413 1274
1414 /* 1275 /*
1415 * Consider @p for a wait by @parent. 1276 * Consider @p for a wait by @parent.
1416 * 1277 *
1417 * -ECHILD should be in ->notask_error before 1278 * -ECHILD should be in ->notask_error before the first call.
1418 * Returns nonzero for a final return, when w 1279 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1419 * Returns zero if the search for a child sho 1280 * Returns zero if the search for a child should continue;
1420 * then ->notask_error is 0 if @p is an eligi 1281 * then ->notask_error is 0 if @p is an eligible child,
1421 * or still -ECHILD. 1282 * or still -ECHILD.
1422 */ 1283 */
1423 static int wait_consider_task(struct wait_opt 1284 static int wait_consider_task(struct wait_opts *wo, int ptrace,
1424 struct task_s 1285 struct task_struct *p)
1425 { 1286 {
1426 /* 1287 /*
1427 * We can race with wait_task_zombie( 1288 * We can race with wait_task_zombie() from another thread.
1428 * Ensure that EXIT_ZOMBIE -> EXIT_DE 1289 * Ensure that EXIT_ZOMBIE -> EXIT_DEAD/EXIT_TRACE transition
1429 * can't confuse the checks below. 1290 * can't confuse the checks below.
1430 */ 1291 */
1431 int exit_state = READ_ONCE(p->exit_st 1292 int exit_state = READ_ONCE(p->exit_state);
1432 int ret; 1293 int ret;
1433 1294
1434 if (unlikely(exit_state == EXIT_DEAD) 1295 if (unlikely(exit_state == EXIT_DEAD))
1435 return 0; 1296 return 0;
1436 1297
1437 ret = eligible_child(wo, ptrace, p); 1298 ret = eligible_child(wo, ptrace, p);
1438 if (!ret) 1299 if (!ret)
1439 return ret; 1300 return ret;
1440 1301
1441 if (unlikely(exit_state == EXIT_TRACE 1302 if (unlikely(exit_state == EXIT_TRACE)) {
1442 /* 1303 /*
1443 * ptrace == 0 means we are t 1304 * ptrace == 0 means we are the natural parent. In this case
1444 * we should clear notask_err 1305 * we should clear notask_error, debugger will notify us.
1445 */ 1306 */
1446 if (likely(!ptrace)) 1307 if (likely(!ptrace))
1447 wo->notask_error = 0; 1308 wo->notask_error = 0;
1448 return 0; 1309 return 0;
1449 } 1310 }
1450 1311
1451 if (likely(!ptrace) && unlikely(p->pt 1312 if (likely(!ptrace) && unlikely(p->ptrace)) {
1452 /* 1313 /*
1453 * If it is traced by its rea 1314 * If it is traced by its real parent's group, just pretend
1454 * the caller is ptrace_do_wa 1315 * the caller is ptrace_do_wait() and reap this child if it
1455 * is zombie. 1316 * is zombie.
1456 * 1317 *
1457 * This also hides group stop 1318 * This also hides group stop state from real parent; otherwise
1458 * a single stop can be repor 1319 * a single stop can be reported twice as group and ptrace stop.
1459 * If a ptracer wants to dist 1320 * If a ptracer wants to distinguish these two events for its
1460 * own children it should cre 1321 * own children it should create a separate process which takes
1461 * the role of real parent. 1322 * the role of real parent.
1462 */ 1323 */
1463 if (!ptrace_reparented(p)) 1324 if (!ptrace_reparented(p))
1464 ptrace = 1; 1325 ptrace = 1;
1465 } 1326 }
1466 1327
1467 /* slay zombie? */ 1328 /* slay zombie? */
1468 if (exit_state == EXIT_ZOMBIE) { 1329 if (exit_state == EXIT_ZOMBIE) {
1469 /* we don't reap group leader 1330 /* we don't reap group leaders with subthreads */
1470 if (!delay_group_leader(p)) { 1331 if (!delay_group_leader(p)) {
1471 /* 1332 /*
1472 * A zombie ptracee i 1333 * A zombie ptracee is only visible to its ptracer.
1473 * Notification and r 1334 * Notification and reaping will be cascaded to the
1474 * real parent when t 1335 * real parent when the ptracer detaches.
1475 */ 1336 */
1476 if (unlikely(ptrace) 1337 if (unlikely(ptrace) || likely(!p->ptrace))
1477 return wait_t 1338 return wait_task_zombie(wo, p);
1478 } 1339 }
1479 1340
1480 /* 1341 /*
1481 * Allow access to stopped/co 1342 * Allow access to stopped/continued state via zombie by
1482 * falling through. Clearing 1343 * falling through. Clearing of notask_error is complex.
1483 * 1344 *
1484 * When !@ptrace: 1345 * When !@ptrace:
1485 * 1346 *
1486 * If WEXITED is set, notask_ 1347 * If WEXITED is set, notask_error should naturally be
1487 * cleared. If not, subset o 1348 * cleared. If not, subset of WSTOPPED|WCONTINUED is set,
1488 * so, if there are live subt 1349 * so, if there are live subthreads, there are events to
1489 * wait for. If all subthrea 1350 * wait for. If all subthreads are dead, it's still safe
1490 * to clear - this function w 1351 * to clear - this function will be called again in finite
1491 * amount time once all the s 1352 * amount time once all the subthreads are released and
1492 * will then return without c 1353 * will then return without clearing.
1493 * 1354 *
1494 * When @ptrace: 1355 * When @ptrace:
1495 * 1356 *
1496 * Stopped state is per-task 1357 * Stopped state is per-task and thus can't change once the
1497 * target task dies. Only co 1358 * target task dies. Only continued and exited can happen.
1498 * Clear notask_error if WCON 1359 * Clear notask_error if WCONTINUED | WEXITED.
1499 */ 1360 */
1500 if (likely(!ptrace) || (wo->w 1361 if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED)))
1501 wo->notask_error = 0; 1362 wo->notask_error = 0;
1502 } else { 1363 } else {
1503 /* 1364 /*
1504 * @p is alive and it's gonna 1365 * @p is alive and it's gonna stop, continue or exit, so
1505 * there always is something 1366 * there always is something to wait for.
1506 */ 1367 */
1507 wo->notask_error = 0; 1368 wo->notask_error = 0;
1508 } 1369 }
1509 1370
1510 /* 1371 /*
1511 * Wait for stopped. Depending on @p 1372 * Wait for stopped. Depending on @ptrace, different stopped state
1512 * is used and the two don't interact 1373 * is used and the two don't interact with each other.
1513 */ 1374 */
1514 ret = wait_task_stopped(wo, ptrace, p 1375 ret = wait_task_stopped(wo, ptrace, p);
1515 if (ret) 1376 if (ret)
1516 return ret; 1377 return ret;
1517 1378
1518 /* 1379 /*
1519 * Wait for continued. There's only 1380 * Wait for continued. There's only one continued state and the
1520 * ptracer can consume it which can c 1381 * ptracer can consume it which can confuse the real parent. Don't
1521 * use WCONTINUED from ptracer. You 1382 * use WCONTINUED from ptracer. You don't need or want it.
1522 */ 1383 */
1523 return wait_task_continued(wo, p); 1384 return wait_task_continued(wo, p);
1524 } 1385 }
1525 1386
1526 /* 1387 /*
1527 * Do the work of do_wait() for one thread in 1388 * Do the work of do_wait() for one thread in the group, @tsk.
1528 * 1389 *
1529 * -ECHILD should be in ->notask_error before 1390 * -ECHILD should be in ->notask_error before the first call.
1530 * Returns nonzero for a final return, when w 1391 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1531 * Returns zero if the search for a child sho 1392 * Returns zero if the search for a child should continue; then
1532 * ->notask_error is 0 if there were any elig 1393 * ->notask_error is 0 if there were any eligible children,
1533 * or still -ECHILD. 1394 * or still -ECHILD.
1534 */ 1395 */
1535 static int do_wait_thread(struct wait_opts *w 1396 static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk)
1536 { 1397 {
1537 struct task_struct *p; 1398 struct task_struct *p;
1538 1399
1539 list_for_each_entry(p, &tsk->children 1400 list_for_each_entry(p, &tsk->children, sibling) {
1540 int ret = wait_consider_task( 1401 int ret = wait_consider_task(wo, 0, p);
1541 1402
1542 if (ret) 1403 if (ret)
1543 return ret; 1404 return ret;
1544 } 1405 }
1545 1406
1546 return 0; 1407 return 0;
1547 } 1408 }
1548 1409
1549 static int ptrace_do_wait(struct wait_opts *w 1410 static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk)
1550 { 1411 {
1551 struct task_struct *p; 1412 struct task_struct *p;
1552 1413
1553 list_for_each_entry(p, &tsk->ptraced, 1414 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
1554 int ret = wait_consider_task( 1415 int ret = wait_consider_task(wo, 1, p);
1555 1416
1556 if (ret) 1417 if (ret)
1557 return ret; 1418 return ret;
1558 } 1419 }
1559 1420
1560 return 0; 1421 return 0;
1561 } 1422 }
1562 1423
1563 bool pid_child_should_wake(struct wait_opts * <<
1564 { <<
1565 if (!eligible_pid(wo, p)) <<
1566 return false; <<
1567 <<
1568 if ((wo->wo_flags & __WNOTHREAD) && w <<
1569 return false; <<
1570 <<
1571 return true; <<
1572 } <<
1573 <<
1574 static int child_wait_callback(wait_queue_ent 1424 static int child_wait_callback(wait_queue_entry_t *wait, unsigned mode,
1575 int sync, voi 1425 int sync, void *key)
1576 { 1426 {
1577 struct wait_opts *wo = container_of(w 1427 struct wait_opts *wo = container_of(wait, struct wait_opts,
1578 1428 child_wait);
1579 struct task_struct *p = key; 1429 struct task_struct *p = key;
1580 1430
1581 if (pid_child_should_wake(wo, p)) !! 1431 if (!eligible_pid(wo, p))
1582 return default_wake_function( !! 1432 return 0;
>> 1433
>> 1434 if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent)
>> 1435 return 0;
1583 1436
1584 return 0; !! 1437 return default_wake_function(wait, mode, sync, key);
1585 } 1438 }
1586 1439
1587 void __wake_up_parent(struct task_struct *p, 1440 void __wake_up_parent(struct task_struct *p, struct task_struct *parent)
1588 { 1441 {
1589 __wake_up_sync_key(&parent->signal->w 1442 __wake_up_sync_key(&parent->signal->wait_chldexit,
1590 TASK_INTERRUPTIBLE 1443 TASK_INTERRUPTIBLE, p);
1591 } 1444 }
1592 1445
1593 static bool is_effectively_child(struct wait_ 1446 static bool is_effectively_child(struct wait_opts *wo, bool ptrace,
1594 struct task_ 1447 struct task_struct *target)
1595 { 1448 {
1596 struct task_struct *parent = 1449 struct task_struct *parent =
1597 !ptrace ? target->real_parent 1450 !ptrace ? target->real_parent : target->parent;
1598 1451
1599 return current == parent || (!(wo->wo 1452 return current == parent || (!(wo->wo_flags & __WNOTHREAD) &&
1600 same_thr 1453 same_thread_group(current, parent));
1601 } 1454 }
1602 1455
1603 /* 1456 /*
1604 * Optimization for waiting on PIDTYPE_PID. N 1457 * Optimization for waiting on PIDTYPE_PID. No need to iterate through child
1605 * and tracee lists to find the target task. 1458 * and tracee lists to find the target task.
1606 */ 1459 */
1607 static int do_wait_pid(struct wait_opts *wo) 1460 static int do_wait_pid(struct wait_opts *wo)
1608 { 1461 {
1609 bool ptrace; 1462 bool ptrace;
1610 struct task_struct *target; 1463 struct task_struct *target;
1611 int retval; 1464 int retval;
1612 1465
1613 ptrace = false; 1466 ptrace = false;
1614 target = pid_task(wo->wo_pid, PIDTYPE 1467 target = pid_task(wo->wo_pid, PIDTYPE_TGID);
1615 if (target && is_effectively_child(wo 1468 if (target && is_effectively_child(wo, ptrace, target)) {
1616 retval = wait_consider_task(w 1469 retval = wait_consider_task(wo, ptrace, target);
1617 if (retval) 1470 if (retval)
1618 return retval; 1471 return retval;
1619 } 1472 }
1620 1473
1621 ptrace = true; 1474 ptrace = true;
1622 target = pid_task(wo->wo_pid, PIDTYPE 1475 target = pid_task(wo->wo_pid, PIDTYPE_PID);
1623 if (target && target->ptrace && 1476 if (target && target->ptrace &&
1624 is_effectively_child(wo, ptrace, 1477 is_effectively_child(wo, ptrace, target)) {
1625 retval = wait_consider_task(w 1478 retval = wait_consider_task(wo, ptrace, target);
1626 if (retval) 1479 if (retval)
1627 return retval; 1480 return retval;
1628 } 1481 }
1629 1482
1630 return 0; 1483 return 0;
1631 } 1484 }
1632 1485
1633 long __do_wait(struct wait_opts *wo) !! 1486 static long do_wait(struct wait_opts *wo)
1634 { 1487 {
1635 long retval; !! 1488 int retval;
1636 1489
>> 1490 trace_sched_process_wait(wo->wo_pid);
>> 1491
>> 1492 init_waitqueue_func_entry(&wo->child_wait, child_wait_callback);
>> 1493 wo->child_wait.private = current;
>> 1494 add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait);
>> 1495 repeat:
1637 /* 1496 /*
1638 * If there is nothing that can match 1497 * If there is nothing that can match our criteria, just get out.
1639 * We will clear ->notask_error to ze 1498 * We will clear ->notask_error to zero if we see any child that
1640 * might later match our criteria, ev 1499 * might later match our criteria, even if we are not able to reap
1641 * it yet. 1500 * it yet.
1642 */ 1501 */
1643 wo->notask_error = -ECHILD; 1502 wo->notask_error = -ECHILD;
1644 if ((wo->wo_type < PIDTYPE_MAX) && 1503 if ((wo->wo_type < PIDTYPE_MAX) &&
1645 (!wo->wo_pid || !pid_has_task(wo-> 1504 (!wo->wo_pid || !pid_has_task(wo->wo_pid, wo->wo_type)))
1646 goto notask; 1505 goto notask;
1647 1506
>> 1507 set_current_state(TASK_INTERRUPTIBLE);
1648 read_lock(&tasklist_lock); 1508 read_lock(&tasklist_lock);
1649 1509
1650 if (wo->wo_type == PIDTYPE_PID) { 1510 if (wo->wo_type == PIDTYPE_PID) {
1651 retval = do_wait_pid(wo); 1511 retval = do_wait_pid(wo);
1652 if (retval) 1512 if (retval)
1653 return retval; !! 1513 goto end;
1654 } else { 1514 } else {
1655 struct task_struct *tsk = cur 1515 struct task_struct *tsk = current;
1656 1516
1657 do { 1517 do {
1658 retval = do_wait_thre 1518 retval = do_wait_thread(wo, tsk);
1659 if (retval) 1519 if (retval)
1660 return retval !! 1520 goto end;
1661 1521
1662 retval = ptrace_do_wa 1522 retval = ptrace_do_wait(wo, tsk);
1663 if (retval) 1523 if (retval)
1664 return retval !! 1524 goto end;
1665 1525
1666 if (wo->wo_flags & __ 1526 if (wo->wo_flags & __WNOTHREAD)
1667 break; 1527 break;
1668 } while_each_thread(current, 1528 } while_each_thread(current, tsk);
1669 } 1529 }
1670 read_unlock(&tasklist_lock); 1530 read_unlock(&tasklist_lock);
1671 1531
1672 notask: 1532 notask:
1673 retval = wo->notask_error; 1533 retval = wo->notask_error;
1674 if (!retval && !(wo->wo_flags & WNOHA !! 1534 if (!retval && !(wo->wo_flags & WNOHANG)) {
1675 return -ERESTARTSYS; !! 1535 retval = -ERESTARTSYS;
1676 !! 1536 if (!signal_pending(current)) {
1677 return retval; !! 1537 schedule();
1678 } !! 1538 goto repeat;
1679 !! 1539 }
1680 static long do_wait(struct wait_opts *wo) !! 1540 }
1681 { !! 1541 end:
1682 int retval; <<
1683 <<
1684 trace_sched_process_wait(wo->wo_pid); <<
1685 <<
1686 init_waitqueue_func_entry(&wo->child_ <<
1687 wo->child_wait.private = current; <<
1688 add_wait_queue(¤t->signal->wait <<
1689 <<
1690 do { <<
1691 set_current_state(TASK_INTERR <<
1692 retval = __do_wait(wo); <<
1693 if (retval != -ERESTARTSYS) <<
1694 break; <<
1695 if (signal_pending(current)) <<
1696 break; <<
1697 schedule(); <<
1698 } while (1); <<
1699 <<
1700 __set_current_state(TASK_RUNNING); 1542 __set_current_state(TASK_RUNNING);
1701 remove_wait_queue(¤t->signal->w 1543 remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait);
1702 return retval; 1544 return retval;
1703 } 1545 }
1704 1546
1705 int kernel_waitid_prepare(struct wait_opts *w !! 1547 static long kernel_waitid(int which, pid_t upid, struct waitid_info *infop,
1706 struct waitid_info !! 1548 int options, struct rusage *ru)
1707 struct rusage *ru) <<
1708 { 1549 {
1709 unsigned int f_flags = 0; !! 1550 struct wait_opts wo;
1710 struct pid *pid = NULL; 1551 struct pid *pid = NULL;
1711 enum pid_type type; 1552 enum pid_type type;
>> 1553 long ret;
>> 1554 unsigned int f_flags = 0;
1712 1555
1713 if (options & ~(WNOHANG|WNOWAIT|WEXIT 1556 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED|
1714 __WNOTHREAD|__WCLONE| 1557 __WNOTHREAD|__WCLONE|__WALL))
1715 return -EINVAL; 1558 return -EINVAL;
1716 if (!(options & (WEXITED|WSTOPPED|WCO 1559 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1717 return -EINVAL; 1560 return -EINVAL;
1718 1561
1719 switch (which) { 1562 switch (which) {
1720 case P_ALL: 1563 case P_ALL:
1721 type = PIDTYPE_MAX; 1564 type = PIDTYPE_MAX;
1722 break; 1565 break;
1723 case P_PID: 1566 case P_PID:
1724 type = PIDTYPE_PID; 1567 type = PIDTYPE_PID;
1725 if (upid <= 0) 1568 if (upid <= 0)
1726 return -EINVAL; 1569 return -EINVAL;
1727 1570
1728 pid = find_get_pid(upid); 1571 pid = find_get_pid(upid);
1729 break; 1572 break;
1730 case P_PGID: 1573 case P_PGID:
1731 type = PIDTYPE_PGID; 1574 type = PIDTYPE_PGID;
1732 if (upid < 0) 1575 if (upid < 0)
1733 return -EINVAL; 1576 return -EINVAL;
1734 1577
1735 if (upid) 1578 if (upid)
1736 pid = find_get_pid(up 1579 pid = find_get_pid(upid);
1737 else 1580 else
1738 pid = get_task_pid(cu 1581 pid = get_task_pid(current, PIDTYPE_PGID);
1739 break; 1582 break;
1740 case P_PIDFD: 1583 case P_PIDFD:
1741 type = PIDTYPE_PID; 1584 type = PIDTYPE_PID;
1742 if (upid < 0) 1585 if (upid < 0)
1743 return -EINVAL; 1586 return -EINVAL;
1744 1587
1745 pid = pidfd_get_pid(upid, &f_ 1588 pid = pidfd_get_pid(upid, &f_flags);
1746 if (IS_ERR(pid)) 1589 if (IS_ERR(pid))
1747 return PTR_ERR(pid); 1590 return PTR_ERR(pid);
1748 1591
1749 break; 1592 break;
1750 default: 1593 default:
1751 return -EINVAL; 1594 return -EINVAL;
1752 } 1595 }
1753 1596
1754 wo->wo_type = type; !! 1597 wo.wo_type = type;
1755 wo->wo_pid = pid; !! 1598 wo.wo_pid = pid;
1756 wo->wo_flags = options; !! 1599 wo.wo_flags = options;
1757 wo->wo_info = infop; !! 1600 wo.wo_info = infop;
1758 wo->wo_rusage = ru; !! 1601 wo.wo_rusage = ru;
1759 if (f_flags & O_NONBLOCK) 1602 if (f_flags & O_NONBLOCK)
1760 wo->wo_flags |= WNOHANG; !! 1603 wo.wo_flags |= WNOHANG;
1761 <<
1762 return 0; <<
1763 } <<
1764 <<
1765 static long kernel_waitid(int which, pid_t up <<
1766 int options, struct <<
1767 { <<
1768 struct wait_opts wo; <<
1769 long ret; <<
1770 <<
1771 ret = kernel_waitid_prepare(&wo, whic <<
1772 if (ret) <<
1773 return ret; <<
1774 1604
1775 ret = do_wait(&wo); 1605 ret = do_wait(&wo);
1776 if (!ret && !(options & WNOHANG) && ( !! 1606 if (!ret && !(options & WNOHANG) && (f_flags & O_NONBLOCK))
1777 ret = -EAGAIN; 1607 ret = -EAGAIN;
1778 1608
1779 put_pid(wo.wo_pid); !! 1609 put_pid(pid);
1780 return ret; 1610 return ret;
1781 } 1611 }
1782 1612
1783 SYSCALL_DEFINE5(waitid, int, which, pid_t, up 1613 SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *,
1784 infop, int, options, struct r 1614 infop, int, options, struct rusage __user *, ru)
1785 { 1615 {
1786 struct rusage r; 1616 struct rusage r;
1787 struct waitid_info info = {.status = 1617 struct waitid_info info = {.status = 0};
1788 long err = kernel_waitid(which, upid, 1618 long err = kernel_waitid(which, upid, &info, options, ru ? &r : NULL);
1789 int signo = 0; 1619 int signo = 0;
1790 1620
1791 if (err > 0) { 1621 if (err > 0) {
1792 signo = SIGCHLD; 1622 signo = SIGCHLD;
1793 err = 0; 1623 err = 0;
1794 if (ru && copy_to_user(ru, &r 1624 if (ru && copy_to_user(ru, &r, sizeof(struct rusage)))
1795 return -EFAULT; 1625 return -EFAULT;
1796 } 1626 }
1797 if (!infop) 1627 if (!infop)
1798 return err; 1628 return err;
1799 1629
1800 if (!user_write_access_begin(infop, s 1630 if (!user_write_access_begin(infop, sizeof(*infop)))
1801 return -EFAULT; 1631 return -EFAULT;
1802 1632
1803 unsafe_put_user(signo, &infop->si_sig 1633 unsafe_put_user(signo, &infop->si_signo, Efault);
1804 unsafe_put_user(0, &infop->si_errno, 1634 unsafe_put_user(0, &infop->si_errno, Efault);
1805 unsafe_put_user(info.cause, &infop->s 1635 unsafe_put_user(info.cause, &infop->si_code, Efault);
1806 unsafe_put_user(info.pid, &infop->si_ 1636 unsafe_put_user(info.pid, &infop->si_pid, Efault);
1807 unsafe_put_user(info.uid, &infop->si_ 1637 unsafe_put_user(info.uid, &infop->si_uid, Efault);
1808 unsafe_put_user(info.status, &infop-> 1638 unsafe_put_user(info.status, &infop->si_status, Efault);
1809 user_write_access_end(); 1639 user_write_access_end();
1810 return err; 1640 return err;
1811 Efault: 1641 Efault:
1812 user_write_access_end(); 1642 user_write_access_end();
1813 return -EFAULT; 1643 return -EFAULT;
1814 } 1644 }
1815 1645
1816 long kernel_wait4(pid_t upid, int __user *sta 1646 long kernel_wait4(pid_t upid, int __user *stat_addr, int options,
1817 struct rusage *ru) 1647 struct rusage *ru)
1818 { 1648 {
1819 struct wait_opts wo; 1649 struct wait_opts wo;
1820 struct pid *pid = NULL; 1650 struct pid *pid = NULL;
1821 enum pid_type type; 1651 enum pid_type type;
1822 long ret; 1652 long ret;
1823 1653
1824 if (options & ~(WNOHANG|WUNTRACED|WCO 1654 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1825 __WNOTHREAD|__WCLONE| 1655 __WNOTHREAD|__WCLONE|__WALL))
1826 return -EINVAL; 1656 return -EINVAL;
1827 1657
1828 /* -INT_MIN is not defined */ 1658 /* -INT_MIN is not defined */
1829 if (upid == INT_MIN) 1659 if (upid == INT_MIN)
1830 return -ESRCH; 1660 return -ESRCH;
1831 1661
1832 if (upid == -1) 1662 if (upid == -1)
1833 type = PIDTYPE_MAX; 1663 type = PIDTYPE_MAX;
1834 else if (upid < 0) { 1664 else if (upid < 0) {
1835 type = PIDTYPE_PGID; 1665 type = PIDTYPE_PGID;
1836 pid = find_get_pid(-upid); 1666 pid = find_get_pid(-upid);
1837 } else if (upid == 0) { 1667 } else if (upid == 0) {
1838 type = PIDTYPE_PGID; 1668 type = PIDTYPE_PGID;
1839 pid = get_task_pid(current, P 1669 pid = get_task_pid(current, PIDTYPE_PGID);
1840 } else /* upid > 0 */ { 1670 } else /* upid > 0 */ {
1841 type = PIDTYPE_PID; 1671 type = PIDTYPE_PID;
1842 pid = find_get_pid(upid); 1672 pid = find_get_pid(upid);
1843 } 1673 }
1844 1674
1845 wo.wo_type = type; 1675 wo.wo_type = type;
1846 wo.wo_pid = pid; 1676 wo.wo_pid = pid;
1847 wo.wo_flags = options | WEXITED; 1677 wo.wo_flags = options | WEXITED;
1848 wo.wo_info = NULL; 1678 wo.wo_info = NULL;
1849 wo.wo_stat = 0; 1679 wo.wo_stat = 0;
1850 wo.wo_rusage = ru; 1680 wo.wo_rusage = ru;
1851 ret = do_wait(&wo); 1681 ret = do_wait(&wo);
1852 put_pid(pid); 1682 put_pid(pid);
1853 if (ret > 0 && stat_addr && put_user( 1683 if (ret > 0 && stat_addr && put_user(wo.wo_stat, stat_addr))
1854 ret = -EFAULT; 1684 ret = -EFAULT;
1855 1685
1856 return ret; 1686 return ret;
1857 } 1687 }
1858 1688
1859 int kernel_wait(pid_t pid, int *stat) 1689 int kernel_wait(pid_t pid, int *stat)
1860 { 1690 {
1861 struct wait_opts wo = { 1691 struct wait_opts wo = {
1862 .wo_type = PIDTYPE_PID 1692 .wo_type = PIDTYPE_PID,
1863 .wo_pid = find_get_pi 1693 .wo_pid = find_get_pid(pid),
1864 .wo_flags = WEXITED, 1694 .wo_flags = WEXITED,
1865 }; 1695 };
1866 int ret; 1696 int ret;
1867 1697
1868 ret = do_wait(&wo); 1698 ret = do_wait(&wo);
1869 if (ret > 0 && wo.wo_stat) 1699 if (ret > 0 && wo.wo_stat)
1870 *stat = wo.wo_stat; 1700 *stat = wo.wo_stat;
1871 put_pid(wo.wo_pid); 1701 put_pid(wo.wo_pid);
1872 return ret; 1702 return ret;
1873 } 1703 }
1874 1704
1875 SYSCALL_DEFINE4(wait4, pid_t, upid, int __use 1705 SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr,
1876 int, options, struct rusage _ 1706 int, options, struct rusage __user *, ru)
1877 { 1707 {
1878 struct rusage r; 1708 struct rusage r;
1879 long err = kernel_wait4(upid, stat_ad 1709 long err = kernel_wait4(upid, stat_addr, options, ru ? &r : NULL);
1880 1710
1881 if (err > 0) { 1711 if (err > 0) {
1882 if (ru && copy_to_user(ru, &r 1712 if (ru && copy_to_user(ru, &r, sizeof(struct rusage)))
1883 return -EFAULT; 1713 return -EFAULT;
1884 } 1714 }
1885 return err; 1715 return err;
1886 } 1716 }
1887 1717
1888 #ifdef __ARCH_WANT_SYS_WAITPID 1718 #ifdef __ARCH_WANT_SYS_WAITPID
1889 1719
1890 /* 1720 /*
1891 * sys_waitpid() remains for compatibility. w 1721 * sys_waitpid() remains for compatibility. waitpid() should be
1892 * implemented by calling sys_wait4() from li 1722 * implemented by calling sys_wait4() from libc.a.
1893 */ 1723 */
1894 SYSCALL_DEFINE3(waitpid, pid_t, pid, int __us 1724 SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options)
1895 { 1725 {
1896 return kernel_wait4(pid, stat_addr, o 1726 return kernel_wait4(pid, stat_addr, options, NULL);
1897 } 1727 }
1898 1728
1899 #endif 1729 #endif
1900 1730
1901 #ifdef CONFIG_COMPAT 1731 #ifdef CONFIG_COMPAT
1902 COMPAT_SYSCALL_DEFINE4(wait4, 1732 COMPAT_SYSCALL_DEFINE4(wait4,
1903 compat_pid_t, pid, 1733 compat_pid_t, pid,
1904 compat_uint_t __user *, stat_addr, 1734 compat_uint_t __user *, stat_addr,
1905 int, options, 1735 int, options,
1906 struct compat_rusage __user *, ru) 1736 struct compat_rusage __user *, ru)
1907 { 1737 {
1908 struct rusage r; 1738 struct rusage r;
1909 long err = kernel_wait4(pid, stat_add 1739 long err = kernel_wait4(pid, stat_addr, options, ru ? &r : NULL);
1910 if (err > 0) { 1740 if (err > 0) {
1911 if (ru && put_compat_rusage(& 1741 if (ru && put_compat_rusage(&r, ru))
1912 return -EFAULT; 1742 return -EFAULT;
1913 } 1743 }
1914 return err; 1744 return err;
1915 } 1745 }
1916 1746
1917 COMPAT_SYSCALL_DEFINE5(waitid, 1747 COMPAT_SYSCALL_DEFINE5(waitid,
1918 int, which, compat_pid_t, pid 1748 int, which, compat_pid_t, pid,
1919 struct compat_siginfo __user 1749 struct compat_siginfo __user *, infop, int, options,
1920 struct compat_rusage __user * 1750 struct compat_rusage __user *, uru)
1921 { 1751 {
1922 struct rusage ru; 1752 struct rusage ru;
1923 struct waitid_info info = {.status = 1753 struct waitid_info info = {.status = 0};
1924 long err = kernel_waitid(which, pid, 1754 long err = kernel_waitid(which, pid, &info, options, uru ? &ru : NULL);
1925 int signo = 0; 1755 int signo = 0;
1926 if (err > 0) { 1756 if (err > 0) {
1927 signo = SIGCHLD; 1757 signo = SIGCHLD;
1928 err = 0; 1758 err = 0;
1929 if (uru) { 1759 if (uru) {
1930 /* kernel_waitid() ov 1760 /* kernel_waitid() overwrites everything in ru */
1931 if (COMPAT_USE_64BIT_ 1761 if (COMPAT_USE_64BIT_TIME)
1932 err = copy_to 1762 err = copy_to_user(uru, &ru, sizeof(ru));
1933 else 1763 else
1934 err = put_com 1764 err = put_compat_rusage(&ru, uru);
1935 if (err) 1765 if (err)
1936 return -EFAUL 1766 return -EFAULT;
1937 } 1767 }
1938 } 1768 }
1939 1769
1940 if (!infop) 1770 if (!infop)
1941 return err; 1771 return err;
1942 1772
1943 if (!user_write_access_begin(infop, s 1773 if (!user_write_access_begin(infop, sizeof(*infop)))
1944 return -EFAULT; 1774 return -EFAULT;
1945 1775
1946 unsafe_put_user(signo, &infop->si_sig 1776 unsafe_put_user(signo, &infop->si_signo, Efault);
1947 unsafe_put_user(0, &infop->si_errno, 1777 unsafe_put_user(0, &infop->si_errno, Efault);
1948 unsafe_put_user(info.cause, &infop->s 1778 unsafe_put_user(info.cause, &infop->si_code, Efault);
1949 unsafe_put_user(info.pid, &infop->si_ 1779 unsafe_put_user(info.pid, &infop->si_pid, Efault);
1950 unsafe_put_user(info.uid, &infop->si_ 1780 unsafe_put_user(info.uid, &infop->si_uid, Efault);
1951 unsafe_put_user(info.status, &infop-> 1781 unsafe_put_user(info.status, &infop->si_status, Efault);
1952 user_write_access_end(); 1782 user_write_access_end();
1953 return err; 1783 return err;
1954 Efault: 1784 Efault:
1955 user_write_access_end(); 1785 user_write_access_end();
1956 return -EFAULT; 1786 return -EFAULT;
1957 } 1787 }
1958 #endif 1788 #endif
1959 1789
1960 /* !! 1790 /**
1961 * This needs to be __function_aligned as GCC !! 1791 * thread_group_exited - check that a thread group has exited
1962 * implementation of abort() cold and drops a !! 1792 * @pid: tgid of thread group to be checked.
1963 * -falign-functions=N. <<
1964 * 1793 *
1965 * See https://gcc.gnu.org/bugzilla/show_bug. !! 1794 * Test if the thread group represented by tgid has exited (all
>> 1795 * threads are zombies, dead or completely gone).
>> 1796 *
>> 1797 * Return: true if the thread group has exited. false otherwise.
1966 */ 1798 */
1967 __weak __function_aligned void abort(void) !! 1799 bool thread_group_exited(struct pid *pid)
>> 1800 {
>> 1801 struct task_struct *task;
>> 1802 bool exited;
>> 1803
>> 1804 rcu_read_lock();
>> 1805 task = pid_task(pid, PIDTYPE_PID);
>> 1806 exited = !task ||
>> 1807 (READ_ONCE(task->exit_state) && thread_group_empty(task));
>> 1808 rcu_read_unlock();
>> 1809
>> 1810 return exited;
>> 1811 }
>> 1812 EXPORT_SYMBOL(thread_group_exited);
>> 1813
>> 1814 __weak void abort(void)
1968 { 1815 {
1969 BUG(); 1816 BUG();
1970 1817
1971 /* if that doesn't kill us, halt */ 1818 /* if that doesn't kill us, halt */
1972 panic("Oops failed to kill thread"); 1819 panic("Oops failed to kill thread");
1973 } 1820 }
1974 EXPORT_SYMBOL(abort); 1821 EXPORT_SYMBOL(abort);
1975 1822
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.