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