Commit 0af83220a4e41797edb0915b177dae84e6e6b57f
Committed by
Greg Kroah-Hartman
1 parent
de661dfb1d
Exists in
ti-linux-3.14.y
and in
2 other branches
exit: call disassociate_ctty() before exit_task_namespaces()
commit c39df5fa37b0623589508c95515b4aa1531c524e upstream. Commit 8aac62706ada ("move exit_task_namespaces() outside of exit_notify()") breaks pppd and the exiting service crashes the kernel: BUG: unable to handle kernel NULL pointer dereference at 0000000000000028 IP: ppp_register_channel+0x13/0x20 [ppp_generic] Call Trace: ppp_asynctty_open+0x12b/0x170 [ppp_async] tty_ldisc_open.isra.2+0x27/0x60 tty_ldisc_hangup+0x1e3/0x220 __tty_hangup+0x2c4/0x440 disassociate_ctty+0x61/0x270 do_exit+0x7f2/0xa50 ppp_register_channel() needs ->net_ns and current->nsproxy == NULL. Move disassociate_ctty() before exit_task_namespaces(), it doesn't make sense to delay it after perf_event_exit_task() or cgroup_exit(). This also allows to use task_work_add() inside the (nontrivial) code paths in disassociate_ctty(). Investigated by Peter Hurley. Signed-off-by: Oleg Nesterov <oleg@redhat.com> Reported-by: Sree Harsha Totakura <sreeharsha@totakura.in> Cc: Peter Hurley <peter@hurleysoftware.com> Cc: Sree Harsha Totakura <sreeharsha@totakura.in> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Cc: Jeff Dike <jdike@addtoit.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Andrey Vagin <avagin@openvz.org> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Showing 1 changed file with 2 additions and 4 deletions Inline Diff
kernel/exit.c
1 | /* | 1 | /* |
2 | * linux/kernel/exit.c | 2 | * linux/kernel/exit.c |
3 | * | 3 | * |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | 4 | * Copyright (C) 1991, 1992 Linus Torvalds |
5 | */ | 5 | */ |
6 | 6 | ||
7 | #include <linux/mm.h> | 7 | #include <linux/mm.h> |
8 | #include <linux/slab.h> | 8 | #include <linux/slab.h> |
9 | #include <linux/interrupt.h> | 9 | #include <linux/interrupt.h> |
10 | #include <linux/module.h> | 10 | #include <linux/module.h> |
11 | #include <linux/capability.h> | 11 | #include <linux/capability.h> |
12 | #include <linux/completion.h> | 12 | #include <linux/completion.h> |
13 | #include <linux/personality.h> | 13 | #include <linux/personality.h> |
14 | #include <linux/tty.h> | 14 | #include <linux/tty.h> |
15 | #include <linux/iocontext.h> | 15 | #include <linux/iocontext.h> |
16 | #include <linux/key.h> | 16 | #include <linux/key.h> |
17 | #include <linux/security.h> | 17 | #include <linux/security.h> |
18 | #include <linux/cpu.h> | 18 | #include <linux/cpu.h> |
19 | #include <linux/acct.h> | 19 | #include <linux/acct.h> |
20 | #include <linux/tsacct_kern.h> | 20 | #include <linux/tsacct_kern.h> |
21 | #include <linux/file.h> | 21 | #include <linux/file.h> |
22 | #include <linux/fdtable.h> | 22 | #include <linux/fdtable.h> |
23 | #include <linux/freezer.h> | 23 | #include <linux/freezer.h> |
24 | #include <linux/binfmts.h> | 24 | #include <linux/binfmts.h> |
25 | #include <linux/nsproxy.h> | 25 | #include <linux/nsproxy.h> |
26 | #include <linux/pid_namespace.h> | 26 | #include <linux/pid_namespace.h> |
27 | #include <linux/ptrace.h> | 27 | #include <linux/ptrace.h> |
28 | #include <linux/profile.h> | 28 | #include <linux/profile.h> |
29 | #include <linux/mount.h> | 29 | #include <linux/mount.h> |
30 | #include <linux/proc_fs.h> | 30 | #include <linux/proc_fs.h> |
31 | #include <linux/kthread.h> | 31 | #include <linux/kthread.h> |
32 | #include <linux/mempolicy.h> | 32 | #include <linux/mempolicy.h> |
33 | #include <linux/taskstats_kern.h> | 33 | #include <linux/taskstats_kern.h> |
34 | #include <linux/delayacct.h> | 34 | #include <linux/delayacct.h> |
35 | #include <linux/cgroup.h> | 35 | #include <linux/cgroup.h> |
36 | #include <linux/syscalls.h> | 36 | #include <linux/syscalls.h> |
37 | #include <linux/signal.h> | 37 | #include <linux/signal.h> |
38 | #include <linux/posix-timers.h> | 38 | #include <linux/posix-timers.h> |
39 | #include <linux/cn_proc.h> | 39 | #include <linux/cn_proc.h> |
40 | #include <linux/mutex.h> | 40 | #include <linux/mutex.h> |
41 | #include <linux/futex.h> | 41 | #include <linux/futex.h> |
42 | #include <linux/pipe_fs_i.h> | 42 | #include <linux/pipe_fs_i.h> |
43 | #include <linux/audit.h> /* for audit_free() */ | 43 | #include <linux/audit.h> /* for audit_free() */ |
44 | #include <linux/resource.h> | 44 | #include <linux/resource.h> |
45 | #include <linux/blkdev.h> | 45 | #include <linux/blkdev.h> |
46 | #include <linux/task_io_accounting_ops.h> | 46 | #include <linux/task_io_accounting_ops.h> |
47 | #include <linux/tracehook.h> | 47 | #include <linux/tracehook.h> |
48 | #include <linux/fs_struct.h> | 48 | #include <linux/fs_struct.h> |
49 | #include <linux/init_task.h> | 49 | #include <linux/init_task.h> |
50 | #include <linux/perf_event.h> | 50 | #include <linux/perf_event.h> |
51 | #include <trace/events/sched.h> | 51 | #include <trace/events/sched.h> |
52 | #include <linux/hw_breakpoint.h> | 52 | #include <linux/hw_breakpoint.h> |
53 | #include <linux/oom.h> | 53 | #include <linux/oom.h> |
54 | #include <linux/writeback.h> | 54 | #include <linux/writeback.h> |
55 | #include <linux/shm.h> | 55 | #include <linux/shm.h> |
56 | 56 | ||
57 | #include <asm/uaccess.h> | 57 | #include <asm/uaccess.h> |
58 | #include <asm/unistd.h> | 58 | #include <asm/unistd.h> |
59 | #include <asm/pgtable.h> | 59 | #include <asm/pgtable.h> |
60 | #include <asm/mmu_context.h> | 60 | #include <asm/mmu_context.h> |
61 | 61 | ||
62 | static void exit_mm(struct task_struct * tsk); | 62 | static void exit_mm(struct task_struct * tsk); |
63 | 63 | ||
64 | static void __unhash_process(struct task_struct *p, bool group_dead) | 64 | static void __unhash_process(struct task_struct *p, bool group_dead) |
65 | { | 65 | { |
66 | nr_threads--; | 66 | nr_threads--; |
67 | detach_pid(p, PIDTYPE_PID); | 67 | detach_pid(p, PIDTYPE_PID); |
68 | if (group_dead) { | 68 | if (group_dead) { |
69 | detach_pid(p, PIDTYPE_PGID); | 69 | detach_pid(p, PIDTYPE_PGID); |
70 | detach_pid(p, PIDTYPE_SID); | 70 | detach_pid(p, PIDTYPE_SID); |
71 | 71 | ||
72 | list_del_rcu(&p->tasks); | 72 | list_del_rcu(&p->tasks); |
73 | list_del_init(&p->sibling); | 73 | list_del_init(&p->sibling); |
74 | __this_cpu_dec(process_counts); | 74 | __this_cpu_dec(process_counts); |
75 | } | 75 | } |
76 | list_del_rcu(&p->thread_group); | 76 | list_del_rcu(&p->thread_group); |
77 | list_del_rcu(&p->thread_node); | 77 | list_del_rcu(&p->thread_node); |
78 | } | 78 | } |
79 | 79 | ||
80 | /* | 80 | /* |
81 | * This function expects the tasklist_lock write-locked. | 81 | * This function expects the tasklist_lock write-locked. |
82 | */ | 82 | */ |
83 | static void __exit_signal(struct task_struct *tsk) | 83 | static void __exit_signal(struct task_struct *tsk) |
84 | { | 84 | { |
85 | struct signal_struct *sig = tsk->signal; | 85 | struct signal_struct *sig = tsk->signal; |
86 | bool group_dead = thread_group_leader(tsk); | 86 | bool group_dead = thread_group_leader(tsk); |
87 | struct sighand_struct *sighand; | 87 | struct sighand_struct *sighand; |
88 | struct tty_struct *uninitialized_var(tty); | 88 | struct tty_struct *uninitialized_var(tty); |
89 | cputime_t utime, stime; | 89 | cputime_t utime, stime; |
90 | 90 | ||
91 | sighand = rcu_dereference_check(tsk->sighand, | 91 | sighand = rcu_dereference_check(tsk->sighand, |
92 | lockdep_tasklist_lock_is_held()); | 92 | lockdep_tasklist_lock_is_held()); |
93 | spin_lock(&sighand->siglock); | 93 | spin_lock(&sighand->siglock); |
94 | 94 | ||
95 | posix_cpu_timers_exit(tsk); | 95 | posix_cpu_timers_exit(tsk); |
96 | if (group_dead) { | 96 | if (group_dead) { |
97 | posix_cpu_timers_exit_group(tsk); | 97 | posix_cpu_timers_exit_group(tsk); |
98 | tty = sig->tty; | 98 | tty = sig->tty; |
99 | sig->tty = NULL; | 99 | sig->tty = NULL; |
100 | } else { | 100 | } else { |
101 | /* | 101 | /* |
102 | * This can only happen if the caller is de_thread(). | 102 | * This can only happen if the caller is de_thread(). |
103 | * FIXME: this is the temporary hack, we should teach | 103 | * FIXME: this is the temporary hack, we should teach |
104 | * posix-cpu-timers to handle this case correctly. | 104 | * posix-cpu-timers to handle this case correctly. |
105 | */ | 105 | */ |
106 | if (unlikely(has_group_leader_pid(tsk))) | 106 | if (unlikely(has_group_leader_pid(tsk))) |
107 | posix_cpu_timers_exit_group(tsk); | 107 | posix_cpu_timers_exit_group(tsk); |
108 | 108 | ||
109 | /* | 109 | /* |
110 | * If there is any task waiting for the group exit | 110 | * If there is any task waiting for the group exit |
111 | * then notify it: | 111 | * then notify it: |
112 | */ | 112 | */ |
113 | if (sig->notify_count > 0 && !--sig->notify_count) | 113 | if (sig->notify_count > 0 && !--sig->notify_count) |
114 | wake_up_process(sig->group_exit_task); | 114 | wake_up_process(sig->group_exit_task); |
115 | 115 | ||
116 | if (tsk == sig->curr_target) | 116 | if (tsk == sig->curr_target) |
117 | sig->curr_target = next_thread(tsk); | 117 | sig->curr_target = next_thread(tsk); |
118 | /* | 118 | /* |
119 | * Accumulate here the counters for all threads but the | 119 | * Accumulate here the counters for all threads but the |
120 | * group leader as they die, so they can be added into | 120 | * group leader as they die, so they can be added into |
121 | * the process-wide totals when those are taken. | 121 | * the process-wide totals when those are taken. |
122 | * The group leader stays around as a zombie as long | 122 | * The group leader stays around as a zombie as long |
123 | * as there are other threads. When it gets reaped, | 123 | * as there are other threads. When it gets reaped, |
124 | * the exit.c code will add its counts into these totals. | 124 | * the exit.c code will add its counts into these totals. |
125 | * We won't ever get here for the group leader, since it | 125 | * We won't ever get here for the group leader, since it |
126 | * will have been the last reference on the signal_struct. | 126 | * will have been the last reference on the signal_struct. |
127 | */ | 127 | */ |
128 | task_cputime(tsk, &utime, &stime); | 128 | task_cputime(tsk, &utime, &stime); |
129 | sig->utime += utime; | 129 | sig->utime += utime; |
130 | sig->stime += stime; | 130 | sig->stime += stime; |
131 | sig->gtime += task_gtime(tsk); | 131 | sig->gtime += task_gtime(tsk); |
132 | sig->min_flt += tsk->min_flt; | 132 | sig->min_flt += tsk->min_flt; |
133 | sig->maj_flt += tsk->maj_flt; | 133 | sig->maj_flt += tsk->maj_flt; |
134 | sig->nvcsw += tsk->nvcsw; | 134 | sig->nvcsw += tsk->nvcsw; |
135 | sig->nivcsw += tsk->nivcsw; | 135 | sig->nivcsw += tsk->nivcsw; |
136 | sig->inblock += task_io_get_inblock(tsk); | 136 | sig->inblock += task_io_get_inblock(tsk); |
137 | sig->oublock += task_io_get_oublock(tsk); | 137 | sig->oublock += task_io_get_oublock(tsk); |
138 | task_io_accounting_add(&sig->ioac, &tsk->ioac); | 138 | task_io_accounting_add(&sig->ioac, &tsk->ioac); |
139 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; | 139 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; |
140 | } | 140 | } |
141 | 141 | ||
142 | sig->nr_threads--; | 142 | sig->nr_threads--; |
143 | __unhash_process(tsk, group_dead); | 143 | __unhash_process(tsk, group_dead); |
144 | 144 | ||
145 | /* | 145 | /* |
146 | * Do this under ->siglock, we can race with another thread | 146 | * Do this under ->siglock, we can race with another thread |
147 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. | 147 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. |
148 | */ | 148 | */ |
149 | flush_sigqueue(&tsk->pending); | 149 | flush_sigqueue(&tsk->pending); |
150 | tsk->sighand = NULL; | 150 | tsk->sighand = NULL; |
151 | spin_unlock(&sighand->siglock); | 151 | spin_unlock(&sighand->siglock); |
152 | 152 | ||
153 | __cleanup_sighand(sighand); | 153 | __cleanup_sighand(sighand); |
154 | clear_tsk_thread_flag(tsk,TIF_SIGPENDING); | 154 | clear_tsk_thread_flag(tsk,TIF_SIGPENDING); |
155 | if (group_dead) { | 155 | if (group_dead) { |
156 | flush_sigqueue(&sig->shared_pending); | 156 | flush_sigqueue(&sig->shared_pending); |
157 | tty_kref_put(tty); | 157 | tty_kref_put(tty); |
158 | } | 158 | } |
159 | } | 159 | } |
160 | 160 | ||
161 | static void delayed_put_task_struct(struct rcu_head *rhp) | 161 | static void delayed_put_task_struct(struct rcu_head *rhp) |
162 | { | 162 | { |
163 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); | 163 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
164 | 164 | ||
165 | perf_event_delayed_put(tsk); | 165 | perf_event_delayed_put(tsk); |
166 | trace_sched_process_free(tsk); | 166 | trace_sched_process_free(tsk); |
167 | put_task_struct(tsk); | 167 | put_task_struct(tsk); |
168 | } | 168 | } |
169 | 169 | ||
170 | 170 | ||
171 | void release_task(struct task_struct * p) | 171 | void release_task(struct task_struct * p) |
172 | { | 172 | { |
173 | struct task_struct *leader; | 173 | struct task_struct *leader; |
174 | int zap_leader; | 174 | int zap_leader; |
175 | repeat: | 175 | repeat: |
176 | /* don't need to get the RCU readlock here - the process is dead and | 176 | /* don't need to get the RCU readlock here - the process is dead and |
177 | * can't be modifying its own credentials. But shut RCU-lockdep up */ | 177 | * can't be modifying its own credentials. But shut RCU-lockdep up */ |
178 | rcu_read_lock(); | 178 | rcu_read_lock(); |
179 | atomic_dec(&__task_cred(p)->user->processes); | 179 | atomic_dec(&__task_cred(p)->user->processes); |
180 | rcu_read_unlock(); | 180 | rcu_read_unlock(); |
181 | 181 | ||
182 | proc_flush_task(p); | 182 | proc_flush_task(p); |
183 | 183 | ||
184 | write_lock_irq(&tasklist_lock); | 184 | write_lock_irq(&tasklist_lock); |
185 | ptrace_release_task(p); | 185 | ptrace_release_task(p); |
186 | __exit_signal(p); | 186 | __exit_signal(p); |
187 | 187 | ||
188 | /* | 188 | /* |
189 | * If we are the last non-leader member of the thread | 189 | * If we are the last non-leader member of the thread |
190 | * group, and the leader is zombie, then notify the | 190 | * group, and the leader is zombie, then notify the |
191 | * group leader's parent process. (if it wants notification.) | 191 | * group leader's parent process. (if it wants notification.) |
192 | */ | 192 | */ |
193 | zap_leader = 0; | 193 | zap_leader = 0; |
194 | leader = p->group_leader; | 194 | leader = p->group_leader; |
195 | if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) { | 195 | if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) { |
196 | /* | 196 | /* |
197 | * If we were the last child thread and the leader has | 197 | * If we were the last child thread and the leader has |
198 | * exited already, and the leader's parent ignores SIGCHLD, | 198 | * exited already, and the leader's parent ignores SIGCHLD, |
199 | * then we are the one who should release the leader. | 199 | * then we are the one who should release the leader. |
200 | */ | 200 | */ |
201 | zap_leader = do_notify_parent(leader, leader->exit_signal); | 201 | zap_leader = do_notify_parent(leader, leader->exit_signal); |
202 | if (zap_leader) | 202 | if (zap_leader) |
203 | leader->exit_state = EXIT_DEAD; | 203 | leader->exit_state = EXIT_DEAD; |
204 | } | 204 | } |
205 | 205 | ||
206 | write_unlock_irq(&tasklist_lock); | 206 | write_unlock_irq(&tasklist_lock); |
207 | release_thread(p); | 207 | release_thread(p); |
208 | call_rcu(&p->rcu, delayed_put_task_struct); | 208 | call_rcu(&p->rcu, delayed_put_task_struct); |
209 | 209 | ||
210 | p = leader; | 210 | p = leader; |
211 | if (unlikely(zap_leader)) | 211 | if (unlikely(zap_leader)) |
212 | goto repeat; | 212 | goto repeat; |
213 | } | 213 | } |
214 | 214 | ||
215 | /* | 215 | /* |
216 | * This checks not only the pgrp, but falls back on the pid if no | 216 | * This checks not only the pgrp, but falls back on the pid if no |
217 | * satisfactory pgrp is found. I dunno - gdb doesn't work correctly | 217 | * satisfactory pgrp is found. I dunno - gdb doesn't work correctly |
218 | * without this... | 218 | * without this... |
219 | * | 219 | * |
220 | * The caller must hold rcu lock or the tasklist lock. | 220 | * The caller must hold rcu lock or the tasklist lock. |
221 | */ | 221 | */ |
222 | struct pid *session_of_pgrp(struct pid *pgrp) | 222 | struct pid *session_of_pgrp(struct pid *pgrp) |
223 | { | 223 | { |
224 | struct task_struct *p; | 224 | struct task_struct *p; |
225 | struct pid *sid = NULL; | 225 | struct pid *sid = NULL; |
226 | 226 | ||
227 | p = pid_task(pgrp, PIDTYPE_PGID); | 227 | p = pid_task(pgrp, PIDTYPE_PGID); |
228 | if (p == NULL) | 228 | if (p == NULL) |
229 | p = pid_task(pgrp, PIDTYPE_PID); | 229 | p = pid_task(pgrp, PIDTYPE_PID); |
230 | if (p != NULL) | 230 | if (p != NULL) |
231 | sid = task_session(p); | 231 | sid = task_session(p); |
232 | 232 | ||
233 | return sid; | 233 | return sid; |
234 | } | 234 | } |
235 | 235 | ||
236 | /* | 236 | /* |
237 | * Determine if a process group is "orphaned", according to the POSIX | 237 | * Determine if a process group is "orphaned", according to the POSIX |
238 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | 238 | * definition in 2.2.2.52. Orphaned process groups are not to be affected |
239 | * by terminal-generated stop signals. Newly orphaned process groups are | 239 | * by terminal-generated stop signals. Newly orphaned process groups are |
240 | * to receive a SIGHUP and a SIGCONT. | 240 | * to receive a SIGHUP and a SIGCONT. |
241 | * | 241 | * |
242 | * "I ask you, have you ever known what it is to be an orphan?" | 242 | * "I ask you, have you ever known what it is to be an orphan?" |
243 | */ | 243 | */ |
244 | static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task) | 244 | static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task) |
245 | { | 245 | { |
246 | struct task_struct *p; | 246 | struct task_struct *p; |
247 | 247 | ||
248 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { | 248 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
249 | if ((p == ignored_task) || | 249 | if ((p == ignored_task) || |
250 | (p->exit_state && thread_group_empty(p)) || | 250 | (p->exit_state && thread_group_empty(p)) || |
251 | is_global_init(p->real_parent)) | 251 | is_global_init(p->real_parent)) |
252 | continue; | 252 | continue; |
253 | 253 | ||
254 | if (task_pgrp(p->real_parent) != pgrp && | 254 | if (task_pgrp(p->real_parent) != pgrp && |
255 | task_session(p->real_parent) == task_session(p)) | 255 | task_session(p->real_parent) == task_session(p)) |
256 | return 0; | 256 | return 0; |
257 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); | 257 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
258 | 258 | ||
259 | return 1; | 259 | return 1; |
260 | } | 260 | } |
261 | 261 | ||
262 | int is_current_pgrp_orphaned(void) | 262 | int is_current_pgrp_orphaned(void) |
263 | { | 263 | { |
264 | int retval; | 264 | int retval; |
265 | 265 | ||
266 | read_lock(&tasklist_lock); | 266 | read_lock(&tasklist_lock); |
267 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); | 267 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
268 | read_unlock(&tasklist_lock); | 268 | read_unlock(&tasklist_lock); |
269 | 269 | ||
270 | return retval; | 270 | return retval; |
271 | } | 271 | } |
272 | 272 | ||
273 | static bool has_stopped_jobs(struct pid *pgrp) | 273 | static bool has_stopped_jobs(struct pid *pgrp) |
274 | { | 274 | { |
275 | struct task_struct *p; | 275 | struct task_struct *p; |
276 | 276 | ||
277 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { | 277 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
278 | if (p->signal->flags & SIGNAL_STOP_STOPPED) | 278 | if (p->signal->flags & SIGNAL_STOP_STOPPED) |
279 | return true; | 279 | return true; |
280 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); | 280 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
281 | 281 | ||
282 | return false; | 282 | return false; |
283 | } | 283 | } |
284 | 284 | ||
285 | /* | 285 | /* |
286 | * Check to see if any process groups have become orphaned as | 286 | * Check to see if any process groups have become orphaned as |
287 | * a result of our exiting, and if they have any stopped jobs, | 287 | * a result of our exiting, and if they have any stopped jobs, |
288 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | 288 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) |
289 | */ | 289 | */ |
290 | static void | 290 | static void |
291 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | 291 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) |
292 | { | 292 | { |
293 | struct pid *pgrp = task_pgrp(tsk); | 293 | struct pid *pgrp = task_pgrp(tsk); |
294 | struct task_struct *ignored_task = tsk; | 294 | struct task_struct *ignored_task = tsk; |
295 | 295 | ||
296 | if (!parent) | 296 | if (!parent) |
297 | /* exit: our father is in a different pgrp than | 297 | /* exit: our father is in a different pgrp than |
298 | * we are and we were the only connection outside. | 298 | * we are and we were the only connection outside. |
299 | */ | 299 | */ |
300 | parent = tsk->real_parent; | 300 | parent = tsk->real_parent; |
301 | else | 301 | else |
302 | /* reparent: our child is in a different pgrp than | 302 | /* reparent: our child is in a different pgrp than |
303 | * we are, and it was the only connection outside. | 303 | * we are, and it was the only connection outside. |
304 | */ | 304 | */ |
305 | ignored_task = NULL; | 305 | ignored_task = NULL; |
306 | 306 | ||
307 | if (task_pgrp(parent) != pgrp && | 307 | if (task_pgrp(parent) != pgrp && |
308 | task_session(parent) == task_session(tsk) && | 308 | task_session(parent) == task_session(tsk) && |
309 | will_become_orphaned_pgrp(pgrp, ignored_task) && | 309 | will_become_orphaned_pgrp(pgrp, ignored_task) && |
310 | has_stopped_jobs(pgrp)) { | 310 | has_stopped_jobs(pgrp)) { |
311 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | 311 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); |
312 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | 312 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); |
313 | } | 313 | } |
314 | } | 314 | } |
315 | 315 | ||
316 | /* | 316 | /* |
317 | * Let kernel threads use this to say that they allow a certain signal. | 317 | * Let kernel threads use this to say that they allow a certain signal. |
318 | * Must not be used if kthread was cloned with CLONE_SIGHAND. | 318 | * Must not be used if kthread was cloned with CLONE_SIGHAND. |
319 | */ | 319 | */ |
320 | int allow_signal(int sig) | 320 | int allow_signal(int sig) |
321 | { | 321 | { |
322 | if (!valid_signal(sig) || sig < 1) | 322 | if (!valid_signal(sig) || sig < 1) |
323 | return -EINVAL; | 323 | return -EINVAL; |
324 | 324 | ||
325 | spin_lock_irq(¤t->sighand->siglock); | 325 | spin_lock_irq(¤t->sighand->siglock); |
326 | /* This is only needed for daemonize()'ed kthreads */ | 326 | /* This is only needed for daemonize()'ed kthreads */ |
327 | sigdelset(¤t->blocked, sig); | 327 | sigdelset(¤t->blocked, sig); |
328 | /* | 328 | /* |
329 | * Kernel threads handle their own signals. Let the signal code | 329 | * Kernel threads handle their own signals. Let the signal code |
330 | * know it'll be handled, so that they don't get converted to | 330 | * know it'll be handled, so that they don't get converted to |
331 | * SIGKILL or just silently dropped. | 331 | * SIGKILL or just silently dropped. |
332 | */ | 332 | */ |
333 | current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2; | 333 | current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2; |
334 | recalc_sigpending(); | 334 | recalc_sigpending(); |
335 | spin_unlock_irq(¤t->sighand->siglock); | 335 | spin_unlock_irq(¤t->sighand->siglock); |
336 | return 0; | 336 | return 0; |
337 | } | 337 | } |
338 | 338 | ||
339 | EXPORT_SYMBOL(allow_signal); | 339 | EXPORT_SYMBOL(allow_signal); |
340 | 340 | ||
341 | int disallow_signal(int sig) | 341 | int disallow_signal(int sig) |
342 | { | 342 | { |
343 | if (!valid_signal(sig) || sig < 1) | 343 | if (!valid_signal(sig) || sig < 1) |
344 | return -EINVAL; | 344 | return -EINVAL; |
345 | 345 | ||
346 | spin_lock_irq(¤t->sighand->siglock); | 346 | spin_lock_irq(¤t->sighand->siglock); |
347 | current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN; | 347 | current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN; |
348 | recalc_sigpending(); | 348 | recalc_sigpending(); |
349 | spin_unlock_irq(¤t->sighand->siglock); | 349 | spin_unlock_irq(¤t->sighand->siglock); |
350 | return 0; | 350 | return 0; |
351 | } | 351 | } |
352 | 352 | ||
353 | EXPORT_SYMBOL(disallow_signal); | 353 | EXPORT_SYMBOL(disallow_signal); |
354 | 354 | ||
355 | #ifdef CONFIG_MM_OWNER | 355 | #ifdef CONFIG_MM_OWNER |
356 | /* | 356 | /* |
357 | * A task is exiting. If it owned this mm, find a new owner for the mm. | 357 | * A task is exiting. If it owned this mm, find a new owner for the mm. |
358 | */ | 358 | */ |
359 | void mm_update_next_owner(struct mm_struct *mm) | 359 | void mm_update_next_owner(struct mm_struct *mm) |
360 | { | 360 | { |
361 | struct task_struct *c, *g, *p = current; | 361 | struct task_struct *c, *g, *p = current; |
362 | 362 | ||
363 | retry: | 363 | retry: |
364 | /* | 364 | /* |
365 | * If the exiting or execing task is not the owner, it's | 365 | * If the exiting or execing task is not the owner, it's |
366 | * someone else's problem. | 366 | * someone else's problem. |
367 | */ | 367 | */ |
368 | if (mm->owner != p) | 368 | if (mm->owner != p) |
369 | return; | 369 | return; |
370 | /* | 370 | /* |
371 | * The current owner is exiting/execing and there are no other | 371 | * The current owner is exiting/execing and there are no other |
372 | * candidates. Do not leave the mm pointing to a possibly | 372 | * candidates. Do not leave the mm pointing to a possibly |
373 | * freed task structure. | 373 | * freed task structure. |
374 | */ | 374 | */ |
375 | if (atomic_read(&mm->mm_users) <= 1) { | 375 | if (atomic_read(&mm->mm_users) <= 1) { |
376 | mm->owner = NULL; | 376 | mm->owner = NULL; |
377 | return; | 377 | return; |
378 | } | 378 | } |
379 | 379 | ||
380 | read_lock(&tasklist_lock); | 380 | read_lock(&tasklist_lock); |
381 | /* | 381 | /* |
382 | * Search in the children | 382 | * Search in the children |
383 | */ | 383 | */ |
384 | list_for_each_entry(c, &p->children, sibling) { | 384 | list_for_each_entry(c, &p->children, sibling) { |
385 | if (c->mm == mm) | 385 | if (c->mm == mm) |
386 | goto assign_new_owner; | 386 | goto assign_new_owner; |
387 | } | 387 | } |
388 | 388 | ||
389 | /* | 389 | /* |
390 | * Search in the siblings | 390 | * Search in the siblings |
391 | */ | 391 | */ |
392 | list_for_each_entry(c, &p->real_parent->children, sibling) { | 392 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
393 | if (c->mm == mm) | 393 | if (c->mm == mm) |
394 | goto assign_new_owner; | 394 | goto assign_new_owner; |
395 | } | 395 | } |
396 | 396 | ||
397 | /* | 397 | /* |
398 | * Search through everything else. We should not get | 398 | * Search through everything else. We should not get |
399 | * here often | 399 | * here often |
400 | */ | 400 | */ |
401 | do_each_thread(g, c) { | 401 | do_each_thread(g, c) { |
402 | if (c->mm == mm) | 402 | if (c->mm == mm) |
403 | goto assign_new_owner; | 403 | goto assign_new_owner; |
404 | } while_each_thread(g, c); | 404 | } while_each_thread(g, c); |
405 | 405 | ||
406 | read_unlock(&tasklist_lock); | 406 | read_unlock(&tasklist_lock); |
407 | /* | 407 | /* |
408 | * We found no owner yet mm_users > 1: this implies that we are | 408 | * We found no owner yet mm_users > 1: this implies that we are |
409 | * most likely racing with swapoff (try_to_unuse()) or /proc or | 409 | * most likely racing with swapoff (try_to_unuse()) or /proc or |
410 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. | 410 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
411 | */ | 411 | */ |
412 | mm->owner = NULL; | 412 | mm->owner = NULL; |
413 | return; | 413 | return; |
414 | 414 | ||
415 | assign_new_owner: | 415 | assign_new_owner: |
416 | BUG_ON(c == p); | 416 | BUG_ON(c == p); |
417 | get_task_struct(c); | 417 | get_task_struct(c); |
418 | /* | 418 | /* |
419 | * The task_lock protects c->mm from changing. | 419 | * The task_lock protects c->mm from changing. |
420 | * We always want mm->owner->mm == mm | 420 | * We always want mm->owner->mm == mm |
421 | */ | 421 | */ |
422 | task_lock(c); | 422 | task_lock(c); |
423 | /* | 423 | /* |
424 | * Delay read_unlock() till we have the task_lock() | 424 | * Delay read_unlock() till we have the task_lock() |
425 | * to ensure that c does not slip away underneath us | 425 | * to ensure that c does not slip away underneath us |
426 | */ | 426 | */ |
427 | read_unlock(&tasklist_lock); | 427 | read_unlock(&tasklist_lock); |
428 | if (c->mm != mm) { | 428 | if (c->mm != mm) { |
429 | task_unlock(c); | 429 | task_unlock(c); |
430 | put_task_struct(c); | 430 | put_task_struct(c); |
431 | goto retry; | 431 | goto retry; |
432 | } | 432 | } |
433 | mm->owner = c; | 433 | mm->owner = c; |
434 | task_unlock(c); | 434 | task_unlock(c); |
435 | put_task_struct(c); | 435 | put_task_struct(c); |
436 | } | 436 | } |
437 | #endif /* CONFIG_MM_OWNER */ | 437 | #endif /* CONFIG_MM_OWNER */ |
438 | 438 | ||
439 | /* | 439 | /* |
440 | * Turn us into a lazy TLB process if we | 440 | * Turn us into a lazy TLB process if we |
441 | * aren't already.. | 441 | * aren't already.. |
442 | */ | 442 | */ |
443 | static void exit_mm(struct task_struct * tsk) | 443 | static void exit_mm(struct task_struct * tsk) |
444 | { | 444 | { |
445 | struct mm_struct *mm = tsk->mm; | 445 | struct mm_struct *mm = tsk->mm; |
446 | struct core_state *core_state; | 446 | struct core_state *core_state; |
447 | 447 | ||
448 | mm_release(tsk, mm); | 448 | mm_release(tsk, mm); |
449 | if (!mm) | 449 | if (!mm) |
450 | return; | 450 | return; |
451 | sync_mm_rss(mm); | 451 | sync_mm_rss(mm); |
452 | /* | 452 | /* |
453 | * Serialize with any possible pending coredump. | 453 | * Serialize with any possible pending coredump. |
454 | * We must hold mmap_sem around checking core_state | 454 | * We must hold mmap_sem around checking core_state |
455 | * and clearing tsk->mm. The core-inducing thread | 455 | * and clearing tsk->mm. The core-inducing thread |
456 | * will increment ->nr_threads for each thread in the | 456 | * will increment ->nr_threads for each thread in the |
457 | * group with ->mm != NULL. | 457 | * group with ->mm != NULL. |
458 | */ | 458 | */ |
459 | down_read(&mm->mmap_sem); | 459 | down_read(&mm->mmap_sem); |
460 | core_state = mm->core_state; | 460 | core_state = mm->core_state; |
461 | if (core_state) { | 461 | if (core_state) { |
462 | struct core_thread self; | 462 | struct core_thread self; |
463 | up_read(&mm->mmap_sem); | 463 | up_read(&mm->mmap_sem); |
464 | 464 | ||
465 | self.task = tsk; | 465 | self.task = tsk; |
466 | self.next = xchg(&core_state->dumper.next, &self); | 466 | self.next = xchg(&core_state->dumper.next, &self); |
467 | /* | 467 | /* |
468 | * Implies mb(), the result of xchg() must be visible | 468 | * Implies mb(), the result of xchg() must be visible |
469 | * to core_state->dumper. | 469 | * to core_state->dumper. |
470 | */ | 470 | */ |
471 | if (atomic_dec_and_test(&core_state->nr_threads)) | 471 | if (atomic_dec_and_test(&core_state->nr_threads)) |
472 | complete(&core_state->startup); | 472 | complete(&core_state->startup); |
473 | 473 | ||
474 | for (;;) { | 474 | for (;;) { |
475 | set_task_state(tsk, TASK_UNINTERRUPTIBLE); | 475 | set_task_state(tsk, TASK_UNINTERRUPTIBLE); |
476 | if (!self.task) /* see coredump_finish() */ | 476 | if (!self.task) /* see coredump_finish() */ |
477 | break; | 477 | break; |
478 | freezable_schedule(); | 478 | freezable_schedule(); |
479 | } | 479 | } |
480 | __set_task_state(tsk, TASK_RUNNING); | 480 | __set_task_state(tsk, TASK_RUNNING); |
481 | down_read(&mm->mmap_sem); | 481 | down_read(&mm->mmap_sem); |
482 | } | 482 | } |
483 | atomic_inc(&mm->mm_count); | 483 | atomic_inc(&mm->mm_count); |
484 | BUG_ON(mm != tsk->active_mm); | 484 | BUG_ON(mm != tsk->active_mm); |
485 | /* more a memory barrier than a real lock */ | 485 | /* more a memory barrier than a real lock */ |
486 | task_lock(tsk); | 486 | task_lock(tsk); |
487 | tsk->mm = NULL; | 487 | tsk->mm = NULL; |
488 | up_read(&mm->mmap_sem); | 488 | up_read(&mm->mmap_sem); |
489 | enter_lazy_tlb(mm, current); | 489 | enter_lazy_tlb(mm, current); |
490 | task_unlock(tsk); | 490 | task_unlock(tsk); |
491 | mm_update_next_owner(mm); | 491 | mm_update_next_owner(mm); |
492 | mmput(mm); | 492 | mmput(mm); |
493 | } | 493 | } |
494 | 494 | ||
495 | /* | 495 | /* |
496 | * When we die, we re-parent all our children, and try to: | 496 | * When we die, we re-parent all our children, and try to: |
497 | * 1. give them to another thread in our thread group, if such a member exists | 497 | * 1. give them to another thread in our thread group, if such a member exists |
498 | * 2. give it to the first ancestor process which prctl'd itself as a | 498 | * 2. give it to the first ancestor process which prctl'd itself as a |
499 | * child_subreaper for its children (like a service manager) | 499 | * child_subreaper for its children (like a service manager) |
500 | * 3. give it to the init process (PID 1) in our pid namespace | 500 | * 3. give it to the init process (PID 1) in our pid namespace |
501 | */ | 501 | */ |
502 | static struct task_struct *find_new_reaper(struct task_struct *father) | 502 | static struct task_struct *find_new_reaper(struct task_struct *father) |
503 | __releases(&tasklist_lock) | 503 | __releases(&tasklist_lock) |
504 | __acquires(&tasklist_lock) | 504 | __acquires(&tasklist_lock) |
505 | { | 505 | { |
506 | struct pid_namespace *pid_ns = task_active_pid_ns(father); | 506 | struct pid_namespace *pid_ns = task_active_pid_ns(father); |
507 | struct task_struct *thread; | 507 | struct task_struct *thread; |
508 | 508 | ||
509 | thread = father; | 509 | thread = father; |
510 | while_each_thread(father, thread) { | 510 | while_each_thread(father, thread) { |
511 | if (thread->flags & PF_EXITING) | 511 | if (thread->flags & PF_EXITING) |
512 | continue; | 512 | continue; |
513 | if (unlikely(pid_ns->child_reaper == father)) | 513 | if (unlikely(pid_ns->child_reaper == father)) |
514 | pid_ns->child_reaper = thread; | 514 | pid_ns->child_reaper = thread; |
515 | return thread; | 515 | return thread; |
516 | } | 516 | } |
517 | 517 | ||
518 | if (unlikely(pid_ns->child_reaper == father)) { | 518 | if (unlikely(pid_ns->child_reaper == father)) { |
519 | write_unlock_irq(&tasklist_lock); | 519 | write_unlock_irq(&tasklist_lock); |
520 | if (unlikely(pid_ns == &init_pid_ns)) { | 520 | if (unlikely(pid_ns == &init_pid_ns)) { |
521 | panic("Attempted to kill init! exitcode=0x%08x\n", | 521 | panic("Attempted to kill init! exitcode=0x%08x\n", |
522 | father->signal->group_exit_code ?: | 522 | father->signal->group_exit_code ?: |
523 | father->exit_code); | 523 | father->exit_code); |
524 | } | 524 | } |
525 | 525 | ||
526 | zap_pid_ns_processes(pid_ns); | 526 | zap_pid_ns_processes(pid_ns); |
527 | write_lock_irq(&tasklist_lock); | 527 | write_lock_irq(&tasklist_lock); |
528 | } else if (father->signal->has_child_subreaper) { | 528 | } else if (father->signal->has_child_subreaper) { |
529 | struct task_struct *reaper; | 529 | struct task_struct *reaper; |
530 | 530 | ||
531 | /* | 531 | /* |
532 | * Find the first ancestor marked as child_subreaper. | 532 | * Find the first ancestor marked as child_subreaper. |
533 | * Note that the code below checks same_thread_group(reaper, | 533 | * Note that the code below checks same_thread_group(reaper, |
534 | * pid_ns->child_reaper). This is what we need to DTRT in a | 534 | * pid_ns->child_reaper). This is what we need to DTRT in a |
535 | * PID namespace. However we still need the check above, see | 535 | * PID namespace. However we still need the check above, see |
536 | * http://marc.info/?l=linux-kernel&m=131385460420380 | 536 | * http://marc.info/?l=linux-kernel&m=131385460420380 |
537 | */ | 537 | */ |
538 | for (reaper = father->real_parent; | 538 | for (reaper = father->real_parent; |
539 | reaper != &init_task; | 539 | reaper != &init_task; |
540 | reaper = reaper->real_parent) { | 540 | reaper = reaper->real_parent) { |
541 | if (same_thread_group(reaper, pid_ns->child_reaper)) | 541 | if (same_thread_group(reaper, pid_ns->child_reaper)) |
542 | break; | 542 | break; |
543 | if (!reaper->signal->is_child_subreaper) | 543 | if (!reaper->signal->is_child_subreaper) |
544 | continue; | 544 | continue; |
545 | thread = reaper; | 545 | thread = reaper; |
546 | do { | 546 | do { |
547 | if (!(thread->flags & PF_EXITING)) | 547 | if (!(thread->flags & PF_EXITING)) |
548 | return reaper; | 548 | return reaper; |
549 | } while_each_thread(reaper, thread); | 549 | } while_each_thread(reaper, thread); |
550 | } | 550 | } |
551 | } | 551 | } |
552 | 552 | ||
553 | return pid_ns->child_reaper; | 553 | return pid_ns->child_reaper; |
554 | } | 554 | } |
555 | 555 | ||
556 | /* | 556 | /* |
557 | * Any that need to be release_task'd are put on the @dead list. | 557 | * Any that need to be release_task'd are put on the @dead list. |
558 | */ | 558 | */ |
559 | static void reparent_leader(struct task_struct *father, struct task_struct *p, | 559 | static void reparent_leader(struct task_struct *father, struct task_struct *p, |
560 | struct list_head *dead) | 560 | struct list_head *dead) |
561 | { | 561 | { |
562 | list_move_tail(&p->sibling, &p->real_parent->children); | 562 | list_move_tail(&p->sibling, &p->real_parent->children); |
563 | /* | 563 | /* |
564 | * If this is a threaded reparent there is no need to | 564 | * If this is a threaded reparent there is no need to |
565 | * notify anyone anything has happened. | 565 | * notify anyone anything has happened. |
566 | */ | 566 | */ |
567 | if (same_thread_group(p->real_parent, father)) | 567 | if (same_thread_group(p->real_parent, father)) |
568 | return; | 568 | return; |
569 | 569 | ||
570 | /* | 570 | /* |
571 | * We don't want people slaying init. | 571 | * We don't want people slaying init. |
572 | * | 572 | * |
573 | * Note: we do this even if it is EXIT_DEAD, wait_task_zombie() | 573 | * Note: we do this even if it is EXIT_DEAD, wait_task_zombie() |
574 | * can change ->exit_state to EXIT_ZOMBIE. If this is the final | 574 | * can change ->exit_state to EXIT_ZOMBIE. If this is the final |
575 | * state, do_notify_parent() was already called and ->exit_signal | 575 | * state, do_notify_parent() was already called and ->exit_signal |
576 | * doesn't matter. | 576 | * doesn't matter. |
577 | */ | 577 | */ |
578 | p->exit_signal = SIGCHLD; | 578 | p->exit_signal = SIGCHLD; |
579 | 579 | ||
580 | if (p->exit_state == EXIT_DEAD) | 580 | if (p->exit_state == EXIT_DEAD) |
581 | return; | 581 | return; |
582 | 582 | ||
583 | /* If it has exited notify the new parent about this child's death. */ | 583 | /* If it has exited notify the new parent about this child's death. */ |
584 | if (!p->ptrace && | 584 | if (!p->ptrace && |
585 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { | 585 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
586 | if (do_notify_parent(p, p->exit_signal)) { | 586 | if (do_notify_parent(p, p->exit_signal)) { |
587 | p->exit_state = EXIT_DEAD; | 587 | p->exit_state = EXIT_DEAD; |
588 | list_move_tail(&p->sibling, dead); | 588 | list_move_tail(&p->sibling, dead); |
589 | } | 589 | } |
590 | } | 590 | } |
591 | 591 | ||
592 | kill_orphaned_pgrp(p, father); | 592 | kill_orphaned_pgrp(p, father); |
593 | } | 593 | } |
594 | 594 | ||
595 | static void forget_original_parent(struct task_struct *father) | 595 | static void forget_original_parent(struct task_struct *father) |
596 | { | 596 | { |
597 | struct task_struct *p, *n, *reaper; | 597 | struct task_struct *p, *n, *reaper; |
598 | LIST_HEAD(dead_children); | 598 | LIST_HEAD(dead_children); |
599 | 599 | ||
600 | write_lock_irq(&tasklist_lock); | 600 | write_lock_irq(&tasklist_lock); |
601 | /* | 601 | /* |
602 | * Note that exit_ptrace() and find_new_reaper() might | 602 | * Note that exit_ptrace() and find_new_reaper() might |
603 | * drop tasklist_lock and reacquire it. | 603 | * drop tasklist_lock and reacquire it. |
604 | */ | 604 | */ |
605 | exit_ptrace(father); | 605 | exit_ptrace(father); |
606 | reaper = find_new_reaper(father); | 606 | reaper = find_new_reaper(father); |
607 | 607 | ||
608 | list_for_each_entry_safe(p, n, &father->children, sibling) { | 608 | list_for_each_entry_safe(p, n, &father->children, sibling) { |
609 | struct task_struct *t = p; | 609 | struct task_struct *t = p; |
610 | do { | 610 | do { |
611 | t->real_parent = reaper; | 611 | t->real_parent = reaper; |
612 | if (t->parent == father) { | 612 | if (t->parent == father) { |
613 | BUG_ON(t->ptrace); | 613 | BUG_ON(t->ptrace); |
614 | t->parent = t->real_parent; | 614 | t->parent = t->real_parent; |
615 | } | 615 | } |
616 | if (t->pdeath_signal) | 616 | if (t->pdeath_signal) |
617 | group_send_sig_info(t->pdeath_signal, | 617 | group_send_sig_info(t->pdeath_signal, |
618 | SEND_SIG_NOINFO, t); | 618 | SEND_SIG_NOINFO, t); |
619 | } while_each_thread(p, t); | 619 | } while_each_thread(p, t); |
620 | reparent_leader(father, p, &dead_children); | 620 | reparent_leader(father, p, &dead_children); |
621 | } | 621 | } |
622 | write_unlock_irq(&tasklist_lock); | 622 | write_unlock_irq(&tasklist_lock); |
623 | 623 | ||
624 | BUG_ON(!list_empty(&father->children)); | 624 | BUG_ON(!list_empty(&father->children)); |
625 | 625 | ||
626 | list_for_each_entry_safe(p, n, &dead_children, sibling) { | 626 | list_for_each_entry_safe(p, n, &dead_children, sibling) { |
627 | list_del_init(&p->sibling); | 627 | list_del_init(&p->sibling); |
628 | release_task(p); | 628 | release_task(p); |
629 | } | 629 | } |
630 | } | 630 | } |
631 | 631 | ||
632 | /* | 632 | /* |
633 | * Send signals to all our closest relatives so that they know | 633 | * Send signals to all our closest relatives so that they know |
634 | * to properly mourn us.. | 634 | * to properly mourn us.. |
635 | */ | 635 | */ |
636 | static void exit_notify(struct task_struct *tsk, int group_dead) | 636 | static void exit_notify(struct task_struct *tsk, int group_dead) |
637 | { | 637 | { |
638 | bool autoreap; | 638 | bool autoreap; |
639 | 639 | ||
640 | /* | 640 | /* |
641 | * This does two things: | 641 | * This does two things: |
642 | * | 642 | * |
643 | * A. Make init inherit all the child processes | 643 | * A. Make init inherit all the child processes |
644 | * B. Check to see if any process groups have become orphaned | 644 | * B. Check to see if any process groups have become orphaned |
645 | * as a result of our exiting, and if they have any stopped | 645 | * as a result of our exiting, and if they have any stopped |
646 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | 646 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) |
647 | */ | 647 | */ |
648 | forget_original_parent(tsk); | 648 | forget_original_parent(tsk); |
649 | 649 | ||
650 | write_lock_irq(&tasklist_lock); | 650 | write_lock_irq(&tasklist_lock); |
651 | if (group_dead) | 651 | if (group_dead) |
652 | kill_orphaned_pgrp(tsk->group_leader, NULL); | 652 | kill_orphaned_pgrp(tsk->group_leader, NULL); |
653 | 653 | ||
654 | if (unlikely(tsk->ptrace)) { | 654 | if (unlikely(tsk->ptrace)) { |
655 | int sig = thread_group_leader(tsk) && | 655 | int sig = thread_group_leader(tsk) && |
656 | thread_group_empty(tsk) && | 656 | thread_group_empty(tsk) && |
657 | !ptrace_reparented(tsk) ? | 657 | !ptrace_reparented(tsk) ? |
658 | tsk->exit_signal : SIGCHLD; | 658 | tsk->exit_signal : SIGCHLD; |
659 | autoreap = do_notify_parent(tsk, sig); | 659 | autoreap = do_notify_parent(tsk, sig); |
660 | } else if (thread_group_leader(tsk)) { | 660 | } else if (thread_group_leader(tsk)) { |
661 | autoreap = thread_group_empty(tsk) && | 661 | autoreap = thread_group_empty(tsk) && |
662 | do_notify_parent(tsk, tsk->exit_signal); | 662 | do_notify_parent(tsk, tsk->exit_signal); |
663 | } else { | 663 | } else { |
664 | autoreap = true; | 664 | autoreap = true; |
665 | } | 665 | } |
666 | 666 | ||
667 | tsk->exit_state = autoreap ? EXIT_DEAD : EXIT_ZOMBIE; | 667 | tsk->exit_state = autoreap ? EXIT_DEAD : EXIT_ZOMBIE; |
668 | 668 | ||
669 | /* mt-exec, de_thread() is waiting for group leader */ | 669 | /* mt-exec, de_thread() is waiting for group leader */ |
670 | if (unlikely(tsk->signal->notify_count < 0)) | 670 | if (unlikely(tsk->signal->notify_count < 0)) |
671 | wake_up_process(tsk->signal->group_exit_task); | 671 | wake_up_process(tsk->signal->group_exit_task); |
672 | write_unlock_irq(&tasklist_lock); | 672 | write_unlock_irq(&tasklist_lock); |
673 | 673 | ||
674 | /* If the process is dead, release it - nobody will wait for it */ | 674 | /* If the process is dead, release it - nobody will wait for it */ |
675 | if (autoreap) | 675 | if (autoreap) |
676 | release_task(tsk); | 676 | release_task(tsk); |
677 | } | 677 | } |
678 | 678 | ||
679 | #ifdef CONFIG_DEBUG_STACK_USAGE | 679 | #ifdef CONFIG_DEBUG_STACK_USAGE |
680 | static void check_stack_usage(void) | 680 | static void check_stack_usage(void) |
681 | { | 681 | { |
682 | static DEFINE_SPINLOCK(low_water_lock); | 682 | static DEFINE_SPINLOCK(low_water_lock); |
683 | static int lowest_to_date = THREAD_SIZE; | 683 | static int lowest_to_date = THREAD_SIZE; |
684 | unsigned long free; | 684 | unsigned long free; |
685 | 685 | ||
686 | free = stack_not_used(current); | 686 | free = stack_not_used(current); |
687 | 687 | ||
688 | if (free >= lowest_to_date) | 688 | if (free >= lowest_to_date) |
689 | return; | 689 | return; |
690 | 690 | ||
691 | spin_lock(&low_water_lock); | 691 | spin_lock(&low_water_lock); |
692 | if (free < lowest_to_date) { | 692 | if (free < lowest_to_date) { |
693 | printk(KERN_WARNING "%s (%d) used greatest stack depth: " | 693 | printk(KERN_WARNING "%s (%d) used greatest stack depth: " |
694 | "%lu bytes left\n", | 694 | "%lu bytes left\n", |
695 | current->comm, task_pid_nr(current), free); | 695 | current->comm, task_pid_nr(current), free); |
696 | lowest_to_date = free; | 696 | lowest_to_date = free; |
697 | } | 697 | } |
698 | spin_unlock(&low_water_lock); | 698 | spin_unlock(&low_water_lock); |
699 | } | 699 | } |
700 | #else | 700 | #else |
701 | static inline void check_stack_usage(void) {} | 701 | static inline void check_stack_usage(void) {} |
702 | #endif | 702 | #endif |
703 | 703 | ||
704 | void do_exit(long code) | 704 | void do_exit(long code) |
705 | { | 705 | { |
706 | struct task_struct *tsk = current; | 706 | struct task_struct *tsk = current; |
707 | int group_dead; | 707 | int group_dead; |
708 | 708 | ||
709 | profile_task_exit(tsk); | 709 | profile_task_exit(tsk); |
710 | 710 | ||
711 | WARN_ON(blk_needs_flush_plug(tsk)); | 711 | WARN_ON(blk_needs_flush_plug(tsk)); |
712 | 712 | ||
713 | if (unlikely(in_interrupt())) | 713 | if (unlikely(in_interrupt())) |
714 | panic("Aiee, killing interrupt handler!"); | 714 | panic("Aiee, killing interrupt handler!"); |
715 | if (unlikely(!tsk->pid)) | 715 | if (unlikely(!tsk->pid)) |
716 | panic("Attempted to kill the idle task!"); | 716 | panic("Attempted to kill the idle task!"); |
717 | 717 | ||
718 | /* | 718 | /* |
719 | * If do_exit is called because this processes oopsed, it's possible | 719 | * If do_exit is called because this processes oopsed, it's possible |
720 | * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before | 720 | * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before |
721 | * continuing. Amongst other possible reasons, this is to prevent | 721 | * continuing. Amongst other possible reasons, this is to prevent |
722 | * mm_release()->clear_child_tid() from writing to a user-controlled | 722 | * mm_release()->clear_child_tid() from writing to a user-controlled |
723 | * kernel address. | 723 | * kernel address. |
724 | */ | 724 | */ |
725 | set_fs(USER_DS); | 725 | set_fs(USER_DS); |
726 | 726 | ||
727 | ptrace_event(PTRACE_EVENT_EXIT, code); | 727 | ptrace_event(PTRACE_EVENT_EXIT, code); |
728 | 728 | ||
729 | validate_creds_for_do_exit(tsk); | 729 | validate_creds_for_do_exit(tsk); |
730 | 730 | ||
731 | /* | 731 | /* |
732 | * We're taking recursive faults here in do_exit. Safest is to just | 732 | * We're taking recursive faults here in do_exit. Safest is to just |
733 | * leave this task alone and wait for reboot. | 733 | * leave this task alone and wait for reboot. |
734 | */ | 734 | */ |
735 | if (unlikely(tsk->flags & PF_EXITING)) { | 735 | if (unlikely(tsk->flags & PF_EXITING)) { |
736 | printk(KERN_ALERT | 736 | printk(KERN_ALERT |
737 | "Fixing recursive fault but reboot is needed!\n"); | 737 | "Fixing recursive fault but reboot is needed!\n"); |
738 | /* | 738 | /* |
739 | * We can do this unlocked here. The futex code uses | 739 | * We can do this unlocked here. The futex code uses |
740 | * this flag just to verify whether the pi state | 740 | * this flag just to verify whether the pi state |
741 | * cleanup has been done or not. In the worst case it | 741 | * cleanup has been done or not. In the worst case it |
742 | * loops once more. We pretend that the cleanup was | 742 | * loops once more. We pretend that the cleanup was |
743 | * done as there is no way to return. Either the | 743 | * done as there is no way to return. Either the |
744 | * OWNER_DIED bit is set by now or we push the blocked | 744 | * OWNER_DIED bit is set by now or we push the blocked |
745 | * task into the wait for ever nirwana as well. | 745 | * task into the wait for ever nirwana as well. |
746 | */ | 746 | */ |
747 | tsk->flags |= PF_EXITPIDONE; | 747 | tsk->flags |= PF_EXITPIDONE; |
748 | set_current_state(TASK_UNINTERRUPTIBLE); | 748 | set_current_state(TASK_UNINTERRUPTIBLE); |
749 | schedule(); | 749 | schedule(); |
750 | } | 750 | } |
751 | 751 | ||
752 | exit_signals(tsk); /* sets PF_EXITING */ | 752 | exit_signals(tsk); /* sets PF_EXITING */ |
753 | /* | 753 | /* |
754 | * tsk->flags are checked in the futex code to protect against | 754 | * tsk->flags are checked in the futex code to protect against |
755 | * an exiting task cleaning up the robust pi futexes. | 755 | * an exiting task cleaning up the robust pi futexes. |
756 | */ | 756 | */ |
757 | smp_mb(); | 757 | smp_mb(); |
758 | raw_spin_unlock_wait(&tsk->pi_lock); | 758 | raw_spin_unlock_wait(&tsk->pi_lock); |
759 | 759 | ||
760 | if (unlikely(in_atomic())) | 760 | if (unlikely(in_atomic())) |
761 | printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n", | 761 | printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n", |
762 | current->comm, task_pid_nr(current), | 762 | current->comm, task_pid_nr(current), |
763 | preempt_count()); | 763 | preempt_count()); |
764 | 764 | ||
765 | acct_update_integrals(tsk); | 765 | acct_update_integrals(tsk); |
766 | /* sync mm's RSS info before statistics gathering */ | 766 | /* sync mm's RSS info before statistics gathering */ |
767 | if (tsk->mm) | 767 | if (tsk->mm) |
768 | sync_mm_rss(tsk->mm); | 768 | sync_mm_rss(tsk->mm); |
769 | group_dead = atomic_dec_and_test(&tsk->signal->live); | 769 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
770 | if (group_dead) { | 770 | if (group_dead) { |
771 | hrtimer_cancel(&tsk->signal->real_timer); | 771 | hrtimer_cancel(&tsk->signal->real_timer); |
772 | exit_itimers(tsk->signal); | 772 | exit_itimers(tsk->signal); |
773 | if (tsk->mm) | 773 | if (tsk->mm) |
774 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); | 774 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); |
775 | } | 775 | } |
776 | acct_collect(code, group_dead); | 776 | acct_collect(code, group_dead); |
777 | if (group_dead) | 777 | if (group_dead) |
778 | tty_audit_exit(); | 778 | tty_audit_exit(); |
779 | audit_free(tsk); | 779 | audit_free(tsk); |
780 | 780 | ||
781 | tsk->exit_code = code; | 781 | tsk->exit_code = code; |
782 | taskstats_exit(tsk, group_dead); | 782 | taskstats_exit(tsk, group_dead); |
783 | 783 | ||
784 | exit_mm(tsk); | 784 | exit_mm(tsk); |
785 | 785 | ||
786 | if (group_dead) | 786 | if (group_dead) |
787 | acct_process(); | 787 | acct_process(); |
788 | trace_sched_process_exit(tsk); | 788 | trace_sched_process_exit(tsk); |
789 | 789 | ||
790 | exit_sem(tsk); | 790 | exit_sem(tsk); |
791 | exit_shm(tsk); | 791 | exit_shm(tsk); |
792 | exit_files(tsk); | 792 | exit_files(tsk); |
793 | exit_fs(tsk); | 793 | exit_fs(tsk); |
794 | if (group_dead) | ||
795 | disassociate_ctty(1); | ||
794 | exit_task_namespaces(tsk); | 796 | exit_task_namespaces(tsk); |
795 | exit_task_work(tsk); | 797 | exit_task_work(tsk); |
796 | check_stack_usage(); | 798 | check_stack_usage(); |
797 | exit_thread(); | 799 | exit_thread(); |
798 | 800 | ||
799 | /* | 801 | /* |
800 | * Flush inherited counters to the parent - before the parent | 802 | * Flush inherited counters to the parent - before the parent |
801 | * gets woken up by child-exit notifications. | 803 | * gets woken up by child-exit notifications. |
802 | * | 804 | * |
803 | * because of cgroup mode, must be called before cgroup_exit() | 805 | * because of cgroup mode, must be called before cgroup_exit() |
804 | */ | 806 | */ |
805 | perf_event_exit_task(tsk); | 807 | perf_event_exit_task(tsk); |
806 | 808 | ||
807 | cgroup_exit(tsk, 1); | 809 | cgroup_exit(tsk, 1); |
808 | 810 | ||
809 | if (group_dead) | ||
810 | disassociate_ctty(1); | ||
811 | |||
812 | module_put(task_thread_info(tsk)->exec_domain->module); | 811 | module_put(task_thread_info(tsk)->exec_domain->module); |
813 | 812 | ||
814 | proc_exit_connector(tsk); | 813 | proc_exit_connector(tsk); |
815 | |||
816 | /* | 814 | /* |
817 | * FIXME: do that only when needed, using sched_exit tracepoint | 815 | * FIXME: do that only when needed, using sched_exit tracepoint |
818 | */ | 816 | */ |
819 | flush_ptrace_hw_breakpoint(tsk); | 817 | flush_ptrace_hw_breakpoint(tsk); |
820 | 818 | ||
821 | exit_notify(tsk, group_dead); | 819 | exit_notify(tsk, group_dead); |
822 | #ifdef CONFIG_NUMA | 820 | #ifdef CONFIG_NUMA |
823 | task_lock(tsk); | 821 | task_lock(tsk); |
824 | mpol_put(tsk->mempolicy); | 822 | mpol_put(tsk->mempolicy); |
825 | tsk->mempolicy = NULL; | 823 | tsk->mempolicy = NULL; |
826 | task_unlock(tsk); | 824 | task_unlock(tsk); |
827 | #endif | 825 | #endif |
828 | #ifdef CONFIG_FUTEX | 826 | #ifdef CONFIG_FUTEX |
829 | if (unlikely(current->pi_state_cache)) | 827 | if (unlikely(current->pi_state_cache)) |
830 | kfree(current->pi_state_cache); | 828 | kfree(current->pi_state_cache); |
831 | #endif | 829 | #endif |
832 | /* | 830 | /* |
833 | * Make sure we are holding no locks: | 831 | * Make sure we are holding no locks: |
834 | */ | 832 | */ |
835 | debug_check_no_locks_held(); | 833 | debug_check_no_locks_held(); |
836 | /* | 834 | /* |
837 | * We can do this unlocked here. The futex code uses this flag | 835 | * We can do this unlocked here. The futex code uses this flag |
838 | * just to verify whether the pi state cleanup has been done | 836 | * just to verify whether the pi state cleanup has been done |
839 | * or not. In the worst case it loops once more. | 837 | * or not. In the worst case it loops once more. |
840 | */ | 838 | */ |
841 | tsk->flags |= PF_EXITPIDONE; | 839 | tsk->flags |= PF_EXITPIDONE; |
842 | 840 | ||
843 | if (tsk->io_context) | 841 | if (tsk->io_context) |
844 | exit_io_context(tsk); | 842 | exit_io_context(tsk); |
845 | 843 | ||
846 | if (tsk->splice_pipe) | 844 | if (tsk->splice_pipe) |
847 | free_pipe_info(tsk->splice_pipe); | 845 | free_pipe_info(tsk->splice_pipe); |
848 | 846 | ||
849 | if (tsk->task_frag.page) | 847 | if (tsk->task_frag.page) |
850 | put_page(tsk->task_frag.page); | 848 | put_page(tsk->task_frag.page); |
851 | 849 | ||
852 | validate_creds_for_do_exit(tsk); | 850 | validate_creds_for_do_exit(tsk); |
853 | 851 | ||
854 | preempt_disable(); | 852 | preempt_disable(); |
855 | if (tsk->nr_dirtied) | 853 | if (tsk->nr_dirtied) |
856 | __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied); | 854 | __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied); |
857 | exit_rcu(); | 855 | exit_rcu(); |
858 | 856 | ||
859 | /* | 857 | /* |
860 | * The setting of TASK_RUNNING by try_to_wake_up() may be delayed | 858 | * The setting of TASK_RUNNING by try_to_wake_up() may be delayed |
861 | * when the following two conditions become true. | 859 | * when the following two conditions become true. |
862 | * - There is race condition of mmap_sem (It is acquired by | 860 | * - There is race condition of mmap_sem (It is acquired by |
863 | * exit_mm()), and | 861 | * exit_mm()), and |
864 | * - SMI occurs before setting TASK_RUNINNG. | 862 | * - SMI occurs before setting TASK_RUNINNG. |
865 | * (or hypervisor of virtual machine switches to other guest) | 863 | * (or hypervisor of virtual machine switches to other guest) |
866 | * As a result, we may become TASK_RUNNING after becoming TASK_DEAD | 864 | * As a result, we may become TASK_RUNNING after becoming TASK_DEAD |
867 | * | 865 | * |
868 | * To avoid it, we have to wait for releasing tsk->pi_lock which | 866 | * To avoid it, we have to wait for releasing tsk->pi_lock which |
869 | * is held by try_to_wake_up() | 867 | * is held by try_to_wake_up() |
870 | */ | 868 | */ |
871 | smp_mb(); | 869 | smp_mb(); |
872 | raw_spin_unlock_wait(&tsk->pi_lock); | 870 | raw_spin_unlock_wait(&tsk->pi_lock); |
873 | 871 | ||
874 | /* causes final put_task_struct in finish_task_switch(). */ | 872 | /* causes final put_task_struct in finish_task_switch(). */ |
875 | tsk->state = TASK_DEAD; | 873 | tsk->state = TASK_DEAD; |
876 | tsk->flags |= PF_NOFREEZE; /* tell freezer to ignore us */ | 874 | tsk->flags |= PF_NOFREEZE; /* tell freezer to ignore us */ |
877 | schedule(); | 875 | schedule(); |
878 | BUG(); | 876 | BUG(); |
879 | /* Avoid "noreturn function does return". */ | 877 | /* Avoid "noreturn function does return". */ |
880 | for (;;) | 878 | for (;;) |
881 | cpu_relax(); /* For when BUG is null */ | 879 | cpu_relax(); /* For when BUG is null */ |
882 | } | 880 | } |
883 | 881 | ||
884 | EXPORT_SYMBOL_GPL(do_exit); | 882 | EXPORT_SYMBOL_GPL(do_exit); |
885 | 883 | ||
886 | void complete_and_exit(struct completion *comp, long code) | 884 | void complete_and_exit(struct completion *comp, long code) |
887 | { | 885 | { |
888 | if (comp) | 886 | if (comp) |
889 | complete(comp); | 887 | complete(comp); |
890 | 888 | ||
891 | do_exit(code); | 889 | do_exit(code); |
892 | } | 890 | } |
893 | 891 | ||
894 | EXPORT_SYMBOL(complete_and_exit); | 892 | EXPORT_SYMBOL(complete_and_exit); |
895 | 893 | ||
896 | SYSCALL_DEFINE1(exit, int, error_code) | 894 | SYSCALL_DEFINE1(exit, int, error_code) |
897 | { | 895 | { |
898 | do_exit((error_code&0xff)<<8); | 896 | do_exit((error_code&0xff)<<8); |
899 | } | 897 | } |
900 | 898 | ||
901 | /* | 899 | /* |
902 | * Take down every thread in the group. This is called by fatal signals | 900 | * Take down every thread in the group. This is called by fatal signals |
903 | * as well as by sys_exit_group (below). | 901 | * as well as by sys_exit_group (below). |
904 | */ | 902 | */ |
905 | void | 903 | void |
906 | do_group_exit(int exit_code) | 904 | do_group_exit(int exit_code) |
907 | { | 905 | { |
908 | struct signal_struct *sig = current->signal; | 906 | struct signal_struct *sig = current->signal; |
909 | 907 | ||
910 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ | 908 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
911 | 909 | ||
912 | if (signal_group_exit(sig)) | 910 | if (signal_group_exit(sig)) |
913 | exit_code = sig->group_exit_code; | 911 | exit_code = sig->group_exit_code; |
914 | else if (!thread_group_empty(current)) { | 912 | else if (!thread_group_empty(current)) { |
915 | struct sighand_struct *const sighand = current->sighand; | 913 | struct sighand_struct *const sighand = current->sighand; |
916 | spin_lock_irq(&sighand->siglock); | 914 | spin_lock_irq(&sighand->siglock); |
917 | if (signal_group_exit(sig)) | 915 | if (signal_group_exit(sig)) |
918 | /* Another thread got here before we took the lock. */ | 916 | /* Another thread got here before we took the lock. */ |
919 | exit_code = sig->group_exit_code; | 917 | exit_code = sig->group_exit_code; |
920 | else { | 918 | else { |
921 | sig->group_exit_code = exit_code; | 919 | sig->group_exit_code = exit_code; |
922 | sig->flags = SIGNAL_GROUP_EXIT; | 920 | sig->flags = SIGNAL_GROUP_EXIT; |
923 | zap_other_threads(current); | 921 | zap_other_threads(current); |
924 | } | 922 | } |
925 | spin_unlock_irq(&sighand->siglock); | 923 | spin_unlock_irq(&sighand->siglock); |
926 | } | 924 | } |
927 | 925 | ||
928 | do_exit(exit_code); | 926 | do_exit(exit_code); |
929 | /* NOTREACHED */ | 927 | /* NOTREACHED */ |
930 | } | 928 | } |
931 | 929 | ||
932 | /* | 930 | /* |
933 | * this kills every thread in the thread group. Note that any externally | 931 | * this kills every thread in the thread group. Note that any externally |
934 | * wait4()-ing process will get the correct exit code - even if this | 932 | * wait4()-ing process will get the correct exit code - even if this |
935 | * thread is not the thread group leader. | 933 | * thread is not the thread group leader. |
936 | */ | 934 | */ |
937 | SYSCALL_DEFINE1(exit_group, int, error_code) | 935 | SYSCALL_DEFINE1(exit_group, int, error_code) |
938 | { | 936 | { |
939 | do_group_exit((error_code & 0xff) << 8); | 937 | do_group_exit((error_code & 0xff) << 8); |
940 | /* NOTREACHED */ | 938 | /* NOTREACHED */ |
941 | return 0; | 939 | return 0; |
942 | } | 940 | } |
943 | 941 | ||
944 | struct wait_opts { | 942 | struct wait_opts { |
945 | enum pid_type wo_type; | 943 | enum pid_type wo_type; |
946 | int wo_flags; | 944 | int wo_flags; |
947 | struct pid *wo_pid; | 945 | struct pid *wo_pid; |
948 | 946 | ||
949 | struct siginfo __user *wo_info; | 947 | struct siginfo __user *wo_info; |
950 | int __user *wo_stat; | 948 | int __user *wo_stat; |
951 | struct rusage __user *wo_rusage; | 949 | struct rusage __user *wo_rusage; |
952 | 950 | ||
953 | wait_queue_t child_wait; | 951 | wait_queue_t child_wait; |
954 | int notask_error; | 952 | int notask_error; |
955 | }; | 953 | }; |
956 | 954 | ||
957 | static inline | 955 | static inline |
958 | struct pid *task_pid_type(struct task_struct *task, enum pid_type type) | 956 | struct pid *task_pid_type(struct task_struct *task, enum pid_type type) |
959 | { | 957 | { |
960 | if (type != PIDTYPE_PID) | 958 | if (type != PIDTYPE_PID) |
961 | task = task->group_leader; | 959 | task = task->group_leader; |
962 | return task->pids[type].pid; | 960 | return task->pids[type].pid; |
963 | } | 961 | } |
964 | 962 | ||
965 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) | 963 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) |
966 | { | 964 | { |
967 | return wo->wo_type == PIDTYPE_MAX || | 965 | return wo->wo_type == PIDTYPE_MAX || |
968 | task_pid_type(p, wo->wo_type) == wo->wo_pid; | 966 | task_pid_type(p, wo->wo_type) == wo->wo_pid; |
969 | } | 967 | } |
970 | 968 | ||
971 | static int eligible_child(struct wait_opts *wo, struct task_struct *p) | 969 | static int eligible_child(struct wait_opts *wo, struct task_struct *p) |
972 | { | 970 | { |
973 | if (!eligible_pid(wo, p)) | 971 | if (!eligible_pid(wo, p)) |
974 | return 0; | 972 | return 0; |
975 | /* Wait for all children (clone and not) if __WALL is set; | 973 | /* Wait for all children (clone and not) if __WALL is set; |
976 | * otherwise, wait for clone children *only* if __WCLONE is | 974 | * otherwise, wait for clone children *only* if __WCLONE is |
977 | * set; otherwise, wait for non-clone children *only*. (Note: | 975 | * set; otherwise, wait for non-clone children *only*. (Note: |
978 | * A "clone" child here is one that reports to its parent | 976 | * A "clone" child here is one that reports to its parent |
979 | * using a signal other than SIGCHLD.) */ | 977 | * using a signal other than SIGCHLD.) */ |
980 | if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) | 978 | if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) |
981 | && !(wo->wo_flags & __WALL)) | 979 | && !(wo->wo_flags & __WALL)) |
982 | return 0; | 980 | return 0; |
983 | 981 | ||
984 | return 1; | 982 | return 1; |
985 | } | 983 | } |
986 | 984 | ||
987 | static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p, | 985 | static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p, |
988 | pid_t pid, uid_t uid, int why, int status) | 986 | pid_t pid, uid_t uid, int why, int status) |
989 | { | 987 | { |
990 | struct siginfo __user *infop; | 988 | struct siginfo __user *infop; |
991 | int retval = wo->wo_rusage | 989 | int retval = wo->wo_rusage |
992 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | 990 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; |
993 | 991 | ||
994 | put_task_struct(p); | 992 | put_task_struct(p); |
995 | infop = wo->wo_info; | 993 | infop = wo->wo_info; |
996 | if (infop) { | 994 | if (infop) { |
997 | if (!retval) | 995 | if (!retval) |
998 | retval = put_user(SIGCHLD, &infop->si_signo); | 996 | retval = put_user(SIGCHLD, &infop->si_signo); |
999 | if (!retval) | 997 | if (!retval) |
1000 | retval = put_user(0, &infop->si_errno); | 998 | retval = put_user(0, &infop->si_errno); |
1001 | if (!retval) | 999 | if (!retval) |
1002 | retval = put_user((short)why, &infop->si_code); | 1000 | retval = put_user((short)why, &infop->si_code); |
1003 | if (!retval) | 1001 | if (!retval) |
1004 | retval = put_user(pid, &infop->si_pid); | 1002 | retval = put_user(pid, &infop->si_pid); |
1005 | if (!retval) | 1003 | if (!retval) |
1006 | retval = put_user(uid, &infop->si_uid); | 1004 | retval = put_user(uid, &infop->si_uid); |
1007 | if (!retval) | 1005 | if (!retval) |
1008 | retval = put_user(status, &infop->si_status); | 1006 | retval = put_user(status, &infop->si_status); |
1009 | } | 1007 | } |
1010 | if (!retval) | 1008 | if (!retval) |
1011 | retval = pid; | 1009 | retval = pid; |
1012 | return retval; | 1010 | return retval; |
1013 | } | 1011 | } |
1014 | 1012 | ||
1015 | /* | 1013 | /* |
1016 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | 1014 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold |
1017 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | 1015 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold |
1018 | * the lock and this task is uninteresting. If we return nonzero, we have | 1016 | * the lock and this task is uninteresting. If we return nonzero, we have |
1019 | * released the lock and the system call should return. | 1017 | * released the lock and the system call should return. |
1020 | */ | 1018 | */ |
1021 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) | 1019 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) |
1022 | { | 1020 | { |
1023 | unsigned long state; | 1021 | unsigned long state; |
1024 | int retval, status, traced; | 1022 | int retval, status, traced; |
1025 | pid_t pid = task_pid_vnr(p); | 1023 | pid_t pid = task_pid_vnr(p); |
1026 | uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p)); | 1024 | uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
1027 | struct siginfo __user *infop; | 1025 | struct siginfo __user *infop; |
1028 | 1026 | ||
1029 | if (!likely(wo->wo_flags & WEXITED)) | 1027 | if (!likely(wo->wo_flags & WEXITED)) |
1030 | return 0; | 1028 | return 0; |
1031 | 1029 | ||
1032 | if (unlikely(wo->wo_flags & WNOWAIT)) { | 1030 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
1033 | int exit_code = p->exit_code; | 1031 | int exit_code = p->exit_code; |
1034 | int why; | 1032 | int why; |
1035 | 1033 | ||
1036 | get_task_struct(p); | 1034 | get_task_struct(p); |
1037 | read_unlock(&tasklist_lock); | 1035 | read_unlock(&tasklist_lock); |
1038 | if ((exit_code & 0x7f) == 0) { | 1036 | if ((exit_code & 0x7f) == 0) { |
1039 | why = CLD_EXITED; | 1037 | why = CLD_EXITED; |
1040 | status = exit_code >> 8; | 1038 | status = exit_code >> 8; |
1041 | } else { | 1039 | } else { |
1042 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; | 1040 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; |
1043 | status = exit_code & 0x7f; | 1041 | status = exit_code & 0x7f; |
1044 | } | 1042 | } |
1045 | return wait_noreap_copyout(wo, p, pid, uid, why, status); | 1043 | return wait_noreap_copyout(wo, p, pid, uid, why, status); |
1046 | } | 1044 | } |
1047 | 1045 | ||
1048 | /* | 1046 | /* |
1049 | * Try to move the task's state to DEAD | 1047 | * Try to move the task's state to DEAD |
1050 | * only one thread is allowed to do this: | 1048 | * only one thread is allowed to do this: |
1051 | */ | 1049 | */ |
1052 | state = xchg(&p->exit_state, EXIT_DEAD); | 1050 | state = xchg(&p->exit_state, EXIT_DEAD); |
1053 | if (state != EXIT_ZOMBIE) { | 1051 | if (state != EXIT_ZOMBIE) { |
1054 | BUG_ON(state != EXIT_DEAD); | 1052 | BUG_ON(state != EXIT_DEAD); |
1055 | return 0; | 1053 | return 0; |
1056 | } | 1054 | } |
1057 | 1055 | ||
1058 | traced = ptrace_reparented(p); | 1056 | traced = ptrace_reparented(p); |
1059 | /* | 1057 | /* |
1060 | * It can be ptraced but not reparented, check | 1058 | * It can be ptraced but not reparented, check |
1061 | * thread_group_leader() to filter out sub-threads. | 1059 | * thread_group_leader() to filter out sub-threads. |
1062 | */ | 1060 | */ |
1063 | if (likely(!traced) && thread_group_leader(p)) { | 1061 | if (likely(!traced) && thread_group_leader(p)) { |
1064 | struct signal_struct *psig; | 1062 | struct signal_struct *psig; |
1065 | struct signal_struct *sig; | 1063 | struct signal_struct *sig; |
1066 | unsigned long maxrss; | 1064 | unsigned long maxrss; |
1067 | cputime_t tgutime, tgstime; | 1065 | cputime_t tgutime, tgstime; |
1068 | 1066 | ||
1069 | /* | 1067 | /* |
1070 | * The resource counters for the group leader are in its | 1068 | * The resource counters for the group leader are in its |
1071 | * own task_struct. Those for dead threads in the group | 1069 | * own task_struct. Those for dead threads in the group |
1072 | * are in its signal_struct, as are those for the child | 1070 | * are in its signal_struct, as are those for the child |
1073 | * processes it has previously reaped. All these | 1071 | * processes it has previously reaped. All these |
1074 | * accumulate in the parent's signal_struct c* fields. | 1072 | * accumulate in the parent's signal_struct c* fields. |
1075 | * | 1073 | * |
1076 | * We don't bother to take a lock here to protect these | 1074 | * We don't bother to take a lock here to protect these |
1077 | * p->signal fields, because they are only touched by | 1075 | * p->signal fields, because they are only touched by |
1078 | * __exit_signal, which runs with tasklist_lock | 1076 | * __exit_signal, which runs with tasklist_lock |
1079 | * write-locked anyway, and so is excluded here. We do | 1077 | * write-locked anyway, and so is excluded here. We do |
1080 | * need to protect the access to parent->signal fields, | 1078 | * need to protect the access to parent->signal fields, |
1081 | * as other threads in the parent group can be right | 1079 | * as other threads in the parent group can be right |
1082 | * here reaping other children at the same time. | 1080 | * here reaping other children at the same time. |
1083 | * | 1081 | * |
1084 | * We use thread_group_cputime_adjusted() to get times for the thread | 1082 | * We use thread_group_cputime_adjusted() to get times for the thread |
1085 | * group, which consolidates times for all threads in the | 1083 | * group, which consolidates times for all threads in the |
1086 | * group including the group leader. | 1084 | * group including the group leader. |
1087 | */ | 1085 | */ |
1088 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); | 1086 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); |
1089 | spin_lock_irq(&p->real_parent->sighand->siglock); | 1087 | spin_lock_irq(&p->real_parent->sighand->siglock); |
1090 | psig = p->real_parent->signal; | 1088 | psig = p->real_parent->signal; |
1091 | sig = p->signal; | 1089 | sig = p->signal; |
1092 | psig->cutime += tgutime + sig->cutime; | 1090 | psig->cutime += tgutime + sig->cutime; |
1093 | psig->cstime += tgstime + sig->cstime; | 1091 | psig->cstime += tgstime + sig->cstime; |
1094 | psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime; | 1092 | psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime; |
1095 | psig->cmin_flt += | 1093 | psig->cmin_flt += |
1096 | p->min_flt + sig->min_flt + sig->cmin_flt; | 1094 | p->min_flt + sig->min_flt + sig->cmin_flt; |
1097 | psig->cmaj_flt += | 1095 | psig->cmaj_flt += |
1098 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | 1096 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; |
1099 | psig->cnvcsw += | 1097 | psig->cnvcsw += |
1100 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | 1098 | p->nvcsw + sig->nvcsw + sig->cnvcsw; |
1101 | psig->cnivcsw += | 1099 | psig->cnivcsw += |
1102 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | 1100 | p->nivcsw + sig->nivcsw + sig->cnivcsw; |
1103 | psig->cinblock += | 1101 | psig->cinblock += |
1104 | task_io_get_inblock(p) + | 1102 | task_io_get_inblock(p) + |
1105 | sig->inblock + sig->cinblock; | 1103 | sig->inblock + sig->cinblock; |
1106 | psig->coublock += | 1104 | psig->coublock += |
1107 | task_io_get_oublock(p) + | 1105 | task_io_get_oublock(p) + |
1108 | sig->oublock + sig->coublock; | 1106 | sig->oublock + sig->coublock; |
1109 | maxrss = max(sig->maxrss, sig->cmaxrss); | 1107 | maxrss = max(sig->maxrss, sig->cmaxrss); |
1110 | if (psig->cmaxrss < maxrss) | 1108 | if (psig->cmaxrss < maxrss) |
1111 | psig->cmaxrss = maxrss; | 1109 | psig->cmaxrss = maxrss; |
1112 | task_io_accounting_add(&psig->ioac, &p->ioac); | 1110 | task_io_accounting_add(&psig->ioac, &p->ioac); |
1113 | task_io_accounting_add(&psig->ioac, &sig->ioac); | 1111 | task_io_accounting_add(&psig->ioac, &sig->ioac); |
1114 | spin_unlock_irq(&p->real_parent->sighand->siglock); | 1112 | spin_unlock_irq(&p->real_parent->sighand->siglock); |
1115 | } | 1113 | } |
1116 | 1114 | ||
1117 | /* | 1115 | /* |
1118 | * Now we are sure this task is interesting, and no other | 1116 | * Now we are sure this task is interesting, and no other |
1119 | * thread can reap it because we set its state to EXIT_DEAD. | 1117 | * thread can reap it because we set its state to EXIT_DEAD. |
1120 | */ | 1118 | */ |
1121 | read_unlock(&tasklist_lock); | 1119 | read_unlock(&tasklist_lock); |
1122 | 1120 | ||
1123 | retval = wo->wo_rusage | 1121 | retval = wo->wo_rusage |
1124 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | 1122 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; |
1125 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) | 1123 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1126 | ? p->signal->group_exit_code : p->exit_code; | 1124 | ? p->signal->group_exit_code : p->exit_code; |
1127 | if (!retval && wo->wo_stat) | 1125 | if (!retval && wo->wo_stat) |
1128 | retval = put_user(status, wo->wo_stat); | 1126 | retval = put_user(status, wo->wo_stat); |
1129 | 1127 | ||
1130 | infop = wo->wo_info; | 1128 | infop = wo->wo_info; |
1131 | if (!retval && infop) | 1129 | if (!retval && infop) |
1132 | retval = put_user(SIGCHLD, &infop->si_signo); | 1130 | retval = put_user(SIGCHLD, &infop->si_signo); |
1133 | if (!retval && infop) | 1131 | if (!retval && infop) |
1134 | retval = put_user(0, &infop->si_errno); | 1132 | retval = put_user(0, &infop->si_errno); |
1135 | if (!retval && infop) { | 1133 | if (!retval && infop) { |
1136 | int why; | 1134 | int why; |
1137 | 1135 | ||
1138 | if ((status & 0x7f) == 0) { | 1136 | if ((status & 0x7f) == 0) { |
1139 | why = CLD_EXITED; | 1137 | why = CLD_EXITED; |
1140 | status >>= 8; | 1138 | status >>= 8; |
1141 | } else { | 1139 | } else { |
1142 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | 1140 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; |
1143 | status &= 0x7f; | 1141 | status &= 0x7f; |
1144 | } | 1142 | } |
1145 | retval = put_user((short)why, &infop->si_code); | 1143 | retval = put_user((short)why, &infop->si_code); |
1146 | if (!retval) | 1144 | if (!retval) |
1147 | retval = put_user(status, &infop->si_status); | 1145 | retval = put_user(status, &infop->si_status); |
1148 | } | 1146 | } |
1149 | if (!retval && infop) | 1147 | if (!retval && infop) |
1150 | retval = put_user(pid, &infop->si_pid); | 1148 | retval = put_user(pid, &infop->si_pid); |
1151 | if (!retval && infop) | 1149 | if (!retval && infop) |
1152 | retval = put_user(uid, &infop->si_uid); | 1150 | retval = put_user(uid, &infop->si_uid); |
1153 | if (!retval) | 1151 | if (!retval) |
1154 | retval = pid; | 1152 | retval = pid; |
1155 | 1153 | ||
1156 | if (traced) { | 1154 | if (traced) { |
1157 | write_lock_irq(&tasklist_lock); | 1155 | write_lock_irq(&tasklist_lock); |
1158 | /* We dropped tasklist, ptracer could die and untrace */ | 1156 | /* We dropped tasklist, ptracer could die and untrace */ |
1159 | ptrace_unlink(p); | 1157 | ptrace_unlink(p); |
1160 | /* | 1158 | /* |
1161 | * If this is not a sub-thread, notify the parent. | 1159 | * If this is not a sub-thread, notify the parent. |
1162 | * If parent wants a zombie, don't release it now. | 1160 | * If parent wants a zombie, don't release it now. |
1163 | */ | 1161 | */ |
1164 | if (thread_group_leader(p) && | 1162 | if (thread_group_leader(p) && |
1165 | !do_notify_parent(p, p->exit_signal)) { | 1163 | !do_notify_parent(p, p->exit_signal)) { |
1166 | p->exit_state = EXIT_ZOMBIE; | 1164 | p->exit_state = EXIT_ZOMBIE; |
1167 | p = NULL; | 1165 | p = NULL; |
1168 | } | 1166 | } |
1169 | write_unlock_irq(&tasklist_lock); | 1167 | write_unlock_irq(&tasklist_lock); |
1170 | } | 1168 | } |
1171 | if (p != NULL) | 1169 | if (p != NULL) |
1172 | release_task(p); | 1170 | release_task(p); |
1173 | 1171 | ||
1174 | return retval; | 1172 | return retval; |
1175 | } | 1173 | } |
1176 | 1174 | ||
1177 | static int *task_stopped_code(struct task_struct *p, bool ptrace) | 1175 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
1178 | { | 1176 | { |
1179 | if (ptrace) { | 1177 | if (ptrace) { |
1180 | if (task_is_stopped_or_traced(p) && | 1178 | if (task_is_stopped_or_traced(p) && |
1181 | !(p->jobctl & JOBCTL_LISTENING)) | 1179 | !(p->jobctl & JOBCTL_LISTENING)) |
1182 | return &p->exit_code; | 1180 | return &p->exit_code; |
1183 | } else { | 1181 | } else { |
1184 | if (p->signal->flags & SIGNAL_STOP_STOPPED) | 1182 | if (p->signal->flags & SIGNAL_STOP_STOPPED) |
1185 | return &p->signal->group_exit_code; | 1183 | return &p->signal->group_exit_code; |
1186 | } | 1184 | } |
1187 | return NULL; | 1185 | return NULL; |
1188 | } | 1186 | } |
1189 | 1187 | ||
1190 | /** | 1188 | /** |
1191 | * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED | 1189 | * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED |
1192 | * @wo: wait options | 1190 | * @wo: wait options |
1193 | * @ptrace: is the wait for ptrace | 1191 | * @ptrace: is the wait for ptrace |
1194 | * @p: task to wait for | 1192 | * @p: task to wait for |
1195 | * | 1193 | * |
1196 | * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED. | 1194 | * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED. |
1197 | * | 1195 | * |
1198 | * CONTEXT: | 1196 | * CONTEXT: |
1199 | * read_lock(&tasklist_lock), which is released if return value is | 1197 | * read_lock(&tasklist_lock), which is released if return value is |
1200 | * non-zero. Also, grabs and releases @p->sighand->siglock. | 1198 | * non-zero. Also, grabs and releases @p->sighand->siglock. |
1201 | * | 1199 | * |
1202 | * RETURNS: | 1200 | * RETURNS: |
1203 | * 0 if wait condition didn't exist and search for other wait conditions | 1201 | * 0 if wait condition didn't exist and search for other wait conditions |
1204 | * should continue. Non-zero return, -errno on failure and @p's pid on | 1202 | * should continue. Non-zero return, -errno on failure and @p's pid on |
1205 | * success, implies that tasklist_lock is released and wait condition | 1203 | * success, implies that tasklist_lock is released and wait condition |
1206 | * search should terminate. | 1204 | * search should terminate. |
1207 | */ | 1205 | */ |
1208 | static int wait_task_stopped(struct wait_opts *wo, | 1206 | static int wait_task_stopped(struct wait_opts *wo, |
1209 | int ptrace, struct task_struct *p) | 1207 | int ptrace, struct task_struct *p) |
1210 | { | 1208 | { |
1211 | struct siginfo __user *infop; | 1209 | struct siginfo __user *infop; |
1212 | int retval, exit_code, *p_code, why; | 1210 | int retval, exit_code, *p_code, why; |
1213 | uid_t uid = 0; /* unneeded, required by compiler */ | 1211 | uid_t uid = 0; /* unneeded, required by compiler */ |
1214 | pid_t pid; | 1212 | pid_t pid; |
1215 | 1213 | ||
1216 | /* | 1214 | /* |
1217 | * Traditionally we see ptrace'd stopped tasks regardless of options. | 1215 | * Traditionally we see ptrace'd stopped tasks regardless of options. |
1218 | */ | 1216 | */ |
1219 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) | 1217 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
1220 | return 0; | 1218 | return 0; |
1221 | 1219 | ||
1222 | if (!task_stopped_code(p, ptrace)) | 1220 | if (!task_stopped_code(p, ptrace)) |
1223 | return 0; | 1221 | return 0; |
1224 | 1222 | ||
1225 | exit_code = 0; | 1223 | exit_code = 0; |
1226 | spin_lock_irq(&p->sighand->siglock); | 1224 | spin_lock_irq(&p->sighand->siglock); |
1227 | 1225 | ||
1228 | p_code = task_stopped_code(p, ptrace); | 1226 | p_code = task_stopped_code(p, ptrace); |
1229 | if (unlikely(!p_code)) | 1227 | if (unlikely(!p_code)) |
1230 | goto unlock_sig; | 1228 | goto unlock_sig; |
1231 | 1229 | ||
1232 | exit_code = *p_code; | 1230 | exit_code = *p_code; |
1233 | if (!exit_code) | 1231 | if (!exit_code) |
1234 | goto unlock_sig; | 1232 | goto unlock_sig; |
1235 | 1233 | ||
1236 | if (!unlikely(wo->wo_flags & WNOWAIT)) | 1234 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
1237 | *p_code = 0; | 1235 | *p_code = 0; |
1238 | 1236 | ||
1239 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); | 1237 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
1240 | unlock_sig: | 1238 | unlock_sig: |
1241 | spin_unlock_irq(&p->sighand->siglock); | 1239 | spin_unlock_irq(&p->sighand->siglock); |
1242 | if (!exit_code) | 1240 | if (!exit_code) |
1243 | return 0; | 1241 | return 0; |
1244 | 1242 | ||
1245 | /* | 1243 | /* |
1246 | * Now we are pretty sure this task is interesting. | 1244 | * Now we are pretty sure this task is interesting. |
1247 | * Make sure it doesn't get reaped out from under us while we | 1245 | * Make sure it doesn't get reaped out from under us while we |
1248 | * give up the lock and then examine it below. We don't want to | 1246 | * give up the lock and then examine it below. We don't want to |
1249 | * keep holding onto the tasklist_lock while we call getrusage and | 1247 | * keep holding onto the tasklist_lock while we call getrusage and |
1250 | * possibly take page faults for user memory. | 1248 | * possibly take page faults for user memory. |
1251 | */ | 1249 | */ |
1252 | get_task_struct(p); | 1250 | get_task_struct(p); |
1253 | pid = task_pid_vnr(p); | 1251 | pid = task_pid_vnr(p); |
1254 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; | 1252 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
1255 | read_unlock(&tasklist_lock); | 1253 | read_unlock(&tasklist_lock); |
1256 | 1254 | ||
1257 | if (unlikely(wo->wo_flags & WNOWAIT)) | 1255 | if (unlikely(wo->wo_flags & WNOWAIT)) |
1258 | return wait_noreap_copyout(wo, p, pid, uid, why, exit_code); | 1256 | return wait_noreap_copyout(wo, p, pid, uid, why, exit_code); |
1259 | 1257 | ||
1260 | retval = wo->wo_rusage | 1258 | retval = wo->wo_rusage |
1261 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | 1259 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; |
1262 | if (!retval && wo->wo_stat) | 1260 | if (!retval && wo->wo_stat) |
1263 | retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat); | 1261 | retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat); |
1264 | 1262 | ||
1265 | infop = wo->wo_info; | 1263 | infop = wo->wo_info; |
1266 | if (!retval && infop) | 1264 | if (!retval && infop) |
1267 | retval = put_user(SIGCHLD, &infop->si_signo); | 1265 | retval = put_user(SIGCHLD, &infop->si_signo); |
1268 | if (!retval && infop) | 1266 | if (!retval && infop) |
1269 | retval = put_user(0, &infop->si_errno); | 1267 | retval = put_user(0, &infop->si_errno); |
1270 | if (!retval && infop) | 1268 | if (!retval && infop) |
1271 | retval = put_user((short)why, &infop->si_code); | 1269 | retval = put_user((short)why, &infop->si_code); |
1272 | if (!retval && infop) | 1270 | if (!retval && infop) |
1273 | retval = put_user(exit_code, &infop->si_status); | 1271 | retval = put_user(exit_code, &infop->si_status); |
1274 | if (!retval && infop) | 1272 | if (!retval && infop) |
1275 | retval = put_user(pid, &infop->si_pid); | 1273 | retval = put_user(pid, &infop->si_pid); |
1276 | if (!retval && infop) | 1274 | if (!retval && infop) |
1277 | retval = put_user(uid, &infop->si_uid); | 1275 | retval = put_user(uid, &infop->si_uid); |
1278 | if (!retval) | 1276 | if (!retval) |
1279 | retval = pid; | 1277 | retval = pid; |
1280 | put_task_struct(p); | 1278 | put_task_struct(p); |
1281 | 1279 | ||
1282 | BUG_ON(!retval); | 1280 | BUG_ON(!retval); |
1283 | return retval; | 1281 | return retval; |
1284 | } | 1282 | } |
1285 | 1283 | ||
1286 | /* | 1284 | /* |
1287 | * Handle do_wait work for one task in a live, non-stopped state. | 1285 | * Handle do_wait work for one task in a live, non-stopped state. |
1288 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | 1286 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold |
1289 | * the lock and this task is uninteresting. If we return nonzero, we have | 1287 | * the lock and this task is uninteresting. If we return nonzero, we have |
1290 | * released the lock and the system call should return. | 1288 | * released the lock and the system call should return. |
1291 | */ | 1289 | */ |
1292 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) | 1290 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) |
1293 | { | 1291 | { |
1294 | int retval; | 1292 | int retval; |
1295 | pid_t pid; | 1293 | pid_t pid; |
1296 | uid_t uid; | 1294 | uid_t uid; |
1297 | 1295 | ||
1298 | if (!unlikely(wo->wo_flags & WCONTINUED)) | 1296 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
1299 | return 0; | 1297 | return 0; |
1300 | 1298 | ||
1301 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) | 1299 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1302 | return 0; | 1300 | return 0; |
1303 | 1301 | ||
1304 | spin_lock_irq(&p->sighand->siglock); | 1302 | spin_lock_irq(&p->sighand->siglock); |
1305 | /* Re-check with the lock held. */ | 1303 | /* Re-check with the lock held. */ |
1306 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | 1304 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { |
1307 | spin_unlock_irq(&p->sighand->siglock); | 1305 | spin_unlock_irq(&p->sighand->siglock); |
1308 | return 0; | 1306 | return 0; |
1309 | } | 1307 | } |
1310 | if (!unlikely(wo->wo_flags & WNOWAIT)) | 1308 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
1311 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; | 1309 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
1312 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); | 1310 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
1313 | spin_unlock_irq(&p->sighand->siglock); | 1311 | spin_unlock_irq(&p->sighand->siglock); |
1314 | 1312 | ||
1315 | pid = task_pid_vnr(p); | 1313 | pid = task_pid_vnr(p); |
1316 | get_task_struct(p); | 1314 | get_task_struct(p); |
1317 | read_unlock(&tasklist_lock); | 1315 | read_unlock(&tasklist_lock); |
1318 | 1316 | ||
1319 | if (!wo->wo_info) { | 1317 | if (!wo->wo_info) { |
1320 | retval = wo->wo_rusage | 1318 | retval = wo->wo_rusage |
1321 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | 1319 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; |
1322 | put_task_struct(p); | 1320 | put_task_struct(p); |
1323 | if (!retval && wo->wo_stat) | 1321 | if (!retval && wo->wo_stat) |
1324 | retval = put_user(0xffff, wo->wo_stat); | 1322 | retval = put_user(0xffff, wo->wo_stat); |
1325 | if (!retval) | 1323 | if (!retval) |
1326 | retval = pid; | 1324 | retval = pid; |
1327 | } else { | 1325 | } else { |
1328 | retval = wait_noreap_copyout(wo, p, pid, uid, | 1326 | retval = wait_noreap_copyout(wo, p, pid, uid, |
1329 | CLD_CONTINUED, SIGCONT); | 1327 | CLD_CONTINUED, SIGCONT); |
1330 | BUG_ON(retval == 0); | 1328 | BUG_ON(retval == 0); |
1331 | } | 1329 | } |
1332 | 1330 | ||
1333 | return retval; | 1331 | return retval; |
1334 | } | 1332 | } |
1335 | 1333 | ||
1336 | /* | 1334 | /* |
1337 | * Consider @p for a wait by @parent. | 1335 | * Consider @p for a wait by @parent. |
1338 | * | 1336 | * |
1339 | * -ECHILD should be in ->notask_error before the first call. | 1337 | * -ECHILD should be in ->notask_error before the first call. |
1340 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. | 1338 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1341 | * Returns zero if the search for a child should continue; | 1339 | * Returns zero if the search for a child should continue; |
1342 | * then ->notask_error is 0 if @p is an eligible child, | 1340 | * then ->notask_error is 0 if @p is an eligible child, |
1343 | * or another error from security_task_wait(), or still -ECHILD. | 1341 | * or another error from security_task_wait(), or still -ECHILD. |
1344 | */ | 1342 | */ |
1345 | static int wait_consider_task(struct wait_opts *wo, int ptrace, | 1343 | static int wait_consider_task(struct wait_opts *wo, int ptrace, |
1346 | struct task_struct *p) | 1344 | struct task_struct *p) |
1347 | { | 1345 | { |
1348 | int ret = eligible_child(wo, p); | 1346 | int ret = eligible_child(wo, p); |
1349 | if (!ret) | 1347 | if (!ret) |
1350 | return ret; | 1348 | return ret; |
1351 | 1349 | ||
1352 | ret = security_task_wait(p); | 1350 | ret = security_task_wait(p); |
1353 | if (unlikely(ret < 0)) { | 1351 | if (unlikely(ret < 0)) { |
1354 | /* | 1352 | /* |
1355 | * If we have not yet seen any eligible child, | 1353 | * If we have not yet seen any eligible child, |
1356 | * then let this error code replace -ECHILD. | 1354 | * then let this error code replace -ECHILD. |
1357 | * A permission error will give the user a clue | 1355 | * A permission error will give the user a clue |
1358 | * to look for security policy problems, rather | 1356 | * to look for security policy problems, rather |
1359 | * than for mysterious wait bugs. | 1357 | * than for mysterious wait bugs. |
1360 | */ | 1358 | */ |
1361 | if (wo->notask_error) | 1359 | if (wo->notask_error) |
1362 | wo->notask_error = ret; | 1360 | wo->notask_error = ret; |
1363 | return 0; | 1361 | return 0; |
1364 | } | 1362 | } |
1365 | 1363 | ||
1366 | /* dead body doesn't have much to contribute */ | 1364 | /* dead body doesn't have much to contribute */ |
1367 | if (unlikely(p->exit_state == EXIT_DEAD)) { | 1365 | if (unlikely(p->exit_state == EXIT_DEAD)) { |
1368 | /* | 1366 | /* |
1369 | * But do not ignore this task until the tracer does | 1367 | * But do not ignore this task until the tracer does |
1370 | * wait_task_zombie()->do_notify_parent(). | 1368 | * wait_task_zombie()->do_notify_parent(). |
1371 | */ | 1369 | */ |
1372 | if (likely(!ptrace) && unlikely(ptrace_reparented(p))) | 1370 | if (likely(!ptrace) && unlikely(ptrace_reparented(p))) |
1373 | wo->notask_error = 0; | 1371 | wo->notask_error = 0; |
1374 | return 0; | 1372 | return 0; |
1375 | } | 1373 | } |
1376 | 1374 | ||
1377 | /* slay zombie? */ | 1375 | /* slay zombie? */ |
1378 | if (p->exit_state == EXIT_ZOMBIE) { | 1376 | if (p->exit_state == EXIT_ZOMBIE) { |
1379 | /* | 1377 | /* |
1380 | * A zombie ptracee is only visible to its ptracer. | 1378 | * A zombie ptracee is only visible to its ptracer. |
1381 | * Notification and reaping will be cascaded to the real | 1379 | * Notification and reaping will be cascaded to the real |
1382 | * parent when the ptracer detaches. | 1380 | * parent when the ptracer detaches. |
1383 | */ | 1381 | */ |
1384 | if (likely(!ptrace) && unlikely(p->ptrace)) { | 1382 | if (likely(!ptrace) && unlikely(p->ptrace)) { |
1385 | /* it will become visible, clear notask_error */ | 1383 | /* it will become visible, clear notask_error */ |
1386 | wo->notask_error = 0; | 1384 | wo->notask_error = 0; |
1387 | return 0; | 1385 | return 0; |
1388 | } | 1386 | } |
1389 | 1387 | ||
1390 | /* we don't reap group leaders with subthreads */ | 1388 | /* we don't reap group leaders with subthreads */ |
1391 | if (!delay_group_leader(p)) | 1389 | if (!delay_group_leader(p)) |
1392 | return wait_task_zombie(wo, p); | 1390 | return wait_task_zombie(wo, p); |
1393 | 1391 | ||
1394 | /* | 1392 | /* |
1395 | * Allow access to stopped/continued state via zombie by | 1393 | * Allow access to stopped/continued state via zombie by |
1396 | * falling through. Clearing of notask_error is complex. | 1394 | * falling through. Clearing of notask_error is complex. |
1397 | * | 1395 | * |
1398 | * When !@ptrace: | 1396 | * When !@ptrace: |
1399 | * | 1397 | * |
1400 | * If WEXITED is set, notask_error should naturally be | 1398 | * If WEXITED is set, notask_error should naturally be |
1401 | * cleared. If not, subset of WSTOPPED|WCONTINUED is set, | 1399 | * cleared. If not, subset of WSTOPPED|WCONTINUED is set, |
1402 | * so, if there are live subthreads, there are events to | 1400 | * so, if there are live subthreads, there are events to |
1403 | * wait for. If all subthreads are dead, it's still safe | 1401 | * wait for. If all subthreads are dead, it's still safe |
1404 | * to clear - this function will be called again in finite | 1402 | * to clear - this function will be called again in finite |
1405 | * amount time once all the subthreads are released and | 1403 | * amount time once all the subthreads are released and |
1406 | * will then return without clearing. | 1404 | * will then return without clearing. |
1407 | * | 1405 | * |
1408 | * When @ptrace: | 1406 | * When @ptrace: |
1409 | * | 1407 | * |
1410 | * Stopped state is per-task and thus can't change once the | 1408 | * Stopped state is per-task and thus can't change once the |
1411 | * target task dies. Only continued and exited can happen. | 1409 | * target task dies. Only continued and exited can happen. |
1412 | * Clear notask_error if WCONTINUED | WEXITED. | 1410 | * Clear notask_error if WCONTINUED | WEXITED. |
1413 | */ | 1411 | */ |
1414 | if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED))) | 1412 | if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED))) |
1415 | wo->notask_error = 0; | 1413 | wo->notask_error = 0; |
1416 | } else { | 1414 | } else { |
1417 | /* | 1415 | /* |
1418 | * If @p is ptraced by a task in its real parent's group, | 1416 | * If @p is ptraced by a task in its real parent's group, |
1419 | * hide group stop/continued state when looking at @p as | 1417 | * hide group stop/continued state when looking at @p as |
1420 | * the real parent; otherwise, a single stop can be | 1418 | * the real parent; otherwise, a single stop can be |
1421 | * reported twice as group and ptrace stops. | 1419 | * reported twice as group and ptrace stops. |
1422 | * | 1420 | * |
1423 | * If a ptracer wants to distinguish the two events for its | 1421 | * If a ptracer wants to distinguish the two events for its |
1424 | * own children, it should create a separate process which | 1422 | * own children, it should create a separate process which |
1425 | * takes the role of real parent. | 1423 | * takes the role of real parent. |
1426 | */ | 1424 | */ |
1427 | if (likely(!ptrace) && p->ptrace && !ptrace_reparented(p)) | 1425 | if (likely(!ptrace) && p->ptrace && !ptrace_reparented(p)) |
1428 | return 0; | 1426 | return 0; |
1429 | 1427 | ||
1430 | /* | 1428 | /* |
1431 | * @p is alive and it's gonna stop, continue or exit, so | 1429 | * @p is alive and it's gonna stop, continue or exit, so |
1432 | * there always is something to wait for. | 1430 | * there always is something to wait for. |
1433 | */ | 1431 | */ |
1434 | wo->notask_error = 0; | 1432 | wo->notask_error = 0; |
1435 | } | 1433 | } |
1436 | 1434 | ||
1437 | /* | 1435 | /* |
1438 | * Wait for stopped. Depending on @ptrace, different stopped state | 1436 | * Wait for stopped. Depending on @ptrace, different stopped state |
1439 | * is used and the two don't interact with each other. | 1437 | * is used and the two don't interact with each other. |
1440 | */ | 1438 | */ |
1441 | ret = wait_task_stopped(wo, ptrace, p); | 1439 | ret = wait_task_stopped(wo, ptrace, p); |
1442 | if (ret) | 1440 | if (ret) |
1443 | return ret; | 1441 | return ret; |
1444 | 1442 | ||
1445 | /* | 1443 | /* |
1446 | * Wait for continued. There's only one continued state and the | 1444 | * Wait for continued. There's only one continued state and the |
1447 | * ptracer can consume it which can confuse the real parent. Don't | 1445 | * ptracer can consume it which can confuse the real parent. Don't |
1448 | * use WCONTINUED from ptracer. You don't need or want it. | 1446 | * use WCONTINUED from ptracer. You don't need or want it. |
1449 | */ | 1447 | */ |
1450 | return wait_task_continued(wo, p); | 1448 | return wait_task_continued(wo, p); |
1451 | } | 1449 | } |
1452 | 1450 | ||
1453 | /* | 1451 | /* |
1454 | * Do the work of do_wait() for one thread in the group, @tsk. | 1452 | * Do the work of do_wait() for one thread in the group, @tsk. |
1455 | * | 1453 | * |
1456 | * -ECHILD should be in ->notask_error before the first call. | 1454 | * -ECHILD should be in ->notask_error before the first call. |
1457 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. | 1455 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1458 | * Returns zero if the search for a child should continue; then | 1456 | * Returns zero if the search for a child should continue; then |
1459 | * ->notask_error is 0 if there were any eligible children, | 1457 | * ->notask_error is 0 if there were any eligible children, |
1460 | * or another error from security_task_wait(), or still -ECHILD. | 1458 | * or another error from security_task_wait(), or still -ECHILD. |
1461 | */ | 1459 | */ |
1462 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) | 1460 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) |
1463 | { | 1461 | { |
1464 | struct task_struct *p; | 1462 | struct task_struct *p; |
1465 | 1463 | ||
1466 | list_for_each_entry(p, &tsk->children, sibling) { | 1464 | list_for_each_entry(p, &tsk->children, sibling) { |
1467 | int ret = wait_consider_task(wo, 0, p); | 1465 | int ret = wait_consider_task(wo, 0, p); |
1468 | if (ret) | 1466 | if (ret) |
1469 | return ret; | 1467 | return ret; |
1470 | } | 1468 | } |
1471 | 1469 | ||
1472 | return 0; | 1470 | return 0; |
1473 | } | 1471 | } |
1474 | 1472 | ||
1475 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) | 1473 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) |
1476 | { | 1474 | { |
1477 | struct task_struct *p; | 1475 | struct task_struct *p; |
1478 | 1476 | ||
1479 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { | 1477 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
1480 | int ret = wait_consider_task(wo, 1, p); | 1478 | int ret = wait_consider_task(wo, 1, p); |
1481 | if (ret) | 1479 | if (ret) |
1482 | return ret; | 1480 | return ret; |
1483 | } | 1481 | } |
1484 | 1482 | ||
1485 | return 0; | 1483 | return 0; |
1486 | } | 1484 | } |
1487 | 1485 | ||
1488 | static int child_wait_callback(wait_queue_t *wait, unsigned mode, | 1486 | static int child_wait_callback(wait_queue_t *wait, unsigned mode, |
1489 | int sync, void *key) | 1487 | int sync, void *key) |
1490 | { | 1488 | { |
1491 | struct wait_opts *wo = container_of(wait, struct wait_opts, | 1489 | struct wait_opts *wo = container_of(wait, struct wait_opts, |
1492 | child_wait); | 1490 | child_wait); |
1493 | struct task_struct *p = key; | 1491 | struct task_struct *p = key; |
1494 | 1492 | ||
1495 | if (!eligible_pid(wo, p)) | 1493 | if (!eligible_pid(wo, p)) |
1496 | return 0; | 1494 | return 0; |
1497 | 1495 | ||
1498 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) | 1496 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) |
1499 | return 0; | 1497 | return 0; |
1500 | 1498 | ||
1501 | return default_wake_function(wait, mode, sync, key); | 1499 | return default_wake_function(wait, mode, sync, key); |
1502 | } | 1500 | } |
1503 | 1501 | ||
1504 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) | 1502 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) |
1505 | { | 1503 | { |
1506 | __wake_up_sync_key(&parent->signal->wait_chldexit, | 1504 | __wake_up_sync_key(&parent->signal->wait_chldexit, |
1507 | TASK_INTERRUPTIBLE, 1, p); | 1505 | TASK_INTERRUPTIBLE, 1, p); |
1508 | } | 1506 | } |
1509 | 1507 | ||
1510 | static long do_wait(struct wait_opts *wo) | 1508 | static long do_wait(struct wait_opts *wo) |
1511 | { | 1509 | { |
1512 | struct task_struct *tsk; | 1510 | struct task_struct *tsk; |
1513 | int retval; | 1511 | int retval; |
1514 | 1512 | ||
1515 | trace_sched_process_wait(wo->wo_pid); | 1513 | trace_sched_process_wait(wo->wo_pid); |
1516 | 1514 | ||
1517 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); | 1515 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); |
1518 | wo->child_wait.private = current; | 1516 | wo->child_wait.private = current; |
1519 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); | 1517 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
1520 | repeat: | 1518 | repeat: |
1521 | /* | 1519 | /* |
1522 | * If there is nothing that can match our critiera just get out. | 1520 | * If there is nothing that can match our critiera just get out. |
1523 | * We will clear ->notask_error to zero if we see any child that | 1521 | * We will clear ->notask_error to zero if we see any child that |
1524 | * might later match our criteria, even if we are not able to reap | 1522 | * might later match our criteria, even if we are not able to reap |
1525 | * it yet. | 1523 | * it yet. |
1526 | */ | 1524 | */ |
1527 | wo->notask_error = -ECHILD; | 1525 | wo->notask_error = -ECHILD; |
1528 | if ((wo->wo_type < PIDTYPE_MAX) && | 1526 | if ((wo->wo_type < PIDTYPE_MAX) && |
1529 | (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type]))) | 1527 | (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type]))) |
1530 | goto notask; | 1528 | goto notask; |
1531 | 1529 | ||
1532 | set_current_state(TASK_INTERRUPTIBLE); | 1530 | set_current_state(TASK_INTERRUPTIBLE); |
1533 | read_lock(&tasklist_lock); | 1531 | read_lock(&tasklist_lock); |
1534 | tsk = current; | 1532 | tsk = current; |
1535 | do { | 1533 | do { |
1536 | retval = do_wait_thread(wo, tsk); | 1534 | retval = do_wait_thread(wo, tsk); |
1537 | if (retval) | 1535 | if (retval) |
1538 | goto end; | 1536 | goto end; |
1539 | 1537 | ||
1540 | retval = ptrace_do_wait(wo, tsk); | 1538 | retval = ptrace_do_wait(wo, tsk); |
1541 | if (retval) | 1539 | if (retval) |
1542 | goto end; | 1540 | goto end; |
1543 | 1541 | ||
1544 | if (wo->wo_flags & __WNOTHREAD) | 1542 | if (wo->wo_flags & __WNOTHREAD) |
1545 | break; | 1543 | break; |
1546 | } while_each_thread(current, tsk); | 1544 | } while_each_thread(current, tsk); |
1547 | read_unlock(&tasklist_lock); | 1545 | read_unlock(&tasklist_lock); |
1548 | 1546 | ||
1549 | notask: | 1547 | notask: |
1550 | retval = wo->notask_error; | 1548 | retval = wo->notask_error; |
1551 | if (!retval && !(wo->wo_flags & WNOHANG)) { | 1549 | if (!retval && !(wo->wo_flags & WNOHANG)) { |
1552 | retval = -ERESTARTSYS; | 1550 | retval = -ERESTARTSYS; |
1553 | if (!signal_pending(current)) { | 1551 | if (!signal_pending(current)) { |
1554 | schedule(); | 1552 | schedule(); |
1555 | goto repeat; | 1553 | goto repeat; |
1556 | } | 1554 | } |
1557 | } | 1555 | } |
1558 | end: | 1556 | end: |
1559 | __set_current_state(TASK_RUNNING); | 1557 | __set_current_state(TASK_RUNNING); |
1560 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); | 1558 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
1561 | return retval; | 1559 | return retval; |
1562 | } | 1560 | } |
1563 | 1561 | ||
1564 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, | 1562 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
1565 | infop, int, options, struct rusage __user *, ru) | 1563 | infop, int, options, struct rusage __user *, ru) |
1566 | { | 1564 | { |
1567 | struct wait_opts wo; | 1565 | struct wait_opts wo; |
1568 | struct pid *pid = NULL; | 1566 | struct pid *pid = NULL; |
1569 | enum pid_type type; | 1567 | enum pid_type type; |
1570 | long ret; | 1568 | long ret; |
1571 | 1569 | ||
1572 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED)) | 1570 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED)) |
1573 | return -EINVAL; | 1571 | return -EINVAL; |
1574 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | 1572 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) |
1575 | return -EINVAL; | 1573 | return -EINVAL; |
1576 | 1574 | ||
1577 | switch (which) { | 1575 | switch (which) { |
1578 | case P_ALL: | 1576 | case P_ALL: |
1579 | type = PIDTYPE_MAX; | 1577 | type = PIDTYPE_MAX; |
1580 | break; | 1578 | break; |
1581 | case P_PID: | 1579 | case P_PID: |
1582 | type = PIDTYPE_PID; | 1580 | type = PIDTYPE_PID; |
1583 | if (upid <= 0) | 1581 | if (upid <= 0) |
1584 | return -EINVAL; | 1582 | return -EINVAL; |
1585 | break; | 1583 | break; |
1586 | case P_PGID: | 1584 | case P_PGID: |
1587 | type = PIDTYPE_PGID; | 1585 | type = PIDTYPE_PGID; |
1588 | if (upid <= 0) | 1586 | if (upid <= 0) |
1589 | return -EINVAL; | 1587 | return -EINVAL; |
1590 | break; | 1588 | break; |
1591 | default: | 1589 | default: |
1592 | return -EINVAL; | 1590 | return -EINVAL; |
1593 | } | 1591 | } |
1594 | 1592 | ||
1595 | if (type < PIDTYPE_MAX) | 1593 | if (type < PIDTYPE_MAX) |
1596 | pid = find_get_pid(upid); | 1594 | pid = find_get_pid(upid); |
1597 | 1595 | ||
1598 | wo.wo_type = type; | 1596 | wo.wo_type = type; |
1599 | wo.wo_pid = pid; | 1597 | wo.wo_pid = pid; |
1600 | wo.wo_flags = options; | 1598 | wo.wo_flags = options; |
1601 | wo.wo_info = infop; | 1599 | wo.wo_info = infop; |
1602 | wo.wo_stat = NULL; | 1600 | wo.wo_stat = NULL; |
1603 | wo.wo_rusage = ru; | 1601 | wo.wo_rusage = ru; |
1604 | ret = do_wait(&wo); | 1602 | ret = do_wait(&wo); |
1605 | 1603 | ||
1606 | if (ret > 0) { | 1604 | if (ret > 0) { |
1607 | ret = 0; | 1605 | ret = 0; |
1608 | } else if (infop) { | 1606 | } else if (infop) { |
1609 | /* | 1607 | /* |
1610 | * For a WNOHANG return, clear out all the fields | 1608 | * For a WNOHANG return, clear out all the fields |
1611 | * we would set so the user can easily tell the | 1609 | * we would set so the user can easily tell the |
1612 | * difference. | 1610 | * difference. |
1613 | */ | 1611 | */ |
1614 | if (!ret) | 1612 | if (!ret) |
1615 | ret = put_user(0, &infop->si_signo); | 1613 | ret = put_user(0, &infop->si_signo); |
1616 | if (!ret) | 1614 | if (!ret) |
1617 | ret = put_user(0, &infop->si_errno); | 1615 | ret = put_user(0, &infop->si_errno); |
1618 | if (!ret) | 1616 | if (!ret) |
1619 | ret = put_user(0, &infop->si_code); | 1617 | ret = put_user(0, &infop->si_code); |
1620 | if (!ret) | 1618 | if (!ret) |
1621 | ret = put_user(0, &infop->si_pid); | 1619 | ret = put_user(0, &infop->si_pid); |
1622 | if (!ret) | 1620 | if (!ret) |
1623 | ret = put_user(0, &infop->si_uid); | 1621 | ret = put_user(0, &infop->si_uid); |
1624 | if (!ret) | 1622 | if (!ret) |
1625 | ret = put_user(0, &infop->si_status); | 1623 | ret = put_user(0, &infop->si_status); |
1626 | } | 1624 | } |
1627 | 1625 | ||
1628 | put_pid(pid); | 1626 | put_pid(pid); |
1629 | return ret; | 1627 | return ret; |
1630 | } | 1628 | } |
1631 | 1629 | ||
1632 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, | 1630 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
1633 | int, options, struct rusage __user *, ru) | 1631 | int, options, struct rusage __user *, ru) |
1634 | { | 1632 | { |
1635 | struct wait_opts wo; | 1633 | struct wait_opts wo; |
1636 | struct pid *pid = NULL; | 1634 | struct pid *pid = NULL; |
1637 | enum pid_type type; | 1635 | enum pid_type type; |
1638 | long ret; | 1636 | long ret; |
1639 | 1637 | ||
1640 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | 1638 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| |
1641 | __WNOTHREAD|__WCLONE|__WALL)) | 1639 | __WNOTHREAD|__WCLONE|__WALL)) |
1642 | return -EINVAL; | 1640 | return -EINVAL; |
1643 | 1641 | ||
1644 | if (upid == -1) | 1642 | if (upid == -1) |
1645 | type = PIDTYPE_MAX; | 1643 | type = PIDTYPE_MAX; |
1646 | else if (upid < 0) { | 1644 | else if (upid < 0) { |
1647 | type = PIDTYPE_PGID; | 1645 | type = PIDTYPE_PGID; |
1648 | pid = find_get_pid(-upid); | 1646 | pid = find_get_pid(-upid); |
1649 | } else if (upid == 0) { | 1647 | } else if (upid == 0) { |
1650 | type = PIDTYPE_PGID; | 1648 | type = PIDTYPE_PGID; |
1651 | pid = get_task_pid(current, PIDTYPE_PGID); | 1649 | pid = get_task_pid(current, PIDTYPE_PGID); |
1652 | } else /* upid > 0 */ { | 1650 | } else /* upid > 0 */ { |
1653 | type = PIDTYPE_PID; | 1651 | type = PIDTYPE_PID; |
1654 | pid = find_get_pid(upid); | 1652 | pid = find_get_pid(upid); |
1655 | } | 1653 | } |
1656 | 1654 | ||
1657 | wo.wo_type = type; | 1655 | wo.wo_type = type; |
1658 | wo.wo_pid = pid; | 1656 | wo.wo_pid = pid; |
1659 | wo.wo_flags = options | WEXITED; | 1657 | wo.wo_flags = options | WEXITED; |
1660 | wo.wo_info = NULL; | 1658 | wo.wo_info = NULL; |
1661 | wo.wo_stat = stat_addr; | 1659 | wo.wo_stat = stat_addr; |
1662 | wo.wo_rusage = ru; | 1660 | wo.wo_rusage = ru; |
1663 | ret = do_wait(&wo); | 1661 | ret = do_wait(&wo); |
1664 | put_pid(pid); | 1662 | put_pid(pid); |
1665 | 1663 | ||
1666 | return ret; | 1664 | return ret; |
1667 | } | 1665 | } |
1668 | 1666 | ||
1669 | #ifdef __ARCH_WANT_SYS_WAITPID | 1667 | #ifdef __ARCH_WANT_SYS_WAITPID |
1670 | 1668 | ||
1671 | /* | 1669 | /* |
1672 | * sys_waitpid() remains for compatibility. waitpid() should be | 1670 | * sys_waitpid() remains for compatibility. waitpid() should be |
1673 | * implemented by calling sys_wait4() from libc.a. | 1671 | * implemented by calling sys_wait4() from libc.a. |
1674 | */ | 1672 | */ |
1675 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) | 1673 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
1676 | { | 1674 | { |
1677 | return sys_wait4(pid, stat_addr, options, NULL); | 1675 | return sys_wait4(pid, stat_addr, options, NULL); |
1678 | } | 1676 | } |
1679 | 1677 |