Doug / smarc-fsl-linux-kernel

Download as
Browse Code ยป

Commit 45cdf5cc0703c537194588c63d53bad1f2539d36

Authored by Oleg Nesterov
1 parent 53c8f9f199
Exists in master and in 6 other branches imx_3.10.17_1.0.1_ga, imx_3.10.53_1.1.0_ga, imx_3.14.28_1.0.0_ga, smarc-imx_3.10.53_1.1.0_ga, smarc-imx_3.14.28_1.0.0_ga, smarc-l5.0.0_1.0.0-ga

kill tracehook_notify_death() 

Kill tracehook_notify_death(), reimplement the logic in its caller,
exit_notify().

Also, change the exec_id's check to use thread_group_leader() instead
of task_detached(), this is more clear. This logic only applies to
the exiting leader, a sub-thread must never change its exit_signal.

Note: when the traced group leader exits the exit_signal-or-SIGCHLD
logic looks really strange:

	- we notify the tracer even if !thread_group_empty() but
	   do_wait(WEXITED) can't work until all threads exit

	- if the tracer is real_parent, it is not clear why can't
	  we use ->exit_signal event if !thread_group_empty()

-v2: do not try to fix the 2nd oddity to avoid the subtle behavior
     change mixed with reorganization, suggested by Tejun.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Reviewed-by: Tejun Heo <tj@kernel.org>

Showing 2 changed files with 13 additions and 42 deletions Inline Diff

include/linux/tracehook.h
Diff comments   View file @ 45cdf5c
1 /* 1 /*
2 * Tracing hooks 2 * Tracing hooks
3 * 3 *
4 * Copyright (C) 2008-2009 Red Hat, Inc. All rights reserved. 4 * Copyright (C) 2008-2009 Red Hat, Inc. All rights reserved.
5 * 5 *
6 * This copyrighted material is made available to anyone wishing to use, 6 * This copyrighted material is made available to anyone wishing to use,
7 * modify, copy, or redistribute it subject to the terms and conditions 7 * modify, copy, or redistribute it subject to the terms and conditions
8 * of the GNU General Public License v.2. 8 * of the GNU General Public License v.2.
9 * 9 *
10 * This file defines hook entry points called by core code where 10 * This file defines hook entry points called by core code where
11 * user tracing/debugging support might need to do something. These 11 * user tracing/debugging support might need to do something. These
12 * entry points are called tracehook_*(). Each hook declared below 12 * entry points are called tracehook_*(). Each hook declared below
13 * has a detailed kerneldoc comment giving the context (locking et 13 * has a detailed kerneldoc comment giving the context (locking et
14 * al) from which it is called, and the meaning of its return value. 14 * al) from which it is called, and the meaning of its return value.
15 * 15 *
16 * Each function here typically has only one call site, so it is ok 16 * Each function here typically has only one call site, so it is ok
17 * to have some nontrivial tracehook_*() inlines. In all cases, the 17 * to have some nontrivial tracehook_*() inlines. In all cases, the
18 * fast path when no tracing is enabled should be very short. 18 * fast path when no tracing is enabled should be very short.
19 * 19 *
20 * The purpose of this file and the tracehook_* layer is to consolidate 20 * The purpose of this file and the tracehook_* layer is to consolidate
21 * the interface that the kernel core and arch code uses to enable any 21 * the interface that the kernel core and arch code uses to enable any
22 * user debugging or tracing facility (such as ptrace). The interfaces 22 * user debugging or tracing facility (such as ptrace). The interfaces
23 * here are carefully documented so that maintainers of core and arch 23 * here are carefully documented so that maintainers of core and arch
24 * code do not need to think about the implementation details of the 24 * code do not need to think about the implementation details of the
25 * tracing facilities. Likewise, maintainers of the tracing code do not 25 * tracing facilities. Likewise, maintainers of the tracing code do not
26 * need to understand all the calling core or arch code in detail, just 26 * need to understand all the calling core or arch code in detail, just
27 * documented circumstances of each call, such as locking conditions. 27 * documented circumstances of each call, such as locking conditions.
28 * 28 *
29 * If the calling core code changes so that locking is different, then 29 * If the calling core code changes so that locking is different, then
30 * it is ok to change the interface documented here. The maintainer of 30 * it is ok to change the interface documented here. The maintainer of
31 * core code changing should notify the maintainers of the tracing code 31 * core code changing should notify the maintainers of the tracing code
32 * that they need to work out the change. 32 * that they need to work out the change.
33 * 33 *
34 * Some tracehook_*() inlines take arguments that the current tracing 34 * Some tracehook_*() inlines take arguments that the current tracing
35 * implementations might not necessarily use. These function signatures 35 * implementations might not necessarily use. These function signatures
36 * are chosen to pass in all the information that is on hand in the 36 * are chosen to pass in all the information that is on hand in the
37 * caller and might conceivably be relevant to a tracer, so that the 37 * caller and might conceivably be relevant to a tracer, so that the
38 * core code won't have to be updated when tracing adds more features. 38 * core code won't have to be updated when tracing adds more features.
39 * If a call site changes so that some of those parameters are no longer 39 * If a call site changes so that some of those parameters are no longer
40 * already on hand without extra work, then the tracehook_* interface 40 * already on hand without extra work, then the tracehook_* interface
41 * can change so there is no make-work burden on the core code. The 41 * can change so there is no make-work burden on the core code. The
42 * maintainer of core code changing should notify the maintainers of the 42 * maintainer of core code changing should notify the maintainers of the
43 * tracing code that they need to work out the change. 43 * tracing code that they need to work out the change.
44 */ 44 */
45 45
46 #ifndef _LINUX_TRACEHOOK_H 46 #ifndef _LINUX_TRACEHOOK_H
47 #define _LINUX_TRACEHOOK_H 1 47 #define _LINUX_TRACEHOOK_H 1
48 48
49 #include <linux/sched.h> 49 #include <linux/sched.h>
50 #include <linux/ptrace.h> 50 #include <linux/ptrace.h>
51 #include <linux/security.h> 51 #include <linux/security.h>
52 struct linux_binprm; 52 struct linux_binprm;
53 53
54 /* 54 /*
55 * ptrace report for syscall entry and exit looks identical. 55 * ptrace report for syscall entry and exit looks identical.
56 */ 56 */
57 static inline void ptrace_report_syscall(struct pt_regs *regs) 57 static inline void ptrace_report_syscall(struct pt_regs *regs)
58 { 58 {
59 int ptrace = current->ptrace; 59 int ptrace = current->ptrace;
60 60
61 if (!(ptrace & PT_PTRACED)) 61 if (!(ptrace & PT_PTRACED))
62 return; 62 return;
63 63
64 ptrace_notify(SIGTRAP | ((ptrace & PT_TRACESYSGOOD) ? 0x80 : 0)); 64 ptrace_notify(SIGTRAP | ((ptrace & PT_TRACESYSGOOD) ? 0x80 : 0));
65 65
66 /* 66 /*
67 * this isn't the same as continuing with a signal, but it will do 67 * this isn't the same as continuing with a signal, but it will do
68 * for normal use. strace only continues with a signal if the 68 * for normal use. strace only continues with a signal if the
69 * stopping signal is not SIGTRAP. -brl 69 * stopping signal is not SIGTRAP. -brl
70 */ 70 */
71 if (current->exit_code) { 71 if (current->exit_code) {
72 send_sig(current->exit_code, current, 1); 72 send_sig(current->exit_code, current, 1);
73 current->exit_code = 0; 73 current->exit_code = 0;
74 } 74 }
75 } 75 }
76 76
77 /** 77 /**
78 * tracehook_report_syscall_entry - task is about to attempt a system call 78 * tracehook_report_syscall_entry - task is about to attempt a system call
79 * @regs: user register state of current task 79 * @regs: user register state of current task
80 * 80 *
81 * This will be called if %TIF_SYSCALL_TRACE has been set, when the 81 * This will be called if %TIF_SYSCALL_TRACE has been set, when the
82 * current task has just entered the kernel for a system call. 82 * current task has just entered the kernel for a system call.
83 * Full user register state is available here. Changing the values 83 * Full user register state is available here. Changing the values
84 * in @regs can affect the system call number and arguments to be tried. 84 * in @regs can affect the system call number and arguments to be tried.
85 * It is safe to block here, preventing the system call from beginning. 85 * It is safe to block here, preventing the system call from beginning.
86 * 86 *
87 * Returns zero normally, or nonzero if the calling arch code should abort 87 * Returns zero normally, or nonzero if the calling arch code should abort
88 * the system call. That must prevent normal entry so no system call is 88 * the system call. That must prevent normal entry so no system call is
89 * made. If @task ever returns to user mode after this, its register state 89 * made. If @task ever returns to user mode after this, its register state
90 * is unspecified, but should be something harmless like an %ENOSYS error 90 * is unspecified, but should be something harmless like an %ENOSYS error
91 * return. It should preserve enough information so that syscall_rollback() 91 * return. It should preserve enough information so that syscall_rollback()
92 * can work (see asm-generic/syscall.h). 92 * can work (see asm-generic/syscall.h).
93 * 93 *
94 * Called without locks, just after entering kernel mode. 94 * Called without locks, just after entering kernel mode.
95 */ 95 */
96 static inline __must_check int tracehook_report_syscall_entry( 96 static inline __must_check int tracehook_report_syscall_entry(
97 struct pt_regs *regs) 97 struct pt_regs *regs)
98 { 98 {
99 ptrace_report_syscall(regs); 99 ptrace_report_syscall(regs);
100 return 0; 100 return 0;
101 } 101 }
102 102
103 /** 103 /**
104 * tracehook_report_syscall_exit - task has just finished a system call 104 * tracehook_report_syscall_exit - task has just finished a system call
105 * @regs: user register state of current task 105 * @regs: user register state of current task
106 * @step: nonzero if simulating single-step or block-step 106 * @step: nonzero if simulating single-step or block-step
107 * 107 *
108 * This will be called if %TIF_SYSCALL_TRACE has been set, when the 108 * This will be called if %TIF_SYSCALL_TRACE has been set, when the
109 * current task has just finished an attempted system call. Full 109 * current task has just finished an attempted system call. Full
110 * user register state is available here. It is safe to block here, 110 * user register state is available here. It is safe to block here,
111 * preventing signals from being processed. 111 * preventing signals from being processed.
112 * 112 *
113 * If @step is nonzero, this report is also in lieu of the normal 113 * If @step is nonzero, this report is also in lieu of the normal
114 * trap that would follow the system call instruction because 114 * trap that would follow the system call instruction because
115 * user_enable_block_step() or user_enable_single_step() was used. 115 * user_enable_block_step() or user_enable_single_step() was used.
116 * In this case, %TIF_SYSCALL_TRACE might not be set. 116 * In this case, %TIF_SYSCALL_TRACE might not be set.
117 * 117 *
118 * Called without locks, just before checking for pending signals. 118 * Called without locks, just before checking for pending signals.
119 */ 119 */
120 static inline void tracehook_report_syscall_exit(struct pt_regs *regs, int step) 120 static inline void tracehook_report_syscall_exit(struct pt_regs *regs, int step)
121 { 121 {
122 if (step) { 122 if (step) {
123 siginfo_t info; 123 siginfo_t info;
124 user_single_step_siginfo(current, regs, &info); 124 user_single_step_siginfo(current, regs, &info);
125 force_sig_info(SIGTRAP, &info, current); 125 force_sig_info(SIGTRAP, &info, current);
126 return; 126 return;
127 } 127 }
128 128
129 ptrace_report_syscall(regs); 129 ptrace_report_syscall(regs);
130 } 130 }
131 131
132 /** 132 /**
133 * tracehook_signal_handler - signal handler setup is complete 133 * tracehook_signal_handler - signal handler setup is complete
134 * @sig: number of signal being delivered 134 * @sig: number of signal being delivered
135 * @info: siginfo_t of signal being delivered 135 * @info: siginfo_t of signal being delivered
136 * @ka: sigaction setting that chose the handler 136 * @ka: sigaction setting that chose the handler
137 * @regs: user register state 137 * @regs: user register state
138 * @stepping: nonzero if debugger single-step or block-step in use 138 * @stepping: nonzero if debugger single-step or block-step in use
139 * 139 *
140 * Called by the arch code after a signal handler has been set up. 140 * Called by the arch code after a signal handler has been set up.
141 * Register and stack state reflects the user handler about to run. 141 * Register and stack state reflects the user handler about to run.
142 * Signal mask changes have already been made. 142 * Signal mask changes have already been made.
143 * 143 *
144 * Called without locks, shortly before returning to user mode 144 * Called without locks, shortly before returning to user mode
145 * (or handling more signals). 145 * (or handling more signals).
146 */ 146 */
147 static inline void tracehook_signal_handler(int sig, siginfo_t *info, 147 static inline void tracehook_signal_handler(int sig, siginfo_t *info,
148 const struct k_sigaction *ka, 148 const struct k_sigaction *ka,
149 struct pt_regs *regs, int stepping) 149 struct pt_regs *regs, int stepping)
150 { 150 {
151 if (stepping) 151 if (stepping)
152 ptrace_notify(SIGTRAP); 152 ptrace_notify(SIGTRAP);
153 } 153 }
154 154
155 #define DEATH_REAP -1
156 #define DEATH_DELAYED_GROUP_LEADER -2
157
158 /**
159 * tracehook_notify_death - task is dead, ready to notify parent
160 * @task: @current task now exiting
161 * @death_cookie: value to pass to tracehook_report_death()
162 * @group_dead: nonzero if this was the last thread in the group to die
163 *
164 * A return value >= 0 means call do_notify_parent() with that signal
165 * number. Negative return value can be %DEATH_REAP to self-reap right
166 * now, or %DEATH_DELAYED_GROUP_LEADER to a zombie without notifying our
167 * parent. Note that a return value of 0 means a do_notify_parent() call
168 * that sends no signal, but still wakes up a parent blocked in wait*().
169 *
170 * Called with write_lock_irq(&tasklist_lock) held.
171 */
172 static inline int tracehook_notify_death(struct task_struct *task,
173 void **death_cookie, int group_dead)
174 {
175 if (task_detached(task))
176 return task->ptrace ? SIGCHLD : DEATH_REAP;
177
178 /*
179 * If something other than our normal parent is ptracing us, then
180 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
181 * only has special meaning to our real parent.
182 */
183 if (thread_group_empty(task) && !ptrace_reparented(task))
184 return task->exit_signal;
185
186 return task->ptrace ? SIGCHLD : DEATH_DELAYED_GROUP_LEADER;
187 }
188
189 #ifdef TIF_NOTIFY_RESUME 155 #ifdef TIF_NOTIFY_RESUME
190 /** 156 /**
191 * set_notify_resume - cause tracehook_notify_resume() to be called 157 * set_notify_resume - cause tracehook_notify_resume() to be called
192 * @task: task that will call tracehook_notify_resume() 158 * @task: task that will call tracehook_notify_resume()
193 * 159 *
194 * Calling this arranges that @task will call tracehook_notify_resume() 160 * Calling this arranges that @task will call tracehook_notify_resume()
195 * before returning to user mode. If it's already running in user mode, 161 * before returning to user mode. If it's already running in user mode,
196 * it will enter the kernel and call tracehook_notify_resume() soon. 162 * it will enter the kernel and call tracehook_notify_resume() soon.
197 * If it's blocked, it will not be woken. 163 * If it's blocked, it will not be woken.
198 */ 164 */
199 static inline void set_notify_resume(struct task_struct *task) 165 static inline void set_notify_resume(struct task_struct *task)
200 { 166 {
201 if (!test_and_set_tsk_thread_flag(task, TIF_NOTIFY_RESUME)) 167 if (!test_and_set_tsk_thread_flag(task, TIF_NOTIFY_RESUME))
202 kick_process(task); 168 kick_process(task);
203 } 169 }
204 170
205 /** 171 /**
206 * tracehook_notify_resume - report when about to return to user mode 172 * tracehook_notify_resume - report when about to return to user mode
207 * @regs: user-mode registers of @current task 173 * @regs: user-mode registers of @current task
208 * 174 *
209 * This is called when %TIF_NOTIFY_RESUME has been set. Now we are 175 * This is called when %TIF_NOTIFY_RESUME has been set. Now we are
210 * about to return to user mode, and the user state in @regs can be 176 * about to return to user mode, and the user state in @regs can be
211 * inspected or adjusted. The caller in arch code has cleared 177 * inspected or adjusted. The caller in arch code has cleared
212 * %TIF_NOTIFY_RESUME before the call. If the flag gets set again 178 * %TIF_NOTIFY_RESUME before the call. If the flag gets set again
213 * asynchronously, this will be called again before we return to 179 * asynchronously, this will be called again before we return to
214 * user mode. 180 * user mode.
215 * 181 *
216 * Called without locks. 182 * Called without locks.
217 */ 183 */
218 static inline void tracehook_notify_resume(struct pt_regs *regs) 184 static inline void tracehook_notify_resume(struct pt_regs *regs)
219 { 185 {
220 } 186 }
221 #endif /* TIF_NOTIFY_RESUME */ 187 #endif /* TIF_NOTIFY_RESUME */
222 188
223 #endif /* <linux/tracehook.h> */ 189 #endif /* <linux/tracehook.h> */
224 190
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/binfmts.h> 23 #include <linux/binfmts.h>
24 #include <linux/nsproxy.h> 24 #include <linux/nsproxy.h>
25 #include <linux/pid_namespace.h> 25 #include <linux/pid_namespace.h>
26 #include <linux/ptrace.h> 26 #include <linux/ptrace.h>
27 #include <linux/profile.h> 27 #include <linux/profile.h>
28 #include <linux/mount.h> 28 #include <linux/mount.h>
29 #include <linux/proc_fs.h> 29 #include <linux/proc_fs.h>
30 #include <linux/kthread.h> 30 #include <linux/kthread.h>
31 #include <linux/mempolicy.h> 31 #include <linux/mempolicy.h>
32 #include <linux/taskstats_kern.h> 32 #include <linux/taskstats_kern.h>
33 #include <linux/delayacct.h> 33 #include <linux/delayacct.h>
34 #include <linux/freezer.h> 34 #include <linux/freezer.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 54
55 #include <asm/uaccess.h> 55 #include <asm/uaccess.h>
56 #include <asm/unistd.h> 56 #include <asm/unistd.h>
57 #include <asm/pgtable.h> 57 #include <asm/pgtable.h>
58 #include <asm/mmu_context.h> 58 #include <asm/mmu_context.h>
59 59
60 static void exit_mm(struct task_struct * tsk); 60 static void exit_mm(struct task_struct * tsk);
61 61
62 static void __unhash_process(struct task_struct *p, bool group_dead) 62 static void __unhash_process(struct task_struct *p, bool group_dead)
63 { 63 {
64 nr_threads--; 64 nr_threads--;
65 detach_pid(p, PIDTYPE_PID); 65 detach_pid(p, PIDTYPE_PID);
66 if (group_dead) { 66 if (group_dead) {
67 detach_pid(p, PIDTYPE_PGID); 67 detach_pid(p, PIDTYPE_PGID);
68 detach_pid(p, PIDTYPE_SID); 68 detach_pid(p, PIDTYPE_SID);
69 69
70 list_del_rcu(&p->tasks); 70 list_del_rcu(&p->tasks);
71 list_del_init(&p->sibling); 71 list_del_init(&p->sibling);
72 __this_cpu_dec(process_counts); 72 __this_cpu_dec(process_counts);
73 } 73 }
74 list_del_rcu(&p->thread_group); 74 list_del_rcu(&p->thread_group);
75 } 75 }
76 76
77 /* 77 /*
78 * This function expects the tasklist_lock write-locked. 78 * This function expects the tasklist_lock write-locked.
79 */ 79 */
80 static void __exit_signal(struct task_struct *tsk) 80 static void __exit_signal(struct task_struct *tsk)
81 { 81 {
82 struct signal_struct *sig = tsk->signal; 82 struct signal_struct *sig = tsk->signal;
83 bool group_dead = thread_group_leader(tsk); 83 bool group_dead = thread_group_leader(tsk);
84 struct sighand_struct *sighand; 84 struct sighand_struct *sighand;
85 struct tty_struct *uninitialized_var(tty); 85 struct tty_struct *uninitialized_var(tty);
86 86
87 sighand = rcu_dereference_check(tsk->sighand, 87 sighand = rcu_dereference_check(tsk->sighand,
88 rcu_read_lock_held() || 88 rcu_read_lock_held() ||
89 lockdep_tasklist_lock_is_held()); 89 lockdep_tasklist_lock_is_held());
90 spin_lock(&sighand->siglock); 90 spin_lock(&sighand->siglock);
91 91
92 posix_cpu_timers_exit(tsk); 92 posix_cpu_timers_exit(tsk);
93 if (group_dead) { 93 if (group_dead) {
94 posix_cpu_timers_exit_group(tsk); 94 posix_cpu_timers_exit_group(tsk);
95 tty = sig->tty; 95 tty = sig->tty;
96 sig->tty = NULL; 96 sig->tty = NULL;
97 } else { 97 } else {
98 /* 98 /*
99 * This can only happen if the caller is de_thread(). 99 * This can only happen if the caller is de_thread().
100 * FIXME: this is the temporary hack, we should teach 100 * FIXME: this is the temporary hack, we should teach
101 * posix-cpu-timers to handle this case correctly. 101 * posix-cpu-timers to handle this case correctly.
102 */ 102 */
103 if (unlikely(has_group_leader_pid(tsk))) 103 if (unlikely(has_group_leader_pid(tsk)))
104 posix_cpu_timers_exit_group(tsk); 104 posix_cpu_timers_exit_group(tsk);
105 105
106 /* 106 /*
107 * If there is any task waiting for the group exit 107 * If there is any task waiting for the group exit
108 * then notify it: 108 * then notify it:
109 */ 109 */
110 if (sig->notify_count > 0 && !--sig->notify_count) 110 if (sig->notify_count > 0 && !--sig->notify_count)
111 wake_up_process(sig->group_exit_task); 111 wake_up_process(sig->group_exit_task);
112 112
113 if (tsk == sig->curr_target) 113 if (tsk == sig->curr_target)
114 sig->curr_target = next_thread(tsk); 114 sig->curr_target = next_thread(tsk);
115 /* 115 /*
116 * Accumulate here the counters for all threads but the 116 * Accumulate here the counters for all threads but the
117 * group leader as they die, so they can be added into 117 * group leader as they die, so they can be added into
118 * the process-wide totals when those are taken. 118 * the process-wide totals when those are taken.
119 * The group leader stays around as a zombie as long 119 * The group leader stays around as a zombie as long
120 * as there are other threads. When it gets reaped, 120 * as there are other threads. When it gets reaped,
121 * the exit.c code will add its counts into these totals. 121 * the exit.c code will add its counts into these totals.
122 * We won't ever get here for the group leader, since it 122 * We won't ever get here for the group leader, since it
123 * will have been the last reference on the signal_struct. 123 * will have been the last reference on the signal_struct.
124 */ 124 */
125 sig->utime = cputime_add(sig->utime, tsk->utime); 125 sig->utime = cputime_add(sig->utime, tsk->utime);
126 sig->stime = cputime_add(sig->stime, tsk->stime); 126 sig->stime = cputime_add(sig->stime, tsk->stime);
127 sig->gtime = cputime_add(sig->gtime, tsk->gtime); 127 sig->gtime = cputime_add(sig->gtime, tsk->gtime);
128 sig->min_flt += tsk->min_flt; 128 sig->min_flt += tsk->min_flt;
129 sig->maj_flt += tsk->maj_flt; 129 sig->maj_flt += tsk->maj_flt;
130 sig->nvcsw += tsk->nvcsw; 130 sig->nvcsw += tsk->nvcsw;
131 sig->nivcsw += tsk->nivcsw; 131 sig->nivcsw += tsk->nivcsw;
132 sig->inblock += task_io_get_inblock(tsk); 132 sig->inblock += task_io_get_inblock(tsk);
133 sig->oublock += task_io_get_oublock(tsk); 133 sig->oublock += task_io_get_oublock(tsk);
134 task_io_accounting_add(&sig->ioac, &tsk->ioac); 134 task_io_accounting_add(&sig->ioac, &tsk->ioac);
135 sig->sum_sched_runtime += tsk->se.sum_exec_runtime; 135 sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
136 } 136 }
137 137
138 sig->nr_threads--; 138 sig->nr_threads--;
139 __unhash_process(tsk, group_dead); 139 __unhash_process(tsk, group_dead);
140 140
141 /* 141 /*
142 * Do this under ->siglock, we can race with another thread 142 * Do this under ->siglock, we can race with another thread
143 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. 143 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
144 */ 144 */
145 flush_sigqueue(&tsk->pending); 145 flush_sigqueue(&tsk->pending);
146 tsk->sighand = NULL; 146 tsk->sighand = NULL;
147 spin_unlock(&sighand->siglock); 147 spin_unlock(&sighand->siglock);
148 148
149 __cleanup_sighand(sighand); 149 __cleanup_sighand(sighand);
150 clear_tsk_thread_flag(tsk,TIF_SIGPENDING); 150 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
151 if (group_dead) { 151 if (group_dead) {
152 flush_sigqueue(&sig->shared_pending); 152 flush_sigqueue(&sig->shared_pending);
153 tty_kref_put(tty); 153 tty_kref_put(tty);
154 } 154 }
155 } 155 }
156 156
157 static void delayed_put_task_struct(struct rcu_head *rhp) 157 static void delayed_put_task_struct(struct rcu_head *rhp)
158 { 158 {
159 struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); 159 struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
160 160
161 perf_event_delayed_put(tsk); 161 perf_event_delayed_put(tsk);
162 trace_sched_process_free(tsk); 162 trace_sched_process_free(tsk);
163 put_task_struct(tsk); 163 put_task_struct(tsk);
164 } 164 }
165 165
166 166
167 void release_task(struct task_struct * p) 167 void release_task(struct task_struct * p)
168 { 168 {
169 struct task_struct *leader; 169 struct task_struct *leader;
170 int zap_leader; 170 int zap_leader;
171 repeat: 171 repeat:
172 /* don't need to get the RCU readlock here - the process is dead and 172 /* don't need to get the RCU readlock here - the process is dead and
173 * can't be modifying its own credentials. But shut RCU-lockdep up */ 173 * can't be modifying its own credentials. But shut RCU-lockdep up */
174 rcu_read_lock(); 174 rcu_read_lock();
175 atomic_dec(&__task_cred(p)->user->processes); 175 atomic_dec(&__task_cred(p)->user->processes);
176 rcu_read_unlock(); 176 rcu_read_unlock();
177 177
178 proc_flush_task(p); 178 proc_flush_task(p);
179 179
180 write_lock_irq(&tasklist_lock); 180 write_lock_irq(&tasklist_lock);
181 ptrace_release_task(p); 181 ptrace_release_task(p);
182 __exit_signal(p); 182 __exit_signal(p);
183 183
184 /* 184 /*
185 * If we are the last non-leader member of the thread 185 * If we are the last non-leader member of the thread
186 * group, and the leader is zombie, then notify the 186 * group, and the leader is zombie, then notify the
187 * group leader's parent process. (if it wants notification.) 187 * group leader's parent process. (if it wants notification.)
188 */ 188 */
189 zap_leader = 0; 189 zap_leader = 0;
190 leader = p->group_leader; 190 leader = p->group_leader;
191 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) { 191 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
192 BUG_ON(task_detached(leader)); 192 BUG_ON(task_detached(leader));
193 do_notify_parent(leader, leader->exit_signal); 193 do_notify_parent(leader, leader->exit_signal);
194 /* 194 /*
195 * If we were the last child thread and the leader has 195 * If we were the last child thread and the leader has
196 * exited already, and the leader's parent ignores SIGCHLD, 196 * exited already, and the leader's parent ignores SIGCHLD,
197 * then we are the one who should release the leader. 197 * then we are the one who should release the leader.
198 * 198 *
199 * do_notify_parent() will have marked it self-reaping in 199 * do_notify_parent() will have marked it self-reaping in
200 * that case. 200 * that case.
201 */ 201 */
202 zap_leader = task_detached(leader); 202 zap_leader = task_detached(leader);
203 203
204 /* 204 /*
205 * This maintains the invariant that release_task() 205 * This maintains the invariant that release_task()
206 * only runs on a task in EXIT_DEAD, just for sanity. 206 * only runs on a task in EXIT_DEAD, just for sanity.
207 */ 207 */
208 if (zap_leader) 208 if (zap_leader)
209 leader->exit_state = EXIT_DEAD; 209 leader->exit_state = EXIT_DEAD;
210 } 210 }
211 211
212 write_unlock_irq(&tasklist_lock); 212 write_unlock_irq(&tasklist_lock);
213 release_thread(p); 213 release_thread(p);
214 call_rcu(&p->rcu, delayed_put_task_struct); 214 call_rcu(&p->rcu, delayed_put_task_struct);
215 215
216 p = leader; 216 p = leader;
217 if (unlikely(zap_leader)) 217 if (unlikely(zap_leader))
218 goto repeat; 218 goto repeat;
219 } 219 }
220 220
221 /* 221 /*
222 * This checks not only the pgrp, but falls back on the pid if no 222 * This checks not only the pgrp, but falls back on the pid if no
223 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly 223 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
224 * without this... 224 * without this...
225 * 225 *
226 * The caller must hold rcu lock or the tasklist lock. 226 * The caller must hold rcu lock or the tasklist lock.
227 */ 227 */
228 struct pid *session_of_pgrp(struct pid *pgrp) 228 struct pid *session_of_pgrp(struct pid *pgrp)
229 { 229 {
230 struct task_struct *p; 230 struct task_struct *p;
231 struct pid *sid = NULL; 231 struct pid *sid = NULL;
232 232
233 p = pid_task(pgrp, PIDTYPE_PGID); 233 p = pid_task(pgrp, PIDTYPE_PGID);
234 if (p == NULL) 234 if (p == NULL)
235 p = pid_task(pgrp, PIDTYPE_PID); 235 p = pid_task(pgrp, PIDTYPE_PID);
236 if (p != NULL) 236 if (p != NULL)
237 sid = task_session(p); 237 sid = task_session(p);
238 238
239 return sid; 239 return sid;
240 } 240 }
241 241
242 /* 242 /*
243 * Determine if a process group is "orphaned", according to the POSIX 243 * Determine if a process group is "orphaned", according to the POSIX
244 * definition in 2.2.2.52. Orphaned process groups are not to be affected 244 * definition in 2.2.2.52. Orphaned process groups are not to be affected
245 * by terminal-generated stop signals. Newly orphaned process groups are 245 * by terminal-generated stop signals. Newly orphaned process groups are
246 * to receive a SIGHUP and a SIGCONT. 246 * to receive a SIGHUP and a SIGCONT.
247 * 247 *
248 * "I ask you, have you ever known what it is to be an orphan?" 248 * "I ask you, have you ever known what it is to be an orphan?"
249 */ 249 */
250 static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task) 250 static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task)
251 { 251 {
252 struct task_struct *p; 252 struct task_struct *p;
253 253
254 do_each_pid_task(pgrp, PIDTYPE_PGID, p) { 254 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
255 if ((p == ignored_task) || 255 if ((p == ignored_task) ||
256 (p->exit_state && thread_group_empty(p)) || 256 (p->exit_state && thread_group_empty(p)) ||
257 is_global_init(p->real_parent)) 257 is_global_init(p->real_parent))
258 continue; 258 continue;
259 259
260 if (task_pgrp(p->real_parent) != pgrp && 260 if (task_pgrp(p->real_parent) != pgrp &&
261 task_session(p->real_parent) == task_session(p)) 261 task_session(p->real_parent) == task_session(p))
262 return 0; 262 return 0;
263 } while_each_pid_task(pgrp, PIDTYPE_PGID, p); 263 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
264 264
265 return 1; 265 return 1;
266 } 266 }
267 267
268 int is_current_pgrp_orphaned(void) 268 int is_current_pgrp_orphaned(void)
269 { 269 {
270 int retval; 270 int retval;
271 271
272 read_lock(&tasklist_lock); 272 read_lock(&tasklist_lock);
273 retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); 273 retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
274 read_unlock(&tasklist_lock); 274 read_unlock(&tasklist_lock);
275 275
276 return retval; 276 return retval;
277 } 277 }
278 278
279 static int has_stopped_jobs(struct pid *pgrp) 279 static int has_stopped_jobs(struct pid *pgrp)
280 { 280 {
281 int retval = 0; 281 int retval = 0;
282 struct task_struct *p; 282 struct task_struct *p;
283 283
284 do_each_pid_task(pgrp, PIDTYPE_PGID, p) { 284 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
285 if (!task_is_stopped(p)) 285 if (!task_is_stopped(p))
286 continue; 286 continue;
287 retval = 1; 287 retval = 1;
288 break; 288 break;
289 } while_each_pid_task(pgrp, PIDTYPE_PGID, p); 289 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
290 return retval; 290 return retval;
291 } 291 }
292 292
293 /* 293 /*
294 * Check to see if any process groups have become orphaned as 294 * Check to see if any process groups have become orphaned as
295 * a result of our exiting, and if they have any stopped jobs, 295 * a result of our exiting, and if they have any stopped jobs,
296 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) 296 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
297 */ 297 */
298 static void 298 static void
299 kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) 299 kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
300 { 300 {
301 struct pid *pgrp = task_pgrp(tsk); 301 struct pid *pgrp = task_pgrp(tsk);
302 struct task_struct *ignored_task = tsk; 302 struct task_struct *ignored_task = tsk;
303 303
304 if (!parent) 304 if (!parent)
305 /* exit: our father is in a different pgrp than 305 /* exit: our father is in a different pgrp than
306 * we are and we were the only connection outside. 306 * we are and we were the only connection outside.
307 */ 307 */
308 parent = tsk->real_parent; 308 parent = tsk->real_parent;
309 else 309 else
310 /* reparent: our child is in a different pgrp than 310 /* reparent: our child is in a different pgrp than
311 * we are, and it was the only connection outside. 311 * we are, and it was the only connection outside.
312 */ 312 */
313 ignored_task = NULL; 313 ignored_task = NULL;
314 314
315 if (task_pgrp(parent) != pgrp && 315 if (task_pgrp(parent) != pgrp &&
316 task_session(parent) == task_session(tsk) && 316 task_session(parent) == task_session(tsk) &&
317 will_become_orphaned_pgrp(pgrp, ignored_task) && 317 will_become_orphaned_pgrp(pgrp, ignored_task) &&
318 has_stopped_jobs(pgrp)) { 318 has_stopped_jobs(pgrp)) {
319 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); 319 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
320 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); 320 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
321 } 321 }
322 } 322 }
323 323
324 /** 324 /**
325 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd 325 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
326 * 326 *
327 * If a kernel thread is launched as a result of a system call, or if 327 * If a kernel thread is launched as a result of a system call, or if
328 * it ever exits, it should generally reparent itself to kthreadd so it 328 * it ever exits, it should generally reparent itself to kthreadd so it
329 * isn't in the way of other processes and is correctly cleaned up on exit. 329 * isn't in the way of other processes and is correctly cleaned up on exit.
330 * 330 *
331 * The various task state such as scheduling policy and priority may have 331 * The various task state such as scheduling policy and priority may have
332 * been inherited from a user process, so we reset them to sane values here. 332 * been inherited from a user process, so we reset them to sane values here.
333 * 333 *
334 * NOTE that reparent_to_kthreadd() gives the caller full capabilities. 334 * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
335 */ 335 */
336 static void reparent_to_kthreadd(void) 336 static void reparent_to_kthreadd(void)
337 { 337 {
338 write_lock_irq(&tasklist_lock); 338 write_lock_irq(&tasklist_lock);
339 339
340 ptrace_unlink(current); 340 ptrace_unlink(current);
341 /* Reparent to init */ 341 /* Reparent to init */
342 current->real_parent = current->parent = kthreadd_task; 342 current->real_parent = current->parent = kthreadd_task;
343 list_move_tail(&current->sibling, &current->real_parent->children); 343 list_move_tail(&current->sibling, &current->real_parent->children);
344 344
345 /* Set the exit signal to SIGCHLD so we signal init on exit */ 345 /* Set the exit signal to SIGCHLD so we signal init on exit */
346 current->exit_signal = SIGCHLD; 346 current->exit_signal = SIGCHLD;
347 347
348 if (task_nice(current) < 0) 348 if (task_nice(current) < 0)
349 set_user_nice(current, 0); 349 set_user_nice(current, 0);
350 /* cpus_allowed? */ 350 /* cpus_allowed? */
351 /* rt_priority? */ 351 /* rt_priority? */
352 /* signals? */ 352 /* signals? */
353 memcpy(current->signal->rlim, init_task.signal->rlim, 353 memcpy(current->signal->rlim, init_task.signal->rlim,
354 sizeof(current->signal->rlim)); 354 sizeof(current->signal->rlim));
355 355
356 atomic_inc(&init_cred.usage); 356 atomic_inc(&init_cred.usage);
357 commit_creds(&init_cred); 357 commit_creds(&init_cred);
358 write_unlock_irq(&tasklist_lock); 358 write_unlock_irq(&tasklist_lock);
359 } 359 }
360 360
361 void __set_special_pids(struct pid *pid) 361 void __set_special_pids(struct pid *pid)
362 { 362 {
363 struct task_struct *curr = current->group_leader; 363 struct task_struct *curr = current->group_leader;
364 364
365 if (task_session(curr) != pid) 365 if (task_session(curr) != pid)
366 change_pid(curr, PIDTYPE_SID, pid); 366 change_pid(curr, PIDTYPE_SID, pid);
367 367
368 if (task_pgrp(curr) != pid) 368 if (task_pgrp(curr) != pid)
369 change_pid(curr, PIDTYPE_PGID, pid); 369 change_pid(curr, PIDTYPE_PGID, pid);
370 } 370 }
371 371
372 static void set_special_pids(struct pid *pid) 372 static void set_special_pids(struct pid *pid)
373 { 373 {
374 write_lock_irq(&tasklist_lock); 374 write_lock_irq(&tasklist_lock);
375 __set_special_pids(pid); 375 __set_special_pids(pid);
376 write_unlock_irq(&tasklist_lock); 376 write_unlock_irq(&tasklist_lock);
377 } 377 }
378 378
379 /* 379 /*
380 * Let kernel threads use this to say that they allow a certain signal. 380 * Let kernel threads use this to say that they allow a certain signal.
381 * Must not be used if kthread was cloned with CLONE_SIGHAND. 381 * Must not be used if kthread was cloned with CLONE_SIGHAND.
382 */ 382 */
383 int allow_signal(int sig) 383 int allow_signal(int sig)
384 { 384 {
385 if (!valid_signal(sig) || sig < 1) 385 if (!valid_signal(sig) || sig < 1)
386 return -EINVAL; 386 return -EINVAL;
387 387
388 spin_lock_irq(&current->sighand->siglock); 388 spin_lock_irq(&current->sighand->siglock);
389 /* This is only needed for daemonize()'ed kthreads */ 389 /* This is only needed for daemonize()'ed kthreads */
390 sigdelset(&current->blocked, sig); 390 sigdelset(&current->blocked, sig);
391 /* 391 /*
392 * Kernel threads handle their own signals. Let the signal code 392 * Kernel threads handle their own signals. Let the signal code
393 * know it'll be handled, so that they don't get converted to 393 * know it'll be handled, so that they don't get converted to
394 * SIGKILL or just silently dropped. 394 * SIGKILL or just silently dropped.
395 */ 395 */
396 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2; 396 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
397 recalc_sigpending(); 397 recalc_sigpending();
398 spin_unlock_irq(&current->sighand->siglock); 398 spin_unlock_irq(&current->sighand->siglock);
399 return 0; 399 return 0;
400 } 400 }
401 401
402 EXPORT_SYMBOL(allow_signal); 402 EXPORT_SYMBOL(allow_signal);
403 403
404 int disallow_signal(int sig) 404 int disallow_signal(int sig)
405 { 405 {
406 if (!valid_signal(sig) || sig < 1) 406 if (!valid_signal(sig) || sig < 1)
407 return -EINVAL; 407 return -EINVAL;
408 408
409 spin_lock_irq(&current->sighand->siglock); 409 spin_lock_irq(&current->sighand->siglock);
410 current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN; 410 current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN;
411 recalc_sigpending(); 411 recalc_sigpending();
412 spin_unlock_irq(&current->sighand->siglock); 412 spin_unlock_irq(&current->sighand->siglock);
413 return 0; 413 return 0;
414 } 414 }
415 415
416 EXPORT_SYMBOL(disallow_signal); 416 EXPORT_SYMBOL(disallow_signal);
417 417
418 /* 418 /*
419 * Put all the gunge required to become a kernel thread without 419 * Put all the gunge required to become a kernel thread without
420 * attached user resources in one place where it belongs. 420 * attached user resources in one place where it belongs.
421 */ 421 */
422 422
423 void daemonize(const char *name, ...) 423 void daemonize(const char *name, ...)
424 { 424 {
425 va_list args; 425 va_list args;
426 sigset_t blocked; 426 sigset_t blocked;
427 427
428 va_start(args, name); 428 va_start(args, name);
429 vsnprintf(current->comm, sizeof(current->comm), name, args); 429 vsnprintf(current->comm, sizeof(current->comm), name, args);
430 va_end(args); 430 va_end(args);
431 431
432 /* 432 /*
433 * If we were started as result of loading a module, close all of the 433 * If we were started as result of loading a module, close all of the
434 * user space pages. We don't need them, and if we didn't close them 434 * user space pages. We don't need them, and if we didn't close them
435 * they would be locked into memory. 435 * they would be locked into memory.
436 */ 436 */
437 exit_mm(current); 437 exit_mm(current);
438 /* 438 /*
439 * We don't want to have TIF_FREEZE set if the system-wide hibernation 439 * We don't want to have TIF_FREEZE set if the system-wide hibernation
440 * or suspend transition begins right now. 440 * or suspend transition begins right now.
441 */ 441 */
442 current->flags |= (PF_NOFREEZE | PF_KTHREAD); 442 current->flags |= (PF_NOFREEZE | PF_KTHREAD);
443 443
444 if (current->nsproxy != &init_nsproxy) { 444 if (current->nsproxy != &init_nsproxy) {
445 get_nsproxy(&init_nsproxy); 445 get_nsproxy(&init_nsproxy);
446 switch_task_namespaces(current, &init_nsproxy); 446 switch_task_namespaces(current, &init_nsproxy);
447 } 447 }
448 set_special_pids(&init_struct_pid); 448 set_special_pids(&init_struct_pid);
449 proc_clear_tty(current); 449 proc_clear_tty(current);
450 450
451 /* Block and flush all signals */ 451 /* Block and flush all signals */
452 sigfillset(&blocked); 452 sigfillset(&blocked);
453 sigprocmask(SIG_BLOCK, &blocked, NULL); 453 sigprocmask(SIG_BLOCK, &blocked, NULL);
454 flush_signals(current); 454 flush_signals(current);
455 455
456 /* Become as one with the init task */ 456 /* Become as one with the init task */
457 457
458 daemonize_fs_struct(); 458 daemonize_fs_struct();
459 exit_files(current); 459 exit_files(current);
460 current->files = init_task.files; 460 current->files = init_task.files;
461 atomic_inc(&current->files->count); 461 atomic_inc(&current->files->count);
462 462
463 reparent_to_kthreadd(); 463 reparent_to_kthreadd();
464 } 464 }
465 465
466 EXPORT_SYMBOL(daemonize); 466 EXPORT_SYMBOL(daemonize);
467 467
468 static void close_files(struct files_struct * files) 468 static void close_files(struct files_struct * files)
469 { 469 {
470 int i, j; 470 int i, j;
471 struct fdtable *fdt; 471 struct fdtable *fdt;
472 472
473 j = 0; 473 j = 0;
474 474
475 /* 475 /*
476 * It is safe to dereference the fd table without RCU or 476 * It is safe to dereference the fd table without RCU or
477 * ->file_lock because this is the last reference to the 477 * ->file_lock because this is the last reference to the
478 * files structure. But use RCU to shut RCU-lockdep up. 478 * files structure. But use RCU to shut RCU-lockdep up.
479 */ 479 */
480 rcu_read_lock(); 480 rcu_read_lock();
481 fdt = files_fdtable(files); 481 fdt = files_fdtable(files);
482 rcu_read_unlock(); 482 rcu_read_unlock();
483 for (;;) { 483 for (;;) {
484 unsigned long set; 484 unsigned long set;
485 i = j * __NFDBITS; 485 i = j * __NFDBITS;
486 if (i >= fdt->max_fds) 486 if (i >= fdt->max_fds)
487 break; 487 break;
488 set = fdt->open_fds->fds_bits[j++]; 488 set = fdt->open_fds->fds_bits[j++];
489 while (set) { 489 while (set) {
490 if (set & 1) { 490 if (set & 1) {
491 struct file * file = xchg(&fdt->fd[i], NULL); 491 struct file * file = xchg(&fdt->fd[i], NULL);
492 if (file) { 492 if (file) {
493 filp_close(file, files); 493 filp_close(file, files);
494 cond_resched(); 494 cond_resched();
495 } 495 }
496 } 496 }
497 i++; 497 i++;
498 set >>= 1; 498 set >>= 1;
499 } 499 }
500 } 500 }
501 } 501 }
502 502
503 struct files_struct *get_files_struct(struct task_struct *task) 503 struct files_struct *get_files_struct(struct task_struct *task)
504 { 504 {
505 struct files_struct *files; 505 struct files_struct *files;
506 506
507 task_lock(task); 507 task_lock(task);
508 files = task->files; 508 files = task->files;
509 if (files) 509 if (files)
510 atomic_inc(&files->count); 510 atomic_inc(&files->count);
511 task_unlock(task); 511 task_unlock(task);
512 512
513 return files; 513 return files;
514 } 514 }
515 515
516 void put_files_struct(struct files_struct *files) 516 void put_files_struct(struct files_struct *files)
517 { 517 {
518 struct fdtable *fdt; 518 struct fdtable *fdt;
519 519
520 if (atomic_dec_and_test(&files->count)) { 520 if (atomic_dec_and_test(&files->count)) {
521 close_files(files); 521 close_files(files);
522 /* 522 /*
523 * Free the fd and fdset arrays if we expanded them. 523 * Free the fd and fdset arrays if we expanded them.
524 * If the fdtable was embedded, pass files for freeing 524 * If the fdtable was embedded, pass files for freeing
525 * at the end of the RCU grace period. Otherwise, 525 * at the end of the RCU grace period. Otherwise,
526 * you can free files immediately. 526 * you can free files immediately.
527 */ 527 */
528 rcu_read_lock(); 528 rcu_read_lock();
529 fdt = files_fdtable(files); 529 fdt = files_fdtable(files);
530 if (fdt != &files->fdtab) 530 if (fdt != &files->fdtab)
531 kmem_cache_free(files_cachep, files); 531 kmem_cache_free(files_cachep, files);
532 free_fdtable(fdt); 532 free_fdtable(fdt);
533 rcu_read_unlock(); 533 rcu_read_unlock();
534 } 534 }
535 } 535 }
536 536
537 void reset_files_struct(struct files_struct *files) 537 void reset_files_struct(struct files_struct *files)
538 { 538 {
539 struct task_struct *tsk = current; 539 struct task_struct *tsk = current;
540 struct files_struct *old; 540 struct files_struct *old;
541 541
542 old = tsk->files; 542 old = tsk->files;
543 task_lock(tsk); 543 task_lock(tsk);
544 tsk->files = files; 544 tsk->files = files;
545 task_unlock(tsk); 545 task_unlock(tsk);
546 put_files_struct(old); 546 put_files_struct(old);
547 } 547 }
548 548
549 void exit_files(struct task_struct *tsk) 549 void exit_files(struct task_struct *tsk)
550 { 550 {
551 struct files_struct * files = tsk->files; 551 struct files_struct * files = tsk->files;
552 552
553 if (files) { 553 if (files) {
554 task_lock(tsk); 554 task_lock(tsk);
555 tsk->files = NULL; 555 tsk->files = NULL;
556 task_unlock(tsk); 556 task_unlock(tsk);
557 put_files_struct(files); 557 put_files_struct(files);
558 } 558 }
559 } 559 }
560 560
561 #ifdef CONFIG_MM_OWNER 561 #ifdef CONFIG_MM_OWNER
562 /* 562 /*
563 * Task p is exiting and it owned mm, lets find a new owner for it 563 * Task p is exiting and it owned mm, lets find a new owner for it
564 */ 564 */
565 static inline int 565 static inline int
566 mm_need_new_owner(struct mm_struct *mm, struct task_struct *p) 566 mm_need_new_owner(struct mm_struct *mm, struct task_struct *p)
567 { 567 {
568 /* 568 /*
569 * If there are other users of the mm and the owner (us) is exiting 569 * If there are other users of the mm and the owner (us) is exiting
570 * we need to find a new owner to take on the responsibility. 570 * we need to find a new owner to take on the responsibility.
571 */ 571 */
572 if (atomic_read(&mm->mm_users) <= 1) 572 if (atomic_read(&mm->mm_users) <= 1)
573 return 0; 573 return 0;
574 if (mm->owner != p) 574 if (mm->owner != p)
575 return 0; 575 return 0;
576 return 1; 576 return 1;
577 } 577 }
578 578
579 void mm_update_next_owner(struct mm_struct *mm) 579 void mm_update_next_owner(struct mm_struct *mm)
580 { 580 {
581 struct task_struct *c, *g, *p = current; 581 struct task_struct *c, *g, *p = current;
582 582
583 retry: 583 retry:
584 if (!mm_need_new_owner(mm, p)) 584 if (!mm_need_new_owner(mm, p))
585 return; 585 return;
586 586
587 read_lock(&tasklist_lock); 587 read_lock(&tasklist_lock);
588 /* 588 /*
589 * Search in the children 589 * Search in the children
590 */ 590 */
591 list_for_each_entry(c, &p->children, sibling) { 591 list_for_each_entry(c, &p->children, sibling) {
592 if (c->mm == mm) 592 if (c->mm == mm)
593 goto assign_new_owner; 593 goto assign_new_owner;
594 } 594 }
595 595
596 /* 596 /*
597 * Search in the siblings 597 * Search in the siblings
598 */ 598 */
599 list_for_each_entry(c, &p->real_parent->children, sibling) { 599 list_for_each_entry(c, &p->real_parent->children, sibling) {
600 if (c->mm == mm) 600 if (c->mm == mm)
601 goto assign_new_owner; 601 goto assign_new_owner;
602 } 602 }
603 603
604 /* 604 /*
605 * Search through everything else. We should not get 605 * Search through everything else. We should not get
606 * here often 606 * here often
607 */ 607 */
608 do_each_thread(g, c) { 608 do_each_thread(g, c) {
609 if (c->mm == mm) 609 if (c->mm == mm)
610 goto assign_new_owner; 610 goto assign_new_owner;
611 } while_each_thread(g, c); 611 } while_each_thread(g, c);
612 612
613 read_unlock(&tasklist_lock); 613 read_unlock(&tasklist_lock);
614 /* 614 /*
615 * We found no owner yet mm_users > 1: this implies that we are 615 * We found no owner yet mm_users > 1: this implies that we are
616 * most likely racing with swapoff (try_to_unuse()) or /proc or 616 * most likely racing with swapoff (try_to_unuse()) or /proc or
617 * ptrace or page migration (get_task_mm()). Mark owner as NULL. 617 * ptrace or page migration (get_task_mm()). Mark owner as NULL.
618 */ 618 */
619 mm->owner = NULL; 619 mm->owner = NULL;
620 return; 620 return;
621 621
622 assign_new_owner: 622 assign_new_owner:
623 BUG_ON(c == p); 623 BUG_ON(c == p);
624 get_task_struct(c); 624 get_task_struct(c);
625 /* 625 /*
626 * The task_lock protects c->mm from changing. 626 * The task_lock protects c->mm from changing.
627 * We always want mm->owner->mm == mm 627 * We always want mm->owner->mm == mm
628 */ 628 */
629 task_lock(c); 629 task_lock(c);
630 /* 630 /*
631 * Delay read_unlock() till we have the task_lock() 631 * Delay read_unlock() till we have the task_lock()
632 * to ensure that c does not slip away underneath us 632 * to ensure that c does not slip away underneath us
633 */ 633 */
634 read_unlock(&tasklist_lock); 634 read_unlock(&tasklist_lock);
635 if (c->mm != mm) { 635 if (c->mm != mm) {
636 task_unlock(c); 636 task_unlock(c);
637 put_task_struct(c); 637 put_task_struct(c);
638 goto retry; 638 goto retry;
639 } 639 }
640 mm->owner = c; 640 mm->owner = c;
641 task_unlock(c); 641 task_unlock(c);
642 put_task_struct(c); 642 put_task_struct(c);
643 } 643 }
644 #endif /* CONFIG_MM_OWNER */ 644 #endif /* CONFIG_MM_OWNER */
645 645
646 /* 646 /*
647 * Turn us into a lazy TLB process if we 647 * Turn us into a lazy TLB process if we
648 * aren't already.. 648 * aren't already..
649 */ 649 */
650 static void exit_mm(struct task_struct * tsk) 650 static void exit_mm(struct task_struct * tsk)
651 { 651 {
652 struct mm_struct *mm = tsk->mm; 652 struct mm_struct *mm = tsk->mm;
653 struct core_state *core_state; 653 struct core_state *core_state;
654 654
655 mm_release(tsk, mm); 655 mm_release(tsk, mm);
656 if (!mm) 656 if (!mm)
657 return; 657 return;
658 /* 658 /*
659 * Serialize with any possible pending coredump. 659 * Serialize with any possible pending coredump.
660 * We must hold mmap_sem around checking core_state 660 * We must hold mmap_sem around checking core_state
661 * and clearing tsk->mm. The core-inducing thread 661 * and clearing tsk->mm. The core-inducing thread
662 * will increment ->nr_threads for each thread in the 662 * will increment ->nr_threads for each thread in the
663 * group with ->mm != NULL. 663 * group with ->mm != NULL.
664 */ 664 */
665 down_read(&mm->mmap_sem); 665 down_read(&mm->mmap_sem);
666 core_state = mm->core_state; 666 core_state = mm->core_state;
667 if (core_state) { 667 if (core_state) {
668 struct core_thread self; 668 struct core_thread self;
669 up_read(&mm->mmap_sem); 669 up_read(&mm->mmap_sem);
670 670
671 self.task = tsk; 671 self.task = tsk;
672 self.next = xchg(&core_state->dumper.next, &self); 672 self.next = xchg(&core_state->dumper.next, &self);
673 /* 673 /*
674 * Implies mb(), the result of xchg() must be visible 674 * Implies mb(), the result of xchg() must be visible
675 * to core_state->dumper. 675 * to core_state->dumper.
676 */ 676 */
677 if (atomic_dec_and_test(&core_state->nr_threads)) 677 if (atomic_dec_and_test(&core_state->nr_threads))
678 complete(&core_state->startup); 678 complete(&core_state->startup);
679 679
680 for (;;) { 680 for (;;) {
681 set_task_state(tsk, TASK_UNINTERRUPTIBLE); 681 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
682 if (!self.task) /* see coredump_finish() */ 682 if (!self.task) /* see coredump_finish() */
683 break; 683 break;
684 schedule(); 684 schedule();
685 } 685 }
686 __set_task_state(tsk, TASK_RUNNING); 686 __set_task_state(tsk, TASK_RUNNING);
687 down_read(&mm->mmap_sem); 687 down_read(&mm->mmap_sem);
688 } 688 }
689 atomic_inc(&mm->mm_count); 689 atomic_inc(&mm->mm_count);
690 BUG_ON(mm != tsk->active_mm); 690 BUG_ON(mm != tsk->active_mm);
691 /* more a memory barrier than a real lock */ 691 /* more a memory barrier than a real lock */
692 task_lock(tsk); 692 task_lock(tsk);
693 tsk->mm = NULL; 693 tsk->mm = NULL;
694 up_read(&mm->mmap_sem); 694 up_read(&mm->mmap_sem);
695 enter_lazy_tlb(mm, current); 695 enter_lazy_tlb(mm, current);
696 /* We don't want this task to be frozen prematurely */ 696 /* We don't want this task to be frozen prematurely */
697 clear_freeze_flag(tsk); 697 clear_freeze_flag(tsk);
698 if (tsk->signal->oom_score_adj == OOM_SCORE_ADJ_MIN) 698 if (tsk->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
699 atomic_dec(&mm->oom_disable_count); 699 atomic_dec(&mm->oom_disable_count);
700 task_unlock(tsk); 700 task_unlock(tsk);
701 mm_update_next_owner(mm); 701 mm_update_next_owner(mm);
702 mmput(mm); 702 mmput(mm);
703 } 703 }
704 704
705 /* 705 /*
706 * When we die, we re-parent all our children. 706 * When we die, we re-parent all our children.
707 * Try to give them to another thread in our thread 707 * Try to give them to another thread in our thread
708 * group, and if no such member exists, give it to 708 * group, and if no such member exists, give it to
709 * the child reaper process (ie "init") in our pid 709 * the child reaper process (ie "init") in our pid
710 * space. 710 * space.
711 */ 711 */
712 static struct task_struct *find_new_reaper(struct task_struct *father) 712 static struct task_struct *find_new_reaper(struct task_struct *father)
713 __releases(&tasklist_lock) 713 __releases(&tasklist_lock)
714 __acquires(&tasklist_lock) 714 __acquires(&tasklist_lock)
715 { 715 {
716 struct pid_namespace *pid_ns = task_active_pid_ns(father); 716 struct pid_namespace *pid_ns = task_active_pid_ns(father);
717 struct task_struct *thread; 717 struct task_struct *thread;
718 718
719 thread = father; 719 thread = father;
720 while_each_thread(father, thread) { 720 while_each_thread(father, thread) {
721 if (thread->flags & PF_EXITING) 721 if (thread->flags & PF_EXITING)
722 continue; 722 continue;
723 if (unlikely(pid_ns->child_reaper == father)) 723 if (unlikely(pid_ns->child_reaper == father))
724 pid_ns->child_reaper = thread; 724 pid_ns->child_reaper = thread;
725 return thread; 725 return thread;
726 } 726 }
727 727
728 if (unlikely(pid_ns->child_reaper == father)) { 728 if (unlikely(pid_ns->child_reaper == father)) {
729 write_unlock_irq(&tasklist_lock); 729 write_unlock_irq(&tasklist_lock);
730 if (unlikely(pid_ns == &init_pid_ns)) 730 if (unlikely(pid_ns == &init_pid_ns))
731 panic("Attempted to kill init!"); 731 panic("Attempted to kill init!");
732 732
733 zap_pid_ns_processes(pid_ns); 733 zap_pid_ns_processes(pid_ns);
734 write_lock_irq(&tasklist_lock); 734 write_lock_irq(&tasklist_lock);
735 /* 735 /*
736 * We can not clear ->child_reaper or leave it alone. 736 * We can not clear ->child_reaper or leave it alone.
737 * There may by stealth EXIT_DEAD tasks on ->children, 737 * There may by stealth EXIT_DEAD tasks on ->children,
738 * forget_original_parent() must move them somewhere. 738 * forget_original_parent() must move them somewhere.
739 */ 739 */
740 pid_ns->child_reaper = init_pid_ns.child_reaper; 740 pid_ns->child_reaper = init_pid_ns.child_reaper;
741 } 741 }
742 742
743 return pid_ns->child_reaper; 743 return pid_ns->child_reaper;
744 } 744 }
745 745
746 /* 746 /*
747 * Any that need to be release_task'd are put on the @dead list. 747 * Any that need to be release_task'd are put on the @dead list.
748 */ 748 */
749 static void reparent_leader(struct task_struct *father, struct task_struct *p, 749 static void reparent_leader(struct task_struct *father, struct task_struct *p,
750 struct list_head *dead) 750 struct list_head *dead)
751 { 751 {
752 list_move_tail(&p->sibling, &p->real_parent->children); 752 list_move_tail(&p->sibling, &p->real_parent->children);
753 753
754 if (task_detached(p)) 754 if (task_detached(p))
755 return; 755 return;
756 /* 756 /*
757 * If this is a threaded reparent there is no need to 757 * If this is a threaded reparent there is no need to
758 * notify anyone anything has happened. 758 * notify anyone anything has happened.
759 */ 759 */
760 if (same_thread_group(p->real_parent, father)) 760 if (same_thread_group(p->real_parent, father))
761 return; 761 return;
762 762
763 /* We don't want people slaying init. */ 763 /* We don't want people slaying init. */
764 p->exit_signal = SIGCHLD; 764 p->exit_signal = SIGCHLD;
765 765
766 /* If it has exited notify the new parent about this child's death. */ 766 /* If it has exited notify the new parent about this child's death. */
767 if (!p->ptrace && 767 if (!p->ptrace &&
768 p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { 768 p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) {
769 do_notify_parent(p, p->exit_signal); 769 do_notify_parent(p, p->exit_signal);
770 if (task_detached(p)) { 770 if (task_detached(p)) {
771 p->exit_state = EXIT_DEAD; 771 p->exit_state = EXIT_DEAD;
772 list_move_tail(&p->sibling, dead); 772 list_move_tail(&p->sibling, dead);
773 } 773 }
774 } 774 }
775 775
776 kill_orphaned_pgrp(p, father); 776 kill_orphaned_pgrp(p, father);
777 } 777 }
778 778
779 static void forget_original_parent(struct task_struct *father) 779 static void forget_original_parent(struct task_struct *father)
780 { 780 {
781 struct task_struct *p, *n, *reaper; 781 struct task_struct *p, *n, *reaper;
782 LIST_HEAD(dead_children); 782 LIST_HEAD(dead_children);
783 783
784 write_lock_irq(&tasklist_lock); 784 write_lock_irq(&tasklist_lock);
785 /* 785 /*
786 * Note that exit_ptrace() and find_new_reaper() might 786 * Note that exit_ptrace() and find_new_reaper() might
787 * drop tasklist_lock and reacquire it. 787 * drop tasklist_lock and reacquire it.
788 */ 788 */
789 exit_ptrace(father); 789 exit_ptrace(father);
790 reaper = find_new_reaper(father); 790 reaper = find_new_reaper(father);
791 791
792 list_for_each_entry_safe(p, n, &father->children, sibling) { 792 list_for_each_entry_safe(p, n, &father->children, sibling) {
793 struct task_struct *t = p; 793 struct task_struct *t = p;
794 do { 794 do {
795 t->real_parent = reaper; 795 t->real_parent = reaper;
796 if (t->parent == father) { 796 if (t->parent == father) {
797 BUG_ON(t->ptrace); 797 BUG_ON(t->ptrace);
798 t->parent = t->real_parent; 798 t->parent = t->real_parent;
799 } 799 }
800 if (t->pdeath_signal) 800 if (t->pdeath_signal)
801 group_send_sig_info(t->pdeath_signal, 801 group_send_sig_info(t->pdeath_signal,
802 SEND_SIG_NOINFO, t); 802 SEND_SIG_NOINFO, t);
803 } while_each_thread(p, t); 803 } while_each_thread(p, t);
804 reparent_leader(father, p, &dead_children); 804 reparent_leader(father, p, &dead_children);
805 } 805 }
806 write_unlock_irq(&tasklist_lock); 806 write_unlock_irq(&tasklist_lock);
807 807
808 BUG_ON(!list_empty(&father->children)); 808 BUG_ON(!list_empty(&father->children));
809 809
810 list_for_each_entry_safe(p, n, &dead_children, sibling) { 810 list_for_each_entry_safe(p, n, &dead_children, sibling) {
811 list_del_init(&p->sibling); 811 list_del_init(&p->sibling);
812 release_task(p); 812 release_task(p);
813 } 813 }
814 } 814 }
815 815
816 /* 816 /*
817 * Send signals to all our closest relatives so that they know 817 * Send signals to all our closest relatives so that they know
818 * to properly mourn us.. 818 * to properly mourn us..
819 */ 819 */
820 static void exit_notify(struct task_struct *tsk, int group_dead) 820 static void exit_notify(struct task_struct *tsk, int group_dead)
821 { 821 {
822 int signal;
823 bool autoreap; 822 bool autoreap;
824 void *cookie;
825 823
826 /* 824 /*
827 * This does two things: 825 * This does two things:
828 * 826 *
829 * A. Make init inherit all the child processes 827 * A. Make init inherit all the child processes
830 * B. Check to see if any process groups have become orphaned 828 * B. Check to see if any process groups have become orphaned
831 * as a result of our exiting, and if they have any stopped 829 * as a result of our exiting, and if they have any stopped
832 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) 830 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
833 */ 831 */
834 forget_original_parent(tsk); 832 forget_original_parent(tsk);
835 exit_task_namespaces(tsk); 833 exit_task_namespaces(tsk);
836 834
837 write_lock_irq(&tasklist_lock); 835 write_lock_irq(&tasklist_lock);
838 if (group_dead) 836 if (group_dead)
839 kill_orphaned_pgrp(tsk->group_leader, NULL); 837 kill_orphaned_pgrp(tsk->group_leader, NULL);
840 838
841 /* Let father know we died 839 /* Let father know we died
842 * 840 *
843 * Thread signals are configurable, but you aren't going to use 841 * Thread signals are configurable, but you aren't going to use
844 * that to send signals to arbitrary processes. 842 * that to send signals to arbitrary processes.
845 * That stops right now. 843 * That stops right now.
846 * 844 *
847 * If the parent exec id doesn't match the exec id we saved 845 * If the parent exec id doesn't match the exec id we saved
848 * when we started then we know the parent has changed security 846 * when we started then we know the parent has changed security
849 * domain. 847 * domain.
850 * 848 *
851 * If our self_exec id doesn't match our parent_exec_id then 849 * If our self_exec id doesn't match our parent_exec_id then
852 * we have changed execution domain as these two values started 850 * we have changed execution domain as these two values started
853 * the same after a fork. 851 * the same after a fork.
854 */ 852 */
855 if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) && 853 if (thread_group_leader(tsk) && tsk->exit_signal != SIGCHLD &&
856 (tsk->parent_exec_id != tsk->real_parent->self_exec_id || 854 (tsk->parent_exec_id != tsk->real_parent->self_exec_id ||
857 tsk->self_exec_id != tsk->parent_exec_id)) 855 tsk->self_exec_id != tsk->parent_exec_id))
858 tsk->exit_signal = SIGCHLD; 856 tsk->exit_signal = SIGCHLD;
859 857
860 signal = tracehook_notify_death(tsk, &cookie, group_dead); 858 if (unlikely(tsk->ptrace)) {
861 if (signal >= 0) 859 int sig = thread_group_leader(tsk) &&
862 autoreap = do_notify_parent(tsk, signal); 860 thread_group_empty(tsk) &&
863 else 861 !ptrace_reparented(tsk) ?
864 autoreap = (signal == DEATH_REAP); 862 tsk->exit_signal : SIGCHLD;
863 autoreap = do_notify_parent(tsk, sig);
864 } else if (thread_group_leader(tsk)) {
865 autoreap = thread_group_empty(tsk) &&
866 do_notify_parent(tsk, tsk->exit_signal);
867 } else {
868 autoreap = true;
869 }
865 870
866 tsk->exit_state = autoreap ? EXIT_DEAD : EXIT_ZOMBIE; 871 tsk->exit_state = autoreap ? EXIT_DEAD : EXIT_ZOMBIE;
867 872
868 /* mt-exec, de_thread() is waiting for group leader */ 873 /* mt-exec, de_thread() is waiting for group leader */
869 if (unlikely(tsk->signal->notify_count < 0)) 874 if (unlikely(tsk->signal->notify_count < 0))
870 wake_up_process(tsk->signal->group_exit_task); 875 wake_up_process(tsk->signal->group_exit_task);
871 write_unlock_irq(&tasklist_lock); 876 write_unlock_irq(&tasklist_lock);
872 877
873 /* If the process is dead, release it - nobody will wait for it */ 878 /* If the process is dead, release it - nobody will wait for it */
874 if (autoreap) 879 if (autoreap)
875 release_task(tsk); 880 release_task(tsk);
876 } 881 }
877 882
878 #ifdef CONFIG_DEBUG_STACK_USAGE 883 #ifdef CONFIG_DEBUG_STACK_USAGE
879 static void check_stack_usage(void) 884 static void check_stack_usage(void)
880 { 885 {
881 static DEFINE_SPINLOCK(low_water_lock); 886 static DEFINE_SPINLOCK(low_water_lock);
882 static int lowest_to_date = THREAD_SIZE; 887 static int lowest_to_date = THREAD_SIZE;
883 unsigned long free; 888 unsigned long free;
884 889
885 free = stack_not_used(current); 890 free = stack_not_used(current);
886 891
887 if (free >= lowest_to_date) 892 if (free >= lowest_to_date)
888 return; 893 return;
889 894
890 spin_lock(&low_water_lock); 895 spin_lock(&low_water_lock);
891 if (free < lowest_to_date) { 896 if (free < lowest_to_date) {
892 printk(KERN_WARNING "%s used greatest stack depth: %lu bytes " 897 printk(KERN_WARNING "%s used greatest stack depth: %lu bytes "
893 "left\n", 898 "left\n",
894 current->comm, free); 899 current->comm, free);
895 lowest_to_date = free; 900 lowest_to_date = free;
896 } 901 }
897 spin_unlock(&low_water_lock); 902 spin_unlock(&low_water_lock);
898 } 903 }
899 #else 904 #else
900 static inline void check_stack_usage(void) {} 905 static inline void check_stack_usage(void) {}
901 #endif 906 #endif
902 907
903 NORET_TYPE void do_exit(long code) 908 NORET_TYPE void do_exit(long code)
904 { 909 {
905 struct task_struct *tsk = current; 910 struct task_struct *tsk = current;
906 int group_dead; 911 int group_dead;
907 912
908 profile_task_exit(tsk); 913 profile_task_exit(tsk);
909 914
910 WARN_ON(atomic_read(&tsk->fs_excl)); 915 WARN_ON(atomic_read(&tsk->fs_excl));
911 WARN_ON(blk_needs_flush_plug(tsk)); 916 WARN_ON(blk_needs_flush_plug(tsk));
912 917
913 if (unlikely(in_interrupt())) 918 if (unlikely(in_interrupt()))
914 panic("Aiee, killing interrupt handler!"); 919 panic("Aiee, killing interrupt handler!");
915 if (unlikely(!tsk->pid)) 920 if (unlikely(!tsk->pid))
916 panic("Attempted to kill the idle task!"); 921 panic("Attempted to kill the idle task!");
917 922
918 /* 923 /*
919 * If do_exit is called because this processes oopsed, it's possible 924 * If do_exit is called because this processes oopsed, it's possible
920 * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before 925 * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before
921 * continuing. Amongst other possible reasons, this is to prevent 926 * continuing. Amongst other possible reasons, this is to prevent
922 * mm_release()->clear_child_tid() from writing to a user-controlled 927 * mm_release()->clear_child_tid() from writing to a user-controlled
923 * kernel address. 928 * kernel address.
924 */ 929 */
925 set_fs(USER_DS); 930 set_fs(USER_DS);
926 931
927 ptrace_event(PTRACE_EVENT_EXIT, code); 932 ptrace_event(PTRACE_EVENT_EXIT, code);
928 933
929 validate_creds_for_do_exit(tsk); 934 validate_creds_for_do_exit(tsk);
930 935
931 /* 936 /*
932 * We're taking recursive faults here in do_exit. Safest is to just 937 * We're taking recursive faults here in do_exit. Safest is to just
933 * leave this task alone and wait for reboot. 938 * leave this task alone and wait for reboot.
934 */ 939 */
935 if (unlikely(tsk->flags & PF_EXITING)) { 940 if (unlikely(tsk->flags & PF_EXITING)) {
936 printk(KERN_ALERT 941 printk(KERN_ALERT
937 "Fixing recursive fault but reboot is needed!\n"); 942 "Fixing recursive fault but reboot is needed!\n");
938 /* 943 /*
939 * We can do this unlocked here. The futex code uses 944 * We can do this unlocked here. The futex code uses
940 * this flag just to verify whether the pi state 945 * this flag just to verify whether the pi state
941 * cleanup has been done or not. In the worst case it 946 * cleanup has been done or not. In the worst case it
942 * loops once more. We pretend that the cleanup was 947 * loops once more. We pretend that the cleanup was
943 * done as there is no way to return. Either the 948 * done as there is no way to return. Either the
944 * OWNER_DIED bit is set by now or we push the blocked 949 * OWNER_DIED bit is set by now or we push the blocked
945 * task into the wait for ever nirwana as well. 950 * task into the wait for ever nirwana as well.
946 */ 951 */
947 tsk->flags |= PF_EXITPIDONE; 952 tsk->flags |= PF_EXITPIDONE;
948 set_current_state(TASK_UNINTERRUPTIBLE); 953 set_current_state(TASK_UNINTERRUPTIBLE);
949 schedule(); 954 schedule();
950 } 955 }
951 956
952 exit_irq_thread(); 957 exit_irq_thread();
953 958
954 exit_signals(tsk); /* sets PF_EXITING */ 959 exit_signals(tsk); /* sets PF_EXITING */
955 /* 960 /*
956 * tsk->flags are checked in the futex code to protect against 961 * tsk->flags are checked in the futex code to protect against
957 * an exiting task cleaning up the robust pi futexes. 962 * an exiting task cleaning up the robust pi futexes.
958 */ 963 */
959 smp_mb(); 964 smp_mb();
960 raw_spin_unlock_wait(&tsk->pi_lock); 965 raw_spin_unlock_wait(&tsk->pi_lock);
961 966
962 if (unlikely(in_atomic())) 967 if (unlikely(in_atomic()))
963 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n", 968 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
964 current->comm, task_pid_nr(current), 969 current->comm, task_pid_nr(current),
965 preempt_count()); 970 preempt_count());
966 971
967 acct_update_integrals(tsk); 972 acct_update_integrals(tsk);
968 /* sync mm's RSS info before statistics gathering */ 973 /* sync mm's RSS info before statistics gathering */
969 if (tsk->mm) 974 if (tsk->mm)
970 sync_mm_rss(tsk, tsk->mm); 975 sync_mm_rss(tsk, tsk->mm);
971 group_dead = atomic_dec_and_test(&tsk->signal->live); 976 group_dead = atomic_dec_and_test(&tsk->signal->live);
972 if (group_dead) { 977 if (group_dead) {
973 hrtimer_cancel(&tsk->signal->real_timer); 978 hrtimer_cancel(&tsk->signal->real_timer);
974 exit_itimers(tsk->signal); 979 exit_itimers(tsk->signal);
975 if (tsk->mm) 980 if (tsk->mm)
976 setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); 981 setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm);
977 } 982 }
978 acct_collect(code, group_dead); 983 acct_collect(code, group_dead);
979 if (group_dead) 984 if (group_dead)
980 tty_audit_exit(); 985 tty_audit_exit();
981 if (unlikely(tsk->audit_context)) 986 if (unlikely(tsk->audit_context))
982 audit_free(tsk); 987 audit_free(tsk);
983 988
984 tsk->exit_code = code; 989 tsk->exit_code = code;
985 taskstats_exit(tsk, group_dead); 990 taskstats_exit(tsk, group_dead);
986 991
987 exit_mm(tsk); 992 exit_mm(tsk);
988 993
989 if (group_dead) 994 if (group_dead)
990 acct_process(); 995 acct_process();
991 trace_sched_process_exit(tsk); 996 trace_sched_process_exit(tsk);
992 997
993 exit_sem(tsk); 998 exit_sem(tsk);
994 exit_files(tsk); 999 exit_files(tsk);
995 exit_fs(tsk); 1000 exit_fs(tsk);
996 check_stack_usage(); 1001 check_stack_usage();
997 exit_thread(); 1002 exit_thread();
998 1003
999 /* 1004 /*
1000 * Flush inherited counters to the parent - before the parent 1005 * Flush inherited counters to the parent - before the parent
1001 * gets woken up by child-exit notifications. 1006 * gets woken up by child-exit notifications.
1002 * 1007 *
1003 * because of cgroup mode, must be called before cgroup_exit() 1008 * because of cgroup mode, must be called before cgroup_exit()
1004 */ 1009 */
1005 perf_event_exit_task(tsk); 1010 perf_event_exit_task(tsk);
1006 1011
1007 cgroup_exit(tsk, 1); 1012 cgroup_exit(tsk, 1);
1008 1013
1009 if (group_dead) 1014 if (group_dead)
1010 disassociate_ctty(1); 1015 disassociate_ctty(1);
1011 1016
1012 module_put(task_thread_info(tsk)->exec_domain->module); 1017 module_put(task_thread_info(tsk)->exec_domain->module);
1013 1018
1014 proc_exit_connector(tsk); 1019 proc_exit_connector(tsk);
1015 1020
1016 /* 1021 /*
1017 * FIXME: do that only when needed, using sched_exit tracepoint 1022 * FIXME: do that only when needed, using sched_exit tracepoint
1018 */ 1023 */
1019 ptrace_put_breakpoints(tsk); 1024 ptrace_put_breakpoints(tsk);
1020 1025
1021 exit_notify(tsk, group_dead); 1026 exit_notify(tsk, group_dead);
1022 #ifdef CONFIG_NUMA 1027 #ifdef CONFIG_NUMA
1023 task_lock(tsk); 1028 task_lock(tsk);
1024 mpol_put(tsk->mempolicy); 1029 mpol_put(tsk->mempolicy);
1025 tsk->mempolicy = NULL; 1030 tsk->mempolicy = NULL;
1026 task_unlock(tsk); 1031 task_unlock(tsk);
1027 #endif 1032 #endif
1028 #ifdef CONFIG_FUTEX 1033 #ifdef CONFIG_FUTEX
1029 if (unlikely(current->pi_state_cache)) 1034 if (unlikely(current->pi_state_cache))
1030 kfree(current->pi_state_cache); 1035 kfree(current->pi_state_cache);
1031 #endif 1036 #endif
1032 /* 1037 /*
1033 * Make sure we are holding no locks: 1038 * Make sure we are holding no locks:
1034 */ 1039 */
1035 debug_check_no_locks_held(tsk); 1040 debug_check_no_locks_held(tsk);
1036 /* 1041 /*
1037 * We can do this unlocked here. The futex code uses this flag 1042 * We can do this unlocked here. The futex code uses this flag
1038 * just to verify whether the pi state cleanup has been done 1043 * just to verify whether the pi state cleanup has been done
1039 * or not. In the worst case it loops once more. 1044 * or not. In the worst case it loops once more.
1040 */ 1045 */
1041 tsk->flags |= PF_EXITPIDONE; 1046 tsk->flags |= PF_EXITPIDONE;
1042 1047
1043 if (tsk->io_context) 1048 if (tsk->io_context)
1044 exit_io_context(tsk); 1049 exit_io_context(tsk);
1045 1050
1046 if (tsk->splice_pipe) 1051 if (tsk->splice_pipe)
1047 __free_pipe_info(tsk->splice_pipe); 1052 __free_pipe_info(tsk->splice_pipe);
1048 1053
1049 validate_creds_for_do_exit(tsk); 1054 validate_creds_for_do_exit(tsk);
1050 1055
1051 preempt_disable(); 1056 preempt_disable();
1052 exit_rcu(); 1057 exit_rcu();
1053 /* causes final put_task_struct in finish_task_switch(). */ 1058 /* causes final put_task_struct in finish_task_switch(). */
1054 tsk->state = TASK_DEAD; 1059 tsk->state = TASK_DEAD;
1055 schedule(); 1060 schedule();
1056 BUG(); 1061 BUG();
1057 /* Avoid "noreturn function does return". */ 1062 /* Avoid "noreturn function does return". */
1058 for (;;) 1063 for (;;)
1059 cpu_relax(); /* For when BUG is null */ 1064 cpu_relax(); /* For when BUG is null */
1060 } 1065 }
1061 1066
1062 EXPORT_SYMBOL_GPL(do_exit); 1067 EXPORT_SYMBOL_GPL(do_exit);
1063 1068
1064 NORET_TYPE void complete_and_exit(struct completion *comp, long code) 1069 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
1065 { 1070 {
1066 if (comp) 1071 if (comp)
1067 complete(comp); 1072 complete(comp);
1068 1073
1069 do_exit(code); 1074 do_exit(code);
1070 } 1075 }
1071 1076
1072 EXPORT_SYMBOL(complete_and_exit); 1077 EXPORT_SYMBOL(complete_and_exit);
1073 1078
1074 SYSCALL_DEFINE1(exit, int, error_code) 1079 SYSCALL_DEFINE1(exit, int, error_code)
1075 { 1080 {
1076 do_exit((error_code&0xff)<<8); 1081 do_exit((error_code&0xff)<<8);
1077 } 1082 }
1078 1083
1079 /* 1084 /*
1080 * Take down every thread in the group. This is called by fatal signals 1085 * Take down every thread in the group. This is called by fatal signals
1081 * as well as by sys_exit_group (below). 1086 * as well as by sys_exit_group (below).
1082 */ 1087 */
1083 NORET_TYPE void 1088 NORET_TYPE void
1084 do_group_exit(int exit_code) 1089 do_group_exit(int exit_code)
1085 { 1090 {
1086 struct signal_struct *sig = current->signal; 1091 struct signal_struct *sig = current->signal;
1087 1092
1088 BUG_ON(exit_code & 0x80); /* core dumps don't get here */ 1093 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
1089 1094
1090 if (signal_group_exit(sig)) 1095 if (signal_group_exit(sig))
1091 exit_code = sig->group_exit_code; 1096 exit_code = sig->group_exit_code;
1092 else if (!thread_group_empty(current)) { 1097 else if (!thread_group_empty(current)) {
1093 struct sighand_struct *const sighand = current->sighand; 1098 struct sighand_struct *const sighand = current->sighand;
1094 spin_lock_irq(&sighand->siglock); 1099 spin_lock_irq(&sighand->siglock);
1095 if (signal_group_exit(sig)) 1100 if (signal_group_exit(sig))
1096 /* Another thread got here before we took the lock. */ 1101 /* Another thread got here before we took the lock. */
1097 exit_code = sig->group_exit_code; 1102 exit_code = sig->group_exit_code;
1098 else { 1103 else {
1099 sig->group_exit_code = exit_code; 1104 sig->group_exit_code = exit_code;
1100 sig->flags = SIGNAL_GROUP_EXIT; 1105 sig->flags = SIGNAL_GROUP_EXIT;
1101 zap_other_threads(current); 1106 zap_other_threads(current);
1102 } 1107 }
1103 spin_unlock_irq(&sighand->siglock); 1108 spin_unlock_irq(&sighand->siglock);
1104 } 1109 }
1105 1110
1106 do_exit(exit_code); 1111 do_exit(exit_code);
1107 /* NOTREACHED */ 1112 /* NOTREACHED */
1108 } 1113 }
1109 1114
1110 /* 1115 /*
1111 * this kills every thread in the thread group. Note that any externally 1116 * this kills every thread in the thread group. Note that any externally
1112 * wait4()-ing process will get the correct exit code - even if this 1117 * wait4()-ing process will get the correct exit code - even if this
1113 * thread is not the thread group leader. 1118 * thread is not the thread group leader.
1114 */ 1119 */
1115 SYSCALL_DEFINE1(exit_group, int, error_code) 1120 SYSCALL_DEFINE1(exit_group, int, error_code)
1116 { 1121 {
1117 do_group_exit((error_code & 0xff) << 8); 1122 do_group_exit((error_code & 0xff) << 8);
1118 /* NOTREACHED */ 1123 /* NOTREACHED */
1119 return 0; 1124 return 0;
1120 } 1125 }
1121 1126
1122 struct wait_opts { 1127 struct wait_opts {
1123 enum pid_type wo_type; 1128 enum pid_type wo_type;
1124 int wo_flags; 1129 int wo_flags;
1125 struct pid *wo_pid; 1130 struct pid *wo_pid;
1126 1131
1127 struct siginfo __user *wo_info; 1132 struct siginfo __user *wo_info;
1128 int __user *wo_stat; 1133 int __user *wo_stat;
1129 struct rusage __user *wo_rusage; 1134 struct rusage __user *wo_rusage;
1130 1135
1131 wait_queue_t child_wait; 1136 wait_queue_t child_wait;
1132 int notask_error; 1137 int notask_error;
1133 }; 1138 };
1134 1139
1135 static inline 1140 static inline
1136 struct pid *task_pid_type(struct task_struct *task, enum pid_type type) 1141 struct pid *task_pid_type(struct task_struct *task, enum pid_type type)
1137 { 1142 {
1138 if (type != PIDTYPE_PID) 1143 if (type != PIDTYPE_PID)
1139 task = task->group_leader; 1144 task = task->group_leader;
1140 return task->pids[type].pid; 1145 return task->pids[type].pid;
1141 } 1146 }
1142 1147
1143 static int eligible_pid(struct wait_opts *wo, struct task_struct *p) 1148 static int eligible_pid(struct wait_opts *wo, struct task_struct *p)
1144 { 1149 {
1145 return wo->wo_type == PIDTYPE_MAX || 1150 return wo->wo_type == PIDTYPE_MAX ||
1146 task_pid_type(p, wo->wo_type) == wo->wo_pid; 1151 task_pid_type(p, wo->wo_type) == wo->wo_pid;
1147 } 1152 }
1148 1153
1149 static int eligible_child(struct wait_opts *wo, struct task_struct *p) 1154 static int eligible_child(struct wait_opts *wo, struct task_struct *p)
1150 { 1155 {
1151 if (!eligible_pid(wo, p)) 1156 if (!eligible_pid(wo, p))
1152 return 0; 1157 return 0;
1153 /* Wait for all children (clone and not) if __WALL is set; 1158 /* Wait for all children (clone and not) if __WALL is set;
1154 * otherwise, wait for clone children *only* if __WCLONE is 1159 * otherwise, wait for clone children *only* if __WCLONE is
1155 * set; otherwise, wait for non-clone children *only*. (Note: 1160 * set; otherwise, wait for non-clone children *only*. (Note:
1156 * A "clone" child here is one that reports to its parent 1161 * A "clone" child here is one that reports to its parent
1157 * using a signal other than SIGCHLD.) */ 1162 * using a signal other than SIGCHLD.) */
1158 if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) 1163 if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE))
1159 && !(wo->wo_flags & __WALL)) 1164 && !(wo->wo_flags & __WALL))
1160 return 0; 1165 return 0;
1161 1166
1162 return 1; 1167 return 1;
1163 } 1168 }
1164 1169
1165 static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p, 1170 static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p,
1166 pid_t pid, uid_t uid, int why, int status) 1171 pid_t pid, uid_t uid, int why, int status)
1167 { 1172 {
1168 struct siginfo __user *infop; 1173 struct siginfo __user *infop;
1169 int retval = wo->wo_rusage 1174 int retval = wo->wo_rusage
1170 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; 1175 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1171 1176
1172 put_task_struct(p); 1177 put_task_struct(p);
1173 infop = wo->wo_info; 1178 infop = wo->wo_info;
1174 if (infop) { 1179 if (infop) {
1175 if (!retval) 1180 if (!retval)
1176 retval = put_user(SIGCHLD, &infop->si_signo); 1181 retval = put_user(SIGCHLD, &infop->si_signo);
1177 if (!retval) 1182 if (!retval)
1178 retval = put_user(0, &infop->si_errno); 1183 retval = put_user(0, &infop->si_errno);
1179 if (!retval) 1184 if (!retval)
1180 retval = put_user((short)why, &infop->si_code); 1185 retval = put_user((short)why, &infop->si_code);
1181 if (!retval) 1186 if (!retval)
1182 retval = put_user(pid, &infop->si_pid); 1187 retval = put_user(pid, &infop->si_pid);
1183 if (!retval) 1188 if (!retval)
1184 retval = put_user(uid, &infop->si_uid); 1189 retval = put_user(uid, &infop->si_uid);
1185 if (!retval) 1190 if (!retval)
1186 retval = put_user(status, &infop->si_status); 1191 retval = put_user(status, &infop->si_status);
1187 } 1192 }
1188 if (!retval) 1193 if (!retval)
1189 retval = pid; 1194 retval = pid;
1190 return retval; 1195 return retval;
1191 } 1196 }
1192 1197
1193 /* 1198 /*
1194 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold 1199 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1195 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold 1200 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1196 * the lock and this task is uninteresting. If we return nonzero, we have 1201 * the lock and this task is uninteresting. If we return nonzero, we have
1197 * released the lock and the system call should return. 1202 * released the lock and the system call should return.
1198 */ 1203 */
1199 static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) 1204 static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p)
1200 { 1205 {
1201 unsigned long state; 1206 unsigned long state;
1202 int retval, status, traced; 1207 int retval, status, traced;
1203 pid_t pid = task_pid_vnr(p); 1208 pid_t pid = task_pid_vnr(p);
1204 uid_t uid = __task_cred(p)->uid; 1209 uid_t uid = __task_cred(p)->uid;
1205 struct siginfo __user *infop; 1210 struct siginfo __user *infop;
1206 1211
1207 if (!likely(wo->wo_flags & WEXITED)) 1212 if (!likely(wo->wo_flags & WEXITED))
1208 return 0; 1213 return 0;
1209 1214
1210 if (unlikely(wo->wo_flags & WNOWAIT)) { 1215 if (unlikely(wo->wo_flags & WNOWAIT)) {
1211 int exit_code = p->exit_code; 1216 int exit_code = p->exit_code;
1212 int why; 1217 int why;
1213 1218
1214 get_task_struct(p); 1219 get_task_struct(p);
1215 read_unlock(&tasklist_lock); 1220 read_unlock(&tasklist_lock);
1216 if ((exit_code & 0x7f) == 0) { 1221 if ((exit_code & 0x7f) == 0) {
1217 why = CLD_EXITED; 1222 why = CLD_EXITED;
1218 status = exit_code >> 8; 1223 status = exit_code >> 8;
1219 } else { 1224 } else {
1220 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; 1225 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1221 status = exit_code & 0x7f; 1226 status = exit_code & 0x7f;
1222 } 1227 }
1223 return wait_noreap_copyout(wo, p, pid, uid, why, status); 1228 return wait_noreap_copyout(wo, p, pid, uid, why, status);
1224 } 1229 }
1225 1230
1226 /* 1231 /*
1227 * Try to move the task's state to DEAD 1232 * Try to move the task's state to DEAD
1228 * only one thread is allowed to do this: 1233 * only one thread is allowed to do this:
1229 */ 1234 */
1230 state = xchg(&p->exit_state, EXIT_DEAD); 1235 state = xchg(&p->exit_state, EXIT_DEAD);
1231 if (state != EXIT_ZOMBIE) { 1236 if (state != EXIT_ZOMBIE) {
1232 BUG_ON(state != EXIT_DEAD); 1237 BUG_ON(state != EXIT_DEAD);
1233 return 0; 1238 return 0;
1234 } 1239 }
1235 1240
1236 traced = ptrace_reparented(p); 1241 traced = ptrace_reparented(p);
1237 /* 1242 /*
1238 * It can be ptraced but not reparented, check 1243 * It can be ptraced but not reparented, check
1239 * !task_detached() to filter out sub-threads. 1244 * !task_detached() to filter out sub-threads.
1240 */ 1245 */
1241 if (likely(!traced) && likely(!task_detached(p))) { 1246 if (likely(!traced) && likely(!task_detached(p))) {
1242 struct signal_struct *psig; 1247 struct signal_struct *psig;
1243 struct signal_struct *sig; 1248 struct signal_struct *sig;
1244 unsigned long maxrss; 1249 unsigned long maxrss;
1245 cputime_t tgutime, tgstime; 1250 cputime_t tgutime, tgstime;
1246 1251
1247 /* 1252 /*
1248 * The resource counters for the group leader are in its 1253 * The resource counters for the group leader are in its
1249 * own task_struct. Those for dead threads in the group 1254 * own task_struct. Those for dead threads in the group
1250 * are in its signal_struct, as are those for the child 1255 * are in its signal_struct, as are those for the child
1251 * processes it has previously reaped. All these 1256 * processes it has previously reaped. All these
1252 * accumulate in the parent's signal_struct c* fields. 1257 * accumulate in the parent's signal_struct c* fields.
1253 * 1258 *
1254 * We don't bother to take a lock here to protect these 1259 * We don't bother to take a lock here to protect these
1255 * p->signal fields, because they are only touched by 1260 * p->signal fields, because they are only touched by
1256 * __exit_signal, which runs with tasklist_lock 1261 * __exit_signal, which runs with tasklist_lock
1257 * write-locked anyway, and so is excluded here. We do 1262 * write-locked anyway, and so is excluded here. We do
1258 * need to protect the access to parent->signal fields, 1263 * need to protect the access to parent->signal fields,
1259 * as other threads in the parent group can be right 1264 * as other threads in the parent group can be right
1260 * here reaping other children at the same time. 1265 * here reaping other children at the same time.
1261 * 1266 *
1262 * We use thread_group_times() to get times for the thread 1267 * We use thread_group_times() to get times for the thread
1263 * group, which consolidates times for all threads in the 1268 * group, which consolidates times for all threads in the
1264 * group including the group leader. 1269 * group including the group leader.
1265 */ 1270 */
1266 thread_group_times(p, &tgutime, &tgstime); 1271 thread_group_times(p, &tgutime, &tgstime);
1267 spin_lock_irq(&p->real_parent->sighand->siglock); 1272 spin_lock_irq(&p->real_parent->sighand->siglock);
1268 psig = p->real_parent->signal; 1273 psig = p->real_parent->signal;
1269 sig = p->signal; 1274 sig = p->signal;
1270 psig->cutime = 1275 psig->cutime =
1271 cputime_add(psig->cutime, 1276 cputime_add(psig->cutime,
1272 cputime_add(tgutime, 1277 cputime_add(tgutime,
1273 sig->cutime)); 1278 sig->cutime));
1274 psig->cstime = 1279 psig->cstime =
1275 cputime_add(psig->cstime, 1280 cputime_add(psig->cstime,
1276 cputime_add(tgstime, 1281 cputime_add(tgstime,
1277 sig->cstime)); 1282 sig->cstime));
1278 psig->cgtime = 1283 psig->cgtime =
1279 cputime_add(psig->cgtime, 1284 cputime_add(psig->cgtime,
1280 cputime_add(p->gtime, 1285 cputime_add(p->gtime,
1281 cputime_add(sig->gtime, 1286 cputime_add(sig->gtime,
1282 sig->cgtime))); 1287 sig->cgtime)));
1283 psig->cmin_flt += 1288 psig->cmin_flt +=
1284 p->min_flt + sig->min_flt + sig->cmin_flt; 1289 p->min_flt + sig->min_flt + sig->cmin_flt;
1285 psig->cmaj_flt += 1290 psig->cmaj_flt +=
1286 p->maj_flt + sig->maj_flt + sig->cmaj_flt; 1291 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1287 psig->cnvcsw += 1292 psig->cnvcsw +=
1288 p->nvcsw + sig->nvcsw + sig->cnvcsw; 1293 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1289 psig->cnivcsw += 1294 psig->cnivcsw +=
1290 p->nivcsw + sig->nivcsw + sig->cnivcsw; 1295 p->nivcsw + sig->nivcsw + sig->cnivcsw;
1291 psig->cinblock += 1296 psig->cinblock +=
1292 task_io_get_inblock(p) + 1297 task_io_get_inblock(p) +
1293 sig->inblock + sig->cinblock; 1298 sig->inblock + sig->cinblock;
1294 psig->coublock += 1299 psig->coublock +=
1295 task_io_get_oublock(p) + 1300 task_io_get_oublock(p) +
1296 sig->oublock + sig->coublock; 1301 sig->oublock + sig->coublock;
1297 maxrss = max(sig->maxrss, sig->cmaxrss); 1302 maxrss = max(sig->maxrss, sig->cmaxrss);
1298 if (psig->cmaxrss < maxrss) 1303 if (psig->cmaxrss < maxrss)
1299 psig->cmaxrss = maxrss; 1304 psig->cmaxrss = maxrss;
1300 task_io_accounting_add(&psig->ioac, &p->ioac); 1305 task_io_accounting_add(&psig->ioac, &p->ioac);
1301 task_io_accounting_add(&psig->ioac, &sig->ioac); 1306 task_io_accounting_add(&psig->ioac, &sig->ioac);
1302 spin_unlock_irq(&p->real_parent->sighand->siglock); 1307 spin_unlock_irq(&p->real_parent->sighand->siglock);
1303 } 1308 }
1304 1309
1305 /* 1310 /*
1306 * Now we are sure this task is interesting, and no other 1311 * Now we are sure this task is interesting, and no other
1307 * thread can reap it because we set its state to EXIT_DEAD. 1312 * thread can reap it because we set its state to EXIT_DEAD.
1308 */ 1313 */
1309 read_unlock(&tasklist_lock); 1314 read_unlock(&tasklist_lock);
1310 1315
1311 retval = wo->wo_rusage 1316 retval = wo->wo_rusage
1312 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; 1317 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1313 status = (p->signal->flags & SIGNAL_GROUP_EXIT) 1318 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1314 ? p->signal->group_exit_code : p->exit_code; 1319 ? p->signal->group_exit_code : p->exit_code;
1315 if (!retval && wo->wo_stat) 1320 if (!retval && wo->wo_stat)
1316 retval = put_user(status, wo->wo_stat); 1321 retval = put_user(status, wo->wo_stat);
1317 1322
1318 infop = wo->wo_info; 1323 infop = wo->wo_info;
1319 if (!retval && infop) 1324 if (!retval && infop)
1320 retval = put_user(SIGCHLD, &infop->si_signo); 1325 retval = put_user(SIGCHLD, &infop->si_signo);
1321 if (!retval && infop) 1326 if (!retval && infop)
1322 retval = put_user(0, &infop->si_errno); 1327 retval = put_user(0, &infop->si_errno);
1323 if (!retval && infop) { 1328 if (!retval && infop) {
1324 int why; 1329 int why;
1325 1330
1326 if ((status & 0x7f) == 0) { 1331 if ((status & 0x7f) == 0) {
1327 why = CLD_EXITED; 1332 why = CLD_EXITED;
1328 status >>= 8; 1333 status >>= 8;
1329 } else { 1334 } else {
1330 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; 1335 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1331 status &= 0x7f; 1336 status &= 0x7f;
1332 } 1337 }
1333 retval = put_user((short)why, &infop->si_code); 1338 retval = put_user((short)why, &infop->si_code);
1334 if (!retval) 1339 if (!retval)
1335 retval = put_user(status, &infop->si_status); 1340 retval = put_user(status, &infop->si_status);
1336 } 1341 }
1337 if (!retval && infop) 1342 if (!retval && infop)
1338 retval = put_user(pid, &infop->si_pid); 1343 retval = put_user(pid, &infop->si_pid);
1339 if (!retval && infop) 1344 if (!retval && infop)
1340 retval = put_user(uid, &infop->si_uid); 1345 retval = put_user(uid, &infop->si_uid);
1341 if (!retval) 1346 if (!retval)
1342 retval = pid; 1347 retval = pid;
1343 1348
1344 if (traced) { 1349 if (traced) {
1345 write_lock_irq(&tasklist_lock); 1350 write_lock_irq(&tasklist_lock);
1346 /* We dropped tasklist, ptracer could die and untrace */ 1351 /* We dropped tasklist, ptracer could die and untrace */
1347 ptrace_unlink(p); 1352 ptrace_unlink(p);
1348 /* 1353 /*
1349 * If this is not a detached task, notify the parent. 1354 * If this is not a detached task, notify the parent.
1350 * If it's still not detached after that, don't release 1355 * If it's still not detached after that, don't release
1351 * it now. 1356 * it now.
1352 */ 1357 */
1353 if (!task_detached(p)) { 1358 if (!task_detached(p)) {
1354 do_notify_parent(p, p->exit_signal); 1359 do_notify_parent(p, p->exit_signal);
1355 if (!task_detached(p)) { 1360 if (!task_detached(p)) {
1356 p->exit_state = EXIT_ZOMBIE; 1361 p->exit_state = EXIT_ZOMBIE;
1357 p = NULL; 1362 p = NULL;
1358 } 1363 }
1359 } 1364 }
1360 write_unlock_irq(&tasklist_lock); 1365 write_unlock_irq(&tasklist_lock);
1361 } 1366 }
1362 if (p != NULL) 1367 if (p != NULL)
1363 release_task(p); 1368 release_task(p);
1364 1369
1365 return retval; 1370 return retval;
1366 } 1371 }
1367 1372
1368 static int *task_stopped_code(struct task_struct *p, bool ptrace) 1373 static int *task_stopped_code(struct task_struct *p, bool ptrace)
1369 { 1374 {
1370 if (ptrace) { 1375 if (ptrace) {
1371 if (task_is_stopped_or_traced(p) && 1376 if (task_is_stopped_or_traced(p) &&
1372 !(p->jobctl & JOBCTL_LISTENING)) 1377 !(p->jobctl & JOBCTL_LISTENING))
1373 return &p->exit_code; 1378 return &p->exit_code;
1374 } else { 1379 } else {
1375 if (p->signal->flags & SIGNAL_STOP_STOPPED) 1380 if (p->signal->flags & SIGNAL_STOP_STOPPED)
1376 return &p->signal->group_exit_code; 1381 return &p->signal->group_exit_code;
1377 } 1382 }
1378 return NULL; 1383 return NULL;
1379 } 1384 }
1380 1385
1381 /** 1386 /**
1382 * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED 1387 * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED
1383 * @wo: wait options 1388 * @wo: wait options
1384 * @ptrace: is the wait for ptrace 1389 * @ptrace: is the wait for ptrace
1385 * @p: task to wait for 1390 * @p: task to wait for
1386 * 1391 *
1387 * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED. 1392 * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED.
1388 * 1393 *
1389 * CONTEXT: 1394 * CONTEXT:
1390 * read_lock(&tasklist_lock), which is released if return value is 1395 * read_lock(&tasklist_lock), which is released if return value is
1391 * non-zero. Also, grabs and releases @p->sighand->siglock. 1396 * non-zero. Also, grabs and releases @p->sighand->siglock.
1392 * 1397 *
1393 * RETURNS: 1398 * RETURNS:
1394 * 0 if wait condition didn't exist and search for other wait conditions 1399 * 0 if wait condition didn't exist and search for other wait conditions
1395 * should continue. Non-zero return, -errno on failure and @p's pid on 1400 * should continue. Non-zero return, -errno on failure and @p's pid on
1396 * success, implies that tasklist_lock is released and wait condition 1401 * success, implies that tasklist_lock is released and wait condition
1397 * search should terminate. 1402 * search should terminate.
1398 */ 1403 */
1399 static int wait_task_stopped(struct wait_opts *wo, 1404 static int wait_task_stopped(struct wait_opts *wo,
1400 int ptrace, struct task_struct *p) 1405 int ptrace, struct task_struct *p)
1401 { 1406 {
1402 struct siginfo __user *infop; 1407 struct siginfo __user *infop;
1403 int retval, exit_code, *p_code, why; 1408 int retval, exit_code, *p_code, why;
1404 uid_t uid = 0; /* unneeded, required by compiler */ 1409 uid_t uid = 0; /* unneeded, required by compiler */
1405 pid_t pid; 1410 pid_t pid;
1406 1411
1407 /* 1412 /*
1408 * Traditionally we see ptrace'd stopped tasks regardless of options. 1413 * Traditionally we see ptrace'd stopped tasks regardless of options.
1409 */ 1414 */
1410 if (!ptrace && !(wo->wo_flags & WUNTRACED)) 1415 if (!ptrace && !(wo->wo_flags & WUNTRACED))
1411 return 0; 1416 return 0;
1412 1417
1413 if (!task_stopped_code(p, ptrace)) 1418 if (!task_stopped_code(p, ptrace))
1414 return 0; 1419 return 0;
1415 1420
1416 exit_code = 0; 1421 exit_code = 0;
1417 spin_lock_irq(&p->sighand->siglock); 1422 spin_lock_irq(&p->sighand->siglock);
1418 1423
1419 p_code = task_stopped_code(p, ptrace); 1424 p_code = task_stopped_code(p, ptrace);
1420 if (unlikely(!p_code)) 1425 if (unlikely(!p_code))
1421 goto unlock_sig; 1426 goto unlock_sig;
1422 1427
1423 exit_code = *p_code; 1428 exit_code = *p_code;
1424 if (!exit_code) 1429 if (!exit_code)
1425 goto unlock_sig; 1430 goto unlock_sig;
1426 1431
1427 if (!unlikely(wo->wo_flags & WNOWAIT)) 1432 if (!unlikely(wo->wo_flags & WNOWAIT))
1428 *p_code = 0; 1433 *p_code = 0;
1429 1434
1430 uid = task_uid(p); 1435 uid = task_uid(p);
1431 unlock_sig: 1436 unlock_sig:
1432 spin_unlock_irq(&p->sighand->siglock); 1437 spin_unlock_irq(&p->sighand->siglock);
1433 if (!exit_code) 1438 if (!exit_code)
1434 return 0; 1439 return 0;
1435 1440
1436 /* 1441 /*
1437 * Now we are pretty sure this task is interesting. 1442 * Now we are pretty sure this task is interesting.
1438 * Make sure it doesn't get reaped out from under us while we 1443 * Make sure it doesn't get reaped out from under us while we
1439 * give up the lock and then examine it below. We don't want to 1444 * give up the lock and then examine it below. We don't want to
1440 * keep holding onto the tasklist_lock while we call getrusage and 1445 * keep holding onto the tasklist_lock while we call getrusage and
1441 * possibly take page faults for user memory. 1446 * possibly take page faults for user memory.
1442 */ 1447 */
1443 get_task_struct(p); 1448 get_task_struct(p);
1444 pid = task_pid_vnr(p); 1449 pid = task_pid_vnr(p);
1445 why = ptrace ? CLD_TRAPPED : CLD_STOPPED; 1450 why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
1446 read_unlock(&tasklist_lock); 1451 read_unlock(&tasklist_lock);
1447 1452
1448 if (unlikely(wo->wo_flags & WNOWAIT)) 1453 if (unlikely(wo->wo_flags & WNOWAIT))
1449 return wait_noreap_copyout(wo, p, pid, uid, why, exit_code); 1454 return wait_noreap_copyout(wo, p, pid, uid, why, exit_code);
1450 1455
1451 retval = wo->wo_rusage 1456 retval = wo->wo_rusage
1452 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; 1457 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1453 if (!retval && wo->wo_stat) 1458 if (!retval && wo->wo_stat)
1454 retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat); 1459 retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat);
1455 1460
1456 infop = wo->wo_info; 1461 infop = wo->wo_info;
1457 if (!retval && infop) 1462 if (!retval && infop)
1458 retval = put_user(SIGCHLD, &infop->si_signo); 1463 retval = put_user(SIGCHLD, &infop->si_signo);
1459 if (!retval && infop) 1464 if (!retval && infop)
1460 retval = put_user(0, &infop->si_errno); 1465 retval = put_user(0, &infop->si_errno);
1461 if (!retval && infop) 1466 if (!retval && infop)
1462 retval = put_user((short)why, &infop->si_code); 1467 retval = put_user((short)why, &infop->si_code);
1463 if (!retval && infop) 1468 if (!retval && infop)
1464 retval = put_user(exit_code, &infop->si_status); 1469 retval = put_user(exit_code, &infop->si_status);
1465 if (!retval && infop) 1470 if (!retval && infop)
1466 retval = put_user(pid, &infop->si_pid); 1471 retval = put_user(pid, &infop->si_pid);
1467 if (!retval && infop) 1472 if (!retval && infop)
1468 retval = put_user(uid, &infop->si_uid); 1473 retval = put_user(uid, &infop->si_uid);
1469 if (!retval) 1474 if (!retval)
1470 retval = pid; 1475 retval = pid;
1471 put_task_struct(p); 1476 put_task_struct(p);
1472 1477
1473 BUG_ON(!retval); 1478 BUG_ON(!retval);
1474 return retval; 1479 return retval;
1475 } 1480 }
1476 1481
1477 /* 1482 /*
1478 * Handle do_wait work for one task in a live, non-stopped state. 1483 * Handle do_wait work for one task in a live, non-stopped state.
1479 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold 1484 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1480 * the lock and this task is uninteresting. If we return nonzero, we have 1485 * the lock and this task is uninteresting. If we return nonzero, we have
1481 * released the lock and the system call should return. 1486 * released the lock and the system call should return.
1482 */ 1487 */
1483 static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) 1488 static int wait_task_continued(struct wait_opts *wo, struct task_struct *p)
1484 { 1489 {
1485 int retval; 1490 int retval;
1486 pid_t pid; 1491 pid_t pid;
1487 uid_t uid; 1492 uid_t uid;
1488 1493
1489 if (!unlikely(wo->wo_flags & WCONTINUED)) 1494 if (!unlikely(wo->wo_flags & WCONTINUED))
1490 return 0; 1495 return 0;
1491 1496
1492 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) 1497 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1493 return 0; 1498 return 0;
1494 1499
1495 spin_lock_irq(&p->sighand->siglock); 1500 spin_lock_irq(&p->sighand->siglock);
1496 /* Re-check with the lock held. */ 1501 /* Re-check with the lock held. */
1497 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { 1502 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1498 spin_unlock_irq(&p->sighand->siglock); 1503 spin_unlock_irq(&p->sighand->siglock);
1499 return 0; 1504 return 0;
1500 } 1505 }
1501 if (!unlikely(wo->wo_flags & WNOWAIT)) 1506 if (!unlikely(wo->wo_flags & WNOWAIT))
1502 p->signal->flags &= ~SIGNAL_STOP_CONTINUED; 1507 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1503 uid = task_uid(p); 1508 uid = task_uid(p);
1504 spin_unlock_irq(&p->sighand->siglock); 1509 spin_unlock_irq(&p->sighand->siglock);
1505 1510
1506 pid = task_pid_vnr(p); 1511 pid = task_pid_vnr(p);
1507 get_task_struct(p); 1512 get_task_struct(p);
1508 read_unlock(&tasklist_lock); 1513 read_unlock(&tasklist_lock);
1509 1514
1510 if (!wo->wo_info) { 1515 if (!wo->wo_info) {
1511 retval = wo->wo_rusage 1516 retval = wo->wo_rusage
1512 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; 1517 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1513 put_task_struct(p); 1518 put_task_struct(p);
1514 if (!retval && wo->wo_stat) 1519 if (!retval && wo->wo_stat)
1515 retval = put_user(0xffff, wo->wo_stat); 1520 retval = put_user(0xffff, wo->wo_stat);
1516 if (!retval) 1521 if (!retval)
1517 retval = pid; 1522 retval = pid;
1518 } else { 1523 } else {
1519 retval = wait_noreap_copyout(wo, p, pid, uid, 1524 retval = wait_noreap_copyout(wo, p, pid, uid,
1520 CLD_CONTINUED, SIGCONT); 1525 CLD_CONTINUED, SIGCONT);
1521 BUG_ON(retval == 0); 1526 BUG_ON(retval == 0);
1522 } 1527 }
1523 1528
1524 return retval; 1529 return retval;
1525 } 1530 }
1526 1531
1527 /* 1532 /*
1528 * Consider @p for a wait by @parent. 1533 * Consider @p for a wait by @parent.
1529 * 1534 *
1530 * -ECHILD should be in ->notask_error before the first call. 1535 * -ECHILD should be in ->notask_error before the first call.
1531 * Returns nonzero for a final return, when we have unlocked tasklist_lock. 1536 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1532 * Returns zero if the search for a child should continue; 1537 * Returns zero if the search for a child should continue;
1533 * then ->notask_error is 0 if @p is an eligible child, 1538 * then ->notask_error is 0 if @p is an eligible child,
1534 * or another error from security_task_wait(), or still -ECHILD. 1539 * or another error from security_task_wait(), or still -ECHILD.
1535 */ 1540 */
1536 static int wait_consider_task(struct wait_opts *wo, int ptrace, 1541 static int wait_consider_task(struct wait_opts *wo, int ptrace,
1537 struct task_struct *p) 1542 struct task_struct *p)
1538 { 1543 {
1539 int ret = eligible_child(wo, p); 1544 int ret = eligible_child(wo, p);
1540 if (!ret) 1545 if (!ret)
1541 return ret; 1546 return ret;
1542 1547
1543 ret = security_task_wait(p); 1548 ret = security_task_wait(p);
1544 if (unlikely(ret < 0)) { 1549 if (unlikely(ret < 0)) {
1545 /* 1550 /*
1546 * If we have not yet seen any eligible child, 1551 * If we have not yet seen any eligible child,
1547 * then let this error code replace -ECHILD. 1552 * then let this error code replace -ECHILD.
1548 * A permission error will give the user a clue 1553 * A permission error will give the user a clue
1549 * to look for security policy problems, rather 1554 * to look for security policy problems, rather
1550 * than for mysterious wait bugs. 1555 * than for mysterious wait bugs.
1551 */ 1556 */
1552 if (wo->notask_error) 1557 if (wo->notask_error)
1553 wo->notask_error = ret; 1558 wo->notask_error = ret;
1554 return 0; 1559 return 0;
1555 } 1560 }
1556 1561
1557 /* dead body doesn't have much to contribute */ 1562 /* dead body doesn't have much to contribute */
1558 if (p->exit_state == EXIT_DEAD) 1563 if (p->exit_state == EXIT_DEAD)
1559 return 0; 1564 return 0;
1560 1565
1561 /* slay zombie? */ 1566 /* slay zombie? */
1562 if (p->exit_state == EXIT_ZOMBIE) { 1567 if (p->exit_state == EXIT_ZOMBIE) {
1563 /* 1568 /*
1564 * A zombie ptracee is only visible to its ptracer. 1569 * A zombie ptracee is only visible to its ptracer.
1565 * Notification and reaping will be cascaded to the real 1570 * Notification and reaping will be cascaded to the real
1566 * parent when the ptracer detaches. 1571 * parent when the ptracer detaches.
1567 */ 1572 */
1568 if (likely(!ptrace) && unlikely(p->ptrace)) { 1573 if (likely(!ptrace) && unlikely(p->ptrace)) {
1569 /* it will become visible, clear notask_error */ 1574 /* it will become visible, clear notask_error */
1570 wo->notask_error = 0; 1575 wo->notask_error = 0;
1571 return 0; 1576 return 0;
1572 } 1577 }
1573 1578
1574 /* we don't reap group leaders with subthreads */ 1579 /* we don't reap group leaders with subthreads */
1575 if (!delay_group_leader(p)) 1580 if (!delay_group_leader(p))
1576 return wait_task_zombie(wo, p); 1581 return wait_task_zombie(wo, p);
1577 1582
1578 /* 1583 /*
1579 * Allow access to stopped/continued state via zombie by 1584 * Allow access to stopped/continued state via zombie by
1580 * falling through. Clearing of notask_error is complex. 1585 * falling through. Clearing of notask_error is complex.
1581 * 1586 *
1582 * When !@ptrace: 1587 * When !@ptrace:
1583 * 1588 *
1584 * If WEXITED is set, notask_error should naturally be 1589 * If WEXITED is set, notask_error should naturally be
1585 * cleared. If not, subset of WSTOPPED|WCONTINUED is set, 1590 * cleared. If not, subset of WSTOPPED|WCONTINUED is set,
1586 * so, if there are live subthreads, there are events to 1591 * so, if there are live subthreads, there are events to
1587 * wait for. If all subthreads are dead, it's still safe 1592 * wait for. If all subthreads are dead, it's still safe
1588 * to clear - this function will be called again in finite 1593 * to clear - this function will be called again in finite
1589 * amount time once all the subthreads are released and 1594 * amount time once all the subthreads are released and
1590 * will then return without clearing. 1595 * will then return without clearing.
1591 * 1596 *
1592 * When @ptrace: 1597 * When @ptrace:
1593 * 1598 *
1594 * Stopped state is per-task and thus can't change once the 1599 * Stopped state is per-task and thus can't change once the
1595 * target task dies. Only continued and exited can happen. 1600 * target task dies. Only continued and exited can happen.
1596 * Clear notask_error if WCONTINUED | WEXITED. 1601 * Clear notask_error if WCONTINUED | WEXITED.
1597 */ 1602 */
1598 if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED))) 1603 if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED)))
1599 wo->notask_error = 0; 1604 wo->notask_error = 0;
1600 } else { 1605 } else {
1601 /* 1606 /*
1602 * If @p is ptraced by a task in its real parent's group, 1607 * If @p is ptraced by a task in its real parent's group,
1603 * hide group stop/continued state when looking at @p as 1608 * hide group stop/continued state when looking at @p as
1604 * the real parent; otherwise, a single stop can be 1609 * the real parent; otherwise, a single stop can be
1605 * reported twice as group and ptrace stops. 1610 * reported twice as group and ptrace stops.
1606 * 1611 *
1607 * If a ptracer wants to distinguish the two events for its 1612 * If a ptracer wants to distinguish the two events for its
1608 * own children, it should create a separate process which 1613 * own children, it should create a separate process which
1609 * takes the role of real parent. 1614 * takes the role of real parent.
1610 */ 1615 */
1611 if (likely(!ptrace) && p->ptrace && 1616 if (likely(!ptrace) && p->ptrace &&
1612 same_thread_group(p->parent, p->real_parent)) 1617 same_thread_group(p->parent, p->real_parent))
1613 return 0; 1618 return 0;
1614 1619
1615 /* 1620 /*
1616 * @p is alive and it's gonna stop, continue or exit, so 1621 * @p is alive and it's gonna stop, continue or exit, so
1617 * there always is something to wait for. 1622 * there always is something to wait for.
1618 */ 1623 */
1619 wo->notask_error = 0; 1624 wo->notask_error = 0;
1620 } 1625 }
1621 1626
1622 /* 1627 /*
1623 * Wait for stopped. Depending on @ptrace, different stopped state 1628 * Wait for stopped. Depending on @ptrace, different stopped state
1624 * is used and the two don't interact with each other. 1629 * is used and the two don't interact with each other.
1625 */ 1630 */
1626 ret = wait_task_stopped(wo, ptrace, p); 1631 ret = wait_task_stopped(wo, ptrace, p);
1627 if (ret) 1632 if (ret)
1628 return ret; 1633 return ret;
1629 1634
1630 /* 1635 /*
1631 * Wait for continued. There's only one continued state and the 1636 * Wait for continued. There's only one continued state and the
1632 * ptracer can consume it which can confuse the real parent. Don't 1637 * ptracer can consume it which can confuse the real parent. Don't
1633 * use WCONTINUED from ptracer. You don't need or want it. 1638 * use WCONTINUED from ptracer. You don't need or want it.
1634 */ 1639 */
1635 return wait_task_continued(wo, p); 1640 return wait_task_continued(wo, p);
1636 } 1641 }
1637 1642
1638 /* 1643 /*
1639 * Do the work of do_wait() for one thread in the group, @tsk. 1644 * Do the work of do_wait() for one thread in the group, @tsk.
1640 * 1645 *
1641 * -ECHILD should be in ->notask_error before the first call. 1646 * -ECHILD should be in ->notask_error before the first call.
1642 * Returns nonzero for a final return, when we have unlocked tasklist_lock. 1647 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1643 * Returns zero if the search for a child should continue; then 1648 * Returns zero if the search for a child should continue; then
1644 * ->notask_error is 0 if there were any eligible children, 1649 * ->notask_error is 0 if there were any eligible children,
1645 * or another error from security_task_wait(), or still -ECHILD. 1650 * or another error from security_task_wait(), or still -ECHILD.
1646 */ 1651 */
1647 static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) 1652 static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk)
1648 { 1653 {
1649 struct task_struct *p; 1654 struct task_struct *p;
1650 1655
1651 list_for_each_entry(p, &tsk->children, sibling) { 1656 list_for_each_entry(p, &tsk->children, sibling) {
1652 int ret = wait_consider_task(wo, 0, p); 1657 int ret = wait_consider_task(wo, 0, p);
1653 if (ret) 1658 if (ret)
1654 return ret; 1659 return ret;
1655 } 1660 }
1656 1661
1657 return 0; 1662 return 0;
1658 } 1663 }
1659 1664
1660 static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) 1665 static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk)
1661 { 1666 {
1662 struct task_struct *p; 1667 struct task_struct *p;
1663 1668
1664 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { 1669 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
1665 int ret = wait_consider_task(wo, 1, p); 1670 int ret = wait_consider_task(wo, 1, p);
1666 if (ret) 1671 if (ret)
1667 return ret; 1672 return ret;
1668 } 1673 }
1669 1674
1670 return 0; 1675 return 0;
1671 } 1676 }
1672 1677
1673 static int child_wait_callback(wait_queue_t *wait, unsigned mode, 1678 static int child_wait_callback(wait_queue_t *wait, unsigned mode,
1674 int sync, void *key) 1679 int sync, void *key)
1675 { 1680 {
1676 struct wait_opts *wo = container_of(wait, struct wait_opts, 1681 struct wait_opts *wo = container_of(wait, struct wait_opts,
1677 child_wait); 1682 child_wait);
1678 struct task_struct *p = key; 1683 struct task_struct *p = key;
1679 1684
1680 if (!eligible_pid(wo, p)) 1685 if (!eligible_pid(wo, p))
1681 return 0; 1686 return 0;
1682 1687
1683 if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) 1688 if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent)
1684 return 0; 1689 return 0;
1685 1690
1686 return default_wake_function(wait, mode, sync, key); 1691 return default_wake_function(wait, mode, sync, key);
1687 } 1692 }
1688 1693
1689 void __wake_up_parent(struct task_struct *p, struct task_struct *parent) 1694 void __wake_up_parent(struct task_struct *p, struct task_struct *parent)
1690 { 1695 {
1691 __wake_up_sync_key(&parent->signal->wait_chldexit, 1696 __wake_up_sync_key(&parent->signal->wait_chldexit,
1692 TASK_INTERRUPTIBLE, 1, p); 1697 TASK_INTERRUPTIBLE, 1, p);
1693 } 1698 }
1694 1699
1695 static long do_wait(struct wait_opts *wo) 1700 static long do_wait(struct wait_opts *wo)
1696 { 1701 {
1697 struct task_struct *tsk; 1702 struct task_struct *tsk;
1698 int retval; 1703 int retval;
1699 1704
1700 trace_sched_process_wait(wo->wo_pid); 1705 trace_sched_process_wait(wo->wo_pid);
1701 1706
1702 init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); 1707 init_waitqueue_func_entry(&wo->child_wait, child_wait_callback);
1703 wo->child_wait.private = current; 1708 wo->child_wait.private = current;
1704 add_wait_queue(&current->signal->wait_chldexit, &wo->child_wait); 1709 add_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
1705 repeat: 1710 repeat:
1706 /* 1711 /*
1707 * If there is nothing that can match our critiera just get out. 1712 * If there is nothing that can match our critiera just get out.
1708 * We will clear ->notask_error to zero if we see any child that 1713 * We will clear ->notask_error to zero if we see any child that
1709 * might later match our criteria, even if we are not able to reap 1714 * might later match our criteria, even if we are not able to reap
1710 * it yet. 1715 * it yet.
1711 */ 1716 */
1712 wo->notask_error = -ECHILD; 1717 wo->notask_error = -ECHILD;
1713 if ((wo->wo_type < PIDTYPE_MAX) && 1718 if ((wo->wo_type < PIDTYPE_MAX) &&
1714 (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type]))) 1719 (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type])))
1715 goto notask; 1720 goto notask;
1716 1721
1717 set_current_state(TASK_INTERRUPTIBLE); 1722 set_current_state(TASK_INTERRUPTIBLE);
1718 read_lock(&tasklist_lock); 1723 read_lock(&tasklist_lock);
1719 tsk = current; 1724 tsk = current;
1720 do { 1725 do {
1721 retval = do_wait_thread(wo, tsk); 1726 retval = do_wait_thread(wo, tsk);
1722 if (retval) 1727 if (retval)
1723 goto end; 1728 goto end;
1724 1729
1725 retval = ptrace_do_wait(wo, tsk); 1730 retval = ptrace_do_wait(wo, tsk);
1726 if (retval) 1731 if (retval)
1727 goto end; 1732 goto end;
1728 1733
1729 if (wo->wo_flags & __WNOTHREAD) 1734 if (wo->wo_flags & __WNOTHREAD)
1730 break; 1735 break;
1731 } while_each_thread(current, tsk); 1736 } while_each_thread(current, tsk);
1732 read_unlock(&tasklist_lock); 1737 read_unlock(&tasklist_lock);
1733 1738
1734 notask: 1739 notask:
1735 retval = wo->notask_error; 1740 retval = wo->notask_error;
1736 if (!retval && !(wo->wo_flags & WNOHANG)) { 1741 if (!retval && !(wo->wo_flags & WNOHANG)) {
1737 retval = -ERESTARTSYS; 1742 retval = -ERESTARTSYS;
1738 if (!signal_pending(current)) { 1743 if (!signal_pending(current)) {
1739 schedule(); 1744 schedule();
1740 goto repeat; 1745 goto repeat;
1741 } 1746 }
1742 } 1747 }
1743 end: 1748 end:
1744 __set_current_state(TASK_RUNNING); 1749 __set_current_state(TASK_RUNNING);
1745 remove_wait_queue(&current->signal->wait_chldexit, &wo->child_wait); 1750 remove_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
1746 return retval; 1751 return retval;
1747 } 1752 }
1748 1753
1749 SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, 1754 SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *,
1750 infop, int, options, struct rusage __user *, ru) 1755 infop, int, options, struct rusage __user *, ru)
1751 { 1756 {
1752 struct wait_opts wo; 1757 struct wait_opts wo;
1753 struct pid *pid = NULL; 1758 struct pid *pid = NULL;
1754 enum pid_type type; 1759 enum pid_type type;
1755 long ret; 1760 long ret;
1756 1761
1757 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED)) 1762 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1758 return -EINVAL; 1763 return -EINVAL;
1759 if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) 1764 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1760 return -EINVAL; 1765 return -EINVAL;
1761 1766
1762 switch (which) { 1767 switch (which) {
1763 case P_ALL: 1768 case P_ALL:
1764 type = PIDTYPE_MAX; 1769 type = PIDTYPE_MAX;
1765 break; 1770 break;
1766 case P_PID: 1771 case P_PID:
1767 type = PIDTYPE_PID; 1772 type = PIDTYPE_PID;
1768 if (upid <= 0) 1773 if (upid <= 0)
1769 return -EINVAL; 1774 return -EINVAL;
1770 break; 1775 break;
1771 case P_PGID: 1776 case P_PGID:
1772 type = PIDTYPE_PGID; 1777 type = PIDTYPE_PGID;
1773 if (upid <= 0) 1778 if (upid <= 0)
1774 return -EINVAL; 1779 return -EINVAL;
1775 break; 1780 break;
1776 default: 1781 default:
1777 return -EINVAL; 1782 return -EINVAL;
1778 } 1783 }
1779 1784
1780 if (type < PIDTYPE_MAX) 1785 if (type < PIDTYPE_MAX)
1781 pid = find_get_pid(upid); 1786 pid = find_get_pid(upid);
1782 1787
1783 wo.wo_type = type; 1788 wo.wo_type = type;
1784 wo.wo_pid = pid; 1789 wo.wo_pid = pid;
1785 wo.wo_flags = options; 1790 wo.wo_flags = options;
1786 wo.wo_info = infop; 1791 wo.wo_info = infop;
1787 wo.wo_stat = NULL; 1792 wo.wo_stat = NULL;
1788 wo.wo_rusage = ru; 1793 wo.wo_rusage = ru;
1789 ret = do_wait(&wo); 1794 ret = do_wait(&wo);
1790 1795
1791 if (ret > 0) { 1796 if (ret > 0) {
1792 ret = 0; 1797 ret = 0;
1793 } else if (infop) { 1798 } else if (infop) {
1794 /* 1799 /*
1795 * For a WNOHANG return, clear out all the fields 1800 * For a WNOHANG return, clear out all the fields
1796 * we would set so the user can easily tell the 1801 * we would set so the user can easily tell the
1797 * difference. 1802 * difference.
1798 */ 1803 */
1799 if (!ret) 1804 if (!ret)
1800 ret = put_user(0, &infop->si_signo); 1805 ret = put_user(0, &infop->si_signo);
1801 if (!ret) 1806 if (!ret)
1802 ret = put_user(0, &infop->si_errno); 1807 ret = put_user(0, &infop->si_errno);
1803 if (!ret) 1808 if (!ret)
1804 ret = put_user(0, &infop->si_code); 1809 ret = put_user(0, &infop->si_code);
1805 if (!ret) 1810 if (!ret)
1806 ret = put_user(0, &infop->si_pid); 1811 ret = put_user(0, &infop->si_pid);
1807 if (!ret) 1812 if (!ret)
1808 ret = put_user(0, &infop->si_uid); 1813 ret = put_user(0, &infop->si_uid);
1809 if (!ret) 1814 if (!ret)
1810 ret = put_user(0, &infop->si_status); 1815 ret = put_user(0, &infop->si_status);
1811 } 1816 }
1812 1817
1813 put_pid(pid); 1818 put_pid(pid);
1814 1819
1815 /* avoid REGPARM breakage on x86: */ 1820 /* avoid REGPARM breakage on x86: */
1816 asmlinkage_protect(5, ret, which, upid, infop, options, ru); 1821 asmlinkage_protect(5, ret, which, upid, infop, options, ru);
1817 return ret; 1822 return ret;
1818 } 1823 }
1819 1824
1820 SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, 1825 SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr,
1821 int, options, struct rusage __user *, ru) 1826 int, options, struct rusage __user *, ru)
1822 { 1827 {
1823 struct wait_opts wo; 1828 struct wait_opts wo;
1824 struct pid *pid = NULL; 1829 struct pid *pid = NULL;
1825 enum pid_type type; 1830 enum pid_type type;
1826 long ret; 1831 long ret;
1827 1832
1828 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| 1833 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1829 __WNOTHREAD|__WCLONE|__WALL)) 1834 __WNOTHREAD|__WCLONE|__WALL))
1830 return -EINVAL; 1835 return -EINVAL;
1831 1836
1832 if (upid == -1) 1837 if (upid == -1)
1833 type = PIDTYPE_MAX; 1838 type = PIDTYPE_MAX;
1834 else if (upid < 0) { 1839 else if (upid < 0) {
1835 type = PIDTYPE_PGID; 1840 type = PIDTYPE_PGID;
1836 pid = find_get_pid(-upid); 1841 pid = find_get_pid(-upid);
1837 } else if (upid == 0) { 1842 } else if (upid == 0) {
1838 type = PIDTYPE_PGID; 1843 type = PIDTYPE_PGID;
1839 pid = get_task_pid(current, PIDTYPE_PGID); 1844 pid = get_task_pid(current, PIDTYPE_PGID);
1840 } else /* upid > 0 */ { 1845 } else /* upid > 0 */ {
1841 type = PIDTYPE_PID; 1846 type = PIDTYPE_PID;
1842 pid = find_get_pid(upid); 1847 pid = find_get_pid(upid);
1843 } 1848 }
1844 1849
1845 wo.wo_type = type; 1850 wo.wo_type = type;
1846 wo.wo_pid = pid; 1851 wo.wo_pid = pid;
1847 wo.wo_flags = options | WEXITED; 1852 wo.wo_flags = options | WEXITED;
1848 wo.wo_info = NULL; 1853 wo.wo_info = NULL;
1849 wo.wo_stat = stat_addr; 1854 wo.wo_stat = stat_addr;
1850 wo.wo_rusage = ru; 1855 wo.wo_rusage = ru;
1851 ret = do_wait(&wo); 1856 ret = do_wait(&wo);
1852 put_pid(pid); 1857 put_pid(pid);
1853 1858
1854 /* avoid REGPARM breakage on x86: */ 1859 /* avoid REGPARM breakage on x86: */
1855 asmlinkage_protect(4, ret, upid, stat_addr, options, ru); 1860 asmlinkage_protect(4, ret, upid, stat_addr, options, ru);
1856 return ret; 1861 return ret;
1857 } 1862 }
1858 1863
1859 #ifdef __ARCH_WANT_SYS_WAITPID 1864 #ifdef __ARCH_WANT_SYS_WAITPID
1860 1865
1861 /* 1866 /*
1862 * sys_waitpid() remains for compatibility. waitpid() should be 1867 * sys_waitpid() remains for compatibility. waitpid() should be
1863 * implemented by calling sys_wait4() from libc.a. 1868 * implemented by calling sys_wait4() from libc.a.
1864 */ 1869 */
1865 SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) 1870 SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options)
1866 { 1871 {
1867 return sys_wait4(pid, stat_addr, options, NULL); 1872 return sys_wait4(pid, stat_addr, options, NULL);
1868 } 1873 }
1869 1874