diff options
Diffstat (limited to 'kernel')
-rw-r--r-- | kernel/exit.c | 129 | ||||
-rw-r--r-- | kernel/fork.c | 121 | ||||
-rw-r--r-- | kernel/kmod.c | 2 | ||||
-rw-r--r-- | kernel/pid.c | 40 | ||||
-rw-r--r-- | kernel/ptrace.c | 8 | ||||
-rw-r--r-- | kernel/signal.c | 344 | ||||
-rw-r--r-- | kernel/sys.c | 73 |
7 files changed, 245 insertions, 472 deletions
diff --git a/kernel/exit.c b/kernel/exit.c index a8c7efc7a681..bc0ec674d3f4 100644 --- a/kernel/exit.c +++ b/kernel/exit.c | |||
@@ -29,6 +29,7 @@ | |||
29 | #include <linux/cpuset.h> | 29 | #include <linux/cpuset.h> |
30 | #include <linux/syscalls.h> | 30 | #include <linux/syscalls.h> |
31 | #include <linux/signal.h> | 31 | #include <linux/signal.h> |
32 | #include <linux/posix-timers.h> | ||
32 | #include <linux/cn_proc.h> | 33 | #include <linux/cn_proc.h> |
33 | #include <linux/mutex.h> | 34 | #include <linux/mutex.h> |
34 | #include <linux/futex.h> | 35 | #include <linux/futex.h> |
@@ -50,15 +51,80 @@ static void __unhash_process(struct task_struct *p) | |||
50 | { | 51 | { |
51 | nr_threads--; | 52 | nr_threads--; |
52 | detach_pid(p, PIDTYPE_PID); | 53 | detach_pid(p, PIDTYPE_PID); |
53 | detach_pid(p, PIDTYPE_TGID); | ||
54 | if (thread_group_leader(p)) { | 54 | if (thread_group_leader(p)) { |
55 | detach_pid(p, PIDTYPE_PGID); | 55 | detach_pid(p, PIDTYPE_PGID); |
56 | detach_pid(p, PIDTYPE_SID); | 56 | detach_pid(p, PIDTYPE_SID); |
57 | if (p->pid) | 57 | |
58 | __get_cpu_var(process_counts)--; | 58 | list_del_init(&p->tasks); |
59 | __get_cpu_var(process_counts)--; | ||
60 | } | ||
61 | list_del_rcu(&p->thread_group); | ||
62 | remove_parent(p); | ||
63 | } | ||
64 | |||
65 | /* | ||
66 | * This function expects the tasklist_lock write-locked. | ||
67 | */ | ||
68 | static void __exit_signal(struct task_struct *tsk) | ||
69 | { | ||
70 | struct signal_struct *sig = tsk->signal; | ||
71 | struct sighand_struct *sighand; | ||
72 | |||
73 | BUG_ON(!sig); | ||
74 | BUG_ON(!atomic_read(&sig->count)); | ||
75 | |||
76 | rcu_read_lock(); | ||
77 | sighand = rcu_dereference(tsk->sighand); | ||
78 | spin_lock(&sighand->siglock); | ||
79 | |||
80 | posix_cpu_timers_exit(tsk); | ||
81 | if (atomic_dec_and_test(&sig->count)) | ||
82 | posix_cpu_timers_exit_group(tsk); | ||
83 | else { | ||
84 | /* | ||
85 | * If there is any task waiting for the group exit | ||
86 | * then notify it: | ||
87 | */ | ||
88 | if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) { | ||
89 | wake_up_process(sig->group_exit_task); | ||
90 | sig->group_exit_task = NULL; | ||
91 | } | ||
92 | if (tsk == sig->curr_target) | ||
93 | sig->curr_target = next_thread(tsk); | ||
94 | /* | ||
95 | * Accumulate here the counters for all threads but the | ||
96 | * group leader as they die, so they can be added into | ||
97 | * the process-wide totals when those are taken. | ||
98 | * The group leader stays around as a zombie as long | ||
99 | * as there are other threads. When it gets reaped, | ||
100 | * the exit.c code will add its counts into these totals. | ||
101 | * We won't ever get here for the group leader, since it | ||
102 | * will have been the last reference on the signal_struct. | ||
103 | */ | ||
104 | sig->utime = cputime_add(sig->utime, tsk->utime); | ||
105 | sig->stime = cputime_add(sig->stime, tsk->stime); | ||
106 | sig->min_flt += tsk->min_flt; | ||
107 | sig->maj_flt += tsk->maj_flt; | ||
108 | sig->nvcsw += tsk->nvcsw; | ||
109 | sig->nivcsw += tsk->nivcsw; | ||
110 | sig->sched_time += tsk->sched_time; | ||
111 | sig = NULL; /* Marker for below. */ | ||
59 | } | 112 | } |
60 | 113 | ||
61 | REMOVE_LINKS(p); | 114 | __unhash_process(tsk); |
115 | |||
116 | tsk->signal = NULL; | ||
117 | tsk->sighand = NULL; | ||
118 | spin_unlock(&sighand->siglock); | ||
119 | rcu_read_unlock(); | ||
120 | |||
121 | __cleanup_sighand(sighand); | ||
122 | clear_tsk_thread_flag(tsk,TIF_SIGPENDING); | ||
123 | flush_sigqueue(&tsk->pending); | ||
124 | if (sig) { | ||
125 | flush_sigqueue(&sig->shared_pending); | ||
126 | __cleanup_signal(sig); | ||
127 | } | ||
62 | } | 128 | } |
63 | 129 | ||
64 | void release_task(struct task_struct * p) | 130 | void release_task(struct task_struct * p) |
@@ -67,21 +133,14 @@ void release_task(struct task_struct * p) | |||
67 | task_t *leader; | 133 | task_t *leader; |
68 | struct dentry *proc_dentry; | 134 | struct dentry *proc_dentry; |
69 | 135 | ||
70 | repeat: | 136 | repeat: |
71 | atomic_dec(&p->user->processes); | 137 | atomic_dec(&p->user->processes); |
72 | spin_lock(&p->proc_lock); | 138 | spin_lock(&p->proc_lock); |
73 | proc_dentry = proc_pid_unhash(p); | 139 | proc_dentry = proc_pid_unhash(p); |
74 | write_lock_irq(&tasklist_lock); | 140 | write_lock_irq(&tasklist_lock); |
75 | if (unlikely(p->ptrace)) | 141 | ptrace_unlink(p); |
76 | __ptrace_unlink(p); | ||
77 | BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children)); | 142 | BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children)); |
78 | __exit_signal(p); | 143 | __exit_signal(p); |
79 | /* | ||
80 | * Note that the fastpath in sys_times depends on __exit_signal having | ||
81 | * updated the counters before a task is removed from the tasklist of | ||
82 | * the process by __unhash_process. | ||
83 | */ | ||
84 | __unhash_process(p); | ||
85 | 144 | ||
86 | /* | 145 | /* |
87 | * If we are the last non-leader member of the thread | 146 | * If we are the last non-leader member of the thread |
@@ -116,21 +175,6 @@ repeat: | |||
116 | goto repeat; | 175 | goto repeat; |
117 | } | 176 | } |
118 | 177 | ||
119 | /* we are using it only for SMP init */ | ||
120 | |||
121 | void unhash_process(struct task_struct *p) | ||
122 | { | ||
123 | struct dentry *proc_dentry; | ||
124 | |||
125 | spin_lock(&p->proc_lock); | ||
126 | proc_dentry = proc_pid_unhash(p); | ||
127 | write_lock_irq(&tasklist_lock); | ||
128 | __unhash_process(p); | ||
129 | write_unlock_irq(&tasklist_lock); | ||
130 | spin_unlock(&p->proc_lock); | ||
131 | proc_pid_flush(proc_dentry); | ||
132 | } | ||
133 | |||
134 | /* | 178 | /* |
135 | * This checks not only the pgrp, but falls back on the pid if no | 179 | * This checks not only the pgrp, but falls back on the pid if no |
136 | * satisfactory pgrp is found. I dunno - gdb doesn't work correctly | 180 | * satisfactory pgrp is found. I dunno - gdb doesn't work correctly |
@@ -238,10 +282,10 @@ static void reparent_to_init(void) | |||
238 | 282 | ||
239 | ptrace_unlink(current); | 283 | ptrace_unlink(current); |
240 | /* Reparent to init */ | 284 | /* Reparent to init */ |
241 | REMOVE_LINKS(current); | 285 | remove_parent(current); |
242 | current->parent = child_reaper; | 286 | current->parent = child_reaper; |
243 | current->real_parent = child_reaper; | 287 | current->real_parent = child_reaper; |
244 | SET_LINKS(current); | 288 | add_parent(current); |
245 | 289 | ||
246 | /* Set the exit signal to SIGCHLD so we signal init on exit */ | 290 | /* Set the exit signal to SIGCHLD so we signal init on exit */ |
247 | current->exit_signal = SIGCHLD; | 291 | current->exit_signal = SIGCHLD; |
@@ -538,13 +582,13 @@ static void exit_mm(struct task_struct * tsk) | |||
538 | mmput(mm); | 582 | mmput(mm); |
539 | } | 583 | } |
540 | 584 | ||
541 | static inline void choose_new_parent(task_t *p, task_t *reaper, task_t *child_reaper) | 585 | static inline void choose_new_parent(task_t *p, task_t *reaper) |
542 | { | 586 | { |
543 | /* | 587 | /* |
544 | * Make sure we're not reparenting to ourselves and that | 588 | * Make sure we're not reparenting to ourselves and that |
545 | * the parent is not a zombie. | 589 | * the parent is not a zombie. |
546 | */ | 590 | */ |
547 | BUG_ON(p == reaper || reaper->exit_state >= EXIT_ZOMBIE); | 591 | BUG_ON(p == reaper || reaper->exit_state); |
548 | p->real_parent = reaper; | 592 | p->real_parent = reaper; |
549 | } | 593 | } |
550 | 594 | ||
@@ -569,9 +613,9 @@ static void reparent_thread(task_t *p, task_t *father, int traced) | |||
569 | * anyway, so let go of it. | 613 | * anyway, so let go of it. |
570 | */ | 614 | */ |
571 | p->ptrace = 0; | 615 | p->ptrace = 0; |
572 | list_del_init(&p->sibling); | 616 | remove_parent(p); |
573 | p->parent = p->real_parent; | 617 | p->parent = p->real_parent; |
574 | list_add_tail(&p->sibling, &p->parent->children); | 618 | add_parent(p); |
575 | 619 | ||
576 | /* If we'd notified the old parent about this child's death, | 620 | /* If we'd notified the old parent about this child's death, |
577 | * also notify the new parent. | 621 | * also notify the new parent. |
@@ -645,7 +689,7 @@ static void forget_original_parent(struct task_struct * father, | |||
645 | 689 | ||
646 | if (father == p->real_parent) { | 690 | if (father == p->real_parent) { |
647 | /* reparent with a reaper, real father it's us */ | 691 | /* reparent with a reaper, real father it's us */ |
648 | choose_new_parent(p, reaper, child_reaper); | 692 | choose_new_parent(p, reaper); |
649 | reparent_thread(p, father, 0); | 693 | reparent_thread(p, father, 0); |
650 | } else { | 694 | } else { |
651 | /* reparent ptraced task to its real parent */ | 695 | /* reparent ptraced task to its real parent */ |
@@ -666,7 +710,7 @@ static void forget_original_parent(struct task_struct * father, | |||
666 | } | 710 | } |
667 | list_for_each_safe(_p, _n, &father->ptrace_children) { | 711 | list_for_each_safe(_p, _n, &father->ptrace_children) { |
668 | p = list_entry(_p,struct task_struct,ptrace_list); | 712 | p = list_entry(_p,struct task_struct,ptrace_list); |
669 | choose_new_parent(p, reaper, child_reaper); | 713 | choose_new_parent(p, reaper); |
670 | reparent_thread(p, father, 1); | 714 | reparent_thread(p, father, 1); |
671 | } | 715 | } |
672 | } | 716 | } |
@@ -807,7 +851,7 @@ fastcall NORET_TYPE void do_exit(long code) | |||
807 | panic("Aiee, killing interrupt handler!"); | 851 | panic("Aiee, killing interrupt handler!"); |
808 | if (unlikely(!tsk->pid)) | 852 | if (unlikely(!tsk->pid)) |
809 | panic("Attempted to kill the idle task!"); | 853 | panic("Attempted to kill the idle task!"); |
810 | if (unlikely(tsk->pid == 1)) | 854 | if (unlikely(tsk == child_reaper)) |
811 | panic("Attempted to kill init!"); | 855 | panic("Attempted to kill init!"); |
812 | 856 | ||
813 | if (unlikely(current->ptrace & PT_TRACE_EXIT)) { | 857 | if (unlikely(current->ptrace & PT_TRACE_EXIT)) { |
@@ -920,13 +964,6 @@ asmlinkage long sys_exit(int error_code) | |||
920 | do_exit((error_code&0xff)<<8); | 964 | do_exit((error_code&0xff)<<8); |
921 | } | 965 | } |
922 | 966 | ||
923 | task_t fastcall *next_thread(const task_t *p) | ||
924 | { | ||
925 | return pid_task(p->pids[PIDTYPE_TGID].pid_list.next, PIDTYPE_TGID); | ||
926 | } | ||
927 | |||
928 | EXPORT_SYMBOL(next_thread); | ||
929 | |||
930 | /* | 967 | /* |
931 | * Take down every thread in the group. This is called by fatal signals | 968 | * Take down every thread in the group. This is called by fatal signals |
932 | * as well as by sys_exit_group (below). | 969 | * as well as by sys_exit_group (below). |
@@ -941,7 +978,6 @@ do_group_exit(int exit_code) | |||
941 | else if (!thread_group_empty(current)) { | 978 | else if (!thread_group_empty(current)) { |
942 | struct signal_struct *const sig = current->signal; | 979 | struct signal_struct *const sig = current->signal; |
943 | struct sighand_struct *const sighand = current->sighand; | 980 | struct sighand_struct *const sighand = current->sighand; |
944 | read_lock(&tasklist_lock); | ||
945 | spin_lock_irq(&sighand->siglock); | 981 | spin_lock_irq(&sighand->siglock); |
946 | if (sig->flags & SIGNAL_GROUP_EXIT) | 982 | if (sig->flags & SIGNAL_GROUP_EXIT) |
947 | /* Another thread got here before we took the lock. */ | 983 | /* Another thread got here before we took the lock. */ |
@@ -951,7 +987,6 @@ do_group_exit(int exit_code) | |||
951 | zap_other_threads(current); | 987 | zap_other_threads(current); |
952 | } | 988 | } |
953 | spin_unlock_irq(&sighand->siglock); | 989 | spin_unlock_irq(&sighand->siglock); |
954 | read_unlock(&tasklist_lock); | ||
955 | } | 990 | } |
956 | 991 | ||
957 | do_exit(exit_code); | 992 | do_exit(exit_code); |
@@ -1281,7 +1316,7 @@ bail_ref: | |||
1281 | 1316 | ||
1282 | /* move to end of parent's list to avoid starvation */ | 1317 | /* move to end of parent's list to avoid starvation */ |
1283 | remove_parent(p); | 1318 | remove_parent(p); |
1284 | add_parent(p, p->parent); | 1319 | add_parent(p); |
1285 | 1320 | ||
1286 | write_unlock_irq(&tasklist_lock); | 1321 | write_unlock_irq(&tasklist_lock); |
1287 | 1322 | ||
diff --git a/kernel/fork.c b/kernel/fork.c index c49bd193b058..b3f7a1bb5e55 100644 --- a/kernel/fork.c +++ b/kernel/fork.c | |||
@@ -84,7 +84,7 @@ static kmem_cache_t *task_struct_cachep; | |||
84 | #endif | 84 | #endif |
85 | 85 | ||
86 | /* SLAB cache for signal_struct structures (tsk->signal) */ | 86 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
87 | kmem_cache_t *signal_cachep; | 87 | static kmem_cache_t *signal_cachep; |
88 | 88 | ||
89 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | 89 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ |
90 | kmem_cache_t *sighand_cachep; | 90 | kmem_cache_t *sighand_cachep; |
@@ -786,14 +786,6 @@ int unshare_files(void) | |||
786 | 786 | ||
787 | EXPORT_SYMBOL(unshare_files); | 787 | EXPORT_SYMBOL(unshare_files); |
788 | 788 | ||
789 | void sighand_free_cb(struct rcu_head *rhp) | ||
790 | { | ||
791 | struct sighand_struct *sp; | ||
792 | |||
793 | sp = container_of(rhp, struct sighand_struct, rcu); | ||
794 | kmem_cache_free(sighand_cachep, sp); | ||
795 | } | ||
796 | |||
797 | static inline int copy_sighand(unsigned long clone_flags, struct task_struct * tsk) | 789 | static inline int copy_sighand(unsigned long clone_flags, struct task_struct * tsk) |
798 | { | 790 | { |
799 | struct sighand_struct *sig; | 791 | struct sighand_struct *sig; |
@@ -806,12 +798,17 @@ static inline int copy_sighand(unsigned long clone_flags, struct task_struct * t | |||
806 | rcu_assign_pointer(tsk->sighand, sig); | 798 | rcu_assign_pointer(tsk->sighand, sig); |
807 | if (!sig) | 799 | if (!sig) |
808 | return -ENOMEM; | 800 | return -ENOMEM; |
809 | spin_lock_init(&sig->siglock); | ||
810 | atomic_set(&sig->count, 1); | 801 | atomic_set(&sig->count, 1); |
811 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); | 802 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); |
812 | return 0; | 803 | return 0; |
813 | } | 804 | } |
814 | 805 | ||
806 | void __cleanup_sighand(struct sighand_struct *sighand) | ||
807 | { | ||
808 | if (atomic_dec_and_test(&sighand->count)) | ||
809 | kmem_cache_free(sighand_cachep, sighand); | ||
810 | } | ||
811 | |||
815 | static inline int copy_signal(unsigned long clone_flags, struct task_struct * tsk) | 812 | static inline int copy_signal(unsigned long clone_flags, struct task_struct * tsk) |
816 | { | 813 | { |
817 | struct signal_struct *sig; | 814 | struct signal_struct *sig; |
@@ -881,6 +878,22 @@ static inline int copy_signal(unsigned long clone_flags, struct task_struct * ts | |||
881 | return 0; | 878 | return 0; |
882 | } | 879 | } |
883 | 880 | ||
881 | void __cleanup_signal(struct signal_struct *sig) | ||
882 | { | ||
883 | exit_thread_group_keys(sig); | ||
884 | kmem_cache_free(signal_cachep, sig); | ||
885 | } | ||
886 | |||
887 | static inline void cleanup_signal(struct task_struct *tsk) | ||
888 | { | ||
889 | struct signal_struct *sig = tsk->signal; | ||
890 | |||
891 | atomic_dec(&sig->live); | ||
892 | |||
893 | if (atomic_dec_and_test(&sig->count)) | ||
894 | __cleanup_signal(sig); | ||
895 | } | ||
896 | |||
884 | static inline void copy_flags(unsigned long clone_flags, struct task_struct *p) | 897 | static inline void copy_flags(unsigned long clone_flags, struct task_struct *p) |
885 | { | 898 | { |
886 | unsigned long new_flags = p->flags; | 899 | unsigned long new_flags = p->flags; |
@@ -1095,6 +1108,7 @@ static task_t *copy_process(unsigned long clone_flags, | |||
1095 | * We dont wake it up yet. | 1108 | * We dont wake it up yet. |
1096 | */ | 1109 | */ |
1097 | p->group_leader = p; | 1110 | p->group_leader = p; |
1111 | INIT_LIST_HEAD(&p->thread_group); | ||
1098 | INIT_LIST_HEAD(&p->ptrace_children); | 1112 | INIT_LIST_HEAD(&p->ptrace_children); |
1099 | INIT_LIST_HEAD(&p->ptrace_list); | 1113 | INIT_LIST_HEAD(&p->ptrace_list); |
1100 | 1114 | ||
@@ -1118,16 +1132,6 @@ static task_t *copy_process(unsigned long clone_flags, | |||
1118 | !cpu_online(task_cpu(p)))) | 1132 | !cpu_online(task_cpu(p)))) |
1119 | set_task_cpu(p, smp_processor_id()); | 1133 | set_task_cpu(p, smp_processor_id()); |
1120 | 1134 | ||
1121 | /* | ||
1122 | * Check for pending SIGKILL! The new thread should not be allowed | ||
1123 | * to slip out of an OOM kill. (or normal SIGKILL.) | ||
1124 | */ | ||
1125 | if (sigismember(¤t->pending.signal, SIGKILL)) { | ||
1126 | write_unlock_irq(&tasklist_lock); | ||
1127 | retval = -EINTR; | ||
1128 | goto bad_fork_cleanup_namespace; | ||
1129 | } | ||
1130 | |||
1131 | /* CLONE_PARENT re-uses the old parent */ | 1135 | /* CLONE_PARENT re-uses the old parent */ |
1132 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) | 1136 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) |
1133 | p->real_parent = current->real_parent; | 1137 | p->real_parent = current->real_parent; |
@@ -1136,6 +1140,23 @@ static task_t *copy_process(unsigned long clone_flags, | |||
1136 | p->parent = p->real_parent; | 1140 | p->parent = p->real_parent; |
1137 | 1141 | ||
1138 | spin_lock(¤t->sighand->siglock); | 1142 | spin_lock(¤t->sighand->siglock); |
1143 | |||
1144 | /* | ||
1145 | * Process group and session signals need to be delivered to just the | ||
1146 | * parent before the fork or both the parent and the child after the | ||
1147 | * fork. Restart if a signal comes in before we add the new process to | ||
1148 | * it's process group. | ||
1149 | * A fatal signal pending means that current will exit, so the new | ||
1150 | * thread can't slip out of an OOM kill (or normal SIGKILL). | ||
1151 | */ | ||
1152 | recalc_sigpending(); | ||
1153 | if (signal_pending(current)) { | ||
1154 | spin_unlock(¤t->sighand->siglock); | ||
1155 | write_unlock_irq(&tasklist_lock); | ||
1156 | retval = -ERESTARTNOINTR; | ||
1157 | goto bad_fork_cleanup_namespace; | ||
1158 | } | ||
1159 | |||
1139 | if (clone_flags & CLONE_THREAD) { | 1160 | if (clone_flags & CLONE_THREAD) { |
1140 | /* | 1161 | /* |
1141 | * Important: if an exit-all has been started then | 1162 | * Important: if an exit-all has been started then |
@@ -1148,17 +1169,9 @@ static task_t *copy_process(unsigned long clone_flags, | |||
1148 | retval = -EAGAIN; | 1169 | retval = -EAGAIN; |
1149 | goto bad_fork_cleanup_namespace; | 1170 | goto bad_fork_cleanup_namespace; |
1150 | } | 1171 | } |
1151 | p->group_leader = current->group_leader; | ||
1152 | 1172 | ||
1153 | if (current->signal->group_stop_count > 0) { | 1173 | p->group_leader = current->group_leader; |
1154 | /* | 1174 | list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group); |
1155 | * There is an all-stop in progress for the group. | ||
1156 | * We ourselves will stop as soon as we check signals. | ||
1157 | * Make the new thread part of that group stop too. | ||
1158 | */ | ||
1159 | current->signal->group_stop_count++; | ||
1160 | set_tsk_thread_flag(p, TIF_SIGPENDING); | ||
1161 | } | ||
1162 | 1175 | ||
1163 | if (!cputime_eq(current->signal->it_virt_expires, | 1176 | if (!cputime_eq(current->signal->it_virt_expires, |
1164 | cputime_zero) || | 1177 | cputime_zero) || |
@@ -1181,23 +1194,25 @@ static task_t *copy_process(unsigned long clone_flags, | |||
1181 | */ | 1194 | */ |
1182 | p->ioprio = current->ioprio; | 1195 | p->ioprio = current->ioprio; |
1183 | 1196 | ||
1184 | SET_LINKS(p); | 1197 | if (likely(p->pid)) { |
1185 | if (unlikely(p->ptrace & PT_PTRACED)) | 1198 | add_parent(p); |
1186 | __ptrace_link(p, current->parent); | 1199 | if (unlikely(p->ptrace & PT_PTRACED)) |
1187 | 1200 | __ptrace_link(p, current->parent); | |
1188 | if (thread_group_leader(p)) { | 1201 | |
1189 | p->signal->tty = current->signal->tty; | 1202 | if (thread_group_leader(p)) { |
1190 | p->signal->pgrp = process_group(current); | 1203 | p->signal->tty = current->signal->tty; |
1191 | p->signal->session = current->signal->session; | 1204 | p->signal->pgrp = process_group(current); |
1192 | attach_pid(p, PIDTYPE_PGID, process_group(p)); | 1205 | p->signal->session = current->signal->session; |
1193 | attach_pid(p, PIDTYPE_SID, p->signal->session); | 1206 | attach_pid(p, PIDTYPE_PGID, process_group(p)); |
1194 | if (p->pid) | 1207 | attach_pid(p, PIDTYPE_SID, p->signal->session); |
1208 | |||
1209 | list_add_tail(&p->tasks, &init_task.tasks); | ||
1195 | __get_cpu_var(process_counts)++; | 1210 | __get_cpu_var(process_counts)++; |
1211 | } | ||
1212 | attach_pid(p, PIDTYPE_PID, p->pid); | ||
1213 | nr_threads++; | ||
1196 | } | 1214 | } |
1197 | attach_pid(p, PIDTYPE_TGID, p->tgid); | ||
1198 | attach_pid(p, PIDTYPE_PID, p->pid); | ||
1199 | 1215 | ||
1200 | nr_threads++; | ||
1201 | total_forks++; | 1216 | total_forks++; |
1202 | spin_unlock(¤t->sighand->siglock); | 1217 | spin_unlock(¤t->sighand->siglock); |
1203 | write_unlock_irq(&tasklist_lock); | 1218 | write_unlock_irq(&tasklist_lock); |
@@ -1212,9 +1227,9 @@ bad_fork_cleanup_mm: | |||
1212 | if (p->mm) | 1227 | if (p->mm) |
1213 | mmput(p->mm); | 1228 | mmput(p->mm); |
1214 | bad_fork_cleanup_signal: | 1229 | bad_fork_cleanup_signal: |
1215 | exit_signal(p); | 1230 | cleanup_signal(p); |
1216 | bad_fork_cleanup_sighand: | 1231 | bad_fork_cleanup_sighand: |
1217 | exit_sighand(p); | 1232 | __cleanup_sighand(p->sighand); |
1218 | bad_fork_cleanup_fs: | 1233 | bad_fork_cleanup_fs: |
1219 | exit_fs(p); /* blocking */ | 1234 | exit_fs(p); /* blocking */ |
1220 | bad_fork_cleanup_files: | 1235 | bad_fork_cleanup_files: |
@@ -1261,7 +1276,7 @@ task_t * __devinit fork_idle(int cpu) | |||
1261 | if (!task) | 1276 | if (!task) |
1262 | return ERR_PTR(-ENOMEM); | 1277 | return ERR_PTR(-ENOMEM); |
1263 | init_idle(task, cpu); | 1278 | init_idle(task, cpu); |
1264 | unhash_process(task); | 1279 | |
1265 | return task; | 1280 | return task; |
1266 | } | 1281 | } |
1267 | 1282 | ||
@@ -1353,11 +1368,21 @@ long do_fork(unsigned long clone_flags, | |||
1353 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | 1368 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 |
1354 | #endif | 1369 | #endif |
1355 | 1370 | ||
1371 | static void sighand_ctor(void *data, kmem_cache_t *cachep, unsigned long flags) | ||
1372 | { | ||
1373 | struct sighand_struct *sighand = data; | ||
1374 | |||
1375 | if ((flags & (SLAB_CTOR_VERIFY | SLAB_CTOR_CONSTRUCTOR)) == | ||
1376 | SLAB_CTOR_CONSTRUCTOR) | ||
1377 | spin_lock_init(&sighand->siglock); | ||
1378 | } | ||
1379 | |||
1356 | void __init proc_caches_init(void) | 1380 | void __init proc_caches_init(void) |
1357 | { | 1381 | { |
1358 | sighand_cachep = kmem_cache_create("sighand_cache", | 1382 | sighand_cachep = kmem_cache_create("sighand_cache", |
1359 | sizeof(struct sighand_struct), 0, | 1383 | sizeof(struct sighand_struct), 0, |
1360 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL); | 1384 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU, |
1385 | sighand_ctor, NULL); | ||
1361 | signal_cachep = kmem_cache_create("signal_cache", | 1386 | signal_cachep = kmem_cache_create("signal_cache", |
1362 | sizeof(struct signal_struct), 0, | 1387 | sizeof(struct signal_struct), 0, |
1363 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL); | 1388 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL); |
diff --git a/kernel/kmod.c b/kernel/kmod.c index 51a892063aaa..20a997c73c3d 100644 --- a/kernel/kmod.c +++ b/kernel/kmod.c | |||
@@ -170,7 +170,7 @@ static int wait_for_helper(void *data) | |||
170 | sa.sa.sa_handler = SIG_IGN; | 170 | sa.sa.sa_handler = SIG_IGN; |
171 | sa.sa.sa_flags = 0; | 171 | sa.sa.sa_flags = 0; |
172 | siginitset(&sa.sa.sa_mask, sigmask(SIGCHLD)); | 172 | siginitset(&sa.sa.sa_mask, sigmask(SIGCHLD)); |
173 | do_sigaction(SIGCHLD, &sa, (struct k_sigaction *)0); | 173 | do_sigaction(SIGCHLD, &sa, NULL); |
174 | allow_signal(SIGCHLD); | 174 | allow_signal(SIGCHLD); |
175 | 175 | ||
176 | pid = kernel_thread(____call_usermodehelper, sub_info, SIGCHLD); | 176 | pid = kernel_thread(____call_usermodehelper, sub_info, SIGCHLD); |
diff --git a/kernel/pid.c b/kernel/pid.c index 1acc07246991..a9f2dfd006d2 100644 --- a/kernel/pid.c +++ b/kernel/pid.c | |||
@@ -218,36 +218,6 @@ task_t *find_task_by_pid_type(int type, int nr) | |||
218 | EXPORT_SYMBOL(find_task_by_pid_type); | 218 | EXPORT_SYMBOL(find_task_by_pid_type); |
219 | 219 | ||
220 | /* | 220 | /* |
221 | * This function switches the PIDs if a non-leader thread calls | ||
222 | * sys_execve() - this must be done without releasing the PID. | ||
223 | * (which a detach_pid() would eventually do.) | ||
224 | */ | ||
225 | void switch_exec_pids(task_t *leader, task_t *thread) | ||
226 | { | ||
227 | __detach_pid(leader, PIDTYPE_PID); | ||
228 | __detach_pid(leader, PIDTYPE_TGID); | ||
229 | __detach_pid(leader, PIDTYPE_PGID); | ||
230 | __detach_pid(leader, PIDTYPE_SID); | ||
231 | |||
232 | __detach_pid(thread, PIDTYPE_PID); | ||
233 | __detach_pid(thread, PIDTYPE_TGID); | ||
234 | |||
235 | leader->pid = leader->tgid = thread->pid; | ||
236 | thread->pid = thread->tgid; | ||
237 | |||
238 | attach_pid(thread, PIDTYPE_PID, thread->pid); | ||
239 | attach_pid(thread, PIDTYPE_TGID, thread->tgid); | ||
240 | attach_pid(thread, PIDTYPE_PGID, thread->signal->pgrp); | ||
241 | attach_pid(thread, PIDTYPE_SID, thread->signal->session); | ||
242 | list_add_tail(&thread->tasks, &init_task.tasks); | ||
243 | |||
244 | attach_pid(leader, PIDTYPE_PID, leader->pid); | ||
245 | attach_pid(leader, PIDTYPE_TGID, leader->tgid); | ||
246 | attach_pid(leader, PIDTYPE_PGID, leader->signal->pgrp); | ||
247 | attach_pid(leader, PIDTYPE_SID, leader->signal->session); | ||
248 | } | ||
249 | |||
250 | /* | ||
251 | * The pid hash table is scaled according to the amount of memory in the | 221 | * The pid hash table is scaled according to the amount of memory in the |
252 | * machine. From a minimum of 16 slots up to 4096 slots at one gigabyte or | 222 | * machine. From a minimum of 16 slots up to 4096 slots at one gigabyte or |
253 | * more. | 223 | * more. |
@@ -277,16 +247,8 @@ void __init pidhash_init(void) | |||
277 | 247 | ||
278 | void __init pidmap_init(void) | 248 | void __init pidmap_init(void) |
279 | { | 249 | { |
280 | int i; | ||
281 | |||
282 | pidmap_array->page = (void *)get_zeroed_page(GFP_KERNEL); | 250 | pidmap_array->page = (void *)get_zeroed_page(GFP_KERNEL); |
251 | /* Reserve PID 0. We never call free_pidmap(0) */ | ||
283 | set_bit(0, pidmap_array->page); | 252 | set_bit(0, pidmap_array->page); |
284 | atomic_dec(&pidmap_array->nr_free); | 253 | atomic_dec(&pidmap_array->nr_free); |
285 | |||
286 | /* | ||
287 | * Allocate PID 0, and hash it via all PID types: | ||
288 | */ | ||
289 | |||
290 | for (i = 0; i < PIDTYPE_MAX; i++) | ||
291 | attach_pid(current, i, 0); | ||
292 | } | 254 | } |
diff --git a/kernel/ptrace.c b/kernel/ptrace.c index d95a72c9279d..86a7f6c60cb2 100644 --- a/kernel/ptrace.c +++ b/kernel/ptrace.c | |||
@@ -35,9 +35,9 @@ void __ptrace_link(task_t *child, task_t *new_parent) | |||
35 | if (child->parent == new_parent) | 35 | if (child->parent == new_parent) |
36 | return; | 36 | return; |
37 | list_add(&child->ptrace_list, &child->parent->ptrace_children); | 37 | list_add(&child->ptrace_list, &child->parent->ptrace_children); |
38 | REMOVE_LINKS(child); | 38 | remove_parent(child); |
39 | child->parent = new_parent; | 39 | child->parent = new_parent; |
40 | SET_LINKS(child); | 40 | add_parent(child); |
41 | } | 41 | } |
42 | 42 | ||
43 | /* | 43 | /* |
@@ -77,9 +77,9 @@ void __ptrace_unlink(task_t *child) | |||
77 | child->ptrace = 0; | 77 | child->ptrace = 0; |
78 | if (!list_empty(&child->ptrace_list)) { | 78 | if (!list_empty(&child->ptrace_list)) { |
79 | list_del_init(&child->ptrace_list); | 79 | list_del_init(&child->ptrace_list); |
80 | REMOVE_LINKS(child); | 80 | remove_parent(child); |
81 | child->parent = child->real_parent; | 81 | child->parent = child->real_parent; |
82 | SET_LINKS(child); | 82 | add_parent(child); |
83 | } | 83 | } |
84 | 84 | ||
85 | ptrace_untrace(child); | 85 | ptrace_untrace(child); |
diff --git a/kernel/signal.c b/kernel/signal.c index 75f7341b0c39..4922928d91f6 100644 --- a/kernel/signal.c +++ b/kernel/signal.c | |||
@@ -22,7 +22,6 @@ | |||
22 | #include <linux/security.h> | 22 | #include <linux/security.h> |
23 | #include <linux/syscalls.h> | 23 | #include <linux/syscalls.h> |
24 | #include <linux/ptrace.h> | 24 | #include <linux/ptrace.h> |
25 | #include <linux/posix-timers.h> | ||
26 | #include <linux/signal.h> | 25 | #include <linux/signal.h> |
27 | #include <linux/audit.h> | 26 | #include <linux/audit.h> |
28 | #include <linux/capability.h> | 27 | #include <linux/capability.h> |
@@ -147,6 +146,8 @@ static kmem_cache_t *sigqueue_cachep; | |||
147 | #define sig_kernel_stop(sig) \ | 146 | #define sig_kernel_stop(sig) \ |
148 | (((sig) < SIGRTMIN) && T(sig, SIG_KERNEL_STOP_MASK)) | 147 | (((sig) < SIGRTMIN) && T(sig, SIG_KERNEL_STOP_MASK)) |
149 | 148 | ||
149 | #define sig_needs_tasklist(sig) ((sig) == SIGCONT) | ||
150 | |||
150 | #define sig_user_defined(t, signr) \ | 151 | #define sig_user_defined(t, signr) \ |
151 | (((t)->sighand->action[(signr)-1].sa.sa_handler != SIG_DFL) && \ | 152 | (((t)->sighand->action[(signr)-1].sa.sa_handler != SIG_DFL) && \ |
152 | ((t)->sighand->action[(signr)-1].sa.sa_handler != SIG_IGN)) | 153 | ((t)->sighand->action[(signr)-1].sa.sa_handler != SIG_IGN)) |
@@ -292,7 +293,7 @@ static void __sigqueue_free(struct sigqueue *q) | |||
292 | kmem_cache_free(sigqueue_cachep, q); | 293 | kmem_cache_free(sigqueue_cachep, q); |
293 | } | 294 | } |
294 | 295 | ||
295 | static void flush_sigqueue(struct sigpending *queue) | 296 | void flush_sigqueue(struct sigpending *queue) |
296 | { | 297 | { |
297 | struct sigqueue *q; | 298 | struct sigqueue *q; |
298 | 299 | ||
@@ -307,9 +308,7 @@ static void flush_sigqueue(struct sigpending *queue) | |||
307 | /* | 308 | /* |
308 | * Flush all pending signals for a task. | 309 | * Flush all pending signals for a task. |
309 | */ | 310 | */ |
310 | 311 | void flush_signals(struct task_struct *t) | |
311 | void | ||
312 | flush_signals(struct task_struct *t) | ||
313 | { | 312 | { |
314 | unsigned long flags; | 313 | unsigned long flags; |
315 | 314 | ||
@@ -321,109 +320,6 @@ flush_signals(struct task_struct *t) | |||
321 | } | 320 | } |
322 | 321 | ||
323 | /* | 322 | /* |
324 | * This function expects the tasklist_lock write-locked. | ||
325 | */ | ||
326 | void __exit_sighand(struct task_struct *tsk) | ||
327 | { | ||
328 | struct sighand_struct * sighand = tsk->sighand; | ||
329 | |||
330 | /* Ok, we're done with the signal handlers */ | ||
331 | tsk->sighand = NULL; | ||
332 | if (atomic_dec_and_test(&sighand->count)) | ||
333 | sighand_free(sighand); | ||
334 | } | ||
335 | |||
336 | void exit_sighand(struct task_struct *tsk) | ||
337 | { | ||
338 | write_lock_irq(&tasklist_lock); | ||
339 | rcu_read_lock(); | ||
340 | if (tsk->sighand != NULL) { | ||
341 | struct sighand_struct *sighand = rcu_dereference(tsk->sighand); | ||
342 | spin_lock(&sighand->siglock); | ||
343 | __exit_sighand(tsk); | ||
344 | spin_unlock(&sighand->siglock); | ||
345 | } | ||
346 | rcu_read_unlock(); | ||
347 | write_unlock_irq(&tasklist_lock); | ||
348 | } | ||
349 | |||
350 | /* | ||
351 | * This function expects the tasklist_lock write-locked. | ||
352 | */ | ||
353 | void __exit_signal(struct task_struct *tsk) | ||
354 | { | ||
355 | struct signal_struct * sig = tsk->signal; | ||
356 | struct sighand_struct * sighand; | ||
357 | |||
358 | if (!sig) | ||
359 | BUG(); | ||
360 | if (!atomic_read(&sig->count)) | ||
361 | BUG(); | ||
362 | rcu_read_lock(); | ||
363 | sighand = rcu_dereference(tsk->sighand); | ||
364 | spin_lock(&sighand->siglock); | ||
365 | posix_cpu_timers_exit(tsk); | ||
366 | if (atomic_dec_and_test(&sig->count)) { | ||
367 | posix_cpu_timers_exit_group(tsk); | ||
368 | tsk->signal = NULL; | ||
369 | __exit_sighand(tsk); | ||
370 | spin_unlock(&sighand->siglock); | ||
371 | flush_sigqueue(&sig->shared_pending); | ||
372 | } else { | ||
373 | /* | ||
374 | * If there is any task waiting for the group exit | ||
375 | * then notify it: | ||
376 | */ | ||
377 | if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) { | ||
378 | wake_up_process(sig->group_exit_task); | ||
379 | sig->group_exit_task = NULL; | ||
380 | } | ||
381 | if (tsk == sig->curr_target) | ||
382 | sig->curr_target = next_thread(tsk); | ||
383 | tsk->signal = NULL; | ||
384 | /* | ||
385 | * Accumulate here the counters for all threads but the | ||
386 | * group leader as they die, so they can be added into | ||
387 | * the process-wide totals when those are taken. | ||
388 | * The group leader stays around as a zombie as long | ||
389 | * as there are other threads. When it gets reaped, | ||
390 | * the exit.c code will add its counts into these totals. | ||
391 | * We won't ever get here for the group leader, since it | ||
392 | * will have been the last reference on the signal_struct. | ||
393 | */ | ||
394 | sig->utime = cputime_add(sig->utime, tsk->utime); | ||
395 | sig->stime = cputime_add(sig->stime, tsk->stime); | ||
396 | sig->min_flt += tsk->min_flt; | ||
397 | sig->maj_flt += tsk->maj_flt; | ||
398 | sig->nvcsw += tsk->nvcsw; | ||
399 | sig->nivcsw += tsk->nivcsw; | ||
400 | sig->sched_time += tsk->sched_time; | ||
401 | __exit_sighand(tsk); | ||
402 | spin_unlock(&sighand->siglock); | ||
403 | sig = NULL; /* Marker for below. */ | ||
404 | } | ||
405 | rcu_read_unlock(); | ||
406 | clear_tsk_thread_flag(tsk,TIF_SIGPENDING); | ||
407 | flush_sigqueue(&tsk->pending); | ||
408 | if (sig) { | ||
409 | /* | ||
410 | * We are cleaning up the signal_struct here. | ||
411 | */ | ||
412 | exit_thread_group_keys(sig); | ||
413 | kmem_cache_free(signal_cachep, sig); | ||
414 | } | ||
415 | } | ||
416 | |||
417 | void exit_signal(struct task_struct *tsk) | ||
418 | { | ||
419 | atomic_dec(&tsk->signal->live); | ||
420 | |||
421 | write_lock_irq(&tasklist_lock); | ||
422 | __exit_signal(tsk); | ||
423 | write_unlock_irq(&tasklist_lock); | ||
424 | } | ||
425 | |||
426 | /* | ||
427 | * Flush all handlers for a task. | 323 | * Flush all handlers for a task. |
428 | */ | 324 | */ |
429 | 325 | ||
@@ -695,9 +591,7 @@ static int check_kill_permission(int sig, struct siginfo *info, | |||
695 | } | 591 | } |
696 | 592 | ||
697 | /* forward decl */ | 593 | /* forward decl */ |
698 | static void do_notify_parent_cldstop(struct task_struct *tsk, | 594 | static void do_notify_parent_cldstop(struct task_struct *tsk, int why); |
699 | int to_self, | ||
700 | int why); | ||
701 | 595 | ||
702 | /* | 596 | /* |
703 | * Handle magic process-wide effects of stop/continue signals. | 597 | * Handle magic process-wide effects of stop/continue signals. |
@@ -747,7 +641,7 @@ static void handle_stop_signal(int sig, struct task_struct *p) | |||
747 | p->signal->group_stop_count = 0; | 641 | p->signal->group_stop_count = 0; |
748 | p->signal->flags = SIGNAL_STOP_CONTINUED; | 642 | p->signal->flags = SIGNAL_STOP_CONTINUED; |
749 | spin_unlock(&p->sighand->siglock); | 643 | spin_unlock(&p->sighand->siglock); |
750 | do_notify_parent_cldstop(p, (p->ptrace & PT_PTRACED), CLD_STOPPED); | 644 | do_notify_parent_cldstop(p, CLD_STOPPED); |
751 | spin_lock(&p->sighand->siglock); | 645 | spin_lock(&p->sighand->siglock); |
752 | } | 646 | } |
753 | rm_from_queue(SIG_KERNEL_STOP_MASK, &p->signal->shared_pending); | 647 | rm_from_queue(SIG_KERNEL_STOP_MASK, &p->signal->shared_pending); |
@@ -788,7 +682,7 @@ static void handle_stop_signal(int sig, struct task_struct *p) | |||
788 | p->signal->flags = SIGNAL_STOP_CONTINUED; | 682 | p->signal->flags = SIGNAL_STOP_CONTINUED; |
789 | p->signal->group_exit_code = 0; | 683 | p->signal->group_exit_code = 0; |
790 | spin_unlock(&p->sighand->siglock); | 684 | spin_unlock(&p->sighand->siglock); |
791 | do_notify_parent_cldstop(p, (p->ptrace & PT_PTRACED), CLD_CONTINUED); | 685 | do_notify_parent_cldstop(p, CLD_CONTINUED); |
792 | spin_lock(&p->sighand->siglock); | 686 | spin_lock(&p->sighand->siglock); |
793 | } else { | 687 | } else { |
794 | /* | 688 | /* |
@@ -1120,27 +1014,37 @@ void zap_other_threads(struct task_struct *p) | |||
1120 | /* | 1014 | /* |
1121 | * Must be called under rcu_read_lock() or with tasklist_lock read-held. | 1015 | * Must be called under rcu_read_lock() or with tasklist_lock read-held. |
1122 | */ | 1016 | */ |
1017 | struct sighand_struct *lock_task_sighand(struct task_struct *tsk, unsigned long *flags) | ||
1018 | { | ||
1019 | struct sighand_struct *sighand; | ||
1020 | |||
1021 | for (;;) { | ||
1022 | sighand = rcu_dereference(tsk->sighand); | ||
1023 | if (unlikely(sighand == NULL)) | ||
1024 | break; | ||
1025 | |||
1026 | spin_lock_irqsave(&sighand->siglock, *flags); | ||
1027 | if (likely(sighand == tsk->sighand)) | ||
1028 | break; | ||
1029 | spin_unlock_irqrestore(&sighand->siglock, *flags); | ||
1030 | } | ||
1031 | |||
1032 | return sighand; | ||
1033 | } | ||
1034 | |||
1123 | int group_send_sig_info(int sig, struct siginfo *info, struct task_struct *p) | 1035 | int group_send_sig_info(int sig, struct siginfo *info, struct task_struct *p) |
1124 | { | 1036 | { |
1125 | unsigned long flags; | 1037 | unsigned long flags; |
1126 | struct sighand_struct *sp; | ||
1127 | int ret; | 1038 | int ret; |
1128 | 1039 | ||
1129 | retry: | ||
1130 | ret = check_kill_permission(sig, info, p); | 1040 | ret = check_kill_permission(sig, info, p); |
1131 | if (!ret && sig && (sp = rcu_dereference(p->sighand))) { | 1041 | |
1132 | spin_lock_irqsave(&sp->siglock, flags); | 1042 | if (!ret && sig) { |
1133 | if (p->sighand != sp) { | 1043 | ret = -ESRCH; |
1134 | spin_unlock_irqrestore(&sp->siglock, flags); | 1044 | if (lock_task_sighand(p, &flags)) { |
1135 | goto retry; | 1045 | ret = __group_send_sig_info(sig, info, p); |
1136 | } | 1046 | unlock_task_sighand(p, &flags); |
1137 | if ((atomic_read(&sp->count) == 0) || | ||
1138 | (atomic_read(&p->usage) == 0)) { | ||
1139 | spin_unlock_irqrestore(&sp->siglock, flags); | ||
1140 | return -ESRCH; | ||
1141 | } | 1047 | } |
1142 | ret = __group_send_sig_info(sig, info, p); | ||
1143 | spin_unlock_irqrestore(&sp->siglock, flags); | ||
1144 | } | 1048 | } |
1145 | 1049 | ||
1146 | return ret; | 1050 | return ret; |
@@ -1189,7 +1093,7 @@ kill_proc_info(int sig, struct siginfo *info, pid_t pid) | |||
1189 | struct task_struct *p; | 1093 | struct task_struct *p; |
1190 | 1094 | ||
1191 | rcu_read_lock(); | 1095 | rcu_read_lock(); |
1192 | if (unlikely(sig_kernel_stop(sig) || sig == SIGCONT)) { | 1096 | if (unlikely(sig_needs_tasklist(sig))) { |
1193 | read_lock(&tasklist_lock); | 1097 | read_lock(&tasklist_lock); |
1194 | acquired_tasklist_lock = 1; | 1098 | acquired_tasklist_lock = 1; |
1195 | } | 1099 | } |
@@ -1405,12 +1309,10 @@ void sigqueue_free(struct sigqueue *q) | |||
1405 | __sigqueue_free(q); | 1309 | __sigqueue_free(q); |
1406 | } | 1310 | } |
1407 | 1311 | ||
1408 | int | 1312 | int send_sigqueue(int sig, struct sigqueue *q, struct task_struct *p) |
1409 | send_sigqueue(int sig, struct sigqueue *q, struct task_struct *p) | ||
1410 | { | 1313 | { |
1411 | unsigned long flags; | 1314 | unsigned long flags; |
1412 | int ret = 0; | 1315 | int ret = 0; |
1413 | struct sighand_struct *sh; | ||
1414 | 1316 | ||
1415 | BUG_ON(!(q->flags & SIGQUEUE_PREALLOC)); | 1317 | BUG_ON(!(q->flags & SIGQUEUE_PREALLOC)); |
1416 | 1318 | ||
@@ -1424,48 +1326,17 @@ send_sigqueue(int sig, struct sigqueue *q, struct task_struct *p) | |||
1424 | */ | 1326 | */ |
1425 | rcu_read_lock(); | 1327 | rcu_read_lock(); |
1426 | 1328 | ||
1427 | if (unlikely(p->flags & PF_EXITING)) { | 1329 | if (!likely(lock_task_sighand(p, &flags))) { |
1428 | ret = -1; | 1330 | ret = -1; |
1429 | goto out_err; | 1331 | goto out_err; |
1430 | } | 1332 | } |
1431 | 1333 | ||
1432 | retry: | ||
1433 | sh = rcu_dereference(p->sighand); | ||
1434 | |||
1435 | spin_lock_irqsave(&sh->siglock, flags); | ||
1436 | if (p->sighand != sh) { | ||
1437 | /* We raced with exec() in a multithreaded process... */ | ||
1438 | spin_unlock_irqrestore(&sh->siglock, flags); | ||
1439 | goto retry; | ||
1440 | } | ||
1441 | |||
1442 | /* | ||
1443 | * We do the check here again to handle the following scenario: | ||
1444 | * | ||
1445 | * CPU 0 CPU 1 | ||
1446 | * send_sigqueue | ||
1447 | * check PF_EXITING | ||
1448 | * interrupt exit code running | ||
1449 | * __exit_signal | ||
1450 | * lock sighand->siglock | ||
1451 | * unlock sighand->siglock | ||
1452 | * lock sh->siglock | ||
1453 | * add(tsk->pending) flush_sigqueue(tsk->pending) | ||
1454 | * | ||
1455 | */ | ||
1456 | |||
1457 | if (unlikely(p->flags & PF_EXITING)) { | ||
1458 | ret = -1; | ||
1459 | goto out; | ||
1460 | } | ||
1461 | |||
1462 | if (unlikely(!list_empty(&q->list))) { | 1334 | if (unlikely(!list_empty(&q->list))) { |
1463 | /* | 1335 | /* |
1464 | * If an SI_TIMER entry is already queue just increment | 1336 | * If an SI_TIMER entry is already queue just increment |
1465 | * the overrun count. | 1337 | * the overrun count. |
1466 | */ | 1338 | */ |
1467 | if (q->info.si_code != SI_TIMER) | 1339 | BUG_ON(q->info.si_code != SI_TIMER); |
1468 | BUG(); | ||
1469 | q->info.si_overrun++; | 1340 | q->info.si_overrun++; |
1470 | goto out; | 1341 | goto out; |
1471 | } | 1342 | } |
@@ -1481,7 +1352,7 @@ retry: | |||
1481 | signal_wake_up(p, sig == SIGKILL); | 1352 | signal_wake_up(p, sig == SIGKILL); |
1482 | 1353 | ||
1483 | out: | 1354 | out: |
1484 | spin_unlock_irqrestore(&sh->siglock, flags); | 1355 | unlock_task_sighand(p, &flags); |
1485 | out_err: | 1356 | out_err: |
1486 | rcu_read_unlock(); | 1357 | rcu_read_unlock(); |
1487 | 1358 | ||
@@ -1613,14 +1484,14 @@ void do_notify_parent(struct task_struct *tsk, int sig) | |||
1613 | spin_unlock_irqrestore(&psig->siglock, flags); | 1484 | spin_unlock_irqrestore(&psig->siglock, flags); |
1614 | } | 1485 | } |
1615 | 1486 | ||
1616 | static void do_notify_parent_cldstop(struct task_struct *tsk, int to_self, int why) | 1487 | static void do_notify_parent_cldstop(struct task_struct *tsk, int why) |
1617 | { | 1488 | { |
1618 | struct siginfo info; | 1489 | struct siginfo info; |
1619 | unsigned long flags; | 1490 | unsigned long flags; |
1620 | struct task_struct *parent; | 1491 | struct task_struct *parent; |
1621 | struct sighand_struct *sighand; | 1492 | struct sighand_struct *sighand; |
1622 | 1493 | ||
1623 | if (to_self) | 1494 | if (tsk->ptrace & PT_PTRACED) |
1624 | parent = tsk->parent; | 1495 | parent = tsk->parent; |
1625 | else { | 1496 | else { |
1626 | tsk = tsk->group_leader; | 1497 | tsk = tsk->group_leader; |
@@ -1695,7 +1566,7 @@ static void ptrace_stop(int exit_code, int nostop_code, siginfo_t *info) | |||
1695 | !(current->ptrace & PT_ATTACHED)) && | 1566 | !(current->ptrace & PT_ATTACHED)) && |
1696 | (likely(current->parent->signal != current->signal) || | 1567 | (likely(current->parent->signal != current->signal) || |
1697 | !unlikely(current->signal->flags & SIGNAL_GROUP_EXIT))) { | 1568 | !unlikely(current->signal->flags & SIGNAL_GROUP_EXIT))) { |
1698 | do_notify_parent_cldstop(current, 1, CLD_TRAPPED); | 1569 | do_notify_parent_cldstop(current, CLD_TRAPPED); |
1699 | read_unlock(&tasklist_lock); | 1570 | read_unlock(&tasklist_lock); |
1700 | schedule(); | 1571 | schedule(); |
1701 | } else { | 1572 | } else { |
@@ -1744,25 +1615,17 @@ void ptrace_notify(int exit_code) | |||
1744 | static void | 1615 | static void |
1745 | finish_stop(int stop_count) | 1616 | finish_stop(int stop_count) |
1746 | { | 1617 | { |
1747 | int to_self; | ||
1748 | |||
1749 | /* | 1618 | /* |
1750 | * If there are no other threads in the group, or if there is | 1619 | * If there are no other threads in the group, or if there is |
1751 | * a group stop in progress and we are the last to stop, | 1620 | * a group stop in progress and we are the last to stop, |
1752 | * report to the parent. When ptraced, every thread reports itself. | 1621 | * report to the parent. When ptraced, every thread reports itself. |
1753 | */ | 1622 | */ |
1754 | if (stop_count < 0 || (current->ptrace & PT_PTRACED)) | 1623 | if (stop_count == 0 || (current->ptrace & PT_PTRACED)) { |
1755 | to_self = 1; | 1624 | read_lock(&tasklist_lock); |
1756 | else if (stop_count == 0) | 1625 | do_notify_parent_cldstop(current, CLD_STOPPED); |
1757 | to_self = 0; | 1626 | read_unlock(&tasklist_lock); |
1758 | else | 1627 | } |
1759 | goto out; | ||
1760 | |||
1761 | read_lock(&tasklist_lock); | ||
1762 | do_notify_parent_cldstop(current, to_self, CLD_STOPPED); | ||
1763 | read_unlock(&tasklist_lock); | ||
1764 | 1628 | ||
1765 | out: | ||
1766 | schedule(); | 1629 | schedule(); |
1767 | /* | 1630 | /* |
1768 | * Now we don't run again until continued. | 1631 | * Now we don't run again until continued. |
@@ -1776,12 +1639,10 @@ out: | |||
1776 | * Returns nonzero if we've actually stopped and released the siglock. | 1639 | * Returns nonzero if we've actually stopped and released the siglock. |
1777 | * Returns zero if we didn't stop and still hold the siglock. | 1640 | * Returns zero if we didn't stop and still hold the siglock. |
1778 | */ | 1641 | */ |
1779 | static int | 1642 | static int do_signal_stop(int signr) |
1780 | do_signal_stop(int signr) | ||
1781 | { | 1643 | { |
1782 | struct signal_struct *sig = current->signal; | 1644 | struct signal_struct *sig = current->signal; |
1783 | struct sighand_struct *sighand = current->sighand; | 1645 | int stop_count; |
1784 | int stop_count = -1; | ||
1785 | 1646 | ||
1786 | if (!likely(sig->flags & SIGNAL_STOP_DEQUEUED)) | 1647 | if (!likely(sig->flags & SIGNAL_STOP_DEQUEUED)) |
1787 | return 0; | 1648 | return 0; |
@@ -1791,86 +1652,37 @@ do_signal_stop(int signr) | |||
1791 | * There is a group stop in progress. We don't need to | 1652 | * There is a group stop in progress. We don't need to |
1792 | * start another one. | 1653 | * start another one. |
1793 | */ | 1654 | */ |
1794 | signr = sig->group_exit_code; | ||
1795 | stop_count = --sig->group_stop_count; | 1655 | stop_count = --sig->group_stop_count; |
1796 | current->exit_code = signr; | 1656 | } else { |
1797 | set_current_state(TASK_STOPPED); | ||
1798 | if (stop_count == 0) | ||
1799 | sig->flags = SIGNAL_STOP_STOPPED; | ||
1800 | spin_unlock_irq(&sighand->siglock); | ||
1801 | } | ||
1802 | else if (thread_group_empty(current)) { | ||
1803 | /* | ||
1804 | * Lock must be held through transition to stopped state. | ||
1805 | */ | ||
1806 | current->exit_code = current->signal->group_exit_code = signr; | ||
1807 | set_current_state(TASK_STOPPED); | ||
1808 | sig->flags = SIGNAL_STOP_STOPPED; | ||
1809 | spin_unlock_irq(&sighand->siglock); | ||
1810 | } | ||
1811 | else { | ||
1812 | /* | 1657 | /* |
1813 | * There is no group stop already in progress. | 1658 | * There is no group stop already in progress. |
1814 | * We must initiate one now, but that requires | 1659 | * We must initiate one now. |
1815 | * dropping siglock to get both the tasklist lock | ||
1816 | * and siglock again in the proper order. Note that | ||
1817 | * this allows an intervening SIGCONT to be posted. | ||
1818 | * We need to check for that and bail out if necessary. | ||
1819 | */ | 1660 | */ |
1820 | struct task_struct *t; | 1661 | struct task_struct *t; |
1821 | 1662 | ||
1822 | spin_unlock_irq(&sighand->siglock); | 1663 | sig->group_exit_code = signr; |
1823 | |||
1824 | /* signals can be posted during this window */ | ||
1825 | 1664 | ||
1826 | read_lock(&tasklist_lock); | 1665 | stop_count = 0; |
1827 | spin_lock_irq(&sighand->siglock); | 1666 | for (t = next_thread(current); t != current; t = next_thread(t)) |
1828 | |||
1829 | if (!likely(sig->flags & SIGNAL_STOP_DEQUEUED)) { | ||
1830 | /* | 1667 | /* |
1831 | * Another stop or continue happened while we | 1668 | * Setting state to TASK_STOPPED for a group |
1832 | * didn't have the lock. We can just swallow this | 1669 | * stop is always done with the siglock held, |
1833 | * signal now. If we raced with a SIGCONT, that | 1670 | * so this check has no races. |
1834 | * should have just cleared it now. If we raced | ||
1835 | * with another processor delivering a stop signal, | ||
1836 | * then the SIGCONT that wakes us up should clear it. | ||
1837 | */ | 1671 | */ |
1838 | read_unlock(&tasklist_lock); | 1672 | if (!t->exit_state && |
1839 | return 0; | 1673 | !(t->state & (TASK_STOPPED|TASK_TRACED))) { |
1840 | } | 1674 | stop_count++; |
1841 | 1675 | signal_wake_up(t, 0); | |
1842 | if (sig->group_stop_count == 0) { | 1676 | } |
1843 | sig->group_exit_code = signr; | 1677 | sig->group_stop_count = stop_count; |
1844 | stop_count = 0; | ||
1845 | for (t = next_thread(current); t != current; | ||
1846 | t = next_thread(t)) | ||
1847 | /* | ||
1848 | * Setting state to TASK_STOPPED for a group | ||
1849 | * stop is always done with the siglock held, | ||
1850 | * so this check has no races. | ||
1851 | */ | ||
1852 | if (!t->exit_state && | ||
1853 | !(t->state & (TASK_STOPPED|TASK_TRACED))) { | ||
1854 | stop_count++; | ||
1855 | signal_wake_up(t, 0); | ||
1856 | } | ||
1857 | sig->group_stop_count = stop_count; | ||
1858 | } | ||
1859 | else { | ||
1860 | /* A race with another thread while unlocked. */ | ||
1861 | signr = sig->group_exit_code; | ||
1862 | stop_count = --sig->group_stop_count; | ||
1863 | } | ||
1864 | |||
1865 | current->exit_code = signr; | ||
1866 | set_current_state(TASK_STOPPED); | ||
1867 | if (stop_count == 0) | ||
1868 | sig->flags = SIGNAL_STOP_STOPPED; | ||
1869 | |||
1870 | spin_unlock_irq(&sighand->siglock); | ||
1871 | read_unlock(&tasklist_lock); | ||
1872 | } | 1678 | } |
1873 | 1679 | ||
1680 | if (stop_count == 0) | ||
1681 | sig->flags = SIGNAL_STOP_STOPPED; | ||
1682 | current->exit_code = sig->group_exit_code; | ||
1683 | __set_current_state(TASK_STOPPED); | ||
1684 | |||
1685 | spin_unlock_irq(¤t->sighand->siglock); | ||
1874 | finish_stop(stop_count); | 1686 | finish_stop(stop_count); |
1875 | return 1; | 1687 | return 1; |
1876 | } | 1688 | } |
@@ -1990,7 +1802,7 @@ relock: | |||
1990 | continue; | 1802 | continue; |
1991 | 1803 | ||
1992 | /* Init gets no signals it doesn't want. */ | 1804 | /* Init gets no signals it doesn't want. */ |
1993 | if (current->pid == 1) | 1805 | if (current == child_reaper) |
1994 | continue; | 1806 | continue; |
1995 | 1807 | ||
1996 | if (sig_kernel_stop(signr)) { | 1808 | if (sig_kernel_stop(signr)) { |
@@ -2430,8 +2242,7 @@ sys_rt_sigqueueinfo(int pid, int sig, siginfo_t __user *uinfo) | |||
2430 | return kill_proc_info(sig, &info, pid); | 2242 | return kill_proc_info(sig, &info, pid); |
2431 | } | 2243 | } |
2432 | 2244 | ||
2433 | int | 2245 | int do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact) |
2434 | do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact) | ||
2435 | { | 2246 | { |
2436 | struct k_sigaction *k; | 2247 | struct k_sigaction *k; |
2437 | sigset_t mask; | 2248 | sigset_t mask; |
@@ -2457,6 +2268,7 @@ do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact) | |||
2457 | if (act) { | 2268 | if (act) { |
2458 | sigdelsetmask(&act->sa.sa_mask, | 2269 | sigdelsetmask(&act->sa.sa_mask, |
2459 | sigmask(SIGKILL) | sigmask(SIGSTOP)); | 2270 | sigmask(SIGKILL) | sigmask(SIGSTOP)); |
2271 | *k = *act; | ||
2460 | /* | 2272 | /* |
2461 | * POSIX 3.3.1.3: | 2273 | * POSIX 3.3.1.3: |
2462 | * "Setting a signal action to SIG_IGN for a signal that is | 2274 | * "Setting a signal action to SIG_IGN for a signal that is |
@@ -2469,19 +2281,8 @@ do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact) | |||
2469 | * be discarded, whether or not it is blocked" | 2281 | * be discarded, whether or not it is blocked" |
2470 | */ | 2282 | */ |
2471 | if (act->sa.sa_handler == SIG_IGN || | 2283 | if (act->sa.sa_handler == SIG_IGN || |
2472 | (act->sa.sa_handler == SIG_DFL && | 2284 | (act->sa.sa_handler == SIG_DFL && sig_kernel_ignore(sig))) { |
2473 | sig_kernel_ignore(sig))) { | ||
2474 | /* | ||
2475 | * This is a fairly rare case, so we only take the | ||
2476 | * tasklist_lock once we're sure we'll need it. | ||
2477 | * Now we must do this little unlock and relock | ||
2478 | * dance to maintain the lock hierarchy. | ||
2479 | */ | ||
2480 | struct task_struct *t = current; | 2285 | struct task_struct *t = current; |
2481 | spin_unlock_irq(&t->sighand->siglock); | ||
2482 | read_lock(&tasklist_lock); | ||
2483 | spin_lock_irq(&t->sighand->siglock); | ||
2484 | *k = *act; | ||
2485 | sigemptyset(&mask); | 2286 | sigemptyset(&mask); |
2486 | sigaddset(&mask, sig); | 2287 | sigaddset(&mask, sig); |
2487 | rm_from_queue_full(&mask, &t->signal->shared_pending); | 2288 | rm_from_queue_full(&mask, &t->signal->shared_pending); |
@@ -2490,12 +2291,7 @@ do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact) | |||
2490 | recalc_sigpending_tsk(t); | 2291 | recalc_sigpending_tsk(t); |
2491 | t = next_thread(t); | 2292 | t = next_thread(t); |
2492 | } while (t != current); | 2293 | } while (t != current); |
2493 | spin_unlock_irq(¤t->sighand->siglock); | ||
2494 | read_unlock(&tasklist_lock); | ||
2495 | return 0; | ||
2496 | } | 2294 | } |
2497 | |||
2498 | *k = *act; | ||
2499 | } | 2295 | } |
2500 | 2296 | ||
2501 | spin_unlock_irq(¤t->sighand->siglock); | 2297 | spin_unlock_irq(¤t->sighand->siglock); |
diff --git a/kernel/sys.c b/kernel/sys.c index c93d37f71aef..7ef7f6054c28 100644 --- a/kernel/sys.c +++ b/kernel/sys.c | |||
@@ -1202,69 +1202,24 @@ asmlinkage long sys_times(struct tms __user * tbuf) | |||
1202 | */ | 1202 | */ |
1203 | if (tbuf) { | 1203 | if (tbuf) { |
1204 | struct tms tmp; | 1204 | struct tms tmp; |
1205 | struct task_struct *tsk = current; | ||
1206 | struct task_struct *t; | ||
1205 | cputime_t utime, stime, cutime, cstime; | 1207 | cputime_t utime, stime, cutime, cstime; |
1206 | 1208 | ||
1207 | #ifdef CONFIG_SMP | 1209 | spin_lock_irq(&tsk->sighand->siglock); |
1208 | if (thread_group_empty(current)) { | 1210 | utime = tsk->signal->utime; |
1209 | /* | 1211 | stime = tsk->signal->stime; |
1210 | * Single thread case without the use of any locks. | 1212 | t = tsk; |
1211 | * | 1213 | do { |
1212 | * We may race with release_task if two threads are | 1214 | utime = cputime_add(utime, t->utime); |
1213 | * executing. However, release task first adds up the | 1215 | stime = cputime_add(stime, t->stime); |
1214 | * counters (__exit_signal) before removing the task | 1216 | t = next_thread(t); |
1215 | * from the process tasklist (__unhash_process). | 1217 | } while (t != tsk); |
1216 | * __exit_signal also acquires and releases the | ||
1217 | * siglock which results in the proper memory ordering | ||
1218 | * so that the list modifications are always visible | ||
1219 | * after the counters have been updated. | ||
1220 | * | ||
1221 | * If the counters have been updated by the second thread | ||
1222 | * but the thread has not yet been removed from the list | ||
1223 | * then the other branch will be executing which will | ||
1224 | * block on tasklist_lock until the exit handling of the | ||
1225 | * other task is finished. | ||
1226 | * | ||
1227 | * This also implies that the sighand->siglock cannot | ||
1228 | * be held by another processor. So we can also | ||
1229 | * skip acquiring that lock. | ||
1230 | */ | ||
1231 | utime = cputime_add(current->signal->utime, current->utime); | ||
1232 | stime = cputime_add(current->signal->utime, current->stime); | ||
1233 | cutime = current->signal->cutime; | ||
1234 | cstime = current->signal->cstime; | ||
1235 | } else | ||
1236 | #endif | ||
1237 | { | ||
1238 | 1218 | ||
1239 | /* Process with multiple threads */ | 1219 | cutime = tsk->signal->cutime; |
1240 | struct task_struct *tsk = current; | 1220 | cstime = tsk->signal->cstime; |
1241 | struct task_struct *t; | 1221 | spin_unlock_irq(&tsk->sighand->siglock); |
1242 | 1222 | ||
1243 | read_lock(&tasklist_lock); | ||
1244 | utime = tsk->signal->utime; | ||
1245 | stime = tsk->signal->stime; | ||
1246 | t = tsk; | ||
1247 | do { | ||
1248 | utime = cputime_add(utime, t->utime); | ||
1249 | stime = cputime_add(stime, t->stime); | ||
1250 | t = next_thread(t); | ||
1251 | } while (t != tsk); | ||
1252 | |||
1253 | /* | ||
1254 | * While we have tasklist_lock read-locked, no dying thread | ||
1255 | * can be updating current->signal->[us]time. Instead, | ||
1256 | * we got their counts included in the live thread loop. | ||
1257 | * However, another thread can come in right now and | ||
1258 | * do a wait call that updates current->signal->c[us]time. | ||
1259 | * To make sure we always see that pair updated atomically, | ||
1260 | * we take the siglock around fetching them. | ||
1261 | */ | ||
1262 | spin_lock_irq(&tsk->sighand->siglock); | ||
1263 | cutime = tsk->signal->cutime; | ||
1264 | cstime = tsk->signal->cstime; | ||
1265 | spin_unlock_irq(&tsk->sighand->siglock); | ||
1266 | read_unlock(&tasklist_lock); | ||
1267 | } | ||
1268 | tmp.tms_utime = cputime_to_clock_t(utime); | 1223 | tmp.tms_utime = cputime_to_clock_t(utime); |
1269 | tmp.tms_stime = cputime_to_clock_t(stime); | 1224 | tmp.tms_stime = cputime_to_clock_t(stime); |
1270 | tmp.tms_cutime = cputime_to_clock_t(cutime); | 1225 | tmp.tms_cutime = cputime_to_clock_t(cutime); |