Merge branch 'linus' into x86/irq
[pandora-kernel.git] / kernel / exit.c
index 97f609f..8f6185e 100644 (file)
@@ -19,6 +19,7 @@
 #include <linux/acct.h>
 #include <linux/tsacct_kern.h>
 #include <linux/file.h>
+#include <linux/fdtable.h>
 #include <linux/binfmts.h>
 #include <linux/nsproxy.h>
 #include <linux/pid_namespace.h>
 
 static void exit_mm(struct task_struct * tsk);
 
+static inline int task_detached(struct task_struct *p)
+{
+       return p->exit_signal == -1;
+}
+
 static void __unhash_process(struct task_struct *p)
 {
        nr_threads--;
@@ -120,6 +126,12 @@ static void __exit_signal(struct task_struct *tsk)
 
        __unhash_process(tsk);
 
+       /*
+        * Do this under ->siglock, we can race with another thread
+        * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
+        */
+       flush_sigqueue(&tsk->pending);
+
        tsk->signal = NULL;
        tsk->sighand = NULL;
        spin_unlock(&sighand->siglock);
@@ -127,7 +139,6 @@ static void __exit_signal(struct task_struct *tsk)
 
        __cleanup_sighand(sighand);
        clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
-       flush_sigqueue(&tsk->pending);
        if (sig) {
                flush_sigqueue(&sig->shared_pending);
                taskstats_tgid_free(sig);
@@ -160,7 +171,7 @@ repeat:
        zap_leader = 0;
        leader = p->group_leader;
        if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
-               BUG_ON(leader->exit_signal == -1);
+               BUG_ON(task_detached(leader));
                do_notify_parent(leader, leader->exit_signal);
                /*
                 * If we were the last child thread and the leader has
@@ -170,7 +181,7 @@ repeat:
                 * do_notify_parent() will have marked it self-reaping in
                 * that case.
                 */
-               zap_leader = (leader->exit_signal == -1);
+               zap_leader = task_detached(leader);
        }
 
        write_unlock_irq(&tasklist_lock);
@@ -329,13 +340,11 @@ void __set_special_pids(struct pid *pid)
        pid_t nr = pid_nr(pid);
 
        if (task_session(curr) != pid) {
-               detach_pid(curr, PIDTYPE_SID);
-               attach_pid(curr, PIDTYPE_SID, pid);
+               change_pid(curr, PIDTYPE_SID, pid);
                set_task_session(curr, nr);
        }
        if (task_pgrp(curr) != pid) {
-               detach_pid(curr, PIDTYPE_PGID);
-               attach_pid(curr, PIDTYPE_PGID, pid);
+               change_pid(curr, PIDTYPE_PGID, pid);
                set_task_pgrp(curr, nr);
        }
 }
@@ -557,6 +566,88 @@ void exit_fs(struct task_struct *tsk)
 
 EXPORT_SYMBOL_GPL(exit_fs);
 
+#ifdef CONFIG_MM_OWNER
+/*
+ * Task p is exiting and it owned mm, lets find a new owner for it
+ */
+static inline int
+mm_need_new_owner(struct mm_struct *mm, struct task_struct *p)
+{
+       /*
+        * If there are other users of the mm and the owner (us) is exiting
+        * we need to find a new owner to take on the responsibility.
+        */
+       if (!mm)
+               return 0;
+       if (atomic_read(&mm->mm_users) <= 1)
+               return 0;
+       if (mm->owner != p)
+               return 0;
+       return 1;
+}
+
+void mm_update_next_owner(struct mm_struct *mm)
+{
+       struct task_struct *c, *g, *p = current;
+
+retry:
+       if (!mm_need_new_owner(mm, p))
+               return;
+
+       read_lock(&tasklist_lock);
+       /*
+        * Search in the children
+        */
+       list_for_each_entry(c, &p->children, sibling) {
+               if (c->mm == mm)
+                       goto assign_new_owner;
+       }
+
+       /*
+        * Search in the siblings
+        */
+       list_for_each_entry(c, &p->parent->children, sibling) {
+               if (c->mm == mm)
+                       goto assign_new_owner;
+       }
+
+       /*
+        * Search through everything else. We should not get
+        * here often
+        */
+       do_each_thread(g, c) {
+               if (c->mm == mm)
+                       goto assign_new_owner;
+       } while_each_thread(g, c);
+
+       read_unlock(&tasklist_lock);
+       return;
+
+assign_new_owner:
+       BUG_ON(c == p);
+       get_task_struct(c);
+       /*
+        * The task_lock protects c->mm from changing.
+        * We always want mm->owner->mm == mm
+        */
+       task_lock(c);
+       /*
+        * Delay read_unlock() till we have the task_lock()
+        * to ensure that c does not slip away underneath us
+        */
+       read_unlock(&tasklist_lock);
+       if (c->mm != mm) {
+               task_unlock(c);
+               put_task_struct(c);
+               goto retry;
+       }
+       cgroup_mm_owner_callbacks(mm->owner, c);
+       mm->owner = c;
+       task_unlock(c);
+       put_task_struct(c);
+}
+#endif /* CONFIG_MM_OWNER */
+
 /*
  * Turn us into a lazy TLB process if we
  * aren't already..
@@ -596,6 +687,7 @@ static void exit_mm(struct task_struct * tsk)
        /* We don't want this task to be frozen prematurely */
        clear_freeze_flag(tsk);
        task_unlock(tsk);
+       mm_update_next_owner(mm);
        mmput(mm);
 }
 
@@ -610,7 +702,7 @@ reparent_thread(struct task_struct *p, struct task_struct *father, int traced)
        if (unlikely(traced)) {
                /* Preserve ptrace links if someone else is tracing this child.  */
                list_del_init(&p->ptrace_list);
-               if (p->parent != p->real_parent)
+               if (ptrace_reparented(p))
                        list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
        } else {
                /* If this child is being traced, then we're the one tracing it
@@ -634,18 +726,18 @@ reparent_thread(struct task_struct *p, struct task_struct *father, int traced)
        /* If this is a threaded reparent there is no need to
         * notify anyone anything has happened.
         */
-       if (p->real_parent->group_leader == father->group_leader)
+       if (same_thread_group(p->real_parent, father))
                return;
 
        /* We don't want people slaying init.  */
-       if (p->exit_signal != -1)
+       if (!task_detached(p))
                p->exit_signal = SIGCHLD;
 
        /* If we'd notified the old parent about this child's death,
         * also notify the new parent.
         */
        if (!traced && p->exit_state == EXIT_ZOMBIE &&
-           p->exit_signal != -1 && thread_group_empty(p))
+           !task_detached(p) && thread_group_empty(p))
                do_notify_parent(p, p->exit_signal);
 
        kill_orphaned_pgrp(p, father);
@@ -698,18 +790,18 @@ static void forget_original_parent(struct task_struct *father)
                } else {
                        /* reparent ptraced task to its real parent */
                        __ptrace_unlink (p);
-                       if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
+                       if (p->exit_state == EXIT_ZOMBIE && !task_detached(p) &&
                            thread_group_empty(p))
                                do_notify_parent(p, p->exit_signal);
                }
 
                /*
-                * if the ptraced child is a zombie with exit_signal == -1
-                * we must collect it before we exit, or it will remain
-                * zombie forever since we prevented it from self-reap itself
-                * while it was being traced by us, to be able to see it in wait4.
+                * if the ptraced child is a detached zombie we must collect
+                * it before we exit, or it will remain zombie forever since
+                * we prevented it from self-reap itself while it was being
+                * traced by us, to be able to see it in wait4.
                 */
-               if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
+               if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && task_detached(p)))
                        list_add(&p->ptrace_list, &ptrace_dead);
        }
 
@@ -766,29 +858,30 @@ static void exit_notify(struct task_struct *tsk, int group_dead)
         * we have changed execution domain as these two values started
         * the same after a fork.
         */
-       if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
+       if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) &&
            (tsk->parent_exec_id != tsk->real_parent->self_exec_id ||
-            tsk->self_exec_id != tsk->parent_exec_id)
-           && !capable(CAP_KILL))
+            tsk->self_exec_id != tsk->parent_exec_id) &&
+           !capable(CAP_KILL))
                tsk->exit_signal = SIGCHLD;
 
-
        /* If something other than our normal parent is ptracing us, then
         * send it a SIGCHLD instead of honoring exit_signal.  exit_signal
         * only has special meaning to our real parent.
         */
-       if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
-               int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
+       if (!task_detached(tsk) && thread_group_empty(tsk)) {
+               int signal = ptrace_reparented(tsk) ?
+                               SIGCHLD : tsk->exit_signal;
                do_notify_parent(tsk, signal);
        } else if (tsk->ptrace) {
                do_notify_parent(tsk, SIGCHLD);
        }
 
        state = EXIT_ZOMBIE;
-       if (tsk->exit_signal == -1 && likely(!tsk->ptrace))
+       if (task_detached(tsk) && likely(!tsk->ptrace))
                state = EXIT_DEAD;
        tsk->exit_state = state;
 
+       /* mt-exec, de_thread() is waiting for us */
        if (thread_group_leader(tsk) &&
            tsk->signal->notify_count < 0 &&
            tsk->signal->group_exit_task)
@@ -967,7 +1060,7 @@ NORET_TYPE void do_exit(long code)
        proc_exit_connector(tsk);
        exit_notify(tsk, group_dead);
 #ifdef CONFIG_NUMA
-       mpol_free(tsk->mempolicy);
+       mpol_put(tsk->mempolicy);
        tsk->mempolicy = NULL;
 #endif
 #ifdef CONFIG_FUTEX
@@ -1032,12 +1125,13 @@ asmlinkage long sys_exit(int error_code)
 NORET_TYPE void
 do_group_exit(int exit_code)
 {
+       struct signal_struct *sig = current->signal;
+
        BUG_ON(exit_code & 0x80); /* core dumps don't get here */
 
-       if (current->signal->flags & SIGNAL_GROUP_EXIT)
-               exit_code = current->signal->group_exit_code;
+       if (signal_group_exit(sig))
+               exit_code = sig->group_exit_code;
        else if (!thread_group_empty(current)) {
-               struct signal_struct *const sig = current->signal;
                struct sighand_struct *const sighand = current->sighand;
                spin_lock_irq(&sighand->siglock);
                if (signal_group_exit(sig))
@@ -1089,7 +1183,7 @@ static int eligible_child(enum pid_type type, struct pid *pid, int options,
         * Do not consider detached threads that are
         * not ptraced:
         */
-       if (p->exit_signal == -1 && !p->ptrace)
+       if (task_detached(p) && !p->ptrace)
                return 0;
 
        /* Wait for all children (clone and not) if __WALL is set;
@@ -1179,8 +1273,7 @@ static int wait_task_zombie(struct task_struct *p, int noreap,
                return 0;
        }
 
-       /* traced means p->ptrace, but not vice versa */
-       traced = (p->real_parent != p->parent);
+       traced = ptrace_reparented(p);
 
        if (likely(!traced)) {
                struct signal_struct *psig;
@@ -1281,9 +1374,9 @@ static int wait_task_zombie(struct task_struct *p, int noreap,
                 * If it's still not detached after that, don't release
                 * it now.
                 */
-               if (p->exit_signal != -1) {
+               if (!task_detached(p)) {
                        do_notify_parent(p, p->exit_signal);
-                       if (p->exit_signal != -1) {
+                       if (!task_detached(p)) {
                                p->exit_state = EXIT_ZOMBIE;
                                p = NULL;
                        }