Fix small typo in drivers/serial/icom.c
[safe/jmp/linux-2.6] / kernel / exit.c
index 22399ca..122fadb 100644 (file)
@@ -4,7 +4,6 @@
  *  Copyright (C) 1991, 1992  Linus Torvalds
  */
 
-#include <linux/config.h>
 #include <linux/mm.h>
 #include <linux/slab.h>
 #include <linux/interrupt.h>
 #include <linux/completion.h>
 #include <linux/personality.h>
 #include <linux/tty.h>
-#include <linux/namespace.h>
+#include <linux/mnt_namespace.h>
 #include <linux/key.h>
 #include <linux/security.h>
 #include <linux/cpu.h>
 #include <linux/acct.h>
+#include <linux/tsacct_kern.h>
 #include <linux/file.h>
 #include <linux/binfmts.h>
+#include <linux/nsproxy.h>
+#include <linux/pid_namespace.h>
 #include <linux/ptrace.h>
 #include <linux/profile.h>
 #include <linux/mount.h>
 #include <linux/proc_fs.h>
 #include <linux/mempolicy.h>
+#include <linux/taskstats_kern.h>
+#include <linux/delayacct.h>
 #include <linux/cpuset.h>
 #include <linux/syscalls.h>
 #include <linux/signal.h>
 #include <linux/mutex.h>
 #include <linux/futex.h>
 #include <linux/compat.h>
+#include <linux/pipe_fs_i.h>
+#include <linux/audit.h> /* for audit_free() */
+#include <linux/resource.h>
+#include <linux/blkdev.h>
 
 #include <asm/uaccess.h>
 #include <asm/unistd.h>
@@ -41,9 +49,6 @@
 #include <asm/mmu_context.h>
 
 extern void sem_exit (void);
-extern struct task_struct *child_reaper;
-
-int getrusage(struct task_struct *, int, struct rusage __user *);
 
 static void exit_mm(struct task_struct * tsk);
 
@@ -55,7 +60,7 @@ static void __unhash_process(struct task_struct *p)
                detach_pid(p, PIDTYPE_PGID);
                detach_pid(p, PIDTYPE_SID);
 
-               list_del_init(&p->tasks);
+               list_del_rcu(&p->tasks);
                __get_cpu_var(process_counts)--;
        }
        list_del_rcu(&p->thread_group);
@@ -114,28 +119,31 @@ static void __exit_signal(struct task_struct *tsk)
        __unhash_process(tsk);
 
        tsk->signal = NULL;
-       cleanup_sighand(tsk);
+       tsk->sighand = NULL;
        spin_unlock(&sighand->siglock);
        rcu_read_unlock();
 
+       __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);
                __cleanup_signal(sig);
        }
 }
 
+static void delayed_put_task_struct(struct rcu_head *rhp)
+{
+       put_task_struct(container_of(rhp, struct task_struct, rcu));
+}
+
 void release_task(struct task_struct * p)
 {
+       struct task_struct *leader;
        int zap_leader;
-       task_t *leader;
-       struct dentry *proc_dentry;
-
 repeat:
        atomic_dec(&p->user->processes);
-       spin_lock(&p->proc_lock);
-       proc_dentry = proc_pid_unhash(p);
        write_lock_irq(&tasklist_lock);
        ptrace_unlink(p);
        BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
@@ -164,10 +172,9 @@ repeat:
 
        sched_exit(p);
        write_unlock_irq(&tasklist_lock);
-       spin_unlock(&p->proc_lock);
-       proc_pid_flush(proc_dentry);
+       proc_flush_task(p);
        release_thread(p);
-       put_task_struct(p);
+       call_rcu(&p->rcu, delayed_put_task_struct);
 
        p = leader;
        if (unlikely(zap_leader))
@@ -182,21 +189,18 @@ repeat:
 int session_of_pgrp(int pgrp)
 {
        struct task_struct *p;
-       int sid = -1;
+       int sid = 0;
 
        read_lock(&tasklist_lock);
-       do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
-               if (p->signal->session > 0) {
-                       sid = p->signal->session;
-                       goto out;
-               }
-       } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
-       p = find_task_by_pid(pgrp);
-       if (p)
-               sid = p->signal->session;
-out:
+
+       p = find_task_by_pid_type(PIDTYPE_PGID, pgrp);
+       if (p == NULL)
+               p = find_task_by_pid(pgrp);
+       if (p != NULL)
+               sid = process_session(p);
+
        read_unlock(&tasklist_lock);
-       
+
        return sid;
 }
 
@@ -208,7 +212,7 @@ out:
  *
  * "I ask you, have you ever known what it is to be an orphan?"
  */
-static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
+static int will_become_orphaned_pgrp(int pgrp, struct task_struct *ignored_task)
 {
        struct task_struct *p;
        int ret = 1;
@@ -216,10 +220,10 @@ static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
        do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
                if (p == ignored_task
                                || p->exit_state
-                               || p->real_parent->pid == 1)
+                               || is_init(p->real_parent))
                        continue;
-               if (process_group(p->real_parent) != pgrp
-                           && p->real_parent->signal->session == p->signal->session) {
+               if (process_group(p->real_parent) != pgrp &&
+                   process_session(p->real_parent) == process_session(p)) {
                        ret = 0;
                        break;
                }
@@ -246,17 +250,6 @@ static int has_stopped_jobs(int pgrp)
        do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
                if (p->state != TASK_STOPPED)
                        continue;
-
-               /* If p is stopped by a debugger on a signal that won't
-                  stop it, then don't count p as stopped.  This isn't
-                  perfect but it's a good approximation.  */
-               if (unlikely (p->ptrace)
-                   && p->exit_code != SIGSTOP
-                   && p->exit_code != SIGTSTP
-                   && p->exit_code != SIGTTOU
-                   && p->exit_code != SIGTTIN)
-                       continue;
-
                retval = 1;
                break;
        } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
@@ -264,7 +257,8 @@ static int has_stopped_jobs(int pgrp)
 }
 
 /**
- * reparent_to_init - Reparent the calling kernel thread to the init task.
+ * reparent_to_init - Reparent the calling kernel thread to the init task
+ * of the pid space that the thread belongs to.
  *
  * If a kernel thread is launched as a result of a system call, or if
  * it ever exits, it should generally reparent itself to init so that
@@ -282,16 +276,14 @@ static void reparent_to_init(void)
        ptrace_unlink(current);
        /* Reparent to init */
        remove_parent(current);
-       current->parent = child_reaper;
-       current->real_parent = child_reaper;
+       current->parent = child_reaper(current);
+       current->real_parent = child_reaper(current);
        add_parent(current);
 
        /* Set the exit signal to SIGCHLD so we signal init on exit */
        current->exit_signal = SIGCHLD;
 
-       if ((current->policy == SCHED_NORMAL ||
-                       current->policy == SCHED_BATCH)
-                               && (task_nice(current) < 0))
+       if (!has_rt_policy(current) && (task_nice(current) < 0))
                set_user_nice(current, 0);
        /* cpus_allowed? */
        /* rt_priority? */
@@ -308,9 +300,9 @@ void __set_special_pids(pid_t session, pid_t pgrp)
 {
        struct task_struct *curr = current->group_leader;
 
-       if (curr->signal->session != session) {
+       if (process_session(curr) != session) {
                detach_pid(curr, PIDTYPE_SID);
-               curr->signal->session = session;
+               set_signal_session(curr->signal, session);
                attach_pid(curr, PIDTYPE_SID, session);
        }
        if (process_group(curr) != pgrp) {
@@ -320,7 +312,7 @@ void __set_special_pids(pid_t session, pid_t pgrp)
        }
 }
 
-void set_special_pids(pid_t session, pid_t pgrp)
+static void set_special_pids(pid_t session, pid_t pgrp)
 {
        write_lock_irq(&tasklist_lock);
        __set_special_pids(session, pgrp);
@@ -390,9 +382,7 @@ void daemonize(const char *name, ...)
        exit_mm(current);
 
        set_special_pids(1, 1);
-       mutex_lock(&tty_mutex);
-       current->signal->tty = NULL;
-       mutex_unlock(&tty_mutex);
+       proc_clear_tty(current);
 
        /* Block and flush all signals */
        sigfillset(&blocked);
@@ -405,9 +395,11 @@ void daemonize(const char *name, ...)
        fs = init_task.fs;
        current->fs = fs;
        atomic_inc(&fs->count);
-       exit_namespace(current);
-       current->namespace = init_task.namespace;
-       get_namespace(current->namespace);
+
+       exit_task_namespaces(current);
+       current->nsproxy = init_task.nsproxy;
+       get_task_namespaces(current);
+
        exit_files(current);
        current->files = init_task.files;
        atomic_inc(&current->files->count);
@@ -433,7 +425,7 @@ static void close_files(struct files_struct * files)
        for (;;) {
                unsigned long set;
                i = j * __NFDBITS;
-               if (i >= fdt->max_fdset || i >= fdt->max_fds)
+               if (i >= fdt->max_fds)
                        break;
                set = fdt->open_fds->fds_bits[j++];
                while (set) {
@@ -474,16 +466,26 @@ void fastcall put_files_struct(struct files_struct *files)
                 * you can free files immediately.
                 */
                fdt = files_fdtable(files);
-               if (fdt == &files->fdtab)
-                       fdt->free_files = files;
-               else
+               if (fdt != &files->fdtab)
                        kmem_cache_free(files_cachep, files);
-               free_fdtable(fdt);
+               call_rcu(&fdt->rcu, free_fdtable_rcu);
        }
 }
 
 EXPORT_SYMBOL(put_files_struct);
 
+void reset_files_struct(struct task_struct *tsk, struct files_struct *files)
+{
+       struct files_struct *old;
+
+       old = tsk->files;
+       task_lock(tsk);
+       tsk->files = files;
+       task_unlock(tsk);
+       put_files_struct(old);
+}
+EXPORT_SYMBOL(reset_files_struct);
+
 static inline void __exit_files(struct task_struct *tsk)
 {
        struct files_struct * files = tsk->files;
@@ -571,7 +573,7 @@ static void exit_mm(struct task_struct * tsk)
                down_read(&mm->mmap_sem);
        }
        atomic_inc(&mm->mm_count);
-       if (mm != tsk->active_mm) BUG();
+       BUG_ON(mm != tsk->active_mm);
        /* more a memory barrier than a real lock */
        task_lock(tsk);
        tsk->mm = NULL;
@@ -581,7 +583,8 @@ static void exit_mm(struct task_struct * tsk)
        mmput(mm);
 }
 
-static inline void choose_new_parent(task_t *p, task_t *reaper)
+static inline void
+choose_new_parent(struct task_struct *p, struct task_struct *reaper)
 {
        /*
         * Make sure we're not reparenting to ourselves and that
@@ -591,7 +594,8 @@ static inline void choose_new_parent(task_t *p, task_t *reaper)
        p->real_parent = reaper;
 }
 
-static void reparent_thread(task_t *p, task_t *father, int traced)
+static void
+reparent_thread(struct task_struct *p, struct task_struct *father, int traced)
 {
        /* We don't want people slaying init.  */
        if (p->exit_signal != -1)
@@ -639,10 +643,11 @@ static void reparent_thread(task_t *p, task_t *father, int traced)
         * outside, so the child pgrp is now orphaned.
         */
        if ((process_group(p) != process_group(father)) &&
-           (p->signal->session == father->signal->session)) {
+           (process_session(p) == process_session(father))) {
                int pgrp = process_group(p);
 
-               if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
+               if (will_become_orphaned_pgrp(pgrp, NULL) &&
+                   has_stopped_jobs(pgrp)) {
                        __kill_pg_info(SIGHUP, SEND_SIG_PRIV, pgrp);
                        __kill_pg_info(SIGCONT, SEND_SIG_PRIV, pgrp);
                }
@@ -653,10 +658,11 @@ static void reparent_thread(task_t *p, task_t *father, int traced)
  * When we die, we re-parent all our children.
  * Try to give them to another thread in our thread
  * group, and if no such member exists, give it to
- * the global child reaper process (ie "init")
+ * the child reaper process (ie "init") in our pid
+ * space.
  */
-static void forget_original_parent(struct task_struct * father,
-                                         struct list_head *to_release)
+static void
+forget_original_parent(struct task_struct *father, struct list_head *to_release)
 {
        struct task_struct *p, *reaper = father;
        struct list_head *_p, *_n;
@@ -664,7 +670,7 @@ static void forget_original_parent(struct task_struct * father,
        do {
                reaper = next_thread(reaper);
                if (reaper == father) {
-                       reaper = child_reaper;
+                       reaper = child_reaper(father);
                        break;
                }
        } while (reaper->exit_state);
@@ -679,7 +685,7 @@ static void forget_original_parent(struct task_struct * father,
         */
        list_for_each_safe(_p, _n, &father->children) {
                int ptrace;
-               p = list_entry(_p,struct task_struct,sibling);
+               p = list_entry(_p, struct task_struct, sibling);
 
                ptrace = p->ptrace;
 
@@ -708,7 +714,7 @@ static void forget_original_parent(struct task_struct * father,
                        list_add(&p->ptrace_list, to_release);
        }
        list_for_each_safe(_p, _n, &father->ptrace_children) {
-               p = list_entry(_p,struct task_struct,ptrace_list);
+               p = list_entry(_p, struct task_struct, ptrace_list);
                choose_new_parent(p, reaper);
                reparent_thread(p, father, 1);
        }
@@ -776,7 +782,7 @@ static void exit_notify(struct task_struct *tsk)
        t = tsk->real_parent;
        
        if ((process_group(t) != process_group(tsk)) &&
-           (t->signal->session == tsk->signal->session) &&
+           (process_session(t) == process_session(tsk)) &&
            will_become_orphaned_pgrp(process_group(tsk), tsk) &&
            has_stopped_jobs(process_group(tsk))) {
                __kill_pg_info(SIGHUP, SEND_SIG_PRIV, process_group(tsk));
@@ -828,7 +834,7 @@ static void exit_notify(struct task_struct *tsk)
 
        list_for_each_safe(_p, _n, &ptrace_dead) {
                list_del_init(_p);
-               t = list_entry(_p,struct task_struct,ptrace_list);
+               t = list_entry(_p, struct task_struct, ptrace_list);
                release_task(t);
        }
 
@@ -850,8 +856,13 @@ fastcall NORET_TYPE void do_exit(long code)
                panic("Aiee, killing interrupt handler!");
        if (unlikely(!tsk->pid))
                panic("Attempted to kill the idle task!");
-       if (unlikely(tsk == child_reaper))
-               panic("Attempted to kill init!");
+       if (unlikely(tsk == child_reaper(tsk))) {
+               if (tsk->nsproxy->pid_ns != &init_pid_ns)
+                       tsk->nsproxy->pid_ns->child_reaper = init_pid_ns.child_reaper;
+               else
+                       panic("Attempted to kill init!");
+       }
+
 
        if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
                current->ptrace_message = code;
@@ -873,14 +884,6 @@ fastcall NORET_TYPE void do_exit(long code)
 
        tsk->flags |= PF_EXITING;
 
-       /*
-        * Make sure we don't try to process any timer firings
-        * while we are already exiting.
-        */
-       tsk->it_virt_expires = cputime_zero;
-       tsk->it_prof_expires = cputime_zero;
-       tsk->it_sched_expires = 0;
-
        if (unlikely(in_atomic()))
                printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
                                current->comm, current->pid,
@@ -895,20 +898,26 @@ fastcall NORET_TYPE void do_exit(long code)
        if (group_dead) {
                hrtimer_cancel(&tsk->signal->real_timer);
                exit_itimers(tsk->signal);
-               acct_process(code);
        }
+       acct_collect(code, group_dead);
        if (unlikely(tsk->robust_list))
                exit_robust_list(tsk);
-#ifdef CONFIG_COMPAT
+#if defined(CONFIG_FUTEX) && defined(CONFIG_COMPAT)
        if (unlikely(tsk->compat_robust_list))
                compat_exit_robust_list(tsk);
 #endif
+       if (unlikely(tsk->audit_context))
+               audit_free(tsk);
+
+       taskstats_exit(tsk, group_dead);
+
        exit_mm(tsk);
 
+       if (group_dead)
+               acct_process();
        exit_sem(tsk);
        __exit_files(tsk);
        __exit_fs(tsk);
-       exit_namespace(tsk);
        exit_thread();
        cpuset_exit(tsk);
        exit_keys(tsk);
@@ -923,27 +932,39 @@ fastcall NORET_TYPE void do_exit(long code)
        tsk->exit_code = code;
        proc_exit_connector(tsk);
        exit_notify(tsk);
+       exit_task_namespaces(tsk);
 #ifdef CONFIG_NUMA
        mpol_free(tsk->mempolicy);
        tsk->mempolicy = NULL;
 #endif
        /*
-        * If DEBUG_MUTEXES is on, make sure we are holding no locks:
+        * This must happen late, after the PID is not
+        * hashed anymore:
+        */
+       if (unlikely(!list_empty(&tsk->pi_state_list)))
+               exit_pi_state_list(tsk);
+       if (unlikely(current->pi_state_cache))
+               kfree(current->pi_state_cache);
+       /*
+        * Make sure we are holding no locks:
         */
-       mutex_debug_check_no_locks_held(tsk);
+       debug_check_no_locks_held(tsk);
 
        if (tsk->io_context)
                exit_io_context();
 
-       /* PF_DEAD causes final put_task_struct after we schedule. */
+       if (tsk->splice_pipe)
+               __free_pipe_info(tsk->splice_pipe);
+
        preempt_disable();
-       BUG_ON(tsk->flags & PF_DEAD);
-       tsk->flags |= PF_DEAD;
+       /* causes final put_task_struct in finish_task_switch(). */
+       tsk->state = TASK_DEAD;
 
        schedule();
        BUG();
        /* Avoid "noreturn function does return".  */
-       for (;;) ;
+       for (;;)
+               cpu_relax();    /* For when BUG is null */
 }
 
 EXPORT_SYMBOL_GPL(do_exit);
@@ -952,7 +973,7 @@ NORET_TYPE void complete_and_exit(struct completion *comp, long code)
 {
        if (comp)
                complete(comp);
-       
+
        do_exit(code);
 }
 
@@ -1002,7 +1023,7 @@ asmlinkage void sys_exit_group(int error_code)
        do_group_exit((error_code & 0xff) << 8);
 }
 
-static int eligible_child(pid_t pid, int options, task_t *p)
+static int eligible_child(pid_t pid, int options, struct task_struct *p)
 {
        if (pid > 0) {
                if (p->pid != pid)
@@ -1034,7 +1055,7 @@ static int eligible_child(pid_t pid, int options, task_t *p)
         * Do not consider thread group leaders that are
         * in a non-empty thread group:
         */
-       if (current->tgid != p->tgid && delay_group_leader(p))
+       if (delay_group_leader(p))
                return 2;
 
        if (security_task_wait(p))
@@ -1043,12 +1064,13 @@ static int eligible_child(pid_t pid, int options, task_t *p)
        return 1;
 }
 
-static int wait_noreap_copyout(task_t *p, pid_t pid, uid_t uid,
+static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid,
                               int why, int status,
                               struct siginfo __user *infop,
                               struct rusage __user *rusagep)
 {
        int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
+
        put_task_struct(p);
        if (!retval)
                retval = put_user(SIGCHLD, &infop->si_signo);
@@ -1073,7 +1095,7 @@ static int wait_noreap_copyout(task_t *p, pid_t pid, uid_t uid,
  * the lock and this task is uninteresting.  If we return nonzero, we have
  * released the lock and the system call should return.
  */
-static int wait_task_zombie(task_t *p, int noreap,
+static int wait_task_zombie(struct task_struct *p, int noreap,
                            struct siginfo __user *infop,
                            int __user *stat_addr, struct rusage __user *ru)
 {
@@ -1235,8 +1257,8 @@ static int wait_task_zombie(task_t *p, int noreap,
  * the lock and this task is uninteresting.  If we return nonzero, we have
  * released the lock and the system call should return.
  */
-static int wait_task_stopped(task_t *p, int delayed_group_leader, int noreap,
-                            struct siginfo __user *infop,
+static int wait_task_stopped(struct task_struct *p, int delayed_group_leader,
+                            int noreap, struct siginfo __user *infop,
                             int __user *stat_addr, struct rusage __user *ru)
 {
        int retval, exit_code;
@@ -1350,7 +1372,7 @@ bail_ref:
  * the lock and this task is uninteresting.  If we return nonzero, we have
  * released the lock and the system call should return.
  */
-static int wait_task_continued(task_t *p, int noreap,
+static int wait_task_continued(struct task_struct *p, int noreap,
                               struct siginfo __user *infop,
                               int __user *stat_addr, struct rusage __user *ru)
 {
@@ -1436,7 +1458,7 @@ repeat:
                int ret;
 
                list_for_each(_p,&tsk->children) {
-                       p = list_entry(_p,struct task_struct,sibling);
+                       p = list_entry(_p, struct task_struct, sibling);
 
                        ret = eligible_child(pid, options, p);
                        if (!ret)
@@ -1525,8 +1547,7 @@ check_continued:
                if (options & __WNOTHREAD)
                        break;
                tsk = next_thread(tsk);
-               if (tsk->signal != current->signal)
-                       BUG();
+               BUG_ON(tsk->signal != current->signal);
        } while (tsk != current);
 
        read_unlock(&tasklist_lock);