Merge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
[pandora-kernel.git] / kernel / fork.c
index ac8100e..56e4e07 100644 (file)
@@ -11,7 +11,6 @@
  * management can be a bitch. See 'mm/memory.c': 'copy_page_range()'
  */
 
-#include <linux/config.h>
 #include <linux/slab.h>
 #include <linux/init.h>
 #include <linux/unistd.h>
@@ -104,6 +103,7 @@ static kmem_cache_t *mm_cachep;
 void free_task(struct task_struct *tsk)
 {
        free_thread_info(tsk->thread_info);
+       rt_mutex_debug_task_free(tsk);
        free_task_struct(tsk);
 }
 EXPORT_SYMBOL(free_task);
@@ -193,7 +193,10 @@ static inline int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
 
        down_write(&oldmm->mmap_sem);
        flush_cache_mm(oldmm);
-       down_write(&mm->mmap_sem);
+       /*
+        * Not linked in yet - no deadlock potential:
+        */
+       down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
 
        mm->locked_vm = 0;
        mm->mmap = NULL;
@@ -368,6 +371,8 @@ void fastcall __mmdrop(struct mm_struct *mm)
  */
 void mmput(struct mm_struct *mm)
 {
+       might_sleep();
+
        if (atomic_dec_and_test(&mm->mm_users)) {
                exit_aio(mm);
                exit_mmap(mm);
@@ -623,6 +628,7 @@ out:
 /*
  * Allocate a new files structure and copy contents from the
  * passed in files structure.
+ * errorp will be valid only when the returned files_struct is NULL.
  */
 static struct files_struct *dup_fd(struct files_struct *oldf, int *errorp)
 {
@@ -631,6 +637,7 @@ static struct files_struct *dup_fd(struct files_struct *oldf, int *errorp)
        int open_files, size, i, expand;
        struct fdtable *old_fdt, *new_fdt;
 
+       *errorp = -ENOMEM;
        newf = alloc_files();
        if (!newf)
                goto out;
@@ -744,7 +751,6 @@ static int copy_files(unsigned long clone_flags, struct task_struct * tsk)
         * break this.
         */
        tsk->files = NULL;
-       error = -ENOMEM;
        newf = dup_fd(oldf, &error);
        if (!newf)
                goto out;
@@ -871,6 +877,7 @@ static inline int copy_signal(unsigned long clone_flags, struct task_struct * ts
                tsk->it_prof_expires =
                        secs_to_cputime(sig->rlim[RLIMIT_CPU].rlim_cur);
        }
+       acct_init_pacct(&sig->pacct);
 
        return 0;
 }
@@ -909,6 +916,15 @@ asmlinkage long sys_set_tid_address(int __user *tidptr)
        return current->pid;
 }
 
+static inline void rt_mutex_init_task(struct task_struct *p)
+{
+#ifdef CONFIG_RT_MUTEXES
+       spin_lock_init(&p->pi_lock);
+       plist_head_init(&p->pi_waiters, &p->pi_lock);
+       p->pi_blocked_on = NULL;
+#endif
+}
+
 /*
  * This creates a new process as a copy of the old one,
  * but does not actually start it yet.
@@ -917,13 +933,13 @@ asmlinkage long sys_set_tid_address(int __user *tidptr)
  * parts of the process environment (as per the clone
  * flags). The actual kick-off is left to the caller.
  */
-static task_t *copy_process(unsigned long clone_flags,
-                                unsigned long stack_start,
-                                struct pt_regs *regs,
-                                unsigned long stack_size,
-                                int __user *parent_tidptr,
-                                int __user *child_tidptr,
-                                int pid)
+static struct task_struct *copy_process(unsigned long clone_flags,
+                                       unsigned long stack_start,
+                                       struct pt_regs *regs,
+                                       unsigned long stack_size,
+                                       int __user *parent_tidptr,
+                                       int __user *child_tidptr,
+                                       int pid)
 {
        int retval;
        struct task_struct *p = NULL;
@@ -955,6 +971,10 @@ static task_t *copy_process(unsigned long clone_flags,
        if (!p)
                goto fork_out;
 
+#ifdef CONFIG_TRACE_IRQFLAGS
+       DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
+       DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
+#endif
        retval = -EAGAIN;
        if (atomic_read(&p->user->processes) >=
                        p->signal->rlim[RLIMIT_NPROC].rlim_cur) {
@@ -989,13 +1009,10 @@ static task_t *copy_process(unsigned long clone_flags,
                if (put_user(p->pid, parent_tidptr))
                        goto bad_fork_cleanup;
 
-       p->proc_dentry = NULL;
-
        INIT_LIST_HEAD(&p->children);
        INIT_LIST_HEAD(&p->sibling);
        p->vfork_done = NULL;
        spin_lock_init(&p->alloc_lock);
-       spin_lock_init(&p->proc_lock);
 
        clear_tsk_thread_flag(p, TIF_SIGPENDING);
        init_sigpending(&p->pending);
@@ -1032,6 +1049,28 @@ static task_t *copy_process(unsigned long clone_flags,
        }
        mpol_fix_fork_child_flag(p);
 #endif
+#ifdef CONFIG_TRACE_IRQFLAGS
+       p->irq_events = 0;
+       p->hardirqs_enabled = 0;
+       p->hardirq_enable_ip = 0;
+       p->hardirq_enable_event = 0;
+       p->hardirq_disable_ip = _THIS_IP_;
+       p->hardirq_disable_event = 0;
+       p->softirqs_enabled = 1;
+       p->softirq_enable_ip = _THIS_IP_;
+       p->softirq_enable_event = 0;
+       p->softirq_disable_ip = 0;
+       p->softirq_disable_event = 0;
+       p->hardirq_context = 0;
+       p->softirq_context = 0;
+#endif
+#ifdef CONFIG_LOCKDEP
+       p->lockdep_depth = 0; /* no locks held yet */
+       p->curr_chain_key = 0;
+       p->lockdep_recursion = 0;
+#endif
+
+       rt_mutex_init_task(p);
 
 #ifdef CONFIG_DEBUG_MUTEXES
        p->blocked_on = NULL; /* not blocked yet */
@@ -1075,6 +1114,9 @@ static task_t *copy_process(unsigned long clone_flags,
 #ifdef CONFIG_COMPAT
        p->compat_robust_list = NULL;
 #endif
+       INIT_LIST_HEAD(&p->pi_state_list);
+       p->pi_state_cache = NULL;
+
        /*
         * sigaltstack should be cleared when sharing the same VM
         */
@@ -1155,18 +1197,6 @@ static task_t *copy_process(unsigned long clone_flags,
        }
 
        if (clone_flags & CLONE_THREAD) {
-               /*
-                * Important: if an exit-all has been started then
-                * do not create this new thread - the whole thread
-                * group is supposed to exit anyway.
-                */
-               if (current->signal->flags & SIGNAL_GROUP_EXIT) {
-                       spin_unlock(&current->sighand->siglock);
-                       write_unlock_irq(&tasklist_lock);
-                       retval = -EAGAIN;
-                       goto bad_fork_cleanup_namespace;
-               }
-
                p->group_leader = current->group_leader;
                list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group);
 
@@ -1264,9 +1294,9 @@ struct pt_regs * __devinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
        return regs;
 }
 
-task_t * __devinit fork_idle(int cpu)
+struct task_struct * __devinit fork_idle(int cpu)
 {
-       task_t *task;
+       struct task_struct *task;
        struct pt_regs regs;
 
        task = copy_process(CLONE_VM, 0, idle_regs(&regs), 0, NULL, NULL, 0);