upgrade to linux 2.6.10-1.12_FC2
[linux-2.6.git] / kernel / fork.c
index 7430f87..d19d14e 100644 (file)
@@ -24,6 +24,7 @@
 #include <linux/mempolicy.h>
 #include <linux/sem.h>
 #include <linux/file.h>
+#include <linux/key.h>
 #include <linux/binfmts.h>
 #include <linux/mman.h>
 #include <linux/fs.h>
@@ -108,131 +109,6 @@ void __put_task_struct(struct task_struct *tsk)
                free_task(tsk);
 }
 
-void fastcall add_wait_queue(wait_queue_head_t *q, wait_queue_t * wait)
-{
-       unsigned long flags;
-
-       wait->flags &= ~WQ_FLAG_EXCLUSIVE;
-       spin_lock_irqsave(&q->lock, flags);
-       __add_wait_queue(q, wait);
-       spin_unlock_irqrestore(&q->lock, flags);
-}
-
-EXPORT_SYMBOL(add_wait_queue);
-
-void fastcall add_wait_queue_exclusive(wait_queue_head_t *q, wait_queue_t * wait)
-{
-       unsigned long flags;
-
-       wait->flags |= WQ_FLAG_EXCLUSIVE;
-       spin_lock_irqsave(&q->lock, flags);
-       __add_wait_queue_tail(q, wait);
-       spin_unlock_irqrestore(&q->lock, flags);
-}
-
-EXPORT_SYMBOL(add_wait_queue_exclusive);
-
-void fastcall remove_wait_queue(wait_queue_head_t *q, wait_queue_t * wait)
-{
-       unsigned long flags;
-
-       spin_lock_irqsave(&q->lock, flags);
-       __remove_wait_queue(q, wait);
-       spin_unlock_irqrestore(&q->lock, flags);
-}
-
-EXPORT_SYMBOL(remove_wait_queue);
-
-
-/*
- * Note: we use "set_current_state()" _after_ the wait-queue add,
- * because we need a memory barrier there on SMP, so that any
- * wake-function that tests for the wait-queue being active
- * will be guaranteed to see waitqueue addition _or_ subsequent
- * tests in this thread will see the wakeup having taken place.
- *
- * The spin_unlock() itself is semi-permeable and only protects
- * one way (it only protects stuff inside the critical region and
- * stops them from bleeding out - it would still allow subsequent
- * loads to move into the the critical region).
- */
-void fastcall prepare_to_wait(wait_queue_head_t *q, wait_queue_t *wait, int state)
-{
-       unsigned long flags;
-
-       wait->flags &= ~WQ_FLAG_EXCLUSIVE;
-       spin_lock_irqsave(&q->lock, flags);
-       if (list_empty(&wait->task_list))
-               __add_wait_queue(q, wait);
-       /*
-        * don't alter the task state if this is just going to
-        * queue an async wait queue callback
-        */
-       if (is_sync_wait(wait))
-               set_current_state(state);
-       spin_unlock_irqrestore(&q->lock, flags);
-}
-
-EXPORT_SYMBOL(prepare_to_wait);
-
-void fastcall
-prepare_to_wait_exclusive(wait_queue_head_t *q, wait_queue_t *wait, int state)
-{
-       unsigned long flags;
-
-       wait->flags |= WQ_FLAG_EXCLUSIVE;
-       spin_lock_irqsave(&q->lock, flags);
-       if (list_empty(&wait->task_list))
-               __add_wait_queue_tail(q, wait);
-       /*
-        * don't alter the task state if this is just going to
-        * queue an async wait queue callback
-        */
-       if (is_sync_wait(wait))
-               set_current_state(state);
-       spin_unlock_irqrestore(&q->lock, flags);
-}
-
-EXPORT_SYMBOL(prepare_to_wait_exclusive);
-
-void fastcall finish_wait(wait_queue_head_t *q, wait_queue_t *wait)
-{
-       unsigned long flags;
-
-       __set_current_state(TASK_RUNNING);
-       /*
-        * We can check for list emptiness outside the lock
-        * IFF:
-        *  - we use the "careful" check that verifies both
-        *    the next and prev pointers, so that there cannot
-        *    be any half-pending updates in progress on other
-        *    CPU's that we haven't seen yet (and that might
-        *    still change the stack area.
-        * and
-        *  - all other users take the lock (ie we can only
-        *    have _one_ other CPU that looks at or modifies
-        *    the list).
-        */
-       if (!list_empty_careful(&wait->task_list)) {
-               spin_lock_irqsave(&q->lock, flags);
-               list_del_init(&wait->task_list);
-               spin_unlock_irqrestore(&q->lock, flags);
-       }
-}
-
-EXPORT_SYMBOL(finish_wait);
-
-int autoremove_wake_function(wait_queue_t *wait, unsigned mode, int sync, void *key)
-{
-       int ret = default_wake_function(wait, mode, sync, key);
-
-       if (ret)
-               list_del_init(&wait->task_list);
-       return ret;
-}
-
-EXPORT_SYMBOL(autoremove_wake_function);
-
 void __init fork_init(unsigned long mempages)
 {
 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
@@ -250,15 +126,16 @@ void __init fork_init(unsigned long mempages)
         * value: the thread structures can take up at most half
         * of memory.
         */
-       max_threads = mempages / (THREAD_SIZE/PAGE_SIZE) / 8;
+       max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE);
+
        /*
         * we need to allow at least 20 threads to boot a system
         */
        if(max_threads < 20)
                max_threads = 20;
 
-       init_task.rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
-       init_task.rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
+       init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
+       init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
 }
 
 static struct task_struct *dup_task_struct(struct task_struct *orig)
@@ -316,17 +193,6 @@ static inline int dup_mmap(struct mm_struct * mm, struct mm_struct * oldmm)
        rb_parent = NULL;
        pprev = &mm->mmap;
 
-       /*
-        * Add it to the mmlist after the parent.
-        * Doing it this way means that we can order the list,
-        * and fork() won't mess up the ordering significantly.
-        * Add it first so that swapoff can see any swap entries.
-        */
-       spin_lock(&mmlist_lock);
-       list_add(&mm->mmlist, &current->mm->mmlist);
-       mmlist_nr++;
-       spin_unlock(&mmlist_lock);
-
        for (mpnt = current->mm->mmap ; mpnt ; mpnt = mpnt->vm_next) {
                struct file *file;
 
@@ -426,7 +292,6 @@ static inline void mm_free_pgd(struct mm_struct * mm)
 #endif /* CONFIG_MMU */
 
 spinlock_t mmlist_lock __cacheline_aligned_in_smp = SPIN_LOCK_UNLOCKED;
-int mmlist_nr;
 
 #define allocate_mm()  (kmem_cache_alloc(mm_cachep, SLAB_KERNEL))
 #define free_mm(mm)    (kmem_cache_free(mm_cachep, (mm)))
@@ -438,9 +303,11 @@ static struct mm_struct * mm_init(struct mm_struct * mm)
        atomic_set(&mm->mm_users, 1);
        atomic_set(&mm->mm_count, 1);
        init_rwsem(&mm->mmap_sem);
+       INIT_LIST_HEAD(&mm->mmlist);
        mm->core_waiters = 0;
-       mm->page_table_lock = SPIN_LOCK_UNLOCKED;
-       mm->ioctx_list_lock = RW_LOCK_UNLOCKED;
+       mm->nr_ptes = 0;
+       spin_lock_init(&mm->page_table_lock);
+       rwlock_init(&mm->ioctx_list_lock);
        mm->ioctx_list = NULL;
        mm->default_kioctx = (struct kioctx)INIT_KIOCTX(mm->default_kioctx, *mm);
        mm->free_area_cache = TASK_UNMAPPED_BASE;
@@ -503,12 +370,14 @@ void fastcall __mmdrop(struct mm_struct *mm)
  */
 void mmput(struct mm_struct *mm)
 {
-       if (atomic_dec_and_lock(&mm->mm_users, &mmlist_lock)) {
-               list_del(&mm->mmlist);
-               mmlist_nr--;
-               spin_unlock(&mmlist_lock);
+       if (atomic_dec_and_test(&mm->mm_users)) {
                exit_aio(mm);
                exit_mmap(mm);
+               if (!list_empty(&mm->mmlist)) {
+                       spin_lock(&mmlist_lock);
+                       list_del(&mm->mmlist);
+                       spin_unlock(&mmlist_lock);
+               }
                put_swap_token(mm);
                mmdrop(mm);
        }
@@ -518,15 +387,11 @@ EXPORT_SYMBOL_GPL(mmput);
 /**
  * get_task_mm - acquire a reference to the task's mm
  *
- * Returns %NULL if the task has no mm.  Checks if the use count
- * of the mm is non-zero and if so returns a reference to it, after
+ * Returns %NULL if the task has no mm.  Checks PF_BORROWED_MM (meaning
+ * this kernel workthread has transiently adopted a user mm with use_mm,
+ * to do its AIO) is not set and if so returns a reference to it, after
  * bumping up the use count.  User must release the mm via mmput()
  * after use.  Typically used by /proc and ptrace.
- *
- * If the use count is zero, it means that this mm is going away,
- * so return %NULL.  This only happens in the case of an AIO daemon
- * which has temporarily adopted an mm (see use_mm), in the course
- * of its final mmput, before exit_aio has completed.
  */
 struct mm_struct *get_task_mm(struct task_struct *task)
 {
@@ -535,12 +400,10 @@ struct mm_struct *get_task_mm(struct task_struct *task)
        task_lock(task);
        mm = task->mm;
        if (mm) {
-               spin_lock(&mmlist_lock);
-               if (!atomic_read(&mm->mm_users))
+               if (task->flags & PF_BORROWED_MM)
                        mm = NULL;
                else
                        atomic_inc(&mm->mm_users);
-               spin_unlock(&mmlist_lock);
        }
        task_unlock(task);
        return mm;
@@ -663,7 +526,7 @@ static inline struct fs_struct *__copy_fs_struct(struct fs_struct *old)
        /* We don't need to lock fs - think why ;-) */
        if (fs) {
                atomic_set(&fs->count, 1);
-               fs->lock = RW_LOCK_UNLOCKED;
+               rwlock_init(&fs->lock);
                fs->umask = old->umask;
                read_lock(&old->lock);
                fs->rootmnt = mntget(old->rootmnt);
@@ -745,7 +608,7 @@ static int copy_files(unsigned long clone_flags, struct task_struct * tsk)
 
        atomic_set(&newf->count, 1);
 
-       newf->file_lock     = SPIN_LOCK_UNLOCKED;
+       spin_lock_init(&newf->file_lock);
        newf->next_fd       = 0;
        newf->max_fds       = NR_OPEN_DEFAULT;
        newf->max_fdset     = __FD_SETSIZE;
@@ -794,8 +657,17 @@ static int copy_files(unsigned long clone_flags, struct task_struct * tsk)
 
        for (i = open_files; i != 0; i--) {
                struct file *f = *old_fds++;
-               if (f)
+               if (f) {
                        get_file(f);
+               } else {
+                       /*
+                        * The fd may be claimed in the fd bitmap but not yet
+                        * instantiated in the files array if a sibling thread
+                        * is partway through open().  So make sure that this
+                        * fd is available to the new process.
+                        */
+                       FD_CLR(open_files - i, newf->open_fds);
+               }
                *new_fds++ = f;
        }
        spin_unlock(&oldf->file_lock);
@@ -879,6 +751,7 @@ static inline int copy_signal(unsigned long clone_flags, struct task_struct * ts
 
        if (clone_flags & CLONE_THREAD) {
                atomic_inc(&current->signal->count);
+               atomic_inc(&current->signal->live);
                return 0;
        }
        sig = kmem_cache_alloc(signal_cachep, GFP_KERNEL);
@@ -886,10 +759,12 @@ static inline int copy_signal(unsigned long clone_flags, struct task_struct * ts
        if (!sig)
                return -ENOMEM;
        atomic_set(&sig->count, 1);
+       atomic_set(&sig->live, 1);
        sig->group_exit = 0;
        sig->group_exit_code = 0;
        sig->group_exit_task = NULL;
        sig->group_stop_count = 0;
+       sig->stop_state = 0;
        sig->curr_target = NULL;
        init_sigpending(&sig->shared_pending);
        INIT_LIST_HEAD(&sig->posix_timers);
@@ -904,6 +779,10 @@ static inline int copy_signal(unsigned long clone_flags, struct task_struct * ts
        sig->nvcsw = sig->nivcsw = sig->cnvcsw = sig->cnivcsw = 0;
        sig->min_flt = sig->maj_flt = sig->cmin_flt = sig->cmaj_flt = 0;
 
+       task_lock(current->group_leader);
+       memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
+       task_unlock(current->group_leader);
+
        return 0;
 }
 
@@ -995,7 +874,7 @@ static task_t *copy_process(unsigned long clone_flags,
                goto bad_fork_cleanup_vm;
 
        if (atomic_read(&p->user->processes) >=
-                       p->rlim[RLIMIT_NPROC].rlim_cur) {
+                       p->signal->rlim[RLIMIT_NPROC].rlim_cur) {
                if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) &&
                                p->user != &root_user)
                        goto bad_fork_cleanup_vm;
@@ -1061,6 +940,10 @@ static task_t *copy_process(unsigned long clone_flags,
        }
 #endif
 
+       p->tgid = p->pid;
+       if (clone_flags & CLONE_THREAD)
+               p->tgid = current->tgid;
+
        if ((retval = security_task_alloc(p)))
                goto bad_fork_cleanup_policy;
        if ((retval = audit_alloc(p)))
@@ -1078,8 +961,10 @@ static task_t *copy_process(unsigned long clone_flags,
                goto bad_fork_cleanup_sighand;
        if ((retval = copy_mm(clone_flags, p)))
                goto bad_fork_cleanup_signal;
-       if ((retval = copy_namespace(clone_flags, p)))
+       if ((retval = copy_keys(clone_flags, p)))
                goto bad_fork_cleanup_mm;
+       if ((retval = copy_namespace(clone_flags, p)))
+               goto bad_fork_cleanup_keys;
        retval = copy_thread(0, clone_flags, stack_start, stack_size, p, regs);
        if (retval)
                goto bad_fork_cleanup_namespace;
@@ -1113,7 +998,6 @@ static task_t *copy_process(unsigned long clone_flags,
         * Ok, make it visible to the rest of the system.
         * We dont wake it up yet.
         */
-       p->tgid = p->pid;
        p->group_leader = p;
        INIT_LIST_HEAD(&p->ptrace_children);
        INIT_LIST_HEAD(&p->ptrace_list);
@@ -1161,7 +1045,6 @@ static task_t *copy_process(unsigned long clone_flags,
                        retval = -EAGAIN;
                        goto bad_fork_cleanup_namespace;
                }
-               p->tgid = current->tgid;
                p->group_leader = current->group_leader;
 
                if (current->signal->group_stop_count > 0) {
@@ -1208,6 +1091,8 @@ fork_out:
 
 bad_fork_cleanup_namespace:
        exit_namespace(p);
+bad_fork_cleanup_keys:
+       exit_keys(p);
 bad_fork_cleanup_mm:
        if (p->mm)
                mmput(p->mm);