1 files changed, 70 insertions, 167 deletions

diff --git a/fs/file.c b/fs/file.c
index 7cb71b99260..66923fe3176 100644
--- a/fs/file.c
+++ b/fs/file.c
@@ -23,23 +23,12 @@
 #include <linux/rcupdate.h>
 #include <linux/workqueue.h>
 
-struct fdtable_defer {
-	spinlock_t lock;
-	struct work_struct wq;
-	struct fdtable *next;
-};
-
 int sysctl_nr_open __read_mostly = 1024*1024;
 int sysctl_nr_open_min = BITS_PER_LONG;
-int sysctl_nr_open_max = 1024 * 1024; /* raised later */
-
-/*
- * We use this list to defer free fdtables that have vmalloced
- * sets/arrays. By keeping a per-cpu list, we avoid having to embed
- * the work_struct in fdtable itself which avoids a 64 byte (i386) increase in
- * this per-task structure.
- */
-static DEFINE_PER_CPU(struct fdtable_defer, fdtable_defer_list);
+/* our max() is unusable in constant expressions ;-/ */
+#define __const_max(x, y) ((x) < (y) ? (x) : (y))
+int sysctl_nr_open_max = __const_max(INT_MAX, ~(size_t)0/sizeof(void *)) &
+			 -BITS_PER_LONG;
 
 static void *alloc_fdmem(size_t size)
 {
@@ -48,65 +37,23 @@ static void *alloc_fdmem(size_t size)
 	 * vmalloc() if the allocation size will be considered "large" by the VM.
 	 */
 	if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER)) {
-		void *data = kmalloc(size, GFP_KERNEL|__GFP_NOWARN);
+		void *data = kmalloc(size, GFP_KERNEL|__GFP_NOWARN|__GFP_NORETRY);
 		if (data != NULL)
 			return data;
 	}
 	return vmalloc(size);
 }
 
-static void free_fdmem(void *ptr)
-{
-	is_vmalloc_addr(ptr) ? vfree(ptr) : kfree(ptr);
-}
-
 static void __free_fdtable(struct fdtable *fdt)
 {
-	free_fdmem(fdt->fd);
-	free_fdmem(fdt->open_fds);
+	kvfree(fdt->fd);
+	kvfree(fdt->open_fds);
 	kfree(fdt);
 }
 
-static void free_fdtable_work(struct work_struct *work)
-{
-	struct fdtable_defer *f =
-		container_of(work, struct fdtable_defer, wq);
-	struct fdtable *fdt;
-
-	spin_lock_bh(&f->lock);
-	fdt = f->next;
-	f->next = NULL;
-	spin_unlock_bh(&f->lock);
-	while(fdt) {
-		struct fdtable *next = fdt->next;
-
-		__free_fdtable(fdt);
-		fdt = next;
-	}
-}
-
 static void free_fdtable_rcu(struct rcu_head *rcu)
 {
-	struct fdtable *fdt = container_of(rcu, struct fdtable, rcu);
-	struct fdtable_defer *fddef;
-
-	BUG_ON(!fdt);
-	BUG_ON(fdt->max_fds <= NR_OPEN_DEFAULT);
-
-	if (!is_vmalloc_addr(fdt->fd) && !is_vmalloc_addr(fdt->open_fds)) {
-		kfree(fdt->fd);
-		kfree(fdt->open_fds);
-		kfree(fdt);
-	} else {
-		fddef = &get_cpu_var(fdtable_defer_list);
-		spin_lock(&fddef->lock);
-		fdt->next = fddef->next;
-		fddef->next = fdt;
-		/* vmallocs are handled from the workqueue context */
-		schedule_work(&fddef->wq);
-		spin_unlock(&fddef->lock);
-		put_cpu_var(fdtable_defer_list);
-	}
+	__free_fdtable(container_of(rcu, struct fdtable, rcu));
 }
 
 /*
@@ -174,12 +121,11 @@ static struct fdtable * alloc_fdtable(unsigned int nr)
 	fdt->open_fds = data;
 	data += nr / BITS_PER_BYTE;
 	fdt->close_on_exec = data;
-	fdt->next = NULL;
 
 	return fdt;
 
 out_arr:
-	free_fdmem(fdt->fd);
+	kvfree(fdt->fd);
 out_fdt:
 	kfree(fdt);
 out:
@@ -221,7 +167,7 @@ static int expand_fdtable(struct files_struct *files, int nr)
 		/* Continue as planned */
 		copy_fdtable(new_fdt, cur_fdt);
 		rcu_assign_pointer(files->fdt, new_fdt);
-		if (cur_fdt->max_fds > NR_OPEN_DEFAULT)
+		if (cur_fdt != &files->fdtab)
 			call_rcu(&cur_fdt->rcu, free_fdtable_rcu);
 	} else {
 		/* Somebody else expanded, so undo our attempt */
@@ -316,7 +262,6 @@ struct files_struct *dup_fd(struct files_struct *oldf, int *errorp)
 	new_fdt->close_on_exec = newf->close_on_exec_init;
 	new_fdt->open_fds = newf->open_fds_init;
 	new_fdt->fd = &newf->fd_array[0];
-	new_fdt->next = NULL;
 
 	spin_lock(&oldf->file_lock);
 	old_fdt = files_fdtable(oldf);
@@ -401,21 +346,16 @@ out:
 	return NULL;
 }
 
-static void close_files(struct files_struct * files)
+static struct fdtable *close_files(struct files_struct * files)
 {
-	int i, j;
-	struct fdtable *fdt;
-
-	j = 0;
-
 	/*
 	 * It is safe to dereference the fd table without RCU or
 	 * ->file_lock because this is the last reference to the
-	 * files structure.  But use RCU to shut RCU-lockdep up.
+	 * files structure.
 	 */
-	rcu_read_lock();
-	fdt = files_fdtable(files);
-	rcu_read_unlock();
+	struct fdtable *fdt = rcu_dereference_raw(files->fdt);
+	int i, j = 0;
+
 	for (;;) {
 		unsigned long set;
 		i = j * BITS_PER_LONG;
@@ -434,6 +374,8 @@ static void close_files(struct files_struct * files)
 			set >>= 1;
 		}
 	}
+
+	return fdt;
 }
 
 struct files_struct *get_files_struct(struct task_struct *task)
@@ -451,14 +393,9 @@ struct files_struct *get_files_struct(struct task_struct *task)
 
 void put_files_struct(struct files_struct *files)
 {
-	struct fdtable *fdt;
-
 	if (atomic_dec_and_test(&files->count)) {
-		close_files(files);
-		/* not really needed, since nobody can see us */
-		rcu_read_lock();
-		fdt = files_fdtable(files);
-		rcu_read_unlock();
+		struct fdtable *fdt = close_files(files);
+
 		/* free the arrays if they are not embedded */
 		if (fdt != &files->fdtab)
 			__free_fdtable(fdt);
@@ -490,23 +427,6 @@ void exit_files(struct task_struct *tsk)
 	}
 }
 
-static void __devinit fdtable_defer_list_init(int cpu)
-{
-	struct fdtable_defer *fddef = &per_cpu(fdtable_defer_list, cpu);
-	spin_lock_init(&fddef->lock);
-	INIT_WORK(&fddef->wq, free_fdtable_work);
-	fddef->next = NULL;
-}
-
-void __init files_defer_init(void)
-{
-	int i;
-	for_each_possible_cpu(i)
-		fdtable_defer_list_init(i);
-	sysctl_nr_open_max = min((size_t)INT_MAX, ~(size_t)0/sizeof(void *)) &
-			     -BITS_PER_LONG;
-}
-
 struct files_struct init_files = {
 	.count		= ATOMIC_INIT(1),
 	.fdt		= &init_files.fdtab,
@@ -516,15 +436,9 @@ struct files_struct init_files = {
 		.close_on_exec	= init_files.close_on_exec_init,
 		.open_fds	= init_files.open_fds_init,
 	},
-	.file_lock	= __SPIN_LOCK_UNLOCKED(init_task.file_lock),
+	.file_lock	= __SPIN_LOCK_UNLOCKED(init_files.file_lock),
 };
 
-void daemonize_descriptors(void)
-{
-	atomic_inc(&init_files.count);
-	reset_files_struct(&init_files);
-}
-
 /*
  * allocate a file descriptor, mark it busy.
  */
@@ -575,7 +489,7 @@ repeat:
 	error = fd;
 #if 1
 	/* Sanity check */
-	if (rcu_dereference_raw(fdt->fd[fd]) != NULL) {
+	if (rcu_access_pointer(fdt->fd[fd]) != NULL) {
 		printk(KERN_WARNING "alloc_fd: slot %d not NULL!\n", fd);
 		rcu_assign_pointer(fdt->fd[fd], NULL);
 	}
@@ -715,16 +629,16 @@ void do_close_on_exec(struct files_struct *files)
 	spin_unlock(&files->file_lock);
 }
 
-struct file *fget(unsigned int fd)
+static struct file *__fget(unsigned int fd, fmode_t mask)
 {
-	struct file *file;
 	struct files_struct *files = current->files;
+	struct file *file;
 
 	rcu_read_lock();
 	file = fcheck_files(files, fd);
 	if (file) {
 		/* File object ref couldn't be taken */
-		if (file->f_mode & FMODE_PATH ||
+		if ((file->f_mode & mask) ||
 		    !atomic_long_inc_not_zero(&file->f_count))
 			file = NULL;
 	}
@@ -733,25 +647,16 @@ struct file *fget(unsigned int fd)
 	return file;
 }
 
+struct file *fget(unsigned int fd)
+{
+	return __fget(fd, FMODE_PATH);
+}
 EXPORT_SYMBOL(fget);
 
 struct file *fget_raw(unsigned int fd)
 {
-	struct file *file;
-	struct files_struct *files = current->files;
-
-	rcu_read_lock();
-	file = fcheck_files(files, fd);
-	if (file) {
-		/* File object ref couldn't be taken */
-		if (!atomic_long_inc_not_zero(&file->f_count))
-			file = NULL;
-	}
-	rcu_read_unlock();
-
-	return file;
+	return __fget(fd, 0);
 }
-
 EXPORT_SYMBOL(fget_raw);
 
 /*
@@ -770,58 +675,54 @@ EXPORT_SYMBOL(fget_raw);
  * The fput_needed flag returned by fget_light should be passed to the
  * corresponding fput_light.
  */
-struct file *fget_light(unsigned int fd, int *fput_needed)
+static unsigned long __fget_light(unsigned int fd, fmode_t mask)
 {
-	struct file *file;
 	struct files_struct *files = current->files;
+	struct file *file;
 
-	*fput_needed = 0;
 	if (atomic_read(&files->count) == 1) {
-		file = fcheck_files(files, fd);
-		if (file && (file->f_mode & FMODE_PATH))
-			file = NULL;
+		file = __fcheck_files(files, fd);
+		if (!file || unlikely(file->f_mode & mask))
+			return 0;
+		return (unsigned long)file;
 	} else {
-		rcu_read_lock();
-		file = fcheck_files(files, fd);
-		if (file) {
-			if (!(file->f_mode & FMODE_PATH) &&
-			    atomic_long_inc_not_zero(&file->f_count))
-				*fput_needed = 1;
-			else
-				/* Didn't get the reference, someone's freed */
-				file = NULL;
-		}
-		rcu_read_unlock();
+		file = __fget(fd, mask);
+		if (!file)
+			return 0;
+		return FDPUT_FPUT | (unsigned long)file;
 	}
+}
+unsigned long __fdget(unsigned int fd)
+{
+	return __fget_light(fd, FMODE_PATH);
+}
+EXPORT_SYMBOL(__fdget);
 
-	return file;
+unsigned long __fdget_raw(unsigned int fd)
+{
+	return __fget_light(fd, 0);
 }
-EXPORT_SYMBOL(fget_light);
 
-struct file *fget_raw_light(unsigned int fd, int *fput_needed)
+unsigned long __fdget_pos(unsigned int fd)
 {
-	struct file *file;
-	struct files_struct *files = current->files;
+	unsigned long v = __fdget(fd);
+	struct file *file = (struct file *)(v & ~3);
 
-	*fput_needed = 0;
-	if (atomic_read(&files->count) == 1) {
-		file = fcheck_files(files, fd);
-	} else {
-		rcu_read_lock();
-		file = fcheck_files(files, fd);
-		if (file) {
-			if (atomic_long_inc_not_zero(&file->f_count))
-				*fput_needed = 1;
-			else
-				/* Didn't get the reference, someone's freed */
-				file = NULL;
+	if (file && (file->f_mode & FMODE_ATOMIC_POS)) {
+		if (file_count(file) > 1) {
+			v |= FDPUT_POS_UNLOCK;
+			mutex_lock(&file->f_pos_lock);
 		}
-		rcu_read_unlock();
 	}
-
-	return file;
+	return v;
 }
 
+/*
+ * We only lock f_pos if we have threads or if the file might be
+ * shared with another process. In both cases we'll have an elevated
+ * file count (done either by fdget() or by fork()).
+ */
+
 void set_close_on_exec(unsigned int fd, int flag)
 {
 	struct files_struct *files = current->files;
@@ -994,16 +895,18 @@ int iterate_fd(struct files_struct *files, unsigned n,
 		const void *p)
 {
 	struct fdtable *fdt;
-	struct file *file;
 	int res = 0;
 	if (!files)
 		return 0;
 	spin_lock(&files->file_lock);
-	fdt = files_fdtable(files);
-	while (!res && n < fdt->max_fds) {
-		file = rcu_dereference_check_fdtable(files, fdt->fd[n++]);
-		if (file)
-			res = f(p, file, n);
+	for (fdt = files_fdtable(files); n < fdt->max_fds; n++) {
+		struct file *file;
+		file = rcu_dereference_check_fdtable(files, fdt->fd[n]);
+		if (!file)
+			continue;
+		res = f(p, file, n);
+		if (res)
+			break;
 	}
 	spin_unlock(&files->file_lock);
 	return res;