Merge branch 'splice' of git://brick.kernel.dk/data/git/linux-2.6-block

* 'splice' of git://brick.kernel.dk/data/git/linux-2.6-block:
  [PATCH] vfs: add splice_write and splice_read to documentation
  [PATCH] Remove sys_ prefix of new syscalls from __NR_sys_*
  [PATCH] splice: warning fix
  [PATCH] another round of fs/pipe.c cleanups
  [PATCH] splice: comment styles
  [PATCH] splice: add Ingo as addition copyright holder
  [PATCH] splice: unlikely() optimizations
  [PATCH] splice: speedups and optimizations
  [PATCH] pipe.c/fifo.c code cleanups
  [PATCH] get rid of the PIPE_*() macros
  [PATCH] splice: speedup __generic_file_splice_read
  [PATCH] splice: add direct fd <-> fd splicing support
  [PATCH] splice: add optional input and output offsets
  [PATCH] introduce a "kernel-internal pipe object" abstraction
  [PATCH] splice: be smarter about calling do_page_cache_readahead()
  [PATCH] splice: optimize the splice buffer mapping
  [PATCH] splice: cleanup __generic_file_splice_read()
  [PATCH] splice: only call wake_up_interruptible() when we really have to
  [PATCH] splice: potential !page dereference
  [PATCH] splice: mark the io page as accessed

8 files changed, 555 insertions, 334 deletions

diff --git a/fs/fifo.c b/fs/fifo.c
index 889f722ee36..49035b174b4 100644
--- a/fs/fifo.c
+++ b/fs/fifo.c
@@ -15,30 +15,35 @@
 #include <linux/fs.h>
 #include <linux/pipe_fs_i.h>
 
-static void wait_for_partner(struct inode* inode, unsigned int* cnt)
+static void wait_for_partner(struct inode* inode, unsigned int *cnt)
 {
 	int cur = *cnt;	
-	while(cur == *cnt) {
-		pipe_wait(inode);
-		if(signal_pending(current))
+
+	while (cur == *cnt) {
+		pipe_wait(inode->i_pipe);
+		if (signal_pending(current))
 			break;
 	}
 }
 
 static void wake_up_partner(struct inode* inode)
 {
-	wake_up_interruptible(PIPE_WAIT(*inode));
+	wake_up_interruptible(&inode->i_pipe->wait);
 }
 
 static int fifo_open(struct inode *inode, struct file *filp)
 {
+	struct pipe_inode_info *pipe;
 	int ret;
 
-	mutex_lock(PIPE_MUTEX(*inode));
-	if (!inode->i_pipe) {
+	mutex_lock(&inode->i_mutex);
+	pipe = inode->i_pipe;
+	if (!pipe) {
 		ret = -ENOMEM;
-		if(!pipe_new(inode))
+		pipe = alloc_pipe_info(inode);
+		if (!pipe)
 			goto err_nocleanup;
+		inode->i_pipe = pipe;
 	}
 	filp->f_version = 0;
 
@@ -53,18 +58,18 @@ static int fifo_open(struct inode *inode, struct file *filp)
 	 *  opened, even when there is no process writing the FIFO.
 	 */
 		filp->f_op = &read_fifo_fops;
-		PIPE_RCOUNTER(*inode)++;
-		if (PIPE_READERS(*inode)++ == 0)
+		pipe->r_counter++;
+		if (pipe->readers++ == 0)
 			wake_up_partner(inode);
 
-		if (!PIPE_WRITERS(*inode)) {
+		if (!pipe->writers) {
 			if ((filp->f_flags & O_NONBLOCK)) {
 				/* suppress POLLHUP until we have
 				 * seen a writer */
-				filp->f_version = PIPE_WCOUNTER(*inode);
+				filp->f_version = pipe->w_counter;
 			} else 
 			{
-				wait_for_partner(inode, &PIPE_WCOUNTER(*inode));
+				wait_for_partner(inode, &pipe->w_counter);
 				if(signal_pending(current))
 					goto err_rd;
 			}
@@ -78,16 +83,16 @@ static int fifo_open(struct inode *inode, struct file *filp)
 	 *  errno=ENXIO when there is no process reading the FIFO.
 	 */
 		ret = -ENXIO;
-		if ((filp->f_flags & O_NONBLOCK) && !PIPE_READERS(*inode))
+		if ((filp->f_flags & O_NONBLOCK) && !pipe->readers)
 			goto err;
 
 		filp->f_op = &write_fifo_fops;
-		PIPE_WCOUNTER(*inode)++;
-		if (!PIPE_WRITERS(*inode)++)
+		pipe->w_counter++;
+		if (!pipe->writers++)
 			wake_up_partner(inode);
 
-		if (!PIPE_READERS(*inode)) {
-			wait_for_partner(inode, &PIPE_RCOUNTER(*inode));
+		if (!pipe->readers) {
+			wait_for_partner(inode, &pipe->r_counter);
 			if (signal_pending(current))
 				goto err_wr;
 		}
@@ -102,11 +107,11 @@ static int fifo_open(struct inode *inode, struct file *filp)
 	 */
 		filp->f_op = &rdwr_fifo_fops;
 
-		PIPE_READERS(*inode)++;
-		PIPE_WRITERS(*inode)++;
-		PIPE_RCOUNTER(*inode)++;
-		PIPE_WCOUNTER(*inode)++;
-		if (PIPE_READERS(*inode) == 1 || PIPE_WRITERS(*inode) == 1)
+		pipe->readers++;
+		pipe->writers++;
+		pipe->r_counter++;
+		pipe->w_counter++;
+		if (pipe->readers == 1 || pipe->writers == 1)
 			wake_up_partner(inode);
 		break;
 
@@ -116,27 +121,27 @@ static int fifo_open(struct inode *inode, struct file *filp)
 	}
 
 	/* Ok! */
-	mutex_unlock(PIPE_MUTEX(*inode));
+	mutex_unlock(&inode->i_mutex);
 	return 0;
 
 err_rd:
-	if (!--PIPE_READERS(*inode))
-		wake_up_interruptible(PIPE_WAIT(*inode));
+	if (!--pipe->readers)
+		wake_up_interruptible(&pipe->wait);
 	ret = -ERESTARTSYS;
 	goto err;
 
 err_wr:
-	if (!--PIPE_WRITERS(*inode))
-		wake_up_interruptible(PIPE_WAIT(*inode));
+	if (!--pipe->writers)
+		wake_up_interruptible(&pipe->wait);
 	ret = -ERESTARTSYS;
 	goto err;
 
 err:
-	if (!PIPE_READERS(*inode) && !PIPE_WRITERS(*inode))
+	if (!pipe->readers && !pipe->writers)
 		free_pipe_info(inode);
 
 err_nocleanup:
-	mutex_unlock(PIPE_MUTEX(*inode));
+	mutex_unlock(&inode->i_mutex);
 	return ret;
 }
 
diff --git a/fs/pipe.c b/fs/pipe.c
index 795df987cd3..e984beb93a0 100644
--- a/fs/pipe.c
+++ b/fs/pipe.c
@@ -36,7 +36,7 @@
  */
 
 /* Drop the inode semaphore and wait for a pipe event, atomically */
-void pipe_wait(struct inode * inode)
+void pipe_wait(struct pipe_inode_info *pipe)
 {
 	DEFINE_WAIT(wait);
 
@@ -44,11 +44,14 @@ void pipe_wait(struct inode * inode)
 	 * Pipes are system-local resources, so sleeping on them
 	 * is considered a noninteractive wait:
 	 */
-	prepare_to_wait(PIPE_WAIT(*inode), &wait, TASK_INTERRUPTIBLE|TASK_NONINTERACTIVE);
-	mutex_unlock(PIPE_MUTEX(*inode));
+	prepare_to_wait(&pipe->wait, &wait,
+			TASK_INTERRUPTIBLE | TASK_NONINTERACTIVE);
+	if (pipe->inode)
+		mutex_unlock(&pipe->inode->i_mutex);
 	schedule();
-	finish_wait(PIPE_WAIT(*inode), &wait);
-	mutex_lock(PIPE_MUTEX(*inode));
+	finish_wait(&pipe->wait, &wait);
+	if (pipe->inode)
+		mutex_lock(&pipe->inode->i_mutex);
 }
 
 static int
@@ -91,7 +94,8 @@ pipe_iov_copy_to_user(struct iovec *iov, const void *from, unsigned long len)
 	return 0;
 }
 
-static void anon_pipe_buf_release(struct pipe_inode_info *info, struct pipe_buffer *buf)
+static void anon_pipe_buf_release(struct pipe_inode_info *pipe,
+				  struct pipe_buffer *buf)
 {
 	struct page *page = buf->page;
 
@@ -100,30 +104,27 @@ static void anon_pipe_buf_release(struct pipe_inode_info *info, struct pipe_buff
 	/*
 	 * If nobody else uses this page, and we don't already have a
 	 * temporary page, let's keep track of it as a one-deep
-	 * allocation cache
+	 * allocation cache. (Otherwise just release our reference to it)
 	 */
-	if (page_count(page) == 1 && !info->tmp_page) {
-		info->tmp_page = page;
-		return;
-	}
-
-	/*
-	 * Otherwise just release our reference to it
-	 */
-	page_cache_release(page);
+	if (page_count(page) == 1 && !pipe->tmp_page)
+		pipe->tmp_page = page;
+	else
+		page_cache_release(page);
 }
 
-static void *anon_pipe_buf_map(struct file *file, struct pipe_inode_info *info, struct pipe_buffer *buf)
+static void * anon_pipe_buf_map(struct file *file, struct pipe_inode_info *pipe,
+				struct pipe_buffer *buf)
 {
 	return kmap(buf->page);
 }
 
-static void anon_pipe_buf_unmap(struct pipe_inode_info *info, struct pipe_buffer *buf)
+static void anon_pipe_buf_unmap(struct pipe_inode_info *pipe,
+				struct pipe_buffer *buf)
 {
 	kunmap(buf->page);
 }
 
-static int anon_pipe_buf_steal(struct pipe_inode_info *info,
+static int anon_pipe_buf_steal(struct pipe_inode_info *pipe,
 			       struct pipe_buffer *buf)
 {
 	buf->flags |= PIPE_BUF_FLAG_STOLEN;
@@ -143,7 +144,7 @@ pipe_readv(struct file *filp, const struct iovec *_iov,
 	   unsigned long nr_segs, loff_t *ppos)
 {
 	struct inode *inode = filp->f_dentry->d_inode;
-	struct pipe_inode_info *info;
+	struct pipe_inode_info *pipe;
 	int do_wakeup;
 	ssize_t ret;
 	struct iovec *iov = (struct iovec *)_iov;
@@ -156,13 +157,13 @@ pipe_readv(struct file *filp, const struct iovec *_iov,
 
 	do_wakeup = 0;
 	ret = 0;
-	mutex_lock(PIPE_MUTEX(*inode));
-	info = inode->i_pipe;
+	mutex_lock(&inode->i_mutex);
+	pipe = inode->i_pipe;
 	for (;;) {
-		int bufs = info->nrbufs;
+		int bufs = pipe->nrbufs;
 		if (bufs) {
-			int curbuf = info->curbuf;
-			struct pipe_buffer *buf = info->bufs + curbuf;
+			int curbuf = pipe->curbuf;
+			struct pipe_buffer *buf = pipe->bufs + curbuf;
 			struct pipe_buf_operations *ops = buf->ops;
 			void *addr;
 			size_t chars = buf->len;
@@ -171,16 +172,17 @@ pipe_readv(struct file *filp, const struct iovec *_iov,
 			if (chars > total_len)
 				chars = total_len;
 
-			addr = ops->map(filp, info, buf);
+			addr = ops->map(filp, pipe, buf);
 			if (IS_ERR(addr)) {
 				if (!ret)
 					ret = PTR_ERR(addr);
 				break;
 			}
 			error = pipe_iov_copy_to_user(iov, addr + buf->offset, chars);
-			ops->unmap(info, buf);
+			ops->unmap(pipe, buf);
 			if (unlikely(error)) {
-				if (!ret) ret = -EFAULT;
+				if (!ret)
+					ret = -EFAULT;
 				break;
 			}
 			ret += chars;
@@ -188,10 +190,10 @@ pipe_readv(struct file *filp, const struct iovec *_iov,
 			buf->len -= chars;
 			if (!buf->len) {
 				buf->ops = NULL;
-				ops->release(info, buf);
+				ops->release(pipe, buf);
 				curbuf = (curbuf + 1) & (PIPE_BUFFERS-1);
-				info->curbuf = curbuf;
-				info->nrbufs = --bufs;
+				pipe->curbuf = curbuf;
+				pipe->nrbufs = --bufs;
 				do_wakeup = 1;
 			}
 			total_len -= chars;
@@ -200,9 +202,9 @@ pipe_readv(struct file *filp, const struct iovec *_iov,
 		}
 		if (bufs)	/* More to do? */
 			continue;
-		if (!PIPE_WRITERS(*inode))
+		if (!pipe->writers)
 			break;
-		if (!PIPE_WAITING_WRITERS(*inode)) {
+		if (!pipe->waiting_writers) {
 			/* syscall merging: Usually we must not sleep
 			 * if O_NONBLOCK is set, or if we got some data.
 			 * But if a writer sleeps in kernel space, then
@@ -216,20 +218,22 @@ pipe_readv(struct file *filp, const struct iovec *_iov,
 			}
 		}
 		if (signal_pending(current)) {
-			if (!ret) ret = -ERESTARTSYS;
+			if (!ret)
+				ret = -ERESTARTSYS;
 			break;
 		}
 		if (do_wakeup) {
-			wake_up_interruptible_sync(PIPE_WAIT(*inode));
- 			kill_fasync(PIPE_FASYNC_WRITERS(*inode), SIGIO, POLL_OUT);
+			wake_up_interruptible_sync(&pipe->wait);
+ 			kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
 		}
-		pipe_wait(inode);
+		pipe_wait(pipe);
 	}
-	mutex_unlock(PIPE_MUTEX(*inode));
-	/* Signal writers asynchronously that there is more room.  */
+	mutex_unlock(&inode->i_mutex);
+
+	/* Signal writers asynchronously that there is more room. */
 	if (do_wakeup) {
-		wake_up_interruptible(PIPE_WAIT(*inode));
-		kill_fasync(PIPE_FASYNC_WRITERS(*inode), SIGIO, POLL_OUT);
+		wake_up_interruptible(&pipe->wait);
+		kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
 	}
 	if (ret > 0)
 		file_accessed(filp);
@@ -240,6 +244,7 @@ static ssize_t
 pipe_read(struct file *filp, char __user *buf, size_t count, loff_t *ppos)
 {
 	struct iovec iov = { .iov_base = buf, .iov_len = count };
+
 	return pipe_readv(filp, &iov, 1, ppos);
 }
 
@@ -248,7 +253,7 @@ pipe_writev(struct file *filp, const struct iovec *_iov,
 	    unsigned long nr_segs, loff_t *ppos)
 {
 	struct inode *inode = filp->f_dentry->d_inode;
-	struct pipe_inode_info *info;
+	struct pipe_inode_info *pipe;
 	ssize_t ret;
 	int do_wakeup;
 	struct iovec *iov = (struct iovec *)_iov;
@@ -262,10 +267,10 @@ pipe_writev(struct file *filp, const struct iovec *_iov,
 
 	do_wakeup = 0;
 	ret = 0;
-	mutex_lock(PIPE_MUTEX(*inode));
-	info = inode->i_pipe;
+	mutex_lock(&inode->i_mutex);
+	pipe = inode->i_pipe;
 
-	if (!PIPE_READERS(*inode)) {
+	if (!pipe->readers) {
 		send_sig(SIGPIPE, current, 0);
 		ret = -EPIPE;
 		goto out;
@@ -273,23 +278,25 @@ pipe_writev(struct file *filp, const struct iovec *_iov,
 
 	/* We try to merge small writes */
 	chars = total_len & (PAGE_SIZE-1); /* size of the last buffer */
-	if (info->nrbufs && chars != 0) {
-		int lastbuf = (info->curbuf + info->nrbufs - 1) & (PIPE_BUFFERS-1);
-		struct pipe_buffer *buf = info->bufs + lastbuf;
+	if (pipe->nrbufs && chars != 0) {
+		int lastbuf = (pipe->curbuf + pipe->nrbufs - 1) &
+							(PIPE_BUFFERS-1);
+		struct pipe_buffer *buf = pipe->bufs + lastbuf;
 		struct pipe_buf_operations *ops = buf->ops;
 		int offset = buf->offset + buf->len;
+
 		if (ops->can_merge && offset + chars <= PAGE_SIZE) {
 			void *addr;
 			int error;
 
-			addr = ops->map(filp, info, buf);
+			addr = ops->map(filp, pipe, buf);
 			if (IS_ERR(addr)) {
 				error = PTR_ERR(addr);
 				goto out;
 			}
 			error = pipe_iov_copy_from_user(offset + addr, iov,
 							chars);
-			ops->unmap(info, buf);
+			ops->unmap(pipe, buf);
 			ret = error;
 			do_wakeup = 1;
 			if (error)
@@ -304,16 +311,18 @@ pipe_writev(struct file *filp, const struct iovec *_iov,
 
 	for (;;) {
 		int bufs;
-		if (!PIPE_READERS(*inode)) {
+
+		if (!pipe->readers) {
 			send_sig(SIGPIPE, current, 0);
-			if (!ret) ret = -EPIPE;
+			if (!ret)
+				ret = -EPIPE;
 			break;
 		}
-		bufs = info->nrbufs;
+		bufs = pipe->nrbufs;
 		if (bufs < PIPE_BUFFERS) {
-			int newbuf = (info->curbuf + bufs) & (PIPE_BUFFERS-1);
-			struct pipe_buffer *buf = info->bufs + newbuf;
-			struct page *page = info->tmp_page;
+			int newbuf = (pipe->curbuf + bufs) & (PIPE_BUFFERS-1);
+			struct pipe_buffer *buf = pipe->bufs + newbuf;
+			struct page *page = pipe->tmp_page;
 			int error;
 
 			if (!page) {
@@ -322,9 +331,9 @@ pipe_writev(struct file *filp, const struct iovec *_iov,
 					ret = ret ? : -ENOMEM;
 					break;
 				}
-				info->tmp_page = page;
+				pipe->tmp_page = page;
 			}
-			/* Always wakeup, even if the copy fails. Otherwise
+			/* Always wake up, even if the copy fails. Otherwise
 			 * we lock up (O_NONBLOCK-)readers that sleep due to
 			 * syscall merging.
 			 * FIXME! Is this really true?
@@ -337,7 +346,8 @@ pipe_writev(struct file *filp, const struct iovec *_iov,
 			error = pipe_iov_copy_from_user(kmap(page), iov, chars);
 			kunmap(page);
 			if (unlikely(error)) {
-				if (!ret) ret = -EFAULT;
+				if (!ret)
+					ret = -EFAULT;
 				break;
 			}
 			ret += chars;
@@ -347,8 +357,8 @@ pipe_writev(struct file *filp, const struct iovec *_iov,
 			buf->ops = &anon_pipe_buf_ops;
 			buf->offset = 0;
 			buf->len = chars;
-			info->nrbufs = ++bufs;
-			info->tmp_page = NULL;
+			pipe->nrbufs = ++bufs;
+			pipe->tmp_page = NULL;
 
 			total_len -= chars;
 			if (!total_len)
@@ -357,27 +367,29 @@ pipe_writev(struct file *filp, const struct iovec *_iov,
 		if (bufs < PIPE_BUFFERS)
 			continue;
 		if (filp->f_flags & O_NONBLOCK) {
-			if (!ret) ret = -EAGAIN;
+			if (!ret)
+				ret = -EAGAIN;
 			break;
 		}
 		if (signal_pending(current)) {
-			if (!ret) ret = -ERESTARTSYS;
+			if (!ret)
+				ret = -ERESTARTSYS;
 			break;
 		}
 		if (do_wakeup) {
-			wake_up_interruptible_sync(PIPE_WAIT(*inode));
-			kill_fasync(PIPE_FASYNC_READERS(*inode), SIGIO, POLL_IN);
+			wake_up_interruptible_sync(&pipe->wait);
+			kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
 			do_wakeup = 0;
 		}
-		PIPE_WAITING_WRITERS(*inode)++;
-		pipe_wait(inode);
-		PIPE_WAITING_WRITERS(*inode)--;
+		pipe->waiting_writers++;
+		pipe_wait(pipe);
+		pipe->waiting_writers--;
 	}
 out:
-	mutex_unlock(PIPE_MUTEX(*inode));
+	mutex_unlock(&inode->i_mutex);
 	if (do_wakeup) {
-		wake_up_interruptible(PIPE_WAIT(*inode));
-		kill_fasync(PIPE_FASYNC_READERS(*inode), SIGIO, POLL_IN);
+		wake_up_interruptible(&pipe->wait);
+		kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
 	}
 	if (ret > 0)
 		file_update_time(filp);
@@ -389,6 +401,7 @@ pipe_write(struct file *filp, const char __user *buf,
 	   size_t count, loff_t *ppos)
 {
 	struct iovec iov = { .iov_base = (void __user *)buf, .iov_len = count };
+
 	return pipe_writev(filp, &iov, 1, ppos);
 }
 
@@ -399,7 +412,8 @@ bad_pipe_r(struct file *filp, char __user *buf, size_t count, loff_t *ppos)
 }
 
 static ssize_t
-bad_pipe_w(struct file *filp, const char __user *buf, size_t count, loff_t *ppos)
+bad_pipe_w(struct file *filp, const char __user *buf, size_t count,
+	   loff_t *ppos)
 {
 	return -EBADF;
 }
@@ -409,21 +423,22 @@ pipe_ioctl(struct inode *pino, struct file *filp,
 	   unsigned int cmd, unsigned long arg)
 {
 	struct inode *inode = filp->f_dentry->d_inode;
-	struct pipe_inode_info *info;
+	struct pipe_inode_info *pipe;
 	int count, buf, nrbufs;
 
 	switch (cmd) {
 		case FIONREAD:
-			mutex_lock(PIPE_MUTEX(*inode));
-			info =  inode->i_pipe;
+			mutex_lock(&inode->i_mutex);
+			pipe = inode->i_pipe;
 			count = 0;
-			buf = info->curbuf;
-			nrbufs = info->nrbufs;
+			buf = pipe->curbuf;
+			nrbufs = pipe->nrbufs;
 			while (--nrbufs >= 0) {
-				count += info->bufs[buf].len;
+				count += pipe->bufs[buf].len;
 				buf = (buf+1) & (PIPE_BUFFERS-1);
 			}
-			mutex_unlock(PIPE_MUTEX(*inode));
+			mutex_unlock(&inode->i_mutex);
+
 			return put_user(count, (int __user *)arg);
 		default:
 			return -EINVAL;
@@ -436,17 +451,17 @@ pipe_poll(struct file *filp, poll_table *wait)
 {
 	unsigned int mask;
 	struct inode *inode = filp->f_dentry->d_inode;
-	struct pipe_inode_info *info = inode->i_pipe;
+	struct pipe_inode_info *pipe = inode->i_pipe;
 	int nrbufs;
 
-	poll_wait(filp, PIPE_WAIT(*inode), wait);
+	poll_wait(filp, &pipe->wait, wait);
 
 	/* Reading only -- no need for acquiring the semaphore.  */
-	nrbufs = info->nrbufs;
+	nrbufs = pipe->nrbufs;
 	mask = 0;
 	if (filp->f_mode & FMODE_READ) {
 		mask = (nrbufs > 0) ? POLLIN | POLLRDNORM : 0;
-		if (!PIPE_WRITERS(*inode) && filp->f_version != PIPE_WCOUNTER(*inode))
+		if (!pipe->writers && filp->f_version != pipe->w_counter)
 			mask |= POLLHUP;
 	}
 
@@ -456,7 +471,7 @@ pipe_poll(struct file *filp, poll_table *wait)
 		 * Most Unices do not set POLLERR for FIFOs but on Linux they
 		 * behave exactly like pipes for poll().
 		 */
-		if (!PIPE_READERS(*inode))
+		if (!pipe->readers)
 			mask |= POLLERR;
 	}
 
@@ -466,17 +481,21 @@ pipe_poll(struct file *filp, poll_table *wait)
 static int
 pipe_release(struct inode *inode, int decr, int decw)
 {
-	mutex_lock(PIPE_MUTEX(*inode));
-	PIPE_READERS(*inode) -= decr;
-	PIPE_WRITERS(*inode) -= decw;
-	if (!PIPE_READERS(*inode) && !PIPE_WRITERS(*inode)) {
+	struct pipe_inode_info *pipe;
+
+	mutex_lock(&inode->i_mutex);
+	pipe = inode->i_pipe;
+	pipe->readers -= decr;
+	pipe->writers -= decw;
+
+	if (!pipe->readers && !pipe->writers) {
 		free_pipe_info(inode);
 	} else {
-		wake_up_interruptible(PIPE_WAIT(*inode));
-		kill_fasync(PIPE_FASYNC_READERS(*inode), SIGIO, POLL_IN);
-		kill_fasync(PIPE_FASYNC_WRITERS(*inode), SIGIO, POLL_OUT);
+		wake_up_interruptible(&pipe->wait);
+		kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
+		kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
 	}
-	mutex_unlock(PIPE_MUTEX(*inode));
+	mutex_unlock(&inode->i_mutex);
 
 	return 0;
 }
@@ -487,9 +506,9 @@ pipe_read_fasync(int fd, struct file *filp, int on)
 	struct inode *inode = filp->f_dentry->d_inode;
 	int retval;
 
-	mutex_lock(PIPE_MUTEX(*inode));
-	retval = fasync_helper(fd, filp, on, PIPE_FASYNC_READERS(*inode));
-	mutex_unlock(PIPE_MUTEX(*inode));
+	mutex_lock(&inode->i_mutex);
+	retval = fasync_helper(fd, filp, on, &inode->i_pipe->fasync_readers);
+	mutex_unlock(&inode->i_mutex);
 
 	if (retval < 0)
 		return retval;
@@ -504,9 +523,9 @@ pipe_write_fasync(int fd, struct file *filp, int on)
 	struct inode *inode = filp->f_dentry->d_inode;
 	int retval;
 
-	mutex_lock(PIPE_MUTEX(*inode));
-	retval = fasync_helper(fd, filp, on, PIPE_FASYNC_WRITERS(*inode));
-	mutex_unlock(PIPE_MUTEX(*inode));
+	mutex_lock(&inode->i_mutex);
+	retval = fasync_helper(fd, filp, on, &inode->i_pipe->fasync_writers);
+	mutex_unlock(&inode->i_mutex);
 
 	if (retval < 0)
 		return retval;
@@ -519,16 +538,17 @@ static int
 pipe_rdwr_fasync(int fd, struct file *filp, int on)
 {
 	struct inode *inode = filp->f_dentry->d_inode;
+	struct pipe_inode_info *pipe = inode->i_pipe;
 	int retval;
 
-	mutex_lock(PIPE_MUTEX(*inode));
+	mutex_lock(&inode->i_mutex);
 
-	retval = fasync_helper(fd, filp, on, PIPE_FASYNC_READERS(*inode));
+	retval = fasync_helper(fd, filp, on, &pipe->fasync_readers);
 
 	if (retval >= 0)
-		retval = fasync_helper(fd, filp, on, PIPE_FASYNC_WRITERS(*inode));
+		retval = fasync_helper(fd, filp, on, &pipe->fasync_writers);
 
-	mutex_unlock(PIPE_MUTEX(*inode));
+	mutex_unlock(&inode->i_mutex);
 
 	if (retval < 0)
 		return retval;
@@ -567,9 +587,9 @@ pipe_read_open(struct inode *inode, struct file *filp)
 {
 	/* We could have perhaps used atomic_t, but this and friends
 	   below are the only places.  So it doesn't seem worthwhile.  */
-	mutex_lock(PIPE_MUTEX(*inode));
-	PIPE_READERS(*inode)++;
-	mutex_unlock(PIPE_MUTEX(*inode));
+	mutex_lock(&inode->i_mutex);
+	inode->i_pipe->readers++;
+	mutex_unlock(&inode->i_mutex);
 
 	return 0;
 }
@@ -577,9 +597,9 @@ pipe_read_open(struct inode *inode, struct file *filp)
 static int
 pipe_write_open(struct inode *inode, struct file *filp)
 {
-	mutex_lock(PIPE_MUTEX(*inode));
-	PIPE_WRITERS(*inode)++;
-	mutex_unlock(PIPE_MUTEX(*inode));
+	mutex_lock(&inode->i_mutex);
+	inode->i_pipe->writers++;
+	mutex_unlock(&inode->i_mutex);
 
 	return 0;
 }
@@ -587,12 +607,12 @@ pipe_write_open(struct inode *inode, struct file *filp)
 static int
 pipe_rdwr_open(struct inode *inode, struct file *filp)
 {
-	mutex_lock(PIPE_MUTEX(*inode));
+	mutex_lock(&inode->i_mutex);
 	if (filp->f_mode & FMODE_READ)
-		PIPE_READERS(*inode)++;
+		inode->i_pipe->readers++;
 	if (filp->f_mode & FMODE_WRITE)
-		PIPE_WRITERS(*inode)++;
-	mutex_unlock(PIPE_MUTEX(*inode));
+		inode->i_pipe->writers++;
+	mutex_unlock(&inode->i_mutex);
 
 	return 0;
 }
@@ -675,37 +695,38 @@ static struct file_operations rdwr_pipe_fops = {
 	.fasync		= pipe_rdwr_fasync,
 };
 
-void free_pipe_info(struct inode *inode)
+struct pipe_inode_info * alloc_pipe_info(struct inode *inode)
+{
+	struct pipe_inode_info *pipe;
+
+	pipe = kzalloc(sizeof(struct pipe_inode_info), GFP_KERNEL);
+	if (pipe) {
+		init_waitqueue_head(&pipe->wait);
+		pipe->r_counter = pipe->w_counter = 1;
+		pipe->inode = inode;
+	}
+
+	return pipe;
+}
+
+void __free_pipe_info(struct pipe_inode_info *pipe)
 {
 	int i;
-	struct pipe_inode_info *info = inode->i_pipe;
 
-	inode->i_pipe = NULL;
 	for (i = 0; i < PIPE_BUFFERS; i++) {
-		struct pipe_buffer *buf = info->bufs + i;
+		struct pipe_buffer *buf = pipe->bufs + i;
 		if (buf->ops)
-			buf->ops->release(info, buf);
+			buf->ops->release(pipe, buf);
 	}
-	if (info->tmp_page)
-		__free_page(info->tmp_page);
-	kfree(info);
+	if (pipe->tmp_page)
+		__free_page(pipe->tmp_page);
+	kfree(pipe);
 }
 
-struct inode* pipe_new(struct inode* inode)
+void free_pipe_info(struct inode *inode)
 {
-	struct pipe_inode_info *info;
-
-	info = kzalloc(sizeof(struct pipe_inode_info), GFP_KERNEL);
-	if (!info)
-		goto fail_page;
-	inode->i_pipe = info;
-
-	init_waitqueue_head(PIPE_WAIT(*inode));
-	PIPE_RCOUNTER(*inode) = PIPE_WCOUNTER(*inode) = 1;
-
-	return inode;
-fail_page:
-	return NULL;
+	__free_pipe_info(inode->i_pipe);
+	inode->i_pipe = NULL;
 }
 
 static struct vfsmount *pipe_mnt __read_mostly;
@@ -713,6 +734,7 @@ static int pipefs_delete_dentry(struct dentry *dentry)
 {
 	return 1;
 }
+
 static struct dentry_operations pipefs_dentry_operations = {
 	.d_delete	= pipefs_delete_dentry,
 };
@@ -720,13 +742,17 @@ static struct dentry_operations pipefs_dentry_operations = {
 static struct inode * get_pipe_inode(void)
 {
 	struct inode *inode = new_inode(pipe_mnt->mnt_sb);
+	struct pipe_inode_info *pipe;
 
 	if (!inode)
 		goto fail_inode;
 
-	if(!pipe_new(inode))
+	pipe = alloc_pipe_info(inode);
+	if (!pipe)
 		goto fail_iput;
-	PIPE_READERS(*inode) = PIPE_WRITERS(*inode) = 1;
+	inode->i_pipe = pipe;
+
+	pipe->readers = pipe->writers = 1;
 	inode->i_fop = &rdwr_pipe_fops;
 
 	/*
@@ -741,10 +767,12 @@ static struct inode * get_pipe_inode(void)
 	inode->i_gid = current->fsgid;
 	inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
 	inode->i_blksize = PAGE_SIZE;
+
 	return inode;
 
 fail_iput:
 	iput(inode);
+
 fail_inode:
 	return NULL;
 }
@@ -757,7 +785,7 @@ int do_pipe(int *fd)
 	struct inode * inode;
 	struct file *f1, *f2;
 	int error;
-	int i,j;
+	int i, j;
 
 	error = -ENFILE;
 	f1 = get_empty_filp();
@@ -790,6 +818,7 @@ int do_pipe(int *fd)
 	dentry = d_alloc(pipe_mnt->mnt_sb->s_root, &this);
 	if (!dentry)
 		goto close_f12_inode_i_j;
+
 	dentry->d_op = &pipefs_dentry_operations;
 	d_add(dentry, inode);
 	f1->f_vfsmnt = f2->f_vfsmnt = mntget(mntget(pipe_mnt));
@@ -813,6 +842,7 @@ int do_pipe(int *fd)
 	fd_install(j, f2);
 	fd[0] = i;
 	fd[1] = j;
+
 	return 0;
 
 close_f12_inode_i_j:
@@ -837,8 +867,9 @@ no_files:
  * d_name - pipe: will go nicely and kill the special-casing in procfs.
  */
 
-static struct super_block *pipefs_get_sb(struct file_system_type *fs_type,
-	int flags, const char *dev_name, void *data)
+static struct super_block *
+pipefs_get_sb(struct file_system_type *fs_type, int flags,
+	      const char *dev_name, void *data)
 {
 	return get_sb_pseudo(fs_type, "pipe:", NULL, PIPEFS_MAGIC);
 }
@@ -852,6 +883,7 @@ static struct file_system_type pipe_fs_type = {
 static int __init init_pipe_fs(void)
 {
 	int err = register_filesystem(&pipe_fs_type);
+
 	if (!err) {
 		pipe_mnt = kern_mount(&pipe_fs_type);
 		if (IS_ERR(pipe_mnt)) {
diff --git a/fs/read_write.c b/fs/read_write.c
index 6256ca81a71..5bc0e9234f9 100644
--- a/fs/read_write.c
+++ b/fs/read_write.c
@@ -202,7 +202,7 @@ int rw_verify_area(int read_write, struct file *file, loff_t *ppos, size_t count
 		goto Einval;
 
 	inode = file->f_dentry->d_inode;
-	if (inode->i_flock && MANDATORY_LOCK(inode)) {
+	if (unlikely(inode->i_flock && MANDATORY_LOCK(inode))) {
 		int retval = locks_mandatory_area(
 			read_write == READ ? FLOCK_VERIFY_READ : FLOCK_VERIFY_WRITE,
 			inode, file, pos, count);
diff --git a/fs/splice.c b/fs/splice.c
index bfa42a277bb..e50a460239d 100644
--- a/fs/splice.c
+++ b/fs/splice.c
@@ -9,11 +9,12 @@
  * that transfers data buffers to or from a pipe buffer.
  *
  * Named by Larry McVoy, original implementation from Linus, extended by
- * Jens to support splicing to files and fixing the initial implementation
- * bugs.
+ * Jens to support splicing to files, network, direct splicing, etc and
+ * fixing lots of bugs.
  *
- * Copyright (C) 2005 Jens Axboe <axboe@suse.de>
- * Copyright (C) 2005 Linus Torvalds <torvalds@osdl.org>
+ * Copyright (C) 2005-2006 Jens Axboe <axboe@suse.de>
+ * Copyright (C) 2005-2006 Linus Torvalds <torvalds@osdl.org>
+ * Copyright (C) 2006 Ingo Molnar <mingo@elte.hu>
  *
  */
 #include <linux/fs.h>
@@ -84,26 +85,43 @@ static void *page_cache_pipe_buf_map(struct file *file,
 				     struct pipe_buffer *buf)
 {
 	struct page *page = buf->page;
-
-	lock_page(page);
+	int err;
 
 	if (!PageUptodate(page)) {
-		unlock_page(page);
-		return ERR_PTR(-EIO);
-	}
+		lock_page(page);
+
+		/*
+		 * Page got truncated/unhashed. This will cause a 0-byte
+		 * splice, if this is the first page.
+		 */
+		if (!page->mapping) {
+			err = -ENODATA;
+			goto error;
+		}
 
-	if (!page->mapping) {
+		/*
+		 * Uh oh, read-error from disk.
+		 */
+		if (!PageUptodate(page)) {
+			err = -EIO;
+			goto error;
+		}
+
+		/*
+		 * Page is ok afterall, fall through to mapping.
+		 */
 		unlock_page(page);
-		return ERR_PTR(-ENODATA);
 	}
 
-	return kmap(buf->page);
+	return kmap(page);
+error:
+	unlock_page(page);
+	return ERR_PTR(err);
 }
 
 static void page_cache_pipe_buf_unmap(struct pipe_inode_info *info,
 				      struct pipe_buffer *buf)
 {
-	unlock_page(buf->page);
 	kunmap(buf->page);
 }
 
@@ -119,34 +137,30 @@ static struct pipe_buf_operations page_cache_pipe_buf_ops = {
  * Pipe output worker. This sets up our pipe format with the page cache
  * pipe buffer operations. Otherwise very similar to the regular pipe_writev().
  */
-static ssize_t move_to_pipe(struct inode *inode, struct page **pages,
+static ssize_t move_to_pipe(struct pipe_inode_info *pipe, struct page **pages,
 			    int nr_pages, unsigned long offset,
 			    unsigned long len, unsigned int flags)
 {
-	struct pipe_inode_info *info;
 	int ret, do_wakeup, i;
 
 	ret = 0;
 	do_wakeup = 0;
 	i = 0;
 
-	mutex_lock(PIPE_MUTEX(*inode));
+	if (pipe->inode)
+		mutex_lock(&pipe->inode->i_mutex);
 
-	info = inode->i_pipe;
 	for (;;) {
-		int bufs;
-
-		if (!PIPE_READERS(*inode)) {
+		if (!pipe->readers) {
 			send_sig(SIGPIPE, current, 0);
 			if (!ret)
 				ret = -EPIPE;
 			break;
 		}
 
-		bufs = info->nrbufs;
-		if (bufs < PIPE_BUFFERS) {
-			int newbuf = (info->curbuf + bufs) & (PIPE_BUFFERS - 1);
-			struct pipe_buffer *buf = info->bufs + newbuf;
+		if (pipe->nrbufs < PIPE_BUFFERS) {
+			int newbuf = (pipe->curbuf + pipe->nrbufs) & (PIPE_BUFFERS - 1);
+			struct pipe_buffer *buf = pipe->bufs + newbuf;
 			struct page *page = pages[i++];
 			unsigned long this_len;
 
@@ -158,8 +172,9 @@ static ssize_t move_to_pipe(struct inode *inode, struct page **pages,
 			buf->offset = offset;
 			buf->len = this_len;
 			buf->ops = &page_cache_pipe_buf_ops;
-			info->nrbufs = ++bufs;
-			do_wakeup = 1;
+			pipe->nrbufs++;
+			if (pipe->inode)
+				do_wakeup = 1;
 
 			ret += this_len;
 			len -= this_len;
@@ -168,7 +183,7 @@ static ssize_t move_to_pipe(struct inode *inode, struct page **pages,
 				break;
 			if (!len)
 				break;
-			if (bufs < PIPE_BUFFERS)
+			if (pipe->nrbufs < PIPE_BUFFERS)
 				continue;
 
 			break;
@@ -187,22 +202,26 @@ static ssize_t move_to_pipe(struct inode *inode, struct page **pages,
 		}
 
 		if (do_wakeup) {
-			wake_up_interruptible_sync(PIPE_WAIT(*inode));
-			kill_fasync(PIPE_FASYNC_READERS(*inode), SIGIO,
-				    POLL_IN);
+			smp_mb();
+			if (waitqueue_active(&pipe->wait))
+				wake_up_interruptible_sync(&pipe->wait);
+			kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
 			do_wakeup = 0;
 		}
 
-		PIPE_WAITING_WRITERS(*inode)++;
-		pipe_wait(inode);
-		PIPE_WAITING_WRITERS(*inode)--;
+		pipe->waiting_writers++;
+		pipe_wait(pipe);
+		pipe->waiting_writers--;
 	}
 
-	mutex_unlock(PIPE_MUTEX(*inode));
+	if (pipe->inode)
+		mutex_unlock(&pipe->inode->i_mutex);
 
 	if (do_wakeup) {
-		wake_up_interruptible(PIPE_WAIT(*inode));
-		kill_fasync(PIPE_FASYNC_READERS(*inode), SIGIO, POLL_IN);
+		smp_mb();
+		if (waitqueue_active(&pipe->wait))
+			wake_up_interruptible(&pipe->wait);
+		kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
 	}
 
 	while (i < nr_pages)
@@ -211,15 +230,16 @@ static ssize_t move_to_pipe(struct inode *inode, struct page **pages,
 	return ret;
 }
 
-static int __generic_file_splice_read(struct file *in, struct inode *pipe,
-				      size_t len, unsigned int flags)
+static int
+__generic_file_splice_read(struct file *in, struct pipe_inode_info *pipe,
+			   size_t len, unsigned int flags)
 {
 	struct address_space *mapping = in->f_mapping;
 	unsigned int offset, nr_pages;
-	struct page *pages[PIPE_BUFFERS], *shadow[PIPE_BUFFERS];
+	struct page *pages[PIPE_BUFFERS];
 	struct page *page;
-	pgoff_t index, pidx;
-	int i, j;
+	pgoff_t index;
+	int i, error;
 
 	index = in->f_pos >> PAGE_CACHE_SHIFT;
 	offset = in->f_pos & ~PAGE_CACHE_MASK;
@@ -229,78 +249,94 @@ static int __generic_file_splice_read(struct file *in, struct inode *pipe,
 		nr_pages = PIPE_BUFFERS;
 
 	/*
-	 * initiate read-ahead on this page range
-	 */
-	do_page_cache_readahead(mapping, in, index, nr_pages);
-
-	/*
-	 * Get as many pages from the page cache as possible..
-	 * Start IO on the page cache entries we create (we
-	 * can assume that any pre-existing ones we find have
-	 * already had IO started on them).
+	 * Initiate read-ahead on this page range. however, don't call into
+	 * read-ahead if this is a non-zero offset (we are likely doing small
+	 * chunk splice and the page is already there) for a single page.
 	 */
-	i = find_get_pages(mapping, index, nr_pages, pages);
+	if (!offset || nr_pages > 1)
+		do_page_cache_readahead(mapping, in, index, nr_pages);
 
 	/*
-	 * common case - we found all pages and they are contiguous,
-	 * kick them off
+	 * Now fill in the holes:
 	 */
-	if (i && (pages[i - 1]->index == index + i - 1))
-		goto splice_them;
+	error = 0;
+	for (i = 0; i < nr_pages; i++, index++) {
+find_page:
+		/*
+		 * lookup the page for this index
+		 */
+		page = find_get_page(mapping, index);
+		if (!page) {
+			/*
+			 * If in nonblock mode then dont block on
+			 * readpage (we've kicked readahead so there
+			 * will be asynchronous progress):
+			 */
+			if (flags & SPLICE_F_NONBLOCK)
+				break;
 
-	/*
-	 * fill shadow[] with pages at the right locations, so we only
-	 * have to fill holes
-	 */
-	memset(shadow, 0, nr_pages * sizeof(struct page *));
-	for (j = 0; j < i; j++)
-		shadow[pages[j]->index - index] = pages[j];
+			/*
+			 * page didn't exist, allocate one
+			 */
+			page = page_cache_alloc_cold(mapping);
+			if (!page)
+				break;
 
-	/*
-	 * now fill in the holes
-	 */
-	for (i = 0, pidx = index; i < nr_pages; pidx++, i++) {
-		int error;
+			error = add_to_page_cache_lru(page, mapping, index,
+						mapping_gfp_mask(mapping));
+			if (unlikely(error)) {
+				page_cache_release(page);
+				break;
+			}
 
-		if (shadow[i])
-			continue;
+			goto readpage;
+		}
 
 		/*
-		 * no page there, look one up / create it
+		 * If the page isn't uptodate, we may need to start io on it
 		 */
-		page = find_or_create_page(mapping, pidx,
-						   mapping_gfp_mask(mapping));
-		if (!page)
-			break;
+		if (!PageUptodate(page)) {
+			lock_page(page);
+
+			/*
+			 * page was truncated, stop here. if this isn't the
+			 * first page, we'll just complete what we already
+			 * added
+			 */
+			if (!page->mapping) {
+				unlock_page(page);
+				page_cache_release(page);
+				break;
+			}
+			/*
+			 * page was already under io and is now done, great
+			 */
+			if (PageUptodate(page)) {
+				unlock_page(page);
+				goto fill_it;
+			}
 
-		if (PageUptodate(page))
-			unlock_page(page);
-		else {
+readpage:
+			/*
+			 * need to read in the page
+			 */
 			error = mapping->a_ops->readpage(in, page);
 
 			if (unlikely(error)) {
 				page_cache_release(page);
+				if (error == AOP_TRUNCATED_PAGE)
+					goto find_page;
 				break;
 			}
 		}
-		shadow[i] = page;
+fill_it:
+		pages[i] = page;
 	}
 
-	if (!i) {
-		for (i = 0; i < nr_pages; i++) {
-			 if (shadow[i])
-				page_cache_release(shadow[i]);
-		}
-		return 0;
-	}
+	if (i)
+		return move_to_pipe(pipe, pages, i, offset, len, flags);
 
-	memcpy(pages, shadow, i * sizeof(struct page *));
-
-	/*
-	 * Now we splice them into the pipe..
-	 */
-splice_them:
-	return move_to_pipe(pipe, pages, i, offset, len, flags);
+	return error;
 }
 
 /**
@@ -311,9 +347,8 @@ splice_them:
  * @flags:	splice modifier flags
  *
  * Will read pages from given file and fill them into a pipe.
- *
  */
-ssize_t generic_file_splice_read(struct file *in, struct inode *pipe,
+ssize_t generic_file_splice_read(struct file *in, struct pipe_inode_info *pipe,
 				 size_t len, unsigned int flags)
 {
 	ssize_t spliced;
@@ -321,6 +356,7 @@ ssize_t generic_file_splice_read(struct file *in, struct inode *pipe,
 
 	ret = 0;
 	spliced = 0;
+
 	while (len) {
 		ret = __generic_file_splice_read(in, pipe, len, flags);
 
@@ -360,10 +396,10 @@ static int pipe_to_sendpage(struct pipe_inode_info *info,
 	int more;
 
 	/*
-	 * sub-optimal, but we are limited by the pipe ->map. we don't
+	 * Sub-optimal, but we are limited by the pipe ->map. We don't
 	 * need a kmap'ed buffer here, we just want to make sure we
 	 * have the page pinned if the pipe page originates from the
-	 * page cache
+	 * page cache.
 	 */
 	ptr = buf->ops->map(file, info, buf);
 	if (IS_ERR(ptr))
@@ -414,7 +450,7 @@ static int pipe_to_file(struct pipe_inode_info *info, struct pipe_buffer *buf,
 	int ret;
 
 	/*
-	 * after this, page will be locked and unmapped
+	 * make sure the data in this buffer is uptodate
 	 */
 	src = buf->ops->map(file, info, buf);
 	if (IS_ERR(src))
@@ -424,7 +460,7 @@ static int pipe_to_file(struct pipe_inode_info *info, struct pipe_buffer *buf,
 	offset = sd->pos & ~PAGE_CACHE_MASK;
 
 	/*
-	 * reuse buf page, if SPLICE_F_MOVE is set
+	 * Reuse buf page, if SPLICE_F_MOVE is set.
 	 */
 	if (sd->flags & SPLICE_F_MOVE) {
 		/*
@@ -434,6 +470,9 @@ static int pipe_to_file(struct pipe_inode_info *info, struct pipe_buffer *buf,
 		if (buf->ops->steal(info, buf))
 			goto find_page;
 
+		/*
+		 * this will also set the page locked
+		 */
 		page = buf->page;
 		if (add_to_page_cache(page, mapping, index, gfp_mask))
 			goto find_page;
@@ -445,7 +484,7 @@ find_page:
 		ret = -ENOMEM;
 		page = find_or_create_page(mapping, index, gfp_mask);
 		if (!page)
-			goto out;
+			goto out_nomem;
 
 		/*
 		 * If the page is uptodate, it is also locked. If it isn't
@@ -462,7 +501,7 @@ find_page:
 
 				if (!PageUptodate(page)) {
 					/*
-					 * page got invalidated, repeat
+					 * Page got invalidated, repeat.
 					 */
 					if (!page->mapping) {
 						unlock_page(page);
@@ -501,12 +540,14 @@ find_page:
 	} else if (ret)
 		goto out;
 
+	mark_page_accessed(page);
 	balance_dirty_pages_ratelimited(mapping);
 out:
 	if (!(buf->flags & PIPE_BUF_FLAG_STOLEN)) {
 		page_cache_release(page);
 		unlock_page(page);
 	}
+out_nomem:
 	buf->ops->unmap(info, buf);
 	return ret;
 }
@@ -519,11 +560,10 @@ typedef int (splice_actor)(struct pipe_inode_info *, struct pipe_buffer *,
  * key here is the 'actor' worker passed in that actually moves the data
  * to the wanted destination. See pipe_to_file/pipe_to_sendpage above.
  */
-static ssize_t move_from_pipe(struct inode *inode, struct file *out,
+static ssize_t move_from_pipe(struct pipe_inode_info *pipe, struct file *out,
 			      size_t len, unsigned int flags,
 			      splice_actor *actor)
 {
-	struct pipe_inode_info *info;
 	int ret, do_wakeup, err;
 	struct splice_desc sd;
 
@@ -535,22 +575,19 @@ static ssize_t move_from_pipe(struct inode *inode, struct file *out,
 	sd.file = out;
 	sd.pos = out->f_pos;
 
-	mutex_lock(PIPE_MUTEX(*inode));
+	if (pipe->inode)
+		mutex_lock(&pipe->inode->i_mutex);
 
-	info = inode->i_pipe;
 	for (;;) {
-		int bufs = info->nrbufs;
-
-		if (bufs) {
-			int curbuf = info->curbuf;
-			struct pipe_buffer *buf = info->bufs + curbuf;
+		if (pipe->nrbufs) {
+			struct pipe_buffer *buf = pipe->bufs + pipe->curbuf;
 			struct pipe_buf_operations *ops = buf->ops;
 
 			sd.len = buf->len;
 			if (sd.len > sd.total_len)
 				sd.len = sd.total_len;
 
-			err = actor(info, buf, &sd);
+			err = actor(pipe, buf, &sd);
 			if (err) {
 				if (!ret && err != -ENODATA)
 					ret = err;
@@ -561,13 +598,14 @@ static ssize_t move_from_pipe(struct inode *inode, struct file *out,
 			ret += sd.len;
 			buf->offset += sd.len;
 			buf->len -= sd.len;
+
 			if (!buf->len) {
 				buf->ops = NULL;
-				ops->release(info, buf);
-				curbuf = (curbuf + 1) & (PIPE_BUFFERS - 1);
-				info->curbuf = curbuf;
-				info->nrbufs = --bufs;
-				do_wakeup = 1;
+				ops->release(pipe, buf);
+				pipe->curbuf = (pipe->curbuf + 1) & (PIPE_BUFFERS - 1);
+				pipe->nrbufs--;
+				if (pipe->inode)
+					do_wakeup = 1;
 			}
 
 			sd.pos += sd.len;
@@ -576,11 +614,11 @@ static ssize_t move_from_pipe(struct inode *inode, struct file *out,
 				break;
 		}
 
-		if (bufs)
+		if (pipe->nrbufs)
 			continue;
-		if (!PIPE_WRITERS(*inode))
+		if (!pipe->writers)
 			break;
-		if (!PIPE_WAITING_WRITERS(*inode)) {
+		if (!pipe->waiting_writers) {
 			if (ret)
 				break;
 		}
@@ -598,31 +636,34 @@ static ssize_t move_from_pipe(struct inode *inode, struct file *out,
 		}
 
 		if (do_wakeup) {
-			wake_up_interruptible_sync(PIPE_WAIT(*inode));
-			kill_fasync(PIPE_FASYNC_WRITERS(*inode),SIGIO,POLL_OUT);
+			smp_mb();
+			if (waitqueue_active(&pipe->wait))
+				wake_up_interruptible_sync(&pipe->wait);
+			kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
 			do_wakeup = 0;
 		}
 
-		pipe_wait(inode);
+		pipe_wait(pipe);
 	}
 
-	mutex_unlock(PIPE_MUTEX(*inode));
+	if (pipe->inode)
+		mutex_unlock(&pipe->inode->i_mutex);
 
 	if (do_wakeup) {
-		wake_up_interruptible(PIPE_WAIT(*inode));
-		kill_fasync(PIPE_FASYNC_WRITERS(*inode), SIGIO, POLL_OUT);
+		smp_mb();
+		if (waitqueue_active(&pipe->wait))
+			wake_up_interruptible(&pipe->wait);
+		kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
 	}
 
-	mutex_lock(&out->f_mapping->host->i_mutex);
 	out->f_pos = sd.pos;
-	mutex_unlock(&out->f_mapping->host->i_mutex);
 	return ret;
 
 }
 
 /**
  * generic_file_splice_write - splice data from a pipe to a file
- * @inode:	pipe inode
+ * @pipe:	pipe info
  * @out:	file to write to
  * @len:	number of bytes to splice
  * @flags:	splice modifier flags
@@ -631,14 +672,17 @@ static ssize_t move_from_pipe(struct inode *inode, struct file *out,
  * the given pipe inode to the given file.
  *
  */
-ssize_t generic_file_splice_write(struct inode *inode, struct file *out,
-				  size_t len, unsigned int flags)
+ssize_t
+generic_file_splice_write(struct pipe_inode_info *pipe, struct file *out,
+			  size_t len, unsigned int flags)
 {
 	struct address_space *mapping = out->f_mapping;
-	ssize_t ret = move_from_pipe(inode, out, len, flags, pipe_to_file);
+	ssize_t ret;
+
+	ret = move_from_pipe(pipe, out, len, flags, pipe_to_file);
 
 	/*
-	 * if file or inode is SYNC and we actually wrote some data, sync it
+	 * If file or inode is SYNC and we actually wrote some data, sync it.
 	 */
 	if (unlikely((out->f_flags & O_SYNC) || IS_SYNC(mapping->host))
 	    && ret > 0) {
@@ -647,7 +691,7 @@ ssize_t generic_file_splice_write(struct inode *inode, struct file *out,
 
 		mutex_lock(&inode->i_mutex);
 		err = generic_osync_inode(mapping->host, mapping,
-						OSYNC_METADATA|OSYNC_DATA);
+					  OSYNC_METADATA|OSYNC_DATA);
 		mutex_unlock(&inode->i_mutex);
 
 		if (err)
@@ -670,10 +714,10 @@ EXPORT_SYMBOL(generic_file_splice_write);
  * is involved.
  *
  */
-ssize_t generic_splice_sendpage(struct inode *inode, struct file *out,
+ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe, struct file *out,
 				size_t len, unsigned int flags)
 {
-	return move_from_pipe(inode, out, len, flags, pipe_to_sendpage);
+	return move_from_pipe(pipe, out, len, flags, pipe_to_sendpage);
 }
 
 EXPORT_SYMBOL(generic_splice_sendpage);
@@ -681,19 +725,20 @@ EXPORT_SYMBOL(generic_splice_sendpage);
 /*
  * Attempt to initiate a splice from pipe to file.
  */
-static long do_splice_from(struct inode *pipe, struct file *out, size_t len,
-			   unsigned int flags)
+static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
+			   size_t len, unsigned int flags)
 {
 	loff_t pos;
 	int ret;
 
-	if (!out->f_op || !out->f_op->splice_write)
+	if (unlikely(!out->f_op || !out->f_op->splice_write))
 		return -EINVAL;
 
-	if (!(out->f_mode & FMODE_WRITE))
+	if (unlikely(!(out->f_mode & FMODE_WRITE)))
 		return -EBADF;
 
 	pos = out->f_pos;
+
 	ret = rw_verify_area(WRITE, out, &pos, len);
 	if (unlikely(ret < 0))
 		return ret;
@@ -704,19 +749,20 @@ static long do_splice_from(struct inode *pipe, struct file *out, size_t len,
 /*
  * Attempt to initiate a splice from a file to a pipe.
  */
-static long do_splice_to(struct file *in, struct inode *pipe, size_t len,
-			 unsigned int flags)
+static long do_splice_to(struct file *in, struct pipe_inode_info *pipe,
+			 size_t len, unsigned int flags)
 {
 	loff_t pos, isize, left;
 	int ret;
 
-	if (!in->f_op || !in->f_op->splice_read)
+	if (unlikely(!in->f_op || !in->f_op->splice_read))
 		return -EINVAL;
 
-	if (!(in->f_mode & FMODE_READ))
+	if (unlikely(!(in->f_mode & FMODE_READ)))
 		return -EBADF;
 
 	pos = in->f_pos;
+
 	ret = rw_verify_area(READ, in, &pos, len);
 	if (unlikely(ret < 0))
 		return ret;
@@ -726,32 +772,168 @@ static long do_splice_to(struct file *in, struct inode *pipe, size_t len,
 		return 0;
 	
 	left = isize - in->f_pos;
-	if (left < len)
+	if (unlikely(left < len))
 		len = left;
 
 	return in->f_op->splice_read(in, pipe, len, flags);
 }
 
+long do_splice_direct(struct file *in, struct file *out, size_t len,
+		      unsigned int flags)
+{
+	struct pipe_inode_info *pipe;
+	long ret, bytes;
+	umode_t i_mode;
+	int i;
+
+	/*
+	 * We require the input being a regular file, as we don't want to
+	 * randomly drop data for eg socket -> socket splicing. Use the
+	 * piped splicing for that!
+	 */
+	i_mode = in->f_dentry->d_inode->i_mode;
+	if (unlikely(!S_ISREG(i_mode) && !S_ISBLK(i_mode)))
+		return -EINVAL;
+
+	/*
+	 * neither in nor out is a pipe, setup an internal pipe attached to
+	 * 'out' and transfer the wanted data from 'in' to 'out' through that
+	 */
+	pipe = current->splice_pipe;
+	if (unlikely(!pipe)) {
+		pipe = alloc_pipe_info(NULL);
+		if (!pipe)
+			return -ENOMEM;
+
+		/*
+		 * We don't have an immediate reader, but we'll read the stuff
+		 * out of the pipe right after the move_to_pipe(). So set
+		 * PIPE_READERS appropriately.
+		 */
+		pipe->readers = 1;
+
+		current->splice_pipe = pipe;
+	}
+
+	/*
+	 * Do the splice.
+	 */
+	ret = 0;
+	bytes = 0;
+
+	while (len) {
+		size_t read_len, max_read_len;
+
+		/*
+		 * Do at most PIPE_BUFFERS pages worth of transfer:
+		 */
+		max_read_len = min(len, (size_t)(PIPE_BUFFERS*PAGE_SIZE));
+
+		ret = do_splice_to(in, pipe, max_read_len, flags);
+		if (unlikely(ret < 0))
+			goto out_release;
+
+		read_len = ret;
+
+		/*
+		 * NOTE: nonblocking mode only applies to the input. We
+		 * must not do the output in nonblocking mode as then we
+		 * could get stuck data in the internal pipe:
+		 */
+		ret = do_splice_from(pipe, out, read_len,
+				     flags & ~SPLICE_F_NONBLOCK);
+		if (unlikely(ret < 0))
+			goto out_release;
+
+		bytes += ret;
+		len -= ret;
+
+		/*
+		 * In nonblocking mode, if we got back a short read then
+		 * that was due to either an IO error or due to the
+		 * pagecache entry not being there. In the IO error case
+		 * the _next_ splice attempt will produce a clean IO error
+		 * return value (not a short read), so in both cases it's
+		 * correct to break out of the loop here:
+		 */
+		if ((flags & SPLICE_F_NONBLOCK) && (read_len < max_read_len))
+			break;
+	}
+
+	pipe->nrbufs = pipe->curbuf = 0;
+
+	return bytes;
+
+out_release:
+	/*
+	 * If we did an incomplete transfer we must release
+	 * the pipe buffers in question:
+	 */
+	for (i = 0; i < PIPE_BUFFERS; i++) {
+		struct pipe_buffer *buf = pipe->bufs + i;
+
+		if (buf->ops) {
+			buf->ops->release(pipe, buf);
+			buf->ops = NULL;
+		}
+	}
+	pipe->nrbufs = pipe->curbuf = 0;
+
+	/*
+	 * If we transferred some data, return the number of bytes:
+	 */
+	if (bytes > 0)
+		return bytes;
+
+	return ret;
+}
+
+EXPORT_SYMBOL(do_splice_direct);
+
 /*
  * Determine where to splice to/from.
  */
-static long do_splice(struct file *in, struct file *out, size_t len,
-		      unsigned int flags)
+static long do_splice(struct file *in, loff_t __user *off_in,
+		      struct file *out, loff_t __user *off_out,
+		      size_t len, unsigned int flags)
 {
-	struct inode *pipe;
+	struct pipe_inode_info *pipe;
+
+	pipe = in->f_dentry->d_inode->i_pipe;
+	if (pipe) {
+		if (off_in)
+			return -ESPIPE;
+		if (off_out) {
+			if (out->f_op->llseek == no_llseek)
+				return -EINVAL;
+			if (copy_from_user(&out->f_pos, off_out,
+					   sizeof(loff_t)))
+				return -EFAULT;
+		}
 
-	pipe = in->f_dentry->d_inode;
-	if (pipe->i_pipe)
 		return do_splice_from(pipe, out, len, flags);
+	}
+
+	pipe = out->f_dentry->d_inode->i_pipe;
+	if (pipe) {
+		if (off_out)
+			return -ESPIPE;
+		if (off_in) {
+			if (in->f_op->llseek == no_llseek)
+				return -EINVAL;
+			if (copy_from_user(&in->f_pos, off_in, sizeof(loff_t)))
+				return -EFAULT;
+		}
 
-	pipe = out->f_dentry->d_inode;
-	if (pipe->i_pipe)
 		return do_splice_to(in, pipe, len, flags);
+	}
 
 	return -EINVAL;
 }
 
-asmlinkage long sys_splice(int fdin, int fdout, size_t len, unsigned int flags)
+asmlinkage long sys_splice(int fd_in, loff_t __user *off_in,
+			   int fd_out, loff_t __user *off_out,
+			   size_t len, unsigned int flags)
 {
 	long error;
 	struct file *in, *out;
@@ -761,13 +943,15 @@ asmlinkage long sys_splice(int fdin, int fdout, size_t len, unsigned int flags)
 		return 0;
 
 	error = -EBADF;
-	in = fget_light(fdin, &fput_in);
+	in = fget_light(fd_in, &fput_in);
 	if (in) {
 		if (in->f_mode & FMODE_READ) {
-			out = fget_light(fdout, &fput_out);
+			out = fget_light(fd_out, &fput_out);
 			if (out) {
 				if (out->f_mode & FMODE_WRITE)
-					error = do_splice(in, out, len, flags);
+					error = do_splice(in, off_in,
+							  out, off_out,
+							  len, flags);
 				fput_light(out, fput_out);
 			}
 		}
diff --git a/fs/xfs/linux-2.6/xfs_file.c b/fs/xfs/linux-2.6/xfs_file.c
index ae4c4754ed3..269721af02f 100644
--- a/fs/xfs/linux-2.6/xfs_file.c
+++ b/fs/xfs/linux-2.6/xfs_file.c
@@ -252,7 +252,7 @@ xfs_file_sendfile_invis(
 STATIC ssize_t
 xfs_file_splice_read(
 	struct file		*infilp,
-	struct inode		*pipe,
+	struct pipe_inode_info	*pipe,
 	size_t			len,
 	unsigned int		flags)
 {
@@ -266,7 +266,7 @@ xfs_file_splice_read(
 STATIC ssize_t
 xfs_file_splice_read_invis(
 	struct file		*infilp,
-	struct inode		*pipe,
+	struct pipe_inode_info	*pipe,
 	size_t			len,
 	unsigned int		flags)
 {
@@ -279,7 +279,7 @@ xfs_file_splice_read_invis(
 
 STATIC ssize_t
 xfs_file_splice_write(
-	struct inode		*pipe,
+	struct pipe_inode_info	*pipe,
 	struct file		*outfilp,
 	size_t			len,
 	unsigned int		flags)
@@ -293,7 +293,7 @@ xfs_file_splice_write(
 
 STATIC ssize_t
 xfs_file_splice_write_invis(
-	struct inode		*pipe,
+	struct pipe_inode_info	*pipe,
 	struct file		*outfilp,
 	size_t			len,
 	unsigned int		flags)
diff --git a/fs/xfs/linux-2.6/xfs_lrw.c b/fs/xfs/linux-2.6/xfs_lrw.c
index 90cd314acba..74a52937f20 100644
--- a/fs/xfs/linux-2.6/xfs_lrw.c
+++ b/fs/xfs/linux-2.6/xfs_lrw.c
@@ -338,7 +338,7 @@ ssize_t
 xfs_splice_read(
 	bhv_desc_t		*bdp,
 	struct file		*infilp,
-	struct inode		*pipe,
+	struct pipe_inode_info	*pipe,
 	size_t			count,
 	int			flags,
 	int			ioflags,
@@ -380,7 +380,7 @@ xfs_splice_read(
 ssize_t
 xfs_splice_write(
 	bhv_desc_t		*bdp,
-	struct inode		*pipe,
+	struct pipe_inode_info	*pipe,
 	struct file		*outfilp,
 	size_t			count,
 	int			flags,
diff --git a/fs/xfs/linux-2.6/xfs_lrw.h b/fs/xfs/linux-2.6/xfs_lrw.h
index eaa5659713f..55c689a86ad 100644
--- a/fs/xfs/linux-2.6/xfs_lrw.h
+++ b/fs/xfs/linux-2.6/xfs_lrw.h
@@ -94,9 +94,9 @@ extern ssize_t xfs_sendfile(struct bhv_desc *, struct file *,
 				loff_t *, int, size_t, read_actor_t,
 				void *, struct cred *);
 extern ssize_t xfs_splice_read(struct bhv_desc *, struct file *,
-				struct inode *, size_t, int, int,
+				struct pipe_inode_info *, size_t, int, int,
 				struct cred *);
-extern ssize_t xfs_splice_write(struct bhv_desc *, struct inode *,
+extern ssize_t xfs_splice_write(struct bhv_desc *, struct pipe_inode_info *,
 				struct file *, size_t, int, int,
 				struct cred *);
 
diff --git a/fs/xfs/linux-2.6/xfs_vnode.h b/fs/xfs/linux-2.6/xfs_vnode.h
index 6f1c79a28f8..88b09f18628 100644
--- a/fs/xfs/linux-2.6/xfs_vnode.h
+++ b/fs/xfs/linux-2.6/xfs_vnode.h
@@ -174,9 +174,9 @@ typedef ssize_t (*vop_sendfile_t)(bhv_desc_t *, struct file *,
 				loff_t *, int, size_t, read_actor_t,
 				void *, struct cred *);
 typedef ssize_t (*vop_splice_read_t)(bhv_desc_t *, struct file *,
-				struct inode *, size_t, int, int,
+				struct pipe_inode_info *, size_t, int, int,
 				struct cred *);
-typedef ssize_t (*vop_splice_write_t)(bhv_desc_t *, struct inode *,
+typedef ssize_t (*vop_splice_write_t)(bhv_desc_t *, struct pipe_inode_info *,
 				struct file *, size_t, int, int,
 				struct cred *);
 typedef int	(*vop_ioctl_t)(bhv_desc_t *, struct inode *, struct file *,