1 files changed, 659 insertions, 207 deletions

diff --git a/fs/splice.c b/fs/splice.c
index bfa42a2..0559e75 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>
@@ -49,7 +50,8 @@ static int page_cache_pipe_buf_steal(struct pipe_inode_info *info,
 	struct page *page = buf->page;
 	struct address_space *mapping = page_mapping(page);
 
-	WARN_ON(!PageLocked(page));
+	lock_page(page);
+
 	WARN_ON(!PageUptodate(page));
 
 	/*
@@ -64,8 +66,10 @@ static int page_cache_pipe_buf_steal(struct pipe_inode_info *info,
 	if (PagePrivate(page))
 		try_to_release_page(page, mapping_gfp_mask(mapping));
 
-	if (!remove_mapping(mapping, page))
+	if (!remove_mapping(mapping, page)) {
+		unlock_page(page);
 		return 1;
+	}
 
 	buf->flags |= PIPE_BUF_FLAG_STOLEN | PIPE_BUF_FLAG_LRU;
 	return 0;
@@ -84,69 +88,89 @@ 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);
 
-	if (!page->mapping) {
+		/*
+		 * Page got truncated/unhashed. This will cause a 0-byte
+		 * splice, if this is the first page.
+		 */
+		if (!page->mapping) {
+			err = -ENODATA;
+			goto error;
+		}
+
+		/*
+		 * 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);
 }
 
+static void page_cache_pipe_buf_get(struct pipe_inode_info *info,
+				    struct pipe_buffer *buf)
+{
+	page_cache_get(buf->page);
+}
+
 static struct pipe_buf_operations page_cache_pipe_buf_ops = {
 	.can_merge = 0,
 	.map = page_cache_pipe_buf_map,
 	.unmap = page_cache_pipe_buf_unmap,
 	.release = page_cache_pipe_buf_release,
 	.steal = page_cache_pipe_buf_steal,
+	.get = page_cache_pipe_buf_get,
 };
 
 /*
  * 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,
-			    int nr_pages, unsigned long offset,
-			    unsigned long len, unsigned int flags)
+static ssize_t move_to_pipe(struct pipe_inode_info *pipe, struct page **pages,
+			    int nr_pages, unsigned long len,
+			    unsigned int offset, 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 +182,9 @@ 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 +193,7 @@ 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 +212,26 @@ 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)
@@ -211,96 +240,155 @@ 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, loff_t *ppos,
+			   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];
+	unsigned int loff, offset, nr_pages;
+	struct page *pages[PIPE_BUFFERS];
 	struct page *page;
-	pgoff_t index, pidx;
-	int i, j;
+	pgoff_t index, end_index;
+	loff_t isize;
+	size_t bytes;
+	int i, error;
 
-	index = in->f_pos >> PAGE_CACHE_SHIFT;
-	offset = in->f_pos & ~PAGE_CACHE_MASK;
+	index = *ppos >> PAGE_CACHE_SHIFT;
+	loff = offset = *ppos & ~PAGE_CACHE_MASK;
 	nr_pages = (len + offset + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
 
 	if (nr_pages > PIPE_BUFFERS)
 		nr_pages = PIPE_BUFFERS;
 
 	/*
-	 * initiate read-ahead on this page range
+	 * 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.
 	 */
-	do_page_cache_readahead(mapping, in, index, nr_pages);
+	if (!offset || nr_pages > 1)
+		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).
+	 * Now fill in the holes:
 	 */
-	i = find_get_pages(mapping, index, nr_pages, pages);
+	error = 0;
+	bytes = 0;
+	for (i = 0; i < nr_pages; i++, index++) {
+		unsigned int this_len;
 
-	/*
-	 * common case - we found all pages and they are contiguous,
-	 * kick them off
-	 */
-	if (i && (pages[i - 1]->index == index + i - 1))
-		goto splice_them;
+		if (!len)
+			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];
+		/*
+		 * this_len is the max we'll use from this page
+		 */
+		this_len = min(len, PAGE_CACHE_SIZE - loff);
+find_page:
+		/*
+		 * lookup the page for this index
+		 */
+		page = find_get_page(mapping, index);
+		if (!page) {
+			/*
+			 * 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)) {
+			/*
+			 * If in nonblock mode then dont block on waiting
+			 * for an in-flight io page
+			 */
+			if (flags & SPLICE_F_NONBLOCK)
+				break;
+
+			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;
-	}
 
-	if (!i) {
-		for (i = 0; i < nr_pages; i++) {
-			 if (shadow[i])
-				page_cache_release(shadow[i]);
+			/*
+			 * i_size must be checked after ->readpage().
+			 */
+			isize = i_size_read(mapping->host);
+			end_index = (isize - 1) >> PAGE_CACHE_SHIFT;
+			if (unlikely(!isize || index > end_index)) {
+				page_cache_release(page);
+				break;
+			}
+
+			/*
+			 * if this is the last page, see if we need to shrink
+			 * the length and stop
+			 */
+			if (end_index == index) {
+				loff = PAGE_CACHE_SIZE - (isize & ~PAGE_CACHE_MASK);
+				if (bytes + loff > isize) {
+					page_cache_release(page);
+					break;
+				}
+				/*
+				 * force quit after adding this page
+				 */
+				nr_pages = i;
+				this_len = min(this_len, loff);
+			}
 		}
-		return 0;
+fill_it:
+		pages[i] = page;
+		bytes += this_len;
+		len -= this_len;
+		loff = 0;
 	}
 
-	memcpy(pages, shadow, i * sizeof(struct page *));
+	if (i)
+		return move_to_pipe(pipe, pages, i, bytes, offset, flags);
 
-	/*
-	 * Now we splice them into the pipe..
-	 */
-splice_them:
-	return move_to_pipe(pipe, pages, i, offset, len, flags);
+	return error;
 }
 
 /**
@@ -311,30 +399,34 @@ 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,
-				 size_t len, unsigned int flags)
+ssize_t generic_file_splice_read(struct file *in, loff_t *ppos,
+				 struct pipe_inode_info *pipe, size_t len,
+				 unsigned int flags)
 {
 	ssize_t spliced;
 	int ret;
 
 	ret = 0;
 	spliced = 0;
+
 	while (len) {
-		ret = __generic_file_splice_read(in, pipe, len, flags);
+		ret = __generic_file_splice_read(in, ppos, pipe, len, flags);
 
-		if (ret <= 0)
+		if (ret < 0)
 			break;
+		else if (!ret) {
+			if (spliced)
+				break;
+			if (flags & SPLICE_F_NONBLOCK) {
+				ret = -EAGAIN;
+				break;
+			}
+		}
 
-		in->f_pos += ret;
+		*ppos += ret;
 		len -= ret;
 		spliced += ret;
-
-		if (!(flags & SPLICE_F_NONBLOCK))
-			continue;
-		ret = -EAGAIN;
-		break;
 	}
 
 	if (spliced)
@@ -360,10 +452,10 @@ 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 +506,7 @@ 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))
@@ -424,12 +516,13 @@ 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) {
 		/*
 		 * If steal succeeds, buf->page is now pruned from the vm
-		 * side (LRU and page cache) and we can reuse it.
+		 * side (LRU and page cache) and we can reuse it. The page
+		 * will also be looked on successful return.
 		 */
 		if (buf->ops->steal(info, buf))
 			goto find_page;
@@ -442,15 +535,27 @@ static int pipe_to_file(struct pipe_inode_info *info, struct pipe_buffer *buf,
 			lru_cache_add(page);
 	} else {
 find_page:
-		ret = -ENOMEM;
-		page = find_or_create_page(mapping, index, gfp_mask);
-		if (!page)
-			goto out;
+		page = find_lock_page(mapping, index);
+		if (!page) {
+			ret = -ENOMEM;
+			page = page_cache_alloc_cold(mapping);
+			if (unlikely(!page))
+				goto out_nomem;
+
+			/*
+			 * This will also lock the page
+			 */
+			ret = add_to_page_cache_lru(page, mapping, index,
+						    gfp_mask);
+			if (unlikely(ret))
+				goto out;
+		}
 
 		/*
-		 * If the page is uptodate, it is also locked. If it isn't
-		 * uptodate, we can mark it uptodate if we are filling the
-		 * full page. Otherwise we need to read it in first...
+		 * We get here with the page locked. If the page is also
+		 * uptodate, we don't need to do more. If it isn't, we
+		 * may need to bring it in if we are not going to overwrite
+		 * the full page.
 		 */
 		if (!PageUptodate(page)) {
 			if (sd->len < PAGE_CACHE_SIZE) {
@@ -462,7 +567,7 @@ find_page:
 
 				if (!PageUptodate(page)) {
 					/*
-					 * page got invalidated, repeat
+					 * Page got invalidated, repeat.
 					 */
 					if (!page->mapping) {
 						unlock_page(page);
@@ -472,10 +577,8 @@ find_page:
 					ret = -EIO;
 					goto out;
 				}
-			} else {
-				WARN_ON(!PageLocked(page));
+			} else
 				SetPageUptodate(page);
-			}
 		}
 	}
 
@@ -501,12 +604,14 @@ find_page:
 	} else if (ret)
 		goto out;
 
+	mark_page_accessed(page);
 	balance_dirty_pages_ratelimited(mapping);
 out:
-	if (!(buf->flags & PIPE_BUF_FLAG_STOLEN)) {
+	if (!(buf->flags & PIPE_BUF_FLAG_STOLEN))
 		page_cache_release(page);
-		unlock_page(page);
-	}
+
+	unlock_page(page);
+out_nomem:
 	buf->ops->unmap(info, buf);
 	return ret;
 }
@@ -519,11 +624,10 @@ 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,
-			      size_t len, unsigned int flags,
+static ssize_t move_from_pipe(struct pipe_inode_info *pipe, struct file *out,
+			      loff_t *ppos, size_t len, unsigned int flags,
 			      splice_actor *actor)
 {
-	struct pipe_inode_info *info;
 	int ret, do_wakeup, err;
 	struct splice_desc sd;
 
@@ -533,24 +637,21 @@ static ssize_t move_from_pipe(struct inode *inode, struct file *out,
 	sd.total_len = len;
 	sd.flags = flags;
 	sd.file = out;
-	sd.pos = out->f_pos;
+	sd.pos = *ppos;
 
-	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 +662,14 @@ 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 +678,11 @@ 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 +700,32 @@ 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
@@ -631,27 +734,34 @@ 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,
+			  loff_t *ppos, 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;
 
-	/*
-	 * 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) {
+	ret = move_from_pipe(pipe, out, ppos, len, flags, pipe_to_file);
+	if (ret > 0) {
 		struct inode *inode = mapping->host;
-		int err;
 
-		mutex_lock(&inode->i_mutex);
-		err = generic_osync_inode(mapping->host, mapping,
-						OSYNC_METADATA|OSYNC_DATA);
-		mutex_unlock(&inode->i_mutex);
+		*ppos += ret;
+
+		/*
+		 * If file or inode is SYNC and we actually wrote some data,
+		 * sync it.
+		 */
+		if (unlikely((out->f_flags & O_SYNC) || IS_SYNC(inode))) {
+			int err;
+
+			mutex_lock(&inode->i_mutex);
+			err = generic_osync_inode(inode, mapping,
+						  OSYNC_METADATA|OSYNC_DATA);
+			mutex_unlock(&inode->i_mutex);
 
-		if (err)
-			ret = err;
+			if (err)
+				ret = err;
+		}
 	}
 
 	return ret;
@@ -670,10 +780,10 @@ EXPORT_SYMBOL(generic_file_splice_write);
  * is involved.
  *
  */
-ssize_t generic_splice_sendpage(struct inode *inode, struct file *out,
-				size_t len, unsigned int flags)
+ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe, struct file *out,
+				loff_t *ppos, size_t len, unsigned int flags)
 {
-	return move_from_pipe(inode, out, len, flags, pipe_to_sendpage);
+	return move_from_pipe(pipe, out, ppos, len, flags, pipe_to_sendpage);
 }
 
 EXPORT_SYMBOL(generic_splice_sendpage);
@@ -681,77 +791,228 @@ 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,
+			   loff_t *ppos, 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);
+	ret = rw_verify_area(WRITE, out, ppos, len);
 	if (unlikely(ret < 0))
 		return ret;
 
-	return out->f_op->splice_write(pipe, out, len, flags);
+	return out->f_op->splice_write(pipe, out, ppos, len, flags);
 }
 
 /*
  * 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,
+static long do_splice_to(struct file *in, loff_t *ppos,
+			 struct pipe_inode_info *pipe, size_t len,
 			 unsigned int flags)
 {
-	loff_t pos, isize, left;
+	loff_t 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);
+	ret = rw_verify_area(READ, in, ppos, len);
 	if (unlikely(ret < 0))
 		return ret;
 
 	isize = i_size_read(in->f_mapping->host);
-	if (unlikely(in->f_pos >= isize))
+	if (unlikely(*ppos >= isize))
 		return 0;
 	
-	left = isize - in->f_pos;
-	if (left < len)
+	left = isize - *ppos;
+	if (unlikely(left < len))
 		len = left;
 
-	return in->f_op->splice_read(in, pipe, len, flags);
+	return in->f_op->splice_read(in, ppos, pipe, len, flags);
+}
+
+long do_splice_direct(struct file *in, loff_t *ppos, struct file *out,
+		      size_t len, unsigned int flags)
+{
+	struct pipe_inode_info *pipe;
+	long ret, bytes;
+	loff_t out_off;
+	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;
+	out_off = 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, ppos, 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, &out_off, 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;
+	loff_t offset, *off;
+	long ret;
+
+	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(&offset, off_out, sizeof(loff_t)))
+				return -EFAULT;
+			off = &offset;
+		} else
+			off = &out->f_pos;
+
+		ret = do_splice_from(pipe, out, off, len, flags);
+
+		if (off_out && copy_to_user(off_out, off, sizeof(loff_t)))
+			ret = -EFAULT;
 
-	pipe = in->f_dentry->d_inode;
-	if (pipe->i_pipe)
-		return do_splice_from(pipe, out, len, flags);
+		return ret;
+	}
+
+	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(&offset, off_in, sizeof(loff_t)))
+				return -EFAULT;
+			off = &offset;
+		} else
+			off = &in->f_pos;
 
-	pipe = out->f_dentry->d_inode;
-	if (pipe->i_pipe)
-		return do_splice_to(in, pipe, len, flags);
+		ret = do_splice_to(in, off, pipe, len, flags);
+
+		if (off_in && copy_to_user(off_in, off, sizeof(loff_t)))
+			ret = -EFAULT;
+
+		return ret;
+	}
 
 	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 +1022,15 @@ 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);
 			}
 		}
@@ -777,3 +1040,192 @@ asmlinkage long sys_splice(int fdin, int fdout, size_t len, unsigned int flags)
 
 	return error;
 }
+
+/*
+ * Link contents of ipipe to opipe.
+ */
+static int link_pipe(struct pipe_inode_info *ipipe,
+		     struct pipe_inode_info *opipe,
+		     size_t len, unsigned int flags)
+{
+	struct pipe_buffer *ibuf, *obuf;
+	int ret, do_wakeup, i, ipipe_first;
+
+	ret = do_wakeup = ipipe_first = 0;
+
+	/*
+	 * Potential ABBA deadlock, work around it by ordering lock
+	 * grabbing by inode address. Otherwise two different processes
+	 * could deadlock (one doing tee from A -> B, the other from B -> A).
+	 */
+	if (ipipe->inode < opipe->inode) {
+		ipipe_first = 1;
+		mutex_lock(&ipipe->inode->i_mutex);
+		mutex_lock(&opipe->inode->i_mutex);
+	} else {
+		mutex_lock(&opipe->inode->i_mutex);
+		mutex_lock(&ipipe->inode->i_mutex);
+	}
+
+	for (i = 0;; i++) {
+		if (!opipe->readers) {
+			send_sig(SIGPIPE, current, 0);
+			if (!ret)
+				ret = -EPIPE;
+			break;
+		}
+		if (ipipe->nrbufs - i) {
+			ibuf = ipipe->bufs + ((ipipe->curbuf + i) & (PIPE_BUFFERS - 1));
+
+			/*
+			 * If we have room, fill this buffer
+			 */
+			if (opipe->nrbufs < PIPE_BUFFERS) {
+				int nbuf = (opipe->curbuf + opipe->nrbufs) & (PIPE_BUFFERS - 1);
+
+				/*
+				 * Get a reference to this pipe buffer,
+				 * so we can copy the contents over.
+				 */
+				ibuf->ops->get(ipipe, ibuf);
+
+				obuf = opipe->bufs + nbuf;
+				*obuf = *ibuf;
+
+				if (obuf->len > len)
+					obuf->len = len;
+
+				opipe->nrbufs++;
+				do_wakeup = 1;
+				ret += obuf->len;
+				len -= obuf->len;
+
+				if (!len)
+					break;
+				if (opipe->nrbufs < PIPE_BUFFERS)
+					continue;
+			}
+
+			/*
+			 * We have input available, but no output room.
+			 * If we already copied data, return that. If we
+			 * need to drop the opipe lock, it must be ordered
+			 * last to avoid deadlocks.
+			 */
+			if ((flags & SPLICE_F_NONBLOCK) || !ipipe_first) {
+				if (!ret)
+					ret = -EAGAIN;
+				break;
+			}
+			if (signal_pending(current)) {
+				if (!ret)
+					ret = -ERESTARTSYS;
+				break;
+			}
+			if (do_wakeup) {
+				smp_mb();
+				if (waitqueue_active(&opipe->wait))
+					wake_up_interruptible(&opipe->wait);
+				kill_fasync(&opipe->fasync_readers, SIGIO, POLL_IN);
+				do_wakeup = 0;
+			}
+
+			opipe->waiting_writers++;
+			pipe_wait(opipe);
+			opipe->waiting_writers--;
+			continue;
+		}
+
+		/*
+		 * No input buffers, do the usual checks for available
+		 * writers and blocking and wait if necessary
+		 */
+		if (!ipipe->writers)
+			break;
+		if (!ipipe->waiting_writers) {
+			if (ret)
+				break;
+		}
+		/*
+		 * pipe_wait() drops the ipipe mutex. To avoid deadlocks
+		 * with another process, we can only safely do that if
+		 * the ipipe lock is ordered last.
+		 */
+		if ((flags & SPLICE_F_NONBLOCK) || ipipe_first) {
+			if (!ret)
+				ret = -EAGAIN;
+			break;
+		}
+		if (signal_pending(current)) {
+			if (!ret)
+				ret = -ERESTARTSYS;
+			break;
+		}
+
+		if (waitqueue_active(&ipipe->wait))
+			wake_up_interruptible_sync(&ipipe->wait);
+		kill_fasync(&ipipe->fasync_writers, SIGIO, POLL_OUT);
+
+		pipe_wait(ipipe);
+	}
+
+	mutex_unlock(&ipipe->inode->i_mutex);
+	mutex_unlock(&opipe->inode->i_mutex);
+
+	if (do_wakeup) {
+		smp_mb();
+		if (waitqueue_active(&opipe->wait))
+			wake_up_interruptible(&opipe->wait);
+		kill_fasync(&opipe->fasync_readers, SIGIO, POLL_IN);
+	}
+
+	return ret;
+}
+
+/*
+ * This is a tee(1) implementation that works on pipes. It doesn't copy
+ * any data, it simply references the 'in' pages on the 'out' pipe.
+ * The 'flags' used are the SPLICE_F_* variants, currently the only
+ * applicable one is SPLICE_F_NONBLOCK.
+ */
+static long do_tee(struct file *in, struct file *out, size_t len,
+		   unsigned int flags)
+{
+	struct pipe_inode_info *ipipe = in->f_dentry->d_inode->i_pipe;
+	struct pipe_inode_info *opipe = out->f_dentry->d_inode->i_pipe;
+
+	/*
+	 * Link ipipe to the two output pipes, consuming as we go along.
+	 */
+	if (ipipe && opipe)
+		return link_pipe(ipipe, opipe, len, flags);
+
+	return -EINVAL;
+}
+
+asmlinkage long sys_tee(int fdin, int fdout, size_t len, unsigned int flags)
+{
+	struct file *in;
+	int error, fput_in;
+
+	if (unlikely(!len))
+		return 0;
+
+	error = -EBADF;
+	in = fget_light(fdin, &fput_in);
+	if (in) {
+		if (in->f_mode & FMODE_READ) {
+			int fput_out;
+			struct file *out = fget_light(fdout, &fput_out);
+
+			if (out) {
+				if (out->f_mode & FMODE_WRITE)
+					error = do_tee(in, out, len, flags);
+				fput_light(out, fput_out);
+			}
+		}
+ 		fput_light(in, fput_in);
+ 	}
+
+	return error;
+}