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] vmsplice: allow user to pass in gift pages
  [PATCH] pipe: enable atomic copying of pipe data to/from user space
  [PATCH] splice: call handle_ra_miss() on failure to lookup page
  [PATCH] Add ->splice_read/splice_write to def_blk_fops
  [PATCH] pipe: introduce ->pin() buffer operation
  [PATCH] splice: fix bugs in pipe_to_file()
  [PATCH] splice: fix bugs with stealing regular pipe pages

fs/block_dev.c

@@ -1104,6 +1104,8 @@ const struct file_operations def_blk_fops = {
 	.readv		= generic_file_readv,
 	.writev		= generic_file_write_nolock,
 	.sendfile	= generic_file_sendfile,
+	.splice_read	= generic_file_splice_read,
+	.splice_write	= generic_file_splice_write,
 };
 
 int ioctl_by_bdev(struct block_device *bdev, unsigned cmd, unsigned long arg)

fs/pipe.c

@@ -55,7 +55,8 @@ void pipe_wait(struct pipe_inode_info *pipe)
 }
 
 static int
-pipe_iov_copy_from_user(void *to, struct iovec *iov, unsigned long len)
+pipe_iov_copy_from_user(void *to, struct iovec *iov, unsigned long len,
+			int atomic)
 {
 	unsigned long copy;
 
@@ -64,8 +65,13 @@ pipe_iov_copy_from_user(void *to, struct iovec *iov, unsigned long len)
 			iov++;
 		copy = min_t(unsigned long, len, iov->iov_len);
 
+		if (atomic) {
+			if (__copy_from_user_inatomic(to, iov->iov_base, copy))
+				return -EFAULT;
+		} else {
 			if (copy_from_user(to, iov->iov_base, copy))
 				return -EFAULT;
+		}
 		to += copy;
 		len -= copy;
 		iov->iov_base += copy;
@@ -75,7 +81,8 @@ pipe_iov_copy_from_user(void *to, struct iovec *iov, unsigned long len)
 }
 
 static int
-pipe_iov_copy_to_user(struct iovec *iov, const void *from, unsigned long len)
+pipe_iov_copy_to_user(struct iovec *iov, const void *from, unsigned long len,
+		      int atomic)
 {
 	unsigned long copy;
 
@@ -84,8 +91,13 @@ pipe_iov_copy_to_user(struct iovec *iov, const void *from, unsigned long len)
 			iov++;
 		copy = min_t(unsigned long, len, iov->iov_len);
 
+		if (atomic) {
+			if (__copy_to_user_inatomic(iov->iov_base, from, copy))
+				return -EFAULT;
+		} else {
 			if (copy_to_user(iov->iov_base, from, copy))
 				return -EFAULT;
+		}
 		from += copy;
 		len -= copy;
 		iov->iov_base += copy;
@@ -94,13 +106,52 @@ pipe_iov_copy_to_user(struct iovec *iov, const void *from, unsigned long len)
 	return 0;
 }
 
+/*
+ * Attempt to pre-fault in the user memory, so we can use atomic copies.
+ * Returns the number of bytes not faulted in.
+ */
+static int iov_fault_in_pages_write(struct iovec *iov, unsigned long len)
+{
+	while (!iov->iov_len)
+		iov++;
+
+	while (len > 0) {
+		unsigned long this_len;
+
+		this_len = min_t(unsigned long, len, iov->iov_len);
+		if (fault_in_pages_writeable(iov->iov_base, this_len))
+			break;
+
+		len -= this_len;
+		iov++;
+	}
+
+	return len;
+}
+
+/*
+ * Pre-fault in the user memory, so we can use atomic copies.
+ */
+static void iov_fault_in_pages_read(struct iovec *iov, unsigned long len)
+{
+	while (!iov->iov_len)
+		iov++;
+
+	while (len > 0) {
+		unsigned long this_len;
+
+		this_len = min_t(unsigned long, len, iov->iov_len);
+		fault_in_pages_readable(iov->iov_base, this_len);
+		len -= this_len;
+		iov++;
+	}
+}
+
 static void anon_pipe_buf_release(struct pipe_inode_info *pipe,
 				  struct pipe_buffer *buf)
 {
 	struct page *page = buf->page;
 
-	buf->flags &= ~PIPE_BUF_FLAG_STOLEN;
-
 	/*
 	 * 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
@@ -112,38 +163,58 @@ static void anon_pipe_buf_release(struct pipe_inode_info *pipe,
 		page_cache_release(page);
 }
 
-static void * anon_pipe_buf_map(struct file *file, struct pipe_inode_info *pipe,
-				struct pipe_buffer *buf)
+void *generic_pipe_buf_map(struct pipe_inode_info *pipe,
+			   struct pipe_buffer *buf, int atomic)
 {
+	if (atomic) {
+		buf->flags |= PIPE_BUF_FLAG_ATOMIC;
+		return kmap_atomic(buf->page, KM_USER0);
+	}
+
 	return kmap(buf->page);
 }
 
-static void anon_pipe_buf_unmap(struct pipe_inode_info *pipe,
-				struct pipe_buffer *buf)
+void generic_pipe_buf_unmap(struct pipe_inode_info *pipe,
+			    struct pipe_buffer *buf, void *map_data)
 {
+	if (buf->flags & PIPE_BUF_FLAG_ATOMIC) {
+		buf->flags &= ~PIPE_BUF_FLAG_ATOMIC;
+		kunmap_atomic(map_data, KM_USER0);
+	} else
 		kunmap(buf->page);
 }
 
 static int anon_pipe_buf_steal(struct pipe_inode_info *pipe,
 			       struct pipe_buffer *buf)
 {
-	buf->flags |= PIPE_BUF_FLAG_STOLEN;
+	struct page *page = buf->page;
+
+	if (page_count(page) == 1) {
+		lock_page(page);
 		return 0;
+	}
+
+	return 1;
 }
 
-static void anon_pipe_buf_get(struct pipe_inode_info *info,
-			      struct pipe_buffer *buf)
+void generic_pipe_buf_get(struct pipe_inode_info *info, struct pipe_buffer *buf)
 {
 	page_cache_get(buf->page);
 }
 
+int generic_pipe_buf_pin(struct pipe_inode_info *info, struct pipe_buffer *buf)
+{
+	return 0;
+}
+
 static struct pipe_buf_operations anon_pipe_buf_ops = {
 	.can_merge = 1,
-	.map = anon_pipe_buf_map,
-	.unmap = anon_pipe_buf_unmap,
+	.map = generic_pipe_buf_map,
+	.unmap = generic_pipe_buf_unmap,
+	.pin = generic_pipe_buf_pin,
 	.release = anon_pipe_buf_release,
 	.steal = anon_pipe_buf_steal,
-	.get = anon_pipe_buf_get,
+	.get = generic_pipe_buf_get,
 };
 
 static ssize_t
@@ -174,22 +245,33 @@ pipe_readv(struct file *filp, const struct iovec *_iov,
 			struct pipe_buf_operations *ops = buf->ops;
 			void *addr;
 			size_t chars = buf->len;
-			int error;
+			int error, atomic;
 
 			if (chars > total_len)
 				chars = total_len;
 
-			addr = ops->map(filp, pipe, buf);
-			if (IS_ERR(addr)) {
+			error = ops->pin(pipe, buf);
+			if (error) {
 				if (!ret)
-					ret = PTR_ERR(addr);
+					error = ret;
 				break;
 			}
-			error = pipe_iov_copy_to_user(iov, addr + buf->offset, chars);
-			ops->unmap(pipe, buf);
+
+			atomic = !iov_fault_in_pages_write(iov, chars);
+redo:
+			addr = ops->map(pipe, buf, atomic);
+			error = pipe_iov_copy_to_user(iov, addr + buf->offset, chars, atomic);
+			ops->unmap(pipe, buf, addr);
 			if (unlikely(error)) {
+				/*
+				 * Just retry with the slow path if we failed.
+				 */
+				if (atomic) {
+					atomic = 0;
+					goto redo;
+				}
 				if (!ret)
-					ret = -EFAULT;
+					ret = error;
 				break;
 			}
 			ret += chars;
@@ -293,21 +375,28 @@ pipe_writev(struct file *filp, const struct iovec *_iov,
 		int offset = buf->offset + buf->len;
 
 		if (ops->can_merge && offset + chars <= PAGE_SIZE) {
+			int error, atomic = 1;
 			void *addr;
-			int error;
 
-			addr = ops->map(filp, pipe, buf);
-			if (IS_ERR(addr)) {
-				error = PTR_ERR(addr);
+			error = ops->pin(pipe, buf);
+			if (error)
 				goto out;
-			}
+
+			iov_fault_in_pages_read(iov, chars);
+redo1:
+			addr = ops->map(pipe, buf, atomic);
 			error = pipe_iov_copy_from_user(offset + addr, iov,
-							chars);
-			ops->unmap(pipe, buf);
+							chars, atomic);
+			ops->unmap(pipe, buf, addr);
 			ret = error;
 			do_wakeup = 1;
-			if (error)
+			if (error) {
+				if (atomic) {
+					atomic = 0;
+					goto redo1;
+				}
 				goto out;
+			}
 			buf->len += chars;
 			total_len -= chars;
 			ret = chars;
@@ -330,7 +419,8 @@ pipe_writev(struct file *filp, const struct iovec *_iov,
 			int newbuf = (pipe->curbuf + bufs) & (PIPE_BUFFERS-1);
 			struct pipe_buffer *buf = pipe->bufs + newbuf;
 			struct page *page = pipe->tmp_page;
-			int error;
+			char *src;
+			int error, atomic = 1;
 
 			if (!page) {
 				page = alloc_page(GFP_HIGHUSER);
@@ -350,11 +440,27 @@ pipe_writev(struct file *filp, const struct iovec *_iov,
 			if (chars > total_len)
 				chars = total_len;
 
-			error = pipe_iov_copy_from_user(kmap(page), iov, chars);
+			iov_fault_in_pages_read(iov, chars);
+redo2:
+			if (atomic)
+				src = kmap_atomic(page, KM_USER0);
+			else
+				src = kmap(page);
+
+			error = pipe_iov_copy_from_user(src, iov, chars,
+							atomic);
+			if (atomic)
+				kunmap_atomic(src, KM_USER0);
+			else
 				kunmap(page);
+
 			if (unlikely(error)) {
+				if (atomic) {
+					atomic = 0;
+					goto redo2;
+				}
 				if (!ret)
-					ret = -EFAULT;
+					ret = error;
 				break;
 			}
 			ret += chars;

fs/splice.c

@@ -78,7 +78,7 @@ static int page_cache_pipe_buf_steal(struct pipe_inode_info *info,
 		return 1;
 	}
 
-	buf->flags |= PIPE_BUF_FLAG_STOLEN | PIPE_BUF_FLAG_LRU;
+	buf->flags |= PIPE_BUF_FLAG_LRU;
 	return 0;
 }
 
@@ -87,11 +87,10 @@ static void page_cache_pipe_buf_release(struct pipe_inode_info *info,
 {
 	page_cache_release(buf->page);
 	buf->page = NULL;
-	buf->flags &= ~(PIPE_BUF_FLAG_STOLEN | PIPE_BUF_FLAG_LRU);
+	buf->flags &= ~PIPE_BUF_FLAG_LRU;
 }
 
-static void *page_cache_pipe_buf_map(struct file *file,
-				     struct pipe_inode_info *info,
+static int page_cache_pipe_buf_pin(struct pipe_inode_info *info,
 				   struct pipe_buffer *buf)
 {
 	struct page *page = buf->page;
@@ -118,64 +117,44 @@ static void *page_cache_pipe_buf_map(struct file *file,
 		}
 
 		/*
-		 * Page is ok afterall, fall through to mapping.
+		 * Page is ok afterall, we are done.
 		 */
 		unlock_page(page);
 	}
 
-	return kmap(page);
+	return 0;
 error:
 	unlock_page(page);
-	return ERR_PTR(err);
-}
-
-static void page_cache_pipe_buf_unmap(struct pipe_inode_info *info,
-				      struct pipe_buffer *buf)
-{
-	kunmap(buf->page);
-}
-
-static void *user_page_pipe_buf_map(struct file *file,
-				    struct pipe_inode_info *pipe,
-				    struct pipe_buffer *buf)
-{
-	return kmap(buf->page);
-}
-
-static void user_page_pipe_buf_unmap(struct pipe_inode_info *pipe,
-				     struct pipe_buffer *buf)
-{
-	kunmap(buf->page);
-}
-
-static void page_cache_pipe_buf_get(struct pipe_inode_info *info,
-				    struct pipe_buffer *buf)
-{
-	page_cache_get(buf->page);
+	return err;
 }
 
 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,
+	.map = generic_pipe_buf_map,
+	.unmap = generic_pipe_buf_unmap,
+	.pin = page_cache_pipe_buf_pin,
 	.release = page_cache_pipe_buf_release,
 	.steal = page_cache_pipe_buf_steal,
-	.get = page_cache_pipe_buf_get,
+	.get = generic_pipe_buf_get,
 };
 
 static int user_page_pipe_buf_steal(struct pipe_inode_info *pipe,
 				    struct pipe_buffer *buf)
 {
+	if (!(buf->flags & PIPE_BUF_FLAG_GIFT))
 		return 1;
+
+	return 0;
 }
 
 static struct pipe_buf_operations user_page_pipe_buf_ops = {
 	.can_merge = 0,
-	.map = user_page_pipe_buf_map,
-	.unmap = user_page_pipe_buf_unmap,
+	.map = generic_pipe_buf_map,
+	.unmap = generic_pipe_buf_unmap,
+	.pin = generic_pipe_buf_pin,
 	.release = page_cache_pipe_buf_release,
 	.steal = user_page_pipe_buf_steal,
-	.get = page_cache_pipe_buf_get,
+	.get = generic_pipe_buf_get,
 };
 
 /*
@@ -210,6 +189,9 @@ static ssize_t splice_to_pipe(struct pipe_inode_info *pipe,
 			buf->offset = spd->partial[page_nr].offset;
 			buf->len = spd->partial[page_nr].len;
 			buf->ops = spd->ops;
+			if (spd->flags & SPLICE_F_GIFT)
+				buf->flags |= PIPE_BUF_FLAG_GIFT;
+
 			pipe->nrbufs++;
 			page_nr++;
 			ret += buf->len;
@@ -325,6 +307,12 @@ __generic_file_splice_read(struct file *in, loff_t *ppos,
 		 */
 		page = find_get_page(mapping, index);
 		if (!page) {
+			/*
+			 * Make sure the read-ahead engine is notified
+			 * about this failure.
+			 */
+			handle_ra_miss(mapping, &in->f_ra, index);
+
 			/*
 			 * page didn't exist, allocate one.
 			 */
@@ -517,26 +505,16 @@ static int pipe_to_sendpage(struct pipe_inode_info *info,
 {
 	struct file *file = sd->file;
 	loff_t pos = sd->pos;
-	ssize_t ret;
-	void *ptr;
-	int more;
-
-	/*
-	 * 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.
-	 */
-	ptr = buf->ops->map(file, info, buf);
-	if (IS_ERR(ptr))
-		return PTR_ERR(ptr);
+	int ret, more;
 
+	ret = buf->ops->pin(info, buf);
+	if (!ret) {
 		more = (sd->flags & SPLICE_F_MORE) || sd->len < sd->total_len;
 
-	ret = file->f_op->sendpage(file, buf->page, buf->offset, sd->len,
-				   &pos, more);
+		ret = file->f_op->sendpage(file, buf->page, buf->offset,
+					   sd->len, &pos, more);
+	}
 
-	buf->ops->unmap(info, buf);
 	return ret;
 }
 
@@ -569,15 +547,14 @@ static int pipe_to_file(struct pipe_inode_info *info, struct pipe_buffer *buf,
 	unsigned int offset, this_len;
 	struct page *page;
 	pgoff_t index;
-	char *src;
 	int ret;
 
 	/*
 	 * make sure the data in this buffer is uptodate
 	 */
-	src = buf->ops->map(file, info, buf);
-	if (IS_ERR(src))
-		return PTR_ERR(src);
+	ret = buf->ops->pin(info, buf);
+	if (unlikely(ret))
+		return ret;
 
 	index = sd->pos >> PAGE_CACHE_SHIFT;
 	offset = sd->pos & ~PAGE_CACHE_MASK;
@@ -587,9 +564,10 @@ static int pipe_to_file(struct pipe_inode_info *info, struct pipe_buffer *buf,
 		this_len = PAGE_CACHE_SIZE - offset;
 
 	/*
-	 * Reuse buf page, if SPLICE_F_MOVE is set.
+	 * Reuse buf page, if SPLICE_F_MOVE is set and we are doing a full
+	 * page.
 	 */
-	if (sd->flags & SPLICE_F_MOVE) {
+	if ((sd->flags & SPLICE_F_MOVE) && this_len == PAGE_CACHE_SIZE) {
 		/*
 		 * If steal succeeds, buf->page is now pruned from the vm
 		 * side (LRU and page cache) and we can reuse it. The page
@@ -599,8 +577,12 @@ static int pipe_to_file(struct pipe_inode_info *info, struct pipe_buffer *buf,
 			goto find_page;
 
 		page = buf->page;
-		if (add_to_page_cache(page, mapping, index, gfp_mask))
+		if (add_to_page_cache(page, mapping, index, gfp_mask)) {
+			unlock_page(page);
 			goto find_page;
+		}
+
+		page_cache_get(page);
 
 		if (!(buf->flags & PIPE_BUF_FLAG_LRU))
 			lru_cache_add(page);
@@ -660,34 +642,36 @@ static int pipe_to_file(struct pipe_inode_info *info, struct pipe_buffer *buf,
 	} else if (ret)
 		goto out;
 
-	if (!(buf->flags & PIPE_BUF_FLAG_STOLEN)) {
-		char *dst = kmap_atomic(page, KM_USER0);
+	if (buf->page != page) {
+		/*
+		 * Careful, ->map() uses KM_USER0!
+		 */
+		char *src = buf->ops->map(info, buf, 1);
+		char *dst = kmap_atomic(page, KM_USER1);
 
 		memcpy(dst + offset, src + buf->offset, this_len);
 		flush_dcache_page(page);
-		kunmap_atomic(dst, KM_USER0);
+		kunmap_atomic(dst, KM_USER1);
+		buf->ops->unmap(info, buf, src);
 	}
 
 	ret = mapping->a_ops->commit_write(file, page, offset, offset+this_len);
-	if (ret == AOP_TRUNCATED_PAGE) {
-		page_cache_release(page);
-		goto find_page;
-	} else if (ret)
-		goto out;
-
+	if (!ret) {
 		/*
-	 * Return the number of bytes written.
+		 * Return the number of bytes written and mark page as
+		 * accessed, we are now done!
 		 */
 		ret = this_len;
 		mark_page_accessed(page);
 		balance_dirty_pages_ratelimited(mapping);
+	} else if (ret == AOP_TRUNCATED_PAGE) {
+		page_cache_release(page);
+		goto find_page;
+	}
 out:
-	if (!(buf->flags & PIPE_BUF_FLAG_STOLEN))
 	page_cache_release(page);
-
 	unlock_page(page);
 out_nomem:
-	buf->ops->unmap(info, buf);
 	return ret;
 }
 
@@ -1095,7 +1079,7 @@ static long do_splice(struct file *in, loff_t __user *off_in,
  */
 static int get_iovec_page_array(const struct iovec __user *iov,
 				unsigned int nr_vecs, struct page **pages,
-				struct partial_page *partial)
+				struct partial_page *partial, int aligned)
 {
 	int buffers = 0, error = 0;
 
@@ -1135,6 +1119,15 @@ static int get_iovec_page_array(const struct iovec __user *iov,
 		 * in the user pages.
 		 */
 		off = (unsigned long) base & ~PAGE_MASK;
+
+		/*
+		 * If asked for alignment, the offset must be zero and the
+		 * length a multiple of the PAGE_SIZE.
+		 */
+		error = -EINVAL;
+		if (aligned && (off || len & ~PAGE_MASK))
+			break;
+
 		npages = (off + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
 		if (npages > PIPE_BUFFERS - buffers)
 			npages = PIPE_BUFFERS - buffers;
@@ -1228,7 +1221,8 @@ static long do_vmsplice(struct file *file, const struct iovec __user *iov,
 	else if (unlikely(!nr_segs))
 		return 0;
 
-	spd.nr_pages = get_iovec_page_array(iov, nr_segs, pages, partial);
+	spd.nr_pages = get_iovec_page_array(iov, nr_segs, pages, partial,
+					    flags & SPLICE_F_GIFT);
 	if (spd.nr_pages <= 0)
 		return spd.nr_pages;
 
@@ -1336,6 +1330,12 @@ static int link_pipe(struct pipe_inode_info *ipipe,
 				obuf = opipe->bufs + nbuf;
 				*obuf = *ibuf;
 
+				/*
+				 * Don't inherit the gift flag, we need to
+				 * prevent multiple steals of this page.
+				 */
+				obuf->flags &= ~PIPE_BUF_FLAG_GIFT;
+
 				if (obuf->len > len)
 					obuf->len = len;
 

include/linux/pipe_fs_i.h

@@ -5,8 +5,9 @@
 
 #define PIPE_BUFFERS (16)
 
-#define PIPE_BUF_FLAG_STOLEN	0x01
-#define PIPE_BUF_FLAG_LRU	0x02
+#define PIPE_BUF_FLAG_LRU	0x01	/* page is on the LRU */
+#define PIPE_BUF_FLAG_ATOMIC	0x02	/* was atomically mapped */
+#define PIPE_BUF_FLAG_GIFT	0x04	/* page is a gift */
 
 struct pipe_buffer {
 	struct page *page;
@@ -15,10 +16,23 @@ struct pipe_buffer {
 	unsigned int flags;
 };
 
+/*
+ * Note on the nesting of these functions:
+ *
+ * ->pin()
+ *	->steal()
+ *	...
+ *	->map()
+ *	...
+ *	->unmap()
+ *
+ * That is, ->map() must be called on a pinned buffer, same goes for ->steal().
+ */
 struct pipe_buf_operations {
 	int can_merge;
-	void * (*map)(struct file *, struct pipe_inode_info *, struct pipe_buffer *);
-	void (*unmap)(struct pipe_inode_info *, struct pipe_buffer *);
+	void * (*map)(struct pipe_inode_info *, struct pipe_buffer *, int);
+	void (*unmap)(struct pipe_inode_info *, struct pipe_buffer *, void *);
+	int (*pin)(struct pipe_inode_info *, struct pipe_buffer *);
 	void (*release)(struct pipe_inode_info *, struct pipe_buffer *);
 	int (*steal)(struct pipe_inode_info *, struct pipe_buffer *);
 	void (*get)(struct pipe_inode_info *, struct pipe_buffer *);
@@ -51,6 +65,12 @@ struct pipe_inode_info * alloc_pipe_info(struct inode * inode);
 void free_pipe_info(struct inode * inode);
 void __free_pipe_info(struct pipe_inode_info *);
 
+/* Generic pipe buffer ops functions */
+void *generic_pipe_buf_map(struct pipe_inode_info *, struct pipe_buffer *, int);
+void generic_pipe_buf_unmap(struct pipe_inode_info *, struct pipe_buffer *, void *);
+void generic_pipe_buf_get(struct pipe_inode_info *, struct pipe_buffer *);
+int generic_pipe_buf_pin(struct pipe_inode_info *, struct pipe_buffer *);
+
 /*
  * splice is tied to pipes as a transport (at least for now), so we'll just
  * add the splice flags here.
@@ -60,6 +80,7 @@ void __free_pipe_info(struct pipe_inode_info *);
 				 /* we may still block on the fd we splice */
 				 /* from/to, of course */
 #define SPLICE_F_MORE	(0x04)	/* expect more data */
+#define SPLICE_F_GIFT	(0x08)	/* pages passed in are a gift */
 
 /*
  * Passed to the actors