fs: introduce write_begin, write_end, and perform_write aops

These are intended to replace prepare_write and commit_write with more
flexible alternatives that are also able to avoid the buffered write
deadlock problems efficiently (which prepare_write is unable to do).

[mark.fasheh@oracle.com: API design contributions, code review and fixes]
[akpm@linux-foundation.org: various fixes]
[dmonakhov@sw.ru: new aop block_write_begin fix]
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Signed-off-by: Dmitriy Monakhov <dmonakhov@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

4 files changed, 231 insertions, 129 deletions

diff --git a/fs/buffer.c b/fs/buffer.c
index 9ece6c2..68b8fbd 100644
--- a/fs/buffer.c
+++ b/fs/buffer.c
@@ -1770,6 +1770,48 @@ recover:
 	goto done;
 }
 
+/*
+ * If a page has any new buffers, zero them out here, and mark them uptodate
+ * and dirty so they'll be written out (in order to prevent uninitialised
+ * block data from leaking). And clear the new bit.
+ */
+void page_zero_new_buffers(struct page *page, unsigned from, unsigned to)
+{
+	unsigned int block_start, block_end;
+	struct buffer_head *head, *bh;
+
+	BUG_ON(!PageLocked(page));
+	if (!page_has_buffers(page))
+		return;
+
+	bh = head = page_buffers(page);
+	block_start = 0;
+	do {
+		block_end = block_start + bh->b_size;
+
+		if (buffer_new(bh)) {
+			if (block_end > from && block_start < to) {
+				if (!PageUptodate(page)) {
+					unsigned start, size;
+
+					start = max(from, block_start);
+					size = min(to, block_end) - start;
+
+					zero_user_page(page, start, size, KM_USER0);
+					set_buffer_uptodate(bh);
+				}
+
+				clear_buffer_new(bh);
+				mark_buffer_dirty(bh);
+			}
+		}
+
+		block_start = block_end;
+		bh = bh->b_this_page;
+	} while (bh != head);
+}
+EXPORT_SYMBOL(page_zero_new_buffers);
+
 static int __block_prepare_write(struct inode *inode, struct page *page,
 		unsigned from, unsigned to, get_block_t *get_block)
 {
@@ -1854,38 +1896,8 @@ static int __block_prepare_write(struct inode *inode, struct page *page,
 		if (!buffer_uptodate(*wait_bh))
 			err = -EIO;
 	}
-	if (!err) {
-		bh = head;
-		do {
-			if (buffer_new(bh))
-				clear_buffer_new(bh);
-		} while ((bh = bh->b_this_page) != head);
-		return 0;
-	}
-	/* Error case: */
-	/*
-	 * Zero out any newly allocated blocks to avoid exposing stale
-	 * data.  If BH_New is set, we know that the block was newly
-	 * allocated in the above loop.
-	 */
-	bh = head;
-	block_start = 0;
-	do {
-		block_end = block_start+blocksize;
-		if (block_end <= from)
-			goto next_bh;
-		if (block_start >= to)
-			break;
-		if (buffer_new(bh)) {
-			clear_buffer_new(bh);
-			zero_user_page(page, block_start, bh->b_size, KM_USER0);
-			set_buffer_uptodate(bh);
-			mark_buffer_dirty(bh);
-		}
-next_bh:
-		block_start = block_end;
-		bh = bh->b_this_page;
-	} while (bh != head);
+	if (unlikely(err))
+		page_zero_new_buffers(page, from, to);
 	return err;
 }
 
@@ -1910,6 +1922,7 @@ static int __block_commit_write(struct inode *inode, struct page *page,
 			set_buffer_uptodate(bh);
 			mark_buffer_dirty(bh);
 		}
+		clear_buffer_new(bh);
 	}
 
 	/*
@@ -1924,6 +1937,130 @@ static int __block_commit_write(struct inode *inode, struct page *page,
 }
 
 /*
+ * block_write_begin takes care of the basic task of block allocation and
+ * bringing partial write blocks uptodate first.
+ *
+ * If *pagep is not NULL, then block_write_begin uses the locked page
+ * at *pagep rather than allocating its own. In this case, the page will
+ * not be unlocked or deallocated on failure.
+ */
+int block_write_begin(struct file *file, struct address_space *mapping,
+			loff_t pos, unsigned len, unsigned flags,
+			struct page **pagep, void **fsdata,
+			get_block_t *get_block)
+{
+	struct inode *inode = mapping->host;
+	int status = 0;
+	struct page *page;
+	pgoff_t index;
+	unsigned start, end;
+	int ownpage = 0;
+
+	index = pos >> PAGE_CACHE_SHIFT;
+	start = pos & (PAGE_CACHE_SIZE - 1);
+	end = start + len;
+
+	page = *pagep;
+	if (page == NULL) {
+		ownpage = 1;
+		page = __grab_cache_page(mapping, index);
+		if (!page) {
+			status = -ENOMEM;
+			goto out;
+		}
+		*pagep = page;
+	} else
+		BUG_ON(!PageLocked(page));
+
+	status = __block_prepare_write(inode, page, start, end, get_block);
+	if (unlikely(status)) {
+		ClearPageUptodate(page);
+
+		if (ownpage) {
+			unlock_page(page);
+			page_cache_release(page);
+			*pagep = NULL;
+
+			/*
+			 * prepare_write() may have instantiated a few blocks
+			 * outside i_size.  Trim these off again. Don't need
+			 * i_size_read because we hold i_mutex.
+			 */
+			if (pos + len > inode->i_size)
+				vmtruncate(inode, inode->i_size);
+		}
+		goto out;
+	}
+
+out:
+	return status;
+}
+EXPORT_SYMBOL(block_write_begin);
+
+int block_write_end(struct file *file, struct address_space *mapping,
+			loff_t pos, unsigned len, unsigned copied,
+			struct page *page, void *fsdata)
+{
+	struct inode *inode = mapping->host;
+	unsigned start;
+
+	start = pos & (PAGE_CACHE_SIZE - 1);
+
+	if (unlikely(copied < len)) {
+		/*
+		 * The buffers that were written will now be uptodate, so we
+		 * don't have to worry about a readpage reading them and
+		 * overwriting a partial write. However if we have encountered
+		 * a short write and only partially written into a buffer, it
+		 * will not be marked uptodate, so a readpage might come in and
+		 * destroy our partial write.
+		 *
+		 * Do the simplest thing, and just treat any short write to a
+		 * non uptodate page as a zero-length write, and force the
+		 * caller to redo the whole thing.
+		 */
+		if (!PageUptodate(page))
+			copied = 0;
+
+		page_zero_new_buffers(page, start+copied, start+len);
+	}
+	flush_dcache_page(page);
+
+	/* This could be a short (even 0-length) commit */
+	__block_commit_write(inode, page, start, start+copied);
+
+	return copied;
+}
+EXPORT_SYMBOL(block_write_end);
+
+int generic_write_end(struct file *file, struct address_space *mapping,
+			loff_t pos, unsigned len, unsigned copied,
+			struct page *page, void *fsdata)
+{
+	struct inode *inode = mapping->host;
+
+	copied = block_write_end(file, mapping, pos, len, copied, page, fsdata);
+
+	/*
+	 * No need to use i_size_read() here, the i_size
+	 * cannot change under us because we hold i_mutex.
+	 *
+	 * But it's important to update i_size while still holding page lock:
+	 * page writeout could otherwise come in and zero beyond i_size.
+	 */
+	if (pos+copied > inode->i_size) {
+		i_size_write(inode, pos+copied);
+		mark_inode_dirty(inode);
+	}
+
+	unlock_page(page);
+	page_cache_release(page);
+
+	return copied;
+}
+EXPORT_SYMBOL(generic_write_end);
+
+/*
  * Generic "read page" function for block devices that have the normal
  * get_block functionality. This is most of the block device filesystems.
  * Reads the page asynchronously --- the unlock_buffer() and
diff --git a/fs/libfs.c b/fs/libfs.c
index 5294de1..f2b32d3 100644
--- a/fs/libfs.c
+++ b/fs/libfs.c
@@ -351,6 +351,26 @@ int simple_prepare_write(struct file *file, struct page *page,
 	return 0;
 }
 
+int simple_write_begin(struct file *file, struct address_space *mapping,
+			loff_t pos, unsigned len, unsigned flags,
+			struct page **pagep, void **fsdata)
+{
+	struct page *page;
+	pgoff_t index;
+	unsigned from;
+
+	index = pos >> PAGE_CACHE_SHIFT;
+	from = pos & (PAGE_CACHE_SIZE - 1);
+
+	page = __grab_cache_page(mapping, index);
+	if (!page)
+		return -ENOMEM;
+
+	*pagep = page;
+
+	return simple_prepare_write(file, page, from, from+len);
+}
+
 int simple_commit_write(struct file *file, struct page *page,
 			unsigned from, unsigned to)
 {
@@ -369,6 +389,28 @@ int simple_commit_write(struct file *file, struct page *page,
 	return 0;
 }
 
+int simple_write_end(struct file *file, struct address_space *mapping,
+			loff_t pos, unsigned len, unsigned copied,
+			struct page *page, void *fsdata)
+{
+	unsigned from = pos & (PAGE_CACHE_SIZE - 1);
+
+	/* zero the stale part of the page if we did a short copy */
+	if (copied < len) {
+		void *kaddr = kmap_atomic(page, KM_USER0);
+		memset(kaddr + from + copied, 0, len - copied);
+		flush_dcache_page(page);
+		kunmap_atomic(kaddr, KM_USER0);
+	}
+
+	simple_commit_write(file, page, from, from+copied);
+
+	unlock_page(page);
+	page_cache_release(page);
+
+	return copied;
+}
+
 /*
  * the inodes created here are not hashed. If you use iunique to generate
  * unique inode values later for this filesystem, then you must take care
@@ -642,6 +684,8 @@ EXPORT_SYMBOL(dcache_dir_open);
 EXPORT_SYMBOL(dcache_readdir);
 EXPORT_SYMBOL(generic_read_dir);
 EXPORT_SYMBOL(get_sb_pseudo);
+EXPORT_SYMBOL(simple_write_begin);
+EXPORT_SYMBOL(simple_write_end);
 EXPORT_SYMBOL(simple_commit_write);
 EXPORT_SYMBOL(simple_dir_inode_operations);
 EXPORT_SYMBOL(simple_dir_operations);
diff --git a/fs/namei.c b/fs/namei.c
index a83160a..b40b808 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -2729,53 +2729,29 @@ int __page_symlink(struct inode *inode, const char *symname, int len,
 {
 	struct address_space *mapping = inode->i_mapping;
 	struct page *page;
+	void *fsdata;
 	int err;
 	char *kaddr;
 
 retry:
-	err = -ENOMEM;
-	page = find_or_create_page(mapping, 0, gfp_mask);
-	if (!page)
-		goto fail;
-	err = mapping->a_ops->prepare_write(NULL, page, 0, len-1);
-	if (err == AOP_TRUNCATED_PAGE) {
-		page_cache_release(page);
-		goto retry;
-	}
+	err = pagecache_write_begin(NULL, mapping, 0, len-1,
+				AOP_FLAG_UNINTERRUPTIBLE, &page, &fsdata);
 	if (err)
-		goto fail_map;
+		goto fail;
+
 	kaddr = kmap_atomic(page, KM_USER0);
 	memcpy(kaddr, symname, len-1);
 	kunmap_atomic(kaddr, KM_USER0);
-	err = mapping->a_ops->commit_write(NULL, page, 0, len-1);
-	if (err == AOP_TRUNCATED_PAGE) {
-		page_cache_release(page);
-		goto retry;
-	}
-	if (err)
-		goto fail_map;
-	/*
-	 * Notice that we are _not_ going to block here - end of page is
-	 * unmapped, so this will only try to map the rest of page, see
-	 * that it is unmapped (typically even will not look into inode -
-	 * ->i_size will be enough for everything) and zero it out.
-	 * OTOH it's obviously correct and should make the page up-to-date.
-	 */
-	if (!PageUptodate(page)) {
-		err = mapping->a_ops->readpage(NULL, page);
-		if (err != AOP_TRUNCATED_PAGE)
-			wait_on_page_locked(page);
-	} else {
-		unlock_page(page);
-	}
-	page_cache_release(page);
+
+	err = pagecache_write_end(NULL, mapping, 0, len-1, len-1,
+							page, fsdata);
 	if (err < 0)
 		goto fail;
+	if (err < len-1)
+		goto retry;
+
 	mark_inode_dirty(inode);
 	return 0;
-fail_map:
-	unlock_page(page);
-	page_cache_release(page);
 fail:
 	return err;
 }
diff --git a/fs/splice.c b/fs/splice.c
index 2df6be4..a7568bc 100644
--- a/fs/splice.c
+++ b/fs/splice.c
@@ -563,7 +563,7 @@ static int pipe_to_file(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
 	struct address_space *mapping = file->f_mapping;
 	unsigned int offset, this_len;
 	struct page *page;
-	pgoff_t index;
+	void *fsdata;
 	int ret;
 
 	/*
@@ -573,49 +573,16 @@ static int pipe_to_file(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
 	if (unlikely(ret))
 		return ret;
 
-	index = sd->pos >> PAGE_CACHE_SHIFT;
 	offset = sd->pos & ~PAGE_CACHE_MASK;
 
 	this_len = sd->len;
 	if (this_len + offset > PAGE_CACHE_SIZE)
 		this_len = PAGE_CACHE_SIZE - offset;
 
-find_page:
-	page = find_lock_page(mapping, index);
-	if (!page) {
-		ret = -ENOMEM;
-		page = page_cache_alloc_cold(mapping);
-		if (unlikely(!page))
-			goto out_ret;
-
-		/*
-		 * This will also lock the page
-		 */
-		ret = add_to_page_cache_lru(page, mapping, index,
-					    GFP_KERNEL);
-		if (unlikely(ret))
-			goto out_release;
-	}
-
-	ret = mapping->a_ops->prepare_write(file, page, offset, offset+this_len);
-	if (unlikely(ret)) {
-		loff_t isize = i_size_read(mapping->host);
-
-		if (ret != AOP_TRUNCATED_PAGE)
-			unlock_page(page);
-		page_cache_release(page);
-		if (ret == AOP_TRUNCATED_PAGE)
-			goto find_page;
-
-		/*
-		 * prepare_write() may have instantiated a few blocks
-		 * outside i_size.  Trim these off again.
-		 */
-		if (sd->pos + this_len > isize)
-			vmtruncate(mapping->host, isize);
-
-		goto out_ret;
-	}
+	ret = pagecache_write_begin(file, mapping, sd->pos, this_len,
+				AOP_FLAG_UNINTERRUPTIBLE, &page, &fsdata);
+	if (unlikely(ret))
+		goto out;
 
 	if (buf->page != page) {
 		/*
@@ -629,31 +596,9 @@ find_page:
 		kunmap_atomic(dst, KM_USER1);
 		buf->ops->unmap(pipe, buf, src);
 	}
-
-	ret = mapping->a_ops->commit_write(file, page, offset, offset+this_len);
-	if (ret) {
-		if (ret == AOP_TRUNCATED_PAGE) {
-			page_cache_release(page);
-			goto find_page;
-		}
-		if (ret < 0)
-			goto out;
-		/*
-		 * Partial write has happened, so 'ret' already initialized by
-		 * number of bytes written, Where is nothing we have to do here.
-		 */
-	} else
-		ret = this_len;
-	/*
-	 * Return the number of bytes written and mark page as
-	 * accessed, we are now done!
-	 */
-	mark_page_accessed(page);
+	ret = pagecache_write_end(file, mapping, sd->pos, this_len, this_len,
+				page, fsdata);
 out:
-	unlock_page(page);
-out_release:
-	page_cache_release(page);
-out_ret:
 	return ret;
 }