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>
torvalds · Oct 16, 2007 · afddba4 · afddba4
1 parent 637aff4
commit afddba4
Show file tree

Hide file tree

Showing 11 changed files with 575 additions and 206 deletions.
diff --git a/Documentation/filesystems/Locking b/Documentation/filesystems/Locking
@@ -178,15 +178,18 @@ prototypes:
 locking rules:
 	All except set_page_dirty may block
 
-			BKL	PageLocked(page)
+			BKL	PageLocked(page)	i_sem
 writepage:		no	yes, unlocks (see below)
 readpage:		no	yes, unlocks
 sync_page:		no	maybe
 writepages:		no
 set_page_dirty		no	no
 readpages:		no
-prepare_write:		no	yes
-commit_write:		no	yes
+prepare_write:		no	yes			yes
+commit_write:		no	yes			yes
+write_begin:		no	locks the page		yes
+write_end:		no	yes, unlocks		yes
+perform_write:		no	n/a			yes
 bmap:			yes
 invalidatepage:		no	yes
 releasepage:		no	yes

diff --git a/Documentation/filesystems/vfs.txt b/Documentation/filesystems/vfs.txt
@@ -537,6 +537,12 @@ struct address_space_operations {
 			struct list_head *pages, unsigned nr_pages);
 	int (*prepare_write)(struct file *, struct page *, unsigned, unsigned);
 	int (*commit_write)(struct file *, struct page *, unsigned, unsigned);
+	int (*write_begin)(struct file *, struct address_space *mapping,
+				loff_t pos, unsigned len, unsigned flags,
+				struct page **pagep, void **fsdata);
+	int (*write_end)(struct file *, struct address_space *mapping,
+				loff_t pos, unsigned len, unsigned copied,
+				struct page *page, void *fsdata);
 	sector_t (*bmap)(struct address_space *, sector_t);
 	int (*invalidatepage) (struct page *, unsigned long);
 	int (*releasepage) (struct page *, int);
@@ -633,6 +639,45 @@ struct address_space_operations {
         operations.  It should avoid returning an error if possible -
         errors should have been handled by prepare_write.
 
+  write_begin: This is intended as a replacement for prepare_write. The
+	key differences being that:
+		- it returns a locked page (in *pagep) rather than being
+		  given a pre locked page;
+		- it must be able to cope with short writes (where the
+		  length passed to write_begin is greater than the number
+		  of bytes copied into the page).
+
+	Called by the generic buffered write code to ask the filesystem to
+	prepare to write len bytes at the given offset in the file. The
+	address_space should check that the write will be able to complete,
+	by allocating space if necessary and doing any other internal
+	housekeeping.  If the write will update parts of any basic-blocks on
+	storage, then those blocks should be pre-read (if they haven't been
+	read already) so that the updated blocks can be written out properly.
+
+        The filesystem must return the locked pagecache page for the specified
+	offset, in *pagep, for the caller to write into.
+
+	flags is a field for AOP_FLAG_xxx flags, described in
+	include/linux/fs.h.
+
+        A void * may be returned in fsdata, which then gets passed into
+        write_end.
+
+        Returns 0 on success; < 0 on failure (which is the error code), in
+	which case write_end is not called.
+
+  write_end: After a successful write_begin, and data copy, write_end must
+        be called. len is the original len passed to write_begin, and copied
+        is the amount that was able to be copied (copied == len is always true
+	if write_begin was called with the AOP_FLAG_UNINTERRUPTIBLE flag).
+
+        The filesystem must take care of unlocking the page and releasing it
+        refcount, and updating i_size.
+
+        Returns < 0 on failure, otherwise the number of bytes (<= 'copied')
+        that were able to be copied into pagecache.
+
   bmap: called by the VFS to map a logical block offset within object to
   	physical block number. This method is used by the FIBMAP
   	ioctl and for working with swap-files.  To be able to swap to

diff --git a/drivers/block/loop.c b/drivers/block/loop.c
@@ -204,14 +204,13 @@ lo_do_transfer(struct loop_device *lo, int cmd,
  * do_lo_send_aops - helper for writing data to a loop device
  *
  * This is the fast version for backing filesystems which implement the address
- * space operations prepare_write and commit_write.
+ * space operations write_begin and write_end.
  */
 static int do_lo_send_aops(struct loop_device *lo, struct bio_vec *bvec,
-		int bsize, loff_t pos, struct page *page)
+		int bsize, loff_t pos, struct page *unused)
 {
 	struct file *file = lo->lo_backing_file; /* kudos to NFsckingS */
 	struct address_space *mapping = file->f_mapping;
-	const struct address_space_operations *aops = mapping->a_ops;
 	pgoff_t index;
 	unsigned offset, bv_offs;
 	int len, ret;
@@ -223,63 +222,47 @@ static int do_lo_send_aops(struct loop_device *lo, struct bio_vec *bvec,
 	len = bvec->bv_len;
 	while (len > 0) {
 		sector_t IV;
-		unsigned size;
+		unsigned size, copied;
 		int transfer_result;
+		struct page *page;
+		void *fsdata;
 
 		IV = ((sector_t)index << (PAGE_CACHE_SHIFT - 9))+(offset >> 9);
 		size = PAGE_CACHE_SIZE - offset;
 		if (size > len)
 			size = len;
-		page = grab_cache_page(mapping, index);
-		if (unlikely(!page))
+
+		ret = pagecache_write_begin(file, mapping, pos, size, 0,
+							&page, &fsdata);
+		if (ret)
 			goto fail;
-		ret = aops->prepare_write(file, page, offset,
-					  offset + size);
-		if (unlikely(ret)) {
-			if (ret == AOP_TRUNCATED_PAGE) {
-				page_cache_release(page);
-				continue;
-			}
-			goto unlock;
-		}
+
 		transfer_result = lo_do_transfer(lo, WRITE, page, offset,
 				bvec->bv_page, bv_offs, size, IV);
-		if (unlikely(transfer_result)) {
-			/*
-			 * The transfer failed, but we still write the data to
-			 * keep prepare/commit calls balanced.
-			 */
-			printk(KERN_ERR "loop: transfer error block %llu\n",
-			       (unsigned long long)index);
-			zero_user_page(page, offset, size, KM_USER0);
-		}
-		flush_dcache_page(page);
-		ret = aops->commit_write(file, page, offset,
-					 offset + size);
-		if (unlikely(ret)) {
-			if (ret == AOP_TRUNCATED_PAGE) {
-				page_cache_release(page);
-				continue;
-			}
-			goto unlock;
-		}
+		copied = size;
 		if (unlikely(transfer_result))
-			goto unlock;
-		bv_offs += size;
-		len -= size;
+			copied = 0;
+
+		ret = pagecache_write_end(file, mapping, pos, size, copied,
+							page, fsdata);
+		if (ret < 0)
+			goto fail;
+		if (ret < copied)
+			copied = ret;
+
+		if (unlikely(transfer_result))
+			goto fail;
+
+		bv_offs += copied;
+		len -= copied;
 		offset = 0;
 		index++;
-		pos += size;
-		unlock_page(page);
-		page_cache_release(page);
+		pos += copied;
 	}
 	ret = 0;
 out:
 	mutex_unlock(&mapping->host->i_mutex);
 	return ret;
-unlock:
-	unlock_page(page);
-	page_cache_release(page);
 fail:
 	ret = -1;
 	goto out;
@@ -313,7 +296,7 @@ static int __do_lo_send_write(struct file *file,
  * do_lo_send_direct_write - helper for writing data to a loop device
  *
  * This is the fast, non-transforming version for backing filesystems which do
- * not implement the address space operations prepare_write and commit_write.
+ * not implement the address space operations write_begin and write_end.
  * It uses the write file operation which should be present on all writeable
  * filesystems.
  */
@@ -332,7 +315,7 @@ static int do_lo_send_direct_write(struct loop_device *lo,
  * do_lo_send_write - helper for writing data to a loop device
  *
  * This is the slow, transforming version for filesystems which do not
- * implement the address space operations prepare_write and commit_write.  It
+ * implement the address space operations write_begin and write_end.  It
  * uses the write file operation which should be present on all writeable
  * filesystems.
  *
@@ -780,7 +763,7 @@ static int loop_set_fd(struct loop_device *lo, struct file *lo_file,
 		 */
 		if (!file->f_op->splice_read)
 			goto out_putf;
-		if (aops->prepare_write && aops->commit_write)
+		if (aops->prepare_write || aops->write_begin)
 			lo_flags |= LO_FLAGS_USE_AOPS;
 		if (!(lo_flags & LO_FLAGS_USE_AOPS) && !file->f_op->write)
 			lo_flags |= LO_FLAGS_READ_ONLY;