[PATCH 16/18] orangefs: implement write through the page cache

[Martin Brandenburg]

From: Martin Brandenburg <martin@xxxxxxxxxxxx>

With this and the previous commit, OrangeFS is capable of writing
through the page cache.

Signed-off-by: Martin Brandenburg <martin@xxxxxxxxxxxx>
---
 fs/orangefs/file.c  | 132 +++++++++++++++++++++++-----------------------------
 fs/orangefs/inode.c |   2 +
 fs/orangefs/super.c |   8 ++++
 3 files changed, 67 insertions(+), 75 deletions(-)

diff --git a/fs/orangefs/file.c b/fs/orangefs/file.c
index d80e1e6c1d95..200c72c21b0b 100644
--- a/fs/orangefs/file.c
+++ b/fs/orangefs/file.c
@@ -397,69 +397,11 @@ static ssize_t orangefs_file_read_iter(struct kiocb *iocb,
 	return generic_file_read_iter(iocb, iter);
 }
 
-static ssize_t orangefs_file_write_iter(struct kiocb *iocb, struct iov_iter *iter)
+static ssize_t orangefs_file_write_iter(struct kiocb *iocb,
+    struct iov_iter *iter)
 {
-	struct file *file = iocb->ki_filp;
-	loff_t pos;
-	ssize_t rc;
-
-	BUG_ON(iocb->private);
-
-	gossip_debug(GOSSIP_FILE_DEBUG, "orangefs_file_write_iter\n");
-
-	inode_lock(file->f_mapping->host);
-
-	/* Make sure generic_write_checks sees an up to date inode size. */
-	if (file->f_flags & O_APPEND) {
-		rc = orangefs_inode_getattr(file->f_mapping->host, 0, 1,
-		    STATX_SIZE);
-		if (rc == -ESTALE)
-			rc = -EIO;
-		if (rc) {
-			gossip_err("%s: orangefs_inode_getattr failed, "
-			    "rc:%zd:.\n", __func__, rc);
-			goto out;
-		}
-	}
-
-	if (file->f_pos > i_size_read(file->f_mapping->host))
-		orangefs_i_size_write(file->f_mapping->host, file->f_pos);
-
-	rc = generic_write_checks(iocb, iter);
-
-	if (rc <= 0) {
-		gossip_err("%s: generic_write_checks failed, rc:%zd:.\n",
-			   __func__, rc);
-		goto out;
-	}
-
-	/*
-	 * if we are appending, generic_write_checks would have updated
-	 * pos to the end of the file, so we will wait till now to set
-	 * pos...
-	 */
-	pos = *(&iocb->ki_pos);
-
-	rc = do_readv_writev(ORANGEFS_IO_WRITE,
-			     file,
-			     &pos,
-			     iter);
-	if (rc < 0) {
-		gossip_err("%s: do_readv_writev failed, rc:%zd:.\n",
-			   __func__, rc);
-		goto out;
-	}
-
-	iocb->ki_pos = pos;
 	orangefs_stats.writes++;
-
-	if (pos > i_size_read(file->f_mapping->host))
-		orangefs_i_size_write(file->f_mapping->host, pos);
-
-out:
-
-	inode_unlock(file->f_mapping->host);
-	return rc;
+	return generic_file_write_iter(iocb, iter);
 }
 
 /*
@@ -554,9 +496,6 @@ static int orangefs_file_mmap(struct file *file, struct vm_area_struct *vma)
 			(char *)file->f_path.dentry->d_name.name :
 			(char *)"Unknown"));
 
-	if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_MAYWRITE))
-		return -EINVAL;
-
 	/* set the sequential readahead hint */
 	vma->vm_flags |= VM_SEQ_READ;
 	vma->vm_flags &= ~VM_RAND_READ;
@@ -579,9 +518,8 @@ static int orangefs_file_release(struct inode *inode, struct file *file)
 		     file);
 
 	/*
-	 * remove all associated inode pages from the page cache and
-	 * readahead cache (if any); this forces an expensive refresh of
-	 * data for the next caller of mmap (or 'get_block' accesses)
+	 * remove all associated inode pages from the readahead cache
+	 * (if any)
 	 */
 	if (file_inode(file) &&
 	    file_inode(file)->i_mapping &&
@@ -594,8 +532,6 @@ static int orangefs_file_release(struct inode *inode, struct file *file)
 			gossip_debug(GOSSIP_INODE_DEBUG,
 			    "flush_racache finished\n");
 		}
-		truncate_inode_pages(file_inode(file)->i_mapping,
-				     0);
 	}
 	return 0;
 }
@@ -707,6 +643,41 @@ const struct file_operations orangefs_file_operations = {
 	.fsync		= orangefs_fsync,
 };
 
+
+static int orangefs_writepage(struct page *page,
+    struct writeback_control *wbc)
+{
+	struct inode *inode = page->mapping->host;
+	struct iov_iter iter;
+	struct bio_vec bv;
+	loff_t off;
+	size_t len;
+	ssize_t r;
+
+	off = page_offset(page);
+	len = i_size_read(inode);
+	if (off + PAGE_SIZE > len)
+		len = len - off;
+	else
+		len = PAGE_SIZE;
+
+	bv.bv_page = page;
+	bv.bv_len = len;
+	bv.bv_offset = 0;
+	iov_iter_bvec(&iter, ITER_BVEC | WRITE, &bv, 1, len);
+
+	set_page_writeback(page);
+
+	r = wait_for_direct_io(ORANGEFS_IO_WRITE, inode, &off, &iter,
+	    len, 0);
+	if (r < 0)
+		mapping_set_error(inode->i_mapping, r);
+
+	end_page_writeback(page);
+	unlock_page(page);
+	return 0;
+}
+
 static int orangefs_readpage(struct file *file, struct page *page)
 {
 	int ret;
@@ -752,6 +723,17 @@ static int orangefs_readpage(struct file *file, struct page *page)
 	return ret;
 }
 
+static int orangefs_write_end(struct file *file,
+    struct address_space *mapping, loff_t pos, unsigned len,
+    unsigned copied, struct page *page, void *fsdata)
+{
+	int r;
+	r = simple_write_end(file, mapping, pos, len, copied, page,
+	    fsdata);
+	mark_inode_dirty_sync(file_inode(file));
+	return r;
+}
+
 static void orangefs_invalidatepage(struct page *page,
 				 unsigned int offset,
 				 unsigned int length)
@@ -781,17 +763,17 @@ static ssize_t orangefs_direct_IO(struct kiocb *iocb,
 {
 	struct file *file = iocb->ki_filp;
 	loff_t pos = *(&iocb->ki_pos);
-	/*
-	 * This cannot happen until write_iter becomes
-	 * generic_file_write_iter.
-	 */
-	BUG_ON(iov_iter_rw(iter) != READ);
-	return do_readv_writev(ORANGEFS_IO_READ, file, &pos, iter);
+	return do_readv_writev(iov_iter_rw(iter) == WRITE ?
+	    ORANGEFS_IO_WRITE : ORANGEFS_IO_READ, file, &pos, iter);
 }
 
 /** ORANGEFS2 implementation of address space operations */
 const struct address_space_operations orangefs_address_operations = {
+	.writepage = orangefs_writepage,
 	.readpage = orangefs_readpage,
+	.set_page_dirty = __set_page_dirty_nobuffers,
+	.write_begin = simple_write_begin,
+	.write_end = orangefs_write_end,
 	.invalidatepage = orangefs_invalidatepage,
 	.releasepage = orangefs_releasepage,
 	.direct_IO = orangefs_direct_IO,
diff --git a/fs/orangefs/inode.c b/fs/orangefs/inode.c
index 2c4f57c5d830..231242a4856e 100644
--- a/fs/orangefs/inode.c
+++ b/fs/orangefs/inode.c
@@ -65,6 +65,8 @@ int orangefs_setattr(struct dentry *dentry, struct iattr *iattr)
 	r = setattr_prepare(dentry, iattr);
 	if (r)
 		return r;
+	if (d_is_reg(dentry))
+		filemap_write_and_wait(d_inode(dentry)->i_mapping);
 	if (iattr->ia_valid & ATTR_SIZE)
 		if (i_size_read(d_inode(dentry)) != iattr->ia_size)
 			iattr->ia_valid |= ATTR_CTIME | ATTR_MTIME;
diff --git a/fs/orangefs/super.c b/fs/orangefs/super.c
index 5c17709862ab..5c1a343ba026 100644
--- a/fs/orangefs/super.c
+++ b/fs/orangefs/super.c
@@ -315,11 +315,19 @@ static int orangefs_write_inode(struct inode *inode,
 	return r;
 }
 
+static void orangefs_evict_inode(struct inode *inode)
+{
+	truncate_inode_pages_final(&inode->i_data);
+	clear_inode(inode);
+	filemap_fdatawrite(&inode->i_data);
+}
+
 static const struct super_operations orangefs_s_ops = {
 	.alloc_inode = orangefs_alloc_inode,
 	.destroy_inode = orangefs_destroy_inode,
 	.write_inode = orangefs_write_inode,
 	.drop_inode = generic_drop_inode,
+	.evict_inode = orangefs_evict_inode,
 	.statfs = orangefs_statfs,
 	.remount_fs = orangefs_remount_fs,
 	.show_options = orangefs_show_options,
-- 
2.15.1