diff --git a/fs/hmdfs/file_cloud.c b/fs/hmdfs/file_cloud.c
index b939b109913c9b29ddd99007880658c4a796413a..427f4d1a7b3a5ba73744e1f2285f05bdaa7aa129 100644
--- a/fs/hmdfs/file_cloud.c
+++ b/fs/hmdfs/file_cloud.c
@@ -31,14 +31,6 @@ static const struct vm_operations_struct hmdfs_cloud_vm_ops = {
 	.page_mkwrite = NULL,
 };
 
-struct cloud_readpages_work {
-	struct file *filp;
-	loff_t pos;
-	int cnt;
-	struct work_struct work;
-	struct page *pages[0];
-};
-
 int hmdfs_file_open_cloud(struct inode *inode, struct file *file)
 {
 	const char *dir_path;
@@ -139,79 +131,56 @@ int hmdfs_file_mmap_cloud(struct file *file, struct vm_area_struct *vma)
 	return ret;
 }
 
-static int cloud_readpages_work_func(struct work_struct *work)
+static int hmdfs_do_readpages_cloud(struct file *filp, int cnt,
+				    struct page **vec)
 {
-	void *pages_buf;
+	struct hmdfs_file_info *gfi = filp->private_data;
+	struct file *lower_filp;
+	loff_t pos = (loff_t)(vec[0]->index) << HMDFS_PAGE_OFFSET;
+	void *pages_buf = NULL;
 	int idx, ret;
-	ssize_t read_len;
-	struct cloud_readpages_work *cr_work;
 
-	cr_work = container_of(work, struct cloud_readpages_work, work);
+	if (gfi) {
+		lower_filp = gfi->lower_file;
+	}
+	else {
+		ret = -EINVAL;
+		goto out_err;
+	}
 
-	read_len = cr_work->cnt * HMDFS_PAGE_SIZE;
-	pages_buf = vmap(cr_work->pages, cr_work->cnt, VM_MAP, PAGE_KERNEL);
-	if (!pages_buf)
+	pages_buf = vmap(vec, cnt, VM_MAP, PAGE_KERNEL);
+	if (!pages_buf) {
+		ret = -ENOMEM;
 		goto out_err;
+	}
 
-	trace_hmdfs_readpages_cloud_work_begin(cr_work->cnt, cr_work->pos);
-	ret = kernel_read(cr_work->filp, pages_buf, read_len, &cr_work->pos);
-	trace_hmdfs_readpages_cloud_work_end(cr_work->cnt, cr_work->pos, ret);
+	trace_hmdfs_do_readpages_cloud_begin(cnt, pos);
+	ret = kernel_read(lower_filp, pages_buf, cnt * HMDFS_PAGE_SIZE, &pos);
+	trace_hmdfs_do_readpages_cloud_end(cnt, pos, ret);
 
-	if (ret >= 0 && ret <= read_len)
-		memset(pages_buf + ret, 0, read_len - ret);
+	if (ret >= 0)
+		memset(pages_buf + ret, 0, cnt * HMDFS_PAGE_SIZE - ret);
 	else
 		goto out_err;
 
 	vunmap(pages_buf);
-	for (idx = 0; idx < cr_work->cnt; ++idx) {
-		SetPageUptodate(cr_work->pages[idx]);
-		unlock_page(cr_work->pages[idx]);
+	for (idx = 0; idx < cnt; ++idx) {
+		SetPageUptodate(vec[idx]);
+		unlock_page(vec[idx]);
 	}
-	goto out_free;
+	goto out;
 
 out_err:
 	if (pages_buf)
 		vunmap(pages_buf);
-	for (idx = 0; idx < cr_work->cnt; ++idx) {
-		ClearPageUptodate(cr_work->pages[idx]);
-		SetPageError(cr_work->pages[idx]);
-		unlock_page(cr_work->pages[idx]);
-	}
-out_free:
-	kfree(cr_work);
-	return ret;
-}
-
-static int prepare_cloud_readpage_work(struct file *filp, int cnt,
-				       struct page **vec)
-{
-	struct cloud_readpages_work *cr_work;
-	struct hmdfs_file_info *gfi = filp->private_data;
-	int idx = 0;
-
-	cr_work = kzalloc(sizeof(*cr_work) +
-			  sizeof(cr_work->pages[0]) * cnt,
-			  GFP_KERNEL);
-	if (!cr_work) {
-		hmdfs_warning("cannot alloc work");
-		goto unlock;
+	for (idx = 0; idx < cnt; ++idx) {
+		ClearPageUptodate(vec[idx]);
+		delete_from_page_cache(vec[idx]);
+		unlock_page(vec[idx]);
+		put_page(vec[idx]);
 	}
-
-	if (gfi)
-		cr_work->filp = gfi->lower_file;
-	else
-		goto out;
-
-	cr_work->pos = (loff_t)(vec[0]->index) << HMDFS_PAGE_OFFSET;
-	cr_work->cnt = cnt;
-	memcpy(cr_work->pages, vec, cnt * sizeof(*vec));
-	return cloud_readpages_work_func(&cr_work->work);
 out:
-	kfree(cr_work);
-unlock:
-	for (idx = 0; idx < cnt; ++idx)
-		unlock_page(vec[idx]);
-	return -ENOMEM;
+	return ret;
 }
 
 static int hmdfs_readpages_cloud(struct file *filp,
@@ -238,23 +207,24 @@ static int hmdfs_readpages_cloud(struct file *filp,
 		struct page *page = lru_to_page(pages);
 
 		list_del(&page->lru);
-		if (add_to_page_cache_lru(page, mapping, page->index, gfp))
+		if (add_to_page_cache_lru(page, mapping, page->index, gfp)) {
+			unlock_page(page);
+			put_page(page);
 			continue;
+		}
 
 		if (cnt && (cnt >= limit || page->index != next_index)) {
-			ret = prepare_cloud_readpage_work(filp, cnt, vec);
+			ret = hmdfs_do_readpages_cloud(filp, cnt, vec);
 			cnt = 0;
 			if (ret)
 				break;
 		}
 		next_index = page->index + 1;
 		vec[cnt++] = page;
-next_page:
-		put_page(page);
 	}
 
 	if (cnt)
-		ret = prepare_cloud_readpage_work(filp, cnt, vec);
+		ret = hmdfs_do_readpages_cloud(filp, cnt, vec);
 
 	kfree(vec);
 	trace_hmdfs_readpages_cloud(nr_pages, pages);
@@ -279,13 +249,13 @@ static int hmdfs_readpage(struct file *file, struct page *page)
 		goto out;
 	ret = kernel_read(lower_file, page_buf, PAGE_SIZE, &offset);
 
-	if (ret >= 0 && ret <= PAGE_SIZE)
+	if (ret >= 0 && ret <= PAGE_SIZE) {
+		ret = 0;
 		memset(page_buf + ret, 0, PAGE_SIZE - ret);
+	}
 
 	kunmap(page);
-	if (ret < 0)
-		SetPageError(page);
-	else
+	if (ret == 0)
 		SetPageUptodate(page);
 out:
 	unlock_page(page);
diff --git a/fs/hmdfs/hmdfs_trace.h b/fs/hmdfs/hmdfs_trace.h
index 02688daacc1cbf75b29b0eb44f2ee2824c82f0b1..0660d06404d2e5cc3c60e288c14aa092163ca3e4 100644
--- a/fs/hmdfs/hmdfs_trace.h
+++ b/fs/hmdfs/hmdfs_trace.h
@@ -546,7 +546,7 @@ TRACE_EVENT(hmdfs_readpages_cloud,
 		  __entry->nr_pages, __entry->err)
 );
 
-TRACE_EVENT(hmdfs_readpages_cloud_work_begin,
+TRACE_EVENT(hmdfs_do_readpages_cloud_begin,
 
 	TP_PROTO(int cnt, loff_t pos),
 
@@ -566,7 +566,7 @@ TRACE_EVENT(hmdfs_readpages_cloud_work_begin,
 		  __entry->cnt, __entry->pos)
 );
 
-TRACE_EVENT(hmdfs_readpages_cloud_work_end,
+TRACE_EVENT(hmdfs_do_readpages_cloud_end,
 
 	TP_PROTO(int cnt, loff_t pos, int ret),