[PATCH 1/4] f2fs: fix extent corrupotion during directIO in LFS mode

[PATCH 1/4] f2fs: fix extent corrupotion during directIO in LFS mode

[Chao Yu]

In LFS mode, por_fsstress testcase reports a bug as below:

[ASSERT] (fsck_chk_inode_blk: 931)  --> ino: 0x12fe has wrong ext: [pgofs:142, blk:215424, len:16]

Since commit f847c699cff3 ("f2fs: allow out-place-update for direct
IO in LFS mode"), we start to allow OPU mode for direct IO, however,
we missed to update extent cache in __allocate_data_block(), finally,
it cause extent field being inconsistent with physical block address,
fix it.

Fixes: f847c699cff3 ("f2fs: allow out-place-update for direct IO in LFS mode")
Signed-off-by: Chao Yu <yuchao0@huawei.com>
---
 fs/f2fs/data.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index bf648c8c50ad..727df32382c6 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -1018,7 +1018,7 @@ static int __allocate_data_block(struct dnode_of_data *dn, int seg_type)
 	if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO)
 		invalidate_mapping_pages(META_MAPPING(sbi),
 					old_blkaddr, old_blkaddr);
-	f2fs_set_data_blkaddr(dn);
+	f2fs_update_data_blkaddr(dn, dn->data_blkaddr);
 
 	/*
 	 * i_size will be updated by direct_IO. Otherwise, we'll get stale
-- 
2.18.0.rc1

[PATCH 2/4] f2fs: fix to handle error path correctly in f2fs_map_blocks

[Chao Yu]

In f2fs_map_blocks(), we should bail out once __allocate_data_block()
failed.

Fixes: f847c699cff3 ("f2fs: allow out-place-update for direct IO in LFS mode")
Signed-off-by: Chao Yu <yuchao0@huawei.com>
---
 fs/f2fs/data.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index 727df32382c6..777888ba171a 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -1195,10 +1195,10 @@ int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map,
 		if (test_opt(sbi, LFS) && flag == F2FS_GET_BLOCK_DIO &&
 							map->m_may_create) {
 			err = __allocate_data_block(&dn, map->m_seg_type);
-			if (!err) {
-				blkaddr = dn.data_blkaddr;
-				set_inode_flag(inode, FI_APPEND_WRITE);
-			}
+			if (err)
+				goto sync_out;
+			blkaddr = dn.data_blkaddr;
+			set_inode_flag(inode, FI_APPEND_WRITE);
 		}
 	} else {
 		if (create) {
-- 
2.18.0.rc1

[PATCH 3/4] f2fs: fix to fallback to buffered IO in IO aligned mode

[Chao Yu]

In LFS mode, we allow OPU for direct IO, however, we didn't consider
IO alignment feature, so direct IO can trigger unaligned IO, let's
just fallback to buffered IO to keep correct IO alignment semantics
in all places.

Fixes: f847c699cff3 ("f2fs: allow out-place-update for direct IO in LFS mode")
Signed-off-by: Chao Yu <yuchao0@huawei.com>
---
 fs/f2fs/f2fs.h | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index 4b21ac42d44e..0522641c0c17 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -3733,9 +3733,12 @@ static inline bool f2fs_force_buffered_io(struct inode *inode,
 	 */
 	if (f2fs_sb_has_blkzoned(sbi))
 		return true;
-	if (test_opt(sbi, LFS) && (rw == WRITE) &&
-				block_unaligned_IO(inode, iocb, iter))
-		return true;
+	if (test_opt(sbi, LFS) && (rw == WRITE)) {
+		if (block_unaligned_IO(inode, iocb, iter))
+			return true;
+		if (F2FS_IO_ALIGNED(sbi))
+			return true;
+	}
 	if (is_sbi_flag_set(F2FS_I_SB(inode), SBI_CP_DISABLED) &&
 					!IS_SWAPFILE(inode))
 		return true;
-- 
2.18.0.rc1

[PATCH 4/4] f2fs: fix to add missing F2FS_IO_ALIGNED() condition

[Chao Yu]

In f2fs_allocate_data_block(), we will reset fio.retry for IO
alignment feature instead of IO serialization feature.

In addition, spread F2FS_IO_ALIGNED() to check IO alignment
feature status explicitly.

Signed-off-by: Chao Yu <yuchao0@huawei.com>
---
 fs/f2fs/data.c    | 6 +++++-
 fs/f2fs/segment.c | 4 +++-
 2 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index 777888ba171a..4764fdcb34d0 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -294,6 +294,9 @@ static inline void __submit_bio(struct f2fs_sb_info *sbi,
 		if (test_opt(sbi, LFS) && current->plug)
 			blk_finish_plug(current->plug);
 
+		if (F2FS_IO_ALIGNED(sbi))
+			goto submit_io;
+
 		start = bio->bi_iter.bi_size >> F2FS_BLKSIZE_BITS;
 		start %= F2FS_IO_SIZE(sbi);
 
@@ -607,7 +610,8 @@ void f2fs_submit_page_write(struct f2fs_io_info *fio)
 		__submit_merged_bio(io);
 alloc_new:
 	if (io->bio == NULL) {
-		if ((fio->type == DATA || fio->type == NODE) &&
+		if (F2FS_IO_ALIGNED(sbi) &&
+				(fio->type == DATA || fio->type == NODE) &&
 				fio->new_blkaddr & F2FS_IO_SIZE_MASK(sbi)) {
 			dec_page_count(sbi, WB_DATA_TYPE(bio_page));
 			fio->retry = true;
diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
index 18584d4c078a..4887e2ff0ff8 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -3129,12 +3129,14 @@ void f2fs_allocate_data_block(struct f2fs_sb_info *sbi, struct page *page,
 		f2fs_inode_chksum_set(sbi, page);
 	}
 
+	if (F2FS_IO_ALIGNED(sbi))
+		fio->retry = false;
+
 	if (add_list) {
 		struct f2fs_bio_info *io;
 
 		INIT_LIST_HEAD(&fio->list);
 		fio->in_list = true;
-		fio->retry = false;
 		io = sbi->write_io[fio->type] + fio->temp;
 		spin_lock(&io->io_lock);
 		list_add_tail(&fio->list, &io->io_list);
-- 
2.18.0.rc1