f2fs: introduce F2FS_IPU_HONOR_OPU_WRITE ipu policy

Once F2FS_IPU_FORCE policy is enabled in some cases:
a) f2fs forces to use F2FS_IPU_FORCE in a small-sized volume
b) user sets F2FS_IPU_FORCE policy via sysfs

Then we may fail to defragment file due to IPU policy check, it doesn't
make sense, let's introduce a new IPU policy to allow OPU during file
defragmentation.

In small-sized volume, let's enable F2FS_IPU_HONOR_OPU_WRITE policy
by default.

Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>

5 files changed, 32 insertions, 15 deletions

diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index 0f124e8de1d4..6b5f389ba998 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -2460,6 +2460,9 @@ static inline bool check_inplace_update_policy(struct inode *inode,
 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 	unsigned int policy = SM_I(sbi)->ipu_policy;
 
+	if (policy & (0x1 << F2FS_IPU_HONOR_OPU_WRITE) &&
+			is_inode_flag_set(inode, FI_OPU_WRITE))
+		return false;
 	if (policy & (0x1 << F2FS_IPU_FORCE))
 		return true;
 	if (policy & (0x1 << F2FS_IPU_SSR) && f2fs_need_SSR(sbi))
@@ -2530,6 +2533,9 @@ bool f2fs_should_update_outplace(struct inode *inode, struct f2fs_io_info *fio)
 	if (is_inode_flag_set(inode, FI_ALIGNED_WRITE))
 		return true;
 
+	if (is_inode_flag_set(inode, FI_OPU_WRITE))
+		return true;
+
 	if (fio) {
 		if (page_private_gcing(fio->page))
 			return true;
@@ -3154,8 +3160,8 @@ static int __f2fs_write_data_pages(struct address_space *mapping,
 			f2fs_available_free_memory(sbi, DIRTY_DENTS))
 		goto skip_write;
 
-	/* skip writing during file defragment */
-	if (is_inode_flag_set(inode, FI_DO_DEFRAG))
+	/* skip writing in file defragment preparing stage */
+	if (is_inode_flag_set(inode, FI_SKIP_WRITES))
 		goto skip_write;
 
 	trace_f2fs_writepages(mapping->host, wbc, DATA);
@@ -3725,6 +3731,7 @@ static int f2fs_migrate_blocks(struct inode *inode, block_t start_blk,
 	filemap_invalidate_lock(inode->i_mapping);
 
 	set_inode_flag(inode, FI_ALIGNED_WRITE);
+	set_inode_flag(inode, FI_OPU_WRITE);
 
 	for (; secidx < end_sec; secidx++) {
 		f2fs_down_write(&sbi->pin_sem);
@@ -3733,7 +3740,7 @@ static int f2fs_migrate_blocks(struct inode *inode, block_t start_blk,
 		f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
 		f2fs_unlock_op(sbi);
 
-		set_inode_flag(inode, FI_DO_DEFRAG);
+		set_inode_flag(inode, FI_SKIP_WRITES);
 
 		for (blkofs = 0; blkofs < blk_per_sec; blkofs++) {
 			struct page *page;
@@ -3750,7 +3757,7 @@ static int f2fs_migrate_blocks(struct inode *inode, block_t start_blk,
 			f2fs_put_page(page, 1);
 		}
 
-		clear_inode_flag(inode, FI_DO_DEFRAG);
+		clear_inode_flag(inode, FI_SKIP_WRITES);
 
 		ret = filemap_fdatawrite(inode->i_mapping);
 
@@ -3761,7 +3768,8 @@ static int f2fs_migrate_blocks(struct inode *inode, block_t start_blk,
 	}
 
 done:
-	clear_inode_flag(inode, FI_DO_DEFRAG);
+	clear_inode_flag(inode, FI_SKIP_WRITES);
+	clear_inode_flag(inode, FI_OPU_WRITE);
 	clear_inode_flag(inode, FI_ALIGNED_WRITE);
 
 	filemap_invalidate_unlock(inode->i_mapping);
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index 51c1392708e6..3b4bf1c3f1ed 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -740,7 +740,8 @@ enum {
 	FI_DROP_CACHE,		/* drop dirty page cache */
 	FI_DATA_EXIST,		/* indicate data exists */
 	FI_INLINE_DOTS,		/* indicate inline dot dentries */
-	FI_DO_DEFRAG,		/* indicate defragment is running */
+	FI_SKIP_WRITES,		/* should skip data page writeback */
+	FI_OPU_WRITE,		/* used for opu per file */
 	FI_DIRTY_FILE,		/* indicate regular/symlink has dirty pages */
 	FI_PREALLOCATED_ALL,	/* all blocks for write were preallocated */
 	FI_HOT_DATA,		/* indicate file is hot */
diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index 6ccdd6e347e2..42fbdcf0ccc9 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -2559,10 +2559,6 @@ static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
 	bool fragmented = false;
 	int err;
 
-	/* if in-place-update policy is enabled, don't waste time here */
-	if (f2fs_should_update_inplace(inode, NULL))
-		return -EINVAL;
-
 	pg_start = range->start >> PAGE_SHIFT;
 	pg_end = (range->start + range->len) >> PAGE_SHIFT;
 
@@ -2570,6 +2566,13 @@ static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
 
 	inode_lock(inode);
 
+	/* if in-place-update policy is enabled, don't waste time here */
+	set_inode_flag(inode, FI_OPU_WRITE);
+	if (f2fs_should_update_inplace(inode, NULL)) {
+		err = -EINVAL;
+		goto out;
+	}
+
 	/* writeback all dirty pages in the range */
 	err = filemap_write_and_wait_range(inode->i_mapping, range->start,
 						range->start + range->len - 1);
@@ -2651,7 +2654,7 @@ do_map:
 			goto check;
 		}
 
-		set_inode_flag(inode, FI_DO_DEFRAG);
+		set_inode_flag(inode, FI_SKIP_WRITES);
 
 		idx = map.m_lblk;
 		while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
@@ -2676,15 +2679,16 @@ check:
 		if (map.m_lblk < pg_end && cnt < blk_per_seg)
 			goto do_map;
 
-		clear_inode_flag(inode, FI_DO_DEFRAG);
+		clear_inode_flag(inode, FI_SKIP_WRITES);
 
 		err = filemap_fdatawrite(inode->i_mapping);
 		if (err)
 			goto out;
 	}
 clear_out:
-	clear_inode_flag(inode, FI_DO_DEFRAG);
+	clear_inode_flag(inode, FI_SKIP_WRITES);
 out:
+	clear_inode_flag(inode, FI_OPU_WRITE);
 	inode_unlock(inode);
 	if (!err)
 		range->len = (u64)total << PAGE_SHIFT;
diff --git a/fs/f2fs/segment.h b/fs/f2fs/segment.h
index 0291cd55cf09..5c94caf0c0a1 100644
--- a/fs/f2fs/segment.h
+++ b/fs/f2fs/segment.h
@@ -651,7 +651,9 @@ static inline int utilization(struct f2fs_sb_info *sbi)
  *                     pages over min_fsync_blocks. (=default option)
  * F2FS_IPU_ASYNC - do IPU given by asynchronous write requests.
  * F2FS_IPU_NOCACHE - disable IPU bio cache.
- * F2FS_IPUT_DISABLE - disable IPU. (=default option in LFS mode)
+ * F2FS_IPU_HONOR_OPU_WRITE - use OPU write prior to IPU write if inode has
+ *                            FI_OPU_WRITE flag.
+ * F2FS_IPU_DISABLE - disable IPU. (=default option in LFS mode)
  */
 #define DEF_MIN_IPU_UTIL	70
 #define DEF_MIN_FSYNC_BLOCKS	8
@@ -667,6 +669,7 @@ enum {
 	F2FS_IPU_FSYNC,
 	F2FS_IPU_ASYNC,
 	F2FS_IPU_NOCACHE,
+	F2FS_IPU_HONOR_OPU_WRITE,
 };
 
 static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index 9af6c20532ec..806836184ebc 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -3957,7 +3957,8 @@ static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
 		F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
 		if (f2fs_block_unit_discard(sbi))
 			sm_i->dcc_info->discard_granularity = 1;
-		sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
+		sm_i->ipu_policy = 1 << F2FS_IPU_FORCE |
+					1 << F2FS_IPU_HONOR_OPU_WRITE;
 	}
 
 	sbi->readdir_ra = 1;