Skip to contentSkip to conversation

f2fs: do in batches truncation in truncate_hole

PATCH Superseded Chao Yu 2 participants 3 messages Sep 11, 2015 – Sep 17, 2015 F2FS file system

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Chao Yu

author
#0 lore raw

PATCH 2/7 f2fs: do in batches truncation in truncate_hole

truncate_data_blocks_range can do in batches truncation which makes all changes in dnode page content, dnode page status, extent cache, block count updating together.

But previously, truncate_hole() always truncates one block in dnode page at a time by invoking truncate_data_blocks_range(,1), which make thing slow.

This patch changes truncate_hole() to do in batches truncation for all target blocks in one direct node inside truncate_data_blocks_range, which can make our punch hole operation in ->fallocate more efficent.

---
 fs/f2fs/file.c | 20 +++++++++++++++-----
 1 file changed, 15 insertions(+), 5 deletions(-)
fs/f2fs/file.c +15 -5
index b2fab9e..ac97f78 100644
@@ -738,23 +738,33 @@ static int fill_zero(struct inode *inode, pgoff_t index,
 
 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
 {
-	pgoff_t index;
+	pgoff_t count, index = pg_start;
 	int err;
 
-	for (index = pg_start; index < pg_end; index++) {
+	while (index < pg_end) {
 		struct dnode_of_data dn;
+		pgoff_t end_offset;
 
 		set_new_dnode(&dn, inode, NULL, NULL, 0);
 		err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
 		if (err) {
-			if (err == -ENOENT)
+			if (err == -ENOENT) {
+				index = PGOFS_OF_NEXT_DNODE(index,
+							F2FS_I(inode));
 				continue;
+			}
 			return err;
 		}
 
-		if (dn.data_blkaddr != NULL_ADDR)
-			truncate_data_blocks_range(&dn, 1);
+		end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
+		count = min(end_offset - dn.ofs_in_node, pg_end - index);
+
+		f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
+
+		truncate_data_blocks_range(&dn, count);
 		f2fs_put_dnode(&dn);
+
+		index += count;
 	}
 	return 0;
 }
Signature · 1 line
2.4.2
maintainer
#1 lore raw

Hi Chao,

How about changing the original patch slightly like below?

o don't have to use index variable
o should not skip the entire dnode indices, when one of them is -ENOENT

---
 fs/f2fs/file.c | 13 ++++++-------
 1 file changed, 6 insertions(+), 7 deletions(-)
fs/f2fs/file.c +6 -7
index 180b838..e8536a8 100644
@@ -739,33 +739,32 @@ static int fill_zero(struct inode *inode, pgoff_t index,
 
 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
 {
-	pgoff_t count, index = pg_start;
+	pgoff_t count;
 	int err;
 
-	while (index < pg_end) {
+	while (pg_start < pg_end) {
 		struct dnode_of_data dn;
 		pgoff_t end_offset;
 
 		set_new_dnode(&dn, inode, NULL, NULL, 0);
-		err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
+		err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
 		if (err) {
 			if (err == -ENOENT) {
-				index = PGOFS_OF_NEXT_DNODE(index,
-							F2FS_I(inode));
+				pg_start++;
 				continue;
 			}
 			return err;
 		}
 
 		end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
-		count = min(end_offset - dn.ofs_in_node, pg_end - index);
+		count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
 
 		f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
 
 		truncate_data_blocks_range(&dn, count);
 		f2fs_put_dnode(&dn);
 
-		index += count;
+		pg_start += count;
 	}
 	return 0;
 }
Signature · 1 line
2.1.1
Quoted 62 lines · Chao Yu

truncate_data_blocks_range can do in batches truncation which makes all changes in dnode page content, dnode page status, extent cache, block count updating together.

But previously, truncate_hole() always truncates one block in dnode page at a time by invoking truncate_data_blocks_range(,1), which make thing slow.

This patch changes truncate_hole() to do in batches truncation for all target blocks in one direct node inside truncate_data_blocks_range, which can make our punch hole operation in ->fallocate more efficent.

---
 fs/f2fs/file.c | 20 +++++++++++++++-----
 1 file changed, 15 insertions(+), 5 deletions(-)
fs/f2fs/file.c +15 -5
index b2fab9e..ac97f78 100644
@@ -738,23 +738,33 @@ static int fill_zero(struct inode *inode, pgoff_t index,
 
 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
 {
-	pgoff_t index;
+	pgoff_t count, index = pg_start;
 	int err;
 
-	for (index = pg_start; index < pg_end; index++) {
+	while (index < pg_end) {
 		struct dnode_of_data dn;
+		pgoff_t end_offset;
 
 		set_new_dnode(&dn, inode, NULL, NULL, 0);
 		err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
 		if (err) {
-			if (err == -ENOENT)
+			if (err == -ENOENT) {
+				index = PGOFS_OF_NEXT_DNODE(index,
+							F2FS_I(inode));
 				continue;
+			}
 			return err;
 		}
 
-		if (dn.data_blkaddr != NULL_ADDR)
-			truncate_data_blocks_range(&dn, 1);
+		end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
+		count = min(end_offset - dn.ofs_in_node, pg_end - index);
+
+		f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
+
+		truncate_data_blocks_range(&dn, count);
 		f2fs_put_dnode(&dn);
+
+		index += count;
 	}
 	return 0;
 }
Signature · 1 line
2.4.2
author
#2 lore raw
Quoted 7 lines

-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Thursday, September 17, 2015 2:01 AM
To: Chao Yu
Cc: [email protected]; [email protected]
Subject: Re: [PATCH 2/7] f2fs: do in batches truncation in truncate_hole

Quoted 6 lines

Hi Chao,

How about changing the original patch slightly like below?

o don't have to use index variable
o should not skip the entire dnode indices, when one of them is -ENOENT

Agree, thanks for your review! :)

Thanks,