Skip to contentSkip to conversation

f2fs: allocate blocks for pinned file

PATCH In review Jaegeuk Kim 2 participants 6 messages Jun 27, 2019 – Jun 29, 2019

Subsystems (2)
lore
Share & tools
lore

Jaegeuk Kim

author maintainer
#0 lore raw

PATCH 1/1 f2fs: allocate blocks for pinned file

This patch allows fallocate to allocate physical blocks for pinned file.

---
 fs/f2fs/file.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
fs/f2fs/file.c +6 -1
index e7c368db8185..cdfd4338682d 100644
@@ -1528,7 +1528,12 @@ static int expand_inode_data(struct inode *inode, loff_t offset,
 	if (off_end)
 		map.m_len++;
 
-	err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
+	if (f2fs_is_pinned_file(inode))
+		map.m_seg_type = CURSEG_COLD_DATA;
+
+	err = f2fs_map_blocks(inode, &map, 1, (f2fs_is_pinned_file(inode) ?
+						F2FS_GET_BLOCK_PRE_DIO :
+						F2FS_GET_BLOCK_PRE_AIO));
 	if (err) {
 		pgoff_t last_off;
 
Signature · 1 line
2.19.0.605.g01d371f741-goog
#1 lore raw

Hi Jaegeuk,

Quoted 1 line · Jaegeuk Kim

This patch allows fallocate to allocate physical blocks for pinned file.

Quoted from manual of fallocate(2):
"
Any subregion within the range specified by offset and len that did not contain data before the call will be initialized to zero.

If the FALLOC_FL_KEEP_SIZE flag is specified in mode .... Preallocating zeroed blocks beyond the end of the file in this manner is useful for optimizing append workloads.
"

As quoted description, our change may break the rule of fallocate(, mode = 0), because with after this change, we can't guarantee that preallocated physical block contains zeroed data

Should we introduce an additional ioctl for this case? Or maybe add one more flag in fallocate() for unzeroed block preallocation, not sure.

Thanks,

Quoted 25 lines
---
 fs/f2fs/file.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
fs/f2fs/file.c +6 -1
index e7c368db8185..cdfd4338682d 100644
@@ -1528,7 +1528,12 @@ static int expand_inode_data(struct inode *inode, loff_t offset,
 	if (off_end)
 		map.m_len++;
 
-	err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
+	if (f2fs_is_pinned_file(inode))
+		map.m_seg_type = CURSEG_COLD_DATA;
+
+	err = f2fs_map_blocks(inode, &map, 1, (f2fs_is_pinned_file(inode) ?
+						F2FS_GET_BLOCK_PRE_DIO :
+						F2FS_GET_BLOCK_PRE_AIO));
 	if (err) {
 		pgoff_t last_off;
 
author maintainer
#2 lore raw
Quoted 15 lines · Chao Yu

Hi Jaegeuk,

Quoted 1 line · Jaegeuk Kim

This patch allows fallocate to allocate physical blocks for pinned file.

Quoted from manual of fallocate(2):
"
Any subregion within the range specified by offset and len that did not contain data before the call will be initialized to zero.

If the FALLOC_FL_KEEP_SIZE flag is specified in mode .... Preallocating zeroed blocks beyond the end of the file in this manner is useful for optimizing append workloads.
"

Quoted 6 lines

As quoted description, our change may break the rule of fallocate(, mode = 0), because with after this change, we can't guarantee that preallocated physical block contains zeroed data

Should we introduce an additional ioctl for this case? Or maybe add one more flag in fallocate() for unzeroed block preallocation, not sure.

I thought like that, but this is a very corner case for the pinned file only in f2fs. And, the pinned file is also indeed used by this purpose.

Thanks,

Quoted 28 lines

Thanks,

Quoted 25 lines
---
 fs/f2fs/file.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
fs/f2fs/file.c +6 -1
index e7c368db8185..cdfd4338682d 100644
@@ -1528,7 +1528,12 @@ static int expand_inode_data(struct inode *inode, loff_t offset,
 	if (off_end)
 		map.m_len++;
 
-	err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
+	if (f2fs_is_pinned_file(inode))
+		map.m_seg_type = CURSEG_COLD_DATA;
+
+	err = f2fs_map_blocks(inode, &map, 1, (f2fs_is_pinned_file(inode) ?
+						F2FS_GET_BLOCK_PRE_DIO :
+						F2FS_GET_BLOCK_PRE_AIO));
 	if (err) {
 		pgoff_t last_off;
 
#3 lore raw
Quoted 20 lines · Jaegeuk Kim
Quoted 19 lines · Chao Yu

Hi Jaegeuk,

Quoted 1 line · Jaegeuk Kim

This patch allows fallocate to allocate physical blocks for pinned file.

Quoted from manual of fallocate(2):
"
Any subregion within the range specified by offset and len that did not contain data before the call will be initialized to zero.

If the FALLOC_FL_KEEP_SIZE flag is specified in mode .... Preallocating zeroed blocks beyond the end of the file in this manner is useful for optimizing append workloads.
"

As quoted description, our change may break the rule of fallocate(, mode = 0), because with after this change, we can't guarantee that preallocated physical block contains zeroed data

Quoted 5 lines
Quoted 2 lines

Should we introduce an additional ioctl for this case? Or maybe add one more flag in fallocate() for unzeroed block preallocation, not sure.

I thought like that, but this is a very corner case for the pinned file only in f2fs. And, the pinned file is also indeed used by this purpose.

Okay, I think we need to find one place to document such behavior that is not matching regular poxis fallocate's sematic, user should be noticed about it.

Thanks,

Quoted 33 lines

Thanks,

Quoted 17 lines

Thanks,

Quoted 14 lines
---
 fs/f2fs/file.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
fs/f2fs/file.c +0 -0
index e7c368db8185..cdfd4338682d 100644
@@ -1528,7 +1528,12 @@ static int expand_inode_data(struct inode *inode, loff_t offset,
 	if (off_end)
 		map.m_len++;
 
Quoted 11 lines

fs/f2fs/file.c static int expand_inode_data(struct inode *inode, loff_t offset, · line 1539

Quoted 11 lines
fs/f2fs/file.c +6 -1
-	err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
+	if (f2fs_is_pinned_file(inode))
+		map.m_seg_type = CURSEG_COLD_DATA;
+
+	err = f2fs_map_blocks(inode, &map, 1, (f2fs_is_pinned_file(inode) ?
+						F2FS_GET_BLOCK_PRE_DIO :
+						F2FS_GET_BLOCK_PRE_AIO));
 	if (err) {
 		pgoff_t last_off;
 

.

author maintainer
#4 lore raw
In reply to Jaegeuk Kim

This patch allows fallocate to allocate physical blocks for pinned file.

---
 Documentation/filesystems/f2fs.txt | 25 +++++++++++++++++++++++++
 fs/f2fs/file.c                     |  7 ++++++-
 2 files changed, 31 insertions(+), 1 deletion(-)
Documentation/filesystems/f2fs.txt +25 -0
index bebd1be3ba49..496fa28b2492 100644
@@ -804,3 +804,28 @@ WRITE_LIFE_NOT_SET    WARM_DATA                WRITE_LIFE_NOT_SET
 WRITE_LIFE_NONE       "                        WRITE_LIFE_NONE
 WRITE_LIFE_MEDIUM     "                        WRITE_LIFE_MEDIUM
 WRITE_LIFE_LONG       "                        WRITE_LIFE_LONG
+
+Fallocate(2) Policy
+-------------------
+
+The default policy follows the below posix rule.
+
+Allocating disk space
+    The default operation (i.e., mode is zero) of fallocate() allocates
+    the disk space within the range specified by offset and len.  The
+    file size (as reported by stat(2)) will be changed if offset+len is
+    greater than the file size.  Any subregion within the range specified
+    by offset and len that did not contain data before the call will be
+    initialized to zero.  This default behavior closely resembles the
+    behavior of the posix_fallocate(3) library function, and is intended
+    as a method of optimally implementing that function.
+
+However, once F2FS receives ioctl(fd, F2FS_IOC_SET_PIN_FILE) in prior to
+fallocate(fd, DEFAULT_MODE), it allocates on-disk blocks addressess having
+zero or random data, which is useful to the below scenario where:
+ 1. create(fd)
+ 2. ioctl(fd, F2FS_IOC_SET_PIN_FILE)
+ 3. fallocate(fd, 0, 0, size)
+ 4. address = fibmap(fd, offset)
+ 5. open(blkdev)
+ 6. write(blkdev, address)
fs/f2fs/file.c +6 -1
index e7c368db8185..cdfd4338682d 100644
@@ -1528,7 +1528,12 @@ static int expand_inode_data(struct inode *inode, loff_t offset,
 	if (off_end)
 		map.m_len++;
 
-	err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
+	if (f2fs_is_pinned_file(inode))
+		map.m_seg_type = CURSEG_COLD_DATA;
+
+	err = f2fs_map_blocks(inode, &map, 1, (f2fs_is_pinned_file(inode) ?
+						F2FS_GET_BLOCK_PRE_DIO :
+						F2FS_GET_BLOCK_PRE_AIO));
 	if (err) {
 		pgoff_t last_off;
 
Signature · 1 line
2.19.0.605.g01d371f741-goog