ext4: avoid starting handle when dio writing an unwritten extent

Since we have deferred the split of the unwritten extent until after I/O
completion, it is not necessary to initiate the journal handle when
submitting the I/O.

This can improve the write performance of concurrent DIO for multiple
files. The fio tests below show a ~25% performance improvement when
wirting to unwritten files on my VM with a mem disk.

  [unwritten]
  direct=1
  ioengine=psync
  numjobs=16
  rw=write     # write/randwrite
  bs=4K
  iodepth=1
  directory=/mnt
  size=5G
  runtime=30s
  overwrite=0
  norandommap=1
  fallocate=native
  ramp_time=5s
  group_reporting=1

 [w/o]
  w:  IOPS=62.5k, BW=244MiB/s
  rw: IOPS=56.7k, BW=221MiB/s

 [w]
  w:  IOPS=79.6k, BW=311MiB/s
  rw: IOPS=70.2k, BW=274MiB/s

Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: Baokun Li <libaokun1@huawei.com>
Reviewed-by: Ojaswin Mujoo <ojaswin@linux.ibm.com>
Link: https://patch.msgid.link/20260105014522.1937690-4-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

diff --git a/fs/ext4/file.c b/fs/ext4/file.c
index 484cb73888024879c70d1104f495fc635e42af07..4fadc476d64568b85380248cefb3dbeea67e8a1d 100644 (file)
--- a/fs/ext4/file.c
+++ b/fs/ext4/file.c
@@ -418,9 +418,7 @@ static const struct iomap_dio_ops ext4_dio_write_ops = {
  *   updating inode i_disksize and/or orphan handling with exclusive lock.
  *
  * - shared locking will only be true mostly with overwrites, including
- *   initialized blocks and unwritten blocks. For overwrite unwritten blocks
- *   we protect splitting extents by i_data_sem in ext4_inode_info, so we can
- *   also release exclusive i_rwsem lock.
+ *   initialized blocks and unwritten blocks.
  *
  * - Otherwise we will switch to exclusive i_rwsem lock.
  */

diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index b86ddc9e85ce80a29e11aa0bbf423fcdd22f52ae..1cb7775c5e41ea11090451f63d956b9f4da8f222 100644 (file)
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -3813,9 +3813,14 @@ static int ext4_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
                        ret = ext4_map_blocks(NULL, inode, &map, 0);
                        /*
                         * For atomic writes the entire requested length should
-                        * be mapped.
+                        * be mapped. For DAX we convert extents to initialized
+                        * ones before copying the data, otherwise we do it
+                        * after I/O so there's no need to call into
+                        * ext4_iomap_alloc().
                         */
-                       if (map.m_flags & EXT4_MAP_MAPPED) {
+                       if ((map.m_flags & EXT4_MAP_MAPPED) ||
+                           (!(flags & IOMAP_DAX) &&
+                            (map.m_flags & EXT4_MAP_UNWRITTEN))) {
                                if ((!(flags & IOMAP_ATOMIC) && ret > 0) ||
                                   (flags & IOMAP_ATOMIC && ret >= orig_mlen))
                                        goto out;