__send_empty_flush() sends empty flush bios to every target in the
dm_table. However, if the num_targets exceeds the number of block
devices in the dm_table's device list, it could lead to multiple
invocations of __send_duplicate_bios() for the same block device.
Typically, a single thread sending numerous empty flush bios to one
block device is redundant, as these bios are likely to be merged by the
flush state machine. In scenarios where num_targets significantly
outweighs the number of block devices, such behavior may result in a
noteworthy decrease in performance.
This issue can be reproduced using this command line:
for i in {0..1023}; do
echo $((8000*$i)) 8000 linear /dev/sda2 $((16384*$i))
done | dmsetup create example
With this fix, a random write with fsync workload executed with the
following fio command:
fio --group_reporting --name=benchmark --filename=/dev/mapper/example \
--ioengine=sync --invalidate=1 --numjobs=16 --rw=randwrite \
--blocksize=4k --size=2G --time_based --runtime=30 --fdatasync=1
results in an increase from 857 KB/s to 30.8 MB/s of the write
throughput (3580% increase).
Signed-off-by: Yang Yang <yang.yang@xxxxxxxx>
---
drivers/md/dm-linear.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/drivers/md/dm-linear.c b/drivers/md/dm-linear.c
index 2d3e186ca87e..3e1a33b4d289 100644
--- a/drivers/md/dm-linear.c
+++ b/drivers/md/dm-linear.c
@@ -62,6 +62,7 @@ static int linear_ctr(struct dm_target *ti, unsigned int argc, char **argv)
ti->num_discard_bios = 1;
ti->num_secure_erase_bios = 1;
ti->num_write_zeroes_bios = 1;
+ ti->flush_pass_around = 1;
ti->private = lc;
return 0;
--
2.34.1
