[Question] generic/064 fail on f2fs

[Sun Ke <sunke32@xxxxxxxxxx>]

I run generic/064 on f2fs:

[root@localhost xfstests-dev]# ./check tests/generic/064
FSTYP         -- f2fs
PLATFORM      -- Linux/x86_64 localhost 5.12.0-rc5-next-20210330 #8 SMP 
Tue Jun 8 11:09:37 CST 2021
MKFS_OPTIONS  -- /dev/sdb
MOUNT_OPTIONS -- -o acl,user_xattr /dev/sdb /tmp/scratch

generic/064 - output mismatch (see 
/root/work/xfstests-dev/results//generic/064.out.bad)
    --- tests/generic/064.out   2021-05-31 15:37:40.000000000 +0800
    +++ /root/work/xfstests-dev/results//generic/064.out.bad 
2021-06-10 17:19:09.469000000 +0800
    @@ -1,2 +1,3 @@
     QA output created by 064
     Extent count after inserts is in range
    +extents mismatched before = 1 after = 50
    ...
    (Run 'diff -u /root/work/xfstests-dev/tests/generic/064.out 
/root/work/xfstests-dev/results//generic/064.out.bad'  to see the entire 
diff)
Ran: generic/064
Failures: generic/064
Failed 1 of 1 tests


***

I added some test code and  tried on xfs and ext4:

root@localhost xfstests-dev]# git diff
diff --git a/tests/generic/064 b/tests/generic/064
index 7098fcd3..1baacb06 100755
--- a/tests/generic/064
+++ b/tests/generic/064
@@ -45,6 +45,8 @@ length=$(($BLOCKS * $BSIZE))
 _do "$XFS_IO_PROG -f -c \"pwrite 0 $length\" -c fsync $src"
 cp $src $dest
 extent_before=`_count_extents $dest`
+echo "extent_before: ${extent_before}" > log.txt
+echo "filesize: $(du -sh $dest)" >> log.txt

 # Insert alternate blocks
 for (( j=0; j < $(($BLOCKS/2)); j++ )); do
@@ -55,6 +57,8 @@ done
 # Check if 50 extents are present, allowing some slop for file systems
 # that don't have ideal allocation behavior
 num_extents=`_count_extents $dest`
+echo "num_extents: ${num_extents}" >> log.txt
+echo "filesize: $(du -sh $dest)" >> log.txt
 _within_tolerance "Extent count after inserts" $num_extents 50 0 6% -v

 _check_scratch_fs
@@ -70,6 +74,8 @@ for (( j=0; j < $(($BLOCKS/2)); j++ )); do
 done

 extent_after=`_count_extents $dest`
+echo "extent_after: ${extent_after}" >> log.txt
+echo "filesize: $(du -sh $dest)" >> log.txt
 if [ $extent_before -ne $extent_after ]; then
        echo "extents mismatched before = $extent_before after = 
$extent_after"
 fi


ext4:

extent_before: 1
filesize: 400K  /tmp/scratch/testfile.dest
num_extents: 50
filesize: 404K  /tmp/scratch/testfile.dest
extent_after: 1
filesize: 404K  /tmp/scratch/testfile.dest

xfs:

extent_before: 1
filesize: 400K  /tmp/scratch/testfile.dest
num_extents: 50
filesize: 404K  /tmp/scratch/testfile.dest 


extent_after: 1
filesize: 400K  /tmp/scratch/testfile.dest


f2fs:

extent_before: 1
filesize: 400K  /tmp/scratch/testfile.dest
num_extents: 50
filesize: 400K  /tmp/scratch/testfile.dest
extent_after: 50
filesize: 400K  /tmp/scratch/testfile.dest


***
The filesize is always 400K and extent_after is still 50. Is it a bug in 
f2fs or I need adjust the case for f2fs?

I also wondering that why after insert 50 holes, the filesize is 404K? 
I think it should be 400K + 4K * 50 = 600K.

The fallocate(2) manual page say: Inserting a hole inside a file 
increases the file size by len bytes.

BLOCKS=100
BSIZE=4096

for (( j=0; j < $(($BLOCKS/2)); j++ )); do
    offset=$((($j*$BSIZE)*2))
    _do "$XFS_IO_PROG -c \"finsert $offset $BSIZE\" $dest"
done