Re: [v2] ext4: Add a test for rename with RENAME_WHITEOUT

[Zorro Lang <zlang@redhat.com>]

On Thu, Jan 28, 2021 at 02:12:02PM +0800, Sun Ke wrote:
> Fill the disk space, try to create some files and rename a file, mount
> again, list directory contents and triggers some errors. It is a
> regression test for kernel commit 6b4b8e6b4ad8 ("ext4: ext4: fix bug for
> rename with RENAME_WHITEOUT")
> 
> Signed-off-by: Sun Ke <sunke32@huawei.com>
> ---
> v1 -> v3:
> Use the original in src and modify 048.out
> ---
> ---
>  tests/ext4/048     | 76 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
>  tests/ext4/048.out |  2 ++
>  tests/ext4/group   |  1 +
>  3 files changed, 79 insertions(+)
>  create mode 100755 tests/ext4/048
>  create mode 100644 tests/ext4/048.out
> 
> diff --git a/tests/ext4/048 b/tests/ext4/048
> new file mode 100755
> index 00000000..0311d1a2
> --- /dev/null
> +++ b/tests/ext4/048
> @@ -0,0 +1,76 @@
> +#! /bin/bash
> +# SPDX-License-Identifier: GPL-2.0
> +# Copyright (c) 2021 HUAWEI.  All Rights Reserved.
> +#
> +# FS QA Test 048
> +#
> +# This is a regression test for kernel patch:
> +# commit 6b4b8e6b4ad8 ("ext4: ext4: fix bug for rename with RENAME_WHITEOUT")
> +
> +seq=`basename $0`
> +seqres=$RESULT_DIR/$seq
> +echo "QA output created by $seq"
> +
> +here=`pwd`
> +tmp=/tmp/$$
> +status=1	# failure is the default!
> +trap "_cleanup; exit \$status" 0 1 2 3 15
> +
> +_cleanup()
> +{
> +	cd /
> +	rm -f $tmp.*
> +}
> +
> +# get standard environment, filters and checks
> +. ./common/rc
> +. ./common/filter
> +
> +# remove previous $seqres.full before test
> +rm -f $seqres.full
> +
> +# real QA test starts here
> +
> +# Modify as appropriate.
> +_supported_fs ext4
> +_supported_fs generic

ext4 or generic? Tell the truth, according to below testing steps, I think it
can be a generic test case, not only for ext4.

> +_require_scratch
> +_require_xfs_io_command "falloc"
> +
> +dmesg -c > /dev/null
> +
> +_scratch_mkfs > $seqres.full 2>&1
> +_scratch_mount >> $seqres.full 2>&1
> +
> +testdir=$SCRATCH_MNT
> +cd ${testdir}
> +
> +mkdir test
> +$XFS_IO_PROG -f -c "falloc 0 128M" img >> $seqres.full
> +$MKFS_EXT4_PROG  img > /dev/null 2>&1
> +$MOUNT_PROG img test

Can we reproduce this bug by $SCRATCH_DEV directly, not through a loopdev.
Likes: 

_scratch_mkfs_sized 128M (maybe bigger ?)
_scratch_mount

> +
> +# fill the disk space
> +dd if=/dev/zero of=test/foo bs=1M count=128 > /dev/null 2>&1

Can _fill_fs() helper help you that?

> +
> +# create 1000 files, not all the files will be created successfully
> +mkdir test/dir
> +cd test/dir
> +for ((i = 0; i < 1000; i++))
> +do
> +	touch file$i > /dev/null 2>&1
> +done

I don't know if it can be 100% sure there's at least 1 file will be created at
here, after you tried to fill the fs.

So I think maybe you can create a file with known name, then use _fill_fs()
to fill the whole fs. Then try to rename the file you know its name. Can that
reproduce the bug?

> +
> +# try to rename, but now no space left on device
> +$here/src/renameat2 -w $testdir/test/dir/file1 $testdir/test/dir/dst_file
> +
> +cd $testdir
> +$UMOUNT_PROG test
> +$MOUNT_PROG img test
> +ls -l test/dir/file1 > /dev/null 2>&1
> +$UMOUNT_PROG test
> +
> +dmesg -c | grep "deleted inode referenced"

Can _check_dmesg_for() help you?

> +# success, all done
> +status=0
> +exit
> diff --git a/tests/ext4/048.out b/tests/ext4/048.out
> new file mode 100644
> index 00000000..5b3990da
> --- /dev/null
> +++ b/tests/ext4/048.out
> @@ -0,0 +1,2 @@
> +QA output created by 048
> +No space left on device
> diff --git a/tests/ext4/group b/tests/ext4/group
> index ceda2ba6..6140dd7e 100644
> --- a/tests/ext4/group
> +++ b/tests/ext4/group
> @@ -50,6 +50,7 @@
>  045 auto dir
>  046 auto prealloc quick
>  047 auto quick dax
> +048 other

I think your test case is a general test case, "auto"and "rename" groups
are good to it, if it's quick enough, add "quick" group.

>  271 auto rw quick
>  301 aio auto ioctl rw stress defrag
>  302 aio auto ioctl rw stress defrag
> -- 
> 2.13.6
>