On Mon, Feb 14, 2022 at 03:34:09PM -0500, Masayoshi Mizuma wrote:
> From: Masayoshi Mizuma <m.mizuma@xxxxxxxxxxxxxx>
>
> commit 25195eb ("xfsdump: handle bind mount target") introduced
> a bug of xfsdump which doesn't store the files to the dump file
> correctly when the root inode number is changed.
>
> The commit 25195eb is reverted, and commit 0717c1c ("xfsdump: intercept
> bind mount targets") which is in xfsdump v3.1.10 fixes the bug to reject
> the filesystem if it's bind-mounted.
>
> Test that xfsdump can reject the bind-mounted filesystem.
>
> Signed-off-by: Masayoshi Mizuma <m.mizuma@xxxxxxxxxxxxxx>
> ---
> tests/xfs/544 | 63 +++++++++++++++++++++++++++++++++++++++++++++++
> tests/xfs/544.out | 2 ++
> 2 files changed, 65 insertions(+)
> create mode 100755 tests/xfs/544
> create mode 100644 tests/xfs/544.out
>
> diff --git a/tests/xfs/544 b/tests/xfs/544
> new file mode 100755
> index 00000000..1d586ebc
> --- /dev/null
> +++ b/tests/xfs/544
> @@ -0,0 +1,63 @@
> +#! /bin/bash
> +# SPDX-License-Identifier: GPL-2.0
> +# Copyright (c) 2022 Fujitsu Limited. All Rights Reserved.
> +#
> +# FS QA Test 544
> +#
> +# Regression test for commit:
> +# 0717c1c ("xfsdump: intercept bind mount targets")
> +
> +. ./common/preamble
> +_begin_fstest auto dump
> +
> +_cleanup()
> +{
> + _cleanup_dump
Is this really needed?
> + cd /
> + rm -r -f $tmp.*
> + $UMOUNT_PROG $mntpnt2 2> /dev/null
> + rmdir $mntpnt1/dir 2> /dev/null
> + $UMOUNT_PROG $mntpnt1 2> /dev/null
> + rmdir $mntpnt2 2> /dev/null
> + rmdir $mntpnt1 2> /dev/null
> + [ -n "$loopdev" ] && _destroy_loop_device $loopdev
> + rm -f "$TEST_DIR"/fsfile
> +}
> +
> +# Import common functions.
> +. ./common/filter
> +. ./common/dump
> +
> +# real QA test starts here
> +
> +_supported_fs xfs
> +
> +mntpnt1=$TEST_DIR/MNT1
> +mntpnt2=$TEST_DIR/MNT2
> +
> +
> +# Set up
> +$MKFS_XFS_PROG -s size=4096 -b size=4096 \
> + -dfile,name=$TEST_DIR/fsfile,size=8649728b,sunit=1024,swidth=2048 \
> + >> $seqres.full 2>&1 || _fail "mkfs failed"
Are "-s size=4096 -b size=4096 -d sunit=1024,swidth=2048" necessary? If so, better
to add comments to explain why they're needed.
> +
> +loopdev=$(_create_loop_device "$TEST_DIR"/fsfile)
Is loopdev necessary? If you need mkfs, can we just use SCRATCH_DEV, and bind $SCRATCH_MNT/someonedir
on $TEST_DIR/someonedir ? E.g:
_cleanup()
{
$UMOUNT_PROG $TEST_DIR/dest 2> /dev/null
cd /
rm -r -f $tmp.*
}
_scratch_mkfs > $seqres.full
_scratch_mount
mkdir $SCRATCH_MNT/src
mkdir $TEST_DIR/dest
$MOUNT_PROG --bind $SCRATCH_MNT/src $TEST_DIR/dest || _fail "Bind mount failed"
$XFSDUMP_PROG -L session -M test -f $tmp.dump $TEST_DIR/dest >> $seqres.full 2>&1 && \
echo "dump with bind-mounted should be failed, but passed."
echo "Silence is golden"
...
If you don't need mkfs, you can use $TEST_DIR (e.g. $TEST_DIR/src and $TEST_DIR/dest) to
be bind mount source and target. Due to this case just trys to test xfsdump a bind mount,
and make sure dump failed as expected. Am I right?
Thanks,
Zorro
> +
> +mkdir $mntpnt1 >> $seqres.full 2>&1 || _fail "mkdir \"$mntpnt1\" failed"
> +
> +_mount $loopdev $mntpnt1
> +mkdir $mntpnt1/dir >> $seqres.full 2>&1 || _fail "mkdir \"$mntpnt1/dir\" failed"
> +mkdir $mntpnt2 >> $seqres.full 2>&1 || _fail "mkdir \"$mntpnt2\" failed"
> +
> +
> +# Test
> +echo "*** dump with bind-mounted test ***" >> $seqres.full
> +
> +mount -o bind $mntpnt1/dir $mntpnt2
> +
> +$XFSDUMP_PROG -L session -M test -f $tmp.dump $mntpnt2 \
> + >> $seqres.full 2>&1 && echo "dump with bind-mounted should be failed, but passed."
> +
> +echo "Silence is golden"
> +status=0
> +exit
> diff --git a/tests/xfs/544.out b/tests/xfs/544.out
> new file mode 100644
> index 00000000..fc7ebff3
> --- /dev/null
> +++ b/tests/xfs/544.out
> @@ -0,0 +1,2 @@
> +QA output created by 544
> +Silence is golden
> --
> 2.31.1
>
