Received-SPF: pass (google.com: domain of linux-ext4-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Mon, 21 Feb 2022 11:30:46 +0530
From:   Ojaswin Mujoo <ojaswin@linux.ibm.com>
To:     Eryu Guan <guan@eryu.me>
Cc:     fstests@vger.kernel.org, linux-ext4@vger.kernel.org,
        riteshh@linux.ibm.com
Subject: Re: [PATCH 2/2] ext4: Test to ensure resize with sparse_super2 is
 handled correctly
Message-ID: <YhMqjlb+qrgHTOVs@li-bb2b2a4c-3307-11b2-a85c-8fa5c3a69313.ibm.com>
References: <cover.1644217569.git.ojaswin@linux.ibm.com>
 <aead63bfa6cca5a8a1c8225075f48a29d06df1ae.1644217569.git.ojaswin@linux.ibm.com>
 <YhJrWA9VaG73H5KC@desktop>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <YhJrWA9VaG73H5KC@desktop>
Precedence: bulk

Hi Eryu

Thanks for the review.

On Mon, Feb 21, 2022 at 12:24:56AM +0800, Eryu Guan wrote:
> On Mon, Feb 07, 2022 at 01:55:34PM +0530, Ojaswin Mujoo wrote:
> > Kernel currently doesn't support resize of EXT4 mounted
> > with sparse_super2 option enabled. Earlier, it used to leave the resize
> > incomplete and the fs would be left in an inconsistent state, however commit
> > b1489186cc83[1] fixed this to avoid the fs corruption by clearly returning
> > -ENOTSUPP.
> > 
> > Test to ensure that kernel handles resizing with sparse_super2 correctly. Run
> > resize for multiple iterations because this leads to kernel crash due to
> > fs corruption, which we want to detect.
> > 
> > Related commit in mainline:
> > 
> > [1] commit b1489186cc8391e0c1e342f9fbc3eedf6b944c61
> > 
> > 	ext4: add check to prevent attempting to resize an fs with sparse_super2
> > 
> > Signed-off-by: Ojaswin Mujoo <ojaswin@linux.ibm.com>
> > ---
> >  tests/ext4/056     | 102 +++++++++++++++++++++++++++++++++++++++++++++
> >  tests/ext4/056.out |   2 +
> >  2 files changed, 104 insertions(+)
> >  create mode 100755 tests/ext4/056
> >  create mode 100644 tests/ext4/056.out
> > 
> > diff --git a/tests/ext4/056 b/tests/ext4/056
> > new file mode 100755
> > index 00000000..9185621d
> > --- /dev/null
> > +++ b/tests/ext4/056
> > @@ -0,0 +1,102 @@
> > +#! /bin/bash
> > +# SPDX-License-Identifier: GPL-2.0
> > +# Copyright (c) 2022 IBM. All Rights Reserved.
> > +#
> > +# We don't currently support resize of EXT4 filesystems mounted
> > +# with sparse_super2 option enabled. Earlier, kernel used to leave the resize
> > +# incomplete and the fs would be left into an incomplete state, however commit
> > +# b1489186cc83 fixed this to avoid the fs corruption by clearly returning
> > +# -ENOTSUPP.
> > +#
> > +# This test ensures that kernel handles resizing with sparse_super2 correctly
> > +#
> > +# Related commit in mainline:
> > +#
> > +# commit b1489186cc8391e0c1e342f9fbc3eedf6b944c61
> > +# ext4: add check to prevent attempting to resize an fs with sparse_super2
> > +#
> > +
> > +. ./common/preamble
> > +_begin_fstest auto ioctl resize quick
> > +
> > +# real QA test starts here
> > +
> > +INITIAL_FS_SIZE=1G
> > +RESIZED_FS_SIZE=$((2*1024*1024*1024))  # 2G
> > +ONLINE_RESIZE_BLOCK_LIMIT=$((256*1024*1024))
> > +
> > +_supported_fs ext4
> > +_require_scratch_size $(($RESIZED_FS_SIZE/1024))
> > +_require_test_program "ext4_resize"
> > +
> > +_log()
> > +{
> > +	echo "$seq: $*" >> $seqres.full 2>&1
> > +}
> 
> Leading under score is used for common functions, local functions don't
> need it.
> 
Got it.
> > +
> > +do_resize()
> > +{
> > +
> > +	$MKFS_PROG `_scratch_mkfs_options -t ext4 -E resize=$ONLINE_RESIZE_BLOCK_LIMIT \
> > +		-O sparse_super2` $INITIAL_FS_SIZE >> $seqres.full 2>&1 \
> > +		|| _fail "$MKFS_PROG failed. Exiting"
> 
> I think we could use _mkfs_dev here.
> 
Oh right, I missed that. I'll fix this, thanks.
> > +
> > +	_scratch_mount || _fail "Failed to mount scratch partition. Exiting"
> > +
> > +	BS=$(_get_block_size $SCRATCH_MNT)
> > +	NEW_BLOCKS=$(($RESIZED_FS_SIZE/$BS))
> > +
> > +	local RESIZE_RET
> > +	local EOPNOTSUPP=95
> > +
> > +	$here/src/ext4_resize $SCRATCH_MNT $NEW_BLOCKS >> $seqres.full 2>&1
> > +	RESIZE_RET=$?
> > +
> > +	# Use $RESIZE_RET for logging
> > +	if [ $RESIZE_RET = 0 ]
> > +	then
> > +		_log "Resizing succeeded but FS might still be corrupt."
> 
> I don't think _log is needed here, just echo message to stdout and this
> will break golden image and fail the test.
> 
Noted.
> > +	elif [ $RESIZE_RET = $EOPNOTSUPP ]
> > +	then
> > +		_log "Resize operation not supported with sparse_super2"
> > +		_log "Threw expected error $RESIZE_RET (EOPNOTSUPP)"
> > +
> > +	else
> > +		_log "Output of resize = $RESIZE_RET. Expected $EOPNOTSUPP (EOPNOTSUPP)"
> > +		_log "You might be missing kernel patch b1489186cc83"
> > +	fi
> > +
> > +	# unount after ioctl sometimes fails with "device busy" so add a small delay
> > +	sleep 0.1
> > +
> > +	_scratch_unmount >> $seqres.full 2>&1 || _fail "$UMOUNT_PROG failed. Exiting"
> > +}
> > +
> > +run_test()
> > +{
> > +	local FSCK_RET
> > +	local ITERS=8
> > +
> > +	for i in $(seq 1 $ITERS)
> > +	do
> > +		_log "----------- Iteration: $i ------------"
> > +		do_resize
> > +	done
> > +
> > +	_log "-------- Iterations Complete ---------"
> > +	_log "Checking if FS is in consistent state"
> > +	_check_scratch_fs
> > +	FSCK_RET=$?
> > +
> > +	return $FSCK_RET
> > +}
> > +
> > +run_test
> > +status=$?
> > +
> > +if [ "$status" -eq "0" ]
> > +then
> > +	echo "Test Succeeded!" | tee -a $seqres.full
> > +fi
> 
> This is not needed, just echo "Silence is golden" at the beginning of
> the test, so any additional output will fail the test.
> 
> Thanks,
> Eryu

Got it, thanks for the suggestions. I'll work on those and send
a v2.

Thanks and regards,
Ojaswin
> 
> > +
> > +exit
> > diff --git a/tests/ext4/056.out b/tests/ext4/056.out
> > new file mode 100644
> > index 00000000..41706284
> > --- /dev/null
> > +++ b/tests/ext4/056.out
> > @@ -0,0 +1,2 @@
> > +QA output created by 056
> > +Test Succeeded!
> > -- 
> > 2.27.0