Re: [PATCH 10/11] xfs: Check for extent overflow when swapping extents

[Chandan Babu R <chandanrlinux@gmail.com>]

On Saturday 14 November 2020 5:38:35 AM IST Darrick J. Wong wrote:
> On Fri, Nov 13, 2020 at 04:57:02PM +0530, Chandan Babu R wrote:
> > This test verifies that XFS does not cause inode fork's extent count to
> > overflow when swapping forks across two files.
> > 
> > Signed-off-by: Chandan Babu R <chandanrlinux@gmail.com>
> > ---
> >  tests/xfs/530     | 115 ++++++++++++++++++++++++++++++++++++++++++++++
> >  tests/xfs/530.out |  13 ++++++
> >  tests/xfs/group   |   1 +
> >  3 files changed, 129 insertions(+)
> >  create mode 100755 tests/xfs/530
> >  create mode 100644 tests/xfs/530.out
> > 
> > diff --git a/tests/xfs/530 b/tests/xfs/530
> > new file mode 100755
> > index 00000000..fccc6de7
> > --- /dev/null
> > +++ b/tests/xfs/530
> > @@ -0,0 +1,115 @@
> > +#! /bin/bash
> > +# SPDX-License-Identifier: GPL-2.0
> > +# Copyright (c) 2020 Chandan Babu R.  All Rights Reserved.
> > +#
> > +# FS QA Test 530
> > +#
> > +# Verify that XFS does not cause inode fork's extent count to overflow when
> > +# swapping forks between files
> > +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
> > +. ./common/inject
> > +
> > +# remove previous $seqres.full before test
> > +rm -f $seqres.full
> > +
> > +# real QA test starts here
> > +
> > +_supported_fs xfs
> > +_require_scratch
> > +_require_xfs_debug
> > +_require_xfs_scratch_rmapbt
> > +_require_xfs_io_command "fcollapse"
> > +_require_xfs_io_command "swapext"
> 
> FWIW it's going to be a while before the swapext command goes upstream.
> Right now it's a part of the atomic file range exchange patchset.

Sorry, I didn't understand your statement. I have the following from my Linux
machine,

root@debian-guest:~# /sbin/xfs_io 
xfs_io> help swapext
swapext <donorfile> -- Swap extents between files.

 Swaps extents between the open file descriptor and the supplied filename.

The above command causes the following code path to be invoked inside the
kernel (assuming rmapbt feature is enabled),
xfs_ioc_swapext() => xfs_swap_extents() => xfs_swap_extent_rmap().

> 
> Do you want me to try to speed that up?
> 
> --D
> 
> > +_require_xfs_io_error_injection "reduce_max_iextents"
> > +
> > +echo "* Swap extent forks"
> > +
> > +echo "Format and mount fs"
> > +_scratch_mkfs >> $seqres.full
> > +_scratch_mount >> $seqres.full
> > +
> > +bsize=$(_get_block_size $SCRATCH_MNT)
> > +
> > +srcfile=${SCRATCH_MNT}/srcfile
> > +donorfile=${SCRATCH_MNT}/donorfile
> > +
> > +echo "Create \$donorfile having an extent of length 17 blocks"
> > +xfs_io -f -c "pwrite -b $((17 * bsize)) 0 $((17 * bsize))" -c fsync $donorfile \
> > +       >> $seqres.full
> > +
> > +# After the for loop the donor file will have the following extent layout
> > +# | 0-4 | 5 | 6 | 7 | 8 | 9 | 10 |
> > +echo "Fragment \$donorfile"
> > +for i in $(seq 5 10); do
> > +	start_offset=$((i * bsize))
> > +	xfs_io -f -c "fcollapse $start_offset $bsize" $donorfile >> $seqres.full
> > +done
> > +donorino=$(stat -c "%i" $donorfile)
> > +
> > +echo "Create \$srcfile having an extent of length 18 blocks"
> > +xfs_io -f -c "pwrite -b $((18 * bsize)) 0 $((18 * bsize))" -c fsync $srcfile \
> > +       >> $seqres.full
> > +
> > +echo "Fragment \$srcfile"
> > +# After the for loop the src file will have the following extent layout
> > +# | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7-10 |
> > +for i in $(seq 1 7); do
> > +	start_offset=$((i * bsize))
> > +	xfs_io -f -c "fcollapse $start_offset $bsize" $srcfile >> $seqres.full
> > +done
> > +srcino=$(stat -c "%i" $srcfile)
> > +
> > +_scratch_unmount >> $seqres.full
> > +
> > +echo "Collect \$donorfile's extent count"
> > +donor_nr_exts=$(_scratch_get_iext_count $donorino data || \
> > +		_fail "Unable to obtain inode fork's extent count")
> > +
> > +echo "Collect \$srcfile's extent count"
> > +src_nr_exts=$(_scratch_get_iext_count $srcino data || \
> > +		_fail "Unable to obtain inode fork's extent count")
> > +
> > +_scratch_mount >> $seqres.full
> > +
> > +echo "Inject reduce_max_iextents error tag"
> > +xfs_io -x -c 'inject reduce_max_iextents' $SCRATCH_MNT
> > +
> > +echo "Swap \$srcfile's and \$donorfile's extent forks"
> > +xfs_io -f -c "swapext $donorfile" $srcfile >> $seqres.full 2>&1
> > +
> > +_scratch_unmount >> $seqres.full
> > +
> > +echo "Check for \$donorfile's extent count overflow"
> > +nextents=$(_scratch_get_iext_count $donorino data || \
> > +		_fail "Unable to obtain inode fork's extent count")
> > +if (( $nextents == $src_nr_exts )); then
> > +	echo "\$donorfile: Extent count overflow check failed"
> > +fi
> > +
> > +echo "Check for \$srcfile's extent count overflow"
> > +nextents=$(_scratch_get_iext_count $srcino data || \
> > +		_fail "Unable to obtain inode fork's extent count")
> > +if (( $nextents == $donor_nr_exts )); then
> > +	echo "\$srcfile: Extent count overflow check failed"
> > +fi
> > +
> > +# success, all done
> > +status=0
> > +exit
> > diff --git a/tests/xfs/530.out b/tests/xfs/530.out
> > new file mode 100644
> > index 00000000..996af959
> > --- /dev/null
> > +++ b/tests/xfs/530.out
> > @@ -0,0 +1,13 @@
> > +QA output created by 530
> > +* Swap extent forks
> > +Format and mount fs
> > +Create $donorfile having an extent of length 17 blocks
> > +Fragment $donorfile
> > +Create $srcfile having an extent of length 18 blocks
> > +Fragment $srcfile
> > +Collect $donorfile's extent count
> > +Collect $srcfile's extent count
> > +Inject reduce_max_iextents error tag
> > +Swap $srcfile's and $donorfile's extent forks
> > +Check for $donorfile's extent count overflow
> > +Check for $srcfile's extent count overflow
> > diff --git a/tests/xfs/group b/tests/xfs/group
> > index bc3958b3..81a15582 100644
> > --- a/tests/xfs/group
> > +++ b/tests/xfs/group
> > @@ -527,3 +527,4 @@
> >  527 auto quick
> >  528 auto quick reflink
> >  529 auto quick reflink
> > +530 auto quick
> 


-- 
chandan