| Message ID | c9fa8fa558e307a5d0d28545ff69433ae8324f4c.1705964751.git.boris@bur.io (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | btrfs/304: test qgroup deletion | expand |
Reviewed-by: David Disseldorp <ddiss@suse.de> A few minor nits below which should be addressed prior to merge... On Mon, 22 Jan 2024 15:06:28 -0800, Boris Burkov wrote: > When using squotas, an extent's OWNER_REF can long outlive the subvolume > that is the owner, since it could pick up a different reference that > keeps it around, but the subvolume can go away. > > Test this case, as originally, it resulted in a read only btrfs. > > Since we can blow up the subvolume in the same transaction as the extent > is written, we can also increment the usage of a non-existent subvolume. > > This leaves an OWNER_REF behind with no corresponding incremented usage > in a qgroup, so if we re-create that qgroup, we can then underflow its > usage. > > Both of these cases are fixed in the kernel by disallowing > creating subvol qgroups and by disallowing deleting qgroups that still > have usage. > > Signed-off-by: Boris Burkov <boris@bur.io> > --- > common/btrfs | 10 +++++ > tests/btrfs/301 | 14 +------ > tests/btrfs/304 | 90 +++++++++++++++++++++++++++++++++++++++++++++ > tests/btrfs/304.out | 6 +++ btrfs/304 is already taken in fstests v2024.01.14 by 9d812702 ("btrfs: add fstest for stripe-tree metadata with 4k write"). > 4 files changed, 108 insertions(+), 12 deletions(-) > create mode 100755 tests/btrfs/304 > create mode 100644 tests/btrfs/304.out > > diff --git a/common/btrfs b/common/btrfs > index f91f8dd86..c8593c1f9 100644 > --- a/common/btrfs > +++ b/common/btrfs > @@ -775,3 +775,13 @@ _has_btrfs_sysfs_feature_attr() > > test -e /sys/fs/btrfs/features/$feature_attr > } > + > +_enable_quota() > +{ > + local mode=$1 > + > + [ $mode == "n" ] && return > + arg=$([ $mode == "s" ] && echo "--simple") > + > + $BTRFS_UTIL_PROG quota enable $arg $SCRATCH_MNT It looks as though the "n" mode isn't used, and the "s" -> "--simple" mapping is confusing. Can we instead just make this: $BTRFS_UTIL_PROG quota enable $* $SCRATCH_MNT or drop the helper function altogether? > +} > diff --git a/tests/btrfs/301 b/tests/btrfs/301 > index db4697247..b3ee66cd9 100755 > --- a/tests/btrfs/301 > +++ b/tests/btrfs/301 > @@ -157,16 +157,6 @@ do_enospc_falloc() > do_falloc $file $sz > } > > -enable_quota() > -{ > - local mode=$1 > - > - [ $mode == "n" ] && return > - arg=$([ $mode == "s" ] && echo "--simple") > - > - $BTRFS_UTIL_PROG quota enable $arg $SCRATCH_MNT > -} > - > get_subvid() > { > _btrfs_get_subvolid $SCRATCH_MNT subv > @@ -186,7 +176,7 @@ prepare() > { > _scratch_mkfs >> $seqres.full > _scratch_mount > - enable_quota "s" > + _enable_quota "s" ...as mentioned, _enable_quota --simple (or inline $BTRFS_UTIL_PROG ...) > $BTRFS_UTIL_PROG subvolume create $subv >> $seqres.full > local subvid=$(get_subvid) > set_subvol_limit $subvid $limit > @@ -397,7 +387,7 @@ enable_mature() > # Sync before enabling squotas to reliably *not* count the writes > # we did before enabling. > sync > - enable_quota "s" > + _enable_quota "s" > set_subvol_limit $subvid $limit > _scratch_cycle_mount > usage=$(get_subvol_usage $subvid) > diff --git a/tests/btrfs/304 b/tests/btrfs/304 > new file mode 100755 > index 000000000..3fce0591c > --- /dev/null > +++ b/tests/btrfs/304 > @@ -0,0 +1,90 @@ > +#! /bin/bash > +# SPDX-License-Identifier: GPL-2.0 > +# Copyright (c) 2024 Meta Platforms, Inc. All Rights Reserved. > +# > +# FS QA Test 304 > +# > +# Test various race conditions between qgroup deletion and squota writes > +# > +. ./common/preamble > +_begin_fstest auto quick qgroup subvol clone > + > +# Override the default cleanup function. > +# _cleanup() > +# { > +# cd / > +# rm -r -f $tmp.* > +# } nit: doesn't look like there's a need to override the default.
diff --git a/common/btrfs b/common/btrfs index f91f8dd86..c8593c1f9 100644 --- a/common/btrfs +++ b/common/btrfs @@ -775,3 +775,13 @@ _has_btrfs_sysfs_feature_attr() test -e /sys/fs/btrfs/features/$feature_attr } + +_enable_quota() +{ + local mode=$1 + + [ $mode == "n" ] && return + arg=$([ $mode == "s" ] && echo "--simple") + + $BTRFS_UTIL_PROG quota enable $arg $SCRATCH_MNT +} diff --git a/tests/btrfs/301 b/tests/btrfs/301 index db4697247..b3ee66cd9 100755 --- a/tests/btrfs/301 +++ b/tests/btrfs/301 @@ -157,16 +157,6 @@ do_enospc_falloc() do_falloc $file $sz } -enable_quota() -{ - local mode=$1 - - [ $mode == "n" ] && return - arg=$([ $mode == "s" ] && echo "--simple") - - $BTRFS_UTIL_PROG quota enable $arg $SCRATCH_MNT -} - get_subvid() { _btrfs_get_subvolid $SCRATCH_MNT subv @@ -186,7 +176,7 @@ prepare() { _scratch_mkfs >> $seqres.full _scratch_mount - enable_quota "s" + _enable_quota "s" $BTRFS_UTIL_PROG subvolume create $subv >> $seqres.full local subvid=$(get_subvid) set_subvol_limit $subvid $limit @@ -397,7 +387,7 @@ enable_mature() # Sync before enabling squotas to reliably *not* count the writes # we did before enabling. sync - enable_quota "s" + _enable_quota "s" set_subvol_limit $subvid $limit _scratch_cycle_mount usage=$(get_subvol_usage $subvid) diff --git a/tests/btrfs/304 b/tests/btrfs/304 new file mode 100755 index 000000000..3fce0591c --- /dev/null +++ b/tests/btrfs/304 @@ -0,0 +1,90 @@ +#! /bin/bash +# SPDX-License-Identifier: GPL-2.0 +# Copyright (c) 2024 Meta Platforms, Inc. All Rights Reserved. +# +# FS QA Test 304 +# +# Test various race conditions between qgroup deletion and squota writes +# +. ./common/preamble +_begin_fstest auto quick qgroup subvol clone + +# Override the default cleanup function. +# _cleanup() +# { +# cd / +# rm -r -f $tmp.* +# } + +# Import common functions. +. ./common/reflink + +# real QA test starts here + +# Modify as appropriate. +_supported_fs btrfs +_require_scratch_reflink +_require_cp_reflink +_require_scratch_enable_simple_quota +_require_no_compress + +_fixed_by_kernel_commit xxxxxxxxxxxx "btrfs: forbid deleting live subvol qgroup" +_fixed_by_kernel_commit xxxxxxxxxxxx "btrfs: forbid creating subvol qgroups" + +subv1=$SCRATCH_MNT/subv1 +subv2=$SCRATCH_MNT/subv2 + +prepare() +{ + _scratch_mkfs >> $seqres.full + _scratch_mount + _enable_quota "s" + $BTRFS_UTIL_PROG subvolume create $subv1 >> $seqres.full + $BTRFS_UTIL_PROG subvolume create $subv2 >> $seqres.full + $XFS_IO_PROG -fc "pwrite -q 0 128K" $subv1/f + _cp_reflink $subv1/f $subv2/f +} + +# An extent can long outlive its owner. Test this by deleting the owning +# subvolume, committing the transaction, then deleting the reflinked copy. +# Deleting the copy will attempt to free space from the missing owner, which +# should be a no-op. +free_from_deleted_owner() +{ + echo "free from deleted owner" + prepare + subvid1=$(_btrfs_get_subvolid $SCRATCH_MNT subv1) + + $BTRFS_UTIL_PROG filesystem sync $SCRATCH_MNT + $BTRFS_UTIL_PROG subvolume delete $subv1 >> $seqres.full + $BTRFS_UTIL_PROG qgroup destroy 0/$subvid1 $SCRATCH_MNT >> $seqres.full + $BTRFS_UTIL_PROG filesystem sync $SCRATCH_MNT + rm $subv2/f + _scratch_unmount +} + +# A race where we delete the owner in the same transaction as writing the +# extent leads to incrementing the squota usage of the missing qgroup. +# This leaves behind an owner ref with an owner id that cannot exist, so +# freeing the extent now frees from that qgroup, but there has never +# been a corresponding usage to free. +add_to_deleted_owner() +{ + echo "add to deleted owner" + prepare + subvid1=$(_btrfs_get_subvolid $SCRATCH_MNT subv1) + + $BTRFS_UTIL_PROG subvolume delete $subv1 >> $seqres.full + $BTRFS_UTIL_PROG qgroup destroy 0/$subvid1 $SCRATCH_MNT >> $seqres.full + $BTRFS_UTIL_PROG filesystem sync $SCRATCH_MNT + $BTRFS_UTIL_PROG qgroup create 0/$subvid1 $SCRATCH_MNT >> $seqres.full + rm $subv2/f + _scratch_unmount +} + +free_from_deleted_owner +add_to_deleted_owner + +# success, all done +status=0 +exit diff --git a/tests/btrfs/304.out b/tests/btrfs/304.out new file mode 100644 index 000000000..55bbc64f5 --- /dev/null +++ b/tests/btrfs/304.out @@ -0,0 +1,6 @@ +QA output created by 304 +free from deleted owner +ERROR: unable to destroy quota group: Device or resource busy +add to deleted owner +ERROR: unable to destroy quota group: Device or resource busy +ERROR: unable to create quota group: Invalid argument
When using squotas, an extent's OWNER_REF can long outlive the subvolume that is the owner, since it could pick up a different reference that keeps it around, but the subvolume can go away. Test this case, as originally, it resulted in a read only btrfs. Since we can blow up the subvolume in the same transaction as the extent is written, we can also increment the usage of a non-existent subvolume. This leaves an OWNER_REF behind with no corresponding incremented usage in a qgroup, so if we re-create that qgroup, we can then underflow its usage. Both of these cases are fixed in the kernel by disallowing creating subvol qgroups and by disallowing deleting qgroups that still have usage. Signed-off-by: Boris Burkov <boris@bur.io> --- common/btrfs | 10 +++++ tests/btrfs/301 | 14 +------ tests/btrfs/304 | 90 +++++++++++++++++++++++++++++++++++++++++++++ tests/btrfs/304.out | 6 +++ 4 files changed, 108 insertions(+), 12 deletions(-) create mode 100755 tests/btrfs/304 create mode 100644 tests/btrfs/304.out