fstests: new testcase to verify show_devname.
cover fstests: new testcase to verify show_devname.
Mismatch causes mount -a (UUID) to fail with -EBUSY in libmount 2.37.4. In libmount 2.40.2 with Btrfs, it shows 'successfully mounted' instead of 'already mounted', kernel without fix.
Anand Jain (2):
fstests: new test case to verify show_devname() device path
fstests: btrfs: new test: show_devname() on multi-device volumes
tests/btrfs/329 | 107 ++++++++++++++++++++++++++++++++++++++++++ tests/btrfs/329.out | 19 ++++++++ tests/generic/765 | 74 +++++++++++++++++++++++++++++ tests/generic/765.out | 6 +++ 4 files changed, 206 insertions(+) create mode 100755 tests/btrfs/329 create mode 100644 tests/btrfs/329.out create mode 100755 tests/generic/765 create mode 100644 tests/generic/765.out
Signature · 1 line
2.49.0
PATCH 1/2 fstests: new test case to verify show_devname() device path
If the device path under /proc/self/mounts does not match the path shown by findmnt --uuid, commands like mount -a may fail to recognize that the device is already mounted and attempt to mount it again. This test case verifies that the two paths match.
--- tests/generic/765 | 74 +++++++++++++++++++++++++++++++++++++++++++ tests/generic/765.out | 6 ++++ 2 files changed, 80 insertions(+) create mode 100755 tests/generic/765 create mode 100644 tests/generic/765.out
tests/generic/765 new file +74 -0
@@ -0,0 +1,74 @@ +#! /bin/bash +# SPDX-License-Identifier: GPL-2.0 +# Copyright (c) 2025 Oracle. All Rights Reserved. +# +# FS QA Test 765 +# +# Test if the device paths are compatible with systemd udev tools like +# mount -a and findmnt. +# Provide the device dm path to the mount command and verify that it is +# correctly resolved to the mapper path in the kernel. +# +. ./common/preamble +_begin_fstest auto + +node="fstests_${seq}" + +_cleanup() +{ + cd / + rm -r -f $tmp.* + + sed -i "\|[[:space:]]$SCRATCH_MNT[[:space:]]|d" /etc/fstab + _systemd_installed && _systemd_reload + + _unmount $SCRATCH_MNT > /dev/null 2>&1 + udevadm control --start-exec-queue + _dmsetup_remove $node +} + +. ./common/systemd +. ./common/filter + +_require_scratch +_require_block_device $SCRATCH_DEV +_require_dm_target linear + +[ "$FSTYP" = btrfs ] && _fixed_by_kernel_commit XXXXXXXXXXXX \ + "btrfs: pass device path update from mount thread" + +filter_mapper() +{ + sed -e "s,\B$scratch_mapper_dev,MAPPER_PATH,g" |\ + sed -e "s,/dev/dm-[^ ]*,DM_PATH,g" |\ + _filter_scratch +} + +#size doesn't matter just create it to the device size +sectors=$(blockdev --getsz $SCRATCH_DEV) +table="0 $sectors linear $SCRATCH_DEV 0" +_dmsetup_create $node --table "$table" || _fail "setup dm device failed" + +scratch_mapper_dev=/dev/mapper/$node +scratch_dm_dev=$(realpath ${scratch_mapper_dev}) + +# Block external triggers that alter the device path inside the kernel, +# they are unreliable. +udevadm control --stop-exec-queue +_mkfs_dev $scratch_dm_dev + +# mount resolves dm path to its mapper path +_mount --verbose $scratch_dm_dev $SCRATCH_MNT | filter_mapper + +fsid=$(findmnt -n -o UUID ${scratch_dm_dev}) +blkid --uuid ${fsid} | filter_mapper +findmnt --source UUID=${fsid} --noheadings --output SOURCE | filter_mapper +cat /proc/self/mounts | grep ${SCRATCH_MNT} | \ + $AWK_PROG '{print $1" "$2}' | filter_mapper + +echo "UUID=${fsid} ${SCRATCH_MNT} $FSTYP defaults 0 0" >> /etc/fstab +_systemd_installed && _systemd_reload +_mount --verbose -a | grep $SCRATCH_MNT | filter_mapper + +status=0 +exit
tests/generic/765.out new file +6 -0
@@ -0,0 +1,6 @@ +QA output created by 765 +mount: MAPPER_PATH mounted on SCRATCH_MNT. +MAPPER_PATH +MAPPER_PATH +MAPPER_PATH SCRATCH_MNT +SCRATCH_MNT : already mounted
Signature · 1 line
2.49.0
PATCH 2/2 fstests: btrfs: new test: show_devname() on multi-device volumes
In reply to Anand JainIf the device path under /proc/self/mounts does not match the path shown by findmnt --uuid, commands like mount -a may fail to recognize that the device is already mounted and attempt to mount it again. This test case verifies that the two paths match.
--- tests/btrfs/329 | 107 ++++++++++++++++++++++++++++++++++++++++++++ tests/btrfs/329.out | 19 ++++++++ 2 files changed, 126 insertions(+) create mode 100755 tests/btrfs/329 create mode 100644 tests/btrfs/329.out
tests/btrfs/329 new file +107 -0
@@ -0,0 +1,107 @@ +#! /bin/bash +# SPDX-License-Identifier: GPL-2.0 +# Copyright (c) 2025 Oracle. All Rights Reserved. +# +# FS QA Test 329 +# +# Test that show_devname() returns a device path that matches findmnt, +# ensuring that mount -a works correctly for multi-device Btrfs filesystems. + +. ./common/preamble +_begin_fstest auto volume + +node="fstests_${seq}" + +_cleanup() +{ + cd / + rm -r -f $tmp.* + + clean_per_iteration + + udevadm control --start-exec-queue + _dmsetup_remove $node +} + +. ./common/systemd +. ./common/filter + +_require_scratch_dev_pool 2 +_require_btrfs_forget_or_module_loadable +_require_block_device $SCRATCH_DEV +_require_dm_target linear + +_fixed_by_kernel_commit XXXXXXXXXXXX \ + "btrfs: pass device path update from mount thread" + +_scratch_dev_pool_get 3 +scratch_devs=($SCRATCH_DEV_POOL) +#size doesn't matter just create it to the device size +sectors=$(blockdev --getsz ${scratch_devs[0]}) +table="0 $sectors linear ${scratch_devs[0]} 0" +_dmsetup_create $node --table "$table" || _fail "setup dm device failed" +scratch_mapper_dev=/dev/mapper/$node +scratch_dm_dev=$(realpath ${scratch_mapper_dev}) + +clean_per_iteration() +{ + _unmount $SCRATCH_MNT > /dev/null 2>&1 + sed -i "\|[[:space:]]$SCRATCH_MNT[[:space:]]|d" /etc/fstab + _systemd_installed && _systemd_reload +} + +filter_mapper() +{ + sed -e "s,\B$scratch_mapper_dev,MAPPER_PATH,g" |\ + sed -e "s,/dev/dm-[^ ]*,DM_PATH,g" |\ + _filter_scratch_pool +} + +test_dev_path() +{ + clean_per_iteration + _btrfs_forget_or_module_reload + + echo --- ${devs[@]} ---- | filter_mapper + _mkfs_dev "${devs[0]} ${devs[1]}" + + # mount resolves dm path to its mapper path + _mount --verbose ${devs[0]} -o device=${devs[1]} $SCRATCH_MNT | \ + filter_mapper + + fsid=$(findmnt -n -o UUID ${devs[0]}) + blkid --uuid ${fsid} | filter_mapper + findmnt --source UUID=${fsid} --noheadings --output SOURCE | \ + filter_mapper + cat /proc/self/mounts | grep ${SCRATCH_MNT} | \ + $AWK_PROG '{print $1" "$2}' | filter_mapper + + echo "UUID=${fsid} ${SCRATCH_MNT} $FSTYP defaults 0 0" >> /etc/fstab + _systemd_installed && _systemd_reload + + # The error output from `mount -a --verbose` varies. + # In libmount 2.37.4, the `mount -a` command will fail with -EBUSY. + # In libmount 2.40.2, it will report "Successfully mounted" instead + # of "already mounted." + _mount --verbose -a | grep $SCRATCH_MNT | _filter_scratch +} + +# Block external triggers such as (64-btrfs-dm.rules) that alter the device path +# inside the kernel, they are unreliable. +udevadm control --stop-exec-queue +secnarios=( + "$scratch_dm_dev ${scratch_devs[1]}" + "${scratch_devs[2]} ${scratch_devs[1]}" + "${scratch_devs[1]} ${scratch_devs[2]}" +) + +devs=() +for devs_str in "${secnarios[@]}"; do + devs=($devs_str) + test_dev_path +done +udevadm control --start-exec-queue + +_scratch_dev_pool_put +status=0 +exit
tests/btrfs/329.out new file +19 -0
@@ -0,0 +1,19 @@ +QA output created by 329 +--- DM_PATH SCRATCH_DEV ---- +mount: MAPPER_PATH mounted on /mnt/scratch. +MAPPER_PATH +MAPPER_PATH +MAPPER_PATH /mnt/scratch +SCRATCH_MNT : already mounted +--- SCRATCH_DEV SCRATCH_DEV ---- +mount: SCRATCH_DEV mounted on /mnt/scratch. +SCRATCH_DEV +SCRATCH_DEV +SCRATCH_DEV /mnt/scratch +SCRATCH_MNT : already mounted +--- SCRATCH_DEV SCRATCH_DEV ---- +mount: SCRATCH_DEV mounted on /mnt/scratch. +SCRATCH_DEV +SCRATCH_DEV +SCRATCH_DEV /mnt/scratch +SCRATCH_MNT : already mounted
Signature · 1 line
2.49.0