| Message ID | 20220609214223.4608-10-pvorel@suse.cz |
|---|---|
| State | Changes Requested |
| Headers | show |
| Series | shell: nfs: $TST_ALL_FILESYSTEMS (.all_filesystems) | expand |
On Fri, Jun 10, 2022 at 5:44 AM Petr Vorel <pvorel@suse.cz> wrote: > Signed-off-by: Petr Vorel <pvorel@suse.cz> > --- > Hi NFS developers, > > your comments are welcome. This is an effort how to support NFS tests on > all filesystems available on SUT. Using $TST_MNTPOINT means test run in > loop, each time different filesystem is used. It's implemented via > formatted loop device (the same way as in LTP C API). > > Code is also at: > https://github.com/pevik/ltp/tree/shell/all_filesystems.v2 > > But this requires restarting NFS server (otherwise legacy > testcases/lib/daemonlib.sh > would have to be used), which is IMHO not optimal. > > Without that (or only run 'systemctl reload nfs-server' or exportfs -r > on remote side) it cannot be umounted after testing - debugging with > fuser, code at: > > https://github.com/pevik/ltp/commit/3656d035d43445a107154ef397ef1db2fad2c4f0 > > The problem is that loop device is still referenced by nfs server and > thus cannot be unmounted. Can this be somehow fixed? Is it even wanted > to have tests on loop device (the only reasonable way to support more > filesystems)? Also tests will run much longer (we could filter out some > filesystems not supported, if there are any). > > # LTP_SINGLE_FS_TYPE=ext2 PATH="/opt/ltp/testcases/bin:$PATH" nfs07.sh -v > 3 -t tcp -i3 > ## NOTE: testing itself is OK > nfs07 1 TINFO: initialize 'lhost' 'ltp_ns_veth2' interface > nfs07 1 TINFO: add local addr 10.0.0.2/24 > nfs07 1 TINFO: add local addr fd00:1:1:1::2/64 > nfs07 1 TINFO: initialize 'rhost' 'ltp_ns_veth1' interface > nfs07 1 TINFO: add remote addr 10.0.0.1/24 > nfs07 1 TINFO: add remote addr fd00:1:1:1::1/64 > nfs07 1 TINFO: Network config (local -- remote): > nfs07 1 TINFO: ltp_ns_veth2 -- ltp_ns_veth1 > nfs07 1 TINFO: 10.0.0.2/24 -- 10.0.0.1/24 > nfs07 1 TINFO: fd00:1:1:1::2/64 -- fd00:1:1:1::1/64 > tst_device.c:89: TINFO: Found free device 0 '/dev/loop0' > tst_supported_fs_types.c:148: TINFO: WARNING: testing only ext2 > tst_supported_fs_types.c:89: TINFO: Kernel supports ext2 > tst_supported_fs_types.c:51: TINFO: mkfs.ext2 does exist > nfs07 1 TINFO: Testing on ext2 > nfs07 1 TINFO: Formatting ext2 with opts='/dev/loop0' > nfs07 1 TINFO: timeout per run is 0h 5m 0s > nfs07 1 TINFO: mount.nfs: (linux nfs-utils 2.6.1) > nfs07 1 TINFO: setup NFSv3, socket type tcp > nfs07 1 TINFO: Mounting NFS: mount -v -t nfs -o proto=tcp,vers=3 10.0.0.2:/tmp/LTP_nfs07.A3PIB82iUv/mntpoint/3/tcp > /tmp/LTP_nfs07.A3PIB82iUv/mntpoint/3/0 > nfs07 1 TPASS: All files and directories were correctly listed > nfs07 2 TPASS: All files and directories were correctly listed > nfs07 3 TPASS: All files and directories were correctly listed > nfs07 4 TINFO: Cleaning up testcase > > ## DEBUGGING CODE in nfs_cleanup() > nfs07 4 TINFO: fuser -mv /tmp/LTP_nfs07.P1XS9smc5w ($TST_TMPDIR) > USER PID ACCESS COMMAND > /tmp/LTP_nfs07.P1XS9smc5w: > root kernel mount /tmp > root 2125 ..c.. tst_timeout_kil > nfs07 4 TINFO: fuser -mv /tmp/LTP_nfs07.P1XS9smc5w/mntpoint ($TST_MNTPOINT) > USER PID ACCESS COMMAND > /tmp/LTP_nfs07.P1XS9smc5w/mntpoint: > root kernel mount > /tmp/LTP_nfs07.P1XS9smc5w/mntpoint > nfs07 4 TINFO: fuser -mv /tmp/LTP_nfs07.P1XS9smc5w/mntpoint/3/0 > USER PID ACCESS COMMAND > /tmp/LTP_nfs07.P1XS9smc5w/mntpoint/3/0: > root kernel mount > /tmp/LTP_nfs07.P1XS9smc5w/mntpoint/3/0 > > ## from nfs_cleanup() > ## grep -q "$local_dir" /proc/mounts && umount $local_dir > nfs07 4 TINFO: umount /tmp/LTP_nfs07.P1XS9smc5w/mntpoint/3/0 > umount: /tmp/LTP_nfs07.P1XS9smc5w/mntpoint: target is busy. > Maybe try with `umount -fl`? unmount manual says: -f, --force force unmount (in case of an unreachable NFS system) -l, --lazy detach the filesystem now, clean up things later
> On Fri, Jun 10, 2022 at 5:44 AM Petr Vorel <pvorel@suse.cz> wrote: > > Signed-off-by: Petr Vorel <pvorel@suse.cz> > > --- > > Hi NFS developers, > > your comments are welcome. This is an effort how to support NFS tests on > > all filesystems available on SUT. Using $TST_MNTPOINT means test run in > > loop, each time different filesystem is used. It's implemented via > > formatted loop device (the same way as in LTP C API). > > Code is also at: > > https://github.com/pevik/ltp/tree/shell/all_filesystems.v2 > > But this requires restarting NFS server (otherwise legacy > > testcases/lib/daemonlib.sh > > would have to be used), which is IMHO not optimal. > > Without that (or only run 'systemctl reload nfs-server' or exportfs -r > > on remote side) it cannot be umounted after testing - debugging with > > fuser, code at: > > https://github.com/pevik/ltp/commit/3656d035d43445a107154ef397ef1db2fad2c4f0 > > The problem is that loop device is still referenced by nfs server and > > thus cannot be unmounted. Can this be somehow fixed? Is it even wanted > > to have tests on loop device (the only reasonable way to support more > > filesystems)? Also tests will run much longer (we could filter out some > > filesystems not supported, if there are any). > > # LTP_SINGLE_FS_TYPE=ext2 PATH="/opt/ltp/testcases/bin:$PATH" nfs07.sh -v > > 3 -t tcp -i3 > > ## NOTE: testing itself is OK > > nfs07 1 TINFO: initialize 'lhost' 'ltp_ns_veth2' interface > > nfs07 1 TINFO: add local addr 10.0.0.2/24 > > nfs07 1 TINFO: add local addr fd00:1:1:1::2/64 > > nfs07 1 TINFO: initialize 'rhost' 'ltp_ns_veth1' interface > > nfs07 1 TINFO: add remote addr 10.0.0.1/24 > > nfs07 1 TINFO: add remote addr fd00:1:1:1::1/64 > > nfs07 1 TINFO: Network config (local -- remote): > > nfs07 1 TINFO: ltp_ns_veth2 -- ltp_ns_veth1 > > nfs07 1 TINFO: 10.0.0.2/24 -- 10.0.0.1/24 > > nfs07 1 TINFO: fd00:1:1:1::2/64 -- fd00:1:1:1::1/64 > > tst_device.c:89: TINFO: Found free device 0 '/dev/loop0' > > tst_supported_fs_types.c:148: TINFO: WARNING: testing only ext2 > > tst_supported_fs_types.c:89: TINFO: Kernel supports ext2 > > tst_supported_fs_types.c:51: TINFO: mkfs.ext2 does exist > > nfs07 1 TINFO: Testing on ext2 > > nfs07 1 TINFO: Formatting ext2 with opts='/dev/loop0' > > nfs07 1 TINFO: timeout per run is 0h 5m 0s > > nfs07 1 TINFO: mount.nfs: (linux nfs-utils 2.6.1) > > nfs07 1 TINFO: setup NFSv3, socket type tcp > > nfs07 1 TINFO: Mounting NFS: mount -v -t nfs -o proto=tcp,vers=3 10.0.0.2:/tmp/LTP_nfs07.A3PIB82iUv/mntpoint/3/tcp > > /tmp/LTP_nfs07.A3PIB82iUv/mntpoint/3/0 > > nfs07 1 TPASS: All files and directories were correctly listed > > nfs07 2 TPASS: All files and directories were correctly listed > > nfs07 3 TPASS: All files and directories were correctly listed > > nfs07 4 TINFO: Cleaning up testcase > > ## DEBUGGING CODE in nfs_cleanup() > > nfs07 4 TINFO: fuser -mv /tmp/LTP_nfs07.P1XS9smc5w ($TST_TMPDIR) > > USER PID ACCESS COMMAND > > /tmp/LTP_nfs07.P1XS9smc5w: > > root kernel mount /tmp > > root 2125 ..c.. tst_timeout_kil > > nfs07 4 TINFO: fuser -mv /tmp/LTP_nfs07.P1XS9smc5w/mntpoint ($TST_MNTPOINT) > > USER PID ACCESS COMMAND > > /tmp/LTP_nfs07.P1XS9smc5w/mntpoint: > > root kernel mount > > /tmp/LTP_nfs07.P1XS9smc5w/mntpoint > > nfs07 4 TINFO: fuser -mv /tmp/LTP_nfs07.P1XS9smc5w/mntpoint/3/0 > > USER PID ACCESS COMMAND > > /tmp/LTP_nfs07.P1XS9smc5w/mntpoint/3/0: > > root kernel mount > > /tmp/LTP_nfs07.P1XS9smc5w/mntpoint/3/0 > > ## from nfs_cleanup() > > ## grep -q "$local_dir" /proc/mounts && umount $local_dir > > nfs07 4 TINFO: umount /tmp/LTP_nfs07.P1XS9smc5w/mntpoint/3/0 > > umount: /tmp/LTP_nfs07.P1XS9smc5w/mntpoint: target is busy. > Maybe try with `umount -fl`? Thank you! Unfortunately it does not work (I tried this before ending up with restarting server because even -fl did not help). I also consider the need of 'umount -fl' as kind of error just less aggressive then restarting server. But if this is considered ok, why not. Kind regards, Petr > unmount manual says: > -f, --force force unmount (in case of an unreachable NFS system) > -l, --lazy detach the filesystem now, clean up things later
Hi! > + > + systemctl restart nfs-server I wonder do we stil have to care for systems without systemd? We do have daemonlib.sh in LTP that has restart_daemon() function that branches to systemctl|service|/etc/init.d/ that is used by a few tests.
Hi all, > > + systemctl restart nfs-server > I wonder do we stil have to care for systems without systemd? That'd be a question for embedded people. I myself wonder if they all moved to systemd. But at least Alpine and Buildroot distros allow to use Busybox as init, thus I'd still care. > We do have daemonlib.sh in LTP that has restart_daemon() function that > branches to systemctl|service|/etc/init.d/ that is used by a few tests. That's exactly what I'm planning to use. Kind regards, Petr
diff --git a/testcases/network/nfs/nfs_stress/nfs_lib.sh b/testcases/network/nfs/nfs_stress/nfs_lib.sh index af7d46a21..a6557177b 100644 --- a/testcases/network/nfs/nfs_stress/nfs_lib.sh +++ b/testcases/network/nfs/nfs_stress/nfs_lib.sh @@ -28,7 +28,7 @@ NFS_PARSE_ARGS_CALLER="$TST_PARSE_ARGS" TST_OPTS="v:t:$TST_OPTS" TST_PARSE_ARGS=nfs_parse_args TST_USAGE=nfs_usage -TST_NEEDS_TMPDIR=1 +TST_ALL_FILESYSTEMS=1 TST_NEEDS_ROOT=1 TST_NEEDS_CMDS="$TST_NEEDS_CMDS mount exportfs mount.nfs" TST_SETUP="${TST_SETUP:-nfs_setup}" @@ -63,7 +63,7 @@ nfs_get_remote_path() done v=${1:-$v} - echo "$TST_TMPDIR/$v/$type" + echo "$TST_MNTPOINT/$v/$type" } nfs_server_udp_enabled() @@ -162,8 +162,8 @@ nfs_setup() tst_brk TCONF "UDP support disabled on NFS server" fi - local_dir="$TST_TMPDIR/$i/$n" - remote_dir="$TST_TMPDIR/$i/$type" + local_dir="$TST_MNTPOINT/$i/$n" + remote_dir="$TST_MNTPOINT/$i/$type" mkdir -p $local_dir nfs_setup_server $(($$ + n)) @@ -174,7 +174,7 @@ nfs_setup() done if [ "$n" -eq 1 ]; then - cd ${VERSION}/0 + cd $TST_MNTPOINT/$VERSION/0 fi } @@ -190,19 +190,22 @@ nfs_cleanup() local n=0 for i in $VERSION; do - local_dir="$TST_TMPDIR/$i/$n" - grep -q "$local_dir" /proc/mounts && umount $local_dir + type=$(get_socket_type $n) + remote_dir="$TST_MNTPOINT/$i/$type" + tst_rhost_run -c "test -d $remote_dir && exportfs -u *:$remote_dir" + tst_rhost_run -c "test -d $remote_dir && rm -rf $remote_dir" n=$(( n + 1 )) done n=0 for i in $VERSION; do - type=$(get_socket_type $n) - remote_dir="$TST_TMPDIR/$i/$type" - tst_rhost_run -c "test -d $remote_dir && exportfs -u *:$remote_dir" - tst_rhost_run -c "test -d $remote_dir && rm -rf $remote_dir" + local_dir="$TST_MNTPOINT/$i/$n" + + grep -q "$local_dir" /proc/mounts && umount $local_dir n=$(( n + 1 )) done + + systemctl restart nfs-server } . tst_net.sh
Signed-off-by: Petr Vorel <pvorel@suse.cz> --- Hi NFS developers, your comments are welcome. This is an effort how to support NFS tests on all filesystems available on SUT. Using $TST_MNTPOINT means test run in loop, each time different filesystem is used. It's implemented via formatted loop device (the same way as in LTP C API). Code is also at: https://github.com/pevik/ltp/tree/shell/all_filesystems.v2 But this requires restarting NFS server (otherwise legacy testcases/lib/daemonlib.sh would have to be used), which is IMHO not optimal. Without that (or only run 'systemctl reload nfs-server' or exportfs -r on remote side) it cannot be umounted after testing - debugging with fuser, code at: https://github.com/pevik/ltp/commit/3656d035d43445a107154ef397ef1db2fad2c4f0 The problem is that loop device is still referenced by nfs server and thus cannot be unmounted. Can this be somehow fixed? Is it even wanted to have tests on loop device (the only reasonable way to support more filesystems)? Also tests will run much longer (we could filter out some filesystems not supported, if there are any). # LTP_SINGLE_FS_TYPE=ext2 PATH="/opt/ltp/testcases/bin:$PATH" nfs07.sh -v 3 -t tcp -i3 ## NOTE: testing itself is OK nfs07 1 TINFO: initialize 'lhost' 'ltp_ns_veth2' interface nfs07 1 TINFO: add local addr 10.0.0.2/24 nfs07 1 TINFO: add local addr fd00:1:1:1::2/64 nfs07 1 TINFO: initialize 'rhost' 'ltp_ns_veth1' interface nfs07 1 TINFO: add remote addr 10.0.0.1/24 nfs07 1 TINFO: add remote addr fd00:1:1:1::1/64 nfs07 1 TINFO: Network config (local -- remote): nfs07 1 TINFO: ltp_ns_veth2 -- ltp_ns_veth1 nfs07 1 TINFO: 10.0.0.2/24 -- 10.0.0.1/24 nfs07 1 TINFO: fd00:1:1:1::2/64 -- fd00:1:1:1::1/64 tst_device.c:89: TINFO: Found free device 0 '/dev/loop0' tst_supported_fs_types.c:148: TINFO: WARNING: testing only ext2 tst_supported_fs_types.c:89: TINFO: Kernel supports ext2 tst_supported_fs_types.c:51: TINFO: mkfs.ext2 does exist nfs07 1 TINFO: Testing on ext2 nfs07 1 TINFO: Formatting ext2 with opts='/dev/loop0' nfs07 1 TINFO: timeout per run is 0h 5m 0s nfs07 1 TINFO: mount.nfs: (linux nfs-utils 2.6.1) nfs07 1 TINFO: setup NFSv3, socket type tcp nfs07 1 TINFO: Mounting NFS: mount -v -t nfs -o proto=tcp,vers=3 10.0.0.2:/tmp/LTP_nfs07.A3PIB82iUv/mntpoint/3/tcp /tmp/LTP_nfs07.A3PIB82iUv/mntpoint/3/0 nfs07 1 TPASS: All files and directories were correctly listed nfs07 2 TPASS: All files and directories were correctly listed nfs07 3 TPASS: All files and directories were correctly listed nfs07 4 TINFO: Cleaning up testcase ## DEBUGGING CODE in nfs_cleanup() nfs07 4 TINFO: fuser -mv /tmp/LTP_nfs07.P1XS9smc5w ($TST_TMPDIR) USER PID ACCESS COMMAND /tmp/LTP_nfs07.P1XS9smc5w: root kernel mount /tmp root 2125 ..c.. tst_timeout_kil nfs07 4 TINFO: fuser -mv /tmp/LTP_nfs07.P1XS9smc5w/mntpoint ($TST_MNTPOINT) USER PID ACCESS COMMAND /tmp/LTP_nfs07.P1XS9smc5w/mntpoint: root kernel mount /tmp/LTP_nfs07.P1XS9smc5w/mntpoint nfs07 4 TINFO: fuser -mv /tmp/LTP_nfs07.P1XS9smc5w/mntpoint/3/0 USER PID ACCESS COMMAND /tmp/LTP_nfs07.P1XS9smc5w/mntpoint/3/0: root kernel mount /tmp/LTP_nfs07.P1XS9smc5w/mntpoint/3/0 ## from nfs_cleanup() ## grep -q "$local_dir" /proc/mounts && umount $local_dir nfs07 4 TINFO: umount /tmp/LTP_nfs07.P1XS9smc5w/mntpoint/3/0 umount: /tmp/LTP_nfs07.P1XS9smc5w/mntpoint: target is busy. ## The rest is not relevant as the problem is above ## tst_umount from tst_test.sh trying hard to umount nfs07 4 TINFO: umount(/tmp/LTP_nfs07.P1XS9smc5w/mntpoint) failed, try 1 ... nfs07 4 TINFO: Likely gvfsd-trash is probing newly mounted fs, kill it to speed up tests. umount: /tmp/LTP_nfs07.P1XS9smc5w/mntpoint: target is busy. nfs07 4 TINFO: umount(/tmp/LTP_nfs07.P1XS9smc5w/mntpoint) failed, try 2 ... nfs07 4 TINFO: Likely gvfsd-trash is probing newly mounted fs, kill it to speed up tests. ... nfs07 4 TINFO: umount(/tmp/LTP_nfs07.P1XS9smc5w/mntpoint) failed, try 50 ... nfs07 4 TINFO: Likely gvfsd-trash is probing newly mounted fs, kill it to speed up tests. nfs07 4 TWARN: Failed to umount(/tmp/LTP_nfs07.P1XS9smc5w/mntpoint) after 50 retries ## if ! tst_device release "$TST_DEVICE"; then in _tst_do_exit() in tst_test.sh tst_device.c:255: TWARN: ioctl(/dev/loop1, LOOP_CLR_FD, 0) no ENXIO for too long Usage tst_device acquire [size [filename]] tst_device release /path/to/device tst_device clear /path/to/device ## rm -r "$TST_TMPDIR" in _tst_do_exit() in tst_test.sh nfs07 4 TWARN: Failed to release device '/dev/loop1' rm: cannot remove '/tmp/LTP_nfs07.P1XS9smc5w/mntpoint': Device or resource busy rm: cannot remove '/tmp/LTP_nfs07.P1XS9smc5w/mntpoint': Device or resource busy nfs07 4 TINFO: AppArmor enabled, this may affect test results nfs07 4 TINFO: it can be disabled with TST_DISABLE_APPARMOR=1 (requires super/root) nfs07 4 TINFO: loaded AppArmor profiles: none Kind regards, Petr testcases/network/nfs/nfs_stress/nfs_lib.sh | 25 ++++++++++++--------- 1 file changed, 14 insertions(+), 11 deletions(-)