From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eryu Guan Subject: Re: [PATCH 6/7] overlay: test encode/decode overlay file handles Date: Tue, 16 Jan 2018 19:06:51 +0800 Message-ID: <20180116110651.GK3102@eguan.usersys.redhat.com> References: <1515348445-1403-1-git-send-email-amir73il@gmail.com> <1515348445-1403-7-git-send-email-amir73il@gmail.com> <20180116073831.GH3102@eguan.usersys.redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from mx1.redhat.com ([209.132.183.28]:50118 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751295AbeAPLGy (ORCPT ); Tue, 16 Jan 2018 06:06:54 -0500 Content-Disposition: inline In-Reply-To: Sender: linux-unionfs-owner@vger.kernel.org List-Id: linux-unionfs@vger.kernel.org To: Amir Goldstein Cc: Jeff Layton , "J . Bruce Fields" , Miklos Szeredi , fstests , linux-fsdevel , overlayfs On Tue, Jan 16, 2018 at 12:53:38PM +0200, Amir Goldstein wrote: > On Tue, Jan 16, 2018 at 9:38 AM, Eryu Guan wrote: > > On Sun, Jan 07, 2018 at 08:07:24PM +0200, Amir Goldstein wrote: > >> - Check encode/write/decode/read content of lower/upper file handles > >> - Check encode/decode/write/read content of lower/upper file handles > >> - Check decode/read of unlinked lower/upper files and directories > >> - Check decode/read of lower file handles after copy up, link and unlink > >> - Check decode/read of lower file handles after rename of parent and self > > > > I'm wondering that if this should be split into multiple tests somehow, > > e.g. tests on regular files, tests on dirs and tests on hardlinks? It > > might be eaiser to review and debug when there're test failures. But I > > have no strong preference on this. > > > > I prefer not splitting the test, this is a classic test with sub-test cases. > I may end up splitting the dir rename tests (open_by_handle -i/-o) > to conform with a similar split that you requested in the generic test. > > >> > >> This test does not cover connectable file handles of non-directories, > >> because name_to_handle_at() syscall does not support requesting > >> connectable file handles. > >> > >> This test covers only encode/decode of file handles for overlayfs > >> configuration of lower and upper on the same fs. > >> > >> Signed-off-by: Amir Goldstein > >> --- > >> tests/overlay/050 | 291 ++++++++++++++++++++++++++++++++++++++++++++++++++ > >> tests/overlay/050.out | 50 +++++++++ > >> tests/overlay/group | 1 + > >> 3 files changed, 342 insertions(+) > >> create mode 100755 tests/overlay/050 > >> create mode 100644 tests/overlay/050.out > > > > I ran the test on your ovl-nfs-export-v2 branch and saw failures like: > > > > --- tests/overlay/050.out 2018-01-16 14:51:11.350000000 +0800 > > +++ /root/xfstests/results//xfs_4k_crc/overlay/050.out.bad 2018-01-16 15:08:43.487000000 +0800 > > @@ -45,6 +45,9 @@ > > test_file_handles SCRATCH_MNT/lowertestdir/subdir -p -o lower_subdir_file_handles > > test_file_handles SCRATCH_MNT -rp -i upper_file_handles > > test_file_handles SCRATCH_MNT -rp -i lower_file_handles > > +open_by_handle() returned 116 incorrectly on a linked dir! > > test_file_handles SCRATCH_MNT -rp -i upper_subdir_file_handles > > test_file_handles SCRATCH_MNT -rp -i lower_subdir_file_handles > > +open_by_handle() returned 116 incorrectly on a linked dir! > > test_file_handles SCRATCH_MNT/lowertestdir.new -rp -i lower_subdir_file_handles > > +open_by_handle() returned 116 incorrectly on a linked dir! > > > > Are these failures expected? > > > > No. not expected. I wonder which base fs did you test with? > Did you have OVERLAY_FS_VERIFY=y in config or verify=on in MOUNT_OPTIONS? > (Not that I know any of the above should matter) I didn't have OVERLAY_FS_VERIFY set in .config, but I did mount with "-o verify=on", and underlying fs is xfs. Here is the screenshot: [root@bootp-73-5-205 xfstests]# OVERLAY_MOUNT_OPTIONS="-o verify=on" ./check -s xfs_4k_crc -overlay overlay/050 SECTION -- xfs_4k_crc RECREATING -- overlay on /mnt/testarea/test FSTYP -- overlay PLATFORM -- Linux/x86_64 bootp-73-5-205 4.15.0-rc2.ovl+ MKFS_OPTIONS -- -f -b size=4k -m crc=1 /mnt/testarea/scratch MOUNT_OPTIONS -- -o verify=on /mnt/testarea/scratch /mnt/testarea/scratch/ovl-mnt overlay/050 - output mismatch (see /root/xfstests/results//xfs_4k_crc/overlay/050.out.bad) --- tests/overlay/050.out 2018-01-16 14:51:11.350000000 +0800 +++ /root/xfstests/results//xfs_4k_crc/overlay/050.out.bad 2018-01-16 19:01:54.984000000 +0800 @@ -45,6 +45,9 @@ test_file_handles SCRATCH_MNT/lowertestdir/subdir -p -o lower_subdir_file_handles test_file_handles SCRATCH_MNT -rp -i upper_file_handles test_file_handles SCRATCH_MNT -rp -i lower_file_handles +open_by_handle() returned 116 incorrectly on a linked dir! test_file_handles SCRATCH_MNT -rp -i upper_subdir_file_handles test_file_handles SCRATCH_MNT -rp -i lower_subdir_file_handles +open_by_handle() returned 116 incorrectly on a linked dir! ... (Run 'diff -u tests/overlay/050.out /root/xfstests/results//xfs_4k_crc/overlay/050.out.bad' to see the entire diff) Ran: overlay/050 Failures: overlay/050 Failed 1 of 1 tests And I just tried with ext4 as underlying fs and got the same result. > > Do you see any overlayfs warnings in dmesg? No, there's no warnings nor other useful information in dmesg, just mount/umount xfs and drop caches messages. Thanks, Eryu