Can you please share the mount options. Did you observe the same failure with 5.14-rc5-smb3-fixes ? I ran the generic/478 test on 5.14 and didn't see any hang. But, test fails with below generic/478 - output mismatch (see /home/lxsmbadmin/xfstests-dev/results//sm b3/generic/478.out.bad) --- tests/generic/478.out 2021-05-03 19:01:04.767577557 +0000 +++ /home/lxsmbadmin/xfstests-dev/results//smb3/generic/478.out.bad 2021-09- 06 15:39:39.625248615 +0000 @@ -1,13 +1,13 @@ QA output created by 478 get wrlck lock could be placed -get wrlck -get wrlck lock could be placed get wrlck ... (Run 'diff -u /home/lxsmbadmin/xfstests-dev/tests/generic/478.out /home/lxsm badmin/xfstests-dev/results//smb3/generic/478.out.bad' to see the entire diff) Ran: generic/478 Failures: generic/478 Failed 1 of 1 tests Regards, Rohith On Mon, Sep 6, 2021 at 5:16 PM Murphy Zhou <jencce.kernel@xxxxxxxxx> wrote: > > Hi, > > Since this commit: > > commit c3f207ab29f793b8c942ce8067ed123f18d5b81b > Author: Rohith Surabattula <rohiths@xxxxxxxxxxxxx> > Date: Tue Apr 13 00:26:42 2021 -0500 > > cifs: Deferred close for files > > Xfstests generic/478 on CIFS can't finish like before. The test programme > never returns but killable. The kernel does not warn about soft or hard > lockups. So it looks like looping forever at some point. > > It's always reproducible. Without this commit, generic/478 fails the test > because of different lock schema but complete very fast. With this commit, > test hang like forever. > > Sorry that I do not look further here, because I have another bisecting to > do to hunting another regression. > > Thanks, > Murphy