Hi Christoph, I've tried to get your feedback on the lockdep_assert_held in iomap_dio_rw in January and didn't hear back from you. Could you please have a look?
On Tue, 22 Jan 2019 at 22:26, Andreas Gruenbacher <agrue...@redhat.com> wrote: > > Hi Christoph, > > there's an assertion that the inode rwsem is taken in iomap_dio_rw > which currently triggers on gfs2 via gfs2_file_read_iter -> > gfs2_file_direct_read -> iomap_dio_rw in fs/gfs2/file.c. > > We're not using inode_[un]lock_shared on that code path because gfs2 > uses its own cluster-wide locking already. Is there a reason why the > inode needs to be locked locally anyway? If not, I'd like to avoid > adding unnecessary lock taking in gfs2_file_direct_read just to > silence iomap_dio_rw. Thanks, Andreas