On Thu, Feb 20, 2020 at 08:32:32AM -0800, Christoph Hellwig wrote:
> On Thu, Feb 20, 2020 at 05:30:35PM +0100, Pavel Reichl wrote:
> > OK, thanks for the comments.
> >
> > Eric in the following code is WARN_ONCE() used as you suggested or did
> > you have something else in mind?
> >
> > static inline bool
> > __xfs_rwsem_islocked(
> > struct rw_semaphore *rwsem,
> > bool excl)
> > {
> > if (!rwsem_is_locked(rwsem)) {
> > return false;
> > }
> >
> > if (excl) {
> > if (debug_locks) {
> > return lockdep_is_held_type(rwsem, 1);
> > }
> > WARN_ONCE(1,
> > "xfs rwsem lock testing coverage has been reduced\n");
> > }
>
> Yikes, hell no. This means every debug xfs build without lockdep
> will be full of warnings all the time.
Well... once per module load, but if you /were/ going to go down this
route I'd at least gate the warning on IS_ENABLED(CONFIG_LOCKDEP) so
that we only get this one-time warning when lockdep is enabled but dies
anyway, so that we'll know that we're running with half a brain.
--D
