On Fri, Nov 05, 2021 at 11:40:57AM -0500, Eric Sandeen wrote: > On 11/4/21 5:38 PM, Dave Chinner wrote: > > > With those changes, we end up with some new stubs in libxfs_priv.h > > and two places where we need #ifdef __KERNEL__ in xfs_ag.[ch]. Most > > of the mess in this patch goes away.... > > > > Cheers, > > > > Dave. > Ok. > > I will split this up into the right patch granularity, but is this the > endpoint you're looking for? One #ifdef in each of xfs_ag.[ch], two total. Yup. > The delayed work init/cancel assymmetry is a little odd, but I'll > get over it. > > diff --git a/libxfs/libxfs_priv.h b/libxfs/libxfs_priv.h > index 15bae1ff..2ca3b9b2 100644 > --- a/libxfs/libxfs_priv.h > +++ b/libxfs/libxfs_priv.h > @@ -673,6 +673,9 @@ static inline void xfs_iunlink_destroy(struct xfs_perag *pag) { } > xfs_agnumber_t xfs_set_inode_alloc(struct xfs_mount *mp, > xfs_agnumber_t agcount); > +/* Faked up kernel bits */ > +#define cancel_delayed_work_sync(work) do { } while(0) Comment is completely redundant. libxfs_priv.h is entirely for "faked up kernel bits". I'd also put this up near the top of the file near the definition of struct iomap, not place it randomly in amongst a bunch of XFS definitions. Cheers, Dave. -- Dave Chinner david@xxxxxxxxxxxxx