On Mon, Jan 31, 2022 at 04:59:49PM -0600, Eric Sandeen wrote:
> On 1/19/22 6:19 PM, Darrick J. Wong wrote:
> > From: Dave Chinner <dchinner@xxxxxxxxxx>
> >
> > Source kernel commit: 0560f31a09e523090d1ab2bfe21c69d028c2bdf2
> >
> > Replace m_flags feature checks with xfs_has_<feature>() calls and
> > rework the setup code to set flags in m_features.
> >
> > Signed-off-by: Dave Chinner <dchinner@xxxxxxxxxx>
> > Reviewed-by: Darrick J. Wong <djwong@xxxxxxxxxx>
> > Signed-off-by: Darrick J. Wong <djwong@xxxxxxxxxx>
> > ---
> > include/xfs_mount.h | 11 +++++++++++
> > libxfs/xfs_attr.c | 4 ++--
> > libxfs/xfs_attr_leaf.c | 41 +++++++++++++++++++++++------------------
> > libxfs/xfs_bmap.c | 4 ++--
> > libxfs/xfs_ialloc.c | 10 ++++------
> > 5 files changed, 42 insertions(+), 28 deletions(-)
> >
> >
> > diff --git a/include/xfs_mount.h b/include/xfs_mount.h
> > index a995140d..d4b4ccdc 100644
> > --- a/include/xfs_mount.h
> > +++ b/include/xfs_mount.h
> > @@ -197,6 +197,17 @@ __XFS_HAS_FEAT(inobtcounts, INOBTCNT)
> > __XFS_HAS_FEAT(bigtime, BIGTIME)
> > __XFS_HAS_FEAT(needsrepair, NEEDSREPAIR)
> >
> > +/* Kernel mount features that we don't support */
> > +#define __XFS_UNSUPP_FEAT(name) \
> > +static inline bool xfs_has_ ## name (struct xfs_mount *mp) \
> > +{ \
> > + return false; \
> > +}
> > +__XFS_UNSUPP_FEAT(wsync)
> > +__XFS_UNSUPP_FEAT(noattr2)
> > +__XFS_UNSUPP_FEAT(ikeep)
> > +__XFS_UNSUPP_FEAT(swalloc)
>
> So I'd like to add small_inums to the above list:
>
> +__XFS_UNSUPP_FEAT(small_inums)
>
> and use that in xfs_set_inode_alloc() here as was done in kernelspace, i.e.:
>
> diff --git a/libxfs/init.c b/libxfs/init.c
> index 7d94b721..364c3578 100644
> --- a/libxfs/init.c
> +++ b/libxfs/init.c
> @@ -540,7 +540,7 @@ xfs_set_inode_alloc(
> * sufficiently large, set XFS_MOUNT_32BITINODES if we must alter
> * the allocator to accommodate the request.
> */
> - if ((mp->m_flags & XFS_MOUNT_SMALL_INUMS) && ino > XFS_MAXINUMBER_32)
> + if (xfs_has_small_inums(mp) && ino > XFS_MAXINUMBER_32)
> mp->m_flags |= XFS_MOUNT_32BITINODES;
> else
> mp->m_flags &= ~XFS_MOUNT_32BITINODES;
>
>
> That will let us keep this "backport libxfs" series more or less the
> same as kernelspace, without needing to gut xfs_set_inode_alloc() as
> got proposed in
>
> [PATCH v1.1 39/45] libxfs: remove pointless *XFS_MOUNT* flags
>
> and followups to preserve userspace inode allocation behavior.
>
> If we want to remove dead code from the userspace-specific version of
> xfs_set_inode_alloc() I think we can/should do that separately.
Ok, that sounds good to me.
--D
> Thanks,
> -Eric
>
