Re: [PATCH 3/3] xfs: actually check xfs_btree_check_block return in xfs_btree_islastblock

["Darrick J. Wong" <darrick.wong@oracle.com>]

On Thu, Nov 07, 2019 at 11:14:41PM -0800, Christoph Hellwig wrote:
> On Thu, Nov 07, 2019 at 11:05:04PM -0800, Darrick J. Wong wrote:
> > From: Darrick J. Wong <darrick.wong@oracle.com>
> > 
> > Coverity points out that xfs_btree_islastblock calls
> > xfs_btree_check_block, but doesn't act on an error return.  This
> > predicate has no answer if the btree is corrupt, so tweak the helper to
> > be able to return errors, and then modify the one call site.
> 
> Could we just kill xfs_btree_islastblock?  It has pretty trivial, and
> only has a single caller which only uses on of the two branches in the
> function anyway.

I'd rather leave it as a btree primitive, honestly.

That said, "Is this cursor pointing to the last block on $level?" only
makes sense if you've already performed a lookup (or seek) operation.
If you've done that, you've already checked the block, right?  So I
think we could just get rid of the _check_block call on the grounds that
we already did that as part of the lookup (or turn it into an ASSERT),
and then this becomes a short enough function to try to make it a four
line static inline predicate.

Same result, but slightly better encapsulation.

(Yeah yeah, it's C, we're all one big happy family of bits...)

--D