Re: [PATCH 02/12] xfs: make xfs_buf_read return an error code

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

On Thu, Jan 23, 2020 at 04:16:58PM -0800, Darrick J. Wong wrote:
> On Thu, Jan 23, 2020 at 02:20:15PM -0800, Christoph Hellwig wrote:
> > On Wed, Jan 22, 2020 at 11:42:03PM -0800, Darrick J. Wong wrote:
> > > -				return -ENOMEM;
> > > +			error = xfs_buf_read(mp->m_ddev_targp, dblkno, dblkcnt,
> > > +					0, &bp, &xfs_attr3_rmt_buf_ops);
> > > +			if (error)
> > > +				return error;
> > >  			error = bp->b_error;
> > >  			if (error) {
> > >  				xfs_buf_ioerror_alert(bp, __func__);
> > 
> > This still has the bogus b_error check where it should just check the
> > error and report it based on the return value.
> 
> Yes, because at this point midway through the series xfs_buf_read only
> knows how to return -ENOMEM.  I'm changing the *interfaces* here, saving
> most of the behavior changes for the xfs_buf_read_map change...
> 
> > > diff --git a/fs/xfs/xfs_buf.h b/fs/xfs/xfs_buf.h
> > > index 56e081dd1d96..fb60c36a8a5b 100644
> > > --- a/fs/xfs/xfs_buf.h
> > > +++ b/fs/xfs/xfs_buf.h
> > > @@ -213,16 +213,25 @@ xfs_buf_get(
> > >  	return xfs_buf_get_map(target, &map, 1, 0);
> > >  }
> > >  
> > > -static inline struct xfs_buf *
> > > +static inline int
> > >  xfs_buf_read(
> > >  	struct xfs_buftarg	*target,
> > >  	xfs_daddr_t		blkno,
> > >  	size_t			numblks,
> > >  	xfs_buf_flags_t		flags,
> > > +	struct xfs_buf		**bpp,
> > >  	const struct xfs_buf_ops *ops)
> > >  {
> > > +	struct xfs_buf		*bp;
> > >  	DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
> > > -	return xfs_buf_read_map(target, &map, 1, flags, ops);
> > > +
> > > +	*bpp = NULL;
> > > +	bp = xfs_buf_read_map(target, &map, 1, flags, ops);
> > > +	if (!bp)
> > > +		return -ENOMEM;
> > > +
> > > +	*bpp = bp;
> 
> ...because otherwise I end having to migrate all the bp->b_error
> checking into this static inline function, which causes compile errors.
> 
> I could work around /that/ by moving the xfs_buf_read_map conversion
> earlier in the series, but I'd have to rebase the whole series to
> achieve the same end result, which seems pointless.
> 
> But I guess I could go mess around with it and see just how much of a
> pain doing that /actually/ is...

Changing the order to the following:

xfs: make xfs_buf_alloc return an error code
xfs: make xfs_buf_get_map return an error code
xfs: make xfs_buf_read_map return an error code
xfs: make xfs_buf_get return an error code
xfs: make xfs_buf_get_uncached return an error code
xfs: make xfs_buf_read return an error code
xfs: make xfs_trans_get_buf_map return an error code
xfs: make xfs_trans_get_buf return an error code
xfs: remove the xfs_btree_get_buf[ls] functions
xfs: make xfs_*read_agf return EAGAIN to ALLOC_FLAG_TRYLOCK callers
xfs: remove unnecessary null pointer checks from _read_agf callers
xfs: fix xfs_buf_ioerror_alert location reporting

Was a lot less code churn than I thought it would be.

--D

> --D
> 
> > > +	return 0;
> > >  }
> > >  
> > >  static inline void
> > > diff --git a/fs/xfs/xfs_log_recover.c b/fs/xfs/xfs_log_recover.c
> > > index 0d683fb96396..b29806846916 100644
> > > --- a/fs/xfs/xfs_log_recover.c
> > > +++ b/fs/xfs/xfs_log_recover.c
> > > @@ -2745,10 +2745,10 @@ xlog_recover_buffer_pass2(
> > >  	if (buf_f->blf_flags & XFS_BLF_INODE_BUF)
> > >  		buf_flags |= XBF_UNMAPPED;
> > >  
> > > -	bp = xfs_buf_read(mp->m_ddev_targp, buf_f->blf_blkno, buf_f->blf_len,
> > > -			  buf_flags, NULL);
> > > -	if (!bp)
> > > -		return -ENOMEM;
> > > +	error = xfs_buf_read(mp->m_ddev_targp, buf_f->blf_blkno, buf_f->blf_len,
> > > +			  buf_flags, &bp, NULL);
> > > +	if (error)
> > > +		return error;
> > >  	error = bp->b_error;
> > >  	if (error) {
> > 
> > .. and same here.
> >