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

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



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...

--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.
> 



[Index of Archives]     [XFS Filesystem Development (older mail)]     [Linux Filesystem Development]     [Linux Audio Users]     [Yosemite Trails]     [Linux Kernel]     [Linux RAID]     [Linux SCSI]


  Powered by Linux