> +struct xfs_buf *
> +xfs_buf_read_uncached(
> + struct xfs_mount *mp,
> + struct xfs_buftarg *target,
> + xfs_daddr_t daddr,
> + size_t length)
> +{
> + xfs_buf_t *bp;
> + int error;
struct xfs_buf and the same indentation as the parameters, please.
> +
> + bp = xfs_buf_get_noaddr(length, target);
I think both the buf_get and buf_read interfaces for the non-hash
buffers should have the same name. Either your uncached or maybe better
unhashed? (And certainly no noaddr, which is not very useful)
> + if (!bp || XFS_BUF_ISERROR(bp))
> + goto fail;
xfs_buf_get_noaddr never returns an error in the buffer.
> + /* set up the buffer for a read IO */
> + xfs_buf_lock(bp);
> + XFS_BUF_SET_ADDR(bp, daddr);
> + XFS_BUF_READ(bp);
> + XFS_BUF_BUSY(bp);
> +
> + xfsbdstrat(mp, bp);
> + error = xfs_iowait(bp);
> + if (error || XFS_BUF_ISERROR(bp))
> + goto fail;
> +
> + return bp;
> +fail:
> + if (bp)
> + xfs_buf_relse(bp);
again, different labels please.
Also this one returns the buffer locked, while buf_get_noaddr doesn't.
I suspect we should also change buf_get_noaddr to return a locked buffer
to make it consistant with all other buf_read/get interfaces.
> +struct xfs_buf * xfs_buf_read_uncached(struct xfs_mount *mp,
> + struct xfs_buftarg *target,
> + xfs_daddr_t daddr, size_t length);
wrong placement of the *
This patch, or at least the introduction of the new read helper should
be moved before patch 1 so that we don't add code that gets removed a
little later.
_______________________________________________
xfs mailing list
xfs@xxxxxxxxxxx
http://oss.sgi.com/mailman/listinfo/xfs
