On Mon, Aug 16, 2010 at 10:48:34PM -0600, Grant Likely wrote:
quoted text
> > +	/* Map DMA buffers */
> > +	sglen = dma_map_sg(&pl022->adev->dev, pl022->sgt_rx.sgl,
> > +			   pl022->sgt_rx.nents, DMA_FROM_DEVICE);
> > +	if (sglen != pages)
> > +		goto err_rx_sgmap;
> > +
> > +	sglen = dma_map_sg(&pl022->adev->dev, pl022->sgt_tx.sgl,
> > +			   pl022->sgt_tx.nents, DMA_TO_DEVICE);
> > +	if (sglen != pages)
> > +		goto err_tx_sgmap;
> > +
> > +	/* Synchronize the TX scatterlist, invalidate buffers, caches etc */
> > +	dma_sync_sg_for_device(&pl022->adev->dev,
> > +			       pl022->sgt_tx.sgl,
> > +			       pl022->sgt_tx.nents,
> > +			       DMA_TO_DEVICE);

This is wrong.  Mapping a scatterlist already makes it available for
the device to use.  There's no point re-doing that work by using
the sync API.

Let's go over the DMA API one more time.

	CPU owns buffer, device must not access

dma_map_xx		buffer ownership transitions from CPU to device

	device owns buffer, CPU must not access

dma_sync_xx_for_cpu	buffer ownership transitions from device to CPU

	CPU owns buffer, device must not access

dma_sync_xx_for_device	buffer ownership transitions from CPU to device

	device owns buffer, CPU must not access

dma_unmap_xx		buffer ownership transitions from device to CPU

	CPU owns buffer, device must not access
--
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/