spi: spi-fsl-dspi: fix DMA mapping
authorMichael Walle <michael@walle.cc>
Tue, 10 Mar 2020 07:33:13 +0000 (08:33 +0100)
committerMark Brown <broonie@kernel.org>
Tue, 10 Mar 2020 18:34:57 +0000 (18:34 +0000)
Use the correct device to request the DMA mapping. Otherwise the IOMMU
doesn't get the mapping and it will generate a page fault.

The error messages look like:
[    3.008452] arm-smmu 5000000.iommu: Unhandled context fault: fsr=0x402, iova=0xf9800000, fsynr=0x3f0022, cbfrsynra=0x828, cb=8
[    3.020123] arm-smmu 5000000.iommu: Unhandled context fault: fsr=0x402, iova=0xf9800000, fsynr=0x3f0022, cbfrsynra=0x828, cb=8

This was tested on a custom board with a LS1028A SoC.

Signed-off-by: Michael Walle <michael@walle.cc>
Link: https://lore.kernel.org/r/20200310073313.21277-1-michael@walle.cc
Signed-off-by: Mark Brown <broonie@kernel.org>
drivers/spi/spi-fsl-dspi.c

index 0683a3f..50e3382 100644 (file)
@@ -497,14 +497,16 @@ static int dspi_request_dma(struct fsl_dspi *dspi, phys_addr_t phy_addr)
                goto err_tx_channel;
        }
 
-       dma->tx_dma_buf = dma_alloc_coherent(dev, dspi->devtype_data->dma_bufsize,
+       dma->tx_dma_buf = dma_alloc_coherent(dma->chan_tx->device->dev,
+                                            dspi->devtype_data->dma_bufsize,
                                             &dma->tx_dma_phys, GFP_KERNEL);
        if (!dma->tx_dma_buf) {
                ret = -ENOMEM;
                goto err_tx_dma_buf;
        }
 
-       dma->rx_dma_buf = dma_alloc_coherent(dev, dspi->devtype_data->dma_bufsize,
+       dma->rx_dma_buf = dma_alloc_coherent(dma->chan_rx->device->dev,
+                                            dspi->devtype_data->dma_bufsize,
                                             &dma->rx_dma_phys, GFP_KERNEL);
        if (!dma->rx_dma_buf) {
                ret = -ENOMEM;
@@ -541,10 +543,12 @@ static int dspi_request_dma(struct fsl_dspi *dspi, phys_addr_t phy_addr)
        return 0;
 
 err_slave_config:
-       dma_free_coherent(dev, dspi->devtype_data->dma_bufsize,
+       dma_free_coherent(dma->chan_rx->device->dev,
+                         dspi->devtype_data->dma_bufsize,
                          dma->rx_dma_buf, dma->rx_dma_phys);
 err_rx_dma_buf:
-       dma_free_coherent(dev, dspi->devtype_data->dma_bufsize,
+       dma_free_coherent(dma->chan_tx->device->dev,
+                         dspi->devtype_data->dma_bufsize,
                          dma->tx_dma_buf, dma->tx_dma_phys);
 err_tx_dma_buf:
        dma_release_channel(dma->chan_tx);
@@ -560,20 +564,19 @@ err_tx_channel:
 static void dspi_release_dma(struct fsl_dspi *dspi)
 {
        struct fsl_dspi_dma *dma = dspi->dma;
-       struct device *dev = &dspi->pdev->dev;
 
        if (!dma)
                return;
 
        if (dma->chan_tx) {
-               dma_unmap_single(dev, dma->tx_dma_phys,
+               dma_unmap_single(dma->chan_tx->device->dev, dma->tx_dma_phys,
                                 dspi->devtype_data->dma_bufsize,
                                 DMA_TO_DEVICE);
                dma_release_channel(dma->chan_tx);
        }
 
        if (dma->chan_rx) {
-               dma_unmap_single(dev, dma->rx_dma_phys,
+               dma_unmap_single(dma->chan_rx->device->dev, dma->rx_dma_phys,
                                 dspi->devtype_data->dma_bufsize,
                                 DMA_FROM_DEVICE);
                dma_release_channel(dma->chan_rx);