[PATCH v1] dmaengine: idma64: Use actual device for DMA transfers

[PATCH v1] dmaengine: idma64: Use actual device for DMA transfers

[Andy Shevchenko]

Intel IOMMU, when enabled, tries to find the domain of the device,
assuming it's a PCI one, during DMA operations, such as mapping or
unmapping. Since we are splitting the actual PCI device to couple of
children via MFD framework (see drivers/mfd/intel-lpss.c for details),
the DMA device appears to be a platform one, and thus not an actual one
that performs DMA. In a such situation IOMMU can't find or allocate
a proper domain for its operations. As a result, all DMA operations are
failed.

In order to fix this, supply parent of the platform device
to the DMA engine framework and fix filter functions accordingly.

We may rely on the fact that parent is a real PCI device, because no
other configuration is present in the wild.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
---
 drivers/dma/idma64.c              | 6 ++++--
 drivers/dma/idma64.h              | 2 ++
 drivers/spi/spi-pxa2xx.c          | 7 +------
 drivers/tty/serial/8250/8250_dw.c | 4 ++--
 4 files changed, 9 insertions(+), 10 deletions(-)

diff --git a/drivers/dma/idma64.c b/drivers/dma/idma64.c
index 0baf9797cc09..83796a33dc16 100644
--- a/drivers/dma/idma64.c
+++ b/drivers/dma/idma64.c
@@ -592,7 +592,7 @@ static int idma64_probe(struct idma64_chip *chip)
 	idma64->dma.directions = BIT(DMA_DEV_TO_MEM) | BIT(DMA_MEM_TO_DEV);
 	idma64->dma.residue_granularity = DMA_RESIDUE_GRANULARITY_BURST;
 
-	idma64->dma.dev = chip->dev;
+	idma64->dma.dev = chip->sysdev;
 
 	dma_set_max_seg_size(idma64->dma.dev, IDMA64C_CTLH_BLOCK_TS_MASK);
 
@@ -632,6 +632,7 @@ static int idma64_platform_probe(struct platform_device *pdev)
 {
 	struct idma64_chip *chip;
 	struct device *dev = &pdev->dev;
+	struct device *sysdev = dev->parent;
 	struct resource *mem;
 	int ret;
 
@@ -648,11 +649,12 @@ static int idma64_platform_probe(struct platform_device *pdev)
 	if (IS_ERR(chip->regs))
 		return PTR_ERR(chip->regs);
 
-	ret = dma_coerce_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
+	ret = dma_coerce_mask_and_coherent(sysdev, DMA_BIT_MASK(64));
 	if (ret)
 		return ret;
 
 	chip->dev = dev;
+	chip->sysdev = sysdev;
 
 	ret = idma64_probe(chip);
 	if (ret)
diff --git a/drivers/dma/idma64.h b/drivers/dma/idma64.h
index 6b816878e5e7..baa32e1425de 100644
--- a/drivers/dma/idma64.h
+++ b/drivers/dma/idma64.h
@@ -216,12 +216,14 @@ static inline void idma64_writel(struct idma64 *idma64, int offset, u32 value)
 /**
  * struct idma64_chip - representation of iDMA 64-bit controller hardware
  * @dev:		struct device of the DMA controller
+ * @sysdev:		struct device of the physical device that does DMA
  * @irq:		irq line
  * @regs:		memory mapped I/O space
  * @idma64:		struct idma64 that is filed by idma64_probe()
  */
 struct idma64_chip {
 	struct device	*dev;
+	struct device	*sysdev;
 	int		irq;
 	void __iomem	*regs;
 	struct idma64	*idma64;
diff --git a/drivers/spi/spi-pxa2xx.c b/drivers/spi/spi-pxa2xx.c
index b6ddba833d02..5ea70f7d12e7 100644
--- a/drivers/spi/spi-pxa2xx.c
+++ b/drivers/spi/spi-pxa2xx.c
@@ -1487,12 +1487,7 @@ static int pxa2xx_spi_get_port_id(struct acpi_device *adev)
 
 static bool pxa2xx_spi_idma_filter(struct dma_chan *chan, void *param)
 {
-	struct device *dev = param;
-
-	if (dev != chan->device->dev->parent)
-		return false;
-
-	return true;
+	return param == chan->device->dev;
 }
 
 #endif /* CONFIG_PCI */
diff --git a/drivers/tty/serial/8250/8250_dw.c b/drivers/tty/serial/8250/8250_dw.c
index d31b975dd3fd..284e8d052fc3 100644
--- a/drivers/tty/serial/8250/8250_dw.c
+++ b/drivers/tty/serial/8250/8250_dw.c
@@ -365,7 +365,7 @@ static bool dw8250_fallback_dma_filter(struct dma_chan *chan, void *param)
 
 static bool dw8250_idma_filter(struct dma_chan *chan, void *param)
 {
-	return param == chan->device->dev->parent;
+	return param == chan->device->dev;
 }
 
 /*
@@ -434,7 +434,7 @@ static void dw8250_quirks(struct uart_port *p, struct dw8250_data *data)
 		data->uart_16550_compatible = true;
 	}
 
-	/* Platforms with iDMA */
+	/* Platforms with iDMA 64-bit */
 	if (platform_get_resource_byname(to_platform_device(p->dev),
 					 IORESOURCE_MEM, "lpss_priv")) {
 		data->dma.rx_param = p->dev->parent;
-- 
2.20.1

Re: [PATCH v1] dmaengine: idma64: Use actual device for DMA transfers

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 317 bytes --]

On Mon, Mar 18, 2019 at 06:39:30PM +0300, Andy Shevchenko wrote:
> Intel IOMMU, when enabled, tries to find the domain of the device,
> assuming it's a PCI one, during DMA operations, such as mapping or
> unmapping. Since we are splitting the actual PCI device to couple of

Acked-by: Mark Brown <broonie@kernel.org>

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH v1] dmaengine: idma64: Use actual device for DMA transfers

[Greg Kroah-Hartman]

On Mon, Mar 18, 2019 at 06:39:30PM +0300, Andy Shevchenko wrote:
> Intel IOMMU, when enabled, tries to find the domain of the device,
> assuming it's a PCI one, during DMA operations, such as mapping or
> unmapping. Since we are splitting the actual PCI device to couple of
> children via MFD framework (see drivers/mfd/intel-lpss.c for details),
> the DMA device appears to be a platform one, and thus not an actual one
> that performs DMA. In a such situation IOMMU can't find or allocate
> a proper domain for its operations. As a result, all DMA operations are
> failed.
> 
> In order to fix this, supply parent of the platform device
> to the DMA engine framework and fix filter functions accordingly.
> 
> We may rely on the fact that parent is a real PCI device, because no
> other configuration is present in the wild.
> 
> Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

For the tty/serial/ portion:

Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Re: [PATCH v1] dmaengine: idma64: Use actual device for DMA transfers

[Vinod Koul]

On 18-03-19, 18:39, Andy Shevchenko wrote:
> Intel IOMMU, when enabled, tries to find the domain of the device,
> assuming it's a PCI one, during DMA operations, such as mapping or
> unmapping. Since we are splitting the actual PCI device to couple of
> children via MFD framework (see drivers/mfd/intel-lpss.c for details),
> the DMA device appears to be a platform one, and thus not an actual one
> that performs DMA. In a such situation IOMMU can't find or allocate
> a proper domain for its operations. As a result, all DMA operations are
> failed.
> 
> In order to fix this, supply parent of the platform device
> to the DMA engine framework and fix filter functions accordingly.
> 
> We may rely on the fact that parent is a real PCI device, because no
> other configuration is present in the wild.

Applied, thanks

-- 
~Vinod