spi/pxa2xx: add support for DMA engine

To be able to use DMA with this driver on non-PXA platforms we implement
support for the generic DMA engine API. This lets user to use different DMA
engines with little or no modification to the driver.

Request lines and channel numbers can be passed to the driver from the
platform specific data.

The DMA engine implementation will be selected by default even on PXA
platform. User can select the legacy DMA API by enabling Kconfig option
CONFIG_SPI_PXA2XX_PXADMA.

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Tested-by: Lu Cao <lucao@marvell.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>

1 files changed, 392 insertions, 0 deletions

diff --git a/drivers/spi/spi-pxa2xx-dma.c b/drivers/spi/spi-pxa2xx-dma.c
new file mode 100644
index 00000000000..c735c5a008a
--- /dev/null
+++ b/drivers/spi/spi-pxa2xx-dma.c
@@ -0,0 +1,392 @@
+/*
+ * PXA2xx SPI DMA engine support.
+ *
+ * Copyright (C) 2013, Intel Corporation
+ * Author: Mika Westerberg <mika.westerberg@linux.intel.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/init.h>
+#include <linux/device.h>
+#include <linux/dma-mapping.h>
+#include <linux/dmaengine.h>
+#include <linux/pxa2xx_ssp.h>
+#include <linux/scatterlist.h>
+#include <linux/sizes.h>
+#include <linux/spi/spi.h>
+#include <linux/spi/pxa2xx_spi.h>
+
+#include "spi-pxa2xx.h"
+
+static int pxa2xx_spi_map_dma_buffer(struct driver_data *drv_data,
+				     enum dma_data_direction dir)
+{
+	int i, nents, len = drv_data->len;
+	struct scatterlist *sg;
+	struct device *dmadev;
+	struct sg_table *sgt;
+	void *buf, *pbuf;
+
+	/*
+	 * Some DMA controllers have problems transferring buffers that are
+	 * not multiple of 4 bytes. So we truncate the transfer so that it
+	 * is suitable for such controllers, and handle the trailing bytes
+	 * manually after the DMA completes.
+	 *
+	 * REVISIT: It would be better if this information could be
+	 * retrieved directly from the DMA device in a similar way than
+	 * ->copy_align etc. is done.
+	 */
+	len = ALIGN(drv_data->len, 4);
+
+	if (dir == DMA_TO_DEVICE) {
+		dmadev = drv_data->tx_chan->device->dev;
+		sgt = &drv_data->tx_sgt;
+		buf = drv_data->tx;
+		drv_data->tx_map_len = len;
+	} else {
+		dmadev = drv_data->rx_chan->device->dev;
+		sgt = &drv_data->rx_sgt;
+		buf = drv_data->rx;
+		drv_data->rx_map_len = len;
+	}
+
+	nents = DIV_ROUND_UP(len, SZ_2K);
+	if (nents != sgt->nents) {
+		int ret;
+
+		sg_free_table(sgt);
+		ret = sg_alloc_table(sgt, nents, GFP_KERNEL);
+		if (ret)
+			return ret;
+	}
+
+	pbuf = buf;
+	for_each_sg(sgt->sgl, sg, sgt->nents, i) {
+		size_t bytes = min_t(size_t, len, SZ_2K);
+
+		if (buf)
+			sg_set_buf(sg, pbuf, bytes);
+		else
+			sg_set_buf(sg, drv_data->dummy, bytes);
+
+		pbuf += bytes;
+		len -= bytes;
+	}
+
+	nents = dma_map_sg(dmadev, sgt->sgl, sgt->nents, dir);
+	if (!nents)
+		return -ENOMEM;
+
+	return nents;
+}
+
+static void pxa2xx_spi_unmap_dma_buffer(struct driver_data *drv_data,
+					enum dma_data_direction dir)
+{
+	struct device *dmadev;
+	struct sg_table *sgt;
+
+	if (dir == DMA_TO_DEVICE) {
+		dmadev = drv_data->tx_chan->device->dev;
+		sgt = &drv_data->tx_sgt;
+	} else {
+		dmadev = drv_data->rx_chan->device->dev;
+		sgt = &drv_data->rx_sgt;
+	}
+
+	dma_unmap_sg(dmadev, sgt->sgl, sgt->nents, dir);
+}
+
+static void pxa2xx_spi_unmap_dma_buffers(struct driver_data *drv_data)
+{
+	if (!drv_data->dma_mapped)
+		return;
+
+	pxa2xx_spi_unmap_dma_buffer(drv_data, DMA_FROM_DEVICE);
+	pxa2xx_spi_unmap_dma_buffer(drv_data, DMA_TO_DEVICE);
+
+	drv_data->dma_mapped = 0;
+}
+
+static void pxa2xx_spi_dma_transfer_complete(struct driver_data *drv_data,
+					     bool error)
+{
+	struct spi_message *msg = drv_data->cur_msg;
+
+	/*
+	 * It is possible that one CPU is handling ROR interrupt and other
+	 * just gets DMA completion. Calling pump_transfers() twice for the
+	 * same transfer leads to problems thus we prevent concurrent calls
+	 * by using ->dma_running.
+	 */
+	if (atomic_dec_and_test(&drv_data->dma_running)) {
+		void __iomem *reg = drv_data->ioaddr;
+
+		/*
+		 * If the other CPU is still handling the ROR interrupt we
+		 * might not know about the error yet. So we re-check the
+		 * ROR bit here before we clear the status register.
+		 */
+		if (!error) {
+			u32 status = read_SSSR(reg) & drv_data->mask_sr;
+			error = status & SSSR_ROR;
+		}
+
+		/* Clear status & disable interrupts */
+		write_SSCR1(read_SSCR1(reg) & ~drv_data->dma_cr1, reg);
+		write_SSSR_CS(drv_data, drv_data->clear_sr);
+		if (!pxa25x_ssp_comp(drv_data))
+			write_SSTO(0, reg);
+
+		if (!error) {
+			pxa2xx_spi_unmap_dma_buffers(drv_data);
+
+			/* Handle the last bytes of unaligned transfer */
+			drv_data->tx += drv_data->tx_map_len;
+			drv_data->write(drv_data);
+
+			drv_data->rx += drv_data->rx_map_len;
+			drv_data->read(drv_data);
+
+			msg->actual_length += drv_data->len;
+			msg->state = pxa2xx_spi_next_transfer(drv_data);
+		} else {
+			/* In case we got an error we disable the SSP now */
+			write_SSCR0(read_SSCR0(reg) & ~SSCR0_SSE, reg);
+
+			msg->state = ERROR_STATE;
+		}
+
+		tasklet_schedule(&drv_data->pump_transfers);
+	}
+}
+
+static void pxa2xx_spi_dma_callback(void *data)
+{
+	pxa2xx_spi_dma_transfer_complete(data, false);
+}
+
+static struct dma_async_tx_descriptor *
+pxa2xx_spi_dma_prepare_one(struct driver_data *drv_data,
+			   enum dma_transfer_direction dir)
+{
+	struct pxa2xx_spi_master *pdata = drv_data->master_info;
+	struct chip_data *chip = drv_data->cur_chip;
+	enum dma_slave_buswidth width;
+	struct dma_slave_config cfg;
+	struct dma_chan *chan;
+	struct sg_table *sgt;
+	int nents, ret;
+
+	switch (drv_data->n_bytes) {
+	case 1:
+		width = DMA_SLAVE_BUSWIDTH_1_BYTE;
+		break;
+	case 2:
+		width = DMA_SLAVE_BUSWIDTH_2_BYTES;
+		break;
+	default:
+		width = DMA_SLAVE_BUSWIDTH_4_BYTES;
+		break;
+	}
+
+	memset(&cfg, 0, sizeof(cfg));
+	cfg.direction = dir;
+
+	if (dir == DMA_MEM_TO_DEV) {
+		cfg.dst_addr = drv_data->ssdr_physical;
+		cfg.dst_addr_width = width;
+		cfg.dst_maxburst = chip->dma_burst_size;
+		cfg.slave_id = pdata->tx_slave_id;
+
+		sgt = &drv_data->tx_sgt;
+		nents = drv_data->tx_nents;
+		chan = drv_data->tx_chan;
+	} else {
+		cfg.src_addr = drv_data->ssdr_physical;
+		cfg.src_addr_width = width;
+		cfg.src_maxburst = chip->dma_burst_size;
+		cfg.slave_id = pdata->rx_slave_id;
+
+		sgt = &drv_data->rx_sgt;
+		nents = drv_data->rx_nents;
+		chan = drv_data->rx_chan;
+	}
+
+	ret = dmaengine_slave_config(chan, &cfg);
+	if (ret) {
+		dev_warn(&drv_data->pdev->dev, "DMA slave config failed\n");
+		return NULL;
+	}
+
+	return dmaengine_prep_slave_sg(chan, sgt->sgl, nents, dir,
+				       DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
+}
+
+static bool pxa2xx_spi_dma_filter(struct dma_chan *chan, void *param)
+{
+	const struct pxa2xx_spi_master *pdata = param;
+
+	return chan->chan_id == pdata->tx_chan_id ||
+	       chan->chan_id == pdata->rx_chan_id;
+}
+
+bool pxa2xx_spi_dma_is_possible(size_t len)
+{
+	return len <= MAX_DMA_LEN;
+}
+
+int pxa2xx_spi_map_dma_buffers(struct driver_data *drv_data)
+{
+	const struct chip_data *chip = drv_data->cur_chip;
+	int ret;
+
+	if (!chip->enable_dma)
+		return 0;
+
+	/* Don't bother with DMA if we can't do even a single burst */
+	if (drv_data->len < chip->dma_burst_size)
+		return 0;
+
+	ret = pxa2xx_spi_map_dma_buffer(drv_data, DMA_TO_DEVICE);
+	if (ret <= 0) {
+		dev_warn(&drv_data->pdev->dev, "failed to DMA map TX\n");
+		return 0;
+	}
+
+	drv_data->tx_nents = ret;
+
+	ret = pxa2xx_spi_map_dma_buffer(drv_data, DMA_FROM_DEVICE);
+	if (ret <= 0) {
+		pxa2xx_spi_unmap_dma_buffer(drv_data, DMA_TO_DEVICE);
+		dev_warn(&drv_data->pdev->dev, "failed to DMA map RX\n");
+		return 0;
+	}
+
+	drv_data->rx_nents = ret;
+	return 1;
+}
+
+irqreturn_t pxa2xx_spi_dma_transfer(struct driver_data *drv_data)
+{
+	u32 status;
+
+	status = read_SSSR(drv_data->ioaddr) & drv_data->mask_sr;
+	if (status & SSSR_ROR) {
+		dev_err(&drv_data->pdev->dev, "FIFO overrun\n");
+
+		dmaengine_terminate_all(drv_data->rx_chan);
+		dmaengine_terminate_all(drv_data->tx_chan);
+
+		pxa2xx_spi_dma_transfer_complete(drv_data, true);
+		return IRQ_HANDLED;
+	}
+
+	return IRQ_NONE;
+}
+
+int pxa2xx_spi_dma_prepare(struct driver_data *drv_data, u32 dma_burst)
+{
+	struct dma_async_tx_descriptor *tx_desc, *rx_desc;
+
+	tx_desc = pxa2xx_spi_dma_prepare_one(drv_data, DMA_MEM_TO_DEV);
+	if (!tx_desc) {
+		dev_err(&drv_data->pdev->dev,
+			"failed to get DMA TX descriptor\n");
+		return -EBUSY;
+	}
+
+	rx_desc = pxa2xx_spi_dma_prepare_one(drv_data, DMA_DEV_TO_MEM);
+	if (!rx_desc) {
+		dev_err(&drv_data->pdev->dev,
+			"failed to get DMA RX descriptor\n");
+		return -EBUSY;
+	}
+
+	/* We are ready when RX completes */
+	rx_desc->callback = pxa2xx_spi_dma_callback;
+	rx_desc->callback_param = drv_data;
+
+	dmaengine_submit(rx_desc);
+	dmaengine_submit(tx_desc);
+	return 0;
+}
+
+void pxa2xx_spi_dma_start(struct driver_data *drv_data)
+{
+	dma_async_issue_pending(drv_data->rx_chan);
+	dma_async_issue_pending(drv_data->tx_chan);
+
+	atomic_set(&drv_data->dma_running, 1);
+}
+
+int pxa2xx_spi_dma_setup(struct driver_data *drv_data)
+{
+	struct pxa2xx_spi_master *pdata = drv_data->master_info;
+	dma_cap_mask_t mask;
+
+	dma_cap_zero(mask);
+	dma_cap_set(DMA_SLAVE, mask);
+
+	drv_data->dummy = devm_kzalloc(&drv_data->pdev->dev, SZ_2K, GFP_KERNEL);
+	if (!drv_data->dummy)
+		return -ENOMEM;
+
+	drv_data->tx_chan = dma_request_channel(mask, pxa2xx_spi_dma_filter,
+						pdata);
+	if (!drv_data->tx_chan)
+		return -ENODEV;
+
+	drv_data->rx_chan = dma_request_channel(mask, pxa2xx_spi_dma_filter,
+						pdata);
+	if (!drv_data->rx_chan) {
+		dma_release_channel(drv_data->tx_chan);
+		drv_data->tx_chan = NULL;
+		return -ENODEV;
+	}
+
+	return 0;
+}
+
+void pxa2xx_spi_dma_release(struct driver_data *drv_data)
+{
+	if (drv_data->rx_chan) {
+		dmaengine_terminate_all(drv_data->rx_chan);
+		dma_release_channel(drv_data->rx_chan);
+		sg_free_table(&drv_data->rx_sgt);
+		drv_data->rx_chan = NULL;
+	}
+	if (drv_data->tx_chan) {
+		dmaengine_terminate_all(drv_data->tx_chan);
+		dma_release_channel(drv_data->tx_chan);
+		sg_free_table(&drv_data->tx_sgt);
+		drv_data->tx_chan = NULL;
+	}
+}
+
+void pxa2xx_spi_dma_resume(struct driver_data *drv_data)
+{
+}
+
+int pxa2xx_spi_set_dma_burst_and_threshold(struct chip_data *chip,
+					   struct spi_device *spi,
+					   u8 bits_per_word, u32 *burst_code,
+					   u32 *threshold)
+{
+	struct pxa2xx_spi_chip *chip_info = spi->controller_data;
+
+	/*
+	 * If the DMA burst size is given in chip_info we use that,
+	 * otherwise we use the default. Also we use the default FIFO
+	 * thresholds for now.
+	 */
+	*burst_code = chip_info ? chip_info->dma_burst_size : 16;
+	*threshold = SSCR1_RxTresh(RX_THRESH_DFLT)
+		   | SSCR1_TxTresh(TX_THRESH_DFLT);
+
+	return 0;
+}