Message-ID: <20131113135912.GA8834@intel.com>
Date:	Wed, 13 Nov 2013 19:29:12 +0530
From:	Vinod Koul <vinod.koul@...el.com>
To:	Jonas Jensen <jonas.jensen@...il.com>
Cc:	linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
	arm@...nel.org, djbw@...com, arnd@...db.de, linux@....linux.org.uk,
	mark.rutland@....com
Subject: Re: [PATCH v11] dmaengine: Add MOXA ART DMA engine driver

On Tue, Oct 08, 2013 at 10:42:36AM +0200, Jonas Jensen wrote:
> The MOXA ART SoC has a DMA controller capable of offloading expensive
> memory operations, such as large copies. This patch adds support for
> the controller including four channels. Two of these are used to
> handle MMC copy on the UC-7112-LX hardware. The remaining two can be
> used in a future audio driver or client application.
I see this is pending and I first need the AKC on DT parts of the patch before
we can apply this.

Also pls cc dmaengine@...r.kernel.org on this patch

> 
> Signed-off-by: Jonas Jensen <jonas.jensen@...il.com>
> ---
> 
> Notes:
>     Changes since v10:
>     
>     device tree bindings document:
>     1. reformat interrupt description text
>     
>     Applies to next-20130927
> 
>  .../devicetree/bindings/dma/moxa,moxart-dma.txt    |  20 +
>  drivers/dma/Kconfig                                |   7 +
>  drivers/dma/Makefile                               |   1 +
>  drivers/dma/moxart-dma.c                           | 651 +++++++++++++++++++++
>  4 files changed, 679 insertions(+)
>  create mode 100644 Documentation/devicetree/bindings/dma/moxa,moxart-dma.txt
>  create mode 100644 drivers/dma/moxart-dma.c
> 
> diff --git a/Documentation/devicetree/bindings/dma/moxa,moxart-dma.txt b/Documentation/devicetree/bindings/dma/moxa,moxart-dma.txt
> new file mode 100644
> index 0000000..697e3f6
> --- /dev/null
> +++ b/Documentation/devicetree/bindings/dma/moxa,moxart-dma.txt
> @@ -0,0 +1,20 @@
> +MOXA ART DMA Controller
> +
> +See dma.txt first
> +
> +Required properties:
> +
> +- compatible :	Must be "moxa,moxart-dma"
> +- reg :		Should contain registers location and length
> +- interrupts :	Should contain an interrupt-specifier for the sole
> +		interrupt generated by the device
> +- #dma-cells :	Should be 1, a single cell holding a line request number
> +
> +Example:
> +
> +	dma: dma@...00000 {
> +		compatible = "moxa,moxart-dma";
> +		reg = <0x90500080 0x40>;
> +		interrupts = <24 0>;
> +		#dma-cells = <1>;
> +	};
> diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
> index f238cfd..f4ed3a9 100644
> --- a/drivers/dma/Kconfig
> +++ b/drivers/dma/Kconfig
> @@ -318,6 +318,13 @@ config K3_DMA
>  	  Support the DMA engine for Hisilicon K3 platform
>  	  devices.
>  
> +config MOXART_DMA
> +	tristate "MOXART DMA support"
> +	depends on ARCH_MOXART
> +	select DMA_ENGINE
> +	help
> +	  Enable support for the MOXA ART SoC DMA controller.
> +
>  config DMA_ENGINE
>  	bool
>  
> diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
> index db89035..9ef0916 100644
> --- a/drivers/dma/Makefile
> +++ b/drivers/dma/Makefile
> @@ -41,3 +41,4 @@ obj-$(CONFIG_MMP_PDMA) += mmp_pdma.o
>  obj-$(CONFIG_DMA_JZ4740) += dma-jz4740.o
>  obj-$(CONFIG_TI_CPPI41) += cppi41.o
>  obj-$(CONFIG_K3_DMA) += k3dma.o
> +obj-$(CONFIG_MOXART_DMA) += moxart-dma.o
> diff --git a/drivers/dma/moxart-dma.c b/drivers/dma/moxart-dma.c
> new file mode 100644
> index 0000000..edd6de2
> --- /dev/null
> +++ b/drivers/dma/moxart-dma.c
> @@ -0,0 +1,651 @@
> +/*
> + * MOXA ART SoCs DMA Engine support.
> + *
> + * Copyright (C) 2013 Jonas Jensen
> + *
> + * Jonas Jensen <jonas.jensen@...il.com>
> + *
> + * This file is licensed under the terms of the GNU General Public
> + * License version 2.  This program is licensed "as is" without any
> + * warranty of any kind, whether express or implied.
> + */
> +
> +#include <linux/dmaengine.h>
> +#include <linux/dma-mapping.h>
> +#include <linux/err.h>
> +#include <linux/init.h>
> +#include <linux/interrupt.h>
> +#include <linux/list.h>
> +#include <linux/module.h>
> +#include <linux/platform_device.h>
> +#include <linux/slab.h>
> +#include <linux/spinlock.h>
> +#include <linux/irq.h>
> +#include <linux/of_address.h>
> +#include <linux/of_irq.h>
> +#include <linux/of_dma.h>
> +
> +#include <asm/cacheflush.h>
> +
> +#include "dmaengine.h"
> +
> +#define APB_DMA_MAX_CHANNEL			4
> +
> +#define REG_ADDRESS_SOURCE			0
> +#define REG_ADDRESS_DEST			4
> +#define REG_CYCLES				8
> +#define REG_CTRL				12
> +#define REG_CHAN_SIZE				16
> +
> +#define APB_DMA_ENABLE				0x1
> +#define APB_DMA_FIN_INT_STS			0x2
> +#define APB_DMA_FIN_INT_EN			0x4
> +#define APB_DMA_BURST_MODE			0x8
> +#define APB_DMA_ERR_INT_STS			0x10
> +#define APB_DMA_ERR_INT_EN			0x20
> +
> +/*
> + * unset to select APB source
> + * set to select AHB source
> + */
> +#define APB_DMA_SOURCE_SELECT			0x40
> +
> +/*
> + * unset to select APB destination
> + * set to select AHB destination
> + */
> +#define APB_DMA_DEST_SELECT			0x80
> +
> +#define APB_DMA_SOURCE				0x100
> +#define APB_DMA_SOURCE_MASK			0x700
> +/*
> + * 000: no increment
> + * 001: +1 (busrt=0), +4  (burst=1)
> + * 010: +2 (burst=0), +8  (burst=1)
> + * 011: +4 (burst=0), +16 (burst=1)
> + * 101: -1 (burst=0), -4  (burst=1)
> + * 110: -2 (burst=0), -8  (burst=1)
> + * 111: -4 (burst=0), -16 (burst=1)
> + */
> +#define APB_DMA_SOURCE_INC_0			0
> +#define APB_DMA_SOURCE_INC_1_4			0x100
> +#define APB_DMA_SOURCE_INC_2_8			0x200
> +#define APB_DMA_SOURCE_INC_4_16			0x300
> +#define APB_DMA_SOURCE_DEC_1_4			0x500
> +#define APB_DMA_SOURCE_DEC_2_8			0x600
> +#define APB_DMA_SOURCE_DEC_4_16			0x700
> +
> +#define APB_DMA_DEST				0x1000
> +#define APB_DMA_DEST_MASK			0x7000
> +/*
> + * 000: no increment
> + * 001: +1 (busrt=0), +4  (burst=1)
> + * 010: +2 (burst=0), +8  (burst=1)
> + * 011: +4 (burst=0), +16 (burst=1)
> + * 101: -1 (burst=0), -4  (burst=1)
> + * 110: -2 (burst=0), -8  (burst=1)
> + * 111: -4 (burst=0), -16 (burst=1)
> +*/
> +#define APB_DMA_DEST_INC_0			0
> +#define APB_DMA_DEST_INC_1_4			0x1000
> +#define APB_DMA_DEST_INC_2_8			0x2000
> +#define APB_DMA_DEST_INC_4_16			0x3000
> +#define APB_DMA_DEST_DEC_1_4			0x5000
> +#define APB_DMA_DEST_DEC_2_8			0x6000
> +#define APB_DMA_DEST_DEC_4_16			0x7000
> +
> +/*
> + * request signal select of destination
> + * address for DMA hardware handshake
> + *
> + * the request line number is a property of
> + * the DMA controller itself, e.g. MMC must
> + * always request channels where
> + * dma_slave_config->slave_id == 5
> + *
> + * 0:    no request / grant signal
> + * 1-15: request / grant signal
> + */
> +#define APB_DMA_DEST_REQ_NO			0x10000
> +#define APB_DMA_DEST_REQ_NO_MASK		0xf0000
> +
> +#define APB_DMA_DATA_WIDTH			0x100000
> +#define APB_DMA_DATA_WIDTH_MASK			0x300000
> +/*
> + * data width of transfer
> + * 00: word
> + * 01: half
> + * 10: byte
> + */
> +#define APB_DMA_DATA_WIDTH_4			0
> +#define APB_DMA_DATA_WIDTH_2			0x100000
> +#define APB_DMA_DATA_WIDTH_1			0x200000
> +
> +/*
> + * request signal select of source
> + * address for DMA hardware handshake
> + *
> + * the request line number is a property of
> + * the DMA controller itself, e.g. MMC must
> + * always request channels where
> + * dma_slave_config->slave_id == 5
> + *
> + * 0:    no request / grant signal
> + * 1-15: request / grant signal
> + */
> +#define APB_DMA_SOURCE_REQ_NO			0x1000000
> +#define APB_DMA_SOURCE_REQ_NO_MASK		0xf000000
> +#define APB_DMA_CYCLES_MASK			0x00ffffff
> +
> +struct moxart_dma_chan {
> +	struct dma_chan			chan;
> +	int				ch_num;
> +	bool				allocated;
> +	bool				error;
> +	void __iomem			*base;
> +	struct dma_slave_config		cfg;
> +	struct dma_async_tx_descriptor	tx_desc;
> +	unsigned int			line_reqno;
> +};
> +
> +struct moxart_dma_container {
> +	int				ctlr;
> +	struct dma_device		dma_slave;
> +	struct moxart_dma_chan		slave_chans[APB_DMA_MAX_CHANNEL];
> +	spinlock_t			dma_lock;
> +	struct tasklet_struct		tasklet;
> +};
> +
> +struct moxart_dma_filter_data {
> +	struct moxart_dma_container	*mdc;
> +	struct of_phandle_args		*dma_spec;
> +};
> +
> +static struct device *chan2dev(struct dma_chan *chan)
> +{
> +	return &chan->dev->device;
> +}
> +
> +static inline struct moxart_dma_container
> +*to_dma_container(struct dma_device *d)
> +{
> +	return container_of(d, struct moxart_dma_container, dma_slave);
> +}
> +
> +static inline struct moxart_dma_chan *to_moxart_dma_chan(struct dma_chan *c)
> +{
> +	return container_of(c, struct moxart_dma_chan, chan);
> +}
> +
> +static int moxart_terminate_all(struct dma_chan *chan)
> +{
> +	struct moxart_dma_chan *ch = to_moxart_dma_chan(chan);
> +	struct moxart_dma_container *c = to_dma_container(ch->chan.device);
> +	u32 ctrl;
> +	unsigned long flags;
> +
> +	dev_dbg(chan2dev(chan), "%s: ch=%p\n", __func__, ch);
> +
> +	spin_lock_irqsave(&c->dma_lock, flags);
> +
> +	ctrl = readl(ch->base + REG_CTRL);
> +	ctrl &= ~(APB_DMA_ENABLE | APB_DMA_FIN_INT_EN | APB_DMA_ERR_INT_EN);
> +	writel(ctrl, ch->base + REG_CTRL);
> +
> +	spin_unlock_irqrestore(&c->dma_lock, flags);
> +
> +	return 0;
> +}
> +
> +static int moxart_slave_config(struct dma_chan *chan,
> +			       struct dma_slave_config *cfg)
> +{
> +	struct moxart_dma_chan *mchan = to_moxart_dma_chan(chan);
> +	struct moxart_dma_container *mc = to_dma_container(mchan->chan.device);
> +	u32 ctrl;
> +	unsigned long flags;
> +
> +	spin_lock_irqsave(&mc->dma_lock, flags);
> +
> +	memcpy(&mchan->cfg, cfg, sizeof(mchan->cfg));
> +
> +	ctrl = readl(mchan->base + REG_CTRL);
> +	ctrl |= APB_DMA_BURST_MODE;
> +	ctrl &= ~(APB_DMA_DEST_MASK | APB_DMA_SOURCE_MASK);
> +	ctrl &= ~(APB_DMA_DEST_REQ_NO_MASK | APB_DMA_SOURCE_REQ_NO_MASK);
> +
> +	switch (mchan->cfg.src_addr_width) {
> +	case DMA_SLAVE_BUSWIDTH_1_BYTE:
> +		ctrl |= APB_DMA_DATA_WIDTH_1;
> +		if (mchan->cfg.direction != DMA_MEM_TO_DEV)
> +			ctrl |= APB_DMA_DEST_INC_1_4;
> +		else
> +			ctrl |= APB_DMA_SOURCE_INC_1_4;
> +		break;
> +	case DMA_SLAVE_BUSWIDTH_2_BYTES:
> +		ctrl |= APB_DMA_DATA_WIDTH_2;
> +		if (mchan->cfg.direction != DMA_MEM_TO_DEV)
> +			ctrl |= APB_DMA_DEST_INC_2_8;
> +		else
> +			ctrl |= APB_DMA_SOURCE_INC_2_8;
> +		break;
> +	default:
> +		ctrl &= ~APB_DMA_DATA_WIDTH;
> +		if (mchan->cfg.direction != DMA_MEM_TO_DEV)
> +			ctrl |= APB_DMA_DEST_INC_4_16;
> +		else
> +			ctrl |= APB_DMA_SOURCE_INC_4_16;
> +		break;
> +	}
> +
> +	if (mchan->cfg.direction == DMA_MEM_TO_DEV) {
> +		ctrl &= ~APB_DMA_DEST_SELECT;
> +		ctrl |= APB_DMA_SOURCE_SELECT;
> +		ctrl |= (mchan->line_reqno << 16 &
> +			 APB_DMA_DEST_REQ_NO_MASK);
> +	} else {
> +		ctrl |= APB_DMA_DEST_SELECT;
> +		ctrl &= ~APB_DMA_SOURCE_SELECT;
> +		ctrl |= (mchan->line_reqno << 24 &
> +			 APB_DMA_SOURCE_REQ_NO_MASK);
> +	}
> +
> +	writel(ctrl, mchan->base + REG_CTRL);
> +
> +	spin_unlock_irqrestore(&mc->dma_lock, flags);
> +
> +	return 0;
> +}
> +
> +static int moxart_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
> +			  unsigned long arg)
> +{
> +	int ret = 0;
> +	struct dma_slave_config *config;
> +
> +	switch (cmd) {
> +	case DMA_TERMINATE_ALL:
> +		moxart_terminate_all(chan);
> +		break;
> +	case DMA_SLAVE_CONFIG:
> +		config = (struct dma_slave_config *)arg;
> +		ret = moxart_slave_config(chan, config);
> +		break;
> +	default:
> +		ret = -ENOSYS;
> +	}
> +
> +	return ret;
> +}
> +
> +static dma_cookie_t moxart_tx_submit(struct dma_async_tx_descriptor *tx)
> +{
> +	struct moxart_dma_chan *mchan = to_moxart_dma_chan(tx->chan);
> +	struct moxart_dma_container *mc = to_dma_container(mchan->chan.device);
> +	dma_cookie_t cookie;
> +	u32 ctrl;
> +	unsigned long flags;
> +
> +	dev_dbg(chan2dev(tx->chan), "%s: mchan=%p mchan->ch_num=%u mchan->base=%p\n",
> +		__func__, mchan, mchan->ch_num, mchan->base);
> +
> +	spin_lock_irqsave(&mc->dma_lock, flags);
> +
> +	cookie = dma_cookie_assign(tx);
> +
> +	ctrl = readl(mchan->base + REG_CTRL);
> +	ctrl |= (APB_DMA_FIN_INT_EN | APB_DMA_ERR_INT_EN);
> +	writel(ctrl, mchan->base + REG_CTRL);
> +
> +	spin_unlock_irqrestore(&mc->dma_lock, flags);
> +
> +	return cookie;
> +}
> +
> +static struct dma_async_tx_descriptor
> +*moxart_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
> +		      unsigned int sg_len,
> +		      enum dma_transfer_direction direction,
> +		      unsigned long tx_flags, void *context)
> +{
> +	struct moxart_dma_chan *mchan = to_moxart_dma_chan(chan);
> +	struct moxart_dma_container *mc = to_dma_container(mchan->chan.device);
> +	unsigned long flags;
> +	unsigned int size, adr_width;
> +
> +	spin_lock_irqsave(&mc->dma_lock, flags);
> +
> +	if (direction == DMA_MEM_TO_DEV) {
> +		writel(virt_to_phys((void *)sg_dma_address(&sgl[0])),
> +		       mchan->base + REG_ADDRESS_SOURCE);
> +		writel(mchan->cfg.dst_addr, mchan->base + REG_ADDRESS_DEST);
> +
> +		adr_width = mchan->cfg.src_addr_width;
> +	} else {
> +		writel(mchan->cfg.src_addr, mchan->base + REG_ADDRESS_SOURCE);
> +		writel(virt_to_phys((void *)sg_dma_address(&sgl[0])),
> +		       mchan->base + REG_ADDRESS_DEST);
> +
> +		adr_width = mchan->cfg.dst_addr_width;
this is odd. You are not supposed to write to hardware here. You should store
all info, prepare the descriptor and then write to hw in issue_pending.
> +	}
> +
> +	size = sgl->length >> adr_width;
> +
> +	/*
> +	 * size is 4 on 64 bytes copied, i.e. one cycle copies 16 bytes
> +	 * ( when data_width == APB_DMAB_DATA_WIDTH_4 )
> +	 */
> +	writel(size, mchan->base + REG_CYCLES);
> +
> +	dev_dbg(chan2dev(chan), "%s: set %u DMA cycles (sgl->length=%u adr_width=%u)\n",
> +		__func__, size, sgl->length, adr_width);
> +
> +	dma_async_tx_descriptor_init(&mchan->tx_desc, chan);
> +	mchan->tx_desc.tx_submit = moxart_tx_submit;
> +	mchan->error = 0;
> +
> +	spin_unlock_irqrestore(&mc->dma_lock, flags);
> +
> +	return &mchan->tx_desc;
you dont see to store the descriptor anywhere?
> +}
> +
> +bool moxart_dma_filter_fn(struct dma_chan *chan, void *param)
> +{
> +	struct moxart_dma_filter_data *fdata = param;
> +	struct moxart_dma_chan *mchan = to_moxart_dma_chan(chan);
> +
> +	if (chan->device->dev != fdata->mdc->dma_slave.dev ||
> +	    chan->device->dev->of_node != fdata->dma_spec->np) {
> +		dev_dbg(chan2dev(chan), "device not registered to this DMA engine\n");
> +		return 0;
> +	}
> +
> +	dev_dbg(chan2dev(chan), "%s: mchan=%p line_reqno=%u mchan->ch_num=%u\n",
> +		__func__, mchan, fdata->dma_spec->args[0], mchan->ch_num);
> +
> +	mchan->line_reqno = fdata->dma_spec->args[0];
> +
> +	return 1;
1..? true/false makes more sense.

> +}
> +
> +static struct dma_chan *moxart_of_xlate(struct of_phandle_args *dma_spec,
> +					struct of_dma *ofdma)
> +{
> +	struct moxart_dma_container *mdc = ofdma->of_dma_data;
> +	struct moxart_dma_filter_data fdata = {
> +		.mdc = mdc,
> +	};
> +
> +	if (dma_spec->args_count < 1)
> +		return NULL;
> +
> +	fdata.dma_spec = dma_spec;
> +
> +	return dma_request_channel(mdc->dma_slave.cap_mask,
> +				   moxart_dma_filter_fn, &fdata);
> +}
> +
> +static int moxart_alloc_chan_resources(struct dma_chan *chan)
> +{
> +	struct moxart_dma_chan *mchan = to_moxart_dma_chan(chan);
> +
> +	dev_dbg(chan2dev(chan), "%s: allocating channel #%u\n",
> +		__func__, mchan->ch_num);
> +	mchan->allocated = 1;
> +
> +	return 0;
> +}
> +
> +static void moxart_free_chan_resources(struct dma_chan *chan)
> +{
> +	struct moxart_dma_chan *mchan = to_moxart_dma_chan(chan);
> +
> +	dev_dbg(chan2dev(chan), "%s: freeing channel #%u\n",
> +		__func__, mchan->ch_num);
> +	mchan->allocated = 0;
> +}
> +
> +static void moxart_issue_pending(struct dma_chan *chan)
> +{
> +	struct moxart_dma_chan *mchan = to_moxart_dma_chan(chan);
> +	struct moxart_dma_container *mc = to_dma_container(mchan->chan.device);
> +	u32 ctrl;
> +	unsigned long flags;
> +
> +	dev_dbg(chan2dev(chan), "%s: mchan=%p\n", __func__, mchan);
> +
> +	spin_lock_irqsave(&mc->dma_lock, flags);
> +
> +	ctrl = readl(mchan->base + REG_CTRL);
> +	ctrl |= APB_DMA_ENABLE;
> +	writel(ctrl, mchan->base + REG_CTRL);
what about channel configuration. Also what about the case when the dma channel
is already executing, you need to wait for that!

> +	spin_unlock_irqrestore(&mc->dma_lock, flags);
> +}
> +
> +static enum dma_status moxart_tx_status(struct dma_chan *chan,
> +					dma_cookie_t cookie,
> +					struct dma_tx_state *txs)
> +{
> +	struct moxart_dma_chan *ch = to_moxart_dma_chan(chan);
> +	enum dma_status ret;
> +
> +	ret = (ch->error) ? DMA_ERROR : dma_cookie_status(chan, cookie, txs);
You are not filling the residue for the in flight descriptors.

> +
> +	return ret;
> +}
> +
> +static void moxart_dma_init(struct dma_device *dma, struct device *dev)
> +{
> +	dma->device_prep_slave_sg		= moxart_prep_slave_sg;
> +	dma->device_alloc_chan_resources	= moxart_alloc_chan_resources;
> +	dma->device_free_chan_resources		= moxart_free_chan_resources;
> +	dma->device_issue_pending		= moxart_issue_pending;
> +	dma->device_tx_status			= moxart_tx_status;
> +	dma->device_control			= moxart_control;
> +	dma->dev				= dev;
> +
> +	INIT_LIST_HEAD(&dma->channels);
> +}
> +
> +static void moxart_dma_tasklet(unsigned long data)
> +{
> +	struct moxart_dma_container *mc = (void *)data;
> +	struct moxart_dma_chan *ch = &mc->slave_chans[0];
> +	struct dma_async_tx_descriptor *tx_desc;
> +	unsigned int i;
> +	enum dma_status s;
> +	struct dma_tx_state txs;
> +
> +	pr_debug("%s\n", __func__);
> +
> +	for (i = 0; i < APB_DMA_MAX_CHANNEL; i++, ch++) {
> +		if (ch->allocated) {
> +			tx_desc = &ch->tx_desc;
> +
> +			s = mc->dma_slave.device_tx_status(&ch->chan,
> +							   ch->chan.cookie,
> +							   &txs);
> +
> +			switch (s) {
> +			case DMA_ERROR:
> +				printk_ratelimited("%s: DMA error\n",
> +						   __func__);
no log level here?
> +				break;
> +			case DMA_SUCCESS:
> +				break;
> +			case DMA_IN_PROGRESS:
> +			case DMA_PAUSED:
> +				continue;
> +			}
Its odd actually, the channel status of PAUSE makes sense but I am not sure what
you mean by DMA_SUCCESS or DMA_IN_PROGRESS. These make sense for the
descriptors.
> +
> +			if (tx_desc->callback) {
> +				pr_debug("%s: call callback for ch=%p\n",
> +					 __func__, ch);
> +				tx_desc->callback(tx_desc->callback_param);
> +			}
> +		}
> +	> +}
> +
> +static irqreturn_t moxart_dma_interrupt(int irq, void *devid)
> +{
> +	struct moxart_dma_container *mc = devid;
> +	struct moxart_dma_chan *mchan = &mc->slave_chans[0];
> +	unsigned int i;
> +	u32 ctrl;
> +
> +	pr_debug("%s\n", __func__);
> +
> +	for (i = 0; i < APB_DMA_MAX_CHANNEL; i++, mchan++) {
> +		if (mchan->allocated) {
> +			ctrl = readl(mchan->base + REG_CTRL);
> +			pr_debug("%s: ctrl=%x\n", __func__, ctrl);
> +
> +			if (ctrl & APB_DMA_FIN_INT_STS) {
> +				ctrl &= ~APB_DMA_FIN_INT_STS;
> +				dma_cookie_complete(&mchan->tx_desc);
> +			}
> +			if (ctrl & APB_DMA_ERR_INT_STS) {
> +				ctrl &= ~APB_DMA_ERR_INT_STS;
> +				mchan->error = 1;
> +			}
> +			/*
> +			 * bits must be cleared here, this function
> +			 * called in a loop if moved to tasklet
> +			 */
> +			writel(ctrl, mchan->base + REG_CTRL);
> +
> +			tasklet_schedule(&mc->tasklet);
> +		}
> +	}
> +
> +	return IRQ_HANDLED;
> +}
I think you have implemnted that there will be _only_ one descriptor active and
submitted at any point of time.. IMO this shouldnt be done, you cna easily
implement a better way by amnaging multiple transactions in the driver.

Also see the virt-dma layer, using that will help you managing the descriptors
and lists for managing the descriptors

--
~Vinod
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists