Re: [PATCH v3] dma: imx-sdma: add support for sdma memory copy
From: Robin Gong
Date: Wed May 21 2014 - 22:07:32 EST
On Wed, May 21, 2014 at 04:26:52PM +0530, Vinod Koul wrote:
> On Tue, May 06, 2014 at 10:12:48AM +0800, Robin Gong wrote:
> > add "device_prep_dma_memcpy" and "device_prep_dma_sg" for memory copy by sdma.
> >
> > Signed-off-by: Robin Gong <b38343@xxxxxxxxxxxxx>
> >
> > ---
> > change:
> > --v3:
> > 1. remove redundant check for bus width
> >
> > --v2:
> > 1. correct some printk format, such as %pad for dma_addr_t
> > 2. split duplicated code in prep_dma_memcpy and prep_dma_sg to make code clean.
> >
> > Signed-off-by: Robin Gong <b38343@xxxxxxxxxxxxx>
> > ---
> > drivers/dma/imx-sdma.c | 250 ++++++++++++++++++++++++++++++++++++-----------
> > 1 files changed, 191 insertions(+), 59 deletions(-)
> >
> > diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
> > index 4e79183..9abc164 100644
> > --- a/drivers/dma/imx-sdma.c
> > +++ b/drivers/dma/imx-sdma.c
> > @@ -229,6 +229,7 @@ struct sdma_context_data {
> > } __attribute__ ((packed));
> >
> > #define NUM_BD (int)(PAGE_SIZE / sizeof(struct sdma_buffer_descriptor))
> > +#define SDMA_BD_MAX_CNT (0xfffc) /* align with 4 bytes */
> >
> > struct sdma_engine;
> >
> > @@ -260,6 +261,7 @@ struct sdma_channel {
> > unsigned int pc_from_device, pc_to_device;
> > unsigned long flags;
> > dma_addr_t per_address;
> > + unsigned int pc_to_pc;
> > unsigned long event_mask[2];
> > unsigned long watermark_level;
> > u32 shp_addr, per_addr;
> > @@ -694,6 +696,7 @@ static void sdma_get_pc(struct sdma_channel *sdmac,
> >
> > sdmac->pc_from_device = 0;
> > sdmac->pc_to_device = 0;
> > + sdmac->pc_to_pc = 0;
> >
> > switch (peripheral_type) {
> > case IMX_DMATYPE_MEMORY:
> > @@ -763,6 +766,7 @@ static void sdma_get_pc(struct sdma_channel *sdmac,
> >
> > sdmac->pc_from_device = per_2_emi;
> > sdmac->pc_to_device = emi_2_per;
> > + sdmac->pc_to_pc = emi_2_emi;
> > }
> >
> > static int sdma_load_context(struct sdma_channel *sdmac)
> > @@ -775,11 +779,12 @@ static int sdma_load_context(struct sdma_channel *sdmac)
> > int ret;
> > unsigned long flags;
> >
> > - if (sdmac->direction == DMA_DEV_TO_MEM) {
> > + if (sdmac->direction == DMA_DEV_TO_MEM)
> > load_address = sdmac->pc_from_device;
> > - } else {
> > + else if (sdmac->direction == DMA_MEM_TO_MEM)
> > + load_address = sdmac->pc_to_pc;
> > + else
> else is not proper here as enum defines one more field. So else-if would be apt.
>
I only keep the same logic as before here...but I agree with you, will split it into
another patch.
> > load_address = sdmac->pc_to_device;
> > - }
> >
> > if (load_address < 0)
> > return load_address;
> > @@ -1010,99 +1015,203 @@ static void sdma_free_chan_resources(struct dma_chan *chan)
> > clk_disable(sdma->clk_ahb);
> > }
> >
> > -static struct dma_async_tx_descriptor *sdma_prep_slave_sg(
> > - struct dma_chan *chan, struct scatterlist *sgl,
> > - unsigned int sg_len, enum dma_transfer_direction direction,
> > - unsigned long flags, void *context)
> > +static int sdma_transfer_init(struct sdma_channel *sdmac,
> > + enum dma_transfer_direction direction)
> > +{
> > + int ret = 0;
> > +
> > + sdmac->status = DMA_IN_PROGRESS;
> > + sdmac->buf_tail = 0;
> > + sdmac->flags = 0;
> > + sdmac->direction = direction;
> > +
> > + ret = sdma_load_context(sdmac);
> > + if (ret)
> > + return ret;
> > +
> > + sdmac->chn_count = 0;
> > +
> > + return ret;
> > +}
> > +
> > +static int check_bd_buswidth(struct sdma_buffer_descriptor *bd,
> > + struct sdma_channel *sdmac, int count,
> > + dma_addr_t dma_dst, dma_addr_t dma_src)
> > +{
> > + int ret = 0;
> > +
> > + switch (sdmac->word_size) {
> > + case DMA_SLAVE_BUSWIDTH_4_BYTES:
> > + bd->mode.command = 0;
> > + if ((count | dma_dst | dma_src) & 3)
> > + ret = -EINVAL;
> > + break;
> > + case DMA_SLAVE_BUSWIDTH_2_BYTES:
> > + bd->mode.command = 2;
> > + if ((count | dma_dst | dma_src) & 1)
> > + ret = -EINVAL;
> > + break;
> > + case DMA_SLAVE_BUSWIDTH_1_BYTE:
> > + bd->mode.command = 1;
> > + break;
> > + default:
> > + return -EINVAL;
> > + }
> > +
> > + return ret;
> > +}
> This looks like a code reorg change, perhpas another patch?
>
ok, will split it into another patch.
> > +
> > +static struct dma_async_tx_descriptor *sdma_prep_memcpy(
> > + struct dma_chan *chan, dma_addr_t dma_dst,
> > + dma_addr_t dma_src, size_t len, unsigned long flags)
> > {
> > struct sdma_channel *sdmac = to_sdma_chan(chan);
> > struct sdma_engine *sdma = sdmac->sdma;
> > - int ret, i, count;
> > int channel = sdmac->channel;
> > - struct scatterlist *sg;
> > + size_t count;
> > + int i = 0, param;
> > + struct sdma_buffer_descriptor *bd;
> >
> > - if (sdmac->status == DMA_IN_PROGRESS)
> > + if (!chan || !len || sdmac->status == DMA_IN_PROGRESS)
> > return NULL;
> > - sdmac->status = DMA_IN_PROGRESS;
> >
> > - sdmac->flags = 0;
> > -
> > - sdmac->buf_tail = 0;
> > + if (len >= NUM_BD * SDMA_BD_MAX_CNT) {
> > + dev_err(sdma->dev, "channel%d: maximum bytes exceeded:%zu > %d\n"
> > + , channel, len, NUM_BD * SDMA_BD_MAX_CNT);
> > + goto err_out;
> > + }
> >
> > - dev_dbg(sdma->dev, "setting up %d entries for channel %d.\n",
> > - sg_len, channel);
> > + dev_dbg(sdma->dev, "memcpy: %pad->%pad, len=%zu, channel=%d.\n",
> > + &dma_src, &dma_dst, len, channel);
> >
> > - sdmac->direction = direction;
> > - ret = sdma_load_context(sdmac);
> > - if (ret)
> > + if (sdma_transfer_init(sdmac, DMA_MEM_TO_MEM))
> > goto err_out;
> >
> > - if (sg_len > NUM_BD) {
> > + do {
> > + count = min_t(size_t, len, SDMA_BD_MAX_CNT);
> > + bd = &sdmac->bd[i];
> > + bd->buffer_addr = dma_src;
> > + bd->ext_buffer_addr = dma_dst;
> > + bd->mode.count = count;
> > +
> > + if (check_bd_buswidth(bd, sdmac, count, dma_dst, dma_src))
> > + goto err_out;
> > +
> > + dma_src += count;
> > + dma_dst += count;
> > + len -= count;
> > + i++;
> > +
> > + param = BD_DONE | BD_EXTD | BD_CONT;
> > + /* last bd */
> > + if (!len) {
> > + param |= BD_INTR;
> > + param |= BD_LAST;
> > + param &= ~BD_CONT;
> > + }
> > +
> > + dev_dbg(sdma->dev, "entry %d: count: %d dma: 0x%u %s%s\n",
> > + i, count, bd->buffer_addr,
> > + param & BD_WRAP ? "wrap" : "",
> > + param & BD_INTR ? " intr" : "");
> > +
> > + bd->mode.status = param;
> > + sdmac->chn_count += count;
> > + } while (len);
> > +
> > + sdmac->num_bd = i;
> > + sdma->channel_control[channel].current_bd_ptr = sdmac->bd_phys;
> > +
> > + return &sdmac->desc;
> > +err_out:
> > + sdmac->status = DMA_ERROR;
> > + return NULL;
> > +}
> > +
> > +/*
> > + * Please ensure dst_nents no smaller than src_nents , also every sg_len of
> > + * dst_sg node no smaller than src_sg. To simply things, please use the same
> > + * size of dst_sg as src_sg.
> > + */
> > +static struct dma_async_tx_descriptor *sdma_prep_sg(
> > + struct dma_chan *chan,
> > + struct scatterlist *dst_sg, unsigned int dst_nents,
> > + struct scatterlist *src_sg, unsigned int src_nents,
> > + enum dma_transfer_direction direction)
> > +{
> > + struct sdma_channel *sdmac = to_sdma_chan(chan);
> > + struct sdma_engine *sdma = sdmac->sdma;
> > + int ret, i, count;
> > + int channel = sdmac->channel;
> > + struct scatterlist *sg_src = src_sg, *sg_dst = dst_sg;
> > +
> > + if (sdmac->status == DMA_IN_PROGRESS)
> > + return NULL;
> > +
> > + dev_dbg(sdma->dev, "setting up %d entries for channel %d.\n",
> > + src_nents, channel);
> > +
> > + if (src_nents > NUM_BD) {
> > dev_err(sdma->dev, "SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
> > - channel, sg_len, NUM_BD);
> > - ret = -EINVAL;
> > + channel, src_nents, NUM_BD);
> > goto err_out;
> > }
> >
> > - sdmac->chn_count = 0;
> > - for_each_sg(sgl, sg, sg_len, i) {
> > + if (sdma_transfer_init(sdmac, direction))
> > + goto err_out;
> > +
> > + for_each_sg(src_sg, sg_src, src_nents, i) {
> > struct sdma_buffer_descriptor *bd = &sdmac->bd[i];
> > int param;
> >
> > - bd->buffer_addr = sg->dma_address;
> > + bd->buffer_addr = sg_src->dma_address;
> >
> > - count = sg_dma_len(sg);
> > + if (direction == DMA_MEM_TO_MEM) {
> > + BUG_ON(!sg_dst);
> why should this be direction specfic?
>
we only need use ext_buffer_addr in case of MEM_TO_MEM.
> > + bd->ext_buffer_addr = sg_dst->dma_address;
> > + }
> > +
> > + count = sg_dma_len(sg_src);
> >
> > - if (count > 0xffff) {
> > + if (count > SDMA_BD_MAX_CNT) {
> > dev_err(sdma->dev, "SDMA channel %d: maximum bytes for sg entry exceeded: %d > %d\n",
> > - channel, count, 0xffff);
> > - ret = -EINVAL;
> > + channel, count, SDMA_BD_MAX_CNT);
> > goto err_out;
> > }
> >
> > bd->mode.count = count;
> > sdmac->chn_count += count;
> >
> > - if (sdmac->word_size > DMA_SLAVE_BUSWIDTH_4_BYTES) {
> > - ret = -EINVAL;
> > - goto err_out;
> > - }
> > + if (direction == DMA_MEM_TO_MEM)
> > + ret = check_bd_buswidth(bd, sdmac, count,
> > + sg_dst->dma_address,
> > + sg_src->dma_address);
> > + else
> > + ret = check_bd_buswidth(bd, sdmac, count, 0,
> > + sg_src->dma_address);
> >
> > - switch (sdmac->word_size) {
> > - case DMA_SLAVE_BUSWIDTH_4_BYTES:
> > - bd->mode.command = 0;
> > - if (count & 3 || sg->dma_address & 3)
> > - return NULL;
> > - break;
> > - case DMA_SLAVE_BUSWIDTH_2_BYTES:
> > - bd->mode.command = 2;
> > - if (count & 1 || sg->dma_address & 1)
> > - return NULL;
> > - break;
> > - case DMA_SLAVE_BUSWIDTH_1_BYTE:
> > - bd->mode.command = 1;
> > - break;
> > - default:
> > - return NULL;
> > - }
> > + if (ret)
> > + goto err_out;
> >
> > param = BD_DONE | BD_EXTD | BD_CONT;
> >
> > - if (i + 1 == sg_len) {
> > + if (i + 1 == src_nents) {
> > param |= BD_INTR;
> > param |= BD_LAST;
> > param &= ~BD_CONT;
> > }
> >
> > - dev_dbg(sdma->dev, "entry %d: count: %d dma: %#llx %s%s\n",
> > - i, count, (u64)sg->dma_address,
> > + dev_dbg(sdma->dev, "entry %d: count: %d dma: %pad %s%s\n",
> > + i, count, &sg_src->dma_address,
> > param & BD_WRAP ? "wrap" : "",
> > param & BD_INTR ? " intr" : "");
> >
> > bd->mode.status = param;
> > + if (direction == DMA_MEM_TO_MEM)
> > + sg_dst = sg_next(sg_dst);
> > }
> >
> > - sdmac->num_bd = sg_len;
> > + sdmac->num_bd = src_nents;
> > sdma->channel_control[channel].current_bd_ptr = sdmac->bd_phys;
> >
> > return &sdmac->desc;
> > @@ -1111,6 +1220,24 @@ err_out:
> > return NULL;
> > }
> >
> > +static struct dma_async_tx_descriptor *sdma_prep_memcpy_sg(
> > + struct dma_chan *chan,
> > + struct scatterlist *dst_sg, unsigned int dst_nents,
> > + struct scatterlist *src_sg, unsigned int src_nents,
> > + unsigned long flags)
> > +{
> > + return sdma_prep_sg(chan, dst_sg, dst_nents, src_sg, src_nents,
> > + DMA_MEM_TO_MEM);
> > +}
> > +
> > +static struct dma_async_tx_descriptor *sdma_prep_slave_sg(
> > + struct dma_chan *chan, struct scatterlist *sgl,
> > + unsigned int sg_len, enum dma_transfer_direction direction,
> > + unsigned long flags, void *context)
> > +{
> > + return sdma_prep_sg(chan, NULL, 0, sgl, sg_len, direction);
> > +}
> > +
> > static struct dma_async_tx_descriptor *sdma_prep_dma_cyclic(
> > struct dma_chan *chan, dma_addr_t dma_addr, size_t buf_len,
> > size_t period_len, enum dma_transfer_direction direction,
> > @@ -1143,9 +1270,9 @@ static struct dma_async_tx_descriptor *sdma_prep_dma_cyclic(
> > goto err_out;
> > }
> >
> > - if (period_len > 0xffff) {
> > - dev_err(sdma->dev, "SDMA channel %d: maximum period size exceeded: %d > %d\n",
> > - channel, period_len, 0xffff);
> > + if (period_len > SDMA_BD_MAX_CNT) {
> > + dev_err(sdma->dev, "SDMA channel %d: maximum period size exceeded: %zu > %d\n",
> > + channel, period_len, SDMA_BD_MAX_CNT);
> this macro change should have been another patch too...
>
>
ok, will split into another patch too.
> --
> ~Vinod
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/