[PATCH] media: stm32: dcmi: create a dma scatterlist based on DMA max_sg_burst value

From: Alain Volmat
Date: Mon Jan 10 2022 - 05:38:53 EST


Prior to submitting a transfer to the DMA, the client should first check
the capabilities of the DMA channel in term of maximum of each segment.
This is given by the max_sg_burst value reported by dma_get_slave_caps API.
Based on that, if the transfer is larger than what can be handled by the
DMA channel, we split the transfer into several scatterlist elements.

Signed-off-by: Alain Volmat <alain.volmat@xxxxxxxxxxx>
---
drivers/media/platform/stm32/stm32-dcmi.c | 47 ++++++++++++++++++-----
1 file changed, 37 insertions(+), 10 deletions(-)

diff --git a/drivers/media/platform/stm32/stm32-dcmi.c b/drivers/media/platform/stm32/stm32-dcmi.c
index e1b17c05229c..ee170e999a88 100644
--- a/drivers/media/platform/stm32/stm32-dcmi.c
+++ b/drivers/media/platform/stm32/stm32-dcmi.c
@@ -113,7 +113,7 @@ struct dcmi_framesize {
struct dcmi_buf {
struct vb2_v4l2_buffer vb;
bool prepared;
- dma_addr_t paddr;
+ struct sg_table sgt;
size_t size;
struct list_head list;
};
@@ -157,6 +157,7 @@ struct stm32_dcmi {
enum state state;
struct dma_chan *dma_chan;
dma_cookie_t dma_cookie;
+ u32 dma_max_burst;
u32 misr;
int errors_count;
int overrun_count;
@@ -326,13 +327,11 @@ static int dcmi_start_dma(struct stm32_dcmi *dcmi,
mutex_lock(&dcmi->dma_lock);

/* Prepare a DMA transaction */
- desc = dmaengine_prep_slave_single(dcmi->dma_chan, buf->paddr,
- buf->size,
+ desc = dmaengine_prep_slave_sg(dcmi->dma_chan, buf->sgt.sgl, buf->sgt.nents,
DMA_DEV_TO_MEM,
DMA_PREP_INTERRUPT);
if (!desc) {
- dev_err(dcmi->dev, "%s: DMA dmaengine_prep_slave_single failed for buffer phy=%pad size=%zu\n",
- __func__, &buf->paddr, buf->size);
+ dev_err(dcmi->dev, "%s: DMA dmaengine_prep_slave_sg failed\n", __func__);
mutex_unlock(&dcmi->dma_lock);
return -EINVAL;
}
@@ -524,6 +523,10 @@ static int dcmi_buf_prepare(struct vb2_buffer *vb)
struct vb2_v4l2_buffer *vbuf = to_vb2_v4l2_buffer(vb);
struct dcmi_buf *buf = container_of(vbuf, struct dcmi_buf, vb);
unsigned long size;
+ unsigned int num_sgs = 1;
+ dma_addr_t dma_buf;
+ struct scatterlist *sg;
+ int i, ret;

size = dcmi->fmt.fmt.pix.sizeimage;

@@ -537,15 +540,33 @@ static int dcmi_buf_prepare(struct vb2_buffer *vb)

if (!buf->prepared) {
/* Get memory addresses */
- buf->paddr =
- vb2_dma_contig_plane_dma_addr(&buf->vb.vb2_buf, 0);
buf->size = vb2_plane_size(&buf->vb.vb2_buf, 0);
- buf->prepared = true;
+ if (buf->size > dcmi->dma_max_burst)
+ num_sgs = DIV_ROUND_UP(buf->size, dcmi->dma_max_burst);

- vb2_set_plane_payload(&buf->vb.vb2_buf, 0, buf->size);
+ ret = sg_alloc_table(&buf->sgt, num_sgs, GFP_ATOMIC);
+ if (ret) {
+ dev_err(dcmi->dev, "sg table alloc failed\n");
+ return ret;
+ }
+
+ dma_buf = vb2_dma_contig_plane_dma_addr(&buf->vb.vb2_buf, 0);

dev_dbg(dcmi->dev, "buffer[%d] phy=%pad size=%zu\n",
- vb->index, &buf->paddr, buf->size);
+ vb->index, &dma_buf, buf->size);
+
+ for_each_sg(buf->sgt.sgl, sg, num_sgs, i) {
+ size_t bytes = min_t(size_t, size, dcmi->dma_max_burst);
+
+ sg_dma_address(sg) = dma_buf;
+ sg_dma_len(sg) = bytes;
+ dma_buf += bytes;
+ size -= bytes;
+ }
+
+ buf->prepared = true;
+
+ vb2_set_plane_payload(&buf->vb.vb2_buf, 0, buf->size);
}

return 0;
@@ -1866,6 +1887,7 @@ static int dcmi_probe(struct platform_device *pdev)
struct stm32_dcmi *dcmi;
struct vb2_queue *q;
struct dma_chan *chan;
+ struct dma_slave_caps caps;
struct clk *mclk;
int irq;
int ret = 0;
@@ -1953,6 +1975,11 @@ static int dcmi_probe(struct platform_device *pdev)
return ret;
}

+ dcmi->dma_max_burst = UINT_MAX;
+ ret = dma_get_slave_caps(chan, &caps);
+ if (!ret && caps.max_sg_burst)
+ dcmi->dma_max_burst = caps.max_sg_burst * DMA_SLAVE_BUSWIDTH_4_BYTES;
+
spin_lock_init(&dcmi->irqlock);
mutex_init(&dcmi->lock);
mutex_init(&dcmi->dma_lock);
--
2.25.1