dw_dmac: store direction in the custom channel structure

The driver requires a custom slave configuration to be present to be able to make the slave transfers. Nevertheless, in some cases we need only the request line as an additional information to the generic slave configuration. The request line is provided by slave_id parameter of the dma_slave_config structure. That's why the custom slave configuration could be optional for such cases.

Signed-off-by Andy Shevchenko <[email protected]>

Currently the direction value is come from the generic slave configuration

structure and explicitly as a preparation function parameter. The first one is kinda obsoleted. Thus, we have to store the value passed to the preparation function somewhere in our structures to be able to use it later. The best candidate to provide the storage is a custom channel structure. Until now we still keep and check the direction field of the slave config structure as well.

Signed-off-by Andy Shevchenko <[email protected]>

---
 drivers/dma/dw_dmac.c      |   10 ++++++++--
 drivers/dma/dw_dmac_regs.h |    2 ++
 2 files changed, 10 insertions(+), 2 deletions(-)

drivers/dma/dw_dmac.c +0 -0

index 5e2c4dc..58c6d3c 100644

drivers/dma/dw_dmac.c +6 -0

@@ -807,6 +807,8 @@ dwc_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
        if (unlikely(!dws || !sg_len))
                return NULL;
 
+       dwc->direction = direction;
+
        prev = first = NULL;
 
        switch (direction) {

@@ -983,6 +985,7 @@ set_runtime_config(struct dma_chan *chan, struct dma_slave_config *sconfig)
                return -EINVAL;
 
        memcpy(&dwc->dma_sconfig, sconfig, sizeof(*sconfig));
+       dwc->direction = sconfig->direction;
 
        convert_burst(&dwc->dma_sconfig.src_maxburst);
        convert_burst(&dwc->dma_sconfig.dst_maxburst);

@@ -1339,6 +1342,8 @@ struct dw_cyclic_desc *dw_dma_cyclic_prep(struct dma_chan *chan,
        if (unlikely(!is_slave_xfer(direction)))
                goto out_err;
 
+       dwc->direction = direction;
+
        if (direction == DMA_MEM_TO_DEV)
                reg_width = __ffs(sconfig->dst_addr_width);
        else

@@ -1713,6 +1718,7 @@ static int __devinit dw_probe(struct platform_device *pdev)
                channel_clear_bit(dw, CH_EN, dwc->mask);
 
                dwc->dw = dw;
+               dwc->direction = DMA_TRANS_NONE;
 
                /* hardware configuration */
                if (autocfg) {

drivers/dma/dw_dmac_regs.h +1 -0

index 8881e9b..362aba1 100644
@@ -9,6 +9,7 @@
  * published by the Free Software Foundation.
  */
 
+#include <linux/dmaengine.h>
 #include <linux/dw_dmac.h>
 
 #define DW_DMA_MAX_NR_CHANNELS 8

Quoted 1 line · Andy Shevchenko

Currently the direction value is come from the generic slave configuration

s/is come/comes

Quoted 17 lines

structure and explicitly as a preparation function parameter. The first one is kinda obsoleted. Thus, we have to store the value passed to the preparation function somewhere in our structures to be able to use it later. The best candidate to provide the storage is a custom channel structure. Until now we still keep and check the direction field of the slave config structure as well.

Signed-off-by Andy Shevchenko <[email protected]>

---
 drivers/dma/dw_dmac.c      |   10 ++++++++--
 drivers/dma/dw_dmac_regs.h |    2 ++
 2 files changed, 10 insertions(+), 2 deletions(-)

drivers/dma/dw_dmac.c +0 -0

index 5e2c4dc..58c6d3c 100644

drivers/dma/dw_dmac.c static void dwc_initialize(struct dw_dma_chan *dwc) · line 185

Quoted 10 lines

drivers/dma/dw_dmac.c +2 -2

@@ -178,9 +178,9 @@ static void dwc_initialize(struct dw_dma_chan *dwc)
                cfghi = dws->cfg_hi;
                cfglo |= dws->cfg_lo & ~DWC_CFGL_CH_PRIOR_MASK;
        } else {
-               if (dwc->dma_sconfig.direction == DMA_MEM_TO_DEV)
+               if (dwc->direction == DMA_MEM_TO_DEV)
                        cfghi = DWC_CFGH_DST_PER(dwc->dma_sconfig.slave_id);
-               else if (dwc->dma_sconfig.direction == DMA_DEV_TO_MEM)
+               else if (dwc->direction == DMA_DEV_TO_MEM)
                        cfghi = DWC_CFGH_SRC_PER(dwc->dma_sconfig.slave_id);
        }

I have a doubt here. This routine is called both for memcpy and slave transfers. Earlier if (dws), code was used to check slave/non-slave transfer but with your earlier patch, for if (!dws) case too, there can be slave transfers based on direction.

Consider a situation. A channel is used for slave transfers and have direction set
to DMA_DEV_TO_MEM. Now, channel is freed and used for memcpy. Now the direction value is retained as we haven't updated it. So, is this a potential bug?