[PATCH v2 0/3] dmaengine: ti-dma-crossbar: channel reserving and edma3-tpcc support

[PATCH v2 0/3] dmaengine: ti-dma-crossbar: channel reserving and edma3-tpcc support

[Peter Ujfalusi]

Hi,

Changes since v1:
- Fixed issue introduced by the bitops patch: wrong error check, also switch to
  use find_first_zero_bit() instead of find_next_zero_bit()

Cover letter:

This series depends on the eDMA work I have done, which has been now applied:
https://lkml.org/lkml/2015/10/16/64

DRA7 family of chips have both sDMA and eDMA. Currently only sDMA can be used
becasue the old driver stack for eDMA did not allowed integration w/o hacks.

Due to the nature of eDMA the crossbar needs to know which eDMA events it can
use to map incoming events towards the eDMA. In eDMA a channel is wired to be
used with one specific event. For example eDMA event 14 can only be handled by
eDMA channel 14.
The eDMA itself can be shared by different processors in the system (ARM, DSP,
etc) and since ARM/Linux is the master we need to know which channels are used
by other cores. Also we need to mask out channels used for memcpy from the
events we use for HW triggers.

Regards,
Peter
---
Peter Ujfalusi (3):
  dmaengine: ti-dma-crossbar: dra7: Use bitops instead of idr
  dmaengine: ti-dma-crossbar: dra7: Support for reserving DMA event
    ranges
  dmaengine: ti-dma-crossbar: dra7: Support for eDMA with new bindings

 .../devicetree/bindings/dma/ti-dma-crossbar.txt    |  6 ++
 drivers/dma/ti-dma-crossbar.c                      | 81 +++++++++++++++++++---
 2 files changed, 76 insertions(+), 11 deletions(-)

-- 
2.6.1

[PATCH v2 1/3] dmaengine: ti-dma-crossbar: dra7: Use bitops instead of idr

[Peter Ujfalusi]

The use of idr was nice, but it was a bit heavy and we did not need the
features it provides. Using simple bitmap to track allocated DMA channels
is adequate here and it will be easier to add support for reserving
channels later on.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
---
 drivers/dma/ti-dma-crossbar.c | 30 +++++++++++++++++++++++-------
 1 file changed, 23 insertions(+), 7 deletions(-)

diff --git a/drivers/dma/ti-dma-crossbar.c b/drivers/dma/ti-dma-crossbar.c
index a415edbe61b1..463ad6b18a32 100644
--- a/drivers/dma/ti-dma-crossbar.c
+++ b/drivers/dma/ti-dma-crossbar.c
@@ -12,7 +12,6 @@
 #include <linux/init.h>
 #include <linux/list.h>
 #include <linux/io.h>
-#include <linux/idr.h>
 #include <linux/of_address.h>
 #include <linux/of_device.h>
 #include <linux/of_dma.h>
@@ -198,7 +197,8 @@ struct ti_dra7_xbar_data {
 	void __iomem *iomem;
 
 	struct dma_router dmarouter;
-	struct idr map_idr;
+	struct mutex mutex;
+	unsigned long *dma_inuse;
 
 	u16 safe_val; /* Value to rest the crossbar lines */
 	u32 xbar_requests; /* number of DMA requests connected to XBAR */
@@ -225,7 +225,9 @@ static void ti_dra7_xbar_free(struct device *dev, void *route_data)
 		map->xbar_in, map->xbar_out);
 
 	ti_dra7_xbar_write(xbar->iomem, map->xbar_out, xbar->safe_val);
-	idr_remove(&xbar->map_idr, map->xbar_out);
+	mutex_lock(&xbar->mutex);
+	clear_bit(map->xbar_out, xbar->dma_inuse);
+	mutex_unlock(&xbar->mutex);
 	kfree(map);
 }
 
@@ -255,8 +257,17 @@ static void *ti_dra7_xbar_route_allocate(struct of_phandle_args *dma_spec,
 		return ERR_PTR(-ENOMEM);
 	}
 
-	map->xbar_out = idr_alloc(&xbar->map_idr, NULL, 0, xbar->dma_requests,
-				  GFP_KERNEL);
+	mutex_lock(&xbar->mutex);
+	map->xbar_out = find_first_zero_bit(xbar->dma_inuse,
+					    xbar->dma_requests);
+	mutex_unlock(&xbar->mutex);
+	if (map->xbar_out == xbar->dma_requests) {
+		dev_err(&pdev->dev, "Run out of free DMA requests\n");
+		kfree(map);
+		return ERR_PTR(-ENOMEM);
+	}
+	set_bit(map->xbar_out, xbar->dma_inuse);
+
 	map->xbar_in = (u16)dma_spec->args[0];
 
 	dma_spec->args[0] = map->xbar_out + xbar->dma_offset;
@@ -299,8 +310,6 @@ static int ti_dra7_xbar_probe(struct platform_device *pdev)
 	if (!xbar)
 		return -ENOMEM;
 
-	idr_init(&xbar->map_idr);
-
 	dma_node = of_parse_phandle(node, "dma-masters", 0);
 	if (!dma_node) {
 		dev_err(&pdev->dev, "Can't get DMA master node\n");
@@ -322,6 +331,12 @@ static int ti_dra7_xbar_probe(struct platform_device *pdev)
 	}
 	of_node_put(dma_node);
 
+	xbar->dma_inuse = devm_kcalloc(&pdev->dev,
+				       BITS_TO_LONGS(xbar->dma_requests),
+				       sizeof(unsigned long), GFP_KERNEL);
+	if (!xbar->dma_inuse)
+		return -ENOMEM;
+
 	if (of_property_read_u32(node, "dma-requests", &xbar->xbar_requests)) {
 		dev_info(&pdev->dev,
 			 "Missing XBAR input information, using %u.\n",
@@ -343,6 +358,7 @@ static int ti_dra7_xbar_probe(struct platform_device *pdev)
 	xbar->dmarouter.route_free = ti_dra7_xbar_free;
 	xbar->dma_offset = (u32)match->data;
 
+	mutex_init(&xbar->mutex);
 	platform_set_drvdata(pdev, xbar);
 
 	/* Reset the crossbar */
-- 
2.6.1

[PATCH v2 2/3] dmaengine: ti-dma-crossbar: dra7: Support for reserving DMA event ranges

[Peter Ujfalusi]

In eDMA the events are directly mapped to a DMA channel (for example DMA
event 14 can only be handled by DMA channel 14). If the memcpy is enabled
on the eDMA, there is a possibility that the crossbar driver would assign
DMA event number already allocated in eDMA for memcpy. Furthermore the
eDMA can be shared with DSP in which case the crossbar driver should also
avoid mapping xbar events to DSP used event numbers (or channels).

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
---
 .../devicetree/bindings/dma/ti-dma-crossbar.txt    |  6 +++
 drivers/dma/ti-dma-crossbar.c                      | 47 ++++++++++++++++++++--
 2 files changed, 49 insertions(+), 4 deletions(-)

diff --git a/Documentation/devicetree/bindings/dma/ti-dma-crossbar.txt b/Documentation/devicetree/bindings/dma/ti-dma-crossbar.txt
index b152a75dceae..aead5869a28d 100644
--- a/Documentation/devicetree/bindings/dma/ti-dma-crossbar.txt
+++ b/Documentation/devicetree/bindings/dma/ti-dma-crossbar.txt
@@ -14,6 +14,10 @@ The DMA controller node need to have the following poroperties:
 
 Optional properties:
 - ti,dma-safe-map: Safe routing value for unused request lines
+- ti,reserved-dma-request-ranges: DMA request ranges which should not be used
+		when mapping xbar input to DMA request, they are either
+		allocated to be used by for example the DSP or they are used as
+		memcpy channels in eDMA.
 
 Notes:
 When requesting channel via ti,dra7-dma-crossbar, the DMA clinet must request
@@ -46,6 +50,8 @@ sdma_xbar: dma-router@4a002b78 {
 	#dma-cells = <1>;
 	dma-requests = <205>;
 	ti,dma-safe-map = <0>;
+	/* Protect the sDMA request ranges: 10-14 and 100-126 */
+	ti,reserved-dma-request-ranges = <10 5>, <100 27>;
 	dma-masters = <&sdma>;
 };
 
diff --git a/drivers/dma/ti-dma-crossbar.c b/drivers/dma/ti-dma-crossbar.c
index 463ad6b18a32..dac7ae06cc58 100644
--- a/drivers/dma/ti-dma-crossbar.c
+++ b/drivers/dma/ti-dma-crossbar.c
@@ -292,14 +292,22 @@ static const struct of_device_id ti_dra7_master_match[] = {
 	{},
 };
 
+static inline void ti_dra7_xbar_reserve(int offset, int len, unsigned long *p)
+{
+	for (; len > 0; len--)
+		clear_bit(offset + (len - 1), p);
+}
+
 static int ti_dra7_xbar_probe(struct platform_device *pdev)
 {
 	struct device_node *node = pdev->dev.of_node;
 	const struct of_device_id *match;
 	struct device_node *dma_node;
 	struct ti_dra7_xbar_data *xbar;
+	struct property *prop;
 	struct resource *res;
 	u32 safe_val;
+	size_t sz;
 	void __iomem *iomem;
 	int i, ret;
 
@@ -347,6 +355,33 @@ static int ti_dra7_xbar_probe(struct platform_device *pdev)
 	if (!of_property_read_u32(node, "ti,dma-safe-map", &safe_val))
 		xbar->safe_val = (u16)safe_val;
 
+
+	prop = of_find_property(node, "ti,reserved-dma-request-ranges", &sz);
+	if (prop) {
+		const char pname[] = "ti,reserved-dma-request-ranges";
+		u32 (*rsv_events)[2];
+		size_t nelm = sz / sizeof(*rsv_events);
+		int i;
+
+		if (!nelm)
+			return -EINVAL;
+
+		rsv_events = kcalloc(nelm, sizeof(*rsv_events), GFP_KERNEL);
+		if (!rsv_events)
+			return -ENOMEM;
+
+		ret = of_property_read_u32_array(node, pname, (u32 *)rsv_events,
+						 nelm * 2);
+		if (ret)
+			return ret;
+
+		for (i = 0; i < nelm; i++) {
+			ti_dra7_xbar_reserve(rsv_events[i][0], rsv_events[i][1],
+					     xbar->dma_inuse);
+		}
+		kfree(rsv_events);
+	}
+
 	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
 	iomem = devm_ioremap_resource(&pdev->dev, res);
 	if (IS_ERR(iomem))
@@ -362,15 +397,19 @@ static int ti_dra7_xbar_probe(struct platform_device *pdev)
 	platform_set_drvdata(pdev, xbar);
 
 	/* Reset the crossbar */
-	for (i = 0; i < xbar->dma_requests; i++)
-		ti_dra7_xbar_write(xbar->iomem, i, xbar->safe_val);
+	for (i = 0; i < xbar->dma_requests; i++) {
+		if (!test_bit(i, xbar->dma_inuse))
+			ti_dra7_xbar_write(xbar->iomem, i, xbar->safe_val);
+	}
 
 	ret = of_dma_router_register(node, ti_dra7_xbar_route_allocate,
 				     &xbar->dmarouter);
 	if (ret) {
 		/* Restore the defaults for the crossbar */
-		for (i = 0; i < xbar->dma_requests; i++)
-			ti_dra7_xbar_write(xbar->iomem, i, i);
+		for (i = 0; i < xbar->dma_requests; i++) {
+			if (!test_bit(i, xbar->dma_inuse))
+				ti_dra7_xbar_write(xbar->iomem, i, i);
+		}
 	}
 
 	return ret;
-- 
2.6.1

[PATCH v2 3/3] dmaengine: ti-dma-crossbar: dra7: Support for eDMA with new bindings

[Peter Ujfalusi]

Allow the crossbar driver to be used with the eDMA node with non legacy
binding.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
---
 drivers/dma/ti-dma-crossbar.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/dma/ti-dma-crossbar.c b/drivers/dma/ti-dma-crossbar.c
index dac7ae06cc58..e107779b1a2e 100644
--- a/drivers/dma/ti-dma-crossbar.c
+++ b/drivers/dma/ti-dma-crossbar.c
@@ -289,6 +289,10 @@ static const struct of_device_id ti_dra7_master_match[] = {
 		.compatible = "ti,edma3",
 		.data = (void *)TI_XBAR_EDMA_OFFSET,
 	},
+	{
+		.compatible = "ti,edma3-tpcc",
+		.data = (void *)TI_XBAR_EDMA_OFFSET,
+	},
 	{},
 };
 
-- 
2.6.1

Re: [PATCH v2 2/3] dmaengine: ti-dma-crossbar: dra7: Support for reserving DMA event ranges

[Rob Herring]

On Fri, Oct 30, 2015 at 10:00:37AM +0200, Peter Ujfalusi wrote:
> In eDMA the events are directly mapped to a DMA channel (for example DMA
> event 14 can only be handled by DMA channel 14). If the memcpy is enabled
> on the eDMA, there is a possibility that the crossbar driver would assign
> DMA event number already allocated in eDMA for memcpy. Furthermore the
> eDMA can be shared with DSP in which case the crossbar driver should also
> avoid mapping xbar events to DSP used event numbers (or channels).
> 
> Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
> ---
>  .../devicetree/bindings/dma/ti-dma-crossbar.txt    |  6 +++
>  drivers/dma/ti-dma-crossbar.c                      | 47 ++++++++++++++++++++--
>  2 files changed, 49 insertions(+), 4 deletions(-)
> 
> diff --git a/Documentation/devicetree/bindings/dma/ti-dma-crossbar.txt b/Documentation/devicetree/bindings/dma/ti-dma-crossbar.txt
> index b152a75dceae..aead5869a28d 100644
> --- a/Documentation/devicetree/bindings/dma/ti-dma-crossbar.txt
> +++ b/Documentation/devicetree/bindings/dma/ti-dma-crossbar.txt
> @@ -14,6 +14,10 @@ The DMA controller node need to have the following poroperties:
>  
>  Optional properties:
>  - ti,dma-safe-map: Safe routing value for unused request lines
> +- ti,reserved-dma-request-ranges: DMA request ranges which should not be used
> +		when mapping xbar input to DMA request, they are either
> +		allocated to be used by for example the DSP or they are used as
> +		memcpy channels in eDMA.

How many requests are there? I think I'd rather see this as a mask 
value.

Rob

Re: [PATCH v2 2/3] dmaengine: ti-dma-crossbar: dra7: Support for reserving DMA event ranges

[Peter Ujfalusi]

On 11/06/2015 11:53 PM, Rob Herring wrote:
> On Fri, Oct 30, 2015 at 10:00:37AM +0200, Peter Ujfalusi wrote:
>> In eDMA the events are directly mapped to a DMA channel (for example DMA
>> event 14 can only be handled by DMA channel 14). If the memcpy is enabled
>> on the eDMA, there is a possibility that the crossbar driver would assign
>> DMA event number already allocated in eDMA for memcpy. Furthermore the
>> eDMA can be shared with DSP in which case the crossbar driver should also
>> avoid mapping xbar events to DSP used event numbers (or channels).
>>
>> Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
>> ---
>>  .../devicetree/bindings/dma/ti-dma-crossbar.txt    |  6 +++
>>  drivers/dma/ti-dma-crossbar.c                      | 47 ++++++++++++++++++++--
>>  2 files changed, 49 insertions(+), 4 deletions(-)
>>
>> diff --git a/Documentation/devicetree/bindings/dma/ti-dma-crossbar.txt b/Documentation/devicetree/bindings/dma/ti-dma-crossbar.txt
>> index b152a75dceae..aead5869a28d 100644
>> --- a/Documentation/devicetree/bindings/dma/ti-dma-crossbar.txt
>> +++ b/Documentation/devicetree/bindings/dma/ti-dma-crossbar.txt
>> @@ -14,6 +14,10 @@ The DMA controller node need to have the following poroperties:
>>  
>>  Optional properties:
>>  - ti,dma-safe-map: Safe routing value for unused request lines
>> +- ti,reserved-dma-request-ranges: DMA request ranges which should not be used
>> +		when mapping xbar input to DMA request, they are either
>> +		allocated to be used by for example the DSP or they are used as
>> +		memcpy channels in eDMA.
> 
> How many requests are there? I think I'd rather see this as a mask 
> value.

The eDMA on dra7 family have 64 channels. I don't have visibility to customer
kernels, but at least 32 channels need be allocated to be used by the DSP.
This depends on the system design. It could be that only the memcpy channels
need to be reserved, but in some setups we do need more than a few.

-- 
Péter

Re: [PATCH v2 0/3] dmaengine: ti-dma-crossbar: channel reserving and edma3-tpcc support

[Vinod Koul]

On Fri, Oct 30, 2015 at 10:00:35AM +0200, Peter Ujfalusi wrote:
> Hi,
> 
> Changes since v1:
> - Fixed issue introduced by the bitops patch: wrong error check, also switch to
>   use find_first_zero_bit() instead of find_next_zero_bit()
> 
> Cover letter:
> 
> This series depends on the eDMA work I have done, which has been now applied:
> https://lkml.org/lkml/2015/10/16/64
> 
> DRA7 family of chips have both sDMA and eDMA. Currently only sDMA can be used
> becasue the old driver stack for eDMA did not allowed integration w/o hacks.
> 
> Due to the nature of eDMA the crossbar needs to know which eDMA events it can
> use to map incoming events towards the eDMA. In eDMA a channel is wired to be
> used with one specific event. For example eDMA event 14 can only be handled by
> eDMA channel 14.
> The eDMA itself can be shared by different processors in the system (ARM, DSP,
> etc) and since ARM/Linux is the master we need to know which channels are used
> by other cores. Also we need to mask out channels used for memcpy from the
> events we use for HW triggers.

Applied, thanks

-- 
~Vinod