Skip to contentSkip to conversation

k3dma patches to add support for hi3660/HiKey960

PATCH Superseded John Stultz 4 participants 19 messages Jan 16, 2019 – Jan 24, 2019

Subsystems (3)

v4 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

v4 cover k3dma patches to add support for hi3660/HiKey960

This patch series is based on recent work by Tanglei Han, and adds support for hi3660 SoCs as found on the HiKey960 board, along with a few patches I've been carrying.

thanks
-john

New in v4:
* Rework hisi,dma-avail-chans to generic dma-channel-mask, per Rob's suggestion

Cc Tanglei Han <[email protected]>
Cc Zhuangluan Su <[email protected]>
Cc Dan Williams <[email protected]>
Cc Vinod Koul <[email protected]>
Cc Wei Xu <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Rob Herring <[email protected]>
Cc Guodong Xu <[email protected]>
Cc Manivannan Sadhasivam <[email protected]>
Cc Ryan Grachek <[email protected]>

John Stultz (3):
Documentation: bindings: dma: Add binding for dma-channel-mask
arm64: dts: hi3660: Add dma to uart nodes
arm64: dts: hi3660: Fixup unofficial dma-min-chan to dma-channel-mask

Li Yu (2):
dma: k3dma: Delete axi_config
dma: k3dma: Add support for dma-channel-mask

Youlin Wang (3):
Documentation: bindings: k3dma: Extend the k3dma driver binding to
support hisi-asp
dma: k3dma: Upgrade k3dma driver to support hisi_asp_dma hardware
arm64: dts: hi3660: Add hisi asp dma device

 Documentation/devicetree/bindings/dma/dma.txt   |  4 ++
 Documentation/devicetree/bindings/dma/k3dma.txt |  4 +-
 arch/arm64/boot/dts/hisilicon/hi3660.dtsi       | 20 ++++++++-
 drivers/dma/k3dma.c                             | 60 +++++++++++++++++++++----
 4 files changed, 77 insertions(+), 11 deletions(-)
Signature · 1 line
2.7.4
#1 lore raw

PATCH v4 1/8 Documentation: bindings: k3dma: Extend the k3dma driver binding to support hisi-asp

From: Youlin Wang <[email protected]>

Extend the k3dma driver binding to support hisi-asp hardware variants.

---
v2: Simplify patch, removing extranious examples

---
 Documentation/devicetree/bindings/dma/k3dma.txt | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
Documentation/devicetree/bindings/dma/k3dma.txt +3 -1
index 4945aea..10a2f15 100644
@@ -3,7 +3,9 @@
 See dma.txt first
 
 Required properties:
-- compatible: Should be "hisilicon,k3-dma-1.0"
+- compatible: Must be one of
+-              "hisilicon,k3-dma-1.0"
+-              "hisilicon,hisi-pcm-asp-dma-1.0"
 - reg: Should contain DMA registers location and length.
 - interrupts: Should contain one interrupt shared by all channel
 - #dma-cells: see dma.txt, should be 1, para number
#2 lore raw

PATCH v4 2/8 Documentation: bindings: dma: Add binding for dma-channel-mask

In reply to John Stultz

Some dma channels can be reserved for secure mode or other hardware on the SoC, so provide a binding for a bitmask listing the available channels for the kernel to use.

This follows the pre-existing bcm,dma-channel-mask binding.

Cc Vinod Koul <[email protected]>
Cc Rob Herring <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Tanglei Han <[email protected]>
Cc Zhuangluan Su <[email protected]>
Cc Ryan Grachek <[email protected]>
Cc Manivannan Sadhasivam <[email protected]>

---
v3: Renamed to hisi-dma-avail-chan
v4: Reworked to generic dma-channel-mask

---
 Documentation/devicetree/bindings/dma/dma.txt | 4 ++++
 1 file changed, 4 insertions(+)
Documentation/devicetree/bindings/dma/dma.txt +4 -0
index 6312fb0..eeb4e4d 100644
@@ -16,6 +16,9 @@ Optional properties:
 - dma-channels: 	Number of DMA channels supported by the controller.
 - dma-requests: 	Number of DMA request signals supported by the
 			controller.
+- dma-channel-mask:	Bitmask of available DMA channels in ascending order
+			that are not reserved by firmware and are available to
+			the kernel. i.e. first channel corresponds to LSB.
 
 Example:
 
@@ -29,6 +32,7 @@ Example:
 		#dma-cells = <1>;
 		dma-channels = <32>;
 		dma-requests = <127>;
+		dma-channel-mask = <0xfffe>
 	};
 
 * DMA router
#3 lore raw

Re: [PATCH 2/8 v4] Documentation: bindings: dma: Add binding for dma-channel-mask

Quoted 28 lines · John Stultz

Some dma channels can be reserved for secure mode or other hardware on the SoC, so provide a binding for a bitmask listing the available channels for the kernel to use.

This follows the pre-existing bcm,dma-channel-mask binding.

Cc Vinod Koul <[email protected]>
Cc Rob Herring <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Tanglei Han <[email protected]>
Cc Zhuangluan Su <[email protected]>
Cc Ryan Grachek <[email protected]>
Cc Manivannan Sadhasivam <[email protected]>

---
v3: Renamed to hisi-dma-avail-chan
v4: Reworked to generic dma-channel-mask

---
 Documentation/devicetree/bindings/dma/dma.txt | 4 ++++
 1 file changed, 4 insertions(+)
Documentation/devicetree/bindings/dma/dma.txt +0 -0
index 6312fb0..eeb4e4d 100644

Documentation/devicetree/bindings/dma/dma.txt Optional properties: · line 21

Quoted 6 lines
Documentation/devicetree/bindings/dma/dma.txt +3 -0
@@ -16,6 +16,9 @@ Optional properties:
 - dma-channels: 	Number of DMA channels supported by the controller.
 - dma-requests: 	Number of DMA request signals supported by the
 			controller.
+- dma-channel-mask:	Bitmask of available DMA channels in ascending order
+			that are not reserved by firmware and are available to
+			the kernel. i.e. first channel corresponds to LSB.

A general assumption is, "dma-channel-mask" refers to the bit fields of the channels which needs to be masked. But here, it refers to the channels which are available. Doesn't it contradict?

Thanks,
Mani

Quoted 14 lines
Documentation/devicetree/bindings/dma/dma.txt +1 -0
 Example:
 
@@ -29,6 +32,7 @@ Example:
 		#dma-cells = <1>;
 		dma-channels = <32>;
 		dma-requests = <127>;
+		dma-channel-mask = <0xfffe>
 	};
 
 * DMA router
Signature · 1 line
2.7.4
#4 lore raw

Re: [PATCH 2/8 v4] Documentation: bindings: dma: Add binding for dma-channel-mask

Quoted 25 lines · Manivannan Sadhasivam
Quoted 23 lines · John Stultz

Some dma channels can be reserved for secure mode or other hardware on the SoC, so provide a binding for a bitmask listing the available channels for the kernel to use.

This follows the pre-existing bcm,dma-channel-mask binding.

Cc Vinod Koul <[email protected]>
Cc Rob Herring <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Tanglei Han <[email protected]>
Cc Zhuangluan Su <[email protected]>
Cc Ryan Grachek <[email protected]>
Cc Manivannan Sadhasivam <[email protected]>

---
v3: Renamed to hisi-dma-avail-chan
v4: Reworked to generic dma-channel-mask

---
 Documentation/devicetree/bindings/dma/dma.txt | 4 ++++
 1 file changed, 4 insertions(+)
Quoted 15 lines

Documentation/devicetree/bindings/dma/dma.txt Optional properties: · line 21

Quoted 11 lines
Documentation/devicetree/bindings/dma/dma.txt +3 -0
index 6312fb0..eeb4e4d 100644
@@ -16,6 +16,9 @@ Optional properties:
 - dma-channels:      Number of DMA channels supported by the controller.
 - dma-requests:      Number of DMA request signals supported by the
                      controller.
+- dma-channel-mask:  Bitmask of available DMA channels in ascending order
+                     that are not reserved by firmware and are available to
+                     the kernel. i.e. first channel corresponds to LSB.

A general assumption is, "dma-channel-mask" refers to the bit fields of the channels which needs to be masked. But here, it refers to the channels which are available. Doesn't it contradict?

Hrm. So while I can sort of understand the common usage of "mask" as to "hide", thus the desire to have a bitfield mean "the channels we hide" or "don't use", but in my experience bitmasking is more commonly used to keep only a portion of the the bits, so from that perspective its more intuitive that a mask be the channels we keep to use. So I'm not sure if your suggestion makes it more clear.

But I'm not very particular here, so I'll defer to others on this.

thanks
-john

maintainer
#5 lore raw

Re: [PATCH 2/8 v4] Documentation: bindings: dma: Add binding for dma-channel-mask

Quoted 43 lines · John Stultz
Quoted 25 lines · Manivannan Sadhasivam
Quoted 23 lines · John Stultz

Some dma channels can be reserved for secure mode or other hardware on the SoC, so provide a binding for a bitmask listing the available channels for the kernel to use.

This follows the pre-existing bcm,dma-channel-mask binding.

Cc Vinod Koul <[email protected]>
Cc Rob Herring <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Tanglei Han <[email protected]>
Cc Zhuangluan Su <[email protected]>
Cc Ryan Grachek <[email protected]>
Cc Manivannan Sadhasivam <[email protected]>

---
v3: Renamed to hisi-dma-avail-chan
v4: Reworked to generic dma-channel-mask

---
 Documentation/devicetree/bindings/dma/dma.txt | 4 ++++
 1 file changed, 4 insertions(+)
Quoted 15 lines

Documentation/devicetree/bindings/dma/dma.txt Optional properties: · line 21

Quoted 11 lines
Documentation/devicetree/bindings/dma/dma.txt +3 -0
index 6312fb0..eeb4e4d 100644
@@ -16,6 +16,9 @@ Optional properties:
 - dma-channels:      Number of DMA channels supported by the controller.
 - dma-requests:      Number of DMA request signals supported by the
                      controller.
+- dma-channel-mask:  Bitmask of available DMA channels in ascending order
+                     that are not reserved by firmware and are available to
+                     the kernel. i.e. first channel corresponds to LSB.

A general assumption is, "dma-channel-mask" refers to the bit fields of the channels which needs to be masked. But here, it refers to the channels which are available. Doesn't it contradict?

Quoted 8 lines

Hrm. So while I can sort of understand the common usage of "mask" as to "hide", thus the desire to have a bitfield mean "the channels we hide" or "don't use", but in my experience bitmasking is more commonly used to keep only a portion of the the bits, so from that perspective its more intuitive that a mask be the channels we keep to use. So I'm not sure if your suggestion makes it more clear.

But I'm not very particular here, so I'll defer to others on this.

Given the context and documentation which explicitly says it is bitmask of available channels, i think we are fine :)

maintainer Reviewed-by on v4 2/8
#6 lore raw

Re: [PATCH 2/8 v4] Documentation: bindings: dma: Add binding for dma-channel-mask

In reply to John Stultz
Quoted 16 lines · John Stultz

Some dma channels can be reserved for secure mode or other hardware on the SoC, so provide a binding for a bitmask listing the available channels for the kernel to use.

This follows the pre-existing bcm,dma-channel-mask binding.

Cc Vinod Koul <[email protected]>
Cc Rob Herring <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Tanglei Han <[email protected]>
Cc Zhuangluan Su <[email protected]>
Cc Ryan Grachek <[email protected]>
Cc Manivannan Sadhasivam <[email protected]>
Quoted 7 lines

---
v3: Renamed to hisi-dma-avail-chan
v4: Reworked to generic dma-channel-mask

---
 Documentation/devicetree/bindings/dma/dma.txt | 4 ++++
 1 file changed, 4 insertions(+)
#7 lore raw

PATCH v4 3/8 dma: k3dma: Upgrade k3dma driver to support hisi_asp_dma hardware

In reply to John Stultz

From: Youlin Wang <[email protected]>

On the hi3660 hardware there are two (at least) DMA controllers, the DMA-P (Peripherial DMA) and the DMA-A (Audio DMA). The two blocks are similar, but have some slight differences. This resulted in the vendor implementing two separate drivers, which after review, they have been able to condense and re-use the existing k3dma driver.

Thus, this patch adds support for the new "hisi-pcm-asp-dma-1.0" compatible string in the binding.

One difference with the DMA-A controller, is that it does not need to initialize a clock. So we skip this by adding and using soc data flags.

After above this driver will support both k3 and hisi_asp dma hardware.

Cc Dan Williams <[email protected]>
Cc Vinod Koul <[email protected]>
Cc Zhuangluan Su <[email protected]>
Cc Ryan Grachek <[email protected]>
Cc Manivannan Sadhasivam <[email protected]>
[jstultz: Reworked to use of_match_data, commit msg improvements]

---
v2:
* Reworked to use of_match_data
v3:
* Further rework of the commit message

---
 drivers/dma/k3dma.c | 37 ++++++++++++++++++++++++++++++++-----
 1 file changed, 32 insertions(+), 5 deletions(-)
drivers/dma/k3dma.c +32 -5
index fdec2b6..df61406 100644
@@ -116,6 +116,13 @@ struct k3_dma_dev {
 	unsigned int		irq;
 };
 
+
+#define K3_FLAG_NOCLK  (1<<0)
+struct k3dma_soc_data {
+	unsigned long flags;
+};
+
+
 #define to_k3_dma(dmadev) container_of(dmadev, struct k3_dma_dev, slave)
 
 static int k3_dma_config_write(struct dma_chan *chan,
@@ -790,8 +797,21 @@ static int k3_dma_transfer_resume(struct dma_chan *chan)
 	return 0;
 }
 
+static const struct k3dma_soc_data k3_v1_dma_data = {
+	.flags = 0,
+};
+
+static const struct k3dma_soc_data asp_v1_dma_data = {
+	.flags = K3_FLAG_NOCLK,
+};
+
 static const struct of_device_id k3_pdma_dt_ids[] = {
-	{ .compatible = "hisilicon,k3-dma-1.0", },
+	{ .compatible = "hisilicon,k3-dma-1.0",
+	  .data = &k3_v1_dma_data
+	},
+	{ .compatible = "hisilicon,hisi-pcm-asp-dma-1.0",
+	  .data = &asp_v1_dma_data
+	},
 	{}
 };
 MODULE_DEVICE_TABLE(of, k3_pdma_dt_ids);
@@ -810,6 +830,7 @@ static struct dma_chan *k3_of_dma_simple_xlate(struct of_phandle_args *dma_spec,
 
 static int k3_dma_probe(struct platform_device *op)
 {
+	const struct k3dma_soc_data *soc_data;
 	struct k3_dma_dev *d;
 	const struct of_device_id *of_id;
 	struct resource *iores;
@@ -823,6 +844,10 @@ static int k3_dma_probe(struct platform_device *op)
 	if (!d)
 		return -ENOMEM;
 
+	soc_data = device_get_match_data(&op->dev);
+	if (!soc_data)
+		return -EINVAL;
+
 	d->base = devm_ioremap_resource(&op->dev, iores);
 	if (IS_ERR(d->base))
 		return PTR_ERR(d->base);
@@ -835,10 +860,12 @@ static int k3_dma_probe(struct platform_device *op)
 				"dma-requests", &d->dma_requests);
 	}
 
-	d->clk = devm_clk_get(&op->dev, NULL);
-	if (IS_ERR(d->clk)) {
-		dev_err(&op->dev, "no dma clk\n");
-		return PTR_ERR(d->clk);
+	if (!(soc_data->flags & K3_FLAG_NOCLK)) {
+		d->clk = devm_clk_get(&op->dev, NULL);
+		if (IS_ERR(d->clk)) {
+			dev_err(&op->dev, "no dma clk\n");
+			return PTR_ERR(d->clk);
+		}
 	}
 
 	irq = platform_get_irq(op, 0);
maintainer
#8 lore raw

Re: [PATCH 3/8 v4] dma: k3dma: Upgrade k3dma driver to support hisi_asp_dma hardware

Quoted 4 lines · John Stultz

From: Youlin Wang <[email protected]>

On the hi3660 hardware there are two (at least) DMA controllers, the DMA-P (Peripherial DMA) and the DMA-A (Audio DMA). The

                    ^^^

typo

Quoted 8 lines

two blocks are similar, but have some slight differences. This resulted in the vendor implementing two separate drivers, which after review, they have been able to condense and re-use the existing k3dma driver.

Thus, this patch adds support for the new "hisi-pcm-asp-dma-1.0" compatible string in the binding.

Quoted 6 lines

One difference with the DMA-A controller, is that it does not need to initialize a clock. So we skip this by adding and using soc data flags.

After above this driver will support both k3 and hisi_asp dma hardware.

Great thanks for the effort!

Quoted 26 lines
Cc Dan Williams <[email protected]>
Cc Vinod Koul <[email protected]>
Cc Zhuangluan Su <[email protected]>
Cc Ryan Grachek <[email protected]>
Cc Manivannan Sadhasivam <[email protected]>
[jstultz: Reworked to use of_match_data, commit msg improvements]

---
v2:
* Reworked to use of_match_data
v3:
* Further rework of the commit message

---
 drivers/dma/k3dma.c | 37 ++++++++++++++++++++++++++++++++-----
 1 file changed, 32 insertions(+), 5 deletions(-)
drivers/dma/k3dma.c +0 -0
index fdec2b6..df61406 100644

drivers/dma/k3dma.c struct k3_dma_dev { · line 120

Quoted 5 lines
drivers/dma/k3dma.c +2 -0
@@ -116,6 +116,13 @@ struct k3_dma_dev {
 	unsigned int		irq;
 };
 
+
+#define K3_FLAG_NOCLK  (1<<0)

why not use BIT()

space between two please

Quoted 9 lines
drivers/dma/k3dma.c +5 -0
+struct k3dma_soc_data {
+	unsigned long flags;
+};
+
+
 #define to_k3_dma(dmadev) container_of(dmadev, struct k3_dma_dev, slave)
 
 static int k3_dma_config_write(struct dma_chan *chan,

drivers/dma/k3dma.c static int k3_dma_transfer_resume(struct dma_chan *chan) · line 802

Quoted 6 lines
drivers/dma/k3dma.c +3 -0
@@ -790,8 +797,21 @@ static int k3_dma_transfer_resume(struct dma_chan *chan)
 	return 0;
 }
 
+static const struct k3dma_soc_data k3_v1_dma_data = {
+	.flags = 0,
+};

So basically this is default right, do we need to set dma_data and not assume default..

Quoted 25 lines
drivers/dma/k3dma.c +12 -1
+
+static const struct k3dma_soc_data asp_v1_dma_data = {
+	.flags = K3_FLAG_NOCLK,
+};
+
 static const struct of_device_id k3_pdma_dt_ids[] = {
-	{ .compatible = "hisilicon,k3-dma-1.0", },
+	{ .compatible = "hisilicon,k3-dma-1.0",
+	  .data = &k3_v1_dma_data
+	},
+	{ .compatible = "hisilicon,hisi-pcm-asp-dma-1.0",
+	  .data = &asp_v1_dma_data
+	},
 	{}
 };
 MODULE_DEVICE_TABLE(of, k3_pdma_dt_ids);
@@ -810,6 +830,7 @@ static struct dma_chan *k3_of_dma_simple_xlate(struct of_phandle_args *dma_spec,
 
 static int k3_dma_probe(struct platform_device *op)
 {
+	const struct k3dma_soc_data *soc_data;
 	struct k3_dma_dev *d;
 	const struct of_device_id *of_id;
 	struct resource *iores;

drivers/dma/k3dma.c static int k3_dma_probe(struct platform_device *op) · line 849

Quoted 6 lines
drivers/dma/k3dma.c +3 -0
@@ -823,6 +844,10 @@ static int k3_dma_probe(struct platform_device *op)
 	if (!d)
 		return -ENOMEM;
 
+	soc_data = device_get_match_data(&op->dev);
+	if (!soc_data)
+		return -EINVAL;

So this is not optional, either ways fine by me :)

#9 lore raw

Re: [PATCH 3/8 v4] dma: k3dma: Upgrade k3dma driver to support hisi_asp_dma hardware

Quoted 8 lines · Vinod Koul
Quoted 4 lines · John Stultz

From: Youlin Wang <[email protected]>

On the hi3660 hardware there are two (at least) DMA controllers, the DMA-P (Peripherial DMA) and the DMA-A (Audio DMA). The

                    ^^^

typo

Thanks so much for the review!

Fixed!

Quoted 6 lines

drivers/dma/k3dma.c struct k3_dma_dev { · line 120

Quoted 2 lines
drivers/dma/k3dma.c +2 -0
+
+#define K3_FLAG_NOCLK  (1<<0)

why not use BIT()

space between two please

Fixed.

Quoted 6 lines

drivers/dma/k3dma.c static int k3_dma_transfer_resume(struct dma_chan *chan) · line 802

Quoted 3 lines
drivers/dma/k3dma.c +3 -0
+static const struct k3dma_soc_data k3_v1_dma_data = {
+     .flags = 0,
+};

So basically this is default right, do we need to set dma_data and not assume default..

I'm not sure I fully understand you here. Basically I'm just creating a per-variant data structure. The k3_v1_dma_data isn't really the default, but is the first variant supported. There may be future variants that cause some new flag that the k3_v3_dma_data may need to set. But for now that variant doesn't have any flags set.

Quoted 28 lines
Quoted 28 lines
drivers/dma/k3dma.c +12 -1
+
+static const struct k3dma_soc_data asp_v1_dma_data = {
+     .flags = K3_FLAG_NOCLK,
+};
+
 static const struct of_device_id k3_pdma_dt_ids[] = {
-     { .compatible = "hisilicon,k3-dma-1.0", },
+     { .compatible = "hisilicon,k3-dma-1.0",
+       .data = &k3_v1_dma_data
+     },
+     { .compatible = "hisilicon,hisi-pcm-asp-dma-1.0",
+       .data = &asp_v1_dma_data
+     },
      {}
 };
 MODULE_DEVICE_TABLE(of, k3_pdma_dt_ids);
@@ -810,6 +830,7 @@ static struct dma_chan *k3_of_dma_simple_xlate(struct of_phandle_args *dma_spec,
 
 static int k3_dma_probe(struct platform_device *op)
 {
+     const struct k3dma_soc_data *soc_data;
      struct k3_dma_dev *d;
      const struct of_device_id *of_id;
      struct resource *iores;
@@ -823,6 +844,10 @@ static int k3_dma_probe(struct platform_device *op)
      if (!d)
              return -ENOMEM;
 
Quoted 5 lines

drivers/dma/k3dma.c static int k3_dma_probe(struct platform_device *op) · line 849

Quoted 3 lines
drivers/dma/k3dma.c +3 -0
+     soc_data = device_get_match_data(&op->dev);
+     if (!soc_data)
+             return -EINVAL;

So this is not optional, either ways fine by me :)

I think this way makes sense, but maybe I'm missing a better alternative?

Do let me know if there's an example you'd rather I follow.

Thanks so much again for the review!
-john

maintainer
#10 lore raw

Re: [PATCH 3/8 v4] dma: k3dma: Upgrade k3dma driver to support hisi_asp_dma hardware

Quoted 30 lines · John Stultz
Quoted 8 lines · Vinod Koul
Quoted 4 lines · John Stultz

From: Youlin Wang <[email protected]>

On the hi3660 hardware there are two (at least) DMA controllers, the DMA-P (Peripherial DMA) and the DMA-A (Audio DMA). The

                    ^^^

typo

Thanks so much for the review!

Fixed!

Quoted 6 lines

drivers/dma/k3dma.c struct k3_dma_dev { · line 120

Quoted 2 lines
drivers/dma/k3dma.c +2 -0
+
+#define K3_FLAG_NOCLK  (1<<0)

why not use BIT()

space between two please

Fixed.

Quoted 6 lines

drivers/dma/k3dma.c static int k3_dma_transfer_resume(struct dma_chan *chan) · line 802

Quoted 3 lines
drivers/dma/k3dma.c +3 -0
+static const struct k3dma_soc_data k3_v1_dma_data = {
+     .flags = 0,
+};

So basically this is default right, do we need to set dma_data and not assume default..

Quoted 5 lines

I'm not sure I fully understand you here. Basically I'm just creating a per-variant data structure. The k3_v1_dma_data isn't really the default, but is the first variant supported. There may be future variants that cause some new flag that the k3_v3_dma_data may need to set. But for now that variant doesn't have any flags set.

So my point was we can skip this and treat driver data NULL as default i.e. no flags, no big deal though, saves you dummy k3_v1_dma_data.

Quoted 36 lines
Quoted 30 lines
Quoted 30 lines
drivers/dma/k3dma.c +12 -1
+
+static const struct k3dma_soc_data asp_v1_dma_data = {
+     .flags = K3_FLAG_NOCLK,
+};
+
 static const struct of_device_id k3_pdma_dt_ids[] = {
-     { .compatible = "hisilicon,k3-dma-1.0", },
+     { .compatible = "hisilicon,k3-dma-1.0",
+       .data = &k3_v1_dma_data
+     },
+     { .compatible = "hisilicon,hisi-pcm-asp-dma-1.0",
+       .data = &asp_v1_dma_data
+     },
      {}
 };
 MODULE_DEVICE_TABLE(of, k3_pdma_dt_ids);
@@ -810,6 +830,7 @@ static struct dma_chan *k3_of_dma_simple_xlate(struct of_phandle_args *dma_spec,
 
 static int k3_dma_probe(struct platform_device *op)
 {
+     const struct k3dma_soc_data *soc_data;
      struct k3_dma_dev *d;
      const struct of_device_id *of_id;
      struct resource *iores;
@@ -823,6 +844,10 @@ static int k3_dma_probe(struct platform_device *op)
      if (!d)
              return -ENOMEM;
 
Quoted 5 lines

drivers/dma/k3dma.c static int k3_dma_probe(struct platform_device *op) · line 849

Quoted 3 lines
drivers/dma/k3dma.c +3 -0
+     soc_data = device_get_match_data(&op->dev);
+     if (!soc_data)
+             return -EINVAL;

So this is not optional, either ways fine by me :)

Quoted 3 lines

I think this way makes sense, but maybe I'm missing a better alternative?

Do let me know if there's an example you'd rather I follow.

To elaborate I was thinking of alternate scheme with:

        compatible = "hisilicon,k3-dma-1.0", NULL
        compatible = "hisilicon,hisi-pcm-asp-dma-1.0", .data = &asp_v1_dma_data

and

        soc_data = device_get_match_data(&op->dev);
        if (!soc_data) {
                /* no data so flags are null */
                dev_warn(... "no driver data specified, assuming  no flags\n"
                k3_dma->flags = 0;
        }
#11 lore raw

Re: [PATCH 3/8 v4] dma: k3dma: Upgrade k3dma driver to support hisi_asp_dma hardware

Quoted 8 lines · Vinod Koul
Quoted 1 line · John Stultz

Do let me know if there's an example you'd rather I follow.

To elaborate I was thinking of alternate scheme with:

        compatible = "hisilicon,k3-dma-1.0", NULL
        compatible = "hisilicon,hisi-pcm-asp-dma-1.0", .data = &asp_v1_dma_data
Quoted 8 lines

and

        soc_data = device_get_match_data(&op->dev);
        if (!soc_data) {
                /* no data so flags are null */
                dev_warn(... "no driver data specified, assuming  no flags\n"
                k3_dma->flags = 0;
        }

Thanks for the clarification! Hmm. I can do this, though the trouble is all the soc_data-> references have to switch to a struct k3dma_soc_data on the stack, which we have to initialize/copy over depending on the soc_data return, which I'm not sure really simplifies much (other then saving a ulong off the heap). But I'm ok with either.

thanks
-john

#12 lore raw

PATCH v4 4/8 dma: k3dma: Delete axi_config

In reply to John Stultz

From: Li Yu <[email protected]>

Axi_config controls whether DMA resources can be accessed in non-secure mode, such as linux kernel. The register should be set by the bootloader stage and depends on the device.

Thus, this patch removes axi_config from k3dma driver.

Cc Dan Williams <[email protected]>
Cc Vinod Koul <[email protected]>
Cc Tanglei Han <[email protected]>
Cc Zhuangluan Su <[email protected]>
Cc Ryan Grachek <[email protected]>
Cc Manivannan Sadhasivam <[email protected]>
[jstultz: Minor tweaks to commit message]
---
 drivers/dma/k3dma.c | 3 ---
 1 file changed, 3 deletions(-)
drivers/dma/k3dma.c +0 -3
index df61406..b2060bf 100644
@@ -52,8 +52,6 @@
 #define CX_SRC			0x814
 #define CX_DST			0x818
 #define CX_CFG			0x81c
-#define AXI_CFG			0x820
-#define AXI_CFG_DEFAULT		0x201201
 
 #define CX_LLI_CHAIN_EN		0x2
 #define CX_CFG_EN		0x1
@@ -168,7 +166,6 @@ static void k3_dma_set_desc(struct k3_dma_phy *phy, struct k3_desc_hw *hw)
 	writel_relaxed(hw->count, phy->base + CX_CNT0);
 	writel_relaxed(hw->saddr, phy->base + CX_SRC);
 	writel_relaxed(hw->daddr, phy->base + CX_DST);
-	writel_relaxed(AXI_CFG_DEFAULT, phy->base + AXI_CFG);
 	writel_relaxed(hw->config, phy->base + CX_CFG);
 }
 
#13 lore raw

PATCH v4 5/8 dma: k3dma: Add support for dma-channel-mask

In reply to John Stultz

From: Li Yu <[email protected]>

Add dma-channel-mask as a property for k3dma, it defines available dma channels which a non-secure mode driver can use.

One sample usage of this is in Hi3660 SoC. DMA channel 0 is reserved to lpm3, which is a coprocessor for power management. So as a result, any request in kernel (which runs on main processor and in non-secure mode) should start from at least channel 1.

Cc Dan Williams <[email protected]>
Cc Vinod Koul <[email protected]>
Cc Tanglei Han <[email protected]>
Cc Zhuangluan Su <[email protected]>
Cc Ryan Grachek <[email protected]>
Cc Manivannan Sadhasivam <[email protected]>
Cc Guodong Xu <[email protected]>
[jstultz: Reworked to use a channel mask]

---
v3: Rename to hisi-dma-avail-chan
v4: Rename to dma-channel-mask

---
 drivers/dma/k3dma.c | 20 +++++++++++++++++++-
 1 file changed, 19 insertions(+), 1 deletion(-)
drivers/dma/k3dma.c +19 -1
index b2060bf..ed19b1f 100644
@@ -111,6 +111,7 @@ struct k3_dma_dev {
 	struct dma_pool		*pool;
 	u32			dma_channels;
 	u32			dma_requests;
+	u32			dma_channel_mask;
 	unsigned int		irq;
 };
 
@@ -318,6 +319,9 @@ static void k3_dma_tasklet(unsigned long arg)
 	/* check new channel request in d->chan_pending */
 	spin_lock_irq(&d->lock);
 	for (pch = 0; pch < d->dma_channels; pch++) {
+		if (!(d->dma_channel_mask & (1<<pch)))
+			continue;
+
 		p = &d->phy[pch];
 
 		if (p->vchan == NULL && !list_empty(&d->chan_pending)) {
@@ -335,6 +339,9 @@ static void k3_dma_tasklet(unsigned long arg)
 	spin_unlock_irq(&d->lock);
 
 	for (pch = 0; pch < d->dma_channels; pch++) {
+		if (!(d->dma_channel_mask & (1<<pch)))
+			continue;
+
 		if (pch_alloc & (1 << pch)) {
 			p = &d->phy[pch];
 			c = p->vchan;
@@ -855,6 +862,13 @@ static int k3_dma_probe(struct platform_device *op)
 				"dma-channels", &d->dma_channels);
 		of_property_read_u32((&op->dev)->of_node,
 				"dma-requests", &d->dma_requests);
+		ret = of_property_read_u32((&op->dev)->of_node,
+				"dma-channel-mask", &d->dma_channel_mask);
+		if (ret) {
+			dev_warn(&op->dev,
+				 "dma-channel-mask doesn't exist, considering all as available.\n");
+			d->dma_channel_mask = (u32)~0UL;
+		}
 	}
 
 	if (!(soc_data->flags & K3_FLAG_NOCLK)) {
@@ -886,8 +900,12 @@ static int k3_dma_probe(struct platform_device *op)
 		return -ENOMEM;
 
 	for (i = 0; i < d->dma_channels; i++) {
-		struct k3_dma_phy *p = &d->phy[i];
+		struct k3_dma_phy *p;
+
+		if (!(d->dma_channel_mask & (1<<i)))
+			continue;
 
+		p = &d->phy[i];
 		p->idx = i;
 		p->base = d->base + i * 0x40;
 	}
#14 lore raw

Re: [PATCH 5/8 v4] dma: k3dma: Add support for dma-channel-mask

Quoted 41 lines · John Stultz

From: Li Yu <[email protected]>

Add dma-channel-mask as a property for k3dma, it defines available dma channels which a non-secure mode driver can use.

One sample usage of this is in Hi3660 SoC. DMA channel 0 is reserved to lpm3, which is a coprocessor for power management. So as a result, any request in kernel (which runs on main processor and in non-secure mode) should start from at least channel 1.

Cc Dan Williams <[email protected]>
Cc Vinod Koul <[email protected]>
Cc Tanglei Han <[email protected]>
Cc Zhuangluan Su <[email protected]>
Cc Ryan Grachek <[email protected]>
Cc Manivannan Sadhasivam <[email protected]>
Cc Guodong Xu <[email protected]>
[jstultz: Reworked to use a channel mask]

---
v3: Rename to hisi-dma-avail-chan
v4: Rename to dma-channel-mask

---
 drivers/dma/k3dma.c | 20 +++++++++++++++++++-
 1 file changed, 19 insertions(+), 1 deletion(-)
drivers/dma/k3dma.c +1 -0
index b2060bf..ed19b1f 100644
@@ -111,6 +111,7 @@ struct k3_dma_dev {
 	struct dma_pool		*pool;
 	u32			dma_channels;
 	u32			dma_requests;
+	u32			dma_channel_mask;
 	unsigned int		irq;
 };
 

drivers/dma/k3dma.c static void k3_dma_tasklet(unsigned long arg) · line 324

Quoted 6 lines
drivers/dma/k3dma.c +3 -0
@@ -318,6 +319,9 @@ static void k3_dma_tasklet(unsigned long arg)
 	/* check new channel request in d->chan_pending */
 	spin_lock_irq(&d->lock);
 	for (pch = 0; pch < d->dma_channels; pch++) {
+		if (!(d->dma_channel_mask & (1<<pch)))
+			continue;
+

As per my comment in bindings patch, the code here gets the assumption that we are skipping the channels which are not masked. Either the property should be named as "dma-avail-chan" or "dma_channel_mask" should have the bit mask of channels which needs to be masked. Then the code will be,

/* Skip the channels which are masked */
if ((d->dma_channel_mask) & BIT(pch))

	continue;

PS: Use BIT() macro where applicable.

Thanks,
Mani

Quoted 44 lines
drivers/dma/k3dma.c +15 -1
 		p = &d->phy[pch];
 
 		if (p->vchan == NULL && !list_empty(&d->chan_pending)) {
@@ -335,6 +339,9 @@ static void k3_dma_tasklet(unsigned long arg)
 	spin_unlock_irq(&d->lock);
 
 	for (pch = 0; pch < d->dma_channels; pch++) {
+		if (!(d->dma_channel_mask & (1<<pch)))
+			continue;
+
 		if (pch_alloc & (1 << pch)) {
 			p = &d->phy[pch];
 			c = p->vchan;
@@ -855,6 +862,13 @@ static int k3_dma_probe(struct platform_device *op)
 				"dma-channels", &d->dma_channels);
 		of_property_read_u32((&op->dev)->of_node,
 				"dma-requests", &d->dma_requests);
+		ret = of_property_read_u32((&op->dev)->of_node,
+				"dma-channel-mask", &d->dma_channel_mask);
+		if (ret) {
+			dev_warn(&op->dev,
+				 "dma-channel-mask doesn't exist, considering all as available.\n");
+			d->dma_channel_mask = (u32)~0UL;
+		}
 	}
 
 	if (!(soc_data->flags & K3_FLAG_NOCLK)) {
@@ -886,8 +900,12 @@ static int k3_dma_probe(struct platform_device *op)
 		return -ENOMEM;
 
 	for (i = 0; i < d->dma_channels; i++) {
-		struct k3_dma_phy *p = &d->phy[i];
+		struct k3_dma_phy *p;
+
+		if (!(d->dma_channel_mask & (1<<i)))
+			continue;
 
+		p = &d->phy[i];
 		p->idx = i;
 		p->base = d->base + i * 0x40;
 	}
Signature · 1 line
2.7.4
#15 lore raw

Re: [PATCH 5/8 v4] dma: k3dma: Add support for dma-channel-mask

Quoted 4 lines · Manivannan Sadhasivam

/* Skip the channels which are masked */
if ((d->dma_channel_mask) & BIT(pch))

        continue;

Per the discussion w/ Vinod and Rob, I think I'll leave this bit be, so we use the channels in the bitmask.

Quoted 1 line

PS: Use BIT() macro where applicable.

But this suggestions I've integrated for v5.

Thanks so much for the review!
-john

#16 lore raw

PATCH v4 6/8 arm64: dts: hi3660: Add dma to uart nodes

In reply to John Stultz

Try to add DMA support to the uart nodes following the assignments made in the dts from the victoria vendor kernel here:
https://consumer.huawei.com/en/opensource/detail/?siteCode=worldwide&keywords=p10&fileType=openSourceSoftware&pageSize=10&curPage=1

Cc Tanglei Han <[email protected]>
Cc Zhuangluan Su <[email protected]>
Cc Ryan Grachek <[email protected]>
Cc Manivannan Sadhasivam <[email protected]>
Cc Wei Xu <[email protected]>
Cc Rob Herring <[email protected]>
Cc Mark Rutland <[email protected]>

---
v3:
* Remove dma enablment on uart0 which would use reserved channel 0

---
 arch/arm64/boot/dts/hisilicon/hi3660.dtsi | 8 ++++++++
 1 file changed, 8 insertions(+)
arch/arm64/boot/dts/hisilicon/hi3660.dtsi +8 -0
index 20ae40d..4c8d682 100644
@@ -478,6 +478,8 @@
 			compatible = "arm,pl011", "arm,primecell";
 			reg = <0x0 0xfdf00000 0x0 0x1000>;
 			interrupts = <GIC_SPI 75 IRQ_TYPE_LEVEL_HIGH>;
+			dma-names = "rx", "tx";
+			dmas =  <&dma0 2 &dma0 3>;
 			clocks = <&crg_ctrl HI3660_CLK_GATE_UART1>,
 				 <&crg_ctrl HI3660_CLK_GATE_UART1>;
 			clock-names = "uartclk", "apb_pclk";
@@ -490,6 +492,8 @@
 			compatible = "arm,pl011", "arm,primecell";
 			reg = <0x0 0xfdf03000 0x0 0x1000>;
 			interrupts = <GIC_SPI 76 IRQ_TYPE_LEVEL_HIGH>;
+			dma-names = "rx", "tx";
+			dmas =  <&dma0 4 &dma0 5>;
 			clocks = <&crg_ctrl HI3660_CLK_GATE_UART2>,
 				 <&crg_ctrl HI3660_PCLK>;
 			clock-names = "uartclk", "apb_pclk";
@@ -514,6 +518,8 @@
 			compatible = "arm,pl011", "arm,primecell";
 			reg = <0x0 0xfdf01000 0x0 0x1000>;
 			interrupts = <GIC_SPI 77 IRQ_TYPE_LEVEL_HIGH>;
+			dma-names = "rx", "tx";
+			dmas =  <&dma0 6 &dma0 7>;
 			clocks = <&crg_ctrl HI3660_CLK_GATE_UART4>,
 				 <&crg_ctrl HI3660_CLK_GATE_UART4>;
 			clock-names = "uartclk", "apb_pclk";
@@ -526,6 +532,8 @@
 			compatible = "arm,pl011", "arm,primecell";
 			reg = <0x0 0xfdf05000 0x0 0x1000>;
 			interrupts = <GIC_SPI 78 IRQ_TYPE_LEVEL_HIGH>;
+			dma-names = "rx", "tx";
+			dmas =  <&dma0 8 &dma0 9>;
 			clocks = <&crg_ctrl HI3660_CLK_GATE_UART5>,
 				 <&crg_ctrl HI3660_CLK_GATE_UART5>;
 			clock-names = "uartclk", "apb_pclk";
Signature · 1 line
2.7.4
#17 lore raw

PATCH v4 7/8 arm64: dts: hi3660: Add hisi asp dma device

In reply to John Stultz

From: Youlin Wang <[email protected]>

Add asp-dma device to hi3660 dts

---
v2: Removed undocumented bindings

---
 arch/arm64/boot/dts/hisilicon/hi3660.dtsi | 10 ++++++++++
 1 file changed, 10 insertions(+)
arch/arm64/boot/dts/hisilicon/hi3660.dtsi +10 -0
index 4c8d682..77a7135 100644
@@ -567,6 +567,16 @@
 			dma-type = "hi3660_dma";
 		};
 
+		asp_dmac: dma-controller@e804b000 {
+			compatible = "hisilicon,hisi-pcm-asp-dma-1.0";
+			reg = <0x0 0xe804b000 0x0 0x1000>;
+			#dma-cells = <1>;
+			dma-channels = <16>;
+			dma-requests = <32>;
+			interrupts = <GIC_SPI 216 IRQ_TYPE_LEVEL_HIGH>;
+			interrupt-names = "asp_dma_irq";
+		};
+
 		rtc0: rtc@fff04000 {
 			compatible = "arm,pl031", "arm,primecell";
 			reg = <0x0 0Xfff04000 0x0 0x1000>;
Signature · 1 line
2.7.4
#18 lore raw

PATCH v4 8/8 arm64: dts: hi3660: Fixup unofficial dma-min-chan to dma-channel-mask

In reply to John Stultz

A undocumented and unimplemented binding got into the hi3660 dtsi, and this switches that binding to the now documented one.

Cc Tanglei Han <[email protected]>
Cc Zhuangluan Su <[email protected]>
Cc Ryan Grachek <[email protected]>
Cc Manivannan Sadhasivam <[email protected]>
Cc Wei Xu <[email protected]>
Cc Rob Herring <[email protected]>
Cc Mark Rutland <[email protected]>

---
v3: Renamed to hisi-dma-avail-chan
v4: Renamed to dma-channel-mask

---
 arch/arm64/boot/dts/hisilicon/hi3660.dtsi | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm64/boot/dts/hisilicon/hi3660.dtsi +1 -1
index 77a7135..8e48f42 100644
@@ -560,7 +560,7 @@
 			#dma-cells = <1>;
 			dma-channels = <16>;
 			dma-requests = <32>;
-			dma-min-chan = <1>;
+			dma-channel-mask = <0xfffe>;
 			interrupts = <GIC_SPI 143 IRQ_TYPE_LEVEL_HIGH>;
 			clocks = <&crg_ctrl HI3660_CLK_GATE_DMAC>;
 			dma-no-cci;
Signature · 1 line
2.7.4