Skip to contentSkip to conversation

Add ASoC support for AMD Stoney APUs

PATCH Superseded Alex Deucher 7 participants 19 messages Aug 18, 2017 – Sep 1, 2017 2 messages not in the archive

Subsystems (5)

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

v3 cover Add ASoC support for AMD Stoney APUs

This patch set updates the AMD GPU and Audio CoProcessor (ACP) audio drivers and the designware i2s driver for Stoney (ST). ST is an APU similar to Carrizo (CZ) which already has ACP audio support. The i2s controller and ACP audio DMA engine are part of the GPU and both need updating so I would like to upstream the whole patch set via one tree if possible.

The current code is based on drm-next, but I'm happy to rebase on whatever tree this ends up going through if there are any problems applying. These patches touch both audio and drm.

The entire patch set can be viewed here:
https://cgit.freedesktop.org/~agd5f/linux/log/?h=stoney_acp2

Thanks!

Alex

v2:
- Patch 1 is already applied to the audio tree, just including it for completeness since it's required for this patch set and it's not yet in the drm tree.
- New patch to share asic types between gpu and audio drivers
- ACPI ID changed for rt5650 machine driver
- Integrate feedback on other patches

v3:
- resend remaining patches that have not been applied yet

Akshu Agrawal (2):
drm/amdgpu Moving amdgpu asic types to a separate file
ASoC: AMD: Add machine driver for cz rt5650

Vijendar Mukunda (4):
drm/amd/amdgpu: Added asic_type as ACP DMA driver platform data
ASoC: AMD: disabling memory gating in stoney platform
ASoC: AMD: DMA driver changes for Stoney Platform
ASoC: AMD: Audio buffer related changes for Stoney

 drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c  |   2 +
 drivers/gpu/drm/amd/include/amd_shared.h |  29 +---
 include/drm/amd_asic_type.h              |  52 +++++++
 sound/soc/amd/Kconfig                    |   7 +
 sound/soc/amd/Makefile                   |   2 +
 sound/soc/amd/acp-pcm-dma.c              | 243 ++++++++++++++++++++++++-------
 sound/soc/amd/acp-rt5645.c               | 210 ++++++++++++++++++++++++++
 sound/soc/amd/acp.h                      |   9 ++
 8 files changed, 472 insertions(+), 82 deletions(-)
 create mode 100644 include/drm/amd_asic_type.h
 create mode 100644 sound/soc/amd/acp-rt5645.c
Signature · 6 lines
2.5.5

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
#1 lore raw

PATCH v3 1/6 drm/amd/amdgpu: Added asic_type as ACP DMA driver platform data

From: Vijendar Mukunda <[email protected]>

asic_type information is passed to ACP DMA Driver as platform data.

---

v2: Removed asic_type local variable and directly passing asic_type instance to ACP DMA driver as platform data.
v3: Add R-b, resend.

 drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c | 2 ++
 sound/soc/amd/acp-pcm-dma.c             | 8 ++------
 sound/soc/amd/acp.h                     | 7 +++++++
 3 files changed, 11 insertions(+), 6 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c +2 -0
index a52795d..ebca223 100644
@@ -371,6 +371,8 @@ static int acp_hw_init(void *handle)
 	adev->acp.acp_cell[0].name = "acp_audio_dma";
 	adev->acp.acp_cell[0].num_resources = 4;
 	adev->acp.acp_cell[0].resources = &adev->acp.acp_res[0];
+	adev->acp.acp_cell[0].platform_data = &adev->asic_type;
+	adev->acp.acp_cell[0].pdata_size = sizeof(adev->asic_type);
 
 	adev->acp.acp_cell[1].name = "designware-i2s";
 	adev->acp.acp_cell[1].num_resources = 1;
sound/soc/amd/acp-pcm-dma.c +2 -6
index 08b1399..dcbf997 100644
@@ -73,12 +73,6 @@ static const struct snd_pcm_hardware acp_pcm_hardware_capture = {
 	.periods_max = CAPTURE_MAX_NUM_PERIODS,
 };
 
-struct audio_drv_data {
-	struct snd_pcm_substream *play_stream;
-	struct snd_pcm_substream *capture_stream;
-	void __iomem *acp_mmio;
-};
-
 static u32 acp_reg_read(void __iomem *acp_mmio, u32 reg)
 {
 	return readl(acp_mmio + (reg * 4));
@@ -916,6 +910,7 @@ static int acp_audio_probe(struct platform_device *pdev)
 	int status;
 	struct audio_drv_data *audio_drv_data;
 	struct resource *res;
+	const u32 *pdata = pdev->dev.platform_data;
 
 	audio_drv_data = devm_kzalloc(&pdev->dev, sizeof(struct audio_drv_data),
 					GFP_KERNEL);
@@ -932,6 +927,7 @@ static int acp_audio_probe(struct platform_device *pdev)
 
 	audio_drv_data->play_stream = NULL;
 	audio_drv_data->capture_stream = NULL;
+	audio_drv_data->asic_type =  *pdata;
 
 	res = platform_get_resource(pdev, IORESOURCE_IRQ, 0);
 	if (!res) {
sound/soc/amd/acp.h +7 -0
index 330832e..28cf914 100644
@@ -84,6 +84,13 @@ struct audio_substream_data {
 	void __iomem *acp_mmio;
 };
 
+struct audio_drv_data {
+	struct snd_pcm_substream *play_stream;
+	struct snd_pcm_substream *capture_stream;
+	void __iomem *acp_mmio;
+	u32 asic_type;
+};
+
 enum {
 	ACP_TILE_P1 = 0,
 	ACP_TILE_P2,
Signature · 6 lines
2.5.5

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#2 lore raw

PATCH v2 2/6 drm/amdgpu Moving amdgpu asic types to a separate file

In reply to Alex Deucher

From: Akshu Agrawal <[email protected]>

Amdgpu asic types will be required for other drivers too. Hence, its better to keep it in a separate include file.

---

v1: New patch to share asic_type definitions between GPU and audio driver.
v2: add R-b, resend

 drivers/gpu/drm/amd/include/amd_shared.h | 29 ++----------------
 include/drm/amd_asic_type.h              | 52 ++++++++++++++++++++++++++++++++
 2 files changed, 54 insertions(+), 27 deletions(-)
 create mode 100644 include/drm/amd_asic_type.h
drivers/gpu/drm/amd/include/amd_shared.h +2 -27
index 70e8c20..3a49fbd 100644
@@ -23,34 +23,9 @@
 #ifndef __AMD_SHARED_H__
 #define __AMD_SHARED_H__
 
-#define AMD_MAX_USEC_TIMEOUT		200000  /* 200 ms */
+#include <drm/amd_asic_type.h>
 
-/*
- * Supported ASIC types
- */
-enum amd_asic_type {
-	CHIP_TAHITI = 0,
-	CHIP_PITCAIRN,
-	CHIP_VERDE,
-	CHIP_OLAND,
-	CHIP_HAINAN,
-	CHIP_BONAIRE,
-	CHIP_KAVERI,
-	CHIP_KABINI,
-	CHIP_HAWAII,
-	CHIP_MULLINS,
-	CHIP_TOPAZ,
-	CHIP_TONGA,
-	CHIP_FIJI,
-	CHIP_CARRIZO,
-	CHIP_STONEY,
-	CHIP_POLARIS10,
-	CHIP_POLARIS11,
-	CHIP_POLARIS12,
-	CHIP_VEGA10,
-	CHIP_RAVEN,
-	CHIP_LAST,
-};
+#define AMD_MAX_USEC_TIMEOUT		200000  /* 200 ms */
 
 /*
  * Chip flags
include/drm/amd_asic_type.h new file +52 -0

52 lines not on this page · raw message

Signature · 6 lines
2.5.5

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

A message that is not in the archive (sent off-list or not yet received): 1503079830-16072-1-git-send-email-alexander.deucher-5C7GfCeVMHo@public.gmane.org

#4 lore raw

PATCH v3 3/6 ASoC: AMD: disabling memory gating in stoney platform

From: Vijendar Mukunda <[email protected]>

For Stoney platform, Memory gating is disabled.i.e SRAM Banks won't be turned off. By Default, SRAM Bank state set to ON. Added condition checks to skip SRAM Bank state set logic for Stoney platform.

---

v2: Added comments in code and removed locally defined macros
for STONEY and Carrizo.
v3: add R-b, resend

sound/soc/amd/acp-pcm-dma.c | 79 +++++++++++++++++++++++++++++++--------------

 1 file changed, 55 insertions(+), 24 deletions(-)
sound/soc/amd/acp-pcm-dma.c +55 -24
index dcbf997..f00b6b9 100644
@@ -20,7 +20,7 @@
 #include <linux/pm_runtime.h>
 
 #include <sound/soc.h>
-
+#include <drm/amd_asic_type.h>
 #include "acp.h"
 
 #define PLAYBACK_MIN_NUM_PERIODS    2
@@ -419,7 +419,7 @@ static void acp_set_sram_bank_state(void __iomem *acp_mmio, u16 bank,
 }
 
 /* Initialize and bring ACP hardware to default state. */
-static int acp_init(void __iomem *acp_mmio)
+static int acp_init(void __iomem *acp_mmio, u32 asic_type)
 {
 	u16 bank;
 	u32 val, count, sram_pte_offset;
@@ -493,9 +493,14 @@ static int acp_init(void __iomem *acp_mmio)
        /* When ACP_TILE_P1 is turned on, all SRAM banks get turned on.
 	* Now, turn off all of them. This can't be done in 'poweron' of
 	* ACP pm domain, as this requires ACP to be initialized.
+	* For Stoney, Memory gating is disabled,i.e SRAM Banks
+	* won't be turned off. The default state for SRAM banks is ON.
+	* Setting SRAM bank state code skipped for STONEY platform.
 	*/
-	for (bank = 1; bank < 48; bank++)
-		acp_set_sram_bank_state(acp_mmio, bank, false);
+	if (asic_type != CHIP_STONEY) {
+		for (bank = 1; bank < 48; bank++)
+			acp_set_sram_bank_state(acp_mmio, bank, false);
+	}
 
 	return 0;
 }
@@ -646,14 +651,22 @@ static int acp_dma_open(struct snd_pcm_substream *substream)
 
 	if (substream->stream == SNDRV_PCM_STREAM_PLAYBACK) {
 		intr_data->play_stream = substream;
-		for (bank = 1; bank <= 4; bank++)
-			acp_set_sram_bank_state(intr_data->acp_mmio, bank,
-						true);
+		/* For Stoney, Memory gating is disabled,i.e SRAM Banks
+		 * won't be turned off. The default state for SRAM banks is ON.
+		 * Setting SRAM bank state code skipped for STONEY platform.
+		 */
+		if (intr_data->asic_type != CHIP_STONEY) {
+			for (bank = 1; bank <= 4; bank++)
+				acp_set_sram_bank_state(intr_data->acp_mmio,
+							bank, true);
+		}
 	} else {
 		intr_data->capture_stream = substream;
-		for (bank = 5; bank <= 8; bank++)
-			acp_set_sram_bank_state(intr_data->acp_mmio, bank,
-						true);
+		if (intr_data->asic_type != CHIP_STONEY) {
+			for (bank = 5; bank <= 8; bank++)
+				acp_set_sram_bank_state(intr_data->acp_mmio,
+							bank, true);
+		}
 	}
 
 	return 0;
@@ -869,14 +882,23 @@ static int acp_dma_close(struct snd_pcm_substream *substream)
 
 	if (substream->stream == SNDRV_PCM_STREAM_PLAYBACK) {
 		adata->play_stream = NULL;
-		for (bank = 1; bank <= 4; bank++)
-			acp_set_sram_bank_state(adata->acp_mmio, bank,
-						false);
-	} else {
+		/* For Stoney, Memory gating is disabled,i.e SRAM Banks
+		 * won't be turned off. The default state for SRAM banks is ON.
+		 * Setting SRAM bank state code skipped for STONEY platform.
+		 * added condition checks for Carrizo platform only
+		 */
+		if (adata->asic_type != CHIP_STONEY) {
+			for (bank = 1; bank <= 4; bank++)
+				acp_set_sram_bank_state(adata->acp_mmio, bank,
+				false);
+		}
+	} else  {
 		adata->capture_stream = NULL;
-		for (bank = 5; bank <= 8; bank++)
-			acp_set_sram_bank_state(adata->acp_mmio, bank,
-						false);
+		if (adata->asic_type != CHIP_STONEY) {
+			for (bank = 5; bank <= 8; bank++)
+				acp_set_sram_bank_state(adata->acp_mmio, bank,
+						     false);
+		}
 	}
 
 	/* Disable ACP irq, when the current stream is being closed and
@@ -945,7 +967,7 @@ static int acp_audio_probe(struct platform_device *pdev)
 	dev_set_drvdata(&pdev->dev, audio_drv_data);
 
 	/* Initialize the ACP */
-	acp_init(audio_drv_data->acp_mmio);
+	acp_init(audio_drv_data->acp_mmio, audio_drv_data->asic_type);
 
 	status = snd_soc_register_platform(&pdev->dev, &acp_asoc_platform);
 	if (status != 0) {
@@ -976,19 +998,27 @@ static int acp_pcm_resume(struct device *dev)
 	u16 bank;
 	struct audio_drv_data *adata = dev_get_drvdata(dev);
 
-	acp_init(adata->acp_mmio);
+	acp_init(adata->acp_mmio, adata->asic_type);
 
 	if (adata->play_stream && adata->play_stream->runtime) {
-		for (bank = 1; bank <= 4; bank++)
-			acp_set_sram_bank_state(adata->acp_mmio, bank,
+		/* For Stoney, Memory gating is disabled,i.e SRAM Banks
+		 * won't be turned off. The default state for SRAM banks is ON.
+		 * Setting SRAM bank state code skipped for STONEY platform.
+		 */
+		if (adata->asic_type != CHIP_STONEY) {
+			for (bank = 1; bank <= 4; bank++)
+				acp_set_sram_bank_state(adata->acp_mmio, bank,
 						true);
+		}
 		config_acp_dma(adata->acp_mmio,
 				adata->play_stream->runtime->private_data);
 	}
 	if (adata->capture_stream && adata->capture_stream->runtime) {
-		for (bank = 5; bank <= 8; bank++)
-			acp_set_sram_bank_state(adata->acp_mmio, bank,
+		if (adata->asic_type != CHIP_STONEY) {
+			for (bank = 5; bank <= 8; bank++)
+				acp_set_sram_bank_state(adata->acp_mmio, bank,
 						true);
+		}
 		config_acp_dma(adata->acp_mmio,
 				adata->capture_stream->runtime->private_data);
 	}
@@ -1009,7 +1039,7 @@ static int acp_pcm_runtime_resume(struct device *dev)
 {
 	struct audio_drv_data *adata = dev_get_drvdata(dev);
 
-	acp_init(adata->acp_mmio);
+	acp_init(adata->acp_mmio, adata->asic_type);
 	acp_reg_write(1, adata->acp_mmio, mmACP_EXTERNAL_INTR_ENB);
 	return 0;
 }
@@ -1031,6 +1061,7 @@ static struct platform_driver acp_dma_driver = {
 
 module_platform_driver(acp_dma_driver);
 
+MODULE_AUTHOR("[email protected]");
 MODULE_AUTHOR("[email protected]");
 MODULE_DESCRIPTION("AMD ACP PCM Driver");
 MODULE_LICENSE("GPL v2");
Signature · 6 lines
2.5.5

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
maintainer Acked-by on v3 3/6
#5 lore raw

Re: [PATCH 3/6 v3] ASoC: AMD: disabling memory gating in stoney platform

Quoted 6 lines · Alex Deucher

From: Vijendar Mukunda <[email protected]>

For Stoney platform, Memory gating is disabled.i.e SRAM Banks won't be turned off. By Default, SRAM Bank state set to ON. Added condition checks to skip SRAM Bank state set logic for Stoney platform.

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#6 lore raw

PATCH v2 4/6 ASoC: AMD: DMA driver changes for Stoney Platform

In reply to a message not in the archive

From: Vijendar Mukunda <[email protected]>

Added DMA driver changes for Stoney platform. Below are the key differences between Stoney and CZ

In Stoney, Memory Gating is disabled.SRAM Banks won't be turned off.No Of SRAM Banks reduced to 6.
DAGB Garlic Interface used and 16 bit resolution is supported. SRAM bank 1 & SRAM bank 2 will be used for playback scenario. SRAM Bank 3 & SRAM Bank 4 will be used for Capture scenario.

---

v2: Added switch cases for asic type.
v3: Add R-b, resend

 sound/soc/amd/acp-pcm-dma.c | 87 +++++++++++++++++++++++++++++++++------------
 sound/soc/amd/acp.h         |  2 ++
 2 files changed, 67 insertions(+), 22 deletions(-)
sound/soc/amd/acp-pcm-dma.c +65 -22
index f00b6b9..f16e0b8 100644
@@ -137,8 +137,8 @@ static void config_dma_descriptor_in_sram(void __iomem *acp_mmio,
  * system memory <-> ACP SRAM
  */
 static void set_acp_sysmem_dma_descriptors(void __iomem *acp_mmio,
-					   u32 size, int direction,
-					   u32 pte_offset)
+					u32 size, int direction,
+					u32 pte_offset, u32 asic_type)
 {
 	u16 i;
 	u16 dma_dscr_idx = PLAYBACK_START_DMA_DESCR_CH12;
@@ -152,20 +152,42 @@ static void set_acp_sysmem_dma_descriptors(void __iomem *acp_mmio,
 					(size / 2) - (i * (size/2));
 			dmadscr[i].src = ACP_INTERNAL_APERTURE_WINDOW_0_ADDRESS
 				+ (pte_offset * SZ_4K) + (i * (size/2));
-			dmadscr[i].xfer_val |=
-			(ACP_DMA_ATTRIBUTES_DAGB_ONION_TO_SHAREDMEM << 16) |
-			(size / 2);
+			switch (asic_type) {
+			case CHIP_STONEY:
+				dmadscr[i].xfer_val |=
+				(ACP_DMA_ATTRIBUTES_DAGB_GARLIC_TO_SHAREDMEM  << 16) |
+				(size / 2);
+				break;
+			default:
+				dmadscr[i].xfer_val |=
+				(ACP_DMA_ATTRIBUTES_DAGB_ONION_TO_SHAREDMEM  << 16) |
+				(size / 2);
+			}
 		} else {
 			dma_dscr_idx = CAPTURE_START_DMA_DESCR_CH14 + i;
-			dmadscr[i].src = ACP_SHARED_RAM_BANK_5_ADDRESS +
-					(i * (size/2));
-			dmadscr[i].dest = ACP_INTERNAL_APERTURE_WINDOW_0_ADDRESS
-						+ (pte_offset * SZ_4K) +
-						(i * (size/2));
-			dmadscr[i].xfer_val |=
-			BIT(22) |
-			(ACP_DMA_ATTRIBUTES_SHAREDMEM_TO_DAGB_ONION << 16) |
-			(size / 2);
+			switch (asic_type) {
+			case CHIP_STONEY:
+				dmadscr[i].src = ACP_SHARED_RAM_BANK_3_ADDRESS +
+				(i * (size/2));
+				dmadscr[i].dest =
+				ACP_INTERNAL_APERTURE_WINDOW_0_ADDRESS +
+				(pte_offset * SZ_4K) + (i * (size/2));
+				dmadscr[i].xfer_val |=
+				BIT(22) |
+				(ACP_DMA_ATTRIBUTES_SHARED_MEM_TO_DAGB_GARLIC << 16) |
+				(size / 2);
+				break;
+			default:
+				dmadscr[i].src = ACP_SHARED_RAM_BANK_5_ADDRESS +
+				(i * (size/2));
+				dmadscr[i].dest =
+				ACP_INTERNAL_APERTURE_WINDOW_0_ADDRESS +
+				(pte_offset * SZ_4K) + (i * (size/2));
+				dmadscr[i].xfer_val |=
+				BIT(22) |
+				(ACP_DMA_ATTRIBUTES_SHAREDMEM_TO_DAGB_ONION << 16) |
+				(size / 2);
+			}
 		}
 		config_dma_descriptor_in_sram(acp_mmio, dma_dscr_idx,
 						&dmadscr[i]);
@@ -186,7 +208,8 @@ static void set_acp_sysmem_dma_descriptors(void __iomem *acp_mmio,
  * ACP SRAM <-> I2S
  */
 static void set_acp_to_i2s_dma_descriptors(void __iomem *acp_mmio,
-					   u32 size, int direction)
+					u32 size, int direction,
+					u32 asic_type)
 {
 
 	u16 i;
@@ -207,8 +230,17 @@ static void set_acp_to_i2s_dma_descriptors(void __iomem *acp_mmio,
 			dma_dscr_idx = CAPTURE_START_DMA_DESCR_CH15 + i;
 			/* dmadscr[i].src is unused by hardware. */
 			dmadscr[i].src = 0;
-			dmadscr[i].dest = ACP_SHARED_RAM_BANK_5_ADDRESS +
+			switch (asic_type) {
+			case CHIP_STONEY:
+				dmadscr[i].dest =
+					 ACP_SHARED_RAM_BANK_3_ADDRESS +
 					(i * (size / 2));
+				break;
+			default:
+				dmadscr[i].dest =
+					 ACP_SHARED_RAM_BANK_5_ADDRESS +
+					(i * (size / 2));
+			}
 			dmadscr[i].xfer_val |= BIT(22) |
 					(FROM_ACP_I2S_1 << 16) | (size / 2);
 		}
@@ -264,7 +296,8 @@ static void acp_pte_config(void __iomem *acp_mmio, struct page *pg,
 }
 
 static void config_acp_dma(void __iomem *acp_mmio,
-			   struct audio_substream_data *audio_config)
+			struct audio_substream_data *audio_config,
+			u32 asic_type)
 {
 	u32 pte_offset;
 
@@ -278,11 +311,11 @@ static void config_acp_dma(void __iomem *acp_mmio,
 
 	/* Configure System memory <-> ACP SRAM DMA descriptors */
 	set_acp_sysmem_dma_descriptors(acp_mmio, audio_config->size,
-				       audio_config->direction, pte_offset);
+				audio_config->direction, pte_offset, asic_type);
 
 	/* Configure ACP SRAM <-> I2S DMA descriptors */
 	set_acp_to_i2s_dma_descriptors(acp_mmio, audio_config->size,
-					audio_config->direction);
+				audio_config->direction, asic_type);
 }
 
 /* Start a given DMA channel transfer */
@@ -502,6 +535,12 @@ static int acp_init(void __iomem *acp_mmio, u32 asic_type)
 			acp_set_sram_bank_state(acp_mmio, bank, false);
 	}
 
+	/* Stoney supports 16bit resolution */
+	if (asic_type == CHIP_STONEY) {
+		val = acp_reg_read(acp_mmio, mmACP_I2S_16BIT_RESOLUTION_EN);
+		val |= 0x03;
+		acp_reg_write(val, acp_mmio, mmACP_I2S_16BIT_RESOLUTION_EN);
+	}
 	return 0;
 }
 
@@ -680,6 +719,8 @@ static int acp_dma_hw_params(struct snd_pcm_substream *substream,
 	struct page *pg;
 	struct snd_pcm_runtime *runtime;
 	struct audio_substream_data *rtd;
+	struct snd_soc_pcm_runtime *prtd = substream->private_data;
+	struct audio_drv_data *adata = dev_get_drvdata(prtd->platform->dev);
 
 	runtime = substream->runtime;
 	rtd = runtime->private_data;
@@ -707,7 +748,7 @@ static int acp_dma_hw_params(struct snd_pcm_substream *substream,
 		rtd->num_of_pages = PAGE_ALIGN(size) >> PAGE_SHIFT;
 		rtd->direction = substream->stream;
 
-		config_acp_dma(rtd->acp_mmio, rtd);
+		config_acp_dma(rtd->acp_mmio, rtd, adata->asic_type);
 		status = 0;
 	} else {
 		status = -ENOMEM;
@@ -1011,7 +1052,8 @@ static int acp_pcm_resume(struct device *dev)
 						true);
 		}
 		config_acp_dma(adata->acp_mmio,
-				adata->play_stream->runtime->private_data);
+			adata->play_stream->runtime->private_data,
+			adata->asic_type);
 	}
 	if (adata->capture_stream && adata->capture_stream->runtime) {
 		if (adata->asic_type != CHIP_STONEY) {
@@ -1020,7 +1062,8 @@ static int acp_pcm_resume(struct device *dev)
 						true);
 		}
 		config_acp_dma(adata->acp_mmio,
-				adata->capture_stream->runtime->private_data);
+			adata->capture_stream->runtime->private_data,
+			adata->asic_type);
 	}
 	acp_reg_write(1, adata->acp_mmio, mmACP_EXTERNAL_INTR_ENB);
 	return 0;
sound/soc/amd/acp.h +2 -0
index 28cf914..a330a99 100644
@@ -19,6 +19,7 @@
 
 /* Capture SRAM address (as a source in dma descriptor) */
 #define ACP_SHARED_RAM_BANK_5_ADDRESS		0x400A000
+#define ACP_SHARED_RAM_BANK_3_ADDRESS		0x4006000
 
 #define ACP_DMA_RESET_TIME			10000
 #define ACP_CLOCK_EN_TIME_OUT_VALUE		0x000000FF
@@ -67,6 +68,7 @@
 #define CAPTURE_START_DMA_DESCR_CH15 6
 #define CAPTURE_END_DMA_DESCR_CH15 7
 
+#define mmACP_I2S_16BIT_RESOLUTION_EN       0x5209
 enum acp_dma_priority_level {
 	/* 0x0 Specifies the DMA channel is given normal priority */
 	ACP_DMA_PRIORITY_LEVEL_NORMAL = 0x0,
Signature · 6 lines
2.5.5

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
#8 lore raw

PATCH v3 5/6 ASoC: AMD: Audio buffer related changes for Stoney

In reply to a message not in the archive

From: Vijendar Mukunda <[email protected]>

Stoney uses 16kb SRAM memory for playback and 16Kb for capture.Modified Max buffer size to have the correct mapping between System Memory and SRAM.

Added snd_pcm_hardware structures for playback and capture for Stoney.

---

v2: Added switch cases for asic type.
v3: add R-b, resend

 sound/soc/amd/acp-pcm-dma.c | 73 +++++++++++++++++++++++++++++++++++++++++----
 1 file changed, 68 insertions(+), 5 deletions(-)
sound/soc/amd/acp-pcm-dma.c +68 -5
index f16e0b8..a6def3b 100644
@@ -35,6 +35,10 @@
 #define MAX_BUFFER (PLAYBACK_MAX_PERIOD_SIZE * PLAYBACK_MAX_NUM_PERIODS)
 #define MIN_BUFFER MAX_BUFFER
 
+#define ST_PLAYBACK_MAX_PERIOD_SIZE 8192
+#define ST_CAPTURE_MAX_PERIOD_SIZE  8192
+#define ST_MAX_BUFFER (ST_PLAYBACK_MAX_PERIOD_SIZE * PLAYBACK_MAX_NUM_PERIODS)
+#define ST_MIN_BUFFER ST_MAX_BUFFER
 static const struct snd_pcm_hardware acp_pcm_hardware_playback = {
 	.info = SNDRV_PCM_INFO_INTERLEAVED |
 		SNDRV_PCM_INFO_BLOCK_TRANSFER | SNDRV_PCM_INFO_MMAP |
@@ -73,6 +77,44 @@ static const struct snd_pcm_hardware acp_pcm_hardware_capture = {
 	.periods_max = CAPTURE_MAX_NUM_PERIODS,
 };
 
+static const struct snd_pcm_hardware acp_st_pcm_hardware_playback = {
+	.info = SNDRV_PCM_INFO_INTERLEAVED |
+		SNDRV_PCM_INFO_BLOCK_TRANSFER | SNDRV_PCM_INFO_MMAP |
+		SNDRV_PCM_INFO_MMAP_VALID | SNDRV_PCM_INFO_BATCH |
+		SNDRV_PCM_INFO_PAUSE | SNDRV_PCM_INFO_RESUME,
+	.formats = SNDRV_PCM_FMTBIT_S16_LE |
+		SNDRV_PCM_FMTBIT_S24_LE | SNDRV_PCM_FMTBIT_S32_LE,
+	.channels_min = 1,
+	.channels_max = 8,
+	.rates = SNDRV_PCM_RATE_8000_96000,
+	.rate_min = 8000,
+	.rate_max = 96000,
+	.buffer_bytes_max = ST_MAX_BUFFER,
+	.period_bytes_min = PLAYBACK_MIN_PERIOD_SIZE,
+	.period_bytes_max = ST_PLAYBACK_MAX_PERIOD_SIZE,
+	.periods_min = PLAYBACK_MIN_NUM_PERIODS,
+	.periods_max = PLAYBACK_MAX_NUM_PERIODS,
+};
+
+static const struct snd_pcm_hardware acp_st_pcm_hardware_capture = {
+	.info = SNDRV_PCM_INFO_INTERLEAVED |
+		SNDRV_PCM_INFO_BLOCK_TRANSFER | SNDRV_PCM_INFO_MMAP |
+		SNDRV_PCM_INFO_MMAP_VALID | SNDRV_PCM_INFO_BATCH |
+		SNDRV_PCM_INFO_PAUSE | SNDRV_PCM_INFO_RESUME,
+	.formats = SNDRV_PCM_FMTBIT_S16_LE |
+		SNDRV_PCM_FMTBIT_S24_LE | SNDRV_PCM_FMTBIT_S32_LE,
+	.channels_min = 1,
+	.channels_max = 2,
+	.rates = SNDRV_PCM_RATE_8000_48000,
+	.rate_min = 8000,
+	.rate_max = 48000,
+	.buffer_bytes_max = ST_MAX_BUFFER,
+	.period_bytes_min = CAPTURE_MIN_PERIOD_SIZE,
+	.period_bytes_max = ST_CAPTURE_MAX_PERIOD_SIZE,
+	.periods_min = CAPTURE_MIN_NUM_PERIODS,
+	.periods_max = CAPTURE_MAX_NUM_PERIODS,
+};
+
 static u32 acp_reg_read(void __iomem *acp_mmio, u32 reg)
 {
 	return readl(acp_mmio + (reg * 4));
@@ -664,10 +706,23 @@ static int acp_dma_open(struct snd_pcm_substream *substream)
 	if (adata == NULL)
 		return -ENOMEM;
 
-	if (substream->stream == SNDRV_PCM_STREAM_PLAYBACK)
-		runtime->hw = acp_pcm_hardware_playback;
-	else
-		runtime->hw = acp_pcm_hardware_capture;
+	if (substream->stream == SNDRV_PCM_STREAM_PLAYBACK) {
+		switch (intr_data->asic_type) {
+		case CHIP_STONEY:
+			runtime->hw = acp_st_pcm_hardware_playback;
+			break;
+		default:
+			runtime->hw = acp_pcm_hardware_playback;
+		}
+	} else {
+		switch (intr_data->asic_type) {
+		case CHIP_STONEY:
+			runtime->hw = acp_st_pcm_hardware_capture;
+			break;
+		default:
+			runtime->hw = acp_pcm_hardware_capture;
+		}
+	}
 
 	ret = snd_pcm_hw_constraint_integer(runtime,
 					    SNDRV_PCM_HW_PARAM_PERIODS);
@@ -905,7 +960,15 @@ static int acp_dma_trigger(struct snd_pcm_substream *substream, int cmd)
 
 static int acp_dma_new(struct snd_soc_pcm_runtime *rtd)
 {
-	return snd_pcm_lib_preallocate_pages_for_all(rtd->pcm,
+	struct audio_drv_data *adata = dev_get_drvdata(rtd->platform->dev);
+
+	if (adata->asic_type == CHIP_STONEY)
+		return snd_pcm_lib_preallocate_pages_for_all(rtd->pcm,
+							SNDRV_DMA_TYPE_DEV,
+							NULL, ST_MIN_BUFFER,
+							ST_MAX_BUFFER);
+	else
+		return snd_pcm_lib_preallocate_pages_for_all(rtd->pcm,
 							SNDRV_DMA_TYPE_DEV,
 							NULL, MIN_BUFFER,
 							MAX_BUFFER);
Signature · 6 lines
2.5.5

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
maintainer
#9 lore raw

Re: [PATCH 5/6 v3] ASoC: AMD: Audio buffer related changes for Stoney

sound/soc/amd/acp-pcm-dma.c line 40

Quoted 3 lines · Alex Deucher
sound/soc/amd/acp-pcm-dma.c +3 -0
+#define ST_PLAYBACK_MAX_PERIOD_SIZE 8192
+#define ST_CAPTURE_MAX_PERIOD_SIZE  8192
+#define ST_MAX_BUFFER (ST_PLAYBACK_MAX_PERIOD_SIZE * PLAYBACK_MAX_NUM_PERIODS)

These defines will go wrong if the capture size is changed independently of the playback size. Either just use a single define for both or have separate defines for the maximum buffer as well as for the period.

Quoted 2 lines

+#define ST_MIN_BUFFER ST_MAX_BUFFER
static const struct snd_pcm_hardware acp_pcm_hardware_playback = {

Missing blank line between the defines and here.

Review

Quoted 2 lines
(quoted diff) +2 -0
+	if (adata->asic_type == CHIP_STONEY)
+		return snd_pcm_lib_preallocate_pages_for_all(rtd->pcm,

As I've said previously it's better to write these variant checks as switch statements, that way if more variants appear (or special handling is discovered for existing chips) the code can be modified more cleanly.

#10 lore raw

PATCH v3 6/6 ASoC: AMD: Add machine driver for cz rt5650

In reply to a message not in the archive

From: Akshu Agrawal <[email protected]>

The driver is used for AMD board using rt5650 codec.

---

v2: Change ACPI ID to AMDI1002
v3: add R-b, resend

 sound/soc/amd/Kconfig      |   7 ++
 sound/soc/amd/Makefile     |   2 +
 sound/soc/amd/acp-rt5645.c | 210 +++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 219 insertions(+)
 create mode 100644 sound/soc/amd/acp-rt5645.c
sound/soc/amd/Kconfig +7 -0
index 78187eb..eb0ae60 100644
@@ -2,3 +2,10 @@ config SND_SOC_AMD_ACP
 	tristate "AMD Audio Coprocessor support"
 	help
 	 This option enables ACP DMA support on AMD platform.
+config SND_SOC_AMD_CZ_RT5645_MACH
+	tristate "AMD CZ support for RT5645"
+	select SND_SOC_RT5645
+	select SND_SOC_AMD_ACP
+	depends on I2C_DESIGNWARE_PLATFORM
+	help
+	 This option enables machine driver for rt5645.
sound/soc/amd/Makefile +2 -0
index 1a66ec0..eed64ff 100644
@@ -1,3 +1,5 @@
 snd-soc-acp-pcm-objs	:= acp-pcm-dma.o
+snd-soc-acp-rt5645-mach-objs := acp-rt5645.o
 
 obj-$(CONFIG_SND_SOC_AMD_ACP) += snd-soc-acp-pcm.o
+obj-$(CONFIG_SND_SOC_AMD_CZ_RT5645_MACH) += snd-soc-acp-rt5645-mach.o
sound/soc/amd/acp-rt5645.c new file +210 -0

210 lines not on this page · raw message

Signature · 6 lines
2.5.5

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
maintainer
#11 lore raw

Re: [PATCH 6/6 v3] ASoC: AMD: Add machine driver for cz rt5650

sound/soc/amd/Kconfig config SND_SOC_AMD_ACP · line 6

Quoted 7 lines · Alex Deucher

+++ b/sound/soc/amd/Kconfig

sound/soc/amd/Kconfig +2 -0
@@ -2,3 +2,10 @@ config SND_SOC_AMD_ACP
 	tristate "AMD Audio Coprocessor support"
 	help
 	 This option enables ACP DMA support on AMD platform.
+config SND_SOC_AMD_CZ_RT5645_MACH
+	tristate "AMD CZ support for RT5645"

Missing blank line between the stanzas.

sound/soc/amd/Kconfig config SND_SOC_AMD_ACP · line 9

Quoted 3 lines
sound/soc/amd/Kconfig +3 -0
+	select SND_SOC_RT5645
+	select SND_SOC_AMD_ACP
+	depends on I2C_DESIGNWARE_PLATFORM

No system dependencies of any kind? Looking at this I'd expect at least CONFIG_ACPI || COMPILE_TEST. It's also unclear to me how the DesignWare device is going to be instantiated here or if that should be a direct depenency at all here.

sound/soc/amd/acp-rt5645.c line 171

Quoted 1 line
sound/soc/amd/acp-rt5645.c new file +1 -0
+	ret = snd_soc_register_card(card);

devm_snd_soc_register_card() and then you don't need the remove function.

sound/soc/amd/acp-rt5645.c line 195

Quoted 5 lines
sound/soc/amd/acp-rt5645.c new file +5 -0
+static const struct acpi_device_id cz_audio_acpi_match[] = {
+	{ "AMDI1002", 0 },
+	{},
+};
+

Missing MODULE_DEVICE_TABLE().

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#12 lore raw

Re: [PATCH 6/6 v3] ASoC: AMD: Add machine driver for cz rt5650

Quoted 11 lines · Mark Brown

sound/soc/amd/Kconfig config SND_SOC_AMD_ACP · line 6

Quoted 7 lines · Alex Deucher

+++ b/sound/soc/amd/Kconfig

sound/soc/amd/Kconfig +2 -0
@@ -2,3 +2,10 @@ config SND_SOC_AMD_ACP
  	tristate "AMD Audio Coprocessor support"
  	help
  	 This option enables ACP DMA support on AMD platform.
+config SND_SOC_AMD_CZ_RT5645_MACH
+	tristate "AMD CZ support for RT5645"

Missing blank line between the stanzas.

Done. Will push the change in next revision.

Quoted 10 lines

sound/soc/amd/Kconfig config SND_SOC_AMD_ACP · line 9

Quoted 4 lines
sound/soc/amd/Kconfig +3 -0
+	select SND_SOC_RT5645
+	select SND_SOC_AMD_ACP
+	depends on I2C_DESIGNWARE_PLATFORM

No system dependencies of any kind? Looking at this I'd expect at least CONFIG_ACPI || COMPILE_TEST. It's also unclear to me how the DesignWare device is going to be instantiated here or if that should be a direct depenency at all here.

Added I2C for system dependency and removed
I2C_DESIGNWARE_PLATFORM as dependency

Quoted 5 lines

sound/soc/amd/acp-rt5645.c line 171

Quoted 1 line
sound/soc/amd/acp-rt5645.c new file +1 -0
+	ret = snd_soc_register_card(card);

devm_snd_soc_register_card() and then you don't need the remove function.

Done.

Quoted 8 lines

sound/soc/amd/acp-rt5645.c line 195

Quoted 5 lines
sound/soc/amd/acp-rt5645.c new file +5 -0
+static const struct acpi_device_id cz_audio_acpi_match[] = {
+	{ "AMDI1002", 0 },
+	{},
+};
+

Missing MODULE_DEVICE_TABLE().

Done.

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#13 lore raw
In reply to a message not in the archive
Quoted 7 lines · Alex Deucher

This patch set updates the AMD GPU and Audio CoProcessor (ACP) audio drivers and the designware i2s driver for Stoney (ST). ST is an APU similar to Carrizo (CZ) which already has ACP audio support. The i2s controller and ACP audio DMA engine are part of the GPU and both need updating so I would like to upstream the whole patch set via one tree if possible.

Quoted 7 lines

The current code is based on drm-next, but I'm happy to rebase on whatever tree this ends up going through if there are any problems applying. These patches touch both audio and drm.

The entire patch set can be viewed here:
https://cgit.freedesktop.org/~agd5f/linux/log/?h=stoney_acp2

Any comments? Can this patch set go in? This is the second time I've resent it since the addressing the initial feedback. Does anyone have a preference on which tree?

Thanks,

Alex

Quoted 37 lines

Thanks!

Alex

v2:
- Patch 1 is already applied to the audio tree, just including it for completeness since it's required for this patch set and it's not yet in the drm tree.
- New patch to share asic types between gpu and audio drivers
- ACPI ID changed for rt5650 machine driver
- Integrate feedback on other patches

v3:
- resend remaining patches that have not been applied yet

Akshu Agrawal (2):
drm/amdgpu Moving amdgpu asic types to a separate file
ASoC: AMD: Add machine driver for cz rt5650

Vijendar Mukunda (4):
drm/amd/amdgpu: Added asic_type as ACP DMA driver platform data
ASoC: AMD: disabling memory gating in stoney platform
ASoC: AMD: DMA driver changes for Stoney Platform
ASoC: AMD: Audio buffer related changes for Stoney

 drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c  |   2 +
 drivers/gpu/drm/amd/include/amd_shared.h |  29 +---
 include/drm/amd_asic_type.h              |  52 +++++++
 sound/soc/amd/Kconfig                    |   7 +
 sound/soc/amd/Makefile                   |   2 +
 sound/soc/amd/acp-pcm-dma.c              | 243 ++++++++++++++++++++++++-------
 sound/soc/amd/acp-rt5645.c               | 210 ++++++++++++++++++++++++++
 sound/soc/amd/acp.h                      |   9 ++
 8 files changed, 472 insertions(+), 82 deletions(-)
 create mode 100644 include/drm/amd_asic_type.h
 create mode 100644 sound/soc/amd/acp-rt5645.c
Quoted 3 lines

--
2.5.5

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

maintainer
#14 lore raw
Quoted 3 lines · Alex Deucher

Any comments? Can this patch set go in? This is the second time I've resent it since the addressing the initial feedback. Does anyone have a preference on which tree?

You need to get someone from the DRM side to pay attention to the second patch and you need to stop resending the first patch since as has been pointed out a few times now you need to stop sending the first patch which is already in Linus' tree.

A message that is not in the archive (sent off-list or not yet received): 20170830151913.6vknazyhh4se5524-GFdadSzt00ze9xe1eoZjHA@public.gmane.org

#16 lore raw
Quoted 11 lines

-----Original Message-----
From: Mark Brown [mailto:[email protected]]
Sent: Wednesday, August 30, 2017 11:19 AM
To: Alex Deucher
Cc: amd-gfx list; Maling list - DRI developers; Dave Airlie; alsa-devel@alsa- project.org; Mukunda, Vijendar; rajeev kumar; [email protected]; Liam Girdwood; Takashi Iwai; Deucher, Alexander
Subject: Re: [PATCH 0/6 v3] Add ASoC support for AMD Stoney APUs

On Wed, Aug 30, 2017 at 09:33:35AM -0400, Alex Deucher wrote:

Quoted 8 lines
Quoted 3 lines

Any comments? Can this patch set go in? This is the second time I've resent it since the addressing the initial feedback. Does anyone have a preference on which tree?

You need to get someone from the DRM side to pay attention to the second patch and you need to stop resending the first patch since as has been pointed out a few times now you need to stop sending the first patch which is already in Linus' tree.

I didn't resent any patches that are already upstream in v3 of this patch set. The rest of the patches have dependencies between drm and audio. I'd like to get them in all together via one tree or the other.

Alex

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#17 lore raw
In reply to Mark Brown
Quoted 11 lines · Mark Brown

[ Unknown signature status ]

Quoted 3 lines · Alex Deucher

Any comments? Can this patch set go in? This is the second time I've resent it since the addressing the initial feedback. Does anyone have a preference on which tree?

You need to get someone from the DRM side to pay attention to the second patch and you need to stop resending the first patch since as has been pointed out a few times now you need to stop sending the first patch which is already in Linus' tree.

Alex *is* from the DRM side and has reviewed that patch.

maintainer
#18 lore raw
Quoted 8 lines · Eric Anholt
Quoted 4 lines · Mark Brown

You need to get someone from the DRM side to pay attention to the second patch and you need to stop resending the first patch since as has been pointed out a few times now you need to stop sending the first patch which is already in Linus' tree.

Alex *is* from the DRM side and has reviewed that patch.

*sigh* Right, OK in that case it would have been really helpful to see either some mention of this either in text or via a pull request (which would show the patch having been applied). It's difficult to keep track of all the different people who might be DRM maintainers especially when there's process things like the resend of the patch from Linus' tree going on.

#19 lore raw
Quoted 18 lines

-----Original Message-----
From: Mark Brown [mailto:[email protected]]
Sent: Wednesday, August 30, 2017 5:10 PM
To: Eric Anholt
Cc: Alex Deucher; [email protected]; Liam Girdwood; Maling list - DRI developers; rajeev kumar; amd-gfx list; Mukunda, Vijendar; Deucher, Alexander; [email protected]
Subject: Re: [PATCH 0/6 v3] Add ASoC support for AMD Stoney APUs

Quoted 3 lines · Eric Anholt
Quoted 1 line · Mark Brown

You need to get someone from the DRM side to pay attention to the

second

Quoted 3 lines
Quoted 3 lines

patch and you need to stop resending the first patch since as has been pointed out a few times now you need to stop sending the first patch which is already in Linus' tree.

Quoted 8 lines
Quoted 1 line

Alex *is* from the DRM side and has reviewed that patch.

*sigh* Right, OK in that case it would have been really helpful to see either some mention of this either in text or via a pull request (which would show the patch having been applied). It's difficult to keep track of all the different people who might be DRM maintainers especially when there's process things like the resend of the patch from Linus' tree going on.

I sent one patch in the v2 patch set that had already gone upstream because I didn't know that it had landed in Linus' tree yet. I had thought it had just landed in the audio tree. I apologize for that. On the v3 cover page, I mentioned that v3 was a resend of the patches that had not been applied to any tree yet; I did not resend any patches that were already applied. I believe all the previous comments were addressed. Now that we've clarified that, are there an outstanding objections to these patches? The patches touch both drm and audio. My preference would be to take them through the drm tree, but I'm happy to have them go through the audio tree if you prefer.

Thanks,

Alex

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

maintainer
#20 lore raw

On Wed, Aug 30, 2017 at 09:40:07PM +0000, Deucher, Alexander wrote:

Please fix your mail client to word wrap within paragraphs at something substantially less than 80 columns. Doing this makes your messages much easier to read and reply to.

Quoted 10 lines

I sent one patch in the v2 patch set that had already gone upstream because I didn't know that it had landed in Linus' tree yet. I had thought it had just landed in the audio tree. I apologize for that. On the v3 cover page, I mentioned that v3 was a resend of the patches that had not been applied to any tree yet; I did not resend any patches that were already applied. I believe all the previous comments were addressed. Now that we've clarified that, are there an outstanding objections to these patches? The patches touch both drm and audio. My preference would be to take them through the drm tree, but I'm happy to have them go through the audio tree if you prefer.

I haven't really looked at them yet because I was waiting for the DRM side of it to get sorted out (previous experience has been that DRM changes that need review can take a while to figure out). To be a bit clearer about what I said in the mail you're replying to you should just apply the initial DRM bits and send a pull request, this is a standard way of handling cross tree stuff and means that worst case there's fewer dependencies after the merge window.