Skip to contentSkip to conversation

This patch series enables DMA mode on Intel Keem Bay platform

PATCH In review Michael Sit Wei Hong 3 participants 10 messages Nov 16, 2020

Subsystems (5)
lore
Share & tools
lore
#0 lore raw

cover This patch series enables DMA mode on Intel Keem Bay platform

v1: Initial patch version, which contains fix for S24_LE format and also enable
DMA mode on Intel Keembay platform

Michael Sit Wei Hong (5):
ASoC: Intel: KMB: Fix S24_LE configuration
dt-bindings: sound: intel, keembay-i2s: Add info for device to use DMA
ASoC: soc-generic-dmaengine-pcm: Add custom prepare and submit
function
ASoC: dmaengine_pcm: expose functions to header file for custom
functions
ASoC: Intel: KMB: Enable DMA transfer mode

 .../bindings/sound/intel,keembay-i2s.yaml     |  14 ++
 include/sound/dmaengine_pcm.h                 |  21 ++
 sound/core/pcm_dmaengine.c                    |  46 ++--
 sound/soc/intel/Kconfig                       |   2 +
 sound/soc/intel/keembay/kmb_platform.c        | 208 ++++++++++++++++--
 sound/soc/intel/keembay/kmb_platform.h        |   9 +
 sound/soc/soc-generic-dmaengine-pcm.c         |   8 +-
 7 files changed, 270 insertions(+), 38 deletions(-)
Signature · 1 line
2.17.1
#1 lore raw

PATCH 1/5 ASoC: Intel: KMB: Fix S24_LE configuration

S24_LE is 24 bit audio in 32 bit container configuration Fixing the configuration to match the data arrangement of this audio format.

Fixes c5477e966728 ("ASoC: Intel: Add KeemBay platform driver")
---
 sound/soc/intel/keembay/kmb_platform.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
sound/soc/intel/keembay/kmb_platform.c +3 -3
index f54b710ee1c2..291a686568c2 100644
@@ -487,9 +487,9 @@ static int kmb_dai_hw_params(struct snd_pcm_substream *substream,
 		kmb_i2s->xfer_resolution = 0x02;
 		break;
 	case SNDRV_PCM_FORMAT_S24_LE:
-		config->data_width = 24;
-		kmb_i2s->ccr = 0x08;
-		kmb_i2s->xfer_resolution = 0x04;
+		config->data_width = 32;
+		kmb_i2s->ccr = 0x14;
+		kmb_i2s->xfer_resolution = 0x05;
 		break;
 	case SNDRV_PCM_FORMAT_S32_LE:
 		config->data_width = 32;
Signature · 1 line
2.17.1
#2 lore raw

PATCH 2/5 dt-bindings: sound: intel, keembay-i2s: Add info for device to use DMA

In reply to Michael Sit Wei Hong

Add descriptions for entries needed for audio device to use DMA channels for audio playback and capture.

---
 .../bindings/sound/intel,keembay-i2s.yaml          | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
Documentation/devicetree/bindings/sound/intel,keembay-i2s.yaml +14 -0
index d346e61ab708..e0658f122cbb 100644
@@ -45,6 +45,18 @@ properties:
       - const: osc
       - const: apb_clk
 
+  dmas:
+    items:
+      - description: DMA controller phandle and DMA channel
+                     for TX and RX
+
+  dma-names:
+    items:
+      - description: "tx" for the transmit channel
+                     "rx" for the receive channel
+      - const: tx
+      - const: rx
+
 required:
   - compatible
   - "#sound-dai-cells"
@@ -70,4 +82,6 @@ examples:
          interrupts = <GIC_SPI 120 IRQ_TYPE_LEVEL_HIGH>;
          clock-names = "osc", "apb_clk";
          clocks = <&scmi_clk KEEM_BAY_PSS_AUX_I2S3>, <&scmi_clk KEEM_BAY_PSS_I2S3>;
+         dmas = <&axi_dma0 29 &axi_dma0 33>;
+         dma-names = "tx", "rx";
      };
Signature · 1 line
2.17.1
#3 lore raw

PATCH 3/5 ASoC: soc-generic-dmaengine-pcm: Add custom prepare and submit function

In reply to Michael Sit Wei Hong

Enabling custom prepare and submit function to overcome DMA limitation.

In the Intel KeemBay solution, the DW AXI-based DMA has a limitation on the number of DMA blocks per transfer. In the case of 16 bit audio ASoC would allocate blocks exceeding the DMA block limitation.

The ASoC layers are not aware of such DMA limitation, and the DMA engine does not provide an API to set the maximum number of blocks per linked link.

This patch suggests an additional callback to let the caller check and modify the number of blocks per transfer to work-around the limitations.

---
 include/sound/dmaengine_pcm.h         |  6 ++++++
 sound/core/pcm_dmaengine.c            | 30 ++++++++++++++++++++++-----
 sound/soc/soc-generic-dmaengine-pcm.c |  8 ++++++-
 3 files changed, 38 insertions(+), 6 deletions(-)
include/sound/dmaengine_pcm.h +6 -0
index 8c5e38180fb0..9fae56d39ae2 100644
@@ -28,6 +28,9 @@ snd_pcm_substream_to_dma_direction(const struct snd_pcm_substream *substream)
 int snd_hwparams_to_dma_slave_config(const struct snd_pcm_substream *substream,
 	const struct snd_pcm_hw_params *params, struct dma_slave_config *slave_config);
 int snd_dmaengine_pcm_trigger(struct snd_pcm_substream *substream, int cmd);
+int snd_dmaengine_pcm_custom_trigger(struct snd_pcm_substream *substream, int cmd,
+	int (*custom_pcm_prepare_and_submit)(struct snd_pcm_substream *substream));
+
 snd_pcm_uframes_t snd_dmaengine_pcm_pointer(struct snd_pcm_substream *substream);
 snd_pcm_uframes_t snd_dmaengine_pcm_pointer_no_residue(struct snd_pcm_substream *substream);
 
@@ -113,6 +116,8 @@ int snd_dmaengine_pcm_refine_runtime_hwparams(
  *   which do not use devicetree.
  * @process: Callback used to apply processing on samples transferred from/to
  *   user space.
+ * @custom_pcm_prepare_and_submit: Callback used to work-around DMA limitations
+ *   related to link lists.
  * @compat_filter_fn: Will be used as the filter function when requesting a
  *  channel for platforms which do not use devicetree. The filter parameter
  *  will be the DAI's DMA data.
@@ -138,6 +143,7 @@ struct snd_dmaengine_pcm_config {
 	int (*process)(struct snd_pcm_substream *substream,
 		       int channel, unsigned long hwoff,
 		       void *buf, unsigned long bytes);
+	int (*custom_pcm_prepare_and_submit)(struct snd_pcm_substream *substream);
 	dma_filter_fn compat_filter_fn;
 	struct device *dma_dev;
 	const char *chan_names[SNDRV_PCM_STREAM_LAST + 1];
sound/core/pcm_dmaengine.c +25 -5
index 4d059ff2b2e4..cbd1429de509 100644
@@ -170,16 +170,20 @@ static int dmaengine_pcm_prepare_and_submit(struct snd_pcm_substream *substream)
 }
 
 /**
- * snd_dmaengine_pcm_trigger - dmaengine based PCM trigger implementation
+ * snd_dmaengine_pcm_custom_trigger - customized PCM trigger implementation to
+ *  work-around DMA limitations related to link lists.
  * @substream: PCM substream
  * @cmd: Trigger command
+ * @custom_pcm_prepare_and_submit: custom function to deal with DMA limitations
  *
  * Returns 0 on success, a negative error code otherwise.
  *
- * This function can be used as the PCM trigger callback for dmaengine based PCM
- * driver implementations.
+ * This function can be used as the PCM trigger callback for customized dmaengine
+ * based PCM driver implementations.
  */
-int snd_dmaengine_pcm_trigger(struct snd_pcm_substream *substream, int cmd)
+
+int snd_dmaengine_pcm_custom_trigger(struct snd_pcm_substream *substream, int cmd,
+	int (*custom_pcm_prepare_and_submit)(struct snd_pcm_substream *substream))
 {
 	struct dmaengine_pcm_runtime_data *prtd = substream_to_prtd(substream);
 	struct snd_pcm_runtime *runtime = substream->runtime;
@@ -187,7 +191,7 @@ int snd_dmaengine_pcm_trigger(struct snd_pcm_substream *substream, int cmd)
 
 	switch (cmd) {
 	case SNDRV_PCM_TRIGGER_START:
-		ret = dmaengine_pcm_prepare_and_submit(substream);
+		ret = custom_pcm_prepare_and_submit(substream);
 		if (ret)
 			return ret;
 		dma_async_issue_pending(prtd->dma_chan);
@@ -214,6 +218,22 @@ int snd_dmaengine_pcm_trigger(struct snd_pcm_substream *substream, int cmd)
 
 	return 0;
 }
+EXPORT_SYMBOL_GPL(snd_dmaengine_pcm_custom_trigger);
+
+/**
+ * snd_dmaengine_pcm_trigger - dmaengine based PCM trigger implementation
+ * @substream: PCM substream
+ * @cmd: Trigger command
+ *
+ * Returns 0 on success, a negative error code otherwise.
+ *
+ * This function can be used as the PCM trigger callback for dmaengine based PCM
+ * driver implementations.
+ */
+int snd_dmaengine_pcm_trigger(struct snd_pcm_substream *substream, int cmd)
+{
+	return snd_dmaengine_pcm_custom_trigger(substream, cmd, dmaengine_pcm_prepare_and_submit);
+}
 EXPORT_SYMBOL_GPL(snd_dmaengine_pcm_trigger);
 
 /**
sound/soc/soc-generic-dmaengine-pcm.c +7 -1
index 9ef80a48707e..88fca6402a36 100644
@@ -173,7 +173,13 @@ static int dmaengine_pcm_close(struct snd_soc_component *component,
 static int dmaengine_pcm_trigger(struct snd_soc_component *component,
 				 struct snd_pcm_substream *substream, int cmd)
 {
-	return snd_dmaengine_pcm_trigger(substream, cmd);
+	struct dmaengine_pcm *pcm = soc_component_to_pcm(component);
+
+	if (pcm->config && pcm->config->custom_pcm_prepare_and_submit)
+		return snd_dmaengine_pcm_custom_trigger(substream, cmd,
+							pcm->config->custom_pcm_prepare_and_submit);
+	else
+		return snd_dmaengine_pcm_trigger(substream, cmd);
 }
 
 static struct dma_chan *dmaengine_pcm_compat_request_channel(
Signature · 1 line
2.17.1
maintainer
#4 lore raw

Re: [PATCH 3/5] ASoC: soc-generic-dmaengine-pcm: Add custom prepare and submit function

Quoted 6 lines · Michael Sit Wei Hong

In the Intel KeemBay solution, the DW AXI-based DMA has a limitation on the number of DMA blocks per transfer. In the case of 16 bit audio ASoC would allocate blocks exceeding the DMA block limitation.

The ASoC layers are not aware of such DMA limitation, and the DMA engine does not provide an API to set the maximum number of blocks per linked link.

Can we not extend the dmaengine API so that the ASoC layer (and any other users) can become aware of this limitation and handle it appropriately rather than jumping straight to some client driver specific handling?

#5 lore raw

Re: [PATCH 3/5] ASoC: soc-generic-dmaengine-pcm: Add custom prepare and submit function

Quoted 6 lines · Mark Brown
Quoted 3 lines · Michael Sit Wei Hong

In the Intel KeemBay solution, the DW AXI-based DMA has a limitation on the number of DMA blocks per transfer. In the case of 16 bit audio ASoC would allocate blocks exceeding the DMA block limitation.

Quoted 7 lines
Quoted 2 lines

The ASoC layers are not aware of such DMA limitation, and the DMA engine does not provide an API to set the maximum number of blocks per linked link.

Can we not extend the dmaengine API so that the ASoC layer (and any other users) can become aware of this limitation and handle it appropriately rather than jumping straight to some client driver specific handling?

This was supposed to be an RFC, I asked Vinod/Lars to be copied for feedback. Unfortunately the RFC tag is missing and Vinod's email wasn't the right one... (fixed now).

This patchset suggests an ALSA-only quirk, having other more generic means to deal with this limitation would be fine - we just wanted to have a discussion on preferred directions. The IPs used are not Intel-specific so sooner or later someone else will have similar limitations to work-around.

maintainer
#6 lore raw

Re: [PATCH 3/5] ASoC: soc-generic-dmaengine-pcm: Add custom prepare and submit function

Quoted 8 lines · Pierre-Louis Bossart
Quoted 6 lines · Mark Brown

On Mon, Nov 16, 2020 at 02:19:03PM +0800, Michael Sit Wei Hong wrote:

Can we not extend the dmaengine API so that the ASoC layer (and any other users) can become aware of this limitation and handle it appropriately rather than jumping straight to some client driver specific handling?

Quoted 8 lines

This was supposed to be an RFC, I asked Vinod/Lars to be copied for feedback. Unfortunately the RFC tag is missing and Vinod's email wasn't the right one... (fixed now).

This patchset suggests an ALSA-only quirk, having other more generic means to deal with this limitation would be fine - we just wanted to have a discussion on preferred directions. The IPs used are not Intel-specific so sooner or later someone else will have similar limitations to work-around.

Generally with the dmaengine stuff we've added new query APIs to dmaengine and then used those in the ALSA/ASoC code to enumerate things, this certainly sounds like something that another device might have so it seems worth following that approach.

#7 lore raw

PATCH 4/5 ASoC: dmaengine_pcm: expose functions to header file for custom functions

In reply to Michael Sit Wei Hong

Moving some functions to the header file to be used by custom prepare and submit function.

In the Intel KeemBay solution, there is a DMA limitation which requires a custom prepare and submit function to modify the number of blocks per linked link.

This patch exposes some of the functions used in the pcm_dmaengine.c to be used by the custom function.

---
 include/sound/dmaengine_pcm.h | 15 +++++++++++++++
 sound/core/pcm_dmaengine.c    | 16 ++--------------
 2 files changed, 17 insertions(+), 14 deletions(-)
include/sound/dmaengine_pcm.h +15 -0
index 9fae56d39ae2..d45652a27f73 100644
@@ -174,8 +174,23 @@ struct dmaengine_pcm {
 	unsigned int flags;
 };
 
+struct dmaengine_pcm_runtime_data {
+	struct dma_chan *dma_chan;
+	dma_cookie_t cookie;
+
+	unsigned int pos;
+};
+
+static inline struct dmaengine_pcm_runtime_data *substream_to_prtd(
+	const struct snd_pcm_substream *substream)
+{
+	return substream->runtime->private_data;
+}
+
 static inline struct dmaengine_pcm *soc_component_to_pcm(struct snd_soc_component *p)
 {
 	return container_of(p, struct dmaengine_pcm, component);
 }
+
+void dmaengine_pcm_dma_complete(void *arg);
 #endif
sound/core/pcm_dmaengine.c +2 -14
index cbd1429de509..0f99c63964ec 100644
@@ -19,19 +19,6 @@
 
 #include <sound/dmaengine_pcm.h>
 
-struct dmaengine_pcm_runtime_data {
-	struct dma_chan *dma_chan;
-	dma_cookie_t cookie;
-
-	unsigned int pos;
-};
-
-static inline struct dmaengine_pcm_runtime_data *substream_to_prtd(
-	const struct snd_pcm_substream *substream)
-{
-	return substream->runtime->private_data;
-}
-
 struct dma_chan *snd_dmaengine_pcm_get_chan(struct snd_pcm_substream *substream)
 {
 	struct dmaengine_pcm_runtime_data *prtd = substream_to_prtd(substream);
@@ -128,7 +115,7 @@ void snd_dmaengine_pcm_set_config_from_dai_data(
 }
 EXPORT_SYMBOL_GPL(snd_dmaengine_pcm_set_config_from_dai_data);
 
-static void dmaengine_pcm_dma_complete(void *arg)
+void dmaengine_pcm_dma_complete(void *arg)
 {
 	struct snd_pcm_substream *substream = arg;
 	struct dmaengine_pcm_runtime_data *prtd = substream_to_prtd(substream);
@@ -139,6 +126,7 @@ static void dmaengine_pcm_dma_complete(void *arg)
 
 	snd_pcm_period_elapsed(substream);
 }
+EXPORT_SYMBOL_GPL(dmaengine_pcm_dma_complete);
 
 static int dmaengine_pcm_prepare_and_submit(struct snd_pcm_substream *substream)
 {
Signature · 1 line
2.17.1
#8 lore raw

PATCH 5/5 ASoC: Intel: KMB: Enable DMA transfer mode

In reply to Michael Sit Wei Hong

Enable DMA transfer mode for Intel Keem Bay ASoC platform driver.

The driver will search the device tree for DMA resources at boot time to enable DMA transfer mode, and will proceed to use DMA transfer if the resource is available, otherwise the default PIO mode will be used.

Due to DMA Block transfer limitation, a customized function is introduced to check and handle the limitation. Instead of limiting the maximum period bytes to the minimum supported, which forces the period size to less than 10.6ms in the worst case, adjusting the DMA to use a longer linked list will allow more flexible period sizes which does not force applications to use ridiculously small period sizes.

---
 sound/soc/intel/Kconfig                |   2 +
 sound/soc/intel/keembay/kmb_platform.c | 202 +++++++++++++++++++++++--
 sound/soc/intel/keembay/kmb_platform.h |   9 ++
 3 files changed, 198 insertions(+), 15 deletions(-)
sound/soc/intel/Kconfig +2 -0
index a5b446d5af19..6114dadfc52f 100644
@@ -200,6 +200,8 @@ config SND_SOC_INTEL_KEEMBAY
 	tristate "Keembay Platforms"
 	depends on ARM64 || COMPILE_TEST
 	depends on COMMON_CLK
+	select SND_DMAENGINE_PCM
+	select SND_SOC_GENERIC_DMAENGINE_PCM
 	help
 	  If you have a Intel Keembay platform then enable this option
 	  by saying Y or m.
sound/soc/intel/keembay/kmb_platform.c +187 -15

307 lines not on this page · raw message

sound/soc/intel/keembay/kmb_platform.h +9 -0
index 9756b132c12f..fd5341b66279 100644
@@ -12,6 +12,7 @@
 #include <linux/bits.h>
 #include <linux/bitfield.h>
 #include <linux/types.h>
+#include <sound/dmaengine_pcm.h>
 
 /* Register values with reference to KMB databook v1.1 */
 /* common register for all channel */
@@ -103,7 +104,12 @@
 #define DW_I2S_MASTER	BIT(3)
 
 #define I2S_RXDMA	0x01C0
+#define I2S_RRXDMA	0x01C4
 #define I2S_TXDMA	0x01C8
+#define I2S_RTXDMA	0x01CC
+#define I2S_DMACR	0x0200
+#define I2S_DMAEN_RXBLOCK	(1 << 16)
+#define I2S_DMAEN_TXBLOCK	(1 << 17)
 
 /*
  * struct i2s_clk_config_data - represent i2s clk configuration data
@@ -131,6 +137,9 @@ struct kmb_i2s_info {
 	u32 xfer_resolution;
 	u32 fifo_th;
 	bool master;
+	/* data related to DMA transfers b/w i2s and DMAC */
+	struct snd_dmaengine_dai_dma_data play_dma_data;
+	struct snd_dmaengine_dai_dma_data capture_dma_data;
 
 	struct i2s_clk_config_data config;
 	int (*i2s_clk_cfg)(struct i2s_clk_config_data *config);
Signature · 1 line
2.17.1
maintainer
#9 lore raw
In reply to Michael Sit Wei Hong
Quoted 12 lines · Michael Sit Wei Hong

v1: Initial patch version, which contains fix for S24_LE format and also enable
DMA mode on Intel Keembay platform

Michael Sit Wei Hong (5):
ASoC: Intel: KMB: Fix S24_LE configuration
dt-bindings: sound: intel, keembay-i2s: Add info for device to use DMA
ASoC: soc-generic-dmaengine-pcm: Add custom prepare and submit
function
ASoC: dmaengine_pcm: expose functions to header file for custom
functions
ASoC: Intel: KMB: Enable DMA transfer mode

Quoted 1 line

[...]

Applied to

https://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound.git for-next

Thanks!

[1/1] ASoC: Intel: KMB: Fix S24_LE configuration

      commit: 1bd7b0fc0165694897b7d2fb39751a07b98f6bf1

All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.

You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.

If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.

Please add any relevant lists and maintainers to the CCs when replying to this mail.

Thanks,
Mark