[PATCH v2 0/2] ASoC: meson: aiu: two fixes (for 5.16)
In this series I am proposing two fixes for the "aiu" driver, used on Amlogic Meson8, Meson8b, Meson8m2, GXBB, GXL and GXM SoCs.
The first patch is the result of me trying to understand the way how we get the DMA buffer and address for the audio data. I'm not an expert in terms of DMA. From what I understand we need to inform DMA core about the limitations of the hardware. In case of AIU it's DMA address registers only take 32 bits, so DMA core should be aware of this.
The second patch is what I could come up with to fix the infamous I2S buffer underrun issue, also called the "machine gun noise" (MGN) bug. After a lot of testing, debugging and comparing vendor code with the upstream "aiu" driver I have come up with this fix. I have written down my thoughts in the description of that patch. To be clear: these are my thoughts, unfortunately I have no way of proving this other than asking other people to test this patch (off-list I have already received positive feedback along with confirmation that both 2-ch and 6-ch audio are still working fine. Even with Kodi's menu - which is an easy way to reproduce the MGN bug - sound output is fine with this patch).
@Mark, @Liam: It would be great if at least the second patch could be queued as a fix for "for-5.16" as this solves a long standing buffer underrun. This issue is nasty because it can occur at any time and it can even loop forever. Christian provided an example with the speakers on low volume: [1]. Imagine the same sound on "medium" volume after playing a movie.
Changes since v1 at [0]: - Dropped a paragraph about FIFO IRQs from the second patch because Jerome has a valid point that this behavior can either mean that AIU_I2S_MISC is related to the FIFO or the FIFO consumer (= encoder) - Add a bit of documentation explaining why we set the AIU_I2S_MISC_FORCE_LEFT_RIGHT bit (affects patch #2) - Collect Christian's and Geraldo's Tested-by as well as Jerome's Acked-by for patch #2 (thanks to all three of you!) - Cc linux-stable for the second patch - Dropped RFC prefix
[0] https://chewitt.libreelec.tv/testing/wp2_audio_noise.mov [1] https://patchwork.kernel.org/project/linux-amlogic/cover/20211205180816.2083...
Martin Blumenstingl (2): ASoC: meson: aiu: fifo: Add missing dma_coerce_mask_and_coherent() ASoC: meson: aiu: Move AIU_I2S_MISC hold setting to aiu-fifo-i2s
sound/soc/meson/aiu-encoder-i2s.c | 33 ------------------------------- sound/soc/meson/aiu-fifo-i2s.c | 19 ++++++++++++++++++ sound/soc/meson/aiu-fifo.c | 6 ++++++ 3 files changed, 25 insertions(+), 33 deletions(-)
The FIFO registers which take an DMA-able address are only 32-bit wide on AIU. Add dma_coerce_mask_and_coherent() to make the DMA core aware of this limitation.
Fixes: 6ae9ca9ce986bf ("ASoC: meson: aiu: add i2s and spdif support") Signed-off-by: Martin Blumenstingl martin.blumenstingl@googlemail.com --- sound/soc/meson/aiu-fifo.c | 6 ++++++ 1 file changed, 6 insertions(+)
diff --git a/sound/soc/meson/aiu-fifo.c b/sound/soc/meson/aiu-fifo.c index 4ad23267cace..d67ff4cdabd5 100644 --- a/sound/soc/meson/aiu-fifo.c +++ b/sound/soc/meson/aiu-fifo.c @@ -5,6 +5,7 @@
#include <linux/bitfield.h> #include <linux/clk.h> +#include <linux/dma-mapping.h> #include <sound/pcm_params.h> #include <sound/soc.h> #include <sound/soc-dai.h> @@ -179,6 +180,11 @@ int aiu_fifo_pcm_new(struct snd_soc_pcm_runtime *rtd, struct snd_card *card = rtd->card->snd_card; struct aiu_fifo *fifo = dai->playback_dma_data; size_t size = fifo->pcm->buffer_bytes_max; + int ret; + + ret = dma_coerce_mask_and_coherent(card->dev, DMA_BIT_MASK(32)); + if (ret) + return ret;
snd_pcm_set_managed_buffer_all(rtd->pcm, SNDRV_DMA_TYPE_DEV, card->dev, size, size);
The out-of-tree vendor driver uses the following approach to set the AIU_I2S_MISC register: 1) write AIU_MEM_I2S_START_PTR and AIU_MEM_I2S_RD_PTR 2) configure AIU_I2S_MUTE_SWAP[15:0] 3) write AIU_MEM_I2S_END_PTR 4) set AIU_I2S_MISC[2] to 1 (documented as: "put I2S interface in hold mode") 5) set AIU_I2S_MISC[4] to 1 (depending on the driver revision it always stays at 1 while for older drivers this bit is unset in step 4) 6) set AIU_I2S_MISC[2] to 0 7) write AIU_MEM_I2S_MASKS 8) toggle AIU_MEM_I2S_CONTROL[0] 9) toggle AIU_MEM_I2S_BUF_CNTL[0]
Move setting the AIU_I2S_MISC[2] bit to aiu_fifo_i2s_hw_params() so it resembles the flow in the vendor kernel more closely. While here also configure AIU_I2S_MISC[4] (documented as: "force each audio data to left or right according to the bit attached with the audio data") similar to how the vendor driver does this. This fixes the infamous and long-standing "machine gun noise" issue (a buffer underrun issue).
Fixes: 6ae9ca9ce986bf ("ASoC: meson: aiu: add i2s and spdif support") Reported-by: Christian Hewitt christianshewitt@gmail.com Reported-by: Geraldo Nascimento geraldogabriel@gmail.com Tested-by: Christian Hewitt christianshewitt@gmail.com Tested-by: Geraldo Nascimento geraldogabriel@gmail.com Acked-by: Jerome Brunet jbrunet@baylibre.com Cc: stable@vger.kernel.org Signed-off-by: Martin Blumenstingl martin.blumenstingl@googlemail.com --- sound/soc/meson/aiu-encoder-i2s.c | 33 ------------------------------- sound/soc/meson/aiu-fifo-i2s.c | 19 ++++++++++++++++++ 2 files changed, 19 insertions(+), 33 deletions(-)
diff --git a/sound/soc/meson/aiu-encoder-i2s.c b/sound/soc/meson/aiu-encoder-i2s.c index 932224552146..67729de41a73 100644 --- a/sound/soc/meson/aiu-encoder-i2s.c +++ b/sound/soc/meson/aiu-encoder-i2s.c @@ -18,7 +18,6 @@ #define AIU_RST_SOFT_I2S_FAST BIT(0)
#define AIU_I2S_DAC_CFG_MSB_FIRST BIT(2) -#define AIU_I2S_MISC_HOLD_EN BIT(2) #define AIU_CLK_CTRL_I2S_DIV_EN BIT(0) #define AIU_CLK_CTRL_I2S_DIV GENMASK(3, 2) #define AIU_CLK_CTRL_AOCLK_INVERT BIT(6) @@ -36,37 +35,6 @@ static void aiu_encoder_i2s_divider_enable(struct snd_soc_component *component, enable ? AIU_CLK_CTRL_I2S_DIV_EN : 0); }
-static void aiu_encoder_i2s_hold(struct snd_soc_component *component, - bool enable) -{ - snd_soc_component_update_bits(component, AIU_I2S_MISC, - AIU_I2S_MISC_HOLD_EN, - enable ? AIU_I2S_MISC_HOLD_EN : 0); -} - -static int aiu_encoder_i2s_trigger(struct snd_pcm_substream *substream, int cmd, - struct snd_soc_dai *dai) -{ - struct snd_soc_component *component = dai->component; - - switch (cmd) { - case SNDRV_PCM_TRIGGER_START: - case SNDRV_PCM_TRIGGER_RESUME: - case SNDRV_PCM_TRIGGER_PAUSE_RELEASE: - aiu_encoder_i2s_hold(component, false); - return 0; - - case SNDRV_PCM_TRIGGER_STOP: - case SNDRV_PCM_TRIGGER_SUSPEND: - case SNDRV_PCM_TRIGGER_PAUSE_PUSH: - aiu_encoder_i2s_hold(component, true); - return 0; - - default: - return -EINVAL; - } -} - static int aiu_encoder_i2s_setup_desc(struct snd_soc_component *component, struct snd_pcm_hw_params *params) { @@ -353,7 +321,6 @@ static void aiu_encoder_i2s_shutdown(struct snd_pcm_substream *substream, }
const struct snd_soc_dai_ops aiu_encoder_i2s_dai_ops = { - .trigger = aiu_encoder_i2s_trigger, .hw_params = aiu_encoder_i2s_hw_params, .hw_free = aiu_encoder_i2s_hw_free, .set_fmt = aiu_encoder_i2s_set_fmt, diff --git a/sound/soc/meson/aiu-fifo-i2s.c b/sound/soc/meson/aiu-fifo-i2s.c index 2388a2d0b3a6..57e6e7160d2f 100644 --- a/sound/soc/meson/aiu-fifo-i2s.c +++ b/sound/soc/meson/aiu-fifo-i2s.c @@ -20,6 +20,8 @@ #define AIU_MEM_I2S_CONTROL_MODE_16BIT BIT(6) #define AIU_MEM_I2S_BUF_CNTL_INIT BIT(0) #define AIU_RST_SOFT_I2S_FAST BIT(0) +#define AIU_I2S_MISC_HOLD_EN BIT(2) +#define AIU_I2S_MISC_FORCE_LEFT_RIGHT BIT(4)
#define AIU_FIFO_I2S_BLOCK 256
@@ -90,6 +92,10 @@ static int aiu_fifo_i2s_hw_params(struct snd_pcm_substream *substream, unsigned int val; int ret;
+ snd_soc_component_update_bits(component, AIU_I2S_MISC, + AIU_I2S_MISC_HOLD_EN, + AIU_I2S_MISC_HOLD_EN); + ret = aiu_fifo_hw_params(substream, params, dai); if (ret) return ret; @@ -117,6 +123,19 @@ static int aiu_fifo_i2s_hw_params(struct snd_pcm_substream *substream, snd_soc_component_update_bits(component, AIU_MEM_I2S_MASKS, AIU_MEM_I2S_MASKS_IRQ_BLOCK, val);
+ /* + * Most (all?) supported SoCs have this bit set by default. The vendor + * driver however sets it manually (depending on the version either + * while un-setting AIU_I2S_MISC_HOLD_EN or right before that). Follow + * the same approach for consistency with the vendor driver. + */ + snd_soc_component_update_bits(component, AIU_I2S_MISC, + AIU_I2S_MISC_FORCE_LEFT_RIGHT, + AIU_I2S_MISC_FORCE_LEFT_RIGHT); + + snd_soc_component_update_bits(component, AIU_I2S_MISC, + AIU_I2S_MISC_HOLD_EN, 0); + return 0; }
On Mon, 6 Dec 2021 22:08:02 +0100, Martin Blumenstingl wrote:
In this series I am proposing two fixes for the "aiu" driver, used on Amlogic Meson8, Meson8b, Meson8m2, GXBB, GXL and GXM SoCs.
The first patch is the result of me trying to understand the way how we get the DMA buffer and address for the audio data. I'm not an expert in terms of DMA. From what I understand we need to inform DMA core about the limitations of the hardware. In case of AIU it's DMA address registers only take 32 bits, so DMA core should be aware of this.
[...]
Applied to
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound.git for-linus
Thanks!
[1/2] ASoC: meson: aiu: fifo: Add missing dma_coerce_mask_and_coherent() commit: 1bcd326631dc4faa3322d60b4fc45e8b3747993e [2/2] ASoC: meson: aiu: Move AIU_I2S_MISC hold setting to aiu-fifo-i2s commit: ee907afb0c39a41ee74b862882cfe12820c74b98
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
participants (2)
-
Mark Brown -
Martin Blumenstingl