On 20/12/2019 16:40, Dmitry Osipenko wrote:
20.12.2019 18:25, Ben Dooks пишет:
On 20/12/2019 15:02, Dmitry Osipenko wrote:
20.12.2019 17:56, Ben Dooks пишет:
On 20/12/2019 14:43, Dmitry Osipenko wrote:
20.12.2019 16:57, Jon Hunter пишет:
On 20/12/2019 11:38, Ben Dooks wrote: > On 20/12/2019 11:30, Jon Hunter wrote: >> >> On 25/11/2019 17:28, Dmitry Osipenko wrote: >>> 25.11.2019 20:22, Dmitry Osipenko пишет: >>>> 25.11.2019 13:37, Ben Dooks пишет: >>>>> On 23/11/2019 21:09, Dmitry Osipenko wrote: >>>>>> 18.10.2019 18:48, Ben Dooks пишет: >>>>>>> From: Edward Cragg edward.cragg@codethink.co.uk >>>>>>> >>>>>>> The tegra3 audio can support 24 and 32 bit sample sizes so add >>>>>>> the >>>>>>> option to the tegra30_i2s_hw_params to configure the S24_LE or >>>>>>> S32_LE >>>>>>> formats when requested. >>>>>>> >>>>>>> Signed-off-by: Edward Cragg edward.cragg@codethink.co.uk >>>>>>> [ben.dooks@codethink.co.uk: fixup merge of 24 and 32bit] >>>>>>> [ben.dooks@codethink.co.uk: add pm calls around ytdm config] >>>>>>> [ben.dooks@codethink.co.uk: drop debug printing to dev_dbg] >>>>>>> Signed-off-by: Ben Dooks ben.dooks@codethink.co.uk >>>>>>> --- >>>>>>> squash 5aeca5a055fd ASoC: tegra: i2s: pm_runtime_get_sync() is >>>>>>> needed >>>>>>> in tdm code >>>>>>> >>>>>>> ASoC: tegra: i2s: pm_runtime_get_sync() is needed in tdm code >>>>>>> --- >>>>>>> sound/soc/tegra/tegra30_i2s.c | 25 >>>>>>> ++++++++++++++++++++----- >>>>>>> 1 file changed, 20 insertions(+), 5 deletions(-) >>>>>>> >>>>>>> diff --git a/sound/soc/tegra/tegra30_i2s.c >>>>>>> b/sound/soc/tegra/tegra30_i2s.c >>>>>>> index 73f0dddeaef3..063f34c882af 100644 >>>>>>> --- a/sound/soc/tegra/tegra30_i2s.c >>>>>>> +++ b/sound/soc/tegra/tegra30_i2s.c >>>>>>> @@ -127,7 +127,7 @@ static int tegra30_i2s_hw_params(struct >>>>>>> snd_pcm_substream *substream, >>>>>>> struct device *dev = dai->dev; >>>>>>> struct tegra30_i2s *i2s = >>>>>>> snd_soc_dai_get_drvdata(dai); >>>>>>> unsigned int mask, val, reg; >>>>>>> - int ret, sample_size, srate, i2sclock, bitcnt; >>>>>>> + int ret, sample_size, srate, i2sclock, bitcnt, audio_bits; >>>>>>> struct tegra30_ahub_cif_conf cif_conf; >>>>>>> if (params_channels(params) != 2) >>>>>>> @@ -137,8 +137,19 @@ static int tegra30_i2s_hw_params(struct >>>>>>> snd_pcm_substream *substream, >>>>>>> switch (params_format(params)) { >>>>>>> case SNDRV_PCM_FORMAT_S16_LE: >>>>>>> val = TEGRA30_I2S_CTRL_BIT_SIZE_16; >>>>>>> + audio_bits = TEGRA30_AUDIOCIF_BITS_16; >>>>>>> sample_size = 16; >>>>>>> break; >>>>>>> + case SNDRV_PCM_FORMAT_S24_LE: >>>>>>> + val = TEGRA30_I2S_CTRL_BIT_SIZE_24; >>>>>>> + audio_bits = TEGRA30_AUDIOCIF_BITS_24; >>>>>>> + sample_size = 24; >>>>>>> + break; >>>>>>> + case SNDRV_PCM_FORMAT_S32_LE: >>>>>>> + val = TEGRA30_I2S_CTRL_BIT_SIZE_32; >>>>>>> + audio_bits = TEGRA30_AUDIOCIF_BITS_32; >>>>>>> + sample_size = 32; >>>>>>> + break; >>>>>>> default: >>>>>>> return -EINVAL; >>>>>>> } >>>>>>> @@ -170,8 +181,8 @@ static int tegra30_i2s_hw_params(struct >>>>>>> snd_pcm_substream *substream, >>>>>>> cif_conf.threshold = 0; >>>>>>> cif_conf.audio_channels = 2; >>>>>>> cif_conf.client_channels = 2; >>>>>>> - cif_conf.audio_bits = TEGRA30_AUDIOCIF_BITS_16; >>>>>>> - cif_conf.client_bits = TEGRA30_AUDIOCIF_BITS_16; >>>>>>> + cif_conf.audio_bits = audio_bits; >>>>>>> + cif_conf.client_bits = audio_bits; >>>>>>> cif_conf.expand = 0; >>>>>>> cif_conf.stereo_conv = 0; >>>>>>> cif_conf.replicate = 0; >>>>>>> @@ -306,14 +317,18 @@ static const struct snd_soc_dai_driver >>>>>>> tegra30_i2s_dai_template = { >>>>>>> .channels_min = 2, >>>>>>> .channels_max = 2, >>>>>>> .rates = SNDRV_PCM_RATE_8000_96000, >>>>>>> - .formats = SNDRV_PCM_FMTBIT_S16_LE, >>>>>>> + .formats = SNDRV_PCM_FMTBIT_S32_LE | >>>>>>> + SNDRV_PCM_FMTBIT_S24_LE | >>>>>>> + SNDRV_PCM_FMTBIT_S16_LE, >>>>>>> }, >>>>>>> .capture = { >>>>>>> .stream_name = "Capture", >>>>>>> .channels_min = 2, >>>>>>> .channels_max = 2, >>>>>>> .rates = SNDRV_PCM_RATE_8000_96000, >>>>>>> - .formats = SNDRV_PCM_FMTBIT_S16_LE, >>>>>>> + .formats = SNDRV_PCM_FMTBIT_S32_LE | >>>>>>> + SNDRV_PCM_FMTBIT_S24_LE | >>>>>>> + SNDRV_PCM_FMTBIT_S16_LE, >>>>>>> }, >>>>>>> .ops = &tegra30_i2s_dai_ops, >>>>>>> .symmetric_rates = 1, >>>>>>> >>>>>> >>>>>> Hello, >>>>>> >>>>>> This patch breaks audio on Tegra30. I don't see errors >>>>>> anywhere, but >>>>>> there is no audio and reverting this patch helps. Please fix it. >>>>> >>>>> What is the failure mode? I can try and take a look at this some >>>>> time >>>>> this week, but I am not sure if I have any boards with an actual >>>>> useful >>>>> audio output? >>>> >>>> The failure mode is that there no sound. I also noticed that video >>>> playback stutters a lot if movie file has audio track, seems >>>> something >>>> times out during of the audio playback. For now I don't have any >>>> more info. >>>> >>> >>> Oh, I didn't say how to reproduce it.. for example simply playing >>> big_buck_bunny_720p_h264.mov in MPV has the audio problem. >>> >>> https://download.blender.org/peach/bigbuckbunny_movies/big_buck_bunny_720p_h... >>> >>> >>> >> >> Given that the audio drivers uses regmap, it could be good to >> dump the >> I2S/AHUB registers while the clip if playing with and without this >> patch >> to see the differences. I am curious if the audio is now being >> played as >> 24 or 32-bit instead of 16-bit now these are available. >> >> You could also dump the hw_params to see the format while playing as >> well ... >> >> $ /proc/asound/<scard-name>/pcm0p/sub0/hw_params > > I suppose it is also possible that the codec isn't properly doing >16 > bits and the fact we now offer 24 and 32 could be an issue. I've not > got anything with an audio output on it that would be easy to test.
I thought I had tested on a Jetson TK1 (Tegra124) but it was sometime back. However, admittedly I may have only done simple 16-bit testing with speaker-test.
We do verify that all soundcards are detected and registered as expected during daily testing, but at the moment we don't have anything that verifies actual playback.
Please take a look at the attached logs.
I wonder if we are falling into FIFO configuration issues with the non-16 bit case.
Can you try adding the following two patches?
It is much better now! There is no horrible noise and no stuttering, but audio still has a "robotic" effect, like freq isn't correct.
I wonder if there's an issue with FIFO stoking? I should also possibly add the correctly stop FIFOs patch as well.
I'll be happy to try more patches.
Meanwhile sound on v5.5+ is broken and thus the incomplete patches should be reverted.
Have you checked if just removing the 24/32 bits from .formats in the dai template and see if the problem continues? I will try and see if I can find the code that does the fifo level checking and see if the problem is in the FIFO fill or something else in the audio hub setup.