ASoC: Add Multi CPU DAI support
v6 cover ASoC: Add Multi CPU DAI support
As discussed in [1], ASoC core supports multi codec DAIs on a DAI link. However it does not do so for CPU DAIs.
So, add support for multi CPU DAIs on a DAI Link by adding multi CPU DAI in Card instantiation, suspend and resume functions, PCM ops, stream handling functions and DAPM.
[1]: https://www.spinics.net/lists/alsa-devel/msg71369.html
changes in v6:
- Addressed review comments from Pierre in calculating delay
for CPU DAI, add check for compress DAIs and other nits
- add proper exit check in dpcm_prune_path() for CPU DAIs
- rebased on latest for-next branch
Shreyas NC (3):
ASoC: Add initial support for multiple CPU DAIs
ASoC: Add multiple CPU DAI support for PCM ops
ASoC: Add multiple CPU DAI support in DAPM
include/sound/soc.h | 6 + sound/soc/soc-core.c | 290 +++++++++++++++++++++--------- sound/soc/soc-dapm.c | 71 +++++--- sound/soc/soc-pcm.c | 491 ++++++++++++++++++++++++++++++++++----------------- 4 files changed, 589 insertions(+), 269 deletions(-)
Signature · 1 line
2.7.4
PATCH v6 1/3 ASoC: Add initial support for multiple CPU DAIs
ASoC core supports multiple codec DAIs but supports only a CPU DAI. To support multiple cpu DAIs, add cpu_dai and num_cpu_dai in snd_soc_dai_link and snd_soc_pcm_runtime structures similar to support for codec_dai. This is intended as a preparatory patch to eventually support the unification of the Codec and CPU DAI.
Inline with multiple codec DAI approach, add support to allocate, init, bind and probe multiple cpu_dai on init if driver specifies that. Also add support to loop over multiple cpu_dai during suspend and resume.
This is intended as a preparatory patch to eventually unify the CPU and Codec DAI into DAI components.
--- include/sound/soc.h | 6 ++ sound/soc/soc-core.c | 290 ++++++++++++++++++++++++++++++++++++--------------- 2 files changed, 211 insertions(+), 85 deletions(-)
include/sound/soc.h +6 -0
@@ -903,6 +903,9 @@ struct snd_soc_dai_link {struct snd_soc_dai_link_component *codecs;unsigned int num_codecs; +struct snd_soc_dai_link_component *cpu_dai; +unsigned int num_cpu_dai; +/* * You MAY specify the link's platform/PCM/DMA driver, either by * device name, or by DT/OF node, but not both. Some forms of link
@@ -1125,6 +1128,9 @@ struct snd_soc_pcm_runtime {struct snd_soc_dai **codec_dais;unsigned int num_codecs; +struct snd_soc_dai **cpu_dais; +unsigned int num_cpu_dai; +struct delayed_work delayed_work;#ifdef CONFIG_DEBUG_FS struct dentry *debugfs_dpcm_root;
sound/soc/soc-core.c +205 -85
480 lines not on this page · raw message
Signature · 1 line
2.7.4
Re: [PATCH v6 1/3] ASoC: Add initial support for multiple CPU DAIs
Quoted 17 lines
/* close any waiting streams */
sound/soc/soc-core.c +3
-3@@ -552,16 +565,21 @@ int snd_soc_suspend(struct device *dev) }list_for_each_entry (rtd, &card->rtd_list, list) { -struct snd_soc_dai *cpu_dai = rtd->cpu_dai; +struct snd_soc_dai *cpu_dai;if (rtd->dai_link->ignore_suspend)continue ; -if (cpu_dai->driver->suspend && cpu_dai->driver->bus_control) - cpu_dai->driver->suspend (cpu_dai); +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai = rtd->cpu_dais[i];
sound/soc/soc-core.c int snd_soc_suspend(struct device *dev) · line 582
Quoted 10 lines
sound/soc/soc-core.c +7
-2-/* deactivate pins to sleep state */ -pinctrl_pm_select_sleep_state (cpu_dai->dev); +if (cpu_dai->driver->suspend && + cpu_dai->driver->bus_control) + cpu_dai->driver->suspend (cpu_dai); + +/* deactivate pins to sleep state */ +pinctrl_pm_select_sleep_state (cpu_dai->dev); + }
I am not sure I get the relationship between cpu_dai and pins. Is this always safe/ok to play with the pincrtl before all cpu_dais are suspended? Or should you implement this with two loops, one to suspend and one to deactivate pins?
Quoted 25 lines
sound/soc/soc-core.c +6
-2@@ -845,35 +882,47 @@ static int soc_bind_dai_link(struct snd_soc_card *card, {struct snd_soc_pcm_runtime *rtd;struct snd_soc_dai_link_component *codecs = dai_link->codecs; -struct snd_soc_dai_link_component cpu_dai_component; +struct snd_soc_dai_link_component *cpu_dai_component;struct snd_soc_component *component; -struct snd_soc_dai **codec_dais; +struct snd_soc_dai **codec_dais, **cpu_dais;struct device_node *platform_of_node;const char *platform_name;int i;dev_dbg (card->dev,"ASoC: binding %s\n" , dai_link->name); + cpu_dai_component = dai_link->cpu_dai; +if (soc_is_dai_link_bound (card, dai_link)) {dev_dbg (card->dev,"ASoC: dai link %s already bound\n" , dai_link->name);return 0 ; } +if (dai_link->dynamic && dai_link->num_cpu_dai >1 ) { +dev_err (card->dev,"ASoC: Multi CPU DAI not supported for FE" );
sound/soc/soc-core.c static int soc_bind_dai_link(struct snd_soc_card *card, · line 869
Quoted 10 lines
sound/soc/soc-core.c +3
-3+return -EINVAL ; + } + rtd =soc_new_pcm_runtime (card, dai_link);if (!rtd)return -ENOMEM ; - cpu_dai_component.name = dai_link->cpu_name; - cpu_dai_component.of_node = dai_link->cpu_of_node; - cpu_dai_component.dai_name = dai_link->cpu_dai_name;
Have you checked if there are any side effects of using cpu_dai_component = dai_link->cpu_dai;
instead of a local variable?
if you are not sure, it may be worth implementing as a separate patch first before introducing the multi-cpu part, at least to allow for git bisect to identify issues?
Quoted 7 lines
sound/soc/soc-core.c +6
-0+static int snd_soc_init_single_cpu_dai (struct snd_soc_card *card, +struct snd_soc_dai_link *dai_link) +{ +if (dai_link->cpu_name || dai_link->cpu_of_node || + dai_link->cpu_dai_name) { + dai_link->num_cpu_dai =1 ;
sound/soc/soc-core.c static int snd_soc_init_multicodec(struct snd_soc_card *card, · line 1111
Quoted 10 lines
sound/soc/soc-core.c +10
-0+ dai_link->cpu_dai =devm_kzalloc (card->dev, +sizeof (struct snd_soc_dai_link_component ), +GFP_KERNEL ); + +if (!dai_link->cpu_dai) +return -ENOMEM ; + + dai_link->cpu_dai[0 ].name = dai_link->cpu_name; + dai_link->cpu_dai[0 ].of_node = dai_link->cpu_of_node; + dai_link->cpu_dai[0 ].dai_name = dai_link->cpu_dai_name;
Question: is cpu_dai[i].of_node defined for i>0 in the multi cpu_dai case?
Quoted 10 lines
sound/soc/soc-core.c +1
-1@@ -1644,7 +1751,7 @@ int snd_soc_runtime_set_dai_fmt(struct snd_soc_pcm_runtime *rtd, unsigned int dai_fmt) {struct snd_soc_dai **codec_dais = rtd->codec_dais; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai; +struct snd_soc_dai **cpu_dais = rtd->cpu_dais;unsigned int i;int ret;
sound/soc/soc-core.c int snd_soc_runtime_set_dai_fmt(struct snd_soc_pcm_runtime *rtd, · line 1662
Quoted 4 lines
sound/soc/soc-core.c +0
-1@@ -1659,35 +1766,44 @@ int snd_soc_runtime_set_dai_fmt(struct snd_soc_pcm_runtime *rtd, } } -/* Flip the polarity for the "CPU" end of a CODEC<->CODEC link */
why was this comment removed?
sound/soc/soc-core.c int snd_soc_runtime_set_dai_fmt(struct snd_soc_pcm_runtime *rtd, · line 1664
Quoted 2 lines
sound/soc/soc-core.c +0
-1/* the component which has non_legacy_dai_naming is Codec */ -if (cpu_dai->component->driver->non_legacy_dai_naming) {
Not sure if the code refactoring below makes sense in a codec-codec link, you probably wouldn't have multiple cpu_dais then, would you?
Quoted 63 lines
sound/soc/soc-core.c +34
-23-unsigned int inv_dai_fmt; +for (i =0 ; i < rtd->num_cpu_dai; i++) { +struct snd_soc_dai *cpu_dai = cpu_dais[i]; +unsigned int inv_dai_fmt, temp_dai_fmt; - inv_dai_fmt = dai_fmt & ~SND_SOC_DAIFMT_MASTER_MASK ; -switch (dai_fmt &SND_SOC_DAIFMT_MASTER_MASK ) { -case SND_SOC_DAIFMT_CBM_CFM : - inv_dai_fmt |=SND_SOC_DAIFMT_CBS_CFS ; -break ; -case SND_SOC_DAIFMT_CBM_CFS : - inv_dai_fmt |=SND_SOC_DAIFMT_CBS_CFM ; -break ; -case SND_SOC_DAIFMT_CBS_CFM : - inv_dai_fmt |=SND_SOC_DAIFMT_CBM_CFS ; -break ; -case SND_SOC_DAIFMT_CBS_CFS : - inv_dai_fmt |=SND_SOC_DAIFMT_CBM_CFM ; -break ; - } + temp_dai_fmt = dai_fmt; +if (cpu_dai->component->driver->non_legacy_dai_naming) { - dai_fmt = inv_dai_fmt; - } + inv_dai_fmt = dai_fmt & ~SND_SOC_DAIFMT_MASTER_MASK ; - ret =snd_soc_dai_set_fmt (cpu_dai, dai_fmt); -if (ret !=0 && ret != -ENOTSUPP ) { -dev_warn (cpu_dai->dev, -"ASoC: Failed to set DAI format: %d\n" , ret); -return ret; +switch (dai_fmt &SND_SOC_DAIFMT_MASTER_MASK ) { +case SND_SOC_DAIFMT_CBM_CFM : + inv_dai_fmt |=SND_SOC_DAIFMT_CBS_CFS ; +break ; + +case SND_SOC_DAIFMT_CBM_CFS : + inv_dai_fmt |=SND_SOC_DAIFMT_CBS_CFM ; +break ; + +case SND_SOC_DAIFMT_CBS_CFM : + inv_dai_fmt |=SND_SOC_DAIFMT_CBM_CFS ; +break ; + +case SND_SOC_DAIFMT_CBS_CFS : + inv_dai_fmt |=SND_SOC_DAIFMT_CBM_CFM ; +break ; + + } + + temp_dai_fmt = inv_dai_fmt; + } + + ret =snd_soc_dai_set_fmt (cpu_dai, temp_dai_fmt); +if (ret !=0 && ret != -ENOTSUPP ) { +dev_warn (cpu_dai->dev, +"ASoC: Failed to set DAI format: %d\n" , ret); +return ret; + } }
Re: [PATCH v6 1/3] ASoC: Add initial support for multiple CPU DAIs
Quoted 19 lines · Pierre-Louis Bossart
Quoted 19 lines
sound/soc/soc-core.c +6
-5/* close any waiting streams */ @@ -552,16 +565,21 @@ int snd_soc_suspend(struct device *dev) }list_for_each_entry (rtd, &card->rtd_list, list) { -struct snd_soc_dai *cpu_dai = rtd->cpu_dai; +struct snd_soc_dai *cpu_dai;if (rtd->dai_link->ignore_suspend)continue ; -if (cpu_dai->driver->suspend && cpu_dai->driver->bus_control) - cpu_dai->driver->suspend (cpu_dai); +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai = rtd->cpu_dais[i]; -/* deactivate pins to sleep state */ -pinctrl_pm_select_sleep_state (cpu_dai->dev); +if (cpu_dai->driver->suspend && + cpu_dai->driver->bus_control) + cpu_dai->driver->suspend (cpu_dai);
Quoted 8 lines
sound/soc/soc-core.c int snd_soc_suspend(struct device *dev) · line 582
Quoted 4 lines
sound/soc/soc-core.c +4
-0+ +/* deactivate pins to sleep state */ +pinctrl_pm_select_sleep_state (cpu_dai->dev); + }I am not sure I get the relationship between
cpu_daiand pins. Is this always safe/ok to play with the pincrtl before allcpu_daisare suspended? Or should you implement this with two loops, one to suspend and one to deactivate pins?
Hmmm, you have a valid point. Two loops is a right approach IMO..
Quoted 27 lines
Quoted 27 lines
sound/soc/soc-core.c +9
-2@@ -845,35 +882,47 @@ static int soc_bind_dai_link(struct snd_soc_card *card, {struct snd_soc_pcm_runtime *rtd;struct snd_soc_dai_link_component *codecs = dai_link->codecs; -struct snd_soc_dai_link_component cpu_dai_component; +struct snd_soc_dai_link_component *cpu_dai_component;struct snd_soc_component *component; -struct snd_soc_dai **codec_dais; +struct snd_soc_dai **codec_dais, **cpu_dais;struct device_node *platform_of_node;const char *platform_name;int i;dev_dbg (card->dev,"ASoC: binding %s\n" , dai_link->name); + cpu_dai_component = dai_link->cpu_dai; +if (soc_is_dai_link_bound (card, dai_link)) {dev_dbg (card->dev,"ASoC: dai link %s already bound\n" , dai_link->name);return 0 ; } +if (dai_link->dynamic && dai_link->num_cpu_dai >1 ) { +dev_err (card->dev,"ASoC: Multi CPU DAI not supported for FE" ); +return -EINVAL ; + } + rtd =soc_new_pcm_runtime (card, dai_link);if (!rtd)
Quoted 8 lines
sound/soc/soc-core.c static int soc_bind_dai_link(struct snd_soc_card *card, · line 869
Quoted 4 lines
sound/soc/soc-core.c +0
-3return -ENOMEM ; - cpu_dai_component.name = dai_link->cpu_name; - cpu_dai_component.of_node = dai_link->cpu_of_node; - cpu_dai_component.dai_name = dai_link->cpu_dai_name;Have you checked if there are any side effects of using
cpu_dai_component=dai_link->cpu_dai;
instead of a local variable?
Yes, I have checked this. Also, snd_soc_init_single_cpu_dai() ensures that when there is a single DAI these are populated and in case of multi cpu DAIs the DAI Link specifies these ..
Quoted 4 lines
if you are not sure, it may be worth implementing as a separate patch first before introducing the multi-cpu part, at least to allow for git bisect to identify issues?
I have found it very hard to split these patches as the changes are monolithic. Infact, my worry is the patch split itself might break git bisect :(
Quoted 9 lines
Quoted 9 lines
sound/soc/soc-core.c +9
-0+static int snd_soc_init_single_cpu_dai (struct snd_soc_card *card, +struct snd_soc_dai_link *dai_link) +{ +if (dai_link->cpu_name || dai_link->cpu_of_node || + dai_link->cpu_dai_name) { + dai_link->num_cpu_dai =1 ; + dai_link->cpu_dai =devm_kzalloc (card->dev, +sizeof (struct snd_soc_dai_link_component ), +GFP_KERNEL );
Quoted 9 lines
sound/soc/soc-core.c static int snd_soc_init_multicodec(struct snd_soc_card *card, · line 1111
Quoted 7 lines
sound/soc/soc-core.c +7
-0+ +if (!dai_link->cpu_dai) +return -ENOMEM ; + + dai_link->cpu_dai[0 ].name = dai_link->cpu_name; + dai_link->cpu_dai[0 ].of_node = dai_link->cpu_of_node; + dai_link->cpu_dai[0 ].dai_name = dai_link->cpu_dai_name;Question: is cpu_dai[i].of_node defined for i>0 in the multi
cpu_daicase?
Yes, it should be defined.
Quoted 5 lines
Quoted 5 lines
sound/soc/soc-core.c +0
-1@@ -1644,7 +1751,7 @@ int snd_soc_runtime_set_dai_fmt(struct snd_soc_pcm_runtime *rtd, unsigned int dai_fmt) {struct snd_soc_dai **codec_dais = rtd->codec_dais; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai;
Quoted 9 lines
sound/soc/soc-core.c int snd_soc_runtime_set_dai_fmt(struct snd_soc_pcm_runtime *rtd, · line 1661
Quoted 7 lines
sound/soc/soc-core.c +1
-1+struct snd_soc_dai **cpu_dais = rtd->cpu_dais;unsigned int i;int ret;@@ -1659,35 +1766,44 @@ int snd_soc_runtime_set_dai_fmt(struct snd_soc_pcm_runtime *rtd, } } -/* Flip the polarity for the "CPU" end of a CODEC<->CODEC link */ why was this comment removed?
Looks like I messed up resolving ocnflicts while rebase
Quoted 4 lines
sound/soc/soc-core.c int snd_soc_runtime_set_dai_fmt(struct snd_soc_pcm_runtime *rtd, · line 1664
Quoted 2 lines
sound/soc/soc-core.c +0
-1/* the component which has non_legacy_dai_naming is Codec */ -if (cpu_dai->component->driver->non_legacy_dai_naming) {Not sure if the code refactoring below makes sense in a codec-codec link, you probably wouldn't have multiple
cpu_daisthen, would you?
Yes, a valid point. You suggest to leave this piece of code as is ?
Quoted 60 lines
Quoted 59 lines
sound/soc/soc-core.c +34
-23-unsigned int inv_dai_fmt; +for (i =0 ; i < rtd->num_cpu_dai; i++) { +struct snd_soc_dai *cpu_dai = cpu_dais[i]; +unsigned int inv_dai_fmt, temp_dai_fmt; - inv_dai_fmt = dai_fmt & ~SND_SOC_DAIFMT_MASTER_MASK ; -switch (dai_fmt &SND_SOC_DAIFMT_MASTER_MASK ) { -case SND_SOC_DAIFMT_CBM_CFM : - inv_dai_fmt |=SND_SOC_DAIFMT_CBS_CFS ; -break ; -case SND_SOC_DAIFMT_CBM_CFS : - inv_dai_fmt |=SND_SOC_DAIFMT_CBS_CFM ; -break ; -case SND_SOC_DAIFMT_CBS_CFM : - inv_dai_fmt |=SND_SOC_DAIFMT_CBM_CFS ; -break ; -case SND_SOC_DAIFMT_CBS_CFS : - inv_dai_fmt |=SND_SOC_DAIFMT_CBM_CFM ; -break ; - } + temp_dai_fmt = dai_fmt; +if (cpu_dai->component->driver->non_legacy_dai_naming) { - dai_fmt = inv_dai_fmt; - } + inv_dai_fmt = dai_fmt & ~SND_SOC_DAIFMT_MASTER_MASK ; - ret =snd_soc_dai_set_fmt (cpu_dai, dai_fmt); -if (ret !=0 && ret != -ENOTSUPP ) { -dev_warn (cpu_dai->dev, -"ASoC: Failed to set DAI format: %d\n" , ret); -return ret; +switch (dai_fmt &SND_SOC_DAIFMT_MASTER_MASK ) { +case SND_SOC_DAIFMT_CBM_CFM : + inv_dai_fmt |=SND_SOC_DAIFMT_CBS_CFS ; +break ; + +case SND_SOC_DAIFMT_CBM_CFS : + inv_dai_fmt |=SND_SOC_DAIFMT_CBS_CFM ; +break ; + +case SND_SOC_DAIFMT_CBS_CFM : + inv_dai_fmt |=SND_SOC_DAIFMT_CBM_CFS ; +break ; + +case SND_SOC_DAIFMT_CBS_CFS : + inv_dai_fmt |=SND_SOC_DAIFMT_CBM_CFM ; +break ; + + } + + temp_dai_fmt = inv_dai_fmt; + } + + ret =snd_soc_dai_set_fmt (cpu_dai, temp_dai_fmt); +if (ret !=0 && ret != -ENOTSUPP ) { +dev_warn (cpu_dai->dev, +"ASoC: Failed to set DAI format: %d\n" , ret); +return ret; + } }
Re: [PATCH v6 1/3] ASoC: Add initial support for multiple CPU DAIs
Quoted 20 lines
Quoted 10 lines
Quoted 10 lines
sound/soc/soc-core.c +9
-0+static int snd_soc_init_single_cpu_dai (struct snd_soc_card *card, +struct snd_soc_dai_link *dai_link) +{ +if (dai_link->cpu_name || dai_link->cpu_of_node || + dai_link->cpu_dai_name) { + dai_link->num_cpu_dai =1 ; + dai_link->cpu_dai =devm_kzalloc (card->dev, +sizeof (struct snd_soc_dai_link_component ), +GFP_KERNEL );Quoted 9 lines
sound/soc/soc-core.c static int snd_soc_init_multicodec(struct snd_soc_card *card, · line 1111
Quoted 7 lines
sound/soc/soc-core.c +7
-0+ +if (!dai_link->cpu_dai) +return -ENOMEM ; + + dai_link->cpu_dai[0 ].name = dai_link->cpu_name; + dai_link->cpu_dai[0 ].of_node = dai_link->cpu_of_node; + dai_link->cpu_dai[0 ].dai_name = dai_link->cpu_dai_name;Question: is cpu_dai[i].of_node defined for i>0 in the multi
cpu_daicase?
Quoted 1 line
Yes, it should be defined.
I have limited understanding of how cpu_of_node would be handled and if there is any guidance for DT folks on how to deal with multiple cpu_dais.
Quoted 16 lines
Quoted 15 lines
sound/soc/soc-core.c int snd_soc_runtime_set_dai_fmt(struct snd_soc_pcm_runtime *rtd, · line 1661
Quoted 13 lines
sound/soc/soc-core.c +1
-2@@ -1644,7 +1751,7 @@ int snd_soc_runtime_set_dai_fmt(struct snd_soc_pcm_runtime *rtd, unsigned int dai_fmt) {struct snd_soc_dai **codec_dais = rtd->codec_dais; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai; +struct snd_soc_dai **cpu_dais = rtd->cpu_dais;unsigned int i;int ret;@@ -1659,35 +1766,44 @@ int snd_soc_runtime_set_dai_fmt(struct snd_soc_pcm_runtime *rtd, } } -/* Flip the polarity for the "CPU" end of a CODEC<->CODEC link */ why was this comment removed?
Quoted 8 lines
Looks like I messed up resolving ocnflicts while rebase
Quoted 4 lines
sound/soc/soc-core.c int snd_soc_runtime_set_dai_fmt(struct snd_soc_pcm_runtime *rtd, · line 1664
Quoted 2 lines
sound/soc/soc-core.c +0
-1/* the component which has non_legacy_dai_naming is Codec */ -if (cpu_dai->component->driver->non_legacy_dai_naming) {Not sure if the code refactoring below makes sense in a codec-codec link, you probably wouldn't have multiple
cpu_daisthen, would you?Yes, a valid point. You suggest to leave this piece of code as is ?
Not necessarily. I don't understand how the codec-codec and multi cpu_dais intersect, all I am asking for is a check if this change is needed or not.
Quoted 62 lines
Quoted 61 lines
Quoted 60 lines
sound/soc/soc-core.c +34
-23-unsigned int inv_dai_fmt; +for (i =0 ; i < rtd->num_cpu_dai; i++) { +struct snd_soc_dai *cpu_dai = cpu_dais[i]; +unsigned int inv_dai_fmt, temp_dai_fmt; - inv_dai_fmt = dai_fmt & ~SND_SOC_DAIFMT_MASTER_MASK ; -switch (dai_fmt &SND_SOC_DAIFMT_MASTER_MASK ) { -case SND_SOC_DAIFMT_CBM_CFM : - inv_dai_fmt |=SND_SOC_DAIFMT_CBS_CFS ; -break ; -case SND_SOC_DAIFMT_CBM_CFS : - inv_dai_fmt |=SND_SOC_DAIFMT_CBS_CFM ; -break ; -case SND_SOC_DAIFMT_CBS_CFM : - inv_dai_fmt |=SND_SOC_DAIFMT_CBM_CFS ; -break ; -case SND_SOC_DAIFMT_CBS_CFS : - inv_dai_fmt |=SND_SOC_DAIFMT_CBM_CFM ; -break ; - } + temp_dai_fmt = dai_fmt; +if (cpu_dai->component->driver->non_legacy_dai_naming) { - dai_fmt = inv_dai_fmt; - } + inv_dai_fmt = dai_fmt & ~SND_SOC_DAIFMT_MASTER_MASK ; - ret =snd_soc_dai_set_fmt (cpu_dai, dai_fmt); -if (ret !=0 && ret != -ENOTSUPP ) { -dev_warn (cpu_dai->dev, -"ASoC: Failed to set DAI format: %d\n" , ret); -return ret; +switch (dai_fmt &SND_SOC_DAIFMT_MASTER_MASK ) { +case SND_SOC_DAIFMT_CBM_CFM : + inv_dai_fmt |=SND_SOC_DAIFMT_CBS_CFS ; +break ; + +case SND_SOC_DAIFMT_CBM_CFS : + inv_dai_fmt |=SND_SOC_DAIFMT_CBS_CFM ; +break ; + +case SND_SOC_DAIFMT_CBS_CFM : + inv_dai_fmt |=SND_SOC_DAIFMT_CBM_CFS ; +break ; + +case SND_SOC_DAIFMT_CBS_CFS : + inv_dai_fmt |=SND_SOC_DAIFMT_CBM_CFM ; +break ; + + } + + temp_dai_fmt = inv_dai_fmt; + } + + ret =snd_soc_dai_set_fmt (cpu_dai, temp_dai_fmt); +if (ret !=0 && ret != -ENOTSUPP ) { +dev_warn (cpu_dai->dev, +"ASoC: Failed to set DAI format: %d\n" , ret); +return ret; + } }
Re: [PATCH v6 1/3] ASoC: Add initial support for multiple CPU DAIs
Quoted 20 lines · Pierre-Louis Bossart
Quoted 20 lines
Quoted 19 lines
Quoted 7 lines
sound/soc/soc-core.c +6
-0+static int snd_soc_init_single_cpu_dai (struct snd_soc_card *card, +struct snd_soc_dai_link *dai_link) +{ +if (dai_link->cpu_name || dai_link->cpu_of_node || + dai_link->cpu_dai_name) { + dai_link->num_cpu_dai =1 ;sound/soc/soc-core.c static int snd_soc_init_multicodec(struct snd_soc_card *card, · line 1111
Quoted 10 lines
sound/soc/soc-core.c +10
-0+ dai_link->cpu_dai =devm_kzalloc (card->dev, +sizeof (struct snd_soc_dai_link_component ), +GFP_KERNEL ); + +if (!dai_link->cpu_dai) +return -ENOMEM ; + + dai_link->cpu_dai[0 ].name = dai_link->cpu_name; + dai_link->cpu_dai[0 ].of_node = dai_link->cpu_of_node; + dai_link->cpu_dai[0 ].dai_name = dai_link->cpu_dai_name;Question: is cpu_dai[i].of_node defined for i>0 in the multi
cpu_daicase?
Quoted 7 lines
Quoted 1 line
Yes, it should be defined.
I have limited understanding of how
cpu_of_nodewould be handled and if there is any guidance for DT folks on how to deal with multiplecpu_dais.From what I could gather, it looks like we need not add additional code.
But, my understanding of DT is pretty limited as well. Vinod, Liam, can you help us here?
Quoted 23 lines
Quoted 23 lines
Quoted 15 lines
sound/soc/soc-core.c int snd_soc_runtime_set_dai_fmt(struct snd_soc_pcm_runtime *rtd, · line 1661
Quoted 13 lines
sound/soc/soc-core.c +1
-2@@ -1644,7 +1751,7 @@ int snd_soc_runtime_set_dai_fmt(struct snd_soc_pcm_runtime *rtd, unsigned int dai_fmt) {struct snd_soc_dai **codec_dais = rtd->codec_dais; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai; +struct snd_soc_dai **cpu_dais = rtd->cpu_dais;unsigned int i;int ret;@@ -1659,35 +1766,44 @@ int snd_soc_runtime_set_dai_fmt(struct snd_soc_pcm_runtime *rtd, } } -/* Flip the polarity for the "CPU" end of a CODEC<->CODEC link */ why was this comment removed?
Looks like I messed up resolving ocnflicts while rebase
Quoted 4 lines
sound/soc/soc-core.c int snd_soc_runtime_set_dai_fmt(struct snd_soc_pcm_runtime *rtd, · line 1664
Quoted 2 lines
sound/soc/soc-core.c +0
-1/* the component which has non_legacy_dai_naming is Codec */ -if (cpu_dai->component->driver->non_legacy_dai_naming) {Not sure if the code refactoring below makes sense in a codec-codec link, you probably wouldn't have multiple
cpu_daisthen, would you?
Quoted 5 lines
Quoted 1 line
Yes, a valid point. You suggest to leave this piece of code as is ?
Not necessarily. I don't understand how the codec-codec and multi
cpu_daisintersect, all I am asking for is a check if this change is needed or not.
In soc_link_dai_widgets() which is called from probe_dai_links(), we call out that we do not support multi cpu/codec for CODEC - CODEC link. All the operation there are done only on single CPU DAI/Codec DAI.
But, in this function snd_soc_dai_set_fmt() is set on all Codec DAIs. So, for sake of consistency we can do it for CPU DAIs too.
--Shreyas
Re: [PATCH v6 1/3] ASoC: Add initial support for multiple CPU DAIs
In reply to Pierre-Louis BossartQuoted 11 lines · Pierre-Louis Bossart
Quoted 6 lines
Quoted 4 lines
sound/soc/soc-core.c int snd_soc_runtime_set_dai_fmt(struct snd_soc_pcm_runtime *rtd, · line 1664
Quoted 2 lines
sound/soc/soc-core.c +0
-1/* the component which has non_legacy_dai_naming is Codec */ -if (cpu_dai->component->driver->non_legacy_dai_naming) {Not sure if the code refactoring below makes sense in a codec-codec link, you probably wouldn't have multiple
cpu_daisthen, would you?Yes, a valid point. You suggest to leave this piece of code as is ?
Not necessarily. I don't understand how the codec-codec and multi
cpu_daisintersect, all I am asking for is a check if this change is needed or not.
Seems plausible to me you could have multiple cpu_dais on CODEC to CODEC links. Really for CODEC to CODEC links the whole CPU/CODEC distinction doesn't really exist. It doesn't really matter which end you make which, so I guess it makes sense to support multiple CPU DAIs on such a link.
Thanks,
Charles
PATCH v6 2/3 ASoC: Add multiple CPU DAI support for PCM ops
In reply to Shreyas NCAdd support in PCM operations to invoke multiple cpu dais as we do for multiple codec dais. Also the symmetry calculations are updated to reflect multiple cpu dais.
--- sound/soc/soc-pcm.c | 491 +++++++++++++++++++++++++++++++++++----------------- 1 file changed, 332 insertions(+), 159 deletions(-)
sound/soc/soc-pcm.c +332 -159
@@ -64,23 +64,27 @@ static bool snd_soc_dai_stream_valid(struct snd_soc_dai *dai, int stream)*/ void snd_soc_runtime_activate (struct snd_soc_pcm_runtime *rtd,int stream) { -struct snd_soc_dai *cpu_dai = rtd->cpu_dai;int i;lockdep_assert_held (&rtd->pcm_mutex);if (stream ==SNDRV_PCM_STREAM_PLAYBACK ) { - cpu_dai->playback_active++; +for (i =0 ; i < rtd->num_cpu_dai; i++) + rtd->cpu_dais[i]->playback_active++;for (i =0 ; i < rtd->num_codecs; i++) rtd->codec_dais[i]->playback_active++; }else { - cpu_dai->capture_active++; +for (i =0 ; i < rtd->num_cpu_dai; i++) + rtd->cpu_dais[i]->capture_active++;for (i =0 ; i < rtd->num_codecs; i++) rtd->codec_dais[i]->capture_active++; } - cpu_dai->active++; - cpu_dai->component->active++; +for (i =0 ; i < rtd->num_cpu_dai; i++) { + rtd->cpu_dais[i]->component->active++; + rtd->cpu_dais[i]->active++; + } +for (i =0 ; i < rtd->num_codecs; i++) { rtd->codec_dais[i]->active++; rtd->codec_dais[i]->component->active++;
@@ -99,23 +103,27 @@ void snd_soc_runtime_activate(struct snd_soc_pcm_runtime *rtd, int stream)*/ void snd_soc_runtime_deactivate (struct snd_soc_pcm_runtime *rtd,int stream) { -struct snd_soc_dai *cpu_dai = rtd->cpu_dai;int i;lockdep_assert_held (&rtd->pcm_mutex);if (stream ==SNDRV_PCM_STREAM_PLAYBACK ) { - cpu_dai->playback_active--; +for (i =0 ; i < rtd->num_cpu_dai; i++) + rtd->cpu_dais[i]->playback_active--;for (i =0 ; i < rtd->num_codecs; i++) rtd->codec_dais[i]->playback_active--; }else { - cpu_dai->capture_active--; +for (i =0 ; i < rtd->num_cpu_dai; i++) + rtd->cpu_dais[i]->capture_active--;for (i =0 ; i < rtd->num_codecs; i++) rtd->codec_dais[i]->capture_active--; } - cpu_dai->active--; - cpu_dai->component->active--; +for (i =0 ; i < rtd->num_cpu_dai; i++) { + rtd->cpu_dais[i]->component->active--; + rtd->cpu_dais[i]->active--; + } +for (i =0 ; i < rtd->num_codecs; i++) { rtd->codec_dais[i]->component->active--; rtd->codec_dais[i]->active--;
@@ -258,7 +266,6 @@ static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, struct snd_pcm_hw_params *params) {struct snd_soc_pcm_runtime *rtd = substream->private_data; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai;unsigned int rate, channels, sample_bits, symmetry, i; rate =params_rate (params);
@@ -266,41 +273,54 @@ static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, sample_bits =snd_pcm_format_physical_width (params_format (params));/* reject unmatched parameters when applying symmetry */ - symmetry = cpu_dai->driver->symmetric_rates || - rtd->dai_link->symmetric_rates; + symmetry = rtd->dai_link->symmetric_rates; + +for (i =0 ; i < rtd->num_cpu_dai; i++) + symmetry |= rtd->cpu_dais[i]->driver->symmetric_rates;for (i =0 ; i < rtd->num_codecs; i++) symmetry |= rtd->codec_dais[i]->driver->symmetric_rates; -if (symmetry && cpu_dai->rate && cpu_dai->rate != rate) { -dev_err (rtd->dev,"ASoC: unmatched rate symmetry: %d - %d\n" , - cpu_dai->rate, rate); -return -EINVAL ; - } +for (i =0 ; i < rtd->num_cpu_dai; i++) +if (symmetry && rtd->cpu_dais[i]->rate && + rtd->cpu_dais[i]->rate != rate) { +dev_err (rtd->dev,"ASoC: unmatched rate symmetry: %d - %d\n" , + rtd->cpu_dais[i]->rate, rate); +return -EINVAL ; + } - symmetry = cpu_dai->driver->symmetric_channels || - rtd->dai_link->symmetric_channels; + symmetry = rtd->dai_link->symmetric_channels; + +for (i =0 ; i < rtd->num_cpu_dai; i++) + symmetry |= rtd->cpu_dais[i]->driver->symmetric_channels;for (i =0 ; i < rtd->num_codecs; i++) symmetry |= rtd->codec_dais[i]->driver->symmetric_channels; -if (symmetry && cpu_dai->channels && cpu_dai->channels != channels) { -dev_err (rtd->dev,"ASoC: unmatched channel symmetry: %d - %d\n" , - cpu_dai->channels, channels); -return -EINVAL ; - } +for (i =0 ; i < rtd->num_cpu_dai; i++) +if (symmetry && rtd->cpu_dais[i]->channels && + rtd->cpu_dais[i]->channels != channels) { +dev_err (rtd->dev,"ASoC: unmatched channel symmetry: %d - %d\n" , + rtd->cpu_dais[i]->channels, channels); +return -EINVAL ; + } - symmetry = cpu_dai->driver->symmetric_samplebits || - rtd->dai_link->symmetric_samplebits; + symmetry = rtd->dai_link->symmetric_samplebits; + +for (i =0 ; i < rtd->num_cpu_dai; i++) + symmetry |= rtd->cpu_dais[i]->driver->symmetric_samplebits;for (i =0 ; i < rtd->num_codecs; i++) symmetry |= rtd->codec_dais[i]->driver->symmetric_samplebits; -if (symmetry && cpu_dai->sample_bits && cpu_dai->sample_bits != sample_bits) { -dev_err (rtd->dev,"ASoC: unmatched sample bits symmetry: %d - %d\n" , - cpu_dai->sample_bits, sample_bits); -return -EINVAL ; - } +for (i =0 ; i < rtd->num_cpu_dai; i++) +if (symmetry && rtd->cpu_dais[i]->sample_bits && + rtd->cpu_dais[i]->sample_bits != sample_bits) { +dev_err (rtd->dev,"ASoC: unmatched sample bits symmetry: %d - %d\n" , + rtd->cpu_dais[i]->sample_bits, + sample_bits); +return -EINVAL ; + }return 0 ; }
@@ -308,13 +328,18 @@ static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream,static bool soc_pcm_has_symmetry (struct snd_pcm_substream *substream) {struct snd_soc_pcm_runtime *rtd = substream->private_data; -struct snd_soc_dai_driver *cpu_driver = rtd->cpu_dai->driver;struct snd_soc_dai_link *link = rtd->dai_link;unsigned int symmetry, i; - symmetry = cpu_driver->symmetric_rates || link->symmetric_rates || - cpu_driver->symmetric_channels || link->symmetric_channels || - cpu_driver->symmetric_samplebits || link->symmetric_samplebits; + symmetry = link->symmetric_rates || link->symmetric_channels || + link->symmetric_samplebits; + +/* Apply symmetery for multiple cpu dais */ +for (i =0 ; i < rtd->num_cpu_dai; i++) + symmetry = symmetry || + rtd->cpu_dais[i]->driver->symmetric_rates || + rtd->cpu_dais[i]->driver->symmetric_channels || + rtd->cpu_dais[i]->driver->symmetric_samplebits;for (i =0 ; i < rtd->num_codecs; i++) symmetry = symmetry ||
@@ -342,10 +367,10 @@ static void soc_pcm_set_msb(struct snd_pcm_substream *substream, int bits)static void soc_pcm_apply_msb (struct snd_pcm_substream *substream) {struct snd_soc_pcm_runtime *rtd = substream->private_data; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai; +struct snd_soc_dai *cpu_dai;struct snd_soc_dai *codec_dai;int i; -unsigned int bits =0 , cpu_bits; +unsigned int bits =0 , cpu_bits =0 ;if (substream->stream ==SNDRV_PCM_STREAM_PLAYBACK ) {for (i =0 ; i < rtd->num_codecs; i++) {
@@ -356,7 +381,16 @@ static void soc_pcm_apply_msb(struct snd_pcm_substream *substream) } bits =max (codec_dai->driver->playback.sig_bits, bits); } - cpu_bits = cpu_dai->driver->playback.sig_bits; + +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai = rtd->cpu_dais[i]; +if (cpu_dai->driver->playback.sig_bits ==0 ) { + cpu_bits =0 ; +break ; + } + + cpu_bits =max (cpu_dai->driver->playback.sig_bits, cpu_bits); + } }else {for (i =0 ; i < rtd->num_codecs; i++) { codec_dai = rtd->codec_dais[i];
@@ -366,7 +400,15 @@ static void soc_pcm_apply_msb(struct snd_pcm_substream *substream) } bits =max (codec_dai->driver->capture.sig_bits, bits); } - cpu_bits = cpu_dai->driver->capture.sig_bits; + +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai = rtd->cpu_dais[i]; +if (cpu_dai->driver->capture.sig_bits ==0 ) { + cpu_bits =0 ; +break ; + } + cpu_bits =max (cpu_dai->driver->capture.sig_bits, bits); + } }soc_pcm_set_msb (substream, bits);
@@ -378,21 +420,18 @@ static void soc_pcm_init_runtime_hw(struct snd_pcm_substream *substream)struct snd_pcm_runtime *runtime = substream->runtime;struct snd_pcm_hardware *hw = &runtime->hw;struct snd_soc_pcm_runtime *rtd = substream->private_data; -struct snd_soc_dai_driver *cpu_dai_drv = rtd->cpu_dai->driver; +struct snd_soc_dai_driver *cpu_dai_drv;struct snd_soc_dai_driver *codec_dai_drv;struct snd_soc_pcm_stream *codec_stream;struct snd_soc_pcm_stream *cpu_stream;unsigned int chan_min =0 , chan_max =UINT_MAX ; +unsigned int cpu_chan_min =0 , cpu_chan_max =UINT_MAX ;unsigned int rate_min =0 , rate_max =UINT_MAX ; -unsigned int rates =UINT_MAX ; +unsigned int cpu_rate_min =0 , cpu_rate_max =UINT_MAX ; +unsigned int rates =UINT_MAX , cpu_rates =UINT_MAX ;u64 formats =ULLONG_MAX ;int i; -if (substream->stream ==SNDRV_PCM_STREAM_PLAYBACK ) - cpu_stream = &cpu_dai_drv->playback; -else - cpu_stream = &cpu_dai_drv->capture; -/* first calculate min/max only for CODECs in the DAI link */ for (i =0 ; i < rtd->num_codecs; i++) {
@@ -422,30 +461,55 @@ static void soc_pcm_init_runtime_hw(struct snd_pcm_substream *substream) rates =snd_pcm_rate_mask_intersect (codec_stream->rates, rates); } +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai_drv = rtd->cpu_dais[i]->driver; + +if (substream->stream ==SNDRV_PCM_STREAM_PLAYBACK ) + cpu_stream = &cpu_dai_drv->playback; +else + cpu_stream = &cpu_dai_drv->capture; + + cpu_chan_min =max (cpu_chan_min, + cpu_stream->channels_min); + cpu_chan_max =min (cpu_chan_max, + cpu_stream->channels_max); + +if (hw->formats) + hw->formats &= cpu_stream->formats; +else + hw->formats = cpu_stream->formats; + + cpu_rates =snd_pcm_rate_mask_intersect (cpu_rates, + cpu_stream->rates); + + cpu_rate_min =max (cpu_rate_min, cpu_stream->rate_min); + cpu_rate_max =min_not_zero (cpu_rate_max, cpu_stream->rate_max); + } +/* -* chan min/max cannot be enforced if there are multiple CODEC DAIs -* connected to a single CPU DAI, use CPU DAI's directly and let -* channel allocation be fixed up later +* chan min/max cannot be enforced if there are multiple +* CODEC DAIs connected to CPU DAI(s), use CPU DAI's +* directly and let channel allocation be fixed up later */ if (rtd->num_codecs >1 ) { - chan_min = cpu_stream->channels_min; - chan_max = cpu_stream->channels_max; + chan_min = cpu_chan_min; + chan_max = cpu_chan_max; } - hw->channels_min =max (chan_min, cpu_stream->channels_min); - hw->channels_max =min (chan_max, cpu_stream->channels_max); -if (hw->formats) - hw->formats &= formats & cpu_stream->formats; -else - hw->formats = formats & cpu_stream->formats; - hw->rates =snd_pcm_rate_mask_intersect (rates, cpu_stream->rates); + hw->channels_min =max (cpu_chan_min, chan_min); + + hw->channels_max =min (cpu_chan_max, chan_max); + + hw->rate_min =max (cpu_rate_min, rate_min); + hw->rate_max =min_not_zero (cpu_rate_max, rate_max); + hw->rates =snd_pcm_rate_mask_intersect (rates, cpu_rates);snd_pcm_limit_hw_rates (runtime); - hw->rate_min =max (hw->rate_min, cpu_stream->rate_min); - hw->rate_min =max (hw->rate_min, rate_min); - hw->rate_max =min_not_zero (hw->rate_max, cpu_stream->rate_max); - hw->rate_max =min_not_zero (hw->rate_max, rate_max); +if (hw->formats) + hw->formats &= formats; +else + hw->formats = formats; }/*
@@ -459,12 +523,15 @@ static int soc_pcm_open(struct snd_pcm_substream *substream)struct snd_pcm_runtime *runtime = substream->runtime;struct snd_soc_component *component;struct snd_soc_rtdcom_list *rtdcom; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai; +struct snd_soc_dai *cpu_dai;struct snd_soc_dai *codec_dai;const char *codec_dai_name ="multicodec" ; -int i, ret =0 , __ret; +const char *cpu_dai_name ="multicpu" ; +int i, ret =0 , __ret, j; + +for (i =0 ; i < rtd->num_cpu_dai; i++) +pinctrl_pm_select_default_state (rtd->cpu_dais[i]->dev); -pinctrl_pm_select_default_state (cpu_dai->dev);for (i =0 ; i < rtd->num_codecs; i++)pinctrl_pm_select_default_state (rtd->codec_dais[i]->dev);
@@ -477,12 +544,15 @@ static int soc_pcm_open(struct snd_pcm_substream *substream)mutex_lock_nested (&rtd->pcm_mutex, rtd->pcm_subclass);/* startup the audio subsystem */ -if (cpu_dai->driver->ops->startup) { - ret = cpu_dai->driver->ops->startup (substream, cpu_dai); -if (ret <0 ) { -dev_err (cpu_dai->dev,"ASoC: can't open interface" -" %s: %d\n" , cpu_dai->name, ret); - goto out; +for (j =0 ; j < rtd->num_cpu_dai; j++) { + cpu_dai = rtd->cpu_dais[j]; +if (cpu_dai->driver->ops->startup) { + ret = cpu_dai->driver->ops->startup (substream, cpu_dai); +if (ret <0 ) { +dev_err (cpu_dai->dev,"ASoC: can't open interface %s: %d\n" , + cpu_dai->name, ret); + goto interface_err; + } } }
@@ -543,34 +613,40 @@ static int soc_pcm_open(struct snd_pcm_substream *substream)if (rtd->num_codecs ==1 ) codec_dai_name = rtd->codec_dai->name; +if (rtd->num_cpu_dai ==1 ) + cpu_dai_name = rtd->cpu_dai->name; +if (soc_pcm_has_symmetry (substream)) runtime->hw.info |=SNDRV_PCM_INFO_JOINT_DUPLEX ; ret = -EINVAL ;if (!runtime->hw.rates) {printk (KERN_ERR "ASoC: %s <-> %s No matching rates\n" , - codec_dai_name, cpu_dai->name); + codec_dai_name, cpu_dai_name); goto config_err; }if (!runtime->hw.formats) {printk (KERN_ERR "ASoC: %s <-> %s No matching formats\n" , - codec_dai_name, cpu_dai->name); + codec_dai_name, cpu_dai_name); goto config_err; }if (!runtime->hw.channels_min || !runtime->hw.channels_max || runtime->hw.channels_min > runtime->hw.channels_max) {printk (KERN_ERR "ASoC: %s <-> %s No matching channels\n" , - codec_dai_name, cpu_dai->name); + codec_dai_name, cpu_dai_name); goto config_err; }soc_pcm_apply_msb (substream);/* Symmetry only applies if we've already got an active stream. */ -if (cpu_dai->active) { - ret =soc_pcm_apply_symmetry (substream, cpu_dai); -if (ret !=0 ) - goto config_err; +for (i =0 ; i < rtd->num_cpu_dai; i++) { +if (rtd->cpu_dais[i]->active) { + ret =soc_pcm_apply_symmetry (substream, + rtd->cpu_dais[i]); +if (ret !=0 ) + goto config_err; + } }for (i =0 ; i < rtd->num_codecs; i++) {
@@ -583,7 +659,7 @@ static int soc_pcm_open(struct snd_pcm_substream *substream) }pr_debug ("ASoC: %s <-> %s info:\n" , - codec_dai_name, cpu_dai->name); + codec_dai_name, cpu_dai_name);pr_debug ("ASoC: rate mask 0x%x\n" , runtime->hw.rates);pr_debug ("ASoC: min ch %d max ch %d\n" , runtime->hw.channels_min, runtime->hw.channels_max);
@@ -622,9 +698,15 @@ static int soc_pcm_open(struct snd_pcm_substream *substream) component->driver->ops->close (substream); } -if (cpu_dai->driver->ops->shutdown) - cpu_dai->driver->ops->shutdown (substream, cpu_dai); -out: + j = rtd->num_cpu_dai; + +interface_err: +while (--j >=0 ) { + cpu_dai = rtd->cpu_dais[j]; +if (cpu_dai->driver->ops->shutdown) + cpu_dai->driver->ops->shutdown (substream, cpu_dai); + } +mutex_unlock (&rtd->pcm_mutex);for_each_rtdcom (rtd, rtdcom) {
@@ -638,8 +720,11 @@ static int soc_pcm_open(struct snd_pcm_substream *substream)if (!rtd->codec_dais[i]->active)pinctrl_pm_select_sleep_state (rtd->codec_dais[i]->dev); } -if (!cpu_dai->active) -pinctrl_pm_select_sleep_state (cpu_dai->dev); + +for (i =0 ; i < rtd->num_cpu_dai; i++) { +if (!rtd->cpu_dais[i]->active) +pinctrl_pm_select_sleep_state (rtd->cpu_dais[i]->dev); + }return ret; }
@@ -682,7 +767,7 @@ static int soc_pcm_close(struct snd_pcm_substream *substream)struct snd_soc_pcm_runtime *rtd = substream->private_data;struct snd_soc_component *component;struct snd_soc_rtdcom_list *rtdcom; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai; +struct snd_soc_dai *cpu_dai;struct snd_soc_dai *codec_dai;int i;
@@ -691,8 +776,11 @@ static int soc_pcm_close(struct snd_pcm_substream *substream)snd_soc_runtime_deactivate (rtd, substream->stream);/* clear the corresponding DAIs rate when inactive */ -if (!cpu_dai->active) - cpu_dai->rate =0 ; +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai = rtd->cpu_dais[i]; +if (!cpu_dai->active) + cpu_dai->rate =0 ; + }for (i =0 ; i < rtd->num_codecs; i++) { codec_dai = rtd->codec_dais[i];
@@ -700,10 +788,16 @@ static int soc_pcm_close(struct snd_pcm_substream *substream) codec_dai->rate =0 ; } -snd_soc_dai_digital_mute (cpu_dai,1 , substream->stream); +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai = rtd->cpu_dais[i]; +snd_soc_dai_digital_mute (cpu_dai,1 , substream->stream); + } -if (cpu_dai->driver->ops->shutdown) - cpu_dai->driver->ops->shutdown (substream, cpu_dai); +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai = rtd->cpu_dais[i]; +if (cpu_dai->driver->ops->shutdown) + cpu_dai->driver->ops->shutdown (substream, cpu_dai); + }for (i =0 ; i < rtd->num_codecs; i++) { codec_dai = rtd->codec_dais[i];
@@ -756,8 +850,11 @@ static int soc_pcm_close(struct snd_pcm_substream *substream)if (!rtd->codec_dais[i]->active)pinctrl_pm_select_sleep_state (rtd->codec_dais[i]->dev); } -if (!cpu_dai->active) -pinctrl_pm_select_sleep_state (cpu_dai->dev); + +for (i =0 ; i < rtd->num_cpu_dai; i++) { +if (!rtd->cpu_dais[i]->active) +pinctrl_pm_select_sleep_state (rtd->cpu_dais[i]->dev); + }return 0 ; }
@@ -772,7 +869,7 @@ static int soc_pcm_prepare(struct snd_pcm_substream *substream)struct snd_soc_pcm_runtime *rtd = substream->private_data;struct snd_soc_component *component;struct snd_soc_rtdcom_list *rtdcom; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai; +struct snd_soc_dai *cpu_dai;struct snd_soc_dai *codec_dai;int i, ret =0 ;
@@ -816,12 +913,16 @@ static int soc_pcm_prepare(struct snd_pcm_substream *substream) } } -if (cpu_dai->driver->ops->prepare) { - ret = cpu_dai->driver->ops->prepare (substream, cpu_dai); -if (ret <0 ) { -dev_err (cpu_dai->dev, -"ASoC: cpu DAI prepare error: %d\n" , ret); - goto out; +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai = rtd->cpu_dais[i]; +if (cpu_dai->driver->ops->prepare) { + ret = cpu_dai->driver->ops->prepare (substream, cpu_dai); +if (ret <0 ) { +dev_err (cpu_dai->dev, +"ASoC: cpu DAI prepare error: %d\n" , + ret); + goto out; + } } }
@@ -838,7 +939,10 @@ static int soc_pcm_prepare(struct snd_pcm_substream *substream)for (i =0 ; i < rtd->num_codecs; i++)snd_soc_dai_digital_mute (rtd->codec_dais[i],0 , substream->stream); -snd_soc_dai_digital_mute (cpu_dai,0 , substream->stream); + +for (i =0 ; i < rtd->num_cpu_dai; i++) +snd_soc_dai_digital_mute (rtd->cpu_dais[i],0 , + substream->stream); out:mutex_unlock (&rtd->pcm_mutex);
@@ -885,8 +989,8 @@ static int soc_pcm_hw_params(struct snd_pcm_substream *substream,struct snd_soc_pcm_runtime *rtd = substream->private_data;struct snd_soc_component *component;struct snd_soc_rtdcom_list *rtdcom; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai; -int i, ret =0 , __ret; +struct snd_soc_dai *cpu_dai; +int i, ret =0 , __ret, j;mutex_lock_nested (&rtd->pcm_mutex, rtd->pcm_subclass);if (rtd->dai_link->ops->hw_params) {
@@ -940,9 +1044,12 @@ static int soc_pcm_hw_params(struct snd_pcm_substream *substream, params_format(&codec_params)); } - ret =soc_dai_hw_params (substream, params, cpu_dai); -if (ret <0 ) - goto interface_err; +for (j =0 ; j < rtd->num_cpu_dai; j++) { + cpu_dai = rtd->cpu_dais[j]; + ret =soc_dai_hw_params (substream, params, cpu_dai); +if (ret <0 ) + goto interface_err; + } ret =0 ;for_each_rtdcom (rtd, rtdcom) {
@@ -963,11 +1070,14 @@ static int soc_pcm_hw_params(struct snd_pcm_substream *substream,if (ret <0 ) goto component_err; -/* store the parameters for each DAIs */ - cpu_dai->rate =params_rate (params); - cpu_dai->channels =params_channels (params); - cpu_dai->sample_bits = -snd_pcm_format_physical_width (params_format (params)); +for (i =0 ; i < rtd->num_cpu_dai; i++) { +/* store the parameters for each DAIs */ + cpu_dai = rtd->cpu_dais[i]; + cpu_dai->rate =params_rate (params); + cpu_dai->channels =params_channels (params); + cpu_dai->sample_bits = +snd_pcm_format_physical_width (params_format (params)); + } ret =soc_pcm_params_symmetry (substream, params);if (ret)
@@ -987,10 +1097,16 @@ static int soc_pcm_hw_params(struct snd_pcm_substream *substream, component->driver->ops->hw_free (substream); } -if (cpu_dai->driver->ops->hw_free) - cpu_dai->driver->ops->hw_free (substream, cpu_dai); + j = rtd->num_cpu_dai; interface_err: +while (--j >=0 ) { + cpu_dai = rtd->cpu_dais[j]; + +if (cpu_dai->driver->ops && cpu_dai->driver->ops->hw_free) + cpu_dai->driver->ops->hw_free (substream, cpu_dai); + } + i = rtd->num_codecs; codec_err:
@@ -1016,7 +1132,7 @@ static int soc_pcm_hw_free(struct snd_pcm_substream *substream)struct snd_soc_pcm_runtime *rtd = substream->private_data;struct snd_soc_component *component;struct snd_soc_rtdcom_list *rtdcom; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai; +struct snd_soc_dai *cpu_dai;struct snd_soc_dai *codec_dai;bool playback = substream->stream ==SNDRV_PCM_STREAM_PLAYBACK ;int i;
@@ -1024,10 +1140,13 @@ static int soc_pcm_hw_free(struct snd_pcm_substream *substream)mutex_lock_nested (&rtd->pcm_mutex, rtd->pcm_subclass);/* clear the corresponding DAIs parameters when going to be inactive */ -if (cpu_dai->active ==1 ) { - cpu_dai->rate =0 ; - cpu_dai->channels =0 ; - cpu_dai->sample_bits =0 ; +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai = rtd->cpu_dais[i]; +if (cpu_dai->active ==1 ) { + cpu_dai->rate =0 ; + cpu_dai->channels =0 ; + cpu_dai->sample_bits =0 ; + } }for (i =0 ; i < rtd->num_codecs; i++) {
@@ -1069,8 +1188,11 @@ static int soc_pcm_hw_free(struct snd_pcm_substream *substream) codec_dai->driver->ops->hw_free (substream, codec_dai); } -if (cpu_dai->driver->ops->hw_free) - cpu_dai->driver->ops->hw_free (substream, cpu_dai); +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai = rtd->cpu_dais[i]; +if (cpu_dai->driver->ops && cpu_dai->driver->ops->hw_free) + cpu_dai->driver->ops->hw_free (substream, cpu_dai); + }mutex_unlock (&rtd->pcm_mutex);return 0 ;
@@ -1081,7 +1203,7 @@ static int soc_pcm_trigger(struct snd_pcm_substream *substream, int cmd)struct snd_soc_pcm_runtime *rtd = substream->private_data;struct snd_soc_component *component;struct snd_soc_rtdcom_list *rtdcom; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai; +struct snd_soc_dai *cpu_dai;struct snd_soc_dai *codec_dai;int i, ret;
@@ -1107,10 +1229,14 @@ static int soc_pcm_trigger(struct snd_pcm_substream *substream, int cmd)return ret; } -if (cpu_dai->driver->ops->trigger) { - ret = cpu_dai->driver->ops->trigger (substream, cmd, cpu_dai); -if (ret <0 ) -return ret; +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai = rtd->cpu_dais[i]; +if (cpu_dai->driver->ops->trigger) { + ret = cpu_dai->driver->ops->trigger (substream, + cmd, cpu_dai); +if (ret <0 ) +return ret; + } }if (rtd->dai_link->ops->trigger) {
@@ -1126,7 +1252,7 @@ static int soc_pcm_bespoke_trigger(struct snd_pcm_substream *substream, int cmd) {struct snd_soc_pcm_runtime *rtd = substream->private_data; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai; +struct snd_soc_dai *cpu_dai;struct snd_soc_dai *codec_dai;int i, ret;
@@ -1140,10 +1266,14 @@ static int soc_pcm_bespoke_trigger(struct snd_pcm_substream *substream, } } -if (cpu_dai->driver->ops->bespoke_trigger) { - ret = cpu_dai->driver->ops->bespoke_trigger (substream, cmd, cpu_dai); -if (ret <0 ) -return ret; +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai = rtd->cpu_dais[i]; +if (cpu_dai->driver->ops->bespoke_trigger) { + ret = cpu_dai->driver->ops->bespoke_trigger (substream, + cmd, cpu_dai); +if (ret <0 ) +return ret; + } }return 0 ; }
@@ -1157,12 +1287,13 @@ static snd_pcm_uframes_t soc_pcm_pointer(struct snd_pcm_substream *substream)struct snd_soc_pcm_runtime *rtd = substream->private_data;struct snd_soc_component *component;struct snd_soc_rtdcom_list *rtdcom; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai; +struct snd_soc_dai *cpu_dai;struct snd_soc_dai *codec_dai;struct snd_pcm_runtime *runtime = substream->runtime;snd_pcm_uframes_t offset =0 ;snd_pcm_sframes_t delay =0 ;snd_pcm_sframes_t codec_delay =0 ; +snd_pcm_sframes_t cpu_delay =0 ;int i;for_each_rtdcom (rtd, rtdcom) {
@@ -1177,8 +1308,16 @@ static snd_pcm_uframes_t soc_pcm_pointer(struct snd_pcm_substream *substream)break ; } +/* +* Use the delay of cpu_dai[0] as in the case of multi cpu DAIs, the +* CPU DAIs are expected to start together and will have the same +* delay +*/ + cpu_dai = rtd->cpu_dais[0 ];if (cpu_dai->driver->ops->delay) - delay += cpu_dai->driver->ops->delay (substream, cpu_dai); + cpu_delay = cpu_dai->driver->ops->delay (substream, cpu_dai); + + delay += cpu_delay;for (i =0 ; i < rtd->num_codecs; i++) { codec_dai = rtd->codec_dais[i];
@@ -1301,12 +1440,16 @@ static struct snd_soc_pcm_runtime *dpcm_get_be(struct snd_soc_card *card,if (!be->dai_link->no_pcm)continue ; -dev_dbg (card->dev,"ASoC: try BE : %s\n" , - be->cpu_dai->playback_widget ? - be->cpu_dai->playback_widget->name :"(not set)" ); +for (i =0 ; i < be->num_cpu_dai; i++) { +struct snd_soc_dai *cpu_dai = be->cpu_dais[i]; -if (be->cpu_dai->playback_widget == widget) -return be; +dev_dbg (card->dev,"ASoC: try BE : %s\n" , + cpu_dai->playback_widget ? + cpu_dai->playback_widget->name :"(not set)" ); + +if (cpu_dai->playback_widget == widget) +return be; + }for (i =0 ; i < be->num_codecs; i++) {struct snd_soc_dai *dai = be->codec_dais[i];
@@ -1321,12 +1464,16 @@ static struct snd_soc_pcm_runtime *dpcm_get_be(struct snd_soc_card *card,if (!be->dai_link->no_pcm)continue ; -dev_dbg (card->dev,"ASoC: try BE %s\n" , - be->cpu_dai->capture_widget ? - be->cpu_dai->capture_widget->name :"(not set)" ); +for (i =0 ; i < be->num_cpu_dai; i++) { +struct snd_soc_dai *cpu_dai = be->cpu_dais[i]; -if (be->cpu_dai->capture_widget == widget) -return be; +dev_dbg (card->dev,"ASoC: try BE %s\n" , + cpu_dai->capture_widget ? + cpu_dai->capture_widget->name :"(not set)" ); + +if (cpu_dai->capture_widget == widget) +return be; + }for (i =0 ; i < be->num_codecs; i++) {struct snd_soc_dai *dai = be->codec_dais[i];
@@ -1376,8 +1523,12 @@ static bool dpcm_end_walk_at_be(struct snd_soc_dapm_widget *widget,if (!rtd->dai_link->no_pcm)continue ; -if (rtd->cpu_dai->playback_widget == widget) -return true; +for (i =0 ; i < rtd->num_cpu_dai; ++i) { +struct snd_soc_dai *cpu_dai = rtd->cpu_dais[i]; + +if (cpu_dai->playback_widget == widget) +return true; + }for (i =0 ; i < rtd->num_codecs; ++i) {struct snd_soc_dai *dai = rtd->codec_dais[i];
@@ -1390,8 +1541,12 @@ static bool dpcm_end_walk_at_be(struct snd_soc_dapm_widget *widget,if (!rtd->dai_link->no_pcm)continue ; -if (rtd->cpu_dai->capture_widget == widget) -return true; +for (i =0 ; i < rtd->num_cpu_dai; ++i) { +struct snd_soc_dai *cpu_dai = rtd->cpu_dais[i]; + +if (cpu_dai->capture_widget == widget) +return true; + }for (i =0 ; i < rtd->num_codecs; ++i) {struct snd_soc_dai *dai = rtd->codec_dais[i];
@@ -1431,12 +1586,20 @@ static int dpcm_prune_paths(struct snd_soc_pcm_runtime *fe, int stream,/* Destroy any old FE <--> BE connections */ list_for_each_entry (dpcm, &fe->dpcm[stream].be_clients, list_be) {unsigned int i; +int widget_found =0 ;/* is there a valid CPU DAI widget for this BE */ - widget =dai_get_widget (dpcm->be->cpu_dai, stream); +for (i =0 ; i < dpcm->be->num_cpu_dai; i++) { +struct snd_soc_dai *dai = dpcm->be->cpu_dais[i]; + + widget =dai_get_widget (dai, stream); + +/* prune the BE if it's no longer in our active list */ +if (widget &&widget_in_list (list, widget)) + widget_found++; + } -/* prune the BE if it's no longer in our active list */ -if (widget &&widget_in_list (list, widget)) +if (widget_found >0 )continue ;/* is there a valid CODEC DAI widget for this BE */
@@ -1778,11 +1941,13 @@ static int dpcm_apply_symmetry(struct snd_pcm_substream *fe_substream, be_substream->runtime->hw.info |=SNDRV_PCM_INFO_JOINT_DUPLEX ;/* Symmetry only applies if we've got an active stream. */ -if (rtd->cpu_dai->active) { - err =soc_pcm_apply_symmetry (fe_substream, - rtd->cpu_dai); -if (err <0 ) -return err; +for (i =0 ; i < rtd->num_cpu_dai; i++) { +if (rtd->cpu_dais[i]->active) { + err =soc_pcm_apply_symmetry (be_substream, + rtd->cpu_dais[i]); +if (err <0 ) +return err; + } }for (i =0 ; i < rtd->num_codecs; i++) {
@@ -2884,13 +3049,13 @@ static int soc_rtdcom_mmap(struct snd_pcm_substream *substream,int soc_new_pcm (struct snd_soc_pcm_runtime *rtd,int num) {struct snd_soc_dai *codec_dai; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai; +struct snd_soc_dai *cpu_dai;struct snd_soc_component *component;struct snd_soc_rtdcom_list *rtdcom;struct snd_pcm *pcm;char new_name[64 ];int ret =0 , playback =0 , capture =0 ; -int i; +int i, cpu_capture =0 , cpu_playback =0 ;if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { playback = rtd->dai_link->dpcm_playback;
@@ -2904,8 +3069,16 @@ int soc_new_pcm(struct snd_soc_pcm_runtime *rtd, int num) capture =1 ; } - capture = capture && cpu_dai->driver->capture.channels_min; - playback = playback && cpu_dai->driver->playback.channels_min; +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai = rtd->cpu_dais[i]; +if (cpu_dai->driver->playback.channels_min) + cpu_playback =1 ; +if (cpu_dai->driver->capture.channels_min) + cpu_capture =1 ; + } + + playback = playback && cpu_playback; + capture = capture && cpu_capture; }if (rtd->dai_link->playback_only) {
@@ -3026,7 +3199,7 @@ int soc_new_pcm(struct snd_soc_pcm_runtime *rtd, int num) out:dev_info (rtd->card->dev,"%s <-> %s mapping ok\n" , (rtd->num_codecs >1 ) ?"multicodec" : rtd->codec_dai->name, - cpu_dai->name); + (rtd->num_cpu_dai >1 ) ?"multicpu" : rtd->cpu_dai->name);return ret; }
Signature · 1 line
2.7.4
Re: [PATCH v6 2/3] ASoC: Add multiple CPU DAI support for PCM ops
Quoted 3 lines · Shreyas NC
Add support in PCM operations to invoke multiple cpu dais as we do for multiple codec dais. Also the symmetry calculations are updated to reflect multiple cpu dais.
This doesn't apply on Mark's tree?
Looks like you need to rebase on top of 244e293690a6 ("ASoC: pcm: Tidy up open/hw_params handling") contributed by Charles on June 19.
Quoted 31 lines
Reviewed-by Charles Keepax <[email protected]>Signed-off-by Vinod Koul <[email protected]>Signed-off-by Shreyas NC <[email protected]>--- sound/soc/soc-pcm.c | 491 +++++++++++++++++++++++++++++++++++----------------- 1 file changed, 332 insertions(+), 159 deletions(-)sound/soc/soc-pcm.c +4
-3@@ -64,23 +64,27 @@ static bool snd_soc_dai_stream_valid(struct snd_soc_dai *dai, int stream)*/ void snd_soc_runtime_activate (struct snd_soc_pcm_runtime *rtd,int stream) { -struct snd_soc_dai *cpu_dai = rtd->cpu_dai;int i;lockdep_assert_held (&rtd->pcm_mutex);if (stream ==SNDRV_PCM_STREAM_PLAYBACK ) { - cpu_dai->playback_active++; +for (i =0 ; i < rtd->num_cpu_dai; i++) + rtd->cpu_dais[i]->playback_active++;for (i =0 ; i < rtd->num_codecs; i++) rtd->codec_dais[i]->playback_active++; }else { - cpu_dai->capture_active++; +for (i =0 ; i < rtd->num_cpu_dai; i++) + rtd->cpu_dais[i]->capture_active++;
sound/soc/soc-pcm.c static bool snd_soc_dai_stream_valid(struct snd_soc_dai *dai, int stream) · line 86
Quoted 10 lines
sound/soc/soc-pcm.c +4
-2for (i =0 ; i < rtd->num_codecs; i++) rtd->codec_dais[i]->capture_active++; } - cpu_dai->active++; - cpu_dai->component->active++; +for (i =0 ; i < rtd->num_cpu_dai; i++) { + rtd->cpu_dais[i]->component->active++; + rtd->cpu_dais[i]->active++; + }
This is becoming complicated, how many times do we need to ref-count?
Quoted 46 lines
sound/soc/soc-pcm.c +10
-6+for (i =0 ; i < rtd->num_codecs; i++) { rtd->codec_dais[i]->active++; rtd->codec_dais[i]->component->active++;@@ -99,23 +103,27 @@ void snd_soc_runtime_activate(struct snd_soc_pcm_runtime *rtd, int stream)*/ void snd_soc_runtime_deactivate (struct snd_soc_pcm_runtime *rtd,int stream) { -struct snd_soc_dai *cpu_dai = rtd->cpu_dai;int i;lockdep_assert_held (&rtd->pcm_mutex);if (stream ==SNDRV_PCM_STREAM_PLAYBACK ) { - cpu_dai->playback_active--; +for (i =0 ; i < rtd->num_cpu_dai; i++) + rtd->cpu_dais[i]->playback_active--;for (i =0 ; i < rtd->num_codecs; i++) rtd->codec_dais[i]->playback_active--; }else { - cpu_dai->capture_active--; +for (i =0 ; i < rtd->num_cpu_dai; i++) + rtd->cpu_dais[i]->capture_active--;for (i =0 ; i < rtd->num_codecs; i++) rtd->codec_dais[i]->capture_active--; } - cpu_dai->active--; - cpu_dai->component->active--; +for (i =0 ; i < rtd->num_cpu_dai; i++) { + rtd->cpu_dais[i]->component->active--; + rtd->cpu_dais[i]->active--; + } +for (i =0 ; i < rtd->num_codecs; i++) { rtd->codec_dais[i]->component->active--; rtd->codec_dais[i]->active--;@@ -258,7 +266,6 @@ static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, struct snd_pcm_hw_params *params) {struct snd_soc_pcm_runtime *rtd = substream->private_data; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai;unsigned int rate, channels, sample_bits, symmetry, i; rate =params_rate (params);
sound/soc/soc-pcm.c static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, · line 270
Quoted 5 lines
sound/soc/soc-pcm.c +0
-2@@ -266,41 +273,54 @@ static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, sample_bits =snd_pcm_format_physical_width (params_format (params));/* reject unmatched parameters when applying symmetry */ - symmetry = cpu_dai->driver->symmetric_rates || - rtd->dai_link->symmetric_rates;
This was a logical OR, but...
sound/soc/soc-pcm.c static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, · line 279
Quoted 4 lines
sound/soc/soc-pcm.c +4
-0+ symmetry = rtd->dai_link->symmetric_rates; + +for (i =0 ; i < rtd->num_cpu_dai; i++) + symmetry |= rtd->cpu_dais[i]->driver->symmetric_rates;
this is a bitwise OR. is this ok?
sound/soc/soc-pcm.c static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, · line 285
Quoted 11 lines
sound/soc/soc-pcm.c +2
-5for (i =0 ; i < rtd->num_codecs; i++) symmetry |= rtd->codec_dais[i]->driver->symmetric_rates; -if (symmetry && cpu_dai->rate && cpu_dai->rate != rate) { -dev_err (rtd->dev,"ASoC: unmatched rate symmetry: %d - %d\n" , - cpu_dai->rate, rate); -return -EINVAL ; - } +for (i =0 ; i < rtd->num_cpu_dai; i++) + if (symmetry && rtd->cpu_dais[i]->rate &&
you could move the if (symmetry) out of the for loop to simplify
Quoted 13 lines
sound/soc/soc-pcm.c +9
-2+ rtd->cpu_dais[i]->rate != rate) { +dev_err (rtd->dev,"ASoC: unmatched rate symmetry: %d - %d\n" , + rtd->cpu_dais[i]->rate, rate); +return -EINVAL ; + } - symmetry = cpu_dai->driver->symmetric_channels || - rtd->dai_link->symmetric_channels; + symmetry = rtd->dai_link->symmetric_channels; + +for (i =0 ; i < rtd->num_cpu_dai; i++) + symmetry |= rtd->cpu_dais[i]->driver->symmetric_channels;
sound/soc/soc-pcm.c static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, · line 301
Quoted 10 lines
sound/soc/soc-pcm.c +2
-5for (i =0 ; i < rtd->num_codecs; i++) symmetry |= rtd->codec_dais[i]->driver->symmetric_channels; -if (symmetry && cpu_dai->channels && cpu_dai->channels != channels) { -dev_err (rtd->dev,"ASoC: unmatched channel symmetry: %d - %d\n" , - cpu_dai->channels, channels); -return -EINVAL ; - } +for (i =0 ; i < rtd->num_cpu_dai; i++) + if (symmetry && rtd->cpu_dais[i]->channels &&
here as well
Quoted 37 lines
sound/soc/soc-pcm.c +17
-8+ rtd->cpu_dais[i]->channels != channels) { +dev_err (rtd->dev,"ASoC: unmatched channel symmetry: %d - %d\n" , + rtd->cpu_dais[i]->channels, channels); +return -EINVAL ; + } - symmetry = cpu_dai->driver->symmetric_samplebits || - rtd->dai_link->symmetric_samplebits; + symmetry = rtd->dai_link->symmetric_samplebits; + +for (i =0 ; i < rtd->num_cpu_dai; i++) + symmetry |= rtd->cpu_dais[i]->driver->symmetric_samplebits;for (i =0 ; i < rtd->num_codecs; i++) symmetry |= rtd->codec_dais[i]->driver->symmetric_samplebits; -if (symmetry && cpu_dai->sample_bits && cpu_dai->sample_bits != sample_bits) { -dev_err (rtd->dev,"ASoC: unmatched sample bits symmetry: %d - %d\n" , - cpu_dai->sample_bits, sample_bits); -return -EINVAL ; - } +for (i =0 ; i < rtd->num_cpu_dai; i++) +if (symmetry && rtd->cpu_dais[i]->sample_bits && + rtd->cpu_dais[i]->sample_bits != sample_bits) { +dev_err (rtd->dev,"ASoC: unmatched sample bits symmetry: %d - %d\n" , + rtd->cpu_dais[i]->sample_bits, + sample_bits); +return -EINVAL ; + }return 0 ; }@@ -308,13 +328,18 @@ static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream,static bool soc_pcm_has_symmetry (struct snd_pcm_substream *substream) {struct snd_soc_pcm_runtime *rtd = substream->private_data; -struct snd_soc_dai_driver *cpu_driver = rtd->cpu_dai->driver;
sound/soc/soc-pcm.c static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, · line 337
Quoted 10 lines
sound/soc/soc-pcm.c +4
-3struct snd_soc_dai_link *link = rtd->dai_link;unsigned int symmetry, i; - symmetry = cpu_driver->symmetric_rates || link->symmetric_rates || - cpu_driver->symmetric_channels || link->symmetric_channels || - cpu_driver->symmetric_samplebits || link->symmetric_samplebits; + symmetry = link->symmetric_rates || link->symmetric_channels || + link->symmetric_samplebits; + +/* Apply symmetery for multiple cpu dais */
You haven't fixed this comment, is this patch the correct one?
Quoted 12 lines
sound/soc/soc-pcm.c +7
-0@@ -422,30 +461,55 @@ static void soc_pcm_init_runtime_hw(struct snd_pcm_substream *substream) rates =snd_pcm_rate_mask_intersect (codec_stream->rates, rates); } +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai_drv = rtd->cpu_dais[i]->driver; + +if (substream->stream ==SNDRV_PCM_STREAM_PLAYBACK ) + cpu_stream = &cpu_dai_drv->playback; +else + cpu_stream = &cpu_dai_drv->capture;
sound/soc/soc-pcm.c static void soc_pcm_init_runtime_hw(struct snd_pcm_substream *substream) · line 480
Quoted 10 lines
sound/soc/soc-pcm.c +10
-0+ + cpu_chan_min =max (cpu_chan_min, + cpu_stream->channels_min); + cpu_chan_max =min (cpu_chan_max, + cpu_stream->channels_max); + +if (hw->formats) + hw->formats &= cpu_stream->formats; +else + hw->formats = cpu_stream->formats;
Can you double-check the 'else' case? This doesn't seem right, you will always use the format used for the last cpu_dai. If the formats are identical for all cpu_dais, then this can be moved outside of the loop.
Quoted 7 lines
sound/soc/soc-pcm.c +0
-3@@ -963,11 +1070,14 @@ static int soc_pcm_hw_params(struct snd_pcm_substream *substream,if (ret <0 ) goto component_err; -/* store the parameters for each DAIs */ - cpu_dai->rate =params_rate (params); - cpu_dai->channels =params_channels (params);
sound/soc/soc-pcm.c static int soc_pcm_hw_params(struct snd_pcm_substream *substream, · line 1080
Quoted 10 lines
sound/soc/soc-pcm.c +8
-2- cpu_dai->sample_bits = -snd_pcm_format_physical_width (params_format (params)); +for (i =0 ; i < rtd->num_cpu_dai; i++) { +/* store the parameters for each DAIs */ + cpu_dai = rtd->cpu_dais[i]; + cpu_dai->rate =params_rate (params); + cpu_dai->channels =params_channels (params); + cpu_dai->sample_bits = +snd_pcm_format_physical_width (params_format (params)); + }
I think I've asked this before but can't recall the answer: does this mean we assume the same number of channels for each codec_dai[j] and cpu_dai[i]?
sound/soc/soc-pcm.c static int soc_pcm_trigger(struct snd_pcm_substream *substream, int cmd) · line 1239
Quoted 8 lines
sound/soc/soc-pcm.c +8
-0+for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai = rtd->cpu_dais[i]; +if (cpu_dai->driver->ops->trigger) { + ret = cpu_dai->driver->ops->trigger (substream, + cmd, cpu_dai); +if (ret <0 ) +return ret; + }
Maybe add a comment that in the multi_cpu case the triggers are supposed to be aligned, i.e. the first trigger starts the others - that would be consistent with the other comments on delay below.
Quoted 7 lines
sound/soc/soc-pcm.c +0
-3@@ -1778,11 +1941,13 @@ static int dpcm_apply_symmetry(struct snd_pcm_substream *fe_substream, be_substream->runtime->hw.info |=SNDRV_PCM_INFO_JOINT_DUPLEX ;/* Symmetry only applies if we've got an active stream. */ -if (rtd->cpu_dai->active) { - err =soc_pcm_apply_symmetry (fe_substream, - rtd->cpu_dai);
sound/soc/soc-pcm.c static int dpcm_apply_symmetry(struct snd_pcm_substream *fe_substream, · line 1951
Quoted 10 lines
sound/soc/soc-pcm.c +7
-2-if (err <0 ) -return err; +for (i =0 ; i < rtd->num_cpu_dai; i++) { +if (rtd->cpu_dais[i]->active) { + err =soc_pcm_apply_symmetry (be_substream, + rtd->cpu_dais[i]); +if (err <0 ) +return err; + } }
Can you recheck this block? In the original code the symmetry is with the fe_substream, it's now with a be_substream. Looks to me like a major typo having significant impact of the result...
Quoted 38 lines
sound/soc/soc-pcm.c +12
-4for (i =0 ; i < rtd->num_codecs; i++) {@@ -2884,13 +3049,13 @@ static int soc_rtdcom_mmap(struct snd_pcm_substream *substream,int soc_new_pcm (struct snd_soc_pcm_runtime *rtd,int num) {struct snd_soc_dai *codec_dai; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai; +struct snd_soc_dai *cpu_dai;struct snd_soc_component *component;struct snd_soc_rtdcom_list *rtdcom;struct snd_pcm *pcm;char new_name[64 ];int ret =0 , playback =0 , capture =0 ; -int i; +int i, cpu_capture =0 , cpu_playback =0 ;if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { playback = rtd->dai_link->dpcm_playback;@@ -2904,8 +3069,16 @@ int soc_new_pcm(struct snd_soc_pcm_runtime *rtd, int num) capture =1 ; } - capture = capture && cpu_dai->driver->capture.channels_min; - playback = playback && cpu_dai->driver->playback.channels_min; +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai = rtd->cpu_dais[i]; +if (cpu_dai->driver->playback.channels_min) + cpu_playback =1 ; +if (cpu_dai->driver->capture.channels_min) + cpu_capture =1 ; + } + + playback = playback && cpu_playback; + capture = capture && cpu_capture; }if (rtd->dai_link->playback_only) {
sound/soc/soc-pcm.c int soc_new_pcm(struct snd_soc_pcm_runtime *rtd, int num) · line 3205
Quoted 8 lines
sound/soc/soc-pcm.c +1
-1@@ -3026,7 +3199,7 @@ int soc_new_pcm(struct snd_soc_pcm_runtime *rtd, int num) out:dev_info (rtd->card->dev,"%s <-> %s mapping ok\n" , (rtd->num_codecs >1 ) ?"multicodec" : rtd->codec_dai->name, - cpu_dai->name); + (rtd->num_cpu_dai >1 ) ?"multicpu" : rtd->cpu_dai->name);return ret; }
Took me a couple of hours to reach the end of this patch ... I had to use a visual diff to figure it out, the diff format is just too hard to look at.
Re: [PATCH v6 2/3] ASoC: Add multiple CPU DAI support for PCM ops
Quoted 10 lines · Pierre-Louis Bossart
Quoted 3 lines · Shreyas NC
Add support in PCM operations to invoke multiple cpu dais as we do for multiple codec dais. Also the symmetry calculations are updated to reflect multiple cpu dais.
This doesn't apply on Mark's tree?
Looks like you need to rebase on top of244e293690a6("ASoC: pcm: Tidy up open/hw_params handling")
contributed by Charles on June 19.
I did rebase the morning before I sent these patches, may be I should do it just before sending them out. Will take care of that next time :)
Quoted 32 lines
Quoted 32 lines
Reviewed-by Charles Keepax <[email protected]>Signed-off-by Vinod Koul <[email protected]>Signed-off-by Shreyas NC <[email protected]>--- sound/soc/soc-pcm.c | 491 +++++++++++++++++++++++++++++++++++----------------- 1 file changed, 332 insertions(+), 159 deletions(-)sound/soc/soc-pcm.c +4
-3@@ -64,23 +64,27 @@ static bool snd_soc_dai_stream_valid(struct snd_soc_dai *dai, int stream)*/ void snd_soc_runtime_activate (struct snd_soc_pcm_runtime *rtd,int stream) { -struct snd_soc_dai *cpu_dai = rtd->cpu_dai;int i;lockdep_assert_held (&rtd->pcm_mutex);if (stream ==SNDRV_PCM_STREAM_PLAYBACK ) { - cpu_dai->playback_active++; +for (i =0 ; i < rtd->num_cpu_dai; i++) + rtd->cpu_dais[i]->playback_active++;for (i =0 ; i < rtd->num_codecs; i++) rtd->codec_dais[i]->playback_active++; }else { - cpu_dai->capture_active++; +for (i =0 ; i < rtd->num_cpu_dai; i++) + rtd->cpu_dais[i]->capture_active++;for (i =0 ; i < rtd->num_codecs; i++) rtd->codec_dais[i]->capture_active++; }
Quoted 8 lines
sound/soc/soc-pcm.c static bool snd_soc_dai_stream_valid(struct snd_soc_dai *dai, int stream) · line 86
Quoted 6 lines
sound/soc/soc-pcm.c +4
-2- cpu_dai->active++; - cpu_dai->component->active++; +for (i =0 ; i < rtd->num_cpu_dai; i++) { + rtd->cpu_dais[i]->component->active++; + rtd->cpu_dais[i]->active++; + }This is becoming complicated, how many times do we need to ref-count?
Earlier we incremented cpu_dai->playback_active++ and here it is cpu_dais->component->active++
Quoted 48 lines
Quoted 42 lines
sound/soc/soc-pcm.c +10
-6+for (i =0 ; i < rtd->num_codecs; i++) { rtd->codec_dais[i]->active++; rtd->codec_dais[i]->component->active++;@@ -99,23 +103,27 @@ void snd_soc_runtime_activate(struct snd_soc_pcm_runtime *rtd, int stream)*/ void snd_soc_runtime_deactivate (struct snd_soc_pcm_runtime *rtd,int stream) { -struct snd_soc_dai *cpu_dai = rtd->cpu_dai;int i;lockdep_assert_held (&rtd->pcm_mutex);if (stream ==SNDRV_PCM_STREAM_PLAYBACK ) { - cpu_dai->playback_active--; +for (i =0 ; i < rtd->num_cpu_dai; i++) + rtd->cpu_dais[i]->playback_active--;for (i =0 ; i < rtd->num_codecs; i++) rtd->codec_dais[i]->playback_active--; }else { - cpu_dai->capture_active--; +for (i =0 ; i < rtd->num_cpu_dai; i++) + rtd->cpu_dais[i]->capture_active--;for (i =0 ; i < rtd->num_codecs; i++) rtd->codec_dais[i]->capture_active--; } - cpu_dai->active--; - cpu_dai->component->active--; +for (i =0 ; i < rtd->num_cpu_dai; i++) { + rtd->cpu_dais[i]->component->active--; + rtd->cpu_dais[i]->active--; + } +for (i =0 ; i < rtd->num_codecs; i++) { rtd->codec_dais[i]->component->active--; rtd->codec_dais[i]->active--;@@ -258,7 +266,6 @@ static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, struct snd_pcm_hw_params *params) {struct snd_soc_pcm_runtime *rtd = substream->private_data; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai;unsigned int rate, channels, sample_bits, symmetry, i; rate =params_rate (params);sound/soc/soc-pcm.c static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, · line 269
Quoted 4 lines
sound/soc/soc-pcm.c +0
-2@@ -266,41 +273,54 @@ static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, sample_bits =snd_pcm_format_physical_width (params_format (params));/* reject unmatched parameters when applying symmetry */ - symmetry = cpu_dai->driver->symmetric_rates || - rtd->dai_link->symmetric_rates;This was a logical OR, but...
Quoted 5 lines
sound/soc/soc-pcm.c static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, · line 279
Quoted 4 lines
sound/soc/soc-pcm.c +4
-0+ symmetry = rtd->dai_link->symmetric_rates; + +for (i =0 ; i < rtd->num_cpu_dai; i++) + symmetry |= rtd->cpu_dais[i]->driver->symmetric_rates;this is a bitwise OR. is this ok?
I made this change on purpose and I am struggling to remember the reason :( If I dont figure that out, I will go back to logical OR..
Quoted 10 lines
sound/soc/soc-pcm.c static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, · line 285
Quoted 9 lines
sound/soc/soc-pcm.c +2
-5for (i =0 ; i < rtd->num_codecs; i++) symmetry |= rtd->codec_dais[i]->driver->symmetric_rates; -if (symmetry && cpu_dai->rate && cpu_dai->rate != rate) { -dev_err (rtd->dev,"ASoC: unmatched rate symmetry: %d - %d\n" , - cpu_dai->rate, rate); -return -EINVAL ; - } +for (i =0 ; i < rtd->num_cpu_dai; i++) + if (symmetry && rtd->cpu_dais[i]->rate &&you could move the if (symmetry) out of the for loop to simplify
Yes, that makes sense :)
Quoted 7 lines
Quoted 7 lines
sound/soc/soc-pcm.c +0
-1@@ -308,13 +328,18 @@ static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream,static bool soc_pcm_has_symmetry (struct snd_pcm_substream *substream) {struct snd_soc_pcm_runtime *rtd = substream->private_data; -struct snd_soc_dai_driver *cpu_driver = rtd->cpu_dai->driver;struct snd_soc_dai_link *link = rtd->dai_link;unsigned int symmetry, i;
Quoted 8 lines
sound/soc/soc-pcm.c static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, · line 337
Quoted 7 lines
sound/soc/soc-pcm.c +4
-3- symmetry = cpu_driver->symmetric_rates || link->symmetric_rates || - cpu_driver->symmetric_channels || link->symmetric_channels || - cpu_driver->symmetric_samplebits || link->symmetric_samplebits; + symmetry = link->symmetric_rates || link->symmetric_channels || + link->symmetric_samplebits; + +/* Apply symmetery for multiple cpu dais */ You haven't fixed this comment, is this patch the correct one?
I am surprised why this crept in :(
This was the first comment I fixed ..
Quoted 15 lines
Quoted 15 lines
sound/soc/soc-pcm.c +12
-0@@ -422,30 +461,55 @@ static void soc_pcm_init_runtime_hw(struct snd_pcm_substream *substream) rates =snd_pcm_rate_mask_intersect (codec_stream->rates, rates); } +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai_drv = rtd->cpu_dais[i]->driver; + +if (substream->stream ==SNDRV_PCM_STREAM_PLAYBACK ) + cpu_stream = &cpu_dai_drv->playback; +else + cpu_stream = &cpu_dai_drv->capture; + + cpu_chan_min =max (cpu_chan_min, + cpu_stream->channels_min); + cpu_chan_max =min (cpu_chan_max, + cpu_stream->channels_max);
Quoted 8 lines
sound/soc/soc-pcm.c static void soc_pcm_init_runtime_hw(struct snd_pcm_substream *substream) · line 480
Quoted 5 lines
sound/soc/soc-pcm.c +5
-0+ +if (hw->formats) + hw->formats &= cpu_stream->formats; +else + hw->formats = cpu_stream->formats;Can you double-check the 'else' case? This doesn't seem right, you will always use the format used for the last
cpu_dai. If the formats are identical for allcpu_dais, then this can be moved outside of the loop.
In the second iteration, we would always go into the if (hw->formats) case.
Quoted 10 lines
Quoted 10 lines
sound/soc/soc-pcm.c +2
-5@@ -963,11 +1070,14 @@ static int soc_pcm_hw_params(struct snd_pcm_substream *substream,if (ret <0 ) goto component_err; -/* store the parameters for each DAIs */ - cpu_dai->rate =params_rate (params); - cpu_dai->channels =params_channels (params); - cpu_dai->sample_bits = -snd_pcm_format_physical_width (params_format (params)); +for (i =0 ; i < rtd->num_cpu_dai; i++) { +/* store the parameters for each DAIs */
Quoted 9 lines
sound/soc/soc-pcm.c static int soc_pcm_hw_params(struct snd_pcm_substream *substream, · line 1080
Quoted 6 lines
sound/soc/soc-pcm.c +6
-0+ cpu_dai = rtd->cpu_dais[i]; + cpu_dai->rate =params_rate (params); + cpu_dai->channels =params_channels (params); + cpu_dai->sample_bits = +snd_pcm_format_physical_width (params_format (params)); + }I think I've asked this before but can't recall the answer: does this mean we assume the same number of channels for each codec_dai[j] and cpu_dai[i]?
Yes, in hw_params we set the same number channels on all the DAIs as that of the stream. But, as I had mentioned in my previous replies, the
machine driver would set the channel masks on the individual DAIs(like we do for SoundWire Multi link case)
Quoted 4 lines
Quoted 4 lines
sound/soc/soc-pcm.c +4
-0+for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai = rtd->cpu_dais[i]; +if (cpu_dai->driver->ops->trigger) { + ret = cpu_dai->driver->ops->trigger(substream,
Quoted 8 lines
sound/soc/soc-pcm.c static int soc_pcm_trigger(struct snd_pcm_substream *substream, int cmd) · line 1239
Quoted 4 lines
sound/soc/soc-pcm.c +4
-0+ cmd, cpu_dai); +if (ret <0 ) +return ret; + }Maybe add a comment that in the
multi_cpucase the triggers are supposed to be aligned, i.e. the first trigger starts the others - that would be consistent with the other comments on delay below.
Would this be the right place to add that comment?
I am not sure if I got a reply to my question in the previous review cycle:
"Just wondering if there can be a case of multiple CPU DAIs but you would not want them to be triggered at the same time."
Based on the answer to my question , we can add a comment here.
Quoted 11 lines
Quoted 11 lines
sound/soc/soc-pcm.c +3
-5@@ -1778,11 +1941,13 @@ static int dpcm_apply_symmetry(struct snd_pcm_substream *fe_substream, be_substream->runtime->hw.info |=SNDRV_PCM_INFO_JOINT_DUPLEX ;/* Symmetry only applies if we've got an active stream. */ -if (rtd->cpu_dai->active) { - err =soc_pcm_apply_symmetry (fe_substream, - rtd->cpu_dai); -if (err <0 ) -return err; +for (i =0 ; i < rtd->num_cpu_dai; i++) { +if (rtd->cpu_dais[i]->active) { + err = soc_pcm_apply_symmetry(be_substream,
Quoted 9 lines
sound/soc/soc-pcm.c static int dpcm_apply_symmetry(struct snd_pcm_substream *fe_substream, · line 1951
Quoted 5 lines
sound/soc/soc-pcm.c +4
-0+ rtd->cpu_dais[i]); +if (err <0 ) +return err; + } }Can you recheck this block? In the original code the symmetry is with the
fe_substream, it's now with abe_substream. Looks to me like a major typo having significant impact of the result...
This was recently fixed and changed from be_substream to fe_substream :( commit 99bcedbdebc57fe5d02fb470b7265f2208c2cf96
ASoC: dpcm: symmetry constraint on FE substream
So, I need to fix my patch as well. Nice catch :)
Quoted 36 lines
Quoted 36 lines
sound/soc/soc-pcm.c +12
-4for (i =0 ; i < rtd->num_codecs; i++) {@@ -2884,13 +3049,13 @@ static int soc_rtdcom_mmap(struct snd_pcm_substream *substream,int soc_new_pcm (struct snd_soc_pcm_runtime *rtd,int num) {struct snd_soc_dai *codec_dai; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai; +struct snd_soc_dai *cpu_dai;struct snd_soc_component *component;struct snd_soc_rtdcom_list *rtdcom;struct snd_pcm *pcm;char new_name[64 ];int ret =0 , playback =0 , capture =0 ; -int i; +int i, cpu_capture =0 , cpu_playback =0 ;if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { playback = rtd->dai_link->dpcm_playback;@@ -2904,8 +3069,16 @@ int soc_new_pcm(struct snd_soc_pcm_runtime *rtd, int num) capture =1 ; } - capture = capture && cpu_dai->driver->capture.channels_min; - playback = playback && cpu_dai->driver->playback.channels_min; +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai = rtd->cpu_dais[i]; +if (cpu_dai->driver->playback.channels_min) + cpu_playback =1 ; +if (cpu_dai->driver->capture.channels_min) + cpu_capture =1 ; + } + + playback = playback && cpu_playback; + capture = capture && cpu_capture; }if (rtd->dai_link->playback_only) {@@ -3026,7 +3199,7 @@ int soc_new_pcm(struct snd_soc_pcm_runtime *rtd, int num) out: dev_info(rtd->card->dev, "%s <-> %s mapping ok\n",
Quoted 9 lines
sound/soc/soc-pcm.c int soc_new_pcm(struct snd_soc_pcm_runtime *rtd, int num) · line 3204
Quoted 5 lines
sound/soc/soc-pcm.c +1
-1(rtd->num_codecs >1 ) ?"multicodec" : rtd->codec_dai->name, - cpu_dai->name); + (rtd->num_cpu_dai > 1) ? "multicpu" : rtd->cpu_dai->name);return ret; }Took me a couple of hours to reach the end of this patch ... I had to use a visual diff to figure it out, the diff format is just too hard to look at.
Unfortunately yes :(
Re: [PATCH v6 2/3] ASoC: Add multiple CPU DAI support for PCM ops
Quoted 30 lines
Quoted 21 lines
Quoted 21 lines
sound/soc/soc-pcm.c +4
-3@@ -64,23 +64,27 @@ static bool snd_soc_dai_stream_valid(struct snd_soc_dai *dai, int stream)*/ void snd_soc_runtime_activate (struct snd_soc_pcm_runtime *rtd,int stream) { -struct snd_soc_dai *cpu_dai = rtd->cpu_dai;int i;lockdep_assert_held (&rtd->pcm_mutex);if (stream ==SNDRV_PCM_STREAM_PLAYBACK ) { - cpu_dai->playback_active++; +for (i =0 ; i < rtd->num_cpu_dai; i++) + rtd->cpu_dais[i]->playback_active++;for (i =0 ; i < rtd->num_codecs; i++) rtd->codec_dais[i]->playback_active++; }else { - cpu_dai->capture_active++; +for (i =0 ; i < rtd->num_cpu_dai; i++) + rtd->cpu_dais[i]->capture_active++;for (i =0 ; i < rtd->num_codecs; i++) rtd->codec_dais[i]->capture_active++; }Quoted 8 lines
sound/soc/soc-pcm.c static bool snd_soc_dai_stream_valid(struct snd_soc_dai *dai, int stream) · line 86
Quoted 6 lines
sound/soc/soc-pcm.c +4
-2- cpu_dai->active++; - cpu_dai->component->active++; +for (i =0 ; i < rtd->num_cpu_dai; i++) { + rtd->cpu_dais[i]->component->active++; + rtd->cpu_dais[i]->active++; + }This is becoming complicated, how many times do we need to ref-count?
Quoted 2 lines
Earlier we incremented cpu_dai->playback_active++ and here it is cpu_dais->component->active++
my point was whether this can be simplified to use a single counter, even before we do the change to multi-cpu?
Quoted 14 lines
Quoted 14 lines
sound/soc/soc-pcm.c static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, · line 269
Quoted 12 lines
sound/soc/soc-pcm.c +0
-3@@ -258,7 +266,6 @@ static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, struct snd_pcm_hw_params *params) {struct snd_soc_pcm_runtime *rtd = substream->private_data; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai;unsigned int rate, channels, sample_bits, symmetry, i; rate =params_rate (params);@@ -266,41 +273,54 @@ static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, sample_bits =snd_pcm_format_physical_width (params_format (params));/* reject unmatched parameters when applying symmetry */ - symmetry = cpu_dai->driver->symmetric_rates || - rtd->dai_link->symmetric_rates;This was a logical OR, but...
Quoted 8 lines
Quoted 5 lines
sound/soc/soc-pcm.c static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, · line 279
Quoted 4 lines
sound/soc/soc-pcm.c +4
-0+ symmetry = rtd->dai_link->symmetric_rates; + +for (i =0 ; i < rtd->num_cpu_dai; i++) + symmetry |= rtd->cpu_dais[i]->driver->symmetric_rates;this is a bitwise OR. is this ok?
I made this change on purpose and I am struggling to remember the reason :( If I dont figure that out, I will go back to logical OR..
the same change was made for multiple codecs, but it's worth rechecking.
Quoted 24 lines
Quoted 15 lines
Quoted 15 lines
sound/soc/soc-pcm.c +12
-0@@ -422,30 +461,55 @@ static void soc_pcm_init_runtime_hw(struct snd_pcm_substream *substream) rates =snd_pcm_rate_mask_intersect (codec_stream->rates, rates); } +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai_drv = rtd->cpu_dais[i]->driver; + +if (substream->stream ==SNDRV_PCM_STREAM_PLAYBACK ) + cpu_stream = &cpu_dai_drv->playback; +else + cpu_stream = &cpu_dai_drv->capture; + + cpu_chan_min =max (cpu_chan_min, + cpu_stream->channels_min); + cpu_chan_max =min (cpu_chan_max, + cpu_stream->channels_max);Quoted 8 lines
sound/soc/soc-pcm.c static void soc_pcm_init_runtime_hw(struct snd_pcm_substream *substream) · line 480
Quoted 5 lines
sound/soc/soc-pcm.c +5
-0+ +if (hw->formats) + hw->formats &= cpu_stream->formats; +else + hw->formats = cpu_stream->formats;Can you double-check the 'else' case? This doesn't seem right, you will always use the format used for the last
cpu_dai. If the formats are identical for allcpu_dais, then this can be moved outside of the loop.
Quoted 1 line
In the second iteration, we would always go into the if (
hw->formats) case.
Sorry I don't get your comment. This test is now placed inside of a for loop, so why would the hw->formats change between iterations?
Quoted 21 lines
Quoted 11 lines
Quoted 11 lines
sound/soc/soc-pcm.c +2
-5@@ -963,11 +1070,14 @@ static int soc_pcm_hw_params(struct snd_pcm_substream *substream,if (ret <0 ) goto component_err; -/* store the parameters for each DAIs */ - cpu_dai->rate =params_rate (params); - cpu_dai->channels =params_channels (params); - cpu_dai->sample_bits = -snd_pcm_format_physical_width (params_format (params)); +for (i =0 ; i < rtd->num_cpu_dai; i++) { +/* store the parameters for each DAIs */ Quoted 9 lines
sound/soc/soc-pcm.c static int soc_pcm_hw_params(struct snd_pcm_substream *substream, · line 1080
Quoted 6 lines
sound/soc/soc-pcm.c +6
-0+ cpu_dai = rtd->cpu_dais[i]; + cpu_dai->rate =params_rate (params); + cpu_dai->channels =params_channels (params); + cpu_dai->sample_bits = +snd_pcm_format_physical_width (params_format (params)); + }I think I've asked this before but can't recall the answer: does this mean we assume the same number of channels for each codec_dai[j] and cpu_dai[i]?
Quoted 4 lines
Yes, in
hw_paramswe set the same number channels on all the DAIs as that of the stream. But, as I had mentioned in my previous replies, the
machine driver would set the channel masks on the individual DAIs(like we do for SoundWire Multi link case)
so the number of channels is really defined by the stream, and for each cpu_dai and codec_dai there is a channel mask defining what the dais should produce/consume, right?
Quoted 14 lines
Quoted 13 lines
sound/soc/soc-pcm.c static int soc_pcm_trigger(struct snd_pcm_substream *substream, int cmd) · line 1239
Quoted 9 lines
sound/soc/soc-pcm.c +8
-0+for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai = rtd->cpu_dais[i]; +if (cpu_dai->driver->ops->trigger) { + ret = cpu_dai->driver->ops->trigger (substream, + cmd, cpu_dai); +if (ret <0 ) +return ret; + }Maybe add a comment that in the
multi_cpucase the triggers are supposed to be aligned, i.e. the first trigger starts the others - that would be consistent with the other comments on delay below.
Quoted 6 lines
Would this be the right place to add that comment?
I am not sure if I got a reply to my question in the previous review cycle:
"Just wondering if there can be a case of multiple CPU DAIs but you would not want them to be triggered at the same time."Based on the answer to my question , we can add a comment here.
If you use the stream/multi cpu_dai concept you want the data to remain phase locked and triggered at the same time. If you don't care about phase, then you are handling different streams that can be handled with the existing model with a set of runtimes that deal with individual cpu_dais.
Quoted 22 lines
Quoted 12 lines
Quoted 12 lines
sound/soc/soc-pcm.c +3
-5@@ -1778,11 +1941,13 @@ static int dpcm_apply_symmetry(struct snd_pcm_substream *fe_substream, be_substream->runtime->hw.info |=SNDRV_PCM_INFO_JOINT_DUPLEX ;/* Symmetry only applies if we've got an active stream. */ -if (rtd->cpu_dai->active) { - err =soc_pcm_apply_symmetry (fe_substream, - rtd->cpu_dai); -if (err <0 ) -return err; +for (i =0 ; i < rtd->num_cpu_dai; i++) { +if (rtd->cpu_dais[i]->active) { + err = soc_pcm_apply_symmetry(be_substream,Quoted 9 lines
sound/soc/soc-pcm.c static int dpcm_apply_symmetry(struct snd_pcm_substream *fe_substream, · line 1951
Quoted 5 lines
sound/soc/soc-pcm.c +4
-0+ rtd->cpu_dais[i]); +if (err <0 ) +return err; + } }Can you recheck this block? In the original code the symmetry is with the
fe_substream, it's now with abe_substream. Looks to me like a major typo having significant impact of the result...
Quoted 6 lines
This was recently fixed and changed from
be_substreamtofe_substream:( commit99bcedbdebc57fe5d02fb470b7265f2208c2cf96ASoC: dpcm: symmetry constraint on FE substream
So, I need to fix my patch as well. Nice catch :)
I see. We want to double-check all the fixes since this work started (and yes unfortunately playing with patches lets this sort of things go through, I had the same issue yesterday with my SOF renaming patch).
Quoted 49 lines
Quoted 37 lines
Quoted 37 lines
sound/soc/soc-pcm.c +12
-4for (i =0 ; i < rtd->num_codecs; i++) {@@ -2884,13 +3049,13 @@ static int soc_rtdcom_mmap(struct snd_pcm_substream *substream,int soc_new_pcm (struct snd_soc_pcm_runtime *rtd,int num) {struct snd_soc_dai *codec_dai; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai; +struct snd_soc_dai *cpu_dai;struct snd_soc_component *component;struct snd_soc_rtdcom_list *rtdcom;struct snd_pcm *pcm;char new_name[64 ];int ret =0 , playback =0 , capture =0 ; -int i; +int i, cpu_capture =0 , cpu_playback =0 ;if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { playback = rtd->dai_link->dpcm_playback;@@ -2904,8 +3069,16 @@ int soc_new_pcm(struct snd_soc_pcm_runtime *rtd, int num) capture =1 ; } - capture = capture && cpu_dai->driver->capture.channels_min; - playback = playback && cpu_dai->driver->playback.channels_min; +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai = rtd->cpu_dais[i]; +if (cpu_dai->driver->playback.channels_min) + cpu_playback =1 ; +if (cpu_dai->driver->capture.channels_min) + cpu_capture =1 ; + } + + playback = playback && cpu_playback; + capture = capture && cpu_capture; }if (rtd->dai_link->playback_only) {@@ -3026,7 +3199,7 @@ int soc_new_pcm(struct snd_soc_pcm_runtime *rtd, int num) out: dev_info(rtd->card->dev, "%s <-> %s mapping ok\n",Quoted 9 lines
sound/soc/soc-pcm.c int soc_new_pcm(struct snd_soc_pcm_runtime *rtd, int num) · line 3204
Quoted 5 lines
sound/soc/soc-pcm.c +1
-1(rtd->num_codecs >1 ) ?"multicodec" : rtd->codec_dai->name, - cpu_dai->name); + (rtd->num_cpu_dai > 1) ? "multicpu" : rtd->cpu_dai->name);return ret; }Took me a couple of hours to reach the end of this patch ... I had to use a visual diff to figure it out, the diff format is just too hard to look at.
Unfortunately yes :(
Re: [PATCH v6 2/3] ASoC: Add multiple CPU DAI support for PCM ops
Quoted 30 lines · Pierre-Louis Bossart
Quoted 30 lines
Quoted 29 lines
Quoted 17 lines
sound/soc/soc-pcm.c +3
-3@@ -64,23 +64,27 @@ static bool snd_soc_dai_stream_valid(struct snd_soc_dai *dai, int stream)*/ void snd_soc_runtime_activate (struct snd_soc_pcm_runtime *rtd,int stream) { -struct snd_soc_dai *cpu_dai = rtd->cpu_dai;int i;lockdep_assert_held (&rtd->pcm_mutex);if (stream ==SNDRV_PCM_STREAM_PLAYBACK ) { - cpu_dai->playback_active++; +for (i =0 ; i < rtd->num_cpu_dai; i++) + rtd->cpu_dais[i]->playback_active++;for (i =0 ; i < rtd->num_codecs; i++) rtd->codec_dais[i]->playback_active++; }else { - cpu_dai->capture_active++; +for (i =0 ; i < rtd->num_cpu_dai; i++)sound/soc/soc-pcm.c static bool snd_soc_dai_stream_valid(struct snd_soc_dai *dai, int stream) · line 83
Quoted 10 lines
sound/soc/soc-pcm.c +5
-2+ rtd->cpu_dais[i]->capture_active++;for (i =0 ; i < rtd->num_codecs; i++) rtd->codec_dais[i]->capture_active++; } - cpu_dai->active++; - cpu_dai->component->active++; +for (i =0 ; i < rtd->num_cpu_dai; i++) { + rtd->cpu_dais[i]->component->active++; + rtd->cpu_dais[i]->active++; + }This is becoming complicated, how many times do we need to ref-count?
Quoted 6 lines
Quoted 2 lines
Earlier we incremented cpu_dai->playback_active++ and here it is cpu_dais->component->active++
my point was whether this can be simplified to use a single counter, even before we do the change to multi-cpu?
Ok
Quoted 20 lines
Quoted 20 lines
Quoted 19 lines
sound/soc/soc-pcm.c static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, · line 269
Quoted 12 lines
sound/soc/soc-pcm.c +0
-3@@ -258,7 +266,6 @@ static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, struct snd_pcm_hw_params *params) {struct snd_soc_pcm_runtime *rtd = substream->private_data; -struct snd_soc_dai *cpu_dai = rtd->cpu_dai;unsigned int rate, channels, sample_bits, symmetry, i; rate =params_rate (params);@@ -266,41 +273,54 @@ static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, sample_bits =snd_pcm_format_physical_width (params_format (params));/* reject unmatched parameters when applying symmetry */ - symmetry = cpu_dai->driver->symmetric_rates || - rtd->dai_link->symmetric_rates;This was a logical OR, but...
sound/soc/soc-pcm.c static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream, · line 279
Quoted 4 lines
sound/soc/soc-pcm.c +4
-0+ symmetry = rtd->dai_link->symmetric_rates; + +for (i =0 ; i < rtd->num_cpu_dai; i++) + symmetry |= rtd->cpu_dais[i]->driver->symmetric_rates;this is a bitwise OR. is this ok?
Quoted 5 lines
Quoted 2 lines
I made this change on purpose and I am struggling to remember the reason :( If I dont figure that out, I will go back to logical OR..
the same change was made for multiple codecs, but it's worth rechecking.
Yes, I will check this :)
Quoted 24 lines
Quoted 24 lines
Quoted 23 lines
Quoted 10 lines
sound/soc/soc-pcm.c +7
-0@@ -422,30 +461,55 @@ static void soc_pcm_init_runtime_hw(struct snd_pcm_substream *substream) rates =snd_pcm_rate_mask_intersect (codec_stream->rates, rates); } +for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai_drv = rtd->cpu_dais[i]->driver; + +if (substream->stream ==SNDRV_PCM_STREAM_PLAYBACK ) + cpu_stream = &cpu_dai_drv->playback; +else + cpu_stream = &cpu_dai_drv->capture;sound/soc/soc-pcm.c static void soc_pcm_init_runtime_hw(struct snd_pcm_substream *substream) · line 479
Quoted 10 lines
sound/soc/soc-pcm.c +10
-0+ + cpu_chan_min =max (cpu_chan_min, + cpu_stream->channels_min); + cpu_chan_max =min (cpu_chan_max, + cpu_stream->channels_max); + +if (hw->formats) + hw->formats &= cpu_stream->formats; +else + hw->formats = cpu_stream->formats;Can you double-check the 'else' case? This doesn't seem right, you will always use the format used for the last
cpu_dai. If the formats are identical for allcpu_dais, then this can be moved outside of the loop.
Quoted 5 lines
Quoted 1 line
In the second iteration, we would always go into the if (
hw->formats) case.Sorry I don't get your comment. This test is now placed inside of a for loop, so why would the
hw->formatschange between iterations?
Ok, the else case can be hit only for the first time(if hw->formats == 0) when we enter the loop.
I get your point, will fix this :)
Quoted 21 lines
Quoted 21 lines
Quoted 20 lines
Quoted 7 lines
sound/soc/soc-pcm.c +0
-3@@ -963,11 +1070,14 @@ static int soc_pcm_hw_params(struct snd_pcm_substream *substream,if (ret <0 ) goto component_err; -/* store the parameters for each DAIs */ - cpu_dai->rate =params_rate (params); - cpu_dai->channels =params_channels (params);sound/soc/soc-pcm.c static int soc_pcm_hw_params(struct snd_pcm_substream *substream, · line 1079
Quoted 10 lines
sound/soc/soc-pcm.c +8
-2- cpu_dai->sample_bits = -snd_pcm_format_physical_width (params_format (params)); +for (i =0 ; i < rtd->num_cpu_dai; i++) { +/* store the parameters for each DAIs */ + cpu_dai = rtd->cpu_dais[i]; + cpu_dai->rate =params_rate (params); + cpu_dai->channels =params_channels (params); + cpu_dai->sample_bits = +snd_pcm_format_physical_width (params_format (params)); + }I think I've asked this before but can't recall the answer: does this mean we assume the same number of channels for each codec_dai[j] and cpu_dai[i]?
Quoted 9 lines
Quoted 4 lines
Yes, in
hw_paramswe set the same number channels on all the DAIs as that of the stream. But, as I had mentioned in my previous replies, the
machine driver would set the channel masks on the individual DAIs(like we do for SoundWire Multi link case)so the number of channels is really defined by the stream, and for each
cpu_daiandcodec_daithere is a channel mask defining what the dais should produce/consume, right?
Yes
Quoted 19 lines
Quoted 19 lines
Quoted 13 lines
sound/soc/soc-pcm.c static int soc_pcm_trigger(struct snd_pcm_substream *substream, int cmd) · line 1239
Quoted 9 lines
sound/soc/soc-pcm.c +8
-0+for (i =0 ; i < rtd->num_cpu_dai; i++) { + cpu_dai = rtd->cpu_dais[i]; +if (cpu_dai->driver->ops->trigger) { + ret = cpu_dai->driver->ops->trigger (substream, + cmd, cpu_dai); +if (ret <0 ) +return ret; + }Maybe add a comment that in the
multi_cpucase the triggers are supposed to be aligned, i.e. the first trigger starts the others - that would be consistent with the other comments on delay below.Would this be the right place to add that comment?
I am not sure if I got a reply to my question in the previous review cycle:
"Just wondering if there can be a case of multiple CPU DAIs but you would not want them to be triggered at the same time."
Quoted 7 lines
Quoted 1 line
Based on the answer to my question , we can add a comment here.
If you use the stream/multi
cpu_daiconcept you want the data to remain phase locked and triggered at the same time. If you don't care about phase, then you are handling different streams that can be handled with the existing model with a set of runtimes that deal with individualcpu_dais.
Ok, will add a comment here :)
--Shreyas
PATCH v6 3/3 ASoC: Add multiple CPU DAI support in DAPM
In reply to Shreyas NCDAPM handles DAIs during soc_dapm_stream_event() and during addition and creation of DAI widgets i.e., dapm_add_valid_dai_widget() and dapm_connect_dai_link_widgets().
Extend these functions to handle multiple cpu dai.
--- sound/soc/soc-dapm.c | 71 ++++++++++++++++++++++++++++++++++------------------ 1 file changed, 46 insertions(+), 25 deletions(-)
sound/soc/soc-dapm.c +46 -25
@@ -4113,38 +4113,57 @@ int snd_soc_dapm_link_dai_widgets(struct snd_soc_card *card)return 0 ; } -static void dapm_connect_dai_link_widgets (struct snd_soc_card *card, -struct snd_soc_pcm_runtime *rtd) +static void dapm_add_valid_dai_widget (struct snd_soc_card *card, +struct snd_soc_pcm_runtime *rtd, +struct snd_soc_dai *codec_dai, +struct snd_soc_dai *cpu_dai) { -struct snd_soc_dai *cpu_dai = rtd->cpu_dai;struct snd_soc_dapm_widget *sink, *source; -int i; -for (i =0 ; i < rtd->num_codecs; i++) { -struct snd_soc_dai *codec_dai = rtd->codec_dais[i]; +/* connect BE DAI playback if widgets are valid */ +if (codec_dai->playback_widget && cpu_dai->playback_widget) { + source = cpu_dai->playback_widget; + sink = codec_dai->playback_widget; +dev_err (rtd->dev,"connected DAI link %s:%s -> %s:%s\n" , + cpu_dai->component->name, + source->name, + codec_dai->component->name, + sink->name); + +snd_soc_dapm_add_path (&card->dapm, source, sink, +NULL ,NULL ); + } -/* connect BE DAI playback if widgets are valid */ -if (codec_dai->playback_widget && cpu_dai->playback_widget) { - source = cpu_dai->playback_widget; - sink = codec_dai->playback_widget; -dev_dbg (rtd->dev,"connected DAI link %s:%s -> %s:%s\n" , - cpu_dai->component->name, source->name, - codec_dai->component->name, sink->name); +/* connect BE DAI capture if widgets are valid */ +if (codec_dai->capture_widget && cpu_dai->capture_widget) { + source = codec_dai->capture_widget; + sink = cpu_dai->capture_widget; +dev_err (rtd->dev,"connected DAI link %s:%s -> %s:%s\n" , + codec_dai->component->name, + source->name, + cpu_dai->component->name, + sink->name); -snd_soc_dapm_add_path (&card->dapm, source, sink, +snd_soc_dapm_add_path (&card->dapm, source, sink,NULL ,NULL ); - } + } + +} -/* connect BE DAI capture if widgets are valid */ -if (codec_dai->capture_widget && cpu_dai->capture_widget) { - source = codec_dai->capture_widget; - sink = cpu_dai->capture_widget; -dev_dbg (rtd->dev,"connected DAI link %s:%s -> %s:%s\n" , - codec_dai->component->name, source->name, - cpu_dai->component->name, sink->name); +static void dapm_connect_dai_link_widgets (struct snd_soc_card *card, +struct snd_soc_pcm_runtime *rtd) +{ +struct snd_soc_dai *cpu_dai; +int i, j; -snd_soc_dapm_add_path (&card->dapm, source, sink, -NULL ,NULL ); +for (i =0 ; i < rtd->num_codecs; i++) { +struct snd_soc_dai *codec_dai = rtd->codec_dais[i]; + +for (j =0 ; j < rtd->num_cpu_dai; j++) { + cpu_dai = rtd->cpu_dais[j]; + +dapm_add_valid_dai_widget (card, rtd, + codec_dai, cpu_dai); } } }
@@ -4211,7 +4230,9 @@ static void soc_dapm_stream_event(struct snd_soc_pcm_runtime *rtd, int stream, {int i; -soc_dapm_dai_stream_event (rtd->cpu_dai, stream, event); +for (i =0 ; i < rtd->num_cpu_dai; i++) +soc_dapm_dai_stream_event (rtd->cpu_dais[i], stream, event); +for (i =0 ; i < rtd->num_codecs; i++)soc_dapm_dai_stream_event (rtd->codec_dais[i], stream, event);
Signature · 1 line
2.7.4
Re: [PATCH v6 3/3] ASoC: Add multiple CPU DAI support in DAPM
Quoted 3 lines · Shreyas NC
DAPM handles DAIs during
soc_dapm_stream_event()and during addition and creation of DAI widgets i.e.,dapm_add_valid_dai_widget()anddapm_connect_dai_link_widgets().
can you split this patch in two, one where you add dapm_add_valid_dai_widget() and the second one where you add the multi-cpu stuff? the current diff format is really hard to read with the two changes lumped together.
sound/soc/soc-dapm.c int snd_soc_dapm_link_dai_widgets(struct snd_soc_card *card) · line 4166
Quoted 8 lines
sound/soc/soc-dapm.c +8
-0+for (i =0 ; i < rtd->num_codecs; i++) { +struct snd_soc_dai *codec_dai = rtd->codec_dais[i]; + +for (j =0 ; j < rtd->num_cpu_dai; j++) { + cpu_dai = rtd->cpu_dais[j]; + +dapm_add_valid_dai_widget (card, rtd, + codec_dai, cpu_dai);
I didn't click on this earlier, but what makes you think all codec_dais are connected to all cpu_dais?
That doesn't seem quite right.
For the multi-codec case, all the codec_dais hang from a single cpu_dai. it's a stretch for me to have a full M:N connectivity. And that's clearly not the case for SoundWire stream in the multi-link case. Can't we use the dai_link information here to only connect cpu_ and codec_dais that are related?
Quoted 14 lines
} } }sound/soc/soc-dapm.c +3
-1@@ -4211,7 +4230,9 @@ static void soc_dapm_stream_event(struct snd_soc_pcm_runtime *rtd, int stream, {int i; -soc_dapm_dai_stream_event (rtd->cpu_dai, stream, event); +for (i =0 ; i < rtd->num_cpu_dai; i++) +soc_dapm_dai_stream_event (rtd->cpu_dais[i], stream, event); +for (i =0 ; i < rtd->num_codecs; i++)soc_dapm_dai_stream_event (rtd->codec_dais[i], stream, event);
Re: [PATCH v6 3/3] ASoC: Add multiple CPU DAI support in DAPM
Quoted 10 lines · Pierre-Louis Bossart
Quoted 3 lines · Shreyas NC
DAPM handles DAIs during
soc_dapm_stream_event()and during addition and creation of DAI widgets i.e.,dapm_add_valid_dai_widget()anddapm_connect_dai_link_widgets().can you split this patch in two, one where you add
dapm_add_valid_dai_widget()and the second one where you add the multi-cpu stuff? the current diff format is really hard to read with the two changes lumped together.
As I had replied earlier, the change is really moving the same code from one function to a new one. So, a patch split would mean we would have the same duplicated code in one patch which surely is not desirable.
I just realized that in my earlier reply I had excluded the list and replied only to you :)
Quoted 12 lines
sound/soc/soc-dapm.c int snd_soc_dapm_link_dai_widgets(struct snd_soc_card *card) · line 4166
Quoted 9 lines
sound/soc/soc-dapm.c +8
-0+for (i =0 ; i < rtd->num_codecs; i++) { +struct snd_soc_dai *codec_dai = rtd->codec_dais[i]; + +for (j =0 ; j < rtd->num_cpu_dai; j++) { + cpu_dai = rtd->cpu_dais[j]; + +dapm_add_valid_dai_widget (card, rtd, + codec_dai, cpu_dai);I didn't click on this earlier, but what makes you think all
codec_daisare connected to allcpu_dais?
Yes, there need not be a M:N connectivity. But, how do you find that out ?
Quoted 4 lines
That doesn't seem quite right.
For the multi-codec case, all thecodec_daishang from a singlecpu_dai. it's a stretch for me to have a full M:N connectivity. And that's clearly not the case for SoundWire stream in the multi-link case.
I mostly do not disagree with you here..
Quoted 2 lines
Can't we use the
dai_linkinformation here to only connectcpu_andcodec_daisthat are related?
Which DAI Link information are you referring to here ?
Other than the machine driver which sets the audio route, I am unable to figure out how we will find out the connected cpu_dai and codec_dais at ASoC core level.
May be I am missing something :(
--Shreyas
Re: [PATCH v6 3/3] ASoC: Add multiple CPU DAI support in DAPM
Quoted 7 lines · Shreyas NC
Quoted 6 lines · Pierre-Louis Bossart
Quoted 3 lines · Shreyas NC
DAPM handles DAIs during
soc_dapm_stream_event()and during addition and creation of DAI widgets i.e.,dapm_add_valid_dai_widget()anddapm_connect_dai_link_widgets().
Quoted 8 lines
Quoted 4 lines
can you split this patch in two, one where you add
dapm_add_valid_dai_widget()and the second one where you add the multi-cpu stuff? the current diff format is really hard to read with the two changes lumped together.As I had replied earlier, the change is really moving the same code from one function to a new one. So, a patch split would mean we would have the same duplicated code in one patch which surely is not desirable.
I don't understand your answer. It's fine to have a small preparation patch that just moves one piece of code to another function, and they change how that function is called.
Quoted 28 lines
I just realized that in my earlier reply I had excluded the list and replied only to you :)
Quoted 12 lines
sound/soc/soc-dapm.c int snd_soc_dapm_link_dai_widgets(struct snd_soc_card *card) · line 4166
Quoted 9 lines
sound/soc/soc-dapm.c +8
-0+for (i =0 ; i < rtd->num_codecs; i++) { +struct snd_soc_dai *codec_dai = rtd->codec_dais[i]; + +for (j =0 ; j < rtd->num_cpu_dai; j++) { + cpu_dai = rtd->cpu_dais[j]; + +dapm_add_valid_dai_widget (card, rtd, + codec_dai, cpu_dai);I didn't click on this earlier, but what makes you think all
codec_daisare connected to allcpu_dais?Yes, there need not be a M:N connectivity. But, how do you find that out ?
Quoted 4 lines
That doesn't seem quite right.
For the multi-codec case, all thecodec_daishang from a singlecpu_dai. it's a stretch for me to have a full M:N connectivity. And that's clearly not the case for SoundWire stream in the multi-link case.I mostly do not disagree with you here..
Quoted 2 lines
Can't we use the
dai_linkinformation here to only connectcpu_andcodec_daisthat are related?
Quoted 6 lines
Which DAI Link information are you referring to here ?
Other than the machine driver which sets the audio route, I am unable to figure out how we will find out the connectedcpu_daiandcodec_daisat ASoC core level.May be I am missing something :(
How is it different from the multi-codec support? You must have a description somewhere that tells you how the cpu_dai is connected to various codec_dais.
Maybe we should start with the examples you provided for Soundwire and describe how the dailinks would be represented.
With the M:N connectivity you'd end-up having spurious events with non-existent connections, it's not necessarily fatal but certainly not elegant and may or may not work depending on state management in codec drivers.
Re: [PATCH v6 3/3] ASoC: Add multiple CPU DAI support in DAPM
Quoted 8 lines
Quoted 8 lines
Quoted 7 lines
Quoted 3 lines
DAPM handles DAIs during
soc_dapm_stream_event()and during addition and creation of DAI widgets i.e.,dapm_add_valid_dai_widget()anddapm_connect_dai_link_widgets().can you split this patch in two, one where you add
dapm_add_valid_dai_widget()and the second one where you add the multi-cpu stuff? the current diff format is really hard to read with the two changes lumped together.
Quoted 8 lines
Quoted 3 lines
As I had replied earlier, the change is really moving the same code from one function to a new one. So, a patch split would mean we would have the same duplicated code in one patch which surely is not desirable.
I don't understand your answer. It's fine to have a small preparation patch that just moves one piece of code to another function, and they change how that function is called.
Ok, I will give it a try :)
Quoted 33 lines
Quoted 33 lines
I just realized that in my earlier reply I had excluded the list and replied only to you :)
Quoted 12 lines
sound/soc/soc-dapm.c int snd_soc_dapm_link_dai_widgets(struct snd_soc_card *card) · line 4166
Quoted 9 lines
sound/soc/soc-dapm.c +8
-0+for (i =0 ; i < rtd->num_codecs; i++) { +struct snd_soc_dai *codec_dai = rtd->codec_dais[i]; + +for (j =0 ; j < rtd->num_cpu_dai; j++) { + cpu_dai = rtd->cpu_dais[j]; + +dapm_add_valid_dai_widget (card, rtd, + codec_dai, cpu_dai);I didn't click on this earlier, but what makes you think all
codec_daisare connected to allcpu_dais?Yes, there need not be a M:N connectivity. But, how do you find that out ?
Quoted 4 lines
That doesn't seem quite right.
For the multi-codec case, all thecodec_daishang from a singlecpu_dai. it's a stretch for me to have a full M:N connectivity. And that's clearly not the case for SoundWire stream in the multi-link case.I mostly do not disagree with you here..
Quoted 2 lines
Can't we use the
dai_linkinformation here to only connectcpu_andcodec_daisthat are related?Which DAI Link information are you referring to here ?
Other than the machine driver which sets the audio route, I am unable to figure out how we will find out the connectedcpu_daiandcodec_daisat ASoC core level.
Quoted 6 lines
Quoted 1 line
May be I am missing something :(
How is it different from the multi-codec support? You must have a description somewhere that tells you how the
cpu_daiis connected to variouscodec_dais.
No, there is no such description that I could find.
My reference was skl_nau88l25_ssm4567.c machine driver:
/* Back End DAI links */
{
/* SSP0 - Codec */
.name = "SSP0-Codec",
.id = 0,
.cpu_dai_name = "SSP0 Pin",
.platform_name = "0000:00:1f.3",
.no_pcm = 1,
.codecs = ssm4567_codec_components,
.num_codecs = ARRAY_SIZE(ssm4567_codec_components),
..
},The only way I could infer the connection was from DAPM route:
{
/* CODEC BE connections */
{ "Left Playback", NULL, "ssp0 Tx"},
{ "Right Playback", NULL, "ssp0 Tx"},
{ "ssp0 Tx", NULL, "codec0_out"},
}Quoted 3 lines
Maybe we should start with the examples you provided for Soundwire and describe how the dailinks would be represented.
Ok , sure. Let me describe a case where we would want to play a 2 ch stream and we have two Masters "int-sdw.1" and "int-sdw.2"
So, DAI Link would look this way:
Here I specify the multi CPU DAIs
static struct snd_soc_dai_link_component sdw_multi_cpu_comp[] = {
{ /* Left */
.name = "int-sdw.1",
.dai_name = "SDW1 Pin0",
},
{ /* Right */
.name = "int-sdw.2",
.dai_name = "SDW2 Pin0",
},
};static struct snd_soc_codec_conf rt700_codec_conf[] = {
{
.dev_name = "sdw:1:25d:700:0:0",
.name_prefix = "Left",
},
{
.dev_name = "sdw:2:25d:701:0:0",
.name_prefix = "Right",
},
};And the DAI Link as:
struct snd_soc_dai_link cnl_rt700_msic_dailink[] = {
{
...
{
.name = "SDW0-Codec",
.cpu_dai = sdw_multi_cpu_comp,
.num_cpu_dai = ARRAY_SIZE(sdw_multi_cpu_comp),
.platform_name = "0000:00:1f.3",
.codecs = sdw_multi_codec_comp,
.num_codecs = ARRAY_SIZE(sdw_multi_codec_comp),
..
},
}And it is in the DAPM route that I would specify :
{
...
{ "Left DP1 Playback", NULL, "SDW1 Tx0"},
{ "SDW1 Tx0", NULL, "codec0_out"},
{ "Right DP1 Playback", NULL, "SDW2 Tx0"},
{ "SDW2 Tx0", NULL, "codec0_out"},
...
}Quoted 5 lines
With the M:N connectivity you'd end-up having spurious events with non-existent connections, it's not necessarily fatal but certainly not elegant and may or may not work depending on state management in codec drivers.
What are the events that you are referring to here?
The reason I ask that is because my understanding is that we will get these events only for those widgets marked connected by DAPM after parsing the DAPM route map.
I will check further if you can let me know of the events you are suspecting :)
Sorry for the longish mail. I thought it would be better to put in the details rather than go back and forth over them :)
--Shreyas