| Message ID | 87frvj8g2v.wl-kuninori.morimoto.gx@renesas.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | ASoC: Replace dpcm_playback/capture to playback/capture_assertion | expand |
> (X) part is for DPCM, and it checks playback/capture availability > if dai_link has dpcm_playback/capture flag (a)(b). > This availability check should be available not only for DPCM, but for > all connections. But it is not mandatory option. Let's name it as > assertion. I don't follow the 'not mandatory option'. Why not make these 'assertions' mandatory? What happens in case the the option is not present? > In case of having assertion flag, non specific side will be disabled. Not following the wording, multiple negatives and not clear on what 'side' refers to (direction or DPCM/non-DPCM). > For example if it has playback_assertion but not have capture_assertion, > capture will be force disabled. > > dpcm_playback -> playabck_assertion = 1 > > dpcm_capture -> capture_assertion = 1 > > playback_only -> playback_assertion = 1 > capture_assertion = 0 > > capture_only -> playback_assertion = 0 > capture_assertion = 1 > > By expanding this assertion to all connections, we can use same code > for all connections, this means code can be more cleanup. I see a contradiction between "expanding the assertion to all connections" and "not mandatory". > Current validation check on DPCM ignores Codec DAI, but there is no > reason to do it. We should check both CPU/Codec in all connection. "there's no reason to do so" ? > This patch expands validation check to all cases. > > [CPU/xxxx]-[yyyy/Codec] > ***** > > In many case (not all case), above [xxxx][yyyy] part are "dummy" DAI > which is always valid for both playback/capture. > But unfortunately DPCM BE Codec (**** part) had been no validation > check before, and some Codec driver doesn't have necessary settings for > it. This means all cases validation check breaks compatibility on some > vender's drivers. Thus this patch temporary uses dummy DAI at BPCM BE vendor > Codec part, and avoid compatibility error. But it should be removed in > the future. > > Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com> > --- > include/sound/soc.h | 9 +++ > sound/soc/soc-pcm.c | 143 +++++++++++++++++++++++++------------------- > 2 files changed, 92 insertions(+), 60 deletions(-) > > diff --git a/include/sound/soc.h b/include/sound/soc.h > index 0376f7e4c15d..e604d74f6e33 100644 > --- a/include/sound/soc.h > +++ b/include/sound/soc.h > @@ -809,6 +809,15 @@ struct snd_soc_dai_link { > unsigned int dpcm_capture:1; > unsigned int dpcm_playback:1; > > + /* > + * Capture / Playback support assertion. Having assertion flag is not mandatory. > + * In case of having assertion flag, non specific side will be disabled. again the 'not mandatory' and 'non specific side will be disabled' are confusing. > + /* > + * Assertion check > + * > + * playback_assertion = 0 No assertion check. > + * capture_assertion = 0 ASoC will use detected playback/capture as-is. > + * No playback, No capture will be error. did you mean "this combination will be handled as an error" ? It's probably best to have a different presentation, to avoid confusions. Using multiple lines without a separator isn't great. Suggested example: playback_assertion = 0 capture_assertion = 0 this combination will be handled as an error playback_assertion = 1 capture_assertion = 0 the DAIs in this dailink must support playback. ASoC will disable capture. In other words "playback_only" > + * > + * playback_assertion = 1 DAI must playback available. ASoC will disable capture. > + * capture_assertion = 0 In other words "playback_only" > + * > + * playback_assertion = 0 DAI must capture available. ASoC will disable playback. > + * capture_assertion = 1 In other words "capture_only" > + * > + * playback_assertion = 1 DAI must both playback/capture available. > + * capture_assertion = 1 nit-pick: the 'must X available' does not read well, 'must support X' is probably what you meant. > + */ > + if (dai_link->playback_assertion) { > + if (!has_playback) { > + dev_err(rtd->dev, "%s playback assertion check error\n", dai_link->stream_name); "invalid playback_assertion" ? > + return -EINVAL; > + } > + /* makes it plyaback only */ typo: playback > + if (!dai_link->capture_assertion) > + has_capture = 0; > + } > + if (dai_link->capture_assertion) { > + if (!has_capture) { > + dev_err(rtd->dev, "%s capture assertion check error\n", dai_link->stream_name); > + return -EINVAL; > + } > + /* makes it capture only */ > + if (!dai_link->playback_assertion) > + has_playback = 0; > + } we probably want a dev_ log when the has_playback/capture is overridden? > > + /* > + * Detect Mismatch > + */ > if (!has_playback && !has_capture) { > dev_err(rtd->dev, "substream %s has no playback, no capture\n", > dai_link->stream_name); > - > return -EINVAL; > } >
Hi Pierre-Louis > > (X) part is for DPCM, and it checks playback/capture availability > > if dai_link has dpcm_playback/capture flag (a)(b). > > This availability check should be available not only for DPCM, but for > > all connections. But it is not mandatory option. Let's name it as > > assertion. > > I don't follow the 'not mandatory option'. Why not make these > 'assertions' mandatory? What happens in case the the option is not present? The big reason why "assertion flag" is not mandatory is that non-DPCM doesn't have such flag before. I can't add such flags to all of non-DPCM, because I don't know which direction (playback/capture) is available on each DAIs. > > In case of having assertion flag, non specific side will be disabled. > > Not following the wording, multiple negatives and not clear on what > 'side' refers to (direction or DPCM/non-DPCM). How about this ? If either playback or capture assertion flag was presented, not presented direction will be disabled by ASoC even if it was available. > I see a contradiction between "expanding the assertion to all > connections" and "not mandatory". (snip) > again the 'not mandatory' and 'non specific side will be disabled' are > confusing. (snip) > > + /* > > + * Assertion check > > + * > > + * playback_assertion = 0 No assertion check. > > + * capture_assertion = 0 ASoC will use detected playback/capture as-is. > > + * No playback, No capture will be error. > > did you mean "this combination will be handled as an error" ? Hmm... is it word selection issue ?? is "must_support" better ? (playback_xxx, capture_xxx) (0, 0) : Both are not must item. available direction is used as-is. But it will be error if nothing was available. (1, 0) : DAI must support selected direction. (0, 1) Not selected direction will be disabled even though it was available (1, 1) : Both must be supported. > It's probably best to have a different presentation, to avoid > confusions. Using multiple lines without a separator isn't great. > > Suggested example: > > playback_assertion = 0 capture_assertion = 0 > this combination will be handled as an error > > playback_assertion = 1 capture_assertion = 0 > the DAIs in this dailink must support playback. > ASoC will disable capture. > In other words "playback_only" Yeah, I agree > > + * playback_assertion = 1 DAI must playback available. ASoC will disable capture. > > + * capture_assertion = 0 In other words "playback_only" > > + * > > + * playback_assertion = 0 DAI must capture available. ASoC will disable playback. > > + * capture_assertion = 1 In other words "capture_only" > > + * > > + * playback_assertion = 1 DAI must both playback/capture available. > > + * capture_assertion = 1 > > nit-pick: the 'must X available' does not read well, 'must support X' is > probably what you meant. Thanks. will fix in v4 > > + if (dai_link->capture_assertion) { > > + if (!has_capture) { > > + dev_err(rtd->dev, "%s capture assertion check error\n", dai_link->stream_name); > > + return -EINVAL; > > + } > > + /* makes it capture only */ > > + if (!dai_link->playback_assertion) > > + has_playback = 0; > > + } > > we probably want a dev_ log when the has_playback/capture is overridden? OK, will do in v4 Thank you for your help !! Best regards --- Renesas Electronics Ph.D. Kuninori Morimoto
>>> (X) part is for DPCM, and it checks playback/capture availability >>> if dai_link has dpcm_playback/capture flag (a)(b). >>> This availability check should be available not only for DPCM, but for >>> all connections. But it is not mandatory option. Let's name it as >>> assertion. >> >> I don't follow the 'not mandatory option'. Why not make these >> 'assertions' mandatory? What happens in case the the option is not present? > > The big reason why "assertion flag" is not mandatory is that non-DPCM doesn't > have such flag before. I can't add such flags to all of non-DPCM, > because I don't know which direction (playback/capture) is available on > each DAIs. Your explanation seems to contradict the sentence above "This availability check should be available not only for DPCM, but for all connections." Can we actually do this 'availability check' for non-DPCM connections. >>> In case of having assertion flag, non specific side will be disabled. >> >> Not following the wording, multiple negatives and not clear on what >> 'side' refers to (direction or DPCM/non-DPCM). > > How about this ? > > If either playback or capture assertion flag was presented, > not presented direction will be disabled by ASoC even if > it was available. Did you mean " The playback (resp. capture) direction will be disabled by ASoC if the playback_assertion (resp. capture) flag is false - even if this direction was available at the DAI level " > (playback_xxx, capture_xxx) > > (0, 0) : Both are not must item. available direction is used as-is. > But it will be error if nothing was available. That new wording makes me even more confused. What does 'available' refer to and at which level is this? This seems also to contradict the definitions above, "available direction is used as-is" is not aligned with "not presented direction will be disabled by ASoC even if it was available". > (1, 0) : DAI must support selected direction. > (0, 1) Not selected direction will be disabled even though it was available > > (1, 1) : Both must be supported.
Hi Pierre-Louis Cc Mark > Your explanation seems to contradict the sentence above "This > availability check should be available not only for DPCM, but for all > connections." > > Can we actually do this 'availability check' for non-DPCM connections. > > > How about this ? > > > > If either playback or capture assertion flag was presented, > > not presented direction will be disabled by ASoC even if > > it was available. > > Did you mean > > " > The playback (resp. capture) direction will be disabled by ASoC if the > playback_assertion (resp. capture) flag is false - even if this > direction was available at the DAI level > " > > (0, 0) : Both are not must item. available direction is used as-is. > > But it will be error if nothing was available. > > That new wording makes me even more confused. What does 'available' > refer to and at which level is this? > > This seems also to contradict the definitions above, "available > direction is used as-is" is not aligned with "not presented direction > will be disabled by ASoC even if it was available". It is complicated by the attempt to merge dpcm_xxx and xxx_only flags. And I noticed that my one of other attemption was not indicated. Let's cleanup what does this patch-set want to do I still wondering why dpcm_xxx flag itself is needed. (A) Before, it checks channels_min for DPCM, same as current non-DPCM. This is very clear for me. Let's name this as "validation check" if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { if (cpu_dai->driver->playback.channels_min) playback = 1; if (cpu_dai->driver->capture.channels_min) capture = 1; (B) commit 1e9de42f4324b91ce2e9da0939cab8fc6ae93893 ("Explicitly set BE DAI link supported stream directions") force use to dpcm_xxx flag if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { playback = rtd->dai_link->dpcm_playback; capture = rtd->dai_link->dpcm_capture; (C) 9b5db059366ae2087e07892b5fc108f81f4ec189 ("ASoC: soc-pcm: dpcm: Only allow playback/capture if supported") checks channels_min (= validation check) again if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { cpu_dai = asoc_rtd_to_cpu(rtd, 0); ... playback = rtd->dai_link->dpcm_playback && snd_soc_dai_stream_valid(cpu_dai, SNDRV_PCM_STREAM_PLAYBACK); capture = rtd->dai_link->dpcm_capture && snd_soc_dai_stream_valid(cpu_dai, SNDRV_PCM_STREAM_CAPTURE); (D) b73287f0b0745961b14e5ebcce92cc8ed24d4d52 ("ASoC: soc-pcm: dpcm: fix playback/capture checks") expanded it to multi connection. So, I would say nothing has changed, but become more complicated. Or if (B) added dpcm_xxx as "option flag", it was understandable for me. like this if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { playback = (cpu_dai->driver->playback.channels_min > 0) || rtd->dai_link->dpcm_playback; capture = (cpu_dai->driver->capture.channels_min > 0) || rtd->dai_link->dpcm_capture; So my opinion is this dpcm_xxx is unnecessary flag that only complicate matters. I guess almost all Card don't need this flag, this means "validation check" only is veryenough, same as current non-DPCM. But because of these history, dpcm_xxx flag have been used as "passage permit" or "gate way". It doesn't try to "validation check" if dpcm_xxx flag was not set. This is the reason why I try to merge dpcm_xxx and xxx_only flag. These are doing the same things with dirrerent flags, IMO. OTOH, some Card want to detect error if expected direction (playback/capture) was not valid. I guess this is your commitment (?). So, let's keep xxx_only flag as-is, and use dpcm_xxx as "available_check". I'm not sure what is the good naming, but for example "playback_available_check" flag means "owner is expecting playback is valid/available, and want to receive error if not". I'm not sure how many owner want this flag, thus I think "option flag" is very enough (= not mandatory, as I mentioned in the patch-set). If we makes these checks generalize, For DPCM, (for example new DPCM) it can remove/ignore "available_check" flag if it don't need, same as current non-DPCM. And for non-DPCM, it can use "available_check" if needed, same as current DPCM. What do you think ? what is your opinion ? Thank you for your help !! Best regards --- Renesas Electronics Ph.D. Kuninori Morimoto
Hi Morimoto-san, > I still wondering why dpcm_xxx flag itself is needed. That's an excellent question indeed. And since you started a historical review, we can get a lot of information. > (A) Before, it checks channels_min for DPCM, same as current non-DPCM. > This is very clear for me. Let's name this as "validation check" > > if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { > if (cpu_dai->driver->playback.channels_min) > playback = 1; > if (cpu_dai->driver->capture.channels_min) > capture = 1; > > (B) commit 1e9de42f4324b91ce2e9da0939cab8fc6ae93893 > ("Explicitly set BE DAI link supported stream directions") force use to > dpcm_xxx flag > > if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { > playback = rtd->dai_link->dpcm_playback; > capture = rtd->dai_link->dpcm_capture; The reason for this (B) addition is very clear from the commit message " Some BE DAIs can be "dummy" (when the DSP is controlling the DAI) and as such wont have set a minimum number of playback or capture channels required for BE DAI registration (to establish supported stream directions). " So (B) introduced these dpcm_xxx flags as override mechanisms, where the dailink information matters more than the dai information. > (C) 9b5db059366ae2087e07892b5fc108f81f4ec189 > ("ASoC: soc-pcm: dpcm: Only allow playback/capture if supported") > checks channels_min (= validation check) again > > if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { > cpu_dai = asoc_rtd_to_cpu(rtd, 0); > ... > playback = rtd->dai_link->dpcm_playback && > snd_soc_dai_stream_valid(cpu_dai, SNDRV_PCM_STREAM_PLAYBACK); > capture = rtd->dai_link->dpcm_capture && > snd_soc_dai_stream_valid(cpu_dai, SNDRV_PCM_STREAM_CAPTURE); It helps to look at the commit message in detail: " Normally the dpcm_playback/capture parameter should match the capabilities of the CPU DAI. However, there is no way to set that parameter from the device tree (e.g. with simple-audio-card or qcom sound cards). dpcm_playback/capture are always both set to 1. " This is where the direction changed from "dpcm_xxx" being override of DAI capabilities to "dpcm_xxx" being required to match DAI capabilities, because some machine drivers incorrectly did an override that made no sense... (C) is essentially (A) && (B) Clearly there's a contradiction. If (C) was the correct solution, then it would revert the direction in (A) and report an error for dummy dais. It's been my question from the beginning of this thread, when the direction information can come from 2 sources, which one do we trust? > (D) b73287f0b0745961b14e5ebcce92cc8ed24d4d52 > ("ASoC: soc-pcm: dpcm: fix playback/capture checks") expanded it to > multi connection. You really want to look at (E) 4f8721542f7b75954bfad98c51aa59d683d35b50 ("ASoC: core: use less strict tests for dailink capabilities") " This patch modifies the snd_soc_dai_link_set_capabilities() helper so that the dpcm_playback (resp. dpcm_capture) dailink capabilities are set if at least one dai supports playback (resp. capture). Likewise the checks are modified so that an error is reported only when dpcm_playback (resp. dpcm_capture) is set but none of the CPU DAIs support playback (resp. capture). " This one has not fundamentally changed the reasons why (C) was introduced, the requirement is that dpcm_xxx be aligned with at least ONE DAI capability. It's still not clear how dummy-dais would be handled since the number of channels may or may not be set for those dummy-dais. > So, I would say nothing has changed, but become more complicated. It's not really become more complicated, the open is which of (B) or (C) are correct. > Or if (B) added dpcm_xxx as "option flag", it was understandable for me. > like this > > if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { > playback = (cpu_dai->driver->playback.channels_min > 0) || > rtd->dai_link->dpcm_playback; > capture = (cpu_dai->driver->capture.channels_min > 0) || > rtd->dai_link->dpcm_capture; That would essentially revert (C), since the direction would ignore the actual capabilities of the DAI. IMHO, we should really try with this revert and go back to the initial definition of (A), where dpcm_xxx is an optional escape mechanism to allow machine drivers to set the direction. I would guess that would impact mostly DT/simple-card users and Qualcomm, if the commit message of (C) is still relevant. Then we can discuss about merging code and removing flags. For now we have different directions/opinions on something that's 10 years old, last modified 4 years ago. We will break lots of eggs if we don't first agree on what works and what doesn't. This 23-patch code merge/simplification is premature at this point IMHO, we should first land in a state where everyone is happy with how dpcm_xxx flags need to be handled. We can merge dpcm_xxx and xxx_only flags later when we understand what they are supposed to do... And now I need a coffee refill :-) Regards, -Pierre
Hi Pierre-Louis Thank you for your feedback > > (B) commit 1e9de42f4324b91ce2e9da0939cab8fc6ae93893 > > ("Explicitly set BE DAI link supported stream directions") force use to > > dpcm_xxx flag > > > > if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { > > playback = rtd->dai_link->dpcm_playback; > > capture = rtd->dai_link->dpcm_capture; > > The reason for this (B) addition is very clear from the commit message > > " > Some BE DAIs can be "dummy" (when the DSP is controlling the DAI) and as > such wont have set a minimum number of playback or capture channels > required for BE DAI registration (to establish supported stream directions). > " I'm still not yet 100% understand around this "dummy" DAI, but is it *not* soc-utils.c :: dummy_dai, but some original dummy DAI is used somewhere ? I know ${LINUX}/sound/soc/codecs/hda.c :: card_binder_dai is one of the DAI which is used but doesn't have channels_min. I think it is used as BE "Codec", but code is checking "CPU" side. Do you know what does this "BE dummy DAI" specifically means here? > > Or if (B) added dpcm_xxx as "option flag", it was understandable for me. > > like this > > > > if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { > > playback = (cpu_dai->driver->playback.channels_min > 0) || > > rtd->dai_link->dpcm_playback; > > capture = (cpu_dai->driver->capture.channels_min > 0) || > > rtd->dai_link->dpcm_capture; > > That would essentially revert (C), since the direction would ignore the > actual capabilities of the DAI. > > IMHO, we should really try with this revert and go back to the initial > definition of (A), where dpcm_xxx is an optional escape mechanism to > allow machine drivers to set the direction. I would guess that would > impact mostly DT/simple-card users and Qualcomm, if the commit message > of (C) is still relevant. I can agree that above code makes dpcm_xxx flag option, and allow to machine drivers to set direction without thinking actual DAI capabilities. I think it is effective if it was around (C) timing. (A) : checked CPU capabilities (B) : uses dpcm_xxx flag (C) : checks both dpcm_xxx and capabilities ... But *current* ASoC is checking both "actual capabilities" and "dpcm_xxx" flag in the same time, and have no problems today. Around (A)-(B) timing, some DAI had issue (= channels_min was not set). Let's name it as "no_chan_DAI". It should be CPU DAI and used as BE if my understanding was correct. Because "no_chan_DAI" exist, (B) was added. But after that (C) was added, and it checks channels_min again. It continues today. This means "no_chan_DAI" today have channels_min, otherwise it can't work today. If my above understanding were all correct, do we still need dpcm_xxx ? because "no_chan_DAI" is no longer exist. Just remove dpcm_xxx seems no problem, I guess... > Then we can discuss about merging code and removing flags. For now we > have different directions/opinions on something that's 10 years old, > last modified 4 years ago. We will break lots of eggs if we don't first > agree on what works and what doesn't. > > This 23-patch code merge/simplification is premature at this point IMHO, > we should first land in a state where everyone is happy with how > dpcm_xxx flags need to be handled. We can merge dpcm_xxx and xxx_only > flags later when we understand what they are supposed to do... Thank you for your help. The problem becoming more clear. Let's focus to "revert C code" or "remove dpcm_xxx" first. > And now I need a coffee refill :-) Enjoy :) Thank you for your help !! Best regards --- Renesas Electronics Ph.D. Kuninori Morimoto
Hi Pierre-Louis, Mark Because Japanese will dive into long vacation since next week, I want to post mail before that. I will back at 7th May. > > > (B) commit 1e9de42f4324b91ce2e9da0939cab8fc6ae93893 > > > ("Explicitly set BE DAI link supported stream directions") force use to > > > dpcm_xxx flag > > > > > > if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { > > > playback = rtd->dai_link->dpcm_playback; > > > capture = rtd->dai_link->dpcm_capture; > > > > The reason for this (B) addition is very clear from the commit message > > > > " > > Some BE DAIs can be "dummy" (when the DSP is controlling the DAI) and as > > such wont have set a minimum number of playback or capture channels > > required for BE DAI registration (to establish supported stream directions). > > " > > I'm still not yet 100% understand around this "dummy" DAI, but is it > *not* soc-utils.c :: dummy_dai, but some original dummy DAI is used > somewhere ? > > I know ${LINUX}/sound/soc/codecs/hda.c :: card_binder_dai is one of > the DAI which is used but doesn't have channels_min. > I think it is used as BE "Codec", but code is checking "CPU" side. > > Do you know what does this "BE dummy DAI" specifically means here? (A) : checked CPU capabilities (B) : uses dpcm_xxx flag (C) : checks both dpcm_xxx and capabilities ... In my understanding, in summary, this dpcm_xxx flag was added to rescue dummy DAI which is used on DCPM BE as CPU at (B), because some of them might not have channels_min (This "dummy DAI" is not same as soc-utils's dummy DAI). Let's name it as "no_chan_DAI" here. In this patch, it was added as "mandatory flag", not "option flag", thus all DPCM needed to use this dpcm_xxx flag. After that (C) was added, but it was contradiction, because it checks both dpcm_xxx and channels_min. If my understanding was correct, original "no_chan_DAI" was supposed to stop working, because it doesn't have channels_min. But there is no such report after (C), during this 4 years. We don't know which DAI is the "no_chan_DAI" (?) Possibilities are as follows - No one is using "no_chan_DAI" - "no_chan_DAI" is no longer exist : it was removed ? - "no_chan_DAI" is no longer exist : it has channels_min ? If my expectation was correct, we don't need dpcm_xxx anymore. But because we have been used dpcm_xxx for 10 years since (B), I understand to feel anxious to suddenly remove dpcm_xxx. I think it should be removed anyway, but want to have grace time ? If so, the idea is that we can use it as "option flag" instead of "mandatory flag" for a while, like below if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { playback = (cpu_dai->driver->playback.channels_min > 0) || rtd->dai_link->dpcm_playback; capture = (cpu_dai->driver->capture.channels_min > 0) || rtd->dai_link->dpcm_capture; * maybe we want to indicate message like "place re-check the flag and remove it" via dev_info() if dpcm_xxx flag was used ? I think +2 kernel version or so is enough for grace time ? After that, we can remove dpcm_xxx flag. When we consider it very detail, above code can't 100% keep compatibility if the user have been used this dpcm_xxx flag to limit availability, for example in case of DAI can use both playback/capture, but it had dpcm_playback flag only. But it can use playback_only flag, instead. But it is very difficult to find such DAI. Each user need to check. I personally think that remove dpcm_xxx directly is no ploblem, but what do you think ? I'm happy to hear any opinion, and happy to create grace time code if someone want, but if there was no comment during Japanese long vacation, I will create patch to remove dpcm_xxx directly. BTW, I would like to know detail things around this topic. I'm happy if someone knows it. * Why dummy DAI doesn't/can't have channels_min ? * Why it checks CPU side channels_min only when DPCM ? I think it should check both CPU and Codec. I could understand if it checks FE:CPU and BE:Codec (it is assuming other side was dummy), but both (FE/BE) check CPU side only... Thank you for your help !! Best regards --- Renesas Electronics Ph.D. Kuninori Morimoto
> Because Japanese will dive into long vacation since next week, > I want to post mail before that. I will back at 7th May. Enjoy! >>>> (B) commit 1e9de42f4324b91ce2e9da0939cab8fc6ae93893 >>>> ("Explicitly set BE DAI link supported stream directions") force use to >>>> dpcm_xxx flag >>>> >>>> if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { >>>> playback = rtd->dai_link->dpcm_playback; >>>> capture = rtd->dai_link->dpcm_capture; >>> >>> The reason for this (B) addition is very clear from the commit message >>> >>> " >>> Some BE DAIs can be "dummy" (when the DSP is controlling the DAI) and as >>> such wont have set a minimum number of playback or capture channels >>> required for BE DAI registration (to establish supported stream directions). >>> " >> >> I'm still not yet 100% understand around this "dummy" DAI, but is it >> *not* soc-utils.c :: dummy_dai, but some original dummy DAI is used >> somewhere ? >> >> I know ${LINUX}/sound/soc/codecs/hda.c :: card_binder_dai is one of >> the DAI which is used but doesn't have channels_min. >> I think it is used as BE "Codec", but code is checking "CPU" side. >> >> Do you know what does this "BE dummy DAI" specifically means here? > > (A) : checked CPU capabilities > (B) : uses dpcm_xxx flag > (C) : checks both dpcm_xxx and capabilities > ... > > In my understanding, in summary, this dpcm_xxx flag was added to rescue > dummy DAI which is used on DCPM BE as CPU at (B), because some of them > might not have channels_min (This "dummy DAI" is not same as soc-utils's > dummy DAI). Let's name it as "no_chan_DAI" here. > In this patch, it was added as "mandatory flag", not "option flag", > thus all DPCM needed to use this dpcm_xxx flag. > > After that (C) was added, but it was contradiction, because > it checks both dpcm_xxx and channels_min. > If my understanding was correct, original "no_chan_DAI" was supposed to > stop working, because it doesn't have channels_min. But there is no > such report after (C), during this 4 years. > We don't know which DAI is the "no_chan_DAI" (?) > > Possibilities are as follows > - No one is using "no_chan_DAI" > - "no_chan_DAI" is no longer exist : it was removed ? > - "no_chan_DAI" is no longer exist : it has channels_min ? > > If my expectation was correct, we don't need dpcm_xxx anymore. I agree with your analysis. We don't have a clear memory/understanding of which "no_chan_DAI" platforms (B) was supposed to handle, and why no one reported them as broken by (C). > But because we have been used dpcm_xxx for 10 years since (B), > I understand to feel anxious to suddenly remove dpcm_xxx. Indeed we err on the side of paranoia with such changes! > I think it should be removed anyway, but want to have grace time ? > If so, the idea is that we can use it as "option flag" instead of > "mandatory flag" for a while, like below > > if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { > playback = (cpu_dai->driver->playback.channels_min > 0) || > rtd->dai_link->dpcm_playback; > capture = (cpu_dai->driver->capture.channels_min > 0) || > rtd->dai_link->dpcm_capture; > > * maybe we want to indicate message like "place re-check the flag and > remove it" via dev_info() if dpcm_xxx flag was used ? > > I think +2 kernel version or so is enough for grace time ? > After that, we can remove dpcm_xxx flag. I am good with that plan, but I'll need to investigate first why we had a failure with one of the Chromebooks on this v3 patchset. That may give us some insights on "special" uses of those flags. > When we consider it very detail, above code can't 100% keep compatibility > if the user have been used this dpcm_xxx flag to limit availability, > for example in case of DAI can use both playback/capture, but it had > dpcm_playback flag only. But it can use playback_only flag, instead. > But it is very difficult to find such DAI. Each user need to check. > > I personally think that remove dpcm_xxx directly is no ploblem, but what > do you think ? I'm happy to hear any opinion, and happy to create > grace time code if someone want, but if there was no comment during > Japanese long vacation, I will create patch to remove dpcm_xxx directly. > > > BTW, I would like to know detail things around this topic. I'm happy if > someone knows it. > > * Why dummy DAI doesn't/can't have channels_min ? > > * Why it checks CPU side channels_min only when DPCM ? > I think it should check both CPU and Codec. > I could understand if it checks FE:CPU and BE:Codec (it is assuming > other side was dummy), but both (FE/BE) check CPU side only...
>> But because we have been used dpcm_xxx for 10 years since (B), >> I understand to feel anxious to suddenly remove dpcm_xxx. > > Indeed we err on the side of paranoia with such changes! > >> I think it should be removed anyway, but want to have grace time ? >> If so, the idea is that we can use it as "option flag" instead of >> "mandatory flag" for a while, like below >> >> if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { >> playback = (cpu_dai->driver->playback.channels_min > 0) || >> rtd->dai_link->dpcm_playback; >> capture = (cpu_dai->driver->capture.channels_min > 0) || >> rtd->dai_link->dpcm_capture; >> >> * maybe we want to indicate message like "place re-check the flag and >> remove it" via dev_info() if dpcm_xxx flag was used ? >> >> I think +2 kernel version or so is enough for grace time ? >> After that, we can remove dpcm_xxx flag. > > I am good with that plan, but I'll need to investigate first why we had > a failure with one of the Chromebooks on this v3 patchset. That may give > us some insights on "special" uses of those flags. I found the reason why this patchset failed on Intel CI: a dailink direction is deemed supported only if ALL cpu- and codec-dais support it. That is a stricter criterion than in existing code. This was a good intention based on symmetry, but in practice there are exceptions to the rule... On some Chromebooks, we tag a dailink as supporting capture for echo reference, but that echo reference is generated by the Intel firmware. The amplifiers only support playback. see https://github.com/thesofproject/linux/pull/4937 for details, I added a clear log: [ 807.304740] kernel: SSP1-Codec: CPU dai SSP1 Pin supports capture but codec DAI rt1011-aif does not So I think for now we probably want to avoid this stricter criterion and only check the supported direction with the cpu-dais. Or we add an escape mechanism to let the core know that the capture support was intentional. Also we relax the checks to go back to (E) 4f8721542f7b75954bfad98c51aa59d683d35b50 ("ASoC: core: use less strict tests for dailink capabilities") " This patch modifies the snd_soc_dai_link_set_capabilities() helper so that the dpcm_playback (resp. dpcm_capture) dailink capabilities are set if at least one [cpu-]dai supports playback (resp. capture). " We tried checking all cpu-dais before and found issues, so "at least one cpu dai" seems the strictest test to apply without breaking legacy.
Hi Pierre-Louis Thank you for your feedback and report > On some Chromebooks, we tag a dailink as supporting capture for echo > reference, but that echo reference is generated by the Intel firmware. > The amplifiers only support playback. > > see https://github.com/thesofproject/linux/pull/4937 for details, I > added a clear log: > > [ 807.304740] kernel: SSP1-Codec: CPU dai SSP1 Pin supports capture > but codec DAI rt1011-aif does not > > So I think for now we probably want to avoid this stricter criterion and > only check the supported direction with the cpu-dais. Or we add an > escape mechanism to let the core know that the capture support was > intentional. I think my patch have escape mechanism, but it was for BE Codec. If my understand was correct, above is FE Codec ? One question is that is it just a mistake (no one noticed) ? or is there some reasons it can't support capture (but use it ?). If it was just a mistake, is it possible to update driver during the grace time ? Or do you think we need "escape mechanism" permanently ? > I agree with your analysis. We don't have a clear memory/understanding > of which "no_chan_DAI" platforms (B) was supposed to handle, and why no > one reported them as broken by (C). (snip) > I am good with that plan, but I'll need to investigate first why we had > a failure with one of the Chromebooks on this v3 patchset. That may give > us some insights on "special" uses of those flags. Thanks. It seems the code was just complicated, and we have been missing important check. Let's break out my patch-set into small pieces, and go more slowly. I think it will be like below. Can you agree about this ? Step1 dpcm_xxx flag will be "option flag" instead of "mandatory flag" for a while to keep compatibility and avoide confusion. But it will be removed in Step3. To indicate such things, it will have dev_warn() if dpcm_xxx flag was used. like below if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { has_playback = /* at least one of CPU DAI supports playback */ has_capture = ... if (!playback && rtd->dai_link->dpcm_playback) { dev_warn(dev, "Playback is requested, but CPU doesn't support it\n") has_playback = 1; } ... Step2 (In case of "escape mechanism" is not needed) Current DPCM is checking CPU side only. Indicate warning until Step4 if Codec is not match . warning only, not error for a while. if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { ... if (has_playback && /* no Codec DAI supports playback */) dev_warn(dev, "Playback is requested, but Codec doesn't support it\n") ... Step3 Step1 deadline remove dpcm_xxx flag Step4 Step2 deadline CPU / Codec mismatch will be error. It will be "at least one of CPU/Codec pair supports playback/capture" Step5 There is no big diff between DPCM / non-DPCM check. Merge these, and clean-up code (soc_get_playback_capture()) Thank you for your help !! Best regards --- Renesas Electronics Ph.D. Kuninori Morimoto
Hi > Step1 > dpcm_xxx flag will be "option flag" instead of "mandatory flag" > for a while to keep compatibility and avoide confusion. > But it will be removed in Step3. To indicate such things, > it will have dev_warn() if dpcm_xxx flag was used. like below > > if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) { > has_playback = /* at least one of CPU DAI supports playback */ > has_capture = ... > > if (!playback && rtd->dai_link->dpcm_playback) { > dev_warn(dev, "Playback is requested, but CPU doesn't support it\n") > has_playback = 1; > } > ... I noticed that this Step is not needed, because all existing driver has both dpcm_xxx and availability. Instead, We want to have warning to indicate to use xxx_only flag if dpcm_xxx was used as limit availability. Thank you for your help !! Best regards --- Renesas Electronics Ph.D. Kuninori Morimoto
diff --git a/include/sound/soc.h b/include/sound/soc.h index 0376f7e4c15d..e604d74f6e33 100644 --- a/include/sound/soc.h +++ b/include/sound/soc.h @@ -809,6 +809,15 @@ struct snd_soc_dai_link { unsigned int dpcm_capture:1; unsigned int dpcm_playback:1; + /* + * Capture / Playback support assertion. Having assertion flag is not mandatory. + * In case of having assertion flag, non specific side will be disabled. + * see details + * soc_get_playback_capture() + */ + unsigned int capture_assertion:1; + unsigned int playback_assertion:1; + /* DPCM used FE & BE merged format */ unsigned int dpcm_merged_format:1; /* DPCM used FE & BE merged channel */ diff --git a/sound/soc/soc-pcm.c b/sound/soc/soc-pcm.c index b0e1bd7f588b..412e7b7d97f5 100644 --- a/sound/soc/soc-pcm.c +++ b/sound/soc/soc-pcm.c @@ -2794,7 +2794,12 @@ static int soc_get_playback_capture(struct snd_soc_pcm_runtime *rtd, int *playback, int *capture) { struct snd_soc_dai_link *dai_link = rtd->dai_link; + struct snd_soc_dai_link_ch_map *ch_maps; struct snd_soc_dai *cpu_dai; + struct snd_soc_dai *codec_dai; + struct snd_soc_dai *dummy_dai = snd_soc_find_dai(&snd_soc_dummy_dlc); + int cpu_playback; + int cpu_capture; int has_playback = 0; int has_capture = 0; int i; @@ -2804,77 +2809,95 @@ static int soc_get_playback_capture(struct snd_soc_pcm_runtime *rtd, return -EINVAL; } - if (dai_link->dynamic || dai_link->no_pcm) { - int stream; - - if (dai_link->dpcm_playback) { - stream = SNDRV_PCM_STREAM_PLAYBACK; - - for_each_rtd_cpu_dais(rtd, i, cpu_dai) { - if (snd_soc_dai_stream_valid(cpu_dai, stream)) { - has_playback = 1; - break; - } - } - if (!has_playback) { - dev_err(rtd->card->dev, - "No CPU DAIs support playback for stream %s\n", - dai_link->stream_name); - return -EINVAL; - } - } - if (dai_link->dpcm_capture) { - stream = SNDRV_PCM_STREAM_CAPTURE; + /* + * REMOVE ME + * + * dpcm_playback/capture will be used as playback/capture_assertion + */ + if (dai_link->playback_only && dai_link->capture_only) { + dev_err(rtd->dev, "both playback_only / capture_only are set\n"); + return -EINVAL; + } + if (dai_link->playback_only) + dai_link->playback_assertion = 1; + if (dai_link->capture_only) + dai_link->capture_assertion = 1; + if (dai_link->dpcm_playback) + dai_link->playback_assertion = 1; + if (dai_link->dpcm_capture) + dai_link->capture_assertion = 1; - for_each_rtd_cpu_dais(rtd, i, cpu_dai) { - if (snd_soc_dai_stream_valid(cpu_dai, stream)) { - has_capture = 1; - break; - } - } + /* Adapt stream for codec2codec links */ + cpu_playback = snd_soc_get_stream_cpu(dai_link, SNDRV_PCM_STREAM_PLAYBACK); + cpu_capture = snd_soc_get_stream_cpu(dai_link, SNDRV_PCM_STREAM_CAPTURE); - if (!has_capture) { - dev_err(rtd->card->dev, - "No CPU DAIs support capture for stream %s\n", - dai_link->stream_name); - return -EINVAL; - } - } - } else { - struct snd_soc_dai_link_ch_map *ch_maps; - struct snd_soc_dai *codec_dai; - - /* Adapt stream for codec2codec links */ - int cpu_capture = snd_soc_get_stream_cpu(dai_link, SNDRV_PCM_STREAM_CAPTURE); - int cpu_playback = snd_soc_get_stream_cpu(dai_link, SNDRV_PCM_STREAM_PLAYBACK); + /* + * see + * soc.h :: [dai_link->ch_maps Image sample] + */ + for_each_rtd_ch_maps(rtd, i, ch_maps) { + cpu_dai = snd_soc_rtd_to_cpu(rtd, ch_maps->cpu); + codec_dai = snd_soc_rtd_to_codec(rtd, ch_maps->codec); /* - * see - * soc.h :: [dai_link->ch_maps Image sample] + * FIXME + * + * DPCM BE Codec has been no checked before. + * It should be checked, but it breaks compatibility. + * It ignores BE Codec here, so far. */ - for_each_rtd_ch_maps(rtd, i, ch_maps) { - cpu_dai = snd_soc_rtd_to_cpu(rtd, ch_maps->cpu); - codec_dai = snd_soc_rtd_to_codec(rtd, ch_maps->codec); - - if (snd_soc_dai_stream_valid(codec_dai, SNDRV_PCM_STREAM_PLAYBACK) && - snd_soc_dai_stream_valid(cpu_dai, cpu_playback)) - has_playback = 1; - if (snd_soc_dai_stream_valid(codec_dai, SNDRV_PCM_STREAM_CAPTURE) && - snd_soc_dai_stream_valid(cpu_dai, cpu_capture)) - has_capture = 1; - } - } + if (dai_link->no_pcm) + codec_dai = dummy_dai; - if (dai_link->playback_only) - has_capture = 0; + if (snd_soc_dai_stream_valid(cpu_dai, cpu_playback) && + snd_soc_dai_stream_valid(codec_dai, SNDRV_PCM_STREAM_PLAYBACK)) + has_playback = 1; + if (snd_soc_dai_stream_valid(cpu_dai, cpu_capture) && + snd_soc_dai_stream_valid(codec_dai, SNDRV_PCM_STREAM_CAPTURE)) + has_capture = 1; + } - if (dai_link->capture_only) - has_playback = 0; + /* + * Assertion check + * + * playback_assertion = 0 No assertion check. + * capture_assertion = 0 ASoC will use detected playback/capture as-is. + * No playback, No capture will be error. + * + * playback_assertion = 1 DAI must playback available. ASoC will disable capture. + * capture_assertion = 0 In other words "playback_only" + * + * playback_assertion = 0 DAI must capture available. ASoC will disable playback. + * capture_assertion = 1 In other words "capture_only" + * + * playback_assertion = 1 DAI must both playback/capture available. + * capture_assertion = 1 + */ + if (dai_link->playback_assertion) { + if (!has_playback) { + dev_err(rtd->dev, "%s playback assertion check error\n", dai_link->stream_name); + return -EINVAL; + } + /* makes it plyaback only */ + if (!dai_link->capture_assertion) + has_capture = 0; + } + if (dai_link->capture_assertion) { + if (!has_capture) { + dev_err(rtd->dev, "%s capture assertion check error\n", dai_link->stream_name); + return -EINVAL; + } + /* makes it capture only */ + if (!dai_link->playback_assertion) + has_playback = 0; + } + /* + * Detect Mismatch + */ if (!has_playback && !has_capture) { dev_err(rtd->dev, "substream %s has no playback, no capture\n", dai_link->stream_name); - return -EINVAL; }
Current soc_get_playback_capture() (A) is checking playback/capture availability for DPCM (X) / Normal (Y) / Codec2Codec (Z) connections. (A) static int soc_get_playback_capture(...) { ... ^ if (dai_link->dynamic || dai_link->no_pcm) { | ... |(a) if (dai_link->dpcm_playback) { | ... | ^ for_each_rtd_cpu_dais(rtd, i, cpu_dai) { |(*) ... | v } | ... (X) } |(b) if (dai_link->dpcm_capture) { | ... | ^ for_each_rtd_cpu_dais(rtd, i, cpu_dai) { |(*) ... | v } | ... v } } else { ^ ^ /* Adapt stream for codec2codec links */ |(Z) int cpu_capture = ... | v int cpu_playback = ... (Y) | ^ for_each_rtd_ch_maps(rtd, i, ch_maps) { |(*) ... v v } } ... } (*) part is checking each DAI's availability. (X) part is for DPCM, and it checks playback/capture availability if dai_link has dpcm_playback/capture flag (a)(b). This availability check should be available not only for DPCM, but for all connections. But it is not mandatory option. Let's name it as assertion. In case of having assertion flag, non specific side will be disabled. For example if it has playback_assertion but not have capture_assertion, capture will be force disabled. dpcm_playback -> playabck_assertion = 1 dpcm_capture -> capture_assertion = 1 playback_only -> playback_assertion = 1 capture_assertion = 0 capture_only -> playback_assertion = 0 capture_assertion = 1 By expanding this assertion to all connections, we can use same code for all connections, this means code can be more cleanup. Here, current CPU / Codec validation check relationship is like this DPCM [CPU/xxxx]-[yyyy/Codec] ^^^^ ^^^^ non DPCM [CPU/Codec] ^^^^^^^^^^^ DPCM part (X) is checking only CPU DAI, and non DPCM part (Y) is checking both CPU/Codec DAI Current validation check on DPCM ignores Codec DAI, but there is no reason to do it. We should check both CPU/Codec in all connection. This patch expands validation check to all cases. [CPU/xxxx]-[yyyy/Codec] ***** In many case (not all case), above [xxxx][yyyy] part are "dummy" DAI which is always valid for both playback/capture. But unfortunately DPCM BE Codec (**** part) had been no validation check before, and some Codec driver doesn't have necessary settings for it. This means all cases validation check breaks compatibility on some vender's drivers. Thus this patch temporary uses dummy DAI at BPCM BE Codec part, and avoid compatibility error. But it should be removed in the future. Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com> --- include/sound/soc.h | 9 +++ sound/soc/soc-pcm.c | 143 +++++++++++++++++++++++++------------------- 2 files changed, 92 insertions(+), 60 deletions(-)