Add support for Mediatek mt8195 platform #3122

yaochunhung · 2021-08-26T13:02:16Z

Hi,
This PR is for support mt8195 platform on SOF.
MT8195 include Cadence HiFi-4 single core, currently we support 4 DAI interfaces, SOF_DL2/DL3/UL4/UL5(Downlink/Uplink path) on MT8195 Audio Hardware(which we call AFE, the Abbreviation of Audio Front End). Some ASoC codes(the committer is Trevor) in the PR is accepted by ASoC reviewer but not in the sof-linux branch, please ignore them then help to review on SOF part.
Thanks!

Update mediatek common driver to support MT8195 Signed-off-by: Trevor Wu <trevor.wu@mediatek.com> Reported-by: kernel test robot <lkp@intel.com>

This patch adds mt8195 audio cg control. Audio clock gates are registered to CCF for reference count and clock parent management. Signed-off-by: Trevor Wu <trevor.wu@mediatek.com>

This patch adds mt8195 tdm/i2s dai driver. MCLK clock tree is as follows. PLL -> MUX -> DIVIDER -> MCLK For PLL source of MCLK, driver only supports APLL1 and APLL2 now. APLL3 and APLL4 are used to track external clock source, so they are only used when slave input is connected. For example, case 1: (HDMI RX connected) DL memif (a1sys) -> etdm out2 (clk from apll1/apll2) -> codec case 2: (HDMI RX disconnected) HDMI RX -> a3sys -> UL memif (a3sys) -> DL memif (a3sys) -> .... -> etdm out2 (clk from apll3) -> codec We keep all modules in the pipeline working on the same clock domain. MCLK is expected to output the clock generated from the same clock source as the pipeline, so dynamic reparenting is required for MCLK configuration. As a result, clk_set_parent() is used to select PLL source, and clk_set_rate() is used to configure divider to get MCLK output rate. Signed-off-by: Trevor Wu <trevor.wu@mediatek.com>

This patch adds mt8195 adda dai driver. audio_h clock is used by ADSP bus and ADDA module. When ADDA requires audio_h clock, it is switched to APLL1, otherwise it is switched to Xtal_26m so that APLL1 can be turned off when audio feature is not used. ADSP bus only requires that the clock is on, so dynamic reparenting is used for the purpose of lowering power consumption. Signed-off-by: Trevor Wu <trevor.wu@mediatek.com>

This patch adds mt8195 pcm dai driver. Signed-off-by: Trevor Wu <trevor.wu@mediatek.com>

This patch adds mt8195 platform and affiliated driver. Signed-off-by: Trevor Wu <trevor.wu@mediatek.com> Reported-by: kernel test robot <lkp@intel.com>

This patch adds mt8195 audio afe document. In order to support dynamic clock reparenting for ADDA and ETDM, PLL and MUX clocks are requested even though they are not consumed by afe directly. Signed-off-by: Trevor Wu <trevor.wu@mediatek.com>

…5682 This patch adds support for mt8195 board with mt6359, rt1019 and rt5682. Signed-off-by: Trevor Wu <trevor.wu@mediatek.com> Reported-by: kernel test robot <lkp@intel.com>

This patch adds DPTX audio support on mt8195-mt6359-rt1019-rt5682 board. Signed-off-by: Trevor Wu <trevor.wu@mediatek.com>

This patch adds HDMITX audio support on mt8195-mt6359-rt1019-rt5682 board. Signed-off-by: Trevor Wu <trevor.wu@mediatek.com>

This patch adds document for mt8195 board with mt6359, rt1019 and rt5682 Signed-off-by: Trevor Wu <trevor.wu@mediatek.com>

paulstelian97 · 2021-08-26T13:11:08Z

sound/soc/sof/pcm.c

 	pd->pcm_construct = sof_pcm_new;
-	pd->ignore_machine = drv_name;
+	if (!plat_data->desc->no_ignore_machine)
+		pd->ignore_machine = drv_name;


This change doesn't look backward compatible (default for new fields is supposed to be a 0-like value, in this case false).

default is zero and set as true if want to reuse original platform driver. we set it as true in the sof_of_mt8195_desc. Is there anything for my misunderstanding? Please advice.Thanks!

I would go with @dbaluta's suggestion. Call the flag the same in your descriptor as in the ASoC driver definition, this will avoid logical inversions that are unfortunately oh so common in engineering circles.

plbossart · 2021-08-26T14:38:11Z

Some ASoC codes(the committer is Trevor) in the PR is accepted by ASoC reviewer but not in the sof-linux branch, please ignore them then help to review on SOF part.

What patches are you referring to? We merge upstream code weekly (that includes Mark Brown and Takashi Iwai for-next branches), so everything needed should be there?

yaochunhung · 2021-08-26T14:53:44Z

What patches are you referring to? We merge upstream code weekly (that includes Mark Brown and Takashi Iwai for-next branches), so everything needed should be there?
Below are mt8195 ASoC commits in the linux kernel.
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound.git/log/?qt=author&q=Trevor+Wu
Because I add the commit ea950f5 is related with machine driver. So I also add them into the PR. Thanks!

include/sound/sof/dai-mediatek.h

sound/soc/sof/mediatek/mt8195/mt8195-clk.c

ranj063 · 2021-08-26T14:55:20Z

sound/soc/sof/mediatek/mt8195/mt8195.c

+	int ret = 0;
+
+	/* power on DSP */
+	pm_runtime_get_sync(&pdev->dev);


same here, why do you need the get_sync() here?

completely wrong, sorry.

I will remove it first then check how to add PM later. Thanks.

ranj063 · 2021-08-26T14:55:31Z

sound/soc/sof/mediatek/mt8195/mt8195.c

+	ret = adsp_clock_on(&pdev->dev);
+	if (ret) {
+		dev_err(sdev->dev, "adsp_clock_on fail!\n");
+		return -EINVAL;


return ret please

ok. thanks!

ranj063 · 2021-08-26T14:55:38Z

sound/soc/sof/mediatek/mt8195/mt8195.c

+	ret = adsp_sram_power_on(sdev->dev, true);
+	if (ret) {
+		dev_err(sdev->dev, "[ADSP] adsp_sram_power_on fail!\n");
+		return -EINVAL;


remove this. return ret at the end will handle it

ok. I will remove it. thanks!

ranj063 · 2021-08-26T14:58:33Z

sound/soc/sof/mediatek/mt8195/mt8195-ipc.c

+		return ret;
+	}
+
+	// init value;


use /* */ fr comments please

ok. I will fix it. Thanks!

ranj063 · 2021-08-26T14:58:49Z

sound/soc/sof/mediatek/mt8195/mt8195.c

+	if (ret < 0) {
+		dev_err(sdev->dev, "error: failed to init adsp ipi device:%d\n",
+			ret);
+		return ret;


remove this

ok. I will remove it. Thanks!

plbossart

Thanks for this series and welcome aboard.

My main objection is wrt to pm_runtime, it doesn't look like the solution allows for any suspend/resume to happen. There are also tons of style issues, please revisit.

plbossart · 2021-08-26T14:45:56Z

include/sound/sof.h

 	bool use_acpi_target_states;

+	/*set to 1 to not use ignore_machine and still use the original platform driver*/
+	bool no_ignore_machine;


unclear why you need this for?

Because we need to reuse original platform driver in our design, we don't want to overwrite it in the function soc_check_tplg_fes to still use original platform. Then we can route sof link to original(normal sink),
e.g. sof_conn_stream : { "ETDM2_OUT_BE", "AFE_SOF_DL2", SOF_DMA_DL2, SNDRV_PCM_STREAM_PLAYBACK}, to route sof_link : SOF_DMA_DL2 to normal link ETDM2_OUT_BE, then we can use normal link setting for audio hardware.

I think going with ignore_machine name here works better.

@yaochunhung I totally missed this. ignore_machine has nothing to do with ignoring the machine driver selected in SOF. It is more to do with how the core deals with the FE DAI links in the legacy machine driver. In your case, setting this should have no effect at all because there's no legacy driver to deal with right? @plbossart is my understanding correct?

@ranj063 We still use legacy driver because some pcm nodes are still controlled by non-sof driver, not move to DSP. And We also want to keep to legacy BE DAI driver because the power and clock of our audio hardware are controlled by AP side. e.g ETDM2_OUT_BE. If no_ignore_machine is set as true, the legacy driver can be keep and reuse. Thanks!

@plbossart In the machine driver sound/soc/mediatek/mt8195/mt8195-mt6359-rt1019-rt5682.c. we define g_sof_conn_streams about the connection among ETDM2_OUT_BE,SOF_DMA_DL2 and AFE_SOF_DL2.
static const struct sof_conn_stream g_sof_conn_streams[] = {
{ "ETDM2_OUT_BE", "AFE_SOF_DL2", SOF_DMA_DL2, SNDRV_PCM_STREAM_PLAYBACK},
{ "ETDM1_OUT_BE", "AFE_SOF_DL3", SOF_DMA_DL3, SNDRV_PCM_STREAM_PLAYBACK},
{ "UL_SRC1_BE", "AFE_SOF_UL4", SOF_DMA_UL4, SNDRV_PCM_STREAM_CAPTURE},
{ "ETDM2_IN_BE", "AFE_SOF_UL5", SOF_DMA_UL5, SNDRV_PCM_STREAM_CAPTURE},
};
Then we implement .late_probe = mt8195_mt6359_rt1019_rt5682_card_late_probe.
In this function, we search pcm_runtime by sof_link and normal_link then get sof's pcm runtime : sof_rtd and normal pcm runtime : normal_rtd. After that, we search the sof's runtime cpu_dai then check the widget in its path then set dapm route's source and sink then add routes.

for example, ETDM2_OUT_BE is normal_link, AFE_SOF_DL2 is sof_link,SOF_DMA_DL2 is sof_dma, and stream_dir is SNDRV_PCM_STREAM_PLAYBACK.
We will use snd_soc_dapm_widget_for_each_source_path to get the playback_widget is BUFn.x in the diagram above.
then set sink of dpam route as conn->sof_dma(SOF_DMA_DL2 in this case). After that, BUFn.x will be connected to SOF_DMA_DL2.
For SOF_DMA_DL2 to ETDM2_OUT_BE , we define the dapm route path in machine driver.
/* SOF Downlink */
{"I070", NULL, SOF_DMA_DL2},
{"I071", NULL, SOF_DMA_DL2},
And also define dapm route path in mt8195-dai-etdm.c
{"O048", "I070 Switch", "I070"},
{"O049", "I071 Switch", "I071"},
{"ETDM2 Playback", NULL, "O048"},
{"ETDM2 Playback", NULL, "O049"},
I hope this explanation is more clear. Thanks.

I hope this explanation is more clear. Thanks.

Humm, no :-(
I am missing the big picture, you just hammered us with details that are probably correct but don't really provide a sense of what the problem statement is, if you considered alternatives and what led you to pick this solution you're asking us to review.

@plbossart
We set no_ignore_machine as true to prevent ignoring normal FE in soc_check_tplg_fes.

/* ignore this FE */
if (dai_link->dynamic) {
dai_link->ignore = true;
continue;
})

The reason is that we still need to use normal FE for the paths which don't gothrough DSP. The benifit is that we can offload the process of certain audio data(e.g. speaker/headphone) to dsp and we can reuse normal alsa driver for those audio data which don't need dsp offloading.(e.g. Display port). Thanks.

@yaochunhung It makes a bit more sense now. but a naive question. Is there a way for you define the Normal FE-> Normal BE in topology?

@ranj063 For normal FE/BE, I think it need to communicate with DSP if defining them in topology(please correct me if my misunderstanding). It takes lower effort to reuse normal FE/BE without topology. Thanks.

sound/soc/sof/mediatek/Kconfig

plbossart · 2021-08-26T14:48:39Z

sound/soc/sof/mediatek/adsp_helper.h

+#define __ADSP_HELPER_H__
+
+#include <linux/of_irq.h>
+#include <linux/platform_device.h>


move those to where you include this file? I don't see any dependency on platform_device below.

yes, they are unnecessary and I will remove it.

ok. I will remove them. Thanks!

plbossart · 2021-08-26T14:50:00Z

sound/soc/sof/mediatek/mt8195/mt8195-clk.c

+	}
+
+	clk_handle[CLK_TOP_AUDIO_LOCAL_BUS_SEL] = devm_clk_get(dev,
+							       "audio_local_bus");


one line if it fits on 100 chars?

yes, I will modify it

plbossart · 2021-08-26T14:50:10Z

sound/soc/sof/mediatek/mt8195/mt8195-clk.c

+	}
+
+	clk_handle[CLK_SCP_ADSP_AUDIODSP] = devm_clk_get(dev,
+							 "scp_adsp_audiodsp");


ok. I will modify it. thanks

plbossart · 2021-08-26T15:15:05Z

sound/soc/mediatek/mt8195/mt8195-mt6359-rt1019-rt5682.c

+
+static int mt8195_mt6359_rt1019_rt5682_card_late_probe(struct snd_soc_card *card)
+{
+	int i = 0;


initialization not needed, and move two lines below.

ok. I will modify it.

plbossart · 2021-08-26T15:15:24Z

sound/soc/mediatek/mt8195/mt8195-mt6359-rt1019-rt5682.c

+{
+	int i = 0;
+	struct snd_soc_pcm_runtime *runtime = NULL;
+	struct snd_soc_component *sof_comp = NULL;


initializations not needed.

ok. I will modify them.

plbossart · 2021-08-26T15:15:52Z

sound/soc/mediatek/mt8195/mt8195-mt6359-rt1019-rt5682.c

+		const struct sof_conn_stream *conn = &g_sof_conn_streams[i];
+		struct snd_soc_pcm_runtime *sof_rtd = NULL;
+		struct snd_soc_pcm_runtime *normal_rtd = NULL;
+		struct snd_soc_pcm_runtime *rtd = NULL;


revisit all initializations please.

ok. I will modify it.

plbossart · 2021-08-26T15:16:55Z

sound/soc/mediatek/mt8195/mt8195-mt6359-rt1019-rt5682.c

+	for_each_card_prelinks(card, i, dai_link) {
+		if (dai_link->no_pcm && !dai_link->stream_name && dai_link->name)
+			dai_link->stream_name = dai_link->name;
+	}


what is this doing? This looks like ASoC code that's duplicated in the machine driver?

https://elixir.bootlin.com/linux/latest/source/sound/soc/soc-core.c#L1851 here skip assign stream_name in soc_check_tplg_fes function when no_ignore_machineis true, so we assign them in the mt8195_mt6359_rt1019_rt5682_card_probe

Not understanding the explanation, sorry.

plbossart · 2021-08-26T15:18:41Z

sound/soc/mediatek/mt8195/mt8195-mt6359-rt1019-rt5682.c

+#else
 	platform_node = of_parse_phandle(pdev->dev.of_node,
 					 "mediatek,platform", 0);
+#endif


this doesn't look good to me, if there is a single build where you want deal with two platforms, you need to auto-detect the platform then make a selection at run time. Compile-time selections are really problematic for distributions. @cujomalainey please comment here.

Sorry missed this comment. I am just trying to make sure I understand this code correctly. Its a compile time switch to select which driver services the platform correct? @yaochunhung does MTK plan to make SOF the default for this platform? Or do you plan to use the other driver still? @plbossart this seems like a reasonable switch for during the transition period, much like we had on GLK for a while no?

@cujomalainey Yes, we plan to make SOF the default for this platform. thanks!

plbossart · 2021-08-26T15:22:29Z

@yaochunhung You will need to work with @dbaluta to update the ARM64 Kconfig so that we can compile your code in GitHub CI. See https://github.com/thesofproject/kconfig/blob/master/kconfig-sof-arm64.sh

dbaluta · 2021-08-27T07:04:49Z

@yaochunhung I will provide you more feedback on Monday. Care to share parts of your dts file? Also, do you reuse your non-SOF machine drivers right?

yaochunhung

@yaochunhung I will provide you more feedback on Monday. Care to share parts of your dts file? Also, do you reuse your non-SOF machine drivers right?

@dbaluta yes, we reuse non-SOF machine driver and I can share dts file to you. Please tell me your email. Thanks!

include/sound/sof/dai-mediatek.h

yaochunhung · 2021-08-27T06:36:58Z

sound/soc/sof/mediatek/mt8195/mt8195.c

 static struct adsp_chip_info *adsp_info;

+static struct snd_soc_acpi_mach sof_mt8195_mach = {
+	.id = "819501",


No, we use acpi mach for mt8195, this id is mt'8195' with version '01'

sound/soc/sof/mediatek/Kconfig

sound/soc/sof/mediatek/mt8195/mt8195.c

yaochunhung · 2021-08-27T07:30:51Z

sound/soc/sof/mediatek/mt8195/mt8195.c

+	struct platform_device *pdev =
+		container_of(sdev->dev, struct platform_device, dev);
+	struct adsp_priv *priv;
+	int ret = 0, mailbox_type;


ok, I will modify it.

sound/soc/sof/mediatek/mt8195/mt8195.c

yaochunhung · 2021-09-02T07:57:55Z

@yaochunhung I've gone your patch series and added some comments. There is still work to be done. I would suggest you take it slower with simple patches first and then go for more delicate ones.

You can run git log --oneline sound/soc/sof/imx/ and see how we added our initial support for i.MX8 boards. It is not complete, we are now sending our clocks patches, then following by PM support, then the rest of the stuff.

If you want and consider it necessary we could also setup a meeting to discuss more.

include/sound/sof/dai-mediatek.h

dbaluta · 2021-08-31T10:56:56Z

sound/soc/sof/mediatek/mt8195/mt8195-clk.c

+	if (ret) {
+		dev_err(dev, "%s clk_prepare_enable(scp_adsp_audiodsp) fail %d\n",
+			__func__, ret);
+		return ret;


If clk_prepare_enable fails here we should call clk_disable_unprepare for the rest of the clocks.

ok. I will modify it. thanks!

dbaluta · 2021-09-01T14:42:58Z

sound/soc/sof/Kconfig

 	  When selected, the probe is handled in two steps, for example to
 	  avoid lockdeps if request_module is used in the probe.

 source "sound/soc/sof/imx/Kconfig"


@yaochunhung would love to see some more details in the commit message. E.g How is the hardware configuration for your platform, what kind of architecture / cores is on application processor, what kind of DSP do you have.
Also some information about the way these two entities communicate, how are the resources managed (clocks, power domains, pinmuxing, etc).
This is useful for review.

ok. I will add more information. Thanks!

dbaluta · 2021-09-01T14:44:50Z

sound/soc/sof/mediatek/adsp_helper.h

+	u32 (*ap2adsp_addr)(u32 addr);
+	u32 (*adsp2ap_addr)(u32 addr);
+
+	void *private_data;


Also can you provide a short memory map from AP view and DSP view. I suspect there is no 1 to 1 mapping. We have the same scenarios on one of our boards and we've implemented similar functions with ap2adsp_addr but on the firmware side not on the kernel. Not sure which is better. I dont suggest to change this if it fits your needs, but it must be explained in some comments or in the commit message.

There is a memory remapping for DRAM between AP view and DSP view. In DSP view, the start address of EMI is 0x50000000 and in AP view is 0x40000000. In current design, I set dram of SOF between AP and DSP are the same. Both are 0x60000000. the memory remapping is zero, Could you please tell me what's the keyword for your implementation? I will check it. Thanks!

sound/soc/sof/mediatek/mt8195/mt8195-clk.c

dbaluta · 2021-09-01T14:47:53Z

sound/soc/sof/mediatek/mt8195/mt8195.c

-	adsp = get_adsp_chip_data();
-	if (!adsp)
-		return NULL;
-


You shouldn't delete code that was previously added in a patch of the same Pull Request. Just do not add it at all from the beginning.

ok. I will modify it.

dbaluta · 2021-09-01T14:48:43Z

sound/soc/sof/mediatek/mt8195/mt8195.c

+
+	/* get machine node and save it to mach->pdata */
+	mach->pdata = of_get_child_by_name(sdev->dev->of_node, "sound");
+	if (!mach->pdata)


Definitely this needs more explanations. How to you plan to describe your machine. Maybe create finer grained patches with proper patches explanations. This should also go to review to DT maintainer.

dbaluta · 2021-09-01T14:50:08Z

sound/soc/sof/mediatek/mt8195/mt8195.c

+#endif
+	return type;
+}
+


This function here is not related with this commit. Maybe add it in the initial commit with hardware support. Also, need to explain the use of iffdef.

ok. I will move it in the initial commit. Not ADSP_USE_DRAM is only used in early development stage for DSP SRAM only case, But DSP SRAM is only 256KB, I think it won't be used without DRAM. I think the definition can be removed. Thanks.

dbaluta · 2021-09-08T14:53:13Z

@yaochunhung my suggestion for this is to submit only the obvious correct patches and then add the rest later. The amount of patches right now will make the code hard the review.
I would suggest on going with:

adding sound/soc/sof/mtk/ -> stuff (MTK HW support only)
sound-of-dev.c changes

Add PM support later.

yaochunhung · 2021-09-09T02:52:00Z

@yaochunhung my suggestion for this is to submit only the obvious correct patches and then add the rest later. The amount of patches right now will make the code hard the review.
I would suggest on going with:

adding sound/soc/sof/mtk/ -> stuff (MTK HW support only)

sound-of-dev.c changes

Add PM support later.
@dbaluta Thanks for suggestion. Do you mean that I should separate my patches and only summit correct patches? Because PM runtime is also handled in probe function, should I remove it first? Thanks.

ranj063 · 2021-09-09T15:17:30Z

sound/soc/sof/mediatek/mt8195/mt8195-clk.c

+	} else { /* Power down ADSP SRAM */
+		writel(readl(va_dspsysreg) | DSP_SRAM_POOL_PD_MASK,
+		       va_dspsysreg);
+	}


braces not needed

ok. I will remove them. thanks

ranj063 · 2021-09-09T15:18:21Z

sound/soc/sof/mediatek/mt8195/mt8195-clk.c

+			dev_err(dev, "failed to adsp_enable_clock: %d\n", ret);
+			goto TAIL;
+		}
+	} else {


what I meant was having a return inside the if(enable) so you wont need the else. but its only minor

sound/soc/sof/mediatek/mt8195/mt8195.c

ranj063 · 2021-09-09T15:20:02Z

sound/soc/sof/mediatek/mt8195/mt8195.c

+		return ret;
+	}
+
+	dev_info(dev, "[ADSP][DMA] pbase=0x%llx, size=0x%llx\n",


is this really meant to be info? or can it be dbg?

ok. I think it can be dbg. thanks

ranj063 · 2021-09-09T15:20:14Z

sound/soc/sof/mediatek/mt8195/mt8195.c

+		return -EINVAL;
+	}
+
+	dev_info(dev, "[ADSP] dram pbase=%pa, dramsize=0x%x\n",


same here, dbg level maybe?

ok. I will modify it. thanks

ranj063 · 2021-09-09T15:21:08Z

sound/soc/sof/mediatek/mt8195/mt8195.c

+		if (IS_ERR(adsp->va_mboxreg[i]))
+			return PTR_ERR(adsp->va_mboxreg[i]);
+
+		dev_info(dev, "[ADSP]MBOX%d pa_mboxreg:%pa, va_mboxreg:%p\n",


could you please go through all th dev_info's in your series to make sure they need to be info and not dbg?

ok. I think they could be dbg level. thanks

ranj063 · 2021-09-09T15:21:49Z

sound/soc/sof/mediatek/mt8195/mt8195.c

+		 offset);
+	WARN_ON(offset < 0);
+	writel(offset, vaddr_emi_map);
+	dev_info(dev, "After vaddr_emi_map 0x%x\n", readl(vaddr_emi_map));


right im not asking to remove. just wondering if they all need to be info level or can they be dev_dbg()

ranj063 · 2021-09-09T15:24:05Z

sound/soc/sof/mediatek/mt8195/mt8195.c

+static int mt8195_get_bar_index(struct snd_sof_dev *sdev, u32 type)
+{
+	return type;
+}


is this a TODO? why return an argument as is?

It is one-to-one match between type and Bar index so it just return type. thanks

can you please add a comment to say this?

ok. I will add comment for it. thanks

ranj063 · 2021-09-09T15:30:00Z

sound/soc/sof/mediatek/adsp_ipc.c

+	ch_info->ipi_op_val = op;
+	ret = mbox_send_message(priv->ipi[idx].chan, NULL);
+	if (ret < 0)
+		dev_err(sdev->dev, "failed to send message via mbox: %d\n", ret);


if this can fail, should this function return the error?

ok . I will modify it. thanks.

ranj063 · 2021-09-09T15:31:12Z

sound/soc/sof/mediatek/mt8195/mt8195.c

+	sof_mailbox_write(sdev, sdev->host_box.offset, msg->msg_data,
+			  msg->msg_size);
+
+	adsp_ipi_send(sdev, ADSP_IPI_MBOX_REQ, ADSP_IPI_OP_REQ);


this can file, so ideally you should return the error to indicate if the msg TX failed

ok. I will add it. thanks.

Add the definition for Mediatek audio front end(AFE) tokens,include AFE sampling rate, channels, and format. Signed-off-by: YC Hung <yc.hung@mediatek.com>

Add new sof dai and config to pass topology file configuration to SOF firmware running on Mediatek platform DSP core. Add mediatek audio front end(AFE) to the list of supported sof_dais Signed-off-by: YC Hung <yc.hung@mediatek.com>

Add no_ignore_machine to check if ignore machine, set to 1 to not use ignore_machine and still use the original platform driver. Signed-off-by: YC Hung <yc.hung@mediatek.com>

This patch initialize to support SOF on Mediatek mt8195 platform. MT8195 has four Coretex A78 cores paired with four Coretex A55 cores. It also has Cadence HiFi-4 DSP single core. There are shared DRAM and mailbox interrupt between AP and DSP to use for IPC communication. The clocks of bus and dsp core are managed by CCF module. The power domain of DSP will be handled by mtk-scpsys. Signed-off-by: YC Hung <yc.hung@mediatek.com>

Add dsp ops callback to select and register machine driver. Signed-off-by: YC Hung <yc.hung@mediatek.com>

Add dsp ops callback to register AFE DL2/DL3/UL4/UL5 sof dai's with ALSA Signed-off-by: YC Hung <yc.hung@mediatek.com>

Add mt8195-loader module with ops callback to load and run firmware on mt8195 platform. Signed-off-by: YC Hung <yc.hung@mediatek.com>

Support pcm stream callback for hw_params and hw_free Signed-off-by: YC Hung <yc.hung@mediatek.com>

Add DSP PM callback for supsend and resume Signed-off-by: YC Hung <yc.hung@mediatek.com>

Add IPC module for mt8195 and exposed ops callback for to synchronize SOF IPC message between host and DSP Signed-off-by: YC Hung <yc.hung@mediatek.com>

This patch adds document for mt8195 DSP Signed-off-by: YC Hung <yc.hung@mediatek.com>

Add SOF device and DT descriptor for Mediatek mt8195 platform. Signed-off-by: YC Hung <yc.hung@mediatek.com>

Add SOF platform driver and use connection table to connect sof platform driver and normal platform driver. Use no_ignore_machine in sof device descriptor and set as 1 to keep and reuse original platform driver. Signed-off-by: YC Hung <yc.hung@mediatek.com>

yaochunhung · 2021-09-14T09:37:47Z

Per discussion with @dbaluta , I will close this pull and keep it for reference. Then create new PR for basic skeleton for MTK hardware support first. Thanks.

cujomalainey · 2021-09-14T17:10:57Z

@yaochunhung please link the followup PRs here

yaochunhung · 2021-09-15T06:57:14Z

@yaochunhung please link the followup PRs here

@cujomalainey ok. I will link them. thanks!

Basic hardware support for mt8195 : #3160 (merged)
Add PCM stream callback for mt8195 : #3190
ASoC: SOF: pcm: add sof_keep_normal_fes to keep normal fes : #3217

wenliangwu added 11 commits August 26, 2021 11:09

ASoC: mediatek: mt8195: update mediatek common driver

8ed024a

Update mediatek common driver to support MT8195 Signed-off-by: Trevor Wu <trevor.wu@mediatek.com> Reported-by: kernel test robot <lkp@intel.com>

ASoC: mediatek: mt8195: support audsys clock control

ed20695

This patch adds mt8195 audio cg control. Audio clock gates are registered to CCF for reference count and clock parent management. Signed-off-by: Trevor Wu <trevor.wu@mediatek.com>

ASoC: mediatek: mt8195: support pcm in platform driver

cdd32ef

This patch adds mt8195 pcm dai driver. Signed-off-by: Trevor Wu <trevor.wu@mediatek.com>

ASoC: mediatek: mt8195: add platform driver

1db6b58

This patch adds mt8195 platform and affiliated driver. Signed-off-by: Trevor Wu <trevor.wu@mediatek.com> Reported-by: kernel test robot <lkp@intel.com>

ASoC: mediatek: mt8195: add machine driver with mt6359, rt1019 and rt…

d6abd2e

…5682 This patch adds support for mt8195 board with mt6359, rt1019 and rt5682. Signed-off-by: Trevor Wu <trevor.wu@mediatek.com> Reported-by: kernel test robot <lkp@intel.com>

ASoC: mediatek: mt8195: add DPTX audio support

caab8f9

This patch adds DPTX audio support on mt8195-mt6359-rt1019-rt5682 board. Signed-off-by: Trevor Wu <trevor.wu@mediatek.com>

ASoC: mediatek: mt8195: add HDMITX audio support

d4e27ec

This patch adds HDMITX audio support on mt8195-mt6359-rt1019-rt5682 board. Signed-off-by: Trevor Wu <trevor.wu@mediatek.com>

dt-bindings: mediatek: mt8195: add mt8195-mt6359-rt1019-rt5682 document

87b70a9

This patch adds document for mt8195 board with mt6359, rt1019 and rt5682 Signed-off-by: Trevor Wu <trevor.wu@mediatek.com>

yaochunhung requested review from bardliao, dbaluta, juimonen, keyonjie, kv2019i, lgirdwood, lyakh, plbossart, ranj063 and singalsu as code owners August 26, 2021 13:02

paulstelian97 reviewed Aug 26, 2021

View reviewed changes

ranj063 requested changes Aug 26, 2021

View reviewed changes

plbossart requested changes Aug 26, 2021

View reviewed changes

yaochunhung commented Aug 27, 2021

View reviewed changes

yaochunhung force-pushed the mt8195-dev branch from ea950f5 to 9618a9b Compare August 27, 2021 15:40

yaochunhung closed this Sep 2, 2021

yaochunhung reopened this Sep 2, 2021

dbaluta reviewed Sep 6, 2021

View reviewed changes

yaochunhung force-pushed the mt8195-dev branch from 0aef6aa to b7ff292 Compare September 7, 2021 16:26

yaochunhung closed this Sep 9, 2021

yaochunhung reopened this Sep 9, 2021

yaochunhung requested a review from ranj063 September 9, 2021 06:21

ranj063 reviewed Sep 9, 2021

View reviewed changes

yaochunhung added 13 commits September 11, 2021 22:36

ASoC: SOF: tokens: add token for Mediatek AFE

ee0f944

Add the definition for Mediatek audio front end(AFE) tokens,include AFE sampling rate, channels, and format. Signed-off-by: YC Hung <yc.hung@mediatek.com>

ASoC: SOF: topology: Add support for Mediatek AFE DAI

2dc101d

Add new sof dai and config to pass topology file configuration to SOF firmware running on Mediatek platform DSP core. Add mediatek audio front end(AFE) to the list of supported sof_dais Signed-off-by: YC Hung <yc.hung@mediatek.com>

ASoC: SOF: pcm: add no ignore machine checking

6c2ce17

Add no_ignore_machine to check if ignore machine, set to 1 to not use ignore_machine and still use the original platform driver. Signed-off-by: YC Hung <yc.hung@mediatek.com>

ASoC: SOF: mediatek: Add machine driver dsp ops for mt8195 platform

488f3d3

Add dsp ops callback to select and register machine driver. Signed-off-by: YC Hung <yc.hung@mediatek.com>

ASoC: SOF: mediatek: Add dai driver dsp ops callback for mt8195

1659d6d

Add dsp ops callback to register AFE DL2/DL3/UL4/UL5 sof dai's with ALSA Signed-off-by: YC Hung <yc.hung@mediatek.com>

ASoC: SOF: mediatek: Add fw loader and mt8195 dsp ops to load firmware

a85c74f

Add mt8195-loader module with ops callback to load and run firmware on mt8195 platform. Signed-off-by: YC Hung <yc.hung@mediatek.com>

ASoC: SOF: mediatek: Add PCM stream callback for mt8195

b4f7aa7

Support pcm stream callback for hw_params and hw_free Signed-off-by: YC Hung <yc.hung@mediatek.com>

ASoC: SOF: mediatek: Add DSP PM callback for mt8195

583f14f

Add DSP PM callback for supsend and resume Signed-off-by: YC Hung <yc.hung@mediatek.com>

ASoC: SOF: mediatek: Add IPC support for mt8195

2ac0d18

Add IPC module for mt8195 and exposed ops callback for to synchronize SOF IPC message between host and DSP Signed-off-by: YC Hung <yc.hung@mediatek.com>

dt-ndings: mediatek: mt8195: add dsp document

8f38f99

This patch adds document for mt8195 DSP Signed-off-by: YC Hung <yc.hung@mediatek.com>

ASoC: SOF: Add mt8195 device descriptor

c836c0d

Add SOF device and DT descriptor for Mediatek mt8195 platform. Signed-off-by: YC Hung <yc.hung@mediatek.com>

yaochunhung force-pushed the mt8195-dev branch from b7ff292 to 46d26ca Compare September 12, 2021 09:17

yaochunhung closed this Sep 14, 2021

wenliangwu mentioned this pull request Apr 26, 2023

ASoC: SOF: mediatek: add mt8188 audio support #4319

Merged

Add support for Mediatek mt8195 platform #3122

Add support for Mediatek mt8195 platform #3122

Uh oh!

Conversation

yaochunhung commented Aug 26, 2021

Uh oh!

Choose a reason for hiding this comment

Uh oh!

yaochunhung Aug 26, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

plbossart commented Aug 26, 2021

Uh oh!

yaochunhung commented Aug 26, 2021

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

plbossart left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

yaochunhung Sep 9, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ranj063 Sep 9, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

yaochunhung Aug 26, 2021 •

edited

Loading

yaochunhung Sep 9, 2021 •

edited

Loading

ranj063 Sep 9, 2021 •

edited

Loading