[PATCH v6 0/3] ASoC: sun4i-i2s: Updates to the driver

[codek...@gmail.com]

From: Marcus Cooper <codek...@gmail.com>

Hi All,

here is a patch series to add some improvements to the sun4i-i2s driver
which is enough to get HDMI audio working on the A83T, A64, H3 and
H5 platforms.

I've dropped a lot of the functionality that was presented earlier in favour
of getting initial HDMI audio delivered. H6 and multi-channel HDMI will
follow shortly.

My test branch for this can be found at
https://github.com/codekipper/linux-sunxi/commits/upstream-i2s , I've been
using a Pine64 to test with; validating the new SoC block with HDMI audio
and ensuring that I've not broken the old block by making sure that the audio
codec still works.

BR,
CK

---
v6 changes compared to v5 are:
- removed patches for multi-channel and H6 HDMI audio.
- removed patch for 20, 24 and 32 bit (will push support for just 20 and 24bit)
- ditched tdm patches as support has already been added.
- added fix for A83T reg map.

v5 changes compared to v4 are:
- removed delivered patches.
- Added more details to commit messages.
- replaced some reg fields with function calls.
- Added DSP_A and DSP_B support for H3 and later SoCs.
- Added support for the Allwinner H6.

v4 changes compared to v3 are:
- Moved patches around so that the more controversial of patches are
at the top of the stack.
- Added more details to commit messages.
- Fixed 20bit audio PCM format to use 4 bytes.
- Reduced number of flags used to indicate a new SoC.

v3 changes compared to v2 are:
- added back slave mode changes
- added back the use of tdm properties
- changes to regmap and caching
- removed loopback functionality
- fixes to the channel offset mask

v2 changes compared to v1 are:
- removed slave mode changes which didn't set mclk and bclk div.
- removed use of tdm and now use a dedicated property.
- fix commit message to better explain reason for sign extending
- add divider calculations for newer SoCs.
- add support for multi-lane i2s data output.
- add support for 20, 24 and 32 bit samples.
- add loopback property so blocks can be tested without a codec.

---

Marcus Cooper (3):
ASoC: sun4i-i2s: incorrect regmap for A83T
ASoC: sun4i-i2s: Add regmap field to sign extend sample
ASoC: sun4i-i2s: Adjust LRCLK width

sound/soc/sunxi/sun4i-i2s.c | 24 ++++++++++++++++++++++--
1 file changed, 22 insertions(+), 2 deletions(-)

--
2.23.0

[codek...@gmail.com]

From: Marcus Cooper <codek...@gmail.com>

On the newer SoCs such as the H3 and A64 this is set by default
to transfer a 0 after each sample in each slot. However the A10
and A20 SoCs that this driver was developed on had a default
setting where it padded the audio gain with zeros.

This isn't a problem whilst we have only support for 16bit audio
but with larger sample resolution rates in the pipeline then SEXT
bits should be cleared so that they also pad at the LSB. Without
this the audio gets distorted.

Signed-off-by: Marcus Cooper <codek...@gmail.com>
---
sound/soc/sunxi/sun4i-i2s.c | 17 +++++++++++++++++
1 file changed, 17 insertions(+)

diff --git a/sound/soc/sunxi/sun4i-i2s.c b/sound/soc/sunxi/sun4i-i2s.c
index 34575a8aa9f6..056a299c03fb 100644
--- a/sound/soc/sunxi/sun4i-i2s.c
+++ b/sound/soc/sunxi/sun4i-i2s.c
@@ -135,6 +135,7 @@ struct sun4i_i2s;
* @field_clkdiv_mclk_en: regmap field to enable mclk output.
* @field_fmt_wss: regmap field to set word select size.
* @field_fmt_sr: regmap field to set sample resolution.
+ * @field_fmt_sext: regmap field to set the sign extension.
*/
struct sun4i_i2s_quirks {
bool has_reset;
@@ -145,6 +146,7 @@ struct sun4i_i2s_quirks {
struct reg_field field_clkdiv_mclk_en;
struct reg_field field_fmt_wss;
struct reg_field field_fmt_sr;
+ struct reg_field field_fmt_sext;

const struct sun4i_i2s_clk_div *bclk_dividers;
unsigned int num_bclk_dividers;
@@ -177,6 +179,7 @@ struct sun4i_i2s {
struct regmap_field *field_clkdiv_mclk_en;
struct regmap_field *field_fmt_wss;
struct regmap_field *field_fmt_sr;
+ struct regmap_field *field_fmt_sext;

const struct sun4i_i2s_quirks *variant;
};
@@ -354,6 +357,10 @@ static int sun4i_i2s_set_clk_rate(struct snd_soc_dai *dai,

regmap_field_write(i2s->field_clkdiv_mclk_en, 1);

+
+ /* Set sign extension to pad out LSB with 0 */
+ regmap_field_write(i2s->field_fmt_sext, 0);
+
return 0;
}

@@ -1073,6 +1080,7 @@ static const struct sun4i_i2s_quirks sun4i_a10_i2s_quirks = {
.mclk_dividers = sun4i_i2s_mclk_div,
.num_mclk_dividers = ARRAY_SIZE(sun4i_i2s_mclk_div),
.get_bclk_parent_rate = sun4i_i2s_get_bclk_parent_rate,
+ .field_fmt_sext = REG_FIELD(SUN4I_I2S_FMT1_REG, 8, 8),
.get_sr = sun4i_i2s_get_sr,
.get_wss = sun4i_i2s_get_wss,
.set_chan_cfg = sun4i_i2s_set_chan_cfg,
@@ -1091,6 +1099,7 @@ static const struct sun4i_i2s_quirks sun6i_a31_i2s_quirks = {
.mclk_dividers = sun4i_i2s_mclk_div,
.num_mclk_dividers = ARRAY_SIZE(sun4i_i2s_mclk_div),
.get_bclk_parent_rate = sun4i_i2s_get_bclk_parent_rate,
+ .field_fmt_sext = REG_FIELD(SUN4I_I2S_FMT1_REG, 8, 8),
.get_sr = sun4i_i2s_get_sr,
.get_wss = sun4i_i2s_get_wss,
.set_chan_cfg = sun4i_i2s_set_chan_cfg,
@@ -1109,6 +1118,7 @@ static const struct sun4i_i2s_quirks sun8i_a83t_i2s_quirks = {
.mclk_dividers = sun8i_i2s_clk_div,
.num_mclk_dividers = ARRAY_SIZE(sun8i_i2s_clk_div),
.get_bclk_parent_rate = sun8i_i2s_get_bclk_parent_rate,
+ .field_fmt_sext = REG_FIELD(SUN4I_I2S_FMT1_REG, 4, 5),
.get_sr = sun8i_i2s_get_sr_wss,
.get_wss = sun8i_i2s_get_sr_wss,
.set_chan_cfg = sun8i_i2s_set_chan_cfg,
@@ -1127,6 +1137,7 @@ static const struct sun4i_i2s_quirks sun50i_a64_codec_i2s_quirks = {
.mclk_dividers = sun4i_i2s_mclk_div,
.num_mclk_dividers = ARRAY_SIZE(sun4i_i2s_mclk_div),
.get_bclk_parent_rate = sun4i_i2s_get_bclk_parent_rate,
+ .field_fmt_sext = REG_FIELD(SUN4I_I2S_FMT1_REG, 8, 8),
.get_sr = sun4i_i2s_get_sr,
.get_wss = sun4i_i2s_get_wss,
.set_chan_cfg = sun4i_i2s_set_chan_cfg,
@@ -1154,6 +1165,12 @@ static int sun4i_i2s_init_regmap_fields(struct device *dev,
if (IS_ERR(i2s->field_fmt_sr))
return PTR_ERR(i2s->field_fmt_sr);

+ i2s->field_fmt_sext =
+ devm_regmap_field_alloc(dev, i2s->regmap,
+ i2s->variant->field_fmt_sext);
+ if (IS_ERR(i2s->field_fmt_sext))
+ return PTR_ERR(i2s->field_fmt_sext);
+
return 0;
}

--
2.23.0

[codek...@gmail.com]

From: Marcus Cooper <codek...@gmail.com>

The regmap configuration is set up for the legacy block on the
A83T whereas it uses the new block with a larger register map.

Fixes: 21faaea1343f ("ASoC: sun4i-i2s: Add support for A83T")

Signed-off-by: Marcus Cooper <codek...@gmail.com>
---

sound/soc/sunxi/sun4i-i2s.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/sound/soc/sunxi/sun4i-i2s.c b/sound/soc/sunxi/sun4i-i2s.c
index 57bf2a33753e..34575a8aa9f6 100644
--- a/sound/soc/sunxi/sun4i-i2s.c
+++ b/sound/soc/sunxi/sun4i-i2s.c
@@ -1100,7 +1100,7 @@ static const struct sun4i_i2s_quirks sun6i_a31_i2s_quirks = {

static const struct sun4i_i2s_quirks sun8i_a83t_i2s_quirks = {

.has_reset = true,
.reg_offset_txdata = SUN8I_I2S_FIFO_TX_REG,
- .sun4i_i2s_regmap = &sun4i_i2s_regmap_config,
+ .sun4i_i2s_regmap = &sun8i_i2s_regmap_config,
.field_clkdiv_mclk_en = REG_FIELD(SUN4I_I2S_CLK_DIV_REG, 8, 8),
.field_fmt_wss = REG_FIELD(SUN4I_I2S_FMT0_REG, 0, 2),
.field_fmt_sr = REG_FIELD(SUN4I_I2S_FMT0_REG, 4, 6),
--
2.23.0

[codek...@gmail.com]

From: Marcus Cooper <codek...@gmail.com>

Some codecs such as i2s based HDMI audio and the Pine64 DAC require
a different amount of bit clocks per frame than what is calculated
by the sample width. Use the values obtained by the tdm slot bindings
to adjust the LRCLK width accordingly.

Signed-off-by: Marcus Cooper <codek...@gmail.com>
---

sound/soc/sunxi/sun4i-i2s.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/sound/soc/sunxi/sun4i-i2s.c b/sound/soc/sunxi/sun4i-i2s.c
index 056a299c03fb..0965a97c96e5 100644
--- a/sound/soc/sunxi/sun4i-i2s.c
+++ b/sound/soc/sunxi/sun4i-i2s.c
@@ -455,7 +455,10 @@ static int sun8i_i2s_set_chan_cfg(const struct sun4i_i2s *i2s,
break;

case SND_SOC_DAIFMT_I2S:
- lrck_period = params_physical_width(params);
+ if (i2s->slot_width)
+ lrck_period = i2s->slot_width;
+ else
+ lrck_period = params_physical_width(params);
break;

default:
--
2.23.0

[Chen-Yu Tsai]

On Tue, Aug 27, 2019 at 2:07 AM <codek...@gmail.com> wrote:
>
> From: Marcus Cooper <codek...@gmail.com>
>
> The regmap configuration is set up for the legacy block on the
> A83T whereas it uses the new block with a larger register map.

Looking at the code Allwinner previously released [1], that doesn't seem to be
the case. Keep in mind that the register map shown in the user manual is for
the TDM interface, which we don't actually support right now.

The file shows the base address as 0x01c22800, and the last defined register
is SUNXI_RXCHMAP at 0x3c.

The I2S driver [2] also shows that it is the old register map size, but with
TX_FIFO and INT_STA swapped around. This might mean that it would need a
separate regmap_config, as the read/write callbacks need to be changed to
fit the swapped registers.

Finally, the TDM driver [3], which matches the TDM section in the manual, shows
a larger register map.

A83T is SUN8IW6, while SUN8IW7 refers to the H3.

ChenYu

[1] https://github.com/allwinner-zh/linux-3.4-sunxi/blob/master/sound/soc/sunxi/hdmiaudio/sunxi-hdmipcm.h
[2] https://github.com/allwinner-zh/linux-3.4-sunxi/blob/master/sound/soc/sunxi/i2s0/sunxi-i2s0.h
[3] https://github.com/allwinner-zh/linux-3.4-sunxi/blob/master/sound/soc/sunxi/daudio0/sunxi-daudio0.h

> Fixes: 21faaea1343f ("ASoC: sun4i-i2s: Add support for A83T")
> Signed-off-by: Marcus Cooper <codek...@gmail.com>
> ---
> sound/soc/sunxi/sun4i-i2s.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/sound/soc/sunxi/sun4i-i2s.c b/sound/soc/sunxi/sun4i-i2s.c
> index 57bf2a33753e..34575a8aa9f6 100644
> --- a/sound/soc/sunxi/sun4i-i2s.c
> +++ b/sound/soc/sunxi/sun4i-i2s.c
> @@ -1100,7 +1100,7 @@ static const struct sun4i_i2s_quirks sun6i_a31_i2s_quirks = {
> static const struct sun4i_i2s_quirks sun8i_a83t_i2s_quirks = {
> .has_reset = true,
> .reg_offset_txdata = SUN8I_I2S_FIFO_TX_REG,
> - .sun4i_i2s_regmap = &sun4i_i2s_regmap_config,
> + .sun4i_i2s_regmap = &sun8i_i2s_regmap_config,
> .field_clkdiv_mclk_en = REG_FIELD(SUN4I_I2S_CLK_DIV_REG, 8, 8),
> .field_fmt_wss = REG_FIELD(SUN4I_I2S_FMT0_REG, 0, 2),
> .field_fmt_sr = REG_FIELD(SUN4I_I2S_FMT0_REG, 4, 6),
> --
> 2.23.0
>

> --
> You received this message because you are subscribed to the Google Groups "linux-sunxi" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to linux-sunxi...@googlegroups.com.
> To view this discussion on the web, visit https://groups.google.com/d/msgid/linux-sunxi/20190826180734.15801-2-codekipper%40gmail.com.

[Code Kipper]

On Tue, 27 Aug 2019 at 06:13, Chen-Yu Tsai <we...@csie.org> wrote:
>
> On Tue, Aug 27, 2019 at 2:07 AM <codek...@gmail.com> wrote:
> >
> > From: Marcus Cooper <codek...@gmail.com>
> >
> > The regmap configuration is set up for the legacy block on the
> > A83T whereas it uses the new block with a larger register map.
>
> Looking at the code Allwinner previously released [1], that doesn't seem to be
> the case. Keep in mind that the register map shown in the user manual is for
> the TDM interface, which we don't actually support right now.

Should it matter what we support right now?, the block according to the user
manual shows the bigger range. I don't have a A83T device and from what I
gather not many users do. But the compatible for the H3 has been removed
and replaced with the settings for the A83T which also has default settings in
registers further up than SUNXI_RXCHMAP.

>
> The file shows the base address as 0x01c22800, and the last defined register
> is SUNXI_RXCHMAP at 0x3c.
>
> The I2S driver [2] also shows that it is the old register map size, but with
> TX_FIFO and INT_STA swapped around. This might mean that it would need a
> separate regmap_config, as the read/write callbacks need to be changed to
> fit the swapped registers.
>
> Finally, the TDM driver [3], which matches the TDM section in the manual, shows
> a larger register map.
>
> A83T is SUN8IW6, while SUN8IW7 refers to the H3.

Since when have we trusted Allwinner code?, the TDM labelled block
clearly supports
I2S. The biggest use case for this block is getting HDMI audio working
on the newer
devices(LibreELEC nightlies has a user base of over 300) and I've tested this on
numerous set ups over the last couple of years.

Failing that reverting (3e9acd7ac693: "ASoC: sun4i-i2s: Remove
duplicated quirks structure")
would help.

BR,
CK

[Chen-Yu Tsai]

On Tue, Aug 27, 2019 at 1:55 PM Code Kipper <codek...@gmail.com> wrote:
>
> On Tue, 27 Aug 2019 at 06:13, Chen-Yu Tsai <we...@csie.org> wrote:
> >
> > On Tue, Aug 27, 2019 at 2:07 AM <codek...@gmail.com> wrote:
> > >
> > > From: Marcus Cooper <codek...@gmail.com>
> > >
> > > The regmap configuration is set up for the legacy block on the
> > > A83T whereas it uses the new block with a larger register map.
> >
> > Looking at the code Allwinner previously released [1], that doesn't seem to be
> > the case. Keep in mind that the register map shown in the user manual is for
> > the TDM interface, which we don't actually support right now.
>
> Should it matter what we support right now?, the block according to the user
> manual shows the bigger range. I don't have a A83T device and from what I

There are a total of four I2S controllers on the A83T. Currently three of them
are listed in the dtsi file, which are _not_ the one shown in the user manual.
The one shown is the fourth one, which is the TDM controller.

It's not like we haven't seen this before. IIRC the A64 also had two variants
of the I2S interface. The one coupled with the audio codec was different from
the others.

> gather not many users do. But the compatible for the H3 has been removed
> and replaced with the settings for the A83T which also has default settings in
> registers further up than SUNXI_RXCHMAP.

I'll sync up with Maxime on this.

> >
> > The file shows the base address as 0x01c22800, and the last defined register
> > is SUNXI_RXCHMAP at 0x3c.
> >
> > The I2S driver [2] also shows that it is the old register map size, but with
> > TX_FIFO and INT_STA swapped around. This might mean that it would need a
> > separate regmap_config, as the read/write callbacks need to be changed to
> > fit the swapped registers.
> >
> > Finally, the TDM driver [3], which matches the TDM section in the manual, shows
> > a larger register map.
> >
> > A83T is SUN8IW6, while SUN8IW7 refers to the H3.
>
> Since when have we trusted Allwinner code?, the TDM labelled block
> clearly supports

Since they haven't listed the I2S block in the user manual, so that is what we
have to go by.

The TDM section in the user manual only lists the block at 0x1c23000. The memory
map says DAUDIO-[012] for addresses 0x1c22000, 0x1c22400, 0x1c22800, and TDM for
address 0x1c23000. One would assume this meant these are somewhat different.

> I2S. The biggest use case for this block is getting HDMI audio working
> on the newer

I understand that.

> devices(LibreELEC nightlies has a user base of over 300) and I've tested this on
> numerous set ups over the last couple of years.

Tested on the H3, correct?

> Failing that reverting (3e9acd7ac693: "ASoC: sun4i-i2s: Remove
> duplicated quirks structure")
> would help.

I'll take a look. IIRC it worked with the old layout, with the two registers
swapped, playing standard 48 KHz / 16 bit audio when I added supported for
the A83T. Then again maybe the stars were perfectly aligned. At the very least
we could separate A83T and H3 as you suggested.

ChenYu

> To view this discussion on the web, visit https://groups.google.com/d/msgid/linux-sunxi/CAEKpxBmCg4AkqKM-O3C76gto%2BmPWyEdDbviAmRJ8PxLOOMTJ7w%40mail.gmail.com.

[Code Kipper]

On Tue, 27 Aug 2019 at 10:01, Chen-Yu Tsai <we...@csie.org> wrote:
>
> On Tue, Aug 27, 2019 at 1:55 PM Code Kipper <codek...@gmail.com> wrote:
> >
> > On Tue, 27 Aug 2019 at 06:13, Chen-Yu Tsai <we...@csie.org> wrote:
> > >
> > > On Tue, Aug 27, 2019 at 2:07 AM <codek...@gmail.com> wrote:
> > > >
> > > > From: Marcus Cooper <codek...@gmail.com>
> > > >
> > > > The regmap configuration is set up for the legacy block on the
> > > > A83T whereas it uses the new block with a larger register map.
> > >
> > > Looking at the code Allwinner previously released [1], that doesn't seem to be
> > > the case. Keep in mind that the register map shown in the user manual is for
> > > the TDM interface, which we don't actually support right now.
> >
> > Should it matter what we support right now?, the block according to the user
> > manual shows the bigger range. I don't have a A83T device and from what I
>
> There are a total of four I2S controllers on the A83T. Currently three of them
> are listed in the dtsi file, which are _not_ the one shown in the user manual.
> The one shown is the fourth one, which is the TDM controller.

The configuration for the A83T suggests that it's a mixture of old and
new which I don't
think is the case considering it was released around the same time as
the H3. There
is enough similarity between the blocks for it to still work. For
example on the H6
we referenced by mistake the H3 block and we still got audio (with
only slight distortion).
I would suggest to validate all of the i2s blocks we need to test
using the internal loopback
as that will also cover capture.

>
> It's not like we haven't seen this before. IIRC the A64 also had two variants
> of the I2S interface. The one coupled with the audio codec was different from
> the others.

Yes...but the i2s of the audio codec was documented in the audio codec
section. I've used
this device to ensure that I've not broken anything in the old block
with these new changes.

Yes....but only with the additional changes for multi-channel with my
LibreELEC build.
These changes I tested on my pine64 before pushing upstream.

>
> > Failing that reverting (3e9acd7ac693: "ASoC: sun4i-i2s: Remove
> > duplicated quirks structure")
> > would help.
>
> I'll take a look. IIRC it worked with the old layout, with the two registers
> swapped, playing standard 48 KHz / 16 bit audio when I added supported for
> the A83T. Then again maybe the stars were perfectly aligned. At the very least
> we could separate A83T and H3 as you suggested.

Thanks,
CK

[Code Kipper]

On Tue, 27 Aug 2019 at 09:01, Maxime Ripard <mri...@kernel.org> wrote:

> That would be the case with the DSP formats too, right?

Maybe....but I need a TDM test volunteer!,
CK
>
> Maxime
>
> --
> Maxime Ripard, Bootlin
> Embedded Linux and Kernel engineering
> https://bootlin.com

[Chen-Yu Tsai]

On Tue, Aug 27, 2019 at 4:35 PM Code Kipper <codek...@gmail.com> wrote:
>
> On Tue, 27 Aug 2019 at 10:01, Chen-Yu Tsai <we...@csie.org> wrote:
> >
> > On Tue, Aug 27, 2019 at 1:55 PM Code Kipper <codek...@gmail.com> wrote:
> > >
> > > On Tue, 27 Aug 2019 at 06:13, Chen-Yu Tsai <we...@csie.org> wrote:
> > > >
> > > > On Tue, Aug 27, 2019 at 2:07 AM <codek...@gmail.com> wrote:
> > > > >
> > > > > From: Marcus Cooper <codek...@gmail.com>
> > > > >
> > > > > The regmap configuration is set up for the legacy block on the
> > > > > A83T whereas it uses the new block with a larger register map.
> > > >
> > > > Looking at the code Allwinner previously released [1], that doesn't seem to be
> > > > the case. Keep in mind that the register map shown in the user manual is for
> > > > the TDM interface, which we don't actually support right now.
> > >
> > > Should it matter what we support right now?, the block according to the user
> > > manual shows the bigger range. I don't have a A83T device and from what I
> >
> > There are a total of four I2S controllers on the A83T. Currently three of them
> > are listed in the dtsi file, which are _not_ the one shown in the user manual.
> > The one shown is the fourth one, which is the TDM controller.
>
> The configuration for the A83T suggests that it's a mixture of old and
> new which I don't
> think is the case considering it was released around the same time as
> the H3. There
> is enough similarity between the blocks for it to still work. For
> example on the H6
> we referenced by mistake the H3 block and we still got audio (with
> only slight distortion).

The difference with the A83T here is large enough that if you play anything
it will simply stall. I already reported it as broken and Maxime has sent
fixes.

Maxime has sent a patch reverting the merger.

ChenYu

[Code Kipper]

On Tue, 27 Aug 2019 at 11:34, Maxime Ripard <mri...@kernel.org> wrote:

>
> On Mon, Aug 26, 2019 at 08:07:33PM +0200, codek...@gmail.com wrote:
> > From: Marcus Cooper <codek...@gmail.com>
> >
> > On the newer SoCs such as the H3 and A64 this is set by default
> > to transfer a 0 after each sample in each slot. However the A10
> > and A20 SoCs that this driver was developed on had a default
> > setting where it padded the audio gain with zeros.
> >
> > This isn't a problem whilst we have only support for 16bit audio
> > but with larger sample resolution rates in the pipeline then SEXT
> > bits should be cleared so that they also pad at the LSB. Without
> > this the audio gets distorted.
> >
> > Signed-off-by: Marcus Cooper <codek...@gmail.com>
>

> If anything, I'd like to have less regmap_fields rather than more of
> them. This is pretty easy to add to one of the callbacks, especially
> since the field itself has been completely reworked from one
> generation to the other.
>
ACK
That's fine....I've been doing that with the patches which follow this.