How to use multiple ALSA devices from shell ?

Hi Pavel,

I am using tlv320aic3x codec.

Attached are the files I'm using.

It is configured now as 2 codecs (devices) in the same card (this is also what is stated in the proting wiki guide in TI's as how to deal with more than one codec).

According to some threads it seems that the failure I'm dealing with codec relates more to problem in alsa driver when registring the controls of codec. 

I wander if seperating the 2 codec to different cards (one codec for each crad) might resolve this naming issue.

0456.ti81xx-evm.c

2671.devices.c

0020.board-ti8148evm.c

Regards,

Ran

I am using this configuration:
======================

card0:

device 0:
McASP2 <-----> AIC3x

device1:
McASP3 <-----> AIC3x (other codec)

I hope that changing it to the following configuration might reoslve this issue (but haven't tried it yet):

suggested configuration:

============
card0:

device 0:
McASP2 <-----> AIC3x

card1:

device0:
McASP3 <-----> AIC3x (other codec)


Regards,
Ran

Ran Shalit said:

I wander if seperating the 2 codec to different cards (one codec for each crad) might resolve this naming issue.

In ti81xx-evm.c you have:

.codec_name = "tlv320aic3x-codec.2-0019",

This means this aic should be attached to I2C1/0x4802A000 at address 0x19. Could you please double check this?

The codec_name is made up of the name used in the codec driver along with the I2C bus and I2C address (1-0018 for the EVM):

.codec_name = "tlv320aic3x-codec.1-0018",

BR
Pavel

In board-ti8148evm.c you have only one ti81xx_register_mcasp() function invocation, while for register two McASPs, you should have (for example):

ti81xx_register_mcasp(0, &aic3x_snd_data); //McASP2

ti81xx_register_mcasp(1, &aic3x_snd_data_elbit); //McASP3

I see other possible mistakes also. Please refer to my files (davinci-evm.c, board-ti8148evm.c and devices.c) as a reference, where I enable McASP0 and register it in card 0, device 1. I have no error messages during linux kernel boot up. I can not test McASP0 as I have no audio codec attached to it.

5466.davinci-evm.c

2500.board-ti8148evm.c

0842.devices.c

BR
Pavel

I also suspect that this issue might be related to prefix_name bug in alsa code.

According to following thread:

This might be the reason for the failure.

I hope that seperating the devices into 2 codecs, will solve this issue for me, so that I won't need to apply the given patch (in that thread),

Thanks,

Ran

Ran Shalit said:

1. I did notice that in the code you've attached the codec name is identical for both codecs ("tlv320aic3x-codec.1-0018"), when using 2 devices I think that each should have his own codec:

Agree. I made it just as an example how to add additional McASP instance (McASP0). I do not modify the codec name, as I have no audio codec (or anything) attached to McASP0 in the DM814x TI EVM.

Ran Shalit said:

2. As to the mcasp registration, I'm not sure this is it, becuase I see 2 registration, the format is a bit different then given in your message:

platform_device_register(&ti81xx_mcasp_aic_device);

platform_device_register(&ti81xx_mcasp_aic_device_elbit);

I would recommend you to try also with the other approach (provided in my files) to see if you will have any improvement.

BR
Pavel

Hi Pavel,



When working with 2 cards instead of 2 device at the same card, I get no error for this issue of 'ALSA control name already busy'.
I guess it is some ALSA bug. I think that working in such configuration should be good enough , Right? I rather do this simple fix instead of applying the prefix_name patch.

This is the boot log with this configuration:

...

usbcore: registered new interface driver snd-usb-audio
Registered tvp5158 audio codec
PM: Adding info for platform:soc-audio.0
PM: Adding info for No Bus:TLV320AIC3X_2
asoc: tlv320aic3x-hifi <-> davinci-mcasp.2 mapping ok
PM: Adding info for No Bus:card0
PM: Adding info for No Bus:pcmC0D0p
PM: Adding info for No Bus:pcmC0D0c
PM: Adding info for No Bus:controlC0
PM: Adding info for platform:soc-audio.1
snd_soc_instantiate_card 4
PM: Adding info for No Bus:TLV320AIC3X_3
asoc: tlv320aic3x-hifi <-> davinci-mcasp.3 mapping ok
PM: Adding info for No Bus:card1
PM: Adding info for No Bus:pcmC1D0p
PM: Adding info for No Bus:pcmC1D0c
PM: Adding info for No Bus:controlC1
ALSA device list:
#0: TI81XX SOUND0
#1: TI81XX SOUND1
TCP cubic registered
NET: Registered protocol family 17

Regards,

Ran

Ran,

Ran Shalit said:

I think that working in such configuration should be good enough , Right?

Are you able to use aplay with McASP2/3 correctly? If yes, then I think this configuration should be also fine. This configuration is used for other newer devices, like OMAP5, see the below pointers:

Could you please also share the final versions of your files (devices.c, board-ti8148evm.c, ti81xx-evm.c)? And have you made changes in other files?

BR
Pavel

Hi Pavel,

On trying to do aplay, arecord, I get success with mcasp2 , card0, but not with mcasp3, card1.

For now I just try to call this utilities, not yet checking the sound. With card0, mcasp2, the aplay, arecord return and no error is printed.

With mcasp3, card1, it hangs, and does not return. I don't know the internal of this utilities, but it seems that it wait for some events.

I had to make to changed for this to function without errors (but it still hangs):

1. At first it made some errors, but after removing the reserved dma for dma 56-, It continue without errors.

2. asp.h (also attached) was added with dma events for mcasp3 (it was missing in that file):

asp.h

====

added line:

#define TI81XX_ASP3_BASE 0x4A1A2000

/* EDMA channels of ti81xx McASP3 */
#define TI81XX_DMA_MCASP3_AXEVT 56
#define TI81XX_DMA_MCASP3_AREVT 57

devices.c

==========

ti81xx_mcasp_aic_device_elbit.id = 3;
ti81xx_mcasp_aic_device_elbit.resource = ti81xx_mcasp3_resource;
ti81xx_mcasp_aic_device_elbit.num_resources = ARRAY_SIZE(ti81xx_mcasp3_resource);

platform_device_register(&ti81xx_mcasp_aic_device_elbit);

platform_device_register(&ti81xx_mcasp_aic_device);

static struct resource ti81xx_mcasp3_resource[] = {
{
.name = "mcasp",
.start = TI81XX_ASP3_BASE,
.end = TI81XX_ASP3_BASE + (SZ_1K * 12) - 1,
.flags = IORESOURCE_MEM,
},
/* TX event */
{
.start = TI81XX_DMA_MCASP3_AXEVT,
.end = TI81XX_DMA_MCASP3_AXEVT,
.flags = IORESOURCE_DMA,
},
/* RX event */
{
.start = TI81XX_DMA_MCASP3_AREVT,
.end = TI81XX_DMA_MCASP3_AREVT,
.flags = IORESOURCE_DMA,
},
};

static const s16 ti814x_dma_rsv_chans[][2] = {
/* (offset, number) */
{0, 2},
{14, 2},
{26, 6},
{48, 4},
//ranran {56, 8},
{-1, -1}
};

static const s16 ti814x_dma_rsv_slots[][2] = {
/* (offset, number) */
{0, 2},
{14, 2},
{26, 6},
{48, 4},
//ranran {56, 8},
{64, 127},
{248, 264},
{-1, -1}
};

#ifdef CONFIG_ELBIT_BOARD
/* McASP3 <-> AIC3101 */
static struct snd_platform_data aic3x_snd_data_elbit = {
.tx_dma_offset = 0x4A1A5000,
.rx_dma_offset = 0x4A1A5000,
.op_mode = DAVINCI_MCASP_IIS_MODE,
.num_serializer = ARRAY_SIZE(aic3x_iis_serializer_direction_elbit),
.tdm_slots = 2,
.serial_dir = aic3x_iis_serializer_direction_elbit,
.asp_chan_q = EVENTQ_0,
.version = MCASP_VERSION_2,
.txnumevt = 32,
.rxnumevt = 32,
/* McASP21_AHCLKX out to feed CODEC CLK*/
.clk_input_pin = MCASP_AHCLKX_OUT,
};
#endif

1106.shared.zip

I think that the issue now with mcasp3, is not related to the device/cards, but more to DMA or event.

I don't yet understand why it hangs with mcasp3...

Thank you very much,

Ran

Hi Pavel,

It still does not work , I suspect it some kind of HW issue ?
It Probably means that there is wrong HW connectivity in the reciever side between codec->dma->mcasp3 , Right ?

If this is the case, I better focus now on the problem with the more functional device (mcasp2).
The speed issue (the speed in the recorded file is faster then the original captured audio): What can be wrong here ? As far as I understand we work with oscillator 24.5, which is the same as EVM, so I don't yet understand why it works fine in EVM, but not in the custom board.
On opening the recorded file in audacity I see that sample rate is always 44100 (according to audacity and vlc info), even when arecord samples it in 8000. The thigs is that it sound correct in evaluation board but in custom board it sounf faster (speed is faster in rate of ~44.1/36 according to my testing).


Regards,
Ran

Ran,

Ran Shalit said:

It still does not work , I suspect it some kind of HW issue ?
It Probably means that there is wrong HW connectivity in the reciever side between codec->dma->mcasp3 , Right ?

If you suspect HW malfunction, I will recommend you to check your HW with the McASP diagnostic test, it is for DM814x EVM, so it might required some adjustment for you custom board:

aic3106/

This CCS test application configures the McASP2 on the DM814X EVM board and also the AIC3106 for the following.
1. Playback on head phones
2. Playback on Line-Out
3. Record from Microphone and playback on Head-phones
4. Record from Line-In and playback on Head-phones

The test application executes in the above mentioned sequence. The Audio codec is configured to record and playback at 48 KHz sampling. The Codec is kept as I2S master with the Bit-clock & Frame-sync being input the McASP.

You can also try to run aplay/arecord with --verbose option, to check if there will be some informative message before the hang.

Note also that by default audio driver works in McASPx slave, codec master format.

BR
Pavel

Ran Shalit said:

If this is the case, I better focus now on the problem with the more functional device (mcasp2).
The speed issue (the speed in the recorded file is faster then the original captured audio): What can be wrong here ? As far as I understand we work with oscillator 24.5, which is the same as EVM, so I don't yet understand why it works fine in EVM, but not in the custom board.
On opening the recorded file in audacity I see that sample rate is always 44100 (according to audacity and vlc info), even when arecord samples it in 8000. The thigs is that it sound correct in evaluation board but in custom board it sounf faster (speed is faster in rate of ~44.1/36 according to my testing).

I assume you do not refer OSC or AUX but the oscillator providing the clock to the MCLK input of the AIC codec, the frequency there is 24.5760MHz.

I would suggest to measure the frequencies of BCLK, WCLK and MCLK and compare the values between EVM and your custom board.

BR
Pavel

Hi Pavel,

Thank you very much for this assistance.
when we checked the schema of the evm board we see that the oscilator for MCLK is not 24.576Mhz but something else (I think 22.579MHz ), so maybe that explain the issue? The strange thing is that I see in code( Ti81xx_evm.c )that sysclk is explicitly written as 24576000.

Is it correct that evm oscillator is not 24.5Mhz but other value?
Are there any dividers which make 24576000 sysclk from that 26Mhz ? Can this explain the speed issue?

Many Thanks,
Ran

Another thing I would like to ask:

Is this the configuration in evm:
MCA2_ACLKX <------------ AIC_BCLK
MCA2_AFSX <------------ AIC_WCLK

MCA2_AHCLKX -----------> AIC_MCLK
MCA2_AXR0 ------------> AIC_DIN
MCA2_AXR1 <------------ AIC_DOUT

If mcasp2 drives out the MCLK, why is it said that the codec is the master ?

Thank you,
Ran