Hello Caleb,
I go through all [few] patchs we apply to the 4.0 linux tree (didn't
jump to 4.2 yet).
There is one concerning the DMA firmware you can find in the freescale
tree, available at git://git.freescale.com/imx/linux-2.6-imx.git
commit 619bfca89908b90cd6606ed894c180df0c481508
Author: Shawn Guo <shawn.guo@xxxxxxxxxxxxx>
Date: Tue Jul 16 22:53:18 2013 +0800
ENGR00269945: firwmare: imx: add imx6q sdma script
Add imx6q sdma script which will be used by all i.MX6 series.
Signed-off-by: Shawn Guo <shawn.guo@xxxxxxxxxxxxx>
firmware/Makefile | 1 +
firmware/imx/sdma/sdma-imx6q.bin.ihex | 116
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
2 files changed, 117 insertions(+)
I don't know how 4.3rcN catch some of the freescale patch. but you may
check if you need to apply this one.
I know there are some other SDMA firmware available all around in this
freescale tree. I didn't find proper release note or documentation. May
be you can test them also.
Arnaud
Le 19/10/2015 17:55, Caleb Crome a écrit :
Hello Arnaud, all,
I'm trying to get the i.MX6 SSI working in 16-channel TDM mode. FYI,
I'm working with the stock 4.3 kernel at
https://github.com/torvalds/linux.
When I apply the patch below, the SSI does all configure properly and
even starts streaming properly with all the bits in the right place on
the SSI pins. However, given a little bit of time and/or IO to the SD
card, the bit-stream slips by 1 slot, causing all of the channels to
be misaligned.
My changes to the SSI driver are very minimal (shown below), and
amount to forcing it into network (TDM) mode, and changing the maximum
channels, and setting the STCCR DC mask.
There is no indication from user space that anything has slipped, so
the data stream just continues on shifted by 1 slot.
You (Arnaud) mentioned in a previous thread ("Multiple codecs on one
sound card for multi-channel sound card"), that I should just have to
set channels_max (and presumably the other changes I mentioned), and
it'll work mostly. However, it's very unreliable at the moment.
Any thoughts to how I can diagnose this problem would be greatly
appreciated!
So, what happens is this:
The SSI Starts sending data like this:
SLOT 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
DATA 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
But then after time, something slips and without warning it goes to:
SLOT 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
DATA 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
diff --git a/sound/soc/fsl/fsl_ssi.c b/sound/soc/fsl/fsl_ssi.c
index 37c5cd4..73778c2 100644
--- a/sound/soc/fsl/fsl_ssi.c
+++ b/sound/soc/fsl/fsl_ssi.c
@@ -749,7 +749,10 @@ static int fsl_ssi_hw_params(struct
snd_pcm_substream *substream,
CCSR_SSI_SCR_NET | CCSR_SSI_SCR_I2S_MODE_MASK,
channels == 1 ? 0 : i2smode);
}
-
+ ssi_private->i2s_mode = CCSR_SSI_SCR_I2S_MODE_NORMAL |
CCSR_SSI_SCR_NET;
+ regmap_update_bits(regs, CCSR_SSI_SCR,
+ CCSR_SSI_SCR_NET | CCSR_SSI_SCR_I2S_MODE_MASK,
+ ssi_private->i2s_mode);
/*
* FIXME: The documentation says that SxCCR[WL] should not be
* modified while the SSI is enabled. The only time this can
@@ -863,6 +866,15 @@ static int _fsl_ssi_set_dai_fmt(struct device *dev,
return -EINVAL;
}
scr |= ssi_private->i2s_mode;
+ // Set to 16 slots/frame
+ regmap_update_bits(regs, CCSR_SSI_STCCR,
+ CCSR_SSI_SxCCR_DC_MASK,
+ CCSR_SSI_SxCCR_DC(16));
+
+ regmap_update_bits(regs, CCSR_SSI_SRCCR,
+ CCSR_SSI_SxCCR_DC_MASK,
+ CCSR_SSI_SxCCR_DC(16));
+
/* DAI clock inversion */
switch (fmt & SND_SOC_DAIFMT_INV_MASK) {
@@ -1084,14 +1099,14 @@ static struct snd_soc_dai_driver
fsl_ssi_dai_template = {
.playback = {
.stream_name = "CPU-Playback",
.channels_min = 1,
- .channels_max = 2,
+ .channels_max = 16,
.rates = FSLSSI_I2S_RATES,
.formats = FSLSSI_I2S_FORMATS,
},
.capture = {
.stream_name = "CPU-Capture",
.channels_min = 1,
- .channels_max = 2,
+ .channels_max = 16,
.rates = FSLSSI_I2S_RATES,
.formats = FSLSSI_I2S_FORMATS,
},
Another thing I have tried is changing the watermark level for the
fifo to give the DMA interrupt some extra time. The problem still
happens, but seems to be a bit better.
The fifo watermark change is this:
diff --git a/sound/soc/fsl/fsl_ssi.c b/sound/soc/fsl/fsl_ssi.c
index 73778c2..7c2e4b0 100644
--- a/sound/soc/fsl/fsl_ssi.c
+++ b/sound/soc/fsl/fsl_ssi.c
@@ -54,6 +54,8 @@
#include "fsl_ssi.h"
#include "imx-pcm.h"
+#define WATERMARK 8
+
/**
* FSLSSI_I2S_RATES: sample rates supported by the I2S
*
@@ -943,7 +950,7 @@ static int _fsl_ssi_set_dai_fmt(struct device *dev,
* size.
*/
if (ssi_private->use_dma)
- wm = ssi_private->fifo_depth - 2;
+ wm = ssi_private->fifo_depth - WATERMARK;
else
wm = ssi_private->fifo_depth;
@@ -1260,8 +1267,8 @@ static int fsl_ssi_imx_probe(struct
platform_device *pdev,
* We have burstsize be "fifo_depth - 2" to match the SSI
* watermark setting in fsl_ssi_startup().
*/
- ssi_private->dma_params_tx.maxburst = ssi_private->fifo_depth - 2;
- ssi_private->dma_params_rx.maxburst = ssi_private->fifo_depth - 2;
+ ssi_private->dma_params_tx.maxburst = ssi_private->fifo_depth -
WATERMARK;
+ ssi_private->dma_params_rx.maxburst = ssi_private->fifo_depth -
WATERMARK;
ssi_private->dma_params_tx.addr = ssi_private->ssi_phys +
CCSR_SSI_STX0;
ssi_private->dma_params_rx.addr = ssi_private->ssi_phys +
CCSR_SSI_SRX0;
Thanks,
-Caleb
_______________________________________________
Alsa-devel mailing list
Alsa-devel@xxxxxxxxxxxxxxxx
http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
