ESP32 Forum

It looks like there is serious bug in spi-master receive when DMA is used.
I was testing it with ILI9341 display with application based on spi-master example (the example itself works ok).
I've added the function to read line of data from display. Display data format (when read) is:
dummy_byte [R_byte G_byte B_byte] ........ [R_byte G_byte B_byte]
The screen was filled with blue color, so the response should be: 00 00 00 fc 00 00 fc 00 00 fc .....

Here are results of testing:
The first column is sequential number of read, first (dummy) byte is not shown, first 10 RGB values are shown (30 bytes)

When reading less than 33 bytes (no DMA used), the result is as expected, both in duplex and in half duplex mode:

[ 0] Rd line: 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc
[ 1] Rd line: 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc
[ 2] Rd line: 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc
[ 3] Rd line: 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc

When reading more than 32 bytes (DMA used) in duplex mode, no values are received:

[ 0] Rd line: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
[ 1] Rd line: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
[ 2] Rd line: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
[ 3] Rd line: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

When reading more than 32 bytes (DMA used) in half duplex mode (.flags=SPI_DEVICE_HALFDUPLEX), the first read is OK, but for every following read values are shifted by 3 bytes (3, 6, 9, ...) :

[ 0] Rd line: 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc
[ 1] Rd line: 00 80 00 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc
[ 2] Rd line: 00 fc 00 00 28 00 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc
[ 3] Rd line: 00 fc 00 00 fc 00 00 10 00 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc

--------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Nothing can reset this, even after spi_bus_remove_device, spi_bus_free -> spi_bus_initialize, spi_bus_add_device
and reinitializing the display, the receive buffer is still shifted!
Only the first read after booting is correct.
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------

-----------------------------------------------------------------------------------------------------------------------------------------------
With a little change in spi_master.c:
host->hw->dma_conf.val |= SPI_OUT_RST|SPI_AHBM_RST|SPI_AHBM_FIFO_RST; to
host->hw->dma_conf.val |= SPI_IN_RST|SPI_OUT_RST|SPI_AHBM_RST|SPI_AHBM_FIFO_RST;
in 4 places where SPI_OUT_RST appears, the shiffting starts from the end of the receive buffer!
-----------------------------------------------------------------------------------------------------------------------------------------------

I would appreciate if someone could explain this issue.

My guess is that there is still some remaining state in the spi dma controller that does not get reset when the transmission has ended. Thanks for the hint on resetting the IN part as well; I'm working on spi-master anyway and will include this in the next MR for it.

loboris wrote:It looks like there is serious bug in spi-master receive when DMA is used.
I was testing it with ILI9341 display with application based on spi-master example (the example itself works ok).
I've added the function to read line of data from display. Display data format (when read) is:
dummy_byte [R_byte G_byte B_byte] ........ [R_byte G_byte B_byte]
The screen was filled with blue color, so the response should be: 00 00 00 fc 00 00 fc 00 00 fc .....

Here are results of testing:
The first column is sequential number of read, first (dummy) byte is not shown, first 10 RGB values are shown (30 bytes)

When reading less than 33 bytes (no DMA used), the result is as expected, both in duplex and in half duplex mode:

[ 0] Rd line: 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc
[ 1] Rd line: 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc
[ 2] Rd line: 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc
[ 3] Rd line: 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc

When reading more than 32 bytes (DMA used) in duplex mode, no values are received:

[ 0] Rd line: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
[ 1] Rd line: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
[ 2] Rd line: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
[ 3] Rd line: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

When reading more than 32 bytes (DMA used) in half duplex mode (.flags=SPI_DEVICE_HALFDUPLEX), the first read is OK, but for every following read values are shifted by 3 bytes (3, 6, 9, ...) :

[ 0] Rd line: 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc
[ 1] Rd line: 00 80 00 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc
[ 2] Rd line: 00 fc 00 00 28 00 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc
[ 3] Rd line: 00 fc 00 00 fc 00 00 10 00 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc 00 00 fc

--------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Nothing can reset this, even after spi_bus_remove_device, spi_bus_free -> spi_bus_initialize, spi_bus_add_device
and reinitializing the display, the receive buffer is still shifted!
Only the first read after booting is correct.
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------

-----------------------------------------------------------------------------------------------------------------------------------------------
With a little change in spi_master.c:
host->hw->dma_conf.val |= SPI_OUT_RST|SPI_AHBM_RST|SPI_AHBM_FIFO_RST; to
host->hw->dma_conf.val |= SPI_IN_RST|SPI_OUT_RST|SPI_AHBM_RST|SPI_AHBM_FIFO_RST;
in 4 places where SPI_OUT_RST appears, the shiffting starts from the end of the receive buffer!
-----------------------------------------------------------------------------------------------------------------------------------------------

I would appreciate if someone could explain this issue.

Hi,

I have started to work on ILI9488 over HSPI Interface. I have connected following SPI pins in specific numbers.

SPI MISO : Pin 18 : GPIO12
SPI MOSI : Pin 20 : GPIO13
SPI Clock : Pin 17 : GPIO14
SPI CS : Pin 21 : GPIO15

Can you please tell me maximum how much clock HSPI can support in master mode because I have set clock as 500 KHz but still not getting accurate clock which I have measured using CRO.

Please let me know if you have any informations for that.