kywwilson11 opened a new pull request, #16803:
URL: https://github.com/apache/nuttx/pull/16803
## Summary
This patch introduces optional double-buffered DMA support for STM32H5 ADC
drivers when operating in circular mode. When `CONFIG_STM32H5_ADCx_DMA_CFG` is
enabled, the DMA buffer size is doubled and half-transfer interrupts are
enabled to allow the CPU to process one half of the buffer while DMA fills the
other.
In one-shot mode (non-circular), DMA and ADC are restarted after each
complete transfer using the new helper `adc_restart_dma()`. This avoids code
duplication and ensures correctness in single-shot conversions.
The DMA configuration has been improved:
- `ntransfers` now accounts for `circular` mode buffer doubling.
- Interrupt flags are now unconditionally cleared to prevent re-entry or
missed interrupts.
- Removed legacy circular DMA logic using `circ_addr_1`, which was unused
and not compatible with GPDMA on STM32H5.
This also improves error handling in `adc_dmaconvcallback()`. Fatal DMA
error flags are detected (e.g., DTEF, ULEF, USEF), logged, and the DMA is
safely stopped. For non-circular mode, DMA is rearmed after such errors to
preserve recoverability.
## Impact
- Adds double-buffered DMA functionality for STM32H5 ADC in circular mode.
- One-shot DMA mode now correctly restarts transfers.
- Acknowledge DMA errors in ADC DMA callback routine.
- Removes dead code related to legacy circular address register usage.
- No impact on existing platforms unless `ADCx_DMA_CFG` is enabled.
## Testing
Testing was performed on a Nucleo-H563ZI board using the `nucleo-h563zi:adc`
configuration as a base. The following system and build environment were used:
- **Host OS:** Ubuntu 24.04 LTS
- **Compiler:** arm-none-eabi-gcc (gcc-arm-none-eabi-10.3-2021.10)
- **Target:** STM32H563ZI
- **App Used:** `examples/adc`
### Configuration Details:
The following defconfig values were changed to test double-buffered DMA:
```
CONFIG_STM32H5_ADC1_DMA_BATCH=2
CONFIG_STM32H5_ADC1_DMA_CFG=y
CONFIG_STM32H5_ADC1_OVERSAMPLE=256
CONFIG_STM32H5_ADC1_OVSR_SHIFT=4
CONFIG_EXAMPLES_ADC_NSAMPLES=10
CONFIG_EXAMPLES_ADC_GROUPSIZ=4
```
All other settings were left unchanged from the base `nucleo-h563zi:adc`
configuration.
### What was tested:
1. Verified successful data acquisition using the `adc` example app.
2. Stepped through `adc_dmaconvcallback()` using GDB to confirm that:
- HTF and TCF were handled correctly.
- Correct buffer offsets were used in circular mode.
- One-shot DMA mode correctly restarted transfers.
3. Tested several `dmabatch` and `CONFIG_ADC_FIFOSIZE` combinations to
validate correct FIFO handling and sample delivery.
4. Tested **one-shot mode** using software triggers:
- Confirmed that `dmabatch = 1` is required in this mode.
- Verified that DMA and ADC restarted properly using the new
`adc_restart_dma()` helper.
### Example Output
Below is a sample run from the `adc` example in circular double-buffered
mode (2 channels × 2 batch):
```
NuttShell (NSH) NuttX-12.10.0
nsh> adc
adc_main: g_adcstate.count: 10
adc_main: Hardware initialized. Opening the ADC device: /dev/adc0
Sample:
1: channel: 3 value: 19984
2: channel: 10 value: 65416
3: channel: 3 value: 19980
4: channel: 10 value: 65412
Sample:
1: channel: 3 value: 19999
2: channel: 10 value: 65427
3: channel: 3 value: 20005
4: channel: 10 value: 65418
Sample:
1: channel: 3 value: 20008
2: channel: 10 value: 65419
3: channel: 3 value: 19986
4: channel: 10 value: 65419
...
```
Values were consistent and interleaved as expected. No dropped samples were
observed.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscr...@nuttx.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
