2019-10-25 20:39:27

by Fabrice Gasnier

Subject: [PATCH] iio: adc: stm32-adc: fix stopping dma

There maybe a race when using dmaengine_terminate_all(). The predisable
routine may call iio_triggered_buffer_predisable() prior to a pending DMA
callback.
Adopt dmaengine_terminate_sync() to ensure there's no pending DMA request
before calling iio_triggered_buffer_predisable().

Fixes: 2763ea0585c9 ("iio: adc: stm32: add optional dma support")

Signed-off-by: Fabrice Gasnier <[email protected]>
---
drivers/iio/adc/stm32-adc.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/iio/adc/stm32-adc.c b/drivers/iio/adc/stm32-adc.c
index 663f8a5..73aee59 100644
--- a/drivers/iio/adc/stm32-adc.c
+++ b/drivers/iio/adc/stm32-adc.c
@@ -1399,7 +1399,7 @@ static int stm32_adc_dma_start(struct iio_dev *indio_dev)
cookie = dmaengine_submit(desc);
ret = dma_submit_error(cookie);
if (ret) {
- dmaengine_terminate_all(adc->dma_chan);
+ dmaengine_terminate_sync(adc->dma_chan);
return ret;
}

@@ -1477,7 +1477,7 @@ static void __stm32_adc_buffer_predisable(struct iio_dev *indio_dev)
stm32_adc_conv_irq_disable(adc);

if (adc->dma_chan)
- dmaengine_terminate_all(adc->dma_chan);
+ dmaengine_terminate_sync(adc->dma_chan);

if (stm32_adc_set_trig(indio_dev, NULL))
dev_err(&indio_dev->dev, "Can't clear trigger\n");
--
2.7.4

2019-10-27 17:08:32

by Jonathan Cameron

Subject: Re: [PATCH] iio: adc: stm32-adc: fix stopping dma

On Fri, 25 Oct 2019 17:04:20 +0200
Fabrice Gasnier <[email protected]> wrote:

> There maybe a race when using dmaengine_terminate_all(). The predisable
> routine may call iio_triggered_buffer_predisable() prior to a pending DMA
> callback.
> Adopt dmaengine_terminate_sync() to ensure there's no pending DMA request
> before calling iio_triggered_buffer_predisable().
>
> Fixes: 2763ea0585c9 ("iio: adc: stm32: add optional dma support")
>
> Signed-off-by: Fabrice Gasnier <[email protected]>
Applied to the fixes-togreg branch of iio.git and marked for stable.
I'm guessing this is not something that has been observed in practice, but
seems like a safe fix anyway to backport.

Thanks,

Jonathan

> ---
> drivers/iio/adc/stm32-adc.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/iio/adc/stm32-adc.c b/drivers/iio/adc/stm32-adc.c
> index 663f8a5..73aee59 100644
> --- a/drivers/iio/adc/stm32-adc.c
> +++ b/drivers/iio/adc/stm32-adc.c
> @@ -1399,7 +1399,7 @@ static int stm32_adc_dma_start(struct iio_dev *indio_dev)
> cookie = dmaengine_submit(desc);
> ret = dma_submit_error(cookie);
> if (ret) {
> - dmaengine_terminate_all(adc->dma_chan);
> + dmaengine_terminate_sync(adc->dma_chan);
> return ret;
> }
>
> @@ -1477,7 +1477,7 @@ static void __stm32_adc_buffer_predisable(struct iio_dev *indio_dev)
> stm32_adc_conv_irq_disable(adc);
>
> if (adc->dma_chan)
> - dmaengine_terminate_all(adc->dma_chan);
> + dmaengine_terminate_sync(adc->dma_chan);
>
> if (stm32_adc_set_trig(indio_dev, NULL))
> dev_err(&indio_dev->dev, "Can't clear trigger\n");