Re: [PATCH v2 3/4] dmaengine: stm32-dma: add support to trigger STM32 MDMA

[Amelie Delaunay]

Hi Marek,

Thanks for reviewing.

On 7/14/22 21:02, Marek Vasut wrote:
> On 7/13/22 16:21, Amelie Delaunay wrote:
>
> [...]
>
> > diff --git a/drivers/dma/stm32-dma.c b/drivers/dma/stm32-dma.c
> > index adb25a11c70f..3916295fe154 100644
> > --- a/drivers/dma/stm32-dma.c
> > +++ b/drivers/dma/stm32-dma.c
> > @@ -142,6 +142,8 @@
> >   #define STM32_DMA_DIRECT_MODE_GET(n)    (((n) & 
> > STM32_DMA_DIRECT_MODE_MASK) >> 2)
> >   #define STM32_DMA_ALT_ACK_MODE_MASK    BIT(4)
> >   #define STM32_DMA_ALT_ACK_MODE_GET(n)    (((n) & 
> > STM32_DMA_ALT_ACK_MODE_MASK) >> 4)
> > +#define STM32_DMA_MDMA_STREAM_ID_MASK    GENMASK(19, 16)
> > +#define STM32_DMA_MDMA_STREAM_ID_GET(n) (((n) & 
> > STM32_DMA_MDMA_STREAM_ID_MASK) >> 16)
>
> Try FIELD_GET() from include/linux/bitfield.h
>
> [...]

Yes, but not only on this new line. I'll add a prior patch to the 
patchset to use FIELD_{GET,PREP}() helpers every where custom macros are 
used.

> > @@ -1630,6 +1670,20 @@ static int stm32_dma_probe(struct 
> > platform_device *pdev)
> >           chan->id = i;
> >           chan->vchan.desc_free = stm32_dma_desc_free;
> >           vchan_init(&chan->vchan, dd);
> > +
> > +        chan->mdma_config.ifcr = res->start;
> > +        chan->mdma_config.ifcr += (chan->id & 4) ? STM32_DMA_HIFCR : 
> > STM32_DMA_LIFCR;
> > +
> > +        chan->mdma_config.tcf = STM32_DMA_TCI;
> > +        /*
> > +         * bit0 of chan->id represents the need to left shift by 6
> > +         * bit1 of chan->id represents the need to extra left shift 
> > by 16
> > +         * TCIF0, chan->id = b0000; TCIF4, chan->id = b0100: left 
> > shift by 0*6 + 0*16
> > +         * TCIF1, chan->id = b0001; TCIF5, chan->id = b0101: left 
> > shift by 1*6 + 0*16
> > +         * TCIF2, chan->id = b0010; TCIF6, chan->id = b0110: left 
> > shift by 0*6 + 1*16
> > +         * TCIF3, chan->id = b0011; TCIF7, chan->id = b0111: left 
> > shift by 1*6 + 1*16
> > +         */
> > +        chan->mdma_config.tcf <<= (6 * (chan->id & 0x1) + 16 * 
> > ((chan->id & 0x2) >> 1));
>
> Some sort of symbolic macros instead of open-coded constants could help 
> readability here.

I agree. As the same kind of computation is done in 
stm32_dma_irq_status() and stm32_dma_irq_clear(), I'll add another prior 
patch to the patchset to introduce new macro helpers to get ISR/IFCR 
offset depending on channel id, and to get channel flags mask depending 
on channel id.

> [...]

Regards,
Amelie