On Wed, Dec 05, 2018 at 11:42:02PM +0530, Vinod Koul wrote:
> On 05-12-18, 18:28, Andy Shevchenko wrote:
> > Intel integrated DMA 32-bit support multi-block transfers.
> > Add missed setting to the platform data.
> >
> > Fixes: f7c799e950f9 ("dmaengine: dw: we do support Merrifield SoC in PCI mode")
> > Signed-off-by: Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx>
> > Cc: stable@xxxxxxxxxxxxxxx
>
> How is this a stable material? It would improve performance by using multi
> blocks but given the fact that this is used for slow peripherals, do you
> really see a user impact?
Last my testing was done with SPI VGA panel connected and I ran few FB based
tests with more or less good FPS numbers. So, I can't tell about notable user
impact. Feel free not to add for stable.
Btw, I have tested via DMA test, where it's indeed at least visible, but still
memory-to-memory is not what that DMA is used for.
> > ---
> > drivers/dma/dw/pci.c | 1 +
> > 1 file changed, 1 insertion(+)
> >
> > diff --git a/drivers/dma/dw/pci.c b/drivers/dma/dw/pci.c
> > index 7778ed705a1a..313ba10c6224 100644
> > --- a/drivers/dma/dw/pci.c
> > +++ b/drivers/dma/dw/pci.c
> > @@ -25,6 +25,7 @@ static struct dw_dma_platform_data mrfld_pdata = {
> > .block_size = 131071,
> > .nr_masters = 1,
> > .data_width = {4},
> > + .multi_block = {1, 1, 1, 1, 1, 1, 1, 1},
> > };
> >
> > static int dw_pci_probe(struct pci_dev *pdev, const struct pci_device_id *pid)
> > --
> > 2.19.2
>
> --
> ~Vinod
--
With Best Regards,
Andy Shevchenko
