Re: [PATCH v2 1/5] PCI: dwc: Refactor dw_pcie_edma_find_chip() API

From: Manivannan Sadhasivam
Date: Wed Feb 21 2024 - 00:35:17 EST


On Tue, Feb 20, 2024 at 05:41:41PM +0530, Siddharth Vadapalli wrote:
> On 24/02/16 11:04PM, Manivannan Sadhasivam wrote:
> > In order to add support for Hyper DMA (HDMA), let's refactor the existing
> > dw_pcie_edma_find_chip() API by moving the common code to separate
> > functions.
> >
> > No functional change.
> >
> > Suggested-by: Serge Semin <fancer.lancer@xxxxxxxxx>
> > Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@xxxxxxxxxx>
> > ---
> > drivers/pci/controller/dwc/pcie-designware.c | 40 +++++++++++++++++++++++-----
> > 1 file changed, 33 insertions(+), 7 deletions(-)
> >
> > diff --git a/drivers/pci/controller/dwc/pcie-designware.c b/drivers/pci/controller/dwc/pcie-designware.c
> > index 250cf7f40b85..3a26dfc5368f 100644
> > --- a/drivers/pci/controller/dwc/pcie-designware.c
> > +++ b/drivers/pci/controller/dwc/pcie-designware.c
> > @@ -880,7 +880,17 @@ static struct dw_edma_plat_ops dw_pcie_edma_ops = {
> > .irq_vector = dw_pcie_edma_irq_vector,
> > };
> >
> > -static int dw_pcie_edma_find_chip(struct dw_pcie *pci)
> > +static void dw_pcie_edma_init_data(struct dw_pcie *pci)
> > +{
> > + pci->edma.dev = pci->dev;
> > +
> > + if (!pci->edma.ops)
> > + pci->edma.ops = &dw_pcie_edma_ops;
> > +
> > + pci->edma.flags |= DW_EDMA_CHIP_LOCAL;
> > +}
> > +
> > +static int dw_pcie_edma_find_mf(struct dw_pcie *pci)
> > {
> > u32 val;
> >
> > @@ -902,8 +912,6 @@ static int dw_pcie_edma_find_chip(struct dw_pcie *pci)
> >
> > if (val == 0xFFFFFFFF && pci->edma.reg_base) {
> > pci->edma.mf = EDMA_MF_EDMA_UNROLL;
> > -
> > - val = dw_pcie_readl_dma(pci, PCIE_DMA_CTRL);
> > } else if (val != 0xFFFFFFFF) {
> > pci->edma.mf = EDMA_MF_EDMA_LEGACY;
>
> Minor suggestion:
>
> The above section prior to this patch was:
> if (val == 0xFFFFFFFF && pci->edma.reg_base) {
> pci->edma.mf = EDMA_MF_EDMA_UNROLL;
>
> val = dw_pcie_readl_dma(pci, PCIE_DMA_CTRL);
> } else if (val != 0xFFFFFFFF) {
> pci->edma.mf = EDMA_MF_EDMA_LEGACY;
>
> pci->edma.reg_base = pci->dbi_base + PCIE_DMA_VIEWPORT_BASE;
> } else {
> return -ENODEV;
> }
>
> And this patch is removing the call to dw_pcie_readl_dma() in the "if"
> condition above. So the curly braces after this patch will only be
> present because of the "else if" statement. So is the following change a
> good idea?
>

Yes indeed. Actually, EDMA_MF_EDMA_LEGACY is the default mapping format (because
its value is 0x0), but setting it explicitly would also make sense.

- Mani

--
மணிவண்ணன் சதாசிவம்