Re: [PATCH] stmmac: pci: Add support for LS7A bridge chip

[Jiaxun Yang]

å 2020/7/11 9:35, Jiaxun Yang åé:
> å 2020/7/10 16:51, Zhi Li åé:
> > Add gmac platform data to support LS7A bridge chip.
> >
> > Co-developed-by: Hongbin Li <lihongbin@xxxxxxxxxxx>
> > Signed-off-by: Hongbin Li <lihongbin@xxxxxxxxxxx>
> > Signed-off-by: Zhi Li <lizhi01@xxxxxxxxxxx>
> > ---
> > Â drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c | 22 
> > ++++++++++++++++++++++
> > Â 1 file changed, 22 insertions(+)
> >
> > diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c 
> > b/drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c
> > index 272cb47..dab2a40 100644
> > --- a/drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c
> > +++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c
> > @@ -138,6 +138,24 @@ static const struct stmmac_pci_info 
> > snps_gmac5_pci_info = {
> > ÂÂÂÂÂ .setup = snps_gmac5_default_data,
> > Â };
> > Â +static int loongson_default_data(struct pci_dev *pdev, struct 
> > plat_stmmacenent_data *plat)
> > +{
> > +ÂÂÂ common_default_data(plat);
> > +
> > +ÂÂÂ plat->bus_id = pci_dev_id(pdev);
> > +ÂÂÂ plat->phy_addr = 0;
> > +ÂÂÂ plat->interface = PHY_INTERFACE_MODE_GMII;
> > +
> > +ÂÂÂ plat->dma_cfg->pbl = 32;
> > +ÂÂÂ plat->dma_cfg->pblx8 = true;
> > +
> > +ÂÂÂ return 0;
> > +}
> > +
> > +static struct stmmac_pci_info loongson_pci_info = {
> > +ÂÂÂ .setup = loongson_default_data;
> > +};
> > +
> > Â /**
> > ÂÂ * stmmac_pci_probe
> > ÂÂ *
> > @@ -204,6 +222,8 @@ static int stmmac_pci_probe(struct pci_dev *pdev,
> > ÂÂÂÂÂ res.addr = pcim_iomap_table(pdev)[i];
> > ÂÂÂÂÂ res.wol_irq = pdev->irq;
> > ÂÂÂÂÂ res.irq = pdev->irq;
> > +ÂÂÂ if (pdev->vendor == PCI_VENDOR_ID_LOONGSON)
> > +ÂÂÂÂÂÂÂ res.lpi_irq = pdev->irq + 1;
>
> This can never work.
> We're allocating IRQs by irq_domain, not ID.
> Please describe IRQ in DeviceTree, and *DO NOT* sne dout untested patch.
Oops, ^ send out

FYI: Here is my solution of GMAC[1] [2], I was too busy to upstream it.
We're using totally different structure than Loongson's out of tree Kernel,
especially in IRQ management.

Please don't simply copy 'n paste code from your company's internal kernel.

Please try to understand how upstream kernel is working and test your 
patches
with upstrem kernel.

[1]: 
https://github.com/FlyGoat/linux/commit/9d6584c186a8007f14dc8bb2524e48a2fd7d689a
[2]: 
https://github.com/FlyGoat/linux/commit/558a256acfeb022e132113e7952a9df3df375302

> Thanks.
>
> > Â ÂÂÂÂÂ return stmmac_dvr_probe(&pdev->dev, plat, &res);
> > Â }
> > @@ -273,11 +293,13 @@ static SIMPLE_DEV_PM_OPS(stmmac_pm_ops, 
> > stmmac_pci_suspend, stmmac_pci_resume);
> > Â Â #define PCI_DEVICE_ID_STMMAC_STMMACÂÂÂÂÂÂÂ 0x1108
> > Â #define PCI_DEVICE_ID_SYNOPSYS_GMAC5_IDÂÂÂÂÂÂÂ 0x7102
> > +#define PCI_DEVICE_ID_LOONGSON_GMACÂÂÂÂÂÂÂ 0x7a03
> > Â Â static const struct pci_device_id stmmac_id_table[] = {
> > ÂÂÂÂÂ { PCI_DEVICE_DATA(STMMAC, STMMAC, &stmmac_pci_info) },
> > ÂÂÂÂÂ { PCI_DEVICE_DATA(STMICRO, MAC, &stmmac_pci_info) },
> > ÂÂÂÂÂ { PCI_DEVICE_DATA(SYNOPSYS, GMAC5_ID, &snps_gmac5_pci_info) },
> > +ÂÂÂ { PCI_DEVICE_DATA(LOONGSON, GMAC, &loongson_pci_info) },
> > ÂÂÂÂÂ {}
> > Â };
> - Jiaxun