[RFT PATCH v2 00/42] PCI: ARM/ARM64: remove pci_fixup_irqs() usage

[Lorenzo Pieralisi]

Hi Linus,

On Wed, Jun 21, 2017 at 05:14:13PM +0200, Linus Walleij wrote:
> On Wed, Jun 21, 2017 at 12:45 PM, Lorenzo Pieralisi
> <lorenzo.pieralisi at arm.com> wrote:
> 
> > Sorry I miss yet another usage of struct pci_bus before bus is
> > scanned in faraday_pci_parse_map_dma_ranges(), all these config
> > accessors want is to access bus number 0 and writei/read some config
> > values in there, please confirm.
> >
> > Updated patch here:
> 
> I fixed it. This was needed:
> diff --git a/drivers/pci/host/pci-ftpci100.c b/drivers/pci/host/pci-ftpci100.c
> index 86f8a3d1c1da..5162dffc102b 100644
> --- a/drivers/pci/host/pci-ftpci100.c
> +++ b/drivers/pci/host/pci-ftpci100.c
> @@ -450,6 +450,8 @@ static int faraday_pci_probe(struct platform_device *pdev)
>      struct resource *io;
>      struct pci_host_bridge *host;
>      struct clk *clk;
> +    unsigned char max_bus_speed = PCI_SPEED_33MHz;
> +    unsigned char cur_bus_speed = PCI_SPEED_33MHz;

I could not fix it myself since we do not have the same code base,
I can't find any clock handling in the mainline as per v4.12-rc6.

Is the ftpci100.c clock handling being queued for this merge window ?

This patch:

https://patchwork.ozlabs.org/patch/778716/

fixes a commit that is not in v4.12-rc6, I need some input to understand
how to proceed.

Thanks,
Lorenzo

>      int ret;
>      u32 val;
>      LIST_HEAD(res);
> @@ -553,27 +555,27 @@ static int faraday_pci_probe(struct platform_device *pdev)
>          unsigned long rate;
>          u32 val;
> 
> -        faraday_pci_read_config(p->bus, 0,
> -                    FARADAY_PCI_STATUS_CMD, 4, &val);
> +        faraday_raw_pci_read_config(p, 0, 0,
> +                        FARADAY_PCI_STATUS_CMD, 4, &val);
>          rate = clk_get_rate(p->bus_clk);
> 
>          if ((rate == 33000000) && (val & PCI_STATUS_66MHZ_CAPABLE)) {
>              dev_info(dev, "33MHz bus is 66MHz capable\n");
> -            p->bus->max_bus_speed = PCI_SPEED_66MHz;
> +            max_bus_speed = PCI_SPEED_66MHz;
>              ret = clk_set_rate(p->bus_clk, 66000000);
>              if (ret)
>                  dev_err(dev, "failed to set bus clock\n");
>          } else {
>              dev_info(dev, "33MHz only bus\n");
> -            p->bus->max_bus_speed = PCI_SPEED_33MHz;
> +            max_bus_speed = PCI_SPEED_33MHz;
>          }
> 
>          /* Bumping the clock may fail so read back the rate */
>          rate = clk_get_rate(p->bus_clk);
>          if (rate == 33000000)
> -            p->bus->cur_bus_speed = PCI_SPEED_33MHz;
> +            cur_bus_speed = PCI_SPEED_33MHz;
>          if (rate == 66000000)
> -            p->bus->cur_bus_speed = PCI_SPEED_66MHz;
> +            cur_bus_speed = PCI_SPEED_66MHz;
>      }
> 
>      ret = faraday_pci_parse_map_dma_ranges(p, dev->of_node);
> @@ -587,6 +589,8 @@ static int faraday_pci_probe(struct platform_device *pdev)
>          return ret;
>      }
>      p->bus = host->bus;
> +    p->bus->max_bus_speed = max_bus_speed;
> +    p->bus->cur_bus_speed = cur_bus_speed;
> 
>      pci_bus_assign_resources(p->bus);
>      pci_bus_add_devices(p->bus);
> 
> 
> I'll send a modified and tested version of your patch.
> 
> Yours,
> Linus Walleij