On Wednesday 11 December 2013, Mohit Kumar wrote:
> +#include <linux/clk.h>
> +#include <linux/delay.h>
> +#include <linux/interrupt.h>
> +#include <linux/kernel.h>
> +#include <linux/module.h>
> +#include <linux/of.h>
> +#include <linux/pci.h>
> +#include <linux/platform_device.h>
> +#include <linux/resource.h>
> +#include <mach/spear.h>
This won't actually build: you cannot access mach/*.h header files
from outside of mach-spear!
> +struct spear13xx_pcie {
> + void __iomem *phy_base;
> + void __iomem *app_base;
> + struct clk *clk;
> + struct pcie_port pp;
> + int id;
> + int is_gen1;
> +};
The pcie driver shouldn't have direct access to the phy registers,
use a phy driver for that.
> +static int workaround_linkup_spear1340(struct spear13xx_pcie *spear13xx_pcie)
> +{
>
> +}
> +
> +static int spear13xx_pcie_establish_link(struct pcie_port *pp)
> +{
These should move into the phy driver.
> + /* txdetectrx workaround for SPEAr1310 */
> + if (of_machine_is_compatible("st,spear1310"))
> + writeb(0x00, spear13xx_pcie->phy_base + 0x16);
Don't use of_machine_is_compatible() to test for features or bugs of a
particular device. Instead use a binary property in the specific
device node, or key it off of the device's "compatible" value if
there are a lot of differences. You seem to have a couple of these
to test for one or the other PHY implementation, those will
go away when you have a proper driver for them.
Arnd
--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
