Re: [PATCH 3/5] PCI: st: Provide support for the sti PCIe controller

From: Gabriel Fernandez
Date: Mon Jan 19 2015 - 08:08:54 EST


Hi Arnd,

On 17 December 2014 at 23:14, Arnd Bergmann <arnd@xxxxxxxx> wrote:
> On Wednesday 17 December 2014 11:34:44 Gabriel FERNANDEZ wrote:
>> sti pcie is built around a Synopsis Designware PCIe IP.
>>
>> Signed-off-by: Fabrice Gasnier <fabrice.gasnier@xxxxxx>
>> Signed-off-by: Gabriel Fernandez <gabriel.fernandez@xxxxxxxxxx>
>> ---
>> drivers/pci/host/Kconfig | 5 +
>> drivers/pci/host/Makefile | 1 +
>> drivers/pci/host/pci-st.c | 713 ++++++++++++++++++++++++++++++++++++++++++++++
>> 3 files changed, 719 insertions(+)
>> create mode 100644 drivers/pci/host/pci-st.c
>>
>> diff --git a/drivers/pci/host/Kconfig b/drivers/pci/host/Kconfig
>> index c4b6568..999d2b9 100644
>> --- a/drivers/pci/host/Kconfig
>> +++ b/drivers/pci/host/Kconfig
>> @@ -102,4 +102,9 @@ config PCI_LAYERSCAPE
>> help
>> Say Y here if you want PCIe controller support on Layerscape SoCs.
>>
>> +config PCI_ST
>> + bool "ST STiH41x PCIe controller"
>> + depends on ARCH_STI
>> + select PCIE_DW
>
> I'd use 'depends on ARCH_STI || (ARM && COMPILE_TEST)' to enable
> building this on other platforms for test purposes.
>
ok

>> +
>> +#define to_st_pcie(x) container_of(x, struct st_pcie, pp)
>> +
>> +/**
>> + * struct st_pcie_ops - SOC dependent data
>> + * @init: reference to controller power & reset init routine
>> + * @enable_ltssm: reference to controller link enable routine
>> + * @disable_ltssm: reference to controller link disable routine
>> + * @phy_auto: flag when phy automatically configured
>> + */
>> +struct st_pcie_ops {
>> + int (*init)(struct pcie_port *pp);
>> + int (*enable_ltssm)(struct pcie_port *pp);
>> + int (*disable_ltssm)(struct pcie_port *pp);
>> + bool phy_auto;
>> +};
>
ok, i will fix it.

> It would be better not to invent another level of indirection. Try
> turning this around so you have a driver that binds to the specific
> SoC compatible string for the PCIe port while calling into a common
> library module for things that are shared.
>
>> +/*
>> + * On ARM platforms, we actually get a bus error returned when the PCIe IP
>> + * returns a UR or CRS instead of an OK.
>> + */
>> +static int st_pcie_abort_handler(unsigned long addr, unsigned int fsr,
>> + struct pt_regs *regs)
>> +{
>> + return 0;
>> +}
>
> You should check that it's actually PCI that caused the abort. Don't
> just ignore a hard error condition.
>
> Usually there are registers in the PCI core that let you identify what
> happened.
>
We return 0 because abort handler is not activated during boot.

>> +static int st_pcie_rd_other_conf(struct pcie_port *pp, struct pci_bus *bus,
>> + unsigned int devfn, int where, int size,
>> + u32 *val)
>> +{
>> + u32 data;
>> + u32 bdf;
>> + struct st_pcie *pcie = to_st_pcie(pp);
>> + int is_root_bus = pci_is_root_bus(bus);
>> + int retry_count = 0;
>> + int ret;
>> + void __iomem *addr;
>> +
>> + /*
>> + * Prerequisite
>> + * PCI express devices will respond to all config type 0 cycles, since
>> + * they are point to point links. Thus to avoid probing for multiple
>> + * devices on the root, dw-pcie already check for us if it is on the
>> + * root bus / other slots. Also, dw-pcie checks for the link being up
>> + * as we will hang if we issue a config request and the link is down.
>> + * A switch will reject requests for slots it knows do not exist.
>> + */
>> + bdf = bdf_num(bus->number, devfn, is_root_bus);
>> + addr = pcie->config_area + config_addr(where,
>> + bus->parent->number == pp->root_bus_nr);
>> +retry:
>> + /* Set the config packet devfn */
>> + writel_relaxed(bdf, pp->dbi_base + FUNC0_BDF_NUM);
>> + readl_relaxed(pp->dbi_base + FUNC0_BDF_NUM);
>> +
>> + ret = dw_pcie_cfg_read(addr, where, size, &data);
>> +
>> + /*
>> + * This is intended to help with when we are probing the bus. The
>> + * problem is that the wrapper logic doesn't have any way to
>> + * interrogate if the configuration request failed or not.
>> + * On the ARM we actually get a real bus error.
>> + *
>> + * Unfortunately this means it is impossible to tell the difference
>> + * between when a device doesn't exist (the switch will return a UR
>> + * completion) or the device does exist but isn't yet ready to accept
>> + * configuration requests (the device will return a CRS completion)
>> + *
>> + * The result of this is that we will miss devices when probing.
>> + *
>> + * So if we are trying to read the dev/vendor id on devfn 0 and we
>> + * appear to get zero back, then we retry the request. We know that
>> + * zero can never be a valid device/vendor id. The specification says
>> + * we must retry for up to a second before we decide the device is
>> + * dead. If we are still dead then we assume there is nothing there and
>> + * return ~0
>> + *
>> + * The downside of this is that we incur a delay of 1s for every pci
>> + * express link that doesn't have a device connected.
>> + */
>> + if (((where & ~3) == 0) && devfn == 0 && (data == 0 || data == ~0)) {
>> + if (retry_count++ < 1000) {
>> + mdelay(1);
>> + goto retry;
>> + } else {
>> + *val = ~0;
>> + return PCIBIOS_DEVICE_NOT_FOUND;
>> + }
>> + }
>> +
>> + *val = data;
>> + return ret;
>> +}
>
> A busy-loop is extremely nasty. If this is only during the initial bus
> scan, could you use an msleep instead?
>
yes it's during the initial bus scan.
But we can't use msleep because we are under raw_spin_lock_irqsave()
see PCI_OP_READ() macro in drivers/pci/access.c

> Also, it sounds like the error you get is actually the fault that you
> are catching above. If this is correct, then use the fault handler to
> communicate this to the probe function.
>
Same as above the handler is not activated during the boot and initial bus scan.


>> +
>> +static void st_pcie_board_reset(struct pcie_port *pp)
>> +{
>> + struct st_pcie *pcie = to_st_pcie(pp);
>> +
>> + if (!gpio_is_valid(pcie->reset_gpio))
>> + return;
>> +
>> + if (gpio_direction_output(pcie->reset_gpio, 0)) {
>> + dev_err(pp->dev, "Cannot set PERST# (gpio %u) to output\n",
>> + pcie->reset_gpio);
>> + return;
>> + }
>> +
>> + /* From PCIe spec */
>> + usleep_range(1000, 2000);
>> + gpio_direction_output(pcie->reset_gpio, 1);
>> +
>> + /*
>> + * PCIe specification states that you should not issue any config
>> + * requests until 100ms after asserting reset, so we enforce that here
>> + */
>> + usleep_range(100000, 150000);
>> +}
>
> This seems hardly performance critical, just use msleep(2) and
> msleep(100) instead of the usleep_range().
>

ok i will fix it.


>> +static void st_msi_init_one(struct pcie_port *pp)
>> +{
>> + struct st_pcie *pcie = to_st_pcie(pp);
>> +
>> + /*
>> + * Set the magic address the hardware responds to. This has to be in
>> + * the range the PCI controller can write to.
>> + */
>> + dw_pcie_msi_init(pp);
>> +
>> + if ((virt_to_phys((void *)pp->msi_data) < pcie->lmi->start) ||
>> + (virt_to_phys((void *)pp->msi_data) > pcie->lmi->end))
>> + dev_err(pp->dev, "MSI addr miss-configured\n");
>> +}
>
> Why do you call virt_to_phys() here? Isn't msi_data a physical address?
>
msi_data is a virtual address, it's obtained through a
__get_free_pages() function in dw_pcie_msi_init() procedure.

>> +static int __init st_pcie_probe(struct platform_device *pdev)
>
> I'd suggest removing the __init here, as discussed in the review for
> the qualcomm driver.
>
ok

> Arnd

BR

Gabriel.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/