Hi Arnd,

Thanks for the comments!

On 12/12/2014 07:30 PM, Arnd Bergmann wrote:

On Friday 12 December 2014 19:14:00 Stanimir Varbanov wrote:

quoted

The PCIe driver reuse the Designware common code for host
and MSI initialization, and also program the Qualcomm
application specific registers.

Signed-off-by: Stanimir Varbanov <redacted>

Looks nice!

quoted

+static int
+qcom_pcie_rd_own_conf(struct pcie_port *pp, int where, int size, u32 *val)
+{
+	if (where == PCI_CLASS_REVISION && size == 4) {
+		*val = readl(pp->dbi_base + PCI_CLASS_REVISION);
+		*val &= ~(0xffff << 16);
+		*val |= PCI_CLASS_BRIDGE_PCI << 16;
+		return PCIBIOS_SUCCESSFUL;
+	}
+
+	return dw_pcie_cfg_read(pp->dbi_base + (where & ~0x3), where,
+				size, val);
+}

Could you add a comment here to explain what this is for?

Sure I will add a comment. The issue is that the pci device class is not
reported correctly from the register.

quoted

+static int __init qcom_pcie_probe(struct platform_device *pdev)
+{

I think it's a bug to mark this function as __init. It breaks
deferred probing and detaching/reattaching the device trough sysfs.

My bad, I have tried to avoid mismatch section warnings came up from
dw_pcie_host_init() which is annotated as __init. Do you think we need
to remove __init from dw_pcie_host_init() declaration and fix the
drivers accordingly?

After you fix that, you can remove the __refdata below.

quoted

+static struct platform_driver __refdata qcom_pcie_driver = {
+	.probe = qcom_pcie_probe,
+	.remove = qcom_pcie_remove,
+	.driver = {
+		.name = "qcom-pcie",
+		.of_match_table = qcom_pcie_match,
+	},
+};

	Arnd

regards,
Stan