Re: [PATCH v1 6/9] usb: xhci: Add NVIDIA Tegra XHCI host-controller driver

From: Andrew Bresticker
Date: Tue Jul 08 2014 - 17:52:18 EST


On Fri, Jun 20, 2014 at 9:58 AM, Julius Werner <jwerner@xxxxxxxxxxxx> wrote:
>> +static const struct hc_driver tegra_xhci_hc_driver = {
>> + .description = "tegra-xhci-hcd",
>> + .product_desc = "Tegra xHCI Host Controller",
>> + .hcd_priv_size = sizeof(struct xhci_hcd *),
>> +
>> + /*
>> + * generic hardware linkage
>> + */
>> + .irq = xhci_irq,
>> + .flags = HCD_MEMORY | HCD_USB3 | HCD_SHARED,
>> +
>> + /*
>> + * basic lifecycle operations
>> + */
>> + .reset = tegra_xhci_setup,
>> + .start = xhci_run,
>> + .stop = xhci_stop,
>> + .shutdown = xhci_shutdown,
>> +
>> + /*
>> + * managing i/o requests and associated device resources
>> + */
>> + .urb_enqueue = xhci_urb_enqueue,
>> + .urb_dequeue = xhci_urb_dequeue,
>> + .alloc_dev = xhci_alloc_dev,
>> + .free_dev = xhci_free_dev,
>> + .alloc_streams = xhci_alloc_streams,
>> + .free_streams = xhci_free_streams,
>> + .add_endpoint = xhci_add_endpoint,
>> + .drop_endpoint = xhci_drop_endpoint,
>> + .endpoint_reset = xhci_endpoint_reset,
>> + .check_bandwidth = xhci_check_bandwidth,
>> + .reset_bandwidth = xhci_reset_bandwidth,
>> + .address_device = xhci_address_device,
>> + .enable_device = xhci_enable_device,
>> + .update_hub_device = xhci_update_hub_device,
>> + .reset_device = xhci_discover_or_reset_device,
>> +
>> + /*
>> + * scheduling support
>> + */
>> + .get_frame_number = xhci_get_frame,
>> +
>> + /* Root hub support */
>> + .hub_control = xhci_hub_control,
>> + .hub_status_data = xhci_hub_status_data,
>> + .bus_suspend = xhci_bus_suspend,
>> + .bus_resume = xhci_bus_resume,
>> +};
>
> I know I missed the first round of discussion where this was
> suggested, but I don't think it's a good idea to pull the whole
> hc_driver structure out into every platform implementation. It will
> lead to duplication, then to future additions only being applied to
> some of the implementations and everything getting out of sync. This
> is already a problem with the PCI/plat split (e.g. the LPM functions
> were only added to xhci-pci even though they should apply to both).
> Also, if I'm not mistaken this code would fail to compile as a module
> (you are referencing lots of symbols that are internal to the xhci-hcd
> module).

You're right Julius, this won't build as a module without a few EXPORT_SYMBOLs.

> I think at the very least you should add a function
> "xhci_default_driver(struct hc_driver *driver)" to xhci-plat.c (or
> even better to xhci.c and use it for PCI as well) that initializes all
> function pointers to the default (internal) symbols, and can then be
> overridden afterwards.

Currently all XHCI host drivers (PCI, platform, MVEBU) will be built
into the xhci-hcd module. I could append the Tegra driver to that
module or introduce a xhci_init_driver() like EHCI does as Julius
suggests. USB folks, do you have a preference?
--
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/