Re: [PATCH 1/2] usb: gadget: composite: Allow function drivers to defer setup responses

[Roger Quadros <roger.quadros@nokia.com>]

On 04/19/2011 05:23 PM, ext Michal Nazarewicz wrote:
> On Tue, 19 Apr 2011 15:33:52 +0200, Roger Quadros <roger.quadros@nokia.com> wrote:
>> @@ -895,6 +904,14 @@ composite_setup(struct usb_gadget *gadget, const struct
>> usb_ctrlrequest *ctrl)
>> if (w_value && !f->set_alt)
>> break;
>> value = f->set_alt(f, w_index, w_value);
>> + if (value == USB_GADGET_DELAYED_STATUS) {
>> + DBG(cdev,
>> + "%s: interface %d (%s) requested delayed status\n",
>> + __func__, intf, f->name);
>> + cdev->delayed_status++;
>
> Do we need it to be a counter? Won't simple flag do?

Counter because multiple interfaces could return USB_GADGET_DELAYED_STATUS
and we need to know how many of them call usb_composite_setup_continue() before 
we actually complete the control transfer.

>
>> + DBG(cdev, "delayed_status count %d\n",
>> + cdev->delayed_status);
>> + }
>> break;
>> case USB_REQ_GET_INTERFACE:
>> if (ctrl->bRequestType != (USB_DIR_IN|USB_RECIP_INTERFACE))
>
>> @@ -1289,3 +1310,41 @@ void usb_composite_unregister(struct
>> usb_composite_driver *driver)
>> return;
>> usb_gadget_unregister_driver(&composite_driver);
>> }
>> +
>> +/**
>> + * usb_composite_setup_continue() - Continue the delayed setup transfer
>> + * @cdev: the composite device who's setup transfer was delayed
>> + *
>> + * This function must be called by the USB function driver to continue
>> + * with the setup transfer's data/status phase in case it had requested to
>> + * delay the status phase. A USB function's setup handler (e.g. set_alt())
>> + * can request the composite framework to delay the setup request's status phase
>> + * by returning USB_GADGET_DELAYED_STATUS.
>> + */
>> +void usb_composite_setup_continue(struct usb_composite_dev *cdev)
>> +{
>> + int value;
>> + struct usb_request *req = cdev->req;
>> + unsigned long flags;
>> +
>> + DBG(cdev, "%s\n", __func__);
>> + spin_lock_irqsave(&cdev->lock, flags);
>> +
>> + if (cdev->delayed_status == 0) {
>> + WARN(cdev, "%s: Unexpected call\n", __func__);
>> +
>> + } else if (--cdev->delayed_status == 0) {
>> + DBG(cdev, "%s: Completing delayed status\n", __func__);
>> + req->length = 0;
>> + req->zero = 1;
>> + value = usb_ep_queue(cdev->gadget->ep0, req, GFP_ATOMIC);
>> + if (value < 0) {
>> + DBG(cdev, "ep_queue --> %d\n", value);
>> + req->status = 0;
>> + composite_setup_complete(cdev->gadget->ep0, req);
>> + }
>> + }Data/status stage.
>> +
>> + spin_unlock_irqrestore(&cdev->lock, flags);
>> +}
>
> It's just my opinion but I dislike how it looks almost identical to the end
> of composite_setup. I would rather make it more general (take value as
> argument for instance) and let composite_setup() call it.

Yes that could be done. I'll try to do so in the next iteration.

>
>> diff --git a/include/linux/usb/composite.h @@ -37,6 +37,15 @@
>> #include <linux/usb/ch9.h>
>> #include <linux/usb/gadget.h>
>> +/*
>> + * USB function drivers should return USB_GADGET_DELAYED_STATUS if they
>> + * wish to delay the status phase of the setup transfer till they are
>> + * ready. The composite framework will then delay the data/status phase
>> + * of the setup transfer till all the function drivers that requested for
>> + * USB_GADGET_DELAYED_STAUS, invoke usb_composite_setup_continue().
>> + *
>> + */
>> +#define USB_GADGET_DELAYED_STATUS 0x7fff /* Impossibly large value */
>
> I was wondering if we could change it to something like EDELAYED and put it to
> <include/linux/errno.h>. Just thinking, it may be a bad idea.
>

I would not do it.

Thanks for review.

-- 
regards,
-roger