Possible device postinit bug in the USB 5.0 stack

johnstrohm — Wed, 17 Dec 2014 20:48:11 GMT

Routine usb_dci_khci_postinit() enables the D+ pullup resistor, enables the USB module, and clears the Suspend condition on the bus. Presumably, there is an equivalent routine in ehci_dev.c, but I haven't looked for it.

usb_dci_khci_postinit() is called through usb_device_postinit().

Every device class driver's class init routine includes a call to usb_device_postinit().

That specifically includes usb_composite.c, the Composite device class driver.

This means that a composite device is going to issue several calls to usb_device_postinit(). This is probably OK for now, until/unless someone adds code to the routine that should only be executed once.

Suggested fix: Wrap the calls in usb_audio.c, usb_dfu.c, usb_hid.c, usb_msc.c, ..., in #if !USBCFG_DEV_COMPOSITE/#endif pairs. This guarantees that devices compiled as standalone devices (not composite) will execute the call as always, and composite devices will only execute the call in usb_composite.c.

Re: Possible device postinit bug in the USB 5.0 stack

Kan_Li — Thu, 18 Dec 2014 07:44:59 GMT

Hi John,

Thanks for the suggestion! I will forward this to our USB stack develop team.

Thanks again and Best Regards,

Kan

topic Re: Possible device postinit bug in the USB 5.0 stack in Kinetis Microcontrollers

Possible device postinit bug in the USB 5.0 stack

Re: Possible device postinit bug in the USB 5.0 stack