On Tue, Jan 14, 2020 at 06:12:26PM +0100, Jules Maselbas wrote: > On Tue, Jan 14, 2020 at 03:27:04PM +0100, Sascha Hauer wrote: > > Hi Jules, > > > > On Tue, Jan 14, 2020 at 02:21:05PM +0100, Jules Maselbas wrote: > > > Hi Sascha, > > > > > > I've been working on a driver for the dwc2 otg controller, for both host > > > and device mode. Like you I've started from U-Boot driver and I mixed > > > it with some part from Linux. For instance I've removed the register > > > structs and I've been using the same defines for register bit-fields as > > > in Linux. > > > > > > I would like to share my version of the host driver, as the gadget driver > > > one still requires some cleanup. This series is not to be applied on the > > > driver you proposed. However I am willing to propose a new series that can > > > be applied on the driver you proposed. What do you think? > > > > The dwc2 driver was a quick shot, I am all in for improvements. Removing > > the register structs is very nice and I would be interested in gadget > > support as well. > The gadget support is almost ready, I got the dfu gadget working. :) Well, USB gadget support would be a strong argument for your version ;) > > > What SoC are you using the driver on? > I am using this driver on our custom SoC, the Kalray Coolidge MPPA (k1c). > I didn't tried on other SoC. > > > I am not sure if it makes sense to rebase your patches on my driver. If > > you can convince me that yours is better then we could just remove mine > > and merge yours instead. > I don't have a strong argument in favor of my version. Except that I have > already made some changes over the original U-Boot version: > > I've tried to improve the roothub interface making it easier to decode > new requests. > > I have removed the memcpy done in transfer_chunk between the usb data > buffer and the dma buffer :/ > However this mean that all buffer used for usb tranfert must be dma > capable. Not all usb transfers use dma_alloc for it's buffers but > sometime they are allocated on the stack, this can be problematic if > the stack cannot be used by DMA. We do this on other controllers as well, so this shouldn't be a problem. The day will come we either have to fix the callers or create some bounce buffer implementation for non DMA capable memory. Sascha -- Pengutronix e.K. | | Steuerwalder Str. 21 | http://www.pengutronix.de/ | 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 | _______________________________________________ barebox mailing list barebox@xxxxxxxxxxxxxxxxxxx http://lists.infradead.org/mailman/listinfo/barebox
