On Tue, Feb 19, 2019 at 10:40:47AM -0500, Alan Stern wrote: > On Tue, 19 Feb 2019, Stanislaw Gruszka wrote: > > > It would be interesting why urb->num_sgs = 0 & urb->sg cause > > the troubles. This is how usb_sg_init() submit urbs for sg_tablesize = 0 > > controllers. So either are there are some requirement on urb->sg > > mapped via dma_map_page() (which mt76usb does not meet) not needed > > for urb->transfer_buffer mapped via dma_map_single() or there > > is something wrong in dwc2 with sg and this driver will not > > work with urb_sg_init() as well. I don't have hardware to investigate > > this and don't want to bother you with more patches. > > urb->sg != NULL and urb->num_sgs == 0 means that the transfer buffer > must fit into a single page, which is pointed to by the first element > of the scatterlist. I asked about that in other thread https://lore.kernel.org/linux-wireless/2cc5674a-a3a0-d8fe-65f5-4357da9b85d3@xxxxxxx/ the answer was it is weird but valid. However I think do dma_map_page() on buffer not fit in the single page is asking for troubles. I just posted patch that should fix this for mt76usb. > But now that I look at the code in usb_sg_init(), it seems odd that the > !use_sg case doesn't increment sg during each loop iteration. I don't > see how that could possibly work -- it looks like a bug! Looks for me that this is done via for_each_sg(sg, sg, io->entries, i) loop. Stanislaw
