Hi, >> >In fact, this problem starts even _before_ the INQUIRY command is >> >received. The log shows that the UDC driver calls the >> >bulk_out_complete routine over and over, even though no packet was >> >received and ka2000_ack_irq didn't run. The same thing happens after >> >the INQUIRY is received. >> >The UDC driver is not supposed to call bulk_out_complete until a >> >transfer has completed. >> >After the gadget was reset, the problem didn't occur any more. >> >> Looking at the log, the UDC driver queue function is called repeatedly >> by start_transfer() before SCSI_INQUIRY is received. Is this ok? >> >> [start_transfer] ebfffffe e5943010 >> ept1 out queue len 0x200, buffer 0xc133c000 >> bulk_out_complete --> 0, 0/31 > > This is not a good question. You don't have to worry about > start_transfer; it is part of file_storage.c and it works correctly. > > The problem is: Why does bulk_out_complete get called so many times? > The UDC driver should not call bulk_out_complete if no bulk-out packets > have been received. > >> When debugging the UDC driver halt problem, i disable setting the halt >> feature for bulk-in endpoint and does not call fsg_disconnect() when >> reset interrupt occurs. > > Why are you disabling this? You should fix it, not disable it. > >> Now when unknown SCSI command 0xa1 is >> received, the usbmon trace still shows the same error, and subsequent >> TEST_UNIT_READY is not received by UDC driver, and (-104) is >> connection reset error. It seems that UDC and gadget driver somehow >> still reset the USB connection? Driver has to reset the USB connection >> when unknown SCSI command is received? > > The reset is done by the host, not by the UDC or the gadget driver. > > I don't know why the subsequent TEST UNIT READY was not received. I > can't tell what is happening inside the gadget and UDC just by reading > the usbmon log (which was recorded on the host). >From the usbmon trace and driver log, i can only see that TEST UNIT READY command is sent out but UDC driver does not receive it. May i ask, under what circumstances, is gadget driver calling start_transfer() to schedule reading from bulk-out endpoint ? Thanks, victor -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
