On Thu, 15 Apr 2010, Sumit Panchasara wrote: > Hi, > > I have question regarding usb.c file available on > http://www.linux-usb.org/gadget/usb.c. I am using it to debug Ch9 tests > failures from USB20CV software running on Windows host while attaching > dev board configured as device to it. > > The source code has handle_control() routine with case for SET_INTERFACE > control requests from gadgetfs module through event GADGETFS_SETUP. > > Now this particular case has following lines of code: > ---------------------------------------------------------------------------------------------------- > if (ioctl (source_fd, GADGETFS_CLEAR_HALT) < 0) { > status = errno; > perror ("reset source fd"); > } > if (ioctl (sink_fd, GADGETFS_CLEAR_HALT) < 0) { > status = errno; > perror ("reset sink fd"); > } > ---------------------------------------------------------------------------------------------------- > > My question is, if handle_control() is envoked from ep0 thread with > parent fd, what is a reason to pass "source_fd" and "sink_fd" in ioctl > calls above? Why it is not only 'fd'? fd is the file descriptor for endpoint 0. If fd were the first argument to the ioctl's, the kernel would clear a halt condition on endpoint 0. But that isn't what the code wants to accomplish -- it wants to clear halt conditions on the source and sink bulk endpoints. Hence it has to pass source_fd and sink_fd to the ioctl's. > Isn't control request handling > should be done on ep0 fd instead of source/sink fd? Not if handling the control request requires the gadget driver to affect other endpoints. > If I replace 'source/sink fd' with 'fd' coming with handle_control() > call, it gets captured at gadgetfs and subsequently request goes to my > dev board controller driver as well. With this I am able to pass Ch9 > tests from USB20CV software running on Windows Host and configuring my > dev board as device to it. Can you pass the Ch9 tests without changing the original code? > Is it okay to propose change as follows in usb.c file? > ---------------------------------------------------------------------------------------------------- > @@ -1476,13 +1476,9 @@ > > /* just reset toggle/halt for the interface's endpoints */ > status = 0; > - if (ioctl (source_fd, GADGETFS_CLEAR_HALT) < 0) { > + if (ioctl (fd, GADGETFS_CLEAR_HALT) < 0) { > status = errno; > - perror ("reset source fd"); > - } > - if (ioctl (sink_fd, GADGETFS_CLEAR_HALT) < 0) { > - status = errno; > - perror ("reset sink fd"); > + perror ("reset fd"); > } > /* FIXME eventually reset the status endpoint too */ > if (status) > ---------------------------------------------------------------------------------------------------- No, this is wrong. Alan Stern -- 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
