On Tue, 29 Jun 2010, Martinez, Aldo wrote:
> Hello everyone
> I'm trying to make a driver that, for now, just answers depending
> on what is sent OUT from the host.
Like the g_zero gadget in loopback mode?
> The problem is that apparently gadgetfs has a queue with the
> usb_requests.
Why is that a problem? After all, USB data should be sent over the bus
in the order given by the gadget driver, which means it should be
stored in a queue, right?
> So whatever I send to the driver I get the answer after
> asking 2 or 3 times, i.e., I have to make two or three IN bulk transfers to get
> what I want.
Then it sounds like you've got a bug in your driver.
> I did some simple modifications to the usb.c example driver found in
> linux-usb.org. I used a global variable to access it from both threads:
> static char *buffer;
>
> In empty_out_buf function I do a simple switch to assing value to the
> global buffer depending on the value sent out by the host:
>
> switch (((char *)buf)[0]) {
> case 0x01:
> printf("case 0x01\n");
> memset(buffer, '1', nbytes);
> break;
> case 0x02:
> printf("case 0x02\n");
> memset(buffer, '2', nbytes);
> break;
> default:
> printf("case default\n");
> memset(buffer, 'd', nbytes);
> break;
> }
>
> The other modification was in simple_source_thread function. Where
> the "buffer" variable created by me is sent instead of the local
> buf variable:
> status = write(source_fd, buffer, USB_BUFSIZE);
>
> I'm using libusb to test on the host side, with the following functions:
> libusb_bulk_transfer (devHandle, BULK_OUT, buffer, bufferSize,
> &transferred, 500);
> libusb_bulk_transfer (devHandle, BULK_IN, buffer, bufferSize,
> &transferred, 500);
>
> When using a buffer size of 511 bytes or less until the third IN
> bulk transfer I get, e.g., in case 0x01 the '1' I assigned before.
> When the buffer size is of 512 or more I get the assigned value
> in the second transfer.
>
> Here is the usbmon trace, USB_BUFSIZE 511:
> f2adf880 32307.036783 S Bi:1:090:1 - 511 <
> f2adf880 32307.036843 C Bi:1:090:1 0 511 = 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
Here the host gets a reply without sending any data first. That seems
like a bug, since your driver is supposed to answer depending on what
it receives from the host. It may be the cause of your other problems.
> f2a78100 32333.300397 S Bo:1:090:1 - 511 = 01010101 01010101 01010101 01010101 01010101 01010101 01010101 01010101
> f2a78100 32333.300472 C Bo:1:090:1 0 511 >
> f2a30500 32349.364816 S Bi:1:090:1 - 511 <
> f2a30500 32349.364975 C Bi:1:090:1 0 511 = 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
> f2ba3400 32350.468457 S Bi:1:090:1 - 511 <
> f2ba3400 32350.468601 C Bi:1:090:1 0 511 = 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
> f2a97680 32351.268549 S Bi:1:090:1 - 511 <
> f2a97680 32351.268600 C Bi:1:090:1 0 511 = 31313131 31313131 31313131 31313131 31313131 31313131 31313131 31313131
>
> USB_BUFSIZE 512:
> f2b86600 35192.622809 S Bi:1:091:1 - 512 <
> f2b86600 35192.622879 C Bi:1:091:1 0 512 = 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
> f2bd2280 35196.388595 S Bo:1:091:1 - 512 = 01010101 01010101 01010101 01010101 01010101 01010101 01010101 01010101
> f2bd2280 35196.388756 C Bo:1:091:1 0 512 >
> f6612080 35198.996722 S Bi:1:091:1 - 512 <
> f6612080 35198.996754 C Bi:1:091:1 0 512 = 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
> f2922680 35200.028271 S Bi:1:091:1 - 512 <
> f2922680 35200.028381 C Bi:1:091:1 0 512 = 31313131 31313131 31313131 31313131 31313131 31313131 31313131 31313131
>
>
> I've also noticed that when "usb" binary it first initialized it calls
> fill_in_buf function (1 or 2 depending on the buffer size as specified
> before).
>
> Is there a form to flush the queue?
There's GADGETFS_FIFO_FLUSH. (Really, you do need to read through the
relevant #include files before trying to do any serious programming.)
It may not do what you want, however.
> Or how can I access
> the information I want right in the next transfer?
That question is not well defined. What exactly are you trying to
accomplish? Try to be precise.
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
