On Thu, 8 Jan 2009, Guennadi Liakhovetski wrote: > Hi, > > I'm using gadgetfs on an i.MX31 platform, and I took the example from > http://www.linux-usb.org/gadget/ as basis, but put all threads in one, and > I use poll to identify when to read / write from / to an endpoint > descriptor. Then, after I get USB_REQ_SET_CONFIGURATION I open bulk-in and > -out endpoints, on the next poll I add the bulk-out endpoint to the poll > array, get a POLLIN event, do a read(), and it blocks. Here is an excerpt > from strace output: > > poll([{fd=3, events=POLLIN|POLLHUP, revents=POLLIN}], 1, -1) = 1 > read(3, "\x00\x09\x03\x00\x00\x00\x00\x00\x03\x00\x00\x00", 60) = 12 > write(2, "SETUP 00.09 v0003 i0000 0\n", 26) = 26 > write(2, "CONFIG #3\n", 10) = 10 > write(2, "Start bulk EPs\n", 15) = 15 > open("ep1in", O_RDWR) = 4 > write(4, "\x01\x00\x00\x00\x07\x05\x81\x02\x40\x00\x00\x07\x05\x81\x02\x40\x00\x00", 18) = 18 > write(2, "ep1in start fd 4\n", 17) = 17 > open("ep1out", O_RDWR) = 5 > write(5, "\x01\x00\x00\x00\x07\x05\x01\x02\x40\x00\x00\x07\x05\x01\x02\x40\x00\x00", 18) = 18 > write(2, "ep1out start fd 5\n", 18) = 18 > read(3, "", 0) = 0 > write(2, "(1): 2 descriptors: 3, 5\n", 25) = 25 > poll([{fd=3, events=POLLIN|POLLHUP}, {fd=5, events=POLLIN|POLLHUP, revents=POLLIN}], 2, -1) = 1 > read(5, > > This looks to me like a bug in gadgetfs, or am I doing anything wrong? Or > can it be a bug in i.MX31 gadget driver? Ouch, just verified - ep_io_operations doesn't have a poll method at all:-( And O_NONBLOCK doesn't check for actual data availability either:-( Would AIO work?... Thanks Guennadi --- Guennadi Liakhovetski, Ph.D. Freelance Open-Source Software Developer -- 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
