Isochronous transfer missing frames on USB4?

[Engineering]

Hello all. I am having an odd problem getting a UVC webcam to work on
FreeBSD

Little background:
This is for an arcade game - needs to be able to run for weeks unattended at
a bar
Running of Intel D945PLNM board, nVidia 9500GT series video card. Storage is
compact flash, with a CF-IDE adapter
Webcam is no-name Chinese, which is supposed to conform to UVC spec - no
other documentation

'Driver' is modified ugen with Hans Petter's newest stack. Running
high-speed isochronous transfers.

We've been running for months with no problems, but now that production
motherboards and cameras are coming in, something is cropping up.

The webcam is locked at 160x120, 30fps

At this rate, it always gives me 39 frames of data per video frame
37 frames of 1024, 1 frame of 968, 1 frame of 12
Minus the 12 bytes of header on each packet, that is 37*1012+956 = 38400
bytes, which is exactly correct
The last 12 byte header contains no data, but contains the end of frame bit

Cannot confirm, but it only seems to happen when webcam is plugged into the
USB port directly underneath the etherNet jack.

Timing issue? Priority? Any suggestions on where to look?

Thanks,
Sam

I have _tons_ of logs and source, but I'll snip what seems relevant for
now...


Apr 28 17:09:04  kernel: FreeBSD 7.0-RELEASE #9: Wed Mar 18 08:04:55 CDT
2009
Apr 28 17:09:04  kernel: usb4: EHCI version 1.0
Apr 28 17:09:04  kernel: usb4: Intel 82801GB/R (ICH7) USB 2.0 controller
on ehci0

Apr 28 17:09:04  kernel: usb0: 12Mbps Full Speed USB v1.0
Apr 28 17:09:04  kernel: uhub0: Intel UHCI root HUB, class 9/0, rev
1.00/1.00, addr 1 on usb0
Apr 28 17:09:04  kernel: uhub0: 2 ports with 2 removable, self powered
Apr 28 17:09:04  kernel: usb1: 12Mbps Full Speed USB v1.0
Apr 28 17:09:04  kernel: uhub1: Intel UHCI root HUB, class 9/0, rev
1.00/1.00, addr 1 on usb1
Apr 28 17:09:04  kernel: uhub1: 2 ports with 2 removable, self powered
Apr 28 17:09:04  kernel: usb2: 12Mbps Full Speed USB v1.0
Apr 28 17:09:04  kernel: uhub2: Intel UHCI root HUB, class 9/0, rev
1.00/1.00, addr 1 on usb2
Apr 28 17:09:04  kernel: uhub2: 2 ports with 2 removable, self powered
Apr 28 17:09:04  kernel: usb3: 12Mbps Full Speed USB v1.0
Apr 28 17:09:04  kernel: uhub3: Intel UHCI root HUB, class 9/0, rev
1.00/1.00, addr 1 on usb3
Apr 28 17:09:04  kernel: uhub3: 2 ports with 2 removable, self powered
Apr 28 17:09:04  kernel: usb4: 480Mbps High Speed USB v2.0
Apr 28 17:09:04  kernel: uhub4: Intel EHCI root HUB, class 9/0, rev
2.00/1.00, addr 1 on usb4
Apr 28 17:09:04  kernel: uhub4: 8 ports with 8 removable, self powered
Apr 28 17:09:04  kernel: ugen0: Vimicro Corp. Sirius USB2.0 Camera, class
239/2, rev 2.00/1.00, addr 2 on usb4
Apr 28 17:09:04  kernel: make_dev uvc.1

// My debug messages follow
Apr 28 17:09:19  kernel: set config 0
Apr 28 17:09:19  kernel: make_dev uvc.1
Apr 28 17:09:19  kernel: set config 1
Apr 28 17:09:19  kernel: make_dev uvc.1
Apr 28 17:09:19  kernel: make_dev uvc.2
Apr 28 17:09:19  kernel: ugen_get_cdesc -1
Apr 28 17:09:19  kernel: usbd_find_idesc 0 0
Apr 28 17:09:19  kernel: UVC Start read uvc returned 0
Apr 28 17:09:19  kernel: High speed 64 frames
Apr 28 17:09:19  kernel: ugen_open_pipe_read: isoc open done

// Megabytes of this stuff...  Shows the raw byte count of each of the 64
frames coming back on iso_read_callback


Apr 28 17:09:19  kernel: 
 0 1024 0 0 0 0 0 1024 0 0 0 0 0 0 0 1024 0 0 0 0 968 12 0 0 0 0 0 0 0 0 0 0
 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1024 0 0 0 0 0 0 1024 0 0 0 0 0 
Apr 28 17:09:19  kernel: 
 1024 0 0 0 0 0 1024 0 0 0 0 0 0 1024 0 0 0 0 0 0 0 1024 0 0 0 0 0 1024 0 0
0 0
 0 1024 0 0 0 0 0 0 1024 0 0 0 0 0 1024 0 0 0 0 0 1024 0 0 0 0 0 0 0 1024 0
0 0 
Apr 28 17:09:19  kernel: 
 0 0 0 1024 0 0 0 0 0 1024 0 0 0 0 0 1024 0 0 0 0 0 0 1024 0 0 0 0 0 1024 0
0 0
 0 0 0 0 1024 0 0 0 0 0 0 1024 0 0 0 0 0 1024 0 0 0 0 0 1024 0 0 0 0 0 0
1024 0 
Apr 28 17:09:19  kernel: 
 0 0 0 0 1024 0 0 0 0 0 1024 0 0 0 0 0 0 0 1024 0 0 0 0 0 0 1024 0 0 0 0 0
1024
 0 0 0 0 0 1024 0 0 0 0 0 0 1024 0 0 0 0 0 1024 0 0 0 0 0 0 0 1024 0 0 0 0 0

Apr 28 17:09:19  kernel:
 1024 0 0 0 0 0 0 1024 0 0 0 0 0 1024 0 0 0 0 0 1024 0 0 0 0 0 0 1024 0 0 0
0 0
 0 0 1024 0 0 0 968 12 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 

37 frames of 1024, 1 frame of 968, 1 frame of 12
Minus the 12 bytes of header on each packet, that is 37*1012+956 = 38400
bytes, which is exactly correct
The last 12 byte header contains no data, but contains the end of frame bit

After time - problems cropping up - a complete packet goes missing - easiest
to see if the 968 or 12 goes, but it happens to the 1024s too
When missing, the packet is completely gone - 0 bytes

Apr 29 13:40:25  kernel: 
 0 0 0 1024 0 0 0 0 0 1024 0 0 0 0 0 0 0 1024 0 0 0 0 968 12 0 0 0 0 0 0 0 0
 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1024 0 0 0 0 0 1024 0 0 0 
Apr 29 13:40:25  kernel: 
 0 0 1024 0 0 0 0 0 0 1024 0 0 0 0 0 1024 0 0 0 0 0 0 0 1024 0 0 0 0 0 1024
0 0
 0 0 0 0 1024 0 0 0 0 0 1024 0 0 0 0 0 1024 0 0 0 0 0 0 

Re: Isochronous transfer missing frames on USB4?

[Hans Petter Selasky]

On Wednesday 29 April 2009, Engineering wrote:
 Hello all. I am having an odd problem getting a UVC webcam to work on
 FreeBSD

 Little background:
 This is for an arcade game - needs to be able to run for weeks unattended
 at a bar
 Running of Intel D945PLNM board, nVidia 9500GT series video card. Storage
 is compact flash, with a CF-IDE adapter
 Webcam is no-name Chinese, which is supposed to conform to UVC spec - no
 other documentation

 'Driver' is modified ugen with Hans Petter's newest stack.

Are you running stock -current code without any patches? And from which date?

Do you have more than 4GB of RAM?

 Running 
 high-speed isochronous transfers.

 We've been running for months with no problems, but now that production
 motherboards and cameras are coming in, something is cropping up.

 The webcam is locked at 160x120, 30fps

 At this rate, it always gives me 39 frames of data per video frame
 37 frames of 1024, 1 frame of 968, 1 frame of 12
 Minus the 12 bytes of header on each packet, that is 37*1012+956 = 38400
 bytes, which is exactly correct
 The last 12 byte header contains no data, but contains the end of frame bit

 Cannot confirm, but it only seems to happen when webcam is plugged into the
 USB port directly underneath the etherNet jack.

 Timing issue? Priority? Any suggestions on where to look?

Maybe an USB analyser (for example the Beagle Ones will give you an answer).

The same software works on one hardware but not the other?

Could you use the usbconfig tool to dump the descriptors of the USB webcam?

 37 frames of 1024, 1 frame of 968, 1 frame of 12
 Minus the 12 bytes of header on each packet, that is 37*1012+956 = 38400
 bytes, which is exactly correct
 The last 12 byte header contains no data, but contains the end of frame bit

 After time - problems cropping up - a complete packet goes missing -
 easiest to see if the 968 or 12 goes, but it happens to the 1024s too
 When missing, the packet is completely gone - 0 bytes


The new USB stack provides you with timing information about isochronous 
transfers, see xfer-isoc_time_complete. This field is a 16-bit millisecond 
counter which tells exactly at which timeslot the isochronous transfer ended. 
You can use this field to figure out if you missed any frames when the frame 
rate is constant.

You might be aware about it: Isochronous transfers do not have any 
retransmission.

--HPS
___
freebsd-usb@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-usb
To unsubscribe, send any mail to freebsd-usb-unsubscr...@freebsd.org

Re: Isochronous transfer missing frames on USB4?

[Hans Petter Selasky]

Another idea:

Have you checked that the number of isochronous frames you setup is a multiple 
of 8 ?

--HPS
___
freebsd-usb@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-usb
To unsubscribe, send any mail to freebsd-usb-unsubscr...@freebsd.org