[alsa-devel] Boss GT-001

[Ricard Wanderlof]

On Tue, 6 Oct 2015, Keith A. Milner wrote:

> OK, so it seems that this hardware is substantially the same as the GT-100 
> which has previously (and recently) been raised as a problem on ALSA.
> 
> Clemens Ladisch wrote:
> > Aaron Spike wrote:
> > > On Sat, Jan 10, 2015 at 2:41 PM, Clemens Ladisch wrote:
> > > > There is something that cannot be derived from the descriptors.
> > > 
> > > In a nutshell, what would someone who knew what they were doing do in this
> > > situation?
> > 
> > 1. Use some USB sniffer to find out what commands the Windows driver sends.
> > 2. Reverse engineer the firmware.
> > 
> 
> I'm going to attempt to capture some USB data using this device in Windows. If 
> I do this, is there an accepted method for uploading the capture so it can be 
> posted here?
> 
> At this point, I'm not 100% sure what the capture format is going to be...

I just went down this road while researching the Zoom R16. There's a 
Windows application called USBPCap which can sniff an USB bus and dump 
packets in .pcap format. A reasonally recent version of Wireshark can then 
be used to analyze the data, as it supports the USBPCap output format from 
a certain version (FWIW, the version of Wireshark included with Debian 
Jessie works fine).

The corresponding dumps on Linux can be done by loading the usbmon module 
into the kernel. This allows Wireshark to sniff the USB buses.

I learned though that although Wireshark is happy reading the .pcap files 
produced by USBPCap, the resulting dumps can look quite different from 
what Wireshark sniffs in Linux, which initially took me aback. At least 
when it comes to USBPcap I think the reason is that the program actually 
sniffs the data halfway up (if not at the top of) the Windows protocol 
stack, meaning that there's data in the dump that does not actually occur 
on the wire. I'm not sure at what level usbmon sniffs, but it looks closer 
to what I'd expect on the physical cable.

In practice, all data is there in both cases, but it takes a bit of 
practice to actually compare the two successfully. Especially when it 
comes to isochronous packets (used for audio data transfer), the USBPcap 
format decodes the individual packets, whereas a dump generated via usbmon 
just delivered one large blob of data, which included both the isochronous 
packet headers and associated payload. Perhaps a more recent Wireshark 
might fare better; at the time the level of detail was sufficient for me 
so I didn't pursue it any further.

/Ricard
-- 
Ricard Wolf Wanderlöf                           ricardw(at)axis.com
Axis Communications AB, Lund, Sweden            www.axis.com
Phone +46 46 272 2016                           Fax +46 46 13 61 30