26 May
2011
26 May
'11
11:35 a.m.
Ken Restivo wrote:
On Wed, May 25, 2011 at 08:55:48AM +0200, Jeremy
Jongepier wrote:
> http://lists.linuxaudio.org/pipermail/linux-audio-user/2010-June/070353.html
That bug in Intel's Rate Matching Hub would corrupt audio data, but
is unlikely to result in error messages from the driver. (Iso packets
have no error correction; corrupted data is just silently ignored.)
> So not mixing USB1 and USB2 devices but mixing a
device performing
> Full-Speed isochronous-out transactions with a device performing
> asynchronous transactions. I've tested my Edirol UA-25 with this chipset
> too and no matter what I tried, full duplex wouldn't work.
The Linux high speed support is somewhat buggy; this affects full
speed devices connected to high speed hubs.
The Rate Matching Hub is internally organized as a high speed hub.
Wait, what?
I've been doing way too much reading of USB specs (and the hub.c,
message.c, and musb driver source code) for work lately, and even though
I'm definitely suffering from early onset of Alzheimer's, I distinctly
remember reading that Isochronous and Bulk endpoints were only allowed
at High Speed (480Mbps).
At Full speed (12Mbps), IIRC, you get Interrupt and Control endpoints,
and that's pretty much it.
Iso/bulk are forbidden only at Low speed (1.5Mbps).
Regards,
Clemens